Algebra – Abstract

and Modern

Dr U. M. Swamy
Dean, Faculty of Science (Retired)
Department of Mathematics
Andhra University
Visakhapatnam

Dr A. V. S. N. Murty
Professor of Mathematics
Srinivasa Institute of Engineering and Technology
Amalapuram

Q001-Algebra-111001_FM.indd 1 9/23/2011 2:17:40 PM

 Copyright © 2012 Dorling Kindersley (India) Pvt. Ltd
Licensees of Pearson Education in South Asia

No part of this eBook may be used or reproduced in any manner whatsoever

without the publisher’s prior written consent.

This eBook may or may not include all assets that were part of the print version.
The publisher reserves the right to remove any material present in this eBook at
any time.

ISBN 9788131758922
eISBN 9789332509931

Head Office: A-8(A), Sector 62, Knowledge Boulevard, 7th Floor, NOIDA
201 309, India
Registered Office: 11 Local Shopping Centre, Panchsheel Park, New Delhi
110 017, India

Q001-Algebra-111001_FM.indd 2 9/23/2011 2:17:40 PM

 To my grandfather – Late Sri. Akella Srihari,
To my father – Late Sri. Akella Krishna Murty,
To my teacher – Prof. U. M. Swamy

and

My family members

A. V. S. N. Murty

Q001-Algebra-111001_FM.indd 3 9/23/2011 2:17:40 PM

 This page is intentionally left blank.

Q001-Algebra-111001_FM.indd 4 9/23/2011 2:17:40 PM

 Contents

Preface ix

Part I: Preliminaries
1. Sets and Relations 1-3
1.1 Sets and subsets 1-3
1.2 Relations and functions 1-11
1.3 Equivalence relations and partitions 1-21
1.4 The cardinality of a set 1-27

2. Number Systems 2-1

2.1 Integers 2-1
2.2 Congruence modulo n 2-13
2.3 Rational, real and complex numbers 2-23
2.4 Ordering 2-30
2.5 Matrices 2-34
2.6 Determinants 2-43

Part II: Group Theory

3. Groups 3-3
3.1 Binary systems 3-3
3.2 Groups 3-16
3.3 Elementary properties of groups 3-32
3.4 Finite groups and group tables 3-45

4. Subgroups and Quotient Groups 4-1

4.1 Subgroups 4-2
4.2 Cyclic groups 4-12
4.3 Cosets of a subgroup 4-24
4.4 Lagrange’s theorem 4-30
4.5 Normal subgroups 4-39
4.6 Quotient groups 4-45

Q001-Algebra-111001_FM.indd 5 9/23/2011 2:17:40 PM

 vi   Contents

5. Homomorphisms of Groups 5-1

5.1 Definition and examples 5-2
5.2 Fundamental theorem of homomorphisms 5-16
5.3 Isomorphism theorems 5-23
5.4 Automorphisms 5-29

6. Permutation Groups 6-1

   6.1 Cayley’s theorem 6-1
   6.2 The symmetric group Sn 6-7
   6.3 Cycles 6-11
   6.4 Alternating group An and dihedral group Dn 6-23

7. Group Actions on Sets 7-1

7.1 Action of a group on a set 7-1
7.2 Orbits and stabilizers 7-8
7.3 Certain counting techniques 7-19
7.4 Cauchy and Sylow theorems 7-28

8. Structure Theory of Groups 8-1

   8.1 Direct products 8-1
   8.2 Finitely generated abelian groups 8-12
   8.3 Invariants of finite abelian groups 8-29
   8.4 Groups of small order 8-33

Part III: Ring Theory

9. Rings 9-3
   9.1 Examples and elementary properties 9-3
   9.2 Certain special elements in rings 9-16
   9.3 The characteristic of a ring 9-22
   9.4 Subrings 9-25
   9.5 Homomorphisms of rings 9-29
   9.6 Certain special types of rings 9-35
   9.7 Integral domains and fields 9-43
10. Ideals and Quotient Rings 10-1
10.1 Ideals 10-1
10.2 Quotient rings 10-20
10.3 Chinese remainder theorem 10-29

Q001-Algebra-111001_FM.indd 6 9/23/2011 2:17:40 PM

 Contents   vii

10.4 Prime ideals 10-34

10.5 Maximal ideals 10-43
10.6 Embeddings of rings 10-56
11. Polynomial Rings 11-1
11.1 Rings of polynomials 11-1
11.2 The division algorithm 11-15
11.3 Polynomials over a field 11-25
11.4 Irreducible polynomials 11-31

12. Factorization in Integral Domains 12-1

12.1 Divisibility in integral domains 12-2
12.2 Principal ideal domains 12-10
12.3 Unique factorization domains 12-18
12.4 Polynomials over UFDs 12-26
12.5 Euclidean domains 12-36
12.6 Some applications to number theory 12-44

13. Modules and Vector Spaces 13-1

13.1 Modules and submodules 13-2
13.2 Homomorphisms and quotients of modules 13-10
13.3 Direct products and sums 13-18
13.4 Simple and completely reducible modules 13-31
13.5 Free modules 13-35
13.6 Vector spaces 13-42

Part IV: Field Theory

14. Extension Fields 14-3
14.1 Extensions of a field 14-3
14.2 Algebraic extensions 14-8
14.3 Algebraically closed fields 14-20
14.4 Derivatives and multiple roots 14-27
14.5 Finite fields 14-33

15. Galois Theory 15-1

15.1 Separable and normal extensions 15-1
15.2 Automorphism groups and fixed fields 15-10
15.3 Fundamental theorem of Galois theory 15-19

Q001-Algebra-111001_FM.indd 7 9/23/2011 2:17:41 PM

 viii   Contents

16. Selected Applications of Galois Theory 16-1

16.1 Fundamental theorem of algebra 16-2
16.2 Cyclic extensions 16-5
16.3 Solvable groups 16-8
16.4 Polynomials solvable by radicals 16-11
16.5 Constructions by ruler and compass 16-20

 nswers/Hints to Selected Even-Numbered Exercises

A A-1
Index I-1

Q001-Algebra-111001_FM.indd 8 9/23/2011 2:17:41 PM

 Preface

This book is designed for a two-semester sequence as a first course in abstract

algebra for advanced undergraduate and junior post-graduate students. A
glance at the table of contents will reveal the scope of the book; the range of
topics covered is reasonably standard, with no major surprises. Our intention
is to present a text that is logically developed, precise, and in keeping with
the spirit of the times. Guided by the principle that a routine diet of defini-
tions, theorems and results soon becomes unpalatable, we have concentrated
on supplementing the concepts with examples and counter-examples and on
establishing the important and fruitful results in a formal, rigorous fashion.
En route, we have tried to showcase the power and elegance of the abstract –
modern approach in mathematics, particularly in algebra, and chosen the title
‘Algebra – Abstract and Modern’ for this book.
The reader is not presumed to possess any previous knowledge of the con-
cepts of modern algebra, except certain mathematical maturity and a will to
learn abstract thinking. Consequently, the book’s initial chapters are some-
what elementary, with the exposition proceeding at a leisurely pace, filling in
the details of proofs, particularly of basic results. To smoothen the approach,
we have devoted Part I to preliminaries consisting of two chapters, one on
sets, relations, function, partitions and the cardinality of a set and the other
on number systems, matrices and determinants. This part also serves as a
vehicle for introducing some of the notation and terminology concerning the
language of basic mathematics to be used in the later parts. Proofs of most of
the results in Part I are skipped and given as exercises to encourage interested
readers to work on them.
There are three parts in the main text of the book, Part II (Chapter 3–8),
Part III (Chapter 9–13) and Part IV (Chapters 14–16) covering Group
­Theory, Ring Theory and Field Theory respectively. Each chapter is divided
into a suitable number of sections in which definitions of the various con-
cepts are immediately followed by a sufficient number of examples and
­counter-examples. Worked exercises are included in each section in addition
to a set of exercises of varying levels of difficulty at the end of each section.
These exercises are an integral part of the book and require the reader’s
active participation. Some of them introduce a variety of ideas not treated
in the body of the text and impart certain additional information about con-
cepts discussed in chapters. We have given a brief introduction of vector
spaces and linear transformations to the extent necessary for a discussion
on Galois Theory. We have resisted the temptation to use Exercises, except

Q001-Algebra-111001_FM.indd 9 9/23/2011 2:17:41 PM

 x   Preface

those in Part I, to develop results that will be needed thereafter. As a result,

the reader does not need to work on all the exercises to assimilate the ideas
presented in the rest of the book. However, for the benefit of slow learners,
answers/hints for all even-numbered exercises have been provided.
When the publishers approached us with a proposal to take up the project
of writing a text book on Algebra, we considered various opinions on what
should be attempted within the framework of a first course in algebra. In
selecting textual material, we have followed, to a considerable extent, our
own interests, condensing or omitting altogether a number of topics that other
authors might have pursed more vigorously. The measure of success of our
efforts in writing this book is directly proportional to the number of readers
stimulated to expand their horizons in the realms of algebra. Comments and
suggestions for the improvement of the quality of the book are most welcome
and will be acknowledged in later editions. We may be excused for any pos-
sible typos.
We profusely thank all persons who directly or indirectly helped us in
bringing out this book. We are grateful to the people at Pearson Education,
to Mr. King D Charles Fenny in particular, for their encouragement and help
in completing this project. A special word of appreciation and thanks goes to
my wife, Lakshmi, who patiently helped me in the early morning hours on the
days when I was writing this book.
U. M. Swamy
A. V. S. N. Murty

Q001-Algebra-111001_FM.indd 10 9/23/2011 2:17:41 PM

 Part I
Preliminaries

Q001-Algebra-111001_CH 01.indd 1 9/16/2011 10:09:00 AM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 01.indd 2 9/16/2011 10:09:00 AM

 1
Sets and Relations
1.1 Sets and Subsets
1.2 Relations and Functions
1.3 Equivalence Relations and Partitions
1.4 The Cardinality of a Set

The concept of a set was used even by the ancient mankind without having
an exact idea of what it was. In modern mathematics, the notion of a set is
most basic. In fact, almost all the mathematical systems are certain collec-
tion of sets and their theories can be categorised as parts of set theory. We do
not intend to discuss axiomatic development of set theory. But, any person
with an intention of starting to learn the present day algebra must necessarily
possess certain elementary knowledge of set theory. This chapter provides a
fairly good platform to refresh with those elementary notions of sets, rela-
tions, functions and the cardinality of a set.

1.1 Sets and subsets

A set is usually defined as a well-defined collection of objects, in the sense
that, given any object we must be in a position to decide whether the object
belongs to the collection or not. First, let us take up two examples.

Example 1.1.1. Let us call a positive integer, a prime number if it has exactly
two positive divisors, namely 1 and itself. Clearly, 1 is not a prime number,
since 1 has only one positive divisor. Let C be the collection of all prime
numbers. We shall argue that C is a well-defined collection of objects. Let a
be any object. If a is not a positive integer, then we can immediately say that
a does not belong to the collection C. Suppose that a is a positive integer,
we can evaluate all the positive divisors of a and see whether these are two in

Q001-Algebra-111001_CH 01.indd 3 9/16/2011 10:09:00 AM

 1-4   Algebra – Abstract and Modern

number. For example, let a 5 123456789, on simple examination, we can say

9(9 1)
that 3 divides a (since the sum of the digits of a is  45 ) and 3  1
2
and 3  a and hence a is not a prime number, so that a does not belong to the
collection C. On the other hand, let b 5 123457687. It may be difficult for us
to decide whether b is prime or not. However, one thing is certain, it is either
a prime or not a prime, but not both. Therefore, C is a well-defined collection
of objects.

Example 1.1.2. Let C be the collection of all sets A satisfying the property
that A is not an object in A (or A does not belong to A). We shall argue that
C is not a well-defined collection. Suppose on the contrary that C is a well-
defined collection, that is, C is a set. Then, if C is an object in C, it follows
that C is not an object in C. On the other hand, if C is not an object in C,
then it follows that C is an object in C. Either way, it leads to a contradiction.
Therefore, we cannot decide whether C is an object in C. Therefore, C is not
a well-defined collection.

Definition 1.1.1. A well-defined collection of objects is called a set. If S is

a set, then the objects in S are called elements of S. We write a ∈ S and read
‘a belongs to S’, when a is an object in S. We write a ∉ S to say that a does
not belong to S.
Sets are usually denoted by uppercase letters, such as A, B, X, Y, etc. and
the elements of sets are denoted by lowercase letters, such as a, b, x, y, etc.

Example 1.1.3
1. The collection of all intelligent persons in India is not a set, since, if we
select a person from India, we cannot say with certainty whether he/
she belongs to the collection or not, as there is no standard scale for the
evaluation of intelligence.
2. For a similar reason, as detailed above, the collection of all tall persons
in India is not a set.
3. The collection of all prime numbers is a set, as discussed in Example
1.1.1.
4. The collection of all positive integers, which are not prime, is a set.
In this book, it is convenient to represent a set with the help of certain
property or properties satisfied only by the elements of the set. In order to
represent a set by this method, we write between the brackets { } a variable
x which stands for each of the set followed by the property or properties

Q001-Algebra-111001_CH 01.indd 4 9/16/2011 10:09:00 AM

 Sets and Relations   1-5

p­ ossessed by each element of the set and these two are separated by a symbol
‘:’ or ‘|’, read as ‘such that’. Therefore, we write

{x : p(x)} or {x | p(x)}

to represent the set of all objects x that satisfy the statement p(x). For example,
the set of all prime numbers is represented by

{x : x is a prime number}.

The set of all positive odd integers is represented by

{x : x is a positive integer and x is odd}

which is same as the set {1, 3, 5, 7, …}.

Definition 1.1.2. A collection having no objects is clearly a set and is called

the empty set or null set and is denoted by the symbol .

Example 1.1.4. The set {x : x is an even integer and 2 , x , 4} is the empty

set, since there is no even integer x, such that 2 , x , 4. Similarly,

{x : x is an integer and x2 1 2 5 0}

is the empty set.

Notation 1.1.1. The implication symbol ⇒ will be read as ‘implies’. If P and

Q are statements, then P ⇒ Q stands for the statement ‘the truth of P implies
the truth of Q’ or simply ‘P implies Q’. The symbol ⇔ is read as ‘implies and
implied by’. For any statements P and Q, P ⇔ Q stands for ‘P implies and
implied by Q’ or ‘P if and only if Q’.

Example 1.1.5
1. Let P be the statement, ‘x is an integer and x2 5 0’ and Q be the state-
ment, ‘x 5 0’. Then, we have P ⇔ Q since, for any integer x, x2 5 0 if
and only if x 5 0.
2. Let P be the statement, ‘x is a real number and x2 5 x’ and Q be the state-
ment, ‘x 5 0 or x 5 1’. Then, P ⇔ Q since, for any real number x, x2 5
x if and only x 5 0 or x 5 1.

Definition 1.1.3. Let A and B be two sets. Then, we say that

1. A is equal to B and express this by A 5 B if, for any object x,
x ∈ A ⇔ x ∈ B.

Q001-Algebra-111001_CH 01.indd 5 9/16/2011 10:09:00 AM

 1-6   Algebra – Abstract and Modern

2. A is a subset of B (or A is contained in B) and express this by A ⊆ B if,

for any object x,
x ∈ A ⇒ x ∈ B.
For any two sets A and B, clearly A 5 B if and only if A ⊆ B and B ⊆ A.
Whenever we are required to prove that two given sets A and B are equal, we
usually prove that A ⊆ B and B ⊆ A, that is, for any object x,

x∈A⇒x∈B and x ∈ B ⇒ x ∈ A.

Sometimes, we say that a set A is smaller than B (or B is larger than A) if A is

a subset of B. A is said to be a proper subset of B and write A ⊂ B if A ⊆ B and
A  B. Also, instead of writing A ⊆ B or A ⊂ B, some times we write B ⊇ A
or B ⊃ A and say that B contains A (or B is a superset of A) or B properly
contains A, respectively. We write A  B if A is not a subset of B.

Definition 1.1.4. For any set S, the collection of all subsets of S is again a set
and is called the power set of S and is denoted by P(S).
Note that the power set P(S) of any set S is always nonempty, since the
empty set  is a subset of every set S. In fact, if S is the empty set , then

P() 5 {},

a set consisting of only one element. It can be easily proved that, for any non-
negative integer n, a set S has exactly n elements if and only if the power set
P(S) has exactly 2n elements.

Definition 1.1.5. A set whose element are sets is called a class of sets or
f amily of sets.
Class of sets will be usually denoted by script letters, such as !, @, #, etc.
For any set S, the power set P(S) is a class of sets. A class # of sets is called
an indexed class if there exists a set I such that, for each i ∈ I, there is a unique
member Ai in # associated with i and the class # is equal to the class of all Ai,
i ∈ I; in this case, we write

# 5 {Ai : i ∈ I} or # 5 {Ai}i∈I

and I is called the index set.

Example 1.1.6. For any positive integer n, let

 1 
An   x : x is a real number and 0 ≤ x ≤ .
 n 

Q001-Algebra-111001_CH 01.indd 6 9/16/2011 10:09:00 AM

 Sets and Relations   1-7

Then, {An }n∈Z is an indexed class of sets and the set Z1 of positive integers
is the index set.

Definition 1.1.6. For any indexed class of sets {Ai}i∈I , we define the set as

∩ A  {a : a ∈ A
i∈ I
i i for all i ∈ I }.

This set is called the set intersection of Ai’s, i ∈ I. In particular, if A1, A2, …, An
are sets, we define
n

∩ A  {a : a ∈ A
i 1
i i for i  1, 2, … , n}

and is also denoted by A1 ∩ A2 ∩ … ∩ An. For any sets A and B, we have

A ∩ B 5 {x : x ∈ A and x ∈ B}.

Two sets A and B are said to be disjoint if A ∩ B 5 , that is, there are no
common elements of A and B.

Definition 1.1.7. For any indexed class {Ai}i∈I of sets, we define the set as

∪ A  {a : a ∈ A
i ∈I
i i for some i ∈ I }.

This set is called the set union of Ai’s, i ∈ I. In particular, for any sets A1, A2,
…, An, we define
n

∪ A  {a : a ∈ A
iI
i i for some 1 ≤ i ≤ n}

and this is also denoted by A1 ∪ A2 ∪ … ∪ An. For any sets A and B, we have

A ∪ B 5 {x : x ∈ A or x ∈ B}.

Example 1.1.7. For any positive integer n, let

 1 
An   x : x is a real number and 0 ≤ x ≤ .
 n 

Then, we have An ⊃ An11 for any n and therefore

Q001-Algebra-111001_CH 01.indd 7 9/16/2011 10:09:01 AM

 1-8   Algebra – Abstract and Modern

[0, 1] 5 A1 ⊃ A2 ⊃ A3 ⊃ … ⊃ An ⊃ An11 ⊃ …

1
∩ +
An  {x : x is a real number and 0  x 
n
for all n ∈ Z + }  {0}
n∈Z

and ∪ An  A1  [0, 1].

n ∈ Z

The following theorems can be easily proved by straight-forward veri-

fications.

Theorem 1.1.1. The following holds good for any sets A, B and C.
1. A ∪ B ⊆ C ⇔ A ⊆ C and B ⊆ C
2. A ⊆ B ∩ C ⇔ A ⊆ B and A ⊆ C
3. A∩B⊆A⊆A∪B
4. A∪A5A5A∩A
5. A ∪ B 5 B ∪ A and A ∩ B 5 B ∩ A
6. (A ∪ B) ∪ C 5 A ∪ (B ∪ C) and (A ∩ B) ∩ C 5 A ∩ (B ∩ C)
7. A5A∩B⇔A⊆B⇔A∪B5B
8. A ∩ (A ∪ B) 5 A 5 A ∪ (A ∩ B)
9. A ∩ (B ∪ C) 5 (A ∩ B) ∪ (A ∩ C)
10. A ∪ (B ∩ C) 5 (A ∪ B) ∩ (A ∪ C)
11. A ∩ (i∪ A )  i∪ (A ∩ Ai ) for any indexed class {Ai}i∈I of sets.
∈I i ∈I
12. A ∪ ( ∩ Ai )  ∩ (A ∪ Ai ) for any indexed class {Ai}i∈I of sets.
i∈I i∈I
13. A ⊆ B ⇒ A ∩ C ⊆ B ∩ C and A ∪ C ⊆ B ∪ C
14. A ∩ B ⊆ A ∩ C and A ∪ B ⊆ A ∪ C ⇔ B ⊆ C

Definition 1.1.8. For any two sets A and B, the difference of A with B is
defined as

A 2 B 5 {x : x ∈ A and x ∉ B}.

Theorem 1.1.2 (De Morgan Laws). For any indexed class {Bi}i∈I of sets and
for any sets A, B and C, the following holds good.
1. A  ( i ∪ B )  i∩ (A  Bi )
∈I i ∈I
2. A  ( i ∩ B )  i∪ (A  Bi )
∈I i ∈I
3. B ⊆ C ⇒ A 2 C ⊆ A 2 B and B 2 A ⊆ C 2 A
4. ( ∪ Bi )  A  ∪ ( Bi  A)
i∈I i∈I

Q001-Algebra-111001_CH 01.indd 8 9/16/2011 10:09:02 AM

 Sets and Relations   1-9

5. ( i ∩ B )  A  i∩ ( Bi  A)
∈I i ∈I
6. (A ∪ B) 2 C 5 (A 2 C) ∪ (B 2 C)
7. (A ∩ B) 2 C 5 (A 2 C) ∩ (B 2 C)
8. A 2 (B ∪ C) 5 (A 2 B) ∩ (A 2 C)
9. A 2 (B ∩ C) 5 (A 2 B) ∪ (A 2 C)
10. (A 2 B) 2 C 5 A 2 (B ∪ C) 5 (A 2 C) 2 B
11. A 2 (B 2 C) 5 (A 2 B) ∪ (A ∩ C)
12. A∩B5⇔A⊆A2B⇔B⊆B2A
13. A25A
14. 2A5

Definition 1.1.9. For any sets A and B, the symmetric difference of A and B
is defined as

A ⊕ B 5 (A 2 B) ∪ (B 2 A).

That is, A ⊕ B 5 {x : x belongs to exactly one of A and B}.

Theorem 1.1.3. The following holds good for any sets A, B and C.
1. A ⊕ B 5 B ⊕ A
2. (A ⊕ B) ⊕ C 5 A ⊕ (B 1 C)
5 (A ∩ B ∩ C) ∪ ((A 2 B)2C) ∪ ((B 2 C)2A)
∪ ((C 2 A)2B)
3. A ⊕  5 A
4. A ⊕ A 5 

Theorem 1.1.4. For any sets A, B and C,

A ∩ (B ⊕ C) 5 (A ∩ B) ⊕ (A ∩ C).

EXERCISE 1(a)

1. Express each of the following sets in the form {x : P(x)} and specify the prop-
erty P(x).
(i) The set of all rational numbers, whose denominators are not divisible by 5.
(ii) The set of all integer multiples of 5 in between 296 and 96.
(iii) The set of all points in the three-dimensional Euclidean space, whose
distance from (0, 0) is a rational number.

Q001-Algebra-111001_CH 01.indd 9 9/16/2011 10:09:02 AM

 1-10   Algebra – Abstract and Modern

(iv) The set of all pairs of real numbers, whose sum of their squares is nonzero.
(v) The set of all even primes.
(vi) The set of all subsets of {1, 2, 3, 4} not containing 3.

2. Write explicitly all elements in each of the following sets

(i) {a : a is an integer and 0  a2  26}
(ii)  { rs : r and s are nonzero integers and 1  rs  1}
(iii) {A : A ⊆ {a, b, c} and b ∉ A}
(iv) The set of all three-digit positive integers, whose all the digits are even
and are in strictly increasing order.
(v) The set of all pairs of integers, whose sum of the squares is zero.
(vi) The set of all integers, whose squares are in between 10 and 15.

3. Let A 5 {a ∈ Z1 : 3 divides a} and

B 5 {a ∈ Z1 : The sum of the digits in a is divisible by 3}.
Prove that A 5 B.
4. Describe P(X) if X 5 {1, 2, 3}.

5. Let X 5 {a ∈ R : 21  a  1} and
Y 5 {r ∈ R : r 5 sin t 2 cos t for some t ∈ R}.
Is X 5 Y ?
6. Let X 5 {1, 2, 3, …, 100}, A 5 {a ∈ X : a 5 b2, b ∈ Z},
B 5 {a ∈ X : a is odd} and for each 1  i  96,
Ci 5 {i, i 1 1, i 1 2, i 1 3, i 1 4}. Write explicitly all elements in each of the
following sets.
(i) A∩B
(ii) A ∪ B ∪ C2
96
(iii) ( ∪ Ci ) ∩ A
i1
25
(iv) B ∩ ( ∪ Ci )
i20
(v) X 2 (A ∪ B)
90
(vi) X  ( ∪ Ci )
i6
96
(vii) A  ( ∪ Ci )
i1
(viii) A2B
7. For any two sets A and B, prove that
A 5 A ∩ B ⇔ A ⊆ B ⇔ A ∪ B 5 B.

8. Prove Theorem 1.1.1.

Q001-Algebra-111001_CH 01.indd 10 9/16/2011 10:09:03 AM

 Sets and Relations   1-11

9. Prove Theorem 1.1.2.

10. Prove Theorem 1.1.3.

11. Prove Theorem 1.1.4.

12. Prove or disprove each of the following for any sets X and Y.
(i) P(X ∩ Y) 5 P(X) ∩ P(Y)
(ii) P(X ∪ Y) 5 P(X) ∪ P(Y)
(iii) P(X 2 Y) 5 P(X) 2 P(Y)
(iv) P(X) 5 P(Y) ⇔ X 5 Y

1.2 RELATIONS AND FUNCTIONS

Consider the set A of all points in the plane and the set B of all straight
lines in the plane. For any x ∈ A and L ∈ B, let us write
x R L if x lies on L (or L passes through x).
Then, R is a relation between the elements of A and the elements of B. Here,
x R L can be read as ‘x is related to L’ and R denotes the relation ‘lies on’. We
can also consider R as the set of ordered pairs (x, L) such that x lies on L. This
pair is ordered in the sense that x and L cannot be interchanged, because the first
component of the pair is a point of A and the second component is a point of
another set B and because the statement ‘L lies on x’ has no meaning. Therefore,
we can consider R as a set of ordered pairs (x, L) satisfying the property that x
lies on L. This concept is formalised in this section by introducing an abstract
concept of a relation and by discussing the various properties of relations.

Definition 1.2.1. A pair of elements (not necessarily in the same set) writ-
ten in a particular order is called an ordered pair and is written by listing its
elements in a particular order, separated by a comma, and enclosing the pair
in brackets. In the ordered pair (x, L), x is called the first component (or first
coordinate) and L is called the second component (or second coordinate).
The ordered pairs (x, L) and (L, x) are different even though they consist of
the same pair of elements. For example, the pairs (2, 5) and (5, 2) represent
two different points in the plane.

Definition 1.2.2. Let A and B be any two sets. Then, the set of all ordered
pairs (a, b) with a ∈ A and b ∈ B is called the Cartesian product of A and B
and is denoted by A 3 B. That is,

A 3 B 5 {(a, b) : a ∈ A and b ∈ B}.

Q001-Algebra-111001_CH 01.indd 11 9/16/2011 10:09:03 AM

 1-12   Algebra – Abstract and Modern

Example 1.2.1. If A 5 {1, 2} and B 5 {a, b, c}, then

A 3 B 5 {(1, a), (1, b), (1, c), (2, a), (2, b), (2, c)}
A 3 A 5 {(1, 1), (1, 2), (2, 1), (2, 2)}
B 3 B 5 {(a, a), (a, b), (a, c), (b, a), (b, b), (b, c), (c, a), (c, b), (c, c)}.

Note 1.2.1. For any sets A and B,

A 3 B 5  ⇔ A 5  or B 5 
A3B5B3A⇔A5B

Definition 1.2.3. For any sets A1, A2, …, An, we define the Cartesian product
of A1, A2, …, An as the set

A1 3 A2 3…3 An 5 {(a1, a2, …, an) : ai ∈ Ai for all 1  i  n}.

In particular, for any set A and for any positive integer n, we define

An 5 {(a1, a2, …, an) : ai ∈ A for all 1  i  n}.

Definition 1.2.4. Let A and B be any sets. Then, any subset of A 3 B is called
a relation from A to B. For any relation R from A to B (that is, R ⊆ A 3 B), if
(a, b) ∈ R, then we say that ‘a is R-related to b’ or ‘a is related to b with respect
to R’ or ‘a and b have relation with R’ and is usually denoted by a R b.

Definition 1.2.5. Let A be any nonempty set. A relation from A to itself is

called a ‘binary relation on A’.

Example 1.2.2. Let Z be the set of all integers and n a positive integer.
Define

R 5 {(a, b) ∈ Z 3 Z : n divides a 2 b}
S 5 {(a, b) ∈ Z 3 Z : a 5 nb}.

Then, both R and S are binary relations on Z.

Definition 1.2.6. Let A, B, and C be sets, R be a relation from A to B and S be

a relation from B to C, that is, R ⊆ A 3 B and S ⊆ B 3 C. Define

S o R 5 {(a, c) ∈ A 3 C : a R b and b S c for some b ∈ B}.

In other words, for any a ∈ A and c ∈ C,

(a, c) ∈ S o R ⇔ There exists b ∈ B such that (a, b) ∈ R and (b, c) ∈ S.

Q001-Algebra-111001_CH 01.indd 12 9/16/2011 10:09:03 AM

 Sets and Relations   1-13

Note that, when S o R is defined, R o S may not be defined. Even when

S o R and R o S are both defined, they may not be equal. S o R is called the
­composition of R with S.

Definition 1.2.7. For any relation R from a set A to a set B, the inverse of R
is defined by

R21 5 {(b, a) : (a, b) ∈ R}.

Note that R21 is a relation from B to A, R o R21 is a binary relation on B and

R21 o R is a binary relation on A. It can be easily verified that (R−1)−1 5 R and
(S o R)−1 5 R−1 o S−1 for any relation R from A to B and any relation S from
B to C.

Definition 1.2.8. A relation R from a set A to a set B is called a function (or

a mapping) of A into B, if for each a ∈ A, there exists unique b ∈ B such
that (a, b) ∈ R. Usually, functions are denoted by lowercase letters f, g, h,
etc. If f is a function of A into B, then f ⊆ A 3 B satisfying the following
conditions:
(i) For each a ∈ A, there exists b ∈ B such that (a, b) ∈ f .
(ii) If (a, b) ∈ f and (a, b1) ∈ f, then b 5 b1.
If f is a function of A into B and (a, b) ∈ f, then we write a f b or (a)f 5 b
or f (a) 5 b. More popular convention is writing f (a) 5 b. This is reasonable,
since b corresponds to a uniquely. In this case, b is called ‘the image of a
under f ’ and a is called ‘a pre-image of b under f ’. We write simply f : A →
B, to denote that f is a function from A into B. If f : A → B, any element a of
A will have exactly one image f (a) in B, while an element b of B may have
any number of pre-images in A or may not have any pre-image at all. These
circumstances lead to the following.

Definition 1.2.9. Let f : A → B be a function.

1. f is said to be an injection (or a one-one f unction) if each element of B
has at most one pre-image in A; or, equivalently, for any a1, a2 ∈ A,
f (a1) 5 f (a2) ⇒ a1 5 a2.

2. f is said to be a surjection (or an onto f unction) if each element of B has

atleast one pre-image in A; or, equivalently,
b ∈ B ⇒ f (a) 5 b for some a ∈ A.

Q001-Algebra-111001_CH 01.indd 13 9/16/2011 10:09:04 AM

 1-14   Algebra – Abstract and Modern

3. f is said to be a bijection (or a one-to-one f unction) if each element of B

has exactly one pre-image in A, or, equivalently, f is both an injection as
well as a surjection.
Note that, to describe a function f : A → B, it is enough if we prescribe the
image f (a) of each element a in A.

Example 1.2.3. Let Z be the set of all integers.

1. Define f : Z → Z by f (a) 5 a2 for any a ∈ Z. Then, f is a function of
Z into itself and f is neither an injection (since f (−1) 5 1 5 f (1)) nor a
surjection (since there is no pre-image for −1).
2. Define g : Z → Z by g(a) 5 4a for any a ∈ Z. Then, g is an injection, but
not a surjection.
3. Define h : Z → Z by h(a) 5 a 1 2 for any a ∈ Z. Then, h is a bijection.
4. Let N be the set of all nonnegative integers and define m : Z → N by m(a)
5 |a| for any a ∈ Z, where |a| 5 a or –a depending upon a is positive or
not. Then, m is a surjection, but not an injection.
Let f : A → B be a function. Then, A is called ‘the domain of f 9 and is
denoted by Dom(f ) and B is called ‘the codomain of f 9 and is denoted by
Codom(f ). The set {f (a) : a ∈ A} is called ‘the image of f 9 and is denoted by
Im(f ). Note that Im(f ) is a subset of the codomain B and is not necessarily
equal to B. In fact, f is a surjection if and only if Im(f ) 5 Codom(f ).

Definition 1.2.10. Let f : A → B and g : B → C be functions. Then, the com-

position g o f is also a function from A to C. Recall from Definition 1.2.6 that
g o f is defined by

g o f 5 {(a, c) ∈ A 3 C : (a, b) ∈ f and (b, c) ∈ g for some b ∈ B}

5 {(a, c) ∈ A 3 C : f (a) 5 b and g(b) 5 c}
5 {(a, c) ∈ A 3 C : g(f (a)) 5 c}.

Therefore, g o f : A → C is a function defined by

(g o f )(a) 5 g(f (a)) for any a ∈ A.

Note that g o f is defined only when Codom(f ) 5 Dom(g) or Im(f ) ⊆ Dom(g).

In fact, we have

Dom(g o f ) 5 Dom(f )
and          Codom(g o f ) 5 Codom(g).

Q001-Algebra-111001_CH 01.indd 14 9/16/2011 10:09:04 AM

 Sets and Relations   1-15

Two functions f and g are said to be equal if their domains are equal and f (a) 5
g(a) for all the elements a in the common domain. For two functions f and g,
both f o g and g o f may be defined but still they may not be equal, consider
the following example.

Example 1.2.4. Define f : Z → Z and g : Z → Z by f (a) 5 a 1 1 and g(a) 5

a2 for any a ∈ Z.
Then, (f o g)(a) 5 f (g(a)) 5 f (a2) 5 a2 1 1
and (g o f )(a) 5 g(f (a)) 5 g(a 1 1) 5 (a 1 1)2 5 a2 1 2a 1 1.
Therefore, (f o g)(a)  (g o f )(a) for 0  a ∈ Z and hence f o g and g o f
are not equal.
Note that, if f and g are injections (surjections, bijections), then so is f o g.
Further, we have (f o g) o h 5 f o (g o h) for any functions f, g and h, when-
ever the compositions are defined.

Definition 1.2.11. Let A be any nonempty set and define a function IA : A →

A by IA(a) 5 a for all a ∈ A. Then, IA is called the ‘identity function on A’. IA
will also be denoted by IdA or Id on X.
For any function f : A → B, it can be seen that

f o IA 5 f 5 IB o f.

In the following, we give certain characterisation properties for injections,

surjections and bijections.

Theorem 1.2.1. Let f : A → B be a function, then

1. f is an injection if and only if there exists a function g : B → A such that
g o f 5 IA.
2. f is a surjection if and only if there exists a function h : B → A such that
f o h 5 IB.

Theorem 1.2.2. A function f : A → B is a bijection if and only if there exists

a function g : B → A such that
f o g 5 IB and g o f 5 IA.

In this case, g is unique and is called the inverse of f and is denoted by f −1.
Note that, for any a ∈ A and b ∈ B,

f (a) 5 b ⇔ a 5 f −1(b)

and that f −1 is also a bijection.

Q001-Algebra-111001_CH 01.indd 15 9/16/2011 10:09:04 AM

 1-16   Algebra – Abstract and Modern

If f : A → B is a surjection, then usually we say that f is a surjection of A onto

B, instead of A into B, just to mention that f is an onto function (or surjection).
Definition 1.2.3. Let A and B be two sets. A is said to be equivalent or
­equipotent with B if there exists a bijection of A onto B; in this case, we
denote it by A . B.
If f : A → B is a bijection, then f −1 : B → A is also a bijection and therefore
we have
A . B ⇔ B . A.
Also, since the identity function IA : A → A is a bijection, we have
A . A for any set A.
Further, if f : A → B and g : B → C are bijections, then g o f : A → C is also
a bijection and therefore
A . B and B . C ⇒ A . C.
Example 1.2.5. Let E be the set of all even integers and Z be the set of all inte-
gers. Then, E 5 {2a : a ∈ Z} and E is equivalent to Z; for, define f : E → Z by
2b if a  4b
f (a )  
b if a  2b and b is odd.

Then, f is a bijection. Therefore E . Z.

Definition 1.2.12. Let f : X → Y be a function, A ⊆ X and B ⊆ Y. The image
of A under f is defined as the set
f (A) 5 {f (a) : a ∈ A}.
The inverse image of B under f is defined as the set
f −1(B) 5 {x ∈ X : f (x) ∈ B}.
Then, clearly f (A) is a subset of Y for all A ⊆ X and f −1(B) is a subset of X for
all B ⊆ Y. In other words, f induces a function from the power set P(X) into
the power set P(Y) and another function from P(Y) into P(X). In this context,
we have the following.
Theorem 1.2.3. The following holds good for any function f : X → Y and
subsets A1 and A2 of X and B1 and B2 of Y.
1. f () 5  and f (X) 5 Im(f )
2. f −1() 5  and f −1(Y) 5 X
3. f (A1 ∪ A2) 5 f (A1) ∪ f (A2)

Q001-Algebra-111001_CH 01.indd 16 9/16/2011 10:09:04 AM

 Sets and Relations   1-17

4. f −1(B1 ∪ B2) 5 f −1(B1) ∪ f −1(B2)

5. f −1(B1 ∩ B2) 5 f −1(B1) ∩ f −1(B2)
6. f (A1 ∩ A2) ⊆ f (A1) ∩ f (A2)
7. A ⊆ f −1(f (A)) for any A ⊆ X
8. f (f −1(B)) ⊆ B for any B ⊆ Y.
Notice that there are only one-side inclusions in (6), (7) and (8). In gen-
eral, these one-side inclusions cannot be replaced by the equality in these. In
this context, we have the following.

Theorem 1.2.4. The following are equivalent to each other for any function
f:X→Y
1. f is an injection.
2. A 5 f −1(f (A)) for any A ⊆ X.
3. f (A1 ∩ A2) 5 f (A1) ∩ f (A2).

Theorem 1.2.5. A function f : X → Y is a surjection if and only if

f (f −1(B)) 5 B for any subset B of Y.

Definition 1.2.13. Let f : X → Y be a function and Z ⊆ X. Then, (Z 3 Y) ∩ f

is a function of Z into Y and is called the restriction of f to Z and is denoted
by f |Z. Note that

(f |Z)(a) 5 f (a) for any a ∈ Z.

EXERCISE 1(b)

1. Prove each of the following for any sets A, B and C.

(i) (A ∪ B) 3 C 5 (A 3 C) ∪ (B 3 C)
(ii) (A ∩ B) 3 C 5 (A 3 C) ∩ (B 3 C)
(iii) (A 2 B) 3 C 5 (A 3 C) 2 (B 3 C)
(iv) A 3 B 5 A 3 C ⇒ A 5  or B 5 C

2. In each of the following cases, find sets A, B, C and D to disprove the

statement.
(i) A5B⇔A2C5B2C
(ii) A5B⇔A∩C5B∩C
(iii) A5B⇔A∪C5B∪C
(iv) (A 2 B) 3 (C 2 D) 5 (A 3 C) 2 (B 3 D)
(v) (A ∪ B) 3 (C ∪ D) 5 (A 3 C) ∪ (B 3 D)

Q001-Algebra-111001_CH 01.indd 17 9/16/2011 10:09:05 AM

 1-18   Algebra – Abstract and Modern

3. Prove that (A ∩ B) 3 (C ∩ D) 5 (A 3 C) ∩ (B 3 D) for any sets A, B, C and D.

4. If A has n elements and B has m elements, then prove that A 3 B has nm ­elements.
Determine the number of relations from A to B.
5. State whether each of the following is a function and substantiate your answers.
(i) R 5 {(a, b) ∈ Z 3 Z : b2 5 a}
(ii) R 5 {(a, b) ∈ Z 3 Z : a2 5 b}
(iii) R 5 {(a, b) ∈ R 3 R : 2a2 2 1 5 b}
(iv) R 5 {(a, b) ∈ R 3 R : 2b2 2 1 5 a}
(v) R 5 {(a, b) ∈ Q 3 Q : a2 1 b2 is an integer}
(vi) R 5 {(a, b) ∈ Z 3 Q : 2b 5 a}.
6. If A has n elements and B has m elements, then determine the number of func-
tions from A into B.
7. Prove that, for any relation R from A to B and any relation S from B to C, R21 o
S21 5 (S o R)21.
8. Prove that a function f : X → Y is an injection if and only if f |Z is an injection for
every subset Z of X. Is this true if we replace injection with surjection?

9. For any function f : X → Y and A ⊆ B ⊆ X, prove that (f |B)|A 5 f |A.

10. Let  5 A ⊆ X and f : A → Y be a function. Does there exist a function g : X →

Y such that g|A 5 f ? If yes, how many such functions g can be found?
11. Let f : A → B and g : B → C be the functions. If f and g are bijections, prove that
g o f is a bijection. Is the converse true? Substantiate your answer.
12. If f : R → R is defined by f (a) 5 a2 for all a ∈ R, then determine f 21 [22, 8),
f 21(2, 0], f 21(21, 1) and f 21(Z).
13. For any real number a, let [a] be the largest integer less than or equal to a and
define f : R → Z by f (a) 5 [a] for any a ∈ R.
Then, determine the following sets
(i) f ((21, 1))
(ii) f ([21, 1])
(iii) f 21({0, 1})
(iv) f 21(E), where E is the set of even integers.
 2 2
(v) f 1  , 
 3 3
(vi) f 21({0})
14. Prove Theorem 1.2.1.
15. Prove Theorem 1.2.2.

Q001-Algebra-111001_CH 01.indd 18 9/16/2011 10:09:05 AM

 Sets and Relations   1-19

16. Prove Theorem 1.2.3.

17. Give an example of a function f : X → Y and a subset A of X for which A is prop-
erly contained in f 21(f (A)).
18. Give an example of a function f : X → Y and a subset B of Y such that f (f 21(B))
is properly contained in B.
19. If X is an n-element set and Y is an m-element set, how many bijections can be
there from X onto Y?

20. Let f : X → Y be a function,   A ⊆ X and   B ⊆ Y. Then, prove the

following.
(i) f (f 21(f (A))) 5 f (A)
(ii) f 21(f (f 21(B))) 5 f 21(B)
(iii) (f |A)21 (B) 5 f 21(B) ∩ A.
21. Let f : X → Y and g : Y → X be mappings, such that g o f 5 IA. Prove that the
following are equivalent to each other.
(i) f is a surjection.
(ii) g is an injection.
(iii) f is a bijection.
(iv) g is a bijection.
and that, in this case, f o g 5 IB.
22. Let f : X → Y and g : Y → Z be mappings, such that g o f is an injection. Then,
prove that f is an injection and that, when f is a surjection also, then g is an
injection.
23. Let #[0, 1] be the set of all real-valued continuous functions defined on the
closed interval [0, 1] and #9[0, 1] be the set of all differentiable functions f in
#[0, 1], such that f (0) 5 0 and the derivative f 9 is continuous. Prove that the
function
D : #9[0, 1] → #[0, 1] defined by D(f ) 5 f 9

is a bijection.
24. Let A be an n-element set and B be an m-element set. Find the number of injec-
tions of A into B in each of the following cases.
(i) n 5 m, (ii) n . m, (iii) n , m.

25. Define f : Z1 → Z1 by f (a) 5 2a 2 1. Prove that there exist infinitely many func-
tions g : Z1 → Z1, such that g o f 5 IZ1 and there is no function h : Z1 → Z1
such that f o g 5 IZ1.
26. Prove Theorem 1.2.4.
27. Prove Theorem 1.2.5.

Q001-Algebra-111001_CH 01.indd 19 9/16/2011 10:09:05 AM

 1-20   Algebra – Abstract and Modern

28. Let f 1, f 2, …, f n be bijections, such that f 1 o f 2 o … o f n is defined. Then, prove

that f n1 ο f n11 ο  ο f11 is defined and is equal to (f 1 o f 2 o … o f n)21.
29. Let f : X → Y and g : Y → X be functions, such that g o f is an injection and f o g
is a surjection. Then, prove that both g o f and f o g are bijections.
30. Let f : X → Y be a function. If g : Y → X is a function, such that g o f 5 IX (f o g
5 IY), then g is called a left (respectively, right) inverse of f . Prove that the fol-
lowing are equivalent to each other.
(i) f has a unique left inverse.
(ii) f is a bijection.
(iii) f has a unique right inverse.
31. Let n and m be positive integers greater than 1, such that n and m have no com-
mon factors except 1. Let In be the set {1, 2, …, n}. Prove that there is a bijection
f : In 1 m → In 1 m such that f (n 1 m) 5 n 1 m and f (i 1 1) 2 f (i) ∈ {n, 2m} for
all 1 ≤ i , m 1 n.
32. Let X be a nonempty set. Prove that f o g 5 g o f for all bijections f and g of X
onto itself if and only if X has almost two elements in X.
33. Let f : X → Y be a function and F : P(Y) → P(X) be defined by F (A) 5 f 21(A) for
all A ⊆ Y. Then, prove the following.
(i) f is injection if and only if F is surjection.
(ii) f is surjection if and only if F is injection.
(iii) f is a bijection if and only if F is a bijection.
34. Let X be a set and define f : P(P(X)) → P(X) by
f ( )  ∪ A for any  ⊆ P( X ).
A ∈

Then, find two distinct right inverses of f .

35. Define f : Z1 3 Z1 → Z1 by f (a, b) 5 2a21(2b21). Then, prove that f is a bijec-
tion and find f 21.
36. Let X and Y be nonempty sets and YX be the set of all functions of X into Y. Prove
the following for any   A ⊆ X.
(i) The function h : YX → YA, defined by h(f ) 5 f /A, is a surjection.
(ii) If Y has atleast two elements, then h is a bijection ⇔ A 5 X.
37. Let X be any set and 2 5 {0, 1}. Prove that the map x : P(X) → 2X, defined by
1 if x ∈ A
χ( A)( x )  
0 if x ∉ A

for any A ∈ P(X) and x ∈ X, is a bijection.
38. Deduce from the above that if A is an n-element set, then P(A) is a 2n-element set.

Q001-Algebra-111001_CH 01.indd 20 9/16/2011 10:09:05 AM

 Sets and Relations   1-21

1.3 EQUIVALENCE RELATIONS AND PARTITIONS

Dividing a set into disjoint subsets is called a partitioning of the set. In this
section, we discuss a special type of binary relations on a set which induce
partitions of the set.
Definition 1.3.1. Let S be any nonempty set and R, binary relation on S.
1. R is said to be ref lexive on S if (a, a) ∈ R for all a ∈ S.
2. R is said to be symmetric if (a, b) ∈ R ⇒ (b, a) ∈ R.
3. R is said to be transitive if (a, b) ∈ R and (b, c) ∈ R ⇒ (a, c) ∈ R.
4. R is said to be an equivalence relation on S, if it is reflexive on S, sym-
metric and transitive.
Example 1.3.1
1. Let X be any nonempty set and
DX 5 {(x, x) : x ∈ X}.

Then DX is an equivalence relation on S and is called the diagonal on X.

DX can also be defined as
DX 5 {(x, y) ∈ X 3 X : x 5 y}.

2. For any set X, the whole of X 3 X is an equivalence relation on X.

3. For any positive integer n, let
Rn 5 {(a, b) ∈ Z 3 Z : n divides a – b}.

Then, Rn is an equivalence relation on Z.

4. Let R 5 {(a, b) ∈ Z 3 Z : a 5 0 5 b or ab . 0}.
Then, R is an equivalence relation on Z.
Definition 1.3.2. Let R be an equivalence relation on a set X and x ∈ X. The
R-equivalence class of x (or simply, the R-class of x) is defined to be the set
R(x) 5 {y ∈ X : (x, y) ∈ R}.
The following can be proved easily.
Theorem 1.3.1. Let R be an equivalence relation on a set X. Then, the ­following
holds good.
1. For any x and y ∈ X,
(x, y) ∈ R ⇔ R(x) 5 R(y)
and (x, y) ∉ R ⇔ R(x) ∩ R(y) 5 .

Q001-Algebra-111001_CH 01.indd 21 9/16/2011 10:09:06 AM

 1-22   Algebra – Abstract and Modern

2. Any two R-equivalence classes in X are either equal or disjoint.

3. ∪ R(x)  X .
x ∈X

Definition 1.3.3. Let X be a nonempty set. A class # of sets is said to be a

partition of X if the following conditions are satisfied:
1. Each member of # is a nonempty subset of X.
2. For any A, B ∈ #, either A 5 B or A ∩ B 5 .
3. The union of all the members of # is X.
In other words, a class # of nonempty subsets of a set X is called a partition
of X if each element in X is in exactly one member of #.
The following is an immediate consequence of Theorem 1.3.1.

Theorem 1.3.2. For any equivalence relation R on a set X, the class of all
R-equivalence classes in X is a partition of X and is denoted by X/R; that is,

X {R(x ) : x  X }.
R
X {R(x ) : x  X }.
R is called the partition on X induced by R or the quotient of X by R.
The converse of the above result is also true, in the sense that, for any
partition # of X, there exists an equivalence relation R# on X such that the
partition of X induced by R# is precisely equal to the given partition #.

Theorem 1.3.3. Let # be a partition of a nonempty set X. Define

R 5 {(x, y) ∈ X 3 X : both x and y belong to the same member of #}.
Then, R is an equivalence relation on X and X/R 5 #. (In fact, if x ∈ A ∈ #,
then A 5 R(x).)
These two processes R a X/R and # a R# are inverses to each other in the
sense that
R a X/R a RX/R 5 R and # a R# a X/R# 5 #
for any equivalence relation R on X and for any partition # of X. Therefore,
we have the following.
Theorem 1.3.4. For any nonempty set X, let j(X) be the set of all equivalence
relations on X and Part(X) be the set of all partitions of X. Then,

j(X) . Part(X),

that is, there is a bijection of j(X) onto Part(X).

Example 1.3.2. Consider the relation R given in Example 1.3.1 (4), we have
R 5 {(a, b) ∈ Z 3 Z : a 5 0 5 b or ab . 0}.

Q001-Algebra-111001_CH 01.indd 22 9/16/2011 10:09:06 AM

 Sets and Relations   1-23

For any a ∈ Z, R(a) 5 {0} or Z1 or Z2 according as a 5 0 or a . 0 or a , 0,

respectively, where Z2 stands for the set of all negative integers. Therefore,
X
R 5 {{0}, Z , Z }.
{R(x ) : x 1 X2}.

Example 1.3.3. Let n be a positive integer and Rn be the equivalence relation

on Z defined in Example 1.3.1 (3); that is,

Rn 5 {(a, b) ∈ Z 3 Z : n divides a – b}.

Then, for any a ∈ Z, the R-class of a is given by

Rn(a) 5 {a 1 nx : x ∈ Z}.

If a 5 qn 1 r, q, r ∈ Z and 0 # r , n, then Rn(a) 5 Rn(r) and hence Rn(0),

Rn(1), …, Rn(n – 1) are all the distinct Rn-classes in Z. That is, there are
exactly n Rn-classes in Z.

Definition 1.3.4. Let #1 and #2 be two partitions on a set X. Then, #2 is said

to be a ref inement of #1 if every member of #2 is a union of members of #1.

Theorem 1.3.5. Let R and S be two equivalence relations on a set X and X R {R(x ) : x  X }
and X S be partitions corresponding to R and S, respectively. Then, R ⊆ S if
and only if X S is a refinement of X R.{R(x ) : x  X }.

Proof: Suppose that R ⊆ S, then, for any x, y ∈ X,

(x, y) ∈ R ⇒ (x, y) ∈ S
and hence R(x) ⊆ S(x) for all x ∈ X. It can be seen that S (x )  ∪ R( y ).
y ∈ S( x)

Therefore, X S is a refinement of X R.{ R(x ) : x  suppose

Conversely X }. that X S is a
refinement of X R.{LetR((x,
x ) :y)x ∈ Then, S(x) is a member of X S and hence
X }.
 R.
S(x) is a union of members of R.{ X R(x ) : x  X }.
Therefore,

S (x)  ∪ R( z ) for some Z ⊆ X.

z ∈Z

Now, since x ∈ S(x), we get that x ∈ R(z) for some z ∈ Z and hence (x, z) ∈ R.
Since (x, y) ∈ R also, we have that (y, z) ∈ R so that y ∈ R(z) ⊆ S(x). Therefore,
(x, y) ∈ S. Thus, R ⊆ S. b
The following theorem is a simple verification.

Theorem 1.3.6. The intersection of any class of equivalence relations on

a set X is again an equivalence relation on X. In particular, if R and S are

Q001-Algebra-111001_CH 01.indd 23 9/16/2011 10:09:07 AM

 1-24   Algebra – Abstract and Modern

e­ quivalence relations on X and X R {andRX(x ) : are

S
X }.corresponding partitions
x the
of X, then R ∩ S is also an equivalence relation whose corresponding partition
is {R(x) ∩ S(x) | x ∈ X}. In other words,

(R ∩ S)(x) 5 R(x) ∩ S(x) for any x ∈ X.

These can be better understood by the following figures showing partitions

of R, S and R ∩ S.

X
R

X
S

X
R ∩S

Q001-Algebra-111001_CH 01.indd 24 9/16/2011 10:09:08 AM

 Sets and Relations   1-25

Recall that a binary relation R on a set X is an equivalence relation on X if

and only if R 5 R21, R o R ⊆ R and DX ⊆ R, where DX is the diagonal of X; in
fact, in this case, R o R 5 R. In general, for any equivalence relations R and S
on a set X, the composition R o S may not be an equivalence relation. In this
context, we have the following.

Theorem 1.3.7. Let R and S be equivalence relations on a set X. Then, the

following are equivalent to each other:

1. R o S is an equivalence relation on X.
2. R o S is symmetric.
3. R o S is transitive.
4. RoS⊆SoR
5. SoR⊆RoS
6. RoS5SoR
7. S o R is symmetric.
8. S o R is transitive.

Theorem 1.3.8 (Fundamental theorem of functions). Any function f : X → Y

can be expressed as

f5goh

for some injection g and some surjection h.

Proof: Let f : X → Y be a function. Define

R 5 {(a, b) ∈ X 3 X : f (a) 5 f (b)}.

Then, R is an equivalence relation on X. Consider the partition X/R 5 {R(x) :

x ∈ X} and define

h : X → X/R by h(x) 5 R(x) for any x ∈ X.

Also, define
g : X/R → Y by g(R(x)) 5 f (x). If R(x) 5 R(x9), then (x, x9) ∈ R and hence
f (x) 5 f (x9). Therefore, g is a well-defined function and clearly g is an
injection. Also, it is clear that g(h(x)) 5 f (x) for all x ∈ X. Thus, f 5 g o h,
g is an injection and h is a surjection. b

Q001-Algebra-111001_CH 01.indd 25 9/16/2011 10:09:08 AM

 1-26   Algebra – Abstract and Modern

EXERCISE 1(c)

1. Which of the following are equivalence relations?

(i) {(a, b) ∈ R 3 R : a – b is a rational number}
(ii) {(a, b) ∈ Q 3 Q : a – b is an integer}
(iii) Let Z* 5 Z – {0} and
R 5 {(a, b) ∈ Z* 3 Z*: b 5 2na for some n ∈ Z}
(iv) {(a, b), (c, d) ∈ R2 3 R2 : a2 1 b2 5 c2 1 d2}
(v) {(a, b) ∈ Z 3 Z : a 5 nb for some n ∈ Z}
(vi) {(a, b) ∈ Z1 3 Z1 : n divides both a and b for some 1 , n ∈ Z}
(vii) {(a, b) ∈ R 3 R : ab is a rational number}
(vii) Let M(R) be the set of all mappings of R into R and
R 5 {(f , g) ∈ M(R) 3 M(R) : f (a) 5 g(a) for some a ∈ R}
(ix) For any set X,
{(A, B) ∈ P(X) 3 P(X) : A ⊕ B is finite}
(x) Let R* 5 R – {0}: and
R 5 {(a, b) ∈ R* 3 R* : 0 , ab21 ∈ Q}.
2. Give three examples of binary relations showing that a relation can satisfy any
one of reflexivity, symmetricity and transitivity without satisfying the other two.
3. Prove that reflexivity, symmetricity and transitivity of a relation are independent
in the sense that no two of them imply the other.
4. If R is an equivalence relation on a set X and f  Y ⊆ X, then prove that
R ∩ (Y 3 Y) is an equivalence relation on R.
5. Let X 5 Z 3 (Z – {0}) and
R 5 {((a, b), (c, d)) ∈ X 3 X: ad 5 bc}.
Prove that R is an equivalence relation on X.
6. Let X 5 Z1 3 Z1 and R 5 {((a, b), (c, d)) ∈ X 3 X : a 1 d 5 b 1 c}.
Prove that R is an equivalence relation on X.
7. Describe the partitions corresponding to each of the equivalence relations given
in Exercises 5 and 6 above.
8. Prove Theorem 1.3.1.
9. Prove Theorem 1.3.2.
10. Prove Theorems 1.3.3 and 1.3.4 and apply these to the relations given in
Exercise 5 above.
11. Prove Theorems 1.3.5 and 1.3.6.

Q001-Algebra-111001_CH 01.indd 26 9/16/2011 10:09:08 AM

 Sets and Relations   1-27

12. Let R be a binary relation on a nonempty set X. Then, prove that R is an equiva-
lence relation on X if and only if R is reflexive on X and
(a, b) ∈ R and (b, c) ∈ R ⇒ (c, a) ∈ R.
13. Describe the equivalence relations on Z corresponding to the following parti-
tions of Z:
(i) {…, – 5, – 1, 3, 7, …}, {…, – 6, – 2, 2, 6, …},
{…, – 7, – 3, 1, 5, …}, {…, – 8, – 4, 0, 4, …}
(ii) {2n : n ∈ Z}, {2n 1 1: n ∈ Z}
(iii) Z2, {0}, Z1
(iv) {…, – 3, 0, 3, 6, …}, {…, – 2, 1, 4, 7, …}, {…, – 1, 2, 5, 8, …}.

14. Describe the partitions corresponding to the equivalence relations given in

Exercise 1 above.

15. Prove Theorem 1.3.7.

1.4 THE CARDINALITY OF A SET

The concepts of cardinality of a set and of cardinal number are very impor-
tant in the abstract study of any branch of mathematics and, in particular, in
the study of abstract algebra. In this section, we give a brief introduction of
these concepts.

Definition 1.4.1. For any set X, let |X| denote the class of all sets that are
equivalent to X (that is, bijective with X). Then, |X| is called the cardinality of
X or the cardinal number of X or, simply, a cardinal number.
If we define, for any two sets A and B, A . B whenever there is a bijec-
tion of A onto B, then . is actually an equivalence relation on the class of
all sets. The following is a direct consequence of the discussion made after
Definition 1.2.3.

Theorem 1.4.1. Let A, B and C be any sets. Then, the following holds good.
1. |A| 5 |B| ⇔ A . B ⇔ A ∈ |B| ⇔ B ∈ |A|
2. A ∈ |B| and B ∈ |C| ⇒ A ∈ |C|.

Definition 1.4.2. For any nonnegative integer n, let In be the set of positive
integers less than or equal to n. That is,

In 5 {1, 2, 3, …, n}.

Note that, if n 5 0, then In is the empty set.

Q001-Algebra-111001_CH 01.indd 27 9/16/2011 10:09:08 AM

 1-28   Algebra – Abstract and Modern

Theorem 1.4.2. The following are equivalent to each other for any ­nonnegative
integers n and m.
1. |In| 5 |Im|
2. In . Im
3. n 5 m
In view of the above theorem, we denote the cardinality of In by simply n.
Note that, for any set A, |A| 5 n if and only if there is a bijection of A onto the
set {1, 2, …, n} and, for this reason, we say that A has n-elements or A is an
n-element set if |A| 5 n.

Definition 1.4.3. A set A is called a f inite set if the cardinality of A is a

­nonnegative integer. A is called an inf inite set if it is not a finite set.
In other words, a set A is called finite if A is bijective with the set In for
some nonnegative integer n. A is called infinite if it is not bijective with In for
any nonnegative integer n.

Definition 1.4.4. A cardinal number is said to be f inite if any (and hence all)
of its members are finite sets.

Example 1.4.1. The set Z1 of positive integers is an infinite set, for we can
easily check that there cannot be a bijection of Z1 onto In for any nonnegative
integer n. If f : In → Z1 is a function, we can choose m ∈ Z1 such that f (a) ,
m for all a ∈ In.

Theorem 1.4.3. Let n be a nonnegative integer and X be a set, such that |X| 5 n.
Then, for any subset Y of X, |Y| 5 m for some 0  m  n.

Corollary 1.4.1. Every subset of a finite set is finite. Equivalently, any

superset of an infinite set is infinite.
We can identify any nonnegative integer n with the cardinal number |A|,
where A is a set with n elements. It can be easily seen that, for any ­nonnegative
integers n and m, n  m if and only if there is an injection of A into B, where
A and B are sets of cardinalities n and m, respectively. This suggests an exten-
sion of the usual ordering  on the set N of nonnegative integers to that of
cardinal numbers.

Definition 1.4.5. Let a and b be two cardinal numbers and X and Y be sets,
such that |X| 5 a and |Y| 5 b. Then, we define a is less than or equal to b (and
express this by a  b) if there is an injection of X into Y.
First of all, we have to prove that  is a well-defined relation on the cardi-
nals, in the sense of the following.

Q001-Algebra-111001_CH 01.indd 28 9/16/2011 10:09:08 AM

 Sets and Relations   1-29

Theorem 1.4.4. Let X, Y, A and B be sets, such that |X| 5 |A| and |Y| 5 |B|. Then,
there is an injection of X into Y if and only if there is an injection of A into B.

Proof: Since |X| 5 |A| and |Y| 5 |B|, there are bijections f : X → A and g : Y
→ B. If h : X → Y is an injection, then g o h o f 21 is an injection of A into B.
On the other hand, if p : A → B is an injection, then g21 o p o f is an injection
of X into Y. b
Thus,  is a well-defined binary operation on the set of cardinals. Since
A  A for any set A, it follows that  is reflexive on the set of cardinals.
Also, since the composition of injections is again an injection, we have that
 is a transitive relation. In addition to the reflexivity and transitivity of the
relation , we have another important property, namely the ­anti-symmetricity;
that is, a  b and b  a are possible only if a 5 b. The proof of this is not
that straight forward and requires a skilled proof.

Theorem 1.4.5. (Schroeder–Bernstein Theorem). Let X and Y be sets

and f : X → Y and g : Y → X be injections, then there exists a bijection of
X onto Y.

Proof: Put Z 5 g(Y). Then, Z is a subset of X.

Define h : X → Z by h(x) 5 g(f (x)) for any x ∈ X. Then, since g and f are
injections, h is also an injection. We define sequences {Xn} and {Zn} of sets
as follows:

X1 5 X and Z1 5 Z

and, for n . 1, Xn 5 h(Xn–1) and Zn 5 h(Zn–1). Then, Xn 5 hn21(X) and

Zn 5 hn21(Z), where hn21 5 h o h o … o h (n – 1 times) and h0 5 IdX.
We have

X 5 X1 ⊇ Z1 ⊇ X2 ⊇ Z2 ⊇ X3 ⊇ Z3 ⊇ X4 ⊇ …

Define p : X → Z by

h( x ) if x ∈ X n  Z n for some n

p( x )   .
 x otherwise

Then, it can be easily verified that p is a bijection of X onto Z. Now, define

q : X → Y by
q(x) 5 y if g(y) 5 p(x).

Q001-Algebra-111001_CH 01.indd 29 9/16/2011 10:09:09 AM

 1-30   Algebra – Abstract and Modern

Since g is an injection and p(x) ∈ Z 5 g(Y), there will be unique y ∈ Y such

that p(x) 5 g(y). Therefore, q is a well-defined mapping of X into Y. It can be
verified that q is a bijection of X onto Y. b

Corollary 1.4.2. The relation  is an anti-symmetric, transitive and reflexive

relation on the set of cardinals.
One can also define a relation  on the set of cardinals as |Y|  |X| if there is
a surjection of Y onto X. This is also an antisymmetric, transitive and reflexive
relation on the set of cardinals. In fact, this is precisely the relation  in view
of the following.

Theorem 1.4.6. Let X and Y be any nonempty sets. Then, there is an injection
of X into Y if and only if there is a surjection of Y onto X.

Proof: Suppose that f : X → Y is an injection, choose an arbitrary element

x0 ∈ X. Define g : Y → X by

 x if f (x)  y
g ( y )  .
 x0 otherwise (that is, y ∉ f (X ))

Since f is an injection, for each y ∈ f (X), there exists unique x ∈ X such that
f (x) 5 y. Therefore, g is a well-defined function of Y into X. Also, for any x ∈
X, f (x) ∈ Y and g(f (x)) 5 x and hence g is a surjection of Y onto X.
Conversely suppose there is a surjection g : Y → X. For each x ∈ X, consider
the set

Ax 5 g21({x}) 5 {y ∈ Y : g(y) 5 x}.

Since g is a surjection, each Ax, x ∈ X, is a nonempty subset of Y. There-

fore, {Ax : x ∈ X} is a nonempty class of nonempty sets. By an important
axiom of set theory (known as the axiom of choice), there exists a function
c : X → ∪ x∈ X Ax such that c(x) ∈ Ax for each x ∈ X (such a function is called
a choice function). Now, define f : X → Y by f (x) 5 c(x).
Note that Ax ∩ Ax′ 5 Ø for any x  x ∈ X and hence c(x)  c(x) if x x.
Therefore, f is an injection of X into Y. b
If A is a finite set and |A| 5 n, then we know that its power set P(A) (the set
of all subsets of A) is of cardinality 2n and that n , 2n. We can extend this to
arbitrary cardinals. First let us have the following.

Definition 1.4.6. If a is the cardinal of a set A, then the cardinal of the power
set P(A) is denoted by 2a, for the simple reason that P(A) . {0,1}A under the

Q001-Algebra-111001_CH 01.indd 30 9/16/2011 10:09:09 AM

 Sets and Relations   1-31

bijection B  xB, where xB is defined by xB(a) 5 1 or 0 according as a ∈ B or

a ∉ B. xB is called the characteristic map of B.
For any two cardinals a and b, we write a , b if a  b and a  b. Also
write a  b to say that a is not less than or equal to b.

Theorem 1.4.7. For any cardinal a, a , 2a.

Proof: Let a be a cardinal and A be a set such that |A| 5 a. Since the cardinal of
the empty set Ø is 0 and P(Ø) 5 {Ø} which is a nonempty set, we get that |Ø| 5
0 , 1 5 20 5 |P(Ø)| therefore, we can suppose that A is a nonempty set. Define

f : A → P(A) by f (a) 5 {a} for any a ∈ A.

Then, clearly f is an injection and hence

a 5 |A| # |P(A)| 5 2a

Now, we prove that |A|  |P(A)| or, equivalently, |P(A)|  |A|. By Theorem
1.4.6, it is enough if we can prove that there is no surjection of A onto P(A).
Suppose, if possible, that there is a surjection g : A → P(A). Then, for each a
∈ A, g(a) is a subset of A and every subset of A is of the form g(a) for some a
∈ A (since g is a surjection). Now, consider the set B defined by

B 5 {a ∈ A : a ∉ g(a)}.

Then, B is a subset of A and hence B 5 g(a) for some a ∈ A. Now,

a ∈ B ⇒ a ∈ g(a) ⇒ a ∉ B
         and a ∉ B ⇒ a ∉ g(a) ⇒ a ∈ B,

which are contradictions, since exactly one of the statements a ∈ B and a ∉ B

must be valid. Therefore, there is no surjection of A onto P(A) and hence
|P(A)|  |A|. Thus, a 5 |A| , |P(A)| 52a. b
Next, we have a brief discussion on countable cardinals.

Definition 1.4.7. Let X be any set and Z1 be the set of positive integers. Then,
X is said to be a countable set and |X| said to be a countable cardinal if |X| 5
|Z1|; that is, if X is equipotent with Z1 and if f : Z1 → X is bijection, then X can
be expressed as X 5 {f (1), f (2), …, f (n), …} or, simply X 5 {x1, x2, …}.
If X is not a countable set, then X is called an uncountable set and |X| is called
an uncountable cardinal.

Definition 1.4.8. A set X is called at most countable if it is either finite or

countable.

Q001-Algebra-111001_CH 01.indd 31 9/16/2011 10:09:09 AM

 1-32   Algebra – Abstract and Modern

Theorem 1.4.8. The following are equivalent to each other for any non-
empty set X:
1. There is an injection f : X → Z1
2. X is at most countable.
3. X is a subset of a countable set.
4. There is a surjection g : Z1 → X.

Proof: (1) ⇒ (2): Suppose that there is an injection f : X → Z1, put Y 5 f (X).
Then, X . Y ⊆ Z1. Suppose that X is not finite, then Y is an infinite subset of
Z1. Define g : Z1 → Y as follows.
Let g(1) be the least element in Y (use the well-ordering principle in Z1).
Having defined g(1), …, g(n 2 1), let g(n) be the least element in Y – {g(1),
g(2), …, g(n – 1)}, for any n . 1. Since Y is infinite, Y – {g(1), …, g(n – 1)}
  for any n . 1 and hence g is welldefined. Now, we have

g(1) , g(2) , … , g(n) , …

Clearly, g is an injection of Z1 into Y. We prove that g is surjection also. Let y

∈ Y. Then, the number of g(m), such that g(m) # y is finite and hence we can
choose the largest m such that g(m) # y. Then, g(m) # y < g(m 1 1). If g(m)
, y, then we get that g(m 1 1) # y (since g(m 1 1) is the least in Y – {g(1),
…, g(m)} and y ∈ Y – {g(1), …, g(m)}. Therefore, g(m) 5 y and hence g is a
surjection also and hence Z1 . Y so that Y is countable. Thus, X is countable
(g21 o f : X → Z is a bijection).
(2) ⇒ (3): If X . In for some n or X . Z1, in either case, X can be treated as
a subset of a countable set.
(3) ⇒ (4): If X is a subset of a countable set Y, then there is a bijection f : Y →
Z1. Then, the restriction of f to X is an injection of X into Z1. Therefore, by
Theorem 1.4.6, there is a surjection of Z1 into X.
(4) ⇔ (1) follows from Theorem 1.4.6. b

Corollary 1.4.3. The set R of real numbers is uncountable.

Proof: Let X 5 {0  x1x2,… : xi 5 0 or 1 for all i}. Then, X ⊆ R. We prove

that X is uncountable. Suppose, if possible, that X is countable. Let f : Z1 →
X be a bijection. Then, X 5 {f (1), f (2), …}. Let

f (n) 5 0  xn xn xn …, where xn 5 0 or 1.
1 2 3 i

For each n ∈ Z , define yn 5 1 2 xn and consider

1
n

y 5 0  y1y2y3…

Q001-Algebra-111001_CH 01.indd 32 9/16/2011 10:09:10 AM

 Sets and Relations   1-33

Then, since yn ? xn for each n, we get that y ? f (n) for all n and hence y ∉ X,
n
which is a contradiction. Thus, X is uncountable and thus so is R. b
For any cardinal number a, we have proved that a , 2a (Theorem 1.4.7)
Z Z
and, in particular Z� 
 2 which automatically implies that 2 is an
uncountable cardinal. In the following, we prove that the cardinal number of

Z�
the set R of real numbers is precisely 2 . The cardinal number of Z1 will
be usually denoted by N0 and that of R by c.

Theorem 1.4.9

c  2N0


Proof: Recall that P(Z ) . 2Z , the set of all mapping of Z1 into the two-

element set {1, 0}. We prove that there is a bijection of 2 onto R. Define

f : 2Z → R by

f (g) 5 0  g(1) g(2)… for any g ∈ 2 ,

where 0  g(1) g(2)… is the real number in the interval [0, 1) whose decimal
places are g(1), g(2), …. Then, clearly f is an injection. On the other hand,
noted that any real number x can be represented in the binary scale in the
form

x 5 …x7x5x3x1  x2x4x6x8….
+
where each xn 5 0 or 1 for every n ∈ Z1. Now, define h : R → 2Z by

h(x)(n) 5 xn if x 5 …x5x3x1  x2x4x6…

for any n ∈ Z1and x ∈ R. Then, h can be easily verified to be an injection.

Therefore, we have two injections f : 2Z → R and h : R → 2Z . By the
1 1

Schroeder–Bernstein Theorem 1.4.5, it follows that there is a bijection of R


onto 2Z�. Thus,

Z
 2N0 .

c  |R|  2Z�  2 b

∞
Theorem 1.4.10. Let X1, X2, … be finite sets. Then, ∪ X n is at most countable.
n1

Q001-Algebra-111001_CH 01.indd 33 9/16/2011 10:09:11 AM

 1-34   Algebra – Abstract and Modern

∞
Proof: Let X  ∪ X n . Without loss of generality, we can assume that each
n1
Xn is nonempty. Let |Xn| 5 mn and X n  {xn1 , xn2 , …, xnmn }. Then,

X = {x11 , x12 , …, x1n1 , x21 , …, x2 n2 , x31 , …}

which shows that there is a surjection of Z1 onto X. By Theorem 1.4.8, X is at

most countable. b

Corollary 1.4.4. If A and B are finite sets, then A ∪ B, A ∩ B and A 3 B are

all finite sets.

Proof: Clearly, A ∪ B and A ∩ B are finite. Also, for each a ∈ A, {a} 3 B . B

and hence finite and therefore A 3 B is finite, since A  B  ∪a∈ A {a} B. b

Theorem 1.4.11. If X and Y are countable sets, then X ∪ Y and X 3 Y are

countable and X ∩ Y is at most countable.

Proof: Suppose that X and Y are countable, then we can write

X 5 {x1, x2, x3, …} and Y 5 {y1, y2, y3, …}.

Since X ∩ Y is a subset of the countable set X, X ∩ Y is at most countable (by

Theorem 1.4.8). Also, since

X ∪ Y 5 {x1, y1, x2, y2, x3, y3, …},

it follows that X ∪ Y is countable. For each n ∈ Z1, let

An 5 {(xi, xj) : i 1 j 5 n 1 1}.

∞
Then, each An is a finite set and X  Y  ∪ An . By Theorem 1.4.10, X 3 Y is at
n1
most countable. Further, since {x} 3 Y  Y, Y is infinite and {x} 3 Y ⊆ X 3 Y,
it follows that X 3 Y is infinite. Thus, X 3 Y is countable. b
n
Corollary 1.4.5. If X1, X2, …, Xn are countable sets, then so are ∪ X i and
i1
X1 3 X2 3 … 3 Xn.

Corollary 1.4.6. Z, Z 3 Z and Q (the set of rational numbers) are countable.

Corollary 1.4.7. Countable union of countable sets is countable.

Q001-Algebra-111001_CH 01.indd 34 9/16/2011 10:09:11 AM

 Sets and Relations   1-35

Proof: Let {X1, X2, …, Xn, …} be a countable class of countable sets. Then,
∞
for each n ∈ Z1, there exists a bijection f n : Z1 → Xn. Now, let X  n∪ X n and
1
define f : Z1 3 Z1 → X by

f (n, m) 5 f n(m).
Then, clearly f is a surjection. Also, since Z1 3 Z1 is countable, there is a bijec-
tion g : Z1 → Z1  Z1. Now, f o g is a surjection of Z1 onto X. Therefore, X is
at most countable. But, since X is infinite, it follows that X is countable. b
Note that countable product of at most countable sets may not be count-
able. For consider the following.

Example 1.4.2. The set 2 5 {0, 1} is a finite set and Z1 is a countable set.
Here, 2Z1 ( R) is uncountable.

EXERCISE 1(d)

1. Prove that the cardinal numbers of Z1, Z2, Z, Q1 and Q are all equal to each
other.

2. Let f : X → Y and g : Y → Z be injections. Then, prove the following:

(i) |X| 5 |Z| ⇒ |Y| 5 |Z|
(ii) |X| 5 |Y| and f (X) ⊆ A ⊆ Y ⇒ |A| 5 |Y|
(iii) X ⊆ A ⊆ Y and |X| 5 |Y| ⇒ |A| 5 |Y|

3. Prove that any infinite subset of a countable set is countable.

4. If X is a countable set and f : X → Y is a surjection, then prove that Y is at most

countable.

5. Prove that a set X is infinite if and only if |X| 5 |Y| for some proper subset Y of X.

6. For any positive integer n, prove that (Q1)n is countable.

7. If X is a set such that |X| 5 |P(P(X))|, then prove that there exists a surjection
f : X → P(X).

8. Deduce from Exercise 7 above that |X| , |P(P(X))| for any set X.

9. If |X| 5 |A| and |Y|5|B|, then prove that |AB| 5 |XY|.

10. For any sets X and Y, prove that |X| 5 |Y| if and only if |P(X)| 5 |P(Y)|.

11. Prove that |P(Z)| 5 |P(Q)|.

12. Give an example of a set # of circles in the plane such that every circle with
positive radius properly contains a member of #.

Q001-Algebra-111001_CH 01.indd 35 9/16/2011 10:09:12 AM

 1-36   Algebra – Abstract and Modern

13. Prove that |(a, b)| 5 |(c, d)| for any intervals (a, b) and (c, d) in R with a , b and
c , d.

14. Let In 5 {1, 2, …, n} for any n ∈ Z1. Prove that |Z1| 5 |Z1 – In| for any n ∈ Z1.

15. Let X be a countable set and PF(X) be the set of all finite subsets of X. Then,
prove that
∞

∪X
n1
n
 X  PF ( X ) .

16. Prove that the set of polynomials in the indeterminate x over the set of rational
numbers is countable.

17. A real number a is said to be an algebraic number if there exists a nonzero

polynomial
f (x) 5 a0 1 a1x 1 … 1 anxn with ai ∈ Q

such that f (a) 5 0. Prove that the set of algebraic numbers is countable.

18. A real number is said to be transcendental if it is not algebraic. Prove that the set
of transcendental numbers is uncountable.

19. Prove that the set of complex numbers is uncountable.

20. Prove that the set of complex numbers, whose real and imaginary parts are ratio-
nal numbers, is countable.

Q001-Algebra-111001_CH 01.indd 36 9/16/2011 10:09:12 AM

 2
Number Systems
2.1 Integers
2.2 Congruence Modulo n
2.3 Rational, Real and Complex Numbers
2.4 Ordering
2.5 Matrices
2.6 Determinants

This chapter is meant to review some of the important properties of the set of
positive integers, the set of integers, the set of rational numbers, the set of real
numbers and the set of complex numbers. We do not discuss any axiomatic
development of these systems. We simply assume familiarity with addition
and multiplication of these and their usual properties. Also, we briefly dis-
cuss the concept of a partial order on a set in general and the usual ordering
on the real number system, in particular, these facilitate us in facing several
encounters with these throughout this book. Further, we recall the notion of
a matrix over the number systems and some important elementary properties
of the matrices and their determinants.

2.1 INTEGERS
In this section, we review certain important elementary properties of integers,
by assuming familiarity with the addition, subtraction, multiplication and the
usual ordering in these (that is, m # n if and only if n 2 m is nonnegative).
As mentioned in the beginning of the book, we follow the notations given
below.
Z : The set of integers {…, 22, 21, 0, 1, 2, …}
Z1 : The set of positive integers {1, 2, 3, …}

Q001-Algebra-111001_CH 02.indd 1 9/16/2011 10:07:56 AM

 2-2   Algebra – Abstract and Modern

Z2 : The set of negative integers {…, 23, 22, 21}

N : The set of nonnegative integers {0, 1, 2, 3, …}
m # n if and only if n 2 m [ N
m , n if and only if n 2 m [ Z1
For any subset S of Z and a [ S, a is called the smallest (or least) member
of S if a # s for all s [ S. It is well known that, for any given integers m and n
with m , n, there are only finitely many (at most n 2 m number of) integers a,
such that m , a , n.

Theorem 2.1.1 (Well-ordering Property of Z1). Any nonempty set of posi-

tive integers has the smallest member.

Proof: Let S be a nonempty set of positive integers. That is, [  S # Z1.

Suppose, if possible, that S has no smallest member. Since S is nonempty, we
can choose a [ S. Then, a is not smallest in S and hence there exists a1 [ S
such that a  a1; that is, a1 , a. Again since a1 is not smallest in S, we get
a2 [ S such that a2 , a1. Continuing this, we get an infinite set of integers
such that

0 , … , an , an21 , … , a2 , a1 , a

which is a contradiction, since there can be only a finite number of integers

between 0 and a. Thus, S has smallest member.
In fact, the above well-ordered property of Z1 can be extended to Z as
given below whose proof is similar to the above one.

Theorem 2.1.2. Let S be a nonempty subset of Z and b [ Z such that b , s

for all s [ S. Then, S has smallest member.
Now, we derive some more properties of Z1 as consequences of the well-
ordering property.

Theorem 2.1.3 (First Principle of Induction). Let b [ S # Z1 such that

b # n [ S ⇒ n 1 1 [ S.

Then, m [ S for all m $ b.

Proof: Put T 5 {m [ Z1 : b # m and m  S}. It is enough if we can prove

that T is the empty set. Suppose, if possible, that T is not empty. Then, b 2 1
, m for all m [ T. Therefore, by Theorem 2.1.2, T has smallest member, say
m0. Then, since m0 [ T,

Q001-Algebra-111001_CH 02.indd 2 9/16/2011 10:07:56 AM

 Number Systems   2-3

b # m0 and m0  S.

Then, since b [ S, we get that b , m0 and hence b # m0 2 1 and m0 2 1 

S (otherwise b # m0 2 1 [ S, so that m0 [ S). This implies that m0 2 1 [
T, which is a contradiction to the least property of m0. Thus, T is empty and
hence m [ S for all m $ b.

Corollary 2.1.1. Let S # Z1 such that 1 [ S and n 1 1 [ S whenever n [

S. Then, S 5 Z1.
Corollary 2.1.1 is the actual statement of the first principle of induction.
However, this has more general form in Theorem 2.1.3. Often, induction is
stated in terms of a proposition on Z1. A proposition on a set S means that,
for each a [ S, P(s) is a statement about s and P(s) is either true or false. The
above principle of induction is equivalent to the following.
Let P be a proposition on Z1 such that P(b) is true for some b [ Z1 and
P(n 1 1) is true whenever n $ b and P(n) is true. Then, P(n) is true for
all n $ b.

n( n1)
Example 2.1.1. Let us prove that 1 1 2 1 … 1 n 5 for all n [ Z1.
n( n1) 2
Let P(n) be the statement ‘1 2  n  ’ and S 5 {n [ Z1 : P(n) is
2
true}. Then, clearly 1 [ S # Z .
1

n [ S ⇒ P(n) is true
n( n1)
⇒1121…1n5
2
n( n 1) (n 1)(n  2)
⇒ 1 2  n  n 1   n 1 
2 2
⇒ P(n 1 1) is true
⇒ n 1 1 [ S.

Therefore, by the first principle of induction, S 5 Z1 and hence P(n) is true

for all n [ Z1.
The following shows the importance of the general version of the first
principle of induction given in Theorem 2.1.3.

Example 2.1.2. Let us find all positive integers n for which 3n , n!

Let S 5 {n [ Z1 : 3n , n!}.
Clearly 31 . 1!, 32 . 2!, 33 . 3!, 34 . 4!, 35 . 5! and 36 5 729 . 720 5 6!
But 37 5 2187 , 5040 5 7!, 38 , 8! and 39 , 9! which suggests that n [ S

Q001-Algebra-111001_CH 02.indd 3 9/16/2011 10:07:57 AM

 2-4   Algebra – Abstract and Modern

for all n $ 7. To prove this, let us apply the first principle of induction (­Theorem
2.1.3). We have 7 [ S and

7 # n [ S ⇒ 3n , n!
⇒ 3n11 5 3n ? 3 , n!(n11)
⇒ 3n11 , (n11)!
⇒ n 1 1 [ S.

Thus, by Theorem 2.1.3, n [ S for all n $ 7. Since we have already checked

that n  S for 0 , n , 7, we get that

S 5 {n [ Z1 : 7 # n}.

The first principle of induction does not work sometimes when we need to
know the truth of one or more smaller cases and not necessarily the immedi-
ately preceding one. To handle situations like this, we need another form of
induction given below.

Theorem 2.1.4 (Second Principle of Induction). Let S be a set of positive

integers and b [ S such that, for any b # n [ Z1,

b, b 1 1, …, n [ S ⇒ n 1 1 [ S.

Then, m [ S for all m $ b.

Proof: Put T 5 {m [ Z1 : b # m and m  S}. We need to prove that T is

empty. Suppose, if possible, that T is nonempty. Then, by the well-ordering
property of Z1 (Theorem 2.1.1), T has a smallest member, say m0. Since m0 [
T, we have b # m0 and m0  S. But, since b [ S, we get that b , m0. Also, by
the least property of m0, any integer less than m0 cannot be in T. Therefore,

b, b 1 1, …, m0 2 1 [ S.

By the hypothesis, it follows that m0 [ S, which is a contraction. Thus, T is

empty and hence m [ S for all m $ b.

Corollary 2.1.2. Let S # Z1 such that 1 [ S and, for any n [ Z1,

m [ S for all m , n ⇒ n [ S.

Then, S 5 Z1.

Q001-Algebra-111001_CH 02.indd 4 9/16/2011 10:07:57 AM

 Number Systems   2-5

We have actually used the well-ordering property of Z1 to prove both the

first and second principles of induction. However, we can prove the well-
ordering property using either of the induction principles. In the following,
we prove that all three are equivalent.

Theorem 2.1.5. The following are equivalent to each other:

1. Well-ordering property of Z1 (Theorem 2.1.1).
2. The first principle of induction (Theorem 2.1.3).
3. The second principle of induction (Theorem 2.1.4).

Proof: The proof of Theorem 2.1.3 is precisely the proof of (1) ⇒ (2).
(2) ⇒ (3): Assume that the first principle of induction holds.
Let b [ S # Z1 and, for any n $ b,

b, b 1 1, …, n [ S ⇒ n 1 1 [ S.

Put A 5 {a [ Z1 : a $ b and b, b 1 1, …, a [ S}. By our assumption on S,

it follows that b [ A and

b # n [ A ⇒ n 1 1 [ A.

From the first principle of induction, m [ A for all m $ b. In particular,

m [ S for all m $ b.
(3) ⇒ (1): Assume that the second principle of induction holds. Let A be a
nonempty set of positive integers. Suppose, if possible, that A has no smallest
member. Put S 5 Z1 2 A. Then, 1  A and hence 1 [ S. For any n $ 1, if 1,
2, …, n [ S, then 1, 2, …, n  A and therefore n 1 1  A (otherwise n 1 1
becomes the smallest member in A) and hence n 1 1 [ S. From the second
principle of induction, it follows that n [ S for all n $ 1; that is, S 5 Z1 and
hence A is empty, which is a contradiction. Thus, A has a smallest member.
The next result is one of the best applications of the second principle of
induction. Before this, let us recall that a positive integer p . 1 is called a
prime number (or simply, prime) if 1 and p are the only factors of p (a is said
to be a factor of b if ac 5 b for some integer c).

Theorem 2.1.6 (Fundamental Theorem of Arithmetic). Any positive integer

greater than 1 can be uniquely expressed as a product of prime numbers.

Proof: Let S 5 {n [ Z1 : n . 1 and n is a product of primes}. Then 2, being a

prime, is a member of S. Let 2 # n [ Z1 such that 2, 3, …, n [ S. If n 1 1 is a
prime, then clearly n 1 1 [ S. Suppose that n 1 1 is not a prime; then there is

Q001-Algebra-111001_CH 02.indd 5 9/16/2011 10:07:57 AM

 2-6   Algebra – Abstract and Modern

a factor a of n 1 1 such that a  1 and a  n 1 1. Therefore, there are positive

integers a and b such that

n 1 1 5 ab and 1,a,n11 and 1, b , n 1 1.

From the induction hypothesis, a and b [ S and hence a and b can be expressed
as products of primes and therefore so is n 1 1. Thus, n 1 1 [ S. By the second
principle of induction, it follows that m [ S for all m $ 2. Thus, any n . 1 can
be expressed as a product of primes.
We prove the uniqueness of the factorization also by using induction prin-
ciple. Let

p1 p2 … pr 5 n 5 q1 q2 … qs

where pi’s and qj’s are prime numbers. Suppose that pi 5 qj for some i and j.
We can suppose, by renumbering of pi’s and qj’s, that p1 5 q1. Then, p2 p3 …
pr 5 q2 q3 … qs , n and hence, by the induction hypothesis, r 5 s and each
pi is equal to some qj and vice versa. Next suppose that pi  qj for all i and j.
Without loss of generality, we can suppose that p1 . q1. Then,

n . (p1 2 q1) p2 … pr 5 p1 p2 … pr 2 q1 p2 … pr
5 q1 q2 … qs 2 q1 p2 … pr
5 q1 (q2 … qs 2 p2 … pr).

Again by the induction hypothesis, q1 5 pi for some 2 # i # r or q1 divides

p1 2 q1. But q1  pi for all i and hence q1 divides p1 2 q1. Therefore, q1 5 p1,
a contradiction to our assumption. Thus, r 5 s and each pi is equal to some qj
and vice versa.
The following is an important property of integers which we use through-
out this book. The proof of this is again by the well-ordering property of Z1.

Theorem 2.1.7 (The Division Algorithm in Z). Let a and b be any integers
and b . 0. Then, there exist unique integers q and r such that

a 5 bq 1 r and 0 # r , b.

(Here q is called the quotient and r is called the remainder of a modulo b.)

Proof: If a 5 0, we can take q 5 0 5 r and, if b 5 1, then we can take q 5 a

and r 5 0. Therefore, we can assume that b . 1 and a  0. Put

S 5 Z1 ∩ {a 2 bx : x [ Z}.

Q001-Algebra-111001_CH 02.indd 6 9/16/2011 10:07:57 AM

 Number Systems   2-7

Since b . 1, a , b|a| and hence a 2 b(2|a|) [ S. Therefore, S is a nonempty

subset of Z1. By the well-ordering property of Z1, S has a smallest member,
say m0.
Let m0 5 a 2 bx . 0 (since m0 [ S). If b , m0, then 0 , m0 2 b 5 a 2 b
(x11) [ S and m0 2 b , m0 which is a contradiction to the least property of
m0. Therefore, we have m0 # b. If m0 5 b, then a 2 bx 5 b and hence a 5 b
(1 1 x), so that we can take q 5 x 1 1 and r 5 0. Thus, m0 , b and

a 5 bx 1 (a 2 bx) 5 bx 1 m0

so that we can take q 5 x and r 5 m0.

To prove the uniqueness of q and r, let

bq 1 r 5 a 5 bq 1 r, 0 # r , b and 0 # r , b.

Then, b(q 2 q) 5 r 2 r and hence

b|q ] q| 5 |r] r| , b.

This implies that |q 2 q| 5 0; that is, q 5 q and r 5 r.

Definition 2.1.1. For any m and n [ Z1, let CD(m, n) be the set of all com-
mon divisors (factors) of m and n in Z1. That is,

CD(m, n) 5 {c [ Z1 : c divides both m and n}.

Clearly CD(m, n) is a nonempty subset of Z1 for any m and n [ Z1, since

1 is a divisor of any positive integer. Also, for any a and b [ Z1,

a divides b ⇒ a # b

and hence every member of CD(m, n) is less than or equal to both m and
n. This implies that CD(m, n) is finite and has a largest (greatest) member,
which is called the greatest common divisor of m and n and is denoted by
g.c.d.{m, n} or, simply (m, n). The following is an interesting property of the
g.c.d.’s.

Theorem 2.1.8. Let m and n be positive integers. Then, the following are
equivalent to each other for any d [ CD(m, n).
1. d 5 g.c.d.{m, n}
2. d 5 ma 1 nb for some a and b [ Z
3. Every member of CD(m, n) divides d.

Q001-Algebra-111001_CH 02.indd 7 9/16/2011 10:07:58 AM

 2-8   Algebra – Abstract and Modern

Proof: Let d [ CD(m, n); that is, d is a common divisor of m and n.

(1) ⇒ (2): Suppose that d is the greatest member of CD(m, n). Let S 5
Z1 ∩ {ma 1nb : a and b [ Z}. Clearly S is a nonempty subset of Z1 (for
example, m2 1 n2 [ S). By the well-ordering property of Z1, S has a smallest
member, say d0. We prove that d0 5 d. Since d0 [ S, we have

0 , d0 5 ma 1 nb for some a and b [ Z.

We first prove that d0 [ CD(m,n). By the division algorithm (Theorem 2.1.7),

we can write

m 5 d0q 1 r, 0 # r , d0, q and r [ Z.

Now, r 5 m 2 d0q 5 m 2 (ma 1 nb)q 5 m(1 2 aq) 1 n (2bq). If r . 0,

then r [ S and r , d0 which is a contradiction to the least property of d0 in S.
Therefore, r 5 0 and m 5 d0q. Thus, d0 divides m and, similarly d0 divides n
and hence d0 [ CD(m, n). From this, it follows that d0 # d. Also, since d [
CD(m, n), d divides m and n and hence d divides ma 1 nb 5 d0. Therefore, d
# d0. Thus, d 5 d0 5 ma 1 nb.
(2) ⇒ (3) and (3) ⇒ (1) are trivial.

Definition 2.1.2. Two positive integers m and n are said to be relatively prime
(or, prime to each other) if (m, n) 5 1. This is equivalent to saying that CD(m,
n) 5 {1}.
Note that m and n are relatively prime if and only if there is no prime
number dividing both m and n. The following is an important consequence
of Theorem 2.1.8.

Theorem 2.1.9. Let m, n and r [ Z1 such that m divides nr and (m, n) 5 1.

Then, m divides r.

Proof: By Theorem 2.1.8, there exist integers a and b such that ma 1

nb 5 1. Now,

r 5 r1 5 r(ma 1 nb) 5 mra 1 nrb.

Since m divides mra and m divides nr, it follows that m divides r.

Corollary 2.1.3. Let p be a prime number and m and n be positive integers

such that p divides mn. Then, p divides either m or n.

Proof: Suppose that p does not divide m. Then, (p, m) 5 1 and therefore, by
Theorem 2.1.9, p divides n.

Q001-Algebra-111001_CH 02.indd 8 9/16/2011 10:07:58 AM

 Number Systems   2-9

Definition 2.1.3. Let X be any set. A function s : Z1 → X is called a sequence

in X. A sequence s is usually represented by {s(1), s(2), …} or {s1, s2, …}.
Quite often a sequence s is described inductively by giving s(1) and a rule
to find s(n 1 1) from s(n). For example, we define s(n) 5 n! inductively by

1! 5 1

and (n 1 1)! 5 (n 1 1)n! for any n $ 1.

In the next theorem, we prove that this inductive method of defining a
sequence works well in the sense that there exists a unique sequence s : Z1 → X
satisfying the given conditions for determining s(n 1 1) from s(n).

Theorem 2.1.10 (Recursion Theorem). Let X be a set and x1 [ X. Suppose

that f : Z1 3 X → X is a mapping. Then, there exists a unique sequence s :
Z1 → X such that

s(1) 5 x1 and s(n 1 1) 5 f (n, s(n)) for all n [ Z1.

Proof: First we prove the existence of a sequence s : Z1 → X satisfying the

required conditions. Let

P 5 {S # Z1 3 X : (1, x1) [ S and if (n, x) [ S, then (n 1 1, f (n, x)) [ S}.

Then, P is nonempty, since Z1 3 X [ P. Let T be the intersection of all

members in P. Then, clearly T is a member of P and is contained in every
member of P. Put

A 5 {n [ Z1 : there is a unique x [ X such that (n, x) [ T}.

We prove, by induction principle, that A 5 Z1. Suppose, if possible, that 1  A.

Since (1, x1) [ T, there exists x  x1 in X such that (1, x) [ T. Then, T ] {(1, x)}
is a member of P and hence it contains T, which is a contradiction. Therefore,
1 [ A.
Next, let n [ A. Suppose, if possible, that n 1 1  A. Let xn be the unique element
in X such that (n, xn) [ T. Since T [ P, it follows that (n 1 1, f (n, xn)) [ T. Since
n 1 1  A, there exists y  f (n, xn) in X such that (n 1 1, y) [ T. Again, it can
be easily verified that T 2 {(n 1 1, y)} is a member of P and hence

T # T 2 {(n 1 1, y)}, which is a contradiction.

Therefore, n 1 1 [ A. By the first principle of induction, it follows that A 5 Z1.

Thus, for each n [ Z1, there exists unique element, say xn, in X such that (n, xn)
[ T. Now, define s : Z1 → X by s(n) 5 xn. Then,

s(1) 5 x1, since (1, x1) [ T

Q001-Algebra-111001_CH 02.indd 9 9/16/2011 10:07:58 AM

 2-10   Algebra – Abstract and Modern

and s(n 1 1) 5 f (n, s(n)), since (n, s(n)) [ T and hence (n 1 1, f (n, s(n))) [ T.
Thus, s is a sequence satisfying the required properties.
Next, we prove the uniqueness of s. Let s and t be sequences such that

s(1) 5 x1 5 t(1)
and s(n 1 1) 5 f (n, s(n)) and t(n 1 1) 5 f (n, t(n))

for all n [ Z1. Let B 5 {n [ Z1 : s(n) 5 t(n)}. Then, 1 [ B and, if n [ B,

then s(n) 5 t(n) and hence

s(n 1 1) 5 f (n, s(n)) 5 f (n, t(n)) 5 t(n 1 1),

so that n 1 1 [ B. Again, by the first principle of induction, B 5 Z1. Thus,

s(n) 5 t(n) for all n [ Z1 and hence s 5 t.
Before we close this section, let us recall the concept of absolute value of
an integer.

Definition 2.1.4. For any a [ Z, define

| a |
{
a if a0
a if a0
.

Note that |a| $ 0 for all a [ Z. The following can be proved by straight
forward verification.

Theorem 2.1.11. The following holds for any integers a, b and c.

(i) |a| 5 0 ⇔ a 5 0
(ii) |ab| 5 |a||b|
(iii) |a 2 b| 5 |b 2 a|
(iv) |a 1 b|  |a| 1 |b|
(v) |a 1 b| 5 |a| 1 |b| if and only if either both a and b are nonpositive or
nonnegative.
(vi) |a 2 b|  |a 2 c| 1 |c 2 b|
(vii) ||a| 2 |b||  |a 2 b|
(viii) ||a| 2 |b|| 5 |a 2 b| if and only if |a 1 b| 5 |a| 1 |b|.

Q001-Algebra-111001_CH 02.indd 10 9/16/2011 10:07:59 AM

 Number Systems   2-11

EXERCISE 2(a)

1. Prove the following for any n [ Z1:

(i) 12 1 22 1 32 1 … 1 n2 5 n(n 1 1)(2n 1 1)/6

( )
2
n(n1)
(ii) 13 1 23 1 33 1 … 1 n3 5
2
(iii) 1?1! 1 2?2! 1 3?3! 1…1 n?n! 5 (n 1 1)! 2 1
(iv) If X is a set and |X| 5 n, then |P(X)| 5 2n
(v) xn 2 1 5 (x 2 1)(xn21 1 xn22 1 … 1 x 1 1)
n  n
(vi) (x 1 y)n 5    x nr y r
 r 
r0 
n
(vii)  4 nn  3 2n 3
r1 3 3n
n
(viii)  1  n
r1 r (r1) n1
n
n(n1)(n2)
(ix)  r (r 1) 
r1 3
(x) 1 1 3 1 5 1 … 1 (2n 2 1) 5 n2.
2. Find all n [ Z1 for which 2n 1 1 , 2n21.
3. Let X be a set such that |X| 5 n $ 2. Prove that these are exactly n(n1) subsets
2
each with exactly two elements.
4. For any n and r [ Z1 such that 1  r  n, prove that
n  i  n 1
∑ r   r 1
ir

5. Use the Binomial theorem given in Exercise 1 (vi) above to prove the
following.
n  n
(i)     3 n for all n ∈ Z
r0  r 
(ii) (a 1 b)n [ aZ1 1 bn for all a, b and n [ Z1
n n
(iii)    =   
r even   
r  r odd  r 
0 r n 0 r n

6. Prove that the set of prime numbers is infinite.

7. For any a [ Z1 and for any prime number p, prove that there is a largest
­nonnegative integer n for which pn divides a.
8. Prove that any positive integer a can be expressed as
a  p1n1 p2n2 ... prnr

where p1, p2, …, pr are distinct primes and n1, n2, …, nr are nonnegative
integers.

Q001-Algebra-111001_CH 02.indd 11 9/16/2011 10:08:00 AM

 2-12   Algebra – Abstract and Modern

9. Let a, b and c [ Z1. c is said to be a common multiple of a and b if both a and b

divide c. Prove that there is a least common multiple for any a and b [ Z1. This
will be denoted by l.c.m.{a, b} or [a, b].

10. Let a  p1n1 p2n2 ... prnr , where pi’s are distinct primes and ni’s are nonnegative inte-
gers and let b [ Z1. Then, prove that b divides a if and only if
b  p1m1 p2m2 ... prmr , where mi [ Z and 0  mi  ni.
r r
11. Let a   pini and b   pimi , where pi’s are distinct primes and ni’s are non-
i1 i1
negative integers. Then, prove that
r
ki
g.c.d.{a, b} 5  pi , where k 5 minimum of n and m
i1 i i i
r
and l.c.m.{a, b}   pidi , where di 5 maximum of ni and mi.
i1

12. For any positive integers a and b, prove that the product of a and b is equal to the
product of their g.c.d. and l.c.m.
13. Let a, b and c be positive integers such that a divides both b and c. Then, prove
that a divides mb 1 nc for any integers m and n.
14. Let 1 , n [ Z1. Prove that either n is a prime or has a prime divisor which
is  n .
15. Let a1, a2, …, ar [ Z1 and a 5 g.c.d.{a1, a2, …, ar}. Then, prove the following:
(i) d 5 b a 1 b a 1…1b a for some b , b , …, b [ Z
1 1 2 2 r r 1 2 r

(ii) If S 5 {b1a1 1 b2a2 1 … 1 brar : bi [ Z}, then

S 5 aZ and a is the least member of S ∩ Z1.
(iii) If b [ Z1 and b divides ai for all 1  i  r, then b divides a.
16. Let a, b, c, d [ Z1 and a 5 bc 1 d. Then, prove that
g.c.d.{a, b} 5 g.c.d.{b, d}
17. For any positive integer a, prove that 1 1 a 1 a2 and 1 1 a are relatively prime.
18. Prove the following for any a, b, c, d [ Z1.
(i) g.c.d.{a, c} 5 1 5 g.c.d.{b, c} ⇔ g.c.d.{ab, c} 5 1
(ii) g.c.d.{a, b} 5 1 ⇔ g.c.d.{a, bn} 5 1 for any n [ Z1.
19. Establish the Euclidean Algorithm given below to find the greatest common divi-
sor of given positive integers a and b.
* Let a, b [ Z1 with a . b and b not a divisor of a. Use division algorithm
(Theorem 2.1.7) repeatedly as necessary, to write
a 5 bq1 1 r1
b 5 r1q2 1 r2
r1 5 r2q3 1 r3
A

Q001-Algebra-111001_CH 02.indd 12 9/16/2011 10:08:01 AM

 Number Systems   2-13

rn23 5 rn22qn21 1 rn21

rn22 5 rn21qn 1 rn

where rn is the last nonzero reminder. Then,

g.c.d.{a, b} 5 rn 5 xa 1 yb for some x, y [ Z.
20. Find x and y [ Z such that
g.c.d.{969, 1273} 5 969x 1 1273y.
21. For any positive integers a and b, prove that a divides b if and only if 2a 2 1
divides 2b 2 1.
22. For any a and b [ Z1, prove that a and b are relatively prime if and only if ma 1
nb 5 1 for some m and n [ Z and, in this case, |m| and |n| are relatively prime.
23. Let a and b be relatively prime and a, b, c [ Z1, then prove that ab divides c if
and only if both a and b divide c.
24. If a, b, c, d [ Z1 such that a 5 c(a, b) and b 5 d(a, b), then prove that (c, d) 5 1.
25. Let a, b [ Z1 such that b . 1 and (a, b) 5 1. Then, prove that (a 1 bc, b) 5 1
for any c [ Z1 and there is a unique n [ Z1 such that 1  n , b and (n, b) 5 1
and b divides an 2 1.

2.2 CONGRUENCE MODULO n

Here, we briefly discuss an important equivalence relation, namely congru-
ence modulo n, on the set Z of integers. The importance of this is due to the
fact that it is compatible with addition, subtraction and multiplication in Z.
Let us first agree with the following notation.

Definition 2.2.1. For any subsets A and B of Z, let

A 1 B 5 {a 1 b : a [ A and b [ B}

and AB 5 {ab : a [ A and b [ B}.

Note that A 1 B 5 B 1 A, AB 5 BA, (AB)C 5 A(BC) and (A 1 B) 1 C 5
A 1 (B 1 C) for any A, B and C # Z. Also AB is empty if and only if A is
empty or B is empty. If A is a single element set {a}, then we write a 1 B for
{a} 1 B and aB for {a}B.

Theorem 2.2.1. Let n be any positive integer. Then,

nZ, 1 1 nZ, …, (n 2 1) 1 nZ

form a partition of Z.

Q001-Algebra-111001_CH 02.indd 13 9/16/2011 10:08:01 AM

 2-14   Algebra – Abstract and Modern

Proof: For any i and j [ {0, 1, 2, … , n 2 1},

(i 1 nZ) ∩ (j 1 nZ)  [ ⇒ i 1 na 5 j 1 nb for some a and b [ Z

⇒ |i 2 j| 5 n|b 2 a|
⇒ i 2 j 5 0 (since |i 2 j| , n)
⇒i5j

Therefore, for any i  j, i 1 nZ and j 1 nZ are disjoint. Also, for any a [ Z,

we have (by the division algorithm, Theorem 2.1.7)

a 5 nq 1 r

for some q and r [ Z such that 0  r , n and hence

a [ r 1 nZ for some r 5 0, 1, …, n 2 1.

Thus, the sets r 1 nZ, 0  r  n 2 1 form a partition of Z.

Theorem 2.2.2. Let n be the equivalence relation on Z corresponding to the

partition given in the above theorem. Then, for any a and b [ Z

a n b if and only if n divides |a 2 b|.

Proof: For any a and b [ Z, a n b if and only if a and b belong to the same
set in the partition. Therefore,

a n b ⇔ a and b [ i 1 nZ for some 0  i , n

⇔ a 5 i 1 nx and b 5 i 1 ny for some x, y [ Z
⇔ a 2 b 5 n(x 2 y)
⇔ |a 2 b| 5 n|x 2 y|, x, y [ Z
⇔ n divides |a 2 b|.

Definition 2.2.2. The equivalence relation n obtained above is called the

congruence modulo n. Quite often, we write

a  b (mod n) for a n b.

That is, for any integers a and b and for any positive integer n,

a  b (mod n) ⇔ n divides |a 2 b|.

Q001-Algebra-111001_CH 02.indd 14 9/16/2011 10:08:01 AM

 Number Systems   2-15

If n (a) is the equivalence class containing a corresponding to n, then

we have

n (a) 5 a 1 nZ 5 r 1 nZ, where a 5 nq 1 r, 0  r , n.

Also, for any a and b [ Z

n (a) 5 n (b) ⇔ a  b (mod n)

⇔ a [ b 1 nZ
⇔ a 2 b [ nZ
⇔ n divides |a 2 b|.

In the following, we prove that the congruence modulo n is compatible

with the usual arithmetical operations.

Theorem 2.2.3. Let 1, n [ Z. Then, the following holds for any a, b, c and
d [ Z.
(i) a  b (mod n) and c  d (mod n) ⇒ a 1 c  b 1 d (mod n)
(ii) a  b (mod n) ⇒ 2a  2b (mod n)
(iii) a  b (mod n) ⇒ ac  bc (mod n)
(iv) a  b (mod n) and c  d (mod n) ⇒ ac  bd (mod n)
(v) a  b (mod n) ⇒ am  bm (mod n) for all m [ Z1
(vi) ac  bc (mod n) and (c, n) 5 1 ⇒ a  b (mod n)

Proof:
(i) a n b and c n d ⇒ n divides |a 2 b| and |c 2 d|
⇒ nx 5 a 2 b and ny 5 c 2 d for some x, y [ Z
⇒ n(x 1 y) 5 (a 1 c) 2 (b 1 d), x 1 y [ Z
⇒ n divides |(a 1 c) 2 (b 1 d)|
⇒ (a 1 c) n (b 1 d)
(ii) and (iii) can be proved similarly.
(iv) a n b and c n d ⇒ ac n bc and bc n bd (by (iii))
⇒ ac n bd (since n is transitive)
(v) is a simple consequence of (iv) and the principle of induction.
(vi) ac n bc and (c, n) 5 1 ⇒ n divides |ac 2 bc| 5 c|a 2 b|
⇒ n divides |a 2 b| (since (c, n) 5 1)
⇒ a n b.

Q001-Algebra-111001_CH 02.indd 15 9/16/2011 10:08:02 AM

 2-16   Algebra – Abstract and Modern

Corollary 2.2.1. Let 1 , n [ Z and a, b, a1, a2, …, ar, b1, b2, …, br [ Z. Then,
the following holds:
r r r r
(i) ai  bi (mod n) for all 1  i  r ⇒  ai ≡  bi (mod n) and  ai ≡  bi (mod n).
r r i1 i1 i1 i1
 ai ≡  bi (mod n).
i1 i1
(ii) If f (x) is a polynomial with integer coefficients and
a  b (mod n), then f (a)  f (b) (mod n).

Proof: To prove (i), use Theorem 2.2.3 (i) and apply induction on r. (ii) is a
consequence of (i), (iii) and (v) of Theorem 2.2.3.
Observe that, when x  y (mod 8), either of x and y may be replaced by
the other in any polynomial congruence modulo n (by Corollary 2.2.1 (ii))
and this idea can be used in solving linear congruences ax  b (mod n)
when a and n are relatively prime. Before proving this, let us have the
­following.

Theorem 2.2.4. Let 1 , n [ Z and a [ Z such that (|a|, n) 5 1. Then, there

exists unique r [ Z1 such that 1  r , n, (r, n) 5 1 and ar  1 (mod n).

Proof: Since (|a|, n) 5 1, there exists u and v [ Z such that u|a| 1 vn 5 1 (by
Theorem 2.1.8) and hence

sa 1 vn 5 1 (*)

for some s and v [ Z. We use the division algorithm to get r and q [ Z

such that

s 5 nq 1 r and 0  r , n.

If r 5 0, then n divides sa 1 vn 5 1, which is a contradiction to the

hypothesis that n . 1. Therefore, we have 1  r , n. Also, by substituting
s 5 nq 1 r in (*), we get that

(nq 1 r)a 1 vn 51
         or ra 1 (v 1 qa)n 5 1 (**)

which implies that ra 2 1 is a multiple on n. Thus, ra  1 (mod n) and 1  r

, n. Also, from equation (**), we get that (r, n) 5 1. If r is any other integer
such that 1  r , n and ra  1 (mod n), then (r 2 r) a  0 (mod n), so that
|r 2 r||a| is a multiple of n. Since (|a|, n) 5 1, it follows that n divides |r 2 r|.
Since 1  r, r , n, we get that r 2 r 5 0 or r 5 r.

Q001-Algebra-111001_CH 02.indd 16 9/16/2011 10:08:02 AM

 Number Systems   2-17

Theorem 2.2.5. Let n [ Z1 and a and b [ Z such that (|a|, n) 5 1. Then, the
linear congruence equation

ax  b (mod n)

has a unique solution r in {0, 1, 2, …, n 2 1} and the set of all integer solu-
tions of this is precisely equal to the congruence class r 1 nZ.

Proof: By Theorem 2.2.4, there exists unique s such that 1  s , n, (s, n) 5 1

and as  1 (mod n). Choose the unique r such that 0  r , n and sb [ r 1 nZ
(or, equivalently, sb  r (mod n). Then,

ar  asb  b (mod n).

Therefore, r is a solution of ax  b (mod n) in {0, 1, …, n 2 1}. We prove that

the congruence class  n (r) 5 r 1 nZ is equal to the set of integer solutions
of ax  b (mod n). For any x [ Z,

ax  b (mod n) ⇔ sax  sb (mod n)

⇔ x  r (mod n)
⇔ x [ r 1 nZ

The uniqueness of r is clear.

Example 2.2.1. Let us find all integer solutions of 55x  65 (mod 80). First
observe that, if m is a common divisor of a, b and n, then ax  b (mod n) if
and only if ma x ≡ mb (mod mn ). Since 5 is a common divisor of 55, 65 and 80,
55x  65 (mod 80) if and only if 11x  13 (mod 16). Note that (16, 11) 5 1.
A quick check reveals that 311  1 (mod 16). Therefore, the integer solutions
of 11x  13 (mod 16) are all y  313 (mod 16) or, equivalently, all y  7
(mod 16) (since 39  7 (mod 16)). Thus, integer solutions of 55x  65 (mod 80)
are members of 7 1 16Z and those in {0, 1, 2, …, 79} are precisely 7, 23,
39, 55, 71.
The method that is followed in Example 2.2.1 above is formalized in the
following.

Theorem 2.2.6. Let 1 , n [ Z1 and a and b [ Z. Then, ax  b (mod n) has

an integer solution if and only if the g.c.d. (a, n) divides b.

Proof: Let d 5 (a, n). Suppose that ax  b (mod n) has an integer solution. Let s
be an integer solution of ax  b (mod n). Then, as  b (mod n) and hence nr 5
as 2 b for some r [ Z. Now, b 5 as 2 nr, d divides both a and n and hence

Q001-Algebra-111001_CH 02.indd 17 9/16/2011 10:08:02 AM

 2-18   Algebra – Abstract and Modern

d divides b. Conversely suppose that d divides b. Put a 5 a/d, b 5 b/d and n 5

n/d. Then, (a, n) 5 1 and hence, by Theorem 2.2.5, ax  b (mod n) has an
integer solution, say x0. Then, n divides |ax0 2 b| and hence nq 5 ax0 2 b
for some q [ Z. Now

dnq 5 dax0 2 db

and hence nq 5 ax0 2 b, which implies that ax0  b (mod n). Thus, ax  b
(mod n) has an integer solution.
We close this section by developing tests for the divisibility of integers by
various primes. These tests are easy for small primes, but these are not practi-
cal for large primes. The following is easy, since any integer is divisible by 2
if and only if the last digit in it is one of 0, 2, 4, 6, and 8.

Theorem 2.2.7. Let a [ Z1, then a is even if and only if a  r (mod 10) for
some r [ {0, 2, 4, 6, 8}. Also a is divisible by 5 if and only if a  0 (mod 10)
or a  5 (mod 10) (that is, the last digit in a is 0 or 5).
r
Theorem 2.2.8. Let a [ Z1 and a 5 ar ar 2 1 … a1a0 5  ai 10i, where ai’s are
i0
integers such that 0  ai  9. Then,
r r
a  ∑ ai (mod 3), a  ∑ ai (mod 9)
i0 i0

and a  (a0 1 a2 1 a4 1 …) 2 (a1 1 a3 1 a5 1 …) (mod 11).

Proof: Consider the polynomial given by

f (x) 5 a0 1 a1x 1 a2x2 1 … 1 ar xr.

r
Then, f (10) 5  ai 10i 5 a. Since 10  1 (mod 3), it follows from ­Corollary
i0 r
2.2.1 (ii) that f (10)  f (1) (mod 3). Then, a   ai (mod 3). Also, since
i0
r
10  1 (mod 9), we get that a 5 f (10)  f (1) 5  ai (mod 9). Similarly, since
i0
10  21 (mod 11), we get that a 5 f (10)  f (21) 5 (a0 1 a2 1 …) 2 (a1 1
a3 1 …) (mod 11).
r
Corollary 2.2.2. Let a [ Z1 and a 5  ai 10i with 0  ai  9. Then,
i0
r
(i) 3 divides a if and only if 3 divides  ai ,
i0
r
(ii) 9 divides a if and only if 9 divides  ai , and
i0
(iii) 11 divides a if and only if 11 divides (a0 1 a2 1 …) 2 (a1 1 a3 1 …).

Q001-Algebra-111001_CH 02.indd 18 9/16/2011 10:08:04 AM

 Number Systems   2-19

Theorem 2.2.9. Let p . 5 be a prime and a [ Z1 such that a 5 10k 1 a0,

where k and a0 are nonnegative integers and 0  a0  9. Then, there exists
unique mp [ Z1 satisfying the following:
(i) 1  mp , p
(ii) 10mp  1 (mod p)
(iii) For any s  mp (mod p), p divides a ⇔ p divides k 1 sa0 (in particular,
p divides a ⇔ p divides k 1 mpa0).
r
Proof: Notice that, when a is represented in decimal system a 5  ai 10i or
i0
a 5 arar 2 1 …a1a0 with 0  ai  9, then a 5 10k 1 a0 where k 5 ar ar 2 1 …
r
a1 5  ai 10i1. Since p is prime and p . 5, (p, 10) 5 1. By Theorem 2.2.4,
i0
there exists unique mp [ Z1 such that 1  mp , p, (mp, p) 5 1and 10mp  1
(mod p). By Theorem 2.2.3,

amp  10kmp 1 a0mp  k 1 a0mp (since 10mp p 1).

Therefore, p divides amp if and only if p divides k 1 a0mp.

Since (p, mp) 5 1, it follows that
p divides a ⇔ p divides amp
⇔ p divides k1mpa0
⇔ p divides k 1 sa0 for any s  mp (mod p).
The above theorem can be better understood by the following examples,
in which we test certain positive integers for their divisibility by a given
prime p . 5.

Example 2.2.2
1. Let us test the divisibility of 62354 by 7. Here, a 5 62354 5 10k 1 a0,
where k 5 6235 and a0 5 4 since 10.5  1 (mod 7), m7 5 5.
7 divides 62354 ⇔ 7 divides k 1 m7a0
⇔ 7 divides 6235 1 5.4 (5 6255)
⇔ 7 divides 625 1 5.5 (5 650)
⇔ 7 divides 65 1 5.0 (5 65)
Since 7 does not divide 65, it follows that 7 does not divide 62354.
2. Consider a 5 5876438 and test its divisibility by 7 we have 10.5  1
(mod 7) and hence m7 5 5.
7 divides a ⇔ 7 divides 587643 1 5.8 (5 587683)
⇔ 7 divides 58768 1 5.3 (5 58783)

Q001-Algebra-111001_CH 02.indd 19 9/16/2011 10:08:04 AM

 2-20   Algebra – Abstract and Modern

⇔ 7 divides 5878 1 5.3 (5 5893)

⇔ 7 divides 589 1 5.3 (5 604)
⇔ 7 divides 60 1 5.4 (580)
⇔ 7 divides 8 1 5.0 (58)
Since 7 does not divide 8, if follows that 7 does not divide 5876438.

Example 2.2.3
1. Test 7892654 for its divisibility by 11. Since 10.10  1 (mod 11), we
have m11 5 10. Also, since 10  2 1 (mod 11), we can take s 5 21 in
Theorem 2.2.9 (iii).Therefore,
11 divides 7892654 ⇔ 11 divides 789265 1 (21) 4 (5 789261)
⇔ 11 divides 78926 1 (21)1 (5 78925)
⇔ 11 divides 7892 1(21)5 (5 7887)
⇔ 11 divides 788 1 (21)7 (5 781)
⇔ 11 divides 78 1 (21)1(5 77),
which is true.

Thus, 11 divides 7892654. In this context, note that Corollary 2.2.2 (iii)
is a better test for the divisibility by 11.
2. Test 7892654 for the divisibility by 13.
Since 10.4  1 (mod 13), we have m13 5 4. Therefore,
13 divides 7892654 ⇔ 13 divides 789265 1 4.4 (5 789273)
⇔ 13 divides 78927 1 4.3 (5 78939)
⇔ 13 divides 7893 1 4.9 (5 7929)
⇔ 13 divides 792 1 4.9 (5 828)
⇔ 13 divides 82 1 4.8 (5 114)
⇔ 13 divides 11 1 4.4 (5 27)
Since 13 does not divide 27, it follows that the given number 7892654 is
not divisible by 13.
3. Test whether 7892654 is divisible by 23.
Since 10.7  1 (mod 23), m23 5 7.
We have 23 divides 7892654 ⇔ 23 divides 789265 1 7.4 (5 789293)
⇔ 23 divides 78929 1 7.3 (5 78950)
⇔ 23 divides 7895 1 7.0 (5 7895)
⇔ 23 divides 789 1 7.5 (5 824)

Q001-Algebra-111001_CH 02.indd 20 9/16/2011 10:08:04 AM

 Number Systems   2-21

⇔ 23 divides 82 1 7.4 (5 110)

⇔ 23 divides 11 1 7.0 (5 11),
which is not true.
Thus, 23 does not divide 7892654.

EXERCISE 2(b)

1. State whether the following are true or false.

(i) 10n  0 (mod 2n) for any n [ Z1
(ii) 6789453  5987654 (mod 3)
(iii) 237092  236092 (mod 100)
(iv) 132  312 (mod 23)
(v) 786 2 687 (mod 11)

2. Find the set of integer solutions for each of the following.

(i) 15x  25 (mod 35)
(ii) 21x  35 (mod 49)
(iii) 25x  16 (mod 20)
(iv) 27x  21 (mod 24)
(v) 7x  16 (mod 17)
(vi) 9x  14 (mod 15)

3. Prove that, for any prime p, (p 2 1)! 1 1  0 (mod p). (This is known as Wilson’s
theorem.)

4. Test the following divisibilities.

(i) 876453 by 3
(ii) 746538 by 9
(iii) 587642 by 7
(iv) 7896534 by 11
(v) 87965325 by 17
(vi) 97865432 by 19
(vii) 67892345 by 23
(viii) 79862345 by 29

5. For any prime p and for integer a, prove that ap  a (mod p). (This is known as
Fermat’s theorem.)

6. Let n [ Z1. A set {a0, a1, …, an21} of distinct integers is called a transversal for
congruence mod n if ai [ i 1 nZ for each 0  i  n 2 1.

Q001-Algebra-111001_CH 02.indd 21 9/16/2011 10:08:04 AM

 2-22   Algebra – Abstract and Modern

Prove the following for any transversal {a0, a1, …, an21} for congruence
modulo n.
(i) ai  aj (mod n) for any i  j
(ii) ai  i (mod n) for any 0  i  n 2 1
(iii) For any a [ Z, a  ai (mod n) for some 0  i  n 2 1.

7. Find whether the following are transversals.

(i) {0, 3, 2, 1} for congruence mod 5.
(ii) {23, 22, 21, 0, 1, 2, 3} for congruence mod 7.
(iii) {0, 2, 22, 23, …, 210} for congruence mod 11.
(iv) {2, 4, 6, …, 2n} for congruence mod n, if n is odd.
(v) {a, 2a, 3a, …, na} for congruence mod n if (a, n) 5 1
(vi) {5, 10, 15, 20, …, 105} for congruence mod 21.

8. For any n [ Z1, prove that any n consecutive integers form a transversal for
congruence mod n.

9. Characterise all n in each of the following cases that satisfy the given
condition
(i) 1 1 2 1 3 1 … 1 (n 2 1)  0 (mod n)
(ii) 12 1 22 1 32 1 … 1 (n 2 1)2  0 (mod n)
(iii) 13 1 23 1 33 1 … 1 (n 2 1)3  (mod n)

10. For any n [ Z1, prove that 3 divides n implies 3 divides m for m any rearrange-
ment of the digits in n.
n
11. For any n [ Z1, prove that 103 ≡ 1 (mod 3n2 ).

12. Find all the digits x (0  x  9) for which 12x, 527, 846, 531 is divisible by 3;
9; or 11.

13. If a1, a2, …, an [ {0, 1, 2, …, 9} and a1  0, prove that 11 divides a1a2…ananan 2 1…

a2a1.

14. Prove the following for any relatively prime positive integers m and n:
(i) For any a and b [ Z1, a  b (mod mn) ⇔ a  b (mod m)
and a  b (mod n)
(ii) If c and d are both integer solutions of x  a (mod n) and of x  b (mod
m), then c  d (mod mn).

15. For any r [ Z1, let Mr 5 1 1 100 1 1002 1 … 1 100r21.

(i) Prove that Mr 5 101010…01 having exactly r ones.
(ii) Prove that each of 7, 9 and 11 divides Mr for infinitely many r’s.

Q001-Algebra-111001_CH 02.indd 22 9/16/2011 10:08:05 AM

 Number Systems   2-23

2.3 RATIONAL, REAL AND COMPLEX NUMBERS

Although we assume familiarity with the rational, real, and complex num-
bers, we prefer to give the structure of rational numbers for the simple reason
that we imitate this construction later in this book in a more general set up.
We do not attempt the construction of real numbers, since it is outside the
scope of this book.

Definition 2.3.1. Let Z be the set of all integers and Z* 5 Z 2 {0}. We define
a binary relation R on the set Z 3 Z* as follows: For any (a, b) and (c, d) in
Z 3 Z*,
(a, b) R (c, d) ⇔ ad 5 bc (that is, the products ad and bc are equal).
The following is a straight forward verification.

Theorem 2.3.1. R is an equivalence relation on Z 3 Z*.

Definition 2.3.2. For any (a, b) [ Z 3 Z*, the equivalence class of (a, b)
corresponding to R will be denoted by a . That is,
b

a
5 R(a, b) 5 {(c, d) [ Z 3 Z* : (a, b) R (c, d)}
b
i.e., a 5 {(c, d) [ Z 3 Z* : ad 5 bc}.
b
2
For example, represents the set of all pairs (c, d) of integers, with d  0,
3
such that 2d 5 3c. Note that, for any (a, b) and (c, d) [ Z 3 Z*,

a c
 ⇔ (a, b) R (c, d) ⇔ ad 5 bc.
b d

Definition 2.3.3. For any (a, b) [ Z 3 Z*, the R-equivalence class a is called
b
a rational number and the set of all rational numbers is denoted by Q. That is,

a 
   : a and b ∈  and b ≠ 0
 b 

Note that Q is precisely the quotient of Z 3 Z* by R. In the following, we

introduce the arithmetical operations addition, subtraction, multiplication and
division. First of all, observe that the following holds for any a, b, c, d [ Z
with b  0 and d  0.

Q001-Algebra-111001_CH 02.indd 23 9/16/2011 10:08:05 AM

 2-24   Algebra – Abstract and Modern

ad a

bd b
0 0 0
 
b d 1
a 0
 ⇔ a0
b d
ab a

b 1

Definition 2.3.4. For any r  a and s  c in Q, we define the following.

b d
a c ad  bc
rs   
b d bd
 a  a
r   
 b  b
 a  c  ac
r  s     
 b  d  bd
1 b
 if a ≠ 0
r a

r
a ( ) ad
 b  , if c ≠ 0.
s c ( )
d
bc

It can be easily verified that r 1 s, 2 r, r  s and r does not depend on the

s
integers a, b, c and d, but they depend on the classes a and c ; that is,
b d
a a c c ad  bc ad bc a a
 and  ⇒  , 
b b d d bd bd b b
ac ac ad ad
 and  if c  0 and c  0.
bd bd bc bc
That is, the operations 1, 2, ? and / are well-defined. Also, it can be easily
seen that the following arithmetical laws are satisfied in Q.
1. r 1 s 5 s 1 r
2. r 1 (s 1 t) 5 (r 1 s) 1 t
0
3. r 1 0 5 r, where 0 5 .
1

Q001-Algebra-111001_CH 02.indd 24 9/16/2011 10:08:07 AM

 Number Systems   2-25

0
4. r 1 (2 r) 5 0 (5 )
1
5. r ? (s ? t) 5 (r ? s) ? t
6. r ? (s 1 t) 5 r ? s 1 r ? t
7. r?s5s?r
1 b
8. r ? 1 5 r, where 15 (5 for any 0  b [ Z)
1 b
9. r ? ( 1r ) 5 1

Definition 2.3.5. Any nonempty set together with the operations 1, 2, ? and /
satisfying the properties (1) to (9) above is called a field.
We will be discussing about fields in great detail later in Part III and Part
IV of this book. We just want to highlight here that the set Q of rational num-
a
bers is a field. For any integer a, consider the rational number and, we can
a  1 
see that the map a    is an injection of the set Z of integers into the set
 1 
Q of rational numbers. If we identify a with a/1, then we can see that Z is a
subset of Q and, for any a and b [ Z,

a b a  b  a  a a b ab
  ,    and ⋅  .
1 1 1  1  1 1 1 1
These demonstrate that the usual arithmetical operations addition, subtrac-
tion and multiplication on the integers are simply the restrictions of those on
rational numbers to Z. Thus, for all practical purposes, we can treat integers
as rational numbers by means of the identification of a with a .
1
As we have constructed rational numbers from integers, we can construct
real numbers from rational numbers. However, the procedure is not as simple
as the construction of rational numbers. We need some more techniques from
analysis to construct real numbers from rational number. However, for the
benefit of an enthusiastic reader, a brief sketch of the construction of real
numbers is given in the exercises. The proofs are not very difficult, but require
care, attention and some elementary knowledge about sequences, Cauchy
sequences, convergent sequences and their limits. The real number system is
denoted by R and it is known that R is a field.
Next we construct the system of complex numbers. Consider the Cartesian
product R 3 R, where R is the set of real numbers. We define addition, sub-
traction, multiplication and division to make R 3 R a field. For any z 5 (a, b)
and w 5 (c, d) in R 3 R, let us define

z 1 w 5 (a, b) 1 (c, d) 5 (a 1 c, b 1 d)
2 z 5 2(a, b) 5 (2a, 2b)
z ? w 5 (a, b) ? (c, d) 5 (ac 2 bd, ad 1 bc)

Q001-Algebra-111001_CH 02.indd 25 9/16/2011 10:08:08 AM

 2-26   Algebra – Abstract and Modern

1 1  a b 
   ,  if a  0 or b  0
z (a, b)  a 2  b 2 a 2  b 2 
z (a, b)  ac  bd bc  ad 
and     ,  if c  0 or d  0.
w (c, d )  c 2  d 2 c 2  d 2 
Theorem 2.3.2. R 3 R, together with the operations defined above, is a field.
For any a [ R, consider (a, 0) in R 3 R. Then, a  (a, 0) is an injection
of R into R 3 R and satisfies the following for any a and b [ R.

(a 1 b, 0) 5 (a, 0) 1 (b, 0)
(2a, 0) 5 2(a, 0)
(ac, 0) 5 (a, 0)?(c, 0)
 1 
 , 0  1 if a  0
 a  ( a, 0)
 c  (c, 0)
    , 0  if a  0.
 a  ( a, 0)

These show that we can identify R with the subset R 3 {0} of R 3 R, by

means of the injection a  (a, 0) and that the usual arithmetical operations on
R coincide with those on R 3 R restricted to R (5 R 3 {0}).
Now, let us identify another distinguished element, namely (0, 1), in R 3 R.
First observe that

(0, 1) ? (0, 1) 5 (21, 0) 5 21

since we are identifying (a, 0) with a,

Put (0, 1) 5 i.

Then, i2 5 21 and hence i is a root of the polynomial x2 1 1 over R. Further,

any element z 5 (a, b) in R 3 R can be expressed as

z 5 (a, 0) 1 (0, 1)(b, 0) 5 a 1 ib

by identifying (a, 0) with a, (b, 0) with b and (0, 1) with i. Thus,

R 3 R 5 {a 1 ib : a and b [ R}

where i is the element (0, 1). a 1 ib is the usual familiar form of complex
numbers and let us agree to call any element of R 3 R as a complex number.

Q001-Algebra-111001_CH 02.indd 26 9/16/2011 10:08:08 AM

 Number Systems   2-27

R 3 R, together with the arithmetical operations defined above, is denoted

by C. Thus, we have

C 5 {a 1 ib : a and b [ R}

and the arithmetical operations on C take the following form.

(a 1 ib) 1 (c 1 id) 5 (a 1 c) 1 i(b 1 d)

   2(a 1 ib) 5 (2a) 1 i(2b)
    (a 1 ib)?(c 1 id) 5 (ac 2 bd) 1 i(ad 1 bc)
1  a   b 
  2 i 
(a  ib )  a  b 2   a 2  b 2 
      
( a  ib)  ac  bd   bc  ad 
  i  
       (c  id )  c 2  d 2   c 2  d 2 

Now, we have the following number systems.

Z1 5 The set of positive integers
N 5 The set of nonnegative integers
Z 5 The set of integers
Q 5 The set of rational numbers
R 5 The set of real numbers
C 5 The set of complex numbers
These are interrelated by

Z1 ⊂ N ⊂ Z ⊂ Q ⊂ R ⊂ C

in such a way that the usual arithmetic operations addition, subtraction and
multiplication on each of these are precisely restrictions of those on the next
system. Moreover, Q, R and C are fields while the others are not.
We close this section with an additional operation, namely the complex
conjugation, on C.

Definition 2.3.6. For any z 5 a 1 ib [ C with a and b [ R, the complex

conjugate of z is defined as the complex number.

z  a  ib (  a  i(  b)).

If z 5 a 1 ib, and a and b [ R, then a and b are called real part and imaginary
part of z, respectively.
The following are easy verifications.

Q001-Algebra-111001_CH 02.indd 27 9/16/2011 10:08:09 AM

 2-28   Algebra – Abstract and Modern

Theorem 2.3.3. The following holds for any complex numbers z and z:
1. z  z z  z 
2. zz
3. zz z z 
4. zz  The real part of z.
2
5. z−z  The imaginary part of z
2i
6. z  z ⇔ z ∈ R ⇔ The real part of z  0.

Theorem 2.3.4. For any z 5 a 1 ib (a and b [ R) [ C,

z z  a2  b2 ,

The nonnegative square root of zz is called the absolute value of z and is

denoted by |z|; That is,

|z|2 5 z z 5 a2 1 b2.

The map z  |z| satisfies the following properties:

1. |z 1 z| # |z| 1 |z|
2. |zz| 5 |z||z|
3. |z| 5 0 ⇔ z 5 0
4. |rz| 5 |r||z| for any real number r.
r if r  0
5. For any real number r, | r | 
r if r  0

EXERCISE 2(c)

A sequence {an} of rational numbers is said to be a Cauchy sequence if, for each posi-
tive rational number [, there exists n0 [ Z1 such that
|an 2 am| , [ for all n and m $ n0.
A sequence {an} in Q is said to be convergent if there exists r [ R such that, for
each rational [ . 0, there exists n0 [ Z1 such that |an 2 r| , [ for all n $ n0 and in
this case we write an → r and r 5 limit of an.
Prove the following:

1. Q is countable and R and C are uncountable.

2. For any a and b in Q such that a , b, then the set {r [ Q : a , r , b} is bijective

with Q.

Q001-Algebra-111001_CH 02.indd 28 9/16/2011 10:08:10 AM

 Number Systems   2-29

3. If a, b, c and d [ Q with a , b and c , d, then (a, b)Q is bijective with (c, d)Q,
where
(a, b)Q 5 {r [ Q : a , r , b}.

4. For any real numbers a, b, c and d with a , b and c , d,

(a, b)  (c, d)  R and (a, b)Q  (c, d)Q  Q
where (a, b) 5 {s [ R : a , s , b}.

5. Between any two real numbers, there is a rational number.

6. { 1n } is a Cauchy sequence.
7. Every convergent sequence in Q is a Cauchy sequence.

8. Let CS(Q) be the set of all Cauchy sequences in Q.

For any {an} and {bn} in CS(Q), define
{an} ~ {bn} if and only if |an 2 bn| → 0.
Then, ~ is an equivalence relation on CS(Q).

9. For each {an} [ CS(Q), there exists r [ R such that an → r.

10. For each r [ R, there exists {an} [ CS(Q) such that an → r and, if {bn} is
another Cauchy sequence in Q such that bn → r, then {an} ~ {bn}.

11. The quotient CS(Q)/~ is bijective with R.

12. For any a [ Q, the sequence {an}, such that an 5 a for all n, is called a con-
stant sequence and is denoted by {a}. Then, a  {
a} is an injection of Q into
CS(Q)/~.

13. If {an} and {bn} [ CS(Q), then {an 1 bn} and {an bn} [ Q.

14. For any {an}, {bn}, {an} and {bn} [ CS(Q),

{a n} ~ {an} and {b n} ~ {bn} ⇒ {an  bn} ~ {an  bn} and {anbn} ~ {anbn}.

15. For any {an} and {bn} [ CS(Q), define

{
an}{
bn}  {a
n  bn }

{
an}  {
 an}

{an} ⋅ {bn}  {a

n bn },

where {an} is the ~-equivalence class of {an} in CS(Q). Then, the operations 1, 2 and ?
are well-defined on CS(Q)/~.

Q001-Algebra-111001_CH 02.indd 29 9/16/2011 10:08:11 AM

 2-30   Algebra – Abstract and Modern

2.4 ORDERING
The well-ordering property of positive integers is with respect to the natural or
usual ordering. This natural ordering is there on the rational number system and
the real number system also. However, there is no such ordering on the complex
number system. In this section, we introduce the abstract concept of a partial
ordering on a given set and discuss its elementary properties in general and those
of the natural ordering on R in particular. Let us begin with the following.

Definition 2.4.1. Let X be a nonempty set. A binary relation  on X is said

to be a partial order or a partial ordering if  is reflexive, transitive and anti-
symmetric (that is, a  b and b  a only when a 5 b). A pair (X, #) is called
a partially ordered set or, simply, a poset if X is a nonempty set and # is a
partial order on X.
A partial order is usually denoted by the symbol # (which is read as ‘less
than or equal to’). It can be easily verified that, if # is partial order on a set
X, then the inverse of # is also a partial order on X and is denoted $. That is,
for any a and b [ X,

a # b ⇔ b $ a.

If a # b and a  b, then we write a , b.

Example 2.4.1
1. (Z1, #), (Z, #), (Q, #) and (R, #) are all partially ordered sets, where
# is the natural ordering.
2. For any nonempty set X, the equality relation is a partial order on X. That
is, for any a and b [ X, if we define a # b if and only if a 5 b, then # is
a partial order on X. Note that this is the only binary relation on X which
is both an equivalence relation and a partial order on X.
3. Let P(X) be the set of all subsets of a given set X. For any A and B [
P(X), define
A # B if and only if A is a subset of B.

Then, # is a partial order on P(X).

4. Let X be the set of all real valued functions defined on a set A (that is, X
5 RA). For any f and g [ X, define
f # g if and only if f (a) # f (b) in R for all a [ A.

Then, # is a partial order on X.

Q001-Algebra-111001_CH 02.indd 30 9/16/2011 10:08:12 AM

 Number Systems   2-31

5. Let (X1, #), (X2, #), …, (Xn, #) be posets and X 5 X1 3 X2 3 … 3 Xn.

For any a 5 (a1, a2, …, an) and b 5 (b1, b2, …, bn), define
a # b if and only if ai # bi for all 1# i # n.
Then, # is a partial order on X and is called the coordinate-wise ordering.
6. In Example (5) above, define a # b if and only if either a 5 b or there
is i0, 1# i0 # n, such that ai 5 bi for all i , i0 and ai0  bi0 . Then, # is a
partial order on X 5 X1 3 X2 3 … 3 Xn and is called the lexicographic
ordering or dictionary ordering.

Definition 2.4.2. A partial order # on a set X is called a total order if, for
any a and b [ X, either a # b or b # a and, in this case, (X, #) is called a
totally ordered set.
R together with the natural ordering is totally ordered set. In Examples
(2) and (3), the partial orders are not total orders, except when X has at most
one element.

Definition 2.4.3. Let (X, #) be a poset, A # X and x [ X. Then, x is called

a lower bound (upper bound) of A in X if x # a (respectively a # x) for all
a [ A. If A has a lower bound (upper bound) in X, then A is said to be bounded
below (respectively, bounded above). A is said to be bounded if it is both
bounded below and bounded above. A lower bound x of A is called great-
est lower bound if y  x for all lower bounds y of A in X and it is denoted
by glbXA or, simply, glb A when there is no ambiguity about X. Similarly, an
upper bound x of A is called least upper bound and denoted by lubXA or lub A
if x  y for all upper bounds y of A in X.

Example 2.4.2
1. In (R, ), Z1 is bounded below and not bounded above, while the set Z2
of negative integers is bounded above and not bounded below.
2. 1 is the lub of the interval (0, 1) and 0 is the glb of (0, 1) in (R, ).

Definition 2.4.4. Let (X, ) be a poset, A # X and a0 [ A. a0 is said to be the

least (greatest) element of A if a0  a is (respectively a  a0) for all a [ A.
a0 is said to be a maximal element of A if there is no element a in A with a0
, a. Similarly, a0 is said to be a minimal element of A if there is no element
a in A with a , a0.
Clearly any subset of a poset can have at most one least element and at most
one greatest element. Also, the least element (greatest element), if it exists, is
minimal (respectively, maximal). However, a subset A of a poset may possess
more than one minimal (maximal) elements and any minimal (maximal) ele-
ment is not necessarily the least (respectively, greatest) element.

Q001-Algebra-111001_CH 02.indd 31 9/16/2011 10:08:12 AM

 2-32   Algebra – Abstract and Modern

Example 2.4.3
1. The interval (0, 1) in (R, ) has neither a minimal element nor a maxi-
mal element.
2. Let X be a set with more than one element and consider the poset
(P(X), #) of all subsets of X. Let Y be the set of all nonempty subsets
of X. Then, Y has minimal elements; in fact, for any x [ X, {x} is a
minimal element in Y and is not the least element, since {x}  {y} for
any y  x in X. Also, let Z be the set of all proper subsets of X. Then,
Z has maximal elements; in fact, for any x [ X, X 2 {x} is a maximal
element in Z and is not greatest in Z.

Definition 2.4.5. Let (X, ) be a poset. Any nonempty subset A in X, such

that, for any a and b [ A, either a  b or b  a, is called a chain in X.
In other words, any totally ordered subset of a poset is called a chain.

Example 2.4.4
1. Z is a chain in (R, ). In fact, R itself is a chain in (R, ) and hence any
nonempty subset of R is a chain.
2. If X 5 {a, b, c, d}, then
A 5 {, {a}, {a, b}, {a, b, c}, X}

is a chain in the poset (P(X), #).

3. Consider the set Z1 of positive integers and, for any a and b [ Z1, define
a if a divides b (that is, ac 5 b for some c [ Z1). Then, | is a partial
b
order on Z1. For any a [ Z1,
{an : n [ Z1}

is a chain in (Z1, 1).

The following is an important axioms of set theory, though its popular
name is Zorn’s lemma. It is a lemma used to prove some other equivalent
axioms of set theory.

Zorn’s lemma 2.4.1. Let (X, ) be a poset in which each chain has an upper
bound in X. Then, (X, ) has a maximal element.
The following is an equivalent form of Zorn’s lemma and, in this form only
the Zorn’s lemma is used several times in this book.

Corollary 2.4.1 (An equivalent form of Zorn’s lemma). Let  be a class of

subsets of a given set X satisfying the following.

Q001-Algebra-111001_CH 02.indd 32 9/16/2011 10:08:12 AM

 Number Systems   2-33

If # is a subclass of  such that, for any A and B [ #, either A # B or

B # A, then ∪ A [ . (In this case,  is said to be closed under unions of
A∈
chains.) Then,  has a maximal member, in the sense that, there is a member
M in  such that M is not properly contained in any other member of .
Next, we take up a brief discussion of the well-ordering principle which
is also an axiom equivalent to the Zorn’s lemma. First, let us have the
­following.

Definition 2.4.6. Let X be a set. A partial order  on X is called a well-order

if every nonempty subset of X has a least element with respect to .
It can be easily seen that any well-order on a set is a total order; for, the set
{a, b} should possess a least element which must be either a or b. However,
there are total orders which are not well-orders. Consider the examples given
below.

Example 2.4.5
1. The natural order on Z1 is a well-order (by Theorem 2.1.1).
2. The natural order on Q is a total order, but not a well-order; for, the
interval (0, 1) ∩ Q has no least member, since for any 0 , a , 1, there
is a rational number r such that 0 , r , a.
3. The division order | on Z1 (that a|b if a divides b) is not a total order (for
example, if p and q are distinct primes, then p  q and q  p) and hence
not a well-order.

The Principle of Well-ordering 2.4.1. Any nonempty set can be well-

ordered; that is, if X is a given nonempty set, then there is a well-order on X.
We close this section with a mention of another important axiom of set
theory, namely the axiom of choice which is known to be equivalent to each
of the Zorn’s lemma and the principle of well-ordering. First, we have the
following.

Definition 2.4.7. Let {Ai}i[I be an indexed nonempty class of nonempty sets

(that is, I is a nonempty set and, each Ai, i [ I, is a nonempty set). Then, any
function c : I → i ∪∈I
Ai such that c(i) [ Ai for all i [ I is called a choice func-
tion.
This amounts to saying that a choice function c is simply choosing one
element from each Ai, i [ I. If the index set I is a finite set, say I 5 {1, 2,
…, n}, then the choice functions can be easily seen to be just elements of
the Cartesian product A1 3 A2 3 … 3 An and, in this case, the existence of
choice function is precisely equivalent to say that the Cartesian product A1 3
A2 3 … 3 An is nonempty. This idea can be extended to define the Cartesian

Q001-Algebra-111001_CH 02.indd 33 9/16/2011 10:08:13 AM

 2-34   Algebra – Abstract and Modern

­product of infinite class of sets. That is, if {Ai}i[I is an infinite class of non-
empty sets, then their Cartesian product can be defined as

∏ A  {c : I → ∪ A
i∈ I
i
i∈ I
i : c(i ) ∈ Ai for all i ∈ I }.

The axiom of choice, given below, say that the Cartesian product of any non-
empty class of nonempty sets is a nonempty set.

The Axiom of Choice 2.4.1. Given any nonempty class {Ai}i[I of nonempty
sets, there is a choice function c : I → ∪ Ai (that is, c is a function such that
i∈I
c(i) [ Ai for all i [ I).

EXERCISE 2(d)

1. List all the partial orders on a 2-element set, a 3-element set and a 4-element
set.

2. Prove that the number of partial orders on an n-element set is less than or equal
n ( n1)
to 2 2 .

3. Prove that the lexicographic ordering on X1 3 X2 3 … 3 Xn is a total ordering if

and only if the partial orders on each Xi is a total order.

4. Prove that any well-order on any set is a total order.

5. Give an example of total order which is not a well-order.

6. Let (X1, ), …, (Xn, ) be posets and X 5 X1 3 X2 3 … 3 Xn. Prove that the
lexicographic ordering on X is a well-order if and only if the partial order  on
each of the Xi’s is a well-order.

2.5 MATRICES
Though matrices are originated from the study of solutions of certain systems
of linear equations and are later found to be in one-to-one correspondence with
linear transformations of a finite dimensional linear space into another finite
dimensional linear space, but these have acquired an independent status and
form one of the most important areas of study in modern abstract algebra. In
particular, matrices are a rich source of examples and counter examples of sev-
eral concepts in noncommutative algebraic structures which we come across
throughout this book. Actually, we study later in detail about matrices over an
abstract ring. However, in this section, we briefly discuss matrices over the real
number system or complex number system. Let us begin with the following.

Q001-Algebra-111001_CH 02.indd 34 9/16/2011 10:08:13 AM

 Number Systems   2-35

Definition 2.5.1. For any positive integer n, let In denote the set of integers
from 1 to n; that is,

In 5 {1, 2, …, n}.

For any m and n [ Z1, a mapping

A : Im 3 In → R (or C)

is called an m 3 n matrix over R or C, as the case may be or, simply, an m

3 n matrix, when there is no ambiguity about R or C. An m 3 n matrix A is
usually represented by the values A(i, j), which are real or complex numbers,
for each 1 # i # m and 1 # j # n and we express the matrix A by writing

A 5 (aij), where aij 5 A(i, j).

Also, we express an m 3 n matrix A by an array of mn real or complex

numbers aij, 1 # i # m and 1 # j # n, written in m rows and n columns
with aij in the ith row and jth column as exhibited below.

 a11 a12 a13  a1n 

 
 a a22 a23  a2 n 
 21 
  
A   
   ai 3  ain 
 
  
 
am1 am 2 am 3  a  mn

Here, aij is called the ijth entry in the matrix A and m 3 n is called the size
of A. Actually, the size of A is not an integer, but it is a pair (m, n) (which is
usually written as m 3 n) of integers. An m 3 n matrix A 5 (aij) and an r 3
s matrix B 5 (bij) are said to be equal if m 5 r, n 5 s and aij 5 bij for all 1 #
i # m and 1 # j # n; that is, A and B have equal number of rows and equal
number of columns and have the same ijth entry for each i and j. The n-tuple
(ai , aj , …, ai ) is called the ith row and the m-tuple (ai , a2 , …, am ) is called
1 2 n j j j
the jth column of the m 3 n matrix A 5 (aij). A 1 3 n matrix is called a row
matrix and an m 3 1 matrix is called a column matrix.

Definition 2.5.2. An n 3 n matrix is called a square matrix of order n and

the n-tuple (a11 a22 … ann) is called the diagonal of a square matrix A 5 (aij).
A is called a diagonal matrix if aij 5 0 for any i  j; that is, except the entries

Q001-Algebra-111001_CH 02.indd 35 9/16/2011 10:08:13 AM

 2-36   Algebra – Abstract and Modern

on the diagonal, all other entries are 0. A diagonal matrix A 5 (aij) is called a
scalar matrix if a11 5 a22 … 5 ann

Definition 2.5.3. A square matrix A 5 (aij) is called an upper triangular

matrix if aij 5 0 for all i . j and A is called a lower triangular matrix if aij 5
0 for all i , j.
The set of all m 3 n matrices over the real number system (complex num-
ber system) is denoted by Mm3n(R) (Mm3n(C), respectively). The set of all
square matrices of order n is denoted by Mn(R) or Mn(C) as the case may be.

Example 2.5.1
2 3 1 1
 
1. 1 0 2 3  is a 3 3 4 matrix over R (over C also, since R ⊆ C).
 
0 1 1 1 

1 2 1
 
2. 2 3 0  is a square matrix of order 3 over C.
 
2 i i 

1 2 0 
 
3. 0 2 3  is an upper triangular matrix of order 3.
 
0 0 1

2 0 0 0
 
 3 1 0 0
4.   is a lower triangular matrix of order 4.
1 2 2 0
 
 5 2 0 4
2 0 0 
 
5. 0 1 0  is a diagonal matrix.
 
0 0 1

2 0 0
 
6. 0 2 0 is a scalar matrix.
 
0 0 2
Note that, for any positive integer n, if we define f : R → Mn(R) by f(a) 5
(aij), where

a if i  j
aij   ,
0 if i ≠ j

Q001-Algebra-111001_CH 02.indd 36 9/16/2011 10:08:14 AM

 Number Systems   2-37

for any a  R, then f is an injection of R into Mn(R) and therefore, we can

identify a in R with the scalar matrix f(a) in Mn(R) and hence R can be
identified with the subset of Mn(R) consisting of all scalar matrices of order
n. Similarly C can be identified with a subset of Mn(C). In the following, we
extend the arithmetical operations addition, subtraction and multiplication
in the real and complex number systems to Mn(R) or Mn(C). We do this in a
more general set up.

Definition 2.5.4. Let m and n be any positive integers and A 5 (aij) and B 5
(bij) be any m 3 n matrices over R or C. Then, we define
A 1 B 5 (cij), where cij 5 aij 1 bij for all 1  i  m and 1  j  n and define

2A 5 (2aij)

A 1 B will also be expressed as A 1 B 5 (aij 1 bij).

Example 2.5.2
 2 3 1 4 3 1 2 5
   
Let A 5 1 2 0 3 
and B 5  2 3 1 4  M3 3 4(R).
   
 4 5 2 2  1 2 0 1

 4  5
 2  3 31 1 2


Then, A 1 B   1 2 23 0  (1) 3 4
 
4  (1) 5  (2) 2  0 2 1 
5 4 3 1
 
 1 5 1 1
 
3 3 2 3

2 3 1 4 
 
and A   1 2 0 3 .
 
4 5 2 2
Definition 2.5.5. Let m, n and r be any positive integers and A 5 (aij) 
Mm3n(R) and B 5 (bij)  Mn 3 r(R). Then, we define the product AB as an
m 3 r matrix given by
n
A ? B 5 (cij) where cij   aik bkj
k1

for all 1  i  m and 1  j  r. That is,

cij 5 ai1b1j 1 ai2b2j 1 … 1 ain bnr.

Q001-Algebra-111001_CH 02.indd 37 9/16/2011 10:08:15 AM

 2-38   Algebra – Abstract and Modern

Note that for the product AB to be defined it is necessary that the number
of columns in A must be equal to the number of rows in B. Therefore, even
if AB is defined, BA may not be defined. If we define the dot product of two
n-tuples, a 5 (a1, a2, …, an) and b 5 (b1, b2, …, bn) by

a ? b 5 a1b1 1 a2b2 1 … 1 anbn,

then the ijth entry in the product AB is precisely the dot product of ith row in A
and jth column in B. Also, note that the product of any two square matrices of
the same order is always defined.
 3 2 1
2 1 3 2  


  2 0 1 
Example 2.5.3. Let A 3 2 1 0  M3 3 4(R) and B   
  1 1 0 
 M4 3 3(R) 6 4 2 1  
4 2 3 
Then, A ? B 5 (cij), where
C11 5 2 ? 3 1 1 ? 2 1 3 ? 1 1 2 ? 4 5 19
C12 5 2 ? 2 1 1 ? 0 1 3 (‑1) 1 2 2 5 5
C13 5 2 ? (21) 1 1?1 1 3?0 1 2?3 5 5
C21 5 3 ? 3 1 2?2 1 (21)?1 1 0 ? 4 5 12
C22 5 3 ? 2 1 2?0 1 (21) ? (21) 1 0 ? 2 5 7
C23 5 3 ? (21) 1 2 ? 1 1 (21) ? 0 1 0 ? 3 5 21
C31 5 6 ? 3 1 4 ? 2 1 2 ? 1 1 1 ? 4 5 32
C32 5 6 ? 2 1 4 ? 0 1 2?(21) 1 1 ? 2 5 12
and C33 5 6 ? (21) 1 4 ? 1 1 2 ? 0 1 1 ? 3 5 1
19 5 5 
 
and hence AB 5 12 7 1. Similarly, we can compute BA and see that
 
32 12 1 
AB  BA. Here, note that AB is a 3 3 3 matrix and BA is a 4 3 4 matrix.
Even if AB and BA are of same size, they may not be equal; for, consider the
matrices

0 2 1 1 0 2
   
A  0 1 2 and B  2 1 0.
   
3 0 1 3 2 1

7 4 1 6 2 3
   
Then, AB  8 5 2 and BA  0 5 4 and therefore AB  BA.
   
6 2 7 3 8 8

Q001-Algebra-111001_CH 02.indd 38 9/16/2011 10:08:16 AM

 Number Systems   2-39

Definition 2.5.6
1. For any m and n  Z1, the m 3 n matrix all of whose entries are zero is
called the zero matrix and is denoted by Om3n or, simply O, when there is
ambiguity about the size of the matrix.
2. For any n  Z1, the square matrix (dij) is called the identity matrix of
order n, where dij is defined by
1 if i  j
dij   .
0 if i  j

The identity matrix of order n is denoted by In3m or, simply I, when there is
no ambiguity about the order of the matrix.

Theorem 2.5.1. Let m and n  Z1 and A, B and C be m 3 n matrices over R

or C. Then, the following holds:
1. A1B5B1A
2. (A 1 B) 1 C 5 A 1 (B 1 C)
3. A 1 Om3n 5 A
4. A 1 (2A) 5 Om3n

Theorem 2.5.2. The following holds for any matrices A, B and C, in the sense
that whenever one side of an equation is defined, then the other side is also
defined and both sides of that equation are equal.
1. A(BC) 5 (A B)C
2. A(B 1 C) 5 AB 1 AC
3. (A 1 B)C 5 AC 1 BC
4. AI 5 A 5 IA, where I is the identity matrix of appropriate order.
In addition to the operations addition and multiplication of matrices, we
have yet another operation of matrices, namely the scalar multiplications. The
real or complex numbers are called scalars and we multiply any matrix by any
scalar as defined below.

Definition 2.5.7. Let A 5 (aij) be an m 3 n matrix and a be a scalar; that is a

 R or C. Then, the matrix aA is defined as

aA 5 (aaij).

That is, aA is obtained by multiplying each entry aij of A by a to get ijth

entry of aA.

Q001-Algebra-111001_CH 02.indd 39 9/16/2011 10:08:16 AM

 2-40   Algebra – Abstract and Modern

Theorem 2.5.3. The following holds for any matrices A and B and for any
scales a and b.

1. a(A 1 B) 5 aA 1 aB, whenever A and B are of same size.

2. (a 1 b)A 5 aA 1 bA
3. a(AB) 5 (aA)B 5 A 5 (aB), whenever A ? B is defined.
4. a(bA) 5 b(aA) 5 (ab)A

Definition 2.5.8. Let n  Z1, 1  i  n and 1  j  n. Then, the n 3 n

matrix whose ijth entry is 1 and all other entries are 0 is called a matrix unit
and is denoted by Eij. Note that each Eij is a square matrix of order n. In gen-
eral, the order of Eij is not mentioned in the notation of the matrix unit Eij and
the order is to be understood as per the context. However, we call Eij as the n
3 n matrix unit, when it is necessary to mention the order of Eij. Note that, for
any scalar a, aEij is the square matrix whose ijth entry is a and all other entries
are 0 and hence we have the following.

Theorem 2.5.4

1. Any n 3 n matrix A can be expressed as

n n
A  ∑∑ aij Eij
i1 j1

2. EijErs 5 djrEis, where djr 5 1 or 0 according as j 5 r or j  r.

We close this section by introducing another important operation on

matrices.

Definition 2.5.9. For any m 3 n matrix A 5 (aij), the transpose of A is defined

as the n 3 m matrix obtained by interchanging the rows and columns of A.
The transpose of A is denoted by At; that is,

At 5 (aji).

Example 2.5.4
1 2 
1 2 3  

1. If A  , then A  2 1
t
2 1 0  
3 0 

Q001-Algebra-111001_CH 02.indd 40 9/16/2011 10:08:16 AM

 Number Systems   2-41

3 1 4 
 3  
 3 1 2
 1 0 2


2. If A 1 0 2 4, then A  
t .
  2 2 3
 4 2 3 0   
3 4 0 
3. If A is a square matrix, then At is also a square matrix of order same
as of A.

Theorem 2.5.5. The following holds for any m 3 n matrices A and B.

1. (At)t 5 A
2. (aA)t 5 aAt
3. (A 1 B)t 5 At 1 Bt
4. (2A)t 5 2At

Definition 2.5.10. A square matrix A of order n is said to be nonsingular or

invertible if there exists a square matrix B of order n such that

AB 5 I 5 BA.

A matrix is said to be singular if it is not nonsingular.

Theorem 2.5.6
1. If A is an m 3 n matrix and B is an n 3 r matrix, then (AB)t 5 BtAt
2. If A is a nonsingular square matrix, then there exists a unique square
matrix B such that
AB 5 I 5 BA

and this B is called the inverse of A and is denoted by A21.

3. If A and B are nonsingular square matrices of the same order, then AB is
nonsingular and (AB)21 5 B21A21.
4. A square matrix A is nonsingular if and only if its transpose At is nonsingular
and, in this case (At) 21 5 (A21)t.

EXERCISE 2(e)

1. Compute the following for the matrices

 2 3 1  2 1 0  0 1 3
     
A  1 2 3, B  0 2 1 and C  2 3 0
 
  
  
 1 3 2 1 0 2 1 0 2

Q001-Algebra-111001_CH 02.indd 41 9/16/2011 10:08:17 AM

 2-42   Algebra – Abstract and Modern

(i) A1B
(ii) (A 1 B) 1 C
(iii) B1C
(iv) A 1(B 1 C)
(v) AB
(vi) At
(vii) Bt
(viii) BtAt
(ix) BA
(x) AtBt

2. For any two matrices A and B, prove that both AB and BA are defined if and only
if A and Bt are of the same size and that, in this case, both AB and BA are square
matrices.

3. Prove that a square matrix A of order n is a scalar matrix if and only if AB 5 BA

for all square matrices B of order n.

4. Prove Theorem 2.5.2.

5. Prove Theorem 2.5.3.

6. For any scalar a, let Sa be the n 3 n scalar matrix in which all the diagonal entries
are a and other entries are 0. Prove that SaA 5 aA 5 ASa for all n 3 n matrices A.

7. Prove Theorems 2.5.4 and 2.5.5.

8. Prove Theorem 2.5.6.

9. Prove the following for any integer n  0.

n
1 1 1 n
 
(i)    
0 1 0 1

 n( n 1) 
 
1 1 0 1 n
n
  2 
  
(ii) 0 1 1  0 1 n 

0 0 1 0 0 1 
 
 
where, for any square matrix A, An is defined inductively by A0 5 I and An 5
An21 ? A for any n . 0.

10. Prove that the sum and product of two upper triangular matrices are again upper
triangular matrices and that the same statement for lower triangular matrices is
also true.

Q001-Algebra-111001_CH 02.indd 42 9/16/2011 10:08:17 AM

 Number Systems   2-43

11. A matrix A is said to be symmetric if A 5 At, and A is called skew-symmetric if

A 5 2At. Prove the following.
(i) Any symmetric or skew-symmetric matrix is a square matrix.
(ii) If A and B are symmetric matrices, then so is rA 1 sB for any scalars
r and s.
(iii) For any matrix A, A ? At and At ? A are both symmetric.
(iv) For any symmetric matrices A and B, the product AB is symmetric if and
only if AB 5 BA.

12. Prove that the diagonal entries of a skew-symmetric matrix are all zero.

13. For any square matrix A, prove that A 1 At is symmetric.

14. Prove that any square matrix can be expressed as the sum of a symmetric matrix
and a skew-symmetric matrix.

2.6 DETERMINANTS
In this section, we briefly discuss an important function known as determinant
function which maps square matrices into scalars. The term ‘determinant of A’
is conventionally used to call the value of this function at a given square matrix A.
Determinants have definite importance as a theoretical tool, besides their
effectiveness as a device for computations. For example they provide us with
simple criterion for the nonsingularity; namely, a square matrix is nonsingular
if and only if its determinant is nonzero.
There are several ways of defining the determinant function. However, we
prefer the classical definition which uses permutations. In view of this, we
first have a brief discussion on permutations. We begin with the following.

Definition 2.6.1. For any positive integer n, let In 5 {1, 2, …, n}. Any bijec-
tion of In onto itself is called a permutation on In. The set of all permutations
on In is denoted by Sn.
Any permutation f on In can expressed by means of an array (a 2 3 n matrix)

 1 2 3  n 
 
 f (1) f (2) f (3)  f ( n)

symbolising that each i is mapped to f (i). Note that the order of the columns
in this representation of f is immaterial. For example

1 2 3 4 5 3 4 2 1 5
  and  
3 4 1 5 2 1 5 4 3 2

Q001-Algebra-111001_CH 02.indd 43 9/16/2011 10:08:17 AM

 2-44   Algebra – Abstract and Modern

represent the same permutation f which is given by f (1) 5 3, f (2) 5 4, f (3) 5 1,

f (4) 5 5 and f (5) 5 2.
If f and g are permutations on In, the f o g, f21 and g21 are also permutations
on In. The set Sn of all permutations on In has the structure of a group, which
we thoroughly discuss later in chapter 6. Note that |Sn| 5 n! for any n  Z1.
A permutation f is called an r-cycle (or cycle of length r) if f maps r elements
i1, i2, …, ir cyclically, keeping the remaining elements, if any fixed and such
an r-cycle will be denoted by

f 5 (i1 i2 … ir)

That is, f (i1) 5 i2, f (i2) 5 i3, …, f (ir21) 5 ir and f (ir) 5 i1 and f (i) 5 i for all
i  In 2{i1, i2,…, ir}. Observe that (i1 i2 … ir), (i2 i3 … iri1), …, (ir i1 i2 … ir21)
are all represent the same cycle. A 2-cycle is called a transposition. Note that,
if f is an r-cycle, the f r(5 f ο f ο … ο f, r times) is the identity map on In and
r is the least such positive integer. In particular, if f is a transposition, then f 2
is the identity map and f interchanges two elements in In and keeps all other
elements fixed.
Two cycles (a1 a2 … ar) and (b1 b2 … bs) are said to be disjoint if ai  bj for
all 1  i  r and 1  j  s. It can be easily proved that f o g 5 g o f for any
disjoint cycles f and g and that any permutation on In can be expressed, in an
essentially unique way, as a product of disjoint cycles. Further, any cycle is a
product of transpositions (since (a1 a2 … ar) 5 (a1ar) o (a1ar21) o … o (a1a2))
and hence any permutation can be expressed as a product of transpositions,
although not necessarily uniquely. For example (2 4) ? (4 5) ? (1 3) 5 (1 3) o
(2 4) o (1 3) o (4 5) o (1 3). However, it can be proved (see Corollary 6.4.2)
that, if a permutation can be expressed as a product of even number of trans-
positions, then it cannot be expressed as a product of odd number of transpo-
sitions. In view of this, a permutation is called an even (odd) permutation if
it is a product of even (odd, respectively) number of permutations. If f and g
are even permutations, then clearly f o g, f21 and g21 are also even (since f 5
f1 ο f2 ο … ο fr implies that f21 5 fr 21 ο f21r21 ο … ο f221 ο f121). Note that an
r-cycle is even if and only if r is odd.

Definition 2.6.2. For any permutation f on In, the signature of f, denoted by

sgn f, is defined by

 1 if f is even
sgn f   .
1 if f is odd

It can be easily verified that, for any permutation f and g on In,

Q001-Algebra-111001_CH 02.indd 44 9/16/2011 10:08:18 AM

 Number Systems   2-45

sgn(f o g) 5 sgn f ? sgn g and sgn f 5 sgn f21.

In the following, we give a formal definition of the determinant function.

Definition 2.6.3. Let A 5 (aij) be an n 3 n matrix. Then the sum

∑ (sgn f ) a
f ∈Sn
a
1f (1) 2f (2) … anf ( n )

is called the determinant of A and is denoted by det A or |A|.

Examples 2.6.1
a a12 
1. Let A   11  be a 2 3 2 matrix. Since S2 has only two elements,
a21 a22 
namely the identity e which is an even permutation and the transposition
s 5 (1 2) which is odd, we have
det A 5 (sgn e)a11a22 1 (sgn s)a12a21
5 a11a22 2 a12a21
 a11 a12 a13 
 
2. Consider a 3 3 3 matrix A  a21 a22 a23 
 
 a31 a32 a33 
S3 has 3! elements; these are
1 2 3 1 2 3 1 2 3
e    , f   , g   
1 2 3 2 3 1 3 1 2
1 2 3 1 2 3 1 2 3
a   , b   ,   
2 1 3 1 3 2 3 2 1

Note that e is the identity, f 5 (1 2 3), g 5 (1 3 2) a 5 (1 2), b 5 (2 3) and

 5 (1 3) and hence e, f and g are even and a, b and  are odd. Therefore
det A 5 a11a22a33 1 a12a23a31 1 a13a21a322a12a21a33 2 a11a23a32 2 a13a22a31
3. The determinant of the zero matrix is 0.
4. det I 5 1, where I is the identity matrix of any order.
In the following we prove some results that facilitate the evaluation of the
determinant of any square matrix in a less tedious manner. First recall that, if
A 5 (aij) is an n 3 n matrix, R1, …, Rn are the n rows of A and C1, C2, …, Cn
are the n-columns of A, then we express A as

Q001-Algebra-111001_CH 02.indd 45 9/16/2011 10:08:18 AM

 2-46   Algebra – Abstract and Modern

 R1 
 
 R 
A =  2  or A 5 (C1, C2, …,Cn)
  
 
 Rn 

 a1j 
 

where Ri 5 (ai1 ai2 … ain) and Cj 5 a2j  .
 
  
 
anj 

Here, each Ri is a 1 3 n matrix and each Cj is a n 3 1 matrix.

Theorem 2.6.1. Let A 5 (aij) be an n 3 n matrix and R1, R2, …, Rn be the

rows of A. For any fixed i, 1  i  n, let Si 5 (bi1 bi2 … bin) be a 1 3 n matrix.
Then, the following holds.

 R1   R1   R1 
     
 R       
 2     
       Ri1 
     
1. det  Ri  Si   det  Ri  det  Si 
     
 Ri1      Ri1 
 
     
        
     
 Rn   Rn   Rn 

 R1   R1 
   
     
   
 Ri−1   Ri1 
   
2. det  aRi  = a det  Ri 
   
 Ri+1   Ri1 
 
  
     
   
 Rn   Rn 

Proof:
1. The left hand side of the equation is

∑ (sgn f )a
f ∈Sn
1f (1) … ai1f (i1) ( aif (i )  bif (i ) )ai+1f (i+1) … anf (n )

 ∑ (sgn f )a1f (1) … ai1f (i1) aif (i ) … anf (n )

f ∈Sn

Q001-Algebra-111001_CH 02.indd 46 9/16/2011 10:08:19 AM

+ (sgn f )a …a b a …a
 ∑ (sgn f )a
f ∈Sn
1f (1) … ai1f (i1) ( aif (i )  bif (i ) )ai+1f Number Systems   2-47
(i+1) … anf (n )

 ∑ (sgn f )a1f (1) … ai1f (i1) aif (i ) … anf (n )

f ∈Sn

+ ∑ (sgn f )a1f (1) … ai1f (i1) bif (i ) ai+1f (i+1) … anf (n )

f ∈Sn

5The right hand side of (1)

2. This is clear from the definition.

Theorem 2.6.2. For any square matrix A, det A 5 det At

Proof: Let A 5 (aij) be an n 3 n matrix. Then, At 5 (aji) and

det At  ∑ (sgn f ) a f (1)1a f (2)2 … a f (n )n

f ∈Sn

 ∑ (sgn f )a1f 1 (1) a2f 1 (2) … anf 1 (n )

f ∈Sn

 ∑ (sgn f 1 )a1f 1 (1) a2f 1 (2) … anf 1 (n )

f ∈Sn

5det A (since Sn 5 {f21: f  Sn}).

Theorem 2.6.3. If two rows of a square matrix A are equal, then det A 5 0.

Proof: Let A 5 (aij) be an n 3 n matrix and R1, R2, …, Rn be the n rows of

A. Suppose that the rth row and sth row are equal; that is, Rr 5 Rs and r  s.
Without loss of generality, we can assume that r , s. Let g be the transposi-
tion (r, s). Let

A 5{f  Sn : f (r) , f (s)} and B 5 {f  Sn : f (r) . f (s)}.

Then, the map a : A → B defined by a( f ) 5 f o g is a bijection. Also, note that

A ∪ B 5 Sn (since r  s, f (r)  f (s) for any f  Sn) and A ∩ B 5 . Therefore,

det A  ∑ (sgn f )a1f (1) … arf (r ) … asf (s ) … anf (n )

f ∈A

 ∑ (sgn f )a1f (1) … arf (r ) … asf (s ) … anf (n )

f ∈B

 ∑ (sgn f )a1f (1) … arf (r ) … asf (s ) … anf (n )

f ∈A

 ∑ (sgn f o g )a1fg (1) … arfg (r ) … asfg (s ) … anfg (n )

f ∈A

 ∑ (sgn f )[a1f (1) … arf (r ) … asf (s ) … anf (n ) − a1fg (1) … arfg (r ) … asfg (s ) … anfg (n ) ]
f ∈A

Q001-Algebra-111001_CH 02.indd 47 9/16/2011 10:08:19 AM

 (sgn f )[a …a …a …a −a …a …a …a ]
 det A  ∑ (sgn f )a1f (1) … arf (r ) … asf (s ) … anf (n )
f ∈A

 ∑ (sgn f )a1f (1) … arf (r ) … asf (s ) … anf (n )

f ∈B

 ∑ (sgn f )a1f (1) … arf (r ) … asf (s ) … anf (n )

f ∈A
2-48   Algebra – Abstract and Modern
 ∑ (sgn f o g )a1fg (1) … arfg (r ) … asfg (s ) … anfg (n )
f ∈A

 ∑ (sgn f )[a1f (1) … arf (r ) … asf (s ) … anf (n ) − a1fg (1) … arfg (r ) … asfg (s ) … anfg (n ) ]
f ∈A

 ∑ (sgn f )[a1f (1) … arf (r ) … asf (s ) … anf (n ) − a1f (1) … arf (s ) … asf (r ) … anf (n ) ]
f ∈A

5 0 (since arj 5 asj for all 1  j  n).

Theorem 2.6.4. Let A 5 (aij) be an n 3 n matrix and B be the matrix obtained

from A by interchanging the rth row and the sth row, then det A 5 2det B.
 R1 
 
Proof: We can assume that r , s. We have A    , where Ri is the ith row of A.
 
 R1   Rn 
 
  
 
 Rr  Rs 
 
Put C    . Then, the rth row of C 5 sth row of C and all other
 
 Rs + Rr 
 
  
 
 Rn 
ith rows of C are same as those of A. Now, by Theorem 2.6.3, we have
 R1   R1   R1   R1 
       
           
          
 Rr   Rr   Rs   Rs 
       
0  det C  det    det    det    det   
       
 Rs   Rr   Rs   Rr 
       
           
       
R
 n R
 n R
 n  Rn 
5 det A 1 0 1 0 1 det B
Thus, det A 5 2det B.
The following is an immediate consequence of Theorems 2.6.2 and 2.6.4.

Corollary 2.6.1. Let A be an n 3 n matrix and B be the matrix obtained from

A by interchanging two columns of A. Then, det A 5 2det B.

Corollary 2.6.2. If R1, R2, …, Rn are the rows of an n 3 n matrix A and f is a

permutation on {1, 2, …, n}, then

Q001-Algebra-111001_CH 02.indd 48 9/16/2011 10:08:20 AM

 Number Systems   2-49

 R f (1) 
 
det     (sgn f ) det A
 
 R f (n ) 

Proof: Let f be a product of m transpositions. Then the matrix A can be trans-

 R f (1) 
 
formed to    by m interchanges of the rows. Therefore, by Theorem 2.6.4,
 
 R 
 f (n ) 

 R f (1)   R1 
   
 
det     (1) det     (sgn f ) det A.
m

   
 R f (n )   Rn 

The following is one of the most important properties of the determinants of

matrices.

Theorem 2.6.5. For any n 3 n matrices A and B,

det(A ? B) 5 det A ? det B

Proof: Let A 5 (aij) and B 5 (bij) be two n 3 n matrices and AB 5 (cij). Then,
n
cij   air brj for any 1  i, j  n.
r1

det ( AB)  ∑ (sgn f ) c1f (1) c2f (2) … cnf (n )

f ∈Sn

 n   n 
 ∑ (sgn f ) ∑ a1r1 br1 f (1) …∑ anrn brn f ( n ) 
f ∈Sn  r11   r =1 
n

 
 ∑ ( a1r1 a2r2 … anrn )  ∑ (sgn f ) br1 f (1) br2 f (2) … brn f ( n ) 
1 ≤ r1 , r2 ,…, rn ≤ n  f ∈Sn 

In the above summation, if r1, r2, …, rn are not all distinct, then, by
­Theorem 2.6.3,

∑ (sgn f ) b
f ∈Sn
r1 f (1) br2 f (2) … brn f (n )  0

Q001-Algebra-111001_CH 02.indd 49 9/16/2011 10:08:21 AM

 2-50   Algebra – Abstract and Modern

and therefore, we can consider only the summands corresponding to distinct

n-tuples r1, r2, …, rn. Therefore,

 
det(AB)  ∑ ( a1g (1) a2g (2) … ang ( n ) )  ∑ (sgn f ) bg (1) f (1) … bg ( n ) f ( n ) 
g ∈ Sn  f ∈Sn 

 
  ∑ (sgn g ) a1g (1) … ang ( n )  (det B) (by Theorem 2.6.10)
 g ∈Sn 

5 (det A) (det B).

Corollary 2.6.3. If A is a nonsingular n 3 n matrix, then det A  0 and

1
det A1 
det A

Proof: If AA21 5 I (the identity n 3 n matrix), then

det A ? det A21 5 det (AA21) 5 det (I) 5 1.

Next, we discuss an expansion for the determinant of a matrix which pro-

vides us with an inductive algorithm to find the value of det A. First, we have
the following.

Definition 2.6.4. Let A 5 (aij) be an n 3 n matrix and, for any 1  i, j  n, let

Aij  ∑ (sgn f ) ∏ a
f ∈Sn r ≠i
r f (r )

f (i ) j

 ∑ (sgn f ) a
f ∈Sn
1f (1) … ai-1f (i-1) ⋅ ai+1f (i+1) … anf (n ) .
f (i )= j

Then, Aij is called the cofactor of aij in det A.

Theorem 2.6.6. The following holds for any n 3 n matrix A 5 (aij).

n det A if r  i
1. ∑a Aij  
j =1
rj
 0 if r  i
n det A if r  j
2. ∑ air Aij  
i1
 0 if r  j

Q001-Algebra-111001_CH 02.indd 50 9/16/2011 10:08:21 AM

 Number Systems   2-51

Proof: Consider all the summands in the sum

det A  ∑ (sgn f ) a1f (1) … a2f (2) … anf (n )

f ∈S n

that contain a given entry aij as a factor. These are corresponding to those
permutations f for which f (i) 5 j. Therefore, the sum of all the summands in
the summation for det A involving aij as a factor is

∑ (sgn f ) a
f ∈S n
1f (1) … a2f (2)  anf (n ) = aij Aij
f (i ) j

and hence

det A  ∑ ( sgn f ) a1f (1) a2f (2) … anf (n )

f ∈Sn

n
∑ ∑ ( sgn f ) a 1f (1) a2f (2) … anf (n )
j1 f ∈Sn
f (i ) = j

n
 ∑ aij Aij for each 1  i  n.
j1

n
Similarly, det A   aij Aij for each 1  j  n.
i1
Next, let 1  r  i  n and consider

n n

∑a
j1
rj Aij  ∑ arj
j1
∑ (sgnf) ∏ a
f ∈ Sn s ≠i
s f (s )

f (i ) j

 ∑ ( sgn f ) a1f (1) … ai1f (i1) arf (i ) ai1f (i1) … anf (n )

f ∈ Sn

 0 (by Theorem 2.6.7)

n
Similarly  air Aij if r  j.
i1
n
The equation  aij Aij  det A given in (1) above is called the expansion of det
j1 n
A with respect to the ith row and the equation  aij Aij  det A given in (2) above
i1
is called the expansion of det A with respect to the jth column. The following
result provides a method to evaluate the cofactors Aij for the matrix A 5 (aij).

Q001-Algebra-111001_CH 02.indd 51 9/16/2011 10:08:22 AM

 2-52   Algebra – Abstract and Modern

Theorem 2.6.7. Let A 5 (aij) be an n 3 n matrix and, for any 1  i, j  n, let

Aij be the submatrix of A obtained by deleting the ith row and jth column. Then,
the cofactor Aij of aij in det A is given by

Aij 5 (21)i 1 j det Aij

Proof: Let us first find the cofactor A11 of a11 in det A. By Definition 2.6.4,
we have

A11  ∑
f ∈Sn
( sgn f ) a2f (2) a3f (3) … anf (n ) .
f (1) 1

If f  Sn and f (1) 5 1, then the restriction of f to {2, 3, …, n} is a permuta-

tion on {2, 3, …, n} and any permutation on {2, 3, …, n} can be uniquely
extended to a permutation f on {2, 3, …, n} by defining f (1) 5 1. Also, for f
 Sn with f (1) 5 1, f is an even permutation if and only if the restriction of f
to {2, 3, …, n} is even. Thus, the above equation is precisely same as

A11 5 det A11 5 (21) 111 det A11.

Next, to find the value of a general cofactor Aij of aij, let us bring aij to the (1, 1)
position by performing some row and column interchanges on A. To bring aij
to the (1, 1) position, we move the jth column to the left to j21th column (that
is, interchanging jth column and j21th column), then to j22th column, …,
to 1st column, so that after j21 interchanges of the columns, the jth column
becomes the first column. Next, in a similar way, we move the ith row up to
the 1st row in i21 interchanges of the rows. Now, we have a matrix B that
is obtained from A by j21 interchanges of columns and i21 interchanges
of rows. Therefore, by Theorem 2.6.4 and Corollary 2.6.1, the determinant
of the new matrix B is (21)(j21)1(i21) det A 5 (21)ij det A. If B 5 (bij), then
b11 5 aij and the matrix obtained by deleting the 1st row and 1st column in B is
precisely Aij. Thus,

Aij 5 (21)i1j det Aij.

Corollary 2.6.4. For any n 3 n matrix A 5 (aij),

n
det A  ∑ ( 1) i j
aij det Aij , for each 1 ≤ i ≤ n
j1

n
 ∑ ( 1) i j
aij det Aij , for each 1 ≤ j ≤ n.
i1

Q001-Algebra-111001_CH 02.indd 52 9/16/2011 10:08:23 AM

 Number Systems   2-53

Example 2.6.2. Let us determine the determinant of a 3 3 3 matrix A 5 (aij),

using Theorem 2.6.7 and Corollary 2.6.4. We have

a a23  a a23  a a22 

A11   22 , A12   21  and A13   21 .
 a32 a33   a31 a33   a31 a32 

Now,

det A 5 (21)111 a11 det A111 (21)112 a12 det A121 (21)113 a13 det A13
a22 a23 a a23 a a22
 a11  a12 21  a13 21
a32 a33 a31 a33 a31 a32
 a11 ( a22 a33  a32 a23 )  a12 ( a21 a33  a31 a23 )  a13 ( a21 a32  a31 a22 ).

This is the expansion of det A with respect to the 1st row. Notice that we get
the same value for det A by expanding it with respect to any other row or any
column.
The following result characterizes nonsingular matrices in terms of the
value of their determinants.

Theorem 2.6.8. A square matrix A is nonsingular if and only if det A  0

and, in this case, the inverse of A is given by

1
A21 5 (A )t
det A ij
where (Aij) is the matrix whose ijth entry is the cofactor Aij of the ijth entry in A.

Proof: Let A 5 (aij) be an n 3 n matrix and Aij be the cofactor of aij in det A.
Let B be the transpose of (Aij).
That is, B 5 (Aij)t 5 (bij); say.
Then, bij 5 Aji for all i and j. By Theorem 2.6.6,

n n

∑a
j1
rj b ji  ∑ arj Aij  ri det A
j1

where dri 5 1 or 0 according as r 5 i or r  i. This implies that, the rith entry

in the product matrix AB is det A if r 5 i, and 0 if r  i. Therefore,

AB 5 det A?I

Q001-Algebra-111001_CH 02.indd 53 9/16/2011 10:08:23 AM

 2-54   Algebra – Abstract and Modern

where I is the n 3 n identity matrix. Similarly, BA 5 det A?I. Thus, det A  0

implies that

 1   1 
A   B  I  
 det A   det A 
B A

and hence A is nonsingular and A1  1 B  1 ( Aij )t . Converse follows

det A det A
from Corollary 2.6.3.

Definition 2.6.5. For any n 3 n matrix A 5 (aij), the transpose of the matrix
(Aij) is called adjoint of A and is denoted by adj A, where Aij is the cofactor
of aij in det A.

Corollary 2.6.5. For any square matrix A,

A ? (adj A) 5 det A ? I 5 (adj A) ? A.

EXERCISE 2(f)

1. Evaluate the determinants of each of the following matrices

2 4 1 
 
(i)  3 1 2 
 
4 3 2
2 1 3 4
 
1 4 2 3
(ii)  
3 4 1 2
 
4 3 2 1
4 3 2 1 
 
 1 2 3 4 
(iii)  
 2 3 1 4
 
3 4 4 1
2. Prove that the determinant of an upper (or a lower) triangular matrix is equal to
the product of the diagonal entries.
1 a a 2 
 
3. Prove that det 1 b b 2   ( a  b)(b  c)(c  a).
 2 
1 c c 

Q001-Algebra-111001_CH 02.indd 54 9/16/2011 10:08:24 AM

 Number Systems   2-55

1 x1 x12  x1n-1 
 
1 x2 x22  x2n-1 
4. Prove that det    ∏ ( x j − xi ).
  1 ≤ i ≤ j ≤ n
 
i xn xn2  xnn-1 

1 a a3 
 
5. Prove that det 1 b b3   (b  c)(c  a)( a  b)( a  b  c).
 
1 c c 3 
6. If A is a nonsingular square matrix such that A2 5 A, then prove that det A 5 1.

7. Prove the following:

(b  c) 2 a2 a 2 
 
(i) det  b 2 ( c + a) 2 b 2   2 abc ( a  b  c)3
 
 c 2 c2 ( a  b) 2 

a  b  c 2b 2c 
 
(ii) det  2a bc a 2c   ( a  b  c)3
 
 2a 2b c  a  b

8. If A is a n 3 n matrix such that Am 5 On3n for some m  Z1, then prove that A
is singular.

9. Let A 5 (aij) and B 5 (bij) be n 3 n matrices such that bij 5 (21)ij aij for all i
and j. Then, prove that det A 5 det B.

10. Prove that a square matrix A is nonsingular if and only if At is nonsingular.

If A is an n 3 n skew-symmetric matrix, then prove that det A 5 (21)n det A?

11. If n is odd, prove that any skew-symmetric n 3 n matrix is singular.

12. Prove that for any n 3 n nonsingular matrix A, det (adj A) 5 (det A)n21.

13. Prove the following for any n 3 n matrices A and B:

(i) det (AB) 5 det (BA)
(ii) det (A ? At) 5 (det A)2
(iii) If A is nonsingular, then det (ABA21) 5 det B.

Q001-Algebra-111001_CH 02.indd 55 9/16/2011 10:08:25 AM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 02.indd 56 9/16/2011 10:08:25 AM

 Part II
Group Theory

Q001-Algebra-111001_CH 03.indd 1 9/16/2011 10:07:12 AM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 03.indd 2 9/16/2011 10:07:12 AM

 3
Groups
3.1 Binary Systems
3.2 Groups
3.3 Elementary Properties of Groups
3.4 Finite Groups and Group Tables

3.1 BINARY SYSTEMS

It is well known that the product of two integers is again an integer. That is, if
a and b are integers, then the product a ? b is again an integer. Here, the sym-
bol ‘?’ denotes the ‘operation of taking product’ of a and b, in this order. Sim-
ilarly, if A and B are two 2 3 2 matrices over the real number system, then the
product A ? B is again a 2 3 2 matrix over R. Here, the symbol ‘?’ denotes the
‘operation of taking product’ of the matrices A and B, in this order. Further,
if f and g are two mappings of a set X into itself, then the composition g o f is
also a mapping of X into itself. Here, the symbol ‘o’ denotes the ‘operation of
taking composition’ of g and f, in this order. Also, if A and B are subsets of a
given set X, then the union A ∪ B is also a subset of X. Here again, the symbol
∪ denotes the ‘operation of taking union’ of A and B, in this order.
In each of these cases, from any two elements of a given set, we obtain
another element of the same set by performing an operation on the two ele-
ments in a specific order. This is formalized in the following definition.

Definition 3.1.1. Let S be a nonempty set and S 3 S be the set of all ordered
pairs of elements of S. That is,

S 3 S 5 {(a, b) : a  S and b  S}.

A mapping f : S 3 S → S is called a binary operation on S.

Q001-Algebra-111001_CH 03.indd 3 9/16/2011 10:07:13 AM

 3-4   Algebra – Abstract and Modern

If f is a binary operation on a set S and a and b are elements of S, then we

write a f b for f (a, b). This is only for convenience. Recall that we write a ? b
for the product of two integers a and b. By definition, for any set S, any map-
ping of S 3 S into S is a binary operation on S. Certain special binary opera-
tions on certain special sets are important to be mentioned. In the following,
we list several binary operations on certain special sets, as examples.

Example 3.1.1
1. The usual multiplication ‘?’ is a binary operation on the set Z of integers.
Quite often, we simply write ab for a ? b.
2. The usual addition 1 is a binary operation on the set Z of integers.
3. Let us define the mapping 2 : Z 3 Z → Z by 2 (a, b) 5 a 2 b, the
usual difference of b with a, for any integers a and b. Then, 2 is a binary
operation on Z. Note that 2 is not a binary operation on the set Z1 of
positive integers, since a 2 b need not be positive for any two positive
integers a and b. Likewise, 2 is not a binary operation on the set Z2 of
negative integers. Note that both the multiplication and addition given in
(1) and (2), respectively are binary operations on Z1. The operation 2
is called the difference operation. Note that each of addition, difference
and multiplication is a binary operation on the set Q of rational numbers
and on the set R of real numbers.
4. Let R be the set of real numbers and, for any real numbers a and b,
define
a ∧ b 5 The minimum of a and b
and a ∨ b 5 The maximum of a and b.
Then, both ∧ and ∨ are binary operations on R. In fact, these are binary
operations on any nonempty subset of R and, in particular, on Q, Z and Z+.
5. Consider the set Z+ of positive integers. For any a and b in Z+, define
a g b 5 (a, b), the greatest common divisor of a and b
and  a  b 5 [a, b], the least common multiple of a and b.
Then, g and  are both binary operations of Z+. Usually, we write (a, b)
and [a, b] to denote respectively the greatest common divisor and the
least common multiple of any positive integers a and b.
6. Let X be any set and P(X), the power set of X; that is, P(X) is the set of
all subsets of X. For any A and B  P(X), define
A ∩ B 5 {x : x  A and x  B}
A ∪ B 5 {x : x  A or x  B}
A 2 B 5 {x : x  A and x  B}
A 1 B 5 (A 2 B) ∪ (B 2 A).

Q001-Algebra-111001_CH 03.indd 4 9/16/2011 10:07:13 AM

 Groups   3-5

Then, ∩, ∪, 2 and 1 are all binary operations on the set P(X) and are
respectively called intersection, union, difference and symmetric differ-
ence. Note that P(X) is not empty even if X is empty.
7. For any positive integers m and n, let Mm3n(R) be the set of all m 3 n
matrices over the real number system R. For any A 5 (aij) and B 5 (bij)
in Mm3n(R), define
A 1 B 5 (cij),

where cij 5 aij 1 bij and 1 is the usual addition of real numbers. Then, 1
is a binary operation on Mm3n(R) and is called the addition of matrices
(of same order).
8. For any positive integer n, let Mn(R) be the set of all n 3 n matrices
(square matrices of order n). For any A 5 (aij) and B 5 (bij) in Mn(R),
define
n
A ? B 5 (dij), where dij 5 ∑ aik bkj
k =1
for any 1 # i, j # n. That is,
dij 5 ai1blj 1 ai2b2j 1 … 1 ainbnj.

(The operations involved in defining dij above are the usual addition and
multiplication of real numbers.) Then, ‘?’ is a binary operation on Mn(R)
and is called the multiplication of square matrices (of the same order).
Note that, we can define addition and multiplication (as in (7) and (8)
above) on the sets Mm3n(Z) and Mn(Z) of matrices over the set Z of
integers or on the sets Mm3n(Q) and Mn(Q) of matrices over the set Q of
rational numbers.
9. Let X be any set and M(X) be the set of all mappings of X into itself. For
any mappings f and g in M(X), define
f o g : X → X by (f o g)(x) 5 f (g(x)) for any x  X.

Then, o is a binary operation on M(X) and is called the composition of

mappings.
10. Let C be the set of complex numbers; that is C is the set of expressions
of the form a 1 bi, where a and b are real numbers.
C 5 {a 1 bi : a and b are real numbers}.

For any a 1 bi and c 1 di in C, define

(a 1 bi) 1 (c 1 di) 5 (a 1 c) 1 (b 1 d)i
and     (a 1 bi) ? (c 1 di) 5 (ac 2 bd) 1 (ad 1 bc)i.

Q001-Algebra-111001_CH 03.indd 5 9/16/2011 10:07:13 AM

 3-6   Algebra – Abstract and Modern

Then, 1 and ‘?’ are binary operations on C and are called the usual addi-
tion and multiplication of complex numbers, respectively.
11. Let n be a positive integer and
Zn 5 {0, 1, 2, …, n 2 1}.

For any a and b in Zn, define

the usual sum a  b, if a  b  n
a n b   .
 a  b  n, if a  b  n

Then, 1n is a binary operation on Zn and is called the addition modulo n.

12. Let Zn be the set considered above. For any a and b in Zn, define
a ?n b 5 r,

where r is the remainder obtained by dividing the usual product ab with

n; that is r is the integer such that
ab 5 qn 1 r, q and r  Z and 0 # r , n.

For example, 7 ?8 6 5 2, 5 ?10 8 5 0 and 8 ?9 8 5 1. Then, ?n is a binary

operation on Zn and is called the multiplication modulo n. Note that, in
the example given in (11) above, for any a and b in Zn, a 1n b can also
be viewed as the remainder obtained by dividing the usual sum a 1 b
with n, since 0 # a 1 b , 2n. a 1n b and a ?n b are respectively called
the sum and product of a and b modulo n.
13. Let n be any integer and define, for any integers a and b,
a * b 5 a 1 b 1 n.
Then, * is a binary operation on the set Z of integers.
14. Let E be the set of all English words (whether meaningfull or not); that
is, E is the set of all finite sequences
a1 a2 … an, n . 0 and ai’s are in alphabet of English.
For any a 5 a1 a2 … an and b 5 b1 b2 … bm, define
a * b 5 a1bm.
Then, * is a binary operation on E. Note that a * b is the two letter word
consisting of the first letter of a followed by the last letter of b.
15. Let X be any nonempty set and RX be the set of all mappings from X into
R. For any f and g in RX, define f 1 g and f ? g : X → R by
(f 1 g)(x) 5 f (x) 1g(x)
       and (f ? g)(x) 5 f (x) ? g(x), for all x  X.

Q001-Algebra-111001_CH 03.indd 6 9/16/2011 10:07:13 AM

 Groups   3-7

Note that the 1 and ? on the right hand sides of the above are the usual
addition and multiplication in the real number system R. Then, 1 and ?
are binary operations on RX and are respectively called the point-wise
addition and point-wise multiplication.
16. The above example can be generalized as follows. Let * be a binary oper-
ation on a nonempty set S and X be a nonempty set. Let SX be the set of
all mappings of X into S. For any f and g in SX, define f * g : X → S by
(f * g)(x) 5 f (x) * g(x) for all x  X.
Then, * is a binary operation on SX and is called the point-wise operation
on SX with respect to the operation * on S.
Note that, in (16) above (and so is in (15)), we have denoted the operations
on SX and in S with the same symbol *. There should not be any confusion.
The * on the left sides is the one we are defining on SX and that on the right
sides is the given binary operation on S.

Note 3.1.1. In defining a binary operation on a set S, one should observe the
following:
(i) For each ordered pair of elements in S, the element assigned to it must
be again an element of S.
(ii) Exactly one element of S must be assigned to each ordered pair of
elements in S.
For example, consider the set R of real numbers and, for any a and b in R,
define a * b  a . Then, * is not a binary operation on R, since * is not defined
b
for all ordered pairs of elements in R. Note that 2 * 0 is not defined, while 0
* 2 is defined. However, this * is a binary operation on a smallest set, namely,
the set R 2 {0} of nonzero real numbers.
Let us consider another example. Let S be the set of all people in a par-
ticular village and define, for any a and b in S, a * b 5 c where c is a person
whose height is equal to the minimum of the heights of a and b. Then, * is
not a binary operation on S, since a * b may not be an unique element in S;
there can be more than one person in the village whose height is equal to the
minimum of those of a and b.

Note 3.1.2. Let S be a finite set with n elements. Then, the number of elements
in S 3 S is n2. Any binary operation S is simply a mapping
2
of S 3 S into S; that is,
an element of Ss3s. Therefore, there are exactly n n many binary operations on S.

Definition 3.1.2. A pair (S, *) is said to be a binary system if S is a nonempty

set and * is a binary operation on S. Here, S is called the underlying set in the
binary system (S, *).

Q001-Algebra-111001_CH 03.indd 7 9/16/2011 10:07:14 AM

 3-8   Algebra – Abstract and Modern

Definition 3.1.3. A binary system (S, *) is said to be finite if the underlying

set S is finite.
A finite binary system (S, *) can be represented by means of a table as
detailed below. Let S 5 {a1, a2, …, an}. These elements a1, a2, …, an are to be
listed across the top of the table and at the left of the table, both in the same
order. The element ai * aj is written in the ith row and jth column as given in the
table given below and * is to written on the extreme left of the top.

* a1 a2 a3 … aj … an

a1 a1 * a1 a1 * a2 a1 * a3 …a1 * aj … a1 * an

a2 a2 * a1 a2 * a2 a2 * a3 … a2 * aj… a2 * an

a3 a3 * a1 a3 * a2 a3 * a3 … a3 * aj … a3 * an

       
ai ai * a1 ai * a2 ai * a3 ai * a j  ai * an
       

an an * a1 an* a2 an * a3 …an * aj … an * an

In the example given below, we shall construct the table representing the
binary system (Z9, 19) where Z9 5 {0, 1, 2, …, 8} and 19 is the addition
modulo 9 (see Example 3.1.1 (11)).

Example 3.1.2. Let Z9 5 {0, 1, 2, …, 8} 5 {a  Z : 0 # a , 9} and 19 the

addition modulo 9. 19 is the binary operation on Z9 defined by

the usual sum a  b, if a  b  9

a 9 b   .
 a  b  9, if a  b  9

for any a and b in Z9. The following table represents the binary system (Z9, 19).

19 0 1 2 3 4 5 6 7 8
0 0 1 2 3 4 5 6 7 8
1 1 2 3 4 5 6 7 8 0
2 2 3 4 5 6 7 8 0 1
3 3 4 5 6 7 8 0 1 2
4 4 5 6 7 8 0 1 2 3

Q001-Algebra-111001_CH 03.indd 8 9/16/2011 10:07:15 AM

 Groups   3-9

5 5 6 7 8 0 1 2 3 4
6 6 7 8 0 1 2 3 4 5
7 7 8 0 1 2 3 4 5 6
8 8 0 1 2 3 4 5 6 7

Here, 2 19 7 5 0, 2 19 5 5 7, 4 19 5 5 0, 5 19 7 5 3, 6 19 8 5 5 , 8 19
8 5 7, etc.

Example 3.1.3. Let S be the set of all positive divisors of 36 and, for any a
and b in S, define

a g b 5 (a, b), the greatest common divisor (GCD) of a and b

(see Example 3.1.1 (5)). Then, (S, g) is a binary system which is represented
by the table given below. We have

S 5 {1, 2, 3, 4, 6, 9, 12 ,18 ,36}

g 1 2 3 4 6 9 12 18 36
1 1 1 1 1 1 1 1 1 1
2 1 2 1 2 2 1 2 2 2
3 1 1 3 1 3 3 3 3 3
4 1 2 1 4 2 1 4 2 4
6 1 2 3 2 6 3 6 6 6
9 1 1 3 1 3 9 3 9 9
12 1 2 3 4 6 3 12 6 12
18 1 2 3 2 6 9 6 18 18
36 1 2 3 4 6 9 12 18 36

Let (S, *) be a binary system. For any elements a, b and c in S, the expres-
sion a * b * c has no meaning, since * is a binary operation and hence * is
defined for pairs of elements. For example, 1 2 2 2 3 has no meaning and we
should specify whether it is (1 2 2) 2 3 (this is what we usually take) or 1 2
(2 2 3). Note that (1 2 2) 2 3 ≠ 1 2 (2 2 3). For arbitrary elements a, b and
c in a binary system (S, *), a * b and c are two elements in S and hence (a * b)
* c is defined and so is a * (b * c). In general, (a * b) * c and a * (b * c) may
be different. In this context, we have the following definition.

Q001-Algebra-111001_CH 03.indd 9 9/16/2011 10:07:15 AM

 3-10   Algebra – Abstract and Modern

Definition 3.1.4. A binary operation * on a set S is said to be associative if

(a * b) * c 5 a * (b * c)

for all elements a, b and c in S.

Definition 3.1.5. A pair (S, *) is said to be a semigroup if S is a nonempty set

and * is an associative binary operation on S.

Example 3.1.4. The binary operations given in Example 3.1.1, except those
in (3), (6) and (16), are all associative and therefore, these together with the
corresponding underlying sets, are semigroups. (Z, 2) is not a semigroup,
since 2 is not associative. For any set X, ( P(X), ∩), (P(X), ∪) and (P(X), 1),
given in Example 3.1.1 (6), are all semigroups. However, (P(X), 2) is not a
semigroup; For, consider X 5 {a, b, c, d}, A 5 {a, b, c}, B 5 {c, d} and c 5
{b}. Then, A, B and C  P(X) and

(A 2 B) 2 C 5 ({a, b, c} 2 {c, d}) 2 {b} 5 {a, b} 2 {b} 5 {a}

and
A 2 (B 2 C) 5 {a, b, c} 2 ({c, d} 2 {b}) 5 {a, b, c} 2 {c, d} 5 {a, b}

and therefore (A 2 B) 2 C ≠ A 2 (B 2 C), so that 2 is not associative. Note

that the operation * on SX given in Example 3.1.1 (16) is associative if and
only if the operation * on S is associative.
The associativity of a binary operation involves three arbitrary elements in
the underlying set. If we take four elements a, b, c and d in a binary system
(S, *), then we get several expressions involving * and the elements a, b, c and
d, in this order. These are given below.

(a * b) * (c * d)
a * (b * (c *d))
a * ((b *c) * d)
((a * b) * c) * d
(a * (b * c)) * d.

One can easily prove that these expressions represent one single element, if *
is associative. In fact, we can generalize and extend the associativity for any
finite sequence of elements. First, let us have the following definition.

Definition 3.1.6. Let (S, *) be a binary system and a1, a 2, …, an  S. A

meaningful expression involving * and a1, a2, …, an, in this order, is called a

Q001-Algebra-111001_CH 03.indd 10 9/16/2011 10:07:15 AM

 Groups   3-11

meaningful product of a1, a2, …, an, in this order. The one given in the follow-
ing definition is a meaningful product.

Definition 3.1.7. Let (S, *) be a binary system and a1, a2, …, an  S. The
standard product in=1ai of a1, a2, …, an, in this order, is defined inductively
as follows.

 a1 if n  1
n
 n1
∏ ai    
∏ ai  * an if n  1
.
 i1 
i =1

For example,

∏a  a * a
i =1
i 1 2

3  2 
∏ a  ∏ a  * a
i =1
i
i =1
i 3  ( a1 * a2 ) * a3

∏ a  (((a * a ) * a ) * a ) * a .
i =1
i 1 2 3 4 5

Theorem 3.1.1 (Generalised Associative Law). Let (S, *) be a semigroup and

a1, a2, ..., an elements of S. Then, all meaningful products of a1, a2, …, an, in
this order, are equal to each other.

Proof: We shall use induction on n to prove that each meaningful product

of a1, a2, …, an is equal to their standard product. If n 5 1 or 2, the theorem
is trivial. Suppose that n . 2 and assume that any meaningful product of b1,
b2, ..., bm, with m , n, is equal to the standard product of b1, b2, ..., bm, in this
order. Let x be any meaningful product of a1, a2, …, an, in this order. Then,
there exists r such that 1 # r , n and

x5s*t

where s and t are meaningful products of a1, a2, …, ar and ar+1, ar+2, ..., an, in
these orders, respectively. By the induction hypothesis, we get that

r n
s  ∏ ai and t  a j
i1 jr1

Q001-Algebra-111001_CH 03.indd 11 9/16/2011 10:07:16 AM

 3-12   Algebra – Abstract and Modern

Now, x  s * t
 r   n 
 ∏ ai  *  ∏ a j 
 i =1   jr1 

 r   n1  
 ∏ ai  *  ∏ a j  * an 
 i1   jr1  

 r   n − 1 
 ∏ ai  *  ∏ a j  * an (since * is associative)

 i1   j = r + 1 
 n1 
 ∏ ai  * an (by the induction hypothesis)
 i1

n
 ∏ ai
i1

Thus, x is the standard product of a1, a2, ..., an, in this order. b

Definition 3.1.8. A binary operation * on a set S is said to be commutative if

a * b 5 b * a for all a and b in S.

Example 3.1.5. Except in (3), (6), (8), (9), (14) and (16), all other binary oper-
ations given in Example 3.1.1 are commutative. The operations ∩, ∪ and 1 on
P(X), given in Example 3.1.1 (6), are all commutative and the operation 2 on
P(X) is not commutative. Also, the operation * on SX, given in Example 3.1.1
(16) is commutative if and only if the operation * on S is commutative.

Theorem 3.1.2 (Generalised Commutative Law). Let * be a commutative

and associative binary operation on a set S and a1, a2, …, an  S. Then, for
any permutation s of {1, 2, …, n}, any meaningful product of a1, a2, …, an is
equal to any meaningful product of as(1), as(2), …, as(n).

Proof: Since * is associative, it is enough to prove that

n n

∏a  ∏a
i1
i
i1
s (i )

for any permutation s of {1, 2, …, n} (by Theorem 3.1.1). We shall prove theo-
rem using induction on n. If n 5 1, the theorem is trivial, if n 5 2, the theorem
follows from the commutativity of *. Let n . 2 and assume that the theorem is

Q001-Algebra-111001_CH 03.indd 12 9/16/2011 10:07:17 AM

 Groups   3-13

true for any n 2 1 elements in S. Let s be a permutation of {1, 2, …, n}. That

is, s is a bijection of {1, 2, …, n} onto itself. Let s(n) 5 k. Then, we have

{s (1), s (2), …, s ( n 1)}  {1, 2, …, k 1, k 1, k  2, …, n}.

Consider the standard product

n  n1 
∏a  ∏ as ( i )  * as ( n )
 i1 
s (i )
i1

 k1   n 
 ∏ ai  *  ∏ ai  * ak (by induction hypothesis)
 i1  ik1 
 k1   n 
 ∏ ai  * ak * ∏ ai  (by associativity and commutativity of *)
 i1   ik 1 

n
 ∏ ai b
i1 

Note that the operation * in a finite binary system (S, *) represented by

the corresponding table is commutative if and only if the entries in the table
are symmetric with respect to the diagonal that starts at the upper left corner
of the table and terminates at the lower right corner (let us call this left–right
diagonal). There is no such single technique to check the associativity of a
given binary operation.

Worked Exercise 3.1.1. Determine the number of commutative binary opera-

tions on a set with n elements.

Answer: Let S be a set with n elements. S 3 S has n2 elements. A binary

operation
2
on S is just a mapping of S 3 S into S. Therefore, there are exactly
n n number of binary operations on S. As mentioned above, a binary opera-
tion is commutative if and only if the entries in the corresponding table are
symmetric with respect to the left–right diagonal. The number of pairs (a, b)
in S 3 S with a ≠ b, is n2 2n. If * is a commutative binary operation on S,
then a * b 5 b * a for all a, b  S and hence, we can consider only half of the
number of pairs (a, b) with a  b. These together with the pair (a, a) , a  S,
constitute a set X consisting of

n2  n  n2  n 
 n 
2  2 

Q001-Algebra-111001_CH 03.indd 13 9/16/2011 10:07:17 AM

 3-14   Algebra – Abstract and Modern

elements. A commutative binary operation on S can be identified with a map-

ping of X into S. Therefore, the number of commutative binary operations on
S is equal to

n2 n
| S X | n 2
.

Worked Exercise 3.1.2. Let X be a nonempty set and M(X) be the set of all
mappings of X into itself. Let o be the composition of mappings on M(X).
Then prove that o is a commutative operation on M(X) if and only if X has
exactly one element.

Answer: If X has exactly one element, then M(X) also has only one element
and the result is trivial. Conversely suppose that X has more than one element,
choose a ≠ b  X and define f and g : X → X by

f (x) 5 a for all x  X

and     g(a) 5 b, g(b) 5 a and g(x) 5 x for all x  {a, b}.
Then, (f o g)(a) 5 f (g(a)) 5 f (b) 5 a
and   (g o f )(a) 5 g(f (a)) 5 g(a) 5 b.

Therefore, (f o g)(a) ≠ (g o f )(a) and hence f o g ≠ g o f.

Worked Exercise 3.1.3. Determine the number of noncommutative binary

operations on a 5-element set.

Answer: Let S be a 5-element set. The total number of binary operations on

S is 525. The number of commutative binary operations on S is

52 5
5 2
 515.

Therefore, the number of noncommutative binary operations on S is

525  515  515 (510 1).

EXERCISE 3(a)

1. Construct tables representing the following binary systems:

(a) (Z4, 14)
(b) (Z10, ?10)

Q001-Algebra-111001_CH 03.indd 14 9/16/2011 10:07:18 AM

 Groups   3-15

(c) (P(X), ∪), where X 5 {a, b, c}

(d) (S, ), where S is the set of positive divisors of 100 and  is defined by
a  b 5 LCM of {a, b}
(e) (M(X), o), where X 5 {a, b}, M(X) is the set of mappings of X into itself
and o is the composition of mappings.
(f ) (S, ?), where S 5 {1, i, 21, 2i} and ‘?’ is the usual multiplication of
complex numbers.

2. Compare the tables in (a) and (f ) above.

3. Fill in the blanks in the following table such that the binary operation * repre-
sented by the table is commutative.

* a b c d e f
a d e c b
b f d f
c a c a d
d c e d
e b b d
f d c b a

4. Compute the following from the table given in (3) above.

(b * (d * a)) * (c * (b * a))
((a * b) * c) * d
(a * b) * (c * d)
5. Is the operation * given in (3) above associative?

6. Give an example of an associative binary operation which is not commutative.

7. Prove that the associativity and the commutativity are independent of each
other.

8. Prove or disprove the statement:

Every commutative binary operation on a 2-element set is associative.

9. Which of the following binary operations are associative or commutative?

(a) On the set Z of integers, a * b 5 (a 1 3)(b 1 2).
(b) On the set Z, a * b 5 a|b|.
(c) On the set R of real numbers, a * b 5 a 2 b.
(d) On the set R, a * b 5 a 1 ab.

Q001-Algebra-111001_CH 03.indd 15 9/16/2011 10:07:18 AM

 3-16   Algebra – Abstract and Modern

(e) On the power set P(X) of a set X, A * B 5 (X 2 A) ∪ (X 2 B).

(f ) On any set X, a * b 5 a.
(g) On the set Z, a * b 5 0.
(h) On the set Z+ of positive integers,
a * b 5 5a+b
10. Prove or disprove the following statement:
Every binary operation on a set S is both commutative and associative if and
only if S has exactly one element.

11. Compute the number of commutative binary operations on a 4-element set.

12. Compute the number of noncommutative binary operations on a 3-element set.

13. Let (S, *) be a binary system and

A 5 {x  S : (x * b) * c 5 x * (b * c)} for all b and c  S}.
If A is nonempty, prove that (A, * ) is a semigroup.
14. Let (S, *) be a semigroup and e be any element not in S and S9 5 S ∪ {e}. Define
a binary operation 1 on S9 as follows. For any a and b S9, define

a*b if both a and b  S


a  b  a if be .
b if ae

Then prove that (S9, 1) is a semigroup and e 1 x 5 x 5 x 1 e for all x  S9.

3.2 GROUPS
The integer 0 has a special property in the binary system (Z, 1) and is unique
satisfying this property; namely

a 10 5 a 5 0 1 a for all a  Z.

Similarly, the integer 1 is the unique element in the binary system (Z, ?)
satisfying the property

a?15a51?a for all a  Z

Likewise, the identity map IX, defined on any set X by IX(x) 5 x for all
x  X, is the unique element in the binary system (M(X), o) satisfying the
property

f o IX 5 f 5 IX o f for all f  M(X),

Q001-Algebra-111001_CH 03.indd 16 9/16/2011 10:07:18 AM

 Groups   3-17

where M(X) is the set of mappings of X into itself and o is the composition of
mapping. An abstraction of these ideas is made in the following definition.

Definition 3.2.1. Let (S, *) be a binary system and e be an element of S.

1. e is said to be a right identity in (S, *) if a * e 5 a for all a  S.
2. e is said to be left identity in (S, *) if e * a 5 a for all a  S.
3. e is said to be an identity in (S, *) if e is both a left identity and a right
identity; that is, a * e 5 a 5 e * a for all a  S.

Example 3.2.1
1. 0 is the only identity in (Z, 1).
2. 1 is the only identity in (Z, ?), where ‘?’ is the usual multiplication of
integers.
3. Let X be any nonempty set and IX : X → X be defined by IX(x) 5 x for all
x  X. Then, IX is the identity in (M(X), o).
4. Let m and n be any positive integers and Mm3n(R) be the set of all m3n
matrices over R. Let Om3n be the m 3 n matrix in which each entry is the
number 0. Then,
O mn  A  A  A O mn

for all matrices A and hence Om3n is the identity in (Mm3n(R), 1), where
1 is the usual addition of matrices.
5. Let S be any nonempty set and define a * b 5 b for all a and b in S. Then,
every element of S is a left identity in the binary system (S, *).
6. Let S be any nonempty set and define a * b 5 a for all a and b  S. Then,
every element of S is a right identity in (S, *).
7. For any set X, the empty set [ is the identity in (P(X), ∪) and also in
(P(X), 1), where ∪ is union operation and 1 is the symmetric difference
operation.
8. Also for any set X, the whole set X is the identity in (P(X), ∩), where ∩
is the intersection operation.
From examples (7) and (8) above, the concept of identity is depending on
the binary operation of the system. Also, it depends on the underlying set. For
example, 0 is the identity in the system (Z, 1) where as it is not the identity
in (Z+, 1), since 0 is not an element in the underlying set Z+.
Also, from examples (5) and (6), observe that a binary system can possess
any number of right identities without having any left identity and vice versa.
However, if e is a right identity and f is a left identity, then e must be equal to
f. This is proved in the following theorem.

Q001-Algebra-111001_CH 03.indd 17 9/16/2011 10:07:19 AM

 3-18   Algebra – Abstract and Modern

Theorem 3.2.1. Let (S, *) be a binary system and e be a left identity in (S, *).
Then, every right identity in (S, *) is an identity and coincides with e.

Proof: Let f be a right identity in (S, *). Then,

f 5 e * f (since e is left identity)

5 e (since f is right identity)

Therefore, f 5 e and hence f is a left identity also. Thus, f is an identity

and f 5 e.

Corollary 3.2.1. There can be almost one identity in any binary system.

Proof: Let (S, *) be a binary system and e and f be identities in (S, *). Since e
is a left identity and f is a right identity, e 5 f by the above theorem.

Example 3.2.2
1. Let E be the set of even integers and is the usual multiplication of inte-
gers. Then, there is no identity in the binary system (E, ?).
2. Let S be any set and define a * b 5 b for all a, b  S. Then, every ele-
ment of S is a left identity in (S, *). However, (S, *) has no right identity,
unless S is a singleton set (one element set).
3. Similarly, if we define a * b 5 a for all a, b  S then every element of S
is a right identity in (S, *) and there are no left identities in (S, *), unless
S is a singleton set.
4. The integer 0 is the only identity in (Z, 1).

Definition 3.2.2. A semigroup (S, *) is called a monoid if it has identity. That

is, a binary system (S, *) is called a monoid if * is associative and the identity
exists in (S, *).

Example 3.2.3
1. The set Mn(R) of all n 3 n square matrices over R together with the
matrix multiplication is a monoid. Here, the matrix In, in which all the
diagonal entries are 1 and the others are 0, is the identity element in
Mn(R). In is called the identity matrix. Note that In 5 (aij), where aij 5 1
or 0 according as i 5 j or i ≠ j.
2. (Z, 1), (Z, ?) and (Z+ , ?) are all monoids, where 1 and ‘?’ are the usual
addition and multiplication, respectively. 0 is the identity in (Z, 1) and
1 is the identity in (Z, ?) and in (Z+, ?).

Q001-Algebra-111001_CH 03.indd 18 9/16/2011 10:07:19 AM

 Groups   3-19

3. (Z+, 1) is a semigroup, but not a monoid.

4. The set M(X) of all mappings of a set X into itself together with the com-
position of mappings is a monoid. The identity in (M(X), o), is precisely
the identity map IX defined by IX(x) 5 x for all x  X.
Next, we shall take up the solvability of linear equations of type a * x 5 b,
where a and b are given elements in a binary system (S, *). It is well known
that, for any real numbers a and b, there is a unique real number x satisfying
the equation

a 1 x 5 b.

Our usual procedure of finding x is the following.

Consider a 1 x 5 b

(2a) 1 (a 1 x) 5 (2a) 1 b (by adding 2a)

(2a 1 a) 1 x 5 2a 1 b (by associativity)
0 1 x 5 2a 1 b (2a 1 a 5 0)
x 5 2a 1 b (0 is the identity in (R, 1)).

Also, if we substitute 2a 1 b for x in a 1 x 5 b, we get that

a 1 (2a 1 b) 5 (a 1 (2a)) 1 b 5 0 1 b 5 b.

This is to say that 2a 1 b is the unique real number satisfying the equa-
tion a 1 x 5 b. In this process finding the unique solution of a 1 x 5 b, we
have skipped one step by not explaining what 2a is. It is obvious that 2a is
the unique real number x satisfying the equation x 1 a 5 0. This concept is
abstracted in the following definition.

Definition 3.2.3. Let (S, *) be a monoid in which e is the identity and a  S.

1. An element a  S is called a left inverse of a if
a * a 5 e
2. An element ar  S is called a right inverse of a if
a * ar 5 e
3. An element a9  S is called an inverse of a if a9 is both a left inverse and
a right inverse of a; that is,
a9 * a 5 e 5 a * a9.
4. a is called invertible, if there exists an inverse of a.

Q001-Algebra-111001_CH 03.indd 19 9/16/2011 10:07:19 AM

 3-20   Algebra – Abstract and Modern

Example 3.2.4
1. In the monoid (R, 1), the number 0 is the identity and every element of
R has inverse.
2. In the monoid (Z, ?), 1 is the identity, where ‘?’ is the usual multiplica-
tion. Here, 1 and 21 are the only elements having inverses.
3. Consider the set M(Z) of all mappings of Z into itself. Then, (M(Z), 0)
is a monoid, in which the identity map I, defined by I(x) 5 x for all x 
Z, is the identity element. Define f : Z → Z by f (x) 5 2x for all x  Z.
For each integer a, define ga : Z → Z by

 x if x is an even integer

g a (x )   2

a if x is an odd integer.

Then, (ga o f )(x) 5 ga(f (x)) 5 ga(2x) 5 2x

2 5 x for all x  Z. Therefore,
ga o f 5 I

and hence ga is a left inverse of f, for each integer a.

4. In (3) above, note that f has no right inverse; for, if g is a right inverse of
f, then
fοg5I

and, in particular, (f ο g)(1) 5 I(1) 5 1 and hence

2∙g(1) 5 1

which is false, since we cannot get an integer g(1) such that 2g(1) 5 1.
From the examples (3) and (4) above, we have noticed that an element in
a monoid can have several left inverses without having any right inverses.
However, if an element has both left inverse and right inverse, then they must
be equal. This is proved in the following theorem.

Theorem 3.2.2. Let (S, *) be a monoid in which e is the identity and a  S.

Let a and ar be left inverse and right inverse of a, respectively. Then, a 5 ar
and a is invertible.

Proof: We are given that a * a 5 e 5 a * ar. Now, consider

a 5 a * e (since e is the identity)

5 a * (a * ar) (since ar is right inverse of a)

Q001-Algebra-111001_CH 03.indd 20 9/16/2011 10:07:20 AM

 Groups   3-21

5 (a * a) * ar (by associativity)

5 e * ar (since a is left inverse of a)


5 ar (since e is identity)

Therefore, a 5 ar and hence a * a 5 e 5 a * a so that a (5 ar) is inverse of a.

Corollary 3.2.2. Any element in a monoid has at most one inverse.

Proof: Let (S, *) be a monoid in which e is the identity and a  S. Suppose

a9 and a0 are inverses of a. Then, a9 is a left inverse and a0 is a right inverse
of a and hence, by the above theorem a9 5 a0.
Note that in a monoid, certain elements may be invertible and other
elements may not be invertible. The identity element e in any monoid is
always invertible and, since e * e5 e, e is the inverse of itself. In the fol-
lowing, we give an example of a monoid in which each element is inverse
of itself.

Example 3.2.5. Let X be any set and P(X) be the power set of X. For any A
and B  P(X), define

A 1 B 5 (A 2 B) ∪ (B2A).

Then, (P(X), 1) is a monoid with the empty set [ as the identity element.
Here, for any A  P(X),

A 1 A 5 (A2A) ∪ (A2A) 5 [ ∪ [ 5 [

and hence A is the inverse of itself.

In certain monoids, some elements may have left inverses, some elements
may have right inverses, some may have both left and right inverses and some
may have neither left inverses nor right inverses. In the following theorem, we
give one such example.

Theorem 3.2.3. Let X be any nonempty set and M(X) be the set of all map-
pings of X into itself. Then, (M(X), o) is a monoid in which the following
holds for any f  M(X).
1. f has a left inverse in M(X) if and only if f is an injection.
2. f has a right inverse in M(X) if and only if f is a surjection.
3. f is invertible in M(X) if and only if f is a bijection.

Q001-Algebra-111001_CH 03.indd 21 9/16/2011 10:07:20 AM

 3-22   Algebra – Abstract and Modern

Proof: We know that (M(X), o) is a monoid (see Example 3.2.3 (4)) in which
o is the composition of mapping and the map IX : X → X, defined by IX(x) 5 x
for all x  X, is the identity. Let f be an arbitrary element of M(X); that is,
f : X → X is a mapping.
1. Suppose that f has a left inverse in M(X). Then, there exists g  M(X)
such that
g o f 5 IX.
For any a, b  X, we have
f (a) 5 f (b) ⇒ g(f (a)) 5 g(f (b))
⇒ (g o f )(a) 5 (g o f )(b)
⇒ IX(a) 5 IX(b)
⇒ a 5 b.
Therefore, f is an injection.
Conversely, suppose that f is an injection. Define
g : X → X by
a if x  f (a) for some a ∈ X
g (x )   ,
s otherwise

where s is an arbitrarily chosen fixed element of X. Note that, since f is

an injection, there can be at most one a  X such that x 5 f (a) and hence
g is welldefined. Now, for any a  X,
(g o f )(a) 5 g(f (a)) 5 a 5 IX(a)
and hence g o f 5 IX, so that g is a left inverse of f.
2. Suppose that f has a right inverse in M(X), Then, there exists h  M(X)
such that
f o h 5 IX.
For any x  X, we have h(x)  X and
f (h(x)) 5 (f o h)(x) 5 IX(x) 5 x
and therefore f is a surjection.
Conversely suppose that f is a surjection. Define h : X → X as follows.
For any x  X, choose one element ax  X such that f (ax) 5 x (since f is
a surjection, f –1{x} is a nonempty subset of X, for each x  X and now,
we have to the axiom of choice). Now, define
h(x) 5 ax, for each x  X.

Q001-Algebra-111001_CH 03.indd 22 9/16/2011 10:07:20 AM

 Groups   3-23

Since (f o h)(x) 5 f (h(x)) 5 f (ax) 5 x for all x  X, we have f o h 5 IX

and hence h is a right inverse of f.
3. This follows from (1) and (2) and from Theorem 3.2.2.

Definition 3.2.4. A monoid (G, *) is called a group if every element of S is

invertible.
To be more elaborate, A pair (G, *) is called a group if the following are
satisfied:
1. G is a nonempty set and * is a binary operation on G.
2. a * (b * c) 5 (a * b) * c for all a, b and c  G.
3. There exists e  G such that
a*e5a5e*a for all a  G.

4. For each a  G, there exists a9  G such that

a9 * a 5 e 5 a * a9.
Recall that a pair (G, *) is called a binary system if (1) is satisfied, semi-
group if (1) and (2) are satisfied, monoid if (1), (2) and (3) are satisfied and is
called a group if all the four (1), (2), (3), and (4) are satisfied.
Also, recall that the element e in (3) is unique and is called the identity in
(G , *). Further, the element a9 in (4) is unique and is called the inverse of a.
The inverse of a is usually denoted by a–1.

Example 3.2.6
1. (Z, 1), (Q, 1), (R, 1) and (C, 1) are all groups in which 1 is the usual
addition, 0 is the identity and 2a is the inverse of any element a.
2. (Q 2{0}, ?), (R 2{0}, ?) and (C 2{0}, ?) are all groups in which ‘?’ is the
1
usual multiplication, 1 is the identity and a is the inverse of any element a.
3. Neither (Z, ?) nor (Z2{0}, ?) are groups, since not all elements are
invertible.
4. For any set X, (P(X), 1) is a group in which 1 is the symmetric differ-
ence operation,  is the identity and, every element is inverse of itself.
5. Let X be a nonempty set and S(X) the set of all bijections of X onto itself.
Then, (S(X), o) is a group in which o is the composition of mappings, IX
is the identity and f –1 is the inverse of any bijection f. Recall that, for any
bijections f and g, the composition f o g is also a bijection.
6. The set Mm3n(R) of all m 3 n matrices over R together with the addition
of matrices is a group. Here the zero matrix, in which all the entries are 0,
is the identity and, for any A 5 (aij) , the matrix (2aij) is the inverse of A.

Q001-Algebra-111001_CH 03.indd 23 9/16/2011 10:07:20 AM

 3-24   Algebra – Abstract and Modern

7. For any positive integer n, (Zn, 1n) is a group where Zn 5 {0, 1, 2, … ,

n 2 1} and 1n is the addition modulo n. Here, 0 is the identity and, for
any a  Zn, n 2 a is the inverse of a. This group (Zn, 1n) is called the
additive group of integers modulo n.
8. Let n be a positive integer. A n 3 n square matrix A is called nonsingu-
lar if its determinant is not zero. The set NSMn(R) of all nonsingular
n 3 n matrices over R together with the matrix multiplication is a
group. Here, the identity matrix In, in which all the diagonal entries
are 1 and other entries are 0, is the identity. It is well known that a
n 3 n square matrix A is nonsingular if and only if there exists n 3 n
matrix B such that
AB 5 In 5 BA.

9. For any points a 5 (a1, a2) and b 5 (b1, b2) in the two-dimensional
Euclidean space R 3 R, d(a, b) be the usual Euclidean distance between
a and b; that is,

d (a, b)  (a1  b1 ) 2  (a2  b2 ) 2 .

Let X be the set of all points on a given geometrical figure (in fact, X may
be any nonempty subset of R 3 R).
A bijection f of X onto itself is called a symmetry of X if
d(f (a), f (b)) 5 d(a, b) for all a, b  X.

Let Sym(X) be the set of all symmetries of X. Then, (Sym(X), o) is a

group in which o is the composition of mappings, IX is the identity and
f–1 is the inverse of any f. This group (Sym(X), o) is called the group of
symmetries of X.
10. Let p be any prime number and Gp 5 {1, 2, …, p 2 1}. Then, (Gp,?p) is
a group in which ?p is the multiplication modulo p (see Example 3.1.1
(12)) defined by
a ?p b 5 the remainder obtained by dividing
the usual product ab with p.

Here, 1 is the identity and, for any 0 , a , p, the GCD of a and p is 1

and hence, there exists integers a and b such that
aa 1 bb 51.

If we divide a by p, we get the remainder b. Then, 0 , b , p and b ?p

a 5 1and hence b is the inverse of a in (Gp, ?p).

Q001-Algebra-111001_CH 03.indd 24 9/16/2011 10:07:21 AM

 Groups   3-25

Theorem 3.2.4. Let (M, *) be a monoid and G be the set of all invertible ele-
ments in (M, *). Then, (G, *) is a group.

Proof: First we shall observe that G is a nonempty set, since the identity e in
(M, *) is always invertible and hence e  G. Also, if a and b  G and a9 and
b9 are inverses of a and G, respectively, then

(a * b) * (b9 * a9) 5 a * (b * b9) * a9

5 a * e * a9
5 a * a9 5 e
and (b9 * a9) * (a * b) 5 b9 * (a9 * a) * b
5 b9 * e * b
5 b9 * b 5 e

and hence a * b is invertible and b9 * a9 is the inverse of a * b, so that a * b

 G. Therefore, * becomes a binary operation on G. Also, since * satisfies
associativity on M, * is associative on G also. Since e  G, (G, *) is a monoid
and clearly, the inverse of any invertible element is also invertible. Therefore,
every element in (G, *) is invertible. Thus, (G, *) is a group.

Definition 3.2.5. Let (M, *) be a monoid and a  M. For any nonnegative

integer n, define

e, the identity if n  0

a n   n1 .
a * a if n 0

If a is invertible and n is a negative integer, define

an 5 (a9)–n , when a9 is the inverse of a.

Note that a0 5 e, a1 5 a, a2 5 a * a, a3 5 (a * a) * a, etc. and, if a is invert-

ible, then

a–1 5 a9, the inverse of a

a–2 5 (a9)2
and a–n 5 (a9)n for any n  Z+.

This justifies the notation a–1 for the inverse of a.

Worked Exercise 3.2.1. Let a be an invertible element in a monoid (M, *) and

a9 be the inverse of a. Then prove the following for any integers m and n.

Q001-Algebra-111001_CH 03.indd 25 9/16/2011 10:07:21 AM

 3-26   Algebra – Abstract and Modern

1. an+m 5 an * am
2. (a9)n 5 a–n
3. (an)m 5 anm 5 (am)n

Answer: We shall fix n  Z and use induction on m.

1. Case (i): Suppose that m ≥ 0.
If m 5 0, then an+m 5 an 5 an * e 5 an * am.
Let m . 0 and assume that an+(m–1) 5 an * am–1. Then,
an * am 5 an * (am–1 * a) (by definition of am)
5 (an * am–1) * a (by associativity)
5 (an+(m–1)) * a (by induction hypothesis)
5 an + m
Thus, an+m 5 an * am for all n, m  Z with m $ 0.
Case (ii): Suppose that m , 0.
Consider, an * am 5 an * (a9)–m (by definition of am)
 a n(m ) if n m
  mn
( a) if  m  n
 a nm

2. This is trivial if n 5 0, since (a9)0 5 e 5 a0

If n . 0, then 2n , 0 and hence, by definition,
a–n 5 (a9)–(–n) 5 (a9)n.

If n , 0, then, again by definition,

(a9)n 5 a–n , since a is the inverse of a9.

3. This is trivial if m 5 0 or n 5 0. Therefore, we can assume that mn ≠ 0.

Case (i): Suppose that mn . 0.
If both m and n are positive, then, by (1),
(an)m 5 anm 5 amn 5 (am)n.

If both m and n are negative, then

(an)m 5 ((an)9)–m
5 (((a9)–n)9)–m
5 (((a9)9)–n)–m
5 (a–n)–m
5 a(–n)(–m) 5 anm

Q001-Algebra-111001_CH 03.indd 26 9/16/2011 10:07:21 AM

 Groups   3-27

Case (ii): Suppose that mn ,0. To be specific, suppose that n , 0 and

m . 0. Then,
anm 5 (a9)–nm
5 (a9)(–n)m
5 ((a9)–n)m (since –n . 0 and m . 0)
5 (an)m.

So is the case when n . 0 and m , 0.

Thus, (an)m 5 anm 5 amn 5 (am)n.

Worked Exercise 3.2.2. Let G be the set of all rotations of the plane about
the origin in the plane and o the composition of mappings. Thus, prove that
(G, o) is a group.

Answer: The rotation about the origin through an angle u can be represented
analytically as the map fu : R 3 R → R 3 R defined by

f u ( x, y )  ( x cos u  y sin u, x sin u  y cos u)

Therefore, G 5 {fu : 0 ≤ u , 2p}. Note that

 f u if u  2p
f u ο f   
 f u2p if u  2p


It can be easily verified that o is an associative binary operation on G and that

f0 ( 5 the identity map) is the identity in (G, o). Also for any u, f2p−u is the
inverse of fu (considering f2p 5 f0). Thus, (G, o) is a group.

Worked Exercise 3.2.3. For any real numbers a and b with a ≠ 0, define

fa,b : R → R by fa,b(x) 5 ax 1 b for all x  R,

let G 5 {fa,b : 0 ≠ a  R and b  R}. Then prove that (G, o) is a group, where
o is the composition of mappings.

Answer: Note that, for any a, b, c and d  R,

( fa,b o fc,d)(x) 5 fa,b (cx 1 d)

5 a(cx 1 d) 1 b
5 acx 1 ad 1 b
5 fac,ad1b(x)

Q001-Algebra-111001_CH 03.indd 27 9/16/2011 10:07:22 AM

 3-28   Algebra – Abstract and Modern

and hence fa,b o fc,d 5 fac, ad1b. Also, if a ≠ 0 and c ≠ 0, then ac ≠ 0. ­Therefore,
0 is an associative binary operation on G. Further, f1,0 is the identity in
(G, o) , since f1,0(x) 5 1  x 1 0 5 x for all x R. Also, f 1 ,− b is the inverse
a a
of fa,b. Thus, (G, o) is a group.
We shall conclude this section with two more important examples of
groups given below.

Example 3.2.7. Let G 5 {f1, f2, f3, f4, f5, f6}, where each f i is a function of
R 2 {0, 1} into itself as defined below.

1
f1 (x )  x, f 2 (x )  , f 3 (x )  1 x,
x
1 x 1 x
f 4 (x )  , f 5 (x )  and f 6 (x )  .
1 x x x 1

Then, (G, o) is a group, where o is the composition of mappings. The following

table represents the binary operation o on G.

0 f1 f2 f3 f4 f5 f6
f1 f1 f2 f3 f4 f5 f6
f2 f2 f1 f4 f3 f6 f5
f3 f3 f5 f1 f6 f2 f4
f4 f4 f6 f2 f5 f1 f3
f5 f5 f3 f6 f1 f4 f2
f6 f6 f4 f5 f2 f3 f1

Note that f1 is the identity and the inverses are given by

f11  f1 , f 21  f 2 , f 31  f 3 , f 41  f 5 , f 51  f 4 , f 61  f 6 .

Also note that f2 o f4 5 f3 ≠ f6 5 f4 o f2.

Example 3.2.8. The group discussed here is called the group of symmetries
of the square (see Example 3.2.6 (9)). Let X be the set of all points in a square
of unit side. Recall that a symmetry of X is a bijection f of X into itself such
that

d(a, b) 5 d(f (a), f (b)) for all a and b  X.

Q001-Algebra-111001_CH 03.indd 28 9/16/2011 10:07:22 AM

 Groups   3-29

Let G be the set of all symmetries of X. Then, (G, o) is a group, where o

is the usual composition of mappings. Note that G consists of exactly eight
symmetries and these are listed below.
e 5 The identity function.
r1 5 The clock-wise rotation about the centre of the square through an
angle p 2
.
r2 5 The clock-wise rotation about the centre through an angle p.
r3 5 The clock-wise rotation about the centre through an angle 3p .
2
h 5 The reflection about the horizontal line through the centre of the
square.
v 5 The reflection about the vertical line through the centre of the square.
d1 5 The reflection about the diagonal D1.
d2 5 The reflection about the diagonal D2.

d1

1 2

h
r3 r1
4 3
r2

d2

We have G 5 {e, r1, r2, r3, h, v, d1, d2}.

The binary operation o on G is represented by the following table.

o e r1 r2 r3 h v d1 d2
e e r1 r2 r3 h v d1 d2
r1 r1 r2 r3 e d1 d2 v h
r2 r2 r3 e r1 v h d2 d1
r3 r3 e r1 r2 d2 d1 h v
h h d2 v d1 e r2 r3 r1
v v d1 h d2 r2 e r1 r3
d1 d1 h d2 v r1 r3 e r2
d2 d2 v d1 h r3 r1 r2 e

Q001-Algebra-111001_CH 03.indd 29 9/16/2011 10:07:23 AM

 3-30   Algebra – Abstract and Modern

EXERCISE 3(b)

1. Determine the following in which 1n and ?n are the addition and multiplication
modulo n, for a given positive integer n.
(i) 7 112 11
(ii) 8 110 7
(iii) 7 ?12 11
(iv) 8 ?10 7
(v) 4 ?6 5
(vi) (7 110 6) ?10 8
(vii) 77 in (Z8, ?8)
(viii) 5–6 in (G7,?7)
(ix) 7–8 in (G11,?11)
(x) 68 in (Z9,?9)

2. List all the invertible elements in each of the following monoids.

(i) (Z10, 110)
(ii) (Z10, ?10)
(iii) (Z36, ?36)
(iv) (Z4, ?4)
(v) (M(X), o), where M(X) is the set of all mappings of X into itself
(vi) (P(X), 1).
(vii) (R+ , ?)
(viii) (Z+ , ?)

3. Determine which of the following gives a group structure on the given set.
(i) For any a, b  Z, a * b 5 a 1 b 1 ab.
(ii) For any a, b  R+ , a * b  ba .
(iii) a * b  ab for any a, b  R+.
2
(iv) a * b 5 |ab| for any a, b  C.
(v) a * b 5 a 1 b 2 2 for any a, b  Z.
(vi) For any a, b,  R+ , a * b 5 5ab.
(vii) For any a, b  Q+ , a * b 5 |ab|.
(viii) a * b 5 a 1 b 1 ab, for any a, b  Z.

4. Prove that the matrices

1 0 1 0 1 0   
 ,  ,  , 1 0 
0 1  0 1  0 1 0 1

Q001-Algebra-111001_CH 03.indd 30 9/16/2011 10:07:24 AM

 Groups   3-31

form a group under the matrix multiplication. What is identity element? Determine
the inverse of each element.

5. State whether each of the following is true (T) or false (F ):

(i) Any semigroup has exactly one left identity.
(ii) Any semigroup has at most one right identity.
(iii) Any group has exactly one left identity.
(iv) In a group, each element has exactly one right inverse.
(v) There is a group with exactly one element.
(vi) For each positive integer n, there is a group with exactly n elements.
(vii) Every semigroup has an identity.
(viii) For any positive integer n, there is a semigroup with exactly n elements
in which every element is a right identity.

6. Let (G, *) be a group. Prove that the identity e is the only element satisfying
x * x 5 x.

7 Let n be a positive integer and G be the set of all nth root of unity; that is,

G 5 {z : z is a complex number and zn 5 1}.

Prove that G is group under the usual multiplication of complex numbers.

8. Let (G, *) be a group and X be any nonempty set. Let GX be the set of all map-
pings of X into G. For any f, g  GX, define f * g : X → G by

(f * g) (x) 5 f (x) * g(x) for all x  X.

Prove that (G , *) is a group. What is the identity in this group? Determine the
X

inverse of any f  GX.

9. For any positive integer n, prove that (Zn, ?n) is a monoid, where Zn 5 {0, 1, 2,
…n 2 1} and ?n is the multiplication modulo n.

10. For any n . 1, prove that (Zn, ?n) is never a group.

11. For any 1 # a , n, prove that a is invertible in the monoid (Zn, ?n) if and only if
a is relatively prime to n.

12. For any prime number p, prove in detail that (Gp, ?p) is a group, where

Gp 5 Zp 2 {0}.

13. For any positive integer n, give an example of a group having exactly 2n
elements.

14. Is {1, 2, 3, 4} a group under multiplication modulo 5?

15. Is {1, 2, 3, 4, 5} a group under multiplication modulo 6?

Q001-Algebra-111001_CH 03.indd 31 9/16/2011 10:07:24 AM

 3-32   Algebra – Abstract and Modern

16. Prove that the following are equivalent to each other for any integer n . 1.
(1) ?n is a binary operation on Zn 2 {0}.
(2) (Zn 2 {0}, ?n) is a group.
(3) n is a prime number.
(4) any 1 # a , n is relatively prime with n.

17. Let G be the set of all rotations about the origin in the plane and reflections in the
lines through origin. Then prove that (G, o) is a group, where o is the composi-
tion of mappings.

18. Consider the regular n-gon (polygon of n equal sides and equal internal angles)
inscribed in the unit circle in the plane, so that one of the vertices is (1, 0). Let
Rn be the set of all rotations about the origin which maps this regular n-gon into
itself. Prove that (Rn, o) is a group, where o is the composition of mappings.
How many elements are there in this group?

19. Let Dn be the set of all rotations and reflections which the regular n-gon, given in
18 above, into itself. Then prove that (Dn, o) is a group, where o is the composi-
tion of mappings. How many elements Dn has? The group (Dn, o) is called the
dihedral group of degree n (see Theorem 6.4.8). The elements of Dn are called
the symmetries of the regular n-gon.

20. For any rational numbers r and s, define r ~ s if and only if r 2 s is an integer.
Then prove that ~ is an equivalence relation on the set Q of rational numbers.
Let Q denote the set of equivalence classes w.r.t. ~ in Q and, for any classes
Z
 s. Then prove that (Q , 1) is a group.
r, s, define r  s  r
Z

3.3 ELEMENTARY PROPERTIES OF GROUPS

In this section, we shall derive certain important elementary properties
of groups. In particular, we obtain several sets of equivalent conditions
for a semigroup to become a group. Let us agree to denote the identity
element in an abstract monoid or group by e. We begin with the following
theorem.

Theorem 3.3.1. Let (G, *) be a group and a, b and c  G. Then, the following
holds.
1. a * b 5 e ⇔ a–1 5 b ⇔ b–1 5 a, where e denotes the identity in the
group.
2. (a–1)–1 5 a
3. (a * b) –1 5 b–1 * a–1
4. a * b 5 c ⇔ a 5 c * b–1 ⇔ b 5 a–1 * c

Q001-Algebra-111001_CH 03.indd 32 9/16/2011 10:07:24 AM

 Groups   3-33

Proof:
1. a * b  e ⇒ a1  a1 * e  a1 * ( a * b)  ( a1 * a) * b  e * b  b
a1  b ⇒ a * b  a * a1  e
a * b  e ⇒ b1  e * b1  ( a * b) * b1  a * (b * b1 )  a * e  a
b1  a ⇔ a * b  b1 * b  e
2. Since a–1 * a 5 e, it follows from (1) that (a–1)–1 5 a
3. Since (a * b) * (b–1 * a–1) 5 a * (b * b–1 )* a–1 5 a * e * a–1 5 a * a–1 5 e,
again from (1) it follows that (a * b)–1 5 b–1 * a–1
4. a * b  c ⇒ c * b1  ( a * b) * b1  a * (b * b1 )  a * e  a
a  c * b1 ⇒ a * b  (c * b1 ) * b  c * (b1 * b)  c * e  c
a * b  c ⇒ a1 * c  a1 * ( a * b)  ( a1 * a) * b  e * b  b
b  a−1 * c ⇒ a * b  a * ( a1 * c)  ( a * a1 ) * c  e * c  c
Note that if we take e for c in (4), we get (1).
Let us recall that a semigroup is a pair (S, *) where S is a nonempty set and *
is an associative binary operation on S and that a semigroup with identity is
called a monoid and also that a monoid is called a group if every each of its
elements is invertible.

Theorem 3.3.2. Let (S, *) be a semigroup. Then, (S, *) is a group if and only
if the following conditions are satisfied.
1. (S, *) has a right identity e. That is, there exists e  S such that a * e 5
a for all a  S.
2. For each a  S, there exists a9  S such that a * a9 5 e

Proof: If (S, *) is a group, then clearly (1) and (2) are satisfied. Conversely sup-
pose that the conditions (1) and (2) are satisfied. By (1), there exists e  S, such
that a * e 5 a for all a  S. We shall prove that this e is actually the identity in (S,
*). Let a be an arbitrary element in S, By (2), there exists a9 and x in S such that

a * a e and ( a * a) * x e (i)

Consider

( a * a) * ( a * a)  a * ( a * a) * a
 ( a * e) * a (by (i)) (ii)
 a * a        

Q001-Algebra-111001_CH 03.indd 33 9/16/2011 10:07:25 AM

 3-34   Algebra – Abstract and Modern

Now,

e  ( a * a) * x (by (i))
 (( a * a) * ( a * a)) * x (by (ii))
 ( a * a) * (( a * a) * x ) (by associativity)
 ( a * a) * e (by (i))
 a * a

Thus, for any a  S, there exists a9  S such that

a9 * a 5 e 5 a * a9           (iii)

Also, e * a 5 (a * a9) *a 5 a * (a9 * a) 5 a * e 5 a.

Thus, e is the identity in (S, *) and, for any a  S, a9 is the inverse of a (by
(iii)), Therefore, (S, *) is a group.
On the lines of above proof, one can also prove that a semigroup is a group
if and only if it has left identity with respect to which every element has left
inverse.
Recall that, for any given real numbers a and b, the equation a 1 x 5 b
has a unique solution in R. In fact, this is an important defining property of a
group as proved in the following theorem.

Theorem 3.3.3. A semigroup (S, *) is a group if and only if, for any elements
a and b in S, the equation

a*x5b and y * a 5 b

are solvable in S (in the sense that there are elements x and y in S satisfying
these equations).

Proof: Let (S, *) be a semigroup. If (S, *) is a group, then for any a, b  S,

we have a–1 * b and b * a–1 are elements of S such that

a * (a–1 * b) 5 (a * a–1) * b 5 e * b 5 b
and (b * a–1) * a 5 b * (a–1 * a) 5 b * e 5 b

and therefore the equations a * x 5 b and y * a 5 b have solutions in S.

Conversely suppose that these equations have solutions in S for any
a and b in S.

Q001-Algebra-111001_CH 03.indd 34 9/16/2011 10:07:25 AM

 Groups   3-35

Let a be an arbitrary element in S. Then, there exists e  s such that

a*e5a (since a * x 5 a is solvable in S).

We shall prove that b * e 5 b for all elements b  S. To prove this, let b  S.

Then, choose an element s  S such that

s*a5b (since y * a 5 b is solvable in S).

Now, b * e 5 (s * a) * e 5 s * (a * e) 5 s * a 5 b.
Thus, e is a right identity in (S, *). Also, since a * x 5 e is solvable in S, we
get that, for each a  S, there exists a9  S such that a * a9 5 e. Thus, by the
above Theorem 3.3.2, (S, *) is a group.
Recall that, in the elementary school mathematics, one is used to conclude
b 5 c whenever a 1 b 5 a 1 c for some a and we were used to give reasoning
for this by saying ‘subtracting a from both sides’ which amounts to adding
−a both sides.
This is abstracted in the following theorem.

Theorem 3.3.4. Let (G, *) be a group and a, b and c  S. Then,

a * b  a * c ⇒ b c (left cancellation law)

and b * a  c * a ⇒ b c (right cancellation law).

Proof: Consider

a * b  a * c ⇒ a−1 * (a * b)  a1 * (a * c)
⇒ (a−1 * a) * b  ( a1 * a) * c
⇒ e*be*c
⇒ bc

Also, b * a 5 c * a ⇒ (b * a) * a21 5 (c * a) * a21

⇒ b * (a * a2l) 5 c * (a * a2l)
⇒b*e5c*e
⇒ b 5 c

A semigroup may satisfy both the left and right cancellation laws without
being a group. This is to say that the converse of the above theorem is not
true. For, consider the following examples.

Q001-Algebra-111001_CH 03.indd 35 9/16/2011 10:07:26 AM

 3-36   Algebra – Abstract and Modern

Example 3.3.1
1. Consider the semigroup (Z1, 1), where Z1 is the set of positive integers
and 1 is the usual addition. Since (Z, 1) is a group, (Z, 1) satisfies both
the cancellation laws. Since Z1 is a subset of Z, (Z1, 1) also satisfies
both the cancellation laws. Nevertheless, (Z1, 1) is not a group, since
this has no identity.
2. A monoid may satisfy the cancellation laws without being a group. Con-
sider the set W of all nonnegative integers. Then, for the same reason
given above, (W, 1) is a monoid satisfying both the cancellation laws
and this is not a group, since no element, except 0, has inverse.
Even though the converse of Theorem 3.3.4 is not true in general, we prove
the converse in the case of finite semigroups. Recall that a semigroup (S, *) is
called finite if the underlying set S is finite.

Theorem 3.3.5. Let (S, *) be a finite semigroup satisfying both the cancella-
tion laws. Then, (S, *) is a group.

Proof : Since S is a finite set, we can enumerate the elements of S. Let a1, a2,
…, an be all the distinct elements of S. That is,

S 5 {a1, a2, …, an}.

Let a and b be any arbitrary elements in S and let

a * S 5 {a * a1, a * a2, …, a * an}.

Then, a * ai’s are all distinct elements in S, since

a * ai 5 a * aj ⇒ ai 5 aj (by left cancellation law)

⇒i 5 j (since ai’s are distinct).

Therefore, a * S is an n-element subset of S and S also has n-elements and

hence

a * S 5 S.

In particular, b  S 5 a * S and hence b 5 a * x for some x  S.

Similarly, by using the right cancellation law, we can prove that S * a 5 S
and hence y * a 5 b for some y  S. Therefore, for any elements a and b in
S, the equations

Q001-Algebra-111001_CH 03.indd 36 9/16/2011 10:07:26 AM

 Groups   3-37

a * x 5 b and y * a 5 b

are solvable in S. Thus, by Theorem 3.3.3, (S, *) is a group.

Definition 3.3.1. A binary operation * on a set S is said to be commutative if

a * b 5 b * a for all a and b  S.
A group (G, *) is said to be a commutative group or abelian group (in
honour of a great algebraist Abel) if * is commutative ; that is,

a * b 5 b * a for all a and b  G.

Example 3.3.2
1. (Z, 1), (Q, 1), (R, 1) and (C, 1) are all abelian groups, since the
addition 1 is commutative.
2. (Q2{0},?), (R2{0},?) and (C2{0},?) are abelian groups, since the
multiplication is commutative.
3. For any set X, (P(X), 1) is an abelian group, since, for any A and B in
P(X),
A 1 B 5 (A 2 B) ∪ (B2A) 5 (B2A) ∪ (A 2 B) 5 B 1 A.
4. Let X be a set with atleast three elements and S(X) the set of all bijections
of X onto itself. Then, (S(X), o) is a group which is not abelian. For, con-
sider three distinct elements a, b and c in X and define f and g : X → X by
f (a) 5 b, f (b) 5 a and f (x) 5 x for all x ≠ a, b
and      g(b) 5 c, g(c) 5 b and g(x) 5 x for all x ≠ b, c.
Then,   (f o g)(a) 5 f (g(a)) 5 f (a) 5 b
and    (g o f )(a) 5 g(f (a)) 5 g(b) 5 c ≠ b 5 (fog)(a).
Therefore, f o g ≠ g o f. Thus, (S(X), o) is an abelian group.
5. The matrix multiplication is not commutative. Let NSMn(R) be the set
of all nonsingular n 3 n matrices over R. Then, (NSMn(R), ? ) is a group
which is not abelian if n > 1.
6. The addition of matrices is a commutative operation. (Mm3n(R), 1) is an
abelian group for any positive integers m and n, where Mm3n(R) is the set
of all m 3 n matrices over R.

Theorem 3.3.6. The following are equivalent to each other for any group
(G, *).
1. (G, *) is an abelian group.
2. (a * b)–1 5 a–1 * b–1 for all a and b  G.
3. (a * b)2 5 a2 * b2 for all a and b  G.

Q001-Algebra-111001_CH 03.indd 37 9/16/2011 10:07:26 AM

 3-38   Algebra – Abstract and Modern

Proof:
(1) ⇒ (2): If (G, *) is an abelian group and a and b  G, then, by Theorem
3.3.1 (1),

(a * b)–1 5 b–1 * a–1 5 a–1 * b–1.

(2) ⇒(3): Suppose that (a * b)–1 5 a–1 * b–1 for all a and b  G.
Then, for any a and b  G, we have

(a * b)2 5 (a * b) * (a * b)
5 (a * (b–1)–1) * ((a–1)–1 * b)
5 a * ((b–1)–1 * (a–1)–1) * b
5 a * (a–1 * b–1)–1 * b
5 a * ((a * b)–1)–1 * b
5 a * (a * b) * b
5 (a * a) * (b * b) 5 a2 * b2

(3) ⇒ (1): For any a and b  G

(a * b)2 5 a2 * b2 ⇒ (a * b) * (a * b) 5 (a * a) * (b * b)
⇒a * (b * a) * b 5 a * (a * b) * b
⇒ b * a 5 a * b (by cancellation laws)

Worked Exercise 3.3.1. Let X be any nonempty set and S(X) be the set of all
bijections of X onto itself. Then prove that (S(X), o) is an abelian group if and
only if |X| , 3.

Answer: If |X| ≥ 3, then we have proved in Example 3.3.2 (4) that the group
(S(X), o) is not abelian.
On the other hand, suppose that |X| , 3. Then, |X| 5 1 or 2. If |X| 5 1,
then S(X) has only one element, namely the identity map and S(X) 5 {IX}
is clearly abelian. If |X| 5 2, say X 5 {a, b}, then there are exactly two
bijections, namely the identity map IX and the function f : X → X defined
by f (a) 5 b and f (b) 5 a and therefore S(X) 5 {IX, f} which is clearly an
abelian group.

Worked Exercise 3.3.2. Let (G, *) be a group. Then prove that (G, *) is abe-
lian if and only if there exist three consecutive integers n such that (a * b)n 5
an * bn for all a, b  G.

Q001-Algebra-111001_CH 03.indd 38 9/16/2011 10:07:26 AM

 Groups   3-39

Answer: If (G, *) is an abelian group, then for any a and b  G, we have

(a * b)0 5 e 5 e * e 5 a0 * b0
(a * b)1 5 a * b 5 a1 * b1
and (a * b)2 5 (a * b) * (a * b)
5 a * (b * a) * b
5 a * (a * b) * b
5 (a * a) * (b * b) 5 a2 * b2

and hence, for n 5 0, 1, 2, (a * b)n 5 an * bn for all a, b  S. Conversely sup-

pose that there exists an integer n such that

(a * b)n–1 5 an–1 * bn–1 (i)

(a * b)n 5 an * bn (ii)
and (a * b)n+1 5 an+1 * bn+1 (iii)

for all a and b  G. From (i) and (ii) we have

a * (an–1 * b) * bn–1 5 an * bn
5 (a * b)n
5 (a * b) * (a * b)n–1
5 (a * b) * an–1 * bn–1
5 a * (b * an–1) * bn–1

and hence, from the cancellation laws, we get that

an–1 * b 5 b * an–1 for all a, b  G        (iv)

Similarly, by using (ii) and (iii) we get that

an * b 5 b * an for all a, b  G          (v)

Now, for any a and b  G, consider

bn–1 * (a * b) 5 (bn–1 * a) * b
5 (a * bn–1) * b       (by (iv))
5 a * (bn–1 * b)
5 a * bn
5 bn * a           (by (v))
5 (bn–1 * b) * a
5 bn–1 * (b * a)

Q001-Algebra-111001_CH 03.indd 39 9/16/2011 10:07:27 AM

 3-40   Algebra – Abstract and Modern

By the left cancellation law, we get a * b 5 b * a for all a and b  G, Thus,

the group (G, *) is abelian.

Worked Exercise 3.3.3. Let (G, *) be a group such that a2 5 e for all a  G.
Then prove that (G, *) is an abelian group.

Answer: For any elements a and b  G, we have a * a 5 a2 5 e and b * b 5

b2 5 e and hence a–1 5 a and b–1 5 b. Now

(a * b)–1 5 a * b 5 a–1 * b–1

Thus, by Theorem 3.3.6, (G, *) is an abelian group.

Worked Exercise 3.3.4. Let (G, *) be a group such that x2 ≠ e for all x ≠ e in
G. Then prove that (G, *) is an abelian group if and only if

(a * b)2 5 (b * a)2 for all a and b  G.

Answer: If (G, *) is an abelian group, then clearly

(a * b)2 5 (b * a)2 for all a and b  G.

Conversely suppose that (a * b)2 5 (b * a)2 for all a and b  G.

Let a and b be arbitrary elements of G and consider

a2 5 (a * e)2 5 (a * (b–1 * b))2

5 ((a * b–1) * b)2
5 (b * (a * b–1))2     (by hypothesis)
5 b * a * b–1 * b * a * b–1
5 b * a * e * a * b–1
5 b * a2 * b–1

Therefore, a2 5 b * a2 * b–1 for all a and b  G and hence a2 * b 5 b * a2 *

b–1 * b 5 b * a2.
Therefore, a2 * b 5 b * a2 for all a and b  G.
Now, put x 5 (a * b) *(a–1 * b–1) and consider

x2 5 x * x 5 (a * b) * (a–1 * b–1) * (a * b) * (a–1 * b–1)

5 (a * b) * (a–1 * b–1 * a) * b * (a–1 * b–1)
5 (a * b) * (a * a–2 * b–1 * a) * b * (a–1 * b–1)

Q001-Algebra-111001_CH 03.indd 40 9/16/2011 10:07:27 AM

 Groups   3-41

5 (a * b) * (a * b–1 * a–2 * a) * b * (a–1 * b–1)

5 (a *b) * (a * b–1 * a–1) * b * (a–1 * b–1)
5 (a * b) * (a * b * b–2 * a–1) * (b * a–1 * b–1)
5 (a * b) * (a * b * a–1 * b–2) * (b * a–1 * b–1)
5 (a * b) * (a * b) * a–1 * b–1 * a–1 * b–1
5 (a * b)2 * (a–1 * b–1)2
5 (b * a)2 * (a–1 * b–1)2
5 (b * a)2 * (b * a)–2 5 e

Therefore, x2 5 e and hence, by hypothesis, x 5 e. From this it follows that

a * b * a–1 * b–1 5 e and hence a *b 5 (a–1 * b–1)–1 5 (b–1)–1 * (a–1)–1 5 b * a.
This (G, *) is an abelian group.

Worked Exercise 3.3.5. Let (G, *) be a group, a and b  G and m and n be

relatively prime positive integers such that

am 5 bm and an 5 bn.

Then prove that a 5 b.

Answer: Since m and n are relatively prime there exist integers r and s
such that

rm 1 sn 5 1

Now, consider

a 5 a1 5 arm 1 sn
5 (am)r * (an)s
5 (bm)r * (bn)s
5 brm1sn
5 b1 5 b

Worked Exercise 3.3.6. Let (G1, *), (G2, *), …, (Gn, *) be groups and

G 5 G1 3 G2 3 … 3 Gn.

For any a 5 (a1, a2, …, an) and b 5 (b1, b2, …, bn)  G, define

a * b 5 (a1 * b1, a2 * b2, …, an * bn).

Q001-Algebra-111001_CH 03.indd 41 9/16/2011 10:07:27 AM

 3-42   Algebra – Abstract and Modern

Then prove that (G, *) is a group and (G, *) is abelian if and only if each
(Gn, *) is abelian.

Answer: Clearly * is a binary operation on G and, using the associativity of

the operations on G1, G2, …, Gn, we can prove that * is associative on G. Also,
if e1, e2…, en are identities in G1, G2, …, Gn, respectively, then the element e 5
(e1, e2, …, en) becomes identity in (G, *). Further, for any a 5 (a1, a2, …, an)
in G, (a1–1, a2–1, …, an–1) is the inverse of a in (G, *). Thus, (G, *) is a group.
If each (Gi, *) is abelian, then

a * b 5 (a1, a2,…, an) * (b1, b2, …, bn)

5 (a1 * b1, a2 * b2, …, an * bn)
5 (b1 * a1, b2 * a2, …, bn * an)
5 (b1, b2, …, bn) * (a1, a2, …, an)
5b*a

For all a, b  G and hence (G, *) is abelian. Conversely suppose that (G, *)
is abelian. Fix 1 ≤ i ≤ n. For any ai and bi  Gi, consider

a 5 (e1, …, ei–1, ai, ei+1, …, en)

and b 5 (e1, …, ei–1, bi, ei+1, …, en).

Since (G, *) is abelian, we have a * b 5 b * a and, in particular, their ith coor-

dinates must be equal and therefore ai bi 5 biai. Thus, (Gi, *) is abelian for
all 1≤ i ≤ n.

Worked Exercise 3.3.7. Describe the group of symmetries of the set X of all
points on the perimeter of an equilateral triangle.

Answer: Let 1, 2 and 3 be vertices of an equilateral triangle and its altitudes

be as shown in the adjacent figure. Let X be the set of all points on the perim-
eter of the triangle.
3

5 4

1 6 2

Q001-Algebra-111001_CH 03.indd 42 9/16/2011 10:07:28 AM

 Groups   3-43

Recall that a symmetry of X is a bijection f of X onto itself such that

d(f (a), f (b)) 5 d(a, b)

for all a and b  X, where d(a, b) is the usual Euclidean distance between
a and b. Therefore, a symmetry of X should map each vertex to a vertex only
and hence we can identify the group (Sym(X), o) with the group (S(V), o),
where S(V) is the set of bijections of the set V of vertices onto V 5 {1,
2, 3}. It follows that Sym(X) has exactly b elements which are described
below. Each of these map each of 1, 2, 3 to the number given vertically
below that.

1 2 3 1 2 3 1 2 3

e    , a    , b   
1 2 3 2 3 1  3 1 2

1 2 3 1 2 3 1 2 3

c    , d    and s   .
1 3 2 3 2 1  2 1 3

These six elements e, a, b, c, d and s are related by the following equations.

a o a 5 b; a o a o a 5 e 5 c o c 5 d o d 5 s o s 5 b o b o b
c o d 5 a; d o c 5 a o a 5 b; d o s 5 a
s o d 5 b; c o s 5 b; s o c 5 a
a o c 5 s; c o a 5 d; a o d 5 c
d o a 5 s; s o a 5 c; a o s 5 d

The following table describes the binary operation o on Sym(X) 5 S(V)

o e a b c d s
e e a b c d s
a a b e s c d
b b e a d s c
c c d s e a b
d d s c b e a
s s c d a b e

Note that this group is not abelian, since a o c ≠ c o a.

Q001-Algebra-111001_CH 03.indd 43 9/16/2011 10:07:28 AM

 3-44   Algebra – Abstract and Modern

EXERCISE 3(c)

1. Prove the following for any elements a, b and c in a group G in which e is the
identity.
(i) a*b5e⇔b*a5e
(ii) (a * b) *c 5 e ⇔ (b * c) * a 5 e

2. Give an example of a finite semigroup satisfying the left cancellation law, but
not satisfying the right cancellation law.

3. Give an example of a finite semigroup satisfying the right cancellation law,

which is not a group.

4. Let (G, *) be a semigroup satisfying the following.

(i) (G, *) has a left identity e.
(ii) For each a  G, there exists a9  G such that a * a9 5 e.
Then prove that (G, *) is a group.

5. Prove that a group (G, *) is abelian if and only if (a * b)n 5 an * bn for all a and
b in G and for all integers n.

6. In any finite semigroup, prove that there exists an element e such that e2 5 e.

7. Let m and n be relatively prime positive integers and (G, *) a group such that
am * bm 5 bm * am and an * bn 5 bn * an
for all a and b  G. Then prove that (G, *) is an abelian group.

8. Let (G, *) be a finite group and suppose that the number of elements in G is
even. Then prove that there exists an element a, other than the identity, in G
such that a2 5 e.

9. For any element a in a finite group (G, *), prove that there exists a positive inte-
ger n such that an 5 e, the identity in G.

10. For any finite group (G, *), prove that there exists a positive integer n such that
an 5 e for all a  G
where e is the identity in (G, *).
11. For any elements a and b in a group (G, *) and for any positive integer n, prove that
(a * b * a–1)n 5 a * b * a–1 ⇔ bn 5 b.

12. Let (G, *) be a group and X be any nonempty set. Let GX be the set of all map-
pings of X into G. For any f and g  GX, define f * g : X → G by
(f * g)(x) 5 f (x) * g(x) for all x  G.

Q001-Algebra-111001_CH 03.indd 44 9/16/2011 10:07:28 AM

 Groups   3-45

Then prove that (GX, *) is a group which is abelian if and only if so is G.

13. Give an example of a nonabelian group having exactly six elements.

14. Prove that any group with fewer than six elements is abelian.

15. For any integer n . 1, prove that the set of all nonsingular n 3 n matrices over
R forms a nonabelian group under the multiplication of matrices.

16. Let (G, *) be a group. Define a new binary operation o on G by

aob5b*a
for all a and b in G. Then prove that (G, o) is a group which is abelian if and only
if (G, *) is abelian.

3.4 FINITE GROUPS AND GROUP TABLES

We know that a binary operation on a finite set can be represented by means
of a table. In this section, we shall take up finite groups and the description of
their group structure in terms of the table representing the binary operation.
First consider the smallest group. Any group should contain the identity ele-
ment e and hence {e} is the smallest group. Since e * e 5 e, the table for the
group ({e}, *) is trivial, as given below.

* e
e e

Next, we consider a two element group G. Then, there should be only one
element in G other than the identity e and therefore G 5 {e, a}, where a ≠ e,
we have a * e 5 a 5 e * a and e * e 5 e. What could be a * a? It cannot be a,
for, if a * a 5 a then a * a 5 a * e and, by the cancellation law, a 5 e which
is false. Therefore, the only possibility is a * a 5 e. The table for the group
(G, *), where G 5 {e, a}, is given below.

* e a
e e a
a a e

Next, we shall take up a 3-element group. In this case, there are exactly two
elements, say a ≠ b, in G other than the identity e. That is,

G 5 {e, a, b} and e, a and b are distinct.

Q001-Algebra-111001_CH 03.indd 45 9/16/2011 10:07:28 AM

 3-46   Algebra – Abstract and Modern

The table representing this group (G, *) should be like the one given below

* e a b
e e a b
a a
b b

Let us search for the possible entries for the vacant places in the table. First
observe that a * b ≠ a (since a * b 5 a 5 a * e ⇒ b 5 e, which is false).
Similarly, a * b ≠ b (since a ≠ e). Therefore, the only possibility is a * b 5 e
and b * a 5 e.

* e a b
e e a b
a a e
b b e

Next, we shall search for a * a. First of all a * a ≠ a (since a ≠ e). Also,

a*a5e⇒a*a*b5e*b5b
⇒a*e5b
⇒ a 5 b, which is false.

Therefore, a * a ≠ e and a * a ≠ a and hence the only possibility is a * a 5 b and

similarly b * b 5 a. Now, the table is complete and is given below

* e a b
e e a b
a a b e
b b e a

The above procedure for arriving at the full table representing the group
{e, a, b} yields the fact that the table of any three element group looks like
the same, except the interchanging of the elements a and b or relabeling the
elements a and b as b and a.
Consider the group (Zn, 1n), where Zn 5 {0, 1, 2, …, n21} and 1n is the
addition modulo n. For n 5 2 and n 5 3, the tables representing (Z2, 12) and

Q001-Algebra-111001_CH 03.indd 46 9/16/2011 10:07:28 AM

 Groups   3-47

(Z3, 13) look like exactly the above tables representing a 2-element group
and a 3-element group. In these cases, 0 is the identity, Z2 5 {0, 1} and
Z3 5 {0, 1, 2}.

13 0 1 2
12 0 1 0 0 1 2
0 0 1 1 1 2 0
1 1 0 2 2 0 1
(Z2, 12) (Z3, 13)

Careful examination of these tables reveal certain necessary conditions

that a table representing a binary operation on a finite set must satisfy
certain properties for the operation to give a group structure on the set.
There must be one element of the set which is the identity in the group
and is denoted by e. Since a * e 5 a for all elements a in the set, the col-
umn of the table under e at the very top must contain exactly the elements
appearing at the extreme left in the same order. Also, since e * a 5 a for
all elements a in the set, the row of the table opposite e at the extreme
left must contain exactly the same elements appearing across the very top
of table in the same order. Further, since any element of the set has left
inverse and right inverse, the row having a at the extreme left must contain
e in some place and the column under a at the top must contain e at some
place. Therefore, e must appear in each row and in each column. In fact,
for any elements a and b in the group, the equations a * x 5 b and y * a 5 b
have unique solutions in the group. This is equivalent to saying that, for a
given element a in the group, every element of the group appears exactly
once in the row with a at the extreme left and exactly once in the column
with a at the very top. In the following, we have formulated a converse of
the above argument.

Theorem 3.4.1. Let * be an associative binary operation on a nonempty finite

set G. Then, (G, *) is a group if and only if, for any a  G, every element of
G appears in that row with a at the extreme left and in the column with a at
the very top.

Proof: Consider the table representing the operation * on G. For any ele-
ments a and b in G, the equation a * x 5 b is solvable in G if and only if b
appears in the row with a at the extreme left. Also, the equation y * a 5 b is
solvable in G if and only if b appears in the column, with a at the very top.
Now, the theorem is a direct consequence of Theorem 3.3.3.

Q001-Algebra-111001_CH 03.indd 47 9/16/2011 10:07:29 AM

 3-48   Algebra – Abstract and Modern

Example 3.4.1. The table representing the group (Z9, 19) is given below.
Recall that Z9 5 {0, 1, 2, 3, 4, 5, 6, 7, 8} and 19 is the addition modulo 9.

19 0 1 2 3 4 5 6 7 8
0 0 1 2 3 4 5 6 7 8
1 1 2 3 4 5 6 7 8 0
2 2 3 4 5 6 7 8 0 1
3 3 4 5 6 7 8 0 1 2
4 4 5 6 7 8 0 1 2 3
5 5 6 7 8 0 1 2 3 4
6 6 7 8 0 1 2 3 4 5
7 7 8 0 1 2 3 4 5 6
8 8 0 1 2 3 4 5 6 7

Example 3.4.2. Let X 5 {1, 2, 3} and S(X) be the set of all bijections of X
onto itself. S(X) has six elements and these are

1 2 3 1 2 3 1 2 3

e    , a   , b   ,
1 2 3 2 3 1 3 1 2

1 2 3 1 2 3 1 2 3

c   , d    and s   
1 3 2 3 2 1 2 1 3

Then, (S(X), o) is a group, where o is the composition of mappings and its

table is given below.

o e a b c d s
e e a c d s
a b e c d
b b e a d c
c c d e a b
d d e
s s d a b e

The vacancies in the above table can be filled in using Theorem 3.4.1.

Q001-Algebra-111001_CH 03.indd 48 9/16/2011 10:07:29 AM

 Groups   3-49

Worked Exercise 3.4.1. Let X 5 {1, 2, 3} and P(X) be the set of all subsets
of X. Construct the table representing the group (P(X), 1), where 1 is the
symmetric difference.

Answer: P(X) has 23 (5 8) elements since X is a 3-element set. For conve-

nience, we label them as given below.
e 5 , a 5 {1}, b 5 {2}, c 5 {3},
p 5 {1, 2}, q 5 {2, 3}, r 5 {1, 3}, X5 {1, 2, 3}.

1 e a b c p q r X
e e a b c p q r X
a a e p r b X c q
b b p e q a c X r
c c r q e X b a p
p p b a X e r q c
q q X c b r e p a
r r c X a q p e b
X X q r p c a b e
(P(X), 1).

Worked Exercise 3.4.2. Let G7 = {1, 2, 3, 4, 5, 6} and ?7 be the multiplication

modulo 7. Construct the table representing the group (G7, ?7).

Answer:

?7 1 2 3 4 5 6
1 1 2 3 4 5 6
2 2 4 6 1 3 5
3 3 6 2 5 1 4
4 4 1 5 2 6 3
5 5 3 1 6 4 2
6 6 5 4 3 2 1
(G7, ?7).

Q001-Algebra-111001_CH 03.indd 49 9/16/2011 10:07:29 AM

 3-50   Algebra – Abstract and Modern

Worked Exercise 3.4.3. Examine whether the table below represent a group
structure.

* a b c d e
a b c d e a
b a d c a b
c d c b e c
d c a a b d
e a b c d e

Answer: The underlying set is X 5 {a, b, c, d, e}. If the table represents a

group (X, *), then the set of entries in each row and in each column must be
equal to the set X. The third row opposite to c at extreme left consists of d,
c, b, e and a is absent in this and hence the given table does not represent a
group structure on X. Note that c * x 5 a is not solvable in X. However, the
element e is the identity in the binary system (X, *).

EXERCISE 3(d)

1. Examine the following tables representing binary systems and determine which
of them represent a group structure.
(i) (ii)

* 0 1 2 3 4 * a b c
0 0 0 0 0 0 a a b c
1 0 1 2 3 4 b b c a
2 0 2 4 1 3 c c a b

3 0 3 1 4 2
4 0 4 3 2 1
(iii) (iv)

* a b c d e * 1 2 3 4
a b c d e a 1 1 2 3 4
b c d e a b 2 2 3 4 1
c d e a b c 3 3 4 2 3

d e a b c d 4 4 3 2 1

e a b c d e

Q001-Algebra-111001_CH 03.indd 50 9/16/2011 10:07:29 AM

 Groups   3-51

(v) (vi)

 i 2i 1 1 * 1 2 3 4
i 21 1 2i i 1 1 2 3 4
2i 1 21 i 2i 2 2 3 4 1
21 2i i 1 21 3 3 4 1 2
1 i 2i 21 1 4 4 1 2 3

2. Let G11 5 { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10} and ?11 be the multiplication modulo

11. Then prove that (G11, ?11) is a group and construct the table representing the
group.

3. Let n be a positive integer greater than n and S 5 {1, 2, …, n 2 1}. Prove that
the multiplication modulo n is a binary operation on S if and only if n is a prime
number.

4. Construct tables representing all the 2-element groups, 3-elements groups,

4-element groups and 5-element groups.

5. By observing the tables in 4 above, prove that every 2-element groups, 3-element
groups, 4-element groups and 5-element groups is abelian.

6. Prove that a finite group (G, *) is not abelian only if |G| . 5.

7. Give an example of a nonabelian group with exactly six elements.

8. For any positive integer n, give an example of an abelian group with exactly n
elements and construct a table representing it.

9. Let G be the set of all rational numbers with odd denominators. Prove that (G,
1) is a group, where 1 is the usual addition of rational numbers.
 0 1 0 i 
10. Let A    and B   , where i is the complex number such that i2 5 1.
1 0  i 0
Let
Q8 5 {(AnBm) : n, m  Z}.
Prove that (Q8, ?) is a group, where ‘?’ is the usual multiplication of complex
numbers and that Q8 has exactly 8 elements. This group is called the Quaternion
group of order 8.

Q001-Algebra-111001_CH 03.indd 51 9/16/2011 10:07:30 AM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 03.indd 52 9/16/2011 10:07:30 AM

 4
Subgroups and
Quotient Groups
4.1 Subgroups
4.2 Cyclic Groups
4.3 Cosets of a Subgroup
4.4 Lagrange’s Theorem
4.5 Normal Subgroups
4.6 Quotient Groups

A nonempty set together with certain operations on it is called an algebraic

system. In the study of any algebraic system, the subsets of the underlying set
which are closed under the operations of the algebraic system called subsys-
tems or substructures. In general, we are not interested in arbitrary subsets of
the underlying set A in the algebraic system; for, they do not reflect the fact that
A has an algebraic structure imposed on it. Whatever subsets we do consider
will be those endowed with algebraic properties derived from those of the given
algebraic system. In this chapter, we introduce the notion of a subgroup of a
group and study various properties of subgroups of a several group.
First let us slightly change our notation followed till now. If * is the binary
operation on a given group (G, *) and a and b are arbitrary elements of G, it is
cumbersome to write a * b. Let us agree to write simply ab for a * b, without
writing the specific symbol * in between a and b. Already we are practicing
this; for example, for any two real numbers a and b, we write ab to denote
the product of a and b. Here afterwards, we simply write ab for a * b, except
on specific occasions where the binary operation in the group is a special one
we are familiar with. For example, when we consider the group (R, 1), we
write a 1 b.

Q001-Algebra-111001_CH 04.indd 1 9/22/2011 11:24:44 AM

 4-2   Algebra – Abstract and Modern

Also, instead of saying that ‘(G, *) is a group’, we simply say that ‘G is a

group’ or ‘G is a group under *’. It is not that * is unimportant and need not
be mentioned. Actually the binary operation in a group is like a backbone to
the group structure. However, for convenience and simplicity, we ignore to
mention the binary operation *, when only one operation is under consider-
ation and there is no ambiguity. When we consider two binary operations on
the same set, then we invariably specify the binary operation which is under
consideration. When there is no ambiguity about the binary operation, we use
G to denote the group as well as the underlying set of the group.

4.1 SUBGROUPS
Recall that a binary operation * on a set S is a mapping of S 3 S into S. If
A is a subset of S, then A 3 A is a subset of S 3 S and if the restriction of * to
A 3 A is a binary operation on A, then the restriction also will be denoted by
* and is called the operation A induced by the operation * on S. In particular,
if (G, *) is a group and A is a subset of G such that a * b  A whenever a and
b  A, then * can be treated as a binary on A with respect to which A can be
a group and, in this case, we say that A is a group under *.

Definition 4.1.1. Let (G, *) be a group. A subset H of G is said to be a sub-

group of (G, *) if H on its own becomes a group under *.
Before going, for example, we obtain equivalent conditions on a subset
of a group to be a subgroup. These facilitate us in checking whether a given
subset is a subgroup.

Theorem 4.1.1. The following are equivalent to each other for any nonempty
subset H of a group G:
1. a and b  H ⇒ ab  H and a21  H.
2. a and b  H ⇒ ab21  H.
3. a and b  H ⇒ a21b  H.
4. H is a subgroup of G.

Proof: Let G be a group and H be a nonempty subset of G.

(1) ⇒ (2): a and b  H ⇒ a and b21  H (by (1))

⇒ ab21  H (by (1))

(2) ⇒ (3): Suppose that a and b  H ⇒ ab21  H.

If a  H, then e 5 aa21  H and hence

a  H ⇒ a21 5 ea21  H (since e and a  H)

Q001-Algebra-111001_CH 04.indd 2 9/22/2011 11:24:45 AM

 Subgroups and Quotient Groups    4-3

Now, a and b  H ⇒ a21 and b21  H

⇒ a21(b21)21  H
⇒ a21b  H

(3) ⇒ (4): Suppose that a and b  H ⇒ a21b  H.

Since H is nonempty, we can choose h  H. Then,

e 5 h21h  H

Also, a  H ⇒ a21 5 a21e  H (since a and e  H)

Now, a and b  H ⇒ a21 and b  H
⇒ (a21)21b  H
⇒ ab  H

That is, a * b  H whenever a and b  H. Therefore, the operation * on G,

when restricted to H, becomes a binary operation on H. Since * is associative
on G, so is on H. Also, since e  H and e is the identity in (G, *), e becomes
the identity in (H, *) also. Further, for any a  H, a21  H and aa21 5 e 5
a21a and hence a21 is the inverse of a in (H, *). Thus, (H, *) is a group and
hence H is a subgroup of G.
(4) ⇒ (1): Suppose H is a subgroup of G. Then, (H, *) is a group and
hence a * b  H, whenever a and b  H. Also, for any a  H, the inverse of
a exists in H also. Let a9 and a21 be inverses of a in H and G, respectively.
Then both these are inverses of a in G and hence equal. Therefore, a21 5 a9
 H. Now,

a and b  H ⇒ ab  H and a21  H.

If the symbol 1 is used to denote the binary operation in a group, then we

write 2a for the inverse a21 of a and write a 2 b for ab21 (for psychological
reasons!). In this case, a nonempty subset H of G is a subgroup of G if and
only if a 2 b  H whenever a and b  H. Recall that, in the group (R, 1)
or (Z, 1) or (C, 1), 0 is the identity and 2a is the inverse of an element a.
Also, the element an defined in the Definition 3.2.5 will be denoted by na,
when 1 is used to denote the binary operation. This is only for not violating
our usual practice right from the elementary school stage. Recall that we are
habituated to write

na for a 1 a 1…1 a (n times)

and   an for a ? a ? a ? … ? a (n times)

Q001-Algebra-111001_CH 04.indd 3 9/22/2011 11:24:45 AM

 4-4   Algebra – Abstract and Modern

2 a for the inverse of a in (R, 1)

a 2 b for a 1 (2b)
1
for b21, the inverse of b in (R 2{0}, ?)
b
a
for ab21.
b
When H is a nonempty finite subset of a group G, we get a simpler crite-
rion for H to be a subgroup of G, which gives a simpler procedure to check
whether a given finite subset is a subgroup. This criterion is obtained in the
following theorem.

Theorem 4.1.2. Let H be a nonempty finite subset of a group G. Then, H is a

subgroup of G if and only if a and b  H ⇒ ab  H.

Proof: Suppose that a and b  H ⇒ ab  H. In order to prove that H is a sub-

group of G, we have to only prove that a2l  H whenever a  H (by Theorem
4.1.1). Now, let a  H. Since H is given to be finite, we can write

H 5 {a1, a2, … , an}, ai  aj for i  j.

Consider the set aH 5 {aa1, aa2, … , aan}.

We have aai 5 aaj ⇒ ai 5 aj (by cancellation law)

⇒i5j

and hence aa1, aa2, …, aan are all distinct. Also, since a and ai  H, we have
aH ⊆ H and H and aH have the same number of elements. From the finiteness
of H, it follows that aH 5 H. In particular,

a  H 5 aH

and hence a 5 aai for some 1# i # n.

Since all these are elements in the group G, we get that e 5 ai  H. Now e 
H 5 aH and hence

e 5 aaj for some 1 # j # n.

Therefore, a21 5 aj  H. Thus, H is a subgroup of G.

The finiteness of H in the above theorem is necessary; for, consider the
set Z1 of positive integers. Then, Z1 is a subset of Z, (Z, 1) is a group

Q001-Algebra-111001_CH 04.indd 4 9/22/2011 11:24:52 AM

 Subgroups and Quotient Groups    4-5

and a 1 b  Z1 whenever a and b  Z1 and still Z1 is not a subgroup of

(Z, 1).
Now, we collect certain examples of subgroups of groups. Some of the
subgroups given below are earlier given as examples of groups.

Example 4.1.1
1. If H is a subgroup of a group K and K is a subgroup of a group G,
then clearly H is a subgroup of G. If 1 denotes the usual addition
of numbers, Z is a group of (Q, 1), Q is a subgroup of (R, 1) and
R is a subgroup of (C, 1) and hence Z, Q and R are all subgroups
of (C, 1).
2. If e is the identity in a group G, then clearly {e} and G are subgroups of
G and are called the trivial subgroups of G. Any subgroup other than {e}
and G is called a nontrivial subgroup. A group G is called nontrivial if
G  {e} and trivial if G 5 {e}.
3. If ? is the usual multiplication of numbers, then Q 2 {0} is a subgroup
of (R 2 {0}, ?) and R 2 {0} is a subgroup of (C 2 {0}, ?). Therefore,
both Q 2 {0} and R 2 {0} are subgroups of (C 2 {0}, ?).
4. Let X be any nonempty set and (S(X), o) be the group of bijections of X
onto itself, where o is the composition of mappings. Let x0 be an arbi-
trary element of X and

H x0  { f ∈ S ( X ) : f ( x0 )  x0 }.

Then, H x0 is a subgroup of (S(X), o) Also, for any subset Y of X, the set

HY 5 {f  S(X) : f(y) 5 y for all y  Y}

is a subgroup of (S(X), o).

5. Let n be a positive integer and NSMn(R) be the set of all nonsingular n 3 n
matrices over the real number system R. Then, (NSMn(R), ?) is a group
where ? is the usual multiplication of matrices. Let
H 5 {(aij)  NSMn(R) : aij 5 0 for all i . j}
and    K 5 {(aij)  NSMn(R) : aij 5 0 for all i , j}.

Then, H and K are subgroups of NSMn(R).

1 a 
6. Let A    : a ∈ R. Then, A is a subgroup of (NSM2(R), ?). One
0 1 
can easily verify that, for any a and b  R,

Q001-Algebra-111001_CH 04.indd 5 9/22/2011 11:24:53 AM

 4-6   Algebra – Abstract and Modern

1 a 1 b 1 a  b

  ⋅     
0 1 0 1 0 1 
1
1 a 1 a
and     .
0 1 0 1 

7. Consider the additive group (Z4, 14) of integers modulo 4. We have Z4

5 {0, 1, 2, 3} and 0 is the identity. If H is a nontrivial subgroup of Z4,
then H 5 {0, 2} (since 1  H ⇒ 2 5 1 14 1  H and 3 5 1 14 1 14 1
 H and 3  H ⇒ 1 5 3 14 3 14 3  H ⇒1, 2, 3  H). Therefore, {0},
{0, 2} and Z4 are the only subgroups of (Z4, 14).
8. The group (Z5, 15) has no nontrivial subgroups; for, if H is a subgroup
of Z5, then
1  H ⇒ 1, 2, 3, 4  H ⇒ H 5 Z5
2  H ⇒ 2 15 2 15 2  H ⇒ 1  H ⇒ H 5 Z5
3  H ⇒ 1 5 3 15 3  H ⇒ H 5 Z5
4  H ⇒ 3 5 4 15 4  H ⇒ H 5 Z5.
Since every subgroup of a group G should contain the identity e in G, it
follows that {e} is the smallest subgroup of G. Also, clearly G is the largest
subgroup of G. In the following theorem, we describe the smallest subgroup
of a group G containing a given element of the group.

Theorem 4.1.3. Let G be a group and a  G. Let

<a> 5 {an : n is an integer}.

Then, <a> is the smallest subgroup of G containing a.

Proof: Recall that an is defined as

e if n  0

a  a n1a
n
if n  0
 1 n
( a ) if n  0

and hence an  H for all integers n and for all subgroups H of G containing
a. Also, by Worked Exercise 3.2.18, <a> is a subgroup of G and contains a.
Thus, <a> is the smallest subgroup containing a.

Definition 4.1.2. For any group G and a  G, <a> 5 {an : n is an integer} is

called the cyclic subgroup generated by a in G.

Q001-Algebra-111001_CH 04.indd 6 9/22/2011 11:24:54 AM

 Subgroups and Quotient Groups    4-7

Before going for a detailed discussion on cyclic subgroups generated by

elements of an arbitrary group, we shall first discuss certain elementary prop-
erties of subgroups.

Theorem 4.1.4. The intersection of any class of subgroups of a group G is

again a subgroup of G.

Proof: Let # be a class of subgroups of a group G and

let H  ∩ C  {a : a ∈ C for all C ∈ }.
C ∈
If # is empty class, then, by logical convention, H 5 G and hence H is a sub-
group of G. Therefore, we can assume that # is a nonempty class.
Since the identity element e must be in every subgroup, we get that e  H and
hence H is a nonempty subset of G. Now,

a and b  H ⇒ a and b  C, for all C  #

⇒ ab21  C, for all C  #
⇒ ab21  H.

Therefore, H is a subgroup of G.

Definition 4.1.3. For any subset S of a group G, let <S > be the intersection
of all subgroups of G containing S. Then, by the above theorem, <S > is a sub-
group of G containing S and is called the subgroup generated by S in G.

Note 4.1.1
1. For any subset S of a group G, <S > is the smallest subgroup of G con-
taining S.
2. <> 5 {e} and <G> 5 G, for any group G.
3. <{a}> 5 <a> 5 {an : n  Z} for any a  G.
4. For any nonempty subset S of a group G, S is a subgroup of G if and only
if S 5 <S >.
In the following, we describe the elements of <S >, for any nonempty sub-
set S of a group. The description of elements of <{a}> is already given in
Theorem 4.1.3. This is generalised in the following theorem.

Theorem 4.1.5. Let S be a nonempty subset of a group G and <S > be the
smallest subgroup of G containing S. Then,

 n 
 S  ∏ si : n ∈  and, for each i, si or si1 ∈ S .
 i=1 

Q001-Algebra-111001_CH 04.indd 7 9/22/2011 11:25:01 AM

 4-8   Algebra – Abstract and Modern

Proof: Let A be the set defined on the right side of the required equality. If
H is any subgroup of G containing S, then s and s21  H for all s  S and
therefore any product of elements of S and their inverses must be in H. This
is to say that A ⊆ H for all subgroups H containing S. Also,   S ⊆ A and,
since

( s1 s2 … sn ) (t1 t 2 … t m )1  s1 s2 … snt m1t m11 … t 21t1 .

Therefore, A is also a subgroup of G containing S. Thus, <S > 5 A.

Corollary 4.1.1. For any subset S of a group G,

 S  S 1 , where S 1  {s1 : s ∈ S}.

We have proved in Theorem 4.1.4 that the intersection of any class of sub-
groups is again a subgroup. A similar statement is not true for unions of
subgroups. In this context, we have the following theorem.

Theorem 4.1.6. Let A and B be subgroups of group G. Then, A ∪ B is a sub-

group of G if and only if either A ⊆ B or B ⊆ A.

Proof: If A ⊆ B, then A ∪ B 5 B and, if B ⊆ A then A ∪ B 5 A and hence,

in this case, A ∪ B is a subgroup of G. Conversely suppose that A ∪ B is a
subgroup of G. Assume that A  B. Then, there exists a  A such that a 
B. Now,

b ∈ B ⇒ a and b ∈ A ∪ B (since a ∈ A)
⇒ ab ∈ A ∪ B (since A ∪ B is a subgroup)
⇒ ab ∈ A or ab ∈ B
⇒ ab ∈ A (since ab ∈ B ⇒ a ( ab)b1 ∈ B )
⇒ b a1 ( ab) ∈ A

Therefore, B ⊆ A.
From the above, it follows that, for any subgroups A and B of a group G,
A ∪ B is not a subgroup in general. However, we have noticed earlier that
A ∩ B is always a subgroup and this is the largest subgroup contained in both
A and B. Also, there is a smallest subgroup containing both A and B (which
need not be A ∪ B). In certain cases, we can describe the elements of this
elegantly. First, we have the following definition.

Q001-Algebra-111001_CH 04.indd 8 9/22/2011 11:25:02 AM

 Subgroups and Quotient Groups    4-9

Definition 4.1.4. Let G be a group and A and B be subsets of G. Define

AB 5 {ab : a  A and b  B} and A21 5 {a21 : a  A}.

Note that, from the associativity of the operation in G, we get that (AB)
C 5 A(BC) for any subsets A, B and C of G. Also, observe that a nonempty
subset A of G is a subgroup of G if and only if AA21 5 A.

Theorem 4.1.7. Let A and B be subgroups of a group G. Then, AB is a sub-

group of G if and only if AB 5 BA and, in this case, AB is the smallest sub-
group of G containing both A and B.

Proof: Suppose that AB is a subgroup of G. Then,

AB  ( AB)1  B1 A1  BA.

Conversely, if AB 5 BA, then

( AB )( AB )1  ( AB )( B1 A−1 )

 A( BB1 ) A1
 ( AB) A1
 ( BA) A1
 B( AA1 )  BA  AB

and hence AB is a subgroup of G.

Since e  A and e  B, if follows that

A  A{e} ⊆ AB and B  {e}B ⊆ AB.

If H is any subgroup containing A and B, then clearly AB ⊆ H. Thus, when

AB 5 BA, AB is the smallest subgroup containing both A and B.

Corollary 4.1.2. For any subgroups A and B of an abelian group G, AB is also

a subgroup of G.

Worked Exercise 4.1.1. Prove that any subgroup of the group (Z, 1) is the
subgroup generated by a single nonnegative integer.

Answer: Recall that, when the symbol 1 is used for the binary operation in
a group, then we write na for an and as such

<a> 5 {na : n is an integer}.

Q001-Algebra-111001_CH 04.indd 9 9/22/2011 11:25:02 AM

 4-10   Algebra – Abstract and Modern

Now, let A be a subgroup of (Z, 1). If A 5 {0}, then clearly A 5 <0>. Sup-
pose that A  {0}. Then, there exists a  0 such that a  A. Since A is a
subgroup, 2a also is in A. Since a or 2a is positive, it follows that A ∩ Z1
is a nonempty subset of Z1. By the well-ordering principle, A ∩ Z1 has the
smallest member, say m. Then, since m  A, we get that <m> ⊆ A. On the
other hand, let x  A. By the division algorithm, we can write

x 5 qm 1 r, q, r  Z and 0 # r , m.

Then, r 5 x 2 qm  A (since x and m  A). Since r , m and since m is the

least positive integer in A, it follows that r 5 0 and hence x 5 qm  <m>.
Thus, A 5 <m> 5{nm : m  Z} 5 nZ.

Worked Exercise 4.1.2. Compute all subgroups of the group (nZ, 1) for any
positive integer n.

Answer: Let n be a positive integer and A be a subgroup of (nZ, 1). Then, A

is a subgroup of (Z, 1) and hence A 5 <a> for some a $ 0. Since A ⊆ nZ, we
get that a  nZ and hence a 5 nq for some q  Z. Therefore, a is multiple of
n. Conversely, if a is a multiple of n, then aZ is a subgroup of (nZ, 1). Thus,
the subgroups of (nZ, 1) are precisely of the form aZ, where a is an integral
multiple of n.

Worked Exercise 4.1.3. Compute all the subgroups of (Zn, 1n) for any posi-
tive integer.

Answer: Let n be a positive integer and A be a subgroup of (Zn, 1n). Suppose

that A  {0}. Let m be the least positive integer in A. As in Worked Exercise
4.1.1, we can prove that A 5 <m> and that m is a positive divisor of n. Note
that, since m  A ⊆ Zn, m , n. Therefore, the subgroups of (Zn, 1n) are
precisely of the form <m>, where m 5 0 or a positive divisor of n. Note that,
for m  0,

<m> 5 {0, m, 2m, …, (q 2 1)m},

where qm 5 n.

EXERCISE 4(a)

1. Determine whether the set given is a subgroup of the group in each of the
following.
(i) {0, 1, 2, 3, 4} in (Z8, 18)
(ii) {0, 3, 6, 9} in (Z12, 112)

Q001-Algebra-111001_CH 04.indd 10 9/22/2011 11:25:03 AM

 Subgroups and Quotient Groups    4-11

(iii) R1 in (R, 1)
(iv) R1 in (R2{0}, ?)
(v) Q1 in (R1, ?)
(vi) pQ in (R, 1)
(vii) {z  C | |z| 5 1} in (C 2 {0}, ?)
(viii) 5Z in (8Z, 1)

2. Determine whether the following are subgroups of the group of nonsingular n 3 n

matrices under the usual multiplication of matrices, where n is a given positive
integer.
(i) The set of all n 3 n matrices whose determinant is 2.
(ii) The set of all n 3 n matrices whose determinant is 1.
(iii) The set of n 3 n matrices whose determinant is a positive real number.
(iv) {( aij ) ∈ NSM n ( R) : aij  0 for all i  j}.
(v) {( aij ) ∈ NSM n ( R ) : aij  0 for all i  j}.
(vi) {( aij ) ∈ NSM n ( ) : aij  0 for i  j}.
(vii) The set of all n 3 n matrices whose determinant is a negative real number.
(viii) {( aij ) ∈ NSM n ( R) : aij  0 for i j and aii  a jj for all i and j}.

3. Let # be a nonempty class of subgroups of a group G such that, for any A and B
in #, there is a member C in # containing both A and B. Then prove that ∪ C
C ∈
is a subgroup of G.

4. Let G be a group such that G 5 <a> for exactly, one element a in G. Then prove
that G has at most two elements.

5. Let G be a group having exactly two subgroups. Then prove that G 5 <a> for
some a  G.

6. Let n be positive integer and consider the group (Zn, 1n) of integers modulo n. For
any 0 , d , n, prove that Zn 5 <d> if and only if d is relatively prime with n.

7. Prove that there is a bijection between the set of subgroups of (Zn, 1n) and the
set of positive divisors of n.

8. Let G be a group and a  G. Prove that the set

Ca 5 {x  G : ax 5 xa}
is a subgroup of G. Ca is caller the centralizer of a.

9. Let S be any subset of G. Then prove that the set

CS 5 {x  G : ax 5 xa for all a  S}
is a subgroup of G.

Q001-Algebra-111001_CH 04.indd 11 9/22/2011 11:25:04 AM

 4-12   Algebra – Abstract and Modern

10. For any group G, prove that the set

Z(G) 5 {x  G : ax 5 xa for all a  G}

is a subgroup of G. Z(G) is called the centre of G.

11. Determine all the subgroups of each of the following:

(i) (Z24, 124)
(ii) (Z7, 17)

12. For any positive integer n, prove that (Zn, 1n) has exactly two subgroups if and
only if n is prime.

13. For any subgroup H of a group G and a  G, prove that

aHa21 5 {axa21 : x  H} is also a subgroup of G.

14. For any finite subgroup H of a group G and a  G prove that H and aHa21 has
equal number of elements.

15. Determine all the subgroups of the group (S(X), ?) of bijections of X onto itself,
where X is a 3-element set.

16. Prove that a nonempty subset H of a group G is a subgroup of G if and only if

HH21 ⊆ H.

4.2 CYCLIC GROUPS

The concept of a cyclic group is an important tool in determining the struc-
ture of a finite (or finitely generated) abelian group. In fact, we prove later
that any finitely generated abelian group is a finite product of cyclic groups.
In order to understand the structure of a finitely generated abelian group, one
has to understand cyclic group. In this section, we thoroughly discuss the
various properties of cyclic groups.
First let us recall that, for any element a of a group G, the smallest sub-
group of G containing a is given by

<a> 5{an : n is an integer}

and is called the cyclic subgroup generated by a in G.

If one uses 1 to denote the binary operation in a group, we write na for an
and 2a for the inverse of a.

Definition 4.2.1. A group G is called a cyclic group if G 5 <a> for some

a  G; that is, there exists a  G such that

Q001-Algebra-111001_CH 04.indd 12 9/22/2011 11:25:04 AM

 Subgroups and Quotient Groups    4-13

G 5 {an : n is an integer}.

In this case, a is called a generator of G.

Example 4.2.1

1. (Z, 1) is a cyclic group, since <1> 5 Z. In fact, Z 5 <21> also and

hence both 1 and 21 are generators of the group Z. Later we will be
proving that these are the only generators of Z.
2. For any positive integer n, the group (Zn, 1n) of integers modulo n is a
cyclic group. Here also, 1 and 21 (5 n 2 1) are generators of Zn. Later,
we shall prove that any positive integer less than n and relatively prime
with n is a generator of Zn.
3. Let G 5 {1, 21, i, 2i}. Then, G is a cyclic group under the usual multi-
plication of complex numbers. Here, <i> 5 G 5 <2i > and hence i and
2i are the only generators of G, as we can easily see that <1> 5 {1} 
G and <21> 5 {1, 21}  G.
4. The group Z2 3 Z2 is not cyclic, since, for any element a in this group,
2a 5 0, the identity and hence <a> 5 {0, a}  Z2 3 Z2.

The following is a fundamental tool in the study of cyclic groups and we

might have used this earlier. Here, we offer a proof. First recall that for any
real number a, there exists largest integer, less than or equal to a and it is
denoted by [a]. That is, [a] is the unique integer such that [a] # a , [a] 11.
[a] is called integral part of a.

Theorem 4.2.1 (The Division Algorithm). Let n be a positive integer and a be

any integer. Then, there exist unique integers q and r such that

a 5 qn 1 r and 0#r,n

q and r are respectively called the quotient and the remainder obtained by
dividing a with n.

Proof: Let q [ an ], the integral part of an , and r 5 a 2 qn. Then, q is an

integer such that

a
q   q 1.
n

Q001-Algebra-111001_CH 04.indd 13 9/22/2011 11:25:05 AM

 4-14   Algebra – Abstract and Modern

Since a, q and n are all integers, r is also an integer. We have

qn # a , qn1n

and hence 0 # a 2 qn , n.
Therefore, we have a 5 qn 1 r and 0 # r , n. For proving the uniqueness, let
q0 and r0 be any integers such that a 5 q0n 1 r0 and 0 # r0 , n.
r r
Then, an  q0  n0 and 0  n0  1. Therefore,

a a
q0   q0 1 and hence q0     q
n  n 

and r0 5 a 2 q0n 5 a 2qn 5 r.

Example 4.2.2
1. The quotient and remainder when 46 is divided by 8 are respectively 5
and 6, since
46 5 5 ? 8 1 6 and 0 # 6 , 8.

2. The quotient and remainder when 46 is divided by 8 are respectively 26

and 2, since
246 5 (26)8 1 2 and 0 # 2 , 8.
We shall make use of the division algorithm in proving the following
theorem.

Theorem 4.2.2. Every subgroup of a cyclic group is cyclic.

Proof: Let H be a subgroup of a cyclic group G and let a  G such that

G 5 <a> 5 {an : n is an integer}.

If H 5 {e}, we are through, since, H 5 <e>. Suppose that H  {e}. Since,

for any n,

an  H ⇔ a2n 5 (an)21  H,

it follows that there exists a positive integer n such that an  H. By the well-
ordering principle, there exist least positive integer m such that am  H. Since
H is a subgroup and am  H, we have <am> ⊆ H. On the other hand, let x  H.

Q001-Algebra-111001_CH 04.indd 14 9/22/2011 11:25:06 AM

 Subgroups and Quotient Groups    4-15

Then, x  G and hence x 5 an for some integer n. By the division algorithm,

there exist integers q and r such that

n 5 qm 1 r and 0 # r , m.

Now, ar 5 an2qm 5 an(am)2q  H (since an, am  H). By the least property of

m, r should not be positive and hence r 5 0 and n 5 qm, so that

x 5 an 5 aqm 5 (am)q  <am>.

Therefore, H ⊆ <am> and hence H 5 <am>. Thus, H is a cyclic group.

Theorem 4.2.3. Every cyclic group is abelian.

Proof: Let G be a cyclic group and a  G such that

G 5 <a> 5 {an : n is an integer}.

Let x and y  G. Then, there exist integers m and n such that x 5 am and
y 5 an. Now,

xy 5 aman 5 am1n 5 an1m 5 anam 5 yx.

Thus, G is an abelian group.

The converse of the above theorem is not true; for example, the group
Z2 3 Z2 is an abelian group which is not cyclic (see Example 4.2.1 (4)).

Example 4.2.3. Consider the cyclic group (Z, 1) in which Z 5 <1>. Following
the proof of Theorem 4.2.2, one can prove that any subgroup of (Z, 1) must be
of the form <n> 5 {mn : m  Z} 5 Zn for some nonnegative integer n.

Example 4.2.4. Let n be a positive integer and consider the group (Zn, 1n)
of integer modulo n. Here again one can prove that any subgroup of (Zn, 1n)
must be of form <d> 5 {0, d, 2d, 3d, …, (m 2 1)d}, where d is a positive
divisor of n and md 5 n (see Worked Exercise 4.1.3).

Definition 4.2.2. If G is a finite group, then the number of elements in G is

called the order of G and is denoted by |G|.
Note that |G| is precisely the cardinality of G. If |G| 5 n, then G is called
a group of order n. In the following, we define the concept of the order of an
­element in a group which, in the finite case, turns out to be the order of the
cyclic subgroup <a> generated by a.

Q001-Algebra-111001_CH 04.indd 15 9/22/2011 11:25:06 AM

 4-16   Algebra – Abstract and Modern

Definition 4.2.3. Let G be a group and a  G. If an 5 e for some positive

integer n, then the least positive integer m such that am 5 e is called the order
of a and is denoted by O(a). In this case, a is said to be an element of order m.
If an  e for all positive integers n, then the order of a is defined to be infinity
and a is said to be an element of infinite order.
In other words, a is said to be an element of order m if m is the least ele-
ment in the set {n  Z1: an 5 e}. If this set is empty, then a is said to be of
infinite order.

Theorem 4.2.4. Let a be an element of finite order in a group G, then the

following holds:

1. O(a21) 5 O(a)
2. For any integer n, there exists 0 # r , O(a) such that an 5 ar.
3. For any integer n, an 5 e if and only if O(a) divides n.
4. <a> 5 {e, a, a2, …, am21}, where m 5 O(a).

Proof:

1. This is a direct consequence of the fact that, for any integer n, an 5 e if

and only if (a21)n 5 e (since (a21)n 5 a2n 5 (an)e).
2. Let n be any integer. By the division algorithm, there exist integers q and
r such that
n 5 qO(a) 1 r and 0 # r , O(a).

Now, an 5 aqO(a)1r 5 aqO(a)∙ar 5 (aO(a))qar 5 eqar 5 ear 5 ar.

3. Let n be an integer. If O(a) divides n, then n 5 qO(a) for some integer q
and hence
an 5 aqO(a) 5 (aO(a))q 5 eq 5 e

conversely suppose that an 5 e. Then, as in (2), we can write n 5 qO(a)

1 r for some integers q and r such that 0 # r , O(a). Then, as in (2),
ar 5 an 5 e

Since 0 # r , O(a) and O(a) is the least positive integer such that aO(a)
5 e, it follows that r 5 0 and n 5 qO(a). Thus, O(a) divides n.
4. <a> 5 {an : n is an integer}
5 {ar : 0 # r , O(a)}
5 {e 5 a0, a, a2, ….., aO(a)21}.

Q001-Algebra-111001_CH 04.indd 16 9/22/2011 11:25:06 AM

 Subgroups and Quotient Groups    4-17

Corollary 4.2.1. For any element a of finite order in a group,

O(a) 5 |<a>|.

That is, the order of the element a is precisely equal to the order of the cyclic
group <a>.

Proof: In the above theorem, we have proved that

<a> 5 {e, a, a2, …, aO(a)21}.

Also, if ar 5 as, then ar2s 5 e 5 as2r and hence O(a) divides r 2 s and s 2 r
and; if 0 # r and s , O(a), then r 5 s. Thus, e, a, a2, …, aO(a)21 are distinct.
Thus, |<a>| 5 O(a).
It is well known that the greatest common divisor (g.c.d.) of two positive
integers m and n can be written as a linear combination m and n. We shall
prove this using Theorem 4.2.2.

Theorem 4.2.5. Let m and n be two positive integers and (m, n) be the great-
est common division of m and n. Then, there exist integers r and s such that

(m, n) 5 rm 1 sn.

Proof: Let H 5 {am 1 bn : a and b  Z}. Note that H 5 Zm 1 Zn 5

<m> 1 <n>. Since (Z, 1) is a cyclic group and H is a subgroup of Z (by
­Corollary 4.1.2), we get that H is a cyclic subgroup of (Z, 1) (by Theorem
4.2.2). ­Therefore, there exists a positive integer d such that

H 5 <d> 5 Zd.

In particular, since d  H, d 5 rm 1 sn for some integers r and s. We shall

prove that d is the g.c.d. of m and n. Since m 5 1m 1 0n  H 5 Zd, d is a
divisor of m. Similarly, d is a divisor of n. Also, if q is any common divisor
of m and n, then

m 5 qk and n 5 qt for some k, t  Z

and hence m and n  <q> so that am 1 bn  <q> for all integers a and b and,
in particular, d 5 rm 1 sn  <q> 5 Zq. Therefore, q divides d. Thus, d is the
greatest common divisor of m and n.
The converse of the above therefore is also true, in the sense that, if d is a
common divisor of m and n and d is of the form rm 1 sn, r, s  Z, then d is

Q001-Algebra-111001_CH 04.indd 17 9/22/2011 11:25:06 AM

 4-18   Algebra – Abstract and Modern

the greatest common divisor of m and n. In particular, m and n are relatively

prime if and only if 1 5 rm 1 sn for some r and s  Z.

Theorem 4.2.6. Let G be a group and a  G such that O(a) 5 m , ∞. Then,

for any 0 # r , m,

m
O(a r ) 
( m, r )

where (m, r) is the g.c.d. of m and r.

Proof: Let 0 # r , m be fixed and d 5 (m, r). Then, by Theorem 4.2.5, there
exist integers s and t such that

d 5 sm 1 tr.

Put b 5 ar. Since d divides both m and r, m and r are integers and
d d
m  r 
1  s   t  
 d   d 

and therefore m and r are relatively prime.

d d
We have

m m rm r
b d  ( a r ) d  a d  ( a m ) d  e.

On the other hand, for any integer q,

bq  e ⇒ (ar )q  e
⇒ a rq  e
⇒ O(a) divides rq
⇒ m divides rq
m r
⇒ divides ⋅ q
d d
m    
⇒ divides q since  m , r   1.
d 
  d d  

Therefore, m is the least positive integer k such that bk 5 e.

d

Q001-Algebra-111001_CH 04.indd 18 9/22/2011 11:25:08 AM

 Subgroups and Quotient Groups    4-19

m m
∴ O (a r ) O (b)  .
d (m, r )


Corollary 4.2.2. Let G be a group, a  G and O(a) 5 m , ∞. If d is a posi-

tive divisor of m, then

m O (a)
O (a d )  .
d d

Proof: This follows from the above theorem and the fact that, for any posi-
tive divisor d of m, (m, d) 5 d.
Let us recall that, if G 5 <a>, then a is called a generator of the cyclic
group G. In the next result, we derive formulae to determine the number of
generators of a cyclic group. First, we have the following definition.

Definition 4.2.4. For any positive integer n, let f(n) be the number of
­positive integers less than or equal to n and relatively prime with n. Then,
f : Z1 → Z1 is a function and is called the Euler–Totient function, which is
an important arithmetical function in the theory of numbers.
Note that f(1) 5 1 5 f(2), f (3) 5 2 5 f(4), f(5) 5 4, f(6) 5 2 and
f(7) 5 6. In fact, for any prime number p, f(p) 5 p 2 1, since any positive
integer less than p is relatively prime with p.

Theorem 4.2.7. Let G be a cyclic group and a  G such that G 5 <a>. Then,
G is infinite if and only if an  am for all n  m  Z and, in this case, a and
a21 are the only generators of G.

Proof: We are given that G 5 <a> 5 {an : n  Z}. Suppose that an 5 am

for some n  m  Z. We can assume that n , m. Then, m 2 n is a positive
integer and am2n 5 am(an)21 5 e. Therefore, a is an element of finite order and,
by Theorem 4.2.4 (4),

G 5 <a> 5 {e, a, a2, …, aO(a)21}

and hence G is finite.

Conversely suppose that G is finite. Then, since an  G for each n  Z, an 5 am
for some n  m  Z. Next, suppose that G is infinite. Then, an  am for all
n  m  Z and, in particular, a  a21. Also,

G 5 <a> 5 <a21>

Q001-Algebra-111001_CH 04.indd 19 9/22/2011 11:25:08 AM

 4-20   Algebra – Abstract and Modern

and therefore a and a21 are two distinct generators of G. Now, suppose that b
is any generator of G. Then,

<b> 5 G 5 <a>

and hence a 5 bn and b 5 am for some n and m  Z. Since a 5 bn 5 (am)n 5

amn, it follows that 1 5 mn. Since m and n are integers, we get that m 5 1 5
n or m 5 21 5 n and hence b 5 a or b 5 a21. Thus, a and a21 are the only
generators of G.
Note that, for any group G and a  G, a is a generator of G if and only if
a21 is a generator of G. Also, a group G can have exactly two generators, but
still G may be finite. For consider the following.

Example 4.2.5. In the group (Z3, 13) of integers modulo 3, 1 and 2 (5 21)
are the only generators and (Z3, 13) is a finite group.

Theorem 4.2.8. Let G be a finite cyclic group of order n and a  G such

that G 5 <a>.
1. For any 1 # r , n, ar is a generator of G if and only if r is relatively
prime with n.
2. G has exactly f(n) generators.

Proof: By Corollary 4.2.1, n 5 |G| 5 |<a>| and hence O(a) 5 n so that

G 5 {e, a, a2, …, an21}.

1. Let 1 # r , n. Then, by Theorem 4.2.6,

n
a r is a generator of G ⇔ n  O(a r ) 
(n, r )
⇔ ( n, r )  1.

2. This follows from the definition of f(n) and from (1) above. Note that,
since f(1) 5 1, (2) is trivial when n 5 1.

Example 4.2.6
1. (Z, 1) is a cyclic group with 1 and 21 as the only generators.
2. (Zn, 1n) is a finite cyclic group with f(n) generators.
3. There are exactly two generators in each of the groups (Z3, 13), (Z4, 14)
and (Z6, 16), since f(3) 5 2 5 f(4) 5 f(6).

Q001-Algebra-111001_CH 04.indd 20 9/22/2011 11:25:09 AM

 Subgroups and Quotient Groups    4-21

4. Since f(8) 5 4, there are four generators for (Z8, 18) and these are 1, 3,
5 and 7.
5. For any prime number p, there are p 2 1 generators for the group (Zp, 1p),
since f(p) 5 p 2 1. That is, any nonzero (nonidentity) element Zp is a
generator.

Worked Exercise 4.2.1. Compute the order of 16 in (Z24, 124).

Answer: Z24 5 <1> and O(1) 5 24 in Z24

24 24
O(16)   3.
(16, 24) 8

Worked Exercise 4.2.2. For any positive integers a and b, prove that

aZ 1 bZ 5 (a, b)Z and aZ ∩ bZ 5 [a, b]Z

where (a, b) and [a, b] are respectively the g.c.d. and l.c.m. of a and b.

Answer: These follow from the fact that, for any n and m  Z1,

nZ ⊆ mZ ⇔ n  mZ ⇔ m divides n

and that aZ 1 bZ is the smallest subgroup of Z containing aZ and bZ. Also

aZ ∩ bZ is the largest subgroup of Z contained in aZ and bZ.

Worked Exercise 4.2.3. Determine all the generators of 36Z 1 24Z.

Answer: Since 36Z 1 24Z5 (36, 24)Z 5 12Z and 12Z is an infinite cyclic
group generated by 12, we get that 12 and 212 are the only generators of
12Z 5 36Z 1 24Z.

Worked Exercise 4.2.4. If a cyclic group has exactly two generators, then
what can we say about the order of G?

Answer: If G is an infinite cyclic group, then, by Theorem 4.2.7, G has

exactly two generators. Suppose that G is a finite cyclic group of order n and
that G has exactly two generators. Then, f(n) 5 2 and we have to determine
all n for which f(n) 5 2. Clearly f(3) 5 2 5 f(4) 5 f(6). If n > 6 and n is
odd, then 1, 2, n 2 1 are distinct and relatively prime with n. Let n > 6 and
n be even. Suppose that 2 , p1 , p2 < … , pr are all the distinct primes

Q001-Algebra-111001_CH 04.indd 21 9/22/2011 11:25:09 AM

 4-22   Algebra – Abstract and Modern

dividing n and m 5 p1 p2 … pr 2 2. Then, (m, n) 5 1 and 1, m , n 2 1 and

hence f(n) > 2. Therefore, f(n) 5 2 if and only if n 5 3 or 4 or 6.

EXERCISE 4(b)

1. State whether each of the following are true and substantiate your answer.
(i) Every finite abelian group is cyclic.
(ii) An infinite group is abelian if and only if it is cyclic.
(iii) (Q, 1) is a cyclic group, where Q is the set of rational numbers.
(iv) (R, 1) is a cyclic group.
(v) (C, 1) is a cyclic group.
(vi) If G1 and G2 are cyclic groups, then G1 3 G2 is also a cyclic group.
(vii) (C 2 {0}, ?) is an abelian group which is not cyclic.
(viii) (Q 2 {0}, ?) is a cyclic group.
(ix) Any group of order 5 or 7 is cyclic.
(x) The group S(X) of bijections of a set X onto itself is a cyclic group under
the composition of mappings.
2. Which of the following are cyclic groups? Substantiate your answers.
(i) (R 2 {0}, ?)
(ii) (P(X), 1), where X is a set.
(iii) The group of quaternions which is of order 8.
(iv) (Q1, ?)
(v) (R1 , ?)
(vi) For any positive Integer n, the group of all nth roots of unity of under the
usual multiplication of complex numbers.
3. What can we say about a cyclic group having exactly one generator?
4. List all the elements in each of the following subgroups of the groups mentioned.
(i) <7> in (Z18, 118)
(ii) <5> in (Z20, 120)
(iii) <3> in (Z12, 112)
(iv) <3> in (Z16, 116)
(v) <i> in (C 2 {0}, ?)
(vi) < 2> in (R1 , ?)
(vii) < 2> in (R, 1)
(viii) < 2 > in (C 2 {0}, ?)
(ix) <e> in any group G, where e is the identity in G.
(x) <12> in (3Z, 1).

Q001-Algebra-111001_CH 04.indd 22 9/22/2011 11:25:10 AM

 Subgroups and Quotient Groups    4-23

5. Prove that O(a) is finite for any element a of a finite group and give an example
of an infinite group in which every element is of finite order.
6. If G is a group in which O(a) is finite for all a  G, then can G be finite?
7. Let K and H be finite cyclic subgroups of orders m and n respectively in an abe-
lian group G. If m and n relatively prime, prove that G has a cyclic subgroup of
order mn.
8. In Exercise 7 above, if the least common multiple of m and n is K, then prove that
G has a cyclic subgroup of order K.
9. If A and B are subgroups of a group G and one of A and B is cyclic, then prove
that A ∩ B is a cyclic subgroup of G.
10. If a and b are elements of a group G, such that O(a) and O(b) are relatively prime
positive integers, then prove that <a> ∩ <b> 5 {e}.
11. Let G be a finite cyclic group of order n and d be a positive divisor of n. Then
prove that the equation xd 5 e has exactly d solutions in G.
12. Prove that the set {4, 8, 12, 16} is a group under the multiplication modulo 20. What
is the identity element? Is this a cyclic group? If so, what are its generators?
13. Let G 5 {7, 35, 49, 77}. Then prove that (G, ?84) is a group, where ?84 is the
multiplication modulo 84. What is the identity in G? Is this a cyclic group? If so,
what are its generators?
14. Is Z 3 Z a cyclic group, where the operation is a coordinate-wise addition?
15. Is Z9 3 Z16 cyclic? If so, what are the generators?
16. Let G and H be finite cyclic groups of orders m and n, respectively. Then prove
that G 3 H is cyclic if and only if m and n are relatively prime.
17. If G is a finite cyclic group and H is an infinite cyclic group, then can G 3 H be
cyclic?
18. Let G be a group and a  G such that O(a) 5 24. Then find a generator of the
group <a9> ∩ <a10>. In general, find a formula for the generator of <an> ∩ <am>
for any 1 # n, m , 24.
19. Let G be a finite
n
cyclic group of order n. Then, for each positive divisor nr of n,
prove that < a r > is the only subgroup of order r and that the map r → <a r > is a
bijection of the set of positive divisors of n onto the set of all subgroups of G.
20. Find all the subgroups of (Z24, 124) and (Z30, 130).
21. Given a positive integer m, give an example of a cyclic group having exactly m
subgroups.
22. List all the subgroups of (Z625, 1625).

Q001-Algebra-111001_CH 04.indd 23 9/22/2011 11:25:11 AM

 4-24   Algebra – Abstract and Modern

23. Let Un be the group of all nth-roots of unity under the usual multiplication of
complex numbers. Any generator of Un is called a primitive nth-root of unity.
Determine all the primitive nth-roots of unity for each of n 5 5, 7 and 11.

24. Prove that any group having only a finite number of subgroups must be finite.

25. Give an example of a nonabelian group such that every proper subgroup is cyclic.

26. Let G be an abelian group and n be a positive integer. Then prove that the set {a
 G : O(a) divides n} is a subgroup of G.

27. Let a be an element of order n in a group G. Then prove that O(ar) 5 O(an2r) for
all 1 # r , n.

28. Let G be a cyclic group of order 24 and a  G such that a8  e and a12  e. Then
prove that a is a generator of G.

29. For any elements a and b of a group, prove that O(ab) 5 O(ba).

30. Let G be an abelian group. Prove that the set of elements of finite order in G is a
subgroup of G.

4.3 COSETS OF A SUBGROUP

For a given subgroup of a group G, we consider two important equivalence
relations on G and study the relationship between their equivalence classes
and the given subgroup. Let us begin with the following theorem.

Theorem 4.3.1. Let H be a subgroup of a group G and define two binary rela-
tions LH and RH on G as follows:

(a, b)  LH if a2lb  H

and (a, b)  RH if ab2l  H

for any a and b  G. Then, LH and RH are equivalence relations on G.
Proof: Let us recall that a reflexive, symmetric and transitive relation on G is
called an equivalence relation. For any a  G,
a2la 5 e  H (since H is a subgroup)
and hence (a, a)  LH. Therefore, LH is a reflexive relation on G. Also, for
any a and b  G,

(a, b)  LH ⇒ a2lb  H
⇒ (a2lb)21  H (since H is a subgroup of G)
⇒ b2la  H
⇒ (b, a)  LH

Q001-Algebra-111001_CH 04.indd 24 9/22/2011 11:25:11 AM

 Subgroups and Quotient Groups    4-25

and hence LH is symmetric. Further,

(a, b) and (b, c)  LH ⇒ a2lb and b2lc  H

⇒ (a2lb)(b2lc)  H
⇒ a2lc  H
⇒ (a, c)  LH

and hence LH is transitive. Thus, LH is an equivalence relation on G. Similarly,

RH is an equivalence relation.

Definition 4.3.1. Let H be a subgroup of a group G and a  G. Define

aH 5 {ax : x  H}
and Ha 5 {xa : x  H}.

aH is called a left coset of H corresponding to a in G and Ha is called a right

coset of H corresponding to a in G.
Recall that, for any equivalence relation u on a set X, the equivalence class
of u containing a given element x  X is given by

u(x) 5 {y  X : (x, y)  u}

and that the equivalence class of u form a partition of X in the sense that any
two distinct equivalence classes of u are disjoint and the union of all equiva-
lence classes of u is equal to the whole set X. In the following, we prove that
the equivalence classes of LH (respectively RH) are precisely the left (right)
cosets of H. Note that, if G is an abelian group then aH 5 Ha for all a  G.

Theorem 4.3.2. Let H be a subgroup of a group G and a  G. Then,

LH(a) 5 aH and RH(a) 5 Ha

That is, {b  G : a2lb  H} 5 aH and {b  G : ab2l  H} 5 Ha.

Proof: For any x  G,

x  LH(a) ⇔ (a, x)  LH
⇔ a2lx  H
⇔ x 5 a(a2lx)  aH

and hence LH(a) 5 aH. Similarly,

Q001-Algebra-111001_CH 04.indd 25 9/22/2011 11:25:11 AM

 4-26   Algebra – Abstract and Modern

x  RH(a) ⇔ (a, x)  RH
⇔ ax2l  H
⇔ xa2l 5 (ax2l) 2l  H
⇔ x 5 (xa2l)a  Ha

and hence RH(a) 5 Ha.

Corollary 4.3.1. Let H be a subgroup a group G and a and b  G. Then,

aH 5 bH ⇔ a2lb  H
and Ha 5 Hb ⇔ ab  H. 2l

In particular, aH 5 H ⇔ a  H ⇔ Ha 5 H.

Corollary 4.3.2. For any subgroup H of a group G, any two left (right) cosets
of H in G are either equal or disjoint.

Corollary 4.3.3. For any subgroup H of a group G, the left (right) cosets of H
in G form a partition of G, then note that, if we use 1 to denote the operation
on the group G, then write a 1 H for aH and H 1 a for Ha.

Example 4.3.1
1. If H 5 {e}, then aH 5 {a} 5 Ha for any a  G and any subgroup H of G.
2. Consider the group (Z, 1) of integers under the usual addition and let
H be a subgroup of (Z, 1). Then, by Worked Exercise 4.1.1, H 5 nZ
for some nonnegative integer n. If H 5 {0}, then for any a  Z, the left
coset a 1 H 5 {a} 5 H 1 a. Suppose that H  {0}. Then, n > 0. For
any a  Z, choose q and r  Z such that
a 5 qn 1 r and 0#r,n

and hence r 2 a 5 (2q)n  <n> 5 H and

therefore a 1 H 5 r 1 H.
Thus,
0 1 H (5 H), 1 1 H, 2 1 H, …, (n 2 1) 1 H

are all the left (right) cosets of H in Z. One can observe that these are all
distinct. Thus, there are exactly n cosets of nZ 5 H in Z.
3. Let X 5 {1, 2, 3} be a 3-element set and S(X) be the group of all bijec-
tions of X onto itself under the composition of mappings. We know from

Q001-Algebra-111001_CH 04.indd 26 9/22/2011 11:25:11 AM

 Subgroups and Quotient Groups    4-27

Example 3.4.2 that S(X) 5 {e, a, b, c, d, s} and the group (S(X), o) is

represented by the table given below.

o e a b c d s
e e a b c d s
a a b e s c d
b b e a d s c
c c d s e a b
d d s c b e a
s s c d a b e

Let H 5 {e, s}, then H is a subgroup of (S(X), o). The left and right
cosets of H in S(X) are given below.
eH 5 {ee, es} 5 {e, s} 5 H
aH 5 {ae, as} 5 {a, d} 5 dH
bH 5 {be, bs} 5 {b, c} 5 cH.
Therefore, there are exactly three distinct left cosets of H in S(X) and
each of these contain exactly two elements. Also
He 5 {ee, se} 5 {e, s} 5 H
Ha 5 {ea, sa} 5 {a, c} 5 Hc
Hb 5 {eb, sb} 5 {b, d} 5 Hd.
Therefore, again there are exactly three right cosets of H in S(X) and
each of these contain exactly two elements. Note that, even though the
number of left cosets of H is equal to the number of right cosets, a left
coset may not be a right coset and vice versa. For example, aH is not
equal to any right coset.
4. Consider the group (Z24, 124) of integers modulo 24. Let us compute all
the subgroups of Z24 and their cosets. This being an abelian group, any
left coset of a subgroup is a right coset. We know that the subgroups of
Z24 correspond to the positive divisors of 24 which are precisely 1, 2, 3,
4, 6, 8, 12 and 24.
For any divisor d of 24, let
24
Hd 5 {0, d, 2d, …, (q21)d}, where q  .
d
Then    H1 5 {0, 1, 2, …, 2421}5 Z24
H2 5 {0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22}
H3 5 {0, 3, 6, 9, 12, 15, 18, 21}

Q001-Algebra-111001_CH 04.indd 27 9/22/2011 11:25:11 AM

 4-28   Algebra – Abstract and Modern

H4 5 {0, 4, 8, 12, 16, 20}

H6 5 {0, 6, 12, 18}
H8 5 {0, 8, 16}
H12 5 {0, 12}
H24 5 {0}

The above eight are all the subgroup of Z24. Coming to the cosets, it
is clear that Z24 is the only coset of H1. There are two cosets of H2
namely
0 24 H 2 ( H 2 ) and 124 H 2 ;

For any a  Z24, if a is even, then a  H2 and hence a 24 H 2  0 24 H 2

and, if a is odd, then a 21 is even and hence a 2 1 H2 so that
a 24 H 2  1 24 H 2 .
In general, for any divisor d of 24, there are exactly d cosets of Hd and
these are

0 24 H d  H d , 124 H d , 2 24 H d , …, ( d 1) 24 H d .

Worked Exercise 4.3.1. Let H be a subgroup of a group G such that there are
exactly two left cosets of H in G. Then prove that every left coset of H in G is
a right coset and vice versa.

Answer: Since H(5 eH) is a left coset, it is given that there is only one more
left coset of H in G and let this be aH. Then, aH  H and a  H. Now,

aH ∩ H 5  and aH ∪ H 5 G

and hence aH and H are complements to each other in G. That is, G 2 H 5

aH and xH 5 aH for all x  H. If x  H, then Hx 5 H 5 He. On the other
hand, for any x  H,

Hx 5 H2lx 5 (x2lH)2l 5 (aH)2l 5 H2la2l 5 Ha21 5 Ha.

Thus, Ha and H are the only right cosets of H in G and hence Ha 5

G 2H 5 aH.

EXERCISE 4(c)

1. Determine all the subgroups of each of the following and list their cosets.
(i) (Z50, 150)

Q001-Algebra-111001_CH 04.indd 28 9/22/2011 11:25:12 AM

 Subgroups and Quotient Groups    4-29

(ii) (Z20, 120)

(iii) (Z29, 129)
(iv) Z2 3 Z2

2. Consider the subgroup Z of the group (R, 1). Prove that there is a bijection
between the set of left cosets of Z in R and the interval [0, 1).

3. Let H be a subgroup of a group G such that aha21  H for all h  H and a  G.

Then prove that every left coset aH is equal to the right coset Ha.

4. For any subgroup H of a group G, prove that aH → Ha21 is a bijection of the set
of left cosets of H in G onto the set of right cosets of H in G.

5. Let H be a subgroup of a group G, a  G and K 5 aHa21. Then prove that K is a

subgroup of G and that there is a bijection of the set of left cosets of H in G onto
the set of left cosets of K in G.

6. Let X 5 {1, 2, 3} and (S(X), o) be the group of bijections of X onto itself. Let f
 S(X) be such that f(1) 5 2, f(2) 5 3 and f(3) 5 1. Then prove that H 5 {e, f,
f 21} is a subgroup of S(X). Compute all the left and right cosets of H in S(X).

7. Let A 5 <2p> be the cyclic subgroup generated by 2p in the group (R, 1).
Prove that the trigonometric functions sine and cosine are constant on any left
coset of A in (R, 1).

8. Let (S(X), o) be the group of all bijections of a nonempty set X onto itself and
x  y  X. Let
Ax 5 {f  S(X) : f(x) 5 x}
and         Ax,y 5 {f  S(X) : f(x) 5 y}.

Prove that Ax is a subgroup of S(X) and that Ax,y is not a subgroup of S(X).

9. In Exercise 8 above, prove that Ax,y is a coset of Ax in S(X). Is it a left coset or a

right coset?

10. Let A be a subgroup of a group G and x, y and z  G such that xyA 5 xzA. Then,
prove that yA 5 zA. Also, prove that Ayx 5 Azx implies Ay 5 Az.

11. Give an example of a subgroup A of a group G such that the product of two left
cosets of A in G is not a left coset of A in G.

12. For any subgroup A of a group G, prove that the only left (right) coset of A in G,
which is also a subgroup of G, is A itself.

13. Let A and B be two subgroups of a group G and x and y  G such that Ax 5 By.
Then prove that A 5 B.

14. Prove that a subset S of a group G cannot be a left coset of two distinct sub-
groups of G.

Q001-Algebra-111001_CH 04.indd 29 9/22/2011 11:25:13 AM

 4-30   Algebra – Abstract and Modern

4.4 LAGRANGE’S THEOREM

We have proved in the previous section that the left cosets of a subgroup
H in a group G form a partition of the group G. This provides a counting
technique that the total number of elements of a finite group G is equal
to the sum of those in the cosets of a subgroup. In fact, we prove that the
number of elements in any finite subgroup H is equal to that in any of its
left or right coset and deduce the famous theorem of Lagrange which states
that the order of any subgroup of a finite group is a divisor of the order of
the group. We prove this theorem of Lagrange and obtain certain important
consequences.
First, let us recall that the number of elements in a finite group G is called
the order of G and is denoted by |G| and that, if a group G is infinite, we say
that the order of G is infinite. In general, for any finite set X, |X| denote the
number of elements in X. Note that |X| 5 0 if and only if X is the empty set.

Theorem 4.4.1. Let H be a subgroup of group G and a  G. Then, H, aH and

Ha are all bijective to each other.

Proof: Define f : H → aH by f(x) 5 ax for all x  H. Clearly f is a surjection.

By the left cancellation law, f is an injection also. Therefore, f is a bijection
of H onto aH. Similarly the map x  xa is a bijection of H onto Ha. Thus,
Ha . H . aH.

Theorem 4.4.2 (Lagrange’s Theorem). Let G be a finite group and H be a

subgroup of G. Then, the order of H is a divisor of the order of G. That is, |H|
divides |G|.

Proof: Since G is a finite set and H is a nonempty subset of G, |G| and |H| are
positive integers. Again, since G is finite, the number of left (right) cosets of
H in G is also finite. Let a1H, a2H, …, anH be all the distinct left cosets of H
in G. By the above theorem,

G a1H a2H a3H anH

Q001-Algebra-111001_CH 04.indd 30 9/22/2011 11:25:14 AM

 Subgroups and Quotient Groups    4-31

|aiH| 5 |H| for all 1 # i # n. Also, we have

aiH ∩ ajH 5  for all i  j

and a1H ∪ a2H ∪…∪ anH 5 G.

Therefore, |G| 5 |a1H| 1 |a2H| 1…1 |anH|

5 |H| 1 |H| 1 … 1 |H| (n times)
5 n|H|.

Thus, |G| 5 n|H| and |H| is a divisor of |G|. b

Corollary 4.4.1. For any subgroup H of a finite group G, the number of left
cosets of H in G is equal to the number of right cosets of H in G which is
|G|
same as .
|H |
Proof: In the proof of the above theorem, we have proved that |G| 5 n|H| and
hence

G
5 n 5 The number of left cosets of H in G.
H
In the above proof, we can replace left cosets with right cosets and prove
similarly that |G| 5 m|H|, where m is the number of right cosets of H in
G. Now,

G
n  m. b
H


Definition 4.4.1. For any subgroup H of a finite group G, the number of left
(right) cosets of H in G is called the index of H in G and is denoted by iG(H).

Corollary 4.4.2. For any subgroup H of a finite group G,

G
iG ( H )  .
H

Note that, even when G is an infinite group, a subgroup can have only
finitely many cosets and one can talk about the index of such a subgroup.
Consider the following example.

Q001-Algebra-111001_CH 04.indd 31 9/22/2011 11:25:14 AM

 4-32   Algebra – Abstract and Modern

Example 4.4.2.
1. Let n be a positive integer and H 5 nZ 5 {na : a  Z}. Then, H is a
subgroup of (Z, 1) and, by Example 4.3.1 (2),
H, 1 1 H, 2 1 H, …, (n 2 1)H

are all the cosets of H in Z. Therefore, the index of H in Z is n.

2. Consider the group (Z12, 112) of integers modulo 12. Let H 5 <3> 5 {0,
3, 6, 9}. Then, |H| 5 4 and |Z12| 5 12 and hence 124 (5 3) is the index of
H in Z12. We have 0 112 H 5 3 112 H 5 6 112 H 5 9 112 H 5 H,
1 112 H 5 4 112 H 5 7 112 H 5 10 112 H 5 {1, 4, 7, 10},
2 112 H 5 5 112 H 5 8 112 H 5 11 112 H 5 {2, 5, 8, 11}.

Therefore, H , 1 112 H and 2 112 H are all the distinct left cosets of H
in Z12.
3. Consider the example given in Example 4.3.1 (3) in which H 5 {e, s}
and G 5 the group S(X) of bijection of a 3-element set X onto itself.
|G| 6
Here, |G| 5 3! 5 6, |H| 5 2 and hence iG(H) 5  5 3 and therefore
|H | 2
there are exactly 3 distinct left(right) cosets of H in G.
For any element a in group, recall that the order of a is defined as the least
positive integer n such that an 5 e (if all there is one such) and is denoted
by O(a).

Theorem 4.4.3. Let a be an element in a finite group G. Then, O(a) divides |G|.

Proof: First note that, since G is finite and an  G for all integers n, an 5 am
for some n  m and hence ar 5 e for some positive integer r. Therefore, a
is an element of finite order. Let O(a) 5 n. Then, by Corollary 4.2.1, O(a) is
equal to the order of the subgroup <a> generated by a in G. By the Lagrange’s
Theorem 4.4.2, |<a>| divides |G|. Thus, O(a) divides |G|.

Corollary 4.4.3. In a finite group G, a|G| 5 e for all a  G.

Proof: Let G be a finite group, a  G and O(a) 5 n. Then, by the above

theorem, nm 5 |G| for some m  Z1 and hence

a|G| 5 anm 5 (an)m 5 em 5 e.

Theorem 4.4.4. Let G be a group of order a prime number. Then, G is a

cyclic group and every nonidentity element in G is a generator of G.

Q001-Algebra-111001_CH 04.indd 32 9/22/2011 11:25:14 AM

 Subgroups and Quotient Groups    4-33

Proof: Let |G| 5 p be a prime number. For any a  e in G, O(a) > 1 and O(a)
is a divisor of p. Since p is prime, O(a) 5 p and therefore

<a> ⊆ G and |<a>| 5 O(a) 5 p 5 |G|.

Since G is a finite set, <a> 5 G. Thus, G is a cyclic group and every a  e

in G is a generator of G.
In the next three theorems, we derive two important results in the theory
of numbers, using the Lagrange’s theorem. First, let us have the following
definition.

Definition 4.4.2. Let n be any positive integer. For any integers a and b, a is
said to be congruent to b modulo n if n divides a 2 b and we denote this by

a  b (mod n).

That is, a  b (mod n) if and only if n divides a 2 b or a 2 b  nZ.

Actually, the above is precisely the equivalence relation RH on Z defined
in Theorem 4.3.1, where H 5 nZ, the cyclic subgroup generated by n in the
group (Z, 1). Besides being an equivalence relation, it has some other prop-
erties. Some of these are listed in the following theorem.

Theorem 4.4.5. Let n be a positive integer. The following holds for any inte-
gers a, b and c.
1. a  a (mod n)
2. a  b (mod n) ⇒ b  a (mod n)
3. a  b (mod n) and b  c (mod n) ⇒ a  c (mod n)
4. a  b (mod n) ⇒ a 1 c  b 1 c (mod n)
5. a  b (mod n) ⇒ ac  bc (mod n)
6. a  0 (mod n) ⇔ a is an integral multiple of n.
7. For each a  Z, there exists an integer r such that 0 # r , n and a  r
(mod n).
8. For any 0 # r  s , n, r  s (mod n).

Proof: (1) to (6) are direct implications of the above definition. For any a  Z,
we can use the division algorithm to get integers q and r such that

a 5 qn 1 r and 0#r,n

and now, a 2 r 5 qn and hence a  r (mod n). This proves (7). To prove 8,
consider 0 # r, s , n. Then, |r 2 s| , n and hence r 2 s cannot be a multiple
of n, unless r 5 s.

Q001-Algebra-111001_CH 04.indd 33 9/22/2011 11:25:15 AM

 4-34   Algebra – Abstract and Modern

Recall that, for any n  Z1, we have defined f(n) to be the number of
positive integers which are less than or equal to n and relatively prime with n
and that the function f : Z1 → Z1 is called the Euler–Totient function.

Theorem 4.4.6 (Euler’s Theorem). Let n be a positive integer. Then,

af(n)  1 (mod n)

for all integers a which are relatively prime with n, where f is the Euler–
Totient function.

Proof: The theorem is trivial if n 5 1. Therefore, we can assume that n > 1.

Let S be the set {1, 2, …, n 2 1}. Then, S is a monoid with respect to the
multiplication ?n modulo n in which 1 is the identity. We know that an element
r in S is invertible if and only if r is relatively prime with n. Let G be the set
of all invertible elements in (S, ?n); that is,

G 5 {r  Z1 | r , n and (r, n) 5 1}.

Then, (G, ?n) is a group where ?n is the multiplication modulo n. Also, we

know that |G| 5 f(n) (see Definition 4.2.4). Now, let a be any integer which
is relatively prime with n. Then, by the division algorithm, we can write

a 5 qn 1 r, where q and r  Z and 0 # r , n.

Since (a, n) 5 1, we get that (r, n) 5 1 and r > 0.

Therefore, r  G. Now we have

af(n) 5 (qn 1 r)f(n)

5 sn 1 rf(n) for some s  Z
5 sn 1 r|G|
 1 (mod n) (by Corollary 4.4.3).

Theorem 4.4.7 (Fermat’s Theorem). Let p be a prime number and a be any

integer. Then,

ap  a (mod p).

Proof: We have to prove that p divides a p  a  a( a p1 1).

If p divides a, then clearly p divides a( a p1 1). Suppose that p does not
divide a. Then, since p is prime and (a, p) is a divisor of p, it follows that
(a, p) 5 1. Also, f(p) 5 p 2 1. Now, by the Euler’s theorem proved above,
we get that

Q001-Algebra-111001_CH 04.indd 34 9/22/2011 11:25:15 AM

 Subgroups and Quotient Groups    4-35

ap21 5 af(p)  1 (mod p)

and hence p divides ( a p1 1), so that p divides a( a p1 1). 

In the following, we introduce certain counting techniques which play cru-
cial roles in the proofs of various results in the theory of finite groups. If A
and B are subgroups of a group G, recall that the set AB may not be subgroup
and that AB is a subgroup if and only if AB 5 BA. However, it is quite natural
to think of the number of elements in AB in terms of those in A and B, when A
and B are finite subgroups. The following is a precise answer to this.

Theorem 4.4.8. Let A and B be finite subgroups of a group G and

AB 5 {ab : a  A and b  B}. Then,

| A || B |
| AB |   | BA | .
A∩ B

Proof: First note that it is quite possible that ab 5 a1b1 for distinct elements
a and a1 in A and distinct elements b and b1 in B. We shall find out how often
does an element ab appear as a product of an element in A and an element in
B. For any x  A ∩ B, we have

ab 5 (ax) (x21b)

and ax  A (since a and x  A) and x21b  B (since x and b  B). Also, if ab

5 a1b1, where a and a1  A, and b and b1  B, then

a21a1 5 bb121 5 s, say and s  A ∩ B.

Also, a1 5 as and b1 5 s21b.

This is to say that any representation of x 5 ab as a product of an element of
A and an element of B must be of the form

(as) (s21b), where s  A ∩ B.

Thus, for each a  A and b  B, the product ab appears |A ∩ B| times as a

product of an element of A and an element of B. This implies that the number
of distinct elements in AB is the total number in the listing of AB as a product
of an element in A and an element in B (that is, |A||B|) divided by the number
of times a given element (that is, |A ∩ B|). Thus,

| A || B | | B || A |
| AB |    | BA |,
| A∩ B | | B ∩ A|

since A ∩ B 5 B ∩ A.

Q001-Algebra-111001_CH 04.indd 35 9/22/2011 11:25:15 AM

 4-36   Algebra – Abstract and Modern

Corollary 4.4.4. Let A and B be subgroups of a finite group G such that

A ∩ B 5 {e}. Then, |A| $ | G | or |B| $ | G | .

Proof: Since AB ⊆ G, we have

A B
|G| $ |AB| 5 A B (by the above theorem)
A∩ B
and hence | G |  |A| or | G |  |B|.
For any subgroup A of a finite group G, we have |A|. iG(A) 5 |G| and hence
both the order and index of A in G are divisors of the order of |G|. In the
­following, we derive some more properties of the index of a subgroup of a
finite group. First note that if A and B are subgroups of a group G and A ⊆ B,
then A is a subgroup of B also and we can talk of the index of A in B also.

Worked Exercise 4.4.1. Prove that the following holds for any subgroups
A and B of a finite group G.
1. If A ⊆ B, then iG(A) 5 iB(A) ? iG(B)
2. iA(A ∩ B) # iG(B)
3. iA(A ∩ B) 5 iG(B) if and only if G 5 AB.
4. iG(A ∩ B) 5 iG(A)  iG(B) if and only if G 5 AB.

Answer:
1. Suppose that A ⊆ B. Then, by Corollary 4.4.2, we have
G G B
iG(A) 5  .  iG ( B)  iB ( A)
A B A

A AB G
2. iA(A∩B) 5    iG ( B) .
A∩ B B B
A B
3. Consider |AB| 5  iA ( A ∩ B) B . Now
A∩ B
iA(A∩B) 5 iG(B) ⇒|AB| 5 iG(B)|B| 5 |G| ⇒ G 5 AB

and, conversely, if G 5 AB, then

G AB A
iG(B) 5    iA ( A ∩ B )
B B A∩ B
4. If G 5 AB, then

Q001-Algebra-111001_CH 04.indd 36 9/22/2011 11:25:17 AM

 Subgroups and Quotient Groups    4-37

G G G G
iG(A∩B) 5    iG ( A)  iG ( B)
A∩ B A B AB A B
and, conversely, if iG(A∩B) 5 iG(A)∙iG(B), then
G G G A B
 and hence G   AB
A∩ B A B A∩ B

which implies that G 5 AB.

Worked Exercise 4.4.2. Let G be a group of order prn, where p is a prime not
dividing n. Let A and B be subgroups of G of orders pr and ps, respectively. If
B  A, prove that AB is not a subgroup of G.

Answer: We are given that |A| 5 pr and |B| 5 ps. Since |B| divides |G| and pr
is the largest power of p dividing |G|, we get that 0 # s # r. Since |A ∩ B| is
a divisor of |A|, |A ∩ B| 5 pt for some t $ 0. Suppose, AB is a subgroup of G,
| A||B|
then |AB| is a divisor of |G| 5 prn. But |AB| 5 , since |A|, |B| and |A ∩ B|
A∩B
are all powers of p, |AB| 5 p for some a $ 0 and a # r (since pr is the largest
a

power of p dividing |G|). Therefore,

|AB| 5 pa # pr 5 |A| # |AB| (since A ⊆ AB)

and hence |AB| 5 pr, so that A 5 AB ⊇ B, which is a contradiction. Thus, AB

is not a subgroup of G.

Worked Exercise 4.4.3. Let A and B be finite subgroups of a group such that
|A| 5 pr and |B| 5 qs, where p and q are distinct primes and r and s are positive
integers. Then prove that A ∩ B 5 {e}.

Proof: Since A ∩ B is a subgroup of A as well as B, |A ∩ B| is a common divisor

of |A| and |B| and hence |A ∩ B| is a divisor of (pr, qs), which is equal to 1 since
p and q are distinct primes. Thus, |A ∩ B| 5 1 and hence |A ∩ B| 5 1.

EXERCISE 4(d)

1. State whether the following are true or false and justify your answer.
(i) There is no subgroup of order 9 in (Z24, 14).
(ii) There is a subgroup of order 36 in (Z120, 1120).
(iii) In any group of order 240, there is a subgroup of index 36.
(iv) If X is a 5-element set, then the group (S(X), o) of bijections of X onto
itself has a subgroup of order 24.

Q001-Algebra-111001_CH 04.indd 37 9/22/2011 11:25:17 AM

 4-38   Algebra – Abstract and Modern

(v) For any positive integer n, there is an element of order n in (R, 1).
(vi) If X is a 6-element set, then the group S(X) has an element of order 27.
(vii) The order of any element in a finite group is finite.
(viii) 43 divides 24221.
(ix) 19 divides 91821.
(x) 8 divides 729421.

2. If A and B are subgroups of a group G such that |A| is a prime and A ∩ B  {e},
then prove that A ⊆ B.

3. If A is a subgroup of index 2 in a group G, then prove that x2  A for all x  G.

4. Prove that the following are equivalent to each other for any subgroup A of a
group G.
(i) iG(A) 5 2.
(ii) x21y  A for all x and y  G 2 A.
(iii) xy21  A for all x and y  G 2 A.

5. If G is a group having no nontrivial subgroups, then prove that G is a cyclic

group of order prime or G is trivial.

6. If A and B are subgroups of finite index in a group G, prove that A ∩ B is also of

finite index and that
iG(A ∩ B) # iG(A)  iG(B).

7. If A and B are subgroups of a group G and a and b  G, then prove that Aa ∩ Bb 5 

or Aa ∩ Bb 5 (A ∩ B)c for some c  G.

8. Let A and B be subgroups of finite index in a group G such that AB 5 BA. Then
prove that
iAB(A ∩ B) 5 iAB(A)  iAB(B).

9. If an abelian group has two subgroups of orders n and m, then prove that it has a
subgroup whose order is the least common multiple of n and m.

10. Let a and b be elements of a group such that a5 5 e and ab21a 5 b2, then find O(b).

11. Determine all the subgroups of a group of order 137.

12. Let G be an abelian group of order 2n, where n is an odd positive integer. Prove
that G contains exactly one element of order 2.

13. Prove that any group of order 4 is abelian and give an example of a noncyclic
group of order 4.

14. Prove that any nonabelian group has atleast six elements and give an example of
a 6-element nonabelian group.

Q001-Algebra-111001_CH 04.indd 38 9/22/2011 11:25:17 AM

 Subgroups and Quotient Groups    4-39

4.5 NORMAL SUBGROUPS

For any subgroup A of a group G, the collection of left cosets of A in G is in
general not the same as the collection of right cosets of A in G. Subgroups for
which these two collections are same are of special importance. In this sec-
tion, we discuss certain important properties of these special subgroups.

Definition 4.5.1. A subgroup N of a group G is said to be normal in G if

every left coset of N in G is a right coset of N in G.
Note that if a left coset aN happens to be a right coset, then it must be the
coset Na only; for if aN 5 Nb, then a  aN 5 Nb and hence ab21  N so that
Na 5 Nb. In the following, we obtain several other equivalent conditions for
a subgroup to be normal.

Theorem 4.5.1. The following are equivalent for any subgroup N of a group G.
1. N is a normal subgroup of G.
2. aN 5 Na for all a  G.
3. aNa21 ⊆ N for all a  G.
4. Every right coset of N in G is a left coset of N in G.
5. The product of any two left cosets of N in G is a left coset of N in G.
6. (aN) (bN) 5 abN for all a and b in G.
7. (Na) (Nb) 5 Nab for all a and b in G.
8. LN 5 RN (see Theorem 4.3.1); that is, for any a and b  G, a21b  N if
and only if ab21  N.

Proof:
(1) ⇒ (2): Suppose that N is a normal subgroup of G. Then, for any a  G,
aN 5 Nb for some b  G, aN 5 Nb for some b  G which implies that
a  (Nb) ∩ (Na) and hence Nb 5 Na. Therefore, aN 5 Na.
(2) ⇒ (3): If aN 5 Na, then aNa21 5 N.
(3) ⇒ (4): Suppose that aNa21⊆ N for all a  G. Then,

aN 5 (aNa21)a ⊆ Na

and, since a21N(a21)21 ⊆ N, we get that a21N ⊆ Na21 and hence Na 5 a(a21N)
a ⊆ a(Na21)a 5 aN. Thus, Na 5 aN for all a  G.
(4) ⇒ (5): Suppose that every right coset of N in G is a left coset of N in G.
For any a and b  G, first choose c  G such that Nb 5 cN and now

(aN)(bN) 5 a(Nb)N 5 a(cN)N 5 acNN 5 acN.

Q001-Algebra-111001_CH 04.indd 39 9/22/2011 11:25:18 AM

 4-40   Algebra – Abstract and Modern

Thus, (aN) (bN) is a left coset of N in G.

(5) ⇒ (6): Assume (5). For any a and b  G, there exists x  G such that

(aN)(bN) 5 xN.

Now, ab 5 ae∙be  (aN)(bN) 5 xN and hence (abN) ∩ (xN)   so that

(aN)(bN) 5 xN 5 abN.
(6) ⇒ (7): Let a and b  G. Then, a21 and b21  G and consider

(Na) (Nb) 5 (b21N21 a21N21)21

5 ((b21N) (a21N))21
5 (b21 a21 N)21 (by (6))
5 N21ab
5 Nab.

(7) ⇒ (8): For any a and b  G, consider

a21b  N ⇒ b 5 a(a21b)  aN
⇒ e 5 ebb21 N (aN) b21 5 (Na)(Nb21) 5 Nab21 (by (7))
⇒ e 5 xab21, for some x  N
⇒ ab21 5 x21  N.

Similarly ab21  N ⇒ a21b  N. Thus, LN 5 RN.

(8) ⇒ (1): For any a  G, consider

x  Na ⇔ xa21  N
⇔ x21a  N (by (8))
⇔ a21x 5 (x21a)21 N
⇔ x  aN.

Thus, aN 5 Na for all a  G and hence N is a normal subgroup of G. b

Example 4.5.1
1. In any group G, the trivial subgroup {e} and G are always normal in G.
2. Consider the group (S(X), o) of all bijections of a 3-element set X onto
itself (see Example 3.4.2). Following the notation given in Example
3.4.2, let H 5 {e, s}. Then, H is a subgroup of S(X). Also,
bsb21 5 bsa 5 ca 5 d  H.
Therefore, H is not a normal subgroup of S(X).

Q001-Algebra-111001_CH 04.indd 40 9/22/2011 11:25:18 AM

 Subgroups and Quotient Groups    4-41

Clearly every subgroup of an abelian group is normal. However, the con-

verse of this may not be true in general. That is, a group in which every sub-
group is normal may not be abelian; for, consider the following example.

Example 4.5.2. Consider the set G 5 {1, 21, i, 2i, j, 2j, k, 2k}. Define the
binary operation on G as just multiplication of numbers, treating i, j and k as
numbers satisfying the following rules.

i2 5 j2 5 k2 5 21
ij 5 k and ji 5 2k
jk 5 i and kj 5 2i
ki 5 j and ik 5 2j
i

k j

Then, G is a group under this operation. This group is called the Q

­ uaternion
group of 8 elements. Since ij 5 k  2k 5 ji, this group G is not an abe-
lian group. However, every subgroup of G is normal, as proved in Worked
­Exercise 4.5.2.

Worked Exercise 4.5.1. Prove that any subgroup of index 2 in any group is
normal.

Answer: Let A be a subgroup of a group G and iG(A) 5 2. Then, there are

exactly two left cosets of A in G. By Worked Exercise 4.3.1, A is normal
subgroup of G.

Worked Exercise 4.5.2. Prove that every subgroup of the Quaternion group
of 8 elements is normal.

Answer: Let G be the 8-element Quaternion group and A be a subgroup of

G. Then, the order of A must be a divisor of the order of G (by the Lagrange’s
­Theorem 4.4.2). Therefore, |A| divides |G| 5 8 and hence |A| 5 1 or 2 or 4 or 8.

Q001-Algebra-111001_CH 04.indd 41 9/22/2011 11:25:18 AM

 4-42   Algebra – Abstract and Modern

If |A| 5 1, then A 5 {1} and hence A is normal.

If |A| 5 8, than A 5 G and hence A is normal in G.
|G| 8
If |A| 5 4, than iG(A) 5   2 and hence, by Worked Exercise 4.5.1, A
| A| 4
is normal.
Finally suppose that |A| 5 2, then A 5 {1, a}, where a is an element of G such
that a2 5 1 and a  1. The only such element in G is 21 and hence

A 5 {1, 21}.

Since (21)x 5 2x 5 x(21) and 1x 5 x 5 1x for all x  G, it follows that

xA 5 {x, 2x} 5 Ax

for all x  G. Thus, A is normal in G.

Worked Exercise 4.5.3. Prove that the centre of any group is normal.

Answer: Recall that, for any group G, the centre of G is defined as the set

Z(G)5{a  G : ax 5 xa for all x  G}.

Thus, for any x  G, xZ(G) 5 Z(G)x and hence Z(G) is a normal subgroup of G.

Worked Exercise 4.5.4. Let X be any nonempty set X and (S(X), o) be the
group of bijections of X onto itself. For any x  X, let

Hx 5 {f  S(X) : f(x) 5 x}.

Then prove that Hx is a normal subgroup of S(X) for all x  X if and only if X
has at most two elements.

Answer: Suppose that X has three distinct elements, say x, y and z. Let f and
g : X → X be defined by

f(x) 5 y, f(y) 5 x and f(s) 5 s for all s  X 2 {x, y}

and    g(y) 5 z, g(z) 5 y and g(s) 5 s for all s  X 2 {x, y}.

Then, g  Hx and f  S(X) and

(fgf 2l)(x) 5 f(g(f2l(x)) 5 f(g(y)) 5 f(z) 5 z  x

and hence fg f 2l  Hx. Therefore, fHx f 21  Hx and hence Hx is not normal

in S(X).

Q001-Algebra-111001_CH 04.indd 42 9/22/2011 11:25:19 AM

 Subgroups and Quotient Groups    4-43

On the other hand, suppose X has at most two elements. If |X| 5 1, then
S(X) 5 {e} 5 Hx and hence Hx is normal. If |X| 5 2, then X 5 {x, y}, where
x  y, and Hx 5 {e} 5 Hy, which is clearly normal.

EXERCISE 4(e)

1. Let A and B be subgroups of a group G and one of them be normal in G. Then

prove that AB is a subgroup of G. Is AB normal in G?

2. For any normal subgroup A of a group G and for any subgroup B of G, prove that
A ∩ B is a normal subgroup of B.

3. Let G be the set consisting of the following eight matrices.

1 0 1 0     
 ,  ,  0 1, 0 1
0 1  0 1 1 0 1 0 

 i 0 0 i     
 ,  , i 0 and  0 i
0 i   i 0  0 i  i 0 

Prove that G is a group under the multiplication of matrices over the complex
numbers. Also prove that G is not abelian and that every subgroup of G is nor-
mal. Compare this with the example given in Worked Exercise 4.5.2.

4. Determine all the normal subgroups of the group S(X) of bijections of a 3-ele-
ment set X onto itself.

5. Let G be a group and n  Z1. If A is the unique subgroup of order n in G, then

prove that A is normal in G.

6. Let A be a subgroup of a group G such that, for any x and y  G,

xA 5 yA ⇒ Ax 5 Ay
Then prove that A is normal subgroup of G.

7. Let A and B be two normal subgroups of a group G such that |A ∩ B| 5 1. Then

prove that ab 5 ba for all a  A and b  B.

8. Give an example of a group G and subgroups A and B such that A ⊆ B, A is

normal in B, B is normal in G and A is not normal in G.

9. Let G be a group and a  G. Define

N(a) 5 {x  G : ax 5 xa}.
Prove that N(a) is a subgroup of G containing <a> as a normal subgroup. N(a) is
called the normaliser of a in G.

Q001-Algebra-111001_CH 04.indd 43 9/22/2011 11:25:19 AM

 4-44   Algebra – Abstract and Modern

10. Let A be a normal subgroup of a finite group G such that |A| and iG(A) are rela-
tively prime. For any x  G, prove that x  A if and only if x|A| 5 e.

11. Let G be a group containing a nontrivial subgroup A which is contained in every

nontrivial subgroup of G. Then prove that every element of G is of finite order.
What could be the order of A?

12. Let A be a subgroup of a group G and

C(A) 5 {x  G : xa 5 ax for all a  A}.
Prove that C(A) is a subgroup of G containing A as a normal subgroup. C(A) is
called the centralizer of A in G.

13. For any subgroup A of group G, let

N(A) 5 {x  G : xA 5 Ax}

Prove that N(A) is the largest subgroup of G containing A as a normal subgroup.

N(A) is called the normaliser of A in G.

14. Compare the normaliser and the centralizer of a subgroup of a group.

15. For any subgroup A of a group G, let

N  ∩ ( a A a1 ).
a∈G

Then prove that N is the largest normal subgroup of G containing A.

16. Let A be a subgroup of a group G such that x2  A for all x  G. Then prove that
A is normal in G.

17. For any real numbers a and b, define Tab : R → R by Tab(x) 5 ax1b for all x 
R. Prove that the set {Tab : a and b  R} is a group under the usual composition
of mappings in which {T1b : b  R} is a normal subgroup.

18. Prove that the intersection of any class of normal subgroups of a group G is
again a normal subgroup of G.

19. Prove that, for any group G, the centre Z(G )  ∩ N (a), where N(a) is the
a ∈G
normaliser of a in G.

20. For any set X with |X| $ 3, prove that Z(S(X)) 5 {e}.

21. Let G1 and G2 be any groups and A1 and A2 be subsets of G1 and G2 respectively.
Then prove that A1 3 A2 is a (normal) subgroup of the group G1 3 G2 if and only
if A1 and A2 are (normal) subgroups of G1 and G2, respectively.

22. If G1 and G2 are groups, prove that {e1} 3 G2 and G1 3 {e2} are normal sub-
groups of G1 3 G2, where e1 and e2 are identities in G1 and G2, respectively.

Q001-Algebra-111001_CH 04.indd 44 9/22/2011 11:25:19 AM

 Subgroups and Quotient Groups    4-45

4.6 QUOTIENT GROUPS

For any normal subgroup N of a group G, we have proved earlier that the
product of any two left cosets of N in G is again a left coset of N in G. That
is, taking product can be considered as a binary operation on the set of all left
cosets of N in G, which will be actually a group under this operation. This is
proved in the following definition.

Definition 4.6.1. Let G be a group and A and B be subsets of G. Then, we

define the product A ? B as the set

A ? B 5 {ab : a  A and b  B}.

Then, AB is called the product of A and B, in this order, induced by the binary
operation on G. Then, ? is a binary operation on the set P(G) of all subsets of
G and is called the multiplication of subsets of the group G.

Theorem 4.6.1. Let N be a normal subgroup of a group G and

G/N 5 {aN : a  G}.

Then, G/N is a group under the multiplication of subsets of the group; that is,
for any aN and bN  G/N,

aN∙bN 5 {xy : x  aN and y  bN}.

Proof: First of all recall that, aN∙bN 5 abN for all a and b  G (since N is a
normal subgroup of G) and hence the multiplication of subsets of the group
G is a binary operation on the set G/N of all left cosets of N in G.
For any a, b and c  G, we have

(aN?bN)?cN 5 (abN)?cN
5 ((ab)?c)N
5 (a(bc))N
5 aN?(bN?cN)

and hence the operation on G/N is associative.

Also, the coset eN 5 N satisfies the property that

(eN) ? (aN) 5 eaN 5 aN 5 (ae)N 5 (aN) ? (eN)

Q001-Algebra-111001_CH 04.indd 45 9/22/2011 11:25:19 AM

 4-46   Algebra – Abstract and Modern

for all aN  G/N and hence eN (5 N) is the identity in G/N. Further, for any
aN  G/N with a  G, consider

(a21N)?(aN) 5 a21a?N 5 eN 5 aa21N 5 (aN)(a21N)

and therefore a21N is the inverse of aN in G/N. Thus, G/N is a group under the
multiplication of subsets of the group G.

Definition 4.6.2. For any normal subgroup N of a group G, the group G/N
constructed above is called the quotient group of G by N. Whenever we refer
to a G/N as a group, we only mean the set of left(right) cosets of N in G
together with the multiplication of subsets of the group G.

Corollary 4.6.1. For any subgroup N of an abelian group G, G/N is an abelian

group.

Theorem 4.6.2. Let N be a normal subgroup of a finite group G. Then,

|G| 5 |N| |G/N|

Proof: This follows from the facts that |G/N| 5 iG(N) and |G| 5 |N|?iG(N).

Example 4.6.1. Let (Z, 1) be the group of all integers under the usual addi-
tion and let n be any positive integer and <n> 5 {na : a  Z} be the subgroup
of (Z, 1) generated by n. Since 1 is the operation on the group Z, the ele-
ments of Z/<n> are of the form a 1 <n> with a  Z. From Example 4.3.1
(2), we know that any coset of <n> in Z is of the form r 1 <n>, where 0 #
r , n. Thus,

Z/<n> 5 {N, 1 1 N, 2 1 N, …, (n 2 1) 1 N}, where N 5 <n>.

Example 4.6.2. Let E be the Euclidean plane and pick any point P in E and
fix a coordinate system with P as origin. Then, any point in E can be expressed
uniquely as an ordered pair (x, y) of real numbers and P corresponds to (0, 0).
For any points Q1 5 (x1, y1) and Q2 5 (x2, y2), define

Q1 1 Q2 5 (x1 1 x2, y1 1 y2).

The following diagram illustrates this addition of points by the construc-

tion of parallelogram with adjacent sides PQ1 and PQ2. Then, E is an abelian
group with respect to the above operation.

Q001-Algebra-111001_CH 04.indd 46 9/22/2011 11:25:20 AM

 Subgroups and Quotient Groups    4-47

Q1 + Q2
Q2(x2,y2)

Q1(x1,y1)

Let L be a straight line in E passing through P. Then, one can easily check
that L is a subgroup of E. Let us describe the cosets of L in E. If Q is any point in
E, then one can prove that the coset Q 1 L is precisely the line passing through
Q and parallel to L. Therefore, the quotient group E/L consists precisely all the
straight lines parallel to L, including L. For any points Q1 and Q2 in E, we have

(Q1 1 L) 1 (Q2 1 L) 5 (Q1 1 Q2) 1 L.

This is illustrated in the following figure.

L

Q2

Q1 + Q2

Q1

L Q1 + L Q2 + L Q1 + Q2 + L

Worked Exercise 4.6.1. Let a be an element of finite order in a group G and

N be a normal subgroup of G. Then prove that the order of the element aN in
G/N is a divisor of O(a).

Proof: Let O(a) 5 n. Then,

(aN)n 5 anN 5 eN 5 N (since O(a) 5 n, an 5 e)

N is the identity in the quotient group G/N. Therefore, aN is of finite order in

G/N and O(aN) is a divisor of n 5 O(a) (by Theorem 4.2.4 (3)).

Q001-Algebra-111001_CH 04.indd 47 9/22/2011 11:25:20 AM

 4-48   Algebra – Abstract and Modern

Definition 4.6.3. Let a and b be elements of a group G. Then, the product

aba21b21 is called the commutator of a and b and is usually denoted by
[a, b]; that is,

[a, b] 5 ab a21b21.

One can easily verify that ab 5 [a, b]ba and hence we can view the com-
mutator [a, b] as a measure of the extent to which ab differs from ba. In fact,
the elements a and b commute (that is, ab 5 ba) if and only if [a, b] 5 e,
the identity in G. The commutators of elements of a group G may not form a
subgroup of G, in general. However, we have the following theorem.

Theorem 4.6.3. Let G be a group and consider the set

 n 
[G , G ]  ∏[ai , bi ] : ai and bi ∈G .
 i1 

Then, [G, G] is a subgroup of G. Also, any subgroup of G containing [G, G]

is normal in G.

Proof: For any a and b  G, we have

[a, b]21 5 (ab a21b21)21 5 bab21a21 5 [b, a]

and therefore [G, G] is precisely the subgroup of G generated by the set {[a, b] :
a and b  G} (see Theorem 4.1.5). Thus, [G, G] is a subgroup of G. Next, let
A be any subgroup of G such that [G, G] ⊆ A. For any a  A and x  G,

xax21 5 (xax21a21)a  A,

since a  A and xax21a21 5 [x, a]  [G, G] ⊆ A. Thus, A is a normal sub-

group in G.

Corollary 4.6.2. For any group G, [G, G] is a normal subgroup of G.

Definition 4.6.4. For any group G, the subgroup [G, G] is called the derived
subgroup or commutator subgroup of G and the quotient group G/[G, G] is
called the commutator quotient group or abelianized group. The reason for
the latter terminology is the following theorem.

Theorem 4.6.4. Let N be any normal subgroup of a group G. Then, the

quotient group G/N is abelian if and only if [G, G] ⊆ N.

Q001-Algebra-111001_CH 04.indd 48 9/22/2011 11:25:21 AM

 Subgroups and Quotient Groups    4-49

Proof: Since ab 5 ba ⇔ [a, b] 5 e, the identity, for any elements a and b in

any group, we have

G/N is abelian ⇔ [aN, bN] 5 N for all aN and bN [ G/N

⇔ aN?bN?(aN) (bN)21 5 N for all a, b [ G
21

⇔ (aba21b21)N 5 N for all a, b,  G

⇔ [a, b]  N for all a, b  G
⇔ [G, G] ⊆ N.

The following is a direct consequence of Corollary 4.6.2 and Theorem

4.6.4.

Corollary 4.6.3. For any group G, G/[G, G] is an abelian group.

Theorem 4.6.4 and Corollary 4.6.3 say that the commutator subgroup
[G, G] is the smallest normal subgroup of G having an abelian quotient group.
The transition from a group G to its commutator quotient group G/[G, G] is
referred to as the abelianization of the group G and provides a convenient
procedure to produce abelian groups from nonabelian ones. Note that a group
G is abelian if and only if the commutator subgroup [G, G] 5 {e}.
In the following, we give a procedure to find all subgroups of the quotient
G/N, where N is a given normal subgroup of a group G.

Theorem 4.6.5. Let N be a normal subgroup of a group G and G/N be the

quotient group. For any subgroup A of G containing N, A/N 5 {aN : a  A}
is a subgroup of G/N. Further, A  A/N is a one-to-one correspondence
between the subgroups of G containing N and the subgroups of G/N.

Proof: For any a and b  A,

( aN )(bN 1 )  ( aN )(b1 N )  ab1 N ∈ A / N ,

since A is a subgroup, and hence A/N is a subgroup of G/N. If A and B are

subgroups of G containing N and A/N 5 B/N, then

a  A ⇒ aN  A/N 5 B/N
⇒ aN 5 bN for some b  B
⇒ a21b  N
⇒ ab21  N ⊆ B (since N is normal)
⇒ a 5 (ab21) b  B

Q001-Algebra-111001_CH 04.indd 49 9/22/2011 11:25:21 AM

 4-50   Algebra – Abstract and Modern

and hence A ⊆ B. Similarly, B ⊆ A and therefore A 5 B. Thus, A  A/N is

one-one. Further, if M is any subgroup of G/N, define A 5 {x  G : xN  M}.
Then, A is a subgroup of G containing N and A/N 5 M.
Thus, A  A/N is a one-to-one (bijective) correspondence between the
subgroups of G containing N and the subgroups of G/N.

Remark 4.6.1. Note that, in the above, A/N is normal in G/N if and only if A
is normal in G. Also, for any subgroups A and B containing N, A/N ⊆ B/N if
and if only A ⊆ B.
If N 5 {e}, then G/N  G and, if N 5 G, then G/N 5 {N}, the trivial
group.

EXERCISE 4(f)

1. Describe the quotient group of each of the following in the groups mentioned
against them.
(i) {0, 4, 8, 12} in (Z16, 116).
(ii) The set E of even integers in (Z, 1).
(iii) Z in (Q, 1).
(iv) {1, 21} in ({1, 21, i, 2i}, ∙).
(v) R in (C, 1).
(vi) Q in (R, 1).

2. Let G 5 <a> be a cyclic group of order 15 and A 5 <a3>. Construct multiplica-

tion table representing the quotient group G/A.

3. For any group G, determine the quotient groups of the trivial normal subgroups
{e} and G.

4. Let A be a subgroup of a group G such that x2  A for all x  G. Then prove that
A is normal in G and the quotient group G/A is abelian.

5. Let Z(G) be the centre of a group G. If G/Z(G) is cyclic, prove that G is

abelian.

6. Let N be a normal subgroup of a finite group G. Then prove the following.

(i) |G/N| 5 |G|/|N| 5 iG(N).
(ii) If n 5 iG(N), then xn  N for all x  G.
(iii) The order of aN in G/N is a divisor of the order of a in G, for any a  G.
Can they be equal? Justify your answer.

7. Let N be a normal subgroup of a group G. Then prove that G is finite if and only
if both N and G/N are finite.

Q001-Algebra-111001_CH 04.indd 50 9/22/2011 11:25:25 AM

 Subgroups and Quotient Groups    4-51

8. A group G is said to be finitely generated if G 5 <F> for some finite subset F of

G. Let A be a subgroup of an abelian group such that both A and G/A are finitely
generated. Then prove that G is finitely generated.

9. Let G be a group and S 5 {a2 : a  G}. Then prove that <S > is a normal sub-
group of G and that G/<S > is an abelian group.

10. List all normal subgroups of the group (S(X), o) of bijections of a 3-element set
X onto itself and construct tables representing the quotient group of each normal
subgroup S(X).

11. Let G 5 {(a, b)  R 3 R : a  0} and, for any (a, b) and (c, d)  G, define
(a, b) * (c, d) 5 (ac, ad 1 b).
Then prove that (G, *) is a group. If K 5 {(1, b) : b  R}, then prove K is a
normal subgroup of G.

12. Let G be a group of order 2n, where n is odd. Prove that G contains a normal
subgroup of index 2.

Q001-Algebra-111001_CH 04.indd 51 9/22/2011 11:25:26 AM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 04.indd 52 9/22/2011 11:25:26 AM

 5
Homomorphisms
of Groups
5.1 Definition and Examples
5.2 Fundamental Theorem of Homomorphisms
5.3 Isomorphism Theorems
5.4 Automorphisms

Homomorphisms play a major role in all aspects of modern algebra. One of

the most difficult problems in the theory of groups is to list all finite groups
having the same order. The difficulty is that we may list the same groups in
different forms. The notion of isomorphism builds up an equivalence relation
between groups, so that we may consider two groups belonging to a given
equivalence class as the same. Just as we cannot conclude that two human
beings are same, because they wear an identical set of clothes or the same
person putting on a different set of clothes does not become different, groups
are to be recognized as same or different on the basis information that is not
readily apparent. Consider the group (Z4, 14), where Z4 5 {0, 1, 2, 3} and 14
is the addition modulo 4, and the group (G, ?), where G 5 {1, i, 21, i} and ‘?’
is the usual multiplication of complex numbers. We have

Z4 5 <1> 5 {0, 1, 2, 3} and G 5 {1, i, i2, i3}

and hence these two groups look same as we have the correspondence f : Z4 → G
defined by f (0) 5 0, f (1) 5 i, f (2) 5 i2 and f (3) 5 i3. This correspondence is
compatible with the operations 14 on Z4 and ‘?’ on G and is a bijection of Z4
onto G. This helps us in proving that any property of Z4 gives a similar prop-
erty in G9. Such correspondences are called homomorphisms.

Q001-Algebra-111001_CH 05.indd 1 9/16/2011 10:06:01 AM

 5-2   Algebra – Abstract and Modern

In this chapter, we focus our attention on homomorphisms of groups. A

homomorphism of two groups may reveal some information about one of
the groups as a deduction of known structural properties of the other. If f is a
homomorphism of a group G onto another group G9 and if we know all about
the structure of G, then we can deduce the structure of G9also.

5.1 DEFINITION AND EXAMPLES

For any groups G and G9, we are interested in the mappings f of G into G9
such that the product of images of two elements in G is same as the image of
the product. If * and o are the binary operations in groups G and G9, respec-
tively, then a mapping f : G → G9 is called a homomorphism if the figure
given below is commutative, in the

GxG * G

fxf f

G′ x G′ G′
ο

sense that, the composition of f and * is same as that of o and f 3 f, where

f 3 f : G 3 G → G9 3 G9 is defined as (f 3 f )(x, y) 5 (f (x), f (y)). This is
precisely expressed in the following definition.

Definition 5.1.1. Let G and G9 be groups. Then, a mapping f : G → G9 is

called a homomorphism of G into G9 if

f (ab) 5 f (a)f (b)

for all elements a and b in G.

Here, the product ab which appears on the left side of the above equation cor-
responds to the binary operation on G, whereas the product f (a)f (b) on the
right side corresponds to the binary operation on G9. As we have agreed to
skip the symbol denoting the binary operation in a group and to write simple
ab for a * b, when * is the binary operation of G and a and b are elements of
G, the equation f (ab) 5 f (a)f (b) makes sense since a and b are elements of G
and f (a) and f (b) are those in G9. Strictly speaking, we should have written

Q001-Algebra-111001_CH 05.indd 2 9/16/2011 10:06:01 AM

 Homomorphisms of Groups    5-3

f ( a ∗ b)  f ( a) ο f (b) for all a, b ∈ G ,

where * and o are the binary operations in the groups G and G9, respectively.
Examples given below make these things clear.

Example 5.1.1
1. Consider the groups (Z, 1) and (R1, ?) where 1 and ‘?’ are the usual
addition and multiplication of real numbers. Let m be any positive inte-
ger and define f : Z → R1 by
f (a) 5 ma for all a  Z.

Then, for any a and b  Z, we have

f (a 1 b) 5 ma1b 5 ma  mb 5 f (a) ? f (b)

and hence f is a homomorphism of (Z, 1) into (R1, ?).

2. Consider the groups (R1, ?) and (R, 1) and let m be any positive integer
greater than 1. Define
f : R1 → R by f (a) 5 logma for all a  R1.

Then, for any a and b  R1,

f (a ? b) 5 logm (ab) 5 logma 1 logmb 5 f (a) 1f (b).

Therefore, f is a homomorphism of (R1, ?) into (R, 1).

3. For any group G, define f : G → G by
f (x) 5 x for all x  G.

Then, clearly f is a homomorphism of G into itself and is called the iden-

tity homomorphism.
4. Let G and G9 be groups in which e and e9 are the identities, respectively.
Define
f : G → G9 by f (x) 5 e9 for all x  G.

Then, for any a and b in G, we have

f (ab) 5 e9 5 e9 ? e9 5 f (a) f (b)

and hence f is a homomorphism of G into G9 and is called the trivial

homomorphism.
5. Let m be an arbitrary integer and define f : Z → Z by
f (a) 5 ma for all a  Z.

Q001-Algebra-111001_CH 05.indd 3 9/16/2011 10:06:02 AM

 5-4   Algebra – Abstract and Modern

Then, f (a 1 b) 5 m(a 1 b) 5 ma 1 mb 5 f (a) 1 f (b) for all a and

b  Z. Therefore, f is a homomorphism of (Z, 1) into itself.
6. Consider the groups (Z, 1) and (Z2, 12), where 12 is the addition mod-
ulo 2. Define

0, if a is even
f : Z → Z2 by f ( a)  
1, if a is odd
for any a  Z. Then, since a 1 b is even if and only if both a and b are
even or both a and b are odd, we get that
f (a 1 b) 5 f (a) 12 f (b) for all a and b  G.

Therefore, f is a homomorphism of (Z, 1) into (Z2, 12).

7. Consider the group (R2{0}, ?) of all nonzero real numbers under the
usual multiplication of real numbers and the group NSM2(R) of all 2 3 2
nonsingular matrices over R under the usual multiplication of matrices.
Define
a b 
f : NSM2(R) → R 2 {0} by f    ad bc

c d 
a b 
for any    NSM2(R).
c d 
a b  r s 
Then, for any A    and B    in NSM2(R), we have
 c d   t u

  
ar bt asbu
f ( AB)  f  
 
 crdt csdu 
 ( ar  bt ) (cs  du )  ( as  bu ) (cr  dt )
 arcs  ardu  btcs  btdu  ascr  asdt  bucr  budt
 ardu  btcs  asdt  bucr
 ( ad  bc) ( ru  st )
 f ( A) f ( B).

Therefore, f is a homomorphism of (NSM2(R), ?) into (R 2 {0}, ?).

8. Let N be a normal subgroup of a group G and G/N, the quotient group
of G by N. Define
f : G → G/N by f (a) 5 aN for all a  G.

Q001-Algebra-111001_CH 05.indd 4 9/16/2011 10:06:02 AM

 Homomorphisms of Groups    5-5

Then, for any a and b in G,

f (ab) 5 (ab)N 5 aN ? bN 5 f (a)?f (b)
and hence f is a homomorphism of G into G/N and is called the natural
or cannonical homomorphism.
In the following, we derive certain important elementary properties of
homomorphisms.

Theorem 5.1.1. Let f : G → G9 be a homomorphism of groups. Then, the

following holds:
1. f (e) 5 e9, where e and e9 are the identities in G and G9, respectively.
2. f (a21) 5 f (a)21 for all a  G.

Proof:
1. Let f (e) 5 x. Then, x is an element in G9 and hence
xe9 5 x 5 f (e) 5 f (ee) 5 f (e)f (e) 5 xx.
Therefore, xx 5 xe9 and, by the left cancellation law, x 5 e9. Thus,
f (e) 5 e9.
2. For any a  G, we have
f (a)f (a21) 5 f (aa21) 5 f (e) 5 e9 5 f (e) 5 f (a21a) 5 f (a21) f (a)
and hence f (a21) is the inverse of f (a) in G9; that is,
f (a21) 5 f (a)21.

Theorem 5.1.2. Let f : G → G9 be a homomorphism of groups.

1. For any subgroup A of G, the image f (A) is a subgroup of G9.
2. For any subgroup A9 of G9, the inverse image f21(A9) is a subgroup of G.

Proof:
1. Let A be a subgroup of G. Then,
f (A) 5 {f (a) : a  G} ⊆ G9.
First of all, since A is a subgroup of G, we have e  A and hence
e9 5 f (e)  f (A)
so that f (A) is a nonempty subgroup of G9. Also,
x and y  f (A) ⇒ x 5 f (a) and y 5 f (b) with a and b  A.
⇒ xy21 5 f (a)f (b)21 5 f (a) f (b21) 5 f (ab21)
                f (A) (since A is a subgroup).

Q001-Algebra-111001_CH 05.indd 5 9/16/2011 10:06:03 AM

 5-6   Algebra – Abstract and Modern

Therefore, f (A) is a subgroup of G9.

2. Let A9 be a subgroup of G9. Then,
f21(A9) 5 {a  G : f (a)  A9}.

Since f (e) 5 e9  A9, we get that e  f21(A9) and hence f21(A9) is a

nonempty subset of G. Now,
a and b  f21(A9) ⇒ f (a) and f (b)  A9
⇒ f (ab21) 5 f (a)f (b21) 5 f (a)f (b)21  A9
(since A9 is a subgroup of G9)
⇒ ab21  f21(A9).

Thus, f21(A9) is a subgroup of G.

Theorem 5.1.3. Let f : G → G9 be a homomorphism of groups.

1. If f is a surjection and A is a normal subgroup of G, then f (A) is a normal
subgroup of G9.
2. If A9 is a normal subgroup of G9, then f21(A9) is a normal subgroup of G.

Proof:
1. Let f be a surjection and A be a normal subgroup of G. Then, we have
already proved that f (A) is a subgroup of G9. Now,
x  f (A) and y  G9 ⇒ x 5 f (a) and y 5 f (b) for some a  A
and b  G
⇒ y x y21 5 f (b)f (a)f (b)21 5 f (bab21)  f (A).

Therefore, yf (A)y21 ⊆ f (A) for all y  G9 and hence f (A) is a normal

subgroup of G9.
2. Let A9 be a normal subgroup of G9. By the above theorem, f21(A9) is a
subgroup of G. Now,
a  f21(A9) and x  G ⇒ f (a)  A9 and f (x)  G9
⇒ f (xax21) 5 f (x)f (a)f (x)21  A9
(since A9 is normal in G9)
⇒xax21  f21(A9).

Therefore, f21(A9) is a normal subgroup of G.

Remark 5.1.1. Note that in Theorem 5.1.3 (1), it is necessary that f is a sur-
jection; for, consider a group G9 and a subgroup G of G9 such that G is not
normal in G9. Let f : G → G9 be the inclusion map defined by f (x) 5 x for all

Q001-Algebra-111001_CH 05.indd 6 9/16/2011 10:06:03 AM

 Homomorphisms of Groups    5-7

x  G. Then, f is a homomorphism of G into G9, G is a normal subgroup of

G9, but f (G) (5 G) is not normal in G9.
For any homomorphism f : G → G9, the subgroup f21({e9}) of G is of the
special importance and is called the kernel of f. This is formally defined in
the following definition.

Definition 5.1.2. Let f : G → G9 be a homomorphism of groups. Then, the

kernel of f is defined to be the set

ker f 5 {a  G : f (a) 5 e9},

where e9 is the identity in the group G9.

Theorem 5.1.4. For any homomorphism f : G → G9 of groups, the kernel of

f is a normal subgroup of G.

Proof: By the definition of the kernel of f, we have

ker f 5 f21({e9}),

where e9 is the identity in G9. Since {e9} is a normal subgroup of G9,

it ­follows from Theorems 5.1.2 (2) and 5.1.3 (2) that ker f is a normal
­subgroup of G.
The converse of the above theorem is also true, in the sense that, any nor-
mal subgroup of a group G is the kernel of some homomorphism of G into
some group G9. This is proved in the following theorem.

Theorem 5.1.5. Let N be a normal subgroup of a group G and G/N, the quo-
tient group of G by N. Define

f : G → G/N by f (a) 5 aN for all a  G.

Then, f is a homomorphism, whose kernel is N.

Proof: For any a and b in G,

f (ab) 5 abN
5 aN ? bN (since N is normal in G)
5 f (a)f (b)

and hence f is a homomorphism (see Example 5.1.1 (8) also) and

Q001-Algebra-111001_CH 05.indd 7 9/16/2011 10:06:03 AM

 5-8   Algebra – Abstract and Modern

ker f 5 {a  G : f (a) 5 the identity in G/N}

5 {a  G : aN 5 N 5 eN}
5 {a  G: a  N} 5 N.

Let us compute the kernels of homomorphisms given in Example 5.1.1.

Example 5.1.2

1. f : Z → R1 is defined by f (a) 5 ma for all a  Z, where m is a given

positive integer. Now,
ker f 5 {a  Z : f (a) 5 the identity in R1}
5 {a  Z : ma 5 1}
Z, if m  1
  .
{0}, if m  1

2. f : R1 → R is defined by f (a) 5 logma.

ker f 5 {a  R1: f (a) 5 the identity in (R, 1)}
5 {a  R1 : logma 5 0}
5 {1}.

3. For any group G, define f (x) 5 x for all x  G. Then,

ker f 5 {e}.

4. f : G → G9 is defined by f (x) 5 e9 for all x  G.

Therefore, ker f 5 G.
5. f : Z → Z is defined by f (a) 5 ma for all a  Z, where m is a given
integer. Then,
ker f 5 {a  Z : f (a) 5 0, the identity in (Z, 1)}
5 {a  Z | ma 5 0}


{Z, if m0
{0}, if m0
.

6. f : Z → Z2 is defined by f ( a) 
{ 0, if a is even
1, if a is odd
.

ker f 5 {a  Z : f (a) 5 0, the identity in (Z2, 12)}

5 2Z, the set of all even integers.

Q001-Algebra-111001_CH 05.indd 8 9/16/2011 10:06:03 AM

 Homomorphisms of Groups    5-9

7. f : NSM2(R) → R 2 {0} is defined by

a b 
f    ad  bc.
 c d 
a b  
ker f    ∈ NSM ( R) : ad  bc  1 .
 c d  2


8. The kernel of the natural map f : G → G/N, defined by f (a) 5 aN, is

precisely N.
In the following, we define various types of homomorphisms.

Definition 5.1.3. Let f : G → G9 be a homomorphism of groups. Then, f is

called
1. a monomorphism if f is an injection.
2. an epimorphism if f is a surjection.
3. an isomorphism if f is a bijection.
4. an endomorphism if G 5 G9.
5. an automorphism if f is a bijective endomorphism.

Example 5.1.3
1. If m > 1, then the map f : Z → R1, defined by f (a) 5 ma, is a monomor-
phism of (Z, 1) into (R1, ?).
2. The homomorphism given in (2), (3) and (5) (if m  0) of Example
5.1.2 are monomorphisms, while those given in (2), (3), (6) and (8) are
epimorphisms.
Note that a homomorphism f : G → G9 is an epimorphism if and only if
f (G) 5 G9. In the following, we give a characterization of monomorphisms
in terms of their kernels.

Theorem 5.1.6. Let f : G → G9 be a homomorphism of groups. Then,

f is a monomorphism if and only if the kernel of f is trivial, that is, ker
f 5 {e}.

Proof: Suppose that f is a monomorphism. Then, f is an injection and

therefore

a  ker f ⇒ f (a) 5 e9 5 f (e) ⇒ a 5 e

Q001-Algebra-111001_CH 05.indd 9 9/16/2011 10:06:04 AM

 5-10   Algebra – Abstract and Modern

and hence ker f contains e alone. Thus, ker f 5 {e}. Conversely, suppose that
f is not a monomorphism, then f is not an injection and hence there exists
a  b  G such that f (a) 5 f (b). Now, ab21  e and

f (ab21) 5 f (a)f (b21) 5 f (a)f (b)21 5 f (b)f (b)21 5 e

and hence e  ab21  ker f, so that ker f  {e}.

Recall that a mapping f : A → B is a bijection if and only if there exists a
unique mapping g : B → A such that

f o g 5 IB and g o f 5 IA;

that is, f (g(b)) 5 b for all b  B and g(f (a)) 5 a for all a  A.
This unique g is called the inverse of f and is denoted by f21. Also, in this case,
for any a  A and b  B, we have

f (a) 5 b ⇔ a 5 f21(b).

In the following, we prove that the inverse of a bijective homomorphism

is again a homomorphism.

Theorem 5.1.7. Let f : G → G9 be an isomorphism of groups. Then, f21 : G9

→ G is also an isomorphism.

Proof: We are given that f : G → G9 is a bijective homomorphism. Then,

clearly f21 : G9 → G is a bijection. Now, for any x and y  G9,

f (f21(xy)) 5 xy 5 f (f21(x)) · f (f21(y)) 5 f (f21(x)f21(y))

and since f is a bijection, we have

f21(xy) 5 f21(x)f21(y) for all x and y  G9.

Thus, f21 is a homomorphism and a bijection and hence an isomorphism.

Definition 5.1.4. Two groups G and G9 are said to be isomorphic if there is

an isomorphism f : G → G9 and, in this case, we write G  G9.

Theorem 5.1.8. The following holds for any groups G, G9 and G:
1. G  G
2. G  G9 ⇒ G9  G
3. G  G9 and G9  G ⇒ G  G

Q001-Algebra-111001_CH 05.indd 10 9/16/2011 10:06:04 AM

 Homomorphisms of Groups    5-11

Proof:
1. follows from the fact that the identity map IG on G is an isomorphism of
G onto itself.
2. is consequence of Theorem 5.1.7 and
3. follows from the fact that the composition of two isomorphisms is again
an isomorphism.

Example 5.1.4. Consider the groups (Z4, 14) and (G, ?), where G 5 {1, 21, i,
2i} and ‘?’ is the multiplication of complex numbers. Define f : Z4 → G by

f (0) 5 1, f (1) 5 i, f (2) 5 i2 5 21 and f (3) 5 i3 5 2i.

Then, f is an isomorphism and hence Z4  G.

Worked Exercise 5.1.1. If f : G → G9 and g : G9 → G are homomorphisms

of groups, prove that g o f : G → G is a homomorphism.

Answer: For any a and b  G, we have

(g o f )(ab) 5 g(f (ab))

5 g(f (a)f (b))
5 g(f (a))g(f (b))
5 (g o f )(a)?(g o f )(b)

Therefore, g o f is a homomorphism.

Worked Exercise 5.1.2. Let f : G → G9 be a homomorphism of groups and a

be an element of finite order in G. Prove that the order of f (a) is finite in G9
and that O(f (a)) divides O(a).

Answer: Let a  G be of finite order and O(a) 5 n. Then, an 5 e and (f (a))n 5

f (an) 5 f (e) 5 e9.
Therefore, f (a) is of finite order and, by Theorem 4.2.4 (3), O(f (a))
divides O(a).

Worked Exercise 5.1.3. Let f : G → G9 be a homomorphism of groups and K

5 ker f. Describe the cosets of K in terms of f.

Answer: We know that K 5 ker f 5 {a  G : f (a) 5 e9} and that K is a nor-

mal subgroup of G and hence every left coset of K is a right coset of K in G.
For any a  G, consider

Q001-Algebra-111001_CH 05.indd 11 9/16/2011 10:06:04 AM

 5-12   Algebra – Abstract and Modern

aK 5 {ax : x  K}
5 {ax : f (x) 5 e9}
5 {y  G : f (y) 5 f (a)},

since f (ax) 5 f (a)f (x) 5 f (a)e9 5 f (a) and, if y  G is such that f (y) 5 f (a), then
y 5 ax, where x 5 a21y  ker f 5 K. Thus, aK 5 f21{f (a)} for any a  G.

Worked Exercise 5.1.4. Let n be a positive integer and Zn 5 {0, 1, 2, …,

n 2 1}. Define f : Z → Zn by f (a) 5 r, where r is the remainder obtained by
dividing a with n. Then, prove that f is an epimorphism of (Z, 1) onto (Zn, 1n).

Answer: Note that, for any a  Z, f (a) 5 r, where

a 5 qn 1 r, q and r  Z and 0  r < n

and hence f (a)  Zn. Therefore, f : Z → Zn is a mapping. For any a and

b  Z, let

f (a) 5 r and f (b) 5 s.

Then, a 5 qn 1 r and b 5 q9n 1 s, where q and q9 are integers.

Now,

a  b  ( q  q)n  ( r  s)
( q  q)n  ( r  s), if r  s  n
 
( q  q1)n  ( r  s  n), if r  s  n
5 tn 1 (r 1n s) and 0  r 1n s < n

and hence f (a 1 b) 5 r 1n s 5 f (a) 1n f (b).

Thus, f is a homomorphism. Also, for any 0  a < n, f (a) 5 a and hence f is a
surjection also. Thus, f is an epimorphism of (Z, 1) onto (Zn, 1n).

Worked Exercise 5.1.5. Let G be an abelian group of order m and let n be any
positive integer relatively prime to m. Define f : G → G by f (a) 5 an for all a
 G. Then, prove that f is an automorphism of G.

Answer: Since G is an abelian group, we have

f (ab) 5 (ab)n 5 anbn 5 f (a)f (b)

for all a and b  G and hence f is a homomorphism of G into G. For any a 

G, O(a) divides |G| 5 m and hence

am 5 e for all a  G.

Q001-Algebra-111001_CH 05.indd 12 9/16/2011 10:06:05 AM

 Homomorphisms of Groups    5-13

Since n is relatively prime to m, there exist integers r and s such that rm 1

sn 5 1. Now, for any a and b  G,

f (a) 5 f (b) ⇒ an 5 bm and am 5 e 5 bm

⇒ a 5 arm1sn 5 (am)r(an)s
5 (bm)r(bn)s 5 brm1sn 5 b

and hence f is an injection. Therefore,

m 5 |G| 5 |f (G)|  |G| 5 m

and hence |f (G)| 5 |G| so that f (G) 5 G (since G is finite and f (G) ⊆ G).
Therefore, f is a surjection also. Thus, f is an isomorphism of G onto G; that
is, f is an automorphism.

EXERCISE 5(a)

1. Determine which of the following are homomorphisms between the given

groups:
(i) Consider the group (Z, 1) and define f : Z → Z by f (a) 5 2a for all a  Z.
(ii) Consider the groups (Z6, 16) and (Z8, 18) and define f : Z6 → Z8 by f (a) 5 a
for all a  Z6.
(iii) Consider the groups (Z6, 16) and (Z, 1) and define f : Z6 → Z by f (a) 5 a
for all a  Z6.
0, if a is even
(iv) Define f : Z8 → Z2 by f ( a)   .
1, if a is odd
(v) Define f : Z15 → Z2 as in (iv) above.
(vi) Consider the group (R, 1) and define f : R → R by f (x) 5 cos x for
all x  R.
(vii) Let Y be a nonempty subset of a set X and consider the group P(X) and
P(Y) under the symmetric difference of sets. Define f : P(X) → P(Y) by
f (A) 5 A ∩ Y for any A  P(X).
(viii) Let X and Y be as in (vii) above. Define f : P(X) → P(Y) by f (A) 5 Y 2 A
for any A  P(X).
(ix) Let G be the group of all real valued continuous functions of the
interval [0, 1] under point-wise addition. Define f : G → R by
1
f (a)  ∫ a( x )dx for all a ∈ G.
0
(x) Let X be any nonempty set and RX be the set of all mappings of X into R.
Consider the group (RX, 1) where 1 is the point-wise addition. For any
x0  X, define f : RX → R by f (a) 5 a(x0) for all a  RX.

Q001-Algebra-111001_CH 05.indd 13 9/16/2011 10:06:05 AM

 5-14   Algebra – Abstract and Modern

2. Let G and G9 be groups and f : G → G9 be a mapping. Prove that f is a homomor-

phism if and only if
f (ab21) 5 f (a)f (b)21 for all a and b  G.

3. Let C 2 {0} and R 2 {0} be the groups of nonzero complex numbers and nonzero
real numbers, respectively, under the usual multiplications. Prove that the map
f : C 2 {0} → R 2 {0} defined by f (z) 5 |z| is a homomorphism.
4. Let G be a group and define f : G → G by f (a) 5 a21 for any a  G. Prove that f
is an endomorphism if and only if G is an abelian group.
5. Determine the Kernels of the homomorphisms (if they are) given in Exercise 1
above.
6. Determine all the homomorphisms of (Z, 1) into (Z2, 12).
7. Determine all the endomorphisms of the group (Z, 1) into itself.
8. Prove that every nontrivial endomorphism of (Z, 1) into itself is a
monomorphism.
9. Prove that there is no epimorphism of (Z, 1) onto itself, except the identity map.
10. Consider the groups (Z, 1) and (R, 1) and, for any real number a, define fa : Z → R
by fa(x) 5 ax for all x  Z. Prove that a mapping f : Z → R is a homomorphism
if and only if f 5 fa for some a  R.
11. Let G and G9 be finite groups and f : G → G9 be a homomorphism. Prove that
the index of the Kernel of f in G is a divisor of |f (G)|.
12. Let f : G → G9 be a homomorphism of groups and a  G. If O(a) is finite, then
prove that O(f (a)) is also finite and is a divisor of O(a). Give an example where
O(f (a)) is a proper divisor of O(a).
13. Let f and g : G → G9 be homomorphisms of groups and A 5 {a  G : f (a) 5
g(a)}. Then, prove that A is a subgroup of G.
14. Let f : G → G9 be a homomorphism and G be a finite group of prime order. Then,
prove that f is either trivial or a monomorphism.
15. Let f : G → G9 be a homomorphism of groups and [G, G] be the commutator
subgroup of G (see 4, 6, 12). Then, prove that f (G) is an abelian group if and
only if [G, G] ⊆ ker f.
16. Let G be a group and a and b  G. Consider the group Z 3 Z under coordinate-
wise addition and define f : Z 3 Z → G by
f (m, n) 5 ambn for any (m, n)  Z 3 Z.
Obtain a necessary and sufficient condition, in terms of a and b, for f to be a
homomorphism.

Q001-Algebra-111001_CH 05.indd 14 9/16/2011 10:06:05 AM

 Homomorphisms of Groups    5-15

17. Let a be an element in a group G and n  Z1. Define

f : Zn → G by f (i) 5 ai for 0  i < n.
Obtain a necessary and sufficient condition in terms of a and n for f to be a
homomorphism.
18. For any element a in a group G, define f : Z → G by f (n) 5 an for all n  Z.
Prove that f is a homomorphism and determine the kernel of f.
19. Let n  Z1. and G be the group of all nth roots of unity under the usual multipli-
cation of complex numbers. Prove that G  (Zn, 1n).
20. Let Q8 be the quaternion group with eight elements. Prove that there is a unique
homomorphism f : G → Z2 such that f (i) 5 0 and f (j) 5 1.
21. Let G and G9 be finite groups of the same order and f : G → G9 be a homomor-
phism. Then, prove that the following are equivalent to each other:
(i) f is a monomorphism.
(ii) f is an epimorphism.
(iii) f is an isomorphism.
22. Let R 2 {0} be the group of nonzero real numbers under multiplication and
G 5 {1, 21}. Define
 1, if a  0
f : R → G by f ( a)   .
1, if a  0
Then, prove that f is an epimorphism.
23. Prove that there is no epimorphism of (Z, 1) onto (R, 1).
24. Exhibit a monomorphism of Z8 into Z24.
25. For any positive integers m and n, obtain a necessary and sufficient condition for
having a monomorphism of Zm into Zn.
26. List all the isomorphisms of (Z, 1) onto (Z, 1).
27. How many homomorphisms are there from Z7 into Z15?
28. Let G and H be any groups and H be an abelain group. Let Hom(G, H) be the set
of all homomorphisms of G into H. Prove that Hom(G, H) is a group under the
point-wise operation.
29. Prove that the composition of any two monomorphisms (epimorphisms) is again
a monomorphism (epimorphism).
30. Let f : G → G9 and g : G9→G0 be homomorphisms, such that g o f is a mono-
morphism. Then, prove that f is a monomorphism. If g o f is an epimorphism,
then prove that g is an epimorphism.
31. Let G be a group and f : G → G be defined by f (x) 5 x3 for all x  G. If f is
monomorphism, then prove that G is abelian.

Q001-Algebra-111001_CH 05.indd 15 9/16/2011 10:06:06 AM

 5-16   Algebra – Abstract and Modern

5.2 FUNDAMENTAL THEOREM OF HOMOMORPHISMS

For any homomorphism f of a group G into a group G9, we know that f (G) is a
subgroup of G9 and hence f (G) is a group by itself. f (G) is called a homomor-
phic image of G. For example, if N is any normal subgroup of a group G and
G/N is the corresponding quotient group, then G/N is a homomorphic image
of G, since we have the natural map f : G → G/N which is a homomorphism
and f (G) 5 G/N. In the following, we prove a converse of the above; that is,
any homomorphic image of a group G is isomorphic to a quotient group of G.

Theorem 5.2.1 (Fundamental Theorem of Homomorphisms). Let f : G → G9

be a homomorphism of groups. Then,

G/ker f  f (G)

and, in particular, if f is an epimorphism, then

G/ker f  G9.

Proof: For simplicity, let K 5 ker f 5 {a  G : f (a) 5 e9}. We know that K

is a normal subgroup of G and hence we have the quotient group whose ele-
ments are the cosets of K in G. Also, we know that f (G) is a subgroup of G9
and hence f (G) is a group on its own.

Define g : G/K → f (G) by g(aK) 5 f (a)

for any aK  G/K. There is an apparent ambiguity in the definition of g. This

looks like depending on a. Actually, this does not depend upon the represen-
tative a of the coset aK. For, consider

aK 5 bK ⇒ a21b  K 5 ker f
⇒ f (a)21f (b) 5 f (a21)f (b) 5 f (a21b) 5 e9
⇒ f (a) 5 f (b)

This clears the ambiguity and it follows that g is well defined. For any aK and
bK  G/K, we have

g(aK  bK) 5 g(abK)

5 f (ab)
5 f (a)f (b)
5 g(aK)  g(bK)

Q001-Algebra-111001_CH 05.indd 16 9/16/2011 10:06:06 AM

 Homomorphisms of Groups    5-17

and therefore g is a homomorphism. Since any element of f (G) is of the form

f (a) 5 g(aK) for some a  G and aK  G/K, g is a surjection. Also, for any
aK and bK  G/K,

g(aK) 5 g(bK) ⇒ f (a) 5 f (b)

⇒ f (a21b) 5 f (a)21f (b) 5 e9
⇒ a21b  ker f 5 K
⇒ aK 5 bK.

Therefore, g is an injection also. Thus, g : G/K → f (G) is an isomorphism

and hence G/K  f (G). If f is an epimorphism, then f (G) 5 G9 and hence
G/K  G9.
For any sets A, B and C, if a : A → B is a bijection and B is a subset of C,
then a can be considered as an injection of A into C. On the other hand, if
b : A → C is an injection, then b can be considered as a bijection of A onto
b(A). Now, the following is an important consequence of the Fundamental
Theorem of Homomorphisms.

Theorem 5.2.2 (Factorization Theorem). Any homomorphism of groups can

be expressed as a composition of an epimorphism and a monomorphism.

Proof: Let f : G → G9 be a homomorphism of groups and K 5 ker f. Then,

by the Fundamental Theorem of Homomorphisms (Theorem 5.2.1), there
is an isomorphism g : G/K → f (G) such that g(aK) 5 f (a) for all a  G.
Now, since f (G) is a subgroup of G9, g can be considered as a monomor-
phism of G/K into G9. Also, let h : G → G/K be the natural homomorphism
defined by h(a) 5 aK for all a  G. Then, clearly h is an epimorphism and
we have,

h
G  g
→ G / K  →G′

for any a  G, (g o h) (a) 5 g(h(a)) 5 g(aK) 5f (a) are hence f 5 g o h. There-

fore, f is the composition of the epimorphism h and the monomorphism g.

Example 5.2.1. Let n be a positive integer and (Zn, 1n) be the group of inte-
gers modulo n. As in Worked Exercise 5.1.4, define f : Z→Zn by f (a) 5 r,
where r is the remainder obtained by dividing a with n; that is, if q and r are
integers such that a 5 qn 1 r and 0  r < n, f (a) 5 r.
In Worked Exercise 5.1.4, we have proved that f is an epimorphism. By the
Fundamental Theorem of Homomorphisms, Z/ker f  Zn, we have

Q001-Algebra-111001_CH 05.indd 17 9/16/2011 10:06:06 AM

 5-18   Algebra – Abstract and Modern

ker f 5 {a  Z : f (a) 5 0, the identity in Zn}

5 {a  Z : a 5 nq for some q  Z}
5 nZ
5 ,n., the subgroup of Z generated by n.

Theorem 5.2.3. Let G, G1 and G2 be groups. Then, prove that G  G1 3 G2

if and only if there exist normal subgroups N1 and N2 of G such that N1N2 5
G, N1∩ N2 5 {e}, G1  G/N1 and G2  G/N2.

Proof: Let e, e1 and e2 be identities in G, G1 and G2, respectively. Suppose that

G  G1 3 G2 and let f : G → G1 3 G2 be an isomorphism. Put A1 5 {e1} 3
G2 and A2 5 G1 3{e2}. Then, A1 and A2 are normal subgroups of G1 3 G2.
Now, put

N1 5 f21(A1) and N2 5 f21(A2).

Then, N1 and N2 are normal subgroups of G. For any x  G, we have f (x) 5

(a1, a2)  G1 3 G2.Choose x1 and x2 in G such that

f (x1) 5 (e1, a2) and f (x2) 5 (a1, e2).

Then, x1  f21(A1), x2  f21(A2) and

f (x) 5 (e1, a2) (a1, e2) 5 f (x1) f (x2) 5 f (x1 x2)

and hence x 5 x1x2  N1N2. Therefore, N1N2 5 G, Also,

x  N1 ∩ N2 ⇒ x  f21(A1) and x  f21(A2)

⇒ f (x)  A1 5 {e1} 3 G2 and f (x)  A2 5 G1 3 {e2}
⇒ f (x)  ({e1} 3 G2) ∩ (G1 3 {e2}) 5 {e1} 3 {e2}
⇒ f (x) 5 (e1, e2) 5 f (e) (since f is a homomorphism)
⇒ x 5 e.

Therefore, N1 ∩ N2 5 {e}. Next define f1 : G → G1 and f2 : G → G2 by

f1(x) 5 a1 and f2 (x) 5 a2 if f (x) 5 (a1, a2).

Then, f1 and f2 are epimorphisms and ker f1 5 N1 and ker f2 5 N2; For

x  N1 ⇔ f (x)  A1 ⇔ f (x) 5 (e1, a2) ⇔ f1(x) 5 e1

and    x  N2 ⇔ f (x)  A2 ⇔ f (x) 5 (a1, e2) ⇔ f2(x) 5 e2.

Q001-Algebra-111001_CH 05.indd 18 9/16/2011 10:06:07 AM

 Homomorphisms of Groups    5-19

Therefore, G/N1  G1 and G/N2  G2.

Conversely suppose that N1 and N2 are normal subgroups of G such that N1
N2 5 G, N1 ∩ N2 5 {e}, G/N1  G1 and G/N2  G2. Let a1 : G/N1 → G1 and
a2 : G/N2 → G2 be isomorphisms and b1 : G → G/N1 and b2 : G → G/N2 be
nature homomorphisms. Now define

 : G → G1 3 G2 by (x) 5 ((a1 o b1)(x), (a2 o b2)(x))

for any x  G since a1, b1, a2, b2 are all homomorphisms,  is also a homo-
morphism. Also, for any x  G,

(x) 5 (e1, e2) ⇔ a1(b1(x)) 5 e1 and a2(b2(x)) 5 e2

   ⇔ b1(x) 5 N1 and b2(x) 5 N2
(since a1 and a2 are isomorphism)
⇔ x  N1∩ N2 5 {e}.

Therefore, ker  5 {e} and hence  is a monomorphism. For any (x1, x2) 
G13G2, choose a1, a2  G such that a1(a1N1)5 x1 and a2(a2N2) 5 x2. Since N1
and N2 are normal subgroups and N1 ∩ N2 5 {e}, we get that ab 5 ba for all
a  N1 and b  N2 (for, consider aba21b21  N1 ∩ N2 5 {e}).
Now since a1 and a2  G 5 N1N2, we get that

a1 5 r1r2 and a2 5 s1s2 for some r1, s1  N1 and r2, s2  N2.

Put x 5 r2s1. Then,

a11 x  ( r21r11r2 ) s1 ∈ N1 (since r21r11r2 ∈ N1 and s1  N1)

1 1 1 1 1 1
and   a2 x  s2 s1 ( r2 s1 )  s2 s1 ( s1r2 ) = s2 r2 ∈ N 2
and hence a1N1 5 xN1 and a2N2 5 xN2. Now,

( x )  ((a1 ( b1 (x ))) ,a2 ( b2 (x )))

 (a1 (xN1 ),a2 (xN 2 ))
 (a1 (a1 N1 ),a2 (a2 N 2 ))  (x1 ,x2 ).

Thus,  is a surjection also. Therefore,  is an isomorphism and G 

G1 3 G2.

Theorem 5.2.4. Let N and M be normal subgroups of a group G such that

NM 5 G. Then,

G/N ∩ M  G/N 3 G/M.

Q001-Algebra-111001_CH 05.indd 19 9/16/2011 10:06:07 AM

 5-20   Algebra – Abstract and Modern

Proof: Define f : G → G/N 3 G/M by f (a) 5 (aN, aM) for all a  G.

For any a and b  G, we have

f ( ab)  ( abN , abM )

 ( aN  bN , aM  bM )
 ( aN , aM )  (bN , bM )
 f ( a)  f ( b)

and hence f is a homomorphism. We shall prove that f is a surjection also. Let

(xN, yM)  G/N 3 G/M, where x and y  G. Since NM 5 G, x and y  NM
and hence

x 5 rs, for some r  N and s  M

and   y 5 tu, for some t  N and u  M.

Now, put a 5 st. Then,

a21x 5 (st)21 (rs) 5 t21(s21rs)  N (since t, r  N)

and hence

aN 5 xN, Also,
a21y 5 (st)21(tu) 5 (t21s21t) u  M (since s, u  M)

and hence aM 5 yM. Therefore, f (a) 5 (aN, aM) 5 (xN, yM). Thus, f is an
epimorphism. By the Fundamental Theorem of Homomorphisms,

G/ker f  G/N 3 G/M.

Now, ker f  {a ∈ G : f ( a)  the identity in G / N  G / M }
 {a ∈ G : ( aN , aM )  ( N , M )}
 {a ∈ G : aN  N and aM = M }
     N ∩ M.

Thus, G/N ∩ M  G/N 3 G/M.

Worked Exercise 5.2.1. Let m and n be relatively prime positive integers.

Then, prove that Zmn  Zm 3 Zn.

Answer: Consider the group (Z, 1) of integers and let N 5 nZ and M 5 mZ.
Since (n, m) 5 1,

1 5 an 1 bm for some integers a and b

Q001-Algebra-111001_CH 05.indd 20 9/16/2011 10:06:08 AM

 Homomorphisms of Groups    5-21

and hence 1  N 1 M so that N 1 M 5 Z. By Theorem 5.2.4 (where we have

written NM, since the binary operation in G is taken as ?),

Z/N ∩ M  Z/N 3 Z/M.

Now, N ∩ M 5 kZ, where k is the least common multiple of n and m. Since

(n, m) 51, k 5 mn. Therefore,

Zmn  Z/mnZ (by Example 5.2.1)

5 Z/N ∩ M
 Z/N 3 Z/M
5 Z/nZ 3 Z/mZ  Z/N 3 Z/M.

In the following, we shall classify the cyclic groups and prove that (Z, 1)
and (Zn, 1n) are the only (up to isomorphism) cyclic groups.

Theorem 5.2.5. Let G be a cyclic group. If G is infinite, them G  Z. If G is

finite, them G  Zn where n is the order of G.

Proof: Since G is cyclic, we can choose an element a in G such that

G 5 < a > 5 {an : n  Z}.

Consider the group (Z, 1) of integers and define

f : Z → G by f (m) 5 am for any m  Z.

For any m and n  Z,

f (m 1 n) 5 am1n 5 am ? an 5 f (m) ? f (n)

and hence f is a homomorphism. Since any element of G is of the form

am 5 f (m) for some m  Z, f is a surjection. Therefore, f is an epimorphism.
By the Fundamental Theorem of Homomorphisms,

Z/ker f  G.

Since ker f is a subgroup of (Z, 1), ker f 5 nZ for some nonnegative integer
n (by Worked Exercise 4.1.1),

n5 0 ⇔ ker f 5 {0} ⇔ f is a monomorphism.

Thus, by Theorem 4.2.7, G is infinite if and only if f is an isomorphism and

hence Z  G. Also,

Q001-Algebra-111001_CH 05.indd 21 9/16/2011 10:06:08 AM

 5-22   Algebra – Abstract and Modern

G is finite ⇔ Z/ker f is finite

⇔ ker f 5 nZ, n > 0.

In this case, |G| 5 |Z/nZ| 5 |Zn| 5 n.

Thus, if G is finite, then G  Zn, where n is the order of G.

Worked Exercise 5.2.2. For any positive integers n and m, prove that the fol-
lowing are equivalent to each other.
1. n and m are relatively prime.
2. Zn 3 Zm is cyclic.
3. Zn 3 Zm  Znm

Answer: (1) ⇒ (3) is proved in Worked Exercise 5.2.1 and (3) ⇒ (2) is triv-
ial, since Znm is cyclic and, for any group G and G9 such that G  G9, G is
cyclic if and only if G9 is cyclic. We are left with only (2) ⇒ (1). Assume that
Zm 3 Zn is cyclic. Since the order of Zm 3 Zn is mn, it follows from Theorem
5.2.5 that

Zm 3 Zn  Zmn.

There must be an element (a, b) in Zm 3 Zn such that O(a, b) 5 mn, 0  a < m

and 0  b < n. Let k be the l.c.m. of m and n. Then,

k 5 ms and k 5 nt for some s and t  Z1.

Now, k (a, b) 5 (ka, kb)

5 (msa, ntb)
5 (0, 0) since a  Zm and b  Zn.

Therefore, O(a, b) divides k and hence mn divides k.

k 5 mn 5 kd

where d is the g.c.d. of m and n. Therefore, d 5 1; that is, m and n are rela-
tively prime. b

EXERCISE 5(b)

1. If N and M are normal subgroups of a group G, prove that G/N ∩ M is isomor-

phic to a subgroup of G/N 3 G/M.

Q001-Algebra-111001_CH 05.indd 22 9/16/2011 10:06:08 AM

 Homomorphisms of Groups    5-23

2. For any positive integers m and n whose least common multiple is k, prove that
Zk is isomorphic to a subgroup of Zn 3 Zm.

3. Suppose that G is a finite group and Z10 is a homomorphic image of G. Then,

what can be said about the order of G?

4. Prove that Z27 is not a homomorphic image of Z72.

5. For any finite group G, prove that there exists a prime number p such that Zp is
not a homomorphic image of G.

6. Prove that the order of any homomorphic image of a finite group G must be a
divisor of the order of G.

7. In Theorem 5.2.2, we have proved that any homomorphism can be expressed as

a composition of an epimorphism and a monomorphism. Discuss the uniqueness
of this expression.

8. Let f : G → G9 and g : G9 → G0 be homomorphism of groups. If g o f is a mono-

morphism, prove that f is a monomorphism. If g o f is an epimorphism, prove
that g is an epimorphism

9. Express Z9 as a homomorphic image of Z27.

10. Prove that Z9 is isomorphic to a subgroup of Z27.

5.3 ISOMORPHISM THEOREMS

The Fundamental Theorem of Homomorphisms is also called the First
Isomorphism Theorem. In this section, we present two more isomorphism
theorems. If N is a normal subgroup of a group G and if we are required
to prove that the quotient group G/N is isomorphic to another group G9,
then Fundamental Theorem of Homomorphisms provides a technique. We
simply exhibit an epimorphism of G onto G9 whose kernel is the given nor-
mal subgroup N. We shall use this technique in proving the following two
theorems.

Theorem 5.3.1 (Second Isomorphism Theorem). Let M and N be sub-

groups of a group G and N normal in G. Then, M ∩ N is a normal subgroup
of M and

M/M ∩ N  MN/N.

Proof: Since N is given to be a normal subgroup of G, Na 5 aN for all aG

and, in particular Na 5 aN for all a  M so that NM 5 MN. Therefore, MN
is a subgroup of G and N ⊆ MN. Also, the normality of N in G implies the

Q001-Algebra-111001_CH 05.indd 23 9/16/2011 10:06:08 AM

 5-24   Algebra – Abstract and Modern

normality of N in MN and hence the quotient group MN/N is defined. Also,

clearly M ∩ N is a subgroup of M. For any a  M ∩ N and x  M, we have

xax2l  N (since a  N and N is normal in G)

and xax2l  M (since a  M and x  M)

and therefore xax2l  M ∩ N. Thus, M ∩ N is a normal subgroup of M. Now,

define

f : M → MN/N by f (m) 5 mN for all m  M.

Observe that, since M ⊆ MN, mN  MN/N for all m  M and therefore f is

well defined. For any a and b  M,

f (ab) 5 (ab)N 5 aN ? bN 5 f (a)f (b)

and hence f is a homomorphism. If xN  MN/N, then

xN 5 (mn)N 5 m(nN) 5 mN 5 f (m)

for some m  M and n  N. Therefore, f is an epimorphism.

By the Fundamental Theorem of Homomorphisms,

M/ker f  MN/N.

Now, ker f 5 {m  M : f (m) 5 the identity in MN/N}

5 {m  M : mN 5 N}
5 {m  M : m  N} 5 M ∩ N

and therefore M/M ∩ N  MN/N. b

Example 5.3.1. Consider the group (Z, 1) of integers and let M 5 <3> 5 3Z
and N 5 <5> 5 5Z. Since 1 is commutative on Z, M and N are normal
­subgroups of Z. Also,

M ∩ N 5 15Z and M 1 N 5 Z

and hence, by the above theorem, we have

3Z/15Z  Z/5Z.

Recall that Z/5Z  Z5, the group of integers modulo 5.

Q001-Algebra-111001_CH 05.indd 24 9/16/2011 10:06:08 AM

 Homomorphisms of Groups    5-25

A closer examination of the cosets and operation in the quotient group 3Z/15Z
reveals this to be none other than the group ({0, 3, 6, 9, 12}, 115). The above
isomorphism is a disguised version of the isomorphism

({0, 3, 6, 9, 12}, 115)  (Z5, 15).

Theorem 5.3.2. Let f : G → G9 be an epimorphism of groups and N be a

normal subgroup of G such that ker f ⊆ N. Then,

G/N  G9/f (N).

Proof: Since N is a normal subgroup of G and f is an epimorphism, it fol-

lows that f (N) is a normal subgroup of G9 (see Theorem 5.1.3) and hence the
quotient group G9/f (N) is defined. Now, define

g : G → G9/f (N) by g(a) 5 f (a)f (N)

for any a  G. For any x and y  G,

g(xy) 5 f (xy)f (N)

5 f (x)f (y)f (N)
5 f (x)f (N)?f (y)f (N) (since f (N) is normal in G9)
5 g(x)g(y)

and therefore g is a homomorphism. Also, for any z  G9, there exists a  G

such that f (a) 5 z (since f is an epimorphism) and therefore

g(a) 5 f (a)f (N) 5 z f (N).

This implies that g is an epimorphism. Further, for any a  G,

a  ker g ⇔ g(a) 5 the identity in G9/f (N)

⇔ f (a) f (N) 5 f (N)
⇔ f (a)  f (N)
⇔ f (a) 5 f (x) for some x  N
⇔ f (x21a) 5 f (x) 21 f (a) 5 e9, x  N
⇔ x2la  ker f ⊆ N and xN
⇔ a 5 x(x2la)  N

and therefore ker g 5 N. Thus, by the Fundamental Theorem of Homo-

morphism,
G/N 5 G/ker g  G9/f (N).

Q001-Algebra-111001_CH 05.indd 25 9/16/2011 10:06:08 AM

 5-26   Algebra – Abstract and Modern

While applying the above theorem, one frequently starts with a normal
subgroup of G9 and use its inverse images, rather than starting with a normal
subgroup of G containing the kernel. In this context, recall the one-to-one
correspondence between normal subgroups of G containing the kernel and the
normal subgroups of G9. In view of this, the above theorem can be rephrased
as follows.

Corollary 5.3.1. Let f : G → G9 be an epimorphism of groups and M be a

normal subgroup in G9. Then,

G/f2l(M)  G9/M.

Proof: Put N 5 f2l(M). Since M is a normal subgroup of G9, N is a normal

subgroup of G. Also,

ker f 5 f2l({e9}) ⊆ f2l(M) 5 N.

Further, since f is a surjection,

f (N) 5 f (f2l(M)) 5 M.

Thus, by Theorem 5.3.2,

G/f2l(M) 5 G/N  G9/f (N) 5 G9/M.

The following is another special case which is of interest on its own and
is popularly called the Third Isomorphism Theorem. The reader is cautioned
that there seems to be no universally accepted agreement on the numbering
of these three Isomorphism theorems. However, the Fundamental Theorem
of Homomorphisms deserves to be called as the First Isomorphism Theorem,
since the other two are proved using this.

Theorem 5.3.3 (Third Isomorphism Theorem). Let M and N be normal sub-

groups of a group G such that M ⊆ N. The N/M is a normal subgroup of G/M
and (G/M)/(N/M)  G/N.

Proof: By Remarks 4.6.1, N/M is a normal subgroup of G/M. Define f :

G/M → G/N by f (aM) 5 aN for any a  G. First observe that f is well defined;
for, if a and b  G, then

aM 5 bM ⇒ a2lb  M ⊆ N ⇒ a2lb  N ⇒ aN 5 bN.

Q001-Algebra-111001_CH 05.indd 26 9/16/2011 10:06:09 AM

 Homomorphisms of Groups    5-27

Clearly f is an epimorphism and its kernel is given by

ker f 5 {aM  G/M : a  G and f (aM) 5 the identity in G/N}
5 {aM  G/M : a  G and aN 5 N}
5 {aM  G/M : a  N} 5 N/M.
Thus, by the Fundamental Theorem of Homomorphisms,
G/M G/M
 G / N
N / M ker f

Therefore,
(G / M ) G
 .
(N / M ) N 

Worked Exercise 5.3.1. Let M and N be subgroups of a group G such that N

is normal in G and MN 5 G. Then, prove that
G/N  M if and only if M ∩ N 5 {e}.
Answer: By the Second Isomorphism Theorem 5.3.1,
G/N 5 MN/N  M/M ∩ N.
Therefore, G/N  M ⇔ M/M ∩ N  M
⇔ M ∩ N 5 {e}.

EXERCISE 5(c)

1. Let G, G1 and G2 be groups and f1 : G → G1 and f2 : G → G2 be epimorphisms

such that ker f1 ⊆ ker f2. Then, prove that there exists a unique homomorphism
f : G1 → G2 such that
f o f1 5 f2

f1

G G1

f2 f

G2

Q001-Algebra-111001_CH 05.indd 27 9/16/2011 10:06:09 AM

 5-28   Algebra – Abstract and Modern

2. From the above exercise, deduce the Factorization Theorem 5.2.2.

3. Let G be the group of nonzero real numbers under the usual multiplication and N
5 {1, 21}. Then, prove that N is a normal subgroup of G and the quotient group
G/N is isomorphic to the group of positive real numbers under multiplication.

4. Let f : G → G9 be an epimorphism of groups. Then, prove that A → f 2l(A) is a

one-to-one correspondence between the (normal) subgroups of G9 and the
(normal) subgroups of G containing ker f.

5. Let G8 be the group of symmetries of a square (see Example 3.2.8) and define
f : G8 → Z2 3Z2 by
f (e) 5 f (r2) 5 (0, 0), f (r1) 5 f (r2) 5 (1, 0),
f (h) 5 f (0) 5 f (0, 1) and f (d1) 5 f (d2) 5 (1, 1).
Prove that f is an epimorphism and deduce that
G8/Z(G8)  Z2 3 Z2, where Z(G8) is the centre of G8.

6. Find all (up to isomorphism) homomorphic images of the group G8 of symme-

tries of a square and exhibit the correspondence between the subgroups of G8
containing Z(G8) and the subgroups of G8/Z(G8).

7. For any (a, b)  R 3 R with a  0, define Tab : R→R by Tab (x) 5 ax 1 b for all
x  R and let

G  {Tab : ( a, b) ∈ R  R and a  0}
and        N 5 {T1b: b  R}.
Prove that G is a group under the composition of mappings and N is a normal
subgroup of G. Further, prove that the quotient group G/N is isomorphic to the
group of nonzero real numbers under the multiplication.

8. Prove the following for any epimorphism f : G → G9 of groups:

(i) For any subgroups A and B with ker f ⊆ A ∩ B, f (A ∩ B) 5 f (A) ∩ f (B).
(ii) For any subgroup A of G, A ker f 5 f21 (f (A))
(iii) If [G, G] is the commutator subgroup of G and ker f ⊆ [G, G], then
G /[G , G ]  G /[G, G]

9. Let f : G → Z8 be an epimorphism of a group G onto the group Z8 of integers

modulo 8. Prove that G has normal subgroups of index 2 and 4.

10. Let f : G → G be an epimorphism of groups and A and A9 be subgroups of G

and G, respectively. Then, prove the following:
(i) If A is of finite index in G and ker f ⊆ A, then

iG ( A)  i  ( f ( A))
G

Q001-Algebra-111001_CH 05.indd 28 9/16/2011 10:06:10 AM

 Homomorphisms of Groups    5-29

(ii) If A is of finite index in G, then

iG ( f 1 ( A)) = i  ( A).
G

11. Prove that any group of order 4 is isomorphic to either Z4 or Z2 3 Z2 and hence
is abelian.

12. Prove that any group of prime order p is isomorphic to Zp and hence is cyclic.

13. Let f : G → A and g : G → B be homomorphisms of groups such that g is an epi-

morphism and ker g ⊆ ker f. Then, prove that there is a unique homomorphism
f : B → A such that f ο g  f

f
G A

g f

14. Prove the following in the above Exercise 13.

(i) f is an epimorphism if and only if f is so.
(ii) f is a monomorphism if and only if ker f 5 ker g.
15. Let A, N1 and N2 be subgroups of a group G such that N1 and N2 are normal in
G and A ∩ N1 5 A ∩ N2. Then, prove that AN1/N2  AN2/N1.

16. Let A1, A2 and N be subgroups of a group G such that N is normal in G and A1N
5 A2N. Then, prove that A1 / A1 ∩ N  A2 / A2 ∩ N .

5.4 AUTOMORPHISMS
Let us recall that a homomorphism of a group G into itself is called an
endomorphism of G and a bijective endomorphism of G is called an auto-
morphism of G. Among the endomorphisms of a group G, the automor-
phisms of G need special attention, for the reason that they form a group
on their own under the composition of mappings and that the structure of
this group reveals that of the group G itself. Even though the following is a
repetition, we prefer to give an independent status for convenience and for
its importance.

Q001-Algebra-111001_CH 05.indd 29 9/16/2011 10:06:11 AM

 5-30   Algebra – Abstract and Modern

Definition 5.4.1. For any group G, a bijective homomorphism of G onto

itself is called an automorphism of G. The set of all automorphisms of G will
be denoted by Aut(G).

Theorem 5.4.1. For any group G, the set Aut(G) of all automorphisms of G
forms a group under the composition of mappings.

Proof: Let G be a group, since the identity mapping IG : G → G, defined by

IG(x) 5 x for all x  G, is an automorphism of G, we have IG  Aut(G) and
hence Aut(G) is a nonempty set. If f and g are automorphisms of G, then the
composition f o g is also an automorphism of G. Therefore, o is a binary
operation on Aut(G), which is clearly associative. Also, for any f  Aut(G),
we have

f ο IG  f  IG ο f .

Therefore, the identity map IG is the identity element in the semigroup

(Aut(G), o). Further, we know that (see Theorem 5.1.7), if f is an automor-
phism of G, then the inverse mapping f21 also an automorphism of G and

f o f21 5 IG 5 f21 o f.

Thus, Aut(G) is a group under the composition of mappings.

Example 5.4.1
1. If G is an abelian group, then the map f : G → G, defined by f (x) 5 x21
for any x  G, is an automorphism of G.
2. Consider the group (Z12, 112) of integers modulo 12 and define f : Z12 →
Z12 by
f ( a)  5a ( a 12 a 12 a 12 a 12 a).

Note that f (a) 5 r, where 5a 5 12q 1 r, where q and r are integers and
0  r < 12.
Then, f is clearly an endomorphism of Z12
Also,

ker f  {a ∈ Z12 ; f ( a)  0}
 {a ∈ Z12 ; 5a  12q, q ∈ Z}
 {0} (since (5, 12)  1)

Q001-Algebra-111001_CH 05.indd 30 9/16/2011 10:06:11 AM

 Homomorphisms of Groups    5-31

and therefore f is a monomorphism of Z12 into Z18 .Since Z12 is a finite set, it
follows that f is an surjection also. Thus, f is an automorphism of Z12.
In the following, we discuss about a special subgroup of Aut(G) consisting
of certain special class of automorphisms which are important in the case of
an abelian group.

Theorem 5.4.2. Let G be a group and a  G. Define Ta : G → G by

Ta(x) 5 a x a21 for all x  G.

Then, Ta is an automorphism of G and hence Ta  Aut(G).

Proof: For any x and y  G,

Ta(xy) 5 a(xy)a21 5 (axa21)(aya21) 5 T(x)T(y)

and hence Ta is a homomorphism. Also, for any y  G, we have

a21ya  G and Ta (a21ya) 5 a(a21ya)a21 5 y

and therefore Ta is a surjection. From the cancellation laws, we have

Ta ( x )  Ta ( y ) ⇒ axa1  aya1 ⇒ x  y

and hence Ta is an injection also. Thus, Ta is an automorphism of G.

Definition 5.4.2. For any element a in a group G, the automorphism Ta

defined above is called the inner automorphism of G corresponding to a. The
set of all inner automorphisms of G will be denoted by I(G); that is,

I(G) 5 {Ta : a  G}.

Theorem 5.4.3. Let G be any group and define a : G → Aut(G) by

a(a) 5 Ta for all a  G.

Then, a is a homomorphism and I(G) is a subgroup of Aut(G). Also,

G / Z (G )  I (G ), where Z(G) is the centre of G, defined by

Z(G) 5 {a G : ax 5 xa for all x  G}.

Q001-Algebra-111001_CH 05.indd 31 9/16/2011 10:06:11 AM

 5-32   Algebra – Abstract and Modern

Proof: For any a and b  G, we have

Tab ( x )  ( ab) x( ab)1  a(bxb1 )a1  Ta (Tb ( x ))  (Ta ο Tb )( x )

for any x  G and hence Tab 5 Ta o Tb. Therefore,

a(ab)  Tab  Ta ο Tb  a(a) ο a(b)

for all a and b  G. Therefore, a is a homomorphism of G into Aut(G) and

hence a(G) 5 I(G) is a subgroup of G. Now, let us compute the kernel of a.

a  ker a ⇔ a(a) 5 the identity in Aut(G)

⇔ Ta 5 IG
⇔ Ta(x) 5 IG (x) for all x  G
⇔ a x a21 5 x for all x  G
⇔ ax 5 xa for all x  G
⇔ a  Z(G), the centre of G.

Thus, by the Fundamental Theorem of Homomorphisms,

G / Z (G )  G / ker a  a(G )  I (G ). 

It can be easily seen that a group G is abelian if and only if Ta(x) 5 x for
all a and x  G (that is, Ta 5 IG for all a  G and the group I(G) is trial). If G
is a nonabelian group, then there exists an automorphism Ta  IG. In the fol-
lowing, we prove that a group G has a nonidentity automorphism if and only
if G has atleast 3 elements.

Theorem 5.4.4. Let G be a group. Then,

| Aut(G ) | 1 if and only if | G | 2.

Proof: Suppose that |Aut(G)| > 1. Then, the group Aut(G) has an element
other than its identity. That is, there exists an automorphism f of G such that f
 IG, the identity map. Now, we can choose an element a  G such that f (a)
 IG(a) 5 a. Since f (e) 5 e, it follows that a  e and, since f is injective,
f (a)  f (e) 5 e. Therefore, e, a and f (a) are three distinct elements in G and
hence |G| > 3 > 2.

Q001-Algebra-111001_CH 05.indd 32 9/16/2011 10:06:12 AM

 Homomorphisms of Groups    5-33

Conversely suppose that |G| > 2. If G is not abelian, then ax  xa for some a
and x  G and hence

Ta(x) 5 axa21  x

so that Ta  IG  Aut(G), where Ta is the inner automorphism of G corre-

sponding to a and hence |Aut(G)| > 1. Therefore, we can assume that G is
abelian. Then, the map g : G → G, defined by g(x) 5 x21 for all x  G, is an
automorphism of G. If g  IG, then |Aut(G)| > 1.Therefore, we can assume
that g 5 IG; that is, g(x) 5 x or x21 5 x or x2 5 e for all x  G. since |G| > 2, G
has atleast three distinct elements; that is, G has atleast two distinct elements
other than the identity. Let a and b  G such that a  e, b  e and a  b. Put
A 5 {e, a, b, ab}. Then,

(ab)a 5 ba2 5 be 5 b and (ab)b 5 ab2 5 ae 5 a

and hence A is a subgroup of G. Define f : A → A by

f (e) 5 e, f (a) 5 b, f (b) 5 a and f (ab) 5 ab.

Then, it can be easily verified that f is an automorphism of A. We shall extend

f to an automorphism of the whole of G. Consider the set

 5 {(B, g) : B is a subgroup of G, A ⊆ B, g  Aut(B) and g/A 5 f}.

Note that (A, f )   and hence  is a nonempty set. For any (B, g) and (C,
h)  G, define

(B, g)  (C, h) if and only if B ⊆ C and h/B 5 g.

Then,  is a partial order on . If {(Bi, gi)}iI is a chain in the partially

ordered set (, ) and B 5 ∪i∈I Bi, then B is a subgroup of G and, if we define
g : B → B such that g/Bi 5 gi, then (B, g) is a member in  and is an upper
bound of {(Bi, gi)}iI in . Thus, (, ) satisfies the hypothesis of the Zorn’s
lemma which guarantees the existence of a maximal member, say (M, a), in
(, ). Since A ⊆ M and a/A 5 f, we have a(a) 5 f (a) 5 b  a and hence a
 IdM and therefore it is enough it we prove that M 5 G.
Otherwise, suppose that there is an element s  G such that s  M. Put
B 5 M ∪ Ms and define g : B → B by g(m) 5 a(m) and g(ms) 5 a(m)s for
all m  M. Then, B is a subgroup of G (note that B 5 MS, where S is the
subgroup {e, s}), g  Aut(B) and g/M5 a so that (M, a) < (B, g), which is

Q001-Algebra-111001_CH 05.indd 33 9/16/2011 10:06:12 AM

 5-34   Algebra – Abstract and Modern

a ­contradiction to the maximality of (M, a) in . Thus, M 5 G, a  Aut()

and a  IG so that |Aut()| > 1.
In the following, we completely determine all the automorphisms of a
cyclic group. In fact, each automorphism of a cyclic group corresponds to a
generator of the group and vice versa.

Theorem 5.4.5. Let f be an automorphism of a cyclic group G.

1. For any a  G, G 5 <a> if and only if G 5 <f (a)>.
2. If f and g  Aut(G) and a is a generator of G such that f (a) 5 g(a),
then f 5 g.

Proof: First recall that f21 is also an automorphism of G.

1. Suppose that a  G such that
G 5 <a> 5 {an : n  Z}.
Then, since f (a)  G, we have < f (a)> ⊆ G. On the other hand,
x  G ⇒ f21(x)  G 5<a>
⇒ f21(x) 5 an for some n  Z
⇒ x 5 f (f21(x)) 5 f (an) 5 f (a)n  <f (a)>
and hence G 5 <f (a)>. The converse follows from the fact that f21 is an
automorphism of G.
2. Let f and g  Aut(G) and G 5 <a> such that f (a) 5 g(a). Then, for any
x  G, x 5 an for some n  Z and
f (x) 5 f (an) 5 f (a)n 5 g(a)n 5 g(a)n 5 g(x)
Therefore, f 5 g.

Corollary 5.4.1. For any cyclic group G, the number of automorphisms of G

is finite and is precisely equal to the number of generators of G.

Proof: Let G 5 <a> be a cyclic group. Then, by the above theorem f  f (a)
is an injection of Aut(G) into the set gen(G) of generators of G. Further, if b
is any generator of G, then we can define automorphism f such that f (a) 5 b
(that is, f (an) 5 bn for any n  Z). Thus, f  f (a) is a bijection of Aut(G) onto
gen(G). From Theorems 4.2.7 and 4.2.8, gen(G) is a finite set and hence so is
Aut(G) and |Aut(G)| 5 |gen(G)|.

Corollary 5.4.2
1. For any infinite cyclic group G, |Aut(G)| 5 2.
2. For any finite cyclic group G of order n, |Aut(G)| 5 f (n), where f is the
Euler-Totient function.

Q001-Algebra-111001_CH 05.indd 34 9/16/2011 10:06:12 AM

 Homomorphisms of Groups    5-35

Proof: These follow from Corollary 5.4.1, Theorems 4.2.7 and 4.2.8.

Worked Exercise 5.4.1. For any groups G and H, if G  H, prove that

Aut(G)  Aut(H).

Answer: Suppose that G  H and a; G → H is an isomorphism. Define

 : Aut (G) → Aut(H) by (f ) 5 a o f o a21

for any f  Aut(G), since f, a and a21 are isomorphisms, so is

a o f o a21 : H → H. Therefore, a o f o a21 is an automorphism of H. For
any f and g  Aut(G),

(f o g) 5 a o (f o g) o a21
5 (a o f o a21) o (a o g o a21)
5 (f ) o (g).

Therefore,  is homomorphism. Further,

(f ) 5 (g) ⇒ a o f o a21 5 a o g o a21

⇒ f 5 (a21 o a) o f o (a21 o a) 5 a21 o (a o f o a21) o a
5 a21 o (a o g o a21) o a 5 (a21 o a) o g o (a21 o a)
5g

Therefore,  is an injection. Also,

h ∈ Aut( H ) ⇒ a1 ο h ο a ∈ Aut(G ) and (a21 o h o a) =

a o (a21 o h o a) o a21 = h.

Therefore,  is a surjection. This  is a bijective homomorphism and hence an

isomorphism of Aut(G) onto Aut(H).

Worked Exercise 5.4.2. List all the automorphisms of the group (Zn, 1n) for
any positive integer n and, in particular, of the group (Z12, 112).

Answer: Recall that, for any 1  r < n, r is a generator of Zn if and only if r is

relatively prime to n. Therefore, by Corollary 5.4.2 (2), these are exactly f (n)
automorphisms of Z12 and these are given by

f r : Z12 → Z12 , f r ( m)  mr (mod 12)

for each 1  r < n, such that (r, n) 5 1.

Q001-Algebra-111001_CH 05.indd 35 9/16/2011 10:06:13 AM

 5-36   Algebra – Abstract and Modern

Consider Z12. We have f (n) 5 4 since 1, 5, 7 and 11 are the only integers
r such that 1  r < 12 and (r, 12) 5 1. Therefore, there are exactly four auto-
morphisms of Z12 and are given below.
f1 5 I, the identity map of Z12,
f5 : Z12 → Z12 defined by f5(m) 5 5m (mod 12),
f7 : Z12 → Z12 defined by f7(m) 5 7m (mod 12),
f11 : Z12 → Z12 defined by f11(m) 5 11m (mod 12).
The following table gives a complete description of all the four automor-
phisms of Z12.

0 1 2 3 4 5 6 7 8 9 10 11

f1 0 1 2 3 4 5 6 7 8 9 10 11

f5 0 5 10 3 8 1 6 11 4 9 2 7

f7 0 7 2 9 4 11 6 1 8 3 10 5

f11 0 11 10 9 8 7 6 5 4 3 2 1

EXERCISE 5(d)

1. For any endomorphism f of a finite group G, prove that the following are equiva-
lent to each other:
(i) f is an epimorphism.
(ii) f is an automorphism of G.
(iii) f is a monomorphism.
2. Give an example of an infinite group G and of an isomorphism of G onto a
proper subgroup of G.
3. Let G 5{e, a, b, ab} be a group of order 4 in which a2 5 e 5 b2 and ab 5 ba.
Then, determine Aut(G).
4. Let A be a subgroup of a group G, such that f (A) ⊆ A for all f  Aut(G). Then,
prove that A is a normal subgroup in G.
5. Let G be a group and f  Aut(G). Prove that the set {a G : f (a) 5 a} is a sub-
group of G.
6. For any group G, prove that {a  G : f (a) 5 a for all f  Aut(G)} is a normal
subgroup of G.
7. Let G be a finite group and f  Aut(G), such that, for any xG,
f (x) 5 x if and only if x 5 e.

Q001-Algebra-111001_CH 05.indd 36 9/16/2011 10:06:13 AM

 Homomorphisms of Groups    5-37

Then, prove that any element x of G can be expressed as

x 5 y21f (y) for some y  G.
8. If f 2 5 IG for the f in Exercise 7 above, then prove that G is an abelian group.

9. Let G be a finite group and f  Aut(G) such that f sends more than three-quarters
of the elements of G onto their inverses. Then, prove that f (a) 5 a21 for all a 
G and that G is abelian.

10. Let C be the commutator subgroup of a group G and f  Aut(G). Prove that
f (C) ⊆ C.

11. If a group G has a nonidentity automorphism, then prove that G has atleast three
elements.

12. Prove that any homomorphic image of a cyclic group is cyclic.

13. Prove that any homomorphic image of an abelian group is abelian.

14. Determine Aut(Zp) for any prime number p.

15. Find all the automorphisms of (Z, 1) and (Zn, 1n) for any n  Z1.

16. Let G be a finite cyclic group of order n and define fm : G → G by fm(a) 5 am for
all a  G and m  Z1. Prove that fm is an automorphism of G if and only if m is
relatively prime with n.

17. Let G be a finite group of order n > 2. If a2  e for some a  G, then prove that
G has a nonidentity automorphism.

18. If G is a noncyclic finite abelian group, prove that Aut(G) is not abelian.

19. Let G be a finite group, such that |Aut(G)| 5 p, where p is a prime number. Then,
prove that |G|  3.

20. Prove that Aut(Z2 3 Z3)  Aut(Z2) 3 Aut(Z3).

Q001-Algebra-111001_CH 05.indd 37 9/16/2011 10:06:13 AM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 05.indd 38 9/16/2011 10:06:13 AM

 6
Permutation Groups
6.1 Cayley’s Theorem
6.2 The Symmetric Group Sn
6.3 Cycles
6.4 Alternating Group An and Dihedral Group Dn

For any nonempty set X, the set M(X) of all mappings of X into itself forms a
monoid under the composition of mappings in which the invertible elements
(elements possessing inverses) are precisely the bijections of X onto itself.
In fact, we have observed that a mapping f : X → X has left (right) inverse in
M(X) if and only if f is an injection (respectively, surjection) and, as such, the
set of all bijections of X onto itself forms a group under the composition of
mappings. Before the advent of the abstract form of a group, mathematicians
were only interested in the group structure of certain sets of bijections, which
were also known as permutations, when the set X is finite. In this chapter, we
discuss thoroughly the structure of this type of groups.

6.1 CAYLEY’S THEOREM

Before the formation of the present day abstract concept of a group, most of
the groups were in the form of a set of transformations of a particular math-
ematical structure, like the group of symmetries of square or of an equilateral
triangle. Most finite groups appeared as groups of bijections of an n-element
set onto itself for some positive integer n. The English Mathematician Cayley
first noted that any abstract group can be viewed as a subgroup of the group
S(X) of bijections of X onto itself, for a suitable set X. In this section, we shall
prove this theorem of Cayley and some of its consequences. First, we have
the following definition.

Q001-Algebra-111001_CH 06.indd 1 9/16/2011 10:05:15 AM

 6-2   Algebra – Abstract and Modern

Definition 6.1.1. Let X be a nonempty set. Any bijection of X onto itself

is called a permutation on X. The set S(X) of all permutations on X forms
a group under the composition of mappings. Any subgroup of X is called a
group of permutations on X.

Theorem 6.1.1. Any group is isomorphic to a group of permutations on a

suitable set.

Proof: Let G be a group. For any a  G, define

fa : G → G by fa(x) 5 ax for all x  G.

Then, for any y  G, a21y  G and fa(a21y) 5 a(a21y) 5 y and therefore fa is a

surjection. Also, by the left cancellation law in the group G, fa is an injection.
Therefore, fa is a permutation on the set G. Now, define

G9 5 {fa : a  G}.

We shall prove that G9 is a group of permutations on the set G and that

G  G9. For any a and b  G, we have

fab(x) 5 (ab)x 5 a(bx) 5 fa(fb(x)) 5 (fa ? fb)(x)

for all x  G and hence fab 5 fa ? fb. Also, if e is the identity in the group G,

fe(x) 5 ex 5 x for all x  G

and hence fe 5 IG, the identity map on G. In particular, for any a  G,

f a  f a−1  f aa−1  f e  I G  f e  f a−1a  f a−1  f a

and hence fa21 5 f a−1  G. Therefore, G9 is a subgroup of the group (S(G), o);
that is, G9 is a group of permutations on the set G. Now, define

 : G → G9 by (a) 5 fa for all a  G.

Then, (ab) 5 fab 5 fa ? fb and hence  is a homomorphism. For any a and

b  G,

fa 5 fb ⇒ fa(e) 5 fb(e) ⇒ a 5 ae 5 be 5 b

and therefore  is an injection. Clearly  is a surjection. Thus,  is an isomor-

phism of G onto G9 and therefore G  G9. b

Q001-Algebra-111001_CH 06.indd 2 9/16/2011 10:05:16 AM

 Permutation Groups   6-3

Cayley’s theorem enables us to view any abstract group as a more con-

crete object, namely as a group of mappings and the binary operation on G
as the composition of mappings. However, this has its own disadvantages.
For example, if G is a finite group of order n, then S(G) is a group of order
n! which is far bigger than n. Now, a natural question is that can we find a
smaller set X such that G is isomorphic to a subgroup of S(X). The following
theorem is a step ahead in this direction.

Theorem 6.1.2. Let G be a group and H be a subgroup of G. Let X be the set

of all left cosets of H in G. Then, there exists a homomorphism  : G → S(X)
satisfying the following:
1. ker  is the largest normal subgroup of G contained in H.
2.  is a monomorphism if and only if H contains no nontrivial normal
subgroup of G.

Proof: We have X 5 {aH : a  G}. For any x  G, define

gx : X → X by gx(aH) 5 (xa)H

for all aH  X, a  G. Note that, for any a and b  G,

aH 5 bH ⇒ a21b  H
⇒ (xa)21(xb) 5 a21x21xb 5 a21b  H
⇒ (xa)H 5 (xb)H

and hence gx is well-defined. Also, for any a and b  G,

gx(aH) 5 gx(bH) ⇒ (xa)H 5 (xb)H

⇒ (xa)21(xb)  H
⇒ a21x21xb  H
⇒ a21b  H ⇒ aH 5 bH

and therefore gx is an injection. Further, if bH  X, then (x21b)H  X and

gx((x21b)H) 5 (x(x21b))H 5 bH

and hence gx is a surjection also. Thus, gx is a permutation on X; that is, gx 

S(X) for any x  G. Now, define

 : G → S(X) by (x) 5 gx for all x  G.

Q001-Algebra-111001_CH 06.indd 3 9/16/2011 10:05:16 AM

 6-4   Algebra – Abstract and Modern

For any x and y  G and aH  X (a  G), we have

(xy)(aH) 5 gxy(aH) 5 ((xy)a)H 5 gx(gy(aH)) 5 ((x) ? (y))(aH)

and hence (xy) 5 (x) ? (y). Therefore,  is a homomorphism of G into S(X).

1. Clearly ker  is a normal subgroup of G and
ker  5 {x  G : (x) 5 the identity in S(X)}
5 {x  G : gx 5 IX}
5 {x  G : gx(aH) 5 aH for all a  G}
5 {x  G : xaH 5 aH for all a  G}
5 {x  G : a21xa  H for all a  G}
5 {x  G : x  aHa21 for all a  G}
 ∩ aHa1
a ∈G

Therefore, ker  ⊆ aHa21 for all a  G and, in particular,

ker  ⊆ eHe21 5 H.
If N is any normal subgroup of G contained in H, then
a21Na ⊆ N ⊆ H
and hence N ⊆ aHa21 for all a  G, so that
N⊆ ∩ aHa
a ∈G
1
5 ker .

Thus, ker  is the largest normal subgroup of G contained in H.

2.  is a monomorphism ⇔ ker  5 {e}
⇔ {e} is the largest normal subgroup of G
contained in H.
⇔ H contains no nontrivial normal subgroup
of G. b
Note that the Cayley’s Theorem 6.1.1 can be deduced from the above theo-
rem by taking H 5 {e}. The above theorem is an important tool in determin-
ing the existence of normal subgroups of a group G contained in a given
subgroup of G.

Theorem 6.1.3. Let H be a subgroup of a finite group G such that |G| is not a
divisor of iG(H)! Then, H contains a nontrivial normal subgroup of G.

Proof: Let X be the set of all left cosets of H in G. Then, |X| 5 iG(H). By the
above theorem, there exists a homomorphism  : G → S(X) such that ker  is

Q001-Algebra-111001_CH 06.indd 4 9/16/2011 10:05:16 AM

 Permutation Groups   6-5

the largest normal subgroup of G contained in H. Then, (G) is a subgroup of

S(X) and hence, by the Lagrange’s theorem, |(G)| is a divisor of |S(X)| 5 |X|!
If ker  5 {e}, then  is a monomorphism and

|G| 5 |(G)|, which is a divisor of |S(X)| 5 iG(H)!,

a contradiction to the hypothesis. Therefore, ker  is a nontrivial normal sub-

group of G contained in H. b

Corollary 6.1.1. Let H be a normal subgroup of a finite group G such that

iG(H)! , |G|. Then, H contains a nontrivial normal subgroup of G.

Worked Exercise 6.1.1. Let A be a subgroup of order 9 in a group of order 36.

Prove that A contains a normal subgroup of G whose order is 3 or 9.
| G | 36
Answer: iG ( A)!  !  !  4!  36  | G |
| A| 9
and hence, by Corollary 6.1.1, A contains a nontrivial normal subgroup N
of G. Since N is a subgroup of A, we get by the Lagrange’s theorem that |N|
divides |A| 5 9. Also, since N is nontrivial, |N| 5 3 or 9.

Worked Exercise 6.1.2. Let G be a group of order 187. Prove that any sub-
group of order 17 in G must be normal.

Answer: Let A be a subgroup of order 17 in G. Then,

| G | 187
iG ( A)    11.
17 17

Since 17 is a prime and 17 . 11, we get that 17 does not divide 11! 5 iG(A)!
and hence |G| does not divide iG(A)! By Theorem 6.1.3, A contains a non-
trivial normal subgroup of G. Let N be a nontrivial normal subgroup of G
contained in A. Then, 1 , |N| and, by Lagrange’s theorem, |N| is a divisor of
|A| 5 17. Since 17 is a prime, |N| 5 17 5 |A| and hence N 5 A. Thus, A is a
normal subgroup of G.

Worked Exercise 6.1.3. Let G be a finite group of order n and p be a prime

number such that p . np . Then prove that any subgroup of order p in G is
normal in G.

Answer: Let A be a subgroup of order p in G. Then, p divides n and iG ( A)  np .

Since p . np and p is a prime, it follows that p does not divide iG(A)!

Q001-Algebra-111001_CH 06.indd 5 9/16/2011 10:05:17 AM

 6-6   Algebra – Abstract and Modern

By Theorem 6.1.3, A contains a nontrivial normal subgroup N of G. By the

Lagrange’s theorem, |N| divides |A| 5 p. Since |N| . 1 and p is a prime, |N| 5
p 5 |A| and, since N ⊆ A, N 5 A. Thus, A is a normal subgroup of G.

EXERCISE 6(a)

1. Which of the following mappings f : R → R are permutations on R?

(i) f (x) 5 x3 2 2
(ii) f (x) 5 x2 2 2
(iii) f (x) 5 3x 1 2
(iv) f (x) 5 2x 2 3
(v) f (x) 5 x3 1 6x2 1 12x 1 8
(vi) f (x) 5 |x| 2 2
(vii) f (x) 5 sin x
(viii)f (x) 5 log |x|
 e x if x  0
(ix) f ( x )   − x
e if x  0
2. State whether the following are true or false and substantiate your answers:
(i) Every surjection of Zn onto itself is a permutation, for any n  Z+.
(ii) Every injection of Z into Z is a permutation.
(iii) For any n  Z+, every injection of Zn into Zn is a permutation.
(iv) For any finite set X, every surjection of S(X) into S(X) is a permutation.
(v) For any finite set X, every injection of P(X) into itself is a permutation,
where P(X) is the power set of X.
(vi) Any group G is isomorphic with a subgroup of (S(G), o).

3. If X is a finite set with |X| 5 n, prove that S(X) is a finite group of order n!

4. For any set X, prove that X is finite if and only if S(X) is finite.

5. Let G be a group and define ga : G → G by ga(x) 5 xa for any a and x  G. Then

prove that ga is a permutation on G. Can we replace fa with ga in the proof of the
Cayley’s theorem?

6. Let A be a subgroup of a finite group G and iG(A) 5 m. If A does not contain any
nontrivial normal subgroup of G, then prove that |G| divides m!

7. Let G be a group of order 396. Prove that any group of order 11 in G is normal
in G.

8. Let G be a finite group of order n and p be a prime number such that p2 does not
divide n. Prove that any subgroup of order p in G is normal in G.

Q001-Algebra-111001_CH 06.indd 6 9/16/2011 10:05:17 AM

 Permutation Groups   6-7

9. For any elements a and b in a set X, prove that there is a permutation f on X such
that f (a) 5 b and f (b) 5 a.

10. Construct a permutation f on R such that f (n) 5 n 1 1 for all integers n.

6.2 THE SYMMETRIC GROUP Sn

Recall that a bijection of a set X onto itself is also called a permutation on X
and the permutations on X form a group under the composition of mappings
which is denoted by S(X). It is well-known that a set X is said to be equipotent
or bijective with another set Y if there is a bijection of X onto Y and that, in
this case, we write X . Y. The following can be easily proved by using induc-
tion on the number of elements of X.
Theorem 6.2.1. If X is a finite set and |X| 5 n, then |S(X)| 5 n!
Theorem 6.2.2. For any nonempty finite sets X and Y,
X . Y if and only if S(X)  S(Y).
Proof: Suppose that X . Y. Then, there exists a bijection a : X → Y. Now, define
 : S(X) → S(Y) by (f ) 5 a o f o a21
−1
f
Y →
a
X  → X 
a
→Y

for all f  S(X). For any f and g  S(X),

(f o g) 5 a o (f o g) o a21
5 (a o f o a21) o (a o g o a21) 5 (f ) o (g)
and therefore  is a homomorphism of groups. Also, for any f and g  S(X),
(f ) 5 (g) ⇒ a o f o a21 5 a o g o a21
⇒ a21 o (a o f o a21) o a 5 a21 o (a o g o a21) o a
⇒ f 5 g.
Therefore,  is an injection. Further, for any h  S(Y), a21 o h o a  S(X) and
(a21 o h o a) 5 a o (a21 o h o a) o a21 5 h.
Therefore,  is a surjection also. Thus,  is an isomorphism of S(X) onto S(Y)
and hence S(X)  S(Y). Conversely suppose that S(X)  S(Y). Then, S(X) .
S(Y) and hence
2n 5 |S(X)| 5 |S(Y)| 5 2m, where |X| 5 n and |Y| 5 m.
This implies that n 5 m and hence |X| 5 |Y|, so that X . Y. b

Q001-Algebra-111001_CH 06.indd 7 9/16/2011 10:05:18 AM

 6-8   Algebra – Abstract and Modern

The above theorem says that two finite sets X and Y are equipotent if and only
if the groups (S(X), o) and (S(Y), o) are isomorphic. In particular, if X is a set
with n elements, then S(X)  S(In), where In 5 {1, 2, …, n}. For this reason,
it is enough if we study the permutations on the set {1, 2, …, n}. We begin
this with the following formal definition.

Definition 6.2.1. For only positive integer n, the set In is defined by

In 5 {1, 2, …, n}

and the group (S(In), o) of permutations on In is denoted by Sn and is called

the symmetric group of degree n. Any permutation f on In is usually denoted
by an array

 1 2 3 … n 
f   
 f (1) f (2) f (3) … f ( n)

which symbolizes that each 1 # i # n is mapped onto f (i), the integer that is
written just below i in the array. As usual, let e denote the identity in the group
Sn. Note that e is the identity mapping on In.

Example 6.2.1
1. Consider I6 5 {1, 2, 3, 4, 5, 6} and define f : I6 → I6 by f (1) 5 3, f (2) 5 5,
f (3) 5 1, f (4) 5 2, f (5) 5 4 and f (6) 5 6. Then, f is denoted by
1 2 3 4 5 6
f   .
3 5 1 2 4 6

2. Let f  S9 be given by
1 2 3 4 5 6 7 8 9
f   .
4 6 8 5 2 3 7 9 1 

Then, f is a permutation on {1, 2, 3, 4, 5, 6, 7, 8, 9} defined by f (1) 5 4,

f (2) 5 6, f (3) 5 8, f (4) 5 5, f (5) 5 2, f (6) 5 3, f (7) 5 7, f (8) 5 9 and
f (9) 5 1.
Recall that the order of the symmetric group of degree n is n!

Theorem 6.2.3. Let n and m be positive integers. Sm is isomorphic to a sub-

group of Sn if and only if m # n.

Q001-Algebra-111001_CH 06.indd 8 9/16/2011 10:05:18 AM

 Permutation Groups   6-9

Proof: Suppose that m # n. For any f  Sm, define (f )  Sn as given

below.

 f (i ), if 1  i  m
( f )(i )   .
i, if m  i  n

Then, it can be easily checked that  is a monomorphism of Sm into Sn and

hence Sm  (Sm), which is a subgroup of Sn. Conversely suppose that Sm is
isomorphic to a subgroup of Sn. Then, |Sm| divides |Sn| so that m! divides n!
which happens only when m # n. b
In view of the above theorem, a permutation f in Sn can be identified with
a permutation in Sm for any m $ n with the understanding that f (i) 5 i for all
n , i # m.

Worked Exercise 6.2.1. Construct a table representing the symmetry group

of degree 3.

Answer: There are 3! (56) elements in the group S3, which are given below
(recall Example 3.4.3).

1 2 3 1 2 3 1 2 3

e    , a    , b   ,
1 2 3 2 3 1  3 1 2

1 2 3 1 2 3 1 2 3

c    , d    and s   .
1 3 2 3 2 1  2 1 3

The table representing S3 is given below.

o e a b c d s
e e a b c d s
a a b e s c d
b b e a d s c
c c d s e a b
d d s c b e a
s s c d a b e

Q001-Algebra-111001_CH 06.indd 9 9/16/2011 10:05:19 AM

 6-10   Algebra – Abstract and Modern

EXERCISE 6(b)

1. Consider the following elements in S8 and compute the expressions in (i) to (viii)
given below.
1 2 3 4 5 6 7 8 
a   
3 7 4 6 8 5 1 2

1 2 3 4 5 6 7 8
b   
4 5 6 7 8 1 2 3

and

1 2 3 4 5 6 7 8 
c   
8 7 6 5 1 2 3 4

(i) a2b
(ii) ab2
(iii) abc
(iv) ab2c
(v) a2bc
(vi) abc2
(vii) b2ca
(viii) c3a

2. For any positive integer n, prove that the order of any element in Sn is finite and
is a divisor of n!

3. Find the orders of a, b and c given in the Exercise 1 above.

4. Determine all the elements in the cyclic subgroups ,a., ,b. and ,c., where
a, b and c are as given above.

5. Prove that Sn is abelian if and only if n # 2.

6. Can S8 be a homomorphic image of S12?

7. Can S8 be isomorphic to a subgroup of S12?

8. Compute aba21, bcb21 and cac21 for the elements a, b and c given in Exercise 1.

9. Determine the orders of all the elements in S3.

10. List all the elements of S4.

Q001-Algebra-111001_CH 06.indd 10 9/16/2011 10:05:19 AM

 Permutation Groups   6-11

6.3 CYCLES
Consider the permutation f in S5 given by f (1) 5 3, f (2) 5 1, f (3) 5 5, f (4) 5 2
and f (5) 5 4. In the array form, f can be expressed as

1 2 3 4 5
f   .
3 1 5 2 4

Instead of this, we can write

 1 3 5 4 2
 
f  ↓ ↓ ↓ ↓ ↓
 
 3 5 4 2 1

suggesting that 1 → 3 → 5 → 4 → 2 →1; that is, f maps 1 to 3, 3 to 5, 5 to 4,

4 to 2 and 2 to 1 back. Here, the action of f on the elements of I5 is cyclic.

2 3

4 5

In this case, we can as well denote f by (1 3 5 4 2) hinting each element, except

the last, is mapped onto the next and the last element to the first, completing
the cycle. Permutations like this are called cycles, which play a vital role in
the study of the permutations for the simple reason that any permutation can
be expressed as a product (composition) of cycles. Before going to prove this
fundamental theorem, we first have the following ­definition.

Definition 6.3.1. Let n be a positive integer and i1, i2, …, ir be distinct ele-
ments in the set In 5 {1, 2, …, n}. Define f : In → In by

i j1 , if i  i j , 1  j  r

f (i )  i1 , if i  ir .

i, if i  i j , 1  j  r


Q001-Algebra-111001_CH 06.indd 11 9/16/2011 10:05:20 AM

 6-12   Algebra – Abstract and Modern

That is, f (i1) 5 i2, f (i2) 5 i3, …, f (ir21) 5 ir, f (ir) 5 i1, and f (i) 5 i for all
i  {i1, i2, …, ir}.
The action of f is cyclic on the set {i1, i2, …, ir} and f is identity on the comple-
ment of {i1, i2, …, ir}.

i1

ir
i2

ir1 i3

For this reason, f is called a cycle of length r or simply an r-cycle and is

denoted by (i1, i2, …, ir) which is called the cyclic representation or the
cyclic form of the r-cycle. A 1-cycle is to be interpreted as the identity
permutation.
The cyclic representation (i1 i2 … ir) is not unique. For example, (i2 i3 … ir
i1) represents the same cycle as (i1 i2 … ir). Also, if a 5 (i1 i2 … ir) is an r-cycle
in Sn, then a is an r-cycle in Sm for all m  n. In fact, if m is the maximum of
i1, i2, …, ir, then (i1 i2 … ir) is an r-cycle in Sm.

Example 6.3.1
1. a 5 (2 5 3 4 6) is a 6-cycle in S6 and hence in Sm for any m  6. a is
defined by
a(2) 5 5, a(5) 5 3, a(3) 5 4, a(4) 5 6, a(6) 5 2
and a(i) 5 i for all i  {2, 3, 4, 5, 6}.
1 2 3 4 5 6 7 8
2. Let a 5  . Let us express this in cyclic form.
3 6 5 7 2 4 8 1
We have a(1) 5 3, a(3) 5 5, a(5) 5 2, a(2) 5 6, a(6) 5 4, a(4) 5 7,
a(7) 5 8 and a(8) 5 1. Therefore,
a 5 (1 3 5 2 6 4 7 8)

Note that a(1) 5 3, a2(1) 5 5, a3(1) 5 2, a4(1) 5 6, a5(1) 5 4, a6(1) 5

7, a7(1) 5 8 and a8(1) 5 1.

Q001-Algebra-111001_CH 06.indd 12 9/16/2011 10:05:20 AM

 Permutation Groups   6-13

The example given in (2) above can be extended to any r-cycle as given
in the following.

Theorem 6.3.1. Let a be a cycle in Sn and a(i)  (i) for some i  In. Then,
a 5 (i c(i) c2(i) … cr21(i)) for some r . 1.

Proof: Let a 5 (i1 i2 … ir). Since a(i)  i, we get that i 5 ik for some 1 # k
# r. Then,

a(i) 5 a(ik) 5 ik11

a2(i) 5 a(ik11) 5 ik12
ar2k(i) 5 ik1r2k 5 ir
ar2k+1(i) 5 a(ir) 5 i1

a (i) 5 ik21
r21

and    ar(i) 5 i.

Therefore, a 5 (ik ik11 … ir i1 … ik21)

5 (i a(i) a2(i) … ar21(i)). b

Theorem 6.3.2. Let a be an r-cycle in Sn. Then,

1. O(a) 5 r (that is, length of a is same as the order of a)

2. a21 5 ar21 5 (ir ir21 … i2 i1), where a 5 (i1 i2 … ir)
3. For any positive integer m, am 5 e if and only if r divides m.

Proof: Let a 5 (i1 i2 … ir). Then,

a 5 (ik ik11 … ir i1 i2 … ik21) for any 1 # k # r.

From the above theorem, it follows that ar(ik) 5 ik for all 1 # k # n and hence
ar coincides with the identity permutation. Therefore, ar 5 e. Also, for any
1 # k , r,

ak(i1) 5 ik11  i1

Q001-Algebra-111001_CH 06.indd 13 9/16/2011 10:05:20 AM

 6-14   Algebra – Abstract and Modern

and hence ak  e for all 1 # k , r. Thus, r is the smallest positive integer such
that ar 5 e and hence the order of a is r, which gives (1).
(2) Since ar 5 e, we get that a21 5 ar21. By the above theorem, we get that
ar21(ir) 5 ir21
ar21(ir21) 5 ir22, etc.
and hence ar21 5 (ir ir21 … i2 i1).
(3) This is clear, since O(a) 5 r. b

Note: If a 5 (i1 i2 … ir) is an r-cycle, then ak(ij) 5 is, where s  j 1 k (mod r).

Definition 6.3.2. A 2-cycle in Sn is called a transposition.

Note that any transposition a is a cycle of the form (i j) where i  j  In
and that a interchanges the positions of i and j and fix all the other elements
of In. That is,
a(i) 5 j, a(j) 5 i and a(k) 5 k for all k  In 2 {i, j}.

Theorem 6.3.3. Any r-cycle in Sn is a product of r 2 1 transpositions. In fact,

if a 5 (i1 i2 … ir) is an r-cycle, then
a 5 (i1 ir) o (i1 ir21) o … o (i1 i2).

Proof: Let a 5 (i1 i2 … ir). If k  In 2 {i1, i2, …, ir}, then a(k) 5 k 5 (i1 ij)
(k) for all 2 # j # r and hence
a(k) 5 ((i1, ir) o (i1 ir21) o … o (i1 i2))(k).
On the other hand, for 1 # j , r
a(ij) 5 ij11 5 ((i1 ir) o (i1 ir21) o … o (i1 i2))(ij)
((i1 ir) o (i1 ir21) o … o (i1 i2))(ij) 5 ((i1 ir) o (i1 ir21) o … o (i1 ij)) (ij)
5 (i1 ir) o … o (i1 ij11)(i1)
5 (i1 ir) o … o (i1 ij12)(ij11)
5 ij11 5 a(ij)
and (i1 ir) o (i1 ir21) o … o (i1 i2)(ir) 5 (i1ir)(ir) 5 i1 5 a(ir).

Thus, a 5 (i1 ir) o (i1 ir21) o … o (i1 i2). b

Example 6.3.2. Let a 5 (2 5 3 6 1 4) be a 6-cycle in S6. Then,

a 5 (2 4) o (2 1) o (2 6) o (2 3) o (2 5).

Q001-Algebra-111001_CH 06.indd 14 9/16/2011 10:05:21 AM

 Permutation Groups   6-15

Definition 6.3.3. For any permutation f in Sn, the support of f is defined as the
set {i  Sn : f (i)  i} of all elements in In which are not fixed by f. The support
of f will be denoted by supp(f ).
Note that if a is the r-cycle (i1 i2 … ir), then the support of a is precisely
the set {i1, i2, …, ir}.

Examples 6.3.3
1 2 3 4 5 6 7 8 9
1. Let f 5  .
6 5 3 7 4 1 2 8 9

Then, supp(f ) 5 {1, 2, 4, 5, 6, 7}.

1 2 3 4 5 1 2 3 4 5 6 7
2. Let f 5   and g 5  .
4 2 3 5 1 5 6 7 3 4 2 1

Then, supp(f ) 5 {1, 4, 5} and supp(g) 5 {1, 2, 3, 4, 5, 6, 7}.

1 2 3 4 5 6 7
Also f o g 5  
1 6 7 3 5 2 4
1 2 3 4 5 6 7
and g o f 5  
3 6 7 4 5 2 1
and therefore supp(f o g) 5 {2, 3, 4, 6, 7}
and     supp(g o f ) 5 {1, 2, 3, 6, 7}.

Definition 6.3.4. Two permutations f and g in Sn are said to be disjoint if their

supports are disjoint sets.
Permutations f1, f2, …, fr are said to be disjoint if they are pair-wise dis-
joint; equivalently, if fk(i)  i for some k, then fj(i) 5 (i) for all j  k.

Theorem 6.3.4. Any two disjoint permutations in Sn commute; that is,

f o g 5 g o f if f and g are disjoint.

Proof: Let f and g be disjoint permutations in Sn. Suppose that supp(f ) 5 A

and supp(g) 5 B. Then, A ∩ B 5  and hence, for any 1 # i # n, i  A or i
 B; that is, f (i) 5 i or g(i) 5 i.
If i  A ∪ B, then f (i) 5 i 5 g(i) and hence

(f o g)(i) 5 f (g(i)) 5 f (i) 5 i 5 (g o f )(i).

Q001-Algebra-111001_CH 06.indd 15 9/16/2011 10:05:21 AM

 6-16   Algebra – Abstract and Modern

If i  A, then f (i)  i and g(i) 5 i and hence

(f o g)(i) 5 f (g(i)) 5 f (i) 5 g(f (i)) 5 (g o f )(i)

(since f (i)  i and f is an injection, f (f (i))  f (i) and hence f (i)  A so that
f (i)  B and g(f (i)) 5 f (i)). Similarly, if i  B, then g(i)  i and f (i) 5 i and
hence (f o g)(i) 5 (g o f )(i). Thus, f o g 5 g o f. b
In the following, we prove a fundamental theorem on permutations which
will be an important tool in the study of permutations.

Theorem 6.3.5. Any nonidentity permutation f in Sn can be expressed as

f 5 a1 o a2 o … o as

where a1, a2, …, as are pair-wise disjoint cycles each of length atleast two. This
expression of f is unique except for the order of occurrences of the cycles ai.

Proof: Let e  f  Sn and A 5 supp(f ). For any i and j  A, define i ~ j if and

only if f r(i) 5 j for some r  Z. Note that f 0(i) 5 e(i) 5 i and that f r(i) 5 j
if and only if f2r(j) 5 i. These imply that ~ is an equivalence relation on
A 5 supp(f ). Let i be the equivalence class of ~ containing ~. Then, we shall
prove the following for any i  A.
1. i 5 {f r(i) : r  Z}.
2. There exist r  Z+ such that
i 5 {i, f (i), f 2(i), …, f r21(i)},
f r(i) 5 i and f s(i)  f t(i) for all 0 # s  t , r.
3. The restriction of f to i is an r-cycle, r . 1.
1. Follows from the definition of ~.
2. First note that j  i ⇒ f (j) i. Since In and hence supp(f ) is finite, i is
also a finite set, therefore, there exists least positive integer r such that i,
f (i), f 2(i), …, f r21(i) are all distinct and f r(i) 5 i. Thus, we have (2).
3. The restriction of f to i is the r-cycle
(i f (i) f 2(i) … f r21(i)).

Since f r(i) 5 i and i  supp(f ), it follows that r  1 and hence r . 1.

Now, since supp(f ) is finite, the number of equivalence classes of ~ in
supp(f ) is finite. Let

i1 , i2 , …, is

Q001-Algebra-111001_CH 06.indd 16 9/16/2011 10:05:22 AM

 Permutation Groups   6-17

be all the distinct equivalence classes of ~ in supp(f ). These are pair-wise

­disjoint and their union is the support of f. Corresponding to each ij , 1 # j # s,
define aj : In → In by

 f ( k ), if k ∈ ij
a j ( k )   .
k , if k ∉ 
i
 j

The restrictions of aj and f to ij are equal and hence, by (3) above, aj is a cycle
of length atleast two and clearly

f (k) 5 (a1 o a2 o … o as)(k) for all k  In.

Thus,   f 5 a1 o a2 o … o as.
Also, since distinct equivalence classes are disjoint, a1, a2, …, as are pair-wise
disjoint cycles, each of length atleast two.
The uniqueness of this representation of f is a direct consequence of the
facts that aj and f coincide on ij and aj is the identity outside ij .  b

Corollary 6.3.1. For any integer n . 1, every permutation in Sn is a product

of transpositions.

Proof: This is a consequence of the above theorem and the fact that any cycle
is a product of transpositions (see Theorem 6.3.3). Note that the identity per-
mutation e can be expressed as

e 5 (i j) o (i j)

where (i j) is any transposition in Sn (since n . 1, there is a transposition

in Sn). b

Note 6.3.1. An algorithm is given below for expressing a given permutation

as a product of disjoint cycles.
Let f be a permutation in Sn and

A 5 supp(f ) 5 {i  In : f (i)  i}.

Choose i1  A and consider

i1, f (i1), f 2(i1), f 3(i1), …

There should exist least r1 . 1 (since all these are elements in the finite set A)
at which f r1 (i1 )  i1 .

Q001-Algebra-111001_CH 06.indd 17 9/16/2011 10:05:23 AM

 6-18   Algebra – Abstract and Modern

Now, consider the r1-cycle

a1 5 (i1 f (i1) f 2(i1) … f r11 (i1 ))

Clearly r . 1 and f r1 (i1 ) 5 i1. Next choose i2  A 2 {i1, f (i1), …, f r11 (i1 )}
and consider

i2, f (i2), f 2(i2), f 3(i2), …

By the same argument given by, there exists r2 . 1 such that f r2 (i2 ) 5 i2 and

i2, f (i2), f 2(i2), …, f r21 (i2 )

are all distinct. Now, consider the r2-cycle

a2 5 (i2 f (i2) f 2(i2) … f r21 (i2 )).

Next choose i3  A 2 {i1, f (i1), …, f r11 (i1 ), i2, f (i2), …, f r21 (i2 )} and con-
tinue the above process. This process terminates when all the elements of
the support of f exhaust. Then, we get disjoint cycles a1, a2, …, as such that
f 5 a1 o a2 o … o as. Since ai’s are pair-wise disjoint, they commute with each
other and hence

f 5 as(1) o as(2) o … o as(s)

for any permutation s on {1, 2, …, s}.

1 2 3 4 5 6 7 8 9
Worked Exercise 6.3.1. Express f 5   as a
3 6 5 4 1 2 8 9 7
product of disjoint cycles and as a product of transpositions.

Answer: We have supp(f ) 5 {1, 2, 3, 5, 6, 7, 8, 9}.

Now 1  supp(f ) and consider

1, f (1) 5 3, f 2(1) 5 5, f 3(1) 5 1.

(1 3 5) is the cycle (1 f (1) f 2(1)).

Next choose 2  supp(f ) 2 {1, 3, 5} and consider

2, f (2) 5 6, f 2(2) 5 2.

(2 6) is the cycle (2 f (2)).

Next choose 7  supp(f ) 2 { 1, 3, 5, 2, 6} and consider

7, f (7) 5 8, f 2(7) 5 9, f 3(7) 5 7.

Q001-Algebra-111001_CH 06.indd 18 9/16/2011 10:05:24 AM

 Permutation Groups   6-19

(7 8 9) is the cycle (7 f (7) f 2(7)). We have exhausted all the elements of

supp(f ) and hence

f 5 (1 3 5) o (2 6) o (7 8 9).

By using Theorem 6.3.3, we have

(1 3 5) 5 (1 5) o (1 3)
and (7 8 9) 5 (7 9) o (7 8)
and hence f 5 (1 5) o (1 3) o (2 6) o (7 9) o (7 8).

Worked Exercise 6.3.2. Let i1, i2, …, ir be given distinct elements in In. How
many distinct r-cycles can be formed using all the i1, i2, …, ir.

Answer: Any permutation s on {1, 2, …, r} gives us an r-cycle (is(1) is(2) …

is(r)) and every r-cycle formed using all the i1, i2, …, ir must be of this form.
But, we have discount repetitions, since if (is(1) is(2) .... is(r)) is an r-cycle, then
(is(k) is(k11) … is(r) is(1) is(2) … is(k21)) is the same r-cycle for any 1 # k # r.
Therefore, there are exactly rr! ( ( r 1)!) distinct r-cycles formed by using
all the i1, i2, …, ir.
Expressing a permutation as a product of disjoint cycles is an important
tool in determining the order of that permutation. The following result gives
us a formula for the order of a permutation.

Theorem 6.3.6. Let f1, f2, …, fr be pair-wise disjoint permutations in Sn and

f 5 f1 o f2 o … o fs. Then,

O(f ) 5 l.c.m. of {O(f1), O(f2), …, O(fs)}.

Proof: Let O(f i) 5 ri, r 5 l.c.m. of {r1, r2, ..., rs} and r 5 riti, ti  Z+. Since
f i’s are pair-wise disjoint, we get that

fi o fj 5 fj o fi for all 1 # i, j # s.

Now,

f r 5 (f1 o f2 o … o fs)r
5 f1r o f2r o … o fsr (since fi o fj 5 fj o fi)
 f1r1t1 ο f 2r2t2 ο  ο f srsts
5 e (since O(fi) 5 ri, f i ri 5 e).

Q001-Algebra-111001_CH 06.indd 19 9/16/2011 10:05:25 AM

 6-20   Algebra – Abstract and Modern

On the other hand, for any positive integer t,

f t 5 e ⇒ f1t o f2t o … o fst 5 e

⇒ f1t 5 f2t 5 … 5 fst 5 e (since f i’s are disjoint)
⇒ O(fi) divides t for all 1 # i # s
⇒ ri divides t for all 1 # i # s
⇒ r divides t (since r 5 l.c.m. of {r1, …, rs}).

Thus, O(f ) 5 r 5 l.c.m. of {O(f1), O(f2), …, O(fs)}.

Corollary 6.3.2. Let f 5 a1 o a2 o … o as, where a1, a2, …, as are pair-wise

disjoint cycles of length r1, r2, …, rs, respectively. Then,

O(f ) 5 l.c.m. of {r1, r2, ..., rs}.

Proof: Since ai is a cycle of length ri, we have O(ai) 5 ri and hence

O(f ) 5 l.c.m. of {O(a1), O(a2), …, O(as)}

5 l.c.m. of {r1, r2, …, rs}. b

EXERCISE 6(c)

1. State whether each of the following is true or false:

(i) Every cycle is a transposition.
(ii) Every transposition is a cycle.
(iii) Every cycle is a permutation.
(iv) Every permutation is a cycle.
(v) Every transposition is a permutation.
(vi) Every permutation is a product of disjoint transpositions.
(vii) f 2 5 e for any transposition f.
(viii) f 2 5 e implies that f is a transposition.
(ix) For any f and g in Sn, f o g 5 e ⇒ f 5 e 5 g
(x) supp(f ) 5 supp(g) if and only if f 5 g.

2. Which of the following are cycles? If they are cycles, then express them in cyclic
representation.
(i) 1 2 3 4 5 6 7 8
 
4 6 8 2 3 5 1 7

Q001-Algebra-111001_CH 06.indd 20 9/16/2011 10:05:25 AM

 Permutation Groups   6-21

1 2 3 4 5 6

(ii) 
2 3 4 1 6 5
1 2 3 4 5 6 7 8
(iii)  
2 3 4 5 6 7 8 1
1 2 3 4 5 6 7 8 9
(iv)  
8 6 7 9 5 1 2 3 4
3. Express the following as products of disjoint cycles, each of length atleast two.
Also express each of following as a product of transpositions.
1 2 3 4 5 6 7 8 9
(i)  
4 3 2 1 5 8 6 7 9
1 2 3 4 5 6 1 2 3 4 5 6
(ii)   ο  
3 2 4 1 6 5 5 1 4 6 2 3
1 2 3 4 5 6 7 8 1 2 3 4 5 6
(iii)   ο  
2 3 1 4 6 5 7 8 5 6 2 3 4 1
   
(iv) 1 2 3 ο 1 2 3 4 5 6
2 3 1 3 1 2 5 6 4
(v) (1 3 7 4 6) ο (2 3 5 6 4) ο (8 7 6 2 4 3 5)
(vi) (1 2 3 4) ο (2 3 4 5) ο (3 4 5 6)
4. Determine the order of each of the following permutations.
1 2 3 4 5 6 7 8 9
(i)  
2 4 6 1 7 3 8 9 5
   
(ii) 1 2 3 4 5 6 ο 1 2 3 4 5
4 3 2 1 6 5 2 3 4 1 5
(iii) (5 4 3 2) ο (1 2 3 4 5 6) ο (2 4 6 1 3 5)
(iv) (8 7 6 9 3 4) ο (4 3 9 6 7 8) ο (3 4 5 6 7 8 9)
5. For any permutations f and g in Sn, prove that supp(f o g) ⊆ supp(f ) ∪ supp(g).

6. If f and g are disjoint permutations in Sn, prove that

supp(f o g) 5 supp(f ) ∪ supp(g)
and supp(f ) ∩ supp(g) 5 .
7. If f 5 f1 o f2 o … o fr and fi’s are pair-wise disjoint permutations in Sn, prove that
supp(f1), …, supp(fr) form a partition of supp(f ).
8. For f  Sn and for m  Z, prove that supp(f m) ⊆ supp(f ).
9. For any disjoint permutations f and g, prove that f o g 5 e if and only if f 5 e 5 g.

Q001-Algebra-111001_CH 06.indd 21 9/16/2011 10:05:26 AM

 6-22   Algebra – Abstract and Modern

10. For any permutation f in Sn, prove that supp(f ) 5  if and only if f 5 e.
11. Let f  Sn. Prove that f is a transposition if and only if supp(f ) is a 2-element set.
12. Prove that f is a 3-cycle if and only if supp(f ) is a 3-element set. Can this be
extended for any r-cycle?
13. Let e  f  Sn. Prove that f 2 5 e if and only if f is a product of disjoint
transpositions.
14. For any f and g  Sn, prove that O(f ) 5 O(g f g21).
15. For any 1 # r # n, prove that there is an element in Sn whose order is r.
16. If a is an r-cycle and r is odd, prove that a2 is also a cycle.
17. If r is even in Exercise 1b, then can a2 be a cycle? Substantiate your answer.
18. If a is an r-cycle and 1 # s < r such that r and s are relatively prime, then prove
that as is also an r-cycle.
19. For any r-cycle a, prove that f o a o f21 is also an r-cycle for any permutation f.
20. If a and b are disjoint cycles, prove that f o a o f21 and f o b o f21 are also disjoint
cycles.
21. If f 5 a1 o a2 o … o as is a representation of a permutation f in Sn as a product of
disjoint cycles and g is any permutation in f, then prove that
g o f o g21 5 (g o a1 o g21) o (g o a2 o g21) o … o (g o as o g21)
is a representation of g o g o g21 as a product of disjoint cycles.

22. For any positive integer n, a partition of n is defined to be a finite sequence r1, r2,
…, rs of positive integers such that r1 # r2 # … # rs and r1 1 r2 1 … 1 rs 5 n.
List all the partitions of 4 and 5.

23. For any permutation f in Sn, let |f | denote the number of elements in the support
of f. Let f 5 a1 o a2 o … o as where a1, a2, …, as are disjoint cycles, each of length
greater than 1, such that |a1| # |a2| # … # |as|. Then prove that |a1|, |a2|, …, |as|
is a partition of |f |.

24. Prove that any permutation in Sn determines a partition of n such that f and g
determine the same partition of n if and only if g 5 h o f o h21 for some h  Sn.

25. Prove that Sn is generated by the n 2 1 transpositions (1 2), (1 3), (1 4), …, (1 n).

26. Prove that Sn is generated by (1 2) and (1 2 3 4 … n).

27. Prove that Sn is generated by the transpositions (1 2), (2 3), (3 4), …, (n 2 1 n).
28. If f 5 (i1 i2 … ir) is an r-cycle in Sn, then prove that g f g21 5 (g(i1) g(i2) … g(in))
for all g  Sn.

Q001-Algebra-111001_CH 06.indd 22 9/16/2011 10:05:26 AM

 Permutation Groups   6-23

29. For any f and g  Sn, prove that f is an r-cycle if and only if g f g21 is an r-cycle.

30. Prove that any group of order 6 is isomorphism to either Z6 or S3.

6.4 ALTERNATING GROUP A n AND DIHEDRAL

GROUP Dn
We have proved in the previous section that every permutation can be expressed
as a product of transpositions. This expression is not unique unless the trans-
positions involved are disjoint. For example, for any distinct i, j, k and l in In,

(i j) o (k l) o (j i) 5 (k l)
(i j) o (j i) 5 e 5 (k l) o (l k).

Even though an expression of a permutation as a product of transpositions is

not unique, the number of transpositions involved in any expression of a given
permutation is always even or always odd. That is, if a permutation f can be
expressed as a product of even number of transpositions, then any expression
of f as a product of transpositions contains even number of transpositions. In
order to prove this, we first have the following definition.

Definition 6.4.1. Let f be a nonidentity permutation in Sn and

f 5 a1 o a2 o … o as

where a1, a2, …, as are pair-wise disjoint cycles. Then, the Cauchy index of f
is defined as

O(a1) 1 O(a2) 1 … 1 O(as) 2 s

and is denoted by CI(f ). Also, for the identity permutation e, we define CI(e)
to be 0.
Since any f  e can be uniquely, except for the order of occurrences of the
cycles, expressed as a product of disjoint cycles, the Cauchy index of f is
well-defined and CI(f ) is always a positive integer for any f  e. Note that,
s
CI( f )  ∑ Ο( ai )  s | f |  s
i1

where e  f 5 a1 o a2 o … o as is an expression of f as a product of disjoint

cycles a’i s and |f | is the number of elements in the support of f. Since each ai
is a cycle of length atleast two, O(ai) . 1 for all 1 # i # s and hence |f | . s,
so that CI(f ) . 0.

Q001-Algebra-111001_CH 06.indd 23 9/16/2011 10:05:27 AM

 6-24   Algebra – Abstract and Modern

Examples 6.4.1
1 2 3 4 5 6 7 8 9
1. Let f 5  
4 1 3 2 7 9 8 5 6
Then, f 5 (1 4 2) o (5 7 8) o (6 9) 5 a1 o a2 o a3
Therefore, CI(f ) 5 O(a1) 1 O(a2) 1 O(a3) 2 3
        5 3 1 3 1 2 2 3 5 5
2. Let f 5 (1 2) o (2 5) o (4 5) o (3 7) o (9 3) o (6 9) o (8 9). Then, we
should express f as a product of disjoint cycles, to find the Cauchy index
of f. We have
f 5 (1 2 5 4) o (3 9 8 6 7) 5 a1 o a2
Therefore, CI(f ) 5 O(a1) 1 O(a2) 2 2
54152257

Theorem 6.4.1. If f  Sn is a product of s transpositions, then

CI(f ) 1 s is even.
Proof: Let f  Sn and f 5 a1 o a2 o … o at, where a1, a2, …, at are pair-wise
disjoint cycles of orders r1, r2, …, rt, respectively. Then, we have
CI(f ) 5 r1 1 r2 1 … 1 rt 2 t.
We shall prove that
CI((i j) o f ) 5 CI(f ) 6 1
for any transposition (i, j). To find CI((i, j) o f ), we should first express (i j) of
as a product of disjoint cycles. We do this in the following cases:
Case (1): Suppose that (i j) is disjoint with f. In this case, (i j) is disjoint with
each ai and
(i j) o f 5 (i j) o a1 o a2 o … o at
which is a product of t 1 1 number of pair-wise disjoint cycles and therefore
CI((i j) o f ) 5 2 1 r1 1 r2 1 … 1 rt 2 (t 1 1)
5 (r1 1 r2 1 … 1 rt 2 t) 1 1
5 CI(f ) 1 1
Case (2): Suppose that i  supp(f ) and j  supp(f ). Then, f (i)  i and f (j) 5 j.
We can assume, without loss of generality, that i  supp(a1) and i  supp(ak)
for all 1 , k # t and that

a1 5 (i k1 k2 … kr).

Q001-Algebra-111001_CH 06.indd 24 9/16/2011 10:05:27 AM

 Permutation Groups   6-25

Then, r1 5 O(a1) 5 r 1 1 and

(i j) o f 5 (i j) o (i k1 k2 … kr) o a2 o a3 o … o at
5 (i k1 k2 … kr j) o a2 o a3 o … o at

which is a product of t number of disjoint cycles and therefore

CI ((i j) o f ) 5 (r 1 2) 1 r2 1 … 1 rt 2 t
5 (r1 1 1) 1 r21 … 1 rt 2t
5 (r1 1 r2 1 … 1 rt 2 t) 1 1
5 CI(f ) 1 1

Case (3): Suppose that both i and j belong to supp(f ).

(i) Suppose that both i and j involve in the same cycle, say a1.
Then, a1 5 (i k1 … kr j m1 … mu)
and r1 5 O(a1) 5 r 1 u 1 2.
Now, (i j) o f 5 (i j) o (i k1 … kr j m1 … mu) o a2 o … o at
      5 (i k1 … kr) o (j m1 … mu) o a2 o … o at
which is a product of t 1 1 number of disjoint cycles and therefore
CI((i j) o f ) 5 (r 1 1) 1 (u 1 1) 1 r2 1 … 1 rt 2 (t 1 1)
5 (r 1 u 1 2) 1 r2 1 … 1 rt 2 (t 1 1)
5 (r1 1 r2 1 … 1 rt 2 t) 2 1
5 CI(f ) 2 1.
(ii) Suppose that i and j involve in distinct cycles of f, say
a1 5 (i k1 … kr) and a2 5 (j m1 … mu).

Then, r1 5 O(a1) 5 r 1 1 and r2 5 O(a2) 5 u 1 1

Now,
(i j) o f 5 (i j) o a1 o a2 o … o at
5 (i j) o (i k1 … kr) o (j m1 … mu) o a3 o … o at
5 (i k1 … kr j m1 … mu) o a3 o … o at

which is a product t 2 1 number of disjoint cycles.

Therefore, CI((i j) o f ) 5 (r 1 u 1 2) 1 r3 1 … 1 rt 2 (t 2 1)
          5 (r1 1 r2 1 r3 1 … 1 rt 2 t) 1 1
          5 CI(f ) 1 1.
Thus, in any case, CI((i j) o f ) 5 CI(f ) 6 1 for any transposition (i j).

Q001-Algebra-111001_CH 06.indd 25 9/16/2011 10:05:27 AM

 6-26   Algebra – Abstract and Modern

Now, let f be a product of s transpositions; that is, f 5 c1 o c2 o … o cs, where

c1, c2, …, cs are transpositions. Then consider

0 5 CI(e) 5 CI(f2l o f )
5 CI((c1 o c2 o … o cs)21 o f )
5 CI(cs o cs21 o … o c2 o c1 o f )
5 CI(cs21 o … o c2 o c1 o f ) 6 1
5…
5 CI(f ) 6 1 6 … 6 1
5 CI(f ) 1 p 2 q

where p and q are nonnegative integers such that p 1 q 5 s. Therefore,

CI(f ) 1 p 5 q
and CI(f ) 1 s 5 CI(f ) 1 p 1 q 5 2q.

Since 2q is always even, it follows that CI(f ) 1 s is even. b

Corollary 6.4.1. Let a1 a2, …, ar and b1, b2, …, bs be transpositions such that

a1 o a2 o … o ar 5 b1 o b2 o … o bs.

Then, r 1 s is even and hence either both r and s are even or both r and s
are odd.

Proof: Let f 5 a1 o a2 o … o ar 5 b1 o b2 o … o bs.

Then, by the above theorem, CI(f ) 1 r and CI(f ) 1 s are even and hence

CI(f ) 1 r 1 CI(f ) 1 s is even.

So that r 1 s is even. The later assertion follows from the fact that r 1 s is odd
if and only if one of r and s is odd and the other is even. b

Corollary 6.4.2. If a permutation can be expressed as a product of even num-

ber of transpositions, then it cannot be expressed as a product of odd number
of transpositions.

Definition 6.4.2. A permutation in Sn is called an even permutation if it can

be expressed as a product of even number of transpositions. A permutation
which is not even is called an odd permutation.

Q001-Algebra-111001_CH 06.indd 26 9/16/2011 10:05:28 AM

 Permutation Groups   6-27

Examples 6.4.2
1 2 3 4 5 6 7 8 9
1. The permutation f 5  can be expressed as
4 6 5 7 2 3 1 9 8
f 5 (1 4 7) o (2 6 3 5) o (8 9)
5 (1 7) o (1 4) o (2 5) o (2 3) o (2 6) o (8 9)

which is a product of six transpositions. Therefore, f is an even permu-

tation.
2. Any r-cycle can be expressed as a product of r 2 1 transpositions and
therefore an r-cycle is an even permutation if and only if r is odd.
3. Any three cycle (i j k) 5 (i k) o (i j) is an even permutation.
4. Any transposition is an odd permutation.
5. The identity e in Sn is an even permutation if n . 1.

Theorem 6.4.2. For any integer n . 1, the set of all even permutations in Sn
is a normal subgroup of Sn and is of index 2 in Sn.

Proof: Let n . 1 and

An 5 the set of all even permutations in Sn.

Consider the group G 5 {1, 21} under the usual multiplication of real num-
bers. Define

1, if f is even
 : Sn → G by ( f )   .
1, if f is odd

Since n . 1, we have (1, 2)  Sn and (1, 2) is an odd permutation and hence

 ((1, 2)) 5 21. Also the identity

e 5 (1 2) o (1 2)

and hence e is an even permutation so that (e) 5 1. Therefore,  is a surjection.

For any f and g in Sn, note that f o g is even if and only if either both f and g are
even or both f and g are odd. Also, for any a and b in G, the product ab 5 1
if and only if either a 5 1 5 b or a 5 2 1 5 b. From these, it follows that 
is a homomorphism and hence an epimorphism. Also,

ker  5 {f  Sn : (f ) 5 1} 5 An.

Q001-Algebra-111001_CH 06.indd 27 9/16/2011 10:05:28 AM

 6-28   Algebra – Abstract and Modern

The kernel of any homomorphism is a normal subgroup of the domain group

and hence An is normal subgroup of Sn. Further, by the Fundamental Theorem
of Homomorphisms,

Sn / An  G.

Therefore, we have

| Sn |
iSn ( An )   | Sn An |  | G |  2.
| An |

Thus, the index of An in Sn is 2.

|Sn| n!
Corollary 6.4.3. For any n . 1, | An |   . b
2 2
Definition 6.4.3. For any n . 1, the group An of all even permutations in Sn
is called the alternating group of degree n.

Worked Exercise 6.4.1. Construct a table representing the alternating group

A3 of degree 3.

Answer: First note that |A3| 5 3!  3.

2
The identity e and the 3-cycles a 5 (1 2 3) and b 5 (1 3 2) are the only even
permutations in S3.
Therefore, A3 5 {e, a, b}. Note that

a2 5 b, a3 5 e, b2 5 a and b3 5 e

o e a b
e e a b
a a b e
b b e a

Worked Exercise 6.4.2. List all the elements of A4 and construct a table
­representing the group A4.

Answer: First note that |A4| 5 4!

2
 12.
There are eight 3-cycles, each of which is an even permutation. These are

a1 5 (1 2 3), a2 5 (1 3 2)
b1 5 (2 3 4), b2 5 (2 4 3)
c1 5 (1 2 4), c2 5 (1 4 2)
d1 5 (1 3 4), d2 5 (1 4 3)

Q001-Algebra-111001_CH 06.indd 28 9/16/2011 10:05:28 AM

 Permutation Groups   6-29

Also,  p 5 (1 2) o (3 4)
q 5 (1 3) o (2 4)
and  r 5 (1 4) o (2 3)

are also even permutations. There are only 12 even permutations and there-
fore the above together with the identity form A4. That is,

A4 5 {e, a1, a2, b1, b2 c1, c2 d1, d2 p, q, r}.

Here, we have

a12  a2 , a22  a1 , a13  e  a23

b12  b2 , b22  b1 , b13  e  b23
c12  c2 , c22  c1 , c13  e  c23
d12  d2 , d22  d1 , d13  e  d23

p2  q2  r 2  e
The following table represents the group A4.

o e a1 a2 b1 b2 c1 c2 d1 d2 p q r
e e a1 a2 b1 b2 c1 c2 d1 d2 p q r
a1 a1 a2 e p c1 q d2 b1 r d1 b2 c2
a2 a2 e a1 d1 q b2 r p c2 b1 c1 d2
b1 b1 q c2 b2 e d1 p r a1 a2 d2 c1
b2 b2 d2 p e b1 r a2 c1 q c2 a1 d1
c1 c1 r d1 a1 p c2 e q b2 d2 a2 b1
c2 c2 b1 q r d2 e c1 a2 p b2 d1 a1
d1 d1 c1 r q a2 p b1 d2 e a1 c2 b2
d2 d2 p b2 c2 r a1 q e d1 c1 b1 a2
p p b2 d2 c1 a1 b1 d1 c2 a2 e r q
q q c2 b1 a2 d1 d2 a1 b2 c1 r e p
r r d1 c1 d2 c2 a2 b2 a1 b1 q p e

Alternating group of degree 4

We know that any 3-cycle is an even permutation. In fact, every even permu-
tation is a product of 3-cycles and hence the 3-cycles generate the group An.
This is proved in the following theorem.

Q001-Algebra-111001_CH 06.indd 29 9/16/2011 10:05:29 AM

 6-30   Algebra – Abstract and Modern

Theorem 6.4.3. Let n . 2 and i  j  In 5 {1, 2, …, n}. Then, the

­alternating group An is generated by the 3-cycles of the form (i j k), where
k  In 2 {i, j}.

Proof: Let S 5 {(i j k) : k  In 2 {i, j}} Clearly S ⊆ An. Any even permuta-
tion must be a product of terms of the form

(a b) o (c d) or (a b) o (a c)

where a, b, c and d are distinct elements of In. Since

(a b) o (c d) 5 (a c b) o (a c d)
and (a b) o (a c) 5 (a c b),

it follows that An is generated by the set of all 3-cycles in Sn. Next, we prove
that any 3-cycle can be expressed as product of 3-cycles in S. Any 3-cycle is
of the form (i j a) or (i a j) or (i a b) or (j a b) or (a b c), where a, b and c are
distinct elements In 2 {i, j}. Now, we have

(i a j) 5 (i j a) o (i j a)
(i a b) 5 (i j b) o (i j a) o (i j a)
(j a b) 5 (i j b) o (i j b) o (i j a)
and (a b c) 5 (i j a)2 o (i j c) o (i j b)2 o (i j a)

Thus, every 3-cycle is a product of members of S and hence An is gener-

ated by S. b

Corollary 6.4.4. For any n  2, An is the smallest subgroup of Sn containing

all the 3-cycles in Sn.

Theorem 6.4.4. Let n . 2 and N be a normal subgroup of An. If N contains

a 3-cycle, then N 5 An.

Proof: Suppose that N contains a 3-cycle (i j k), where i, j and k are some
distinct elements in In.
For any a  In 2 {i, j, k}, we have

(i j a) 5 (i j) o (k a) o (i j k)2 o (k a) o (i j)
5 (i j) o (k a) o (i j k)2 o ((i j) o (k a))21  f N f21,
where f 5 (i j) o (k a)  An.

Q001-Algebra-111001_CH 06.indd 30 9/16/2011 10:05:29 AM

 Permutation Groups   6-31

Since N is normal in An, it follows that (i j a)  N. By Theorem 6.4.3,

N 5 An.
In the following, we prove an important property of the alternating groups;
namely, all alternating groups, except A4, are simple in the following sense.

Definition 6.4.4. A nontrivial group G is said to be simple if G does not

contain any nontrivial proper normal subgroups; that is, {e} and G are the
only normal subgroups of G.

Example 6.4.3
1. Any finite group of prime order is simple; for if |G| 5 p, where p is a
prime number and H is a subgroup of G, then, by the Lagrange’s theo-
rem, |H| divides |G| and hence |H| 5 1 or p which implies that H 5 {e}
or H 5 G.
2. A4 is a not a simple group, since
N 5 {e, p q, r}

is a normal subgroup of A4 (see Worked Exercise 6.4.2)

3. In the following, we prove that An is simple for any n  4.

Theorem 6.4.5. Let n . 2. Then, the alternative group An is simple if and

only if n  4.

( )
Proof: Since A3 is a group of order 3  32! , A3 is simple (see Example 6.4.3 (1)).
Also A4 is not simple by Example 6.4.3 (2). Now, let n . 4. We shall prove
that An is simple. Let N be a normal subgroup of An and N  {e}. By
Theorem 6.4.4, it is enough if we can prove that N contains a 3-cycle. We
prove this by distinguishing the following cases.
Case (1): Suppose that N contains an element f such that f is the product of
disjoint cycles, atleast one of which is of length r  4. Then,

f 5 (i1 i2 … ir) o b

where r  4 and (i1 i2 … ir) and b are disjoint. Now, put a 5 (i1 i2 i3). Then, a
 An and, by the normality of N in An, a o f o a21  N. Now, we have

(i1 i2 ir) 5 b2l o (i1 ir ir21 … i2) o a o (i1 i2 … ir) o b o (i1 i3 i2)
5 f21 o (a o f o a21)  N

Thus, N contains a 3-cycle.

Q001-Algebra-111001_CH 06.indd 31 9/16/2011 10:05:30 AM

 6-32   Algebra – Abstract and Modern

Case (2): Suppose that N contains an element f such that f is the product of
disjoint cycles, atleast two of which are 3-cycles. Then,

f 5 (i1 i2 i3) o (i4 i5 i6) o b

where (i1 i2 i3), (i4 i5 i6) and b are disjoint.

Put a 5 (i1 i2 i4). Then, a  An and, by the normality of N in An, a o f o a21  N.
Now, consider

(i1 i4 i2 i6 i3) 5 b21 o (i4 i6 i5) o (i1 i3 i2) o (i1 i2 i4) o (i1 i2 i3)
o (i4 i5 i6) o b o (i1 i4 i2)
5 f21 o (a o f o a21)  N

Therefore, N has a 5-cycle and hence, by Case (1), N contains a 3-cycle.

Case (3): Suppose that N contains an element f which is the disjoint prod-
uct of one 3-cycle and some 2-cycles. Then,

f 5 (i1 i2 i3) o b

where (i1 i2 i3) and b are disjoint and b is a product of disjoint 2-cycles. Now
we have

(i1 i3 i2) 5 (i1 i2 i3)2

5 (i1 i2 i3)2 o b2 (since b2 5 e)
5 (i1 i2 i3) o b o (i1 i2 i3) o b
5 f 2  N.

Thus, N contains a 3-cycle.

Case (4): Suppose that every element of N is the product of (an even
number of ) disjoint 2-cycles.
Let f  N such that f 5 (i1 i2) o (i3 i4)
where (i1,i2), (i3,i4) and N are disjoint. Put a 5 (i1 i2 i3)
Then, a  An and therefore f 21 o (a o f o a21)  N. But

f 1 ο ( a ο f ο a1 )  b1 ο (i3 ⋅ i4 ) ο (i1 ⋅ i2 ) ο (i1 ⋅ i2 ⋅ i3 ) ο (i1 ⋅ i2 ) ο

(i3 i4) o b o (i1 i3 i2)
5 (i1 i3) o (i2 i4)

Now, put  5 (i1 i3) o (i2 i4). We have   N. since n $ 5, we can choose
j  In 2 {i1, i2, i3, i4}. Put  5 (i1 i3 j).

Q001-Algebra-111001_CH 06.indd 32 9/16/2011 10:05:30 AM

 Permutation Groups   6-33

Then,   An and  o ( o  o 21)  N. Also,

 ο ( ο  ο 1 )  (i1i3 ) ο (i2i4 ) ο (i1i3 j ) ο (i1i3 ) ο (i2i4 ) ο (i1 ji3 )

 (i1i3 j ).

Therefore, (i1 i3 j)  N. Thus, N contains a 3-cycle. Thus, in any case, N

contains a 3-cycle and, by Theorem 6.4.4, N 5 An. Thus, An is simple for
all n  5. b

Theorem 6.4.6. For any n . 1, the alternating group An is the only subgroup
of index 2 in the symmetric group Sn.

Proof: Let H be a subgroup of index 2 in Sn. Then, there is only one left coset
of H other than H. Then, fH 5 Sn 2 H and hence fH 5 gH for any f and g 
Sn 2 H. In particular, fH 5 f21H for any f  Sn 2 H (note that f  H if and
only if f21  H) and hence f2  H for all f  Sn.
If a 5 (i j k) is any 3-cycle in Sn, then a3 5 e and hence
a21 5 a2  H and therefore a  H.
Therefore, H contains all 3-cycles. By Corollary 6.4.4, An ⊆ H. But

| Sn | |S |
 iSn ( H )  2  iSn ( An )  n
|H | | An |

and hence |H| 5 |An|. Thus, H 5 An. b

In the following, we shall exhibit another special subgroup of the sym-
metric group Sn and prove a characterization theorem for it.

Definition 6.4.5. Let n $ 3 and  and f  Sn be defined as follows:

 5 the n-cycle (1 2 3 … n)
1 2 3 4 … n 
and    
1 n n 1 n  2 … 2 

Note that f   (i n  2  i ), the product of the transpositions (i n 1

2in2i
2 2 i). Let Dn be the subgroup of Sn generated by {, }. Dn is called the
dihedral group of degree n.
The dihedral group Dn is completely characterized by certain properties of its
generators  and .

Q001-Algebra-111001_CH 06.indd 33 9/16/2011 10:05:31 AM

 6-34   Algebra – Abstract and Modern

Theorem 6.4.7. Let n $ 3. Then, a group G is isomorphic to the dihedral

group Dn of degree n if and only if G is generated by two elements a and b
satisfying the following.
1. O(a) 5 n
2. O(b) 5 2
3. aba 5 b

Proof: Let G be a group. Suppose that G  Dn and f : G → Dn is an isomor-

phism. Recall that Dn is generated by  and  , where

  (1 2 3 … n)
1 2 3 4  n
and f   .
1 n n 1 n  2  2

Being an n-cycle,  is of order n. Also,

f ∏
2in2i
(i, n  2  i )  (2 n) ο (3 n 1) ο (4 n  2) ο 

which is a product of disjoint transpositions. Since the order of any transposi-

tion is 2, we get that O() 5 2. Also, it can be easily verified that

 o  o  5 .

If we choose a and b  G such that f (a) 5  and f (b) 5 , then we get that G
is generated by a and b, O(a) 5 n, O(b) 5 2 and aba 5 b.
Conversely suppose that G is generated by two elements a and b, such that
O(a) 5 n, O(b) 5 2 and aba 5 b (or ab 5 ba21 or ba 5 a21b). Then, for any
integers j and k,

   bak 5 a21bak21 5 a22bak22 5 … 5 a2kb

and b j a k  b j1ak b  b j2 a k b 2  …  a(1) jk b j .

From these relations and from the hypothesis that G is generated by a and b,
it follows that every element of G can be expressed as ak bj for some integers
k and j. Since O(a) 5 n,

{ai : i ∈ Z}  a  {e, a, a 2 , …, a n1}.

Q001-Algebra-111001_CH 06.indd 34 9/16/2011 10:05:31 AM

 Permutation Groups   6-35

Also, since O(b) 5 2, we have b 5 b21 and hence

e, if j is even
b j   .
b, if j is odd

Thus, G 5 {akbj : 0 # k , n and j 5 0 or 1}.

Further, suppose that akbj 5 arbs, where 0 # k, r , n and j, s  {0, 1}.
Then,

a kr  b s j .

If s  j, then s 2 j 5 1 or 21, so that ak2r 5 b and

a kr2  a( a kr )a  aba  b  a kr

and hence a2 5 e, which is a contradiction (since O(a) 5 n . 2). Therefore,

s 5 j and ak2r 5 e so that k 5 r.
Thus, every element of G can be uniquely expressed as ak bj where 0 # k , n
and j 5 0 or 1. Now, define

f : G → Dn by f (ak bj) 5 uk j

for all 0 # k , n and j 5 0 or 1. Using the fact that u and  satisfy the same
conditions (1), (2) and (3) in Dn as a and b in G, it can easily checked that f is
an isomorphism. Thus, G . Dn. b

Corollary 6.4.5. The order of the dihedral group Dn of degree n is 2n.

Proof: This is an immediate consequence of the fact that any element of Dn

can be uniquely expressed as u k j, where 0 # k , n and j 5 0 or 1.
Recall from Example 3.2.8 that the group of symmetries of a square
(a regular 4-gon) is of order 8. In fact, we prove in the following theorem that
the group of symmetries of a square is isomorphic to the dihedral group D4
of degree 4.

Theorem 6.4.8. Let n  3 and D9n be the group of all symmetries of a regular
n-gon (a polygon of n equal sides). Then,

Dn  D9n.

Q001-Algebra-111001_CH 06.indd 35 9/16/2011 10:05:32 AM

 6-36   Algebra – Abstract and Modern

Proof: Consider a regular n-gon. Without loss of generality, we can assume

that one of the vertices of the n-gon lies on the X-axis. First observe that D9n
is generated by the rotation a and the reflection b, where a and b are analyti-
cally given by

 2 2 2 2 
a( x, y )   x cos  y sin , x sin  y cos 
 n n n n 
and b(x, y) 5 (x, 2y).

2�
n

It is clear that O(a) 5 n, O(b) 5 2 and aba 5 b. Therefore, by Theorem 6.4.7,

D9n  Dn.

Corollary 6.4.6. The group of symmetries of a square is isomorphic to the

dihedral group D4 of degree 4 and the group of symmetries of an equilateral
triangle (see Worked Exercise 3.3.7) is isomorphic to D3.

Worked Exercise 6.4.3. Prove that the dihedral group Dn, n  3, is not simple.

Answer: Dn is generated by two elements u and , where O(u)n, O() 5 2

and u o  o u 5 . Consider the subgroup A generated by u in Dn. Then, |A|
5 O(u) 5 n. Since |Dn| 5 2n, the index of A in Dn is 2 and hence A is normal
in Dn, A  {e} and A  Dn. Thus, Dn has a nontrivial proper normal subgroup
and hence Dn is not simple.

EXERCISE 6(d)

1. State whether the following are true or false and substantiate your answers.
(i) The Cauchy index of a permutation f is equal to the number of elements
in the support of f.
(ii) The order of A5 is even.
(iii) A3 is an abelian group.
(iv) A4 is an abelian group.
(v) |A5| 5 120
(vi) A2 is trivial.

Q001-Algebra-111001_CH 06.indd 36 9/16/2011 10:05:32 AM

 Permutation Groups   6-37

(vii) The Cauchy index of any permutation is even.

(viii) The set of all odd permutations is a subgroup of Sn.
(ix) CI(f o g) 5 CI(f ) 1 CI(g) for all f and g  Sn.
(x) For any n . 2, An has a subgroup of order 3.

2. Determine the Cauchy index of each of the following permutation.

1 2 3 4 5 6 7 8 9 
(i)  
3 4 2 1 5 7 8 9 6 
1 2 3 4 5 6 7 8 9 
(ii)  
3 4 6 7 1 9 8 2 5 
(iii) (3 5 2 6 4) o (4 5 6 7 3) o (7 2 8 3 9)
(iv) (7 8 3 4 9 6) o (5 6 7 3) o (4 2 6 8)

3. Which of the following are even permutations?

1 2 3 4 5 6 7 8 
(i)  
7 8 6 5 3 4 2 1 
1 2 3 4 5 6 7 8 9 
(ii)  
9 8 7 6 5 4 3 2 1 
(iii) (2 4 8 6) o (4 3 2 6 1) o (8 5 3)
(iv) (8 4 6 5 3) o (7 8 9 4 5 6) o (1 2 3 4 5 6)
4. Prove that the Cauchy index of any permutation f is equal to that of f21.
5. Prove that CI(f ) is a nonnegative integer for any f in Sn.
6. For any permutations f and g in Sn, prove that that CI(f ) 5 CI(g o f o g21).
7. For any f  Sn, prove that CI(f ) 5 0 ⇔ f is the identity.
8. Prove that CI(f ) 5 1 if and only if f is a transposition.
9. Determine all the permutations in S4 whose Cauchy index is two.
10. Determine all the elements in S6 whose Cauchy index is one.
11. Find all the positive integers n for which An is a cyclic group.
12. Prove that An is abelian if and only if n , 4.
13. List all odd permutations in S3 and S4.
14. For any n . 1, prove that the number of odd permutations in Sn is equal to that
of even permutations in Sn.
15. How many odd permutations are there in S8?
16. Show that A8 contains an element of order 15.

Q001-Algebra-111001_CH 06.indd 37 9/16/2011 10:05:33 AM

 6-38   Algebra – Abstract and Modern

17. What is the maximum order of any element in A10?

18. Prove that f 2 is even for all permutations f in Sn.

19. Find all maximal subgroups M of Sn such that f 2  M for all f  Sn.

20. Prove that there is no proper subgroup of Sn containing An properly.

21. Let H be a subgroup of Sn containing an odd permutation. Then prove that exactly
half of the number of elements in H are even.

22. Prove that the order of any subgroup of Sn containing an odd permutation is
even.

23. If f is an odd permutation in Sn, then prove that fAn is the set of all odd permuta-
tions in Sn.

24. Prove that any element of order 5 in S6 is a 5-cycle.

25. Prove that f o g o f 21 o g21 is even for all f and g  Sn.

26. Prove that there is no subgroup of order 6 in A4. What does this say about the
Lagrange’s theorem.

27. How many elements of order 5 are there in S7?

28. Determine the centralizers of (2 4 1) and (1 2) o (3 4) in A4.

29 If a  S7 such that a4 5 (2 1 4 3 5 6 7), then what is a?

30. Prove that (1 2 3 4) is not a product of 3-cycles in Sn for all n $ 4.

31. If f 5 (1 2 3) o (1 4 5), then express f  99 as a product of disjoint cycles.

32. If f 5 (1 3 5 7 9 8 6) o (2 4 10), what is the smallest positive integer n for which

f n 5 f 25?

33. Prove that {f  S6 : f (3) 5 3 and f (5) 5 5} is a subgroup of S6. What is its
order?

34. How many elements of order 7 are there in A8?

35. If f 5 (9 7 5 3 1) o (6 4 2) o (8 10) and f m is a 5-cycle, then what can we say

about m?

36. If f is a 10-cycle, then find all the integers m between 2 and 10 for which f m is
also a 10-cycle.

37. For any f and g in Sn, prove that g is even if and only if f o g o f 21 is even.

38. Prove that the set of all odd permutations in Sn is a coset of An in Sn.

39. Prove that the centre of the group Sn is trivial for any n # 3.

Q001-Algebra-111001_CH 06.indd 38 9/16/2011 10:05:33 AM

 Permutation Groups   6-39

40. How many 3-cycles are there in S5?

41. For any 1 , r # n, derive a formula for the number of r-cycles in Sn.

42. How many 4-cycles are there in A8.

43. For any 1 , r # n, derive a formula for the number of r-cycles in An.

44. Let f  Sn, such that the order of f in the group Sn is odd. Prove that f is an even
permutation.

45. Let n  3. Let G be the multiplicative group of matrices over complex numbers
generated by
e 2  i n 0  0 1
A    and B   .
 0 e 
2 i n 1 0

Prove that G is isomorphic with the dihedral group Dn of degree n.

46. If a is the generator of order n in Dn, prove that ,a. is normal in Dn and
Dn/,a.  Z2.

47. List all the normal subgroups of Dn.

48. For any n . 1, prove that the alternating group An is the only subgroup of index
2 in Sn.

49. Determine all the subgroups of A4.

50. Prove that Z(Dn), the centre of the dihedral group of degree n, is of the order
1 or 2 according as n is odd or even.

Q001-Algebra-111001_CH 06.indd 39 9/16/2011 10:05:33 AM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 06.indd 40 9/16/2011 10:05:33 AM

 7
Group Actions on Sets
7.1 Action of a Group on a Set
7.2 Orbits and Stabilizers
7.3 Certain Counting Techniques
7.4 Cauchy and Sylow Theorems

Before the concept of an abstract group took its present shape, the theory of
groups dealt only with permutation groups. Abstract groups were introduced
much later in order to focus attention on those properties of permutation
groups that concern the resultant composition (the binary operation in the
permutation groups) only and do not refer to the set on which the permuta-
tions act. However, we have seen that any group can be identified (isomor-
phic) with a group of permutations on some set. Switching back from the
abstract point of view to the concrete case of a permutation group is often
useful in the abstract theory. The use of permutation groups provides cer-
tain counting techniques which play an important role in the theory of finite
groups. In this chapter, we provide a procedure for passing from the abstract
point of view to the concrete case of permutations, by introducing the con-
cept of ‘a group acting on a set’ and develop certain counting techniques in
finite groups.

7.1 ACTION OF A GROUP ON A SET

Cayley’s theorem states that any group is isomorphic with a subgroup of the
group S(X) of permutations on some set X (bijections of X onto itself). Sup-
pose that G is a subgroup of S(X). Then, for any a  G and x  X, there cor-
responds an element a(x) of X and this correspondence satisfies the following
properties.

Q001-Algebra-111001_CH 07.indd 1 9/16/2011 10:04:34 AM

 7-2   Algebra – Abstract and Modern

(i) e(x) 5 x
(ii) (a ? b)(x) 5 a(b(x))
for all x  X and a and b  G, where e is the identity in G. This is abstracted
in the following definition.

Definition 7.1.1. Let G be a group and X be any nonempty set. A mapping

 : G 3 X → X is called an action of G on X if it satisfies the following
­properties.
1. (e, x) 5 x for all x  X, where e is the identity in the group G.
2. (ab, x) 5 (a, (b, x)) for all x  X and a, b  G.
We say that G acts on X if there is an action of G on X.

Example 7.1.1
1. Let X be a nonempty set and G be a subgroup of the group S(X) of per-
mutations on X (bijections of X onto itself). Define
 : G 3 X → X by (a, x) 5 a(x)
for any a  G and x  X. Then, it can be easily verified that  is an action
of G on X. Note that the identity in G is the identity mapping and the
binary operation in the group G is the composition of mappings. This
action is called the natural action of G.
2. Let G be a group and X be the set G itself. Define  : G 3 X → X by (a,
x) 5 a  x, where a  x is the product of a and x in the group G. Then
clearly  is an action of G on itself and is called the action of G on itself
by left translation. The action of an element a in G on an element x in X
(5 G) is simply multiplying x by a on the left.
3. Again let G be a group and X be the set G. Define  : G 3 X → X by
(a, x) 5 xa21. Then,
(e, x) 5 xe21 5 xe 5 x
and (ab, x) 5 x(ab)21 5 x(b21a21) 5 (xb21)a21 5 (a, (b, x))

for all x  X and a, b  G. Therefore,  is an action of G on itself and is

called the action by right translation.
4. By clubbing the above two actions of a group G on itself, we get another
important action of G. Let G be a group and X be the set G itself.
Define
 : G 3 X → X by (a, x) 5 axa21

Q001-Algebra-111001_CH 07.indd 2 9/16/2011 10:04:34 AM

 Group Actions on Sets   7-3

for all a  G and x  X (5 G). axa21 is called the conjugate of x cor-

responding to a. For any a and b  G and x  X(5 G), we have

(e, x) 5 exe21 5 x
and   (ab, x) 5 (ab)x(ab)21 5 a(bxb21)a21 5 (a, (b, x)).

Therefore,  is an action of G on itself and is called the action by conju-

gation.
5. Let H be a subgroup of a group G and X be the set of all left cosets of H
in G. Define

 : G 3 X → X by (a, xH) 5 axH

for all a  G and xH  X with x  G. First note that  is well defined; for

          xH 5 yH ⇒ x21y  H
⇒ (ax)21(ay) 5 x21a21ay 5 x21y  H
⇒ axH 5 ayH

Then,  is an action of G on the set of left cosets of H in G.

6. Let G be a group and X be the set of all subgroups of G. Define  : G 3
X → X by

(a, H) 5 aHa21 for any a  G and H  X.

Then,  is an action of G on the set of all subgroups of G.

As in the case of the binary operation in a group, we simply write ax
for (a, x), where  is an action of G on X, a  G and x  X. This is only
for simplicity and convenience. The defining conditions for an action can be
rewritten as

1. ex 5 x for all x  X
2. (ab)x 5 a(bx) for all a, b  G and x  X.

The condition (2) is not the associative law, for we are not dealing with a
binary operation on a set. a and b are elements of the group G and x is an ele-
ment of X. There should not be any confusion with this notation. One should
understand as per the context.
There may be several actions of the same group on the same set, as given
in the examples (2), (3) and (4) above. In the following, we establish an inter-
relation between the actions of a given group G on a given set X and the
homomorphisms of G into the group S(X) of permutations on X.

Q001-Algebra-111001_CH 07.indd 3 9/16/2011 10:04:35 AM

 7-4   Algebra – Abstract and Modern

Theorem 7.1.1. Let  be an action of a group G on a nonempty set X. For

each a  G, define

f(a) : X → X by f(a)(x) 5 (a, x)

for any x  X. Then, f(a) is a permutation on X and f defines a homo-

morphism of G into S(X). Further,   f is a bijection of the set of all
actions of G on X onto the set Hom(G, S(X)) of homomorphisms of G
into S(X).

Proof: First note that, for any a  G and x, y  X,

      f(a)(x) 5 f(a)(y) ⇒ (a, x) 5 (a, y)

       ⇒ (a21, (a, x)) 5 (a21, (a, y))
    ⇒ (a21a, x) 5 (a21a, y)
⇒ (e, x) 5 (e, y)
             ⇒ x 5 y.

Therefore, f(a) : X → X is an injection. Also, for any y  X, we have (a21, y)

 X and

f(a)((a21, y)) 5 (a, (a21, y)) 5 (aa21, y) 5 (e, y) 5 y

and hence f is a surjection also. Therefore, f(a) is a permutation on X; that

is, f(a)  S(X) and hence f can be considered as a mapping of G into S(X).
Also, for any a and b  G, we have

f(ab)(x) 5 (ab, x) 5 (a, (b, x)) 5 (f(a) ? f(b))(x)

for all x  X and therefore f(a, b) 5 f(a) ? f(b). This means that f is a
homomorphism of G into S(X). If 1 and 2 are two actions of G on X, such
that f 1  f 2 then for any (a, x)  G 3 X,

1 ( a, x )  f 1 ( a)( x )  f 2 ( a)( x ) 2 ( a, x )

and hence 1 5 2. Thus,   f is an injection.

Next, let a  Hom(G, S(X)) and define  : G 3 X → X by

(a, x) 5 a(a)(x) for all a  G and x  X.

Then, it can be verified that  is an action of G on X and f 5 a. Thus,

  f is a bijection of the set of actions of G on X onto the set Hom(G, S(X))
of homomorphisms of G into S(X). b

Q001-Algebra-111001_CH 07.indd 4 9/16/2011 10:04:35 AM

 Group Actions on Sets   7-5

Definition 7.1.2. For any action  of a group G on a set X, the homomor-

phism f : G → S(X) defined above is called the homomorphism associated
with the action . If f is an injection, then  is called an effective action of G
on X and, in this case, we say that G acts on X effectively. The kernel of f is
called the kernel of the action .

That is, ker  5 {a  G : f(a) 5 the identity in S(X)}

5 {a  G : f(a)(x) 5 x for all x  X}
5 {a  G : (a, x) 5 x for all x  X}

Therefore, an action  is effective if and only if ker  5 {e}; that is, e is the
only element a in G such that (a, x) 5 x for all x  X. For example, if  is
the action of G on itself by left translation (see Example 7.1.1 (2)), then  is
effective. On the other hand, if  is the action of G on itself by conjugation
(that is, (a, x) 5 axa21), then  is effective if and only if the centre Z(G) of
G is trivial, since

Z(G) 5 {a  G : ax 5 xa for all x  G}

5 {a  G : axa21 5 x for all x  G}
5 {a  G : (a, x) 5 x for all x  G}
5 ker .

Since a group G is abelian (commutative) if and only if Z(G) 5 G, we can

consider the effectiveness of the conjugacy action as a measure of the com-
mutativity of the group.

Worked Exercise 7.1.1. Let X be the set of all complex number with unit
modulus and G be the additive group of real numbers. Define  : G 3 X → X
by (a, x) 5 eiax for any a  G and x  X. Then prove that  is an action of
G on X. Is  effective?

Answer: Note the 0 is the identity in G (5 (R, 1)) and that the usual addition
1 is the binary operation on G.

(0, x) 5 ei0x 5 1x 5 x for all x  X.

Also, for any a and b  G and x  X,

(a1b, x) 5 ei(a1b)x 5 eia(eibx) 5 (a, (b, x)).

Q001-Algebra-111001_CH 07.indd 5 9/16/2011 10:04:35 AM

 7-6   Algebra – Abstract and Modern

Therefore,  is an action of G on X. Note that,

ker  5 {a  G : (a, x) 5 x for all x  X}

5 {a  G : eiax 5 x for all x  X}
5 {a  G : eia 5 1}
5 {2n : n  Z}

Therefore,  is not effective.

Worked Exercise 7.1.2. Let  be an action of a group on a set X. For any

A ⊆ X, define

(a, A) 5 {(a, x) : x  A}.

Then prove that  is an action of G on the power set of X and that ker
 5 ker .

Answer: For any A ⊆ X,

(e, A) 5 {(e, x) : x  A} 5 A
and (a, (b, A)) 5 {(a, y) : y  (b, A)}
5 {(a, (b, x)) : x  A}
5 {(ab, x) : x A}
5 (ab, A)

Therefore,  is an action of G on P(X), the power set of X. For any a  G,

a  ker  ⇔ (a, x) 5 x for all x  X

⇔ (a, A) 5 A for all A ⊆ X
(note that (a, {x}) 5 {(a, x)})
⇔ a  ker .

Therefore, ker  5 ker ; in particular,  is effective if and only if  is

effective.

EXERCISE 7(a)

1. Let G be a group and define  : G 3 G → G by (a, x) 5 xa. Then prove that 

is an action of G on itself if and only if G is abelian.
2. Let G 5 S5 and X 5 {x1, x2, x3, x4, x5} and define  : G 3 X → X by (f, xi) 5 xf(i) for
any f  S5 and 1  i  5. Then prove that  is an action of G on X. Is it effective?

Q001-Algebra-111001_CH 07.indd 6 9/16/2011 10:04:36 AM

 Group Actions on Sets   7-7

3. Let X be the set of vertices 1, 2, 3, 4 of a square and G 5 D4, the dihedral group
of degree 4. Define  : D4 3 X → X by (f, i) 5 f(i). Then prove that  is an action
of D4 on X.

4. Let H be a subgroup of a group G and X be the set of left cosets of H in G. Define

 : G 3 X → X by

(a, xH) 5 axa21H for a  G and xH  X, x  G.

Then prove that  is an action of G on X. Is it effective? What is the Kernel

of ?

5. Let H be a subgroup of finite index n in a group G. Then prove that there is a

homomorphism f : G → Sn whose kernel is ∩ aHa1.
a ∈G

6. If H is a normal subgroup of finite index n in a group G, then prove that G is

isomorphic to a subgroup of Sn.

7. If G is a simple group and H is a proper subgroup of index n in G, then prove that

G is isomorphic to a subgroup of Sn.

8. Let G be the group of symmetries of a cube. Prove that there are nontrivial
actions of G on each of the set of edges, the set of faces, the set of vertices and
the set of diagonals of the cube.

9. Let  be an action of a group on a set X and define 9 : G 3 (X 3 X) →(X 3

X) by
2(a, (x, y)) 5 ((a, x), (a, y))

for any a  G and x and y  X. Then prove that 2 is an action of G on X 3 X

and that 2 is effective if and only if  is effective.

10. Let G be a group and X1 and X2 be nonempty sets, let 1 and 2 be actions of G
on X1 and X2, respectively. Define (132) : G 3 (X1 3 X2) → X1 3 X2 by

(1 3 2)(a, (x1, x2)) 5 (1(a, x1), 2(a, x2))

for any a  G and (x1, x2)  X1 3 X2. Prove that 1 3 2 is an action of G on X1

3 X2 and that

ker(1 3 2) 5 ker 1 ∩ ker 2.

11. Let a group G act on itself by left translation. Prove that the action is effective
and deduce the Cayley’s theorem.

12. Let H be a subgroup of a group G and X be the set of right cosets of H in G.

Prove that (a, Hx)  Hxa21 is an action of G on X which is effective if and only
if H contains no nontrivial normal subgroup of G.

13. Let H be a proper subgroup of a finite group G. Then prove that G  ∪ aHa1.
a∈G

Q001-Algebra-111001_CH 07.indd 7 9/16/2011 10:04:36 AM

 7-8   Algebra – Abstract and Modern

14. Let p be the smallest prime dividing the order of a finite group G. Then prove
that any subgroup of index p in G is normal in G.

15. Derive from the above that any subgroup of index 2 in a group is normal.

16. Prove that any subgroup of order 539 in a group of order 2695 is normal.

17. Let p and q be distinct primes such that p , q. Prove that any subgroup of order
q in a group of order pq is normal.

18. Let G be a group of odd order. Then prove that any subgroup of index 3 is
­normal in G.

7.2 ORBITS AND STABILIZERS

In this section, we introduce two important concepts regarding the actions of
groups on sets. When  is an arbitrary action of a group G on a set X, we sim-
ply write, as agreed earlier, ax for (a, x) for any a  G and x  X, when there
is no ambiguity about the action . Recall that a group is a pair (G, *) where
G is a nonempty set and * is a binary operation on G satisfying certain condi-
tions. However, we use to simply say that G is a group without specifically
mentioning about the binary operation and further we used to write simply ab
for a * b, where a and b  G. Likewise, when there is an action of a group
G on a set X, we simply say that G acts on X without specifically mentioning
the action of G on X. It is understood that is an action (a, x)  ax sending any
pair (a, x) in G 3 X onto the element ax.

Definition 7.2.1. Let a group G act on a set X and x  X. The orbit of x is

defined to be the set

O(x) 5 {ax : a  G}.

Before going to certain examples, we prove an important elementary prop-

erty of orbits in the following theorem.

Theorem 7.2.1. Let a group G act on a set X. Then, the orbits of elements
of X form a partition of X. That is, any two orbits are either equal or disjoint
subsets of X and the union of all orbits is equal to X.

Proof: For any x  X, we have O(x) 5 {ax : a  G}. Clearly O(x) is a subset
of X for each x  X. Also, since

x 5 ex  O(x)

Q001-Algebra-111001_CH 07.indd 8 9/16/2011 10:04:36 AM

 Group Actions on Sets   7-9

We have X  ∪ O( x ). Now, let x and y  X such that the orbits O(x) and
x∈ X
O(y) are not disjoint. Then, choose z  O(x) ∩ O(y). Thus,
z 5 ax 5 by for some a and b  G
and hence x 5 (a21b)y and y 5 (b21a)x so that O(x) ⊆ O(y) and O(y) ⊆ O(x)
and therefore O(x) 5 O(y). Then,
         O(x) ∩ O(y) 5  or O(x) 5 O(y). b
Definition 7.2.2. Let a group G act on a set X and x  X. The stabilizer of x
is defined to be the set
St(x) 5 {a  G : ax 5 x}.
Theorem 7.2.2. Let a group G act on a set X and x  X. Then, the stabilizer
St(x) is a subgroup of G and there is a bijection of the orbit O(x) onto the set
of left cosets of the stabilizer St(x) in G.
Proof: Since ex 5 x, where e is the identity in the group G,
e  St(x) 5 {a  G : ax 5 x}
and hence St(x) is a nonempty subset of G. Also,
       a and b  St(x) ⇒ ax 5 x 5 bx
⇒ (ab)x 5 a(bx) 5 ax 5 x
⇒ ab  St(x)
and a  St(x) ⇒ ax 5 x
⇒ a21x 5 a21(ax) 5 (a21a)x 5 ex 5 x
⇒ a21  St(x).
Therefore, the stabilizer St(x) is a subgroup of G. Next, let A be the set of all
left cosets of St(x) in G. That is,
A 5 {aSt(x) : a  G}.
Define a : O(x) → A by a(ax) 5 aSt(x).
For any a and b  G, we have
ax 5 bx ⇔ a21bx 5 x
⇔ a21b  St(x)
⇔ aSt(x) 5 bSt(x).
Therefore, a is well defined and is an injection. Clearly a is a surjection also.
Thus, a is a bijection of O(x) onto the set of left cosets of St(x) in G. b

Q001-Algebra-111001_CH 07.indd 9 9/16/2011 10:04:36 AM

 7-10   Algebra – Abstract and Modern

Corollary 7.2.1. Let G be a group acting on a finite set X and x  X. Then,

the number of elements in the orbit of x is equal to the index of the stabilizer
|G|
of x in G. That is, |O(x)| 5 iG(St(x)) 5 .
|st( x )|
Now, we shall take up various examples of actions of groups on sets and
determine the orbits and stabilizers of arbitrary elements of X.

Example 7.2.1
1. Let X be any nonempty set and G be any group. Define  : G 3 X →
X by
(a, x) 5 x for all a  G and x  X.
Then clearly  is an action of G on X and is called the trivial action.
Here, for any x  X,
the orbit O(x) 5 {x}
and the stabilizer St(x) 5 G.
2. Consider the action of any group G on itself by left translation (see
Example 7.1.1 (2)). Here, X 5 G and, for any x  X (5 G),
the orbit O(x) 5 {ax : a  G} 5 X 5 G
since any b  G can be written as b 5 (bx21)x, bx21  G.
Also, the stabilizer St(x) 5 {a  G : ax 5 x} 5 {e}.
3. Consider the action of a group on itself by conjugation (see Example
7.1.1 (4)), where  : G 3 G → G is defined by (a, x) 5 axa21, for all a
and x  G. Here, for any x  X,
the orbit O(x) 5 {(a, x) : a  G}
5 {axa21 : a  G}
and the stabilizer St(x) 5 {a  G : (a, x) 5 x}
5 {a  G : axa21 5 x}
5 {a  G : ax 5 xa}
The orbit O(x) is called the conjugacy class of x in G and is usually
denoted by C(x). Also, the stabilizer St(x) is called the centralizer of x in
G and is usually denoted by CentG(x). That is,
C(x) 5 {axa21 : a  G}
and CentG(x) 5 {a  G : ax 5 xa}.
4. Let H be a subgroup of a group G and X be the set of left cosets of H in
G. Define  : G 3 X → X by
 (a, xH) 5 (axa21)H

Q001-Algebra-111001_CH 07.indd 10 9/16/2011 10:04:37 AM

 Group Actions on Sets   7-11

for any a  G and xH  X, x  G. Then,  is an action of G on X. The

orbit of any xH, x  G, is

O(xH) 5 {(a, xH) : a  G}

5 {axa21H : a  G}

and the stabilizer of xH is

St(xH) 5 {a  G : (a, xH) 5 xH}

5 {a  G : axa21H 5 xH}.

Definition 7.2.3. An action of a group G on a set X is said to transitive

if there is only one orbit in X; that is, O(x) 5 X for all x  X or, equiva-
lently O(x) 5 O(y) for all x and y  X. In this case, we say that G acts
transitively on X.
Clearly an action is transitive if and only if, for any x and y  X, there is
an element a in the group G such that ax 5 y; that is, any element of X can be
transformed to any other element of X by means of the action of an element
of the group.

Example 7.2.2
1. The action of a group G on itself by left (right) translation (see Example
7.1.1 (2) and (3)) is transitive, since, for any x and y  G, yx21  G and
y21x  G and
(yx21)x 5 y and x(y21 x)21 5 y.
2. The action of a group G on itself by conjugation (see Example 7.1.1 (4))
is not transitive, in general. This is transitive if and only if there is only
one conjugacy class or, equivalently, any two elements of the group or
conjugates to each other.

A group may act on the same set (or on two different sets) differently. In
the following, we define equivalence of two such actions in a natural way.

Definition 7.2.4. Let a group G act on two sets X and X. These two actions
are said to equivalent if there is a bijection a : X → X such that

a(ax) 5 aa(x)

for all a  G and x  X; that is, if  : G 3 X → X and  : G 3 X → X are

actions of G on X and X, respectively, then  and  are said to be equivalent if

a ? f(a) 5 f(a) ? a

Q001-Algebra-111001_CH 07.indd 11 9/16/2011 10:04:37 AM

 7-12   Algebra – Abstract and Modern

f�(a)
X X

� �

X′ X′
f�′(a)

for all a  G, where f(a) and f(a) are the permutations on X and X cor-
responding to the actions  and , respectively (see Theorem 7.1.1). In other
words,

a((a, x)) 5 (a, a(x))

for all a  G and x  X.

Example 7.2.3. Consider the actions of a group G on itself by left translation

and right translation defined by

(a, x) 5 ax and (a, x) 5 xa21

(see Example 7.1.1 (2) and (3)). Define a : G → G by a(x) 5 x21 for any x  G.
Then, a is a bijection and, for any a and x  G,

a((a, x)) 5 a(ax) 5 (ax)21 5 x21a21 5 (a, a(x)).

Therefore, the actions  and  are equivalent.

In the following, we obtain an internal characterization of transitive
actions. This is an extension of Theorem 7.2.2 which gives us that the orbit
of any element x in X is bijective with the set of left cosets of the stabilizer
of x in G.

Theorem 7.2.3. Let a group G act transitively on a set X, x  X and

H 5 St(x). Then, the action of G on X is equivalent to the action of G on the
set of left cosets of H in G by left translation.

Proof: Let  be the given transitive action on X and Y be the set of left
cosets of H in G. Let  be the action of G on Y by left translation; that is,
 : G 3 Y → Y is defined by

(a, bH) 5 abH for any a  G and bH  Y.

Q001-Algebra-111001_CH 07.indd 12 9/16/2011 10:04:37 AM

 Group Actions on Sets   7-13

Since the action  is transitive, we get that the orbit O(x) 5 X. Let a : X → Y
be the bijection given in the proof of Theorem 7.2.2; that is,

a(ax) 5 aH for any a  G.

Now, for any a  G and y  X, choose b  G such that y 5 bx (since X is the

orbit of x) and we have

a((a, y)) 5 a(ay) 5 a(abx)

5 abH 5 (a, bH)
5 (a, a(bx)) 5 (a, a(y)).

Thus, the actions  and  are equivalent. b

Among the transitive actions, there is a special class, namely primitive
actions, which deserves an emphasis. In the following, we define primitive
actions and characterize these in terms of the stabilizers.

Definition 7.2.5. Let  be an action of a group G on a set X. An equiva-

lence relation c on X is said to be compatible with the action  if, for any
x and y  X,

(x, y)  c ⇒ ((a, x), (a, y))  c for all a  G.

Clearly the whole of X 3 X and the diagonal X5 {(x, x) : x  X} are equiva-
lence relation X which are compatible with every action of G on X. These two
equivalence relations are called trivial relations.

Definition 7.2.6. An action of a group G on a set X is called primitive if

X 3 X and X are the only equivalence relations on X which are compatible
with the action. An action which is not primitive is called imprimitive.

Theorem 7.2.4. Let  and  be equivalent actions of G on X and X, respec-

tively. Then,  is primitive if and only if so is .

Proof: Since  and  are equivalent, there exists a bijection a : X → X

such that

a((a, x)) 5 (a, a(x))

for any a  G and x  X. For any equivalence relation c on X, let

a(c) 5 {(a(x), a(y): (x, y)  c}.

Q001-Algebra-111001_CH 07.indd 13 9/16/2011 10:04:38 AM

 7-14   Algebra – Abstract and Modern

Then, a(c) is an equivalence relation on X and c  a(c) is a one-to-one cor-

respondence between the equivalence relations on X and those on X. Now,
theorem follows from the fact that c is compatible with the action  on X if
and only if a(c) is compatible with the action  on X. b

Theorem 7.2.5. Let  be an action of a group G on a set X. Then,  is imprim-

itive if and only if there exists a proper subset Y of X with |Y| . 1 such that,
for any a  G, either (a, Y) 5 Y or (a, Y) ∩ Y 5 , where (a, Y) 5 {
(a, y) : y  Y}.

Proof: Suppose that  is imprimitive. Then, there exists an equivalence rela-

tion c on X which is compatible with  such that c  X 3 X and c  X.
Choose x  y  X such that (x, y)  c. Put Y 5 the equivalence class of c
containing x. That is, Y 5 c(x) 5 {z  X : (x, z)  c}. Since x  y  Y, |Y| . 1.
Also, since c  X 3 X, Y is a proper subset of X. Now, let a  G and (a, Y)
∩ Y  . Then, choose z  Y such that (a, z)  Y 5 c(x). Since (x, (a, z))
 c and c is compatible with , we get that

((a, x), (a, (a, z)))  c

and hence ((a, x), (a, z))  c so that (x, (a, x))  c. Now, it can be easily
verified that (a, Y) 5 Y. Conversely suppose that there is a proper subset Y of
X such that |Y| . 1 and, for any a  G, either (a, Y) 5 Y or (a, Y) ∩ Y 5 .
From this it follows that, for any a and b  G,

either (a, Y) 5 (b, Y) or (a, Y) ∩ (b, Y) 5 .

Put Z  X  ( ∪ ( a,Y )). Then,  5 {(a, Y) : a  G} ∪ {Z} is a partition

a∈G
of X and the corresponding equivalence relation c on X is compatible with
the action . Recall that

c 5 {(x, y)  X 3 X : both x and y belong to the same set in }.

Since Y 5 (e, Y) is an equivalence class and Y  X, it follows that c 

X 3 X. Also, since |Y| . 1, c  ∆X. Thus,  is imprimitive. b

Theorem 7.2.6. Let  be an action of a group on a set X and define

c 5 {(x, y)  X 3 X : (a, x) 5 y for some a  G}.

Q001-Algebra-111001_CH 07.indd 14 9/16/2011 10:04:38 AM

 Group Actions on Sets   7-15

Then, c is an equivalence relation on X, that is, compatible with the action .

Proof: Note that (x, y)  c ⇔ y  O(x), the orbit of x

⇔ x  O(y), the orbit of y

and hence c is precisely the equivalence relation corresponding to the parti-

tion of X consisting the orbits of elements of X. Now, for any a  G,

(x, y)  c ⇔ (b, x) 5 y for some b  G

⇔ (b, (a, x)) 5 (ab, x) 5 (a, (b, x)) 5 (a, y)
⇔ ( (a, x), (a, y))  c.

Therefore, c is compatible with the action . b

Corollary 7.2.2. If an action  of G on X is primitive, then either  is trivial

(that is, (a, x) 5 x for all a  G and x  X) or  is transitive.

Proof: If  is primitive, then c 5 X or X 3 X and hence all orbits or

singleton sets or there is only one orbit which mean that either  is trivial or
transitive.
In particular, a nontrivial primitive action must be necessarily transitive
and hence the class of nontrivial primitive actions of a group on a set X is a
subclass of the transitive actions of G on X. But in general a transitive action
need not be primitive. For consider the following example.

Example 7.2.4. Let a group G act on itself by left translation. (see Example
7.1.1 (2)) and H be a nontrivial proper subgroup of G. The action  of G on
itself is defined by (a, x) 5 ax, the product of a and x in G. For any x and y
 G, we have y 5 (yx21)x 5 (yx21, x) and yx21  G and therefore the action
 is transitive. Define a relation c on G by

(x, y)  c if and only if x21y  H.

It can be easily seen that c is an equivalence relation on G. Also for any

a  G,

(x, y)  c ⇒ x21y  H
⇒ (ax)21(ay) 5 x21a21ay 5 x21y  G
⇒ (ax, ay)  c

Q001-Algebra-111001_CH 07.indd 15 9/16/2011 10:04:38 AM

 7-16   Algebra – Abstract and Modern

and hence c is compatible with the action . Also since H  {e}, c  H and
since H  G, c  G 3 G. Therefore, the action  is imprimitive.
Note that in the above example, any equivalence class of c is simply
a left coset of H in G. If G has no nontrivial proper subgroups, then the
above action is primitive and vice versa. In other words, the action of G
on itself by left translation is primitive if and only if {e} is a maximal
subgroup of G.
This is generalised in the following theorem. First, let us call a proper sub-
group K of a group G maximal if there is no proper subgroup of G containing
K properly; that is, for any subgroup H of G, K ⊆ H ⊆ G implies that either
H 5 K or H 5 G.

Theorem 7.2.7. Let a group G act transitively on a set X with |X|  2. Then,
the action is primitive if and only if the stabilizer of any x  X is a maximal
subgroup of G.

Proof: For any x  X, let S 5 St(x), the stabilizer of x in G. By Theorem

7.2.3, the given action G on X is equivalent to the action  of G on the set G/S
of left cosets of S in G by left translation.
Here,  : G 3 G/S → G/S is defined by (a, xS) 5 axS for any a  G
and xS  G/S, x  G, where G/S 5 {xS : x  G} (note that G/S is not
the quotient group, unless S is a normal subgroup of G). Therefore, by
Theorem 7.2.4, the given action of G on X is primitive if and only if  is
primitive.
First suppose that S 5 St(x) is not a maximal subgroup of G for some x  X.
Then, choose a subgroup H of G such that S  H  G.
Let Y 5 {xS : x  H}.
Since H  G, Y is a proper subset of G/S. Also, since S  H, Y has atleast two
elements. Further, for any a  G,

(a, Y) ∩ Y   ⇒ xS 5 (a, yS) for some x, y  H

⇒ xS 5 ayS, for some x, y  H
⇒x21ay  S # H, x, y  H
⇒ a  xHy21 5 H
⇒ (a, Y) 5 {axS : x  H} 5 {yS : y  H} 5 Y.

Thus, by Theorem 7.2.5,  is imprimitive and hence the given action of G

on X is imprimitive. Conversely, suppose that the given action of G on X is

Q001-Algebra-111001_CH 07.indd 16 9/16/2011 10:04:38 AM

 Group Actions on Sets   7-17

imprimitive. Then, again by Theorem 7.2.5, there exists a proper subset Y of

X with |Y| . 1 such that, for any a  G,

either aY 5 Y or aY ∩ Y 5 .

Choose x  Y. We shall prove that the stabilizer St(x) is not a maximal sub-
group of G.

Put H 5 {a  G : aY 5 Y}.

Clearly H is a subgroup of G. Also,

a  St(x) ⇒ ax 5 x
⇒ aY ∩ Y  
⇒ aY 5 Y
⇒ a  H.

Therefore, St(x) ⊆ H ⊆ G. Since |Y| . 1, we can choose y  Y such that

y  x. Since the action of G is transitive, there exists a  G such that ax 5
y  x. Now, a  St(x) and, since ax 5 y  aY ∩ Y, it follows that aY 5 Y and
hence a  H. Therefore, St(x)  H. Further, since Y is a proper subset of X,
we can choose z  X such that z  Y. Again by the transitivity of the action,
there exists a  G such that ax 5 z  Y, so that aY  Y and hence a  H.
Therefore, H  G. Now, since

St(x)  H  G,

if follows that St(x) is not a maximal subgroup of G. b

Worked Exercise 7.2.1. Let G be a finite group of prime order. Suppose that
G acts on a set X and x  X such that ax 5 x for some a  e in G. Then prove
that bx 5 x for all b  G.

Answer: Consider the stabilizer St(x), which is a subgroup of G. Since the

order of G is prime, by the Lagrange’s theorem |St(x)| 5 1 or |G| and hence
St(x) 5 {e} or G. If there is a  e, such that ax 5 x, then St(x)  {e} and
hence St(x) 5 G, so that bx 5 x for all b  G.

Q001-Algebra-111001_CH 07.indd 17 9/16/2011 10:04:39 AM

 7-18   Algebra – Abstract and Modern

EXERCISE 7(b)

1. Let H be a subgroup of a group G and

N  ∩ x H x1.
x ∈G

Then prove that N is largest normal subgroup of G contained in H.

2. Prove that the N given in (1) above is kernel of the action of G on the set of left
cosets of H in G by left translation.
3. Prove that a subgroup H of a group contains no nontrivial normal subgroups of
G if and only if the action of G on the set of cosets of H in G by left translation
is effective.
4. Let H be a subgroup of a group G and X be the set of right cosets of H in G. Then
prove that
(a, Hx)  Hxa21
is an action of G on X whose kernel is the largest normal subgroup of G con-
tained in H.
5. Let a group G act on a set X and Y be a subset of X. Let St(Y) 5 {a  G : ay 5 y
for all y  Y}. Prove that St(Y) is a subgroup of G.
6. Let G be the group (R, 1) and X 5 R2 be the two-dimensional Euclidean space.
For any a  G, let ra be the rotation of the plane about the origin through a
radius. Prove that a  ra is a homomorphism of G into S(X) and hence yields an
action of G on X. Is this action effective? Give a geometrical description of the
orbit of a point P in R2. What is the stabilizer of a point P?
7. Let a group G act on a set X. Prove that the action of G on X is effective if and
only if, for any a and b  G,
ax 5 bx for all x  X ⇒ a 5 b.
8. Let a group G act on a set X primitively and N be a nontrivial normal subgroup
of G. Then prove that the induced action of N on X is transitive.
9. Let two groups G1 and G2 act on sets X1 and X2, respectively and X1 ∩ X2 5 .
Define an action of G1 3 G2 on X1 ∪ X2 by
(a1, a2)x1 5 a1x1 and (a1, a2)x2 5 a2x2

for any (a1, a2)  G1 3 G2, x1  X1 and x2  X2. Prove that this is an action,
which is not transitive.

10. An action of a group G on a set X is called doubly transitive if, for any x1, x2, y1
and y2  X, there exists a  G such that ax1 5 y1 and ax2 5 y2. Prove that any
doubly transitive action is primitive. Is the converse true?

Q001-Algebra-111001_CH 07.indd 18 9/16/2011 10:04:39 AM

 Group Actions on Sets   7-19

7.3 CERTAIN COUNTING TECHNIQUES

We have observed in Theorem 7.2.1 that the orbits form a partition of X,
when a group G acts on X and that, in Theorem 7.2.2, we have proved that the
number of elements in any finite orbit is precisely the index of the stabilizer
of the corresponding element. We consolidate these ideas in proving the fol-
lowing theorem.

Theorem 7.3.1. Let G be group acting on a finite set X and O(x1), O(x2), …,
O(xn) be all the distinct orbits in X. Then, the number of elements in X can be
obtained by the formula

n  n 1 
| X |  ∑ | O( xi ) |  | G | ∑ .
i1
 i1 | St( xi ) | 

Proof: Since the orbits form a partition of X, we have

n
∪ O( xi )  X and O( xi ) ∩ O( x j )   for i  j.
i1

X O(x1) O(x2) O(x3) O(xn)

Therefore, the total number of elements in X is equal to the sum of the num-
bers of elements in the orbits. That is,

|X| 5 |O(x1)| 1 |O(x2)| 1 … 1 |O(xn)|

5 iG(St(x1)) 1 … 1 iG(St(xn)) (by Theorem 7.2.2)
|G | |G |
 
| S ( x1 ) | | S ( xn ) |

 n 1 
 | G | ∑ .
 i1 | St( xi ) |  b


We shall apply the above formula to a special action of a group on itself

and derive an important formula for the order of a finite group, namely,

Q001-Algebra-111001_CH 07.indd 19 9/16/2011 10:04:40 AM

 7-20   Algebra – Abstract and Modern

the class equation. First let us recall that (a, x)  axa21 is an action of a group
G on itself and is called the action of G on itself by conjugation. Here, for any
x  G, the orbit of x is simply the conjugate class C(x) of x in G; that is,

O(x) 5 {axa21 : a  G} 5 C(x).

and the stabilizer of x is given by

St(x) 5 {a  G : axa21 5 x}
5 {a  G : ax 5 xa}

which is known as the centralizer of x in G and is denoted by CentG(x). By

­ orollary 7.2.1, we have |C(x)| 5 iG(CentG(x)); that is, the number of elements
C
in the conjugate class of x is equal to the index of the centralizer of x in G.

Theorem 7.3.2 (The class equation). Let G be a finite group. Then,

n
| G |  ∑ iG (Cent G ( xi ))  | Z (G ) |
i1

where Z(G) is the centre of G and x1, …, xn are elements of G such that C(x1),
C(x2), …, C(xn) are all the distinct conjugacy classes, each with more than one
element. This equation is known as the class equations of G.

Proof: Consider the action of G on itself by conjugation. Then, the orbit of x

is the conjugacy class of x and the stabilizer of x is the centralizer of x. Since
the orbits form a partition of G, the conjugacy classes form a partition of G.
Therefore,
m m
| G |  ∑ | C ( yi ) |  ∑ iG (Cent G ( yi )).
i1 i1

Now, we shall distinguish two types of conjugacy class, namely classes each
with only one element and classes each with more than one element. Note
that, for any x  G, the conjugacy class of x is

C(x) 5 {axa21 : a  G}.

Therefore, |C(x)| 5 1 ⇔ C(x) 5 {x}

⇔ axa21 5 x for all a  G
⇔ ax 5 xa for all a  G
⇔ x  Z(G), the centre of G.

Q001-Algebra-111001_CH 07.indd 20 9/16/2011 10:04:40 AM

 Group Actions on Sets   7-21

Therefore, each element of Z(G) contributes a singleton conjugacy class and

vice versa. If C(x1), C(x2), …, C(xn) are all the distinct conjugacy classes each
with more than one element, then

n
| G |  | Z (G ) | ∑ | C ( xi ) |
i1
n
 | Z (G ) | ∑ iG (Cent G ( xi ))
i1

 n 1 
 | Z (G ) |  | G | ∑ . b
 i1 | Cent( xi ) | 


For any subset A of a group G and for any element x in G, the set

xAx21 5 {xax21 : a  A}

is called the conjugate of A corresponding to x. The map (x, A)  xAx21 is an

action of G on the power set of G, with respect to which the orbit of A is

C(A) 5 {xAx21 : x  G}

and the stabilizer of A is

CentG(A) 5 {x  G : xAx21 5 A}
5 {x  G : xA 5 Ax}.

C(A) and CentG(A) are respectively called the conjugacy class of A in P(G)
and the centralizer or normalizer of A in G. Clearly, for any subset A of a
finite group G,

|G |
| C ( A) |  iG (Cent G ( A))  .
| Cent G ( A) |

In the following, we give certain important applications of the class equa-

tion of a finite group G proved in Theorem 7.3.2. The following is a simple
consequence of the discussion made above.

Theorem 7.3.3. Let K be a subgroup of a finite group G. Then, the number

of subgroups of G conjugate to K is equal to the index of the normalizer of
K in G.

Q001-Algebra-111001_CH 07.indd 21 9/16/2011 10:04:40 AM

 7-22   Algebra – Abstract and Modern

Theorem 7.3.4. Let p be a prime number and n be a positive integer. Let G be

a group of order pn. Suppose that G acts on a finite set X and

X0 5 {x  X : ax 5 x for all a  G}.

Then, |X|  |X0| (mod p).

Proof: We have to prove that |X| 2 |X0| is a multiple of p. Observe that the
orbit of an element x  X is a singleton set if and only if x  X0. Therefore,
there are exactly |X0| number of singleton orbits in X. Let O(x1), O(x2), …,
O(xn) be all the distinct orbits each with more than one element. Then, by
Theorem 7.3.1, we have
n n
| X |  | X 0 | ∑ | O( X i ) |  | X 0 | ∑ iG (St( xi )).
i1 i1

Note that the stabilizer St(xi) is a subgroup of G and |G| 5 pn. By the Lagrange’s
theorem, |St(xi)| is a divisor of pn and

|G |
 iG (St( xi ))  | O( xi ) | 1
| St( xi ) |

for each 1  i  n. Therefore, for each 1  i  n,

iG(St(xi)) 5 p ni for some ni . 0.

n n
Therefore, | X |  | X 0 |  p ni = | X 0 | p (  p ni 1 )
i=1 i=1
and thus |X| 2 |X0| is a multiple of p, so that

|X|  |X0| (mod p). b

Theorem 7.3.5. Let G be a group of order pn, where p is a prime and n be a

positive integer. Then, the centre of G is nontrivial; that is, |Z(G)| . 1.

Proof: Consider the action of G on itself by conjugation. Then, by Theorem

7.3.4,

|G|  |G0| (mod p),

where G0 5 {x  G : axa21 5 x for all a  G} 5 Z(G). Therefore, pn 5 |G| 

|Z(G)|(mod p), which implies that |Z(G)| is a multiple of p. Since e  Z(G),
|Z(G)| . 0 and hence |Z(G)| . 1, so that Z(G)  {e}. b

Q001-Algebra-111001_CH 07.indd 22 9/16/2011 10:04:41 AM

 Group Actions on Sets   7-23

Theorem 7.3.6. Let p be a prime number. Then, any group of order p2 is abelian.

Proof: Let G be a group of order p2 and Z(G) its centre. By Theorem 7.3.5,
|Z(G)| . 1 and, by the Lagrange’s theorem |Z(G)| is a divisor of |G| 5 p2.
Therefore, |Z(G)| 5 p or p2. Suppose that |Z(G)| 5 p. Then, Z(G)|  G and
hence we can choose a  G such that a  Z(G). Then, there exists x  G such
that ax  xa. Consider the centralizer of x

CentG(x) 5 {y  G : xy 5 yx}.

CentG(x) is a subgroup of G containing Z(G) properly (since x  CentG(x)

and x  Z(G)). Then follows that |CentG(x)| is a divisor of p2 and |CentG(x)| .
|Z(G)| 5 p. Therefore, |CentG(x)| 5 p2 5 |G| and hence CentG(x) 5 G which is
a contradiction, since a  CentG(x). Therefore, being |Z(G)| 5 p is impossible
and hence |Z(G)| 5 p2, so that Z(G) 5 G and hence G is abelian. b
In the following, we prove a theorem of Burnside which has applica-
tions in combinatories. When a group G act on a set X, then, for any a  G,
define

Xa 5 {x  X : ax 5 x}.

That is, Xa is the set of all elements of X which are fixed by the action of a.
Note that Xe 5 X, where e is the identity in group G.

Theorem 7.3.7 (Burnside’s theorem). Let a finite group G act on a finite set
X and n be the number of orbits in X. Then,

1
n ∑ |X a |.
|G| a∈G

Proof: Consider the set

A 5 {(a, x)  G 3 X : ax 5 x}.

Note that, for any fixed element x in X, the number of pairs (a, x) in A is pre-
cisely equal to the order of the stabilizer St(x). Also, for any fixed element a
in G, the number of pairs (a, x) in A is exactly equal to |Xa|, where Xa 5 {x 
X : ax 5 x}. Therefore, we have

∑ | St( x) |  | A | ∑ | X
x∈ X a∈G
a |⋅

Q001-Algebra-111001_CH 07.indd 23 9/16/2011 10:04:41 AM

 7-24   Algebra – Abstract and Modern

By Corollary 7.2.1,

|G |
| St( x ) | ,
| O( x ) |

where O(x) is the orbit of x. Note that O(x) 5 O(y) for all y  O(x). Since X
is finite, the number of orbits in X is finite. Let O(x1), O(x2), …, O(xn) be all
the distinct orbits in X. Then,

|G |
∑ | St( x) |  ∑ |O( x) |
x∈ X x∈ X
n
1
| G | ∑ ∑
i1 x∈O( xi ) | O( x) |
n
1
| G | ∑ ∑
i1 x∈O( xi ) | O( xi ) |
n
1
 | G | ∑ | O( xi ) |
i1 | O( xi ) |
| G | ⋅ n

Thus, n  |G1 |  | St( x ) | |G1 |  | X a | .  b

x∈ X a∈G

Worked Exercise 7.3.1. Let p be a prime number and n be a positive integer.

Find the number of different necklaces formed by p beads, where the beads
can have any of n different colours.

Answer: Let (Zp, 1) be the additive group of integers modulo p and X be the
set of all possible necklaces. Since there are p beads in each of the necklaces
and each bead can have any of n different colours, |X| 5 np.

xp x1 xi +1

x2 xi +2
i  =
x3

Let Zp act on X as shown in the figure, where the subscripts are modulo p. The
action of any i  Zp on any given necklace yields the same necklace; only
the beads are permuted cyclically. Therefore, the number of orbits in X is
same as the number of different necklaces, which can be computed by using

Q001-Algebra-111001_CH 07.indd 24 9/16/2011 10:04:42 AM

 Group Actions on Sets   7-25

Theorem 7.3.7. First, let us compute Xi 5 {x  X : ix 5 x}, for any i  Zp.

Clearly, X0 5 X and hence |X0| 5 np. For any 0  i  Zp,

  Xi 5 {x  X : ix 5 x},
5 {x  X : (i 1p j)x 5 x for all 0  j  p}
5{x  X : jx 5 x for all j  Zp},

since Zp is a cyclic group of prime order and hence any nonidentity element
generates Zp. Therefore, for and 0  i  Zp, Xi consists those necklaces which
are unchanged by permutation and hence Xi consists of those necklaces in
which all the beads are of same colour. Since we are given with n different
colours, it follows that |Xi| 5 n for all 0  i  Zp. Thus, by the Burnside’s
theorem, the number of different necklaces (the number of orbits in X) is

1 1 p1
∑
| Z p | i∈ p
| X i | ∑| X i |
p i0
1 p
 ( n  ( n  n))
p
1 p
 n  ( p 1)n
p
n p1
 ( n  p 1)
p

Worked Exercise 7.3.2. Let G be a group and a  G such that O(a) . 1. Sup-
pose that G has exactly two conjugacy classes. Then prove that |G| 5 2.

Answer: Let O(a) 5 n . 1. Then, a  e. Since G has exactly two conjugacy

classes and {e} is a singleton conjugacy class, it follows that {e} and C(a) are
the only conjugacy classes. Therefore,

G 2 {e} 5 C(a) 5 {xax21 : x  G}.

Since O(a) 5 O(xax21), it follows that O(b) 5 n for all b  e and, in particular
bn 5 e for all b  e. Now, we prove that n is a prime. Since n . 1, we can
O( a)
choose a prime p dividing n. Then, O( a p )  p  np  n and hence ap 5 e,
so that p 5 O(a) 5 n. Thus, n is a prime. Next, consider a2. If a2  e, then a2
 C(a) and hence a2 5 xax21 for some x  G, so that (a2)m 5 xmax2m for all
m . 0 and

e 5 a2n 5 (xax21)n 5 xnax2n 5 eae21 5 a,

Q001-Algebra-111001_CH 07.indd 25 9/16/2011 10:04:42 AM

 7-26   Algebra – Abstract and Modern

which is a contradiction. Therefore, a2 5 e and O(a) 5 2. Thus, b2 5 e for all

b  G (since O(b) 5 O(a) for all b  e). Therefore, G is abelian and hence
C(a) 5 {a}. Thus, G 5 {e, a} and |G| 5 2.

Worked Exercise 7.3.3. Let p be a prime number and n be a positive integer.

Let G be a group of order pn and N be a nontrivial normal subgroup of G.
Prove that Z(G) ∩ N is nontrivial, whole Z(G) is the centre of G.

Answer: Since N is a nontrivial subgroup of G, |N| . 1 and |N| is a divisor of

|G| 5 pn. Therefore, |N| 5 pm for some m . 0. Also, since N is normal in G,
axa21  N for all x  N and a  G. Therefore, G acts on N by conjugation
(the action is (a, x) → axa21). By Theorem 7.3.4,

|N|  |N0| (mod p)

where N0 5 {x  N : axa21 5 x for all a  G}

5 {x  N : ax 5 xa for all a  G}
5 N ∩ Z(G)

Since p divides both |N| and |N| 2 |N0|, it follows that |N0| is a multiple of p.
Also, since e  Z(G) ∩ N 5 N0, |N0| . 0 and hence |N0|  p . 1. Thus, Z(G)
∩ N is nontrivial.

EXERCISE 7(c)

1. Determine all the distinct conjugacy classes in each of the following and verify
that the number of elements in each conjugacy class is a divisor of the order of
the group
(i) The symmetric group S3 of degree three.
(ii) The alternating group A4 of degree four.
(iii) The symmetric group S4 of degree four.
(iv) The dihedral group D4 of degree four.

2. Find the number of different (distinguishable) dice that can be made by marking
the faces of a cube using one to six dots.

3. How many different tetrahedral dice can be made by marking the faces of a
regular tetrahedron using one to four dots?

4. How many different ways can seven people be seated at a round table, where
there is no distinguishable leader to the table?

5. Find the number of different ways the edges of an equilateral triangle can be
painted if four different colours of paint are available, assuming only one colour
is used on each edge, and the same colour may be used on different edges.

Q001-Algebra-111001_CH 07.indd 26 9/16/2011 10:04:42 AM

 Group Actions on Sets   7-27

6. Repeat Exercise 5 above with the assumption that a different colour is used on
each edge.

7. For any proper subgroup H of a finite group G, prove that G  ∪ aHa1.

a∈G

8. Let H be a proper subgroup of finite index in a group G. Prove that H contains a

normal subgroup N that is of finite index.

9. Let G be a group such that any proper subgroup is contained in a maximal

subgroup of finite index in G and that any two maximal subgroups of G are
conjugate to each other in G. Then prove that G is cyclic.

10. Let N be a normal subgroup of a finite group G such that the order and the index
of N are relatively prime. If a is an element of G, such that O(a) divides |N|, then
prove that a  N.

11. Let G be a group and H 5 {a  G : C(a) is finite}, where C(a) is the conjugacy
class of a. Then prove that H is a subgroup of G.

12. Let N be a normal subgroup of order 3 in a group G such that N  Z(G). Then
prove that G has a subgroup of index 2.

13. Find the number of different necklaces that can be formed with five beads and
two colours.

14. Determine the number of different necklaces that can be formed with six beads
and two colours.

15. Find the number of neckties having n strips (of equal width) of K distinct
colours.

16. Let S be a subset of a group G and

C(S) 5 {aSa21 : a  G}
         and N(S) 5 {a  G : aSa21 5 S}.

Prove that N(S) is a subgroup of G and that C(S) is bijective with the set of left
cosets of N(S) in G. N(S) is called the normalizer of S in G.

17. Let G be a group of order pn, where p is a prime and n  Z1. If A is a proper
subgroup of G, then prove that A is properly contained in the normalizer of
A in G.

18. If G is a group of order pn (p is a prime and n  Z1) and A is a subgroup of order

pn21 in G, then prove that A is normal in G.

19. Prove that any subgroup of order 343 in a group of order 2401 is normal.

20. Find the number of different necklaces formed by 11 beads, where each bead can
have any of the five given different colours.

Q001-Algebra-111001_CH 07.indd 27 9/16/2011 10:04:43 AM

 7-28   Algebra – Abstract and Modern

7.4 CAUCHY AND SYLOW THEOREMS

The Lagrange’s theorem states that, for any finite group G, if d is the order
of a subgroup of G, then d is a divisor of the order of G. The converse
of this is not true. That is, if d is a divisor of the order of G, we may not
find a subgroup of order d in G. For, consider the alternating group A4 of
degree 4 whose order is 12. Even though 6 is a divisor of the order of A4,
there is no subgroup of order 6 in A4. However, in certain special cases, the
converse of the Lagrange’s theorem is true. In particular, when the divisor
d is a prime or a power of a prime, then there always exists a subgroup of
order d. We prove these and certain important consequences of these in
this section.
Recall that any group of prime order is cyclic and hence, for any group G
and for any prime number p, the existence of a subgroup of order p in G is
equivalent to the existence of an element of order p in G.

Theorem 7.4.1 (Cauchy’s Theorem). Let G be a finite group and p be a

prime number such that p divides the order of G. Then, G has an element of
order p.

Proof: Consider the set

X 5 {(x1, x2, …, xp) : xi  G and x1, x2, …, xp 5 e}

Then, |X | 5 |G|p21, since, for any (x1, x2, …, xp21)  G p21, (x1, x2, …, xp21, xp)
 X, where xp 5 (x1, x2, …, xp21)21 and vice versa. Since p divides |G| and
p − 1 . 0, it follows that p divides |X|.
Consider the group Zp of integers modulo p. We shall define an action of
Zp on X as follows: for any x 5 (x1, x2, …, xp)  X and i  Zp 5 {0, 1, 2, …,
p21}, define

i ? (x1, x2, …, xp) 5 (xi11, xi12, …, xp, x1, x2, …, xi).

Since (x1 x2 … xi) (xi11 xi12 … xp) 5 e, it follows that

(xi11 … xp)  (x1 … xp) 5 e

and hence the above defines a mapping of Zp 3 X into X. It can be easily veri-
fied that this is an action of Zp on X. Therefore, by Theorem 7.3.4,

|X|  |X0|(mod p),

Q001-Algebra-111001_CH 07.indd 28 9/16/2011 10:04:43 AM

 Group Actions on Sets   7-29

where X0 5 {x  X : ix 5 x for all i  Zp}

5 {(x1, x2, …, xp)  X : 1 ? (x1, x2, …, xp) 5 (x1, x2, …, xp)}
5 {(x1, x2, …, xp)  X : (x2, x3, …, xp, x1) 5 (x1, x2, …, xp)}
5 {(x , x , …, x )  X : x 5 x 5 … 5 x }
1 2 p 1 2 p

5 {(x, x, …, x) : x  G and xp 5 e}.

Since p divides |X| and |X| 2 |X0|, it follows that p divides |X0|. Also, since (e,
e, …, e)  X0, we get that |X0| . 0. Therefore, |X0| 5 pn for some n  Z1 and,
in particular,

|X0|  p . 1.

Thus, there exists (x, x, …, x)  X0 other then (e, e, …, e) and hence there
exists x  e in G such that xp 5 e. Then, O(x) . 1 and O(x) is a divisor of p.
Since p is prime, it follows that O(x) 5 p. Thus, x is an element of order
p in G. b

Corollary 7.4.1. Let G be a finite group and p be a prime divisor of the order
of G. Then, G has a subgroup of order p.

Proof: If x  G such that O(x) 5 p, then

,x. 5 {e, x, x2, …, xp−1}

is a subgroup of order p in G.

Definition 7.4.1. Let p be any prime number. A group G is called a p-group

if the order of every element of G is a power of p; that is, for any a  G,
O(a) 5 pn for some integer n $ 0.

Theorem 7.4.2. A finite group G is a p-group (where p is a prime) if and only

if |G| 5 pm for some nonnegative integer m.

Proof: Let G be a finite group and p be a prime number. Suppose that |G| 5 pm,
0  m  Z. For any a  G, O(a) divides |G| 5 pm and hence O(a) 5 pm for
some 0  n  m. Therefore, G is a p-group. Conversely suppose that |G|  pm
for any m  0. Then, there exists a prime q  p such that q divides |G|. By
the Cauchy’s theorem, there exists an element a of order q in G and therefore
G is not a p-group. b
In the proof of the above theorem, the finiteness of G is necessary. For any
prime p, there are infinite p-groups. Consider the following example.

Q001-Algebra-111001_CH 07.indd 29 9/16/2011 10:04:43 AM

 7-30   Algebra – Abstract and Modern

Example 7.4.1. Let p be any prime number and

 a 
H   n : a and n ∈ Z and n  0.
 p 

Then, (H, 1) is a group, where 1 is the usual addition of rational numbers.

Any integer a can be expressed as a / p0  H. Therefore, Z is a subgroup of
H. Now, let

G 5 the quotient group H/Z.

Then, for any positive integers n and m,

1 1 1 1
 ∉ Z and hence n  Z  m  Z.
pn pm p p

Therefore, G is an infinite group. Let x be any element of G. Then,

a
x Z for some a and n  Z, n  0.
pn
Since pnx 5 p n (( a / p n )  Z)  a  Z  Z, it follows that O(x) is a divisor of
pn and hence O(x) 5 pm for some 0  m  n. Thus, G is a p-group.

Definition 7.4.2. A subgroup H of a group G is called a p-subgroup of G if it

is a p-group; that is, every element of H is of order pn for some n  0.
Clearly the trivial subgroup {e} is a p-subgroup of any group G, for any
prime p. Also, every subgroup of a p-group is a p-subgroup. In the follow-
ing, we prove an important result on p-subgroups which plays a crucial
role in the proofs of Sylow theorems. For any subgroup H of a group G,
the set

NG(H) 5 {a  G : aHa21 5 H}

is called the normalizer of H in G. Recall that NG(H) is precisely the stabilizer

of H when G acts on the power set P(G) by conjugation. Also, clearly NG(H)
is the largest subgroup of G containing H as normal subgroup; that is, for any
subgroup A of G containing H,

H is normal in A ⇔ A ⊆ NG(H).

Q001-Algebra-111001_CH 07.indd 30 9/16/2011 10:04:44 AM

 Group Actions on Sets   7-31

Theorem 7.4.3. Let G be a finite group and p be a prime number. Let H be a

p-subgroup of G. Then,

iG ( H ) ≡ iNG ( H ) ( H )(mod p),

where iG(H) is the index of H in G.

Proof: Let X be the set of all left cosets of H in G. Then, H acts on X by

left translation; that is, (h, xH)  hxH is an action of H on X. Then, by
Theorem 7.3.4,

iG(H) 5 |X|  |X0|(mod p)

where  X0 5 {xH  X : hxH 5 xH for all h  H}
5 {xH  X : x21hx  H for all h  H}
5 {xH : x  G and x21Hx ⊆ H}
5 {xH : x  G and xHx21 5 H}
5 {xH : x  NG(H)}

Therefore, |X0| 5 iNG ( H ) ( H ), the index of H in NG(H). Thus, iG(H)  iNG ( H ) ( H )

(mod p). b

Corollary 7.4.2. Let H be a p-subgroup of a finite group G such that p divides

iG(H). Then, NG(H)  H.

Proof: By the above theorem,

iG(H)  iNG ( H ) ( H ) (mod p).

Therefore, p divides iG(H) 2 iNG ( H ) ( H ). By the hypothesis, p divides iG(H)

also. Therefore, p divides iNG ( H ) ( H ) also, and hence,

NG (H )
 iNG ( H ) ( H )  ps for some s ∈ Z .
|H |

which implies that |H| , |NG(H)| so that H is a proper subgroup of NG(H).

When pn divides the order of a finite group G, the following theorem guaran-
tees the existence of a subgroup of order pn in G.

Q001-Algebra-111001_CH 07.indd 31 9/16/2011 10:04:45 AM

 7-32   Algebra – Abstract and Modern

Theorem 7.4.4 (Sylow Theorem 2 I). Let G be a finite group, p be a prime

number and n be a nonnegative integer such that pn divides the order of the
group G. Then, G has a subgroup of order pn.

Proof: We shall prove the theorem by induction on n. The theorem is trivial

for n 5 0 and, for n 5 1, the theorem is a consequence of Corollary 7.4.1.
Now, let n . 1 and pn divides |G|. Then, pn21 divides |G| and hence, by induc-
tion; there exists a subgroup H of order pn21 in G. By Theorem 7.4.3,

iG(H)  iNG ( H ) ( H ) (mod p).

Since pn divides |G| 5 |H| ? iG(H) 5 pn21 ? iG(H), it follows that p divides iG(H)
and therefore p divides iNG ( H ) ( H ). Recall that NG(H) 5 {a  G : aHa21 5
H} and hence H is a normal subgroup of NG(H) and therefore we have the
quotient group NG(H)/H whose order is iNG ( H ) ( H ). Now, NG(H)/H is a group
whose order is divisible by p. Therefore, by Corollary 7.4.1, NG(H)/H has a
subgroup K of order p. Then, K5 A/H where A is a subgroup of NG(H) con-
taining H. Now,

|A| 5 |A/H| ? |H| 5 |K| |H| 5 p ? pn21 5 pn

and, since A is a subgroup of NG(H) which is a subgroup of G, it follows that

A is a subgroup of order pn in G. b

Corollary 7.4.3. Let H be a subgroup of order pn21 in a finite group G, where

p is a prime number and n is a positive integer. If pn divides |G|, then there
exists a subgroup A of order pn such that H is a normal subgroup of A.

Proof: In the proof of the above theorem, we have H ⊆ A ⊆ NG(H) and hence
H is a normal subgroup of A (since a  A ⇒ a  NG(H) ⇒ aHa21 5 H).

Corollary 7.4.4. Let G be a finite group and p be a prime number. Then, every
p-subgroup of G is contained in a maximal p-subgroup.

Proof: Follows from the fact that a subgroup H of G is a p-subgroup if and

only if the order of H is a power of p and from the above corollary.

Definition 7.4.3. For any prime number p, a maximal p-subgroup of a finite

group G is called a Sylow p-subgroup of G.
Note that the order of a Sylow p-subgroup of G must be the largest power
of p dividing the order of G. In fact, a subgroup H of G is a Sylow p-subgroup
of G if and only if |H| 5 pn, where pn divides |G| and pn11 does not divide |G|.

Q001-Algebra-111001_CH 07.indd 32 9/16/2011 10:04:45 AM

 Group Actions on Sets   7-33

By the Sylow Theorem 2 I (7.4.4), for any prime p and a finite group
G, maximal p-subgroups exist in G. If p does not divide |G|, then p0(51)
is the largest power of p dividing |G| and hence {e} is the only Sylow p-sub-
group of G.

Example 7.4.2. Consider the alternating group A4 of degree 4. The order of A4

is 12 5 22  31. For any prime p other than 2 and 3, {e} is the Sylow ­p-subgroup
of A4. Also, any subgroup of order 4 is a Sylow 2-subgroup and any subgroup
of order 3 is a Sylow 3-subgroup. Clearly any Sylow ­3-subgroup is a cyclic
subgroup generated by a 3-cycle in S4. Further, any subgroup of order 4 in
A4 is not cyclic, since a 4-cycle is not an even permutation. It can be easily
checked that any Sylow 2-subgroup of A4 must be necessarily of the form
{e, a, b, ab} where each of a and b is a product of two disjoint ­transpositions,
and that there are three Sylow 2-subgroups of A4.
Note that, for any prime p and for any finite group G, a subgroup H of G
is a Sylow p-subgroup if and only if aHa21 is also a Sylow p-subgroup for
every a  G, since H and aHa21 are of same order. Therefore, if H is a Sylow
p-subgroup of G, then any conjugate of H in G is also a Sylow p-subgroup
and, in the following, we prove that {aHa21 : a  G} is the complete list of
Sylow p-subgroups of G.

Theorem 7.4.5 (Sylow Theorem 2 II). Let p be a prime number and G be a

finite group. If S is a Sylow p-subgroup of G and H is a p-subgroup of G, then
H ⊆ aSa21 for some a  G. In particular, any two Sylow p-subgroups of G
are conjugate to each other.

Proof: Let S be a Sylow p-subgroup and H be any p-subgroup of G and let X

be the set of all left cosets of S in G. Since |S| is the largest power of p divid-
ing |G| and |X| 5 iG(S) 5 |G| / |S|, it follows that p does not divide |X|. Now,
H acts on X by left translation; that is, (h, xS)  hxS is an action of H on X.
Then, by Theorem 7.3.4,

iG(S) 5 |X|  |X0|(mod p),

where X0 5 {xS  X and hxS 5 xS for all h  H}
5 {xS : x  G and x21hx  S for all h  H}
5 {xS : x  G and H ⊆ xSx21}.

Since p divides |X| 2 |X0| and p does not divide |X|, it follows that p does not
divide |X0| and hence |X0|  0; that is, X0 is a nonempty set. Thus, there exists x
 G such that H ⊆ xSx21. In particular, if H is also a Sylow p-subgroup of G,
then |H| 5 |S| and H ⊆ xSx21 and hence H 5 xSx21 for some x  G. b

Q001-Algebra-111001_CH 07.indd 33 9/16/2011 10:04:45 AM

 7-34   Algebra – Abstract and Modern

Corollary 7.4.5. Let S be a Sylow p-subgroup of a finite group G. Then, S is

normal in G if and only if S is the only Sylow p-subgroup of G.

Corollary 7.4.6. Let G be a finite abelian group. Then, for each prime num-
ber p, G has a unique Sylow p-subgroup.
Sylow Theorem 2 II describes all Sylow p-subgroups in terms of a given
Sylow p-subgroup. However, it does not give the exact number of Sylow
p-subgroups. In the following, we derive certain formulae to find the exact
number of Sylow p-subgroups.

Theorem 7.4.6 (Sylow Theorem 2 III). Let G be a finite group. For any prime
p, let np be the number of Sylow p-subgroups of G. Then,
1. np divides |G| / pn, where pn is the largest power of p dividing |G|,
2. np divides |G| and
3. np 5 mp 1 1 for some nonnegative integer m.

Proof:
1. Let X be the set of all subgroups of G. Then, G acts on X by conjugation.
Let P be a Sylow p-subgroup of G. By the Sylow Theorem 2 II (7.4.5),
the orbit of P in X is precisely the set of all Sylow p-subgroups of G. We
know that
np 5 |O(P)| 5 i(NG(P)),
where NG(P) 5 {a  G : aPa21 5 P}. Let pn be the largest power of p
dividing |G|. Then, |P| 5 pn and, by Worked Exercise 4.4.1 (1),
|G |
iNG ( P ) ( P ) ⋅ n p  iNG ( P ) ( P ) ⋅ iG ( N G ( P ))  iG ( P ) 
pn
and hence np is a divisor of |G| / pn.
2. This a single consequence of (1).
3. Let Y be the set of all Sylow p-subgroups of G and S  Y. Let S act on Y
by conjugation. Then, by Theorem 7.3.4,
|Y|  |Y0|(mod p),
where Y0 5 {P  Y : aPa21 5 P for all a  S}
5 {P  Y : S ⊆ NG(P)},
where NG(P) 5 {a  G : aPa21 5 P}. We shall prove that Y0 is a single-
ton set. Consider P  Y0. Then,
aPa21 5 P and hence aP 5 Pa for all a  S

Q001-Algebra-111001_CH 07.indd 34 9/16/2011 10:04:46 AM

 Group Actions on Sets   7-35

so that SP 5 PS which implies that SP is a subgroup of G. Also, since S and

P ⊆ NG(P), it follows that S ⊆ SP ⊆ NG(P). Since S and P are p-subgroups
of G. SP is also a p-subgroup of G. By the maximality of S, it follows that
S 5 SP and hence P ⊆ S. Since P and S are subgroups of the same order
(because both of these are Sylow p-subgroups), we get that P 5 S. Thus,
Y0 5 {S} and |Y0| 5 1 and therefore
np 5 |Y|  1(mod p).
That is, np 2 1 is divisible by p or np 5 mp 1 1 for some nonnegative
integer m. b
In the following theorem, we prove that the converse of the Lagrange’s
theorem holds good for finite abelian groups.

Theorem 7.4.7. Let G be a finite abelian group and d be a positive divisor of

|G|. Then, G has a subgroup of order d.

Proof: The theorem is trivial if |G| 5 1. Therefore, we can assume that |G| . 1.
Let us suppose that

|G|  p1r1 p2r2  pkrk ,

where p1, p2, …, pk are distinct primes and r1, r2, …, rk are positive integers.
Then, since d is a divisor of |G|,

d  p1s1 p2s2  pksk ,

where s1, s2, …, sk are integers such that 0  si  ri for all 1  i  k. Now, for
each i, pisi is a divisor of d and d is a divisor of |G| and hence pisi is a divisor
of |G|. Therefore, by Sylow Theorem 2 I (7.4.4) there exists a subgroup Ai of
G such that |Ai| 5 pisi . For any i  j, Ai ∩ Aj is a subgroup of Ai as well as of Aj
and hence, by Lagrange’s theorem |Ai ∩ Aj| is a common divisor of |Ai| ( pisi )
s s
and |Aj| ( p j j ). Since pi and pj are distinct primes, pisi and p j j are relatively
prime and hence |Ai ∩ Aj| 5 1 for all i  j. Also, since G is an abelian group,
AiAj is a subgroup of G and

| Ai || Aj | s
| Ai Aj |  pisi p j j .
| Ai ∩ Aj |

This argument can be extended inductively to prove that

| A1 A2  Ai || A1 || A2 |  | Ai | p1s1 p2s2  pisi

Q001-Algebra-111001_CH 07.indd 35 9/16/2011 10:04:47 AM

 7-36   Algebra – Abstract and Modern

for any 1  i  k. Now, put A 5 A1A2…Ak. Then, A is a subgroup of order

p1s1 p2s2  pksk 5 d in G. b

Worked Exercise 7.4.1. Prove that any group of order 15 is cyclic.

Answer: Let G be a group of order 15. Let n3 and n5 be the number of

Sylow 3-subgroups and Sylow 5-subgroups, respectively. Then, by Sylow
Theorem 2 III (7.4.6),

|G |
n3 divides 5 5 and n3 5 3m 1 1, m  0
3
|G |
n5 divides 5 3 and n5 5 5s 1 1, s  0.
5
These imply that n3 5 1 5 n5. Therefore, there is a unique subgroup A of
order 3 in G (note that 31 is the largest power of 3 dividing |G| and 51 is the
largest power of 5 dividing |G|) and hence A is normal. Similarly, there is a
normal subgroup B of order 5 in G. Since |A| and |B| are relatively prime, we
get that A ∩ B 5 {e}. From this, we have

| A || B |
| AB |   | A || B | 3  5  15  | G |
| A∩ B |

and hence AB 5 G. Any element of G can be uniquely expressed as ab with a

 A and b  B (a1b1 5 a2b2 ⇒ a2 1a1 5 b2 b11  A ∩ B 5 {e} ⇒ a2 1a1 5 e 5
b2 b11 ⇒ a1 5 a2 and b1 5 b2). Also, for any a  A and b  B, a(ba21b21) 5
(aba21)b21  A ∩ B 5 {e} and hence aba21b21 5 e or ab 5 ba. From these, it
can be verified that (a, b)  ab is an isomorphism of A 3 B onto G. Further,
A > Z3 and B > Z5

 G > A 3 B > Z3 3 Z5 > Z15.

Thus, G is cyclic.
The above result is extended to any groups of order pq, where p and q
are primes, p . q and q does not divide p 2 1. In the above result, we have
15 5 5 ? 3 and 3 does not divide 5 2 1.

Worked Exercise 7.4.2. Let G be a group of order pq, where p and q are dis-
tinct primes, p . q and q does not divide p 2 1. Then prove that G is cyclic.

Q001-Algebra-111001_CH 07.indd 36 9/16/2011 10:04:48 AM

 Group Actions on Sets   7-37

Answer: Let np and nq be the number of Sylow p-subgroups and the number
of Sylow q-subgroups, respectively. Then, we have

np 5 mp 1 1, m  0, nq 5 sq11, s  0,

np divides q and nq divides p. Since p . q and np 5 mp 1 1, m  0, it follows

that np 5 1 and hence there exists unique Sylow p-subgroup A in G and this
A must be normal subgroup of order p. Also, since nq divides p and p is a
prime, nq 5 1 or p; but sq 1 1 5 nq  p (otherwise sq11 5 p and q divides
p 2 1, which is a contradiction to the hypothesis). Therefore, nq 5 1 and
hence G has a unique Sylow q-subgroup B, which becomes a normal sub-
group of order q in G. As in 3.4.19, we get that A ∩ B 5 {e} and AB 5 G.
Therefore, as in the above exercise,

G > A 3 B > Zp 3 Zq > Zpq.

Thus, G is cyclic.

Worked Exercise 7.4.3. Prove that there are no simple groups of order 63.

Answer: Let G be a group of order 63 5 32  7. Let n7 be the number of Sylow

7-subgroups in G. Then, n7 5 7m 1 1, m $ 0 and n7 divides 63. From these
two, we can infer that m 5 0 and n7 5 1. Therefore, there is a unique Sylow
7-subgroup H of G. Then, H is a normal subgroup of order 7 in G and hence
H  {e} and H  G. Thus, G is not simple.

Worked Exercise 7.4.4. Let S be a Sylow p-subgroup of a finite group G. Then

prove that NG(NG(S)) 5 NG(S), where NG(S) is the normalizer of S in G.

Answer: We have NG(S) 5 {a  G : aSa21 5 S} and

NG(NG(S)) 5 {a  G : aNG(S)a21 5 NG(S)}.

For simplicity, let N 5 NG(S).

First note that every conjugate of S is a Sylow p-subgroup of G. Also, if H is
a subgroup of G such that aSa21 ⊆ H, then aSa21 is a Sylow p-subgroup of
H. Clearly S is a Sylow p-subgroup of NG(S) 5 N. Further, if T is any Sylow
p-subgroup of N, then T 5 aSa21 for some a  N and hence T 5 S. Therefore,
S is the only Sylow p-subgroup of N. Now, consider

Q001-Algebra-111001_CH 07.indd 37 9/16/2011 10:04:48 AM

 7-38   Algebra – Abstract and Modern

a  NG(N) ⇒ aNa21 5 N
⇒ aSa21 ⊆ aNa21 5 N
⇒ aSa21 5 S
⇒ a  NG(S) 5 N.

Therefore, NG(N) ⊆ N. Since N is always contained in NG(N), it follows that

NG(N) 5 N.

EXERCISE 7(d)

1. State whether each of the following is true or false and substantiate your
answer.
(i) For any prime p and for any finite group G, there is a Sylow p-subgroup
of G.
(ii) The order of a Sylow 3-subgroup of a group of order 108 is 27.
(iii) Any Sylow 3-subgroup of a group of order 54 is normal.
(iv) There exists a subgroup of order 16 in a group of order 216.
(v) Any group of order 159 is simple.
(vi) Any group of order 159 is cyclic.
(vii) A group of prime power order has no Sylow p-subgroups.
(viii) Every p-subgroup of a finite group is a Sylow p-subgroup.
(ix) Any group of order 121 is abelian.
(x) Any group of order 8 is abelian.

2. Determine all the Sylow p-subgroups of the following groups for all the primes p.
(i) Z24, the group of integers modulo 24.
(ii) S3, the symmetric group of degree 3.
(iii) S4, the symmetric group of degree 4.
(iv) A4, the alternating group of degree 4.
(v) Z3 3 Z3.

3. Prove that any group of order 45 has a normal subgroup of order 9.

4. Prove that there is no simple group of order 56.

5. Show that a normal p-subgroup of a finite group is contained in every Sylow

p-subgroup.

6. For any fixed prime p, prove that the intersection of all Sylow p-subgroups of a
group G is a normal subgroup of G.

7. Prove that there are no simple groups of order 255.

Q001-Algebra-111001_CH 07.indd 38 9/16/2011 10:04:48 AM

 Group Actions on Sets   7-39

8. If p is a prime and r and n are positive integers such that n , p, then prove that
there are no simple groups of order pr  n.

9. Let G be a group of order pn, where p is a prime and n  Z1. Prove that there
are normal subgroups Ai for 0  i  n such that |Ai| 5 pi and Ai ⊂ Ai11 for all 0
 i  n.

10. Deduce from above that there are no simple groups of order pn, for any prime p
and n $ 2.

11. Prove that no group of order 30 or 36 or 48 is simple.

12. Show that any group of order 225 is cyclic.

13. Prove that there is exactly one, up to isomorphism, group of order 323.

14. Prove that any group of order 899 or 961 is cyclic.

15. Prove that no group of order 160 is simple.

16. Prove the following in the symmetric group Sn of degree n.

(i) If a 5 (i1 i2 … ir) is an r-cycle, then f ? a ? f21 5 (f(i1) f(i2) … f(ir)), which
is again an r-cycle.
(ii) Any two cycles of same length are conjugate to each other.
(iii) Two permutations f and g in Sn are conjugates to each other if and only if
f 5 a1 ? a2 ? … ? ak and g 5 b1 ? b2 ? … ? bk, where ai’s are disjoint cycles
and bi’s are disjoint cycles such that O(ai) 5 O(bi) for all 1  i  k.
(iv) A finite sequence 0 , r1  r2  …  rk of positive integers is said to be
a partition of n if r1 1 r2 1 … 1 rk 5 n. Then, the number of conjugate
classes in Sn is equal to the number of partitions of n.

17. Determine all the conjugacy classes in S4 and write down the class equation of S4.

18. Prove that the centre of Sn is trivial for any n . 2.

19. Let G be a group of order 341. Prove that any subgroup of order 31 is normal in G.

20. Let p be a prime and N be a normal subgroup of a group G. Prove that G is a

p-group if and only if both N and G/N are p-groups.

21. Let N be a normal subgroup of order p in a p-group G, where p is prime. Then

prove that N is contained in the centre of G.

22. Let p be a prime and n  Z1 such that p . n. Prove that any subgroup of order
p in a group G of order pn is normal in G.

23. If a group G contains a proper subgroup of finite index, then prove that G con-
tains a proper normal subgroup of finite index.

24. Give an example of an infinite 7-group.

Q001-Algebra-111001_CH 07.indd 39 9/16/2011 10:04:48 AM

 7-40   Algebra – Abstract and Modern

25. Let G be an infinite p-group. Then prove that either G has a subgroup of order
pn for each positive integer n or there exists a positive integer m such that every
finite subgroup of G is of order  pm.

26. Let f be an endomorphism of a finite group G. If S is a normal Sylow p-subgroup

of G, then prove that f(a)  S for all a  S.

27. Let p and q be distinct primes and p . q. Prove that any group of order pnq con-
tains a unique normal subgroup of index q.

28. Prove that any finite abelian group of square-free order is cyclic (an integer is
said to be square-free if it is not divisible by m2 for any integer m . 1).

29. Let G be a finite abelian group and p be a prime number. Let

Sp 5 {a  G : O(a) 5 pr for some r  0}.
Then prove that Sp is the unique Sylow p subgroup of G.

30. Use the above to find all Sylow p-subgroups of Z30, for each prime p.

Q001-Algebra-111001_CH 07.indd 40 9/16/2011 10:04:49 AM

 8
Structure Theory
of Groups
8.1 Direct Products
8.2 Finitely Generated Abelian Groups
8.3 Invariants of Finite Abelian Groups
8.4 Groups of Small Order

It is well known that any cyclic group is abelian and the product of any class
of abelian groups is abelian. In this chapter, we prove the celebrated theo-
rem known as the Fundamental Theorem of finitely generated abelian groups
which states that any finitely generated abelian group is a product of finite
number of cyclic groups. This amounts to saying that the cyclic groups are
like ‘building blocks’ for the finite or finitely generated abelian groups. Since
any cyclic group is isomorphic to the group Z of integers or the group Zn
of integers modulo n for some positive integers, the Fundamental Theorem
implies that any finitely generated abelian group is isomorphic to the product
of a finite number of copies of Z and Zn’s. This facilitates to a great extent
the study of finitely generated abelian groups and, in particular, finite abelian
groups. In fact, we derive a precise formula for the number of abelian groups
of a given order n in terms of the partitions of n.

8.1 DIRECT PRODUCTS

It is well known that, for any groups G1, G2, …, Gn, the Cartesian product
G 5 G1 3 G2 3 … 3 Gn can be made into a group by defining

(a1, a2, …, an), (b1, b2, …, bn) 5 (a1b1, a2b2, …, anbn)

Q001-Algebra-111001_CH 08.indd 1 9/16/2011 10:03:38 AM

 8-2   Algebra – Abstract and Modern

for any (a1, a2, …, an) and (b1, b2, …, bn)  G. In this group, (e1, e2, …, en)
is the identity, where ei is the identity in Gi and, for any (a1, a2, …, an)  G,
( a11 , a2 1 , …, an 1 ) is the inverse of (a1, a2, …, an), where ai−1 is the inverse
of ai in Gi. This group G is called then direct product, or simply, the product
of G1, G2, …, Gn and is denoted by  Gi or G1  G2  Gn . If a group H
i1
is isomorphic to G1 3 G2 3 … 3 Gn, then H is said to be decomposed into
product of groups G1 3 G2 3 … 3 Gn. In this section, we obtain equivalent
conditions for the decompositions of a group G into products of groups in
terms of normal subgroups of G and the corresponding quotient groups.
If A and B are normal subgroups of a group G such that AB 5 G and
A ∩ B 5 {e}, then we have proved (see 7. …) that the map (a, b)  ab is
an isomorphism of A 3 B onto G. This is extended further in the following
theorem.

Theorem 8.1.1. Let G, G1, G2, …, Gn be groups. Then, G  G1  G2  Gn

if and only if there exist normal subgroups N1, N2, …, Nn of G satisfying the
following:
1. N1 N2 … Nn 5 G.
2. For each 1 # i # n, Ni ∩ (N1 … Ni–1Ni11 … Nn) 5 {e}.
3. N i  Gi for each 1  i  n.

Proof: Let H 5 G1 3 G2 3 … 3 Gn. For each 1 # i # n, let

Mi 5 {(a1, a2, …, an)  H : aj 5 ej for all j  i}.

Then, it can be easily verified that each Mi is a subgroup of H. For any

a 5 (a1, …, an)  Mi and x 5 (x1, …, xn)  H, we have,

( xax−1 ) j  x j a j x−j 1  x j e j x−j 1  e j for all j  i

and hence xax–1  Mi. Therefore, Mi is a normal subgroup of H. Further, any

x 5 (x1, x2, …, xn)  H can be expressed as

x 5 (x1, e2, …, en)(e1, x2, e3, …, en) … (e1, e2, …, en–1, xn)

and therefore M1M2…Mn 5 H. Also, for any 1 # i # n, if x ∈ M i ∩ ( M1  M i1 M i1  M n ),

x ∈ M i ∩ ( M1  M i1 M i1  M n ), then xj 5 ej for all j  i, and x  xi  xi1 xi1  xn for some
xj  Mj, j  i, so that xi  ( x1 )i ( xi1 )i ( xi1 )i  ( xn )i  ei  ei ei  ei  ei and
therefore x 5 (e1, e2, …, en) 5 e, the identity in H. Thus,

Hi ∩ (H1…Hi–1 Hi11…Hn) 5{e} for each 1 # i # n.

Q001-Algebra-111001_CH 08.indd 2 9/16/2011 10:03:39 AM

 Structure Theory of Groups    8-3

Also, x  xi is an isomorphism of Mi onto Gi and hence M i  Gi .

Now suppose that G  G1  G2  Gn and f is an isomorphism of G onto
G1 3 G2 3 … 3 Gn. By taking Ni 5 f–1(Mi) for each 1 # i # n, it follows that
N1, N2, …, Nn satisfy all the three required conditions.
Conversely, suppose that N1, N2, …, Nn be normal subgroups of G such that
1. N1 N2…Nn 5 G.
2. N i ∩ ( N1  N i1 N i1  N n )  {e} for each 1  i  n.
3. N i  Gi for each 1  i  n.

Define g : N1 3 N2 3 … 3 Nn → G by

g(a1, a2, …, an) 5 a1a2…an.

We shall prove that g is an isomorphism so that

G  N1  N 2  N n  G1  G2  Gn .

First observe that, for any i  j and a  Ni and b  Nj, we have

( ab)( a1b1 )  ( aba1 )b1  a(ba1b1 ) ∈ N i ∩ N j  {e},

since Ni and Nj are normal subgroups of G and

N i ∩ N j ⊆ N i ∩ ( N1  N i1 N i1  N n )  {e}.

Therefore, ab(ba)–1 5 ab(a–1b–1) 5 e so that ab 5 ba. Now, for any (a1, …, an)
and (b1, …, bn)  N1 3 … 3 Nn,

g (( a1 , …, an )(b1 , …, bn ))  g ( a1b1 , …, an bn )
 ( a1b1 )( a2 b2 )  ( an bn )
 a1 (b1a2 )(b2 a3 )  (bn1an )bn
 a1a2  an b1b2 bn
 g ( a1 , …, an ) g (b1 , …, bn ).

Thus, g is a homomorphism of G into N1 3 … 3 Nn. For any (a1, a2, …, an)

 N1 3 … 3 Nn, we have

g(a1, …, an) 5 e ⇒ a1 a2 … an 5 e

Q001-Algebra-111001_CH 08.indd 3 9/16/2011 10:03:40 AM

 8-4   Algebra – Abstract and Modern

⇒ a2…an 5 a11  N1 ∩ (N2…Nn) 5{e}

⇒ a1 5 e 5 a2…an
⇒ a1 5 e and a2 1 5 a3…an  N2 ∩ (N3…Nn) 5 {e}
⇒a 5e5a 5…5a
1 2 n

⇒ (a1, …, an) 5 (e1, …, en)

and hence ker g 5 {e} so that g is an injection. Since G 5 N1N2…Nn, it follows

that g is a surjection also. Thus, g is an isomorphism of N1 3 … 3 Nn onto G.
Therefore,

G1 3 G2 3 … 3 Gn  N1 3 N2 3 … 3 Nn  G. b

Corollary 8.1.1. Let G, G1 and G2 be groups. Then, G  G1 3 G2 if and only

if there exist normal subgroups N1 and N2 of G such that G 5 N1N2, N1 ∩ N2
5 {e}, G1  N1 and G2  N2.

Corollary 8.1.2. Let N1, N2, …, Nn be subgroups of a group G. Then, the map-
ping (a1, a2, …, an)  a1a2…an is an isomorphism of N1 3 N2 3 … 3 Nn onto
G if and only if the following are satisfied.
1. Each Ni is a normal subgroup of G.
2. N1N2 … Nn 5 G.
3. Ni ∩ (N1…Ni–1 Ni11…Nn) 5 {e} for each 1 # i # n.
We obtain another characterization of a decomposition of a group G in terms
of its quotient groups. First recall that quotient groups of G are precisely (up to
isomorphism) the homomorphic images of G. If N is a normal subgroup of a
group G, then the natural map x  xN is an epimorphism of G onto the quotient
group G/N. If N1 and N2 are two normal subgroups, then clearly the map x 
(xN1, xN2) is a homomorphism of G into G/N1 3 G/N2. We obtain a necessary and
sufficient condition for this map to be a surjection in the following theorem.

Theorem 8.1.2. Let N1 and N2 be normal subgroups of a group G. Define

f : G → G/N1 3 G/N2 by f (x) 5 (xN1, xN2)

for any x  G. Then, f is an epimorphism if and only if N1N2 5 G.

Proof: Clearly f is always a homomorphism. Suppose that N1N2 5 G. Let

(x1N1, x2N2)  G/N1 3 G/N2, where x1 and x2  G 5 N1N2. Then, we can
write

x1 5 a1a2 and x2 5 b1b2,

Q001-Algebra-111001_CH 08.indd 4 9/16/2011 10:03:41 AM

 Structure Theory of Groups    8-5

where a1 and b1  N1 and a2 and b2  N2. Now, put x 5 b1a2. Then, we have

x–1x1 5 (b1a2)–1a1a2 5 a2 1 (b11a1 ) a2 ∈ N1 ,

since b11a1 ∈ N1 and N1 is normal in G. Therefore, xN1 5 x1N1. Also,

x–1x2 5 (b1a2)–1x2 5 a2 1b11 (b1 b2 ) a2 1b2 ∈ N 2

and hence xN2 5 x2N2. Therefore,

f (x) 5 (xN1, xN2) 5 (x1N1, x2N2).

Thus, f is a surjection and hence an epimorphism. Conversely suppose f is

an epimorphism. Let x  G. Consider the element (N1, xN2)  G/N1 3 G/N2.
Since f is a surjection, there exists a  G such that

(aN1, aN2) 5 f (a) 5 (N1, xN2).

Then, aN1 5 N1 and aN2 5 xN2 and hence a  N1 and a–1x  N2 so that
x 5 a(a–1x)  N1N2. Thus, N1N2 5 G. b

Corollary 8.1.3. Let N1 and N2 be normal subgroups of a group G such that

N1N2 5 G. Then,

G N1 ∩ N 2  G N1  G N 2

and in particular, when N1 ∩ N2 5 {e}, G  G/N1 3 G/N2.

Proof: In the above theorem, we have an epimorphism f : G → G/N1 3 G/N2

whose Kernel is given by

ker f 5 {x  G : f (x) 5 (eN1, eN2)

5 {x  G : xN1 5 N1 and xN2 5 N2}
5 N1 ∩ N2

and therefore, by the fundamental theorem of homomorphisms

G/N1 ∩ N2  G/N1 3 G/N2.

These are generalised in the following theorem. b

Q001-Algebra-111001_CH 08.indd 5 9/16/2011 10:03:41 AM

 8-6   Algebra – Abstract and Modern

Theorem 8.1.3. Let N1, N2, …, Nn be normal subgroups of a group G and

f : G → G/N1 3 G/N2 3 … 3 G/Nn be defined by

f (x) 5 (xN1, xN2, …, xNn).

Then, f is an epimorphism if and only if, for each 1 # i # n,

 
N i  ∩ N j   G.
 j≠i 

Proof: Clearly f is a homomorphism. Suppose that f is epimorphism. Fix 1 #

i # n. For any x  G, consider the element

(xN1, …, xNi–1, Ni, xNi11, …, xNn)  G/N1 3 … 3 G/Nn.

Since f is a surjection, there exists a  G such that

(xN1, …, xNi–1, Ni, xNi11, …, Nn) 5 f (a)

5 (aN1, aN2, …, aNn).

Therefore, xNj 5 aNj for all j  i and Ni 5 aNi and hence a  Ni and a–1x
 Nj for all j  i, so that

(
x  a( a−1 x ) ∈ N i . ∩ N j .
j ≠i
)
Thus, G  N i  ( ∩ N j ).
j ≠i
Conversely suppose that N i  ( ∩ N j )  G for each 1 # i # n. Then, NiNj 5 G
j ≠i
for all i  j and

 m−1 
 ∩ N i   N m  G for all 1  m  n
 i1 

(note that, for any subgroups A and B of a group G, AB 5 G if and only if

BA 5 G). We shall use induction on n to prove that f is a surjection. If n 5 1,
it is trivial. If n 5 2, Theorem 8.1.2 gives the result. Let n . 2 and assume
the result for n 2 1. Let

(x1N1, x2N2, …, xnNn)  G/N1 3 G/N2 3 … 3 G/Nn,

Q001-Algebra-111001_CH 08.indd 6 9/16/2011 10:03:42 AM

 Structure Theory of Groups    8-7

where x1, x2, …, xn  G. Then, there exists y  G such that

yNi 5 xiNi for all 1 # i # n 2 1

and hence y–1xi  Ni for all 1 # i # n 2 1.

Now, put M 5 N1 ∩ N2 ∩ … ∩ Nn–1. Then, MNn 5 G and hence, by Theorem
8.1.2, a  (aM, aNn) is an epimorphism of G onto G/M 3 G/Nn. We have
(yM, xnNn)  G/M 3 G/Nn. Therefore, there exists x  G such that

xM 5 yM and xNn 5 xnNn

so that x–1y  M and x–1xn  Nn. For any 1 # i # n 2 1, we have

x–1xi 5 (x–1y)(y–1xi)  Ni for all 1 # i # n 2 1

and hence xNi 5 xiNi for all 1 # i # n, so that

f (x) 5 (xN1, …, xNn) 5 (x1N1, …, xnNn).

Thus, f is a surjection and hence an epimorphism. b

Corollary 8.1.4. Let N1, N2, …, Nn be normal subgroups of a group G and

define

G G G
f: n
→ 
N1 Nn
∩N
i=1
i

by

  n 
f  x  ∩ N i   ( xN1 , , xN n ).
 i =1 

Then, f is an isomorphism if and only if, for each 1 # i # n,

(
N i ⋅ ∩ N j  G.
j ≠i
)

Q001-Algebra-111001_CH 08.indd 7 9/16/2011 10:03:43 AM

 8-8   Algebra – Abstract and Modern

Corollary 8.1.5. Let G, G1, G2, …, Gn be groups. Then, G  G1 3 G2 3

… 3 G if and only if there exist normal subgroups N , N , …, N of G
n 1 2 n
satisfying the following:
1. N ∩ N ∩ … ∩ N 5 {e}.
1 2 n
2. N i ⋅ ( ∩ N j )  G, for all 1 # i # n.
j ≠i
3. G/Ni  Gi, for all 1 # i # n.

Proof: Let H 5 G1 3 G2 3 … 3 Gn and define

pi : H → Gi by pi(x1, x2, …, xn) 5 xi.

Then, it can be easily verified that pi is an epimorphism for each 1 # i # n

and hence H/ker pi  Gi. Put Mi 5 ker pi. Then, Mi is a normal subgroup of H
for each 1 # i # n and H/Mi  Gi. Also, M1 ∩ M2 ∩ … ∩ Mn 5 {e}, where e 5
(e1, e2, …, en) and ei is the identity in Gi. For any 1 # i # n and x 5 (x1, x2,
…, xn)  H, we can write

x 5 (x1, …, xi–1, ei, xi11, …, xn)  (e1, …, ei–1, xi, ei11, .., en)

which is an element in M i ⋅ ( ∩ M j )  H. If f : G → H is an isomorphism and

j ≠i
Ni 5 f–1(Mi) for 1 # i # n, then N1, N2, …, Nn satisfy all the required proper-
ties. Converse follows from Theorem 8.1.3 and Corollary 8.1.4. b

Worked Exercise 8.1.1. Let A1, A2, …, An be subgroups of a group G and

G 5 A1A2…An.
Define f : A1 3 A2 3 … 3 An → G by

f (a1, a2, …, an) 5 a1a2 … an.

Then prove that f is an isomorphism if and only if each Ai is a normal sub-

group of G and any element a  G can be uniquely expressed as a 5 a1a2…
an for some ai  Ai, 1 # i # n.

Answer: Suppose that f is an isomorphism. For each 1 # i # n, let

Bi 5 {(a1, …, an)  A1 3 … 3 An : aj 5 ej for all j  i}.

Then, as in Theorem 8.1.1, we can verify that each Bi is a normal subgroup

of A1 3 … 3 An and f/Bi is an isomorphism of Bi onto Ai. Therefore, Ai is a
normal subgroup of G for each 1 # i # n. Since f is an isomorphism and, in
particular, a bijection, any element of G can be uniquely expressed as a prod-
uct a1 a2…an, where ai Ai.

Q001-Algebra-111001_CH 08.indd 8 9/16/2011 10:03:43 AM

 Structure Theory of Groups    8-9

Conversely, suppose that each Ai is a normal subgroup of G and any element

of G can be uniquely expressed as a1 a2…an, ai  Ai. Then, clearly f is a bijec-
tion. To prove that f is a homomorphism, first observe that Ai ∩ Aj 5 {e} for
any i  j; for, if a  Ai ∩ Aj and i  j, then

f (e, …, e, ath , e, …, e)  f (e, …, e, ath , e, … , e)

i j

and hence a 5 e. From this we get that, for any i  j, ai  Ai and aj  Aj,

ai ( a j ai−1a−j 1 )  ( ai a j ai−1 ) a−j 1 ∈ Ai ∩ Aj  {e}

and therefore aiaj 5 ajai. From this, it follows that f is a homomorphism.

Thus, f is an isomorphism.

Worked Exercise 8.1.2. Let G be a finite nontrivial group such that a2 5 e for
all a  G. Then prove that G  C1 3 C2 3 … 3 Cn, where n . 0 and each Ci
is a cyclic group of order 2 and deduce that |G| 5 2n.

Answer: Since a2 5 e, we have a 5 a21 for all a G and hence ab 5 (ab)–1

5 b21a21 5 ba for all a and b  G. Therefore, G is an abelian group. Since
G is nontrivial, choose a1  e in G and let C1 5 {e, a1}. Then, C1 is a normal
subgroup of G and C1 is a cyclic group of order 2.
If G 5 C1, we are through. Otherwise, choose a2  G such that a2 ∉ C1 and
let C2 5 {e, a2}. Then, C1 ∩ C2 5 {e} and C1C2  C1 3 C2. Also C1  C1 C2 .
Again, if C1C2 5 G, we are through. Otherwise, choose a3  G such that a3
∉ C1C2 and continue the process to get cyclic subgroups C1, C2, C3, …, each
of order 2 and

C1  C1 C2  C1 C2 C3 …

Since G is finite, the process should terminate at a finite stage and then

G  C1 3 C2 3 … 3 Cn.

Worked Exercise 8.1.3. Prove that any group of order p2 where p is a prime,
is isomorphic to either Z p 2 or Zp 3 Zp.

Answer: Let G be a group of order p2, where p is a prime. Then, by

­Theorem 7.3.6, G is abelian and hence every subgroup of G is normal. If G
is cyclic, then by Theorem 7.3.6, G  Z p2 . Suppose that G is not cyclic, then

Q001-Algebra-111001_CH 08.indd 9 9/16/2011 10:03:44 AM

 8-10   Algebra – Abstract and Modern

O(a)  |G| 5 p2 for all a  G. Choose a  e in G and let A 5 ,a.. Then,

O(a) is a divisor of p2 and therefore O(a) 5 p. Since A  G, there exists
b  G such that b ∉ A. Let B 5 ,b.. Then, |A ∩ B| 5 1 or p. If |A ∩ B| 5 p,
then A ∩ B 5 A and hence A ⊆ B so that A 5 B, which is not true. Therefore,
|A ∩ B| 5 1 and hence A ∩ B 5 {e}. Also

| A || B | | A | | B |
| AB | = = = p2 = | G |
| A∩ B | 1

and hence AB 5 G. Therefore, by Corollary 8.1.1,

G  A 3 B  Zp 3 Zp,

since any group of prime order is cyclic.

Worked Exercise 8.1.4. Let G be a cyclic group of order p1r1 p2 r2 … pn rn ,

where p1, p2, …, pn are distinct primes and r1, r2, …, rn are positive integers.
Prove that G is isomorphic to a product of cyclic groups A1, A2, …, An where
| Ai |  pi ri .

Answer: Since G is given to be cyclic, there exists a  G such that G 5 ,a.

r
and O( a)  p1r1 p2 r2 … pn rn . For each 1 # i # n, let mi   p jj and bi  a mi .
j ≠i
O( a)
Then, O(bi )  m  pi ri . Let Ai 5 ,bi.. Then, Ai is a cyclic subgroup of
i
order pi ri . Since p1, p2, …, pn are distinct primes, ( pi ri , mi ) = 1 for each l # i
# n. From this, it follows that

Ai ∩ (A1 … Ai21Ai11 … An) 5 {e}

for each 1 # i # n. Also, A1 A2…An 5 G. Note that G is abelian (being cyclic)

and hence Ai’s are normal subgroups of G. Therefore,

G  A1 3 A2 3 … 3 An.

EXERCISE 8(a)

1. State whether each of the following is true or false and substantiate your
answer.
(i) Any group of order 25 is cyclic.
(ii) If G is a group of order 9 and G is not cyclic, Then, G  Z3 3 Z3.
(iii) Any group of order 121 is abelian.
(iv) Any group of order 8 is abelian.

Q001-Algebra-111001_CH 08.indd 10 9/16/2011 10:03:45 AM

 Structure Theory of Groups    8-11

(v) Any cyclic group of order 180 can be decomposed as a product of non-
trivial groups.
(vi) Z36  Z9 3Z4.
(vii) If G is a group of order 36, than, G  Z9 3 Z4.
(viii)  Z Z .
12Z 4Z 3Z
2. Prove that Z4 cannot be decomposed as a product of groups of order 2.

3. Show that Z8 cannot be decomposed as a product of two nontrivial subgroups.

4. Let A1, A2, …, An be subgroups of a group G such that A1 A2…An 5 G. Prove

that the map f : A1 3 … 3 An → G, defined by f (a1, …, an)5 a1a2…an, is an
isomorphism if and only if each Ai is normal in G and, for any ai  Ai,
a1a2…an 5 e ⇒ a1 5 a2 5 … 5 an 5 e.

5. Let G be a group of order pq, where p and q are distinct primes. If A and B are
normal subgroups of orders p and q, respectively, then prove that G is cyclic.

6. Prove that Z10 is isomorphic to the product of the subgroups A 5 {0, 2, 4, 6, 8}

and B 5 {0, 5}.

7. Let G be a cyclic group of order mn, where m and n are relatively prime positive
integers. Prove that there exist subgroups A and B of orders m and n, respectively
such that G  A 3 B.

8. Is Z 3 Z cyclic?

9. Let G be a finite abelian group of order p1r1 p2 r2 … pn rn , where p1, p2, …, pn are
distinct primes and r1, r2, …, rn are positive integers. For each 1 # i # n, let
Si 5 {a  G : O(a) 5 ps for some 0 # s  Z}.

Prove that each Si is a subgroup of G and G is isomorphic to the product S1 3 S2

3…3S. n

10. Let m  p1r1 p2 r2 … pn rn , where pi’s are distinct primes ri’s are positive integers.
Prove that
Z m  Z pr1  Z pr2  Z prn .
1 2 n

11. A nontrivial group is said to be indecomposable, if it is not isomorphic to the

product of two nontrivial groups. Prove that Zp is indecomposable.

12. Prove that the group of symmetries of a square, the group (Z, 1) and the group
(Q, 1) are all indecomposable.

13. Prove that the group (Zn, 1n) is indecomposable if and only if n 5 pr for some
prime p and r . 0.

Q001-Algebra-111001_CH 08.indd 11 9/16/2011 10:03:46 AM

 8-12   Algebra – Abstract and Modern

14. Let A and B be normal subgroups of a finite group G such that |A| ? |B| 5 |G| and
|A| and |B| are relatively prime. Then prove that G  A 3 B.

15. Let A and B be subgroups of a group G such that G  A 3 B. For any normal
subgroup N of G, prove that either N is contained in the centre of G or N has
nontrivial intersection with A or B.

16. Prove that the symmetric group S3 is indecomposable.

17. Let f : G → G9 be an epimorphism of groups and N be a normal subgroup of G.

If the restriction of f to N is an isomorphism of N onto G9, prove that G  N 3
(ker f ).

18. Let G1, G2, …, Gn be groups and G 5 G1 3 G2 3 … 3 Gn. For any group H,
prove that a mapping f : H → G is a homomorphism if and only if pi ? f : H → Gi
is a homomorphism for each 1 # i # n, where pi : G → Gi is the ith projection.

19. Let A and B be any normal subgroups of a group G. If the natural map f : G →
G (defined be f (x) 5 xA) induces an isomorphism of B onto G/A, then prove
A
that G  A 3 B.

20. Prove the following for any groups G1, G2 and G3:
(i) G1 3 G2  G2 3 G1
(ii) (G1 3 G2) 3 G3  G1 3 (G2 3 G3)
(iii) G1  G2 ⇒ G1 3 G3  G2 3 G3
(iv) G1 3 G2 is abelian ⇔ G1 and G2 are abelian.
(v) G1 3 G2 is cyclic ⇒ G1 and G2 are cyclic.
(vi) The converse of (v) is not true.

8.2 FINITELY GENERATED ABELIAN GROUPS

The study of finitely generated abelian groups and, in particular, finite abe-
lian groups is one of the richest and deepest branches of the whole of group
theory. No other general class of groups is the structure as completely known
or as easily described. The overall strategy in the study of a structure theory
of any algebraic system is to express, in some sense, a complicated algebraic
system in terms of those which are better behaved and whose structure is
well known. For example, the structure of a cyclic group is well known and
in fact, we have proved earlier that any cyclic group is isomorphic to the
additive group Z of integers or to the group Zn of integers modulo n accord-
ing as the group is infinite or finite of order n, respectively. Here, we prove a
fundamental theorem which states that a group is a finitely generated abelian
group if and only if it is isomorphic to the product of a finite number of cyclic
groups.

Q001-Algebra-111001_CH 08.indd 12 9/16/2011 10:03:46 AM

 Structure Theory of Groups    8-13

We first consider the case of finite abelian groups. Let us recall that a
group G is called a p-group, where p is a given prime number, if the order of
any element of G is a power of p and that a finite group is a p-group if and
only if it is of order pn for some nonnegative integer n. The following is a
central topic in the structure theory of finite abelian groups.

Theorem 8.2.1. Any finite abelian group is isomorphic to a product of

p-groups.

Proof: The trivial group {e} is a p-group, for any prime p (since its order is
1 5 p0). Let G be a nontrivial abelian group of order n and n . 1. We can
write

n  p1r1 p2 r2 … pk rk ,

where p1, p2, …, pk are distinct primes and r1, r2, …, rk are positive integers.
From Sylow Theorem I (7.4.4), there exist subgroups A1, A2, …, Ak of G such
that | Ai | pi ri for 1# i # k. Since the group G is abelian, each Ai is a normal
subgroup of G. Also, since pi’s are distinct primes, it can be easily verified
that Ai ∩ Aj 5 {e} (for, the order of Ai ∩ Aj is a common divisor of |Ai| and
|Aj|) for i  j.
Further the order of Ai ∩ (A1…Ai21Ai11…Ak) is a common divisor of
|Ai| 5 (5 piri ) and |A1…Ai21Ai11…Ak| (  p j j ) and, since piri and  p j j are
s s
j ≠i j ≠i
relatively prime, it follows that |Ai ∩ (A1…Ai21Ai11…Ak| 5 1 and hence

Ai ∩ (A1…Ai21Ai11…Ak) 5 {e}.

Also, it can be proved that A1A2 … Ak 5 G.

Therefore, by Theorem 8.1.1,

G  A1 3 A2 3 … 3 Ak

and each Ai is a pi-group (actually Ai is the unique Sylow pi-subgroup of G).

Definition 8.2.1. Let G be a finite abelian group. For any prime p, let

Gp 5 {a  G : O(a) 5 pr for some 0 # r  Z}.

It can be easily verified that Gp is a p-subgroup of G. In fact, Gp is the unique

Sylow p-subgroup of G. The following is a simple consequence of the proof
of the above theorem.

Q001-Algebra-111001_CH 08.indd 13 9/16/2011 10:03:47 AM

 8-14   Algebra – Abstract and Modern

Corollary 8.2.1. Let G be a finite abelian group. Then,

G  G p1  G p2  G pk ,

where p1, p2, …, pk are all the distinct primes dividing the order of G.
The two results proved above reduce the study of arbitrary finite abelian
groups to the study of finite abelian p-groups. The basic result on p-groups
from which the whole structure theory can be pinned down is proved in the
following theorem.

Theorem 8.2.2. Let p be an arbitrarily fixed prime number. Then, any finite
abelian p-group is isomorphic to a product of a finite number of cyclic
p-groups.

Proof: Let G be a finite abelian p-group. Then, |G| 5 pn for some nonnega-
tive integer n. We shall use induction on n. If n 5 0, there is nothing to prove,
since G becomes trivial. If n 5 1, then G is a group of order p, which is a
prime, and hence G is itself a cyclic p-group. Now, let n . 1 and suppose that
the theorem holds good for all groups of order pm with m , n.
Since G is a p-group, the order of any element of G is a power of p. Let a be
an element of maximal order in G and O(a) 5 pk, where k # n. Put H 5 ,a.,
the cyclic subgroup of G generated by a. If k 5 n, then H 5 G and hence G
is itself a cyclic p-group. Suppose that k , n and consider the quotient group
G/H whose order is pn2k and n 2 k , n. By the induction hypothesis,

G/H  G1 3 G2 3 … 3 Gr ,

where Gi, 1# i # r, is a cyclic p-group. By Theorem 8.1.1, there exist sub-

groups A1, A2, …, Ar of G/H such that Ai  Gi for 1 # i # r, A1 A2…Ar 5
G/H and

Ai ∩ (A1…Ai21Ai11…Ar) 5 {H}.

For any 1# i # n (note that H is the identity in G/H).

Therefore, there exist subgroups H1, H2, …, Hr of G such Ai 5 Hi/H for
1# i # r. Now, these subgroups Hi’s satisfy the following:
1. H ⊆ Hi for all 1 # i # r
2. Gi  Hi/H for all 1 # i # r
3. G/H 5 H1/H ? H2/H ? … ? Hr/H
4. Hi ∩ (H1…Hi21Hi11…Hn) 5 H for all 1 # i # n.

Q001-Algebra-111001_CH 08.indd 14 9/16/2011 10:03:47 AM

 Structure Theory of Groups    8-15

Since Gi and hence Hi/H is cyclic, we get a coset biH generating Hi/H. Let |H/
Hi| 5 p ji for each 1 # i # r. Then, the order of biH in G/H is equal to p ji .
In the next step, we produce a representative ci of biH such that the order of ci
in G is equal to the order of biH in G/H. j
For convenience, let us write tempo-
rarily b for bi and j for ji. Since (bH ) p 5 H, we have
j
b p ∈ H  a 
j
and hence b p 5 as for some s. Since kthe order of a is maximal and O( a ) = p k ,
we have O(b) # pk and therefore b p  e which implies that
k− j k− j j k− j k
a sp ( a s ) p  (b p ) p  b p  e.

Therefore, O(a) divides spk2j ; that is, pk divides spk2j and hence pj divides s.
Let s 5 tpj and c 5 ba2t. Then, b21c 5 a2t  ,a. 5 H and
j j j j
c p  (bat ) p b p at p  a s as  e

which implies that O(c) 5 pj.

Restoring the index i, we have the following:
For each 1 # i # r, there is an element ci  G and an integer ni such that

ci H  bi H and cii e.

Let K be the subgroup of G generated by c1, c2, …, cr. We shall prove that
HK 5 G and H ∩ K 5 {e}, so that G  H 3 K.
Consider an arbitrary element x  G. Then, xH  G/H and hence

xH  (b1 H ) n1 (b2 H ) n2 (br H ) nr for some integers ni

 (c1 H ) n1 (c2 H ) n2 (cr H ) nr

 (c1n1 c2n2  crnr ) H .

Therefore, x  (c1n1 c2n2  crnr ) y for some y  H. Thus, x  KH 5 HK.

Thus, HK 5 G.
To prove that H ∩ K 5 {e}, consider an element x in H ∩ K. Since x  K,
there exist integers mi such that

x  c1m1 c2m2  crmr .

Q001-Algebra-111001_CH 08.indd 15 9/16/2011 10:03:48 AM

 8-16   Algebra – Abstract and Modern

Now,

H  xH  (c1 H ) m1 (c2 H ) m2 ( cr H ) mr
 (b1 H ) m1 (b2 H ) m2 (br H ) mr .

From the property (4) of Hi’s above, it follows that

(bi H ) mi  H for all 1  i  r.

Since the order of biH in G/H is equal to the order of ci (which is equal to p ji ),
we get that p ji divides mi and hence (ci ) mi 5 e for all 1 # i # r, so that x 5
e. Thus, H ∩ K 5 {e}. Again by Theorem 8.1.1, we have

G  H 3 K.

From the induction hypothesis, there exist cyclic p-groups K1, K2, …, Kw
such that

K  K1 3 … 3 Kw.

Now, H is also a cyclic p-group and

G  H 3 K  H 3 K1 3 … 3 Kw.

This completes the proof. b

Theorems 8.2.1 and 8.2.2 together yield the following fundamental structure
theorem.

Theorem 8.2.3 (Fundamental structure theorem for finite abelian groups).

Any finite abelian group is isomorphic to a product of cyclic p-groups.
Since any cyclic group of order n is isomorphic to the group Zn of integers
modulo n, we have the following corollary.

Corollary 8.2.2. Let G be any nontrivial finite abelian group. Then,

G  Z pn1  Z pn2  Z pnr ,

1 2 r

where p1, p2, …, pr are (not necessarily distinct) primes and n1, n2, …, nr are
positive integers.
We derive a formula to find the exact number of distinct (nonisomorphic)
abelian groups of a given order n. For example,

Q001-Algebra-111001_CH 08.indd 16 9/16/2011 10:03:49 AM

 Structure Theory of Groups    8-17

Z4 and Z2 3 Z2

are the only distinct abelian groups of order 4 and

Z8, Z4 3 Z2 and Z2 3 Z2 3 Z2

are all the distinct abelian groups of order 8. Before going for the derivation
of the formula, we collect few miscellaneous facts about abelian groups that
will be used in the derivation of the formula.
For any abelian group G and for any integer m, the sets

Gm 5 {am : a  G}

and G(m) 5 {a  G : am 5 e}
are subgroups of G. For any prime p and a positive integer n, it can be easily
verified that
m
Z ppn  Z pnm for any m  n

and Z pn ( p)  Z p .

If G1, G2, …, Gr are groups and f : G → G1 3 … 3 Gr is an isomorphism,

then f induces isomorphisms. For any groups G, G1, G2, …, Gr, if G  G1 3
G2 3 … 3 Gr, then G m  G1m  G2m  Grm

and G ( m)  G1 ( m)  G2 ( m)  Gr ( m)

for any integer m.

Definition 8.2.2. Let n be any positive integer. A finite sequence {n1, n2, …, nr}
of positive integers is said to be a partition of n if

n1 # n2 # … # nr and n1 1 n2 1  1 nr 5 n.

Let P(n) denote the number of partitions of n. For example, P(1) 5 1,

P(2) 5 2, since {1, 1} and {2} are the only partitions of 2.
P(3) 5 3, since {1, 1, 1}, {1, 2} and {3} are the only partitions of 3.
P(4) 5 5 and P(5) 5 7.

Theorem 8.2.4. Let p be a prime and n be a positive integer. Then, there are
exactly P(n) number of distinct (nonisomorphic) abelian groups of order pn,
where P(n) is the number of partitions of n.

Q001-Algebra-111001_CH 08.indd 17 9/16/2011 10:03:50 AM

 8-18   Algebra – Abstract and Modern

Proof: For any partition {n1, n2, …, nr} of n, consider the product
Z pn1  Z pn2  Z pnr

which is an abelian group of order p n1  p n2 … p nr  p n1n2nr  p n .

We shall prove that the correspondence
{n1 , n2 , …, nr }  Z pn1  Z pn2  Z pnr

is a bijection between the set of all partitions on n and the set of all distinct
(nonisomorphic) abelian groups of order pn. If G is an abelian group of order
pn, then, by Theorem 8.2.2,

G   pn1  Z pn2  Z pnr

(since any cyclic p-group must be isomorphic to Z pm for some m . 0) and,

in this case,

p n  | G |  | Z pn1  Z pn2  Z pnr |  p n1n2nr

and hence n 5 n1 1 n2 1 … 1 nr. We can rearrange Z pni ’s such that n1 #

n2 # … # nr and therefore {n1, n2, …, nr} is a partition of n. Therefore, the
above correspondence is a surjection. To prove the injectivity of this correspon-
dence, let {n1, n2, …, nr} and {m1, m2, …, ms} be partitions of n such that

G Z  Z   Z pm2  Z .

pn1 pn2 pnr pm1 pms

Then, we have to prove that r 5 s and ni 5 mi for all 1 # i # s. From the

discussion made before Definition 8.2.2, we have

Z p  Z p  Z p  Z pn1 ( p)  Z pn2 ( p)  Z pnr ( p)

(r times)

 (Z pn1  Z  Z pnr )(p)

pn2

 G (p)
 (Z pm1  Z pm2  Z pms )(p)
 Z p  Z p  Z p
(s times)

and therefore pr 5 ps and hence r 5 s.

Q001-Algebra-111001_CH 08.indd 18 9/16/2011 10:03:51 AM

 Structure Theory of Groups    8-19

Thus, we have proved that if a group G is isomorphic to a product of r number

of Z pni’s, then any expression of G as an isomorphic copy of a product of Z pmi’s
has exactly r number of factors.
Now, suppose ni  mi for some 1 # i # r. Choose j, 1 # j # r such that nj
ni 5
mi for all i , j and nj  mj. We may assume that nj , mj. Since (Z pni ) p  {0}
for i # j (note that 0 is the identity in Z pni ), we have

nj nj
G p  (Z pn1  Z pn2  Z pnr ) p
nj nj nj
 Z ppn1  Z ppn2  Z ppnr
Z n j +1n j ×Z n j +2n j  Z nr n j
p p p

and nj11 2 nj # nj12 2 nj # … # nr 2 nj. Clearly

nj
there are at most r 2 j nonzero
factors in the above decomposition of G p . Similarly, since ni 5 mi for i , j
and nj , mj and since G  Z p m1  Z p mr , we get that
1 r

nj
Gp Z m j n j Z m j +1n j  Z mr n j
p p p

and 1 # mj 2 nj # mj11 2 nj # … # mr 2 nj. Clearly,n there are atleast r 2 j 1 1

1
nonzero factors in the second decomposition
nj
of G p . Therefore, we have two
decompositions of the group G p as a product of cyclic p-groups and the
number of factors in the first decomposition is less than the number of factors
in the second decomposition. This contradicts the conclusion obtained just
after proving that r 5 s. Thus, we must have ni 5 mi for all 1 # i # r. This
proves that there is a bijection between the set of partitions of n and the set of
distinct (nonisomorphic) abelian groups of order pn and hence the numbers of
members in both the sets are same. b

Corollary 8.2.3. Let m  p1n1 p2n2 … prnr , where p1, p2, …, pr are distinct primes
and n1, n2, …, nr are positive integers. Then, the number of distinct (noniso-
morphic) abelian groups of order m is equal to p(n1) p(n2) … p(nr), where
p(ni) is the number of partitions of ni.

Proof: Any abelian group of order m must be isomorphic with a product

G1 3 G2 3 … 3 Gr ,

Q001-Algebra-111001_CH 08.indd 19 9/16/2011 10:03:52 AM

 8-20   Algebra – Abstract and Modern

where each Gi is an abelian group of order pi ni , 1 # i # r, and conversely any

such product is an abelian group of order m. For each 1 # i # r, we know (by
the above theorem) that there are p(ni) number of distinct (nonisomorphic)
abelian groups of order pi ni and hence the corollary. b

Corollary 8.2.4. Let m be any square-free positive integer (that is, m is not
divisible by any perfect square greater than 1). Then, any abelian group of
order m is cyclic and hence Zm is the only (up to isomorphism) abelian group
of order m.

Proof: Since m is square-free, we can write

m 5 p1 p2 … pr

where p1, p2, …, pr and distinct primes. By Corollary 8.2.3, the number of
distinct abelian groups of order m is p(1) p(1)…p(1) 5 1. We know that the
group Zn of integers modulo n is an abelian group of order n. Thus, Zn is the
only (up to isomorphism) abelian group of order n. b

Worked Exercise 8.2.1. Find the number of abelian groups of order 7,200.

Answer: We have to first express 7,200 as a product of primes. We have

7,200 5 25 3 32 3 52 and hence, by Corollary 8.2.3, the number of distinct
abelian groups of order 7,200 is p(5) p(2) p(2), where p(n) denotes the num-
ber of partitions of n. Note that p(2) 5 2, since {1, 1} and {2} are the only
partitions of 2. Coming to p(5), note that

{1, 1, 1, 1, 1}, {1, 1, 2}, {1, 1, 3},

{1, 2, 2}, {1,4}, {2, 3} and {5}

are the only partitions of 5 and hence p(5) 5 7. Thus, there are exactly 7 3 2 3
2 (5 28) abelian groups of order 7,200.

Worked Exercise 8.2.2. Prove that any abelian group of order 2,310 is cyclic.

Answer: Note that 2,310 5 2 3 3 3 5 3 7 3 11, which is a product of

distinct primes and hence square-free. By Corollary 8.2.4, the group Z2,310
of integers modulo 2,310 is the only (up to isomorphism) abelian group of
order 2,310.

Q001-Algebra-111001_CH 08.indd 20 9/16/2011 10:03:52 AM

 Structure Theory of Groups    8-21

Worked Exercise 8.2.3. List all (up to isomorphism) abelian groups each of
order 240.

Answer: First, we have to express 240 as a product of primes.

240 5 24 ? 31 ? 51

Therefore, there are p(4) p(1) p(1) (5 5 3 1 3 1 5 5) abelian groups of order

240. To list these, we have to write down all the partitions of 4 and 1.
Partitions of 4 are {1, 1, 1, 1}, {1, 1, 2}, {1, 3}, {2,2} and {4}. Partitions of
1 is {1}.
Thus, the following five are all (up to isomorphism) the abelian groups of
order 240.

Z2 3 Z2 3 Z2 3 Z2 3 Z3 3 Z5
Z2 3 Z2 3 Z 22 3 Z3 3 Z5
Z2 3 Z 23 3 Z3 3 Z5
Z 22 3 Z 22 3 Z3 3 Z5
Z 24 3 Z3 3 Z5

Note that Zn 3 Zm  Znm if n and m are relatively prime and therefore

Z2 3 Z3 3 Z5  Z30. The above five groups are isomorphic to the following,
respectively.

Z2 3 Z2 3 Z2 3 Z30 ( Z2 3 Z2 3 Z6 3 Z10)
Z2 3 Z2 3 Z60 ( Z2 3 Z6 3 Z20)
Z2 3 Z120 ( Z8 3 Z30)
Z12 3 Z20
Z240

Worked Exercise 8.2.4. List all (up to isomorphism) abelian groups of

order 3,375.

Answer: We have 3,375 5 33 3 53 and hence there are 9 ( 5 p(3)∙p(3) 5 3 3 3)

abelian groups of order 3,375. The partitions of 3 are

{1, 1, 1}, {1, 2} and {3}.

Q001-Algebra-111001_CH 08.indd 21 9/16/2011 10:03:53 AM

 8-22   Algebra – Abstract and Modern

Therefore, the required groups are given below.

Z3 3 Z3 3 Z3 3 Z5 3 Z5 3 Z5 ( Z15 3 Z15 3 Z15)

Z3 3 Z3 3 Z3 3 Z5 3 Z52 ( Z3 3 Z15 3 Z75)
Z3 3 Z3 3 Z3 3 Z53 ( Z3 3 Z3 3 Z375)
Z3 3 Z32 3 Z5 3 Z5 3 Z5 ( Z15 3 Z45 3 Z5)
Z3 3 Z32 3 Z5 3 Z52 ( Z15 3 Z225)
Z3 3 Z32 3 Z53 ( Z9 3 Z375)
Z33 3 Z5 3 Z5 3 Z5 ( Z135 3 Z5 3 Z5)
Z33 3 Z5 3 Z52 ( Z135 3 Z25)
Z33 3 Z53 ( Z3,375)

Next, we prove a most general form of the fundamental structure theorem

for finitely generated abelian groups. For convenience, we shall denote the
binary operation in an abelian group by 1 instead of ? and, as such we write
0 for the identity element of an abelian group (G, 1). The inverse of a will
be denoted by 2a instead of a21 and write na instead of an, where a is an
arbitrary element of an abelian group (G, 1) and n is an integer. That is, na
is defined as

 0 if n  0, the integer zero



na  ( n 1)a  a if n  0 .

(n)(a) if n  0

Definition 8.2.3. Let (G, 1) be an abelian group and S ⊆ G. If G 5 ,S.,

the smallest subgroup of G containing S, then S is called a generating set for
G. G is called finitely generated if there exists a finite generating set for G.
If S 5 {s1, s2, …, sr} is a finite generating set for an abelian group (G, 1),
then any element a of G can be expressed as

a 5 n1s1 1 n2s2 1  1 nrsr

for some integers n1, n2, …, nr. In fact, for any subset S of G, we have

,S. 5 {n1s1 1 n2s2 1  1 nrsr : si  S and ni  Z}.

The expression a 5 n1s1 1 n2s2 1  1 nrsr need not be unique. In this

context, we have the following theorem.

Q001-Algebra-111001_CH 08.indd 22 9/16/2011 10:03:54 AM

 Structure Theory of Groups    8-23

Theorem 8.2.5. Let (G, 1) be an abelian group. Then, G is isomorphic to

the product of a finite number of copies of (Z, 1) if and only if there exists
a finite generating set {s1, s2, …, sn} such that any element a of G can be
uniquely expressed as

a 5 n1s1 1 n2s2 1  1 nrsr

for some integers n1, n2, …, nr.

Proof: Let G  Z 3 Z 3 … 3 Z (r copies of Z) and f : G → Z 3 Z 3 …

3 Z be an isomorphism. For each 1 # i # r, let ti 5 (0, …, 0, 1, 0, …, 0), the
element whose ith component is 1 and all other components are zeros. Then,
any element t  Z 3 … 3 Z can be uniquely expressed as

t 5 (n1, n2, …, nr) 5 n1t1 1 n2t2 1  1 nrtr.

If we consider si 5 f−1(ti) for 1 # i # r, then s1, s2, …, sr satisfy the required

property. Conversely, suppose that s1, s2, …, sr are elements in S satisfying the
given properties. Then, it can be easily verified that the map

(n1, n2, …, nr)  n1s1 1 n2s2 1  1 nrsr

is an isomorphism of Z 3 Z 3  3 Z (r copies of Z) onto G. b

The following result provides a complete characterization of finitely gen-
erated abelian groups. First, let us agree, for convenience, to write
n

∑a
i1
i for a1 1 a2 1  1 an,

where a1, a2, …, an are any elements of an abelian group (G, 1).

Theorem 8.2.6 (Fundamental structure theorem for finitely generated

a­ belian groups). Let (G, 1) be a finitely generated abelian nontrivial group.
Then, G is isomorphic to the product of a finite number of cyclic groups Ai;
that is,

G  A1 3 A2 3  3 Ak,

where each Ai is a nontrivial cyclic group such that either all of the Ai’s are
infinite or for some s, 1 # s # k, A1, A2, …, As are of orders m1, m2, …, ms
respectively with mi divides mi11 for each 1 # i , s and As11, …, Ak are
infinite.

Q001-Algebra-111001_CH 08.indd 23 9/16/2011 10:03:54 AM

 8-24   Algebra – Abstract and Modern

Proof: Since (G, 1) is a finitely generated abelian group, there are finite sets
generating G. Let k be the least positive integer such that G has a k-element
generating set. We shall use induction on k. If k 5 1, then G is generated by
a single element and hence G itself is cyclic. Suppose that k . 1 and assume
that the theorem is true for all abelian groups generated by a set of k 2 1
elements.
If G has a generating set {a1, a2, …, ak} such that

n1a1 1  1 nrar 5 0 ⇒ n1 5 5 nr 5 0 (*)

then any element of G can be uniquely expressed as n1a1 1  1 nrar, ni  Z

and hence by Theorem 8.2.5,

G  Z 3  3 Z (r copies).

Next, suppose that G has no generating set {a1, …, ak} satisfying the property
(*). Then, for any generating set {a1, …, ak}, there exists integers n1, …, nk,
not all zero, such that n1a1 1  1 nkak 5 0. Since

k k k

∑ n a  0 ⇒ ∑ (n )a  ∑ n a  0.
i1
i i
i1
i i
i1
i i

There is an equation of the form n1a1 1  1 nkak 5 0 with one of ni’s posi-
tive. Let T be the set of all positive integers occurring in equations of the form
n1a1 1  1 nkak 5 0, where {a1, …, ak} is a generating set for G. The above
discussion implies that T is a nonempty set of positive integers. Let m1 be the
least positive integer in T. We can assume that

m1a1 1 n2a2 1  1 nkak 5 0 (1)

for some generating set {a1, a2, …, ak} for G and integers n2, …, nk. We shall
prove that m1 divides each ni, 2 # i # k. By the division algorithm, let us
write

ni 5 qim1 1 ri, 0 # ri , m1, qi and ri  Z.

Then, from Equation (1),

m1b1 1 r2a2 1  1 rkak 5 0 (2)

where b1 5 a1 1 q2a2 1  1 qkak. If b1 5 0, then a1 5 2 q2a2 2 q3a3 2  2

qkak and hence {a2, a3, …, ak} is a k 2 1 element generating set for G which
is a contradiction to the least property of k.

Q001-Algebra-111001_CH 08.indd 24 9/16/2011 10:03:55 AM

 Structure Theory of Groups    8-25

Therefore, b1  0. Also,

a1 5 b1 2 q2a2 2  2 qkak

and therefore {b1, a2, …, ak} is a k-element generating set for G and r2, …, rk
are nonnegative integers less than m1 occurring in Equation (2). By the least
property of m1, it follows that r2 5 r3 5  5 rk 5 0 and hence mi 5 qim1 for
2 # i # k. Also, from Equation (2), we have

m1b1 5 0.

Put A1 5 ,b1., the subgroup generated by b1 in G. Put H1 5 ,a2, …, ak.,

the subgroup generated by {a2, …, ak} in G. Since {b1, a2, …, ak} is a generat-
ing set, we get that A1 1 H1 5 G. Also, we prove that A1 ∩ H1 5 {0}. If x 
A1 ∩ H1, then, since m1b1 5 0,

x 5 mb1 for some 0 # m , m1

and x 5 n2a2 1  1 nkak for some n2, …, nk  Z

and therefore mb1 1 (2 n2)a2 1  1 (2nk)ak 5 0. By the least property of

m1, we get that m 5 0 (otherwise m  X and m , m1) and hence x 5 0. There-
fore, A1 ∩ H1 5 {0}. By Theorem 8.1.1, G  A1 3 H1. Also, H1 is generated
by k 2 1 elements a2, …, ak and it is not generated by a set with less than k 2 1
elements (otherwise G would be generated by a set with less than k elements,
which is a contradiction). Therefore, by the induction hypothesis,

H1  A2 3  3 Ak

where each Ai is a cyclic group such that either all the A2, A3, …, Ar are infi-
nite or for some s, 2 # s # r, A2, …, As are of orders m2, …, ms respectively
with mi divides mi11 for all 2 # i , s and As11, …, Ar are infinite. The proof
is complete, if we can show that m1 divides m2 also. To do this, let Ai 5 ,bi.
and bi be of order mi for 2 # i , k. Then, {b1, b2, …, bk} is a generating set
for G and

m1b1 1 m2b2 1 0b3 1  1 0bk 5 0.

By an argument similar to that made after Equation (1), it follows that m1

divides m2. Thus, we have

G  A1 3 H1  A1 3 A2 3  3 Ak

where A1, A2, …, Ak are cyclic groups satisfying the required properties. This
completes the proof of the theorem. b

Q001-Algebra-111001_CH 08.indd 25 9/16/2011 10:03:55 AM

 8-26   Algebra – Abstract and Modern

Since any infinite cyclic group is isomorphic to the group (Z, 1) and any
finite cyclic group of order n is isomorphic to the group (Zn, 1n), the follow-
ing is an immediate consequence of the above theorem.

Corollary 8.2.5. Let G be a finitely generated abelian group. Then,

G  Z  Z  Z m1  Z ms

(r components)

where r and s are nonnegative integers and m1, …, ms are positive integers
such that mi divides mi11 for all 1 # i , s.

Worked Exercise 8.2.5. State and prove the converse of Corollary 8.2.5.

Answer: If a group G is isomorphic to a product

Z     m1  Z ms

then G is a finitely generated abelian group.

In fact, we prove that the product of finite number of cyclic groups is a finitely
generated abelian group.
Let G1, G2, …, Gn be cyclic groups and

G 5 G1 3 G2 3 … 3 Gn

since each Gi is abelian, the product G is also abelian. Let ai be a generator for
Gi and ei be the identity in Gi. Put xi be the element in G whose ith component
is ai and other components are identities. That is,

xi 5 (e1, …, ei21, ai, ei11, …, en).

Then, any element g  G can be written as

g  ( g1 , ..., g n ), gi ∈ Gi
 ( a1r1 , ..., anrn ), ri ∈ 
 x1r1 x2r2 ... xnrn ∈< x1 , ..., xn > .

Therefore, G is generated by {x1, x2, …, xn}. Thus, G is a finitely generated

abelian group.

Q001-Algebra-111001_CH 08.indd 26 9/16/2011 10:03:56 AM

 Structure Theory of Groups    8-27

Worked Exercise 8.2.6. Let G be a finitely generated abelian group. Prove

that G is finite if and only if the order of every element of G is finite.

Answer: If G is finite and a  G, then O(a) # |G| (in fact, O(a) is a divisor of
|G|) and hence O(a) is finite. Conversely, suppose that O(a) is finite for every
a  G. By Corollary 8.2.5,

G  Z  Z  Z m1  Z ms

(r components),

where r and s are nonnegative integers and m1, …, ms are positive integers. If
r . 0, then consider

a 5 (2, 0, …, 0, 0, …, 0)

then a is of infinite order and hence the element x in G corresponding to a is

also of infinite order, which is a contradiction to our hypothesis. Thus, r 5 0
and hence

G  Z m1  Z ms

since each Z mi is finite, so is G.

EXERCISE 8(b)

1. State whether each of the following is True or False. Substantiate your answer.
(i) Z2  Z2  Z4 .
(ii) Z3  Z12 is cyclic.
(iii) Z6  Z 25 is cyclic.
(iv) There is exactly one abelian group of order 105.
(v) There is exactly one group of order 30.
(vi) Any abelian group of order 165 is isomorphic to Z165.
(vii) The number of abelian groups of order 24 is 3.
(viii) The number of groups of order 24 is 3.
(ix) Any abelian group of order divisible by 7 contains a cyclic subgroup of
order 7.
(x) Any abelian group of order divisible by 9 contains a cyclic subgroup of
order 9.

Q001-Algebra-111001_CH 08.indd 27 9/16/2011 10:03:56 AM

 8-28   Algebra – Abstract and Modern

2. Prove that the following are equivalent to each other for any positive integers m
and n.
(a) Z m  n is cyclic.
(b) Z m  Z n  Z mn .
(c) m and n are relatively prime.
(d) Z  m, n , the subgroup generated by m and n in the group (Z, 1).
3. Prove that any abelian group of order 8 is isomorphic to one of the following.
Z8 , Z 4  Z 2 , Z 2  Z 2  Z 2 .

4. Let G1, G2, …, Gn be groups and ai  Gi, 1 # i # n. Let a 5 (a1, a2, …, an). Prove
that O(a) is finite in the product G1 3 G2 3 … 3 Gn if and only if O(ai) is finite
in Gi for each 1 # i # n and, in this case, O(a) 5 l.c.m. {O(a1), …, O(an)}.

5. Prove that, for any n . 1,  n2 is not isomorphic with Z n  Z n .

6. Let Z(G) denote the centre of any group G. For any groups G1, G2, …, Gn, prove
that Z(G1 3 G2 3 … 3 Gn) 5 Z(G1) 3 Z(G2) 3 … 3 Z(Gn).

7. Let [G, G] denote the commutator subgroup of a group G. For G 5 G1 3 … 3

Gn, prove that
[G, G] 5 [G , G ] 3 … 3 [G , G ].
1 1 n n

8. Let N1 and N2 be normal subgroups of groups G1 and G2, respectively. Prove that
N1 3 N2 is a normal subgroup of G 1 3 G2 and that

G1 N1  G2 N 2 ≅ (G1 × G2 ) N1 × N 2 .

9. Let N and M be normal subgroups of a group G such that N ∩ M 5 {e}. Then

prove that G is isomorphic to a subgroup of G / M  G / N .

10. Prove that any cyclic p-group is finite.

11. Prove that any finite abelian p-group is generated by its elements of highest
order.

12. Show that any homomorphic image of a p-group is a p-group and product of
p-groups is also a p-group.

13. Let G be a finite cyclic group and p be a prime dividing the order of G. Prove that
there are exactly p 2 1 elements in G each having order p.

14. Prove that a nontrivial finite abelian group is cyclic if and only if it is isomorphic
to Z p n1  Z p n2  Z p nr for some distinct primes p1, …, pr and positive integers
1 2 r
n1, …, nr.
15. Determine all (up to isomorphism) abelian group of order 144, 625 and
1,94,481.

Q001-Algebra-111001_CH 08.indd 28 9/16/2011 10:03:57 AM

 Structure Theory of Groups    8-29

16. Prove that a cyclic group is indecomposable if and only if it is either infinite or
of prime power order.

17. Describe all the positive integers n for which there is exactly one (up to isomor-
phism) abelian group of order n.

18. Let {a1, a2, …, ar} be a generating set for an abelian group G. Prove that the
following are equivalent to each other.
r
(1) ( n1 , n2 , ..., nr )   ni ai is an isomorphism of Z  Z  Z ( r copies)
i1
onto G.
(2) Any element a of G can be uniquely expressed as
a 5 n1a1 1 … 1 nrar, for integers n1, n2, …, nr.
(3) For any integers n1, n2, …, nr,
n1a1 1 n2a2 1 … 1 nrar 5 0 ⇒ n1 5 n2 5 … 5 nr 5 0
19. Let G be a finitely generated abelian group. Prove that G is finite if and only if G
is isomorphic with a product of finitely many cyclic groups.

20. Let G be a nontrivial abelian group. Prove that G is finite if and only if
G Z m Z m
1 r

for some positive integers m1, …, mr such that mi . 1 and mi divides mi11 for all
1 # i , r.

8.3 INVARIANTS OF FINITE ABELIAN GROUPS

The fundamental structure theorem for finitely generated abelian groups
(Theorem 8.2.6) can be applied to finite abelian groups to associate a unique
finite division sequence of positive integers with each finite abelian group.
First, let us define the following.

Definition 8.3.1. A finite sequence m1, m2, …, mr of positive integers is called

a division sequence if 1 , m1 and mi divides mi11 for each 1 # i , r. A division
sequence is denoted by

1 , m1|m2| … |mr.

In general, if a and b are integers, we write a to say that a divides b; that is,
b
ac 5 b for some integer c.

Theorem 8.3.1. Let G be a nontrivial finite abelian group. Then, there exists
a unique division sequence

1 , m1|m2| … |mr

Q001-Algebra-111001_CH 08.indd 29 9/16/2011 10:03:58 AM

 8-30   Algebra – Abstract and Modern

such that G  Z m1  Z m2  Z mr .

Proof: By Theorem 8.2.6, there exists nontrivial finite cyclic groups G1, G2,
…, Gr of orders m1, m2, …, mr, respectively such that mi divides m1, m2, …, mr
respectively such that mi divides mi11 for each 1 # i , r and

G  G1 3 G2 3 … 3 Gr .

Since any cyclic group of order m is isomorphic to Zm, we get that 1 , m1|m2|
… |mr is a division sequence and

G  Z m1  Z m2  Z mr .  (1)

Suppose that 1 , n1|n2| … |ns is any division sequence such that

G  Z n1  Z n2  Z ns .  (2)

We shall prove that r 5 s and mi 5 ni for each 1 # i # r. First note that the
order of any element of Z mi is a divisor of mi and hence that of G is a divisor
of the l.c.m. {m1, m2, …, mr} which is equal to mr (since mi divides mr for all
1 # i # r). Therefore, the order of any element of G is a divisor of mr. Since
the element (0, …, 0, 1) is of order ns in Z n1  Z ns  G , there exists an
element of order ns in G and therefore ns divides mr. Similarly, by symmetry,
we can prove that mr divides ns and hence mr 5 ns. From the decompositions
(1) and (2), we have

mr−1G  ( mr−1 Z m1 )  ( mr−1 Z mr1 )  ( mr−1 Z mr )

 ( mr−1 Z n1 )  ( mr−1 Z ns1 )( mr−1 Z ns ),

where mA stands for {ma : a  A}. Since mi|mr21 for all 1 # i # r 2 1, it

follows that mr1 Z mi  {0} for each 1 # i # r 2 1 and hence

| mr1G || mr1 Z mr || mr1Z ns | .

This implies that | mr1Z n j |  1 for each 1 # j # s 2 1 and, in particular

| mr1 Zns1 | 1 so that ns21 divides mr21. Similarly, by interchanging mi’s and
nj’s, we get that mr21 divides ns21. Therefore, mr21 5 ns21. Continuing this
process, we can prove that mr−i 5 ns2i for i 5 0, 1, 2, …. Since m1 m2 … mr 5
|G| 5 n1 n2…ns, we get that r 5 s and mi 5 ni for all 1# i # r. b

Q001-Algebra-111001_CH 08.indd 30 9/16/2011 10:03:59 AM

 Structure Theory of Groups    8-31

Definition 8.3.2. If G is an abelian group and 1 , m1|m2| … |mr is a division

sequence such that

G  Z m1  Z m2  Z mr

then G is said to be of type (m1, m2, …, mr) and the integers m1, m2, …, mr
are called the invariants of G. The following is an immediate consequence of
Theorem 8.3.1.

Corollary 8.3.1. Let n be a positive integer greater than 1. Then, the number
of abelian groups of order n is equal to the number of division sequences 1 ,
m1|m2| … |mr such that n 5 m1 m2 … mr.

Example 8.3.1. Let us find the invariants of the group G 5 Z6 3 Z8 3 Z5.

First note that 6, 8, 5 do not form a division sequence. We have

G  Z 6  Z8  Z 5  Z 2  Z 3  Z8  Z 5
 Z 2  Z120

and 2 divides 120. Therefore, the invariants of G are 2, 120.

Worked Exercise 8.3.1. Determine the invariants of the group

G  Z 2  Z 4  Z 9  Z 27  Z81 .

Answer: By inspection, we can see that

G  Z 9 Z54  Z324

and 9|54|324 is a division sequence. Therefore, 9, 54, 324 are the invari-
ants of G.
The following provides an algorithm to find the invariants of a given finite
abelian group. Let us recall that a partition of a positive integer n is a finite
sequence {n1, n2, …, nr} of positive integers such that n1 # n2 # … # nr and
n1 1 n2 1 … 1 nr 5 n. For the convenience and for the purpose of proving
the following result, we relax the definition of a partition of n by including
certain zeros in the beginning. Accordingly, a partition of n is a sequence of
nonnegative integers {n1, n2, …, nr} such that 0 # n1 # n2 # … # nr and
n1 1 n2 1 … 1 nr 5 n.

Q001-Algebra-111001_CH 08.indd 31 9/16/2011 10:04:00 AM

 8-32   Algebra – Abstract and Modern

Theorem 8.3.2. Let G be an abelian group of order n  p1m1 p2m2  pkmk where
p1, p2, …, pk are distinct primes and m1, m2, …, mk are positive integers. For
each 1 # i # k, let {ni1 , ni2 , ..., nir } be a partition of mi such that

G  (Z p n11  Z pn1r )  (Z pnk 1  Z pnkr ).

1 1 k k

n n
For each 1 # j # r, let s j  p1 1 j … pk kj . Then, s1, s2, …, sr are the invari-
ants of G.

Proof: Note that 0  ni1  ni    ni and ni1 1 ni2 1 … 1 nir 5 mi for

2 r
each 1 # i # k. Let

Gj  Z n Z n … Z n , for 1 # j # r.
p11 j p22 j pk kj

n n n
Since p1, p2, …, pk are distinct primes, it follows that p1 1 j , p2 2 j , ..., pk kj are
pair-wise relatively prime and hence

n n n
G j  Z s j , where s j = p1 1 j p2 2 j … pk kj .

Now, G  G1  G2  Gr ≅ Z s1  Z s2  Z sr .

n n
Also, since nij # nij11 for 1 # i # k and 1 # j # r, we get that pi ij divides pi ij1
and hence sj divides sj11 for each 1 # j , r. Thus, s1, s2, …, sr are the invari-
ants of G.

Example 8.3.2. Let G be the set of all positive integers less than 100 and rela-
tively prime to 100. Then, G is an abelian group under multiplication modulo
100. Let us find the invariants of G.
First, we shall list all elements of G and find the order G. We have |G| 5
f (100) 5 40 5 2335

G 5 {1, 3, 7, 9, 11, 13, 17, 19, 21, 23, 27, 29, 31, 33,
37, 39, 41, 43, 47, 49, 51, 53, 57, 59, 61, 63, 67,
69, 71, 73, 77, 79, 81, 83, 87, 89, 91, 93, 97, 99}.

One can easily verify the G 5 ,3., the cyclic subgroup generated by 3.
Therefore, G is a cyclic group of order 40 and hence G  Z40. 40 is the only
invariant of G.

Q001-Algebra-111001_CH 08.indd 32 9/16/2011 10:04:01 AM

 Structure Theory of Groups    8-33

EXERCISE 8(c)

1. Determine the invariants of each of the following groups.

(i) Z2 3 Z2 3 Z5 3 Z5.
(ii) Z9 3 Z25 3 Z7 3 Z4 3 Z121.
(iii) Z8 3 Z7 3 Z49 3 Z9.
(iv) Z2 3 Z4 3 Z8 3Z9 3 Z3 3 Z5 3 Z25.

2. Let X be a set with 5-elements and G 5 (P(X), ⊕), where P(X) is the power set of
X and ⊕ is the symmetric difference operation. Determine the invariants of G.

3. Let G be the group of all positive integers less than 47 and relatively prime to 47,
under the multiplication modulo 47. Then find the invariants of G.

4. Let G be the group of all mappings of a 4-element set into the group (Z4, 14)
under point-wise operation. Determine all the invariants of G.

5. Determine the invariants of each abelian group of order less than or equal to 30.

8.4 GROUPS OF SMALL ORDER

We conclude the discussion on group theory with the complete description
of all groups of order less than or equal to 20. We have derived an exact for-
mula for the number of abelian groups of a given order n and an algorithm
to list all these groups, up to isomorphism. However, there is no precise
formula for the number of all groups (nonabelian groups) of a given order
n. In this section, we describe these groups (up to isomorphism) of order n
for n # 20.
Let us first recall that any group of prime order is cyclic and hence abelian
and that any group of order p2, where p is prime, is abelian and there are only
two such groups, namely Zp or Zp 3 Zp. Further, we have proved that any
group of order pq, where p , q are primes such that p does not divide q 2 1,
is cyclic. In the following, we prove that there is a unique nonabelian group
of order pq, when p divides q 2 1.

Theorem 8.4.1. Let p and q be primes such that p , q and p divides q 2 1

and G be a nonabelian group of order pq. Then, G is a group generated by two
elements a and b satisfying the following:
1. ap 5 1 5 bq, a  e and b  e.
2. a–1ba 5 br, where r  1(mod q) and rp ≡ 1(mod q).
Also, G is the unique (up to isomorphism) nonabelian group of order pq.

Q001-Algebra-111001_CH 08.indd 33 9/16/2011 10:04:01 AM

 8-34   Algebra – Abstract and Modern

Proof: By Sylow Theorem III, the number of Sylow q-subgroups of G is kq

1 1 which divides |G| 5 pq. This implies that kq 1 1 5 1 and hence there
exists exactly one Sylow q-subgroup B which is a normal subgroup of order
q in G. Since q is prime, it follows that B is cyclic and hence B 5 ,b. for
some b  G such that bq 5 e  b. Also, the number of Sylow p-subgroups is
of the form mp 1 1 for some m  0 and is divisor of |G| 5 pq. This implies
that mp 1 1 5 1 or q. If mp 1 1 5 1, then, as above, there exists a unique
subgroup A of order p which becomes normal in G and, in this case A ∩ B 5
{e} and AB 5 G, so that G  A 3 B  Zp 3 Zq  Zpq and hence G is cyclic
and abelian, which is a contradiction to the hypothesis that G is nonabelian.
Therefore, mp 1 1 5 q; that is, the number of Sylow p-subgroups is q.
Let A be a Sylow p-subgroup of G. Then, |A| 5 p and hence A 5 ,a. for
some a  G such that ap 5 e  a. Consider the subgroup ,a, b. generated
by a and b. Since A and B are contained in ,a, b., we have AB ⊆ ,a, b..
Also, A ∩ B 5 {e} and hence

| A || B | p⋅q
| AB |  = | G |
| A∩ B | 1

and hence AB 5 G, so that ,a, b. 5 G.

Since B 5 ,b. is a normal subgroup of G, a–1ba 5 br for integer r, then r 
1 (mod q), (otherwise, a–1ba 5 b and hence ba 5 ab which implies that G is
abelian, a contradiction). Now, we have
a–1ba 5 br ⇒ a–1b2a 5 (a–1ba)(a–1ba) 5 b2r
2
⇒ a–1bra 5 b r (by induction)
2
⇒ a–1(a–1ba)a 5 a–1bra 5 b r
2
⇒ a–2ba2 5 b r
p
⇒ a–pbap 5 b r (by induction)
p
⇒ b 5 ebe 5 a–pbap 5 b r
⇒ rp  1 (mod q) (since O(b) 5 q).
Thus, G is generated by a and b, which satisfy the following:
1. ap 5 e 5 bq, a  e and b  e.
2. a21ba 5 br, r  1 (mod q) and rp  1 (mod q).
On the other hand, if G9 is any nonabelian group of order pq, then G9 is gen-
erated by elements a and b satisfying the above properties (1) and (2). Now,
AB 5 G9, where A 5 ,a. and B 5 ,b. and hence

G9 5 {aibj : 0 # i , p and 0 # j , q}.

Q001-Algebra-111001_CH 08.indd 34 9/16/2011 10:04:02 AM

 Structure Theory of Groups    8-35

Since the solutions of xp  1 (mod q), x  1 (mod q), are r, r2, …, rp21, it fol-
lows that G  G9, since replacing a by aj as a generator of ,a. replaces r by
rj. This completes the proof.

Corollary 8.4.1. Let G be a group of order pq, where p and q are primes such
that p , q. Then, either G is cyclic or G is a nonabelian group generated by
two elements a and b satisfying the properties (1) and (2) above.
Now, we list all groups of order less than or equal to 20. Some of the
proofs involved in the listing are left as exercises to the reader.

1. The trivial group {e} is the only group of order 1.

2. Z2 is the only (up to isomorphism) group of order 2, since 2 is prime.
3. Z3 is the only (up to isomorphism) group of order 3.
4. Any group of order 4(5 22) is abelian and there are only two groups of
order 4, one cyclic and the other noncyclic abelian group; namely Z4 and
Z2 3 Z2.
5. Since 5 is prime, Z5 is the only group of order 5.
6. There are two groups of order 6, one is cyclic and the other is a nonabe-
lian group (see Corollary 8.4.1). These are Z6 and the symmetric group
S3. Since 6 5 2 ? 3, 2 and 3 are primes, there is exactly one abelian group
of order 6 which is the cyclic group Z6. S3 is a nonabelian group of order
6 and S3 is generated by the elements

a 5 (1 2) and b 5 (1 2 3),
a2 5 e 5 b3, a  e, b  e and
a21ba 5 (1 2) (1 2 3) (1 2) 5 (1 3 2) 5 b2 (see Theorem 8.4.1).

7. Since 7 is prime, Z7 is the only group of order 7.

8. There are 5 groups of order 8, 3 abelian and 2 nonabelian. Z2 3 Z2 3 Z2,
Z4 3 Z2 and Z8 are the only abelian groups of order 8.
The quaternion group Q8 5 {1, 2 1, i, 2 i, j, 2 j, k, 2 k} and the dihe-
dral group D4 are the only nonabelian groups of order 8.
9. Any group of order 9 (5 32 and 3 is prime) is abelian. There are only
two groups of order 9, one is cyclic Z9 and the other is noncyclic abelian
Z3 3 Z3.
10. There are only two groups of order 10, one is the cyclic group Z10 and
the other is the dihedral group D5.
11. Since 11 is prime, the cyclic group Z11 is the only group of order 11.
12. There are five groups of order 12, two abelian and three nonabelian. Z12
and Z2 3 Z6 are the abelian groups of order 12.

Q001-Algebra-111001_CH 08.indd 35 9/16/2011 10:04:02 AM

 8-36   Algebra – Abstract and Modern

The alternating group A4 (the group of even permutations in S4), the

dihedral group D6 and the group G described below are the only nonabe-
lian groups of order 12. Note that

D6  S3 3 Z2  D3 3 Z2.

G is the group generated by two elements a and b such that O(a) 5 4,

O(b) 5 3 and ab 5 b21a.
13. Since 13 is a prime, the cyclic group Z13 is the only group of order 13.
14. There are two groups of order 14, one is the cyclic group Z14 and the
other is the dihedral group D7.
15. Any group of order 15 is cyclic (since 15 5 3 ? 5 and 3 does not divide
5 2 1) and hence Z15 is the only group of order 15.
16. There are totally 14 groups of order 16 out of which 5 are abelian and 9
are nonabelian.
The abelian groups of order 16 are Z16, Z8 3 Z2, Z4 3 Z4, Z4 3 Z2 3 Z2
and Z2 3 Z2 3 Z2 3 Z2.
The nonabelian groups of order 16 are given below.
(i) The dihedral group D8.
(ii) D4 3 Z2, where D4 is the dihedral group of degree 4.
(iii) Q8 3 Z2, where Q8 is the eight element quaternion group.
(iv) The group generated by two elements a and b such that O(a) 5 8,
O(a) 5 2 and ab 5 ba3.
(v) G 5 ,a, b., where O(a) 5 8, O(b) 5 2 and ab 5 ba5.
(vi) G 5 ,a, b., where O(a) 5 4 5 O(b) and ab 5 ba3.
(vii) G 5 ,a, b, c., where O(a)5 4, O(b) 5 2 5 O(c), cbca2b 5 1,
bab 5 a and cac 5 a.
(viii) G 5 ,a, b., where O(a) 5 8, a4 5 b2 and aba 5 b.
(ix) G 5 ,a, b., where O(a) 5 4 5 O(b), abab 5 e and ab3 5 ba3.
17. 17 is a prime and hence Z17 is the only group of order 17.
18. There are five groups of order 18, two abelian and three nonabelian. Z18
and Z3 3 Z6 are the abelian groups and the nonabelian groups are D9,
S3 3 Z3 and the group G 5 ,a, b, c. such that O(a) 5 2, O(b) 5 3 5
O(c), bc 5 cb, bab 5 a and cac 5 a.
19. 19 is a prime and hence Z19 is the only group of order 19.
20. There are five groups of order 20, two abelian and three nonabelian. Z20 and
Z2 3 Z10 are the abelian groups, while the nonabelian groups are the dihe-
dral group D10, the group G 5 ,a, b. where O(a) 5 4, O(b) 5 5 and bab
5 a and the group H 5 ,x, y. where O(a) 5 4, O(b) 5 5 and ba 5 ab2.

Q001-Algebra-111001_CH 08.indd 36 9/16/2011 10:04:02 AM

 Part III
Ring Theory

Q001-Algebra-111001_CH 09.indd 1 9/21/2011 4:52:26 PM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 09.indd 2 9/21/2011 4:52:26 PM

 9
Rings
9.1 Examples and Elementary Properties
9.2 Certain Special Elements in Rings
9.3 The Characteristic of a Ring
9.4 Subrings
9.5 Homomorphisms of Rings
9.6 Certain Special Types of Rings
9.7 Integral Domains and Fields

It is well known that there are two familiar binary operations, namely the
addition 1 and the multiplication ? on the set Z of integers and that (Z, 1) is
an abelian group where as (Z, ?) is only a semigroup. We have earlier worked
with algebraic systems, namely semigroups, monoids and groups, where
there is only one binary operation in each. Now, in this chapter, we initiate
the study of abstract algebraic systems having two binary operations as in the
case of integers. Also, we have the rational number system, the real number
system, the complex number system, the set of all n 3 n matrices, where in
each of these cases we have two binary operations satisfying certain connec-
tive properties in addition to the properties satisfied by the individual opera-
tions. We introduce a common abstraction of these in the form of a ring and
develop a general elementary theory of rings. A ring is basically a combina-
tion of an abelian group and a semigroup and therefore a previous knowledge
of groups and semigroups will be of considerable help. Most of the important
concepts in group theory have natural extensions to ring theory.

9.1 EXAMPLES AND ELEMENTARY PROPERTIES

When we have two binary operations say * and o on a set X, in order to get
information about the algebraic system (X, *, o) more than we could obtain by

Q001-Algebra-111001_CH 09.indd 3 9/21/2011 4:52:26 PM

 9-4   Algebra – Abstract and Modern

studying each operation separately, these must be some relationship between

the two operations. The most common requirement is that one of them be
distributive over the other: the operation * is said to left distributive over o if

a * (b o c) 5 (a * b) ο (a * c) for all a, b and c  X

and * is said to be right distributive over ο if

(a o b) * c 5 (a * c) o (b * c) for all a, b and c  X.

We say that * is distributive over o if it is both left and right distributive

over ο.
In this section, we present a formal definition of a ring with several illus-
trative examples and prove certain important elementary properties of rings.

Definition 9.1.1. A triple (R, 1, ?) is called a ring if R is a nonempty set and

1 and ? are binary operations on R satisfying the following:
I. (R, 1) is an abelian group; that is,
(I.1) a 1 b 5 b 1 a for all a and b  R
(I.2) a 1 (b 1 c) 5 (a 1 b) 1 c for all a, b and c  R
(I.3) There is a (unique) element 0 such that
a 1 0 5 a for all a  R and
(I.4) For each a  R, there exists (unique) element 2a  R such that
a 1 (2a) 5 0.
II. (R, ?) is a semigroup; that is,
a ? (b ? c) 5 (a ? b) ? c for all a, b and c  R

III. The operation ? is distributive over the operation 1; that is,

a ? (b 1 c) 5 (a ? b) 1 (a ? c)
and (a 1 b) ? c 5 (a ? c) 1 (b ? c) for all a, b and c  R.
One should clearly understand that 1 and ? are abstract binary opera-
tions and not ordinary addition and multiplication of integers or real num-
bers. However, for convenience, we call the operation 1 as addition and
the operation ? as multiplication. In the light of this terminology, it is
natural then to speak of the abelian group (R, 1) as the additive group of
the ring (R, 1, ?) and of (R, ?) as the multiplicative semigroup of the ring
(R, 1, ?).

Q001-Algebra-111001_CH 09.indd 4 9/21/2011 4:52:26 PM

 Rings   9-5

Recall from the group theory that the element 0 in (I.3) above is unique
and is called the additive identity or the zero element in R. Also, for any
a  R, the element 2a in (I.4) above is unique and is called the additive
inverse of a.
In order to minimize the use of parentheses (brackets) in expressions
involving both the operations 1 and ?, let us stipulate that multiplication is to
be performed before addition. Accordingly, the expression a ? b 1 c stands
for (a ? b) 1 c and not for a ? (b 1 c). Also, because of the generalised asso-
ciative law, parentheses can also be omitted when we write sums or products
of more than two elements. For example, we write

a1 ? b1 1 a2 ? b2 1 a3 1 a4 ? b4
instead of ((a1 ? b1) 1 (a2 ? b2)) 1 (a3 1 (a4 ? b4)).

It is needless to say that a 1 b is called the sum of a and b and a ? b is called

the product of a and b in this order. Also, as usual, we write a 2 b for a 1
(2b) and ab for a ? b, when there is no ambiguity about the multiplication.
Note that, as in the case of groups, we simply say that ‘R is a ring’ instead of
saying that ‘(R, 1, ?) is a ring’ when there is no ambiguity about the opera-
tions 1 and ?.
Further, notice that the operations 1 and ? in a ring R cannot be inter-
changed, for (R, ?) may not be a group at all and 1 may not be distribu-
tive over ?. In fact, except in the trivial case when R 5 {0}, (R, ?) is not a
group. These things will be more clear, when we consider the following
examples.

Example 9.1.1
1. Let Z be the set of all integers, Q be the set of all rational numbers and
R be the set of all real numbers. Then, (Z, 1, ?), (Q, 1, ?) and (R, 1, ?)
are all rings, where 1 and ? are the usual addition and multiplication of
real numbers. In each of these cases, the number 0 is the zero element
(that is, the additive identity).
2. Let (G, 1) be an abelian (commutative) group in which 0 is the identity
element. Define a ? b 5 0 for all a and b  G. Then, (G, 1, ?) is a ring,
since ? is clearly associative and distributive over 1, for,
a ? (b 1 c) 5 0 5 0 1 0 5 (a ? b) 1 (a ? c)
and (a 1 b) ? c 5 0 5 0 1 0 5 (a ? c) 1 (b ? c)
for any a, b and c  G. The multiplication here is called the trivial mul-
tiplication in an abelian group. Rings of this type are called zero rings.

Q001-Algebra-111001_CH 09.indd 5 9/21/2011 4:52:26 PM

 9-6   Algebra – Abstract and Modern

3. Let M2(R) be the set of all 2 3 2 matrices over R (that is, with entries as real
a a  b b 
numbers).For any matrices A   11 12  and B   11 12  in M 2 ( R),
define a21 a22  b21 b22 

a  b a12  b12 
A B   11 11 
a21  b21 a22  b22 
a b  a b a11b12  a12 b22 
and A  B   11 11 12 21 
a21b11  a22 b21 a21b12  a22 b22 

Then, (M2(R), 1, ?) is a ring and is called the ring of 2 3 2 matrices

over R. One should carefully check the validity of the axioms I, II and
III of Definition 9.1.1 carefully. Note that the matrix in which all the
entries are zero is the zero element and is called the zero matrix. Also,
the additive inverse 2A of A is given by
a a12 
A   11 .
a21 a22 

4. Let C denote the set of all complex numbers; that C is the set of all
expressions of the form a 1 ib, where a and b are arbitrary real numbers.
For any x 5 a 1 ib and y 5 c 1 id in C, define
x 5 y ⇔ a 5 c and b 5 d
x 1 y 5 (a 1 b) 1 i(b 1 d)
x ? y 5 (ac 2 bd) 1 i(ad 1 bc),
where a 1 b, ac, etc. are the sums and products in the ring (R, 1, ?).
Then, (C, 1, ?) is a ring and is called the ring of complex numbers. Note
that 0 1 i0 is the zero element in C, which is also denoted simply by 0.
Further, the additive inverse 2x of x 5 a 1 ib is given by
2x 5 (2a) 1 i(2b)
As usual, we denote a 1 i0 by a, 0 1 ib by ib and 0 1 i1 by i. As per this
notation, note that ii 5 21.
5. In the above example, the operation 1 is defined coordinate wise. If
multiplication is also defined as coordinate wise (considering a and b
as first and second coordinates of a 1 ib), then C together with these
coordinate wise addition and multiplication forms a ring.
6. The procedure in 5 above can be generalised as follows. Let (R1, 1, ?),
(R2, 1, ?), …, (Rn, 1, ?) be any rings and
R 5 R1 3 R2 3 … 3 Rn 5 {(a1, a2, …, an) : ai  Ri for 1  i  n}.

Q001-Algebra-111001_CH 09.indd 6 9/21/2011 4:52:27 PM

 Rings   9-7

For any a 5 (a1, a2, …, an) and b 5 (b1, b2, …, bn) in R, Define

a 1 b 5 (a1 1 b1, a2 1 b2, …, an 1 bn)

and a ? b 5 (a1 ? b1, a2 ? b2, …, an ? bn).

Then, (R, 1, ?) is a ring and is called the product of R1, R2, …, Rn and is
n
denoted by  Ri or, simply, R1 3 R2 3 … 3 Rn. Note that, (0, 0, …, 0)
i1
is the zero element in the product, where 0 stands for the zero element in
each Ri. The additive inverse of any a 5 (a1, a2, …, an) is given by

2a 5 (2a1, 2a2, …, 2an).

7. Let (R, 1, ?) be any ring and X be any nonempty set. Let RX be the set
of all mappings of X into R. For any f and g  RX, define f 1 g and f ? g :
X → R by
(f 1 g)(x) 5 f (x) 1 g(x)
and (f ? g)(x) 5 f (x) ? g(x), for all x  X,

where the operations 1 and ? on the right side are those in R. Then, (RX,
1, ?) is a ring. The operations 1 and ? on RX defined above are called
point-wise addition and point-wise multiplication. The constant map
which maps each element of X onto the zero element in R will be the
zero element in RX and the additive inverse of f is defined by (2f )(x) 5
2f (x) for all x  X.
8. Let n be any positive integer and consider the group (Zn, 1n), where

Zn 5 {0, 1, 2, …, n 2 1}

and 1n is the addition modulo n. Recall that 1n is defined on Zn by

a  b if a  b  n
a n b   ,
a  b  n if a  b  n

for any a and b  Zn. Note that a 1n b is precisely the remainder obtained
by dividing the usual sum a 1 b by n. Now, extend this to the multiplica-
tion also, by defining
a n b 5 r, where 0  r , n, ab 5 qn 1 r, q and r  Z,
for any a and b  Zn. Note that a n b is precisely the remainder obtained
by dividing the usual product ab by n. This operation ?n is called the
multiplication modulo n. Zn is a finite set with n elements and the addi-
tion 1n and multiplication ?n modulo n are given in the following table
for n 5 6.

Q001-Algebra-111001_CH 09.indd 7 9/21/2011 4:52:27 PM

 9-8   Algebra – Abstract and Modern

16 0 1 2 3 4 5
0 0 1 2 3 4 5
1 1 2 3 4 5 0
2 2 3 4 5 0 1
3 3 4 5 0 1 2
4 4 5 0 1 2 3
5 5 0 1 2 3 4

(Z6, 16)

?6 0 1 2 3 4 5

0 0 0 0 0 0 0

1 0 1 2 3 4 5

2 0 2 4 0 2 4

3 0 3 0 3 0 3

4 0 4 2 0 4 2

5 0 5 4 3 2 1

(Z6, ?6)

It can be proved that (Zn, 1n, ?n) is a ring and is called the ring of integers
modulo n.
9. Let R be a set consisting of only one element, say R 5 {a}. Then, the
only way of defining binary operation on R is a 1 a 5 a and a ? a 5 a
and (R, 1, ?) is a ring in which a itself is the zero element and 2a 5 a.
This ring is called the trivial ring. When we say that R is a nontrivial
ring, it means that R contains atleast two elements.
10. Let X be any set and P(X) be the set of all subsets of X. For any A and B
in P(X), define

A 1 B 5 (A 2 B) ∪ (B 2 A)
and A ? B 5 A ∩ B.

Then, we shall prove that (P(X), 1, ?) is a ring. Recall that the empty set
 is the zero element and that (P(X), 1) is an abelian group. Clearly ?
is associative. Also, for any A, B and C in P(X), we have

Q001-Algebra-111001_CH 09.indd 8 9/21/2011 4:52:27 PM

 Rings   9-9

A ? (B 1 C) 5 A ∩ ((B 2 C) ∪ (C 2 B))
5 (A ∩ (B 2 C)) ∪ (A ∩ (C 2 B))
5 ((A ∩ B) 2 (A ∩ C)) ∪ ((A ∩ C) 2 (A ∩ B))
5 (A ∩ B) 1 (A ∩ C)
5 (A ? B) 1 (A ? C)
Since ∩ is a commutative operation, there is no need to verify that ? is
right distributive over 1. Thus, (P(X), 1, ?) is a ring.
In the following, we prove certain important elementary properties of
rings.

Theorem 9.1.1. Let (R, 1, ?) be a ring. Then, the following holds good for
any elements a, b and c in R.
1. 0a 5 0 5 a0, where 0 is the zero element in R.
2. a(2b) 5 2(ab) 5 (2a)b
3. (2a)(2b) 5 ab
4. a(b 2 c) 5 ab 2 ac
5. (a 2 b)c 5 ac 2 bc.

Proof:
1. We have
0 1 0a 5 0a 5 (0 1 0)a 5 0a 1 0a
and, by the cancellation law in the group (R, 1), it follows that 0 5 0a.
Also,
0 1 a0 5 a0 5 a(0 1 0) 5 a0 1 a0
and hence 0 5 a0.
2. Consider
ab 1 a(2b) 5 a(b 1 (2b)) 5 a0 5 0 (by (1))
and hence a(2b) is the additive inverse of ab. That is, 2(ab) 5 a(2b).
Similarly
ab 1 (2a)b 5 (a 1 (2a))b 5 0b 5 0 (by (1))
and therefore, (2a)b 5 2(ab).
3. We have (2a)(2b) 5 2((2a)(b))52(2(ab)) 5 ab
4. a(b 2 c) 5 a(b 1 (2c)) 5 ab 1 a(2c) 5 ab 2 ac
5. (a 2 b)c 5 (a 1 (2b))c 5 ac 1 (2b)c 5 ac 2 bc. b

Q001-Algebra-111001_CH 09.indd 9 9/21/2011 4:52:27 PM

 9-10   Algebra – Abstract and Modern

Definition 9.1.2. A ring (R, 1, ?) is said to be commutative if the multiplica-

tion ? is a commutative operation; that is,

a ? b 5 b ? a for all a and b  R.

Note that the additive operation 1 is always commutative in any ring R and
therefore, by a commutative ring R, we only mean that the multiplication is
commutative.

Definition 9.1.3. A ring (R, 1, ?) is said to be a ring with unity or a ring with
identity if there exists an element e in R such that

a ? e 5 a 5 e ? a for all a  R.

The element e, if exists, is unique.

Note that R always has the additive identity, namely the zero element 0. By
a ring with unity, we only mean that R has multiplicative identity also. The
multiplicative identity, if exists, is usually denoted by 1, with due respect to
the convention that the multiplicative identity in the real number system R is
1, and is called the unity or identity in the ring.

Theorem 9.1.2. Let (R, 1, ?) be a ring with unity. Then, R is trivial if and
only if 0 5 1 in R (that is, the additive identity coincides with the multiplica-
tive identity in the ring R).

Proof: If 0 5 1, then, for any a  R,

a 5 1a 5 0a 5 0

and hence R 5 {0}. The converse is trivial. b

Example 9.1.2
1. Each of the rings (Z, 1, ?), (Q, 1, ?), (R, 1, ?), (C, 1, ?), (Zn, 1n, ?n) for
any n  Z1 and (P(X), 1, ∩ ) is a commutative ring with unity. X is the
unity element in (P(X), 1, ∩), while 1 is the unity element in all these
other rings.
2. Any zero ring (see Example 9.1.1 (2)) is commutative and has no unity
element, unless it is trivial.
1 0
3. The ring M2(R) of 2 3 2 matrices is with unity, where  
0 1 
is the unity. However M2(R) is not commutative, for consider
0 2 1 0
A    and B   .
0 1  2 0

Q001-Algebra-111001_CH 09.indd 10 9/21/2011 4:52:28 PM

 Rings   9-11

0 2 1 0 4 0

Then, A  B         
0 1  2 0 2 0
1 0 0 2 0 2
and B  A         
2 0 0 1  0 4
and hence A ? B  B ? A.
4. The ring R 5 R1 3 R2 3 … 3 Rn given in Example 9.1.1 (6) is commu-
tative if and only if each Ri is commutative and R has unity if and only if
each Ri is so.
5. In Example 9.1.1 (7) also, RX is commutative if and only if R is commu-
tative. Further RX has unity if and only if R is so.

Worked Exercise 9.1.1. Prove that (Zn, 1n, ?n) is a commutative ring with
unity for any positive integer n, where 1n and ?n are addition and multiplica-
tion modulo n, respectively.

Answer: Recall that Zn 5 {0, 1, 2, …, n 2 1} and that we have already

proved in group theory that (Zn, 1n) is an abelian group.
Let a, b and c  Zn. Suppose that

a ?n b 5 r, r ?n c 5 s
b ?n c 5 t and a ?n t 5 u.

Then, ab 5 qn 1 r, rc 5 q1n 1 s

bc 5 pn 1 t and at 5 p1n 1 u,

where q, q1, p and p1  Z and r, s, t and u  {0, 1, 2, …, n 2 1}.

Since the usual multiplication of integers is associative, we have (ab)c 5
a(bc) and therefore

(qn 1 r)c 5 a(pn 1 t)

 qnc 1 q1n 1 s 5 apn 1p1n 1 u
i.e., (qc 1 q1)n 1 s 5 (ap 1 p1)n 1 u.

By the uniqueness of the quotient and the remainder in the division algo-
rithm, it follows that

qc 1 q1 5 ap 1 p1 and s 5 u.

In particular, (a ?n b) ?n c 5 r ?n c 5 s 5 u 5 a ?n t 5 a ?n (b ?n c).

Q001-Algebra-111001_CH 09.indd 11 9/21/2011 4:52:28 PM

 9-12   Algebra – Abstract and Modern

Thus, ?n is associative. Also, since the usual multiplication is commutative, ?n

is also commutative.
To prove the distributivity of ?n, let a, b and c  Zn

and b 1n c 5 x, a ?n (b 1n c) 5 y
a ?n b 5 z and a ?n c 5 v
so that b 1 c 5 qn 1 x, ax 5 q9n 1 y,
ab 5 pn 1 z and ac 5 p9n 1 v,

where q, q9, p and p9  Z and x, y, z and v  {0, 1, 2, …, n 2 1}.

Let z 1 v 5 jn 1 t, 0  t , n,
so that z 1n v 5 t.
Now, by the distributivity of the usual multiplication over the usual addition
in Z, we have

a(b 1 c) 5 ab 1 ac

and hence a(qn 1 x) 5 (pn 1 z) 1 (p9n 1 v).

Therefore, aqn 1 q9n 1 y 5 (p 1 p9)n 1 jn 1 t

 (aq 1 q9)n 1 y 5 (p 1 p9 1 j)n 1 t

and hence y 5 t, so that

a ?n (b 1n c) 5 y 5 t 5 z 1n v 5 (a ?n b) 1n (a ?n c).

Thus, ?n is distributive over 1n. Therefore, (Zn, 1n, ?n) is a commutative ring.
If n 5 1, then Zn is trivial. If n  1, then 1 is the unit element in Zn. Thus,
(Zn, 1n, ?n) is a commutative ring with unity.

Worked Exercise 9.1.2. Prove that, for any set X, (P(X), 1, ∩) is a commuta-
tive ring with unity.

Answer: We have already proved that (P(X), 1) is an abelian group, where

1 is the symmetric difference operation. For any subsets A, B and C of X, we
have (A ∩ B) ∩ C 5 A ∩ (B ∩ C) and A ∩ B 5 B ∩ A and hence (P(X), ∩) is a
semigroup. In Example 9.1.1 (10), we have proved the distributivity of ∩ over
1. Thus, (P(X), 1, ∩) is a commutative ring. Also, since

X ∩ A 5 A for all A  P(X),

X is the unity element in P(X). Thus, (P(X), 1, ∩) is a commutative ring with

unity.

Q001-Algebra-111001_CH 09.indd 12 9/21/2011 4:52:28 PM

 Rings   9-13

Worked Exercise 9.1.3. Let Z[i] 5 {a 1 ib : a and b  Z}. Then, prove

that (Z[i], 1, ) is a commutative ring with unity, where 1 and ? are the
usual addition and multiplication of complex numbers, defined in Example
9.1.1 (4).

Answer: Clearly, (Z[i], 1) is an abelian group. Let x, y and z  Z[i]. Then

x 5 a 1 ib, y 5 c 1 id and z 5 r 1 is

where a, b, c, d, r and s  Z. Then,

x ? (y ? z) 5 (a 1 ib) ((cr 2 ds) 1 i(cs 1 dr))

5 (a(cr 2 ds) 2 b(cs 1 dr)) 1 i(a(cs 1 dr) 1 b(cr 2 ds))
(x ? y) ? z 5 ((ac 2 bd) 1 i(ad 1 bc)) ? (r 1 is)
5 ((ac 2 bd)r 2 (ad 1 bc)s) 1 i((ac 2 bd)s 1 (ad 1 bc)r)
5 x ? (y ? z)

Therefore, (Z[i], ?) is a semigroup, clearly ? is commutative.

Also, x  (y 1 z) 5 (a 1 ib)  ((c 1 r) 1 i(d 1 s))
5 (a(c 1 r) 2 b(d 1 s)) 1 i(b(c 1 r) 1 a(d 1 s))
5 [(ac 2 bd) 1 i(ad 1 bc)] 1 [(ar 2 bs) 1 i(as 1 br)]
5x?y1x?z

Thus, ? distributes over 1. Also, 1(5 1 1 i0) is the unity in Z[i]. Thus, (Z[i],
1, ?) is a commutative ring with unity. Z[i] is called the ring of Gaussian
integers.
Recall the following from group theory.

Definition 9.1.4. Let (R, 1, ) be a ring. Then, for any a  R and n  Z, we

define na inductively as follows.

0 if n  0

na  ( n 1)a  a if n  0.

(n)(a) if n  0

That is, 0a 5 0, 1a 5 a, 2a 5 a 1 a, 3a 5 a 1 a 1 a, (22)a 5 2(2a) 5

(2a) 1 (2a), etc.

Q001-Algebra-111001_CH 09.indd 13 9/21/2011 4:52:29 PM

 9-14   Algebra – Abstract and Modern

Definition 9.1.5. For any element a in a ring R and positive integer n, we

define

a if n  1
a n   n1 .
a  a if n  1

That is a2 5 a  a, a3 5 a  a  a, etc. If the ring R has unity 1, then we define

a0 5 1. If the element a is multiplicatively invertible in R and a9 is its inverse
in R, then we define a2n 5 (a9)n for all n  0.

EXERCISES 9(a)

1. Which of the following are rings? Substantiate your answers (here 1 and  are
usual addition and multiplication of numbers).
(i) (Z1, 1, ?)
(ii) (6Z, 1, ?)
(iii) (E, 1, ?), where E is the set of even integers.
(iv) (O, 1, ?), where O is the set of odd integers.
(v) (P(X), ∩, ∪), where P(X) is the power set of a set X.
(vi) (P(X), ∪, ∩)
(vii) (P(X), 1, ∪)
(viii) (Z[ 2], , ), where Z[ 2]  {a  b 2 : a, b  Z}.
(ix) (Q[ 2], , ), where Q[ 2]  {a  b 2 : a and b are rational numbers}.
(x) (Zn, ?n, 1n)
(xi) (Q 2 {0}, 1, ?)
(xii) (R 2 Q, 1, ?)

2. Compute the following in the given rings.

(i) 13 ?15 8 in (Z15, 115, ?15)
(ii) 7 ?10 9 in (Z10, 110, ?10)
1 1
(iii) A ? A ? A in (M2(R), 1, ?) where A  .
1 1
(iv) 34 ?100 67 in (Z100, 1100, ?100)
(v) 8 110 (9 ?10 6) 1 (7 ?10 4) in (Z10, 110, ?10).
(vi) (A 1 B) ∩ C in (P(X), 1, ∩), where A 5 {2, 3, 4, 5}, B 5 {3, 5, 6, 7}
and C 5 {1, 2, 3}.

3. Prove that a ring (R, 1, ?) is commutative if and only if (a 1 b) (a 2 b) 5 a2 2

b2 for all a and b in R.

Q001-Algebra-111001_CH 09.indd 14 9/21/2011 4:52:29 PM

 Rings   9-15

4. Let (R, 1, ?) be a ring. Prove that the following are equivalent to each other.
(i) 1 distributes over ?.
(ii) R is trivial; that is, R 5 {0}.
(iii) (R, ?, 1) is a ring.
(iv) a 1 b 5 ab for all a and b in R.

5. Let a and b be two elements of a ring such that a  b 5 b  a. Prove the following
for any positive integer n.
(a 1 b)n 5 an 1 nc1 an21b 1 nc2 an22b2 1 … 1 ncn21 abn21 1 bn
n
n!
 ∑ ncr a nr b r , where ncr  .
r0 r !( n  r )!

6. For any prime p and a and b  Zp, prove that (a 1 b)p 5 ap 1 bp.

7. Prove that the commutativity of the operation 1 in a ring (R, 1, ?) is a conse-

quence of the other axioms of a ring.

8. Let (R, 1, ?) be a ring with unity 1. Define new operations ⊕ and  on R as

­follows for any a and b  R.

a ⊕ b 5 a 1b 1 1
a  b 5 a ? b 1 a 1 b.

Prove that (R, ⊕, ) is a ring with unity and that (R, 1, ) is commutative if and
only if (R, ⊕, ) is commutative.

9. Let (R, 1, ?) be a ring such that (a2 1 a)x 5 x(a2 1 a) for all a and x  R. Then
prove that (R, 1, ?) is a commutative ring.

10. Let (R, 1, ?) be a ring such that, for any a, b and c  R,

ab 5 ca ⇒ a 5 0 or b 5 c.

Then prove that (R, 1, ?) is a commutative ring.

11. Prove that a ring (R, 1, ?) is commutative if the group (R, 1) is cyclic.

12. Let (R, 1, ?) be a ring and n be an integer such that n  1 and xn 5 x for all x  R.
Then prove that, for any a and b  R,

ab 5 0 ⇔ ba 5 0.

13. Prove that the set {0, 2, 4} is a commutative ring with unity with respect to addi-
tion and multiplication modulo 6.

14. Give an example of a finite noncommutative ring.

Q001-Algebra-111001_CH 09.indd 15 9/21/2011 4:52:30 PM

 9-16   Algebra – Abstract and Modern

9.2 CERTAIN SPECIAL ELEMENTS IN RINGS

In any ring, we have the additive identity 0 and, in certain rings, there is unity
which is the multiplicative identity. In this section, we shall introduce certain
other special elements in a ring and discuss their properties.

Definition 9.2.1. An element a in a ring (R, 1, ?) is said to be an idempotent

if a ? a 5 a.

Example 9.2.1
1. The zero element 0 and the unity, if it exists, in any ring are idempotents.
2. 3 and 4 are idempotents in (Z6, 16, ?6), since 3 ?6 3 5 3 and 4 ?6 4 5 4; 5
is not an idempotent, since 5 ?6 5 5 1 in Z6.
3. In the ring Z of integers, 0 and 1 are the only idempotents.

Definition 9.2.2. A ring (R, 1, ?) in which every element is an idempotent is

called a Boolean ring.

Example 9.2.2
1. (Z2, 12, ?2), the ring of integers modulo 2 is a Boolean ring.
2. For any set X, (P(X), 1, ∩) is a Boolean ring, since A ∩ A 5 A for all
A ⊆ X.
3. For any set X, the set Z 2X of all mappings of X into Z2 is a Boolean ring
under the point-wise operations (see Example 9.1.1 (7)), since Z2 is a
Boolean ring.
Note that examples (2) and (3) given in Example 9.2.2 are not different.
They appear to be the same in the sense given below.

Theorem 9.2.1. Let X be any set. For any A ⊆ X, define

1 if x  A
 A : X → Z2 by  A ( x )   .
0 if x  A

Then, A  A is a bijection of P(X) onto Z 2X such that, for any A and B

in P(X)

A1B 5 A 12 B and A∩B 5 A 2 B.

Proof: If A and B  P(X) and a  A 2 B, then

A(a) 5 1 and B(a) 5 0

Q001-Algebra-111001_CH 09.indd 16 9/21/2011 4:52:30 PM

 Rings   9-17

and hence A  B if A  B. Thus, A  A is an injection. Further, if f  Z 2X

and A 5 f21({1}), then A 5 f. Thus, A  A is a bijection of P(X) onto Z 2X .
The other assertions follow from the definitions of the operations 1 on P(X)
and Z 2X , ∩ on P(X) and the point-wise operation ?2 on Z 2X .  b

Theorem 9.2.2. For any elements a and b in a Boolean ring (R, 1, ?),

a 1 a 5 0; that is, a 5 2a
and ab 5 ba

and hence every Boolean ring is commutative.

Proof: Let (R, 1, ?) be a Boolean ring and a and b  R. Then, consider

a 1 b 5 (a 1 b)2 5 (a 1 b) (a 1 b)
5 a2 1 ab 1 ba 1 b2
5 a 1 ab 1 ba 1 b.

From the cancellation laws in the group (R, 1), we have

0 5 ab 1 ba.

In particular, by taking a 5 b, we have

0 5 aa 1 aa 5 a 1 a

Also, ab 5 2(ba) 5 (2b)a 5 ba.

Thus, (R, 1, ?) is a commutative ring. b

Definition 9.2.3. Let (R, 1, ?) be a ring and a  R. Then, a is called a nilpo-

tent if an 5 0 for some positive integer n.

Example 9.2.3
1. The zero element 0 in any ring R is nilpotent, since 01 5 0
2. 6 is a nilpotent element in Z8, since
63 5 (6 ?8 6) ?8 6 5 4 ?8 6 5 0
3. Except 0, no element in the ring of integers is nilpotent.

Definition 9.2.4. A nonzero element a in a ring R is said to be a zero-divisor

if there exists a nonzero element b such that ab 5 0 5 ba.

Q001-Algebra-111001_CH 09.indd 17 9/21/2011 4:52:31 PM

 9-18   Algebra – Abstract and Modern

Note that, for two elements a and b in a ring R, it is quite possible

that ab 5 0 and ba  0. For consider the example (4) in the following
example.

Example 9.2.4

1. The ring (Z, 1, ?) of integers has no zero divisors, since, for any integers
a and b, ab 5 0 only when a 5 0 or b 5 0.
2. 2 and 3 are zero divisors in (Z6, 16, ?6).
3. If X is a set and A is a nonempty proper subset of X, then A ∩ (X 2 A) 5 ,
A   and X 2 A   and hence A is a zero divisor in (P(X), 1, ∩).
4. Consider the ring M2(R) of 2 3 2 matrices. Let

1 0 0 0
A    and B   .
1 0 1 1 
0 0 0 0 0 0
Then A  B    and B  A    .
0 0 2 0 0 0

However, there is some other C in M2(R) such that A  C 5 0 5 C  A; for,

0 0 
take C   . Then, AC 5 0 5 CA and therefore A is zero divisor.
1 1

Definition 9.2.5. Let (R, 1, ?) be a ring and a  R. If R is a ring with unity

and a has multiplicative inverse in R, then a is called a unit or multiplicatively
invertible or, simply, invertible. That is, a is a unit in R if there exists b  R
such that ab 5 1 5 ba.
Note that, in any ring (R, 1, ?), every element a has additive inverse,
namely 2a. However, an element in a ring with unity may not possess mul-
tiplicative inverse. Elements possessing multiplicative inverse need special
attention. First, Let us consider the following example.

Example 9.2.5

1. In any ring, the unity (if it exists) is a unit.

2. In the ring of integers Z, 1 and 21 are the only units and, for each of
them, the multiplicative inverse is itself.
3. In the ring of rational numbers, or in the ring of real numbers, or in the
ring of complex numbers, every nonzero element is a unit.
4. The zero element is a unit in a ring R if and only if R is trivial; since
0a 5 0 5 a0 for all a  R.

Q001-Algebra-111001_CH 09.indd 18 9/21/2011 4:52:31 PM

 Rings   9-19

If a is a unit in a ring R, then the element b  R such that ab 5 1 5 ba is

unique and is denoted by a21 and is called the inverse of a. Note that the
­additive inverse of a is denoted by 2a, while the multiplicative inverse (if
exists) of a is denoted by a–1.

Theorem 9.2.3. Let (R, 1, ) be a ring with unity. If a and b are units in R,
then so is their product ab and (ab)21 5 b21a21. Also, the set U(R) of all units
in R forms a group under multiplication.

Proof: If aa21 5 1 5 a21a and bb21 5 1 5 b21a, then (ab)(b21a21) 5 a(bb21)

a21 5 a1a21 5 1 and (b21a21)(ab) 5 b21(a21a)b 5 b121b 5 1 and hence ab
is a unit and (ab)21 5 b21a21. Therefore, the multiplication in R, restricted to
U(R), is a binary operation on U(R) and is clearly associative. Also, the unity
1 will be the identity in (U(R), ?). Further, if a is a unit, then so is its inverse
and (a21)21 5 a. Thus, (U(R), ?) is a group. b

Worked Exercise 9.2.1. Let 1 , n  Z and Zn be the ring of integers modulo

n. For any a  Zn, prove that a is a unit in Zn if and only if a and n are rela-
tively prime.

Answer: Consider the following

(a, n) 5 1 ⇔ xa 1 yn 5 1 for some integers x and y

⇔ ba 1 zn 51 for some b, z  Z with 0 , b , n
(use division algorithm to write x 5 qn 1 b)
⇔ b ?n a 5 1, b  Zn
⇔ a is a unit in (Zn, 1n, ?n).

Worked Exercise 9.2.2. For any n  1, prove that any nonzero element in the
ring Zn is either a unit or a zero divisor.

Answer: Let 0 , a , n. Suppose that a is not a unit in Zn. Then, (a, n)  1

and let d 5 (a, n). Then, d is a common divisor for a and n and hence both a
d
and dn are integers and

n a
a   = n  n 0.
d  d 

Therefore, a n ( n )  0 in Zn and n is nonzero in Zn. Therefore, a is a zero

d d
divisor in Zn.

Q001-Algebra-111001_CH 09.indd 19 9/21/2011 4:52:32 PM

 9-20   Algebra – Abstract and Modern

Worked Exercise 9.2.3. Let a and b be elements in a ring R such that ab 5 ba.
Prove the following:
1. a 1 b is a nilpotent if a and b are nilpotents.
2. ab is a nilpotent if a or b is a nilpotent.

Answer:
1. Suppose that a and b are nilpotents. Then, there exist positive integers n
and m such that an 5 0 5 bm. Now, since ab 5 ba, we have
(a 1 b)m1n 5 am1n 1(m 1 n)c1am1n21b 1 … 1 (m 1 n)cm1nbm1n
mn
 ∑ ( m  n)Cr a mnr b r .
r0

since as 5 0 for all s ≥ n and bt 5 0 for all t ≥ m and since, for any 0  r
 m 1 n, either m 1 n 2 r ≥ n or r ≥ m (otherwise (m 1 n 2 r) 1 r ,
n 1 m, an absurd), we get that am1n2r 5 0 or br 5 0 for all 0  r  m 1
n and therefore (a 1 b)m1n 5 0, so that a 1 b is a nilpotent.
2. If an 5 0, then (ab)n 5 anbn 5 0bn 5 0. Therefore, if a is a nilpotent, then
ab (5 ba) is also a nilpotent.

Worked Exercise 9.2.4. Prove that 1, 21, i and 2i are the only units in the
ring Z[i] of Gaussian integers.

Answer: Let x 5 a 1 ib be a unit in Z[i]. Then, there exists y 5 c 1 id in Z[i]

such that xy 5 1. Here a, b, c and d are integers. Then,

1 5 |xy|2 5 |x|2|y|2 5 (a2 1 b2)(c2 1 d2)

Therefore, a2 1 b2 5 1 5 c2 1 d2 (since a2 1 b2 ≥ 1). Again, since a2 ≥ 0 and

b2 ≥ 0, we get that
a 5 1 or 21 and b 5 0
(or) a 5 0 and b 5 1 or 21
so that x 5 1 or 21 or x 5 i or 2i.

EXERCISE 9(b)

1. Determine all the zero divisors, nilpotents, idempotents and units in each of the
following rings
(i) The ring R of real numbers.
(ii) Z 3 R, where Z is the ring of integers.

Q001-Algebra-111001_CH 09.indd 20 9/21/2011 4:52:32 PM

 Rings   9-21

(iii) Z 3 Z.
(iv) The ring Z24 of integers modulo 24.
(v) Z15.
(vi) Z12 3 Z.
(vii) (P(X), 1, ∩), for any set X.
(viii) Z4 3 Z9.
2. In any nontrivial ring with unity, prove that no zero divisor is a unit.
3. Let a be a nonzero element in a commutative ring R. Prove that a is not a zero divi-
sor if and only if a satisfies the following cancellation law for any b and c in R:
ab 5 ac ⇒ b 5 c.
4. Let R be a Boolean ring with unity. Prove that the unity is the only unit in R0 and
the zero is the only nilpotent in R.
5. Let n be any integer greater than 1. The content of n is defined to be the product
of all distinct primes dividing n and is denoted by c(n). Prove that a  Zn is a
nilpotent if and only if c(n) divides a.
6. Using 5 above, derive a formula for the number of nilpotents in the ring Zn of
integers modulo n.
7. For any integers a and n with 0  a , n, prove that a is an idempotent in Zn if
and only if a(a 2 1) is a multiple of n.
8. Let R1, R2, …, Rn be rings and R 5 R1 3 R2 3 … 3 Rn. For any a 5 (a1, a2, …,
an)  R, prove that a is a nilpotent (idempotent) in R if and only if each ai is a
nilpotent (respectively idempotent) in Ri for 1  i  n.
9. In 8 above, when each Ri is a ring with unity, prove that (a1, a2, …, an) is a unit
in R1 3 R2 3 … 3 Rn if and only if ai is a unit in Ri for each 1  i  n.
10. If a and b are idempotents in a commutative ring R, prove that ab is also an
idempotent. Can a 1 b be an idempotent? If not, give a counter example.
11. Find all the solutions of x2 1 2x 1 4 5 0 in the ring Z6.
12. Prove that 0 is the only solution of x2 5 0 in a ring R if and only if R has no
nonzero nilpotents.
13. For any prime number p, prove that the set of all nonzero elements in Zp forms a
group under multiplication modulo p.
14. Prove that any nonzero nilpotent in any ring R is a zero divisor in R.
15. Let f be the Euler-totient function. Prove that the multiplicative group of units
in Zn is of order f (n), for any integer n  1.
16. Let R be a ring with unity and a  R be a nilpotent. Then prove that 1 1 a is a unit.

Q001-Algebra-111001_CH 09.indd 21 9/21/2011 4:52:32 PM

 9-22   Algebra – Abstract and Modern

17. Let R be a ring with unity and without zero divisors. For any a and b in R, prove
that ab 5 1 if and only if ba 5 1 and that a2 5 1 if and only if a 5 1 or 21.

18. Let R be a ring without nonzero nilpotents and a be an idempotent in R. Prove

that ax 5 xa for all x  R.

9.3 THE CHARACTERISTIC OF A RING

It is well known that na  0 for any positive integer n and any nonzero
integer a. That is, there is no positive integer n such that na 5 0 for all ele-
ments a in the ring Z of integers. However, there are rings for which there
exists a positive integer n such that na 5 0 for all elements a in the ring. For
example, consider a positive integer m and the ring Zm of integers modulo
m. Then, ma 5 0 for all a  Zm. In fact if n is any positive integral multiple
of m, then na 5 0 for all a  Zn. Recall that, for any element a in a ring R,
the order of a in the group (R, 1) is precisely the smallest positive integer
(if exists) n such that na 5 0. In this section, we discuss the existence of a
common positive integer n such that na 5 0 for all elements a in the ring.
First recall that, for any positive integer n and for any element a in a ring R,
we have defined na inductively by

a if n  1
na   .
( n −1)a  a if n  1

Definition 9.3.1. Let (R, 1, ?) be a ring. If there is no positive integer n such

that na 5 0 for all a  R, then the characteristic of R is defined to be zero.
Otherwise, the smallest positive integer n such that na 5 0, for all a  R, is
called the characteristic of R and is denoted by char(R).
Note that na 5 0 if n is the integer zero or a is the zero element in the ring
R. If char(R) 5 0, then, for each positive integer n, there exists an element a
in the ring R such that na 5 0. If char(R) 5 n  0, then na 5 0 for all a  R
and n is the least such positive integer.

Theorem 9.3.1. Let (R, 1, ?) be ring and char(R) 5 n  0. Then, for any
integer m,

ma 5 0 for all a  R if and only if n divides m.

Proof: Let m be any integer. If n divides m, then m 5 nr for some integer r

and hence, for any a  R,

ma 5 (nr)a 5 r(na) 5 r0 5 0.

Q001-Algebra-111001_CH 09.indd 22 9/21/2011 4:52:33 PM

 Rings   9-23

Conversely suppose that ma 5 0 for all a  R. By the division algorithm, we

can express m as

m 5 qn 1 r, where q and r  Z and 0  r , n.

Then, for any a  R, we have

0 5 ma 5 (qn 1 r)a 5 q(na) 1 ra 5 0 1 ra 5 ra.

Since n is the least positive integer such that na 5 0 for all a  R and since
r , n, it follows that r cannot be positive. Since 0  r, we get that r 5 0 and
hence m 5 qn. Thus, n divides m. b

Theorem 9.3.2. Let (R, 1, ?) be a ring with unity 1. Then, the characteristic
of R is precisely the order of the unity in the group (R, 1).

Proof: This follows from the fact that, for any a  R and n  Z,

na 5 n(1a) 5 (n1)a

and that na 5 0 for all a  R if and only if n1 5 0. b

Example 9.3.1
1. The characteristic of each of the rings Z, Q, R and C is zero, since for any
integer n  0 and for any nonzero real or complex number a, na  0.
2. char(Zn) 5 n for any positive integer n, where Zn is the ring of integers
modulo n.
3. char(Zn 3 Zm) is the least common multiple of m and n for positive inte-
gers m and n.

Worked Exercise 9.3.1. Let R and S be rings of characteristic m and n, respec-

tively. Then prove that the characteristic of the product ring R 3 S is the least
common multiple of m and n.

Answer: We have char(R) 5 m and char(S) 5 n. First let us assume that m  0

and n  0. Let r be the least common multiple of m and n and

r 5 ms and r 5 nt

for some positive integers s and t. Then, for any element (a, b) in R 3 S,
we have

Q001-Algebra-111001_CH 09.indd 23 9/21/2011 4:52:33 PM

 9-24   Algebra – Abstract and Modern

r(a, b) 5 (ra, rb)

5 ((ms)a, (nt)b)
5 (s(ma), t(nb))
5 (s0, t0) 5 (0, 0)

Therefore, 0 , char(R 3 S)  r. Put char(R 3 S) 5 k.

Then, for any a  R,

(ka, 0) 5 k(a, 0) 5 (0, 0) and hence ka 5 0.

By Theorem 9.3.1, char(R) divides k. Similarly, we can prove that char(S)

divides k. Therefore, k is a common multiple of m and n and hence r  k.
Thus,

char(R 3 S) 5 r 5 l.c.m.{m, n}.

On the other hand, if char(R 3 S) 5 p  0, then

(pa, pb) 5 p(a, b) 5 (0, 0)

and hence pa 5 0 and pb 5 0 for all a  R and b  S so that char(R)  0

and char(S)  0.

Worked Exercise 9.3.2. Determine the characteristic of Z12 3 Z.

Answer: If n is any positive integer, then

n(0, a) 5 (n0, na) 5 (0, na)  (0, 0)

for any a  Z and (0, a)  Z12 3 Z. Therefore, char(Z12 3 Z) 5 0.

EXERCISE 9(c)

1. Find the characteristic of each of the following rings.

(i) The ring E of even integers
(ii) 5Z
(iii) Z12 3 Z18
(iv) Z 3 Z24
(v) (P(X), 1, ∩) for any set X
(vi) Z30
X
under the point-wise operations, where X is any set.

Q001-Algebra-111001_CH 09.indd 24 9/21/2011 4:52:33 PM

 Rings   9-25

(vii) The trivial ring R 5 {0}.

(viii) The ring Z[i] of Gaussian integers.
(ix) ZX under the point-wise operations, where X is any set.
(x) Z6 3 Z20 3 Z14
2. Let R be a commutative ring with unity and char(R) 5 3. For any a and b  R,
compute the following and simplify.
(i) (a 1 b)9
(ii) (a 1 b)3
(iii) (a 1 b)6
(iv) (a 1 b)12
3. Prove that a ring R is trivial if and only if char(R) 5 1.
4. Prove that the characteristic of any finite ring is positive.
5. Let R be a commutative ring of characteristic 2 and E(R) be the set of idempo-
tents in R. Prove that E(R) is a ring under the operations on R.
6. Give an example of a ring R of characteristic 5 such that every nonzero element
in R is a unit.
7. Let R be a commutative ring with unity in which each nonzero element is a unit.
If char(R) 5 2 and R has atleast three elements, then prove that there exist ele-
ments a and b in R such that
(a 1 b)3  a3 1 b3.
8. In Exercise 7 above, suppose that char(R) is a prime number p and
A 5 {a  R : ap 5 a}.
Then prove that A is a ring under the operations on R and that every nonzero
element in A is a unit in A.
9. Let R be a ring with identity and char(R) 5 n  0. If n is not prime, prove that R
has zero divisors.
10. Let R be a finite ring and R 5 {a1, a2, …, ar}. Let O(ai) be the order of ai in the group
(R, 1). Prove that char(R) is the least common multiple of O(a1), O(a2), …, O(ar).
11. Prove the characteristic of a finite ring R that divides |R|.
12. Let R be a finite ring with unity and a and b  R. Prove that ab 5 1 if and only
if ba 5 1.

9.4 SUBRINGS
In this section, we deal with the situation where a subset of a ring constitutes a
ring again. Recall the set Z of integers is a subset of the ring (R, 1, ?) of real

Q001-Algebra-111001_CH 09.indd 25 9/21/2011 4:52:33 PM

 9-26   Algebra – Abstract and Modern

numbers and Z itself is a ring under the addition and multiplication of real
numbers. This is abstracted in the following definition.

Definition 9.4.1. Let (R, 1, ?) be a ring. A nonempty subset S of R is called

a subring of R if S is itself is a ring under the operations 1 and ? on R
restricted to S.
In other words, S is a subring of R if S is a subgroup of (R, 1) and S is a
subsemigroup of (R, ?) (that is, ab  S whenever a and b  S). The reason for
this is that the distributive laws and the associativity of the multiplication ?
hold automatically for the elements of S as a consequence of their validity in
the ring R. The following is a simpler characterization of a subring and whose
proof is trivial.

Theorem 9.4.1. Let S be a nonempty subset of a ring R. Then, S is a subring

of R if and only if

a and b  S ⇒ a 2 b  S and ab  S.

Clearly {0} and the whole of R are subrings of any ring R and are called
trivial subrings and all other subrings (if they exist) are called nontrivial sub-
rings. A subring S of R is called a proper subring if S  R.

Example 9.4.1
1. Z is a subring of the ring (Q, 1, ?) of rational numbers, Q is a subring of
the ring (R, 1, ?) of real numbers and R is a subring of the ring (C, 1, ? )
of complex numbers.
2. Let Y be a subset of a set X. Then, P(Y), the power set of Y, is a subring
of (P(X), 1, ∩).
3. For any nonnegative integer n, the set nZ of all integral multiples of n, is
a subring of the ring (Z, 1, ?) of integers. In particular, the set E of even
integers is a subring of (Z, 1, ?).
Note that a ring R may possess the unity (multiplicative identity) while a
subring may not possess and, even when a subring possesses unity then it may
be different from that of R. Consider the following examples.

Example 9.4.2
1. The set E of even integers is a subring of the (Z, 1, ?) of integers. Z has
unity, while E has no unity (there is no even integer e such that ea 5 a
for all even integers).
2. Let X be a set and Y be a proper subset of X. Then, P(Y) is a subring of (P(X),
1, ∩). X and Y are unit elements in P(X) and P(Y), respectively and Y  X.

Q001-Algebra-111001_CH 09.indd 26 9/21/2011 4:52:33 PM

 Rings   9-27

The following is a routine verification. Observe that any subring S of a ring

R contains the zero element 0, since S is a subgroup of (R, 1).

Theorem 9.4.2. The intersection of any class of subrings of a ring R is again

a subring of R.

Worked Exercise 9.4.1. Let (R, 1, ?) be a ring and define

C(R) 5 {a  R : ax 5 xa for all x  R}.

Then prove that C(R) is a subring of R. C(R) is called the centre of R.

Answer: Since 0x 5 0 5 x0 for all x  R, 0  C(R) and therefore C(R) is a

nonempty subset of R. For any a and b  C(R) and x  R, we have

(a2b)x 5 ax 2 bx 5 xa2xb 5 x(a2b)

and (ab)x 5 a(bx) 5 a(xb) 5 (ax)b 5 (xa)b 5 x(ab)

and hence a 2 b  C(R) and ab  C(R). Thus, C(R) is a subring of R.

Worked Exercise 9.4.2. Let S be a subring of a ring R and char(R)  0. Then

prove that char(S) is a positive divisor of char(R).

Answer: Let char(R) 5 n, we are given that n  0 and na 5 0 for all a  R

and, in particular, na 5 0 for all a  S. Therefore, char(S)  0 and, by ­Theorem
9.3.1, char(S) divides n.

Worked Exercise 9.4.3. Let R be a ring. For any a  R, let

C(a) 5 {x  R : ax 5 xa}.

Prove that C(a) is a subring of R and that C ( R)  ∩ C ( a).

aR

Answer: Clearly a and 0  C(a) and hence C(a) is a nonempty subset of R.

For any x and y  C(a), we have

a(x 2 y) 5 ax 2 ay 5 xa 2 ya 5 (x 2 y)a
and a(xy) 5 (ax)y 5 (xa)y 5 x(ay) 5 x(ya) 5 (xy)a

and there x 2 y  C(a) and xy  C(a). Thus, C(a) is a subring of R. Also,

x  C(R) ⇔ xa 5 ax for all a  R

⇔ x  C(a) for all a  R.

Therefore, C ( R)  ∩ C ( a).
aR

Q001-Algebra-111001_CH 09.indd 27 9/21/2011 4:52:34 PM

 9-28   Algebra – Abstract and Modern

EXERCISE 9(d)

1. Determine all the subrings of each of the following rings.

(i) (Z2, 12, ?2)
(ii) (Z7, 17, ?7)
(iii) (Z10, 110, ?10)
(iv) (Z, 1, ?)

2. If S is a subring of a ring R and T is a subring of S, then prove that T is a subring

of R.

3. Prove that S is a subring of the ring Z of integers if and only if S 5 nZ for some
nonnegative integer n.

4. Let n be a positive integer and Zn be the ring of integers modulo n. Prove that S
is a subring of Zn if and only if S 5 {0, m, 2m, …, (r 2 1)m} for some divisor m
of n and rm 5 n, r  0.

5. Let X be a subset of a ring R and (X) be the intersection of all subrings of R con-
taining X. Prove that (X) is the smallest subring of R containing X. (X) is called
the subring of R generated by X.

6. In Exercise 5 above, describe all the elements of (X).

7. Let S be a subring of a ring with unity 1 and 1  S. If a  C(R), prove that

(S ∪ {a}) 5 {s0 1 s1a 1 … 1 snan : n ≥ 0 and si  S}.
8. Give an example of a ring R without unity and of a subring S of R such that S is
with unity.

9. Let S be a nontrivial subring of a ring R and 19 be the unity in S such that 19 is

not the unity in R. Then prove that 19 is a zero divisor in R.

10. Let S and T be subrings of a ring R. Then prove that S ∪ T is a subring of R if and
only if either S ⊆ T or T ⊆ S.

11. Let # be a class of subrings of a ring R such that, for any S1 and S2  #, there exist
S3  # containing both S1 and S2. Then prove that ∪ S is a subring of R.
S 

12. Let S be a nonempty subset of a finite ring R. Prove that S is a subring of R if and
only if
a and b  S ⇒ a 1 b and ab  S.

13. Let R be a ring which has no nonzero nilpotent elements. Prove that every idem-
potent in R is in the centre C(R).

14. Let R be a ring such that a2 1 a  C(R) for all a  R. Prove that R is a commuta-
tive ring.

Q001-Algebra-111001_CH 09.indd 28 9/21/2011 4:52:34 PM

 Rings   9-29

15. Let (R, 1, ) be a ring of characteristic n  0 and (Zn, 1n, ?n) the ring of integers
modulo n. Define the operations 1 and  on R 3 Zn by

(x, a) 1 (y, b) 5 (x 1 y, a 1n b)
and (x, a)  (y, b) 5 (xy 1 ay 1 bx, a ?n b).

Prove that (R 3 Zn, 1, ) is a ring in which R 3 {0} is a subring.

9.5 HOMOMORPHISMS OF RINGS

A homomorphism from a ring R into a ring R9 is, as one might guess, a
function f : R → R9 which preserves both the ring operations. This amounts
to applying the familiar homomorphism concept to the underlying additive
group (R, 1) and the multiplicative semigroup (R, ). In the following, we
give a precise definition.

Definition 9.5.1. Let R and R9 be rings. A function f : R → R9 is called a

homomorphism of R into R9 if

f (a 1 b) 5 f (a) 1 f (b)
and f (a  b) 5 f (a)  f (b) for all a and b  R.

Note that the symbols 1 and  occurring on the left sides of the above
equations denote the addition and multiplication in R where as 1 and  occur-
ring on the right sides denote those in R9. This use of the same symbols for
the operations of addition and multiplication in two different rings need cause
no confusion provided the reader gives careful attention to the context if the
notation is employed. The following is the usual terminology we apply, as in
the case of group theory.

Definition 9.5.2
1. An injective homomorphism is called a monomorphism or an embedding.
2. A surjective homomorphism is called an epimorphism.
3. A bijective homomorphism is called an isomorphism.
4. A homomorphism of a ring R into itself is called an endomorphism of R.
5. An isomorphism of a ring R onto itself is called an automorphism of R.
6. A ring R is said to be isomorphic with a ring R9 and denote this by
R > R9 if there is an isomorphism of R onto R9. The following examples
should help us for a better understanding of the above concepts.

Q001-Algebra-111001_CH 09.indd 29 9/21/2011 4:52:34 PM

 9-30   Algebra – Abstract and Modern

Example 9.5.1
1. Let R and R9 be any rings and define f : R → R9 by f (x) 5 0 for all x  R.
Then, for any a and b in R,
f (a 1 b) 5 0 5 0 1 0 5 f (a) 1 f (b)
f (ab) 5 0 5 0  0 5 f (a)  f (b)
and therefore f is a homomorphism, which is called the trivial or zero
homomorphism. Note that this is not a monomorphism unless R 5 {0}
and is not an epimorphism unless R9 5 {0}.
2. The identity mapping IR of a ring R onto itself is an automorphism of R.
3. Let R be any ring and X be any nonempty set. Consider the ring RX of
all mappings of X into R under the point-wise operations (refer Example
9.1.1 (7)). For any x  X, define ax : RX → R by

ax(f ) 5 f (x) for all f  RX.

Then, for any f and g  RX, we have

ax(f 1 g) 5 (f 1 g)(x) 5 f (x) 1 g(x) 5 ax(f ) 1 ax(g)

and   ax(f  g) 5 (f  g)(x) 5 f (x)  g(x) 5 ax(f )  ax(g)

and hence ax is a homomorphism of RX into R which is called the evalu-

ation homomorphism at x. It can be verified that ax is an epimorphism.
4. Let n be a positive integer and consider the ring Z of integers and the
ring Zn of integers modulo n. Define f : Z → Zn by

f (a) 5 r, where a 5 qn 1 r, q and r  Z1 and 0 # r , n.

That is, f (a) is precisely the remainder obtained by dividing a with n.

Then, f is an epimorphism (see Worked Exercise 9.5.1).
In the following, we exhibit a few elementary properties of homomor-
phisms of rings and prove that some of the structural features are preserved
under homomorphisms of ring. First of all, the following is a simple conse-
quence of the fact at a homomorphism of rings is a homomorphism of the
underlying additive groups.

Theorem 9.5.1. Let R and R9 be rings and f : R → R9 a homomorphism of

rings. Then, the following holds.
1. f (0) 5 0.
2. f (2a) 52f (a) for all a  R.
3. f (a 2 b) 5 f (a) 2 f (b) for all a and b  R.

Q001-Algebra-111001_CH 09.indd 30 9/21/2011 4:52:34 PM

 Rings   9-31

Theorem 9.5.2. Let f : R → R9 be an epimorphism of rings with unity. Then,

f (1) 5 1 and, for any unit a in R, f (a) is a unit in R9 and f (a21) 5 f (a)21.

Proof: Let x9  R9. Since f is an epimorphism, we can choose x  R such

that f (x) 5 x9. Now,

f (1)x9 5 f (1)f (x) 5 f (1x) 5 f (x) 5 x9

and x9f (1) 5 f (x)f (1) 5 f (x1) 5 f (x) 5 x9

and hence f (1) is the multiplicative identity in R9 so that f (1) 5 1, the unity in
R9. Next, let a be a unit in R. Then, there is an element a21 in R such that aa21
5 1 5 a21a. By applying f to these, we get that

f (a)f (a21) 5 f (aa21) 5 f (1) 5 1

and f (a21)f (a) 5 f (a21a) 5 f (1) 5 1

and therefore f (a) is multiplicatively invertible in R9 and f (a)21 5 f (a21). b

Theorem 9.5.3. Let f : R → R9 be a homomorphism of rings. Then, f (S) is

a subring of R9 for any subring S of R and f21(S9) is a subring of R for any
subring S9 of R9.

Proof: Let S and S9 be subrings of R and R9, respectively since f (0) 5 0 and S
and S9 contain zero elements, it follows that f (S)  f and f21(S9)  f. Now,

x and y  f (S) ⇒ x 5 f (a) and y 5 f (b) for some a and b  S

⇒ x 2 y 5 f (a) 2 f (b) 5 f (a 2 b) and a 2 b  S
and   xy 5 f (a)f (b) 5 f (ab) and ab  S
⇒ x 2 y and xy  f (S)

and therefore f (S) is a subring of R9. Also,

a and b  f21(S9) ⇒ f (a) and f (b)  S9

⇒ f (a 2 b) 5 f (a) 2 f (b)  S9
and f (ab) 5 f (a)f (b)  S9
⇒ a 2 b and ab  f21(S9)

and therefore f21(S9) is a subring of R. b

We discuss some more important properties of homomorphisms of rings
after introducing the concepts of ideals and quotient rings later.

Q001-Algebra-111001_CH 09.indd 31 9/21/2011 4:52:35 PM

 9-32   Algebra – Abstract and Modern

Worked Exercise 9.5.1. Let n be a positive integer and define f : Z → Zn by

f (a) 5 r, where r is the remainder obtained by dividing a with n. Then prove
that f is an epimorphism of rings.

Answer: For any a  Z, we have

f (a) 5 r, where a 5 qn 1 r, q, r  Z and 0 # r , n.

If r  Zn, then 0 # r , n and clearly f (r) 5 r.

Therefore, f is a surjection.
Let a and b  Z and

a 5 qn 1 r and b 5 q9n 1 s,

where q, q9, r, s  Z, 0 # r , n and 0 # s , n. Then, f (a) 5 r and f (b) 5 s.

Now,

a 1 b 5 qn 1 r 1 q9n 1 s 5 (q 1 q9)n 1 (r 1 s)
( q  q)n  ( r n s) if r  s  n
 
( q  q1)n  ( r n s) if r  s  n

and therefore f (a 1 b) 5 r 1n s 5 f (a) 1n f (b). Also,

   ab 5 (qn 1 r)(q9n 1 s)
           5 (qq9n 1 qs 1 q9r)n 1 rs

If rs 5 un 1 t, 0 # t , n, then

ab 5 (qq9n 1 qs 1 q9r 1 u)n 1 t

   and hence f (ab) 5 t 5 r n s 5 f (a) n f (b).

Thus, f is a homomorphism of rings. Since f is a surjection also, it follows that

f is an epimorphism.

Worked Exercise 9.5.2. Prove that the composition of homomorphism of rings

is again a homomorphism.

Answer: Let f : R → R9 and g : R9 → R be homomorphisms of rings. Then,

g o f : R → R is a function and, for any a and b  R, we have

(g o f )(a 1 b) 5 g(f (a 1 b)) 5 g(f (a) 1 f (b))5 g(f (a)) 1 g(f (b))
and (g o f )(ab) 5 g(f (a)f (b)) 5 g(f (a))g(f (b))

and therefore g o f is a homomorphism.

Q001-Algebra-111001_CH 09.indd 32 9/21/2011 4:52:35 PM

 Rings   9-33

Worked Exercise 9.5.3. Define f : Z → Z2 by

0 if a is even
f (a)   .
1 if a is odd

Then prove that f is an epimorphism of rings.

Answer: This is a special case of Worked Exercise 9.5.1 by taking n 5 2.

However, the following is an independent proof.
Since f (0) 5 0, f (1) 5 1 and Z2 5 {0, 1}, f is a surjection. Also, note that,
for any integers a and b, a 1 b is even if and only if both a and b are even or
both a and b are odd.
Since 0 1 0 5 0 5 1 1 1, it follows that

f (a 1 b) 5 f (a) 1 f (b) for all a, b  Z.

Further, ab is even if and only if atleast one of a and b is even. Since

0 ? 1 5 0 5 1 ? 0 5 0 ? 0 and 1 ? 1 5 1, it follows that

f (ab) 5 f (a)f (b) for all a and b  Z.

Thus, f is an epimorphism of rings.

Worked Exercise 9.5.4. Determine all the endomorphisms of the ring Z of

integers.

Answers: Let f : Z → Z be a homomorphism of rings.

Since
f (1) 5 f (1  1) 5 f (1)f (1),
we have f (1)(f (1) 2 1) 5 0 and hence f (1) 5 0 or f (1) 5 1. If f (1) 5 0, then,
for any a  Z,

f (a) 5 f (a ? 1) 5 f (a) ? f (1) 5 f (a)0 5 0

and hence f is the zero homomorphism. On the other hand, if f (1) 5 1, then
for any 0 , a  Z,

f (a) 5 f (1 1 1 1 … 1 1) (a times)
        5 f (1) 1 f (1) 1 … 1 f (1) (a times)
5 1 1 1 1 … 1 1 (a times)
5a

and for 0  a  Z, 2a  0 and

f (a) 5 f (2(2a)) 5 2f (2a) 5 2(2a) 5 a.

Q001-Algebra-111001_CH 09.indd 33 9/21/2011 4:52:35 PM

 9-34   Algebra – Abstract and Modern

Therefore, if f (1) 5 1, then f (a) 5 a for all a  Z.

Thus, the zero homomorphism and the identity homomorphism are the only
homomorphisms of Z into itself. These two are the only endomorphisms of
the ring Z. In this context, recall that there are several endomorphisms of the
group (Z, 1).

EXERCISE 9(e)

For any rings R and R9, let Hom(R, R9) denote the set of all ring homomorphisms of R
into R9 and End(R) denote the set of all endomorphisms of the ring R.

1. Determine all the members of each of the following:

(i) End(Q), where Q is the ring of rational numbers.
(ii) End(Zn), for any positive integer n.
(iii) Hom(R, Z), where R is the ring of real numbers.
(iv) Hom(Z, R)
(v) Hom(Zn, Q)
(vi) Hom(Q, Zn)

2. State which of the following are true. Substantiate your answers.

(i) Every monomorphism of Zn into Zn is an isomorphism.
(ii) For any integers 0  n  m, there exists a monomorphism of Zn into Zm.
(iii) If there is a monomorphism of Zn into Zm, then n divides m.
(iv) If n divides m, then there is an epimorphism of Zm onto Zn.
(v) For any ring R, End(R) has atleast two members.
(vi) The zero map is the only homomorphism of R into Q.
(vii) Hom(Q, R) has exactly two members.
(viii) For any prime number p, End(Zp) has exactly two members.

3. For any ring R, prove that (End(R), 1, ο) is a ring, where 1 is the point-wise
addition and ο is the composition of mappings.

4. Let R be a ring with unity. Prove that f : R → R is an endomorphism of the ring R

if and only if there exists an idempotent e in the centre of R (that is, ee 5 e and
ex 5 xe for all x  R) such that f (a) 5 ea for all a  R.

5. Let f : R → R9 be a homomorphism of rings and a  R. Then prove that

a 1 ker f 5 {x  R : f (x) 5 f (a)},

where ker f 5 {y  R : f (y) 5 0}.

6. If f : R → R9 is an isomorphism of rings, then prove that f21 : R9 → R is also an

isomorphism of rings.

Q001-Algebra-111001_CH 09.indd 34 9/21/2011 4:52:35 PM

 Rings   9-35

7. For any homomorphism f : R → R9 of rings, prove that ker f is a subring of R such

that ax and xa  ker f for all a  ker f and x  R.
8. Let R be a ring with unity and define f : Z → R by f (n) 5 n1 for all n  Z. Prove
that f is a homomorphism of rings and ker f 5 mZ if char(R) 5 m  0 and ker
f 5 {0} if char(R) 5 0.
9. Prove the following for any rings R, R9 and R.
(i) R>R
(ii) R > R9 ⇒ R9 > R
(iii) R > R9 and R9 > R ⇒ R > R

10. Let R be a ring with unity. Prove that there is a subring S of R such that S is iso-
morphic to Z or to Zm depending on whether char(R) 5 0 or m.

11. Prove that the rings R and C are not isomorphic.

12. Determine all the ring epimorphisms of Z onto Z.

9.6 CERTAIN SPECIAL TYPES OF RINGS

In almost every occasion where there is a need for an example of a ring, we
used to refer till now to the ring of integers or of real numbers or of complex
numbers or the ring Zn of integers modulo n. Notice that all these are commu-
tative rings. It is not that noncommutative rings are unimportant. In fact the
knowledge of noncommutative rings is very important in the study of linear
algebra, in particular, of linear transformations from a vector space into itself.
In this section, we discuss mainly three types of noncommutative rings.

Theorem 9.6.1 (Ring of Endomorphisms of an Abelian Group). Let (G, 1) be

an abelian group and End(G) be the set of all endomorphisms of the group G.
Then, (End(G), 1, o) is a ring with unity, where 1 is the point-wise addition
and o is the composition of mappings.

Proof: Note that, for any f and g  End(G), f 1 g and f ο g are defined by

(f 1 g)(x) 5 f (x) 1 g(x)

and (f ο g)(x) 5 f (g(x)).

For any f and g  End(G) and x and y  G, we have

(f 1 g)(x 1 y) 5 f (x 1 y) 1 g(x 1 y)
          5 f (x) 1 f (y) 1 g(x) 1 g(y)
         5 f (x) 1 g(x) 1 f (y) 1 g(y) (since G is abelian)
         5 (f 1 g)(x) 1(f 1 g)(y)

Q001-Algebra-111001_CH 09.indd 35 9/21/2011 4:52:36 PM

 9-36   Algebra – Abstract and Modern

and (f ο g) (x 1 y) 5 f (g(x 1 y))

5 f (g(x) 1 g(y))
                  5 f (g(x)) 1 f (g(y)) 5 (f ο g)(x) 1 (f ο g)(y)

and therefore f 1 g and f ο g are endomorphisms of the group (G, 1). Thus,
1 and o are binary operations on End(G). The associativity of 1 in End(G)
follows from that of 1 in G. The zero endomorphism will be the zero element
in End(G). Also, for any f  End(G), the map 2f, defined by (2f )(x) 5 2f (x)
for all x  G, is the additive inverse of f in End(G). Thus, (End(G), 1) is an
abelian group. Also, clearly the composition o is associative and,

(f ο (g 1 h))(x) 5 f (g 1 h)(x)
5 f (g(x) 1 h(x))
5 f (g(x)) 1 f (h(x))
5 (f ο g 1 f ο h)(x)

        and ((f 1 g) o h)(x) 5 (f 1 g)(h(x))

          5 f (h(x)) 1 g(h(x))
           5 (f o h 1 g o h)(x)

for all x  G and hence

f o (g 1 h) 5 f o g 1 f o h
and (f 1 g) o h 5 f o h 1 g o h

for any f, g and h  End(G) Thus, (End(G), 1, ο) is a ring. Further, the iden-
tity homomorphism IG of the group G is the multiplicative identity in the ring
End(G). Thus, (End(G), 1, ο) is a ring with unity. b
In general the ring End(G) of endomorphisms of an abelian group G is not
commutative. For, consider the following example.

Example 9.6.1. Let G be the product group Z 3 Z under coordinate wise

addition and define f and g : G → G by

f (a1, a2) 5 (a1, a1 2 a2) and g(a1, a2) 5 (2a2, a1).

It can be easily verified that f and g are endomorphisms of G(5 Z 3 Z). Now
consider

(f o g)(1, 1) 5 f (g(1, 1)) 5 f (21, 1) 5 (21, 21 2 1) 5 (21, 2)

and (g o f )(1, 1) 5 g(f (1, 1)) 5 g(1, 1 2 1) 5 g(1, 0) 5 (0, 1).

Q001-Algebra-111001_CH 09.indd 36 9/21/2011 4:52:36 PM

 Rings   9-37

Therefore, (f o g) (1, 1)  (g o f )(1, 1) and hence f o g  g o f. Thus, End(G)

is a noncommutative ring.
Next let us consider the familiar concept of a real matrix; that is, an array
in which all the entries are real numbers. In fact, we can replace the real num-
ber system here with an abstract ring. In the following, we define addition
and multiplication of n 3 n matrices in such a way that the set of all n 3 n
matrices becomes a noncommutative ring.

Definition 9.6.1. Let (R, 1, ) be a ring and n a positive integer. By an n 3 n

matrix over R, we mean an array of n2 elements of the ring R, not necessarily
distinct, arranged in n rows and n columns as given below.

a11 a12 a13 … a1n 

 
a a22 a23 … a2 n 
 21 
    …  
 
an1 an 2 an3 … ann 

where each aij, 1 # i # n and 1 # j # n, is an element of the given ring. The

elements aij, 1 # j # n, constitute the ith row and the elements aij, 1 # i # n
constitute the jth column. As such aij is the element in both the ith row and jth
column. The matrix itself will be denoted by (aij). The set of all n 3 n matri-
ces over a ring R will be denoted by Mn(R). Two matrices (aij) and (bij) are said
to be equal if aij 5 bij for all 1 # i, j # n.

Definition 9.6.2. Let (R, 1, ) be a ring and n  Z1. For any matrices

A 5 (aij) and B 5 (bij) in Mn(R),

we define A 1 B and A  B as follows.

A 1 B 5 (cij), where cij 5 aij 1 bij

and A  B 5 (dij), where dij 5 ai1  b1j 1 ai2  b2j 1 … 1 ain  bnj for any 1 # i,
j # n. Note that the 1 and  on the right sides are those in the ring R.

Theorem 9.6.2. Let n be a positive integer and R be an arbitrary ring. Then,

the set Mn(R) of all n 3 n matrices over R is a ring under the operations 1
and  defined above.

Proof: Let A 5 (aij), B 5 (bij) and C 5 (cij) be arbitrary n 3 n matrices over

R. Since the addition 1 on R is commutative and associative, we have

Q001-Algebra-111001_CH 09.indd 37 9/21/2011 4:52:36 PM

 9-38   Algebra – Abstract and Modern

A 1 B 5 (aij 1 bij) 5 (bij 1 aij) 5 B 1 A

   and (A 1 B) 1 C 5 ((aij 1 bij) 1 cij)
5 (aij 1 (bij 1 cij)) 5 A 1 (B 1 C).

Therefore, 1 is commutative and associative on Mn(R). Let us denote the n 3

n matrix, all of whose entries are 0, by 0 itself. Then, clearly

A105A501A
and A 1 (2A) 5 0 5 (2A) 1 A,

where –A 5 ( –aij). Therefore, (Mn(R), 1) is an abelian group. To prove the

associativity of , let

A  B 5 (sij) and (A  B)  C 5 (tij).

n n
Then, sij  ∑ aik bkj and tij  ∑ sik ckj . Now,
k1 k1

 nn 
tij  ∑ ∑ air brk ckj

k1  r1 
   
n n
 ∑∑ ( air brk )ckj
      k1 r1
n n
 ∑∑ air (brk ckj )
     k1 r1

n  n 
 ∑ air ∑ brk ckj 

 k1 
     r1

     5 the ijth entry in A  (B  C).

Thus, (A  B)  C 5 A  (B  C). Also,

 n 
A  ( B  C )  ∑ air (brj  crj )
 r1 

 n   n 
 ∑ air brj ∑ air crj 
 r1   r1 
             
        5 A  B 1 A  C

and, similarly (A 1 B)  C 5 A  C 1 B  C. Thus, (Mn(R), 1, ) is a ring.

Q001-Algebra-111001_CH 09.indd 38 9/21/2011 4:52:37 PM

 Rings   9-39

Note:
1. If R is a ring with unity 1 and En is the n 3 n matrix defined by
1 if i  j
En 5 (eij), where eij  
0 if i  j
then En  A 5 A 5 A  En for any matrix A in Mn(R). Therefore, if R is
with unity, then Mn(R) is a ring with unity.
2. If n 5 1, then a 1 3 1 matrix (a) can be identified with a itself and, hence
M1(R) is isomorphic with R.
3. If n  1, then Mn(R) may not be commutative even when R is commuta-
tive; for consider the following theorem.

Theorem 9.6.3. For any n  1, the ring Mn(R) of n 3 n matrices over the real
numbers is a noncommutative ring with unity.

Proof: Let n  1. Since the real number system R forms a ring with unity,
by Theorem 9.6.2 (1), Mn(R) is a ring with unity. Let A 5 (aij) and B 5 (bij)
be the matrices defined by

1 if i  1  j
aij  
0 otherwise

1 if j  1 and i  1 or 2
   and bij   .
0 otherwise

1 0 … 0 1 0 … 0

  1 0 0 … 0
0 0 … 01 0 … 0  
   0 0 0 … 0
Then, AB  0 0 … 00 0 … 0   
         
            
0   0 0 0 … 0
0 … 0 0 0 … 0

1 0 … 01 0 … 0 1 0 … 0

   
1 0 … 00 0 … 0 1 0 … 0
    
   and BA 0 0 … 00 0 … 0 = 0 0 … 0
    
        …      …  
    
0 0 … 00 0 … 0 0 0 … 0

and hence AB  BA. Thus, Mn(R) is a noncommutative ring. b

Q001-Algebra-111001_CH 09.indd 39 9/21/2011 4:52:38 PM

 9-40   Algebra – Abstract and Modern

In the following, we give another important example of a noncommutative

ring, namely the ring of real quaternions.

Definition 9.6.3. The algebraic system (R4, 1, ), where R4 is the set of all
quadruples of real numbers and 1 and  are the binary operations defined
as follows, is called the system of real quaternions and each element of this
system is called a real quaternion.
For convenience, let us write a quadruple by (a0, a1, a2, a3). For any

a 5 (a0, a1, a2, a3) and b 5 (b0, b1, b2, b3) in R4,

we define

a 1 b 5 (a0 1 b0, a1 1 b1, a2 1 b2, a3 1 b3)

and a  b 5 (c0, c1, c2, c3), where

c0 5 a0b0 2 a1b1 2 a2b2 2 a3b3

c1 5 a0b1 1 a1b0 1 a2b3 2 a3b2
c2 5 a0b2 1 a2b0 1 a3b1 2 a1b3
and c3 5 a0b3 1 a3b0 1 a1b2 2 a2b1.

Even though 1 is defined coordinate wise, the multiplication  is not coordi-

nate wise and needs special attention. An easy way of remembering the rule for
multiplication is given below. Let us represent a quadruple (a0, a1, a2, a3) by

(a0, a1, a2, a3) 5 a0 1 a1i 1 a2j 1 a3k

(as a complex number (a, b) is represented by a 1 bi).

i

k j

As we go around clockwise, we read off the product; for example,

i  j 5 k, j  k 5 i, k  i 5 j.

Going around anticlockwise, we read off the negatives; for example,

i  k 5 2j, j  i 5 2k, k  j 5 2i.

Q001-Algebra-111001_CH 09.indd 40 9/21/2011 4:52:38 PM

 Rings   9-41

The multiplication in R4 is now defined as if we multiply two sums of real

numbers obeying the following rules.

i  i 5 j  j 5 k  k 5 21
                      i  j 5 k, j  k 5 i, k  i 5 j (A)
       j  i 5 2k, k  j 5 2i, i  k 5 2j

Now, the multiplication of real quaternions can be formally defined as

follows.

(a0 1 a1i 1 a2j 1 a3k)  (b0 1 b1i 1 b2j 1 b3k) 5 c0 1 c1i 1 c2j 1 c3k,

      where c0 5 a0b0 2 a1b1 2 a2b2 2 a3b3

         c1 5 a0b1 1 a1b0 1 a2b3 2 a3b2 (B)
         c2 5 a0b2 1 a2b0 1 a3b1 2 a1b3
        and c3 5 a0b3 1 a3b0 1 a1b2 2 a2b1.

To multiply a0 1 a1i 1 a2j 1 a3k by b0 1 b1i 1 b2j 1 b3k on the right, we first
multiply each ‘term’ in the first quaternion with each term in the second on
the right, use the laws given in (A) and collect the terms with each of i, j and
k and without any of them. Recall that the elements 1, 21, i, 2i, j, 2j, k and
2k form a nonabelian group of order 8 under the above multiplication rules
(A) and is called the group of quaternions. If we write

(1, 0, 0, 0) 5 1, (0, 1, 0, 0) 5 i
(0, 0, 1, 0) 5 j and (0, 0, 0, 1) 5 k,

then, by (B) above,

i2 5 (0, 1, 0, 0)  (0, 1, 0, 0) 5 2(1, 0, 0, 0) 5 21

and, similarly j2 5 21 5 k2 and other rules of multiplication in (A) can be

derived from (B). Now, the following is a routine verification.

Theorem 9.6.4. The real quaternions form a noncommutative ring with unity
under the addition and multiplication given above. This ring is denoted by QR
and is called the ring of real quaternions.

Worked Exercise 9.6.1. Give an example of a noncommutative ring with

exactly 16 elements.

Q001-Algebra-111001_CH 09.indd 41 9/21/2011 4:52:38 PM

 9-42   Algebra – Abstract and Modern

Answer: Consider the ring Z2 of integers modulo 2. Then, Z2 5 {0, 1}. Then,
the ring M2(Z2) of 2 3 2 matrices over Z2 has exactly 2232 (516) elements and
is not commutative. For, consider
1 1  1 0 11 0 0 0
      
0 0 1 0 0 0 0 0
1 0 1 1  1 1
and        .
1 0 0 0 1 1
      

Worked Exercise 9.6.2. Prove that every nonzero element in the ring of real
quaternions is a unit.

Answer: Let 0  a 5 a0 1 a1i 1 a2 j 1 a3k  QR. Put s  a02  a12  a22  a32 .
Since a  0, atleast one ai must be nonzero and hence s  0. Now, consider

a0 a1 a a
b  i  2 j  3 k.
s s s s
s
Then, ab  ba   1. Therefore, a is a unit in QR.
s
EXERCISE 9(f )

1. Evaluate the following products a  b and b  a in the rings mentioned against them.
(i) a, b  End(R3) defined by
a( r1 , r2 , r3 )  ( r1  r2 , r1  r2 ,0) and
.
b( r1 , r2 , r3 )  ( r3  r1 , r3  r2 , r2  r3 )
1 0 0  
  0 2 1 
(ii) a  0 2 3 and b  1 2 3 in M 3 (Z) .
   
2 1 0 3 1 2
(iii) a 5 2 1 3i 1 4j 1 5k and b 5 1 1 2i 1 3j 1 k in QR.
(iv) a 5 1 1 2i 1 5j 2 3k and b 5 3 1 2i 2 2j 1 k in QR.
2. Give an example of a noncommutative ring with unity having exactly 81 elements.

3. Describe the ring End(Z6), where Z6 is the group of integers modulo 6.

4. Prove or disprove that the set

a a  b  
X    : a and b are integers
a  b
 b 

is a subring of the ring M2(R) of 2 3 2 matrices over R.

Q001-Algebra-111001_CH 09.indd 42 9/21/2011 4:52:39 PM

 Rings   9-43

5. Give examples of two matrices A and B in M4(R) such that AB 5 0 and BA  0.

6. Let a 5 a0 1 a1i 1 a2j 1 a3k and b 5 a0 2 a1i 2 a2j 2 a3k, where a0, a1, a2, a3
are real numbers. Evaluate the products a ? b and b ? a in QR.
7. Prove that QR 2 {0} forms a group under the multiplication of real
quaternions.
8. Determine all the nilpotents and all the idempotents in the ring QR of real
quaternions.
0 0 1
 
9. If A 0 0 1 , then compute Am in M3(R) for any positive integer m.
  
0 0 1
a  a 
10. Prove that 
  : a and b are integers is a subring of M (R).
b b  2

11. Prove in detail that QR is a ring under the addition and multiplication of real
quaternions.

12. Find the centre of the ring of Mn(R) of all n 3 n matrices over the real number
system R.

13. Extend the above Exercise 12 for an arbitrary ring R in place of the ring R of real
numbers.

14. Determine the centre of the ring of real quaternions.

15. If S is a subring of a ring R, prove that Mn(S) is a subring of Mn(R) for any posi-
tive integer n.

16. If R is a ring such that Mn(R) is a ring with unity, then prove that R is a ring with
unity.

9.7 INTEGRAL DOMAINS AND FIELDS

Now we turn our attention to certain important special types of commuta-
tive rings. One of the motives of inventing the abstract concept of a ring is
to put the algebraic properties of the integers into an abstract setting. A ring
is not the appropriate abstraction of the integers, because much is lost in
the process. Besides the two obvious properties of commutativity and the
existence of unity, there is one other essential feature of the integers that
rings in general do not satisfy, namely cancellation property for multiplica-
tion. In this section, we introduce a special class of rings, known as integral
domains, which have all the three properties, namely the commutativity,
the existence of unity and the cancellation law for multiplication. Integral
domains play a major role in Algebraic Number Theory and various other
areas of mathematics. Also, a special kind of integral domains, namely

Q001-Algebra-111001_CH 09.indd 43 9/21/2011 4:52:40 PM

 9-44   Algebra – Abstract and Modern

fields, are introduced in this section and several elementary properties of

these are discussed. Fields are abstractions of the rational or real or com-
plex number systems.
First let us recall that a nonzero element a in a ring R is called a zero divi-
sor if there exists a nonzero element b in R such that ab 5 0 5 ba.

Definition 9.7.1. A nontrivial commutative ring with unity and without zero
divisors is called an integral domain.
Recall that a ring R with unity 1 is nontrivial or nonzero (that is, R  {0})
if and only if the additive identity 0 and the multiplicative identity 1 are dif-
ferent. In the following, we obtain some other simple equivalent conditions
for a ring to be an integral domain.

Theorem 9.7.1. The following are equivalent to each other for any nontrivial
commutative ring (R, 1, ) with unity.
1. (R, 1, ) is an integral domain.
2. For any elements a, b and c in R,
ab 5 ac ⇒ a 5 0 or b 5 c.
3. For any elements a and b in R,

ab 5 0 ⇒ a 5 0 or b 5 0.

Proof: (1) ⇒ (2): If ab 5 ac and a  0, then

a(b – c) 5 ab – ac 5 0

and, since a is not a zero divisor, b – c 5 0 or b 5 c.

(2) ⇒ (3): If ab 5 0, then ab 5 a0 and therefore, by (2), a 5 0 or b 5 0.
(3) ⇒ (1) is clear. b

Example 9.7.1
1. The ring Z of integers, the ring Q of rational numbers, the ring R of real
numbers and the ring C of complex numbers are all integral domains
with respect to usual addition and multiplication. In each of these, the
product of any two nonzero elements is again nonzero and hence there
are no zero divisors.
2. The ring Z[i] of Gaussian integers is an integral domain with respect to
the addition and multiplication of complex numbers. Note that Z[i] is a
subring of the ring C of complex numbers.
3. (Zp, 1p, ·p) is an integral domain for any prime number p.

Q001-Algebra-111001_CH 09.indd 44 9/21/2011 4:52:40 PM

 Rings   9-45

4. (Z6, 16, ·6) is not an integral domain, since 2  0, 3  0 and 2 ·6 3 5 0

in Z6. This is a commutative ring with unity and zero divisors.
5. For any n  1, (nZ, 1, ) is a commutative ring without zero divisors
and with no unity and hence not an integral domain.
6. The ring QR of real quaternions is a ring with unity and without zero
divisors and not commutative and hence not an integral domain.
The examples in (4), (5) and (6) above substantiate that the three defining
properties of an integral domain, namely, the nonexistence of zero divisors,
the existence of unity and the commutativity are all independent of others, in
the sense that no two imply the other. In the following, we introduce a special
class of integral domains.

Definition 9.7.2. A nontrivial commutative ring with unity is called a field if

every nonzero element of it is a unit (that is, multiplicatively invertible).

Example 9.7.2
1. The rings Q, R and C are all fields, since for any nonzero number a,
there is 1/a for which a  1a  1 and hence a is a unit.
2. The ring Z of integers is not a field, since 2 is not a unit in Z. In fact, 1
and 21 are the only units in Z.
3. For any prime number p, (Zp, 1p, ·p) is a field, since any 0 , a , p is
relatively prime with p and hence a unit in Zp.
4. The ring QR of real quaternions is not a field, even though every nonzero
element is a unit in it; because it is not commutative.

Theorem 9.7.2. Every field is an integral domain and the converse is not true.

Proof: Let (R, 1, ) be a field. Then, R is a nontrivial commutative ring with

unity and, if 0  a  R, a has multiplicative inverse a21 in R and therefore,
for any b  R,

ab 5 0 ⇒ a21(ab) 5 0 ⇒ (a21a)b 5 0 ⇒ b 5 0.

Therefore, R has no zero divisors and hence R is an integral domain. The

converse is not true, since the ring Z of integers is an integral domain but not
a field. b
However, certain special type of integral domains are fields. In this context,
we have the following theorem.

Theorem 9.7.3. Every finite integral domain is a field.

Q001-Algebra-111001_CH 09.indd 45 9/21/2011 4:52:40 PM

 9-46   Algebra – Abstract and Modern

Proof: Let (R, 1, ) be a finite integral domain. Then, R is a nontrivial com-

mutative ring with unity and without zero divisors. Let 0  a  R. Since R
is finite, we can write

R 2 {0} 5 {a1, a2, …, an},

where a1, a2, …, an are all the distinct nonzero elements in R. Since R is an
integral domain, a  0 and each ai  0, we get that aai  0 for each 1  i  n.
Therefore, the set

S 5 {aa1, aa2, …, aan}

is a subset of R 2 {0}. Further, aai  aaj for all i  j (since a  0 and ai  aj).
S is an n-element subset of R 2 {0}, which also has n elements. Therefore,

R 2 {0} 5 S 5 {aa1, aa2, …, aan}.

In particular, 1  R 2 {0} and hence aai 5 1 for some i. Therefore, a is a unit

in R. Thus, R is a field. b

Corollary 9.7.1. The following are equivalent to each other for any positive
integer n.
1. n is a prime number.
2. (Zn, 1n, ·n) is an integral domain.
3. (Zn, 1n, ·n) is a field.

Proof: First note that, for each of these, n must be necessarily greater than 1
(for, if n 5 1, Zn is trivial).
(1) ⇒ (2) follows from the fact that, for any a and b  Zn,

ab 5 0 in Zp ⇒ n divides ab

and that, if n is prime,

n divides ab ⇔ n divides a or b.

(2) ⇒ (3) follows from Theorem 9.7.3.

(3) ⇒ (1) follows from the fact that, for any 0 , a , n, a is a unit in Zn if and
only if a is relatively prime with n.

Corollary 9.7.2. For any prime number p, (Zp, 1p, ·p) is a field.

Definition 9.7.3. A nontrivial ring with unity is called a division ring if every
nonzero element is a unit.

Q001-Algebra-111001_CH 09.indd 46 9/21/2011 4:52:40 PM

 Rings   9-47

Note that any field is a division ring and the converse is not true, since
the ring QR of real quaternions is a division ring, but not a field (see Worked
Exercise 9.6.2). However, any commutative division ring is a field.

Theorem 9.7.4. The characteristic of any integral domain is either 0 or a

prime number.

Proof: Let R be an integral domain and char(R)  0. Since R is nontrivial,

char(R)  1. Let char(R) 5 n  1. Suppose that n is not a prime. Then, there
exist positive integers r and s such that n 5 rs, r  1 and s  1.
Now, consider

(r1)  (s1) 5 rs1 5 n1 5 0.

Since r , n 5 char(R), r1  0 (otherwise rx 5 0 for all x  R). Similarly,

s1  0. This is a contradiction, since R is an integral domain. Thus, n is
prime. b

Corollary 9.7.3. The characteristic of any field is either 0 or a prime.

Worked Exercise 9.7.1. Prove that any integral domain has exactly two idem-
potents.

Answer: Let R be an integral domain. Then, 0  1 and clearly these two are
idempotents. If a is any idempotent in R, then a2 5 a and hence

a(a 2 1) 5 0 so that a 5 0 or a 5 1

Thus, 0 and 1 are the only idempotents of R.

Worked Exercise 9.7.2. Prove that Z3[i] 5 {a 1 bi : a and b  Z3} is a field

under addition and multiplication modulo 3 by writing the tables representing
the operations 13 and ·3 on Z3[i]. Z3[i] is called the ring of Gaussian integers
modulo 3.

Answer: We have

Z3[i] 5 {0, 1, 2, i, 1 1 i, 2 1 i, 2i, 1 1 2i, 2 1 2i}.

Here, the elements are added and multiplied as in the complex number system,
except that the coefficients are reduced modulo 3. In particular, note that

21 5 2, 2i 5 2i, 2 ·3 2 5 1, 2i ·3 2i 5 2

Q001-Algebra-111001_CH 09.indd 47 9/21/2011 4:52:40 PM

 The table representing 13 and ·3 on Z3[i] are given below.
13 0 1 2 i 11i 21i 2i 1 1 2i 2 1 2i
0 0 1 2 i 11i 21i 2i 1 1 2i 2 1 2i
1 1 2 0 11i 21i i 1 1 2i 2 1 2i 2i

Q001-Algebra-111001_CH 09.indd 48
2 2 0 1 21i i 11i 2 1 2i 2i 1 1 2i
i i 11i 21i 2i 1 1 2i 2 1 2i 0 1 2
11i 11i 21i i 1 1 2i 2 1 2i 2i 1 2 0
21i 21i i 11i 2 1 2i 2i 1 1 2i 2 0 1
2i 2i 1 1 2i 2 1 2i 0 1 2 i 11i 21i
1 1 2i 1 1 2i 2 1 2i 2i 1 2 0 11i 21i i
9-48   Algebra – Abstract and Modern

2 1 2i 2 1 2i 2i 1 1 2i 2 0 1 21i i 11i

3 0 1 2 i 11i 21i 2i 1 1 2i 2 1 2i
0 0 0 0 0 0 0 0 0 0
1 0 1 2 i 11i 21i 2i 1 1 2i 2 1 2i
2 0 2 1 2i 2 1 2i 1 1 2i i 21i 11i
i 0 i 2i 2 21i 2 1 2i 1 11i 1 1 2i
11i 0 11i 2 1 2i 21i 2i 1 1 1 2i 2 i
21i 0 21i 1 1 2i 2 1 2i 1 i 11i 2i 2
2i 0 2i i 1 1 1 2i 11i 2 2 1 2i 21i
1 1 2i 0 1 1 2i 21i 11i 2 2i 2 1 2i i 1
2 1 2i 0 2 1 2i 11i 1 1 2i i 2 21i 1 2i

9/21/2011 4:52:41 PM
 Rings   9-49

It is a straight forward verification that Z3[i] is a commutative ring with unity.

By looking at the multiplication table given above (zero does not appear in any
row and column corresponding to nonzero elements), we can infer that Z3[i] is
an integral domain. Since it is finite (with 9 elements), it is a field also.

Worked Exercise 9.7.3. Prove that Z2[i] 5 {a 1 bi : a and b  Z2} is a com-

mutative ring with unity which is not an integral domain under addition and
multiplication of complex numbers modulo 2.

Answer: We have

Z2[i] 5 {0, 1, i, 1 1 i}.

The tables for 12 and 2 on Z2[i] are given below.

12 0 1 i 11i
0 0 1 i 11i
1 1 0 11i i
i i 11i 0 1
11i 11i i 1 0

·2 0 1 i 11i
0 0 0 0 0
1 0 1 i 11i
i 0 i 1 11i
11i 0 11i 11i 0

It can be easily verified that (Z2[i], 12, 2) is a commutative ring with unity.
Since

(1 1 i)(1 1 i) 5 1 1 2i 1 (21) 5 0,

1 1 i is a zero divisor and hence Z2[i] is not an integral domain.

Worked Exercise 9.7.4. Let R be a nontrivial ring such that, for each 0  a
 R, there exists unique element x in R such that axa 5 a. Prove that R is a
division ring.

Answer: We first prove that R has no zero divisors. Suppose that a and b  R
such that ab 5 0 and a  0. Choose x  R such that axa 5 a. Then,

a(x 1 b)a 5 axa 1 aba 5 axa 1 0 5 axa 5 a.

Q001-Algebra-111001_CH 09.indd 49 9/21/2011 4:52:41 PM

 9-50   Algebra – Abstract and Modern

By the uniqueness of x, it follows that x 1 b 5 x or b 5 0. Thus, ab 5 0 implies

that a 5 0 or b 5 0. From this we get R 2 {0} is closed under multiplication.
Let 0  a  R and x be the unique element in R such that axa 5 a. Then,
ax  ax 5 ax and xa  xa 5 xa and hence ax and xa are idempotents and are
nonzero (since axa 5 a  0). Next, we shall prove that there is only one
nonzero idempotent in R. Let e and f be nonzero idempotents in R. Let g be
the unique element such that

(ef )  g ? (ef ) 5 ef.

Then, g  0, since ef  0. Now,

(ef )(ge)(ef ) 5 ef and (ef )(fg)(ef ) 5 ef.

By the uniqueness of g, we get that g 5 ge 5 fg. Also, (ef )(ge  fg)(ef ) 5 ef

and hence ge  fg 5 g which implies that g2 5 g. Therefore, geg 5 g 5 gfg
and hence e 5 f. Thus, R 2 {0} has exactly one idempotent, say e.
In particular, ax 5 xa 5 e and hence

ae 5 axa 5 a and ea 5 axa 5a.

Thus, e is the unity in R and, since ax 5 e 5 xa, x is the multiplicative inverse

of a. Thus, R is a division ring.

EXERCISE 9(g)

1. Consider the following classes of rings.

FF 5 The class of all finite fields.
F 5 The class of all fields.
ID 5 The class of all integral domains.
R 5 The class of all rings.
RU 5 The class of all rings with unity.
CR 5 The class of all commutative rings.
CRU 5 The class of all commutative rings with unity.
NCR 5 The class of all noncommutative rings.
Draw a Venn diagram representing the above classes.

2. Which of the following are fields or integral domains? Substantiate your

answers.
(i) R 3 R under coordinate wise addition and multiplication.
(ii) Qn, where Q is the field of rationals and n  Z1 under coordinate wise
operations.

Q001-Algebra-111001_CH 09.indd 50 9/21/2011 4:52:41 PM

 Rings   9-51

(iii) Z[ 2]  {a  b 2 : a and b  Z} under the addition and multiplication

of real numbers.
(iv) Z[i], the ring of Gaussian integers.
(v) (Z5[i], 15, ?5).
(vi) (Z4[i], 14, ?4).
(vii) (Z3[i], 13, ?3).
(viii) (Z2[i], 12, ?2).
(ix) Z5 3 Z3.
(x) Z19.

3. Prove that Q[i] 5 {a 1 bi : a and b are rationals} is a field under the addition
and multiplication of complex numbers.

4. Let R be a field. Prove that R is a Boolean ring if and only if R has exactly two
elements.

5. Give an example of a field with exactly 30 nonzero elements.

6. Let n be a positive integer and

Zn[i] 5 {a 1 bi : a and b  Zn}.
Prove that Zn[i] is a ring under addition and multiplication modulo n.
7. Give an example of a positive integer n for which Zn[i] is not an integral domain.

8. Prove that Zn[i] is an integral domain if and only if it is a field.

9. Let Z[ 2]  {a  b 2 : a and b are integers}. Prove that Z[ 2] is an integral

domain under the addition and multiplication of real numbers.

10. Prove that Z[ 2] is not a field.

11. Let p be a prime number and

 a 
R   : a and b  Z and p does not divide b.
 b 
Prove that R is an integral domain under the usual addition and multiplication of
rational numbers.

12. Is the above R a field?

13. Let R be a nontrivial finite ring without zero divisors. Then prove that R is with
unity and that (R 2 {0}, ?) is a group.

14. Prove that any finite commutative ring without zero divisors is a field.

15. For any simple abelian group (G, 1), prove that the ring End(G) of all endomor-
phisms of (G, 1) is a division ring.

Q001-Algebra-111001_CH 09.indd 51 9/21/2011 4:52:42 PM

 9-52   Algebra – Abstract and Modern

16. Prove that the characteristic of a finite field is a prime number.

17. Is there an integral domain having exactly 6 elements?

18. Let R be a nontrivial finite ring without zero divisors. Then prove that R is a divi-
sion ring.

19. Let R and S be two rings. Then prove that the product ring R 3 S is an integral
domain if and only if one of R and S is an integral domain and the other is the
trivial ring.

20. Let R be an integral domain, 0  a  R and n  Z1 such that na 5 0. Prove that

char(R)  0.

Q001-Algebra-111001_CH 09.indd 52 9/21/2011 4:52:42 PM

 10
Ideals and Quotient
Rings
10.1 Ideals
10.2 Quotient Rings
10.3 Chinese Remainder Theorem
10.4 Prime Ideals
10.5 Maximal Ideals
10.6 Embeddings of Rings

In the study of finite groups, we have proved several results using the concept
of a normal subgroup, quotient construction and induction on the group order.
Homomorphic images of groups are identified with quotient groups with the
help of the kernel of the homomorphism which is a normal subgroup. The
role of normal subgroups in groups is played by ideals in rings. The concepts
of ideal and quotient rings are important in the structure theory of rings. A
special kind of subrings, which are most suitable (ideal) for the study of the
structure of rings, are popularly called ideals.

10.1 IDEALS
In this section, we introduce the notion of an ideal in a ring and discuss sev-
eral important elementary properties of ideals.

Definition 10.1.1. Let (R, 1, ?) be a ring and I be a subgroup of (R, 1).

Then, I is called
1. a left ideal of R if ra  I for all a  I and r  R.
2. a right ideal of R if ar  I for all a  I and r  R.
3. an ideal of R if it is both a left ideal and a right ideal of R.

Q001-Algebra-111001_CH 10.indd 1 9/21/2011 4:53:49 PM

 10-2   Algebra – Abstract and Modern

Clearly any left ideal or right ideal of a ring R is a subring of R. But a sub-
ring of R may be neither a left ideal nor a right ideal. For example, the set Z
of integers is a subring of the ring R of real numbers and Z is not an ideal of
R, since 1  1  Z. If R is a commutative ring, there is no difference between
2
a left ideal, a right ideal and an ideal. Sometimes, we refer to an ideal as a
two-sided ideal.

Example 10.1.1

1. For any ring R, clearly {0} and R are ideals of R and are called trivial
ideals. {0} is called the zero ideal. Ideals other than {0} and R are called
proper ideals.
2. If (R, 1, ?) is a ring with trivial multiplication, that is, ab 5 0 for any a
and b in R, then every subgroup of (R, 1) is an ideal of R.
3. Consider the ring M2(R) of 2 3 2 matrices over the real R and let

0 a 
I    : a and b are real numbers .

0 b 

Then, I is a left ideal of M2(R), since

     r s 
r s 0 a  0 ra  sb  I for all    M 2 ( R).
0 b 0 ta  ub 
t u t u

It can be easily verified that I is a subgroup of M2(R). However, I is not

a right ideal of M2(R), since

0 2 1 1  2 2  I .


0 31 1 3 3

a b 
4. Let J    : a and b are real numbers. Then, J is a right ideal of

0 0 
M2(R) and is not a left ideal.
5. For any nonnegative integer n, let

n  {na : a is an integer}.

Then, nZ is an ideal of the ring Z of integers. In fact, any ideal of Z is of

this form nZ for some n  0.

Q001-Algebra-111001_CH 10.indd 2 9/21/2011 4:53:50 PM

 Ideals and Quotient Rings   10-3

a b  
6. Let K    : a, b, c and d are even integers . Then, K is an ideal

c d  
of M2(Z). Later, we shall prove that M2(R) has no nontrivial ideals, since
R is so. However, Z has several ideals and so is M2(Z).
7. Let X be any set and consider the ring (P(X), 1, ∩) of all subsets of X,
where 1 and ∩ are defined by
A + B = ( A − B) ∪ ( B − A) and A ∩ B = The intersection of A and B.

Let I 5 {Y : Y is a finite subset of X}. Then, I is an ideal of (P(X), 1, ∩).

8. Let R be any ring and X be any nonempty set and consider the ring RX of
all mappings of X into R. For any Y ⊆ X, let

IY  { f  R X : f ( y )  0 for all y  Y }.

Then, IY is an ideal of RX.

Theorem 10.1.1. Let R be a ring with unity and U(R) be the set of all units
in R. Then, the following are equivalent to each other for any left (right or
two-sided) ideal I of R.

1. I5R
2. U(R) ⊆ I
3. I ∩ U(R)  
4. 1I

Proof: Let I be a left ideal of R. (1) ⇒ (2) and (2) ⇒ (3) are trivial.
(3) ⇒ (4): Suppose that a  I ∩ U(R). Then, a has multiplicative inverse a21
in R and 1 5 a21 ? a  I (since a  I and I is a left ideal of R).
(4) ⇒ (1): If 1  I, then, for any r  R, r 5 r ? 1  I and therefore R ⊆ I so
that I 5 R.
In the following, our discussion is restricted to ideals of rings. Some of
the results proved for ideals can be extended to left or right ideals easily.
However, we are more interested in two-sided ideals, since these lead to the
construction of quotient rings. First, we discuss certain standard methods of
constructing new ideals from given ones.

Theorem 10.1.2. Let {Ia}a be a nonempty class of ideals of a ring R. Then,

∩ I a is also an ideal of R.
a

Q001-Algebra-111001_CH 10.indd 3 9/21/2011 4:53:50 PM

 10-4   Algebra – Abstract and Modern

Proof: First note that every ideal contains the zero element of the ring; for, an
ideal I is nonempty and hence there exists a  I so that 0 5 0a  I. Therefore,
0  Ia for all a  . Put I  ∩ I a Then, 0  I and hence I is a nonempty
a
subset of R. Since the intersection of any family of subgroups of (R, 1) is
again a subgroup, it follows that I is a subgroup of (R, 1). Also,

a  I and r  R ⇒ a  Ia for all a   and r  R

⇒ ra and ar  Ia for all a  
⇒ ra and ar  I.

Thus, I is an ideal of R.

We have proved above that the intersection of ideals is again an ideal.
However, the union of ideals may not be an ideal. If I and J are ideals, then
they are subgroups of (R, 1). Therefore, I ∪ J is a subgroup of (R, 1) if and
only if I ⊆ J or J ⊆ I (see Theorem 4.1.6). Thus, for any ideals I and J, I ∪ J
is an ideal if and only if I ⊆ J or J ⊆ I. For certain special classes of ideals,
union of the class of ideals is again an ideal.

Theorem 10.1.3. Let {Ia}a be a class of ideals of a ring. Suppose that, for
any a and b  , there exists    such that Ia ⊆ I and Ib ⊆ I and Ib ⊆ I
(such classes are called directed above). Then, ∪ I a is an ideal of R.
a

Proof: Let I  ∪ I a . Then, clearly I is a nonempty subset of R. Now,

a

a and b  I ⇒ a  Ia and b  Ib for some a and b  

⇒ there exists    such that a  Ia ⊆ I and b  Ib ⊆ I
⇒ a and b  I,   
⇒ a 2 b  I ⊆ I
⇒ a 2 b  I.

Therefore, I is a subgroup of (R, 1). Also,

a  I and r  R ⇒ a  Ia for some a   and r  R

⇒ ra and ar  Ia ⊆ I
⇒ ra and ar  I.

Thus, I is an ideal of R. b

Corollary 10.1.1. Let {Ia}a be a chain of ideals of a ring R (that is, given
any two members in the class, one of them is contained in the other). Then,
∪ I a is again an ideal of R.
a

Q001-Algebra-111001_CH 10.indd 4 9/21/2011 4:53:51 PM

 Ideals and Quotient Rings   10-5

Definition 10.1.2. Let R be a ring and S ⊆ R. Let

<S> 5 ∩{I : I is an ideal of R and S ⊆ I}.

By Theorem 10.1.2, <S> is an ideal of R and is called the ideal generated by

S. Note that, for any ideal I of R, S ⊆ I if and only if <S> ⊆ I. For this reason,
we say that <S> is the smallest ideal of R containing S. If I 5 <S>, then we
say that I is the ideal generated by S or S generates I. An ideal I is said to be
finitely generated if I 5 <S> for some finite set S. If S 5 {a}, then <S> will
be denoted by <a> and is called a principal ideal generated by a.
A natural question that arises in one’s mind is about the precise form of
elements in the ideal <S> generated by S. Answer to such a question will be
clear if we can determine the precise form of the elements in a principal ideal
<a>. We do this in the following theorem.

Theorem 10.1.4. Let R be a ring and a  R. Then, any element of <a> is of

the form
m
ra + as + na + ∑ xi ayi ,
i =1

where m is a nonnegative integer, n is an integer and r, s, x1, …, xm, y1, …,

ym  R.

Proof: Let A be the set of all elements of the form given in the theorem. We
shall prove that A is the smallest ideal of R containing a. By taking r 5 0 5 s;
m 5 0 and n 5 1, we have a  A. If I is any ideal of R containing a, then ra,
as, xay and na  I for any r, s, x, y  R and n  Z and hence A ⊆ I. Thus, we
are left with only verifying that A is an ideal of R. Using the commutativity of
1 and the distributivity of the multiplication over the addition, one can easily
prove that A is an ideal of R. Thus, A 5 <a>.
In certain special cases, <a> turns out to be much simpler.

Corollary 10.1.2. Let R be a ring with unity and a  R. Then,

 m 
< a > ∑ xi ayi : 0  m  Z, xi and yi  R.
 i1 

Proof: Let B be the set given on the right hand side. Then, by taking m 5 1,
x1 5 1 5 y1 (the unity in R), we get that a  B. Since B is closed under 1, it
follows that na  B for all n  Z. Also, by taking m 5 1 and x1 5 1, we get

Q001-Algebra-111001_CH 10.indd 5 9/21/2011 4:53:51 PM

 10-6   Algebra – Abstract and Modern

that ay  B for all y  R. Similarly, xa  B for all x  R. Thus, by Theorem

10.1.4, <a> ⊆ B ⊆ <a> and hence <a> 5 B.

Corollary 10.1.3. Let R be a commutative ring and a  R. Then,

<a> 5 {ra 1 na : r  R and n  Z}.

Proof: By Theorem 10.1.4, ra 1 na  <a> for any r  R and n  Z. Also,

since R is commutative,

m
 m 
ra 1 as 1 na 1 ∑ x ay = r  s  ∑ xi ayi  a 1 na.
i=1
i i
 i1


Thus, any element of <a> is of the form ra 1 na for some r  R and n  Z

and hence

<a> 5 {ra 1 na: r  R and n  Z}.

Corollary 10.1.4. Let R be a commutative ring with unity and a  R. Then,

<a> 5 {ra : r  R} = Ra = aR.

Proof: For any r  R and n  Z, we have

ra 1 na 5 ra 1 (n  1)a 5 (r 1 n1)a  Ra,

where 1 is the unity in R. Therefore, by Corollary 10.1.3, <a> 5 Ra 5 aR.

For any ideals I and J of a ring R, clearly I ∩ J is the largest ideal contained
in both I and J. In the following, we describe the smallest ideal containing
both I and J. This may not be I ∪ J, since I ∪ J may not be an ideal at all,
unless I ⊆ J or J ⊆ I.

Theorem 10.1.5. Let I and J be ideals of a ring R and

I 1 J 5 {a 1 b : a  I and b  J}.

Then, I 1 J is the smallest ideal of R containing both I and J; that is, I 1

J 5 <I ∪ J>.

Proof: Since 0  I ∩ J, we have

I 5 I 1 {0} ⊆ I 1 J and J 5 {0} 1 J ⊆ I 1 J.

Q001-Algebra-111001_CH 10.indd 6 9/21/2011 4:53:52 PM

 Ideals and Quotient Rings   10-7

Since I and J are subgroups of (R, 1), so is I 1 J. Also, for any r  R, a 

I and b  J.
Also, for any r  R, a  I and b  J,

r(a 1 b) 5 ra 1 rb  I 1 J
          and (a 1 b)r 5 ar 1 br  I 1 J

and hence I 1 J is an ideal of R. Further, if K is any ideal of R containing both

I and J, then clearly I 1 J ⊆ K. Thus, I 1 J is the smallest ideal of R contain-
ing both I and J.

Corollary 10.1.5. For any ideals I1, I2, …, In of a ring R, let

n

∑I
i1
i  I1  I 2  I n  {a1  a2  an : ai  I i }.

n n
Then,  I i is the smallest ideal containing ∪ I i .
i1 i1

Corollary 10.1.6. Let {Ia}a be a nonempty class of ideals of a ring R and

∑I
a
a  ∪I
a
a .

Then,  I a  {a1  a2  an : ai  I ai for some ai  }.

a

Corollary 10.1.7. Let S be a nonempty subset of a ring R. Then, any element

of <S> is a finite sum of elements of the form
m
ra  as  na  ∑ xi ayi ,
i1

where a  S, n and m  Z, m  0, and r, s, xi, yi  R.

Corollary 10.1.8. Let R be a commutative ring with unity and S be a non-

empty subset of R. Then,

 m 
<S >  ∑ ra : m  0, ri  R and ai  S  .
 i1 i i 

For any ideals I1, I2, …, In of a ring R, we have proved that any element of
n
∪ I i can be expressed as a sum a1 1 a2 1  1 an with ai  Ii, 1 # i # n.
i1

Q001-Algebra-111001_CH 10.indd 7 9/21/2011 4:53:53 PM

 10-8   Algebra – Abstract and Modern

However, there is no guarantee that this expression is unique, unless the ide-
als I1, I2, …, In satisfy certain additional conditions.

Theorem 10.1.6. Let I1, I2, …, In be ideals of a ring R and I 5 I1 1 I2 1 …

1 In. Then, any element of I can be uniquely expressed as a1 1 a2 1  1 an,
with ai  Ii, if and only if

 
I i ∩∑ I j   {0} for all 1  i  n.
 ji 

Proof: Suppose that any element a of I can be uniquely expressed as a 5 a1 1

a2 1  1 an, with ai  Ii. Fix 1 # i # n and let a  I i ∩  I j . Then,
a  I and
j ≠i
( )
a 5 a1 1  1 ai21 1 ai11 1  1 an, aj  Ij for j  i
or a1 1  1 ai21 2 a 1 ai11 1  1 an 5 0 5 0 1 0 1  1 0.

( )
By the uniqueness, a 5 0. Thus, I i ∩  I j {0}. Conversely, suppose that
j ≠i
the given condition is satisfied. Since I 5 I1 1 I2 1  1 In, any element of I
can be expressed as a1 1 a2 1  1 an, with ai  Ii. Now, suppose that

a1 1 a2 1  1 an 5 b1 1 b2 1  1 bn,

where ai, bi  Ii for 1 # i # n. Now, for each i,

 
ai  bi  ∑ (b j  a j )  I i ∩∑ I j   {0}
j ≠i  j≠i 

and hence ai 2 bi 5 0 or ai 5 bi. Thus, any element of I can be uniquely

expressed as a1 1 a2 1  1 an, with ai  Ii.

Corollary 10.1.9. Let I and J be ideals of a ring R. Then, any element of R can
be uniquely expressed as a 1 b with a  I and b  J if and only if

I  J  R and I ∩ J  {0}.

Definition 10.1.3. An ideal I of a ring R is said to be a direct summand of R if

there is an ideal J of R such that I 1 J 5 R and I ∩ J 5 {0}. In this case, R is
said to be the direct sum of I and J and denote this by R 5 I ! J. Also, I and
J are called direct complements to each other if R 5 I ! J.

Q001-Algebra-111001_CH 10.indd 8 9/21/2011 4:53:53 PM

 Ideals and Quotient Rings   10-9

In a ring with unity, we can have yet another beautiful description of direct
summands. Before going to this, let us define the following definition.

Definition 10.1.4. Let R be a ring and a  R. Then, a is called a central

idempotent if a2 5 a and ax 5 xa for all x  R; that is, a is an idempotent
belonging to the centre of R.

Theorem 10.1.7. Let R be a ring with unity and I be an ideal of R. Then, I

is a direct summand of R if and only if I is the principal ideal generated by a
central idempotent of R.

Proof: First note that, for any central idempotent e in R, the principal ideal
generated by e is of the form

<e> 5 eR 5 {ex : x  R}.

Suppose that I 5 <e> for some central idempotent e. Then, put J = (1 2 e)

R={(1 2 e)x : x  R}. Since

(1 2 e)25 1 2 e 2 e 1 e2 = 1 2 e 2 e 1 e = 1 2 e
and (1 2 e)x 5 x 2 ex 5 x 2 xe 5 x(1 2 e)

for all x  R, we get that 1 2 e is also a central idempotent in R and J 5

<1 2 e>, the principal ideal generated by 1 2 e. Now, for any x  R, we
can write

x 5 ex 1 (1 2 e)x  I 1 J

and hence I 1 J 5 R. Also,

a  I ∩ J ⇒ a 5 ex 5 (1 2 e)y for some x, y  R

⇒ a 5 ex 5 e2x 5 ea 5 e(1 2 e)y 5 0

and therefore I ∩ J 5 {0}. Thus, R 5 I ! J and I is a direct summand

of R.
Conversely, suppose that I is a direct summand of R. Then, there is an ideal
J or R such that R 5 I ! J; that is, I 1 J 5 R and I ∩ J 5 {0}. Since R is with
unity, 1  R 5 I 1 J and hence 1 5 e 1 s for some e  I and s  J, clearly
s 5 1 2 e. Now,

es  I ∩ J 5 {0} and hence es 5 0.

Q001-Algebra-111001_CH 10.indd 9 9/21/2011 4:53:54 PM

 10-10   Algebra – Abstract and Modern

From 1 5 e 1 s, we have

e 5 e(e 1 s) 5 e2 1 es 5 e2 (since es 5 0)

and therefore e is an idempotent. Also, for any x  R,

ex 1 sx 5 (e 1 s)x 5 x 5 x(e 1 s) 5 xe 1 xs

and therefore ex 2 xe 5 xs 2 sx  I ∩ J 5 {0} (since e  I and s  J), so

that ex 2 xe 5 0 or ex 5 xe. Thus, e is a central idempotent in R. Now, since
e  I, we get that <e> ⊆ I. Also,

x  I ⇒ x 5 (e 1 s)x 5 ex 1 sx 5 ex (since sx  I ∩ J 5 {0})

⇒ x  <e>.

Thus, I 5 eR 5 <e>.
The central idempotents in any ring have certain nice properties. They form
a Boolean ring under suitable operations, defined in the following theorem.

Theorem 10.1.8. Let (R, 1, ?) be a ring and B(R) be the set of all central
idempotents in R. For any a and b  B(R), define

a ∗ b 5 a 1 b 2 2ab ( = (a 2 ab) 1 (b 2 ab)).

Then, (B(R), ∗, ?) is a Boolean ring. Also, if R is with unity, then so is B(R).

Proof: First observe that, for any a and b  B(R), a ∗ b and a  b are central
idempotents of R; i.e.,

(a ∗ b)2 5 (a 1 b 2 2ab)2 5 a 1 b 2 2ab

        (a  b)2 5 abab 5 ab
      (a ∗ b)x 5 (a 1 b 2 2ab)x 5 x(a 1 b 2 2ab) 5 x(a ∗ b)
    and (a  b)x 5 abx 5 axb 5 xab 5 x(a  b).

It is a straight forward verification to prove that all the axioms of rings are
satisfied in B(R) with ∗ as addition and ? as multiplication. Note that a  a 5
a and a ∗ a 5 0 for all a  B(R) and 0 is the zero element in B(R) also. If R
has unity 1, then 1  B(R) and 1 is the unity in B(R) also.
In addition to the two binary operations ∩ and 1 on the set of ideals of a
ring, we introduce yet another binary operation, which is denoted by juxtapo-
sition, in the following definition.

Q001-Algebra-111001_CH 10.indd 10 9/21/2011 4:53:54 PM

 Ideals and Quotient Rings   10-11

Definition 10.1.5. For any ideals I and J of a ring R, define

 n 
IJ  ∑ ai bi : n  Z+ , ai  I and bi  J .
 i1 

Theorem 10.1.9. Let I and J be ideals of a ring R. Then, IJ is an ideal of R

and IJ ⊆ I ∩ J.
n m
Proof: If x and y  IJ, then x   ai bi and y   c j d j , where n, m  Z1, ai
i1 j1
and cj  I and bi and dj  J. Then,

x 2 y 5 a1b1 1  1 anbn 1 (2c1)d1 1  1 (2cm)dm

and hence x 2 y  IJ, so that IJ is a subgroup of (R, 1).

n
Also, if r  R and x   ai bi  IJ , then
i1

n n
rx  ∑ ( rai )bi and xr  ∑ ai (bi r )
i1 i1

and rai, ai  I and bi, bir  J and hence rx and xr belong to IJ. Thus, IJ is an
ideal of R. Clearly IJ ⊆ I and J.
There is an important observation that an ideal of an ideal need not be an
ideal; that is, if I is an ideal of a ring R, then I can be treated as a ring on its
own (since I is a subring of R) and, if J is an ideal of I, then J need not be an
ideal of the ring R. This is illustrated in the following example.

Example 10.1.2. Let R be the ring of real numbers under the usual addition
and multiplication. Consider the ring RR of all mapping of R into R under the
point-wise addition and multiplication. Let

S 5 {f  RR : f is continuous},

where R is with the usual topology. Then, S is a subring of RR and hence S is

a commutative ring with unity (the constant map 1 is the unity in S). Let i be
the identity map defined by i(x) 5 x for all x  R. Then, i  S. Now, let

I 5 {if : f  S and f (0) 5 0}

      and J 5 {i2f 1 ni2 : f  S, f (0) 5 0 and n  Z}.

Then, I is an ideal of the ring S and J is an ideal of I. However, J fails to

be an ideal of S, since i2  J, 1  S and 1 i2  J, where 1 denotes the
2 2 2

Q001-Algebra-111001_CH 10.indd 11 9/21/2011 4:53:55 PM

 10-12   Algebra – Abstract and Modern

c­ onstant map of R which maps every element of R onto the real number 1 .
2
The assumption 12 i2  J leads to a contradiction; for, let 12 i2  J. Then,

1 2
i 5 i2f 1 ni2 for some f  S with f (0) 5 0 and n  Z.
2
Therefore, fi2 5 ( 12 2 n)i2; that is, f (x)x2 5 ( 12 2 n)x2 for all x  R. If x  0,
f (x) 5 12 2 n. Therefore, f is a nonzero constant function on R 2 {0} and
f (0) 5 0. This is a contradiction to the fact that f is continuous.
Next we discuss a characterization theorem for ideals of a matrix ring
Mn(R), where R is an arbitrary ring with unity.

Theorem 10.1.10. Let R be a ring with unity, n be a positive integer and

Mn(R) be the ring of n 3 n matrices over R. For any I ⊆ Mn(R), I is an ideal of
Mn(R) if and only if I 5 Mn(J) for some ideal J of R.

Proof: It can be easily verified that Mn(J) is an ideal of Mn(R) for any ideal J
of R. Conversely suppose that I is an ideal of Mn(R).
For any 1 # i, j # n, let Eij be the n 3 n matrix over R such that the ijth
entry is 1 and all other entries are 0. Then, any n 3 n matrix (aij) can be
expressed as

n
( aij )  ∑ aij Eij
i , j1

1 if j  k
and EijEkr 5 jkEir, where  jk   .
0 if j  k
Now, consider the given ideal I of Mn(R) and define

J 5 {a  R : a 5 a11 for some (aij)  I}.

That is, J is the set of all 11 entries (entries in the first column and first row) of
the matrices belonging to I. Since I  , J is a nonempty subset of R. Clearly
a 2 b  J for any a and b  J. Next, suppose that a  J and r  R. Then, a
5 a11 for some (aij)  I. Since I is an ideal of Mn(R), we have

 n 
ra E11  r E11 ∑ aij Eij  E11  rE11 ( aij ) E11  I
i , j1 
 n 
   and ar E11  E11 ∑ aij Eij  rE11  E11 ( aij )rE11  I
i , j1 

Q001-Algebra-111001_CH 10.indd 12 9/21/2011 4:53:56 PM

 Ideals and Quotient Rings   10-13

and hence ra and ar  J (since ra is the 11th entry of raE11 and ar is the 11th
entry of arE11). Thus, J is an ideal of R. We shall prove that I 5 Mn(J).
n
Let A 5 (aij)  I. Then, A   aij Eij . For any 1 # i, j # n, consider
i , j1

 n  n
E1i AE j1  E1i ∑ ars Ers  E j1  ∑ ars ir E1s E j1
r ,s1  r ,s1

n
 ∑ ars ir sj E11  aij E11.
r ,s1

Since A  I and aijE11 5 E1iAEj1  I, we get that aij  J. Therefore, I ⊆

Mn(J).
On the other hand, let A 5 (aij)  Mn(J). Then, aij  J for all i and j. Now, for
any 1 # i, j # n, aij  J and hence there exists B 5 (brs)  I such that b11 5
aij. Now,

 n  n
Ei1 BE1 j  Ei1  ∑ brs Ers  E1 j  ∑ brs Ei1 Ers E1 j
 r ,s1  r , s1

n
 ∑ brs 1r Eis E1 j
r , s1
n
 ∑ brs 1r s1 Eij  b11 Eij  aij Eij .
r , s1

Therefore, aijEij 5 Ei1BE1j  I, since I is an ideal and B  I. Now,

n
A   aij Eij  I . Therefore, Mn(J) ⊆ I. Thus, I 5 Mn(J) and J is an ideal of R.
i , j1

The above theorem is false for rings without unity. This is illustrated in the
following example.

Example 10.1.3. Consider the ring 2Z of even integers. Note that 2Z has no
unity. Consider the ring M2(2Z) of 2 3 2 matrices over 2Z. Let

I 5 {(aij)  M2(2Z) : a12  4Z}.

It can be easily checked that I is an ideal of M2(2Z). Note that I  M2(J) for
any ideal J of 2Z.
Theorem 10.1.10 is useful only to the extent that we can describe the ideals
of a ring R, which is not usually easy to do, although it is easy for the ring Z
of integers (recall that any ideal of Z, being a subgroup of (Z, 1), is generated
by a nonnegative integer). However, the ideals of fields are easy to describe.

Q001-Algebra-111001_CH 10.indd 13 9/21/2011 4:53:56 PM

 10-14   Algebra – Abstract and Modern

Theorem 10.1.11. Let R be a nontrivial commutative ring with unity. Then,

R is a field if and only if {0} and R are the only ideals of R.

Proof: Clearly {0} and R are ideals of R and these are distinct, since R is
nontrivial. If R is a field and I  {0} is an ideal of R, then there exists 0 
a  I and hence 1 5 a21a  I, so that I 5 R. Conversely suppose that {0}
and R are the only ideals of R. Let 0  a  R and I 5 aR 5 <a>. Then, I is a
nonzero ideal of R and hence I 5 R. In particular, 1  R 5 I 5 aR and hence
1 5 ab for some b  R. Therefore, a is a unit. Thus, R is a field. b

Corollary 10.1.10. The ring R of real numbers (or the ring C of complex
numbers) has only two ideals, namely {0} and the whole ring.

Corollary 10.1.11. For any positive integer n, the ring Mn(R) of n 3 n matri-
ces over R has exactly two ideals namely {0} and the whole ring Mn(R).
Since Mn(R) is a noncommutative ring for n > 1, Mn(R) is not a field (and
not a division ring) even though it has only two ideals. This says that the
commutativity of the ring R in Theorem 10.1.11 cannot be dropped from the
hypothesis. Nontrivial rings having only two ideals are called simple rings.
Mn(R) is a simple ring for all n  Z1.

Worked Exercise 10.1.1. Let I and J be ideals of a ring R with unity. Then
prove that R 5 I ! J if and only if there are central idempotents a and b in R
such that a 1 b 5 1, ab 5 0, I 5 aR and J 5 bR.

Answer: Suppose that R 5 I ! J. Then, I 1 J 5 R and I ∩ J 5 {0}. Since 1

 R 5 I 1 J, we have

1 5 a 1 b for some a  I and b  J.

Then, since ab  I ∩ J 5 {0}, we have ab 5 0. Therefore, a 5 a  1 5 a(a 1

b) 5 a2 1 ab 5 a2. Similarly, b 5 b2. Also, for any x  R,

ax 1 bx 5 (a 1 b)x 5 x 5 x(a 1 b) 5 xa 1 xb
and hence ax 2 xa 5 xb 2 bx  I ∩ J 5 {0}

so that ax 5 xa and bx 5 xb. Therefore, a and b are central idempotents in

R, Also, since a  I, aR ⊆ I. Further, if x  I, then bx  I ∩ J 5 {0} and
hence

x 5 (a 1 b)x 5 ax 1 bx 5 ax 1 0 5 ax  aR.

Q001-Algebra-111001_CH 10.indd 14 9/21/2011 4:53:57 PM

 Ideals and Quotient Rings   10-15

Thus, I 5 aR. Similarly, J 5 bR.

Conversely suppose that a and b are central idempotents such that a 1 b 5 1,
ab 5 0, aR 5 I and bR 5 J. Then, for any x  R,

x 5 (a1b)x 5 ax 1 bx  I 1 J

and hence I 1 J 5 R. Also,

x  I ∩ J ⇒ x 5 ay 5 bz for some y and z  R

⇒ x 5 a2y 5 ax 5 a(bz) 5 abz 5 0

and hence I ∩ J 5 {0}. Thus, R 5 I ! J.

Worked Exercise 10.1.2. Let I, J and K be ideals of a ring R such that I ⊆ K.

Then prove that

I 1 (J ∩ K) 5 (I 1 J) ∩ K.

This is known as modular law.

Answer: Since I ⊆ I 1 J and I ⊆ K, we have I ⊆ (I 1 J) ∩ K. Also, J ∩ K ⊆

(I 1 J) ∩ K. Therefore, I 1 (J ∩ K) ⊆ (I 1 J) ∩ K. On the other hand, let x 
(I 1 J) ∩ K. Then, x  K and x 5 a 1 b for some a  I and b  J. Now,

b 5 x 2 a  K (since x  K and a  I ⊆ K)

and hence b  J ∩ K and x 5 a 1 b  I 1 (J ∩ K). Therefore, (I 1 J) ∩ K ⊆

I 1 (J ∩ K) Thus, I 1 (J ∩ K) 5 (I 1 J) ∩ K.

Worked Exercise 10.1.3. Consider the ring Z of integers. For any positive
integers n and m, let I 5 nZ and J 5 mZ. Then, compute I 1 J, I ∩ J and IJ.

Answer: Let (m, n) and [m, n] be the greatest common divisor and least com-
mon multiple of m and n, respectively. Note that a  nZ if and only if a is a
multiple of n (or n divides a). Therefore,

a  I 1 J ⇔ a  nZ 1 mZ
⇔ a 5 nx 1 my for some x, y  Z
⇔ (m, n) divides a
⇔ a  (m, n)Z
a  I ∩ J ⇔ a  nZ ∩ mZ

Q001-Algebra-111001_CH 10.indd 15 9/21/2011 4:53:57 PM

 10-16   Algebra – Abstract and Modern

⇔ a is a common multiple of n and m

⇔ a is a multiple of [m, n]
⇔ a  [m, n]Z
r
a  IJ ⇔ a  ∑ xi yi , xi  nZ and yi  mZ
i1
r
⇔ a  ∑ xi yi , n divides xi and m divides yi
i1

⇔ mn divides a
⇔ a  mnZ.

Thus, I 1 J 5 (m, n)Z, I ∩ J 5 [m, n]Z and IJ 5 mnZ.

Worked Exercise 10.1.4. Let f : R → S be a homomorphism of rings. If J is an

ideal of S, then prove that f21(J) is an ideal of R. Also, if f is an epimorphism
and I is an ideal of R, prove that f (I) is an ideal of S.

Answer: Let J be an ideal of S. Then,

f21(J) 5 {a  R : f (a)  J}  , since f (0) 5 0  J.

For any a and b  R, we have

a and b  f21(J) ⇒ f (a) and f (b)  J

⇒ f (a 2 b) 5 f (a) 2 f (b)  J
⇒ a 2 b  f21(J)
and a  f (J) and x  R ⇒ f (a)  J and f (x)  S
21

⇒ f (ax) 5 f (a)f (x)  J

and f (xa) 5 f (x)f (a)  J ⇒ ax and xa  f21(J).

Thus, f 21(J) is an ideal of R.

Next, let f be an epimorphism and I be an ideal of R. Then,

f (I) 5 {f (a): a  I}  , since I  .

Now, we have

x and y  f (I) ⇒ x 5 f (a) and y 5 f (b), a and b  I

⇒ x 2 y 5 f (a) 2 f (b) 5 f (a 2 b)  f (I)

Q001-Algebra-111001_CH 10.indd 16 9/21/2011 4:53:57 PM

 Ideals and Quotient Rings   10-17

    and x  f (I) and s  S ⇒ x 5 f (a) for some a  I and

s 5 f (r) for some r  R
⇒ xs 5 f (a)f (r) 5 f (ar)  f (I)
              and sx 5 f (r)f (a) 5 f (ra)  f (I).

Thus, f (I) is an ideal of S.

Worked Exercise 10.1.5. Let R be a ring with unity and I ⊆ R 3 R. Prove

that I is an ideal of the ring R 3 R if and only if I 5 I1 3 I2 for some ideals
I1 and I2 of R.

Answer: Suppose that I is an ideal of R 3 R. Put

I1  {a  R : ( a, b)  I for some b  R}

     and I 2  {b  R : ( a, b)  I for some a  R}.

Then, I1 and I2 are ideals of R, since I1 5 p1(I) and I2 5 p2(I), where p1 and
p2 : R 3 R → R are the first and second projections, respectively and since
p1 and p2 are epimorphisms of rings. Also,

(a, b)  I ⇒ a  I1 and b  I2
⇒ (a, b)  I1 3 I2.

Therefore, I ⊆ I1 3 I2. On the other hand,

(a, b)  I1 3 I2 ⇒ a  I1 and b  I2
⇒ (a, c)  I1 and (d, b)  I2 for some c, d  R.
⇒ (a, 0) 5 (1, 0)(a, c)  I and (0, b) 5 (0, 1)(d, b)  I
⇒ (a, b) 5 (a, 0) 1 (0, b)  I

and therefore I1 3 I2 ⊆ I. Thus, I 5 I1 3 I2.

Converse can be easily proved.

EXERCISE 10(a)

1. Determine all the ideals in each of the following rings under the operations.
(i) The ring Z of integers.
(ii) The ring Q of rational numbers.

Q001-Algebra-111001_CH 10.indd 17 9/21/2011 4:53:58 PM

 10-18   Algebra – Abstract and Modern

(iii) The ring R of real numbers.

(iv) The ring C of complex numbers.
(v) The ring QR of real quaternions.
(vi) The ring Zn of integers modulo n for any n > 0.
(vii) M2(R), the ring of 2 3 2 real matrices over R.
(viii) The ring Mn(Z) of n 3 n matrices over Z, for any n > 0.
(ix) Z12
(x) Z13 3 Z13

2. Which of the following are true? Substantiate your answers.

(i) Z is an ideal of Q.
(ii) Q is an ideal of R.
(iii) Every subring of a ring R is an ideal of R.
(iv) For any ideal I of a ring R, I 1 I 5 I.
(v) For any ideal I of a ring R, I 5 {a 2 b : a, b  I}.
(vi) There is a finite ideal in any ring.
(vii) There can be infinitely many ideals in a finite ring.
(viii) There is a ring with exactly three ideals.

3. For any subsets S and T of a ring R, prove that

<S ∪ T> 5 <S> 1 <T>.
Is <S ∩ T> 5 <S> ∩ <T> true?

4. Let R be a ring with unity. If R is a division ring, prove that R has only two ideals.
Is the converse true?

5. Let R be a commutative ring and A ⊆ R. Prove that

A* 5 {x  R : xa 5 0 for all a  A}
is an ideal of R. A* is called the annihilator of A in R.
6. For any ideals I and J of a commutative ring R, prove the following:
(i) (I 1 J)* 5 I* ∩ J* 5 (I ∪ J)*
(ii) I ⊆ J ⇒ J* ⊆ I*
(iii) I* 1 J* ⊆ (I ∩ J)*
7. Let I be an ideal of the ring Z of integers and 13  I. Then prove that I 5 13Z or
I 5 Z.
8. Prove the Exercise 7 above with an arbitrary prime number in place of 13.
9. Let R be a commutative ring and I be an ideal of R. Let
r(I) 5 {a  R: an  I for some n  Z1}
Prove that r(I) is an ideal of R containing I.

Q001-Algebra-111001_CH 10.indd 18 9/21/2011 4:53:58 PM

 Ideals and Quotient Rings   10-19

10. For any ideals I and J of a commutative ring R, prove the following:
(i) r(I ∩ J) 5 r(I) ∩ r(J)
(ii) I ⊆ J ⇒ r(I) ⊆ r(J)
(iii) r(I) 1 r(J) ⊆ r(I 1 J)
(iv) r(I 1 J) 5 r(r(I) 1 r(J))
11. For any ideas I and J of a ring R, let
(I : J) 5 {x  R : xa  J for all a  I}.
Prove that (I : J) is a left ideal of R.
12. Express each of the following in the form of nZ for a suitable n.
(i) 8Z ∩ 12Z
(ii) 6Z 1 9Z
(iii) (6Z : 9Z)
(iv) r(12Z)
(v) (12Z)*
(vi) (9Z : 6Z)
13. For any positive integers m and n, express each of the following in the form of
dZ for a suitable integer d:
(i) mZ ∩ nZ
(ii) mZ 1 nZ
(iii) r(mZ)
(iv) (nZ : mZ)
(v) (mZ)*

14. Consider the ring Z[ 3]  {a  b 3 : a and b  Z}, under the usual addition
and multiplication of real numbers. Prove that the set
I  {a  b 3 : a, b  Z and a  b is even}

is an ideal of Z[ 3].
15. Let R be a ring without zero divisions. If every subring of R is an ideal of R, then
prove that R is commutative.

16. Let S be a subset of a ring R and define

Annl(S) 5 {x  R : xs 5 0 for all s  S}.
Then prove that Annl(S) is a left ideal of R. If S is a left ideal of R, then prove that
Annl(S) is an ideal of R. Annl(S) is called the left annihilator of S.

17. Define the notion of the right annihilator Annr(S) of S and formulate and prove a
statement similar to the Exercise 16 for right annihilators.
18. Prove that every ideal of Zn is a principal ideal.

Q001-Algebra-111001_CH 10.indd 19 9/21/2011 4:53:58 PM

 10-20   Algebra – Abstract and Modern

19. Let R be commutative ring in which {0} and R are the only ideals. Then prove that
R is field or R is a finite ring with |R| as a prime and ab 5 0 for all a and b  R.

20. Prove that the set

I  {a  b 2 : a and b  Z and a is even}

is an ideal of the ring Z[ 2].

21. In the ring Z[i] of Gaussian integers, prove that the set
I  {a  bi : a and b are even}
is an ideal and find the annihilator I*.

22. If R is a simple ring with unity, then prove that the ring Mn(R) of all n 3 n matri-
ces is a simple ring.

10.2 QUOTIENT RINGS

The concept and construction of quotient rings are same as for quotient
groups. For an ideal I of a ring R, I is a subgroup of the abelian group (R,
1) and hence I is a normal subgroup of the group (R, 1) and therefore, as in
Theorem 4.6.1, we can construct the quotient group (R, ) as the group of
all cosets a 1 I, a  R under the operation defined by I

(a 1 I) 1 (b 1 I) 5 (a 1 b) 1 I.

Then, (R, ) is an abelian group. Since R is a ring, we have the multipli-

I
cation in R and it is natural to ask whether the quotient group (R, 1)/I has a
corresponding ring structure using (a 1 I)  (b 1 I) 5 ab 1 I as the multipli-
cation. This is answered positively in the following theorem.

Theorem 10.2.1. Let I be an ideal of a ring (R, 1, ) and

R/I 5 {a 1 I : a  R}.

For any a 1 I and b 1 I in R/I, define

(a 1 I) 1 (b 1 I) 5 (a 1 b) 1 I
        and (a 1 I)  (b 1 I) 5 ab 1 I.

Then, (R/I, 1, ?) is a ring.

Proof: Since I is a subgroup of (R, 1), which is an abelian group, I becomes

a normal subgroup of (R, 1). Therefore, by Theorem 4.6.1, (R/I, 1) is an

Q001-Algebra-111001_CH 10.indd 20 9/21/2011 4:53:59 PM

 Ideals and Quotient Rings   10-21

abelian group. Next, with regard to the multiplication in R/I, we should first
prove that the operation on R/I is well defined.
To do this, for any a, b, a9 and b9  R, we have

a  I  a I 
 ⇒ a  a  I and b  b  I
b  I  b I 
⇒ ( a  a) b  I and a(b  b)  I
⇒ ab  ab ( a  a)b  a(b  b)  I
⇒ ab  I  ab I .

Therefore, the multiplication on R/I depends on the cosets, but not on their
representatives. For any a, b and c  R, we have

((a 1 I)  (b 1 I))  (c 1 I) 5 (ab)c 1 I 5 a(bc)I

5 (a 1 I)  ((b1I)  (c 1 I)).

Therefore, ? is associative. Also,

(a 1 I)  ((b 1 I) 1 (c 1 I)) 5 a(b 1 c) 1 I

5 (ab 1 ac) 1 I
5 (ab 1 I) 1 (ac 1 I)
5 ((a 1 I)(b 1 I)) 1 ((a 1 I)(c 1 I)).

Therefore, ? distributes over 1 from left and, similarly from right also. Thus,
(R/I, 1, ?) is a ring.

Definition 10.2.1. For any ideal I of a ring R, the ring (R/I, 1, ?) constructed
above is called quotient ring of R by I or factor ring of R by I.

Note 10.2.1
1. The zero element in the quotient ring R/I is 0 1 I 5 I, where 0 is the zero
element in R.
2. If the ring R has unity 1, then R/I also has unity, namely 1 1 I. The con-
verse may not hold good. That is, R/I may have unity while R has no unity.
For example, R is an ideal of R and the quotient R/R is the trivial ring
which obviously has unity (when a ring has only one element, then that
element is the additive identity as well as the multiplicative identity).
3. If R is a commutative ring, then the quotient R/I is also commutative ring
for any ideal I or R.

Q001-Algebra-111001_CH 10.indd 21 9/21/2011 4:53:59 PM

 10-22   Algebra – Abstract and Modern

In group theory, we have defined the concept of the kernel of a homomor-

phism and proved that it is a normal subgroup of the domain group and con-
versely, any normal subgroup is the kernel of a homomorphism. These results
are extended to the case of rings in the following ­definition.

Definition 10.2.2. Let f : R → S be a homomorphism of rings. Then, the

Kernel of f is defined to be the set

ker f 5 {a  R: f (a) 5 0 in S}.

Theorem 10.2.2. The kernel of any homomorphism of rings is an ideal of

the domain ring.

Proof: Let f : R → S be a homomorphism of rings. Then, f is a homomor-

phism of the group (R, 1) into the group (S, 1) and hence ker f is a subgroup
of (R, 1). Also,

a  ker f and r  R ⇒ f (a) 5 0 and r  R

⇒ f (ra) 5 f (r)f (a) 5 f (r)0 5 0
and f (ar) 5 f (a)f (r) 5 0f (r) 5 0 ⇒ ra and ar  ker f.

Thus, ker f is an ideal of R.

We prove the converse of the above result; that is, any ideal I of a ring R is
the kernel of a homomorphism of R into some ring S.

Theorem 10.2.3. Let I be an ideal of a ring R. Then, there exists a ring S and
a homomorphism f : R → S such that I 5 ker f.

Proof: Consider the quotient ring R/I and define f : R → R/I by f (a) 5 a 1 I
for any a  R. Then, f is a homomorphism of rings; for,

f (a 1 b) 5 (a 1 b) 1 I 5 (a 1 I) 1 (b 1 I) 5 f (a) 1 f (b)
    and f (ab) 5 ab 1 I 5 (a 1 I)(b 1 I) 5 f (a)f (b)

for any a and b  R. Also,

ker f 5 {a  R : f (a) 5 0 in R/I}

5{a  R : a 1 I 5 I} 5 I

Q001-Algebra-111001_CH 10.indd 22 9/21/2011 4:53:59 PM

 Ideals and Quotient Rings   10-23

Definition 10.2.3. For any ideal I of a ring R, the homomorphism f : R → R/I,

defined by f (a) 5 a 1 I for any a  R, is called the canonical homomorphism
or natural homomorphism. Actually, f is an epimorphism, since any element
of R/I is of the form a 1 I for some a  R.
The fundamental theorem of homomorphisms of groups proved in
­Theorem 5.2.1 is extended for rings in the following theorem.

Theorem 10.2.4 (Fundamental Theorem of Homomorphism for Rings). Let

f : R → S be a homomorphism of rings. Then, f (R) is a subring of S and

R / ker f  f ( R).

In particular, if f is an epimorphism, then R/ker f  S.

Proof: The proof is same as that of Theorem 5.2.1, except that the map g :
R/K → f (R) defined by

g(a 1 k) 5 f (a)

is a ring homomorphism also, where k 5 ker f.

This is clear from

g((a1k)(b1k)) 5 g(ab 1 k) 5 f (ab) 5 f (a)f (b).

The above fundamental theorem can be restated as ‘any homomorphic image

of a ring R is isomorphic to a quotient ring of R’. This is not only a fundamen-
tal result but also an important tool in proving several isomorphism theorems
for quotient ring. Some of these are listed below and their proofs are similar
to those proved in Section 5.3.

Theorem 10.2.5. For any ideals I and J of a ring R,

I J
I .
I ∩J J

Theorem 10.2.6. Let f : R → S be an epimorphism of rings and I be an ideal

of R such that ker f ⊆ I. Then, f (I) is an ideal of S and

R S
 .
I f (I )

Q001-Algebra-111001_CH 10.indd 23 9/21/2011 4:54:00 PM

 10-24   Algebra – Abstract and Modern

Theorem 10.2.7. Let f : R → S be an epimorphism of rings and J be an ideal

of S. Then, f21(J) is an ideal of R and

R S
1
 .
f (J ) J

Theorem 10.2.8. Let I and J be ideals of a ring and I ⊆ J. Then, J/I is an ideal
of R/I and

(R/I)/(J/I)  R/J.

Further, the correspondence between the subgroups of a quotient group G/N

and the subgroups of G containing N can be easily extended to rings as given
in the following theorem whose proof is routine.

Theorem 10.2.9. Let I be an ideal of a ring R. Then,

J  J/I

is a one-to-one correspondence between the ideals of R containing I and the

ideals of R/I.

Example 10.2.1. Let n be a positive integer and Zn be the ring of integers

modulo n. We shall prove that Zn is isomorphic to a quotient of the ring Z of
integers, by using the fundamental theorem of homeomorphisms, as in the
case where we have treated these as groups alone (see Example 5.2.1). As
usual, define

f : Z → Zn by f (a) 5 r,

where r is the remainder obtained by dividing a with n; that is, a 5 qn 1 r,

0  r , n. Then, f is a epimorphism of rings (see Example 9.5.1 (4)) and ker
f 5 nZ and hence

Z/nZ  Zn.

Worked Exercise 10.2.1. Prove that any nontrivial homomorphic image of a

field is again a field.

Answer: Let R be a nontrivial homomorphic image of a field F. That is, R is a

nontrivial ring and there is an epimorphism f : F → R of rings. Then, consider

Q001-Algebra-111001_CH 10.indd 24 9/21/2011 4:54:00 PM

 Ideals and Quotient Rings   10-25

ker f which is an ideal of F. Since F is a field, ker f 5 {0} or F. Also, F/ker

f  f (F ) 5 R. If ker f 5 F, then F/ker f is trivial and hence R is trivial, which
is a contradiction to the hypothesis that R is nontrivial. Therefore, ker f 5 {0}
and hence

F  F/ker f  R.

Since F is a field, R is also a field.

Worked Exercise 10.2.2. For any ideal I and J of a ring R, prove that R/I ∩ J
is isomorphic to a subring of R/I 3 R/J.

Answer: Define f : R → R/I 3 R/J by

f (a) 5 (a 1 I, a 1 J) for any a  R.

It can be easily verified that f is a homomorphism of the ring R into the prod-
uct ring R/I 3 R/J. By the fundamental theorem of homomorphisms, f (R) is
a subring of R/I 3 R/J and R/ker f  f (R). Since I 5 (0 1 I) and J 5 (0 1 J)
are the zero elements of R/I and R/J, respectively, (I, J) is the zero element in
the ring R/I 3 R/J. Therefore,

ker f 5 {a  R : f (a) 5 zero element in R/I 3 R/J}

5 {a  R : (a 1 I, a 1 J) 5 (I, J)}
5 {a  R : a 1 I 5 I and a 1 J 5 J}
5 {a  R : a  I and a  J}
5 I ∩ J.

Thus, R/I ∩ J  f (R), which is a subring of R/I 3 R/J.

Worked Exercise 10.2.3. Let I 5 3Z/12Z. Prove that I is isomorphic to an

ideal J of Z12 such that Z12/J  Z3.

Answer: Recall that 12Z ⊆ 3Z ⊆ Z and 12Z and 3Z are ideas of Z. By

Theorem 10.2.8,

(Z/12Z)/(3Z/12Z)  Z/3Z  Z3.

Since 3Z/12Z is an ideal of Z/12Z  Z12 (by Example 10.2.1), there must be
an ideal J of Z12 such that

I 5 3Z/12Z  J and Z12/J  Z3.

Q001-Algebra-111001_CH 10.indd 25 9/21/2011 4:54:00 PM

 10-26   Algebra – Abstract and Modern

Worked Exercise 10.2.4. Let I and J be ideals of array R such that I ∩ J 5 {0}
and I 1 J 5 R. Then prove that

J  R/I, I  R/J.

Answer: Define f : J → R/I by f (a) 5 a 1 I for any a  J. Then, clearly f is a

homomorphism of rings. Also, for any x 1 I  R/I, x  R, we can write x 5
b 1 a, for some b  I and a  J (since R 5 I 1 J) and ­therefore x 1 I 5 (b
1 a) 1 I 5 a 1 (b 1 I) 5 a 1 I (since b  I) and hence f is an epimorphism.
Also, for any a  J,

f (a) 5 0 ⇒ a 1 I 5 I, the zero in R/I

⇒aI
⇒ a  I ∩ J 5 {0}
⇒ a 5 0.

Therefore, f is an injection also and hence f is an isomorphism of J onto R/J.

Thus, J  R/I. Similarly, I  R/J.

Worked Exercise 10.2.5. Let I be an ideal of a ring R and the characteristic

of R be n > 0. Prove that the characteristic of the quotient ring R/I is a divi-
sor of n.

Answer: We are given that char(R) 5 n and hence n is the least positive
integer such that na 5 0 for all a  R. Now, for any a 1 I  R/I, a  R,
we have

n(a 1 I) 5 na 1 I 5 0 1 I 5 I, the zero in R/I

and therefore char(R/I) > 0 and, by Theorem 9.3.1, char(R/I) is a divisor of n.

Worked Exercise 10.2.6. Let P(X) be the power set of any set X. Let Y be a non-
empty proper subset of X. Prove that P(Y) is an ideal of the ring (P(X), 1, ∩)
and describe the quotient ring P(X)/ P(Y).

Answer: We are given that P(Y) 5 {A : A ⊆ Y}. We have

A and B  P(Y) ⇒ A 1 B 5(A 2 B) ∪ (B 2 A) ⊆ A ∪ B ⊆ Y

       and A ∩ Z ⊆ Y for all Z  P(X).

Q001-Algebra-111001_CH 10.indd 26 9/21/2011 4:54:00 PM

 Ideals and Quotient Rings   10-27

Therefore, P(Y) is an ideal of the ring (P(X), 1, ∩). Any element of the quo-
tient ring P(X)/P(Y) is of the form Z 1 P(Y) for some Z  P(X); that is,
Z ⊆ X. Now, we can write

Z 5 (Z ∩ Y) ∪ (Z ∩ (X 2 Y))
5 (Z ∩ Y) 1 (Z ∩ (X 2 Y))
and hence Z 1 P(Y) 5 ((Z ∩ Y) 1 P(Y)) 1 ((Z ∩ (X 2 Y)) 1 P(Y))
5 (Z ∩ (X 2 Y)) 1 P(Y), since Z ∩ Y  P(Y).

Therefore, P(X)/ P(Y) 5 {A 1 P(Y) : A ⊆ X 2 Y}.

Note that, for any A and B ⊆ X 2 Y,

A 1 P(Y) 5 B 1 P(Y) ⇒ A 2 B  P(Y)

⇒A1B#Y
⇒ A 1 B ⊆ Y ∩ (X 2 Y) 5 
⇒A1B5
⇒A52B5B

Thus, A  A 1 P(Y) is a bijective map of P(X 2 Y) onto P(X ) P(Y ). It can

be easily verified that this map preserves the ring operations in P(X 2 Y) and
P(X ) P(X )
P(Y ). Thus, the factor ring P(Y ) is isomorphic to P(X 2 Y). One
can consider the map f : P(X) → P(X 2 Y) defined by f (A) 5 A ∩ (X 2 Y). It
can be verified that f is an epimorphism of rings and ker f 5 P(Y) and hence,
by the fundamental theorem of homomorphisms,

P(X)/ P(Y)  P(X 2 Y).

EXERCISE 10(b)

1. Determine all the elements of each of the following quotient rings.

(i) Z/5Z
(ii) 3Z/6Z
(iii) 2Z/10Z
(iv) 3Z/9Z
 a 2b   
(v)R/I, where R    : a, b, c and d  Z and I   a 2b : a, b, c and d  Z

   
 

3c d   
3c 3d  
   
2b 


  a 2 b  

 
 : a, b, c and d  Z and I    
 : a, b, c and d  Z 
d  

3c 3d 



(vi) P(X)/I, where X 5 {1, 2, 3, 4, 5} and I 5 P({2,4}).

Q001-Algebra-111001_CH 10.indd 27 9/21/2011 4:54:01 PM

 10-28   Algebra – Abstract and Modern

2. State whether each of the following is true or false. Substantiate your answers.
(i) For any ring R, R has unity if and only if R/I has unity for all ideals I of R.
(ii) For any ideal I of a commutative ring, R/I is commutative.
(iii) A ring R is commutative if and only if R/I is commutative for all ideals
I of R.
(iv) For any integral domain R, R/I is an integral domain for any ideal I of R.
(v) A ring R is an integral domain if and only if R/I is an integral domain for
any ideal I of R.
(vi) Any quotient ring of a field is a field.
(vii) For any ideal I of a ring R, R/I is a field if and only if R is a field.
(viii) A nontrivial ring R is a field if R/I is a field for all proper ideal, I of R.
3. Prove that a homomorphism f : R → S of rings is an injection if and only if
ker f 5 {0}.
4. For any subring S of a ring R, prove that the multiplication of additive cosets of
S in R is well defined by the equation
(a 1 S)(b 1 s) 5 ab 1 S
if and only if S is an ideal of R.
5. For any ideals I and J of a ring R, prove that the set
J  {a  I : a  J }

is an ideal of R/I.
6. Determine all the idempotents, nilpotents and units in each of the following
quotient rings
(i) Z/6Z
(ii) Z/8Z
(iii) Z/7Z
7. Let I be an ideal of the ring Z such that Z/I is an integral domain. Then prove that
I 5 {0} or I 5 pZ for some prime number p.
8. Prove that the following are equivalent to each other for any positive integer n.
(i) n is a prime number.
(ii) Z/nZ is a field.
(iii) Z/nZ is an integral domain.
9. Let R be a commutative ring and N be the set of all nilpotents in R. Then prove
that N is an ideal of R and the quotient ring R/N has no nonzero nilpotents.
10. Prove that Zn’s, {0} and Z are the only homomorphic images of the ring Z of
integers.

Q001-Algebra-111001_CH 10.indd 28 9/21/2011 4:54:01 PM

 Ideals and Quotient Rings   10-29

11. Let S be a subring and I be an ideal of a ring R such that S ∩ I 5 {0}. Prove that
S is isomorphic to a subring of the quotient ring R/I.

12. For any two subsets A and B of a ring R, let the product of A and B be defined by
the set
AB 5 {ab : a  A and b  B}.
Give an example of an ideal I of a ring R such that the product (x 1 I) (y 1 I) of
two cosets (x 1 I) and (y 1 I) is properly contained in the coset (xy 1 I).

13. For any ideal I of a ring R, prove that

Mn(R)/Mn(I)  Mn(R/I)
for any positive integer n, where Mn(S) denotes the ring of n 3 n matrices over S.

14. For any pair of relatively prime positive integers m and n, prove that Zm/nmZ
 Z n.

15. Let R1, …, Rn be rings and I1, …, In be ideals of R1, …, Rn, respectively. Then
prove that
R1 3 … 3 Rn/I1 3 … 3 In  R1/I1 3 … 3 Rn/In.

16. For any positive integer n, determine all the ideals of the quotient ring Z/nZ and
all the homomorphic images of Z/nZ.

10.3 CHINESE REMAINDER THEOREM

In this section, we extend a remarkable result known as ‘Chinese Remainder
Theorem’ in the theory of numbers to the ideals of a ring. Recall that the set
Z of integers forms a ring under the usual addition and multiplication of inte-
gers and that the ideals of Z are of the form nZ for some nonnegative integer
n. Note that, if I 5 nZ and a  Z, then

a  I 5 nZ ⇔ n divides a.

The classical version of the Chinese Remainder Theorem is that ‘given dis-
tinct primes p1, p2, …, pn and integers a1, a2, …, an, one can always find an
integer a such that

a  ai(mod pi) for all 1  i  n;

that is, a 2 ai  piZ or a 1 piZ 5 ai 1piZ for all 1  i  n.’ If we take Ii 5

piZ, then the Chinese Remainder Theorem states that, for any elements a1, a2,
…, an in the ring Z, there exists an element a in Z such that

a 1 Ii 5 ai 1 Ii for all 1  i  n.

Q001-Algebra-111001_CH 10.indd 29 9/21/2011 4:54:01 PM

 10-30   Algebra – Abstract and Modern

In the following theorem, we arrive at a necessary and sufficient condition

on ideals I1, I2, …, In of an arbitrary ring for the validity of the above results.

Theorem 10.3.1. Let R be a ring with identity and I1, I2, …, In be ideals of
R. If Ii 1 Ij 5 R for any i  j, then, for any x1, x2, …, xn  R there exists x 
R such that

x 2 xi  Ii; that is, x 1 Ii 5 xi 1 Ii for all 1  i  n.

Proof: Recall that an ideal I of R is the whole of R if and only if the unity 1
belongs to I. Suppose that

Ii 1 Ij 5 R for all i  j.

First, we shall prove that, for each 1  i  n,

 
I i ∩ I j   R.
 ji 

To prove this, fix 1  i  n and put K i  ∩ I j . For each j  i, we have

j ≠i

1  R 5 Ii 1 Ij and hence 1 5 aj 1 bj

for some aj  Ii and bj  Ij. Now, consider

1  ∏ ( a j  b j )  si  ti ,
ji

where ti   b j and si  Ii, since the expansion of  ( a j  b j ) gives a sum

ji ji
in which all the summands, except  b j , are products involving atleast one
ji
aj which is in Ii and Ii is an ideal. Also, ti   b j  I j for all j  i and hence
ji
ti  ∩ I j  K i .
ji

Therefore,

1 5 si 1 ti, si  Ii and ti  Ki (*)

and hence I i  ( ∩ I j )  I i  K i  R for all 1  i  n. Now, let x1, x2, …, xn

ji
be any elements in R. Put

x 5 x1t1 1 x2t2 1 … 1xntn.

Q001-Algebra-111001_CH 10.indd 30 9/21/2011 4:54:03 PM

 Ideals and Quotient Rings   10-31

Then, for each 1  i  n,

n
x  xi  ∑ x j t j  xi (si  ti ) (by (*))
ji

 ∑ x j t j − xi si  I i
ji


since tj  Ii for all j  i and si  Ii.

A converse of the above result is also true and this is proved in the follow-
ing theorem.

Theorem 10.3.2 Let I1, I2, …, In be ideals in a ring R such that, for any ele-
ments x1, x2, ..., xn in R, there exists x  R with x 2 xi  Ii for all 1  i  j.
Then, Ii 1 Ij 5 R for all i  j.

Proof: Fix i  j. Let a  R. Define x1, x2, …, xn by xi 5 a and xk 5 0 for all

k  i. Then, by hypothesis, there exists x  R such that

x 2 a 5 x 2 xi  Ii and x 5 x 2 xj  Ij

and therefore a 5 (a 2 x) 1 x  Ii 1Ij. Thus, Ii 1 Ij 5 R for all i  j. b

Theorem 10.3.3. Let I1, I2, …, In be ideals of a ring with unity and R/I1,
R/I2, …, R/In be the corresponding quotient rings. Define f : R → R/I1 3 R/I2
3 … 3 R/In by

f (x) 5 (x 1 I1, x 1 I2, …, x 1 In)

for any x  R. Then, f is an epimorphism if and only if Ii 1 Ij 5 R for all i  j

n
and, in this case R / ∩ I i is isomorphic to R/I1 3 R/I2 3 … 3 R/In.
i1

Proof: Clearly f is a homomorphism of rings and

ker f 5 {a  R : f (a) 5 zero in R/I1 3 … 3 R/In}

5 {a  R : (a1I1, …, a 1 In) 5 (I1, …, In)}
5 {a  R : a  Ii for 1  i  n}
n
 ∩ Ii .
i1

Q001-Algebra-111001_CH 10.indd 31 9/21/2011 4:54:03 PM

 10-32   Algebra – Abstract and Modern

Also, f is a surjection if and only if, for any x1, x2, …, xn in R, there exists
x  R such that

x 1 Ii 5 xi 1 Ii for all 1  i  n

which in turn, by Theorems 10.3.1 and 10.3.2, is equivalent to saying that Ii 1

Ij 5 R for all i  j. In this case, we have by the fundamental theorem of
homomorphisms, that

R R R R
f : n
→   is an isomorphism.
I1 I 2 In
∩I
i1
i

Corollary 10.3.1. Let R be a ring with unity and I1, I2, …, In be ideals of R
n
such that Ii 1 Ij 5 R for all i  j and ∩ I i  {0}. Then, R  R/I1 3 R/I2 3
i1
… 3 R/I .
n

Corollary 10.3.2. Let R be a ring with unity and R be the direct sum of ideals
I and J. Then,

R  R/I 3 R/J.

Corollary 10.3.3 (Chinese Remainder Theorem). Let m1, m2, …, mn be posi-

tive integers which are pair-wise relatively prime. For any integers a1, a2, …,
an, there exists an integer a such that

a  ai(mod mi) for all 1  i  n.

Further, a is unique modulo the l.c.m. of {m1, m2, …, mn}.

Proof: Put Ii 5 miZ for 1  i  n. Then, Ii is an ideal of Z and Ii 1 Ij 5 g.c.d.

n
{mi, mi}Z 5 1Z 5 Z for all i  j. Also, ∩ I i  mZ, where m 5 l.c.m. of
i1
{m1, …, mn}. Now, let a1, a2, …, an be any integers. Then, by Theorem 10.3.1,
there exist a  Z such that

9.5a 2 ai  Ii 5 miZ for all 1  i  n

and hence mi divides a 2 ai, so that a  ai(mod mi). Also, if b is any other
integer with this property, then

a 2 b 5 (a 2 ai) 2 (b 2 ai)  Ii for all 1  i  n

n
and hence a  b  ∩ I i  mZ, so that a  b(mod m). Thus, a is unique
i1
modulo m.

Q001-Algebra-111001_CH 10.indd 32 9/21/2011 4:54:04 PM

 Ideals and Quotient Rings   10-33

Corollary 10.3.4 (Classical Chinese Remainder Theorem). Let p1, p2, …, pn

be distinct prime numbers and a1, a2, …, an be any integers. Then, there exists
an integer a such that

a  ai(mod pi) for all 1  i  n

and this a is unique modulo the product p1p2  pn.

Corollary 10.3.5. Let m be a positive integer greater than 1 and

m  p1r1 p2r2 … pnrn ,

where p1, p2, …, pn are distinct primes and r1, r2, …, rn are positive integers.
Then,

Z m  Z p r1  Z p r2  Z p rn .
1 2 n

Proof: Z m  Z mZ  Z p1r1 Z  Z pnrn Z  Z p r1 ×× Z p rn .

1 n 
Worked Exercise 10.3.1. Prove that Z12  Z4 3 Z3.

Answer: Since 12 5 22 3 31 and 2 and 3 are distinct primes, it follows from

Corollary 10.3.5 that

Z12  Z 22  Z31  Z 4  Z3 .

Worked Exercise 10.3.2. Can Z12  Z2 3 Z2 3 Z3?

Answer: No; for, char(Z2) 5 2 and char(Z3) 5 3 and hence char(Z2 3 Z2 3

Z3) 5 l.c.m {2, 2, 3} 5 6. But char(Z12) = 12 and hence Z12 cannot be iso-
morphic with Z2 3 Z2 3 Z3.

EXERCISE 10(c)

1. Which of the following are true? Substantiate your answer.

(i) Z16 3 Z5  Z80
(ii) Z6 3 Z5  Z2 3 Z15
(iii) Z16 3 Z20  Z320

Q001-Algebra-111001_CH 10.indd 33 9/21/2011 4:54:04 PM

 10-34   Algebra – Abstract and Modern

(iv) P(X)  P(Y) 3 P(X 2 Y) for any subset Y of a set X, where P(X) is the
ring. (P(X), 1, ∩)
(v) Z  Z3 3 Z2
(vi) Z3ZZ

2. Let R be any ring with unity and

a b 
S    : a, b and c  R.
 
0 c  
 
Then prove that S is a ring under the usual addition and multiplication of matri-
ces over R.

3. Let I and J be ideals of a ring R with unity and S be the ring given in Exercise 2.
Let
a b 
K    : a  I , c  J , b  R.

0 c  
 

Then prove that K is an ideal of S and that

S/K  R/I 3 R/J.
4. Let {Rj}jJ be a class of rings and

∏ R { f
j J
j : J → ∪ R j : f ( j )  R j for all j  J }.
j J

Then prove that  R j is a ring under the point-wise addition and multiplication.
j J

5. Let ⊕ R j  {f   R j : f (j )  0 for all but finite j’s in J }. Then prove that ⊕ R j

j J j J j J
is a subring of the  R j given in 4 above.
j J

6. For any nonempty proper subset K of J, prove that there is a subring S of

⊕ R j given in 5 above such that S  ⊕ R , S is an ideal of ⊕ R and
j J j j
j K j J
⊕ Rj S  ⊕ Rj .
j J j  J K
n
7. For any ideals I1, I2, …, In of a ring R, prove that R / ∩ I j is isomorphic to a sub-
n j1
ring of ⊕ R / I j .
j1

8. For any class {I j } j J of ideals of a ring R, prove that R / ∩ I j is isomorphic to

j J
a subring of  R I j .
j J

10.4 Prime ideals

We have noticed earlier that a quotient ring of a ring may be an integral
domain irrespective of whether the given ring is so. For example, the ring Z

Q001-Algebra-111001_CH 10.indd 34 9/21/2011 4:54:06 PM

 Ideals and Quotient Rings   10-35

of ­integers is an integral domain and the quotient ring Z/6Z( Z6 ) is not an

integral domain, while the quotient ring Z/5Z( Z5 ). is an integral domain.
In this section, we discuss ideals with respect to which the quotient ring
becomes an integral domain.

Definition 10.4.1. Let R be a ring. A proper ideal P of R is said to be a prime

ideal if, for any ideals I and J of R,

IJ ⊆ P ⇒ I ⊆ P or J ⊆ P.

Example 10.4.1
1. {0} is a prime ideal of the ring Z of integers.
2. A nonzero ideal I of Z is prime if and only if I 5 pZ for some prime
number p.
3. Consider the ring Z 3 Z under co-ordinate wise addition and multiplica-
tion. Then, Z 3 {0} and {0} 3 Z are prime ideals of Z 3 Z.
4. {0} is not a prime ideal in the ring M2(R) of 2 3 2 matrices over the real
number system.

Theorem 10.4.1. Let P be a proper ideal of a ring R such that, for any a and
b in R,

ab  P ⇒ a  P or b  P.

Then, P is a prime ideal. The converse holds if R is a commutative ring.

Proof: Let I and J be ideals of R such that I  P and J  P. Then, there exist ele-
ments a and b such that a  I, a  P, b  J and b  P. Then, by the hypothesis,
ab  P. Since ab [ IJ, it follows that IJ  P. Thus, P is a prime ideal of R.
Conversely suppose that R is a commutative ring and P is a prime ideal of R. Let
a and b [ R such that a  P and b  P. Consider the ideals <a> and <b>. We
have <a>  P and <b>  P and hence <a><b>  P. Therefore, there exists an
element x  <a><b> such that x  P. x is a finite sum of elements of the form

(ra 1 na) (sb 1 mb) 5 rsab 1 mrab 1 nsab 1 nmab

(see Corollary 10.1.3) and, since x  P, it follows that ab  P.

Corollary 10.4.1. A proper ideal P of a commutative ring R is prime if and

only if, for any a and b in R,

ab [ P ⇔ a [ P or b [ P.

Q001-Algebra-111001_CH 10.indd 35 9/21/2011 4:54:07 PM

 10-36   Algebra – Abstract and Modern

Theorem 10.4.2. Let P be an ideal of a commutative ring R with unity. Then,

P is prime if and only if the quotient R/P is an integral domain.

Proof: Suppose that P is a prime ideal of R. Since P is proper ideal of R, R/P

is a nontrivial ring. Also, since R is a commutative ring with unity, so is R/P.
Now, for only a 1 P and b 1 P in R/P,

(a 1 P)(b 1 P) 5 zero in R/P ⇒ ab 1 P 5 P

⇒ ab [ P
⇒ a [ P or b [ P
⇒ a 1 P 5 P or b 1 P 5 P.

Thus, R/P is an integral domain.

Conversely, suppose that R/P is an integral domain. Then, R/P is nontrivial
and hence P is a proper ideal of R. Now, for any a and b in R,

ab [ P ⇒ (a 1 P)(b 1 P) 5 ab 1 P 5 P
⇒ (a 1 P)(b 1 P) 5 The zero in R/P
⇒ a 1 P 5 P or b 1 P 5 P
⇒ a [ P or b [ P.

Thus, P is a prime ideal of R.

Let us recall that, for any ideals I and J of a ring R, the set

J
5 {a 1 I : a [ J}
I
is an ideal of the quotient ring R/I and that J  J/I is a one-to-one cor-
respondence between the ideals of R containing I and the ideals of R/I (see
Theorem 10.2.9). This correspondence can be carried to prime ideals also,
as proved in the following theorem.

Theorem 10.4.3. Let I be an ideal of a ring R and P be an ideal of R contain-

ing I. Then, P is a prime ideal of R if and only if P/I is a prime ideal of R/I.

Proof: Suppose that P is a prime ideal of R. Then, P/I is a proper ideal of R/I,
since P is proper in R. Let A and B be ideals of R/I such that AB ⊆ P/I. Then,
A 5 J/I and B 5 K/I for some ideals J and K of R containing I. Also,

JK/I 5 (J/I)(K/I) 5 AB ⊆ P/I

and hence JK ⊆ P, so that J ⊆ P or K ⊆ P, since P is prime. Therefore, A 5

J/I ⊆ P/I or B 5 K/I ⊆ P/I. Thus, P/I is a prime ideal of R/I.

Q001-Algebra-111001_CH 10.indd 36 9/21/2011 4:54:07 PM

 Ideals and Quotient Rings   10-37

Conversely suppose that P/I is a prime ideal of R/I. Then, P/I is proper in R/I
and hence P is a proper ideal of R. Let J and K be ideals of R such that JK ⊆
P, Then, J/I and K/I are ideals of R/I and

(J/I)(K/I) 5 JK/I ⊆ P/I.

Since P/I is a prime ideal of R/I, it follows that J/I ⊆ P/I or K/I ⊆ P/I and
hence J ⊆ P or K ⊆ P. Thus, P is a prime ideal of R.

Corollary 10.4.2. Let I be an ideal of a ring R. Then, P  P/I is a one-to-

one correspondence between the prime ideals of R containing I and the prime
ideals of R/I.

Example 10.4.2. Consider the ring Z of integers in which {0} is a prime

ideal and any nonzero prime ideal is precisely of the form pZ for some prime
number p. Recall that any ideal of Z is a principal ideal. Let I 5 <n> 5 nZ
a a a
and n  p1 1 p2 2 … pr r , where r > 0, p1, p2, …, pr are distinct prime numbers
and a1, a2 …, ar are positive integers. Then, P is a prime ideal of Z containing
nZ if and only if P 5 piZ, for some 1  i  r. Thus, there are exactly r prime
ideals in Z/nZ( Z n ) and these are Pi Z/nZ , 1  i  r.
In the following definition, we introduce a concept which plays an impor-
tant role in many aspects of the theory of ideals in commutative rings.

Definition 10.4.2. Let R be a commutative ring and I be an ideal of R. The nil

radical of I is defined as the set

I  {x  R : x n  I for some n  Z}.

Using the commutativity of the ring R, one can easily prove that I is an
ideal of R containing I. In fact, we prove in the following that I is the inter-
section of all prime ideals of R containing I. Before going to the proof of this,
let us recall an axiom of the theory of sets, which is popularly known as the
Zorns lemma. Though it is called a lemma, it is actually an axiom. There are
several equivalent formulations of this axiom. We present a convenient form
of this in the following lemma.

Zorn’s Lemma 10.4.1. Let X be any set and  be a nonempty class of subsets
of X. Suppose that, for any subclass # of  in which any two members of
# are comparable (that is, for any A and B [ #, either A ⊆ B or B ⊆ A), the
union of all the members of # is a member of . Then,  has a maximal
member; that is, there exists M [  such that M is not properly contained in
any member of .

Q001-Algebra-111001_CH 10.indd 37 9/21/2011 4:54:08 PM

 10-38   Algebra – Abstract and Modern

Theorem 10.4.4. Let I be an ideal of a commutative ring R. Then, the nil

radical of I is equal to the intersection of all prime ideals of R containing I.

Proof: Let I be the nil radical of I and J be the intersection of all prime
ideals of R containing I. We shall prove that I  J . If P is any prime ideal
of R containing I, then

x  I ⇒ x n  I # P for some n  Z
⇒ x n  P , n  Z
⇒ x  P (since P is prime)

and hence I ⊆ P. Therefore, I ⊆ J .

On the other hand, let x [ R such that x  I . Then, xn  I for all n [
Z1. Put

S 5 {xn: n [ Z1}.

Then, x [ S and S ∩ I 5 [. Let

 5 {K : K is an ideal of R, I ⊆ K and S ∩ K 5 [}.
Since I [ ,  is a nonempty class of subsets of R. We shall verify that the
hypothesis in the Zorn’s Lemma is satisfied for . Let # be a subclass of 
such that, for any A and B [ #, either A ⊆ B or B ⊆ A. If K  ∪ A, then K is
A
an ideal of R (by Corollary 10.1.1), I ⊆ K and S ∩ K 5 [ and hence K [ .
Therefore,  satisfies the hypothesis of the Zorn’s Lemma and hence  has a
maximal member, say M. Then, M is an ideal of R, I ⊆ M and x  M (since x
[ S and S ∩ M 5 [). We shall prove that M is a prime ideal, which implies
that x  J. Let a and b [ R such that ab [ M. Suppose, if possible, a  M
and b  M. Then, M  M  <a > and M  M  <b >. By the maximality of
M, M 1 <a> and M 1 <b> cannot be members of . Therefore,

S ∩ (M 1 <a>)  [ and S ∩ (M 1 <b>)  [.

and hence there exist positive integers n and m such that

xn 5 y 1 r1a 1 sa and xm 5 z 1 r2b 1 tb

for some y, z [ M, r1, r2 [ R and s, t [ Z. Now xn1m [ S and

xn1m 5 xnxm 5 (y 1 r1a 1 sa)(z 1 r2b 1 tb)

5 y(z 1 r2b 1 tb) 1 (r1a 1 sa)z 1 r1r2ab 1 tr1ab 1 sr2ab 1 stab.

Q001-Algebra-111001_CH 10.indd 38 9/21/2011 4:54:09 PM

 Ideals and Quotient Rings   10-39

Since y, z and ab [ M, we get that xn1m [ S ∩ M, which is a contradiction to

the fact that S ∩ M 5 [ (since M [ ). Therefore, a [ M or b [ M. Thus,
M is a prime ideal of R. Since x  M and I ⊆ M, we get that x  J. Therefore,
we get that J ⊆ I . Thus, I  J . 

Example 10.4.3. In the ring Z of integers, let n  p1a1 p2a2  prar , where p1, p2,
…, pr are distinct primes and a1, a2, …, ar are positive integers. Then, p1Z,
p2Z, …, prZ are all the prime ideals containing nZ and hence

nZ  ( p1Z) ∩ ( p2 Z) ∩  ∩ ( pr Z)  ( p1 p2  pr )Z.

As concrete illustrations of this, we have

24Z  6Z (since 24  23  31 )

      and 100Z  10Z (since 100  22  52 ).

Corollary 10.4.3. The nil radical of {0} in any commutative ring R is pre-
cisely the set of nilpotents in R; that is,

{0}  {x  R : x n  0 for some n  Z}.

{0} is usually denoted by N(R) and is called the prime radical of R.

Corollary 10.4.4. For any ideal I of a commutative ring R,

 R I
N    .
 I  I

Worked Exercise 10.4.1. Prove that the following are equivalent to each other
for any ideal I of a commutative ring R.
1. I 5 I
2. I 5 J for some ideal J of R.
3. I is the intersection of a class of prime ideals of R.

Answer: (1) ⇒ (2) is trivial.

(2) ⇒ (3) follows from the fact that J is equal to the intersection of prime
ideals of R containing J (by Theorem 10.4.4).
(3) ⇒ (1): Suppose that {Pa}a[D is a class of prime ideals of R such that
I  ∩ Pa , we always have I ⊆ I . On the other hand,
a∆

Q001-Algebra-111001_CH 10.indd 39 9/21/2011 4:54:11 PM

 10-40   Algebra – Abstract and Modern

x  I ⇒ x n  I  ∩ pa for some n  Z
a

⇒ x n  pa for all a  
⇒ x  pa for all a   (since pa is prime)
⇒ x  ∩ pa  I
a

and therefore I ⊆ I . Thus, I  I .

Worked Exercise 10.4.2. Let I be an ideal of a commutative ring R. Prove that

I 5 I if and only if the quotient ring R/I has no nonzero nilpotent ­elements.

Answer: Suppose that I  I . Let a 1 I be a nilpotent element in R/I. Then,

(a 1 I)n 5 I for some n [ Z1 and hence an 1 I 5 I. This implies that an [ I
and hence a [ I 5 I, so that a 1 I 5 I. Therefore, zero element is the only
nilpotent in R/I.
Conversely suppose that R/I has no nonzero nilpotents. Then,

a  I ⇒ a n  I for some n  Z
⇒ ( a  I ) n  a n  I  I , n  Z
R
⇒ a  I is a nilpotent in
I
⇒ a I  I
⇒ a  I.

Therefore, I ⊆I⊆ I and hence I 5 I.

Worked Exercise 10.4.3. Let P and Q be prime ideals of a ring R. Prove that
P ∩ Q is a prime ideal if and only if P ⊆ Q or Q ⊆ P.

Answer: If P ⊆ Q or Q ⊆ P, then P ∩ Q 5 P or Q and hence P ∩ Q is a prime

ideal, conversely suppose that P ∩ Q is a prime ideal and P  Q. Choose an
element a [ P such that a  Q. Then,

b [ Q ⇒ ab [ P ∩ Q
⇒ a [ P ∩ Q or b [ P ∩ Q
⇒ b [ P ∩ Q (since a  Q)
⇒b[P

and hence Q ⊆ P.

Q001-Algebra-111001_CH 10.indd 40 9/21/2011 4:54:12 PM

 Ideals and Quotient Rings   10-41

Worked Exercise 10.4.4. Let R and S be commutative rings with unities and
f : R → S be an epimorphism of rings. Prove that S is an integral domain if
and only if ker f is a prime ideal of R.

Answer: By the fundamental theorem of homomorphisms (Theorem 10.2.4),

R/ker f  S. Note that S is nontrivial ⇔ ker f is a proper ideal and therefore, S
is an integral domain if and only if ker f is a prime ideal of R.

Exercise 10(d)

1. Determine all the prime ideals of each of the following rings

(i) Z
(ii) Z180
(iii) R
(iv) Q
(v) R3Q
(vi) Q3Z
(vii) Z3Z
(viii) Z31
(ix) M2(R)
(x) R3R3R
(xi) M2(Z)
(xii) (P (X), 1, ∩).

2. Let {Pa}a[D be a class of prime ideals of a ring such that, for any a and b [ D,
there is g [ D such that Pg ⊆ Pa and Pg ⊆ Pb. Then prove that ∩ Pa is a prime
a∆
ideal of R.

3. Let {Pa}a[D be class of prime ideals of a ring R with unity such that, for any a and
b [ D, there is g [ D such that Pa ⊆ Pg and Pb ⊆ Pg. Then prove that ∪ Pa is a
a∆
prime ideal of R.

4. Let {Pa}a[D be a chain of prime ideals of a ring R with unity (that is, for any a
and b [ D, Pa ⊆ Pb or Pb ⊆ Pa). Then prove that ∩ Pa and ∪ Pa are prime
a∆ a∆
ideals of R.

5. Let I and J be ideals of a commutative ring R. Then prove the following.

(i) IJ  I ∩ J  I ∩ J
(ii) I  J ⊆ I J  I J
(iii) I ⊆ J if I ⊆ J for some n  Z.
n

(iv) I I

Q001-Algebra-111001_CH 10.indd 41 9/21/2011 4:54:13 PM

 10-42   Algebra – Abstract and Modern

(v) I n  I for all n  Z.

(vi) For any prime ideal P of R, I ⊆ P ⇔ I ⊆ P.
(vii) If I is prime, then I 5 I .
(viii) I5 I if and only if a2 [ I implies a [ I for any a [ R.
6. A proper ideal I of a commutative ring R is called primary if, for any a and b [ R,
ab [ I ⇒ a [ I or bn [ I for some n [ Z1.
Prove that every prime ideal of R is primary. Give an example to prove that the
converse fails.
7. Prove that a nonzero ideal I of the ring Z of integers is primary if and only if
I 5 pnZ for some prime number p and positive integer n.
8. Let I be a primary ideal of a commutative ring R. Prove that I is the smallest
prime ideal of R containing I.
9. Prove that an ideal I of a commutative ring R is primary if and only if every zero
divisor of the quotient ring R/I is nilpotent.
10. For any prime number p and any positive integer n, prove that every zero divisor
in the ring Z n is nilpotent.
p

11. For any ideal I of a ring R, prove that P  P / I is a one-to-one correspondence

between the primary ideals of R containing I and the primary ideals of R/I.
12. A commutative ring R is called regular if, for any a [ R, there exists b [ R such
that aba 5 a. Prove that a commutative ring R is regular if and only if I 5 I
for all ideals I of R.
13. Prove that the prime radical N(R) of a ring R is {0} if and only if, for any ideals
I and J of R,
IJ 5 {0} ⇔ I ∩ J 5 {0}.
14. Prove that the following are equivalent to each other for any commutative ring
R, with unity.
(i) R has a unique prime ideal.
(ii) Every nonunit in R is nilpotent.
(iii) {0} is a primary ideal and every nonunit nonzero element of R is a zero
divisor.
15. Let R be the ring of all mappings of the real number system R into itself under
point-wise addition and multiplication. Let
I 5 {f [ R : f (1) 5 0 5 f (21)}.
Prove that I is an ideal of R. Is I a prime ideal of R?
16. Let R be a commutative ring and char(R) 5 n > 0. If N(R) 5 {0}, then prove that
n is square free (that is, a2 does not divide n for any integer n > 1).

Q001-Algebra-111001_CH 10.indd 42 9/21/2011 4:54:14 PM

 Ideals and Quotient Rings   10-43

10.5 MAXIMAL IDEALS

In this section, we discuss another special type of ideals of rings, namely
maximal ideals. Clearly, for any ring R, the whole ring R is the largest ideal of
R. We search for ideals of R which are maximal among all proper ideals.

Definition 10.5.1. Let R be a nontrivial ring. A proper ideal M of R is called

a maximal ideal if M is not properly contained in any proper ideal of R; that
is, for any ideal I of R,

M⊆I⊆R⇒M5I or I 5 R.

Example 10.5.1

1. Let us recall that any ideal of Z is of the form nZ for some nonnegative
integer n and that nZ ⊆ mZ if and only if m divides n. Therefore, nZ is
a maximal ideal of Z if and only if n is a prime number.
2. In the ring R of real numbers, {0} is a maximal ideal. In fact, {0} is a
maximal ideal in any field, since a field has two ideals, namely {0} and
the whole field.
3. Let R 5 R 3 R under co-ordinate wise addition and multiplication.
Then, R 3 {0} and {0} 3 R are maximal ideals of R.
4. Consider the ring Mn(R) of n 3 n matrices over the real number system
R. Then, {0} is a maximal ideal of Mn(R), since, by Theorem 10.1.10,
Mn(R) is the only nonzero ideal of Mn(R).

Let us recall that a commutative ring R with unity is a field if and only if R
has exactly two ideals, namely {0} and R. This, together with the fact that the
ideals of R/I are in one-to-one correspondence with the ideals of R containing
I, imply the following important result. However, we prefer to give an inde-
pendent proof in view of its technicality.

Theorem 10.5.1. Let M be an ideal of a commutative ring R with unity. Then,

M is a maximal ideal of R if and only it the quotient ring R/M is a field.

Proof: First note that M is a proper ideal of R if and only if R/M is nontrivial.
Suppose that M is a maximal ideal of R. Then, M is a proper ideal and hence
R/M is a nontrivial commutative ring with unity (since so is R). Now, let a 1
M be a nonzero element of R/M. Then, a 1 M  M and hence a  M. Let I
be the ideal defined by

I 5 M 1 <a> 5 {x 1 ra : x  R}.

Q001-Algebra-111001_CH 10.indd 43 9/21/2011 4:54:14 PM

 10-44   Algebra – Abstract and Modern

Then, M  I (since a  I and a  M). By the maximality of M, it follows that

I 5 R. In particular, 1  I 5 x 1 ra for some x  M and r  R.
Now, we have 1 2 ra 5 x  M and hence

(r 1 M)(a 1 M) 5 ra 1 M 5 1 1 M, the unity in R/M.

Therefore, r 1 M is the multiplicative inverse of a 1 M in R/M. Thus, every

nonzero element in R/M is a unit and hence R/M is a field.
Conversely suppose that R/M is a field. Then, R/M is nontrivial and hence
M is a proper ideal of R. Let J be any ideal of R such that M ⊆ I ⊆ R. Suppose
that M  I. Then, there exists a  I such that a  M. Now, a 1 M is a nonzero
element in the field R/M and hence a 1 M is a unit in R/M. Therefore, there
exists b  R such that

ab 1 M 5 (a 1 M)(b 1 M) 5 1 1 M

and hence 1 2 ab  M ⊆ I. Also, since a  I, ab  I and therefore

1 5 (1 2 ab) 1 ab  I

which implies that I 5 R. Thus, M is a maximal ideal of R.

Corollary 10.5.1. A commutative ring R with unity is a field if and only if

{0} is a maximal ideal of R.

Proof: This follows from the fact that R/{0}  R and from Theorem
10.5.3.

Theorem 10.5.2. Let R be a ring with unity. Then, every maximal ideal of R
is a prime ideal and the converse is not true.

Proof: Let M be a maximal ideal of R. Then, M is a proper ideal of R. Let I

and J be ideals of R such that IJ ⊆ M. Suppose that I  M. Choose a  I such
that a  M. Recall that

 n 
a ∑ xi ayi  ar  sa : xi , yi , r , s  R, n  0.
 i1 

Put K 5 M 1 <a>. Then, K is an ideal of R containing M properly, since a 

K and a  M. By the maximality of M, we get that K 5 R. In particular, 1 
K 5 M 1 <a> and hence

Q001-Algebra-111001_CH 10.indd 44 9/21/2011 4:54:14 PM

 Ideals and Quotient Rings   10-45

n
1  x  ∑ xi ayi  ar  sa
i1

for some x  M and xi, yi, r, s  R, n ≥ 0. Now,

 n 
b  J ⇒ b 5 1b 5  x  ∑ xi ayi  ar  sa b
 i1 
n
⇒ b 5 xb 1 ∑ (x a) (y b)  (ar ) b  ( sa) b
i1
i i

⇒ b  M 1 IJ, since x  M, a  I, b  J
⇒ b  M, since IJ ⊆ M and M 1 IJ 5 M.

Therefore, J ⊆ M. Thus, I ⊆ M or J ⊆ M and hence M is a prime ideal. The

converse is not true; that is, an ideal can be prime without being maximal.
For example, {0} is a prime ideal of the ring Z of integers and {0} is not a
maximal ideal.
Note that, in proving the above theorem, the existence of unity in the ring
is essential. For consider the following example.

Example 10.5.2. Consider the ring 2Z of even integers. Let M 5 4Z.

Then, M is a maximal ideal of 2Z; for, let I be an ideal of 2Z such that
M ⊆ I ⊆ 2Z. Suppose that M  I. Then, there exists 2a  I such that 2a
 M 5 4Z. Now, a must be odd and hence a 5 2n 1 1 for some n  Z.
Consider

2 5 2 ? 1 5 2(a 2 2n) 5 2a 2 4n  I

(since 2a  I and 4n  M ⊆ I). This implies that 2Z ⊆ I and hence I 5 2Z.

Thus, M is a maximal ideal of 2Z. However, M is not prime, since

2  2  M and 2  M.

Note that 2Z is a commutative ring without unity.

Theorem 10.5.3. Let M be a proper ideal of a ring R. Then, M is a maximal

ideal if and only if M 1 <a> 5 R for all a  R 2 M.

Proof: Suppose that M is a maximal ideal of R and a  R 2 M. Then, M 1

<a> is an ideal of R containing M properly. Since M is maximal, we get that
M 1 <a> 5 R. Conversely, if M is not maximal, then there exists an ideal

Q001-Algebra-111001_CH 10.indd 45 9/21/2011 4:54:15 PM

 10-46   Algebra – Abstract and Modern

I of R such that M  I  R and therefore there exists a  I 2 M ⊆ R 2 M, so

that M 1 <a> ⊆ I and hence M 1 <a>  R. b

Corollary 10.5.2. Let R be a commutative ring with unity and M be a proper

ideal of R. Then, M is maximal if and only if, for each a  R 2 M, 1 2 ar 
M for some r  R.

Proof: This is a consequence of the above and of the fact that M 1 <a> 5 R
if and only if 1 5 x 1 ar for some x  M and r  R. b
Next, we obtain a general result which assures the existence of suitably
many maximal ideals. The crucial step here is again the Zorn’s Lemma; an
equivalent form of which is given in Zorn’s Lemma 10.4.1.

Theorem 10.5.4. Let R be a ring with unity. Then, any proper ideal of R is
contained in a maximal ideal of R.

Proof: Recall that an ideal of R is proper if and only if it does not contain the
unity of R. Let I be a proper ideal of R. Consider the class

 5 {J : J is a proper ideal of R and I ⊆ J}

since I  ,  is a nonempty class of subsets of R. If {Ja} is a chain in 

(that is, any two members of it are comparable), then by Corollary 10.1.1,
∪ J a is an ideal of R and, since each Ja is a proper ideal of R, 1  Ja for each
a
a  ∆ and hence 1  ∪ J a, so that ∪ J a is a proper ideal of R. Also, clearly
a a
∪ J a is a member of  containing each Ia. Therefore, the hypothesis of the
a
Zorn’s Lemma is satisfied for  and hence  has a maximal member, which
is clearly a maximal ideal containing I. b

Corollary 10.5.3. Let R be a commutative ring with unity and a  R. Then, a

is a unit in R if and only if a does not belong to any maximal ideal of R.

Proof: a is a nonunit in R if and only if <a> (5 aR) is a proper ideal of R and

hence, by the above theorem, <a> is contained in a maximal ideal of R. b

Theorem 10.5.5. Let R be a commutative ring with unity. Suppose that R has
exactly one maximal ideal. Then, 0 and 1 are the only idempotents in R.

Proof: Let M be the unique maximal ideal of R. Let a be an idempotent in R;

that is, a2 5 a  R. Suppose that a  0 and a  1. Then,
a(a 2 1) 5 0

Q001-Algebra-111001_CH 10.indd 46 9/21/2011 4:54:15 PM

 Ideals and Quotient Rings   10-47

and hence a and a 2 1 are zero divisors in R. Therefore, a and a 2 1 are both non-
units and, by Corollary 10.5.3, a  M and a 2 1  M. From this, we get that
1 5 a 2 (a 2 1)  M
which is a contradiction to the fact that M  R. Thus, either a 5 0 or
a 5 1. b
Recall from Corollary 10.4.3, the prime radical of R is defined as the set
of all nilpotents in R and, by Theorem 10.4.4, it is precisely the intersection of
all prime ideals of R. In the following, we introduce another type of radical,
which plays an important role in the structure theory of commutative rings.

Definition 10.5.2. Let R be a nontrivial commutative ring with unity. The

intersection of all maximal ideals of R is called the Jacobson radical of R and
is denoted by J(R) or by Rad(R). R is said to be semisimple if J(R) 5 {0}.
The Jacobson radical always exists, since any nontrivial commutative ring
with unity has atleast one maximal by ideal, by Theorem 10.5.4. Also, from
the definition of J(R), it is immediate that the Jacobson radical of R is an ideal
contained in each maximal ideal of R. Two important examples of semisimple
rings are given below.

Example 10.5.3
1. Recall that the maximal ideals of the ring Z are precisely of the form pZ
for some prime number p. Now, the Jacobson radical of Z is given by
J (Z)  ∩ pZ  {0},
pP

where P is the set of prime numbers. Therefore, Z is a semisimple ring.

2. Let X be a nonempty set and F be a field. Then, the set FX of all map-
pings of X into F forms a commutative ring with unity under the
point-wise operations, with reference to the ring operations in F. For
each x  X, let
Mx 5 {f  FX : f (x) 5 0}
and define ax : FX → F by ax(f ) 5 f (x). It can be easily verified that ax is
an epimorphism of rings and ker ax 5 Mx. By the fundamental theorem
of homomorphisms, F X / M x  F . Since F is a field, it follows from
Theorem 10.5.1 that Mx is a maximal ideal of FX. Also,

∩M
x X
x  { f  F X : f ( x )  0 for all x  X }  {0}

and therefore the Jacobson radical of FX is {0}. Thus, FX is a semisimple

ring.

Q001-Algebra-111001_CH 10.indd 47 9/21/2011 4:54:16 PM

 10-48   Algebra – Abstract and Modern

In the following, we obtain a basic connection between the Jacobson radi-

cal and multiplicative invertibility of the elements of the ring.

Theorem 10.5.6. Let I be an ideal of a commutative ring R with unity. Then,

I ⊆ J(R) if and only if each element of the coset 1 1 I is a unit in R.

Proof: Suppose that I ⊆ J(R). Let 1 1 a  1 1 I, a  I. If 1 1 a is a nonunit,

then by Corollary 10.5.3, there exists a maximal ideal M of R such that 1 1
a  M. Then, since a  I ⊆ J(R) ⊆ M, we get that

1 5 (1 1 a) 2 a  M

which is a contradiction to the fact that M is a proper ideal. Thus, 1 1 a is a

unit for any a  I.
Conversely suppose that 1 1 a is a unit in R for each a  I. Suppose, if
possible, that I  J(R). We can choose an element a  I such that a  J(R).
Then, there exists a maximal ideal M such that a  M. Then, M is properly
contained in M 1 aR. By the maximality of M, we have M 1 aR 5 R. In
particular, 1  R 5 M 1 aR and hence

1 5 x 1 ar for some x  M and r  R.

Now, since a  I, 2 ar  I and 1 2 ar  1 1 I. Therefore, x (5 1 2 ar) is a

unit in R, which is a contradiction since x  M and M is a proper ideal. Thus,
I ⊆ J(R).
The above theorem gives us a characterization of elements of the Jacobson
radical J(R), if we replace I above a principal ideal <a>.

Theorem 10.5.7. Let R be a commutative ring with unity. Then, the Jacobson
radical J(R) is given by

J(R) 5 {a  R : 1 1 ar is a unit in R for all r  R}.

Proof: If a  J(R), then <a> ⊆ J(R) and hence every element of 1 1 <a>
is a unit in R; that is, 1 1 ar is a unit for all r  R. Conversely suppose that
1 1 ar is a unit for all r  R. Then, every element of 1 1 <a> is a unit and
therefore, by the above theorem, <a> ⊆ J(R) which is equivalent to saying
that a  J(R).

Corollary 10.5.4. The following holds for any commutative ring R with
unity.

Q001-Algebra-111001_CH 10.indd 48 9/21/2011 4:54:16 PM

 Ideals and Quotient Rings   10-49

1. 0 is the only idempotent in the Jacobson radical J(R).

2. An element a  R is a unit in R if and only if the coset a 1 J(R) is a unit
in the quotient ring R/J(R).
3. The prime radical N(R) is contained in the Jacobson radical J(R).

Proof:
1. Let a  J(R) be an idempotent. Then, by Theorem 10.5.7, 1 2 a is a unit
in R and hence there exists b  R such that (1 2 a)b 5 1. Now,
a 5 a1 5 a(1 2 a)b 5 (a 2 a2)b 5 0 (since a2 5 a)
Thus, 0 is the only idempotent in J(R).
2. Let a  R. Suppose that a 1 J(R) is a unit in R/J(R). Then, there exists
b  R such that
(a 1 J(R))(b 1 J(R)) 5 1 1 J(R)
and therefore 1 2 ab  J(R). By Theorem 10.5.7, 1 2 (1 2 ab) is a unit
in R and therefore ab is a unit. Thus, a is a unit in R. The converse is
trivial.
3. a  N(R) ⇒ a belongs to every prime ideal of R
⇒ a belongs to every maximal ideal of R (since every maximal
ideal is prime)
⇒ a  J(R).
Thus, N(R) ⊆ J(R). b

Theorem 10.5.8. Let R be a commutative ring with unity and J(R) be the
Jacobson radical of R. Then, the quotient ring R/J(R) is semisimple.

Proof: We have to prove that the Jacobson radical of R/J(R) is trivial. Let a 1
J(R) be an element in the Jacobson radical of R/J(R). Then, by Theorem
15.5.15,

1 1 J(R) 1 (a 1 J(R))(r 1 J(R))(5(1 1 ar) 1 J(R))

is a unit in R/J(R) for all r 1 J(R)  R/J(R) and therefore, by Corollary

10.5.4 (2), 1 1 ar is a unit in R for all r  R. This implies that a  J(R) and
hence a 1 J(R) is the zero in R/J(R). Thus, J(R/J(R)) is trivial and hence
R/J(R) is semisimple. b

Worked Exercise 10.5.1. Let n  Z1 and R 5 R 3 R 3  3 R (n factors).

Then, R is a commutative ring with unity under the co-ordinate wise opera-
tions. Determine all maximal ideals of the ring R.

Q001-Algebra-111001_CH 10.indd 49 9/21/2011 4:54:16 PM

 10-50   Algebra – Abstract and Modern

Answer: Let any element of R be denoted by

a 5 (a1, a2, …, an)

where a1, a2, …, an are real numbers. For each 1  i  n, let

Mi 5 {a  R : ai 5 0}.

Let pi : R → R be the ith projection; that is,

pi(a) 5 ai for all a  R.

Then, pi is an epimorphism of rings and ker pi 5 Mi. Therefore, R M i  R.

Since R is a field, so is R/Mi and hence Mi is a maximal ideal of R. We shall
prove that M1, M2, …, Mn are the only maximal ideals of R.
Let M be any maximal ideal of R. We shall prove that M # Mi for some
1 # i # n and hence M 5 Mi. On the contrary, suppose that M  Mi for all
i. Then, we can choose, for each 1 # i # n, xi  M such that xi  Mi. Then,
xii  0 for each 1 # i # n. Put

a 5 (x1)2 1 (x2)2 1  1 (xn)2.

Then, a  M, since xi  M for all 1 # i # n. Since the ith co-ordinate of a is

ai  (xi1 ) 2  (xi2 ) 2   (xin ) 2 .

Since xii  0, we get that (xii ) 2  0 and hence

ai > 0 for all 1 # i # n.

Now,

1 1 1
(1, 1, …, 1)  ( a1 , a2 , …, an )  , , …, 
 a1 a2 an 

and hence a is a unit in R. This is a contradiction to the fact that a  M and

M is a maximal ideal. Thus, M ⊆ Mi and, by the maximality of M, M 5 Mi for
some 1 # i # n. Thus, M1, M2, …, Mn are all the maximal ideals of R.

Worked Exercise 10.5.2. Let I be a proper ideal of a Boolean ring B (see

Definition 9.2.2). Prove that the following are equivalent to each other.

Q001-Algebra-111001_CH 10.indd 50 9/21/2011 4:54:17 PM

 Ideals and Quotient Rings   10-51

1. I is a prime ideal of B.
2. I is a maximal ideal of B.
3. For any a  B, either a  I or b 2 ab  I for all b  B, but not both.

Answer: (1) ⇒ (2): Suppose that I is a prime ideal of B. Let J be an ideal of

B containing I properly. Choose a  J such that a  I. Then,

a(1 2 a) 5 a 2 a2 5 0  I.

Since I is a prime ideal and a  I, we get that 1 2 a  I # J. Now, since

a  J, we have

1 5 a 1 (1 2 a)  J

and hence J 5 B.
(2) ⇒ (3): Suppose that I is a maximal ideal of B and a  B. Suppose that a
 I. Then, I 1 <a> 5 B and hence, for any b  B,

b 5 x 1 ar for some x  I and r  B

     and ab 5 ax 1 a2r 5 ax 1 ar
and hence b 2 ab 5 x 2 ax  I.

If both a  I and b 2 ab  I for all b  B, then

b 5 (b 2 ab) 1 ab  I for all b  B

and hence I 5 B, which is a contradiction. Thus, either a  I or b 2 ab  I

for all b  B and not both.
(3) ⇒ (1): Suppose (3) is satisfied. Let a and b  B such that ab  I. If a 
I, then b 2 ab  I (by (3)) and hence

b 5 (b 2 ab) 1 ab  I.

Therefore, a  I or b  I. Already we are given that I is a proper ideal of B.

Thus, I is a prime ideal of B.

Worked Exercise 10.5.3. Consider the ring Z[i] of Gaussian integers. Let

I 5 {a 1 bi  Z[i] : a and b are both even}

Prove that I is an ideal of Z[i], which is not a maximal ideal.

Q001-Algebra-111001_CH 10.indd 51 9/21/2011 4:54:17 PM

 10-52   Algebra – Abstract and Modern

Answer: If x 5 a 1 bi and y 5 c 1 di  I, then a, b, c and d are even and

hence a 2 c and b 2 d are even and therefore x 2 y 5 (a 2 c) 1 (b 2 d)i 
I. Also, if x 5 a 1 bi  I and z 5 s 1 ti  Z[i], then

xz 5 (as 2 bt) 1 (at 1 bs)i  I

since a and b are even and hence as, bt, at and bs are all even. Thus, I is an
ideal of Z[i]. Note that Z[i] is a commutative ring with unity. We shall prove
that I is not a maximal ideal of Z[i]. Let

J 5 {a 1 bi  Z[i] : a2 1 b2 is even}.

Observe that a2 1 b2 is even if and only if either both a and b are even or both
a and b are odd. We verify that J is an ideal of Z[i].
Let x 5 a 1 bi and y 5 c 1 di  J. Then, a2 1 b2 and c2 1 d2 are even and

x 2 y 5 (a 2 c) 1 (b 2 d)i  J

since (a 2 c)2 1 (b 2 d)2 5 (a2 1 b2) 1 (c2 1 d2) 2 2(ac 1 bd), which is
even. Also, for any z 5 s 1 ti  Z[i],

xz 5 (a 1 bi)(s 1 ti) 5 (as 2 bt) 1 (at 1 bs)i

and (as 2 bt)2 1 (at 1 bs)2 5 (a2 1 b2)s2 1 (a2 1 b2)t2

which is even, since a2 1 b2 is even. Therefore, xz  J. Thus, J is an ideal of
Z[i]. Further,

I ⊂ J ⊂[i ],
 

since 1 1 i  J 2 I and 1 1 2i  J. Thus, I is not a maximal ideal of Z[i].

Worked Exercise 10.5.4. Let I 5 {a 1 bi  Z[i]: 3 divides both a and b}.

Prove that I is a maximal ideal of the ring Z[i] of Gaussian integers.

Answer: Clearly 3 1 3i  I and hence I is a nonempty subset of Z[i]. Let x 5

a 1 bi and y 5 c 1 id  I. Then, 3 divides a, b, c and d and hence 3 divides
a 2 c and b 2 d, so that

x 2 y 5 (a 2 c) 1 (b 2 d)i  I

Also, for any z 5 s 1 ti  Z[i],

xz 5 (a 1 bi)(s 1 ti)
        5 (as 2 bt) 1 (at 1 bs)i

Q001-Algebra-111001_CH 10.indd 52 9/21/2011 4:54:18 PM

 Ideals and Quotient Rings   10-53

which belongs to I, since 3 divides a and b and hence 3 divides as 2 bt and at 1

bs. Thus, I is an ideal of Z[i]. To prove that I is maximal, let J be any ideal of
Z[i] such that I  J ⊆ Z[i ]. Choose a 1 bi  J such that a 1 bi  I. Then, 3
does not divide a or b or both. We shall distinguish these cases separately and
prove that 1  J in each case.
1. Suppose that 3 divides a and 3 does not divide b.
Then, a 5 a10i  I # J and hence
bi 5 (a 1 bi) 2 a  J.
Now, b 5 (bi)(2bi)  J, since J is an ideal. Since 3 does not divide b,
2

g.c.d. (3, b2) 5 1 and therefore there exist c and d  Z such that 3c 1
b2d 5 1. Now, b2d  J and 3c  I # J and hence
1 5 3c 1 b2d  J
which implies that J 5 Z[i].
2. Suppose that 3 does not divide a and 3 divides b. In this case, using the
technique of (1) above we can prove that J 5 Z[i].
3. Suppose that 3 divides neither a nor b. Then, a 5 3k 1 1 or 3k 1 2 for
some k  Z and b 5 3s 1 1 or 3s 1 2 for some s  Z. Then,
a2 5 3(3k2 1 2k) 1 1 or 3(3k2 1 4k 1 1) 1 1
  and b2 5 3(3s2 1 2s) 1 1 or 3(3s2 1 4s 1 1) 1 1
which imply that a2 1 b2 5 3t 1 2 for some t  Z and hence 3 does not
divide a2 1 b2. Put c 5 a2 1 b2. Then, g.c.d.(3, c) 5 1 and hence there
exist integers n and m such that 3m 1 cn 5 1. Now,
3  I # J and c 5 a2 1 b2 5 (a 1 bi)(a 2 bi)  J
since a 1 bi  J. Therefore, 3 and c  J and hence
1 5 3m 1 cn  J
which implies that J 5 Z[i].
Thus, in all cases, we have proved that J 5 Z[i]. Therefore, I is a maximal
ideal of Z[i].

Worked Exercise 10.5.5. Let M and N be two distinct maximal ideals of a

commutative ring R with unity. Then prove that MN 5 M ∩ N.

Answer: Since M and N are distinct maximal ideals, M  N and N  M and

hence M 1 N  M and M 1 N  N. In particular, M1N is an ideal of R con-
taining M properly. By the maximality of M, it follows that M 1 N 5 R. In
particular, 1  R 5 M 1 N and hence
1 5 a 1 b for some a  M and b  N.

Q001-Algebra-111001_CH 10.indd 53 9/21/2011 4:54:18 PM

 10-54   Algebra – Abstract and Modern

Now, clearly MN # M ∩ N. Also,

x  M ∩ N ⇒ x 5 x1 5 xa 1 xb  MN

since x  N, a  M, x  M and b  N. Therefore, M ∩ N # MN. Thus, MN 5

M ∩ N.

EXERCISES 10(e)

1. Determine all maximal ideals and the Jacobson radicals in each of the following
rings.
(i) R
(ii) Q
(iii) Z
(iv) Z120
(v) Z3Z
(vi) Q3Z
(vii) M3(R)
(viii) Zn, n [ Z1.

2. Give an example of a maximal ideal in the ring 3Z which is not a prime.

3. Let I be a proper ideal of a ring R. Prove that M  M / I is a one-to-one corre-

spondence between the maximal ideals of R containing I and the maximal ideals
of R/I.

4. Let n [ Z1. Prove that an ideal of Zn is prime if and only if it is maximal.

5. Let R be a finite commutative ring with unity. Prove that an ideal of R is prime if
and only if it is maximal.

6. Let X be a nonempty finite set with n elements. Prove that there are exactly n
maximal ideals in the ring (P(X), 1, ∩).

7. Let #(X, R) be the set of all real valued continuous functions defined on a topo-
logical space X. Then prove that #(X, R) is a commutative ring with unity under
the point-wise addition and multiplication.

8. In the ring #(X, R) given in 7 above, prove that the set

Mx 5 {f [ #(X, R) : f (x) 5 0}
is a maximal ideal for each x [ X.

9. Let X be a Compact Hausdorff space and #(X, R) be the ring given in 7 above.
Prove that x  M x is a bijection of X onto the set of maximal ideals of #(X, R),
where Mx is the set given in 8 above.

Q001-Algebra-111001_CH 10.indd 54 9/21/2011 4:54:19 PM

 Ideals and Quotient Rings   10-55

10. Let Z[i] be the ring of Gaussian integers. Prove that the set
I 5 {a 1 bi [ Z[i] : a 2 b is even}.
is a maximal ideal of Z[i] and find the number of elements of the quotient ring
Z[i]/I.

11. Let Z[ 2]  {a  b 2 : a and b  Z}. Prove that Z[ 2] is a commutative ring

with unity under the usual addition and multiplication of real numbers. Prove
that the set

I  {a  b 2  Z 2 : 5 divides both a and b}

is a maximal ideal of Z[ 2] and find the number of elements in Z[ 2] I .

12. Prove that the set

I 5 {a 1 bi [ Z[i] : 5 divides both a and b}

is an ideal of Z[i] and is not maximal. Is this a prime ideal? Estimate the number
of elements in the quotient ring Z[i ] / I .

13. Prove that a proper ideal M of a ring R is maximal if and only if, for any ideal I
of R, either I ⊆ M or I 1 M 5 R.

14. Let f : R → S be an epimorphism of rings.

(i) If M is a maximal (prime) ideal of R containing ker f, prove that f (M) is
a maximal (prime) ideal of S.
(ii) If M9 is a maximal (prime) ideal of S, then prove that f21(M9) is a maxi-
mal (prime) ideal of R.
(iii) Prove that M  f ( M ) is a one-to-one correspondence between the
maximal (prime) ideals of R containing ker f and the maximal (prime)
ideals of S.
15. A nonzero ideal I of a ring R is called minimal if there is no ideal properly in
between {0} and I. Prove that a nonzero ideal I of R is minimal if and only if I 5
<a> for all 0  a [ I. Show that the ring Z of integers has no minimal ideals.
16. Let P be a prime ideal of a commutative ring R such that the quotient ring R/P is
finite. Then prove that P is a maximal ideal.
17. Let R be a commutative ring with unity such that, for each a [ R, there exists an
integer n > 1 such that an 5 a. Prove that an ideal of R is prime if and only if it
is maximal.
18. Let M be a maximal ideal of a commutative ring R with unity and n [ Z1. Prove
that R/Mn has exactly one prime ideal.
19. Prove that the following are equivalent to each other for any commutative ring
with unity.

Q001-Algebra-111001_CH 10.indd 55 9/21/2011 4:54:19 PM

 10-56   Algebra – Abstract and Modern

(i) R has a unique prime ideal.

(ii) R has a unique maximal ideal and the Jacobson radical of R is equal to
the prime radical of R.
(iii) Every nonunit in R is nilpotent.

20. Prove the following for any commutative ring R with unity.
(i) R is semisimple if and only if, for each a [ R, 1 2 ra is a nonunit for
some r [ R.
(ii) If R is regular, then it is semisimple.
(iii) If I is an ideal of R such that R/I is semisimple, then the Jacobson radical
of R is contained in I.

10.6 Embeddings of Rings

Rings without unity lack certain important properties. However, we shall
prove in this section that any ring can be treated as a subring of a ring with
unity. It is well known that any subring with unity of a field is an integral
domain. We prove a converse of this, in the sense that any integral domain can
be treated as a subring of a field. First, let us have the following definition.

Definition 10.6.1. A ring R is said to be embedded in a ring S if R is isomor-

phic to a subring of S.
It can be easily proved that R is embedded in S if and only if there is a
monomorphism of R into S. In the following, we prove that any ring can be
embedded in a ring with unity.

Theorem 10.6.1. Let R be any ring. Then, there exists a ring S with unity
satisfying the following properties:
1. R is embedded in S.
2. R is isomorphic to an ideal of S.
3. R is commutative if and only if S is commutative.
4. char(R) 5 char(S).

Proof: First, we assume that R is of characteristic zero. Let

S 5 Z 3 R.

For any (m, a) and (n, b) [ S, define

(m, a) 1 (n, b) 5 (m 1 n, a 1 b)
       and (m, a) ? (n, b) 5 (mn, mb 1 na 1 ab).

Q001-Algebra-111001_CH 10.indd 56 9/21/2011 4:54:20 PM

 Ideals and Quotient Rings   10-57

Then, 1 and ? are binary operations on S. Since 1 is precisely co-ordinate

wise addition and (Z, 1) and (R, 1) are abelian groups, it follows that (S, 1)
is also an abelian group in which (0, 0) is the identity, where the first 0 is the
integer 0 and the second 0 is the zero element in the ring R. For any (m, a), (n,
b) and (r, c) in S, we have

((m, a) ? (n, b)) ? (r, c) 5 (mn, mb 1 na 1 ab) ? (r, c)

5 ((mn)r, mnc 1 r(mb 1 na 1 ab) 1
(mb 1 na 1 ab)c)
5 (m(nr), m(nc 1 rb 1 bc) 1 nra 1
a(nc 1 rb 1 bc))
5 (m, a) ? ((n, b) ? (r, c)).

Therefore, ? is associative on S. Also,

(m, a) ? ((n, b) 1 (r, c)) 5 (m, a) ? (n 1 r, b 1 c)

5 (m(n 1 r), m(b 1 c) 1 (n 1 r)a 1 a(b 1 c))
5 (mn 1 mr, (mb 1 na 1 ab) 1 (mc 1 ra 1 ac))
5 (mn, mb 1 na 1 ab) 1 (mr, mc 1 ra 1 ac)
5 (m, a) ? (n, b) 1 (m, a) ? (r, c).

Similarly, we can prove the other distributive law.

Thus, (S, 1, ?) is a ring. Consider the element (1, 0) in S. For any (m, a)  S,
we have

(m, a) ? (1, 0) 5 (m1, m0 1 1a 1 a0) 5 (m, a)

     and (1, 0) ? (m, a) 5 (1m, 1a 1 m0 1 0a) 5 (m, a)

and therefore (1, 0) is the unity (multiplicative identity) in S. Thus, S is a

ring with unity. We shall prove that this ring S satisfies all the required three
properties.
1. Define f : R → S by f (a) 5 (0, a) for any a  R. It can be easily verified
that f is a monomorphism of rings. Therefore, R is embedded in S.
2. Put I 5 f (R) 5 {(0, a) : a  R}. Then, for any (0, a)  I and (n, b)  S,
we have
(0, a) ? (n, b) 5 (0n, 0b 1 na 1 ab) (0, na 1 ab)  I
and (n, b) ? (0, a) 5 (n0, na 1 0b 1 ba) 5 (0, na 1 ab)  I.

Also, (0, a) 1 (0, b) 5 (0, a 1 b)  I.

Q001-Algebra-111001_CH 10.indd 57 9/21/2011 4:54:20 PM

 10-58   Algebra – Abstract and Modern

Thus, I is an ideal of S. Clearly the map f defined in (1) above is an iso-

morphism of R onto f (R) 5 I.
3. If S is commutative, then f (R), being a subring of S, is commutative
and hence R is commutative. Conversely suppose that R is commutative.
Then, for any (m, a) and (n, b) in S, we have
(m, a) ? (n, b) 5 (mn, mb 1 na 1 ab)
5 (nm, na 1 mb 1 ba)
5 (n, b) ? (m, a)
and hence S is commutative.
4. Since Z is of characteristic zero, so is S and hence char(R) 5 0 5
char(S).
Next, we assume that char(R) 5 n . 0. In this case, the above construction of
S is of no use, since Z is of characteristic zero and hence so is S, irrespective
of whether char(R) is zero or not. Therefore, we slightly change the construc-
tion at S. Let

S 5 Zn 3 R,

where Zn is the ring of integers modulo n. Define addition and multiplication

in S as follows

(i, a) 1 (j, b) 5 (i 1n j, a 1 b)
        and (i, a) (j, b) 5 (i n j, ib 1 ja 1 ab).

Then, similar to the above case, we can easily prove that (S, 1, ?) is a ring
with unity satisfying the properties (1) through (4). Note that char(S) 5 n 5
char(R).
We have proved earlier that any field is an integral domain and that any
finite integral domain is a field. Also, any subring (with unity) of a field is
an integral domain. In the following theorem, we prove that integral domains
arise as subrings of fields only.

Theorem 10.6.2. Any integral domain can be embedded in a field.

Proof: Let R be an integral domain; that is, R is a nontrivial commutative

ring with unity in which product of any two nonzero elements is again non-
zero. Let

S 5 {(a, b) : a and b  R and b  0}

and define a binary relation U on S by

(a, b) U (c, d) ⇔ ad 5 bc.

Q001-Algebra-111001_CH 10.indd 58 9/21/2011 4:54:20 PM

 Ideals and Quotient Rings   10-59

We shall verify that U is an equivalence relation on S. Since R is commuta-

tive, ab 5 ba and hence (a, b) U (a, b) for any (a, b)  S. Therefore, U is
reflexive. Also,

(a, b) U (c, d) ⇒ ad 5 bc ⇒ cb 5 da ⇒ (c, d) U (a, b)

and therefore U is symmetric. Further,

(a, b) U (c, d) and (c, d) U (s, t) ⇒ ad 5 bc and ct 5 ds

⇒ adt 5 bct 5 bds
⇒ (at)d 5 (bs)d
⇒ at 5 bs since d  0
⇒ (a, b) U (s, t).

Therefore, U is transitive also. Thus, U is an equivalence relation on S. For

any (a, b)  S, let

[a, b] 5 The equivalence class containing (a, b)

5{(c, d)  S : (a, b) U (c, d)}

Recall that [a, b] 5 [c, d] ⇔ (a, b) U (c, d) ⇔ ad 5 bc. Let

F 5 {[a, b] : (a, b)  S}

we shall define addition and multiplication on F by

[a, b] 1 [c, d] 5 [ad 1 bc, bd]

        and [a, b]  [c, d] 5 [ac, bd].

First note that, since R is an integral domain and b  0 and d  0, we have

bd  0 and hence the above definitions of 1 and  make sense. Next, we have
to prove that 1 and ? are well defined, in the sense that they depend on the
classes [a, b] and [c, d], but not on the representative elements a, b, c and d.
Suppose that

[a, b] 5 [a′, b′] and [c, d] 5 [c′, d′].

Then, ab′ 5 ba′ and cd′ 5 dc′ and hence

(ad 1 bc)b′d′ 5 adb′d′ 1 bcb′d′

5 (ab′)dd′ 1 (cd′)bb′
5 (ba′)dd′ 1 (dc′)bb′
5 bd(a′d′ 1 b′c′)

Q001-Algebra-111001_CH 10.indd 59 9/21/2011 4:54:20 PM

 10-60   Algebra – Abstract and Modern

and hence [ad 1 bc, bd] 5 [a′d′ 1 b′c′, b′d′]. Thus, 1 is well defined. Simi-
larly, we can prove that ? is well defined. In the following, let [a, b], [c, d] and
[s, t] be arbitrary elements of F.

([a, b] 1 [c, d]) 1 [s, t] 5 [ad 1 bc, bd] 1 [s, t]

5 [(ad 1 bc)t 1 bds, bdt]
5 [a(dt) 1 b(ct 1 ds), bdt]
5 [a, b] 1 ([c, d] 1 [s, t]).

Therefore, 1 is associative.

[a, b] 1 [c, d] 5 [ad 1 bc, bd]

5 [cb 1 da, db]
5 [c, d] 1 [a, b].

Therefore, 1 is commutative.

[a, b] 1 [0, 1] 5 [a1 1 b0, b1] 5 [a, b]

Therefore, [0, 1] is the additive identity in F

[a, b] 1 [2a, b] 5 [ab 1 b(2a), b2] 5 [0, b2] 5 [0, 1].

Therefore, [2a, b] is the additive inverse of [a, b] in F. Thus, (F, 1) is an

abelian group. One can easily verify that ? is associative and commutative.
Also,

[a, b] ? [1, 1] 5 [a1, b1] 5 [a, b]

and therefore [1, 1] is the multiplicative identity in F.

[a, b] ? ([c, d] 1 [s, t]) 5 [a, b] ? [ct 1 ds, dt]

5 [a(ct 1 ds), bdt]
5 [b(act 1 ads), b(bdt)]
5 [acbt, 1 bdas, bdbt]
5 [ac, bd] 1 [as, bt]
5 [a, b] ? [c, d] 1 [a, b] ? [s, t]

Therefore, ? distributes over 1 . Thus, (F, 1, ?) is a commutative ring

with unity. Also, since R is nontrivial, 1  0 in R and [1, 1]  [0, 1] in F.

Q001-Algebra-111001_CH 10.indd 60 9/21/2011 4:54:20 PM

 Ideals and Quotient Rings   10-61

­Therefore, F is nontrivial. Also for any [a, b]  [0, 1], we have a  0 and
hence [b, a]  F and

[a, b] ? [b, a] 5 [ab, ba] 5 [1, 1].

Therefore, [b, a] is the multiplicative inverse of [a, b]. Thus, (F, 1, ?) is

a field.
Now, define f : R → F by f (a) 5 [a, 1] for all a  R. One can easily verify that
f is a monomorphism of rings. Thus, R is embedded in the field F.

Definition 10.6.2. For any integral domain R, the field F constructed above
is called the field of quotients of R.

Example 10.6.1. The ring Z of integers is an integral domain and the field
of quotients of Z is precisely the field Q of rational numbers. A rational
number is usually written as a/b which is precisely [a, b]. Recall from the
high school mathematics that two rational numbers a/b and c/d are equal if
and only if ad 5 bc and that a/b represents a class of pairs (c, d) for which
ad 5 bc.
By means of the monomorphism f of an integral domain R into the field of
quotients F defined by f (a) 5 [a, 1], we can identify an element a in R with
the element [a, 1] in F. With this identification, we can treat R as a subring
of F. Also, any element [a, b] of F can be expressed as ab21 with a  R and
0  b  R, since

[a, b] 5 [a, 1] ? [1, b] 5 [a, 1][b, 1]21.

Corollary 10.6.1. If R is a field, then the field of quotients of R is isomorphic

to R.

Proof: Let R be a field and F be its field of quotients.

Let f : R → F be the monomorphism defined by

f (a) 5 [a, 1] for any a  R.

Then, for any [a, b]  F, we can write

[a, b] 5 [ab21, 1] 5 f (ab21)

and ab21  R (Note that b  0 and hence b is a unit in R).

Therefore, f is a surjection also and hence f is an isomorphism. Thus,
F  R.

Q001-Algebra-111001_CH 10.indd 61 9/21/2011 4:54:21 PM

 10-62   Algebra – Abstract and Modern

Theorem 10.6.3. Let R be an integral domain and K be a field containing R

as a subring. Let

Q 5 {ab21  K: a and b  R and b  0}.

Then, Q is a subfield of K and is isomorphic to the field of quotients of R.

Proof: Let F be the field of quotients of R and define g : F → Q by

g ([a, b]) 5 ab21 for any [a, b]  F.

For any [a, b] and [c, d]  F, we have

[a, b] 5 [c, d] ⇔ ad 5 bc
⇔ ab21 5 cd21

This shows that g is well defined and is an injection. By the definition of Q, g

is a surjection also. Further,

g([a, b] 1 [c, d]) 5 g([ad 1 bc, bd])

5 (ad 1 bc)(bd)21
5 (ad 1 bc)b21d21
5 ab21 1 cd21
5 g([a, b]) 1 g([c, d])
     and g([a, b] [c, d]) 5 g([ac, bd]) 5 ac(bd)21
5 ab21cd21 5 g([a, b])g([c, d]).

Therefore, g is a homomorphism also. Thus, g is an isomorphism of

F onto Q. This implies that Q is a subfield of K and F  Q.

Corollary 10.6.2. Let R be an integral domain and F be its field of quotients.

Then, F is the smallest field containing R, in the sense that any field contain-
ing R as a subring should contain an isomorphic copy of F.

Worked Exercise 10.6.1. Let F be the field of quotients of an integral domain

R. Then prove that R and F are of same characteristic.

Answer: We can treat R as a subring of F and the unities in R and F are the
same. In Theorem 9.3.2, we have proved that the characteristic of any ring

Q001-Algebra-111001_CH 10.indd 62 9/21/2011 4:54:21 PM

 Ideals and Quotient Rings   10-63

with unity is precisely the order of the unity in the additive group of the ring.
Therefore,

char(R) 5 O(1) 5 char(F ),

where O(1) is the order of 1 in the group (R, 1).

Worked Exercise 10.6.2. Let Z[i] be the ring of Gaussian integers. Deter-
mine the field of quotients of Z[i].

Answer: Recall that

Z[i] 5 {a 1 bi : a and b are integers}

and that Z[i] is an integral domain under the addition and multiplication
of complex numbers. Let C be the field of complex numbers. Then, Z[i] is
a subring of C. By Theorem 10.6.3, the field of quotients of Z[i] is equal
­(isomorphic) to

Q 5 {st21 : s and t  Z[i] and t  0}

If s 5 a 1 bi and t 5 c 1 di  0, then c  0 or d  0 and hence c2 1

d2 > 0. Now,

a  bi ( a  bi )(c  di )
st 1  
c  di (c  di )(c  di )
ac  bd  (bc  ad )i

c2  d 2
ac  bd bc  ad
 2  i
c  d 2 c2  d 2

Since a, b, c and d are all integers, it follows that st21 belongs to the set

Q[i] 5 {p 1 qi : p and q are rational}

and hence Q ⊆ Q[i].

On the other hand, let p 1 qi  Q[i] and p  a and q  c , where a, b, c,
b d
d  Z, b  0 and d  0. Then,

p 5 ab21 and q 5 cd21, a, b, c, d  Z ⊆ Z[i]

Q001-Algebra-111001_CH 10.indd 63 9/21/2011 4:54:21 PM

 10-64   Algebra – Abstract and Modern

and hence p and q  Q. Since i  Z[i] ⊆ Q, it follows that p 1 qi  Q. There-

fore, Q[i] ⊆ Q. Thus, Q 5 Q[i]. That is, the field of quotients of Z[i] is Q[i].
Recall that Q is the field of quotients of Z.

Worked Exercise 10.6.3. Prove that any isomorphism between two integral
domains can be extended to their fields of quotients.

Answer: Let R and R′ be integral domains and f : R → R′ be an isomorphism.

Let F and F′ be fields of quotients of R and R′, respectively. Then,

F 5 {ab21 : a and b  R and b  0}

       and F′ 5 {xy21 : x and y  R′ and y  0}.

Define g : F → F′ by g(ab21) 5 f (a)f (b)21. Note that b  0 in R if and only if

f (r)  0 in R′, since f is an isomorphism. Also,

ab21 5 cd21 ⇔ ad 5 bc
⇔ f (ad) 5 f (bc)
⇔ f (a)f (d) 5 f (b)f (c)
⇔ f (a)f (b)21 5 f (c)f (d)21

This shows that g is well defined and is an injection. Also, if xy21  F′ with x,
y  R′ and y  0, we can choose elements a and b in R such that f (a) 5 x and
f (b) 5 y (since f is a bijection). Then, b  0 and g(ab21) 5 f (a)f (b)21 5 xy21.
Therefore, g is a surjection also. Further, for any ab21, cd21,  F,

g(ab21 1 cd21) 5 g((ad 1 bc)(bd)21)

5 f (ad 1 bc)f (bd)21
5 (f (a)f (d) 1 f (b)f (c))(f (b)f (d))21
5 f (a)f (b)21 1 f (c)f (d)21
5 g(ab21) 1 g(cd21)
   and g((ab ) ? (cd )) 5 g(ac(bd)21)
21 21

5 f (ac)f (bd)21
5 f (a)f (c)(f (b)f (d))21
5 f (a)f (b)21f (c)f (d)21
5 g(ab21)g(cd21).

Therefore, g is a homomorphism also. Thus, g is an isomorphism of F onto

F′. Also, g is an extension of f, since g(a) 5 g(a121) 5 f (a)f (1)21 5 f (a)1 5
f (a) for all a  R.

Q001-Algebra-111001_CH 10.indd 64 9/21/2011 4:54:21 PM

 Ideals and Quotient Rings   10-65

Worked Exercise 10.6.4. Let F be the field of quotients of an integral domain

R and S be a subring of F such that R ⊆ S ⊆ F. Prove that the field of quotients
of S is isomorphic to F.

Answer: Since F is a field, it is an integral domain. Being a subring of F, S is

also an integral domain.
Note that the unity in R is same unity in S as well as in F. Now, by Theorem
10.6.3, the field of quotients of S is given by

Q 5 {ab21: a and b  S and b  0}.

Since F is a field containing S, we get from Corollary 10.6.2 that Q ⊆ F. On

the other hand,

x  F ⇒ x 5 ab21, where a and b  R and b  0

⇒ x 5 ab21, a, b  S, b  0 (since R ⊆ S).
⇒ x  Q.

Thus, F ⊆ Q and hence F 5 Q.

EXERCISE 10(f)

1. Determine the field of quotients of each of the following integral domains.

(i) Z
(ii) Z7
(iii) Z11
(iv) Z2[i] 5 {a 1 bi : a and b  Z2} under addition and multiplication
­modulo 2.
(v) R
(vi) Q[i]
2. Prove in detail that Zn 3 R in the proof of Theorem 10.6.1 is a ring with unity
and is of characteristic n, where R is a ring of characteristic n.
3. Determine the field of quotients of the integral domain
Z[ 3]  {a  b 3 : a and b  Z}.

4. Let R 5 { a : a and b  Z and 5 does not divide b}.

b
Prove that R is a subring of the ring Q of rational numbers and deduce that R is
an integral domain. Determine the field of quotients of R.
5. Let F be a field containing no subfield properly (such fields are called prime
fields). Prove that F is isomorphic either to the field Q of rational numbers or to
the field Zp of integers modulo a prime p.

Q001-Algebra-111001_CH 10.indd 65 9/21/2011 4:54:22 PM

 10-66   Algebra – Abstract and Modern

6. Prove that every field contains a subfield, that is, isomorphic to Q or Zp (such
subfields are called prime subfields).

7. Prove that any automorphism of a field F fixes every element of the prime
subfield.

8. Determine the prime subfields of each of the following fields.

(i) R
(ii) Q
(iii) Z79
(iv) Q[ 5 ] {a  b 5 : a and b  Q}
(v) Q[ 2 ] {a  b 2 : a and b  Q}
(vi) Q[ i ]  {a  bi : a and b  Q}

9. If F is a subfield of a field K, then prove that F and K have the same prime
subfields.

10. If a field F has exactly 9 elements, then prove that the prime subfield of F is
isomorphic to Z3.

11. If Z5 is the prime subfield of a field F, then prove that there exists a  F such
that a 1 a  0.

12. Prove that any automorphism of an integral domain can be extended uniquely to
an automorphism of its field of quotients.

13. Let R be a commutative ring with no zero divisions. Then prove that R can be
embedded in an integral domain.

14. Prove that a commutative ring can be embedded in a field if and only if it has no
zero divisors.

15. Let R be a commutative ring with unity. A subset S of R is said to be multiplica-

tive if 1  S, 0  S and ab  S for any a and b  S. Define a binary relation U
on R 3 S by
(a, s) U (b, t) ⇔ there exists u  S such that u(at 2 bs) 5 0.
Prove that U is an equivalence relation on R 3 S. Let the equivalence containing
(a, s) be denoted by a/s and let

 a 
S 1R   : a ∈ R and s  S .
 s 

Define addition and multiplication on S21R by

a b at  bs a b ab
  and ⋅ 
s t st s t st

Q001-Algebra-111001_CH 10.indd 66 9/21/2011 4:54:22 PM

 Ideals and Quotient Rings   10-67

Then prove that (S21R, 1, ?) is a commutative ring with unity. This ring is called
the ring of fractions of R by S.

16. Prove that the field of quotients of an integral domain R is the ring of fractions
of R by R 2 {0}.

17. Let P be a prime ideal of a commutative ring R with unity. Then prove that
(R 2 P)21R has a unique maximal ideal.

18. Let S be a multiplicative subset of a commutative ring R with unity. Prove that
there is a one-to-one correspondence between the prime ideals of R disjoint with
S and the prime ideals of the ring S21R of fractions of R by S.

Q001-Algebra-111001_CH 10.indd 67 9/21/2011 4:54:23 PM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 10.indd 68 9/21/2011 4:54:23 PM

 11
Polynomial Rings
11.1 Rings of Polynomials
11.2 The Division Algorithm
11.3 Polynomials over a Field
11.4 Irreducible Polynomials

We are very familiar with polynomials which are introduced to us very early
in our mathematical education, in fact, in high school itself, we are thoroughly
drilled in adding, multiplying, dividing, factoring and simplifying them. We
have learnt the remainder theorem in eighth or ninth standard. Later, at higher
level, polynomials appear as functions and we were concerned with their con-
tinuity, derivatives and integrals and their maxima and minima. Now, we are
interested in polynomials, but from neither of the above view points. Here,
polynomials will simply be elements of a certain ring and we shall be con-
cerned with the algebraic properties of this ring.
At the secondary school level, we have studied polynomials with integer
coefficients, rational coefficients, real coefficients and, may be even complex
coefficients. Notice that, in each case, the set of coefficients is a ring and the
set of polynomials also forms a ring under suitable addition and multiplica-
tion, with which we are all familiar. In this chapter, we make an abstraction
of these cases and study polynomials with coefficients from a given abstract
ring.

11.1 RINGS OF POLYNOMIALS

The word ‘polynomial’ reminds us an expression or symbol of the form

a0  a1 x  a2 x 2  an x n .

Q001-Algebra-111001_CH 11.indd 1 9/16/2011 10:17:05 AM

 11-2   Algebra – Abstract and Modern

We used to quickly add, multiply and divide such expressions. Now, it is

appropriate to clarify what different terms of such an expression mean. Of
course, a0, a1, a2, …, an are from some number system. Instead, we can take
these ai’s as elements of an abstract ring. What are a1x, a2x2, …, anxn? What is
the 1 in between them? We used to write

3x 1 x3 1 2x5 for 0 1 3x 1 0x2 1 1x3 1 0x4 1 2x5.

Also, we used to treat two polynomials

a0 1 a1x 1 a2x2 1 … 1 anxn and b0 1 b1x 1 b2x2 1 … 1 bmxm

equal if and only if ai 5 bi for all i  0.

In essence, the coefficients a0, a1, a2, …, an are only important things and
the arithmetics of polynomials only depend on these ai’s. In this section, we
shall formulise these intuitive ideas and arrive at an exact idea of how the
classical arithmetic of polynomials fit into our ring theory. To begin with, we
have following definition.

Definition 11.1.1. Let (R, 1, ?) be any ring. A polynomial over R is defined

to be an infinite sequence

( a0 , a1 , a2 , …, an , …)

of elements of R such that an’s are zero for all but finite number of n’s; equiva-
lently, there exists a nonnegative integer k such that an 5 0 for all n  k. The
set of all polynomials over R will be denoted by Poly(R).
Recall that an infinite sequence of elements of R can be viewed as a
mapping of the set of nonnegative integers into R. Let us agree that two
polynomials

f  ( a0 , a1 , a2 , …) and g  (b0 , b1 , b2 , …)

are considered to be equal if and only if an 5 bn for all n  0.

Often it is convenient to use the notation (a0, a1, a2, …, an, 0, 0, …) for a
polynomial with an as the last nonzero term; when n 5 0, we allow the pos-
sibility that a0 5 0 in order to include the zero polynomial (0, 0, 0, …) each
of whose term is zero. With this notation, we have

Poly( R)  {( a0 , a1 , a2 , …, an , 0, 0, …) : ai  R and n  0}.

Q001-Algebra-111001_CH 11.indd 2 9/16/2011 10:17:05 AM

 Polynomial Rings   11-3

Therefore, (0, 1, 1, 0, 1, 0, 0, 0, …) is a polynomial over Z, where as (0, 1, 0,

1, 0, 1, 0, 1, …) is not. In the following, we introduce suitable operations on
Poly(R) to make it a ring.

Definition 11.1.2. For any polynomials a 5 (a0, a1, a2, …) and b 5 (b0, b1,
b2, …) over a ring (R, 1, ?), define

a  b  ( a0  b0 , a1  b1 , a2  b2 , …)
and a ⋅ b  (c0 , c1 , c2 , …),
n
where cn  a0 bn  a1bn1  an b0   ar bnr   ar bs .
r0 rsn
Note that the additive operation 1 and the multiplicative operation ? on the
right sides of the above defining equations are those in the (R, 1, ?). Since a
and b are in Poly(R), there exist nonnegative integers m and n such that

ai 5 0 for all i  m and bi 5 0 for all i  n

and hence, for any i  max{m, n}, ai 5 0 5 bi so that

ai  bi  0
i
and ci   ar bir  0 for all i  max{m, n}
r0

(since i  m 1 n implies r  m or i 2 r  n for any 0  r  i and therefore,

ar 5 0 or bi2r 5 0). This leads to the fact that a 1 b and a ? b are well
defined.

Theorem 11.1.1. For any ring (R, 1, ?), (Poly(R), 1, ?) is a ring, where 1 and ?
are the operations defined above.

Proof: Let (R, 1, ?) be a ring and a, b and c be arbitrary elements of Poly(R).

Then, a, b and c are polynomials given by

a  ( a0 , a1 , a2 , …)
b  (b0 , b1 , b2 , …)
and c  (c0 , c1 , c2 , …)

with ai’s, bi’s and ci’s are elements in the given ring R. Then, using the asso-
ciatively and commutativity of 1 in R, we can prove that

( a  b)  c  a  (b  c) and a  b  b  a.

Q001-Algebra-111001_CH 11.indd 3 9/16/2011 10:17:06 AM

 11-4   Algebra – Abstract and Modern

Therefore, 1 is associative and commutative in Poly(R). Let us write 0 for the

polynomial (0, 0, 0, …). Then,

a  0  a  0  a for all a  Poly( R).

Therefore, 0 is the identity element for 1. Also, for any a 5 (a0, a1, a2, …) in
Poly(R), the polynomial 2a defined by

a  (a0 ,  a1 ,  a2 , …)

satisfies the property

a  (a)  (0, 0, 0, …)  0 a  a

and hence 2a is the inverse of a with respect to 1. Thus, (Poly(R), 1) is an

abelian group.
To prove the associativity of multiplication, let

a ⋅ b  ( d0 , d1 , d2 , …)
and ( a ⋅ b) ⋅ c  ( x0 , x1 , x2 , …).

Then,

n n
dn  ∑ ar bnr ∑ ar bs
r0 rs

n
and xn  ∑ ds cns  ∑ ds ct
s0 stn

 
∑  ar bu  ct
stn
r∑
us

 ∑
rutn
ar bu ct

 
 ∑ a  ∑ b c 
rsn
r
uts
u t

which is precisely the nth term in a ? (b ? c). Thus,

( a ⋅ b) ⋅ c  a ⋅ (b ⋅ c).

Q001-Algebra-111001_CH 11.indd 4 9/16/2011 10:17:07 AM

 Polynomial Rings   11-5

Also, if a ? (b 1 c) 5 (y0, y1, y2, …), then

yn  ∑ a (b  c )
rsn
r s s

 ∑ab ∑ac
rsn
r s
rsn
r s

which is precisely the nth term in (a ? b) 1 (a ? c). Thus, a ? (b 1 c) 5 a ? b 1

a ? c. Similarly, one can prove that (a 1 b) ? c 5 a ? c 1 b ? c. Thus, (Poly(R),
1, ?) is a ring.

Definition 11.1.3. For any ring R, (Poly(R), 1, ?) is called the ring of polyno-
mials over R and is simply denoted by Poly(R).
In the following, we prove that any given ring R is isomorphic to a subring
of the ring of polynomials over R.

Theorem 11.1.2. Any ring R can be embedded in Poly(R).

Proof: Let (R, 1, ?) be a ring and Poly(R) be the ring of polynomials

over R. Define f : R → Poly(R) by f (a) 5 (a, 0, 0, 0, …). Then, for any
a and b  R,

f ( a  b)  ( a  b, 0, 0, …)
 ( a  b, 0  0, 0  0, …)
 ( a, 0, 0, …)  (b, 0, 0, …)
 f ( a)  f ( b)
and f ( ab)  ( ab, 0, 0, …)
 ( a, 0, 0, …)  (b, 0, 0, …)
 f ( a)  f (b).

Therefore, f is a homomorphism. Also, for any a and b  R,

f ( a)  f (b) ⇒ ( a, 0, 0, …)  (b, 0, 0, …)
⇒ a  b.

Thus, f is an injection also and hence f is a monomorphism of R into Poly(R).

This says that R can be embedded in Poly(R) and R is isomorphic to the sub-
ring f (R) of Poly(R).

Q001-Algebra-111001_CH 11.indd 5 9/16/2011 10:17:07 AM

 11-6   Algebra – Abstract and Modern

Theorem 11.1.3. For any ring R, the ring Poly(R) is commutative if and only
if R is commutative.

Proof: Let R be a ring. If Poly(R) is commutative, then clearly R is commuta-

tive, since R is isomorphic to a subring of Poly(R). Conversely, suppose that
R is commutative. Then, for any

a 5 (a0, a1, a2, …) and b 5 (b0, b1, b2, …),

in Poly(R), we have, for any n  0,

nth term in a ? b  ∑ab

rsn
r s

 ∑ba
srn
s r

th
 n term in b  a

and hence a ? b 5 b ? a. Thus, Poly(R) is commutative.

Theorem 11.1.4. A ring R is with unity if and only if the ring Poly(R) is with
unity.

Proof: Let R be a ring. Suppose that R has unity 1.

Let e 5 (1, 0, 0, 0, …). Then, for any a 5 (a0, a1, a2, …) in Poly(R), we have,
for any n  0,

nth term in a ? e  ∑ae

rsn
r s

 an (since es  0 for all s  0 and e1  1)

 ∑ er as
rsn

 nth term in e ⋅ a

and hence a ? e 5 a 5 e ? a. Thus, e is the unity element in Poly(R).

Note that a ring R is trivial if and only if the ring Poly(R) is trivial. Recall
that a nontrivial commutative ring with unity and without zero divisors is
called an integral domain.

Theorem 11.1.5. Let (R, 1, ?) be a ring R. Then, R is an integral domain if

and only if Poly(R) is an integral domain.

Proof: First note that, from Theorems 11.1.1 and 11.1.4, R is a nontrivial
commutative ring with unity if and only if so is Poly(R). Therefore, we can

Q001-Algebra-111001_CH 11.indd 6 9/16/2011 10:17:07 AM

 Polynomial Rings   11-7

assume that R is a nontrivial commutative ring with unity, which is a neces-

sary qualification for R (and hence for Poly(R)) to be an integral domain.
Suppose that R is an integral domain. Let a 5 (a0, a1, a2, …) and b 5 (b0,
b1, b2, …) be any nonzero elements in Poly(R). Since a  0 and b  0, there
exist nonnegative integers n and m such that an  0, bm  0 and ai 5 0 for
all i  n and bj 5 0 for all j  m. Now, consider the (n 1 m)th term in the
product a ? b. It is given by

∑
rsnm
ar bs  an bm  0,

since R is an integral domain, an  0 and bm  0. (Note that r 1 s 5 n 1 m

⇒ r  n or s  m ⇒ (r 5 n or ar 5 0) or (s 5 m or bs 5 0). Therefore, the
(n 1 m)th term of a ? b is nonzero and hence a ? b  0. Therefore, Poly(R) has
no zero divisors. Thus, Poly(R) is an integral domain.
Conversely, suppose that Poly(R) is an integral domain. Since R is isomorphic
to a subring of Poly(R), it follows that R is an integral domain.
In the following result, we demonstrate that Poly(R) can never be a field,
even when R is a field.

Theorem 11.1.6. For any ring R, Poly(R) can never be a field.

Proof: Let R be a ring and suppose that Poly(R) is a field. Then, Poly(R) is an
integral domain and hence, by Theorem 11.1.5, R is also an integral domain.
Consider the element

x = (0, 1, 0, 0, …) in Poly( R).

Then, x is a nonzero element in Poly(R) and hence x is a unit. Therefore, there

exists a 5 (a0, a1, a2, …) in Poly(R) such that

x  a  a  x  1  (1, 0, 0, …)
hence ( a, a0 , a1 , a2 , …)  (1, 0, 0, …).

Therefore, 0 5 1, which is a contradiction to the fact that R is nontrivial.

Thus, Poly(R) is not a field.
The element x given in the above proof is of special importance. Even from
the high school days, we are well aware that a polynomial over R is an expres-
sion of the form

a0  a1 x1  a2 x 2  an x n ,

Q001-Algebra-111001_CH 11.indd 7 9/16/2011 10:17:08 AM

 11-8   Algebra – Abstract and Modern

where a0, a1, a2, …, an are real numbers and x is an indeterminate. Though we
are familiar with this, we did not know what x is what ax is for any a  R.
Further, we should give a mathematically valid explanation for the operation
symbol 1 in the above expression of a polynomial. We shall give satisfactory
answers to these questions in the following result. Recall that R can be identi-
fied with a subring of Poly(R) and any element a in R can be identified with
the polynomial (a, 0, 0, …).

Theorem 11.1.7. Let (R, 1, ?) be a ring with unity and x be the polynomial
over R given by
x  (0, 1, 0, 0, …).

Then, any polynomial over R can be expressed uniquely as

a0  a1  x  a2  x 2  an  x n ,

where a0, a1, …, an are elements of R, identified with the elements of Poly(R).

Proof: Note that, in the expression a0 1 a1 ? x 1 a2 ? x2 1 … 1 an ? xn, the

operation symbols 1 and ? denote the addition and multiplication in Poly(R).
Using the definitions of x and the multiplication in Poly(R), the following can
be proved easily.

x  (0, 1, 0, 0, …)
x 2  x ⋅ x  (0, 0, 1, 0, 0, …)
x 3  x 2 ⋅ x  (0, 0, 0, 1, 0, 0, …)

x n  x n1 ⋅ x  (0, 0, …, 0, 1, 0, 0, …)
( n 1) th term

for any positive integer n. Also, for any a  R, by identifying a with (a, 0,
0, …) in Poly(R), we have

a ⋅ x  (0, a, 0, 0, …)
a ⋅ x 2  (0, 0, a, 0, 0, …)
a ⋅ x 3  (0, 0, 0, a, 0, 0, …)

n
a ⋅ x  (0, 0, …, 0, a, 0, 0, …)
( n 1) th term

Q001-Algebra-111001_CH 11.indd 8 9/16/2011 10:17:08 AM

 Polynomial Rings   11-9

for any positive integer n. If p is a polynomial over R, then p is a sequence

(a0, a1, a2, …, an, 0, 0, …) with ai’s are in R. These ai’s are unique in R
such that

p  ( a0 , a1 , a2 , …, an , 0, 0, …).

Now, we have

p  ( a0 , 0, 0, …)  (0, a1 , 0, 0, …)  (0, 0, …, 0, an , 0, 0, …)

 a0  a1  x  a2  x 2  an  x n .

Remarks 11.1.1
1. If we identify a  R with (a, 0, 0, …) in Poly(R) and identify R as a
subring of Poly(R), then Poly(R) is the subring generated by R and x. For
this reason, we prefer to write R[x] for Poly(R).
2. The expression a0 1 a1x 1 a2x2 1 … 1 anxn for a polynomial looks
simple and elegant for two reasons. The first one is that we are familiar
with this right from our school days. The second is that we can straight
away multiply two polynomials by treating x also as a real number or as
an element in the ring containing ai’s.
3. As mentioned in the above theorem and its proof, x is not an element
in the ring R and it is an element in Poly(R); that is, x is a polynomial
over R.
4. x is usually called an indeterminate.
5. When we are completely aware of what Poly(R) and R[x] are, we prefer
to use familiar notation for polynomials over a given ring R and for the
ring of polynomials over R. R[x] is the most standard notation used to
denote the ring polynomials over R.
6. The expression a0 1 a1x 1 … 1 anxn for a polynomial is known as poly-
nomial in an indeterminate form. Though x is called an indeterminate, it
is actually a polynomial by itself and the operations 1 and ? in the above
expression are the addition and multiplication of polynomials only.
7. Often, for
n
convenience, a polynomial a0 1 a1x 1 … 1 anxn is also writ-
ten as  ai x i , with the assumption that x0 5 1.
i1

To sum up, the polynomials in an indeterminate form are expressions of

the form a0 1 a1x 1 a2x2 1 … 1 anxn, where a0, a1, …, an are elements of
a given ring R and the set of all such expressions will be denoted by R[x].
The elements of R[x] can be added and multiplied as we do in the number
systems.

Q001-Algebra-111001_CH 11.indd 9 9/16/2011 10:17:09 AM

 11-10   Algebra – Abstract and Modern

Example 11.1.1

1. Z[x] 5 {a0 1 a1x 1 … 1 anxn : n  0, ai  Z}. 3 1 2x 1 x2 1 4x3 is in

Z[x], while 2 1 3x 1 3x2 is not in Z[x]. However, 2 1 3x 1 3x2 is an
element of R[x].
2. Let us compute (3 1 2x 1 x3) 1 (5x 1 3x2 1 2x4) and (3 1 2x 1 x3) ?
(5x 1 3x2 1 2x4).
First write p 5 3 1 2x 1 x3 5 a0 1 a1x 1 a2x2 1 a3x3 and q 5 5x 1 3x2
1 2x4 5 b0 1 b1x 1 b2x2 1 b3x3 1 b4x4, where a0 5 3, a1 5 2, a2 5 0 and
a3 5 1; and b0 5 0, b1 5 5, b2 5 3, b3 5 0 and b4 5 2.
3 4 4
p  q  ∑ ai x i  ∑ bi x i  ∑ (ai  bi ) x i
i0 i0 i0

where ai 5 0 for all i  3 and bi 5 0 for all i  4.

 p 1 q 5 (3 1 0) 1 (2 1 5)x 1 (0 1 3)x2 1 (1 1 0)x3 1 (0 1 2)x4
5 3 1 7x 1 3x2 1 x3 1 2x4.
Also, pq 5 c0 1 c1x 1 c2x2 1 c3x3 1 c4x4 1 c5x5 1 c6x6 1 c7x7,
where c0 5 a0 ? b0 5 3 ? 0 5 0
c1 5 a0 ? b1 1 a1 ? b0 5 3 ? 5 1 2 ? 0 5 15
c2 5 a0 ? b2 1 a1 ? b1 1 a2 ? b0 5 3 ? 3 1 2 ? 5 1 0 ? 0 5 19
c3 5 3i0 ai b3i 5 3 ? 0 1 2 ? 3 1 0 ? 5 1 1 ? 0 5 6
c4 5 3 ? 2 1 2 ? 0 1 0 ? 3 1 1 ? 5 1 0 ? 0 5 11
c5 5 3 ? 0 1 2 ? 2 1 0 ? 0 1 1 ? 3 1 0 ? 5 1 0 ? 0 5 7
c6 5 3 ? 0 1 2 ? 0 1 0 ? 2 1 1 ? 0 1 0 ? 2 1 0 ? 5 1 0 ? 0 5 0
c7 5 3 ? 0 1 2 ? 0 1 0 ? 0 1 3 ? 2 1 0 ? 0 1 0 ? 3 1 0 ? 5 1
0 ? 0 5 6.
Thus, p ? q 5 0 1 15x 1 19x2 1 6x3 1 11x4 1 7x5 1 0x6 1 6x7
5 15x 1 19x2 1 6x3 1 11x4 1 7x5 1 6x7.
3. Let us compute (2 1 x 1 x3) 1 (1 1 2x 1 x2 1 x3) in Z3[x]. The required
sum is
(2 13 1) 1 (1 13 2)x 1 (0 13 1)x2 1 (1 13 1)x3 5 x2 1 2x3
since 2 13 1 5 0 in Z3.
4. 1 1 2x is a unit in Z4[x]; for
(1 1 2x)(1 1 2x) 5 1 1 (2 14 2)x 1 (2 ?4 2)x2 5 1

since 2 14 2 5 0 5 2 ?4 2. Here, 1 1 2x is the multiplicative inverse of

itself.

Q001-Algebra-111001_CH 11.indd 10 9/16/2011 10:17:09 AM

 Polynomial Rings   11-11

Definition 11.1.4. Let p 5 (a0, a1, a2, …) be a nonzero polynomial over a

ring R. Then, ai  0 for some i. The largest n for which an  0 is called the
degree of p and is denoted by deg(p).
Note that we have not defined the degree of the zero polynomial (0, 0, 0, …).
Also, if p is a polynomial of degree n, then p can expressed as

p 5 a0 1 a1x 1 … 1 anxn

The ai’s involved in this expression are called the coefficients in the poly-
nomial p.

Example 11.1.2
1. The degree of 2 1 3x4 in R[x] is 4, since
2 1 3x4 5 2 1 0x 1 0x2 1 0x3 1 4x4
2. In Z4[x], the degree of (1 1 2x)2 is 0, since
(1 1 2x)2 5 1 1 (2 14 2)x 1 (2 ?4 2)x2
5 1 1 0x 1 0x2 5 1.

Definition 11.1.5. The zero polynomial and the polynomials of degree 0 are
called constant polynomials.
For any ring R, the set of constant polynomials over a ring R form a sub-
ring of R[x] and is isomorphic to R. In the following, we discuss how the
degrees of polynomials vary when we take sums and products.

Definition 11.1.6. Let R be a ring and R[x] be the ring of polynomials over R.
The following holds for any nonzero polynomials f and g  R[x]:
1. Either f 1 g 5 0 or deg(f 1 g)  max{deg(f ), deg(g)}.
2. Either f ? g 5 0 or deg(f ? g)  deg(f ) 1 deg(g).
3. If R is an integral domain, then f ? g  0 and deg(f ? g) 5 deg(f ) 1 deg(g).

Proof: Let deg(f ) 5 m and deg(g) 5 n. Then,

   f 5 a0 1 a1x 1 a2x2 1 … 1 amxm, am  0

      and g 5 b0 1 b1x 1 b2x2 1 … 1 bnxn, bn  0.

Let k 5 max{m, n}. Then, we can write

f 5 a0 1 a1x 1 a2x2 1 … 1 akxk

        and g 5 b0 1 b1x 1 b2x2 1 … 1 bkxk,

Q001-Algebra-111001_CH 11.indd 11 9/16/2011 10:17:10 AM

 11-12   Algebra – Abstract and Modern

where ai5 0 for i  m and bj 5 0 for j  n.

1. We have f 1 g 5 (a0 1 b0) 1 (a1 1 b1)x 1 … 1 (ak 1 bk)xk and therefore
either f 1 g 5 0 or
deg(f 1 g)  k 5 max{deg(f ), deg(g)}.
2. Suppose that f ? g  0. Let
f ? g 5 c0 1 c1x 1 c2x2 1 … 1 csxs,
r
where cr   ai bri   ai b j .
i0 i jr
If i 1 j 5 r  m 1 n, then either i  m or j  n and hence ai 5 0 or
bj 5 0 and, in either case aibj 5 0. Therefore,
cr 5 0 for all r  m 1 n
which implies that deg(f ? g)  m 1 n 5 deg(f ) 1 deg(n).
3. Suppose that R is an integral domain. Then,

Cmn  ∑
i jmn
ai b j  am bn  0 (since R is an integral domain)

and, as in (2), cr 5 0 for all r  m 1 n. Therefore,

deg(f ? g) 5 m 1 n 5 deg(f ) 1 deg(g).
In fact, the converse of (3) above is also true, in the sense of the following
corollary.

Corollary 11.1.1. Let R be a nontrivial commutative ring with unity. Then,

the following are equivalent to each other.
1. deg(f ? g) 5 deg(f ) 1 deg(g) for all nonzero f and g  R[x].
2. R is an integral domain.
3. R[x] is an integral domain.

Proof: We have proved (2) ⇔ (3) in Theorem 11.1.5 and (2) ⇒ (1) by
­Definition 11.1.6 (3).
(1) ⇒ (3) follows from the fact that the degree is defined for nonzero poly-
nomials only and hence, by (1), f ? g  0 for all nonzero f and g in R[x]. b

Worked Exercise 11.1.1. In Z6[x], let f 5 2 1 x 1 4x2 1 3x3

and g 5 4 1 3x 1 5x2 1 3x3.

Compute f 1 g and f ? g and their degrees.

Q001-Algebra-111001_CH 11.indd 12 9/16/2011 10:17:10 AM

 Polynomial Rings   11-13

Answer: f 1 g 5 (2 16 4) 1 (1 16 3)x 1 (4 16 5)x2 1 (3 16 3)x3

5 0 1 4x 1 3x2 1 0 ? x3
5 4x 1 3x2

and therefore, deg(f 1 g) 5 2  max{deg(f ), deg(g)}.

Also, f ? g 5 (2 1 x 1 4x2 1 3x3) (4 1 3x 1 5x2 1 3x3)

5 (2 ?6 4) 1 (2 ?6 3 16 1 ?6 4)x 1 (2 ?6 5 16 1 ?6 3 16 4 ?6 4)x2
1 (2 ?6 3 1 6 1 ?6 5 16 4 ?6 3 16 3 ?6 4)x3
1 (1 ?6 3 16 4 ?6 5 16 3 ?6 3)x4 1 (4 ?6 3 16 3 ?6 5)x5
1 (3 ?6 3)x6
5 2 1 4x 1 (4 16 3 16 4)x2 1 (5 16 0 16 0)x3
1 (3 16 2 16 3)x4 1 (0 16 3)x5 1 3x6
5 2 1 4x 1 5x2 1 5x3 1 2x4 1 3x5 1 3x6

and therefore, deg(f ? g) 5 6 5 deg(f ) 1 deg(g).

Worked Exercise 11.1.2. Give examples of two polynomials f and g over a

ring R for which f ? g  0 and

deg(f ? g)  deg(f ) 1 deg(g).

Answer: Let f 5 1 1 2x and g 5 1 1 3x2 in Z6[x].

Then, f ? g 5 (1 1 2x)(1 1 3x2)

5 1 1 2x 1 3x2 1 (2 6 3)x3
5 1 1 2x 1 3x2 1 (since 2 ?6 3 5 0).

Therefore, deg(f ? g) 5 2  deg(f ) 1 deg(g).

Worked Exercise 11.1.3. Let R be a commutative ring with unity. Then, prove
that R and R[x] have the same characteristic.

Answer: First note that both R and R[x] have the same unity, namely 1. Now,
it follows from Theorem 9.3.2 that

char R 5 0(1) in (R, 1)

5 0(1) in (R[x], 1)
5 char R[x].

Q001-Algebra-111001_CH 11.indd 13 9/16/2011 10:17:10 AM

 11-14   Algebra – Abstract and Modern

EXERCISE 11(a)

1. Evaluate the following:

(i) (1, 2, 0, 4, 0, 0, …) 1 (2, 0, 1, 4, 0, 0, …) in Poly(Z5).
(ii) (1, 2, 0, 3, 0, 0, …) ? (3, 1, 0, 4, 2, 0, 0, …) in Poly(Z).
(iii) (4, 3, 2, 1, 0, 0, …) ? (1, 2, 3, 4, 5, 0, 0, …) in Poly(Z6).
(iv) (1, 2, 3, 4, 0, 0, …) 1 (4, 3, 2, 1, 0, 0, …) in Poly(Z5).
(v) (0, 1, 0, 0, …)n in Poly(R) for any n  Z1.
(vi) (1, 1, 0, 0, …)n in Poly(Zn), for each n  Z1.
2. State whether each of the following is true and substantiate your answer.
(i) Z2[x] is an integral domain.
(ii) Z5[x] is a field.
(iii) Z6[x] is an integral domain.
(iv) A ring R is finite if and only if R[x] is finite.
(v) 1 1 x is a unit in Z[x].
(vi) 2 1 2x is a nilpotent in Z4[x].
(vii) 1 1 x is a zero divisor in Z4[x].
(viii) Z2[x] is a finite integral domain.
3. If S is a subring of a ring R, then prove that S[x] is a subring of R[x].
4. If I is an ideal of a ring R, then prove that I[x] is an ideal of R[x].
5. Determine all the units in the ring Z[x].
6. Prove that a polynomial f over R is a unit in R[x] if and only if f is a nonzero
constant polynomial.
7. For any positive integer n, prove that Zn[x] is an integral domain if and only if n
is a prime number.
8. Determine the number of nonzero polynomials of degree  5 in Z2[x].
9. For any positive integers m and n, derive a formula for the number of polynomi-
als of degree less then m in Zn[x].
10. For any ring R, prove that the set
I 5 {a0 1 a1x 1 … 1 anxn  R[x] : a0 5 0}
is an ideal of R[x].
11. For any i  0, let
Ji 5 {a0 1 a1x 1 … 1 anxn  R[x] : ai 5 0}.
Then, for any ring R, is Ji an ideal of R[x].

Q001-Algebra-111001_CH 11.indd 14 9/16/2011 10:17:10 AM

 Polynomial Rings   11-15

12. Let R be a commutative ring with unity. Prove that a0 1 a1x 1 … 1 anxn  R[x] is
a unit in R[x] if and only if a0 is a unit in R and a1, a1, …, an are nilpotents in R.

13. Deduce Exercise 6 above from Exercise 12 above.

14. Let R be a commutative ring with unity and I be a proper ideal of R. Then prove
that I is a prime ideal of R if and only if I[x] is a prime ideal of R[x].

15. Can we replace ‘prime ideal’ in Exercise 14 above by ‘maximal ideal’?

16. Let R be a commutative ring with unity. Then prove that R[x]/x  R, where
x is the ideal generated by x in R[x].

11.2 THE DIVISION ALGORITHM

It is well known that there is an algorithm through which, by dividing any
integer by any nonzero integer, we get the quotient and the remainder, pre-
cisely, if a is any integer and b is a nonzero integer, then there exist unique
integers q and r such that a 5 qb 1 r and |r|  |b|. The algorithm through
which we get q and r is called the division algorithm in Z. In this section,
we extend this algorithm to polynomials over commutative rings with unity.
The degree of a polynomial is used in the derivation of the division algo-
rithm in as much the same way as the absolute value is employed among
integers.
First, let us have a small change in the notation for polynomials. The
elements of R[x] are the polynomials over R in the indeterminable form
and these will be denoted by f (x), g(x), etc., in order to mention the inde-
terminate x also. This will also be helpful later in treating polynomials as
functions.

Definition 11.2.1. Let R be a ring and f (x) 5 a0 1 a1x 1 … 1 anxn be a

nonzero polynomial over R of degree n. Then, an  0 and an is called the
leading coefficient of f (x). If an is the unity in R, then f (x) is called a monic
polynomial.

Example 11.2.1
1. If f (x) 5 2 1 3x 1 x2 1 6x3  Z[x], then degree of f (x) is 3 and 6 is the
leading coefficient in f (x).
2. If f (x) 5 2 2 x2  Z[x], then 21 is the leading coefficient of f (x).
3. 3 1 x 2 2x3 1 x4 is a monic polynomial over Z.

Theorem 11.2.1 (Division Algorithm for Polynomials). Let f (x) and g(x) be
polynomials over a commutative ring R with unity such that g(x)  0 and the

Q001-Algebra-111001_CH 11.indd 15 9/16/2011 10:17:11 AM

 11-16   Algebra – Abstract and Modern

leading coefficient of g(x) is a unit in R. Then, there exists unique polynomi-

als q(x) and r(x) in R[x] such that

f (x) 5 q(x)g(x) 1 r(x)

       and either r(x) 5 0 or deg(r(x))  deg(g(x)).

Proof: If f (x) 5 0 or f (x)  0 such that deg(f (x))  deg(g(x)), then we can
take q(x) 5 0 and r(x) 5 f (x). Therefore, we can assume that f (x)  0 and
deg(f (x)) $ deg(g(x)).
We apply induction on the degree of f (x). First, let deg(f (x)) 5 0. Then, since
deg(f (x))  deg(g(x))  0, it follows that deg(g(x)) 5 0 and hence both f (x)
and g(x) are constant polynomials, so that f (x) and g(x) are elements of R. In
particular, the leading coefficient of g(x) is g(x) itself and is invertible in R (by
the hypothesis). Now, put

q(x) 5 f (x)g(x)21 and r(x) 5 0.

Then, q(x) and r(x) satisfy the required properties.

Next, let deg(f (x)) 5 n  0 and assume that the theorem is true for all
polynomials f1(x) of degree less than n. Let deg(g(x)) 5 m. Then, we have
n $ m.
Let f (x) 5 a0 1 a1x 1 … 1 anxn, an  0
and g(x) 5 b0 1 b1x 1 … 1 bmxm, bm  0.
By hypothesis, bm is a unit in R. Put

f1 ( x )  f ( x )  an bm1 x nm g ( x ).

Then, f1(x)  R[x] and deg(f1(x))  n. Since the coefficient of xn in f1(x) is

an  ( an bm1 )bm  0, it follows that deg(f1(x))  n. By the induction hypoth-
esis, there exists polynomials q1(x) and r(x)  R[x] such that

f1(x) 5 q1(x)g(x) 1 r(x),

where r(x) 5 0 or deg(r(x))  deg(g(x)).

From this, by substituting for f1(x), we get

f ( x )  ( q1 ( x )  an bm1 x nm ) g ( x )  r ( x )
5 q(x)g(x) 1 r(x),

where q( x )  q1 (x )  an bm1 x nm .

Q001-Algebra-111001_CH 11.indd 16 9/16/2011 10:17:11 AM

 Polynomial Rings   11-17

To prove the uniqueness of q(x) and r(x), suppose that

f (x) 5 q(x)g(x) 1 r(x) 5 q9(x)g(x) 1 r9(x),

where r(x) and r9(x) satisfy the requirements of the theorem. Then, we get that

r(x) 2 r9(x) 5 (q9(x) 2 q(x))g(x).

Since the leading coefficient bm of g(x) is a unit in R, it is not a zero divisor.

If q9(x) 2 q(x)  0, then

deg[(q9(x) 2 q(x))g(x)] 5 deg(q9(x) 2 q(x)) 1 deg(g(x))

 deg(g(x))  deg(r(x) 2 r9(x))

which is a contradiction. The last in equality is based on the fact that the
degrees of both r(x) and r9(x) are less than deg(g(x)). Thus, it is necessary
that q9(x) 2 q(x) 5 0 and hence r(x) 2 r9(x) 5 0. Therefore, q(x) 5 q9(x) and
r(x) 5 r9(x).

Definition 11.2.2. The polynomials q(x) and r(x) in the above theorem and
called, respectively, the quotient and remainder on dividing f (x) by g(x). The
proof of the above theorem actually provides an algorithm to find the quotient
and remainder and hence the theorem is called the division algorithm. Let us
take up an example.

Example 11.2.2. In Z[x], let f (x) 5 2 1 3x 2 4x2 1 x3 2 3x4 and g(x) 5 3 1

x 2 x2. The leading coefficient of g(x) is 21 which is a unit in the ring Z. Put
1. f1 ( x )  f ( x )  an bm1 x nm g ( x ) (as in the above proof ).
Then, f1(x) 5 (2 1 3x 2 4x2 1 x3 2 3x4) 2 (23)(21)21x422(3 1 x 2 x2)
5 2 1 3x 2 4x2 1 x3 2 3x4 2 3x2(3 1 x 2 x2)
5 2 1 3x 2 13x2 2 2x3
2. Put f2(x)  f1 ( x )  abm1 x 32 g ( x ), where a is the leading coefficient of
f1(x). Then, f2(x) 5 (2 1 3x 2 13x2 2 2x3) 2 (22)(21)21x(3 1 x 2 x2)
5 (2 1 3x 2 13x2 2 2x3) 2 2x(3 1 x 2 x2)
5 2 2 3x 2 15x2
3. Put f3(x)  f 2 ( x )  bbm1 x 22 g ( x ), where b is the leading coefficient of
f2(x). Then, f3(x) 5 (2 2 3x 2 15x2) 2 (215)(21)21(3 1 x 2 x2)
5 (2 2 3x 2 15x2) 2 15(3 1 x 2 x2)
5 243 2 18x

Q001-Algebra-111001_CH 11.indd 17 9/16/2011 10:17:12 AM

 11-18   Algebra – Abstract and Modern

Now, deg(f3(x)) 5 1  deg(g(x)) and the process stops, we have, from (1),
(2) and (3),

f (x) 5 f1 ( x )  an bm1 x nm g ( x )  (by (1))

5 f1(x) 1 (23)(21)21x422g(x)
 f 2 (x )  a bm1 x 32 g (x )  3x 2 g (x )  (by (2))
5 f2(x) 1 (22)(21)21xg(x) 1 3x2g(x)
 f 3 (x )  b bm1 x 22 g (x )  2x g (x )  3x 2 g ( x )  (by (3))
5 243 2 18x 1 (215)(21)21g(x) 1 2xg(x) 1 3x2g(x)
5 (15 1 2x 1 3x2)g(x) 1 (243 2 18x)
5 q(x)g(x) 1 r(x),

   where q(x) 5 15 1 2x 1 3x2 and r(x) 5 243 2 18x.

Algorithm 11.2.1 (The Process of Division Algorithm). Consider f (x) and

g(x) as in Theorem 11.2.1. Let deg(f (x)) 5 n, deg(g(x)) 5 m  n. Let

f1 ( x )  f ( x )  an bm1 x nm g ( x ),

where an and bm are leading coefficients of f (x) and g(x), respectively. Let c1
be the leading coefficient of f1(x) and n1 5 deg(f1(x)). Let

f 2 ( x )  f1 ( x )  c1bm1 x n1m g ( x )

and continue the process of obtaining polynomials f0(x) 5 f (x), f1(x), f2(x),
f3(x), … of degrees n0 5 n, n1, n2, n3, … with leading coefficients c0 5 an, c1,
c2, c3, …, respectively. Then,

f r1 ( x )  f r ( x )  cr bm1 x nr m g ( x ) for all r  0

and deg(f (x))  deg(f2(x))  deg(f3(x))  …

At some stage, we should get fs(x) such that deg(fs(x))  m 5 deg(g(x)) and
let fs(x) be first such stage. Then,

f (x )  an bm1 x nm g ( x )  f1 ( x )
 c0 bm1 x n0 m g ( x )  c1bm1 x n1m g ( x )  f 2 ( x )
…

Q001-Algebra-111001_CH 11.indd 18 9/16/2011 10:17:13 AM

 Polynomial Rings   11-19

 s1 
 ∑ bm1ct x nt m  g ( x )  f s ( x )
 t0 
 f (x) 5 q(x)g(x) 1 r(x),
s1
where q(x)   bm1ct x nt m and r ( x )  f s ( x ) and these are the quotient and
t0
remainder, respectively.
The following are an immediate consequences of the division algorithm
(Theorem 11.2.1).

Corollary 11.2.1. Let f (x) and g(x) be polynomials over a commutative ring
R with unity. If g(x) is a nonzero monic polynomial, then there exist unique
q(x) and r(x) in R[x] such that

f (x) 5 q(x)g(x) 1 r(x)

      and either r(x) 5 0 or deg(r(x))  deg(g(x)).

Corollary 11.2.2. Let F be a field and f (x) and g(x)  F[x] with g(x)  0.
Then, there exists unique q(x) and r(x) in R[x] such that

f (x) 5 q(x)g(x) 1 r(x)

      and either r(x) 5 0 or deg(r(x))  deg(g(x)).

In the following, we introduce the concept of an evaluation homomor-

phism which is an important tool in the study of solutions of polynomial
equations.

Theorem 11.2.2. Let S be a commutative ring with unity and R be a subring

of S containing the unity of S. For any a  S, define

fa : R[x] → S

by fa(a0 1 a1x 1 a2x2 1 … 1 anxn) 5 a0 1 a1a 1 a2a2 1 … 1 anan for all

a0 1 a1x 1 … 1anxn  R [x]. Then, fa is a homomorphism of R[x] into S
such that fa(x) 5 a and fa(a) 5 a for all a  R.

Proof: First of all, observe that a  S and a0, a1, …, an  R ⊆ S and hence
a0 1 a1a 1 … 1 anan  S. Also, a0 1 a1x 1 … 1 anxn 5 b0 1 b1x 1 … 1
bnxn implies that ai 5 bi for all 1  i  n and hence a0 1 a1a 1 … 1 anan 5
b0 1 b1a 1 … 1 bnan. Therefore, fa is a well-defined mapping of R[x]
into S.

Q001-Algebra-111001_CH 11.indd 19 9/16/2011 10:17:13 AM

 11-20   Algebra – Abstract and Modern

Now, let f 5 a0 1 a1x 1 … 1 amxm

and g 5 b0 1 b1x 1 … 1 bnxn
be arbitrary elements in R[x], where ai, bi  R.
Put r 5 max{m, n}, ai 5 0 5 bj for i  m and j  n. Then, we have

fa(f 1 g) 5 fa((a0 1 b0) 1 (a1 1 b1)x 1 … 1 (ar 1 br)xr)

5 (a0 1 b0) 1 (a1 1 b1)a 1 … 1 (ar 1 br)an
5 (a0 1 a1a 1 … 1 arar) 1 (b0 1 b1a 1 … 1 brar)
5 (a0 1 a1a 1 … 1 amam) 1 (b0 1 b1a 1 … 1 bnan)
5 fa(f ) 1 fa(g).
r
Also, f ? g 5 c0 1 c1x 1 … 1 cm1nxm1n, where cr   ai bri
Now, fa(f ? g) 5 c0 1 c1a 1 … 1 cm1nam1n
i0

5 (a0 1 a1a 1 … 1 amam) (b0 1 b1a 1 … 1 bnan)

(by the commutativity in S)
5 fa(f ) ? fa(g).

Thus, fa is a homomorphism of R[x] into S. By the very definition of fa, it

follows that fa[x] 5 a and fa[a] 5 a for all a  R. b

Definition 11.2.3. The homomorphism fa defined above is called the evalu-

ation homomorphism at a.
The evaluation homomorphism at a is actually unique with respect to its
defining properties, namely fa(x) 5 a and fa(a) 5 a for all a  R, for the
simple reason that R[x] is generated by R and x; that is, R[x] is the only sub-
ring of R[x] containing R and x. This justifies the notation R[x] for the ring of
polynomials over R.

Theorem 11.2.3. Let S be a commutative ring with unity 1 and R be a subring

of S containing 1. For each a  S, there exists a unique homomorphism f :
R[x] → S such that f (x) 5 a and f (a) 5 a for all a  R.

Proof: Let a  S. We have the existence of the required homomorphism,

namely fa, in Definition 11.2.1. To prove the uniqueness, let f : R[x] → S be
a homomorphism such that f (x) 5 a and f (a) 5 a for all a  R. Then, for
any f 5 a0 1 a1x 1 … 1 anxn in R[x], we have
f (f ) 5 f (a0 1 a1x 1 … 1 anxn)
5 f (a0) 1 f (a1)f (x) 1 … 1 f (an)f (x)n
5 a0 1 a1a 1 … 1 anan
5 fa(f ).
   Thus, f 5 fa. b

Q001-Algebra-111001_CH 11.indd 20 9/16/2011 10:17:14 AM

 Polynomial Rings   11-21

For any a  R and the evaluation homomorphism fa, we often write f (a)
for fa(f ), where f 5 f (x) is a polynomial over R. This is to say that fa(f ) is
just the element of R obtained by substituting a for x in the polynomial f 5
f (x). The following is an important result to which we were exposed at school
level itself and now, we supplement a proof of the most general version of the
remainder theorem.

Theorem 11.2.4 (Remainder Theorem). Let f (x) be a polynomial over a

commutative ring R with unity and a  R. Then, there exists unique q(x) 
R[x] such that

f (x) 5 q(x)(a 2 x) 1 f (a),

where f (a) is the element in R obtained by substituting a for x in f (x); that is,
f (a) 5 a(f (x)).

Proof: Let f (x) 5 a0 1 a1x 1 … 1 anxn. Then,

f (a) 5 a0 1 a1a 1 … 1 anan  R.

Consider the polynomial a 2 x whose leading coefficient is 21, which is a

unit in R. Therefore, by Theorem 11.2.1 (the division algorithm), there exist
unique q(x) and r(x) in R[x] such that

f (x) 5 q(x)(a 2 x) 1 r(x) (*)

where r(x) 5 0 or deg(r(x))  deg(a 2 x) 5 1. Since deg(r(x)) is always a

nonnegative integer, it follows that deg(r(x)) 5 0 and hence r(x) is a constant
polynomial over R. Let r(x) 5 b  R.

f (a) 5 q(a)(a 2 a) 1 b 5 0 1 b 5 b 5 r(x).

Therefore, again by (*),

f (x) 5 q(x)(a 2 x) 1 f (a).

Definition 11.2.4. Let f (x) and g(x) be polynomials over a commutative

ring R. Then, g(x) is said to be a divisor (or a factor) of f (x) if g(x)h(x) 5
f (x) for some h(x)  R[x]; in this case, we also say that f (x) is a multiple
of g(x).

Corollary 11.2.3. Let f (x) be a polynomial over a commutative ring R and

a  R. Then, a 2 x divides f (x) if and only if f (a) 5 0.

Q001-Algebra-111001_CH 11.indd 21 9/16/2011 10:17:14 AM

 11-22   Algebra – Abstract and Modern

Definition 11.2.5. For any polynomial f (x) over R and a  R, we say that a is
a root (or zero) of f (x) if f (a) 5 0.
In other words, a is a root of f (x) if and only if f (x) 5 (a 2 x)g(x) for some
g(x)  R[x]. The next theorem is about the number of roots of a polynomial
in a integral domain.

Theorem 11.2.5. Let f (x) be a nonzero polynomial of degree n over an inte-

gral domain R. Then, f (x) can have at most n distinct roots in R.

Proof: We apply induction on the degree of f (x), If deg(f (x)) 5 0, then the
theorem is trivial, since f (x), being nonzero, cannot have any root in R.
Next, suppose that deg(f (x)) 5 n  0 and assume that any nonzero polyno-
mial of degree m  n can have at most m distinct roots in R. If f (x) has a root
in R; that is, if a  R is such that f (a) 5 0, then, by Theorem 11.2.5,

f (x) 5 (a 2 x)g(x) for some g(x)  R[x].

Now, b is a root of f (x) in R implies that

0 5 f (b) 5 (a 2 b)g(b)

and, since R is an integral domain, it follows that b 5 a or g(b) 5 0. Therefore,

the number of roots of f (x) in R other than a cannot exceed the number of
roots of g(x) in R. Since R is an integral domain, we have

n 5 deg(f (x)) 5 deg(a 2 x) 1 deg(g(x)) 5 1 1 deg(g(x))

and hence deg(g(x)) 5 n 2 1. By the induction hypothesis, it follows that

g(x) can have at most n 2 1 district roots in R. Thus, f (x) can have at most n
distinct roots in R.

Corollary 11.2.4. Let f (x) and g(x) be polynomials of degree n over an inte-
gral domain R and a1, a2, …, an11 be distinct elements of R such that f (ai) 5
g(ai) for 1  i  n. Then, f (x) 5 g(x).

Proof: Consider h(x) 5 f (x) 2 g(x)  R[x]. If h(x)  0, then deg (h(x)) 
n and a1, …, an11 are distinct roots of h(x) in R, which is a contradiction to
Theorem 11.2.5. Therefore, h(x) 5 0 and f (x) 5 g(x).

Corollary 11.2.5. Let f (x) be a polynomial over an integral domain R. Sup-

pose that A is an infinite subset of R such that f (a) 5 0 for all a  A. Then,
f (x) is the zero polynomial.

Q001-Algebra-111001_CH 11.indd 22 9/16/2011 10:17:14 AM

 Polynomial Rings   11-23

Proof: If f (x) is a nonzero polynomial, then deg(f (x)) 5 n and f (x) can have
at most n district roots in R which is a contradiction to the assumption that
each element of the infinite set A is a root of f (x).

Worked Exercise 11.2.1. Find the quotient and remainder when f (x) 5 3 1
4x 1 3x2 1 2x3 1 2x4 1 x5 is divided by g(x) 5 5 1 3x 1 4x2 1 x3 in Z6[x].

Answer: We follow the process of division algorithm as given in Algorithm

11.2.1 (the process of division algorithm). Note that the addition and multi-
plication in Z6 are modulo 6.

3 4 x  3 x 2  2 x 3  2 x 4  x 5 2
5 3x  4 x 2  x 3 x
(5 x 2  3x 3  4 x 4  x 5 )
3 4 x  4 x 2  5 x 3  4 x 4
4x
(2 x  0  4 x 3  4 x 4 )
3 2 x  4 x 2  x 3
1
(5  3 x  4 x 2  x 3 )
4  5x

Therefore, the quotient is 1 1 4x 1 x2 and the remainder is 4 1 5x.

Worked Exercise 11.2.2. Let R be a commutative ring with unity and a  R.

Then prove that R[ x ]/  a  x   R, where a 2 x is the principal ideal
generated by a 2 x in R[x].

Answer: Consider the evaluation homomorphism fa from R[x] into R (by

taking S 5 R in Definition 11.2.1) defined by fa(a0 1 a1x 1 … 1 anxn) 5 a0
1 a1a 1 … 1 anan. If a  R, then, for the constant polynomial a in R[x],we
have fa(a) 5 a. Therefore, fa is an epimorphism and hence, by the funda-
mental theorem of homomorphism,

R[x ]  R.
ker fa

If f  a 2 x 5 (a 2 x)R[x], then f 5 (a 2 x)g for some g  R[x] and

hence

fa(f ) 5 fa(a 2 x)fa(g) 5 (a 2 a)fa(g) 5 0

Q001-Algebra-111001_CH 11.indd 23 9/16/2011 10:17:15 AM

 11-24   Algebra – Abstract and Modern

so that f  ker fa. Using the division algorithm (whose proof is given in the
next section), one can prove that fa(f ) 5 0 implies f 5 (a 2 x)g for some g
 R[x]. Therefore,

ker fa 5 a 2 x

Thus, R[x ]/  a  x  R.

Exercise 11(b)

1. Find the remainder and quotient when f (x) is divided by g (x) in the rings men-
tioned in each of the following:
(i) f (x) 5 2 1 3x 1 x2 1 x4 1 2x5 and g(x) 5 2 1 x2 2 x3 in Z[x].
(ii) f (x) 5 1 1 x 1 x2 1 x4 1 x6 and g(x) 5 1 1 x 1 x2 in Z2[x].
1 2 3 1 1
(iii) f ( x )   x  x 2  x 3 and g ( x )   x in Q[ x ].
2 3 4 3 2
(iv) f (x) 5 2 1 3x 1 4x2 1 5x3 and g(x) 5 1 2 x in Z[x].
(v) f (x) 5 1 1 2x 1 3x2 1 4x3 and g(x) 5 1 1 x2 in Z5[x].
(vi) f (x) 5 (1 1 i) 1 (2 1 3i)x 1 (1 2 2i)x2 1 (1 1 3i)x3 and g(x) 5 i 1
(2 1 i)x 2 2x2 in C[x].
2. Evaluate each of the following for the indicated evaluation homomorphism fa :
Z5[x] → Z5.
(i) f0(2 1 3x 1 4x2 1 x3)
(ii) f1(1 1 x 1 x2 1 x3 1 x4 1 x5)
(iii) f2(1 1 2x 1 3x2 1 4x3 1 x4)
(iv) f3(2 1 3x 1 4x2 1 x5 1 x6)
(v) f4(2 1 x 1 3x2 1 x4 1 x7)
3. Find eight elements in the Kernel of the evaluation homomorphism f5: Q[x] → R.
4. For any subfield F of any field E, prove that the Kernel of the evaluation homo-
morphism fa: F[x] → E is an infinite set for each a [ E.
5. Find all the roots in Z5 of the polynomial 2 1 4x 1 3x2 1 4x3 1 x4 in Z5[x].
6. Prove that 1 1 4x is a unit in Z8[x].
7. Let F be an infinite field and f (x) [ F[x]. Prove that f (a) 5 0 for infinitely many
elements a in F if and only if f (x) 5 0.
8. Let R be an integral domain and f (x) and g(x) [ R[x]. Prove that {a [ R : f (a) 5
g(a)} is infinite if and only if f (x) 5 g(x).
9. Determine all the roots of x 2 x5 in Z5.
10. For any prime number p, prove that every element of Zp is a root of the polyno-
mial x 2 xp in Zp[x].

Q001-Algebra-111001_CH 11.indd 24 9/16/2011 10:17:16 AM

 Polynomial Rings   11-25

11.3 POLYNOMIALS OVER A FIELD

The ring F[x] of polynomials over a field F has a rich structure theory. In par-
ticular, F[x] satisfies most of the ring theoretic properties that are satisfied by
the ring Z of integers. For example, we will be proving that any ideal of F[x]
is generated by a single element, as in the case of Z. Also, we have unique
factorizations in F[x]. More so, we have the division algorithm, as mentioned
in Corollary 11.2.2 and Theorem 11.2.4 (Remainder Theorem). Throughout
our discussions in this section, F always denotes an arbitrary field, unless
otherwise stated.
Let us recall from Theorem 11.2.4 (Remainder Theorem) that, for any
a [ F and 0  f (x) [ F[x], f (a) is the remainder obtained by dividing f (x)
with a 2 x and that a is a root of f (x) if and only if a 2 x divides f (x) (or
a 2 x is a factor of f (x)).

Definition 11.3.1. Let F be a field, f (x) [ F[x] and a [ F. If (a 2 x)n is a

factor of f (x) for some n  1, then a is called a multiple root of f (x), and the
least such n is called the multiplicity of the root a.
The following is a slight variation of Theorem 11.2.5, where we have
proved that any nonzero polynomial of degree n over an integral domain R
can have at most n distinct roots.

Theorem 11.3.1. Any nonzero polynomial of degree n over a field F can

have at most n roots in F, including the multiplicity of the roots.

Proof: Let F be a field, 0  f (x) [ F[x] and deg(f (x)) 5 n. We shall use
induction on n. If n 5 0, there is nothing to prove. If n 5 1, then f (x) 5 a0 1
a1 x and a1  0 and hence a0 a11 is the only root of f (x).
Suppose that n  1 and that the theorem is true for all polynomials of degree
less that n. If f (x) has no roots in F, then we are done. Let a be a root of f (x)
in F. Then,

f (x) 5 (a 2 x)mg(x) for some g(x) [ F[x] and m [ Z1.

Comparing the degrees both the sides, we get that

n 5 deg(f (x)) 5 m 1 deg(g(x))

and hence deg(g(x)) 5 deg(f (x)) 2 m  n 2 m  n.

If f (x) has no roots other than a in F, then we are done, since m  n. On the
other hand, if b  a is a root of f (x) in F, then

0 5 f (b) 5 (a 2 b)mg(b)

Q001-Algebra-111001_CH 11.indd 25 9/16/2011 10:17:16 AM

 11-26   Algebra – Abstract and Modern

and hence g(b) 5 0, so that b is a root of g(x). By the induction hypothesis,

g(x) has at most deg(g(x)) roots in F, including the multiplicity of the roots.
Thus, f (x) has at most n 2 m ( 5 deg(g(x))) roots in F other than a. Therefore,
f (x) has at most m 1 (n 2 m) roots in F, including the multiplicity of the
roots.
Let us recall that an ideal I of a ring R is called a principal ideal if I 5 a
for some a [ R; that is, I is generated by a single element of R. It is well
known that every ideal of the ring Z of integers is a principal ideal. This prop-
erty is satisfied by the rings of polynomials over fields also.

Theorem 11.3.2. Let F be a field and F[x] be the ring of polynomials over F.
Then, F[x] is an integral domain in which every ideal is principal.

Proof: By Theorem 11.1.5, F[x] is an integral domain since F is so. Let I be

an ideal of F[x]. If I 5 {0}, then there is nothing to prove, since I 5 0 5
{0}. Suppose that I is a nonzero ideal; that is, I contains atleast one nonzero
polynomial. Consider the set

S 5 {deg(f (x)) : 0  f (x) [ I}.

Since S is a nonempty set of nonnegative integers, S has a least member, say

n. Then, n 5 deg(f (x)) for some f (x) [ I and n  deg(g(x)) for all 0  g(x) [
I. We shall prove that I is generated by f (x). Since f (x) [ I, we have f (x)
⊆ I. On the other hand, suppose that g(x) [ I. Then, by the division algorithm,
there exist q(x) and r(x) [ F[x] such that

g(x) 5 q(x)f (x) 1 r(x),

where r(x) 5 0 or deg(r(x))  deg(f (x)) 5 n. Now,

r(x) 5 g(x) 2 q(x)f (x) [ I

since g(x) and f (x) [ I and I is an ideal. By the least property of n, it follows
that r(x) 5 0 and hence

g(x) 5 q(x)f (x) [ f (x).

Therefore, I ⊆ f (x). Thus, I 5 f (x).

The converse of the above result is also true in the sense of the following
theorem.

Theorem 11.3.3. Let R be a ring such that R[x] is an integral domain in which
every ideal is principal, Then, R is a field.

Q001-Algebra-111001_CH 11.indd 26 9/16/2011 10:17:16 AM

 Polynomial Rings   11-27

Proof: Since R[x] is given to be an integral domain, it follows from

­Theorem 11.1.5 that R is an integral domain. To prove that R is a field, let
0  a [ R. Consider the ideal a, x generated by a and x in R[x]. By
hypothesis, this ideal must be principal and hence there exists f (x) [ R[x]
such that

a, x 5 f (x).

First of all, note that f (x)  0, since 0  a [ f (x). Since both a and x [
f (x), we get that

a 5 f (x)g(x) (1)
and x 5 f (x)h(x) (2)

for some g(x) and h(x) [ R[x]. From (1), we have

0 5 deg(a) 5 deg(f (x)g(x)) 5 deg(f (x)) 1 deg(g(x))

and hence deg f (x) 5 0. Let f (x) 5 a0 [ R.

Then, from (2), we have x 5 a0 h(x) and hence

1 5 deg(x) 5 deg(a0) 1 deg(h(x)) 5 deg(h(x)).

Therefore, h(x) 5 b0 1 b1x for some b0 and 0  b, in R and hence, from (2),
we have

x 5 a0(b0 1 b1 x).

By comparing the coefficient of x on both sides, we get that

1 5 a0b1 [ f (x) 5 a, x.

Therefore, there exist f1(x) and f2(x) in R[x] such that

1 5 af1(x) 1 xf2(x)

which implies that 1 5 ac0 where c0 is the constant term in f1(x). Thus, a is a
unit in R. Therefore, R is a field.

Worked Exercise 11.3.1. Give an example of an ideal of Z[x] which is not

principal.

Q001-Algebra-111001_CH 11.indd 27 9/16/2011 10:17:16 AM

 11-28   Algebra – Abstract and Modern

Answer: Since Z is not a field, it follows from Definition 13.3.2 that there
must be an ideal of Z[x] which is not principal. Let

I 5 {a0 1 a1x 1 … 1 anxn [ Z[x] : a0 is even}.

Then, one can easily verify that I is an ideal of Z[x]. We prove that I is not a
principal ideal. On the contrary, suppose that

I 5 f (x), f (x) [ Z[x].

Then, 2 [ I 5 f (x) and hence

2 5 f (x)g(x), for some g(x) [ Z[x].

Then, 0 5 deg(2) 5 deg(f (x)g(x)) 5 deg(f (x)) 1 deg(g(x)), so that deg(f (x)) 5
0 5 deg(g(x)). Let f (x) 5 b and g(x) 5 c [ Z. Then, 2 5 bc, we can assume
that b  0 and c  0 (since ,2f (x) 5 f (x)). Then, b 5 1 or 2.
But b  1, since 1  I. Therefore, b 5 2. This implies that I 5 2, which
is a contradiction, since 2 1 x [ I and 2 1 x  2. Thus, I  f (x) for
any f (x) [ Z[x]. That is, I is not a principal ideal in Z[x].

Worked Exercise 11.3.2. Let F be a field. For any a [ F, let

Ma 5 {f (x) [ F[x] : f (a) 5 0}.

Then, prove that Ma is a maximal ideal of F[x] and that F[x]/Ma  F. If F is

infinite, prove that ∩ M a  {0}.
a∈F

Answer: Let a [ F. Consider the evaluation homomorphism fa : F[x] → F

defined by fa(f (x)) 5 f (a). Then, fa is an epimorphism (for, if r [ F, then
r [ F[x] and fa(r) 5 r). Also,

ker fa 5 {f (x) [ F[x] : fa(f (x)) 5 0}

5 {f (x) [ F[x] : f (a) 5 0}
5 Ma.

By the fundamental theorem of homomorphisms, Ma is an ideal of F[x] and

F[x]/Ma  F. Since F is a field, so is F[x]/Ma and hence Ma is a maximal ideal
of F[x]. Further, suppose that F is infinite. If f ( x )  ∩ M a , then f (a) 5 0 for
a∈F
infinitely many a and therefore, by Corollary 11.2.5, f (x) 5 0. Thus,

∩M
a∈F
a  {0}.

Q001-Algebra-111001_CH 11.indd 28 9/16/2011 10:17:17 AM

 Polynomial Rings   11-29

Worked Exercise 11.3.3. Let F be a field and Ma be as defined in Worked

Exercise 11.3.2 above for any a [ F. Then prove that

∩M
a∈F
a  {0} ⇔ F is infinite.

Answer: Suppose that F is finite and |F| 5 n. Consider the group F* 5

F 2 {0} under the multiplication in F. Then, F* is a finite group of order
n 2 1 and hence

an21 5 1 for all a [ F*.

Therefore, an 5 a for all a [ F. Put f (x) 5 x 2 xn. Then, 0  f (x) [ F[x] and
f (a) 5 0 for all a [ F. Therefore, 0  f (x) [ ∩ M a . Thus, ∩ M a  {0}.
a∈F a∈F
Converse of this is proved in Worked Exercise 11.3.2.

Worked Exercise 11.3.4. Let R be an integral domain. Prove that R is a field

if and only if x is a maximal ideal of R[x].

Answer: Suppose that R is a field. For any a [ R. Let

Ma 5 {f (x) [ R[x] : f (a) 5 0}.

Then, by Worked Exercise 11.3.2, Ma is a maximal ideal of R[x]. In particular,

M0 is a maximal ideal of R[x]. For any f (x) [ R[x], we have

f (x) [ M0 ⇔ f (0) 5 0
⇔ x divides f (x) (by Corollary 11.2.3)
⇔ f (x) [ x.

Therefore, M0 5 x and hence x is a maximal ideal of R[x]. Con-

versely, suppose that x is a maximal ideal of R[x]. Since x 5 M0 5 ker
f0, where f0 is the evaluation homomorphism at 0, it follows that

R[x]/x  R.

since x is a maximal ideal of R[x], R[x]/x and hence R is a field.

We close this section with a remark that several other properties of rings
of polynomials over fields will be discussed in the next section and in the
next chapter.

Q001-Algebra-111001_CH 11.indd 29 9/16/2011 10:17:17 AM

 11-30   Algebra – Abstract and Modern

Exercise 11(c)

1. Let F be field and f (x) be a polynomial of degree n (n  0) over F. Then prove

that the quotient ring f[x]/f (x) is bijective with Fn.

2. Let R be a commutative ring with unity in which every ideal is principal. Then
prove that an ideal in R is maximal if and only if it is prime.

3. Let F be a field, 0  a [ F and f (x) [ F[x], prove that f (x) 5 af (x).

4. Let F be a subfield of a field E and a [ E. Then prove that {f [ F[x] : a is a root

of f (x)} is an ideal of F[x] and describe a generator of this ideal.

5. Let F be a field and I be the set of all polynomials over F for each of which the
sum of the coefficients is zero. Then prove that I is an ideal of F[x] and determine
a generator of I.

6. Prove that there are infinitely many polynomials f (x) in Z3[x] for each of which
every element of Z3 is a root.

7. Prove that the rings Q[ 2] and Q[ x ]/ 2  x 2 are isomorphic.

8. Let p be a prime number. Then prove that (x 2 1)(x 2 2)(x 2 3) … (x 2 (p 2 1)) 5

xp21 2 1 in Zp[x].

9. Let f (x) [ R[x], a [ R and f 9(x) be the derivative of f with respect to x. Then
prove that f (a) 5 0 5 f 9(a) if and only if (a 2 x)2 divides f (x).

10. In Z[x], prove that x is a prime ideal but not a maximal ideal.

11. Let P be a prime ideal of a commutative ring R with unity. Then prove that P[x]
is a prime ideal of R[x]. If M is a maximal ideal of R, is M[x] a maximal ideal
of R[x]?

12. Prove that R[x]/1 1 x2 is isomorphic to the field of complex numbers.

13. Prove that Z[x]/1 1 x2 is isomorphic with the ring of Gaussian integers
Z[i].

14. Let R be an integral domain. Then prove that the set

I 5 {f (x) [ R[x] : f (x) 50 or deg(f (x)) . 0}

is an ideal of the ring R[x]. Also, for any n . 0, prove that

In5 {f (x) [ R[x] : f (x) 50 or deg(f (x))  n}.
∞
15. In Exercise 14 above, determine ∩ I n .
n1

16. Let f (x) [ Z[x] be a monic polynomial and a be a rational number such that
f (a) 5 0. Then prove that a must be integer.

Q001-Algebra-111001_CH 11.indd 30 9/16/2011 10:17:18 AM

 Polynomial Rings   11-31

11.4 IRREDUCIBLE POLYNOMIALS

Several interesting questions on factorization of polynomials are based on the
idea of irreducibility. In this section, we discuss irreducible polynomials over
mainly integral domains.

Definition 11.4.1. Let R be an integral domain. A polynomial f (x) of

positive degree over R is said to be irreducible over R if f (x) cannot be
expressed as a product of two polynomials of positive degree over R. That
is, f (x) [ R[x] is said to be irreducible over R if deg(f (x))  0 and, for any
g(x) and h(x) [ R[x].

f (x) 5 g(x)h(x) ⇒ deg(g(x)) 5 0 or deg(h(x)) 5 0.

If f (x) is not irreducible and deg(f (x))  0, we say that f (x) is reducible.
Note that the above definition applies only to polynomials of positive
degree and as such the constant polynomials are neither reducible nor
irreducible. Also, the irreducibility of a polynomial f (x) [ R[x] depends
much on the integral domain R; that is, a given polynomial may be irre-
ducible when viewed as a polynomial over one domain, yet reducible when
viewed as a polynomial over another domain. For, consider the following
­example.

Example 11.4.1
1. The polynomial 1 1 x2 is irreducible over R the field R of real numbers;
but it is reducible over the field C of complex numbers, since 1 1 x2 5
(1 1 ix)(1 2 ix) and 1 1 ix and 1 2 ix [ C[x].
2. 1 1 x2 is reducible in Z2[x], since 1 1 x2 5 (1 1 x)(1 1 x) in Z2[x]; but
1 1 x2 is irreducible in Z3[x].
Thus, to ask merely whether a polynomial is irreducible, without specifying
the coefficient ring involved, is incomplete and meaningless. More often, it
is a formidable task to decide when a given polynomial is irreducible over
a specific ring. The following provide certain simple tips in finding a given
polynomial to be irreducible over a given field or an Integral Domain.

Theorem 11.4.1
1. Let R be an integral domain and f (x) [ R[x] with deg(f (x)) 5 1. Then,
f (x) is irreducible over R.
2. Let F be a field and f (x) [ F[x] with deg(f (x)) 5 2 or 3. Then, f (x) is
irreducible if and only if f (x) has no root in F.

Q001-Algebra-111001_CH 11.indd 31 9/16/2011 10:17:18 AM

 11-32   Algebra – Abstract and Modern

Proof:
1. Recall that the degree of any nonzero polynomial is a nonnegative inte-
ger. If f (x) 5 g(x)h(x) and g(x);h(x) [ R[x], then
1 5 deg(f (x)) 5 deg(g(x)) 1 deg(h(x))
and hence deg(g(x)) 5 0 or deg(h(x)) 5 0. Therefore, f (x) cannot be
expressed as a product of two polynomials of positive degree. Thus, f (x)
is irreducible over R.
2. We shall prove that f (x) is reducible over F if and only if f (x) has a root
in F. First note that any polynomial a0 1 a1x, a1  0 of degree over F
one has a root, namely, a11a0 in F. Suppose that f (x) is reducible over
F. Then,

f (x) 5 g(x)h(x) for some g(x) and h(x) [ F[x]

with deg(g(x))  0 and deg(h(x))  0. Then, either deg(g(x)) 5 1 or
deg(h(x)) 5 1 (otherwise, if deg(g(x)) $ 2 and deg(h(x)) $ 2, then
deg(f (x)) 5 deg(g(x)) 1 deg(h(x)) $ 4, which is a contradiction to the
hypothesis that deg(f (x)) 5 2 or 3). If deg(g(x)) 5 1, then, by (1) above,
g(x) has a root in F and hence f (x) has a root in F. Similarly, if deg(h
(x)) 5 1, then h(x) and hence f (x) has a root in F.
Conversely, suppose that f (x) has a root in F. Let a be a root of f (x) in F.
Then, by Corollary 11.2.3, a 2 x divides f (x). Therefore, there exists
g(x) [ F[x] such that f (x) 5 (a 2 x)g(x). Since deg f (x) 5 2 or 3, it follows
that deg(g(x))  0 and hence f (x) is reducible over F.

Theorem 11.4.2. Let f (x) be a nonzero polynomial over a field F. Then, the
following are equivalent
1. f (x) is irreducible over F.
2. f (x) is a maximal ideal of F[x].
3. f (x) is a prime ideal of F[x].

Proof: (1) ⇒ (2): Suppose that f (x) is irreducible over F. Then, f (x) is not
a constant polynomial and hence f (x) is not a unit in F[x], so that f (x)
is a proper ideal of F[x]. Let I be any ideal of F[x] containing f (x). By
Theorem 11.3.2, I 5 g(x) for some g(x) [ F[x]. Since f (x) ⊆ I 5
g(x), we get that f (x) 5 g(x)h(x) for some h(x) [ F[x]. Since f (x) is irre-
ducible, it follows that either g(x) or h(x) is a constant, If g(x) is a constant,
then g(x) is a unit (note that f (x)  0 and h(x)  0, since deg(f (x))  0 and
hence f (x)  0) so that I 5 g(x) 5 F[x]. If h(x) is a constant, then h(x) is

Q001-Algebra-111001_CH 11.indd 32 9/16/2011 10:17:18 AM

 Polynomial Rings   11-33

a unit in F[x] and g(x) 5 f (x)h(x)21 [ f (x) so that I ⊆ f (x) and hence
I 5 f (x). Thus, f (x) is a maximal ideal of F[x].
(2) ⇒ (3): This is trivial, since any maximal ideal of a commutative ring with
unity is a prime ideal.
(3) ⇒ (1): Suppose that f (x) is a prime ideal of F[x]. Then, f (x) is a
proper ideal and hence f (x) is not a unit. Therefore, deg(f (x))  0. For any
g(x) and h(x)  F[x],

f (x) 5 g(x)h(x) ⇒ g(x)h(x)  ,f (x).

⇒ g(x)  ,f (x). or h(x)  ,f (x).
⇒ g (x) 5 f (x)g1(x) or h(x) 5 f (x)h1(x) for some
g1(x) and h1(x)  F[x]
⇒ f (x) 5 f (x)g1(x)h(x) or f (x) 5 g(x)f (x)h1(x)
⇒ g1(x)h(x) 5 1 or 5 g(x)h1(x) 5 1
⇒ h(x) or g(x) is a unit
⇒ deg(h(x)) 5 0 or deg (g(x)) 5 0.

Thus, f (x) is irreducible over F[x].

When we deal with polynomials over the field of complex numbers, the
crucial tool is the fundamental theorem of algebra. There are several proofs
of this theorem but none of these come under the topics covered in this book
and hence proof is omitted here and the reader can simply assume the validity
of the following theorem.

Theorem 11.4.3 (Fundamental Theorem of Algebra). Any nonconstant

polynomial over the field C of complex numbers has atleast one root in C.

Corollary 11.4.1. Let f (x)  C[x] be a polynomial of degree n  0. Then,

f (x) can be expressed as a product of n (not necessarily distinct) polynomials,
each of degree one.

Proof: This follows from Theorem 11.4.3 (fundamental theorem of algebra)

and from the fact that a is a root of f (x) if and only if a 2 x is a factor of f (x)
and also by using induction on the deg(f (x)).

Corollary 11.4.2. The only irreducible polynomials over C are those of

degree one. Since the field R of real numbers can be considered as a sub-
field of the field C of complex numbers, R[x] can be considered as a subring
of C[x]. Therefore, any nonzero polynomial over R can be considered as a
polynomial over C and hence has a root in C. This observation leads to the
following corollary.

Q001-Algebra-111001_CH 11.indd 33 9/16/2011 10:17:19 AM

 11-34   Algebra – Abstract and Modern

Corollary 11.4.3. Let f (x) be a nonconstant monic polynomial over the real
number system R. Then, f (x) can be expressed as a product of polynomials
over R, each of degree 2 or 1.

Proof: Let deg(f (x)) 5 n. Since f (x)  R[x] ⊆ C[x], it follows from Corollary
11.4.1 that

f (x) 5 g1(x)g2(x) … gn(x),

where gj(x) is a polynomial of degree one over C, for each 1 # j # n. Since

f (x) is monic, we can assume that each gj(x) is monic (the leading coefficient
is a unit) and hence

gj(x) 5 s 2 x, s  C.

If s  R, then gj(x)  R[x]. If s ∉ R, then s 5 a 1 bi, where a and b  R

and b ≠ 0. Note that s is a root of f (x) (since gj(x) 5 s 2 x is a divisor
of f (x)).
Also, since f (x)  R[x], all the coefficients of the f (x) are real numbers.
Now, consider the complex conjugate s  a  bi of s.
If f (x) 5 a0 1 a1x 1 … 1 anxn, ai  R, then
f (s )  a0  a1 s  an s n
 a0  a1s  an s n

 f (s)  0  0.

and hence s is a root of f (x), so that s  x is also a factor of f (x), so that

s  x 5 gk(x) for some 1 # k # n. Now,

g j ( x ) g k ( x )  ( s  x )( s  x )
5 (a 1 bi 2x)(a 2 bi 2 x)
5 (a2 1 b2) 2 (a 1 bi 1 a 2 bi)x 1 x2
5 a2 1 b2 2 2ax 1 x2  R[x].

Therefore, gj(x)gk(x) is a factor of f (x) and is a polynomial of degree 2 over R.

Thus, f (x) is a product of polynomials over R, each of degree 1 or 2.
In the next chapter, we prove some more important properties of irreduc-
ible polynomials and, in particular, we prove the Eisenstein criterion to find
the irreducibility of certain polynomials. b

Worked Exercise 11.4.1. Prove that the polynomial f (x) 5 1 1 x 1 x3 is

irreducible over Z2.

Q001-Algebra-111001_CH 11.indd 34 9/16/2011 10:17:19 AM

 Polynomial Rings   11-35

Answer: If there are any factors f (x), then atleast one factor must be of degree
one, say a 1 bx (with a and b  Z2 and b ≠ 0). In this case, 2b21a is a root of
a 1 bx and hence of f (x). Therefore, if f (x) is reducible, then f (x) must have
a root in Z2; but it can be easily verified that f (0) 5 1 5 f (1) (sense 1 1 1 5
0 in Z2) . Therefore, f (x) is irreducible over Z2.

Worked Exercise 11.4.2. Let f (x)  R[x] and s be a complex number. Then,
prove that s is a root of f (x), if and only if s is a root of f (x), where s is the
complex conjugate of s.

Answer: Let s 5 a 1 bi, where a and b  R. Then, s 5 a 2 bi. Let

f (x) 5 a0 1 a1x 1 … 1 anxn,

where a0, a1, …, an  R. Note that r is a real number if and only if r  r.

Now, suppose that s is a root of f (x). Then, f (s) 5 0 and

f (s )  a0  a1 s  an (s ) n
a0  as  a n s n

a0  a1s  an s n
 f (s)  0  0.

Thus, s is a root of f (x). The converse follows from the fact that s 5 s.

Exercise 11(d)

1. Which of the following are true?

(i) The degree of any irreducible polynomial is positive.
(ii) The degree of any reducible polynomial is greater that one.
(iii) 1 1 x 1 x2 1 x3 is irreducible over Z2.
(iv) 1 1 x 1 x2 1 x3 is reducible over Z3.
(v) 3 2 x2 is irreducible over Q.
(vi) The degree of any irreducible polynomial is less that 4.
(vii) The number of irreducible polynomial in Z2[x] is finite.
(viii) ,x. is a maximal ideal in R[x].
(ix) ,x. is a maximal ideal in Z[x].
(x) ,x. is a prime ideal in Z[x].

2. Let f (x) be a nonconstant polynomial over a field F. Then, prove that f (x) is
irreducible if and only if F [ x ]/  f ( x )  is a field.

Q001-Algebra-111001_CH 11.indd 35 9/16/2011 10:17:20 AM

 11-36   Algebra – Abstract and Modern

3. Prove that Z2[x]/1 1 x 1 x2 is a field and determine the number of elements in it.

4. Give an example of a field with exactly nine elements.

5. Let p be a prime number and f (x) be an irreducible polynomial of degree n over

Zp, then prove that Z p [ x ]/  f (x )  has exactly pn elements.

6. Let F be a field, f (x)  F[x] and 0 ≠ a  F. Then prove that f (x) is irreducible
over F if and only if so is f (x).

7. Is 1 1 x 1 x2 1 x3 reducible over Z5[x]?

8. Prove that Theorem 11.4.1 (2) fails if deg(f (x)) $ 4.

9. Prove that 1 1 x 1 x3 1 x4 is reducible over any field F.

10. Let R[x] be the ring of polynomials over a commutative ring R with unity. Then,
the ring of polynomials over R[x] will be denoted by R[x, y]; that is, R[x, y] 5
R[x][y]. By induction, we define
R[x1 x2, …, xn] 5 R[x1 …, xn21][xn]

for any n  1. This is called the ring of polynomials in n indeterminates over R.

Prove that R[x1 x2, …, xn] is an integral domain if and only if so is R.

11. Prove that there is an ideal in R[x, y] which is not principal.

12. For any field F, prove that

F [ x, y ]/  x  y   F[x]  F[y].
13. For any field F, prove that the ideal < x, y > generated by {x, y} in F[x, y] is a
maximal ideal of F[x, y].

14. Prove that Z4[x] has infinitely many units and infinitely many nilpotent elements.

Q001-Algebra-111001_CH 11.indd 36 9/16/2011 10:17:21 AM

 12
Factorization in Integral
Domains
12.1 Divisibility in Integral Domains
12.2 Principal Ideal Domains
12.3 Unique Factorization Domains
12.4 Polynomials over UFDs
12.5 Euclidean Domains
12.6 Some Applications to Number Theory

This chapter is concerned with the problem of factoring elements of an integral

domain. The motivation for this lies in the ring Z of integers, where the Funda-
mental Theorem of Arithmetic states that every integer n > 1 can be written, in
an essentially unique way, as a product of prime numbers; for example,

6,300 5 2 3 2 3 3 3 3 3 5 3 5 3 7

and 2, 3, 5 and 7 are prime numbers. In this chapter, we extend the factoriza-
tion theory of the ring Z and, in particular, the above-mentioned Fundamental
Theorem of Arithmetic, to a more general setting. Naturally, any reasonable
abstraction of these number theoretic ideals depends on a suitable interpreta-
tion of prime elements (the building blocks for the study of divisibility prob-
lems in Z). All the topics discussed in this chapter are more concerned with
integral domains. We proceed from the most general results about divisibility,
prime elements and factorization to stronger results concerning certain spe-
cific classes of integral domains.
First, let us recall that an integral domain is a nontrivial commutative ring
with unity and without zero divisors (or equivalently, product of two nonzero
elements is again nonzero).

Q001-Algebra-111001_CH 12.indd 1 9/16/2011 10:16:08 AM

 12-2   Algebra – Abstract and Modern

12.1 DIVISIBILITY IN INTEGRAL DOMAINS

In this section, we extend the concepts of divisibility, greatest common divisor
(g.c.d.), least common multiple (l.c.m.) and primes in the ring Z of integers to
arbitrary integral domains. Let us begin with the following definition.

Definition 12.1.1. Let R be an integral domain and a and b  R. If a 5 bu

for some unit u in R, then we say that a is an associate of b and denote this
by a ~ b.
Since a 5 a1, it follows that a ~ a for each a  R. Also, if a ~ b, then a 5 bu
for some unit u in R and hence b 5 au21 so that b ~ a. Further, if a 5 bu and b 5
cv for some units u and v in R, then uv is a unit in R and a 5 c(uv) and hence
a ~ c. All these arguments say that ~ is an equivalence relation. If ã stands for
the set of all associates of a, then the following can be easily proved.
1. 0 5 {0}
2. ã 5 {au : u is a unit in R}, for any a  R.
3. 1 5 The set of all units in R.
4. ã 5 b if and only if a ~ b.
5. ã’s form a partition of R.

Example 12.1.1
1. In the ring Z of integers, 1 and 21 are the only units and hence a ~ b if
and only if |a| 5 |b|, for any a and b  Z, where |a| is the absolute value
of a.
2. In a field F, each nonzero element is a unit and hence a ~ b (since a 5
b(b21a)) for any nonzero elements a and b in F. Therefore, ã 5 R 2 {0}
for any 0  a  R.
3. Consider the ring Z[i] of Gaussian integers in which 1, 21, i and 2i are
the only units. For any x 5 a 1 bi  Z[i],
x 5 {x, 2x, ix, 2ix}
5 {a 1 bi, 2a 2 bi, 2b 1 ai, b 2 ai}.
4. Let R[x] be the ring of polynomials over an integral domain R. Then, the
units of R[x] are precisely the units in R. For any f (x)  R[x],

f (x ) = {uf ( x ) : u is a unit in R}.

Definition 12.1.2. Let a and b be any elements of an integral domain R. If

there exists x  R such that ax 5 b, then we say that a divides b (or a is a
divisor of b or b is a multiple of a) and denote this by a|b.

Q001-Algebra-111001_CH 12.indd 2 9/16/2011 10:16:09 AM

 Factorization in Integral Domains   12-3

Virtually all statements about divisibility can be phrased in terms of prin-

cipal ideals. Let us recall that, for any element a in an integral domain R, the
principal ideal generated by a in R is given by

<a> 5 aR 5 {ar : r  R}.

Theorem 12.1.1. The following holds for any elements a, b and c in an

integral domain R.

1. a|b ⇔ b  <a> ⇔ <b> ⊆ <a>

2. a|b and b|a ⇔ a ~ b ⇔ <a> 5 <b>
3. a|0
4. a|1 ⇔ a is a unit in R ⇔ <a> 5 R

Proof:
1. a|b ⇔ ax 5 b for some x  R
⇔ b  aR 5 <a>
⇔ <b> ⊆ <a>.

2. By (1), a|b and b|a ⇔ <a> 5 <b>.

Suppose that a ~ b. Then, a 5 bu and au21 5 b for some unit u in R and
hence a|b and b|a. Conversely, suppose that a|b and b|a. Then, ax 5 b
and by 5 a for some x and y  R and hence
a 5 by 5 (ax)y 5 a(xy).
First note that a 5 0 if and only if b 5 0. Now, if a  0, then xy 5 1
(since a 5 a(xy) and R is an integral domain) and hence x and y are units.
Since ux 5 b and x is a unit, we get that a ~ b.
3. a0 5 0 and hence a|0.
4. a|1 ⇔ ax 5 1 for some x  R.
⇔ a is a unit in R
⇔ <a> 5 R.
When a|b, we often use the phrases ‘a is a factor of b’ or ‘b is divisible
by a’ or ‘b is a multiple of a’. When a does not divide b, we write a  b. The
next result is a routine verification and hence its proof is left as an exercise
to the reader.

Theorem 12.1.2. The following holds for any elements a, b and c of an

integral domain R.

Q001-Algebra-111001_CH 12.indd 3 9/16/2011 10:16:09 AM

 12-4   Algebra – Abstract and Modern

1. a|a and 1|a

2. a|b and b|c ⇒ a|c
3. a|b ⇒ ac|bc; the converse holds if c  0.
4. a|c and a|b ⇒ a|cx 1 by for all x and y  R.
In the following, we introduce the notion of the g.c.d. for a given finite set
of nonzero elements in an integral domain.

Definition 12.1.3. Let a1, a2, …, an be nonzero elements in an integral domain

R. An element d  R is called a g.c.d. of a1, a2, …, an if the following are
satisfied.
(i) d|ai for all 1  i  n.
(ii) If c  R and c|ai for all 1  i  n, then c|d.
The use of the superlative adjective ‘greatest’ in the above definition does
not imply that d has greatest magnitude than any other common divisor c of
a1, a2, …, an; but only that d is a multiple of any such c. A natural question
that arises is whether the elements a1, a2, …, an can possess two different
g.c.d.s. The answer is affirmative; for, in the ring Z of integers, both 2 and
22 are g.c.d.s of 6 and 10, as per the above definition. However, 2 and 22
are associates to each other. The same is true in a general integral domain. If
d and d9 are both g.c.d.s of a1, a2, …, an, then, by (ii) above, d|d9 and d9|d and
hence d and d9 are associates to each other. Thus, the g.c.d. of a1, a2, …, an is
unique up to associates, whenever it exists and is usually denoted by (a1, a2,
…, an). The following theorem deals with the existence of g.c.d.

Theorem 12.1.3. Let a1, a2, …, an be any nonzero elements in an integral

domain R. Then, a1, a2, …, an have g.c.d. d expressible in the form

d 5 a1r1 1 a2r2 1  1 anrn (ri  R)

if and only if the ideal <a1, a2, …, an> generated by the set {a1, a2, …, an} in
R is the principal ideal.

Proof: Suppose that d 5 a1r1 1 a2r2 1…1 anrn (ri  R) is a g.c.d. of a1, a2,
…, an in R. Then, d|ai and hence ai  <d> for all 1  i  n. Therefore,

<a1, a2, …, an> ⊆ <d>,

where <a1, a2, …, an> is the ideal generated by a1, a2, …, an in R. Also, since

d 5 a1r1 1 a2r2 1 … 1 anrn  <a1, a2, …, an>,

Q001-Algebra-111001_CH 12.indd 4 9/16/2011 10:16:09 AM

 Factorization in Integral Domains   12-5

we get that <d> ⊆ <a1, a2, …, an>. Thus,

<a1, a2, …, an> 5 <d>.

Conversely, suppose that <a1, a2, …, an> 5 <d> for some d  R. Then, ai
<d> and hence d|ai for all 1  i  n. If c is a common divisor of a1, a2, …,
an, then c|ai and hence ai  <c> for all 1  i  n, so that

<d> 5 <a1, a2, …, an> ⊆ <c>.

Therefore, c|d. Thus, d is a g.c.d. of a1, a2, …, an. Also, since

d  <d> 5 <a1, a2, …, an> 5 a1R 1 a2R 1…1 anR,

it follows that d 5 a1r1 1 a2r2 1…1 anrn for some r1, r2, …, rn  R.
It is well known that the ring Z of integers is an integral domain in which
every ideal is a principal ideal. This together with the above theorem implies
the following corollary.

Corollary 12.1.1. Any nonzero integers a1, a2, …, an have g.c.d. and

g.c.d. {a1, a2, …, an} 5 a1r1 1 a2r2 1…1 anrn

for some integers r1, r2, …, rn.

Dual to the notation of g.c.d., we have the concept of l.c.m. which is
defined in the following definition.

Definition 12.1.4. Let a1 a2, …, an be any nonzero elements in an integral

domain R. An element d  R is called l.c.m. of a1 a2, …, an if the following
are satisfied.
(i) ai|d for all 1  i  n.
(ii) If c  R and ai|c for all 1  i  n, then d|c.
In other words, a common multiple of a1 a2, …, an is called l.c.m. if it
divides any other common multiple. Note that a l.c.m., if it exists, is unique
apart from the distinction between associates, for, if d and d9 are l.c.m.’s of a1
a2, …, an in R, then by (ii) above, d| d9 and d9|d and hence d and d9 are associ-
ates to each other.

Theorem 12.1.4. For any nonzero elements a1 a2, …, an in an integral domain

n
R, a1, a2, …, an have l.c.m. if and only if the ideal ∩  ai  is principal.
i1

Q001-Algebra-111001_CH 12.indd 5 9/16/2011 10:16:10 AM

 12-6   Algebra – Abstract and Modern

Proof: This follows from the definition of l.c.m. and fromn

the fact that, for
any a and b  R, a|b if and only if <b> ⊆ <a>. Note that ∩  ai  is the larg-
i1
est ideal contained in each of <ai>.
Next, we introduce two new classes of elements, namely prime and irre-
ducible elements in an arbitrary integral domain. When we consider the ring
Z of integers, these two concepts become equivalent and yield the usual
notion of a prime number.

Definition 12.1.5. Let p be a nonzero and nonunit element in an integral

domain R. Then,
1. p is called a prime element if, for any a and b  R,
p|ab ⇒ p|a or p|b.

2. p is called an irreducible element if, for any a and b  R,

p 5 ab ⇒ a is a unit or b is a unit.
In other words, a nonzero and nonunit element p is called irreducible
if it cannot be factored in R in a nontrivial way; that is, the only factors of
p are its associates and units in R. Note that any unit u is a factor of every
element, since u(u21a) 5 a. In fields, where each nonzero element is a
unit, the concepts of prime elements and irreducible elements are of no
significance.

Theorem 12.1.5. Let R be an integral domain. Then, every prime element in

R is irreducible. The converse is false.

Proof: Let p be a prime element in R. Then, p is nonzero and nonunit. To

prove the irreducibility of p, let a and b  R such that p 5 ab. Then, p|ab and,
since p is prime, p|a or p|b. Now,

p|a ⇒ ps 5 a for some s  R

⇒ abs 5 a
⇒ bs 5 1 (since p  0 and hence a  0)
⇒ b is a unit in R.

Similarly, p|b ⇒ a is a unit in R. Thus, either a or b is a unit in R. Therefore,

p is irreducible. The converse fails; for, consider the following example in
which we exhibit an irreducible element which is not prime.

Example 12.1.2. Let Z [ 5 ] 5 {a 1 b 5 ; a and b  Z}. Then, Z [ 5 ]

is an integral domain under the usual addition and multiplication of complex

Q001-Algebra-111001_CH 12.indd 6 9/16/2011 10:16:10 AM

 Factorization in Integral Domains   12-7

numbers. For any x  a  b 5, let |x| be the usual modulus of the complex
number x  a  ib 5; that is,

|x| = a 2 + 5b 2 .

The following can be easily verified for any x and y in Z[ 5].

1. |xy| 5 |x||y|
2. |x| 5 0 ⇔ x 5 0
3. x is a unit in Z [ 5 ] ⇔ |x| 5 1 ⇔ x 5 1.
Now, we shall prove that 2  5 is irreducible, but not prime in [ 5].
Let p  2  5. Then, |p| 5 9 5 3.
Suppose that x and y  Z [ 5 ] such that p 5 xy.
Then, 3 5 |p| 5 |xy| 5 |x||y| and hence |x| 5 1 or |y| 5 1 so that x or y is a unit
in Z[ 5].
Thus, p is irreducible. Now, consider

3 3  9  ( 2  5 )( 2  5 )  p( 2  5 ).

Therefore, p divides 3 3 3. But p does not divide 3, since we cannot find

integers a and b such that ( 2 [ 5 ])( a  b 5 )  3. Thus, p is not prime
in Z[ 5].
Recall that we have introduced the notion of an irreducible polynomial over an
integral domain R as a nonconstant polynomial over R which cannot be expressed
as a product of two nonconstant polynomials. In the integral domain R[x], an
irreducible element may be a constant and hence not an irreducible polynomial.
However, if F is a field, then the ring F[x] of polynomials over F is an integral
domain in which units are precisely nonzero constant polynomials and hence
irreducible polynomials over F and irreducible elements in F[x] are same.
In the following, we establish a relation between primeness (irreducibility)
of an element p and the primeness (maximality) of the ideal <p> generated
by p. First, let us define a principal ideal of R to be a maximal principal ideal
if it is maximal (with respect to the inclusion relation) in the set of proper
principal ideals of R.

Theorem 12.1.6. The following holds for any nonzero and nonunit p in an
integral domain R.
1. p is a prime element in R if and only if <p> is a prime ideal of R.
2. p is an irreducible element in R if and only if <p> is a maximal principal
ideal of R.

Q001-Algebra-111001_CH 12.indd 7 9/16/2011 10:16:12 AM

 12-8   Algebra – Abstract and Modern

Proof:
1. is trial because of the fact that x  <p> if and only if p divides x.
2. Suppose that p is an irreducible element in R and <p> ⊆ <x>, x  R.
Then, <p>  R (since p is a nonunit) and p  <x> and hence p 5 xy for
some y  R. Since p is irreducible, either x or y is a unit in R. If x is a
unit, then <x> 5 R. If y is a unit, then x 5 py21  <p> and hence <x> ⊆
<p>, so that <x> 5 <p>. Thus, <p> is a maximal principal ideal of R.
Conversely, suppose that <p> is a maximal principal ideal of R. Then, <p>  R
and hence p is not a unit. Suppose a and b  R such that p 5 ab. Then, p 
<a> and hence <p> ⊆ <a>. By the maximality of <p>, either <p> 5 <a> or
<a> 5 R. If <a> 5 R, then a is a unit. If <p> 5 <a>, then a 5 pc for some
c  R and hence

p 5 ab 5 (pc)b 5 p(cb)

since R is an integral domain and p  0, it follows that 1 5 cb and hence b is

a unit. Therefore, in any case, either a or b is a unit. Thus, p is an irreducible
element in R. b

Worked Exercise 12.1.1. Let p and q be associates to each other in an integral

domain R. Then, prove the following:
1. p is prime if and only if q is prime.
2. p is irreducible if and only if q is irreducible.

Answer: Since p ~ q, we have p 5 qu for some unit u in R and hence q 5

pu21. Therefore, p 5 0 if and only if q 5 0 and p is a unit if and only if q is a
unit. Also, note that p|q and q|p.
1. Suppose that p is prime. Then, p is nonzero and nonunit and hence so is
q. For any a and b  R,
q|ab ⇒ p|ab (since p|q)
⇒ p|a or p|b (since p is prime)
⇒ q|a or q|b (since q|p).

Thus, q is prime. Converse follows from the fact that p ~ q if and only
if q ~ p.
2. Suppose that p is irreducible. Then, p is nonzero and nonunit and hence
so is q. Let a and b  R such that q 5 ab. Then, p 5 qu 5 abu. Since p
is irreducible, either a or bu is a unit. Therefore, a or b is a unit. Thus, q
is irreducible.

Q001-Algebra-111001_CH 12.indd 8 9/16/2011 10:16:12 AM

 Factorization in Integral Domains   12-9

Worked Exercise 12.1.2. Prove that an integral domain is a field if and only
if there are exactly two associate classes.

Answer: Let R be an integral domain. If R is a field, then every nonzero ele-

ment of R is a unit and hence a ~ 1 for all a  0 in R, so that 0 and 1 (5 R 2
{0}) are the only associate classes in R. Conversely, suppose that there are
exactly two associate classes in R. Since 0 5 {0}, we get that ã 5 R 2 {0}
for any 0  a  R. In particular, 1 ~ a for all 0  a  R and hence, every
nonzero element in R is a unit. Therefore, R is a field.

EXERCISE 12(a)

1. State whether the following are true and justify your answers.
(i) 5 is an irreducible element in Z.
(ii) 10 is an irreducible element in Z[i].
(iii) 13 is a prime element in Z[i].
(iv) Any prime element in Z is a prime element in Z[i].
(v) 2  2 is an associate of 2 in Z[ 2].
(vi) 2 is irreducible in R.
(vii) 5 is a prime element in R.
(viii) 25 is a prime element in Z.

2. Determine all the units in each of the following:

(i) Z
(ii) Z[i]
(iii) Z[ 2]
(iv) Z[x]
(v) R[x]
(vi) Z5[x].

3. Determine all the associates of each of the following in the rings mentioned
against them
(i) 4 in Z
(ii) 1 1 x in R[x]
(iii) 1 1 i in Z[i]
(iv) 2 1 x in Z5[x]
(v) 1 1 2x in Z3[x]
(vi) 1 1 x 1 x2 in Z2[x].

Q001-Algebra-111001_CH 12.indd 9 9/16/2011 10:16:13 AM

 12-10   Algebra – Abstract and Modern

4. Consider the ring Z[ 3]  {a  b 3 : a and b ∈ }. In this, prove that 2  3 is a

unit and 3 2 3 is an associate 3.

5. Let a and b  Z such that a2 1 b2 is prime in Z. Then prove that a 1 bi is prime

in Z[i].

6. Prove that 2 and 1 1 5 are irreducible in Z[ 5] but not prime.

7. Let a and b  Z such that |a2 2 10b2| is prime in Z. Then prove that a  b 10
is irreducible in Z[ 10].

8. Let R be an integral domain and a1, a2, …, an  R. If p is a prime element in R

and p divides the product a1, a2, …, an, then prove that p divides atleast one ai.

9. Let a and b  Z such that a2 1 3b2 is prime in Z. Then prove that a  b 3 is

irreducible in Z[ 3].

10. Let a and b be nonzero elements in an integral domain R and a ~ b. If c  R and

a 5 bc, prove that c is a unit in R.

12.2 PRINCIPAL IDEAL DOMAINS

It is well known that every ideal of the ring Z of integers is a principal ideal.
In this, we discuss integral domains in which every ideal is principal.

Definition 12.2.1. An integral domain in which every ideal is a principal is

called a principal ideal domain (PID).

Example 12.2.1
1. Z is a PID.
2. Any field is a PID, since <0> and <1> are the only ideals of a field.
3. The ring F[x] of polynomials over a field F is a PID (refer Theorem
11.3.2).
4. The ring Z[x] of polynomials over Z is an integral domain, but not a
PID; for, we have exhibited, in Worked Exercise 11.3.1 an ideal of Z[x]
which is not principal.
The following is an immediate consequence of Theorems 12.1.3 and
12.1.4.

Theorem 12.2.1. In a PID, any finite number of nonzero elements have both
g.c.d. and l.c.m.

Corollary 12.2.1. Let R be a PID and a1, a2, …, an  R, Then,

g.c.d. {a1, a2, …, an} 5 a1r1 1 a2r2 1 … 1 anrn

for suitable elements r1, r2, …, rn in R.

Q001-Algebra-111001_CH 12.indd 10 9/16/2011 10:16:14 AM

 Factorization in Integral Domains   12-11

Definition 12.2.2. Let a1, a2, …, an be nonzero elements of an integral domain

such that g.c.d. {a1, a2, …, an} is a unit in R. Then, a1, a2, …, an are said to be
relatively prime and denote this by g.c.d. {a1, a2, …, an} ~ 1.
Note that any nonzero elements a1, a2, …, an of a PID R are relatively
prime if and only if there exist elements r1, r2, …, rn in R such that

a1r1 1 a2r2 1…1 anrn 5 1.

This identity is known as Bezout’s identity. The following is one of the most
useful applications of Bezout’s identity.

Theorem 12.2.2. Let R be a PID and a, b and c  R such that a and b are
relatively prime and a divides bc. Then, a divides c.

Proof: Since a and b are relatively prime, there exist r and s  R such that

ar 1 bs 5 1

(by Bezout’s identity). Now,

c 5 c  1 5 car 1 cbs 5 a(cr) 1 (bc)s.

Since a divides bc, ax 5 bc for some x  R. Therefore,

c 5 acr 1 bcs 5 a(cr 1 xs)

and hence a divides c.

Although prime elements are irreducible in a general integral domain (by
Theorem 12.1.5), we have observed that the converse is not true. ­However,
in a PID, any irreducible element is prime, as proved in the ­following
­theorem.

Theorem 12.2.3. The following are equivalent to each other for any nonzero
element p in a PID R.
1. p is a prime element.
2. p is an irreducible element.
3. <p> is a maximal ideal of R.
4. <p> is a prime ideal of R.

Proof: First observe that, to satisfy any of the conditions (1) through (4)
above, it is necessary that p is a nonunit in R. (1) ⇒ (2) follows from Theorem
12.1.5.

Q001-Algebra-111001_CH 12.indd 11 9/16/2011 10:16:14 AM

 12-12   Algebra – Abstract and Modern

(2) ⇔ (3): By Theorem 12.1.6 (2), p is irreducible if and only if <p> is

maximal among proper principal ideals of R. But R being a PID, every ideal
of R is principal. Therefore, p is irreducible if and only if <p> is a maximal
ideal of R.
(3) ⇒ (4) is trivial and (4) ⇒ (1) follows from Theorem 12.1.6 (1). b

Corollary 12.2.2. A nonzero ideal of a PID is maximal if and only if it is a

prime ideal.
The next results are concerned with the ideal structure of a PID. A sequence
{In} of ideals of a ring R is said to be an ascending (or increasing) sequence if
In ⊆ In11 for all n. A sequence {In} is said to terminate at a finite stage if there
exists n  Z1 such that In 5 In11 5 In12 5 In1k for all k  Z1.

Theorem 12.2.4. The following holds in any PID R.

1. Every ascending sequence of ideals of R terminates at a finite stage.
2. Any nonempty class of ideals of R has a maximal member.

Proof:
1. Let I1 ⊆ I2 ⊆ I3 ⊆ … be an ascending sequence of ideals of R. Put

I  ∪ I n . Then, I is an ideal of R. Since R is a PID, there exists a  R
n1
such that
∞

∪I
n1
n  I  a .

Then, for some n  Z1, a  In and hence

In1k ⊆ I 5 <a> ⊆ In ⊆ In1k

for all k  Z1, so that In 5 In1k for all k  Z1. Thus, the sequence termi-
nates at a finite stage.
2. Let # be a nonempty class of ideals of R and suppose, if possible, that #
has no maximal member. Since # is nonempty, choose I1  # Then, I1 is
not maximal in # and hence there exists I2  # such that I1  I2. Again,
since I2 is not maximal, there exists I3  # such that I2  I3. Continuing
this procedure, we get an ascending sequence
I1  I2  I3  …
of ideals of R which does not terminate at any finite stage. This is a con-
tradiction to (1) above. Thus, # must contain a maximal member. b

Q001-Algebra-111001_CH 12.indd 12 9/16/2011 10:16:15 AM

 Factorization in Integral Domains   12-13

Corollary 12.2.3. Let R be a PID and {an} be a sequence of elements in R

such that an divides an21 for all n > 1. Then, there exists n  Z1 such that

an ~ an11 ~ an1k for all k  Z1.

That is, an and an1k are associates for all k  Z1.

Proof: This is an immediate consequence of Theorem 12.2.4 above and of

the facts that, for any a and b  R,

<a> ⊆ <b> if and only if b divides a

     and <a> 5 <b> if and only if a and b are associates. b

Theorem 12.2.5. Let R be a PID and a be a nonzero nonunit element in R.

Then, there exists a prime element p in R such that p divides a.

Proof: Since a is a nonunit, the principal ideal <a> is a proper ideal and
hence <a> is contained in a maximal ideal M of R. Since R is a PID, there
exists p  R such that M 5 <p>. Also, since a  0. We have

<0>  <a> ⊆ M 5 <p> and hence p  0.

Since M is a maximal ideal, we get from Theorem 12.2.3, that p is a prime

element of R. Since <a> ⊆ <p>, we get that p divides a. b

Theorem 12.2.6. Any nonzero nonunit in a PID can be expressed as a finite

product of prime elements.

Proof: Let R be a PID and a be a nonzero nonunit element in R. By the

above theorem, there exists a prime element p1 in R such that p1 divides a
and hence

a 5 p1a1 for some a1  R.

Then, a1  0 (since a  0). If a1 is a unit, then a is an associate of p1 and

hence a itself is prime. If a1 is not a unit, then again by the above theorem,
there exists a prime element p2 dividing a1 and hence a1 5 p2a2 for some a2 
R. Repeating this process with a2 and so on, we get prime elements p1, p2, …
in R and elements a1, a2, … in R such that an 5 pn11an11. Now,

a 5 p1a1 5 p1p2a2 5 … 5 p1p2p3…pnan.

Q001-Algebra-111001_CH 12.indd 13 9/16/2011 10:16:15 AM

 12-14   Algebra – Abstract and Modern

Then, {an} is a sequence of nonzero elements such that an11 divides

an for all n. By Corollary 12.2.3, there exists n such that an ~ an1k for all
k  Z1.
Let n be the least positive integer such that an ~ an1k for all k  Z1. We claim
that an (and hence all an1k) is a unit. For, otherwise, we have the prime element
pn11 such that an 5 pn11an11 and, since an ~ an11, it follows that pn11 is a unit
which is a contradiction to the primeness of pn11. Thus, an is a unit and hence
pnan is also prime. Now, we have

a 5 p1p2 … pn21pnan

and p1, p2, …, pn21 and (pnan) are primes in R. b

Worked Exercise 12.2.1. Prove that the ring Z[i] of Gaussian integers is
a PID.

Answer: We have Z[i] 5 {a 1 bi : a and b  Z}. Z[i] is an integral domain

under the usual addition and multiplication of complex members. Let I be an
ideal of Z[i]. If I 5 {0} 5 <0>, we are done. Suppose I  {0}. Let

A 5 {a2 1 b2 : 0  a 1 bi  I}.

Then, A is a nonempty subset of Z1 and hence A has a least member. Let 0 

a 1 bi  I be such that a2 1 b2 is least in A. Now, we shall prove that I is the
principal ideal generated by a 1 bi. Put x 5 a 1 bi. Since x  I, we have <x>
⊆ I. On the other hand, let y 5 c 1 di  I. Consider the complex number

y (c  di )( a  bi ) ( ac  bd )  ( ad  bc)i
   a  bi,
x ( a  bi )( a  bi ) a2  b2

where a and b are rational numbers given by

ac  bd ad  bc
a and b .
a2  b2 a2  b2

Choose integers m and n such that

1 1
| ma |  and | n  b |  .
2 2

Now, y 5 (a 1 bi)x 5 (m 1 ni)x 1 ((a 2 m) 1 (b 2 n)i)x

Q001-Algebra-111001_CH 12.indd 14 9/16/2011 10:16:15 AM

 Factorization in Integral Domains   12-15

Put r 5 ((a 2 m) 1 (b 2 n)i)x. Then,

r 5 y 2 (m 1 ni)x  I (since x and y  I).

Also,

|r|2 5 ((a 2 m)2 1 (b 2 n)2) |x|2

1 1 1
    | x |2 | x |2 < a 2  b 2 .
4 4 2

Since a2 1 b2 is least in A, it follows that r 5 0 and hence

y 5 (m 1 ni)x <x>.

Therefore, I ⊆ <x>. Thus, I 5 <x>. Therefore, Z[i] is a PID.

Worked Exercise 12.2.2. Let R be a PID. Then prove that any nonzero proper
ideal of R can be expressed as a finite product of maximal ideals of R.

Answer: Let I be a nonzero proper ideal of R. Since R is a PID, I 5 <a> for

some nonzero nonunit a in R. By Theorem 12.2.6, a can be expressed as a
product of prime elements. Let

a 5 p1p2pn,

where p1, p2, …, pn are prime elements. Put Mi 5 <pi> for each 1  i  n. By
Theorem 12.2.3, each Mi is a maximal ideal of R. Now,

I 5 <a> 5 <p1p2pn>
5 <p1><p2><pn>
5 M1M2Mn.

Worked Exercise 12.2.3. Determine all the units of the ring Z[i] of Gaussian
integers.

Answer: Let a 1 bi be a unit in Z[i]. Then, there exist integers c and d such that

(a 1 bi)(c 1 di) 5 1.

By taking the absolute values, we have

(a2 1 b2)(c2 1 d2) 5 1.

Q001-Algebra-111001_CH 12.indd 15 9/16/2011 10:16:16 AM

 12-16   Algebra – Abstract and Modern

Since a, b, c and d are all integers, it follows that

a2 1 b2 5 1 5 c2 1 d2

and (a 5 0 and b 5 61) or (b 5 0 and a 5 61) and hence a 1 bi 5 1 or 21

or i or 2i. Thus, 1, 21, i and 2i are all the units in Z[i].

Worked Exercise 12.2.4. Let I be a nonzero ideal of the ring Z[i]. Then prove
that the quotient ring Z[i]/I is finite.

Answer: Since Z[i] is a PID, I is a nonzero principal ideal and hence I 5 <x>
for some 0  x  Z[i]. Let x 5 a 1 bi. Then, a  0 or b  0 and hence a2 1 b2
is a positive integer. Consider an element y 1 I  Z[i ] / I with y  Z[i]. As
in Worked Exercise 12.2.1, we can write

y 5 (m 1 ni)x 1 r

for some integers m and n and r  Z[i] such that

|r|2 < |x|2 5 a2 1 b2.

Now, y 2 r 5 (m 1 ni)x  <x> 5 I and hence y 1 I 5 r 1 I, where r 

Z[i] such that |r|2 < a2 1 b2. Since r must be of the form c 1 di with c and d
integers and c2 1 d2 < a2 1 b2 and since there can be only finitely many pairs
(c, d) of integers such that c2 1 d2 < a2 1 b2, it follows that

Z[i]/I 5 {(c 1 di) 1 I : c and d  Z and c2 1 d2 < a2 1 b2}

is finite.

EXERCISE 12(b)

1. Which of the following are PIDs? Justify your answers.

(i) Z
(ii) Q
(iii) R[x]
(iv) C[x]
(v) Z10
(vi) Z 31
(vii) Z2
(viii) Z6

Q001-Algebra-111001_CH 12.indd 16 9/16/2011 10:16:16 AM

 Factorization in Integral Domains   12-17

(ix) F[x], where F is a field.

(x) F[x, y] (5 F[x][y]), where F is a field.
2. Prove that Z[x] is not a PID.
3. For any integral domain R, prove that the polynomial ring R[x] is a PID if and
only if R is a field.
4. Prove that any homomorphic image of a PID is a PID.
5. Prove that Z[ 3] 5 {a 1 b 3 : a and b  Z} is an integral domain under
the usual addition and multiplication of complex numbers and that Z[ 3] is
not a PID.
6. Prove that Z[ 19] is not a PID.
7. For any odd prime number p, prove that Z[  p ] is not a PID.
8. Prove that Z[ 2] is a PID
9. Let D 5 {a0 1 a1x 1  1 anxn  R[x] : a0 is rational}. Prove that D is an integral
domain under the usual addition and multiplication of polynomials.
10. Let R be a PID and S be a multiplicative subset of R. Prove that the ring of
fractions of R by S (refer Exercise 15 of 10(f )) is a PID.
11. Let I be a nonzero ideal of a PID R. Then prove that any descending chain of
ideals of R/I terminates at a finite stage.
12. Prove that an integral domain R is a PID if and only if R/I is a PID for each
ideal I of R.
13. Let R be an integral domain in which any descending chain of ideals terminates
at a finite stage. Then prove that R is a field.
14. Prove that a PID R has a unique maximal ideal if and only if any two irreducible
elements of R are associates.
15. Prove that a nontrivial commutative ring with identity is a field if and only if
every proper ideal is prime.
16. Let R1, R2, …, Rn be PIDs and R 5 R1 3 R2 33 Rn. Then prove that every ideal
of R is principal. Is R a PID?
17. Prove the following for any ideals I, J and K of a PID R.
(i) I(J ∩ K) 5 (IJ) ∩ (IK)
(ii) I ∩ (J 1 K) 5 (I ∩ J) 1 (I ∩ K)
(iii) I 1 (J ∩ K) 5 (I 1 J) ∩ (I 1 K)
(iv) If I 1 J 5 R, then IJ 5 I ∩ J.
18. Let R be a PID. Prove that a nonzero ideal P of R is primary (refer Exercise 6 of
10(d)) if and only if P 5 <pn> for some prime element p in R and n  Z1.

Q001-Algebra-111001_CH 12.indd 17 9/16/2011 10:16:17 AM

 12-18   Algebra – Abstract and Modern

12.3 UNIQUE FACTORIZATION DOMAINS

We have developed the analogues, in an arbitrary integral domain, of the con-
cepts of the divisibility and prime numbers in the ring Z of integers. Recall
that any nonzero nonunit in Z is a product of finite number of prime numbers
(or their associates) and that this factorization is unique, except for the order
of occurrences of the primes and their associates. We shall formalise this in
the following definition.

Definition 12.3.1. Let R be an integral domain and a  R such that

a 5 p1p2 … pn,

where each pi is an irreducible element in R. Then, the equation a 5 p1p2 … pn

is called a factorization of a in R.

Examples 12.3.1
1. 6 5 2 ? 3 and 6 5 (22)(23) are factorizations of 6 in Z, since 2, 3, 22
and 23 are irreducible in Z.
2. 20 5 2 ? 2 ? 5, 20 5 (22)2(25) and 20 5 2(22)(25) are factorizations
of 20 in Z.
3. 1 1 2x 1 x2 5 (1 1 x)(1 1 x) is a factorization of 1 1 2x 1 x2 in Z[x],
since 1 1 x is irreducible in Z[x].
4. 1 1 x2 5 (1 1 x)(1 1 x) is a factorization of 1 1 x2 in Z2[x], since 1 1 x
is an irreducible element in Z2[x].

Definition 12.3.2. An integral domain R is called a factorization domain

(FD) if every nonzero nonunit in R has a factorization in R.

Examples 12.3.2
1. Z is a FD.
2. Any PID is a FD (recall Theorem 12.2.6). Note that an element in a PID
is prime if and only if it is irreducible.
3. The ring Z[i] of Gaussian integers is a FD, since it is a PID (see Worked
Exercise 12.2.1).
4. The ring F[x] of polynomials over a field F is a FD, since F[x] is a PID.
In the following, we give sufficient condition on an integral domain for it
to be a FD. This helps us as a tool to quickly ascertain that a given integral
domain is a FD.

Q001-Algebra-111001_CH 12.indd 18 9/16/2011 10:16:17 AM

 Factorization in Integral Domains   12-19

Theorem 12.3.1. Let R be an integral domain such that there is a map  : R 2

{0} → Z satisfying the following for any elements a and b in R 2 {0}.
1. (a)  0
2. (ab)  (a)
3. (ab) 5 (a) if and only if b is a unit in R.
Then, R is a FD.

Proof: We have to prove that every nonzero nonunit in R has a factorization

in R. Let S be the set of all nonzero nonunits in R which have no factorizations
in R. It is enough if we prove that S is empty. On the contrary, suppose that S
is not empty. Consider the set

A 5 {(a) : a  S}.

Then, A is a nonempty set of nonnegative (by (1)) integers. By the well-

ordering principle, A has a least member. Let n be the least in A. Then, n 5
(a) for some a  S. Then, a is not irreducible (since every irreducible ele-
ment gives a factorization of itself ). Therefore, a 5 bc for some nonzero
nonunits b and c in R. Then, by (2) and (3),

(a) 5 (bc) > (b) and (a) > (c).

By the least property of (a), it follows that b  S and c  S and hence

both b and c have factorizations in R. But then a (5 bc) also has a fac-
torization in R, which is a contradiction to the fact that a  S. Thus, S is
empty and hence any nonzero nonunit in R has a factorization in R. Thus,
R is a FD. b

Corollary 12.3.1. Z is a FD.

Proof: The map a  |a| from Z 2 {0} into Z satisfies the properties men-
tioned in Theorem 12.3.1 and hence Z is a FD. b

Corollary 12.3.2. The ring Z[i] of Gaussian integers is a FD.

Proof: Define  : Z[i] 2 {0} → Z by

(a 1 bi) 5 a2 1 b2.

Then,  satisfies (1), (2) and (3) of Theorem 12.3.1 and hence Z[i] is a FD. b

Q001-Algebra-111001_CH 12.indd 19 9/16/2011 10:16:17 AM

 12-20   Algebra – Abstract and Modern

Corollary 12.3.3. Let n be a positive integer greater than 1 and

{
Z  n   a  b n : a and b ∈  .
  }
Then, Z [ n ] is a FD.

Proof: Define  : Z [ n ] → Z by

( a  b n )  a 2  nb 2 .

Then,  satisfies the conditions in Theorem 12.3.1 and hence Z[ n ]

is a FD. b

Definition 12.3.3. A FD R is called an unique factorization domain (UFD)

if the following is satisfied:
If p1p2pn 5 q1q2qm, where pi’s and qj’s are irreducible elements in R,
then n 5 m and ai ~ bs(i) for some permutation s on {1,2, …, n}.
In other words, an integral domain R is called a UFD if every nonzero and
nonunit in R has a factorization in R which is unique, except for the associates
and order of occurrences of the irreducible factors.
Before going to certain examples of an UFD, we first prove two important
properties of UFD’s which are tools in determining whether a given FD is
a UFD.

Theorem 12.3.2. Let R be an UFD. Then, an element p in R is irreducible if

and only if it is prime.

Proof: We know that every prime element in any integral domain (and hence
in R) is irreducible. Conversely, suppose that p is an irreducible element in R.
Let b and c  R such that p|bc. We can assume that b and c are both nonzero
(since p|0). If b is a unit, then p|c. Similarly, if c is a unit, then p|b. Suppose that
neither b nor c is a unit. Since p|bc, there exists a  R such that pa 5 bc. Then,
a  0 (since b  0 and c  0) and a is nonunit (otherwise, if a is a unit, then
p 5 b(ca21) and, since p is irreducible, b or ca21 is a unit which is not true).
Thus, a, b and c are nonzero nonunits in R. Since R is a UFD, we get that

a 5 p1p2pn, b 5 q1q2qm and c 5 r1r2rt,

where pi’s, qi’s and ri’s are irreducible elements in R. Therefore, we have

pp1p2pn 5 q1q2qm r1r2rt

Q001-Algebra-111001_CH 12.indd 20 9/16/2011 10:16:18 AM

 Factorization in Integral Domains   12-21

By the uniqueness of the factorizations in R, p is an associate of qi or rj and

hence p divides b or c. Thus, p is a prime element in R.

Example 12.3.3. Consider the integral domain Z[ 5]. We have seen in

Corollary 12.3.3 that Z[ 5] is a FD. Also, in Example 12.1.2, we have
proved that 2  5 is irreducible, but not prime in Z[ 5]. Therefore, by
the above theorem, Z[ 5] is not an UFD.
Next result is a converse of Theorem 12.3.2 in the sense that a FD is an
UFD if every irreducible element is prime.

Theorem 12.3.3. Let R be a FD. Then, R is an UFD if and only if every irre-
ducible element in R is prime.

Proof: Suppose that every irreducible element in R is prime. Since R is a FD,

any nonzero nonunit in R can be expressed as a product of finite number of irre-
ducible elements. We have to prove only the uniqueness of the factorizations.
Let p1, p2, …, pn and q1, q2, …, qm be irreducible elements in R such that

p1p2pn 5 q1q2qm.

Suppose, if possible, that n  m. Without loss of generality, we can assume

that n > m. Since all the pi’s and qj’s are irreducible, they are primes. Since
p1 is a prime and p1 divides q1q2  qm, p1 should divide some qj. Let s(1) be
such a j. That is, 1  s(1)  m and p1 divides qs(1). Since qs(1) is irreducible,
it follows that p1 is an associate of qs(1). Therefore, there exists a unit u1 in R
such that p1u1 5 qs(1). Now, we have

p1 p2  pn  qs (1) ( ∏ q ) = p u ( ∏ q ).
j ≠ s (1)
j
1 1
j ≠ s (1)
j

Since R is an integral domain and p1  0, we can cancel p1 on both the sides.

Then, we have

p2  pn  u1 ( ∏ q ).
js (1)
j

We can repeat the above process, with p2 in place of p1, to get s(2) {1, 2, …,
m} 2 {q(1)} such that p2u2 5 qs(2) for some unit u2 in R. Then,

p2  pn  u1 p2 u2 ∏
j ≠ s (1), s ( 2 )
qj

and hence p3  pn  u1u2 ∏

js (1), s ( 2 )
qj
       

Q001-Algebra-111001_CH 12.indd 21 9/16/2011 10:16:19 AM

 12-22   Algebra – Abstract and Modern

This process can be continued for m steps (since n > m) to exhaust all qj’s and
then we get

pm11pn 5 u1u2um,

where u1, u2, …, um are units and hence their product u1u2um is also a unit.
Now, pm11 divides the unit u1u2um and hence pm11 itself is a unit which
is a contradiction to the fact that an irreducible element is necessarily a
nonunit. Thus, n 5 m and we have permutation s on {1, 2, …, n} such
that pi ~ qs(i) for all 1  i  n. Thus, R is an UFD, converse is proved in
­Theorem 12.3.2. b

Corollary 12.3.4. Every PID is an UFD.

Proof: Let R be a PID. In Theorem 12.2.3, we have proved that an element

in R is irreducible if and only if it is prime. Also, in Theorem 12.2.6, we
have proved that R is a FD. Therefore, by the above Theorem 12.3.3, R is
an UFD. b
The converse of the above result fails. That is, there are UFDs which are
not PIDs. For example, Z[x] is not a PID (refer Worked Exercise 11.3.1).
However, we prove in the next section that Z[x] is an UFD.

Corollary 12.3.5. The ring F[x] of polynomials over a field F is an UFD

in which the irreducible elements are precisely the irreducible polynomials
over F.

Proof: We have proved in Theorem 11.3.2 that, for any field F, F[x] is a PID
and hence, by the above corollary, F[x] is an UFD. b
Unique factorizations in an UFD help us in determining the g.c.d. and
l.c.m. of two elements. In this direction, we have the following theorem whose
proof is a routine verification.

Theorem 12.3.4. Let R be an UFD and a and b be nonzero nonunits in R.

Then, we can express a and b by

a  p1a1 p2a2 … pnan

b  p1b1 p2b2 … pnbn ,
           and
where p1, p2, …, pn are pair-wise nonassociate irreducible elements in R (i.e.,
pi is not an associate to any pj, j  i) and ai’s and bi’s are nonnegative integers.
Also, we have the following:

Q001-Algebra-111001_CH 12.indd 22 9/16/2011 10:16:19 AM

 Factorization in Integral Domains   12-23

1. a divides b if and only if ai  bi for all 1  i  n.

n
2. g.c.d .{a, b}   pimin{ai , bi }
i1
n
3. l.c.m.{a, b}   pimax{ai , bi } .
i1
We have proved earlier (see Theorem 12.2.3) that, in a PID, the principal
ideal generated by an irreducible element is a maximal ideal. We prove the
converse in the following theorem.
Theorem 12.3.5. Let R be an UFD. Then, R is a PID if and only if <p> is a
maximal ideal of R for any irreducible element p of R.
Proof: Suppose that <p> is a maximal ideal for any irreducible p in R. Let I
be an ideal of R. If I 5 {0}, then I is principal. Therefore, we can assume that
I  {0}. Also, we can assume that I  R. Put

 r r

m   : m  ∑ ai and ∏ pi  I ,
 ai

A i1 i1 .
 
 where p1 , …, pr are irreducible in R

Since I has atleast one nonzero nonunit, if follows that A is a nonempty set of
positive integers. Let m be the least member in A. Then, there exist irreducible
elements p1, p2, …, pr in R and positive integers a1, a2, …, ar such that
r r
d  ∏ piai  I and m  ∑ ai .
i1 i1

Now, we claim that I 5 <d>. Since d  I, we have <d> ⊆ I. On the other hand,
suppose that 0  x  I. We can assume that each pi is not an associate of any
a
other pj, j  i. It is enough if we can prove that each pi i divides x. Suppose
a1
that p1 does not divide x. Then, we have

x 5 p1ny, where 0  n < a1 and p1  y.

Then, y  <p1> and <p1> is a maximal ideal of R. Therefore, <y> 1

<p1> 5 R and hence
1 5 ay 1 bp1 for some a and b  R.
a 1 a
Put Z  p1 1 p2 2 … pr r . Then,
a

z 5 z1 5 zay 1 zbp1
 p1n y( ap1a11n p2a2 … prar )  db
 x( ap1a11n p2a2 … prar )  db  I
                    

Q001-Algebra-111001_CH 12.indd 23 9/16/2011 10:16:20 AM

 12-24   Algebra – Abstract and Modern

since x  I and d  I. Therefore, z  I which is a contradiction to the least

property of m. Therefore, p1a1  x. Similarly, p1a1  x for all 1 ≤ i ≤ n and
hence d|x. Therefore, x  <d>. Thus, I 5 <d>. Therefore, R is a PID. ­Converse
is proved in Theorem 12.2.3. b

Worked Exercise 12.3.1. Let Z [ 3 ] 5 {a 1 b 3 : a and b  Z}. Then

prove that Z [ 3 ] is a FD.

Answer: Define  : Z [ 3 ] 2 {0} → Z by

(a 1 b 3 ) 5 |a2 2 3b2|.

Then, (x)  0 for all 0  x  Z [ 3 ] . Also, if x 5 a 1 b 3 and y 5

c 1 d 3 , then

(xy) 5 ((ac 1 3bd) 1 (ad 1 bc) 3 )

5 |(ac 1 3bd)2 2 3(ad 1 bc)2|
   5 |a2c2 1 9b2d2 2 3a2d2 2 3b2c2|
   5 |a2 2 3b2| |c2 2 3d2| 5 (x)(y)
 |a2 2 3b2| 5 (x)

    and (x) 5 1 ⇔ |a2 2 3b2| 5 1

⇔ |(a 1 b 3 )(a 2 b 3 )| 5 1

⇔ (a 1 b 3 )(a 2 b 3 ) 5 1

⇔ a 1 b 3 is a unit.

∴ (xy) 5 (x) if and only if y is a unit. Therefore, by Theorem 12.3.1, Z[ 3]

is a FD.

Worked Exercise 12.3.2. Let R be a UFD and P be a nonzero prime ideal of

R. Then prove that there exists an irreducible element in P.

Answer: Since P  {0}, we can choose 0  x  P. Also since P  R, x is a

nonunit. Now, we have

x  p1a1 p2a2 … prar ,

where p1, p2, …, pr are irreducible elements in R and a1, a2, …, ar are positive
integers. Since x  P and P is a prime ideal, pi  P for some i.

Q001-Algebra-111001_CH 12.indd 24 9/16/2011 10:16:22 AM

 Factorization in Integral Domains   12-25

EXERCISE 12(c)

1. Which of the following ring are UFDs? Justify your answers.

(i) Z
(ii) Q
(iii) R
(iv) Z6
(v) Z[i]
(vi) Z [ 5 ]
(vii) R[x]
(viii) Q[x].

2. Prove that the l.c.m. of any finite subset of an UFD exists and is unique up to
associates.

3. Let R be an UFD and F be its field of quotients. For any prime p in R, let

 a 
R( p )    F : g.c.d.{a, b}  1 and p  b .
 b 
Prove that R(p) is a subring of F and that R(p) is a PID and hence an UFD.
4. Let S be a multiplicative set in an UFD R and S21R the ring of fractions of R by
S. Prove that S21R is an UFD.
5. Let R be an UFD and {an} be a sequence of elements in R such that an11 divides
an for all n  Z1. Then prove that there exists n  Z1 such that an is an associate
of an1k for all k  Z1.
6. Let p be a prime element in an UFD R such that any prime element in R is an
associate of p. Prove that every nonzero proper ideal of R is of the form <pn> for
some n  Z1.
7. Let R be an UFD and P be the only nonzero prime ideal of R. Then prove that
any nonzero proper ideal I of R is of the form Pn for some n  Z.
8. Prove that any increasing sequence of principal ideals in an UFD terminates at a
finite stage.
9. If P is a nonzero prime ideal of an UFD R, then is R/P a UFD?
10. Let R be an UFD and (a, b) denote the g.c.d. {a, b} for any a and b  R. Prove
the following for any nonzero elements a, b and c  R.
(i) (a, (b, c)) ~ ((a, b), c)
(ii) (a, 1) ~ 1
(iii) (ca, cb) ~ c(a, b)

Q001-Algebra-111001_CH 12.indd 25 9/16/2011 10:16:23 AM

 12-26   Algebra – Abstract and Modern

(iv) (a, ab) ~ a

(v) (a, b) ~ 1 ~ (a, c) ⇒ (a, bc) ~ 1
(vi) (a, b) ~ 1, a|c and b|c ⇒ ab|c
(vii) (a, b) ~ 1 and a|bc ⇒ a|c
(viii) ab ~ (a, b)[a, b], where [a, b] 5 l.c.m. {a, b}.

12.4 POLYNOMIALS OVER UFDS

Polynomials over an UFD are of special importance in view of the famous
Gauss theorem which states that such polynomials again form an UFD. In this
section, we introduce few more concepts in polynomials which ultimately
lead to the proof of the Gauss theorem. Let us recall that, in any UFD, any
finite number of nonzero elements have g.c.d.

Definition 12.4.1. Let R be an UFD and f (x) 5 a0 1 a1x 1 a2x2 1  1 anxn

be a nonzero polynomial over R. Then, the content of f is defined to be the
g.c.d. of the coefficients a0, a1, a2, …, an and is denoted by c(f (x)); that is,

c(f (x)) 5 g.c.d. {a0, a1, a2, …, an}.

A polynomial f (x) is said to be primitive over R if c(f (x)) is a unit in R.

The content and the primitivity of a polynomial depends on the UFD over
which the polynomial is defined. Consider the following example.

Example 12.4.1. Let f (x) 5 4 1 8x 1 6x2. If we consider f (x) as a polyno-

mial over Z, then

c(f (x)) 5 g.c.d. {4, 8, 6} 5 2

and therefore, f (x) is not primitive over Z. However, if we consider f (x) as a

polynomial over the field Q of rational numbers, then

c(f (x)) 5 1,

since any nonzero in Q is a unit. Therefore, f (x) is primitive over Q.

Note 12.4.1
1. Any monic polynomial over any UFD is primitive.
2. If R is an UFD and 0  f (x)  R[x], then
f (x) 5 c(f (x))g(x)

for some primitive polynomial g(x) in R[x].

Q001-Algebra-111001_CH 12.indd 26 9/16/2011 10:16:23 AM

 Factorization in Integral Domains   12-27

In the following, we prove that primitive polynomials over any UFD are
closed under multiplication.

Theorem 12.4.1. Let R be an UFD and f (x) and g(x) be primitive polynomials
over R. Then, f (x)g(x) is primitive.

Proof: Let f (x) 5 a0 1 a1x 1  1 anxn, an  0

   and g(x) 5 b0 1 b1x 1  1 bmxm, bm  0.

If deg(f (x)) 5 n 5 0, then c(f (x)) 5 a0 and hence a0(5 f (x)) is a unit so that
c(f (x)g(x)) ~ c(g(x)) ~ 1. Therefore, we can assume that n > 0 and m > 0.
We prove that there is no prime element in R that divides all the coefficients
in f (x)g(x). To prove this, let p be a prime in R. Since f (x) is primitive, p does
not divide some coefficient ai in f (x). Let ai be the first coefficient in f (x) which
is not divisible by p. Similarly, let bj be the first coefficient in g(x) which is not
divisible by p. Let ci1j be the coefficient of xi1j in f (x)g(x). Then,

ci1j 5 a0bi1j 1 a1bi1j21 1  1 ai1jb0

 ∑
rsi j
ar bs .

Since p|ar for all 0  r < i and p|bs for all 0  s < j, it follows that ci1j 2 arbs
is divisible by p. Since p does not divide aibj, we get that p does not divide ci1j.
Therefore, there is no prime dividing all the coefficients in f (x)g(x) and hence

c(f (x)g(x)) is a unit.

Thus, f (x)g(x) is primitive. b

Corollary 12.4.1. For any nonzero polynomials f (x) and g(x) over an UFD,
c(f (x)g(x)) 5 c(f (x))c(g(x)).

Proof: Let R be an UFD and f (x) and g(x)  R[x] 2 {0}. Then, there exist
primitive polynomials f1(x) and g1(x) in R[x] such that

f (x) 5 c(f (x))f1(x) and g(x) 5 c(g(x))g1(x).

Thus, by the above theorem, f1(x)g1(x) is primitive and

f (x)g(x) 5 c(f (x))c(g(x))f1(x)g1(x)

and therefore, c(f (x)g(x)) 5 c(f (x))c(g(x)).

Q001-Algebra-111001_CH 12.indd 27 9/16/2011 10:16:23 AM

 12-28   Algebra – Abstract and Modern

Let us recall that an element p in an integral domain R is called irreducible

if it is nonzero, nonunit and not a product of two nonunits. (See Definition
12.1.5). Also, recall that a polynomial over R is called irreducible over R if
it is nonconstant and not a product of two nonconstant polynomials over R.
In the case of polynomial rings over an integral domain R, there is a subtle
difference between irreducible elements in R[x] and irreducible polynomials
over R. However, if R is a field, both these concepts coincide. In the case of
general integral domains, we distinguish these two. Irreducible elements in
R[x] and irreducible polynomials over R are two different concepts. Irreduc-
ible elements in R[x] are often called irreducible in R[x].
Before going further, let us observe that, if F is the field of quotients of an
UFD R and 0  f (x)  F[x], then we can write f (x) 5 ab21g(x) for some a
and b  R and g(x) is primitive in R[x]; for, if

a0 a1 a
f ( x)   x  n x n
b0 b1 bn

and b 5 b0b1 … bn  R and f (x) 5 b21h(x) 5 b21c(h(x))g(x) for some primi-

tive g(x) in R[x]. b

Theorem 12.4.2 (Gauss Lemma). Let R be an UFD and F be the field of

quotients of R. Let f (x)  R[x] be a primitive polynomial of positive degree.
Then, f (x) is irreducible in R[x] if and only if f (x) is irreducible in F[x].

Proof: Suppose that f (x) is irreducible in F[x]. Suppose that f (x) 5 g(x)h(x)
where g(x) and h(x)  R[x]. If g(x) and h(x) are both of positive degree then
they are nonunits in R[x] and in F[x], which is a contradiction to the irreduc-
ibility of f (x) in F[x]. Therefore, g(x) or h(x) is of degree 0. Let deg(g(x)) 5 0.
Then, g(x)  R. Since f (x) is primitive in R[x], we get that g(x) is a unit in R.
Similarly, if deg(h(x)) 5 0, then h(x) is a unit in R. Thus, f (x) is irreducible
in R[x].
Conversely, suppose that f (x) is reducible in F[x]. Then, f (x) 5 g(x)h(x) where
g(x) and h(x) are nonunits in F[x] and hence g(x) and h(x) are polynomials of
positive degree (since F is a field). Let

a0 a1 a
g( x)   x  n x n , an  0, n  0
b0 b1 bn
c0 c1 c
     and h( x )   x  m x m , cm  0, m  0
d0 d1 dm

where ai, bi, cj, dj  R, bi  0 and dj  0.

Q001-Algebra-111001_CH 12.indd 28 9/16/2011 10:16:24 AM

 Factorization in Integral Domains   12-29

Put b 5 b0b1b2 … bn and d 5 d0d1d2 … dm.

Then, bg(x) and dh(x)  R[x]. We can write

bg(x) 5 ag1(x) and dh(x) 5 ch1(x)

for some primitive polynomials g1(x) and h1(x) in R[x]. Now, we have

bdf (x) 5 bg(x)dh(x) 5 acg1(x)h1(x).

By Theorem 12.4.1, g1(x)h1(x) is primitive in R[x]. Since f (x) is also primi-

tive, it follows that bd is an associate of ac and hence bdu 5 ac for some unit
u in R. Now,

bdf (x) 5 bdug1(x)h1(x).

and hence f (x) 5 ug1(x)h1(x), which implies that f (x) is reducible in R[x]. b

Theorem 12.4.3 (Gauss Theorem). The ring R[x] of polynomials over an

UFD R is also an UFD.

Proof: Let R be an UFD. Then, clearly R[x] is an integral domain. Let

0  f (x)  R[x]. If deg(f (x)) 5 0, then f (x)  R and, since R is an UFD, f (x)
has a factorization in R and hence in R[x]. Therefore, we can assume that
deg(f (x)) > 0.
Let F be the quotient field of R. Then, R is a subring of F. We can write
f (x) 5 c(f (x))g(x), where g(x) is a primitive polynomial over R and hence
over F. Now, recall that F[x] is an UFD (by Corollary 12.3.5). Since g(x)
 F[x] and deg(g(x)) > 0, we get that

g(x) 5 g1(x)g2(x)gn(x)

for some irreducible polynomials g1(x), g2(x), …, gn(x) in F[x]. Then, deg(gi(x))
> 0 for each 1  i  n. We can write gi ( x )  ai bi1 hi ( x ) where ai, bi  R
(bi  0) and hi(x) is primitive in R[x]. Then, we have

b1b2bng(x) 5 a1a2anh1(x)h2(x)hn(x).

Since each hi(x) is primitive, so is their product. Also, since g(x) is primitive,
it follows, by taking contents both sides that

b1b2bn ~ a1a2an in R

and hence g(x) 5 uh1(x)h2(x)hn(x),

Q001-Algebra-111001_CH 12.indd 29 9/16/2011 10:16:24 AM

 12-30   Algebra – Abstract and Modern

where u is a unit in R. Further, since each hi(x) is irreducible in F[x] and

primitive in R[x], it follows that each hi(x) is irreducible in R[x] (by the Gauss
Lemma (Theorem 12.4.2)). Thus,

f (x) 5 c(f (x))uh1(x)h2(x)hn(x).

If c(f (x)) is a unit in R, c(f (x))uh1(x), h2(x), …, hn(x) are irreducible elements
in R[x] and hence we have a factorization of f (x) in R[x]. If c(f (x)) is not a unit
in R, then, by the factorization property in R,

c(f (x)) 5 p1 p2pm,

where p1, p2, …, pm are irreducible elements in R and hence in R[x]. Then,

f (x) 5 p1p2pm21(pmu)h1(x)h2(x)hn(x)

is a factorization of f (x) in R[x]. Thus, R[x] is a FD.

To prove the uniqueness of the factorizations in R[x], first observe that any
irreducible polynomial in R[x] of positive degree must be primitive. Any fac-
torization of f (x) in R[x] must be of the form

f (x) 5 c1c2cng1(x)g2(x)gm(x) (1)

where c1, c2, …, cn (n  0) are irreducible elements in R and g1(x), g2(x), …,

gm(x) (m  0) are irreducible in R[x], each of positive degree, and hence g1(x),
g2(x), …, gm(x) are primitive. Now, suppose that

f (x) 5 d1d2drh1(x)h2(x)hs(x) (2)

is another factorization of f (x) in R[x]. Since each of gi(x) and hi(x) are
irreducible polynomials of positive degree in R[x], they are primitive and
hence these are irreducible in F[x]. Also, by taking contents in (1) and (2),
we get that

c1c2cn 5 ud1dr (3)

for some unit u in R. Since R is an UFD, n 5 r and each ci is an associate of

some dj. Further, from (1), (2) and (3), we get that

h1(x)h2(x)hs(x) 5 ug1(x)g2(x)gm(x).

Since F[x] is an UFD, it follows that s 5 m and each hi(x) is an associate of

gj(x). Thus, the factorization (1) is unique. Therefore, R[x] is an UFD.

Q001-Algebra-111001_CH 12.indd 30 9/16/2011 10:16:24 AM

 Factorization in Integral Domains   12-31

Corollary 12.4.2. If R is an UFD, then so is R[x1, x2, …, xn], where R[x1, x2,
…, xn] 5 R[x1, …, xn21][xn].

Corollary 12.4.3. Z[x] is an UFD but not a PID.

Corollary 12.4.4. For any field F, F[x1, x2, …, xn] is an UFD.

Though we have proved in the above theorem that any polynomial over
an UFD R can expressed as a product of irreducible elements in R[x], it is
a difficult task to find such a factorization for a given polynomial. Gauss
theorem only ensures the existence of a factorization. There is no general
explicit method for obtaining such a factorization, not even for deciding
whether a given polynomial is irreducible or not. However, we have certain
sufficient conditions for the irreducibility of a polynomial over an UFD as
given below.

Theorem 12.4.4 (Eisenstein’s Criterion). Let R be an UFD and F be the field

of quotients of R. Let

f (x) 5 a0 1 a1x 1  1 anxn, n > 0, an  0

be a nonconstant polynomial in R[x]. Suppose that there exists a prime

element p in R such that
1. p divides ai for 0  i < n.
2. p does not divide an.
3. p2 does not divide a0.
Then, f (x) is irreducible in F[x]. Also, if f (x) is primitive, then f (x) is irreduc-
ible in R[x].

Proof: First, we assume that f (x) is primitive and prove that f (x) is irreduc-
ible in R[x]. Suppose, if possible, f (x) is not irreducible. Since f (x) is a non-
constant primitive polynomial, there exist two nonzero nonunit polynomials
f1(x) and f2(x) in R[x] such that

f (x) 5 f1(x)f2(x).

Let f1(x) 5 c0 1 c1x 1  1 crxr and f2(x) 5 d0 1 d1x 1  1 dsxs.

Then,

n 5 deg(f (x)) 5 deg(f1(x)) 1 deg(f2(x)) 5 r 1 s.

If r 5 0, then f1(x) 5 c0 and f (x) 5 c0 f2(x) and hence c0 is a nonzero non-

unit (since so is f1(x)) divisor of c(f (x)) which is a unit. Therefore, r  0.

Q001-Algebra-111001_CH 12.indd 31 9/16/2011 10:16:24 AM

 12-32   Algebra – Abstract and Modern

­Similarly, s  0. Since n 5 r 1 s, it follows that 0 < r < n and 0 < s < n. Now,
f (x) 5 f1(x)f2(x) implies that a0 5 c0d0. Since p|a0 and p2a0, it follows that p
divides exactly one of c0 and d0. Without loss of generality, we can assume
that p|c0 and pd0. Also, we have an 5 crds and pan and hence pcr and pds.
Therefore, we have p|c0 and pcr. Let i be the least positive integer such that
pci. Then, p|cj for all 0  j < i. Equating the coefficients of xi in f (x) and f1(x)
f2(x), we get that

ai 5 c0di 1 c1di21 1 … 1 cid0,

where cj 5 0 for all j > r and dj 5 0 for all j > s. Now, since p|c0, p|c1, …,
p|ci21, we get that

p | ai 2 cid0.

Also, since 0 < i  r < n, p|ai, by (1). Therefore, p|cid0 which is a contradiction
since p is prime, pci and pd0.
Thus, f (x) is irreducible in R[x]. Also, since f (x) is primitive, the Gauss
Lemma (Theorem 12.4.2) implies that f (x) is irreducible in F[x].
Now, let us take up the general case. There exists a primitive polynomial g(x)
in R[x] such that

f (x) 5 c(f (x))g(x),

where c(f (x)) is the content of f (x). If c(f (x)) is a unit, then f (x) is primitive
and hence, by the first case, f (x) is irreducible in R[x] and in F[x]. Suppose
that c(f (x)) is a nonunit. Let

c(f (x)) 5 d and g(x) 5 b0 1 b1x 11 bnxn.

Then, deg(f (x)) 5 deg(g(x)) and hence bn  0. Since f (x) 5 dg(x), we get
that
ai 5 dbi for all 0  i  n.

Since pan, we get that pd and pbn. Also, for any 1  i < n, p|ai implies that
p|bi. Further, p2b0, since p2a0. Thus, by the first case, g(x) is irreducible in
R[x] and hence in F[x]. Since d is a unit in F, we get that f (x) is an associate
of g(x) in F[x]. Thus, f (x) is irreducible in F[x]. b
Since Z is a UFD and the rational number field  is the field of quo-
tients of Z, the following is an immediate consequence of the Eisenstein’s
­Criterion.

Q001-Algebra-111001_CH 12.indd 32 9/16/2011 10:16:25 AM

 Factorization in Integral Domains   12-33

Corollary 12.4.5. Let f (x) 5 a0 1 a1x 11 anxn be a constant polynomial in

Z[x] and p be a prime number satisfying the following:
1. p divides ai for all 0  i < n.
2. p does not divide an.
3. p2 does not divide a0.
Then, f (x) is irreducible in [x]. Further, if f (x) is primitive, then f (x) is
irreducible in Z[x].

Corollary 12.4.6. Let p be a prime number and f (x) 5 1 1 x 1 x2 1  1

xp21. Then, f (x) is irreducible in [x] and in Z[x]. This polynomial f (x) is
called the cyclotomic polynomial.

Proof: Since f (x) is a monic polynomial, it is primitive in Z[x].Consider the

polynomial f (x 1 1). We are given that

x p 1
f ( x )  1 x  x 2  x p1  .
x 1
(x 1) p 1
Therefore,   f (x 1) 
(x 1) 1
1
 ( x p  px p1  (p2 )x p2  px )
x
p1
 ∑ ( pr ) x pr 1
r0

 x p1  px p2 ( pi )x pi1  p.

By the Eisenstein’s Criterion (Theorem 12.4.4), f (x 1 1) is irreducible in
[x] and therefore, f (x) is irreducible in [x]. Since f (x) is primitive in Z[x],
it is irreducible in [x]. b
The following theorem is a very useful tool to determine the irreducibility
of certain polynomials in Z[x] and their rational roots.

Theorem 12.4.5. Let f (x) 5 a0 1 a1x 1  1 anxn  Z[x] with a0  0 and

an  0. Let p and q be relatively prime integers and q > 0. Suppose that p/q
is a root of f (x) in . Then, p divides a0 and q divides an.

Proof: We are given that f ( p / q)  0 and therefore

2 n
 p  p  p
a0  a1    a2    an    0.
 q   q   q 

Q001-Algebra-111001_CH 12.indd 33 9/16/2011 10:16:25 AM

 12-34   Algebra – Abstract and Modern

By multiplying with qn, we have

a0qn 1 a1pqn21 1 a2p2qn22 1 … 1 an21pn21q 1 anpn 5 0

and hence

 n 
a0 q n  p ∑ ar p r q nr 
 r1 
 nn1 
aan0 pq n  qp aarr pprrqqnnrr.
and  ∑ 01
rr

 n1 
an p n  q ∑ ar p r q nr .
 r0 

Therefore, p divides a0qn and q divides anpn. Since p and q are relatively
prime, it follows that p divides a0 and q divides an. b

Worked Exercise 12.4.1. Let f (x) 5 5 1 11x 2 7x2 1 9x3  Z[x]. Prove that
f (x) is irreducible over Q as well as over Z.

Answer: Since deg(f (x)) 5 3, f (x) is irreducible over Q if and only if f (x)
has a root in Q. Suppose that p/q is a root of f (x), we can assume that p and q
are relatively prime integers and q > 0. Then, by the above Theorem 12.4.5, p
should divide 5 and q should divide 9. Therefore, p 5  1 or  5 and q 5 
1 or  3 or  9 and hence

p
q
{ 1 5 1 5
 1,  5,  ,  ,  ,  .
3 3 9 9
}
But, none of the elements in this set is a root of f (x). Therefore, f (x) has no root
in Q. Thus, f (x) is irreducible over Q. Since f (x) is primitive in Z[x], it follows
from Gauss Lemma (Theorem 12.4.2), that f (x) is irreducible over Z also.

Worked Exercise 12.4.2. Let f (x) 5 22 1 15x 2 9x2 1 x3  Z[x]. Prove that
f (x) is irreducible over neither Q nor Z.

Answer: Suppose, if possible, (p/q)  Q is a root of f (x). We can assume that

p and q are relatively prime integers and q > 0. Then, by the above Theorem
12.4.5, p should divide 22 and q should divide 1 and hence p 5 1 or 2
and q 5 1. Therefore,

p
 ±1 or ± 2.
q

Q001-Algebra-111001_CH 12.indd 34 9/16/2011 10:16:26 AM

 Factorization in Integral Domains   12-35

By a physical verification, we can see that 2 is a root of f (x); that is, f (2) 5 0.
Thus, x 2 2 is a factor of f (x); in fact,

f (x) 5 (x 2 2)(x2 2 7x 1 1).

Thus, f (x) is reducible over Q as well as over Z.

Exercise 12(d)

1. Which of the following are UFDs? Justify your answers.

(i) Z[x]
(ii) Q[x]
(iii) R[x]
(iv) C[x]
(v) Z3[x]
(vi) Z6[x]
(vii) Z9[x]
(viii) Z13[x].

2. Prove that every field is an UFD.

3. Let R be an UFD and f (x)  R[x]. Prove that f (x) is irreducible over R if and only
if f (x 1 a) is irreducible over R for some a  R.

4. Which of the following polynomials are irreducible over the UFDs mentioned
against them?
(i) 15 2 9x2 1 6x3 1 2x4 over Z
(ii) 3 1 2x2 1 x3 over Q
(iii) 4 1 2x 1 x3 over Z5
(iv) 1 1 x2 1 x5 over Z2
(v) 9 2 x3 over Z31
(vi) 1 1 x3 1 x6 over Q
(vii) 5 1 10x 1 15x3 1 2x5 over Z
(viii) 2 1 2x 1 x4 over Q
(ix) 9 2 x3 over Z11
(x) 14 2 7x 1 10x4 over Q.

5. Prove that any polynomial over R of degree  3 is reducible over R.

6. For any prime number p, prove that p 2 xn is irreducible over Q for any positive
integer n.

Q001-Algebra-111001_CH 12.indd 35 9/16/2011 10:16:26 AM

 12-36   Algebra – Abstract and Modern

7. Prove that 1 1 x4 is irreducible over Q and reducible over Zp for any prime
number p.
8. Determine all irreducible polynomials of degree 2 in Z2[x].
9. Give an example of a polynomial which is irreducible over Z but not irreducible
over Z2.
10. For any prime p, prove that there are exactly (p(p 2 1))/2 irreducible monic
polynomials of degree 2 in Zp[x].
11. Let p be a prime number and
f (x) 5 1 2 x 1 x2 2 x3 1 … 1 (21)p21xp21.
Then prove that f (x) is irreducible over Z.
12. Let R be an UFD and F be its field of quotients. Prove Theorem 12.4.5 with R
and F in place of Z and Q, respectively.

12.5 EUCLIDEAN DOMAINS

Another important class of integral domains, about which we discuss in this
section, is the class of Euclidean domains. These arose out of attempts to
generalize the familiar Euclidean division algorithm for integers to elements
of arbitrary rings. Let us begin our discussion with the following.

Definition 12.5.1. An integral domain R is said to be an Euclidean domain if

there exists a function g : R 2 {0} → Z1 satisfying the following conditions:
1. g(ab) 5 g(a)g(b) for all a and b  R 2 {0}.
2. For any a and b  R with b  0, there exist elements q and r  R such that
a 5 qb 1 r and either r 5 0 or g(r) < g(b).
The function g is called the guage function (or Euclidean valuation).
In other words, an integral domain R is called an Euclidean domain if,
to each nonzero element a of R, there is an associated positive integer g(a),
called the guage of a, satisfying the conditions (1) and (2) above. Condition
(2) is called the Euclidean division algorithm and q and r in (2) are called
quotient and remainder, respectively. We first mention the following simple
examples of Euclidean domains.

Example 12.5.1
1. The ring Z of integers is an Euclidean domain, in which the gauge func-
tion is defined by
g(a) 5 |a|, the absolute value of a,

Q001-Algebra-111001_CH 12.indd 36 9/16/2011 10:16:26 AM

 Factorization in Integral Domains   12-37

for any a  Z 2 {0}. Clearly, |ab| 5 |a||b|. To prove the division algo-
rithm, let a and b  Z and b  0. Without loss of generality, we can
assume that b > 0. Let q be the integral part of the rational number a/b;
that is, q is an integer such that
a a
q and q 1  .
b b

Then, bq  a and bq 1 b > a. Now, put r 5 a 2 bq. Since a, b and q are

all integers, we get that r  Z and
a 5 qb 1 r, where r 5 0 or |r| 5 r < b 5 |b|.
Thus, Z is an Euclidean domain.
2. Every field F is an Euclidean domain; for, define g : F 2 {0} → Z1 by
g(a) 5 1 for all a  F 2 {0}. Then, g(ab) 5 1 5 g(a)g(b) for all a and
b  F 2 {0} and
a 5 (ab21)b 1 0 5 qb 1 r, where q 5 ab21 and r 5 0.
In the following, we exhibit certain elementary properties of the gauge
function of an Euclidean domain.

Theorem 12.5.1. Let R be an Euclidean domain with gauge function g. Then,

the following holds
1. g(1) 5 1.
2. For any 0  a  R, a is a unit in R if and only if g(a) 5 1.
3. If a and b are associates in R 2 {0}, then g(a) 5 g(b).

Proof:
1. follows from the facts that g(1) is a positive integer and g(1) 5 g(1 ? 1)
5 g(1)g(1).
2. Let 0  a  R. If a is a unit in R, then there exists b  R such that ab 5 1.
Then, b  0 and
1 5 g(1) 5 g(ab) 5 g(a)g(b).
Since g(a) and g(b) are positive integers, we get that g(a) 5 1 5 g(b).
Conversely, suppose that g(a) 5 1. By the Euclidean division algorithm,
there exist q and r  R such that
1 5 qa 1 r, where r 5 0 or g(r) < g(a) 5 1.
If r  0, then g(r)  Z1 which is not true, since g(r) < 1. Therefore,
necessarily r 5 0 and 1 5 qa. Thus, a is a unit in R.

Q001-Algebra-111001_CH 12.indd 37 9/16/2011 10:16:26 AM

 12-38   Algebra – Abstract and Modern

3. Suppose a and b are associates. Then, au 5 b for some unit u in R. Now,

g(b) 5 g(au) 5 g(a)g(u) 5 g(a) (since g(u) 5 1). b

Theorem 12.5.2. Every Euclidean domain is a PID.

Proof: Let R be an Euclidean domain with gauge function g. Let I be an ideal

of R. If I 5 {0}, then I 5 <0> and hence I is principal. Suppose that I  {0}.
Consider the set

A 5 {g(a) : 0  a  I}.

Then, I is a nonempty set of positive integers and, by the well-ordering prin-

cipal, A has a least member, say g(a) with 0  a  I. Now we prove that A
5 <a>. Since a  I, we have <a> ⊆ I. On the other hand, let x  I. Then, by
Euclidean division algorithm, there exist q and r  R such that

x 5 qa 1 r, where r 5 0 or g(r) < g(a).

Now, r 5 x 2 qa  I (since x and a  I). Since g(a) is least in A, it follows

that r 5 0 and hence x 5 qa  <a>. Thus, I ⊆ <a> and hence I 5 <a>. Thus,
R is a PID. b

Corollary 12.5.1. Every Euclidean domain is an UFD.

Proof: This follows from the fact that every PID is an UFD and from the
above theorem. b

Theorem 12.5.3. The ring F[x] of polynomials over a field F is an Euclidean

domain and hence a PID and an UFD.

Proof: Let F be a field. For any 0  f (x)  F[x], define

g(f (x)) 5 2deg(f (x)).

Since deg(f (x))  0, g : F[x] 2 {0} → Z1is a function.

Also, for any nonzero f (x) and h(x) in F[x],

g(f (x)h(x)) 5 2deg(f (x)h(x))

5 2deg(f (x)1deg(h(x))
5 2deg(f (x)) ? 2deg(h(x))
5 g(f (x))g(h(x)).

Q001-Algebra-111001_CH 12.indd 38 9/16/2011 10:16:27 AM

 Factorization in Integral Domains   12-39

Also, for any f (x) and h(x)  F[x] with h(x)  0, by the division algorithm for
polynomials, there exist q(x) and r(x)  F[x] such that

f (x) 5 q(x)h(x) 1 r(x),

where r(x) 5 0 or deg(r(x)) < deg(h(x)) and hence r(x) 5 0 or g(r(x)) <
g(h(x)). Thus, F[x] is an Euclidean domain. b
The following is a generalization of the well-known algorithm to find the
g.c.d. of any two positive integers.

Theorem 12.5.4. Let R be an Euclidean domain with gauge function g and a

and b  R 2{0}. Let {qn} and {rn} be sequences of elements in R satisfying
the following:

b 5 q1a 1 r1, where r1 5 0 or g(r1) < g(a)

a 5 q2r1 1 r2, where r2 5 0 or g(r2) < g(r1)
r1 5 q3r2 1 r3, where r3 5 0 or g(r3) < g(r2)
r2 5 q4r3 1 r4, where r4 5 0 or g(r4) < g(r3)
                          

Then, there exists n such that rn11 5 0 and rn21 5 qn11rn. If n is the least such
integer, then

rn 5 g.c.d. {a, b}.

Proof: First observe that, since g(a) > g(r1) > g(r2) > …, the above process
of getting qn’s and rn’s should terminate (at most after g(a) number of steps)
and hence rn11 5 0 for some n. Writing from bottom to top of the above equa-
tions, we have

rn21 5 qn11rn
rn22 5 qnrn21 1 rn, 0  rn 5 rn22 2 qnrn21
rn23 5 qn21rn22 1 rn21, 0  rn21 5 rn23 2 qn21rn22
                      
 r1 5 q3r2 1 r3, 0  r3 5 r1 2 q3r2
a 5 q2r1 1 r2, 0  r2 5 a 2 q2r1
b 5 q1a 1 r1, 0  r1 5 b 2 q1a.

Tracing from top to bottom of the left hand side equations, we get that
rn|rn21, rn|rn22, …, rn|a and rn|b. Therefore, rn is a common divisor of a and b.

Q001-Algebra-111001_CH 12.indd 39 9/16/2011 10:16:27 AM

 12-40   Algebra – Abstract and Modern

Also, if d is any common divisor of a and b, then tracing from bottom to top
of the equations on the right hand side above, we get that

d|a, d|b, d|r1, d|r2, d|r3, …, d|rn21 and d|rn.

Thus, rn is the g.c.d. of a and b. b

Worked Exercise 12.5.1. Let R be an Euclidean domain or a PID and a and

b  R 2 {0}. Then prove that the g.c.d. {a, b} exists and is of the form ax 1 by
for some x and y  R.

Answer: Since every Euclidean domain is a PID, we prove this result in a

PID. Consider the ideal aR 1 bR, Then, there exists d  R such that

aR 1 bR 5 <d> 5 dR.

Then, it can be easily verified that d is the g.c.d.{a, b} and d 5 ax 1 by for

some x and y  R.

Worked Exercise 12.5.2. Let R be an Euclidean domain with gauge function

g and a  R 2 {0}. Then prove that a is a unit in R if and only if g(ab) 5 g(b)
for some 0  b  R.

Answer: If a is a unit in R and 0  b  R, then

g(ab) 5 g(a)g(b) 5 1g(b) 5 g(b).

Also, if g(ab) 5 g(b) for some 0  b  R, then

g(a)g(b) 5 g(ab) 5 g(b).

and hence g(a) 5 1 (since g(b) > 0), so that a is a unit in R.

Worked Exercise 12.5.3. Let Z[ 2]  {a  b 2 : a and b ∈ }. Then prove

that Z[ 2] is an Euclidean domain under the usual addition and multiplica-
tion of real numbers.

Answer: It can be easily proved that Z[ 2] is a nontrivial subring of the

integral domain R and hence Z[ 2] is an integral domain. Define

g : Z[ 2]{0} → Z by g ( a  b 2)  | a 2  2b 2 | .

Q001-Algebra-111001_CH 12.indd 40 9/16/2011 10:16:28 AM

 Factorization in Integral Domains   12-41

Since a  b 2  0 implies a  0 or b  0 and hence |a 2  2b 2 |  0

(for, a2 5 2b2 has no integral solutions). Thus,

g (a  b 2)  0 for all a  b 2  Z[ 2].

Also, for any x 5 a 1 b 2 and y 5 c 1 d 2 in Z[ 2]{0},

g ( xy )  g (( a  b 2) (c  d 2))
 g ( ac  2bd  ( ad  bc) 2)
 | ( ac  2bd ) 2  2( ad  bc) 2 |
 | a 2c 2  4b 2 d 2  4 acbd  2a 2 d 2  2b 2c 2  4 adbc |
 | ( a 2  2b 2 ) || (c 2  2d 2 ) |
 g ( x ) g ( y ).

Next, let x 5 a 1 b 2 and 0  y 5 c 1 b d  Z[ 2].

Then, c  0 or d  0. Now,

x ( a  b 2) (c  d 2)
  a  b 2, say
y c 2  2d 2

where a and b are rational numbers. Choose integers m and n such that

1 1
| m a| and | n b| .
2 2

Then, x  y(a  b 2  ( m  n 2) y [(a  m)  ( b  n) 2] y .

Now, [(am)  ( bn) 2] y  x  ( m  n 2) y  Z[ 2]
Put rr
Put [(aa
[( mm))
((bb
nn)) 22 ]] yy.. Then,
Then, xx yq
 yq rr,,
where qq
where m nn 22 and
m and rr
00 oror

g ( r ) | (a  m) 2  2( b  n) 2 || c 2  2d 2 |
1 2
  | c 2  2d 2 || c 2  2d 2 | g ( y ).
2 4

Thus, Z[ 2] is an Euclidean domain.

Q001-Algebra-111001_CH 12.indd 41 9/16/2011 10:16:29 AM

 12-42   Algebra – Abstract and Modern

Worked Exercise 12.5.4. Determine the g.c.d. of 1 1 x 1 x2 and 1 1 2x 1

3x3 1 x5 in R[x].

Answer: We know that R[x] is an Euclidean domain. Let us follow the algo-
rithm given in Theorem 12.5.4.
Let f (x) 5 1 1 x 1 x2 and g(x) 5 1 1 2x 1 3x3 1 x5. Then,

g(x) 5 (22 1 3x 2 x2 1 x3)(1 1 x 1 x2) 1 x 1 3

1 1 x 1 x2 5 (x 2 2)(x 1 3) 1 7
1 
x 1 3 5  x  7  3
 7 
752?311
3 5 3 ? 1 1 0.

Therefore, 1 is the g.c.d. of 1 1 x 1 x2 and 1 1 2x 1 3x3 1 x5. This is better

understood by the following method.
x 5  3 x 3  2 x 1 3
x 2  x 1 5 3
x  x 2  3x  2
x  3x  x  3
x 2  x 1
x 3 x2
x2  x 6
x 3 1
7 x
x 7
7
3 2
6
3
1 3
3
0.

Before we close this section, let us summarize various classes of integral

domains introduced in this section and discuss their inter relationships. Let
us fix notation for these classes as given below.
ID 5 The class of integral domains
FD 5 The class of factorization domains
UFD 5 The class of unique factorization domains
PID 5 The class of principle ideal domains
ED 5 The class of Euclidean domains
F 5 The class of fields.

Q001-Algebra-111001_CH 12.indd 42 9/16/2011 10:16:29 AM

 Factorization in Integral Domains   12-43

Theorem 12.5.5. We have the following inclusions among the above classes

F ⊂ ED ⊂ PID ⊂ UFD ⊂ FD ⊂ ID

and these are strict inclusions.

Proof:
1. In Example 12.5.1 (2), we have proved that every field is an Euclidean
domain and therefore F ⊆ ED.
Z is an Euclidean domain, but not a field and hence F  ED.
2. In Theorem 12.5.2, we have proved that every Euclidean domain is a
PID and therefore ED ⊆ PID. Consider the ring R given by

{ b
R  a  (1 i 19 ) : a and b  Z .
2
}
Then, R is a PID, but not an Euclidean domain (the proof of this is little
bit involved and hence we skip the proof ). Therefore, ED  PID.
3. In Corollary 12.3.4, we have proved that every PID is an UFD and hence
PID ⊆ UFD. The ring Z[x] is an UFD but not a PID (see Theorem 12.4.3).
Therefore, PID  UFD.

ID
FD
UFD
PID
ED
F

4. Clearly, every unique factorization is a FD. We have seen in 12.3…. that

Z 5 is a FD which is not an UFD. Therefore, UFD  FD.
5. Clearly, every FD is an integral domain. There are integral domains
which are not FDs. That is, FD  ID. b

EXERCISE 12(e)

1. Let R be an Euclidean domain with gauge function g. For any a and b  R 2 {0},
prove that a and b are associates if and only if a divides b and f(a) 5 f(b).

Q001-Algebra-111001_CH 12.indd 43 9/16/2011 10:16:30 AM

 12-44   Algebra – Abstract and Modern

2. Prove that Z[x] is not an Euclidean domain.

3. Prove that Q[x], R[x] and C[x] are all Euclidean domains.

4. If R is an integral domain which is not a field, prove that R[x] is not an Euclidean
domain.

5. Prove that any nonzero prime ideal in an Euclidean domain is maximal.

6. In any integral domain with gauge function g, prove that g(a) 5 g(2a) for any
nonzero element a.

7. Prove that, for any n  Z1, g : Z 2 {0} → Z1 defined by g(a) 5 |a|n is a gauge
function on Z.

8. Prove that Z[ 3]  {a  b 3 : a and b  Z} is a Euclidean domain.

9. Let R be an Euclidean domain and a, b  R with b  0. Let q, r  R such that

a 5 bq 1 r with r  0. Then prove that g.c.d. {a, b} 5 g.c.d. {b, r}.

10. Let x 102 10 3 and y 1 7 3. Find q and r in Z[ 3] such that x 5 yq 1 r
where either r 5 0 or r  a  b 3 with |a2 2 3b2| < 146.

12.6 SOME APPLICATIONS TO NUMBER THEORY

In this section, we apply the general results proved about Euclidean rings
to the ring of Gaussian integers and obtain a relatively difficult theorem
about prime numbers due to the famous mathematician, Fermat. First
recall that

Z[i] 5 {a 1 bi : a and b are integers}

is an integral domain under the usual addition and multiplication of complex

numbers. In fact Z[i] is a subring of the field of complex numbers. Also,
recall that 1, 21, i and 2i are the only units in Z[i] and that Z[i] is called the
ring of Gaussian integers.

Theorem 12.6.1. The ring of Gaussian integers Z[i] is an Euclidean domain.

Proof: Define g : Z[i] 2 {0} → Z1 by

g(a 1 bi) 5 a2 1 b2.

Note that a 1 bi  0 in Z[i] implies that a  0 or b  0 and hence a2 1 b2

is a positive integer.

Q001-Algebra-111001_CH 12.indd 44 9/16/2011 10:16:31 AM

 Factorization in Integral Domains   12-45

For any x 5 a 1 bi and y 5 c 1 di  Z[i]

g(xy) 5 g((a 1 bi)(c 1 di))

5 g((ac 2 bd) 1 (ad 1 bc)i)
5 (ac 2 bd)2 1 (ad 1 bc)2
5 a2c2 1 b2d2 1 a2d2 1 b2c2
5 (a2 1 b2)(c2 1 d2)
5 g(x)g(y).

To prove the Euclidean division algorithm, let x 5 a 1 bi and 0  y 5 c 1

di  Z[i]. Then, c2 1 d2 > 0. Consider the complex number

x (a  bi)(c  di) (ac  bd )(bc  ad )i

   a  bi
y (c  di)(c  di) c2  d 2

for some rational numbers a and b. Now, choose integers m and n such that

1 1
|am |  and |bn | 
2 2

    we have x 5 (a 1 bi)y

   5 (m 1 ni)y 1 ((a 2 m) 1 (b 2 n)i)y.

Put q 5 m 1 ni and r 5 [(a 2 m) 1 (b 2 n) i]y.

Then, clearly q  Z[i] and

r 5 x 2 (m 1 ni)y  Z[i].

Now, x 5 qy 1 r and either r 5 0

        or g(r) 5 ((a 2 m)2 1 (b 2 n)2)g(y)

1 1
    g ( y ) g ( y).
 4 4 

Thus, Z[i] is an Euclidean domain. b

Corollary 12.6.1. Z[i] is a PID and an UFD.

Now, we are free to apply the properties of PIDs and UFDs to Z[i] to prove
the following purely number theoretic results.

Q001-Algebra-111001_CH 12.indd 45 9/16/2011 10:16:31 AM

 12-46   Algebra – Abstract and Modern

Theorem 12.6.2. Let p  Z1 be prime and n  Z such that p does not divide
n. Suppose that we can find integers x and y such that np 5 x2 1 y2. Then, p
can be expressed as a sum of two squares of integers; that is, p 5 a2 1 b2 for
some a and b  Z.

Proof: First observe that any integer m can be treated as a Gaussian inte-
ger m 1 0i and that Z is a subring of Z[i]. Next, we prove that p cannot
be a prime element in Z[i]. Suppose, if possible, that p is prime in Z[i].
Since

pn 5 x2 1 y2 5 (x 1 iy)(x 2 iy),

we get that p divides either x 1 iy or x 2 iy. If p divides x 1 iy then

p(s 1 it) 5 x 1 iy

for some s and t  Z and hence ps 5 x and pt 5 y, so that p(s 2 it) 5 x 2 iy.
Therefore,

pn 5 (x 1 iy)(x 2 iy) 5 p(s 1 it)p(s 2 it) 5 p2(s2 1 t2)

from which it follows that p2 divides pn and hence p divides n, which is a con-
tradiction to our hypothesis. Therefore, p is not a prime in Z[i]. Since Z[i] is
an UFD, an element in Z[i] is prime if and only if it is irreducible. Therefore,
p is not an irreducible element in Z[i]. Therefore, there exists two nonunits
a 1 bi and c 1 di in Z[i] such that

p 5 (a 1 bi)(c 1 di)
and hence p2 5 g(p) 5 g(a 1 bi)g(c 1 di),

where g is the gauge function on Z[i] defined in Theorem 12.6.1. Therefore,

p2 5 (a2 1 b2)(c2 1 d2)

and hence a2 1 b2 is a divisor of p2. Therefore, a2 1 b2 5 1 or p or p2. But,

since a 1 bi and c 1 di are nonunits in Z[i], a2 1 b2 > 1 and c2 1 d2 > 1.
Again, since c2 1 d2 > 1, if follows from p2 5 (a2 1 b2)(c2 1 d2) that
a2 1 b2 5 p. b
The following is a famous theorem in elementary number theory and we
state this without proof. First recall that, for integers a, b and n, we write a ≡ b
(mod n) when n divides a 2 b.

Q001-Algebra-111001_CH 12.indd 46 9/16/2011 10:16:31 AM

 Factorization in Integral Domains   12-47

Theorem 12.6.3 (Wilson’s Theorem). For any prime number p in Z1,

(p 2 1)! ≡ 21 (mod p).

Note that any positive integer m must be of the form 4n or 4n 1 1 or 4n 1 2

or 4n 1 3, for some integer n  0. In particular, when m is a prime > 2, then
m 5 4n 1 1 or 4n 1 3 for some n  0. In the following two results, we prove
that primes of the form 4n 1 1 have certain special properties; in particular,
we prove that each such prime is a sum of two perfect squares.

Theorem 12.6.4. Let p be a prime number of the form 4n 1 1. Then, there

exists an integer x such that

x2 ≡ 21 (mod p).

Proof: First note that, since p 5 4n 1 1, ( p1) / 2 is an even integer. Now,

put x  (( p 1) / 2)!. Then,

 p 1  p 1
x  1  2  3     (1)(2)(3)   
 2   2 

since, (p21)/2 is even. Also, since p 2 a ≡ 2a (mod p) for any a  Z, we have

  p 1   p 1
x 2  1  2  3    (1)(2)(3)   
  2    2 

  p 1   p 1
 1  2  3    ( p 1)( p  2)( p  3)   p   (mod p)
  2    2 

p 1 p 1 p  3
1 2  3  ⋅ ⋅  ( p  2)( p 1)
2 2 2
 ( p 1)!  1 (mod p).

Thus, x2  21 (mod p). b

Theorem 12.6.5 (Fermat’s Theorem). Any prime number of the form 4n 1 1

can be expressed as a sum of two perfect squares.

Proof: Let p be a prime number and p 5 4n 1 1. By the above theorem,

there exists an integer x such that x2  21 (mod p). x can be chosen in such

Q001-Algebra-111001_CH 12.indd 47 9/16/2011 10:16:32 AM

 12-48   Algebra – Abstract and Modern

a way that 0  x  p 2 1; for, consider the remainder of x on division by p.

That is, let

x 5 qp 1 r, where 0  r < p.

Then, r2 5 (x 2 qp)2 5 x2 1 p(q2p 2 2xq) ≡ 21 (mod p). Further, we can

assume that r  (p/2); for, if r > (p/2), then s 5 r 2 p satisfies the property
that s2 ≡ 21 (mod p) and 0 < s  (p/2). Thus, we can find an integer s such
that

s2 ≡ 21 (mod p) and 0  s  (p/2).

Now, p divides s2 1 1 and hence pn 5 s2 1 12.

Also, since pn 5 s2 1 1  (p2/4) 1 1 < p2, we get that p does not divide n.
Thus, by Theorem 12.6.2, it follows that p 5 a2 1 b2 for some integers
a and b. b

Worked Exercise 12.6.1. Prove that any integer of the form 4n 1 3 cannot be
expressed as the sum of two perfect squares.

Answer: Suppose, if possible, that

4n 1 3 5 a2 1 b2

for some n, a, b  Z. Since 4n 1 3 is odd, one of a2 and b2 must be odd and

the other must be even. Therefore, one of a and b must be odd and the other
must be even. Without loss of generality, we can suppose that a 5 2r and b 5
2s 1 1 for some r and s  Z. Now, we have

4n 1 3 5 (2r)2 1 (2s 1 1)2 5 4r2 1 4s2 1 4s 1 1

and therefore, 2 5 4(r2 1 s2 2 s 2 n) so that 1 5 2(r2 1 s2 2 s 2 n), which is

absurd. Thus, 4n 1 3 cannot be expressed as a sum of two squares of integers.

Worked Exercise 12.6.2. Let I be a nonzero ideal of the ring Z[i] of Gaussian
integers. Then prove that the quotient ring Z[i ] / I is finite.

Proof: By Corollary 12.6.1, Z[i] is a PID and hence I 5 <x> for some x 
Z[i]. Let x 5 a 1 bi with a and b  Z. Since I is nonzero, we have x  0
and hence a  0 or b  0, so that g(x) 5 a2 1 b2 > 0, where g is the gauge
function on Z[i]. Any element of Z[i ] / I is of the form y 1 I, y  Z[i]. Now,

Q001-Algebra-111001_CH 12.indd 48 9/16/2011 10:16:32 AM

 Factorization in Integral Domains   12-49

if y  Z[i], then by Euclidean division algorithm, there exist elements q and

r  Z[i] such that

y 5 qx 1 r, where r 5 0 or g(r) < g(x)

and hence y 2 r 5 qx  <x> 5 I, so that

y 1 I 5 r 1 I, r 5 0 or g(r) < g(x).

Z[i ]
Therefore, I  {r  I : r  0 or g ( r )  g ( x )}.
Since there can be only finitely many pairs (c, d) of integers such that c2 1
d2 < g(x) 5 a2 1 b2, it follows that Z[i ] / I is a finite set. b

Exercise 12(f )

1. In each of the following two elements x and y of Z[i] are given. Find q and r in
Z[i] such that x 5 qy 1 r; with r 5 0 or g(r) < g(y).
(i) x 5 3 1 2i and y 5 2 2 3i
(ii) x 5 5 and y 5 2i
(iii) x 5 1 1 i and y 5 2 1 i
(iv) x 5 2 1 3i and y 5 1 2 i
(v) x 5 4 2 5i and y 5 5 1 4i.

2. Prove the neither 2 nor 17 is a prime element in Z[i].

3. Prove that Z 2 and Z 2 are Euclidean domains.

4. Prove that any nonzero prime ideal in Z[i] is maximal.

5. Prove that Z 6 is not a Euclidean domain.

6. Prove that Z 5 is not a Euclidean domain.

7. Determine all prime elements in Z[i].

8. Prove that 2 2 7i and 2 1 11i are relatively prime in Z[i].

9. Find the g.c.d. of 25 1 10i and 3 1 i in Z[i].

10. Find x and y  Z[i] such that g.c.d. {25 1 10i, 3 1 i} 5 (25 1 10i)x 1
(3 1 i)y.

Q001-Algebra-111001_CH 12.indd 49 9/16/2011 10:16:33 AM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 12.indd 50 9/16/2011 10:16:33 AM

 13
Modules and Vector
Spaces
13.1 Modules and Submodules
13.2 Homomorphisms and Quotients of Modules
13.3 Direct Products and Sums
13.4 Simple and Completely Reducible Modules
13.5 Free Modules
13.6 Vector Spaces

Another important algebraic structure is that of a module over a ring, in particu-

lar, a vector space over a division ring or a field. Till now, we have come across
groups and rings, where one or two binary operations are involved. Modules
are concerned with one binary operation and several binary operations, one
corresponding to each element in the ring. Consider an abelian group (M, 1)
and let R 5 End(M, 1), the set of all endomorphisms of (M, 1). It is well
known that R is a ring under point-wise addition and composition of mappings
(as multiplication). For each f  R and x  M, let fx be the image of x under f
in M. Then, one can easily see that the following conditions are satisfied for any
x and y  M and f and g  R:
1. f (x 1 y) 5 fx 1 fy
2. (f 1 g)x 5 fx 1 gx
3. (fg)(x) 5 f (gx)
4. 1x 5 x, where 1 is the unity in R.
This situation is abstracted in this chapter to introduce the concept of a
module over a ring and prove certain elementary properties of modules over
a given ring.

Q001-Algebra-111001_CH 13.indd 1 9/21/2011 4:45:13 PM

 13-2   Algebra – Abstract and Modern

13.1 Modules and Submodules

In this section, we define the notions of modules and submodules and ­discuss cer-
tain examples and properties of these. Let us begin with the following definition.

Definition 13.1.1. Let (R, 1, ?) be a ring and (M, 1) be an abelian group.

Then, M is called a left R-module (or a left module over R) if there exists a
mapping (a, x)  ax of R 3 M into M satisfying the following for any x and
y  M and a and b  R:
1. a(x 1 y) 5 ax 1 ay
2. (a 1 b)x 5 ax 1 bx
3. a(bx) 5 (ab)x
4. 1x 5 x, if 1 is the unity in R.

Example 13.1.1
1. Let (M, 1) be any abelian group and R 5 End(M), the ring of all endo-
morphisms of (M, 1), then M is a left R-module, where the map (f, x) 
fx of R 3 M into M is defined simply by fx 5 f (x), the image of x under f.
2. Let (M, 1) be any abelian group and consider the ring Z of integers. For
any n  Z and x  M, define

0 if n  0


nx  ( n 1) x  x if n  0.

(n)(x ) if n  0

That is, nx 5 0 if n 5 0 and,

nx 5 x 1 x 1 … 1 x (n times) if n > 0
and nx 5 (2x) 1 (2x) 1 … 1 (2x) (2n times) if n < 0.
Then, under this map M is a left Z-module.
3. Any ring R can itself be treated as a left R-module by defining ax to be
the product of a and x as elements of R, for any a  R and x  R.
4. Let m and n be any positive integers and R be any ring. Let Mm3n(R) be
the set of all m 3 n matrices over R. Then, clearly Mm3n(R) is an abelian
group under the usual addition of matrices. Now, for any a  R and A 5
(aij)  Mm3n(R), define
aA 5 (aaij).
Then, under this map ( a, A)  aA, Mm3n(R) is a left R-module.

Q001-Algebra-111001_CH 13.indd 2 9/21/2011 4:45:13 PM

 Modules and Vector Spaces   13-3

5. Let n be any positive integer and R be a ring. Then, M 5 R 3 R 3 … 3 R

(n factors) is an abelian group under the coordinate-wise addition. Also,
for any a  R and x 5 (x1, x2, …, xn)  M, define
ax 5 (ax1, ax2, …, axn).
Under this map, M is a left R-module. Actually, this is a special case of
(4) above, since R 3 … 3 R is precisely the set of all 1 3 n matrices over
R. However, this needs a special mention, in view of its importance.
When M is a left R-module, then we often call the map (a, x)  ax as the
scalar multiplication and ax is called the scalar multiplication or simply the
multiplication of x by a on the left. One can define a right R-module similarly
by considering the mappings (x, a)  xa from M 3 R into M satisfying the
properties similar to (1) through (4) of Definition 13.1.1. When R is a com-
mutative ring and M is a left R-module, then M can be made into a right
R-module by defining xa 5 ax (since (xa)b 5 b(xa) 5 b(ax) 5 (ba)x 5
(ab)x 5 x(ab), the multiplication in R being commutative). In this case, left
R-modules are same as right R-modules. When R is commutative, we do not
distinguish between left R-modules and right R-modules and simply call them
R-modules. The following are certain elementary properties of R-modules.

Theorem 13.1.1. Let M be a left R-module. Then, the following holds for any
x  M and a  R:
1. 0x 5 0, where 0 on the left is the zero element in the ring R and 0 on the
right is the zero element (identity element) in the group (M, 1).
2. a0 5 0, where 0 is the zero element in M.
3. (2a)x 5 2(ax) 5 a(2x).

Proof:
1. Since 0x 1 0 5 0x 5 (0 1 0)x 5 0x 1 0x, it follows that 0x 5 0.
2. Also, since a0 1 0 5 a0 5 a(0 1 0) 5 a0 1 a0, we get that a0 5 0.
3. We have 0 5 0x 5 (a 2 a)x 5 ax 1 (2a)x and hence (2a)x 5 2(ax). Also,
0 5 a0 5 a(x 1(2x)) 5 ax 1 a(2x)
and therefore a(2x) 5 2(ax).
Note that we are using the same symbol 0 to denote the additive identity (zero
element) in the ring as well as the identity element in the group M. This need
not create any ambiguity and we should take it as per the context. Further,
throughout our discussions in this chapter, all modules are assumed to be left
R-modules, unless otherwise stated. Further, a module over R will be denoted
by (M, 1, R) or simply by M when there is no ambiguity about the group
operation 1 on M and about the ring R over which M is a module.

Q001-Algebra-111001_CH 13.indd 3 9/21/2011 4:45:13 PM

 13-4   Algebra – Abstract and Modern

Definition 13.1.2. Let R 5 (R, 1, ?) be a ring and M 5 (M, 1, R) be an

R-module. A nonempty subset N of M is called an R-submodule (or simply,
a submodule) of M if N is a subgroup of (M, 1) and rx  N for all r  R
and x  N.
Before going for certain examples of submodules of modules, let us men-
tion the following whose proof is an easy exercise.

Theorem 13.1.2. Let R be a ring with unity and M be an R-module. Then, the
following are equivalent to each other for any subset N of M:
1. N is an R-submodule of M.
2. N  [ and ax 2 by  N for all a, b  R and x, y  N.
3. 0  N and ax 2 by  N for all a, b  R and x, y  N.

Example 13.1.2
1. As in Example 13.1.1 (3), any ring R can be treated as an R-module and
the R-submodules of R are precisely the left ideals of R.
2. Let R be any ring and R[x] be the set of all polynomials over R. Then,
R[x] is an R-module under the usual addition and multiplication of
polynomials (recall that elements of R can be treated as polynomials of
degree zero). For any n $ 0, let
Rn[x] 5 {f (x)  R[x] : f (x) 5 0 or deg f (x) # n}.

Then, Rn[x] is an R-submodule of R[x].

3. Let M be an R-module, x  M and
Rx 5 {ax : a  R}.

Then, Rx is an R-submodule of M, since 0 5 0x  Rx and, for any r, s, a

and b  R, we have
r(ax) 2 s(bx) 5 (ra 2 sb)x  Rx.

4. Let M be an R-module, x  M and

<x> 5 {ax 1 nx : a  R and n  Z}.
Then, <x> is an R-submodule of M, since
0 5 0x 1 0x  <x>
and, for any r, s, a, b  R and n, m  Z, we have
r(ax 1 nx) 2 s(bx 1 mx) 5 rax 1 rnx 2 sbx 2 smx
5 (ra 1 nr 2 sb 2 ms)x
5 (ra 1 nr 2 sb 2 ms)x 1 0x  <x>.

Q001-Algebra-111001_CH 13.indd 4 9/21/2011 4:45:14 PM

 Modules and Vector Spaces   13-5

5. If R is a ring with unity and M is an R-module, then, for any x  M,

<x> 5 Rx.

Theorem 13.1.3. The intersection of any class of R-submodules of an

R-module M is again an R-submodule of M.

Proof: Let {Mi}iI be a class of R-submodules of an R-module M and let

N  i∩ M i. Since any submodule should contain 0, it follows that 0  Mi for
I
all i  I and hence 0  N. Also,

a, b  R and x, y  N ⇒ a, b  R and x, y  Mi for all i  I

⇒ ax 2 by  Mi for all i  I
⇒ ax 2 by  N.

Thus, N is a submodule of M.
However, as usual, the union of submodules may not be a submodule. In
fact, as in the case of subgroups of a group, for any submodules N and K of an
R-module M, N ∪ K is a submodule of M if and only if either N ⊆ K or K ⊆ N.

Definition 13.1.3. Let M be an R-module and X be a subset of M. Then, the

intersection of all submodules of M containing X is called the submodules
generated by X and is denoted by <X>. Clearly, <X> is the smallest submod-
ule of M containing X. If X 5 {x}, then

<X> 5 <x> 5 {ax 1 nx : a  R and n  Z}.

Theorem 13.1.4. Let {Ni}iI be a class of R-submodules of an R-module

M. Then

∪N
i∈ I
i {
 x1  x2  xn : x j  N i ,
j
}
i1 , … , in  I .

and this will be denoted by  N i . In particular, for any R-submodules N1, …,

i∈ I
Nm of M,

<N1 ∪ … ∪ Nm> 5 {x1 1 … 1 xm : xj  Nj for 1 # j # m}.

{
Proof: Let N  x1  x2  x n : x j  N i , i1 , i2 , …, in  I .
j
}
We shall prove that N is the smallest R-submodule of M containing i ∪
I
Ni .
Clearly, Ni ⊆ N for all i  I and hence i ∪ N i ⊆ N . Also,
I

Q001-Algebra-111001_CH 13.indd 5 9/21/2011 4:45:16 PM

 13-6   Algebra – Abstract and Modern

a, b  R and x, y  N ⇒ a, b  R, x 5 x1 1 … 1 xn and y 5 y1 1 … 1 ym,

 where x j ∈ N i j , yk ∈ N ik, ij and ik  I
⇒ ax 2 by 5 ax1 1 … 1 axn 1 b(2y1) 1 … 1 b(2ym)
⇒ ax 2 by  N.

Therefore, N is an R-submodule of M containing i ∪ I

N i . On the other hand,
let P be any submodule of M containing i ∪ N i . Then, Ni
⊆ P for all i  I and
I
hence N ⊆ P. Thus,

N ∪N
i∈ I
i .

In particular, if N1, N2, …, Nm are submodules of M, then 0  Nj for all 1 #

j # n and hence

<N1 ∪ … ∪ Nm> 5 {x1 1 … 1 xm : xj  Nj for 1# j # m}.

Corollary 13.1.1. Let N and K be R-submodules of an R-module M and

N 1 K 5 {x 1 y : x  N and y  K}.

Then, <N ∪ K> 5 N 1 K, which is the smallest submodule of M containing

both N and K.

Definition 13.1.4. Let M be an R-module. If X 5 {x1, x2, …, xn} is a finite

subset of M, then the submodule of M generated by X is denoted by <x1, x2,
…, xn>; that is,

<X> 5 <x1, x2, …, xn>.

It can be easily verified that

n
 x1 , x2 , …, xn  x1  x2  xn  ∑  xi .
i1

M is said to be finitely generated if M 5 <x1, x2, …, xn> for some x1, x2, …,
xn  M. The elements x1, x2, …, xn are said to be generators of M and the set
X 5 {x1, x2, …, xn} is said to generate M.

Q001-Algebra-111001_CH 13.indd 6 9/21/2011 4:45:17 PM

 Modules and Vector Spaces   13-7

Definition 13.1.5. An R-module M is called cyclic if M 5 <x> for some x 

M. Recall from Example 13.1.2 (4) that M is a cyclic R-module if and only if
there exists x  M such that

M 5 {ax 1 nx : a  R and n  Z}.

If R is a ring with unity, then an R-module M is cyclic if and only if

M 5 Rx 5 {ax : a  R}

for some x  M.
Note that the set of generators of a module need not be unique. For exam-
ple, the Z-module Z is cyclic, since Z 5 <1> and also Z 5 <2, 3>. In fact,
we have following exercise.

Worked Exercise 13.1.1. Consider the ring Z of integers. Then, Z is a mod-

ule over itself. For each positive integer n, prove that the Z-module Z has an
n-element generating set X such that no proper subset of X generates Z.

Answer: Let n be a given positive integer. If n 5 1, then Z 5 <1> 5 <21>

and hence we are through. Let n > 1. Consider any distinct primes p1, p2, …,
pn. For each 1 # i # n, let

xi  p1 p2  pi1 pi1  pn  ∏p
j ≠i
j

and X 5 {x1, x2, …, xn}. Then, x1, x2, …, xn are relatively prime (since there is
no prime dividing all xi’s).
Hence, there exist integers y1, y2, …, yn such that

1 5 y1x1 1 y2x2 1 … 1 ynxn.

Therefore, any a  Z can be expressed as

a 5 (ay1)x1 1 (ay2)x2 1 … 1 (ayn)xn.

Thus, Z 5 <x1, x2, …, xn>. Further, for each 1# i # n, pi divides all xj, j  i
and, in fact

g.c.d. {xj : j  i} 5 pi

and hence   x j  pi  ≠ �

. Therefore, no proper subset of X gener-
j ≠i
ates Z.

Q001-Algebra-111001_CH 13.indd 7 9/21/2011 4:45:17 PM

 13-8   Algebra – Abstract and Modern

Worked Exercise 13.1.2. Let F be any field and n be a positive integer. Let M
be the set of all polynomials over F of degree less than n. Then, prove that M
is a finitely generated F-module and exhibit two distinct sets of generators of
M, each with n elements.

Answer: Under the usual addition of polynomials and scalar multiplication,

M is clearly an F-module. Also, any element of M can be expressed as

f (x) 5 a0 1 a1x 1 … 1 an21xn21

and therefore {1, x, x2, …, xn21} is a generating set for M and has n elements.
Also, for any a  0 in F,

{a, a 1 x, x2, …, xn21}

is a generating set for M; since x 5 (a 1 x) 2 a and 1 5 a21a, we get that

<a, a 1 x, x2, …, xn21> 5 <1, x, x2, …, xn21> 5 M.

EXERCISE 13(a)

1. State whether each of the following is true or false and justify your answer:
(i) Any ring R is a finitely generated R-module.
(ii) Any ring R with unity is a finitely generated R-module.
(iii) If R is a finite ring, then any R-module is finitely generated.
(iv) Any finite R-module is finitely generated.
(v) Any module over a finite ring is finite.
(vi) Any finite R-module has only a finite number of R-submodules.
(vii) Any finitely generated R-module has only a finite number of R-submodules.
(viii) If an R-module M is cyclic and M 5 <x>, then M 5 <ax> for each 0 
a  R.
(ix) If a is a unit in a ring R and M is an R-module, then <x> 5 <ax> for
each x  M.
(x) If M is a finite R-module, then R is finite.

2. Consider the real member field R and the R-module R4. Which of the following
are R-submodules of R4?
(i) {(a1, a2, a3, a4)  R4 : a2 5 0}
(ii) {(a1, a2, a3, a4)  R4 : a1 5 a2 5 a3 5 a4}
(iii) {(a1, a2, a3, a4) : a1 1 a2 5 0}
(iv) {(a1, a2, a3, a4) : a1 5 a2}

Q001-Algebra-111001_CH 13.indd 8 9/21/2011 4:45:17 PM

 Modules and Vector Spaces   13-9

(v) {(a1, a2, a3, a4) : a1  0}

(vi) {(a1, a2, a3, a4) : a1a2 5 0}
(vii) {(a1, a2, a3, a4) : a1 1 a2 # 0}
(viii) {(a1, a2, a3, a4) : a1 1 a2 1 a3 1 a4 $ 0}.

3. For any R-module M, prove that the set

Ann(M) 5 {a  R : ax 5 0 for all x  M}

is an ideal of R.

4. An R-module M is called faithful if Ann(M) 5 {0}. Give an example of a faithful

R-module and of an R-module which is not faithful.

5. Let R be a ring and M be the set of all mappings of R into itself. Prove that M is
an R-module under the operations defined by

(f 1 g)(a) 5 f (a) 1 g(a)

and (af)(b) 5 af (b)

for all f and g  M and a and b  R.

6. For any R-submodules P and Q of an R-module M, prove that

P 1 Q 5 Q ⇔ P ⊆ Q ⇔ P 5 P ∩ Q.
7. Let N, P and Q be R-submodules of an R-module M, such that N ⊆ P. Then,
prove that
N 1 (Q ∩ P) 5 (N 1 Q) ∩ P
(This is called the modular law.)

8. Give an example of three R-submodules N, P and Q of an R-module M such that

N ∩ (P 1 Q)  (N ∩ P) 1 (N ∩ Q).
9. Let M be an R-module. Prove that the set
RM 5 {a1x1 1 a2x2 1 … 1 aixi : ai  R and xi  M}

is an R-submodule of M.

10. Let M be an R-module and N0 be the intersection of all nonzero submodules of

M. If N0  {0}, prove that N0 is cyclic and N0 5 <x> for all 0  x  N0.

11. Let M be an R-module. An equivalence relation  on M is called an R-congru-

ence on M if,

(x, y)   and (s, t)   ⇒ (x 1 s, y 1 t)  

and (x, y)   ⇒ (ax, ay)   for all a  R.

Q001-Algebra-111001_CH 13.indd 9 9/21/2011 4:45:17 PM

 13-10   Algebra – Abstract and Modern

If  is an R-congruence on M and x  M, then the set

(x) 5 {y  M : (x, y)  }
is called the congruence class of x relative to  on M. Prove the following for any
R-congruence  on M:
(i) (0) is an R-submodule of M.
(ii) (x) 5 x 1 (0) for all x  M.
(iii) (x) 5 (y) ⇔ x 2 y  (0).
12. Let M be an R-module. For any R-submodule N of M, let
N 5 {(x, y)  M 3 M : x 2 y  N}.
Then, prove that N is an R-congruence on M and every R-congruence on M is of
the form N for some R-submodule N of M. Also prove that, for any R-submod-
ules A and B of M,
A ⊆ B ⇔ A ⊆ B.

13.2 HOMOMORPHISMS AND QUOTIENTS

OF MODULES
For a fixed ring R, a function from one R-module into another R-module which
preserves the operations is called a homomorphism. Note that with each ele-
ment a of the ring R there corresponds a binary operation on an R-module M
given by x  ax. A homomorphism should preserve these binary operations
also in addition to the group operation 1. To be more precise, we have the
following definition.

Definition 13.2.1. Let R be a ring and M and N be R-modules. A function

f : M → N is called an R-homomorphism (or simply, a homomorphism when
there is no ambiguity about the ring R) if the following are satisfied:
1. f (x 1 y) 5 f (x) 1 f (y) for all x and y  M.
2. f (ax) 5 af (x) for all a  R and x  M.
An R-homomorphism of R-modules is also called an R-linear mapping or,
simply, linear mapping. An R-homomorphism f : M → N is necessarily a
homomorphism of the group (M, 1) into the group (N, 1) and hence we have
following theorem.

Theorem 13.2.1. The following holds for any R-homomorphism f : M → N

of R-modules:
1. f (0) 5 0
2. f (2x) 5 2f (x) for all x  M.
3. f (x 2 y) 5 f (x) 2 f (y) for all x and y  M.

Q001-Algebra-111001_CH 13.indd 10 9/21/2011 4:45:18 PM

 Modules and Vector Spaces   13-11

Definition 13.2.2. The set of all R-homomorphisms of an R-module M into

an R-module N is denoted by HomR(M, N). An R-homomorphism of M into
itself is called an R-endomorphism, or simply, an endomorphism of M and the
set of all R-endomorphisms of M is denoted by EndR(M); that is, HomR(M,
N) 5 EndR(M).
As usual an R-homomorphism is called an R-monomorphism if it is injec-
tive, an R-epimorphism if it is surjective and an R-isomorphism if it is bijec-
tive. A bijective R-endomorphism is called an R-automorphism. If there is an
R-isomorphism of M onto N, then we say that M is R-isomorphic to N and
denote this by M > N. Being R-isomorphic is clearly an equivalence relation
on the set of R-modules, for any given ring R.

Definition 13.2.3. Let f : M → N be an R-homomorphism.

1. The set {x  M : f (x) 5 0} is called the kernel of f and is denoted by ker f.
2. The set {f (x) : x  M} is called the image of M under f and is denoted
by f (M).

Theorem 13.2.2. The following holds for any R-homomorphism f : M → N.

1. ker f is an R-submodule of M.
2. f (M) is an R-submodule of N.

Proof:
1. Since f (0) 5 0, 0  ker f. If a and b  R and x and y  ker f, then f (x)
5 0 5 f (y),
f (x 2 y) 5 f (x) 2 f (y) 5 0 2 0 5 0
      and f (ax) 5 af (x) 5 a0 5 0.

Therefore, ker f is an R-submodule of M.

2. Clearly, f (M) is a nonempty subset of N. If s and t  f (M), then s 5 f (x)
and t 5 f (y) for some x and y  M and hence
s 2 t 5 f (x) 2 f (y) 5 f (x 2 y)  f (M)

and as 5 af (x) 5 f (ax)  f (M) for all a  R and therefore f (M) is an

R-submodule of N.

Example 13.2.1
1. Let R be any ring and M and N be R-modules. Define f : M → N by f (x)
5 0 for all x  M. Then, f is an R-homomorphism and is called the zero
homomorphism. Note that ker f 5 M and f (M) 5 {0}.

Q001-Algebra-111001_CH 13.indd 11 9/21/2011 4:45:18 PM

 13-12   Algebra – Abstract and Modern

2. Let M be any R-module and define f : M → M by f (x) 5 x for all x  M.

Then, f is an R-endomorphism of M and is called the identity endomor-
phism of M. Here, ker f 5 {0} and f (M) 5 M.
3. Let R be any ring. For any positive integers m and n, the set Mm3n(R) of
all m 3 n matrices over R is an R-module under the usual addition and
scalar multiplication of matrices. For any m 3 n matrix A, define fA :
M13m(R) → M13n(R) by
fA(B) 5 BA for any 1 3 m matrix B.

Then, fA is an R-homomorphism. Note that M13m(R) 5 Rm and M13n(R)

5 Rn. Here, ker fA 5 {B  Rm : BA 5 (0)} and fA(Rm) 5 {BA : B  Rm}.
4. Let R be a commutative ring and M be an R-module. Let a be a fixed
elements of R and define fa : M → M by
fa(x) 5 ax for all x  M.

Then, fa is an R-endomorphism of M.
Fundamental theorem of homomorphisms and other isomorphism theo-
rems are analogous to those of groups and rings. Before going to these, we
formally define the notion of quotient module.

Theorem 13.2.3. Let N be an R-submodule of an R-module M and, for any

x  M, let

x 1 N 5 {x 1 s : s  N}.

Then, the set

M
 {x  N : x ∈ M }
N

forms an R-module under the operations defined by

(x 1 N) 1 (y 1 N) 5 (x 1 y) 1 N
and a (x 1 N) 5 ax 1 N

for any x and y  M and a  R.

Proof: As in the case of groups (since N is a subgroup of the abelian group

(M, 1)), two cosets x 1 N and y 1 N are equal if and only if x 2 y  N.
Also, any two cosets are either equal or disjoint. The operation 1 defined

Q001-Algebra-111001_CH 13.indd 12 9/21/2011 4:45:18 PM

 Modules and Vector Spaces   13-13

on M/N is well defined and (M/N, 1) is an abelian group. Also, for any
x and y  M and a  R,

x1N5y1N⇒x2yN
⇒ ax 2 ay 5 a(x 2 y)  N
⇒ ax 1 N 5 ay 1 N

and therefore the scalar multiplication defined in M/N is also well defined. It
is routine to verify all the axioms of an R-module for M/N. Thus, M/N is an
R-module.

Definition 13.2.4. The R-module M/N defined above is called the quotient
R-module (or simply quotient module) of M by N.
The proofs of the following two theorems are similar to those in groups
and rings.

Theorem 13.2.4 (Fundamental Theorem of R-homomorphisms) Let f : M → N

be an R-homomorphism of R-modules.
Then

M
> f ( M ).
ker f

If f is an R-epimorphism, then ( M / ker f ) > N .

Theorem 13.2.5. Let N be an R-submodule of an R-module M and M/N be

the quotient of M by N. Then, any R-submodule of M/N of the form A/N for
some R-submodule A of M containing N.

Worked Exercise 13.2.1. For any R-module M, prove that EndR(M) is a ring
with unity under the point-wise addition and the composition of mappings as
multiplication.

Answer: EndR(M) is the set of all R-endomorphisms of M (R-homomorphisms

of M into itself). For any f and g in EndR(M), we define f 1 g and f ? g by

(f 1 g)(x) 5 f (x) 1 g(x)

        and ( f ⋅ g )( x )  f ( g ( x ))

for all x  M. Since 1 is commutative in M, we have

Q001-Algebra-111001_CH 13.indd 13 9/21/2011 4:45:19 PM

 13-14   Algebra – Abstract and Modern

(f 1 g)(x 1 y) 5 f (x 1 y) 1 g(x 1 y)
5 f (x) 1 f (y) 1 g(x) 1 g(y)
5 f (x) 1 g(x) 1 f (y) 1 g(y)
5 (f 1 g)(x) 1 (f 1 g)(y)

and (f 1 g)(ax) 5 f (ax) 1 g(ax)

5 af (x) 1 ag(x)
5 a(f (x) 1 g(x))
5 a(f 1 g)(x)

for all x and y  M and a  R. Therefore, f 1 g is an R-endomorphism.

Similarly f ? g  EndR(M). The zero homomorphism acts as identity for 1 in
EndR(M). It can be easily verified that (EndR(M), 1) is an abelian group. Also
clearly the composition ο of mappings is associative. Further,

f ο (g 1 h) 5 f ο g 1 f ο h
and (f 1 g) ο h 5 f ο h 1 g ο h

for all f, g and h  EndR(M). Also, the identity homomorphism of M is the

identity for ο in EndR(M). Thus, EndR(M) is a ring with unity.

Worked Exercise 13.2.2. Let A and B be R-submodules of an R-module M

such that A ⊆ B, Then, prove that B/A is an R-submodule of M/A and

( M A) > M .
( B A) B
Proof: By Theorem 13.2.5, B/A is an R-submodule of M/A and hence we can
form the quotient module (M/A)/(B/A). Define f : M/A → M/B by f (x 1 A) 5
x 1 B for any x 1 A  M/A, x  M. For any x and y  M,

x1A5y1A⇒x2yA⊆B
⇒x2yB
⇒ x 1 B 5 y 1 B.

Therefore, f is well defined. It can be easily verified that f is an R-homomor-

phism. Also,

Q001-Algebra-111001_CH 13.indd 14 9/21/2011 4:45:19 PM

 Modules and Vector Spaces   13-15


 M 
M
ker f   x  A ∈ : f ( x  A)  The zero in 


 A B


 M 
 x  A ∈ : x  B  B
 A 
 M 
 x  A ∈ : x ∈ B
 A 
B
 .
A

Also, clearly f is a surjection. Therefore, by the Fundamental Theorem of

R-homomorphisms, it follows that

( M A) > M .
( B A) B
Worked Exercise 13.2.3. For any R-submodules A and B of an R-module M,
prove that ( A B) / A > B /( A ∩ B).
Proof: Clearly, A 1 B is an R-module (being an R-submodule of M) and A
is an R-submodule of A 1 B. Also, A ∩ B is an R-submodule of B. Define
f : B → ( A B) / A by f (b) 5 b 1 A for any b  B.
Since B ⊆ A 1 B, f is well defined. Clearly, f is an R-homomorphism. Also,
for any x 5 a 1 b  A 1 B with a  A and b  B, we have
x2b5aA
and hence x 1 A 5 b 1 A 5 f (b), b  B. Therefore, f is an R-epimorphism.
Further
 A B 
ker f  b ∈ B : f (b)  zero in 
 A 
5 {b  B : b 1 A 5 A}
5 {b  B : b  A}
5 A ∩ B.

By the fundamental theorem of R-homomorphisms,

B B A B
 > .
A ∩ B ker f A

Thus, ( A B) / A > B /( A ∩ B).

Q001-Algebra-111001_CH 13.indd 15 9/21/2011 4:45:20 PM

 13-16   Algebra – Abstract and Modern

Worked Exercise 13.2.4. Let R be a ring with unity and M an R-module.

Prove that M is cyclic if and only if M > ( R / I ) for some left ideal I of R.

Proof: Let I be a left ideal of R. Then, clearly R/I is a left R-module. Put
x 5 1 1 I. Then, for any a  R,

a 1 I 5 a(1 1 I) 5 ax  <x>

Therefore, ( R / I )  x  and hence R/I is a cyclic R-module. If M > ( R / I ),

then M is also cyclic.
Conversely, suppose that M is cyclic and M 5 <x>. Define f : R → M by
f (a) 5 ax for all a  R. Then, f is an R-epimorphism and let I 5 ker f.

ker f 5 {a  R : f (a) 5 0}
5 {a  R : ax 5 0}.

It can be easily verified that ker f is a left ideal of R. By the fundamental theo-
rem of R-homomorphisms,

R R
 > M.
I ker f

EXERCISE 13(b)

1. Which of the following are homomorphisms of modules?

(i) f : R3 → R; f (a1, a2, a3) 5 a1 1 a2
(ii) f : R → R; f (a) 5 2a
(iii) f : R3 → R; f (a1, a2, a3) 5 a3
(iv) A and B are R-submodules of an R-module M and f : A 3 B → M;
f (a, b) 5 a 1 b
(v) I is a left ideal of a ring R and f : R → R / I ; f (a) 5 a 1 I.
(vi) For any ring R and r  R, f : R → R; f (a) 5 ra.

2. Let M1, M2, …, Mn be R-modules. Prove that M1 3 M2 3 … 3 Mn is an R-module

under the coordinate-wise addition and the scalar multiplication defined by
a (x1, x2, …, xn) 5 (ax1, ax2, …, axn).
3. Let M and N be R-modules and P and Q be R-submodules of M and N, respec-
tively. Prove that M 3 N is an R-module and P 3 Q is an R-submodule of M 3 N
and ( M  N ) /( P  Q ) > ( M / P )  ( N / Q ) .

4. For any R-module M, prove that the set EndR(M) of all R-endomorphisms of M
is a subring of the ring End(M) of all endomorphisms of the group (M, 1).

Q001-Algebra-111001_CH 13.indd 16 9/21/2011 4:45:21 PM

 Modules and Vector Spaces   13-17

5. Let R be a ring with unity and consider R as a right R-module. Then, prove that
the ring R is isomorphic to the ring EndR(R) of all R-endomorphisms of R.
6. Let f : M → N be an R-homomorphism of R-modules and A and B be R-submod-
ules of M and N, respectively. Then, prove that f (A) and f21(B) are R-submodules
of N and M, respectively.
7. Prove that an R-homomorphism of R-modules is an R-monomorphism if and
only if its kernel is trivial (zero).
8. Let M be the set of all differentiable real valued functions defined on the set R of
real numbers. Then, prove that M is an R-module under the point-wise addition and
scalar multiplication. Prove that the derivative operator is an R-homomorphism of
M into RR and determine its kernel.
9. If M is a cyclic R-module, prove that the quotient M/N is also a cyclic R-module
for any R-submodule N of M.
10. Let M and N be R-module and M > N. Prove that M is cyclic (finitely generated)
if and only if N is so.
11. A sequence (finite or infinite) of R-modules and R-homomorphisms
f n1 fn f n1 f n2
… → M n1  → M n → M n1   →… is called exact if f i(Mi21) 5
ker fi11 for all i.
Suppose that the following diagram of R-modules and R-homomorphisms is
commutative.

A f B g C

� � �

D � E � F

That is, a ο  5  ο f and  ο  5  ο g. Prove the following:

(i) If ,  and a are injections, then so is d.
(ii) If ,  and g are surjections, then so is d.

12. Consider the group (Q, 1) of rational numbers as a Z-module. Prove that the
ring EndZ(Q) of Z-endomorphisms of Q is isomorphic to the ring Q of rational
numbers.
13. If R is a ring with unity and I is a left ideal of R such that R / I > R as R-modules,
then prove that there is an idempotent e in R such that Re 5 I. Is the converse true?

14. For any R-submodules A and B of an R-module M such that A 1 B 5 M, prove that
M M M
>  .
A∩ B A B

Q001-Algebra-111001_CH 13.indd 17 9/21/2011 4:45:22 PM

 13-18   Algebra – Abstract and Modern

15. Let M be an R-module and A1, A2, …, An be R-submodules of M such that

 
Ai ∩ Aj   M for all 1 # i # n.
 j≠i 
M M M
Prove that n
>  .
A1 An
∩A i1
i

16. Let f : M → N be an R-homomorphism of R-module and let

f 5 {(x, y)  M 3 M : f (x) 5 f (y)}.
Then, prove the following:
(i) f is an R-congruence on M.
(ii) f (0) 5 ker f
(iii) For any s  f (M), {x  M : f (x) 5 s} is a congruence class relative to
f in M.
(iv) f 5 {(x, y)  M 3 N : x 2 y  ker f}.

13.3 DIRECT PRODUCTS AND SUMS

For any given in R, the class of R-modules exhibits a special property that has
direct products and sums of arbitrary subclasses. First, let us introduce the
following definition.

Definition 13.3.1. Let R be a fixed given ring and {Mi}iI be any nonempty
class of R-modules. Let
 
M  a : I → ∪ M i : a(i ) ∈ M i for all i ∈ I .
 i∈ I


By the choice axiom, we observe that M is a nonempty set. For any  and 
 M and a  R, define
( 1 )(i) 5 (i) 1 (i)
and (a)(i) 5 a(i)
for all i  I. Then, M is an R-module under these addition and scalar multipli-
cation. This M is called the direct product of {Mi}iI and is denoted by  M i .
i∈ I
Each Mi is called a direct factor of the direct product M.
The direct product of any class of R-modules satisfies the following cat-
egorical property.

Theorem 13.3.1. Let M be the direct product of a given class of R-modules

{Mi}iI over a given ring R. Then, M satisfies the following properties:
1. There exist R-homomorphisms pi : M → Mi, for each i  I.

Q001-Algebra-111001_CH 13.indd 18 9/21/2011 4:45:22 PM

 Modules and Vector Spaces   13-19

2. If N is any R-module and qi : N → Mi, i  I, are R-homomorphisms, then

there exists a unique R-homomorphism q : N → M such that
pi ο q 5 qi for all i  I.

pi
M
Mi

q
qi

Proof:
1. For each i  I, define pi : M → Mi by
pi(a) 5 a(i) for all a  M.
Then, pi(a 1 ) 5 (a 1 )(i) 5 a(i) 1 (i) 5 pi(a) 1 pi()
and pi(aa) 5 (aa)(i) 5 aa(i) 5 api(a)
for all a and   M and a  R. Therefore, each pi, i  I, is an R-homo-
morphism.
2. Let N be any R-module and, for each i  I, let qi : N → Mi be an
R-homomorphism. Then, define q : N → M by
q(x)(i) 5 qi(x)
for all x  N and i  I. Then, q(x)  M for all x  N and q is a well-
defined mapping of N into M. For any x and y  N and a  R, we have

q(x 1 y)(i) 5 qi(x 1 y) 5 qi(x) 1 qi(y) 5 q(x)(i) 1 q(y)(i) 5 (q(x) 1

q(y))(i)
and q(ax)(i) 5 qi(ax) 5 aqi(x) 1 aq(x)(i) 5 (aq(x))(i)
for all i  I and hence
q(x 1 y) 5 q(x) 1 q(y) and q(ax) 5 aq(x).
Therefore, q is an R-homomorphism. Also, for any x  N,
(pi ο q)(x) 5 pi(q(x)) 5 q(x)(i) 5 qi(x)
and therefore pi ο q 5 qi for all i  I.
To prove the uniqueness of q, let q9 : N → M be any R-homomorphism
such that pi ο q9 5 qi for all i  I. Then, for any x  N,
(q9(x))(i) 5 pi(q9(x)) 5 qi(x) 5 q(x)(i)

Q001-Algebra-111001_CH 13.indd 19 9/21/2011 4:45:23 PM

 13-20   Algebra – Abstract and Modern

for all i  I, and hence q9(x) 5 q(x) for all x  N. Therefore, q9 5 q.

Thus, q is a unique R-homomorphism of N into M such that pi ο q 5 qi
for all i  I.
The converse of the above theorem is also true in the sense of the following.

Theorem 13.3.2. Let {Mi}iI be any class of R-modules and M be an R-module

satisfying the properties (1) and (2) of the above theorem. Then, M >  M i .
i∈ I

Proof: Let pi : M → Mi, i  I, be the R-homomorphisms satisfying (1) and (2)

above. Let N >  M i and, for each i  I, define qi : N → Mi by qi(a) 5 a(i)
iI
for all a  N. Then, by (2) above, there exists a unique R-homomorphism
q : N → M such that pi ο q 5 qi for all i  I.

pi
M
Mi

q
qi

Also, since the direct product N satisfies the properties (1) and (2) (from
­Theorem 13.3.1), there exists a unique R-homomorphism p : M → N such that

qi ο p 5 pi for all i  I.

qi
N
Mi

p
pi

Now, consider q ο p : M → M. We have

pi ο (q ο p) 5 (pi ο q) ο p
   5qi ο p
       5 pi 5 pi ο (IdM)

Q001-Algebra-111001_CH 13.indd 20 9/21/2011 4:45:24 PM

 Modules and Vector Spaces   13-21

pi
M
Mi

qοp
pi

By the uniqueness of the R-homomorphism in (2), we get that

q ο p 5 IdM.

Similarly, we can prove that p ο q 5 IdN. Therefore, p and q are bijections and
inverses to each other. In particular, p : M → N is an R-isomorphism.
Thus,

M > N  ∏ Mi .
i∈ I

Definition 13.3.2. Let {Mi}iI be a class of R-modules and M   M i . For

iI
each i  I, the map pi : M → Mi, defined by

pi(a) 5 a(i) for any a  M

is called the ith projection.

Theorem 13.3.3. Let M   M i be the direct product of R-modules {Mi}iI.

iI
Then, each projection pi : M → Mi is an R-epimorphism and ( M / ker pi ) > M i .

Proof: Choose some element   M. Fix i  I and let xi  Mi.

Define  : I → ∪ M j by
jI

 x if j  i
( j )   i .
a( j ) if j ≠ i

Then,   M and pi() 5 (i) 5 xi. Therefore, pi is an R-epimorphism. By

the fundamental theorem of R-homomorphisms,

M
> Mi .
ker pi

Q001-Algebra-111001_CH 13.indd 21 9/21/2011 4:45:25 PM

 13-22   Algebra – Abstract and Modern

If I 5 {1, 2, …, n} is a finite set and M1, M2, …, Mn are R-modules, then it

can be easily observed that
n

∏M ∏M
i∈ I
i
i1
i  M1  M 2  M n

which is the usual set of n-tuples (x1, x2, …, xn) with xi  Mi.
Next let us turn our attention to the concept of direct sum of a given family
of R-modules over a given fixed ring R.

Definition 13.3.3. Let {Mi}iI be a nonempty family of R-modules, where R

is a given fixed ring. An R-module M is called an external direct sum if the
following are satisfied:
1. For each i  I, these is an R-homomorphism fi : Mi → M.
2. If N is an R-module and, for each i  I, gi : Mi → N is an R-homomor-
phism, then there exists unique R-homomorphism g : M → N such that
g ο f i 5 gi for all i  I.
fi M
Mi

g
gi

In the following, we prove uniqueness (up to isomorphism) of the external

direct sum of a given family of R-modules.

Theorem 13.3.4. Let {Mi}iI be a nonempty family of R-modules, M and M9

be R-modules and {fi}iI and {f i9}iI be R-homomorphisms satisfying the prop-
erties (1) and (2) above. Then, there exists an R-isomorphism f 9: M → M9 such
that f 9 ο fi 5 f i9 for all i  I.

Proof: Since M and {fi}iI satisfy (2) with M9 in place of N and f i9in place of gi,
there exists an R-homomorphism f 9: M → M9 such that f 9 ο fi 5 fi for all i  I.
fi M
Mi

f′
f ′i

M′

Q001-Algebra-111001_CH 13.indd 22 9/21/2011 4:45:26 PM

 Modules and Vector Spaces   13-23

Also, since M9 and {f i9}iI satisfies (2) with M in place of N and f i in place of
gi, there exists an R-homomorphism f : M9 → M such that

f ο f i9 5 fi for all i  I.
f ′i M′
Mi

f
fi

Now, consider f ο f 9 : M → M. We have

(f ο f 9) ο fi 5 f ο f i9 5 fi 5 IdM ο fi for all i  I.

From the uniqueness of the R-homomorphism in (2), with M in place of N and

f i in place of gi, we get that f ο f 9 5 IdM. Similarly, by interchanging the roles
of M and M9, we can prove that f 9 ο f 5 IdM. Therefore, f and f 9 are bijections
and are inverses to each other. Thus, f 9 : M → M9 is an R-isomorphism and f 9 ο
f i 5 f i9 for all i  I.
Before taking up the proof of the existence of the external direct sums, let
us have the following notation.

Definition 13.3.4. Let {Mi}iI be a nonempty family of R-modules and

a   M i . Then, the set
i I
|a| 5 {i  I : a(i)  0 in Mi} is called the support of a.

Theorem 13.3.5. Let {Mi}iI be a nonempty family of R-modules and  M i

i I
be the direct product of Mi’s. Let

 
M  a ∈ ∏ M i : a is finite.
 i∈ I


Then, M is the external direct sum of {Mi}iI.

Proof: First observe that, for any a and    M i ,

i I

|2a|5|a|, |a 1 |⊆|a|∪|| and |aa|⊆|a|

for all a  R and hence M is an R-submodule of the direct product  M i .

i I
Therefore, M is an R-module.

Q001-Algebra-111001_CH 13.indd 23 9/21/2011 4:45:26 PM

 13-24   Algebra – Abstract and Modern

For each i  I, define fi : Mi → M by

 x if j  i
f i ( x )( j )  
0 if j ≠ i

for any x  Mi and j  I. Note that |fi(x)|⊆{i} and hence fi(x)  M for all x  Mi. It
can be easily verified that fi is an R-homomorphism. Now, let N be any R-module
and gi : Mi → N be an R-homomorphism for each i  I. Define g : M → N by

g (a)  ∑ gi (a(i )) for any a ∈ M .

i∈ I

Since || is finite, a(i) 5 0 and hence gi(a(i)) 5 0 for all but finite number of
i’s. Therefore,

g (a)  ∑ gi (a(i ))
i∈|a|

and the summation in the definition of g(a) is meaningful. For any a and  
M and a  R, we have

g (a  )  ∑ gi ((a  )(i ))
         i∈ I

 ∑ gi (a(i )  (i ))
i∈ I

 ∑ gi (a(i )) ∑ gi ((i )) g (a)  g ( )

           i∈ I i∈ I

and g ( aa)  ∑ gi (( aa)(i ))

i∈ I

 ∑ gi (aa(i ))
i∈ I

 ∑ agi (a(i ))
i∈ I

     a∑ gi (a(i )) 5 ag().

i∈ I

Therefore, g is an R-homomorphism. Also, for any x  Mi, we have

(g ο fi )( x )  g ( fi ( x )) ∑ g j ( fi ( x )( j )) gi ( x ),
j ∈I

Q001-Algebra-111001_CH 13.indd 24 9/21/2011 4:45:28 PM

 Modules and Vector Spaces   13-25

since fi(x)(j) 5 0 for j  i and fi(x)(i) 5 x. Therefore, g ο fi 5 gi for all i  I.

To prove the uniqueness of g, let us take an R-homomorphism g9 : M → N
such that g9 ο fi 5 gi for all i  I. Then, for any a  M,

g (a)  ∑ gi (a(i ))
i∈ I

 ∑ (g ′ ο f i )(a(i ))
     i∈ I

 ∑ g ′ ( f i (a(i )))
     i∈ I

 
          g ′ ∑ f i (a(i )) 5 g9(),
 i∈I 

since  f i (a(i ))   f i (a(i ))  a. Thus, g 5 g9 and hence g is the unique

iI i| a|
R-homomorphism such that g ο f i 5 gi for all i  I. Thus, M is the external
direct sum of {Mi}iI.

Corollary 13.3.1. Let M be the external direct sum of a family {Mi}iI of

R-modules. Then, there exists R-submodules {Ni}iI of M satisfying the
following:
1. Mi > Ni for all i  I.
2. Each nonzero element of M can be uniquely expressed as a sum x1 1 x2
1 … 1 xn where 0 ≠ x j ∈ N i j for 1 # j # n and i1, i2, …, in are distinct
members of the index set I.

Proof: From Theorems 13.3.4 and 13.3.5, we can take M as the R-submodule
of the direct product  M i given by
i∈ I

 
M  a ∈ ∏ M i : a is finite
 i∈ I


and fi : Mi → M as the R-homomorphism defined by

 x if j  i
f i ( x )( j )  
0 if j ≠ i

for any x  Mi and j  I.

Q001-Algebra-111001_CH 13.indd 25 9/21/2011 4:45:29 PM

 13-26   Algebra – Abstract and Modern

Now, put Ni 5 f i(Mi) for each i  I. Then, each Ni is an R-submodule of M.

It can be easily observed that each f i is an R-monomorphism and hence f i
can be treated as an R-isomorphism of Mi onto Ni. Therefore, Mi > Ni for
each i  I. Let 0  a  M. Then, the support |a| is a nonempty finite subset
of I. Let

|a| 5 {i1, i2, …, in}.

Put x j  f i j (a(i j ))∈ N i j for 1# j # n. Then,

a 5 x1 1 x2 1 … 1 xn

(by evaluating both sides at each k  I) and clearly this expression of a is

unique.

Definition 13.3.5. Let M be an R-module and {Ni}iI be a nonempty family

of R-submodules of M. If each element x of M can be uniquely expressed as
a sum

x  ∑ xi ,
i∈ I

with xi  Ni for all i  I and xi 5 0 for all but a finite number i’s, then M is
called the internal direct sum of {Ni}iI and denote this by M  ⊕ N i . In this
iI
case, each Ni is called a direct summand of M. If I is a finite set, say I 5 {1,
2, …, n}, then ⊕ N i will be written as N1 ⊕ N2 ⊕ … ⊕ Nn.
iI
If M is the external direct sum of R-modules {Mi}iI, then we have proved in
Corollary 13.3.1 that there are R-submodules {Ni} of M such that Ni > Mi and
M is the internal direct sum of {Ni}iI. On the other hand, if M is the internal
direct sum of R-submodules {Mi}iI, then we prove below that M is (isomorphic
to) the external direct sum of {Mi}iI.

Theorem 13.3.6. Let M be an R-module and M be the internal direct sum

of R-submodules {Mi}. Then, M is isomorphic to the external direct sum of
{Mi}iI.

Proof: Let N be the external direct sum of {Mi}iI. That is, by Theorem 13.3.5,

 
N  a ∈ ∏ M i : a is finite.
 i∈ I


Q001-Algebra-111001_CH 13.indd 26 9/21/2011 4:45:30 PM

 Modules and Vector Spaces   13-27

Define f : M → N as follows.

f(0) 5 0 and, if 0  x  M, then x 5 x1 1 x2 1 … 1 xn

where x j ∈ M i j , i1, i2, …, in  I and define

 x j if i  i j , 1  j  n
f ( x )(i )   .
0 otherwise


Then, it can be verified that f is an R-isomorphism.

Theorem 13.3.7. Let {Mi}iI be a family of R-submodules of an R-module M

such that M   M i . Then, the following are equivalent to each other:
iI
1. M  ⊕ M i .
i∈ I
2. x1 1 x2 1 … 1 xn 5 0, x j ∈ M i j and i1, i2, …, in  I imply that x1 5
x2 5 … 5 xn 5 0.

( )
3. M i ∩  M j {0} for each i ∈ I .
i ≠ j ∈I

Proof: (1) ⇒ (2): If M  ⊕ M i , then any element x of M can be uniquely

i∈ I
expressed as x 5 x1 1 x2 1 … 1 xn with x j ∈ M ij for 1# j # n and i1, i2, …,
in  I. Therefore, if x1 1 x2 1 … 1 xn 5 0 5 0 1 0 1 … 1 0, it follows from
the uniqueness that x1 5 0 5 x2 5 … 5 xn.
(2) ⇒ (3): Let i  I be fixed

 
xi ∈ M i ∩  ∑ M j  ⇒ xi  x j1  x jn , jk ≠ i for 1  k  n and x jk ∈ M jk
i ≠ j ∈I 

⇒ (xi )  x j1  x jn  0.

⇒ xi 5 0 (by (2)).

(
Therefore, M i ∩  M j 5 {0} for all i  I.
i ≠ j ∈I
)
(3) ⇒ (1): We are given that M =  M i .
iI
If x1 1 x2 1 … 1 xn 5 y1 1 y2 1 … 1 yn and xj, y j ∈ M i j , then

 
x1  y1  ( y2  x2 )  ( yn  xn ) ∈ M i1 ∩ ∑ M i   {0}
 i ≠i1 

Q001-Algebra-111001_CH 13.indd 27 9/21/2011 4:45:31 PM

 13-28   Algebra – Abstract and Modern

and hence x1 5 y1. Similarly, x2 5 y2, …, xn 5 yn. Thus, M  ⊕ M i .

iI
Observe that, if N1, N2, …, Nn are finite number of R-submodules of an
n n
R-module M such that M  ⊕ N i , then M >  N i . That is, for any finite
i1 i1
number of R-modules, their direct product and direct sum are equal (iso-
morphic).

Worked Exercise 13.3.1. Let M 5 R3 be consider M as an R-module. Let

x1 5 (2, 0, 0), x2 5 (0, 1, 0) and x3 5 (0, 0, 3). Then, prove that M 5 <x1>
⊕ <x2> ⊕ <x3>.

Answer: If x 5 (a1, a2, a3)  M, then

a1 a
x x1  a2 x2  3 x3   x1  x2  x3 
2 3

Therefore, M 5 <x1> 1 <x2> 1 <x3>.

Suppose yi  <xi> for i 5 1, 2, 3 such that y1 1 y2 1 y3 5 0.
Then yi 5 aixi for some ai  R for i 5 1, 2, 3 and

(0, 0, 0) 5 y1 1 y2 1 y3 5 a1x1 1 a2x2 1 a3x3.

5 a1(2, 0, 0) 1 a2(0, 1, 0) 1 a3(0, 0, 3)
5 (2a1, a2, 3a3)

and hence 2a1 5 0 5 a2 5 3a3 which imply that a1 5 0 5 a2 5 a3 so that y1 5

0 5 y2 5 y3 . Therefore, by Theorem 13.3.7, M 5 <x1> 1 <x2> 1 <x3>.

Worked Exercise 13.3.2. Let e1 and e2 be idempotents in a ring R and regard

R as a left R-module. Prove that

Re1 ⊕ R(e2 2 e2e1) 5 Re1 1 Re2

Answer: Let M 5 Re1 1 Re2, M1 5 Re1 and M2 5 R(e2 2 e2e1). Then, M is an

R-module and M1 and M2 are R-submodules of M. Since

e1 5 e1e1  Re1 ⊆ M
         and e2 5 e2e1 1 e2(e2 2 e2e1)  M1 1 M2,

it follows that Re1 1 Re2 ⊆ M1 1 M2 and therefore

M 5 M1 1 M2. Also,

x  M1 ∩ M2 ⇒ x 5 re1 and x 5 s(e22e2e1) for some r, s  R

       ⇒ x 5 re1e1 5 s(e2 2 e2e1)e1 5 s(e2e1 2 e2e1) 5 0

Q001-Algebra-111001_CH 13.indd 28 9/21/2011 4:45:32 PM

 Modules and Vector Spaces   13-29

and therefore M1 ∩ M2 5 {0}. By Theorem 13.3.7, it follows that

M 5 M1 ⊕ M2.

Worked Exercise 13.3.3. Let R be a ring with unity and I be a left ideal of R.
Consider R as a left R-module. Then, prove that I is a direct summand of R if
and only if I 5 Re for some idempotent e in R.

Answer: Suppose that I 5 Re and e2 5 e  R. Then, put J 5 R(1 2 e). Then,

I and J are R-submodules of R and, any x  R can be written as

x 5 xe 1 x(1 2 e)  I 1 J

and hence I 1 J 5 R. Also,

x  I ∩ J ⇒ x 5 re 5 s(1 2 e) for some r and s  R

⇒ x 5 xe 5 s(1 2 e)e 5 s0 5 0

and therefore I ∩ J 5 {0}. Thus, R 5 I ⊕ J and I is a direct summand of R.

Conversely suppose that I is a direct summand of R. Then, there exists an
R-submodule J of R such that R 5 I ⊕ J. Then, I and J are left ideals of R, R
5 I 1 J and I ∩ J 5 {0}. Since I  R 5 I 1 J, we get that

1 5 e 1 f for some e  I and f  J.

Now, e 5 e(e 1 f) 5 e2 1 ef and hence

e 2 e2 5 ef  I ∩ J 5 {0}.

Therefore, e 2 e2 5 0 or e is an idempotent of R. Also, since I is a left ideal

of R and e  I, we have Re ⊆ I. Further,

x  I ⇒ x 5 x(e 1 f) 5 xe 1 xf
     ⇒ x 2 xe 5 xf  I ∩ J 5 {0}
⇒ x 2 xe 5 0
⇒ x 5 xe  Re.

Thus, I ⊆ Re and hence I 5 Re.

EXERCISE 13(c)

1. Consider the R-module R4 and determine whether R4 is the direct sum of <x1>,
<x2>, <x3> and <x4> in each of the following cases:
(i) x1 5 (0, 2, 0, 3), x2 5 (0, 3, 0, 4), x3 5 (2, 0, 0, 5) and x4 5 (0, 6, 0, 9)

Q001-Algebra-111001_CH 13.indd 29 9/21/2011 4:45:32 PM

 13-30   Algebra – Abstract and Modern

(ii) x1 5 (0, 0, 1, 0), x2 5 (0, 2, 0, 0), x3 5 (3, 0, 0, 0) and x4 5 (0, 0, 0, 4)

(iii) x1 5 (3, 0, 0, 0), x2 5 (3, 3, 0, 0), x3 5 (3, 3, 3, 0) and x4 5 (3, 3, 3, 3)
(iv) x1 5 (0, 2, 2, 2), x2 5 (0, 3, 3, 3), x3 5 (0, 1, 1, 1) and x4 5 (0, 4, 4, 4)

2. Consider R4 as Z-module and determine whether R4 is the direct sum of <x1>,

<x2>, <x3> and <x4> in each of the above cases in Exercise 1.

3. Let I be any nonempty set and R be any ring. Then, RI, the set of all mappings of
I into R, is an R-module under the usual point-wise operations. Prove that there
exists a family {Mi}iI of R-modules such that Mi is R-isomorphic to R for each
i  I and R I >  M i .
i I

4. In the above exercise, let

R(I) 5 {a  RI : |a| is finite}.
Then, prove that there exists R-submodules {Ni}iI of R(I) such that Ni is
R-isomorphic to R for each i  I and R( I )  ⊕ N i .
i I

5. Let {Mi}iI be a family of R-modules and   J  I . If M   M i , A   M j

i I j∈ J
and B   M i , then prove that A and B are (isomorphic to) R-submodules of
i I  J
M and M > A ⊕ B.
6. Let {Mi}iI be class of R-modules and M be an R-module. For each i  I, let
pi : M → Mi be an R-homomorphism satisfying the properties (1) and (2) of
Theorem 13.3.1. Then, prove that each pi, i  I, is an R-epimorphism.

7. Let M   M i . Prove that there exists an R-submodule Ni of M such that

i∈ I
( M / N i ) > M i for each i  I and N i >  M j .
i≠ j∈I

8. Let M be the external direct sum of {Mi}iI and fi : Mi → M be the R-homomorphism

as in Definition 13.3.3 (1). Then, prove that fi is a R-monomorphism for each
i  I.

9. Let {Mi}iI and {Ni}iI be two families of R-modules and, for each i  I, let
f i : Mi → Ni be an R-homomorphism. Then, prove that there is a unique
R-homomorphism f :  M i →  N i such that f()(i) 5 fi((i)) for all a   M i
i∈ I i∈ I i∈ I
and i  I.

10. In the above exercise, if each fi is an R-isomorphism, then prove that f is an

R-isomorphism.

11. State results similar to the above two exercises for external or internal direct
sums of R-modules and prove them.

12. Let R be a ring with unity and I be a left ideal of R such that ( R / I ) > R regarded
as R-modules. Then, prove that I 5 Re for idempotent e in R and deduce that I is
a direct summand of R.

Q001-Algebra-111001_CH 13.indd 30 9/21/2011 4:45:33 PM

 Modules and Vector Spaces   13-31

13.4 SIMPLE AND COMPLETELY REDUCIBLE MODULES

A module with nontrivial scalar multiplication and without nontrivial sub-
modules is called a simple module. A module which can be expressed as a
sum of simple submodules is called completely reducible. In this section, we
briefly discuss about the simple modules and completely reducible modules.
First recall that, for an R-module M, RM is an R-submodule of M defined by

 n 
RM    ri xi : ri  R and xi ∈ M .
i1 

Definition 13.4.1. An R-module M is called simple or irreducible if RM  {0}

and M has no R-submodules except {0} and M.
Note that a simple module M is necessarily nonzero and the scalar multi-
plication is nontrivial, in the sense that, rx  0 for some r  R and x  M.

Example 13.4.1
1. Let R be a field or a division ring and consider R as left R-module. Then,
R has no left ideals except {0} and R. Therefore, R is a simple R-module.
2. Let F be a field and R 5 Mn(F), the ring of all n 3 n matrices over F. For
each 1 # i, j # n, let eij be the matrix whose ijth entry is 1 and all the other
entries are zero. Fix 1# k # n and consider M 5 Rekk. Then, M is a left
R-module. We prove that M is a simple R-module. Clearly, RM  {0}. Let
N  {0} be an R-submodule of M and let 0 ≠ ( aij ) ∈ N .
Then, since
n
( aij )  ∑ aik eik ,
i1

ajk  0 for some 1 # j # n and ajk has multiplicative inverse in F. Now,

 n 
ekk  ajk 1 ekj ∑ aik eik  ∈ N
 i1 

and hence Rekk ⊆ N so that N 5 M. Thus, M is a simple R-module. Note

that M is precisely the set of all n 3 n matrices over F in which every
entry is zero except possibly in the kth column.
3. Let R be a ring with unity and M be a minimal left ideal of R. Then,
clearly M is a simple R-module.
In general, a minimal left ideal of a ring R need not be a simple R-module;
for consider an abelian group (R, 1) of order p, where p is a prime ­number,

Q001-Algebra-111001_CH 13.indd 31 9/21/2011 4:45:34 PM

 13-32   Algebra – Abstract and Modern

and define ab 5 0 for all a and b in R. Then, R is a ring (without unity). R

is a minimal left ideal (in fact, it is the only nonzero left ideal) and not a
simple R-module, since RR 5 {0}. The following is an useful characteriza-
tion of simple modules over rings with unity.

Theorem 13.4.1. The following are equivalent to each other for any module
M over a ring R with unity:
1. M is a simple R-module.
2. There exists a maximal left ideal I of R such that M > (R/I).
3. M  {0} and M 5 Rx for any 0  x  M.

Proof: (1) ⇒ (2): Suppose that M is a simple R-module. Since R is a ring

with unity, M 5 RM  {0}. Choose 0  x  M. Then, Rx is a nonzero
R-submodule of M and, since M is simple, Rx 5 M. Define

f : R→M by f(a) 5 ax for all a  R.

Then, f is an R-epimorphism. Put I 5 ker f 5 {a  R : ax 5 0}. Then, I

is a left ideal of R and, by the fundamental theorem of R-homomorphisms,
( R / I ) > M . Now, since M is simple R-module, so is R/I. The R-submodules
of R/I are in one-to-one correspondence with the left ideals of R containing
I. Therefore, I and R are the only left ideals of R containing I. Thus, I is a
maximal left ideal of R and ( R / I ) > M .
(2) ⇒ (3): Let I be a maximal left ideal of R such that M > ( R / I ); without
loss of generality, we can assume that M5R/I. Since I  R, it follows that
M  {0} and RM 5 M  {0}. If 0  x  M, then

x5a1I for some a  R 2 I

Then, Ra 1 I is a left ideal of R containing I properly. By the maximality of I,

it follows that Ra 1 I 5 R. Therefore, Rx  R( a  I )  Ra  I  ( R / I )  M .
(3) ⇒ (1): If N is any nonzero R-Submodule of M and 0  x  N, then
M 5 Rx ⊆ N and hence N 5 M. Also, RM 5 M  {0}. Thus, M is a simple
R-module.
The following result, which is popularly known as Schur’s lemma, is
an important property of simple R-modules. Let us first recall that, for
any R-module M, the set EndR(M) of all R-endomorphisms of M forms
a ring with unity under the point-wise addition and the composition of
mappings as multiplication. Also note that an R-module M can also be
viewed as a module over the ring EndR(M), where the scalar multiplica-
tion is defined by

Q001-Algebra-111001_CH 13.indd 32 9/21/2011 4:45:35 PM

 Modules and Vector Spaces   13-33

fx 5 f(x)

for any f  EndR(M) and x  M.

Theorem 13.4.2. (Schur’s lemma) EndR(M) is a division ring for any simple
R-module M.

Proof: Let M be a simple R-module. We have already observed that the set
EndR(M) of all R-endomorphisms of M is a ring with unity under the point-
wise addition and composition of mappings as multiplication. Note that the
identity map is the unity element in the ring EndR(M). We have to only prove
that every nonzero element in EndR(M) has multiplicative inverse in EndR(M).
Let 0  f  EndR(M). Then, ker f  M and f(M)  {0}. Both ker f and f(M)
are R-submodules of M. Since M is simple, it follows that ker f 5 {0} and
f(M) 5 M. These imply that f is a bijection and hence f is an R-isomorphism,
so that the inverse map f21 exists and is an R-endomorphism. Therefore, f has
multiplicative inverse in EndR(M). Thus, EndR(M) is a division ring.

Definition 13.4.2. An R-module M is called completely reducible if there

exists a family {M} of simple R-submodules of M such that

M  ∑ M a  { x1  xn : xi ∈ M ai , ai ∈  for 1  i  n}.

a∈

Example 13.4.2
1. Clearly any simple R-module is completely reducible.
2. Let {M} be any family of simple R-modules and M be the (external)
direct sum of {M}. Then, M is completely reducible.
In fact, we prove below that any completely reducible module is neces-
sarily a direct sum of a family of simple R-modules.

Theorem 13.4.3. Let N be an R-submodule of a completely reducible

R-module M and {M} be a family of simple R-submodules of M such
that M   M a . Then, there exists a subset I of  such that  M a is a direct
a aI
sum and M  N ⊕ ( ⊕ M a ).
aI

Proof: Here, we use the Zorn’s lemma. Let

   
P   J ⊆  : ∑ M a is a direct sum and N ∩ ∑ M a   {0} .
 a∈ J

 a∈ J  

Q001-Algebra-111001_CH 13.indd 33 9/21/2011 4:45:35 PM

 13-34   Algebra – Abstract and Modern

If J 5 [, then by  M a we mean {0}. Since the empty set [ is a member of

a∈ J
P, we get that P is a nonempty class of subsets of . It can be proved that P is
closed under unions of chains (totally ordered subsets of P, with respect to the
inclusion ordering). Therefore, by Zorn’s lemma, P has a maximal member.
Let I be a maximal member of P and let

 
 a∈I  (
A  N ⊕ ∑ M a   N ⊕ ⊕ M a .
a∈ I
)
We prove that A 5 M. For this it is enough if we can prove that M ⊆ A for all 
 ∆. Let   ∆ be arbitrarily fixed. If   I, then M ⊆  M a ⊆ A. Suppose
a I
that  ∉ I. Now M ∩ A is an R-submodule of M. Since M is simple, we get
that M ∩ A 5 {0} or M. Suppose, if possible, M ∩ A 5 {0}. Then

( )
M  ∩ ⊕ M a  {0}
a∈I

and hence  M a is a direct sum and has zero intersection with N. This
aI ∪{ }
implies that I ∪ {}  P, which is a contradiction to the fact that I is a maxi-
mal member of P, since  ∉ I and I  I ∪ {}. Therefore, M ∩ A  {0} and
hence M ∩ A 5 M so that M ⊆ A. Thus,  M  ⊆ A and hence A 5 M.

Thus,  M a is a direct sum and M  N ⊕ ( ⊕ M a ).
aI aI

Corollary 13.4.1. Let {M} be a family of simple R-submodules of an

R-module M. Then, there exists a subset I of  such that M is the direct sum
of {M}I.

Proof: This follows from the Theorem 13.4.7 by taking N 5 {0}.

Corollary 13.4.2. An R-module M is completely reducible if and only if it is

a direct sum of a family of simple R-submodules of M.

Worked Exercise 13.4.1. Prove that any nonzero R-submodule of a com-

pletely reducible R-module M is completely reducible and is a direct sum-
mand of M.

Answer: Let M be a completely reducible R-module and N be a nonzero

R-submodule of M. Let {M} be a family of simple R-submodules of
M such that M  ⊕ M a . (by Corollary 13.4.2). By Theorem 13.4.3, there
a
exists a subset I of  such that

Q001-Algebra-111001_CH 13.indd 34 9/21/2011 4:45:37 PM

 Modules and Vector Spaces   13-35

M  N ⊕ ⊕ Ma .( a∈ I
)
Then,

N>
M

(( ⊕ M ) ⊕ ( ⊕ M )) >
a∈ I
a
a∈ − I
a
⊕ Ma .
(⊕ M )
a∈ I
a (⊕ M )
a∈ I
a
a∈I

Thus, N is completely reducible and N is a direct summand of M.

EXERCISE 13(d)

1. Which of the following are simple modules and which of them are completely
reducible?
(i) Z, as a Z-module
(ii) R, as a R-module
(iii) R, as a Q-module
(iv) Q, as a Z-module
(v) Q, as a Q-module
(vi) For any field F, F[x] as an F-module.

2. Let X be any nonempty set and R be the ring of all real numbers. Let
M 5 {a : X → R : |a| is finite}.
Prove that M is a completely reducible R-module and the M is simple if and only
if X is a singleton set.
3. Let R be ring of 2 3 2 matrices over a field F. Prove that R is a completely reduc-
ible R-module.

4. Let M be a completely reducible R-module and N be a proper R-submodule of

M. Then, prove that the quotient M/N is a completely reducible R-module.

5. Let R be a ring with unity. Prove that R as an R-module is completely reducible

if and only if every R-module M is completely reducible.

6. Prove that the direct sum of any family of completely reducible R-modules is
completely reducible.

13.5 FREE MODULES

Modules over fields are best examples of free modules and are called vec-
tor spaces. In the next section, we are going to have a detailed discussion on

Q001-Algebra-111001_CH 13.indd 35 9/21/2011 4:45:37 PM

 13-36   Algebra – Abstract and Modern

v­ ector spaces. In this section, we consider a generalization of vector spaces;

that is, a module over a general ring with unity, which satisfies one of the most
important properties of a vector space. Throughout this section, R stands for a
nonzero ring with unity, unless otherwise stated.

Definition 13.5.1. Let M be an R-module. A finite sequence {x1, x2, …, xn}

of distinct elements in M is said to be linearly independent if, for any a1, a2,
…, an  R,

a1x1 1 a2x21… 1 anxn 5 0 ⇒ a1 5 a2 5 … 5 an 5 0.

A finite sequence of distinct elements of M is said to be linearly dependent if

it is not linearly independent. A subset X of M is said to be linearly indepen-
dent if every finite sequence of distinct elements in X is linearly independent;
otherwise X is called linearly dependent. Clearly, any linearly independent set
does not contain 0.

Example 13.5.1
1. Consider R as an R-module. Then, clearly {1} is linearly independent.
2. For any positive integer n, consider Rn as an R-module. For each 1# i # n,
let ei denote the n-tuple in which the ith coordinate is 1 and all other coor-
dinates are 0. Then, {e1, e2, …, en} is linearly independent.
3. Let X be any nonempty set and consider RX as an R-module. For each x
 X, define ex : X → R by

1 if y  x
ex ( y )   .
0 if y ≠ x

Then, the set X9 5 {ex : x  X} is linearly independent in the R-module RX.

4. Consider the set R[x] of polynomials over R. Then, R[x] is an R-module
in which {1, x, x2, …} is a linearly independent set, since for any poly-
nomial f(x) 5 a0 1 a1x 1 … 1 anxn,

f(x) 5 0 ⇒ ai 5 0 for all i.

Definition 13.5.2. An R-module M is said to be a free R-module if there

exists a subset B of M such that
1. B is linearly independent and
2. M is generated by B as an R-module.
In this case, B is called a basis for M.

Q001-Algebra-111001_CH 13.indd 36 9/21/2011 4:45:37 PM

 Modules and Vector Spaces   13-37

In other words, M is called a free R-module if M has a basis. Of course,

there can be more than one basis. For example, in any field F, as an F-module,
every nonzero element constitutes a basis.

Example 13.5.2
1. Any ring R is a free R-module, since {1} is a basis.
2. Also, as in Example 13.5.1 (2), Rn is a free R-module. The set {e1, e2, …, en}
is a basis for Rn. Any element x 5 (a1, a2, …, an) can be expressed as

x 5 a1e1 1 a2e2 1 … 1 anen.

This basis {e1, e2, …, en} is called the standard basis of Rn.
3. The set R[x] of polynomials over a ring R is a free R-module, since {1, x,
x2, …} is a basis for R[x].
4. Consider the R-module RX as in Example 13.5.1 (3). Then, RX is not a
free R-module. However, the R-submodule M of RX given by
M 5 {f  RX : | f | is finite}

is a free R-module, since {ex : x  X} is a basis for M.

The following is an interesting result on its own and it serves as a good
example for free Z-modules. Recall that any abelian group (G, 1) can be
regarded as a Z-module.

Theorem 13.5.1. Consider a cyclic group (G, 1) and regard it as a Z-module.

Then, G is a free Z-module if and only if G is infinite (i.e., G is isomorphic
to the group Z of integers).

Proof: Suppose that G is finite. Then, there exists a positive integer n such
that na 5 0 for all a  G (for example, we can take n 5 |G|) and hence there
is no linearly independent set in G; in particular, G has no basis. Therefore, G
is not a free Z-module. Conversely, suppose that G is an infinite group. Since
G is given to be cyclic, we get G is isomorphic to Z as Z-module. Since Z is
a free Z-module, it follows that G is also a free Z-module.
The following result suggests an alternate proof of the above theorem.

Theorem 13.5.2. Let M be a free R-module with a basis {x1, x2, …, xn}. Then,
M > Rn, as R-modules.

Proof: Define f : Rn → M by

f(a1, a2, …, an) 5 a1x1 1 a2x2 1…1 anxn.

Q001-Algebra-111001_CH 13.indd 37 9/21/2011 4:45:37 PM

 13-38   Algebra – Abstract and Modern

Since M is generated {x1, x2, …, xn}, any element of M can be expressed as

a1x1 1 a2x2 1 … 1 anxn 5 f(a1, a2, …, an) and hence f is a surjection. It can
be easily verified that f is an R-homomorphism. Also,

(a1, a2, …, an)  ker f ⇒ f(a1, a2, …, an) 5 0

⇒ a1x1 1 a2x2 1 … 1 anxn 5 0
⇒ a1 5 a2 5 … 5 an 5 0.

Therefore, f is an R-monomorphism. Thus, f : Rn → M is an R-isomorphism

and M > Rn, as R-modules.
In the next few results, we exhibit certain special properties of finitely
generated free modules over commutative rings.

Theorem 13.5.3. Let M be a finitely generated free module over a com-

mutative ring R. Then, all basis of M are finite and have the same number of
elements.

Proof: Let B 5 {ei}iI be a basis of M. Since M is finitely generated, there is

finite set {x1, x2, …, xn} generating M. For each 1# j # n, we have

x j  ∑ a ji ei , a ji ∈ R
i∈ I

in which all but finite number of aji are zero. For each 1# j # n, let

Sj 5 {i  I : aji  0}
n
and S  ∩ S j . Then, each Sj and hence S are finite subsets of I.
j1
Let D 5 {ei : i  s}.
Then, D is a linearly independent finite set and generates M and hence D is a
basis. Since D ⊆ B and B is also a basis of M, if follows that D 5 B. Thus, B
is a finite basis of M.
Let C be any other basis of M and |C| 5 m and |B| 5 n. Then, by Theorem
13.5.2, M > Rm and M > Rn. Therefore, there exists an R-isomorphism g :
Rm → Rn and let h 5 g21. Suppose, if possible, that m  n. Without loss of
generality, we can suppose that m < n. Let {e1, e2, …, em} and {f1, f2, …, fn} be
the standard basis of Rm and Rn, respectively. Let

n
g (ei )  ∑ a ji f j for each 1  i  m
j1

Q001-Algebra-111001_CH 13.indd 38 9/21/2011 4:45:38 PM

 Modules and Vector Spaces   13-39

and

m
g1 ( f j )  ∑ bij ei for each 1  j  n.
i1

and let A 5 (aji) and B 5 (bij) be the corresponding n 3 m and m 3 n matrices,

respectively. Then, for each i,

n
(ei )  g −1 g (ei )  ∑ a ji g1 ( f j )
j1

n m
 ∑ a ji ∑ bkj ek .
j1 k1

m  n 
Therefore, ei     bkj a ji  ek for each 1 # i # m. Since {ei} are linearly
k1 j1 
independent, we get that

n 1 if k  i
∑b a ki   .
j1
kj ji
0 if k ≠ i

Thus, the matrix product BA 5 Im.

Consider the augmented matrices

 B
A [ A 0 ] and B   .
0
 
These are n 3 n matrices, where each of the 0 blocks is a matrix of appropri-
ate size. Then,

I 0
A9B9 5 In and B A  m .
0 0

This implies that det(A9B9) 5 1 and det(B9A9) 5 0. This is a contradic-
tion, since A9 and B9 are n 3 n matrices over the commutative ring R
and det(A9B9) 5 det(B9A9). Thus, m < n is impossible. Similarly, n < m is
impossible. Thus, m 5 n.

Definition 13.5.3. Let M be a finitely generated free module over a commu-

tative ring R with unity. The number of elements in any basis of M is called
the rank of M and denoted by rank(M).

Q001-Algebra-111001_CH 13.indd 39 9/21/2011 4:45:39 PM

 13-40   Algebra – Abstract and Modern

Worked Exercise 13.5.1. Let e be an idempotent in a commutative ring with

unity, e  0 and e  1. Then, prove that Re is not a free R-module.

Answer: Clearly, Re is an R-module. If 0  x  Re, then x 5 re for some

r  R and

(1 2 e)x 5 (1 2 e)re 5 r(1 2 e)e 5 r(e 2 e2) 5 0

and 1 2 e  0. Therefore, there are no linearly independent sets in Re. Thus,

the R-module Re is not a free R-module.

Worked Exercise 13.5.2. Let R be a ring with unity, M an R-module and

X ⊆ M. Then, prove that X is a basis of M if and only if, for any R-module
N and for any mapping f : X → N, there exists unique R-homomorphism
f : M → N such that f ( x )  f ( x ) for all x  X.

Answer: Suppose that X is basis of M. Let N be any R-module and f : X → N

any mapping. Define f : M → N by
n n
f ( y )  ∑ ai f ( xi ) if y  ∑ ai xi ai ∈ R, xi ∈ X .
i1 i1

Since any element of M can be uniquely expressed as a1x1 1 … 1 anxn with

ai  R and xi  X, f is well defined. It can be easily verified that f is an
R-homomorphism and f ( x )  f ( x ) for all x  X.
Conversely, suppose that the given condition is satisfied. First, we prove
that X generates M. Put

 n 
M ′  X  ∑ ai xi : ai ∈ R and xi ∈ X .
 i1 

Then, M9 is an R-submodule of M and consider the quotient module M/M9.

Define

M M
f :X → by f(x) 5 M9, the zero element in .
M′ M ′
Then, the natural map g : M → M defined by g(y) 5 y 1 M9 and the zero
M
homomorphism h : M → M defined by h(y) 5 M9 for all y  M are both
M
R-homomorphisms such that

g(x) 5 f(x) 5 h(x) for all x  X.

Q001-Algebra-111001_CH 13.indd 40 9/21/2011 4:45:40 PM

 Modules and Vector Spaces   13-41

By the uniqueness of the R-homomorphism, it follows that g 5 h; that is,

y 1 M9 5 g(y) 5 h(y) 5 M9 for all y  M

so that y  M9 for all y  M and hence M 5 M9 5 <x>.

Thus, X generates M.
Next, we prove that X is linearly independent. Let x1, x2, …, xn be distinct ele-
ments in X and a1, a2, … , an  R such that

a1x1 1 a2x2 1 … 1 anxn 5 0.

We have to prove that ai 5 0 for each 1 # i # n. Let 1 # i # n be fixed. Con-

sider R as an R-module and define

1 if x  xi
f : X → R by f ( x )  
0 if x ≠ xi .

Then, there exists an R-homomorphism f : M → N such that

f ( x )  f ( x ) for all x  X.

Now, we have
0  f ( a1 x1  an xn )  a1 f ( x1 )  an f ( xn )  ai since f ( xi )  1 and
f ( x j )  0 for all j  i. Therefore, ai 5 0 for each 1# i # n. Thus, X is lin-
early independent. Thus, X is a basis of M.

EXERCISE 13(e)

1. Which of the following R-modules M are free? Justify your answers.

(i) R5R5M
(ii) R 5 R and M 5 Rn, n  Z1
(iii) R 5 R2 and M 5 R 3 {0}
(iv) R 5 Z and M 5 Zn, n  Z1
(v) R5Z5M
(vi) R 5 Z2 and M 5 {0} 3 Z
(vii) R 5 Z3 and M 5 Z 3 {0} 3 Z
(viii) R 5 Q and M 5 Qn, n  Z1.

2. Let B be a basis of a free R-module M. Then, prove that M  ⊕ Rx.

x∈ X

3. Prove that the direct sum of a family of free R-modules is again a free R-module.
Is this true for direct products?

Q001-Algebra-111001_CH 13.indd 41 9/21/2011 4:45:41 PM

 13-42   Algebra – Abstract and Modern

4. If M1, M2, …, Mn are free R-modules, prove that the direct product M1 3 M2 3
… 3 M is a free R-module.
n

5. Prove that Q is not a free Z-module.

6. Prove that every ideal of Z is a free Z-module.
7. Let R be an integral domain and x  R. Then, prove that Rx is a free R-module.
8. Prove that every R-module is a homomorphic image of a free R-module.
9. Let M and N be R-modules and M > N. Prove that M is a free R-module if and
only if N is a free R-module.
10. Prove that any finitely generated R-module is isomorphic to a quotient of a free
R-module.
11. If B is a basis for an R-module M, then prove that B is a minimal generating set;
that is, no proper subset of B generates M.
12. Prove that any basis B of a free R-module M is a maximal linearly independent
set in M; that is, any subset of M containing B properly is linearly dependent.
13. Let R be a ring with unity, M be an R-module and X ⊆ M. Prove that X is a basis
of M if and only if any element y of M can be uniquely expressed as
y 5 a1x1 1 … 1 anxn
where ai  R and x1, x2, …, xn are distinct elements of X.
14. Let F be a free R-module and M be another R-module. If N is an R-submodule
of M and f : F → ( M / N ) is an R-homomorphism, then prove that there exists
an R-homomorphism g : F → M such that
f(x) 5 g(x) 1 N for all x  F.
15. If f : M → N is an R-epimorphism of R-modules and N is a free R-module, then
prove that ker f is a direct summand of M.

13.6 VECTOR SPACES

Modules over a field are called vector spaces and these play a vital role in any
branch of mathematics. In particular, the homomorphisms between vector
spaces exhibit rich structural properties. In this, we discuss briefly about the
vector spaces and homomorphisms between these.
Definition 13.6.1. Let F be a field. Then, any F-module V is called a vector
space over F or a F-vector space or, simply, a vector space. The elements of
V are called vectors and elements of the field F are called scalars.
Definition 13.6.2. Let V and W be vector spaces over a field F. Then, an
F-homomorphism f : V → W is called a linear transformation of V into W.

Q001-Algebra-111001_CH 13.indd 42 9/21/2011 4:45:41 PM

 Modules and Vector Spaces   13-43

Definition 13.6.3. Let V be a vector space over a field F. Then, a F-submodule

W of V is called a subspace of V. For any vectors x1, x2, …, xn and for any sca-
lars a1, a2, …, an, the sum

a1x1 1 a2x2 1 … 1 anxn

is called a linear combination of x1, x2, …, xn. For any subset X, the subspace
generated by X is called linear span of X in V and is denoted, as usual, by <X>.
Recall that

<X> 5 {a1x1 1 a2x2 1 … 1 anxn : xi  X, ai  F}.

The following is a special property of vector spaces.

Theorem 13.6.1. Any nonzero vector space over any field F is free (i.e., a
free F-module).

Proof: Let F be a field and V be a nonzero vector space over F. Let

P 5 {X ⊆ V : X is linearly independent}.

First note that for any 0  a  F and x  V,

ax 5 0 ⇒ a21(ax) 5 0 ⇒(a21a)x 5 0⇒ x 5 0

and hence, for any 0  x  V and a  F,

ax 5 0 ⇒ a 5 0.

Therefore, {x} is linearly independent for any 0  x  V and hence P is a

nonempty class of subsets of V. Let {Xi}iI be a chain in P and X  i ∪ ∈I
Xi.
Then, X is linearly independent; for, if a1, a2, …, an  F and x1, x2, …, xn 
X such that a1x1 1 a2x2 1 … 1 anxn 5 0, then x j ∈ X i j for some ij  I and,
since {Xi}iI is a chain, there exists i  I such that, X i j ⊆ X i for all 1 # j #
n and hence x1, x2, …, xn  Xi and by the linear independence of Xi, it follows
that a1 5 a2 5 … 5 an 5 0. Therefore, X  P. That is, P is closed under
unions of chains. By the Zorn’s lemma, P has a maximal member, say B.
Now, we prove that B is a basis of V. Since B  P, B is linearly independent.
Let 0  x  V and x ∉ B. Then, by the maximality of B, B ∪ {x} ∉ P; that is,
B ∪ {x} is linearly dependent and hence there exist x1, x2, …, xn  B and a1a2,
…, an, a  F, with a  0, such that

a1x1 1 a2x2 1 … 1 anxn 1 ax 5 0.

Q001-Algebra-111001_CH 13.indd 43 9/21/2011 4:45:42 PM

 13-44   Algebra – Abstract and Modern

Since 0  a  F and F is a field, a21 exists in F and

x 5 (2a21a1)x1 1 (2a21a2)x2 1 … 1 (2a21an)xn

Therefore, x  <B> for all x  V, so that V 5 <B>. Thus, B generates V.

Thus, B is a basis of V and hence V is a free F-module.

Corollary 13.6.1. Let V be a vector space over a field F. Then, any linearly
independent subset of V can be extended to a basis of V.

Proof: Let S be a linearly independent subset of V and consider the class

P 5 {X ⊆ V : X is linearly independent and S ⊆ X}.

On the lines of the proof given above, we can prove that P has a maximal
member, say B. Then, B is a basis of V and S ⊆ B.

Theorem 13.6.2. Let V be a vector space over a field F. Then, any subspace
of V is a direct summand of V.

Proof: Let W be a subspace of V. Then, W is a vector space over F and, by

Theorem 13.6.1, W has a basis B. Now, since B is a linearly independent sub-
set of V and hence, by Corollary 13.6.1, there exists a basis C of V containing
B. Now, put W9 5 <C 2 B>. We prove that V 5 W ⊕ W9

n m
x  W ∩ W  ⇒ x  ∑ ai xi  ∑ b j y j ,
i1 j1

where ai, bj  F and x1, …, xn  B and y1, …, ym  C 2 B

      ⇒ a1x1 1 … 1 anxn 2 b1y1 2 b2y2 2 … 2bmym 5 0

⇒ a1 5 … 5 an 5 b1 5 … 5 bm 5 0
(since C is linearly independent)
⇒ x 5 0.

Therefore, W ∩ W9 5 {0}. Also, clearly

V 5 <C> 5 <B> 1 <C 2 B> 5 W 1 W9.

Thus, V 5 W ⊕ W9 and hence W is a direct summand of V.

Q001-Algebra-111001_CH 13.indd 44 9/21/2011 4:45:42 PM

 Modules and Vector Spaces   13-45

Theorem 13.6.3. Any vector space is completely reducible.

Proof: Let V be a vector space over a field and let B be a basis of V. Then, it
can be proved that
V  ⊕ Fx.
x∈ B

Also, for any 0  x  V, Fx > F as vector spaces over F and, since F is a

field, F is a single F-module. Therefore, each Fx is a simple F-module. Thus,
V is completely reducible.

Definition 13.6.4. Let V be a finitely generated vector space over a field F.

Then, the rank of V (that is, the number of elements in any basis of V) is called
the dimension of V over F and is denoted by dimFV. If V is not finitely gener-
ated, then V is said to be infinite dimensional. Recall that, if dimFV 5 n < ∞,
then V > Fn, as vector spaces over F.
Recall that, in Fn, the elements e1, e2, …, en form a basis, where ei is
the n-tuple in which the ith coordinate is 1 and the other coordinates are
0. This is called an ordered basis for Fn. Now, let us turn our attention to
linear transformations (F-homomorphisms) of finite dimensional vector
spaces.

Theorem 13.6.4. Let V and W be vector spaces of dimensions m and n,

respectively, over a field F. Then,

HomF(V, W) > Fm3n

as vector spaces over F, where HomF(V, W) is the vector space of all linear
transformations of V into W.

Proof: Fix bases B 5 {e1, e2, …, em} and C 5 {f1, f2, …, fn} for V and W,
respectively. If  : V → W is a linear transformation, then, for each 1# i # m,
(ei)  w 5 <C> and hence
n
a(ei )  ∑ aij f j .
j1

Then, the m 3 n matrix A 5 (aij) is called the matrix of  with respect to the
bases B and C. Conversely, if A 5 (aij) is a m 3 n matrix over F, then we can
define  : V → W by

n
a(ei )  ∑ aij f j for each 1# i # n
j1

Q001-Algebra-111001_CH 13.indd 45 9/21/2011 4:45:42 PM

 13-46   Algebra – Abstract and Modern

m m
     and a( x )  ∑ ai a(ei ) if x  ∑ ai ei .
i1 i1

Then,  is a linear transformation of V into W such that A 5 A. It is a straight

forward verification to prove that   A is an isomorphism of HomF(V, W)
onto Fm3n as vector spaces over F.

Corollary 13.6.2. If V and W are vector spaces of dimensions m and n, respec-

tively, over a field F, then HomF(V, W) is a vector space over F and

dimF(HomF(V, W)) 5 mn.

Corollary 13.6.3. For any n-dimensional vector space V over F,

HomF(V, V) > Mn(F)

where Mn(F) is the vector space of all n 3 n matrices over F.

In fact, both HomF(V, V) and Mn(F) are algebras over F in the sense of the
following definition.

Definition 13.6.5. Let A be a vector space over a field. If there is another

binary operation on A such that (A, 1, ?) is a ring in which the ring multipli-
cation and the scalar multiplication are compatible with each other, then A is
called an algebra over F.

Corollary 13.6.4. For any n-dimensional vector space over a field F,

HomF(V, V) and Mn(F) are isomorphic as algebras over F.
Now, we consider change of basis and their effect on the matrices of linear
transformation. First, we have the following theorem.

Theorem 13.6.5. Let V be a n-dimensional vector space over a field F and

{e1, e2, …, en} be a basis of V. Let A 5 (aij) be an n 3 n matrix over F and

ei′  ai1 e1  ai 2 e2  ain en for each 1  i  n.

Then, {e19, e29, …, e9n} is also a basis of V if and only if A is an invertible

matrix (unit) in the ring Mn(F) of all n 3 n matrices over F.

Proof: Let B 5 {e1, e2, …, en} and B9 5 {e19, e29, …, e9n}.

First, we suppose that B9 is a basis of V. For, each 1# i # n, we have

Q001-Algebra-111001_CH 13.indd 46 9/21/2011 4:45:43 PM

 Modules and Vector Spaces   13-47

n
ei  ∑ bij e j for some bij  F.
j1

Now, consider

n n  n 
ei  ∑ bij ej  ∑ bij ∑ a jk ek 
j1 j1

 k1 

n  n 
∴ ei  ∑ ∑ bij a jk ek .
       k1 
 j1 

Since {e1, e2, …, en} are linearly independent, we get that

k 1 if i  k
∑b a  ik   .
j1
ij jk
0 if i ≠ k

This shows that the matrix product

A9A 5 I where A9 5 (bij) and A 5 (aij).

Similarly, by considering

n n n n  n 
ei  ∑ aij e j ∑ aij ∑ b jk ek  ∑ ∑ aij b jk  ek ,
j1 j1 k1 k1 
 j1 

we get that AA9 5 I. Therefore, A is an invertible matrix and A9 is the

inverse of A.
Conversely suppose that A is an invertible matrix and A9 5 (bij) be the inverse
of A. We prove that B9 is a basis of V. Let

r1e91 1 r2e92 1 … 1 rne9n 5 0, where ri  F.

Then,

n  n 
∑ r ∑ a e   0
j1
j
i1
ji i


 n
n 
 r a e  0.
∑ ∑ j ji  i
i1  j1 

Q001-Algebra-111001_CH 13.indd 47 9/21/2011 4:45:44 PM

 13-48   Algebra – Abstract and Modern

Since e1, e2, …, en are linearly independent, we get that

n

∑r a
j1
j ji  0 for all 1 # i # n.

Let

 r1 
 
r 
P   2 
  
 
rn 

be the n 3 1 matrix. Then,

0
 
0
AP   .
  
 
0

Therefore,

0
 
0
A AP   
  
 
0

and hence

0
 
0
P   
  
 
0

since A9A 5 In.

Thus, r1 5 r2 5 … 5 rn 5 0. Therefore, e19, e29, …, en9 are linearly indepen-
dent and hence {e19, e29, …, en9} forms a basis of V.

Note: The matrix A 5 (aij) given in the above theorem is called the matrix of
transformation from the basis B9 to the basis B.

Q001-Algebra-111001_CH 13.indd 48 9/21/2011 4:45:44 PM

 Modules and Vector Spaces   13-49

The following describes the effect of a change of a basis on the matrix of a lin-
ear transformation. First, recall that if B 5 {e1, e2, …, em} and C 5 {f1, f2, …, fn}
are basis of vector spaces V and W, respectively, and  : V → W is a linear
transformation such that

n
a(ei )  ∑ aij f j for each 1  i  m,
j1

then the m 3 n matrix A 5 (aij) is called the matrix of  with respect to the
bases B and C.

Theorem 13.6.6. Let V and W be vector spaces of dimensions m and n,

respectively, over a field F. Let B 5 {e1, e2, …, em} and C 5 {f1, f2, …, fn}
bases of V and W, respectively, and A 5 (aij) be the matrix of a linear transfor-
mation  : V → W with respect to the bases B and C.

1. Let B9 5 {e19, e29, …, em9} and C9 5 {f19, f29, …, fn9} be new bases of V
and W, respectively. Then, the matrix of  with respect to the bases B9
and C9 is of the form PAQ21 where P and Q are matrices of transforma-
tions from B9 to B and C9 to C, respectively.
2. Conversely, if P and Q are m 3 m and n 3 n invertible matrices, respec-
tively, then there exist bases B9 and C9 of V and W, respectively, such that
PAQ21 is the matrix of  with respect to the bases B9 and C9.
n
Proof: (1) We have a(ei )   aij f j for each 1# i # m.
j1
Let A9 5 (aij)9 be the matrix of  with respect to the bases B9 and C9. Then,
we have

n
a(ei )  ∑ aij f j for each 1# i # m.
j1

Let P 5 (pij) and Q 5 (qij) be matrices of transformations from B9 to B and

C9 to C, respectively.
Also let Q21 5 (q9ij) be the matrix of transformation from C to C9 (by
Theorem 13.6.5). Now, we have

n
f j  ∑ qjk f k for 1 # j # n
k1

m
       and ei  ∑ pil el for 1# i # m.
l1

Q001-Algebra-111001_CH 13.indd 49 9/21/2011 4:45:45 PM

 13-50   Algebra – Abstract and Modern

Therefore,

 m 
a(ei ) a∑ pil el 
 l1 
m
 ∑ pil a(el )
l1

m  n 
 ∑ pil ∑ alj f j 
l1  j1 
m  n  n 
 ∑ pil ∑ alj ∑ q ′jk f k′
l1  j1  k1 
n  m  n 
 ∑ ∑ pil ∑ alj q ′jk  f k′.
 l1  j1
k1 


Thus, the matrix  with respect to the bases B9 and C9 is equal to PAQ21.
This proves (1). The proof of (2) is similar to the above and to the proof
of Theorem 13.6.5.

Worked Exercise 13.6.1. Let F be any field and consider the vector spaces
F3 and F2 over F. Define  : F3 → F2 by

(a, b, c) 5 (a 1 b 1 c, b 1 c)

Then, prove that  is a linear transformation and determine the matrix of 

with respect to the standard basis of F3 and F2.

Answer: For any x 5 (a, b, c) and y 5 (a9, b9, c9)  F3 and r and s  F, we
have

(rx 1 sy) 5 (ra 1 sa9, rb 1 sb9, rc 1 rc9)

5 (ra 1 sa9 1 rb 1 sb9 1 rc 1 sc9, rb 1 sb9 1 rc 1 sc9)
5 r(a 1 b 1 c, b 1 c) 1 s(a9 1 b9 1 c9, b9 1 c9)
5 ra(x) 1 sa(y).

Therefore,  is a linear transformation of F3 into F2. Let {e1, e2, e3} and
{f1, f2} be standard bases of F3 and F2, respectively. Then

e1 5 (1, 0, 0), e2 5 (0, 1, 0) and e3 5 (0, 0, 1)

    and f1 5 (1, 0), f2 5 (0, 1).

Q001-Algebra-111001_CH 13.indd 50 9/21/2011 4:45:45 PM

 Modules and Vector Spaces   13-51

Now,     (e1) 5 (1, 0, 0) 5 (1, 0) 5 1 ? f1 1 0 ? f2

(e2) 5 (0, 1, 0) 5 (1, 1) 5 1 ? f1 1 1 ? f2
(e3) 5 (0, 0, 1) 5 (1, 1) 5 1 ? f1 1 1 ? f2.

Therefore,

1 0
 
1 1
 
1 1

is the matrix of a with respect to the bases {e1, e2, e3} and {f1, f2}.

EXERCISE 13(f)

1. Let V and W be vector spaces over a field F and a : V → W be a mapping. Prove

that a is a linear transformation if and only if a(ax 1 by) 5 aa(x) 1 ba(y) for
all x, y  V and a, b  F.

2. Let V 5 F 5 and F be a field. Let

e1 5 (2, 0, 0, 0, 0), e2 5 (2, 1, 0, 0, 0) and e3 5 (1, 2, 3, 0, 0).
Prove that {e1, e2, e3} is linearly independent in V and extend this to a basis
of V.

3. Define a : F3 → F2 by a(a, b, c) 5 (a 1 b 1 c, b 1 c). Determine the matrix of a

with respect to the bases {(21, 0, 2), (0, 1, 1), (3, 21, 0)} and {(21, 1), (1, 0)} of
F3 and F2, respectively.

4. Let F be a field and V 5 {f(x)  F[x] : deg(f(x)) # 4}. Prove that B 5 {1, x, x2,
x3, x4} and C 5 {1, 1 1 x, 1 1 x 1 x2, 1 1 x 1 x2 1 x3, 1 1 x 1 x2 1 x3 1 x4}
are bases of V. If D is the differentiation operator on V, then determine the matrix
of D with respect to each of the bases B and C. Also determine the matrix of
transformation from B to C and from C to B.

5. For any subset X of a vector space V over a field F, let <X> be the linear span of
X in V. Prove the following for any subsets X and Y in V:
(1) X ⊆ <X>
(2) X 5 <X>
(3) <X ∪ Y> 5 <X> 1 <Y>
(4) <X> 5 ∪ {<Y> : Y is finite subset of X}.

Q001-Algebra-111001_CH 13.indd 51 9/21/2011 4:45:45 PM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 13.indd 52 9/21/2011 4:45:45 PM

 Part IV
Field Theory

Q001-Algebra-111001_CH 14.indd 1 9/19/2011 4:51:08 PM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 14.indd 2 9/19/2011 4:51:08 PM

 14
Extension Fields
14.1 Extensions of a Field
14.2 Algebraic Extensions
14.3 Algebraically Closed Fields
14.4 Derivatives and Multiple Roots
14.5 Finite Fields

It is well known that the field Q of rational numbers is a subfield of the field
R of real numbers and, in this case, we say that R is an extension field of Q.
Likewise, the field C of complex numbers is an extension of R. Let us recall
that the polynomial 1 1 x2 has no root in R. However, there is an extension
field, namely C, containing a root of 1 1 x2. In this chapter, we discuss in
detail about the existence of an extension field containing roots of a given
polynomial over a given field.
The field R of real numbers has a deficit that not all polynomials over R
have roots in R. The field C of complex number is an extension of R con-
taining all the roots of any polynomial over C. Such fields like C are called
algebraically closed fields. We discuss these and similar concepts in the pres-
ent chapter.
Here afterwards F denotes an arbitrary field, unless otherwise stated. Also,
a homomorphism of one field F into another field K is always assumed to be a
ring homomorphism of F into K carrying the unity in F onto the unity in K.

14.1 EXTENSIONS OF A FIELD

Any field K can be considered as a vector space over any of its subfield and
we can discuss their dimensions and related concepts. If a field K has a sub-
field which is an isomorphic copy of another field F, then we can treat F itself
as a subfield of K. Formally, we have the following definition.

Q001-Algebra-111001_CH 14.indd 3 9/19/2011 4:51:08 PM

 14-4   Algebra – Abstract and Modern

Definition 14.1.1. Let F and K be fields. Then, K is said to be a field exten-

sion of F or, F is said to be a subfield of K, if there exists a monomorphism
(or embedding) s : F → K. Note that this s is an injective homomorphism of
rings such that s(1) 5 1.
If s : F → K is a monomorphism, then F is isomorphic to s(F) which is a
subfield of K and hence we can consider a field extension K of F to be a field
containing F as a subfield. We write F ⊆ K when K is an extension of F. Also,
in this case, for any a  F and x  K, we have ax  K (note that a is identified
with s(a) in K). Therefore, we have mapping F 3 K → K given by (a, x) 
ax. With this as the scalar multiplication, we get that the additive group (K, 1)
becomes a vector space over F.

Definition 14.1.2. If K is a field extension of F, then the dimension of the

vector space K over F is called the degree of K over F and is denoted by [K :
F]. K is said to be finite or infinite extension of F according as the degree of
K over F is finite or infinite.

Example 14.1.1
1. For any field F, [F : F] 5 1.
2. The degree of the field C of complex numbers over the field R of real
numbers is 2, since {1, i} is a basis of C over R.
3. Let F be any field and F[x] be the ring of polynomials over F. Let K
be the field of quotient of F[x]. Then, K is a field extension of F. Also,
consider the set {1, x, x2, …}. If a01 1 a1x 1 a2x2 11 anxn 5 0, then
a0 5 a1 5  5 an 5 0 and hence {1, x, x2, …} is an infinite linearly
independent subset of K. Therefore, K is an infinite extension of F.
4. Let Q[ 2] 5 {a  b 2 : a and b  Q}. Then, Q[ 2] is a field extension
of Q and is of degree 2 over Q, since {1, 2} is a basis of Q[ 2] over Q.
The following theorem about the degrees of finite extensions of fields is
very important and useful.

Theorem 14.1.1. Let F be any field, K be a field extension of F and L be a

field extension of K (i.e., F ⊆ K ⊆ L). Then, [L : F] is finite if and only if both
[L : K] and [K : F] are finite and, in this case,

[L : F] 5 [L : K] [K : F].

Proof: Suppose that [L : F] is finite. Since K is a subspace of L over F, we

get that [K : F]  [L : F] and hence [K : F] is finite. Also, if B is a basis of L

Q001-Algebra-111001_CH 14.indd 4 9/19/2011 4:51:09 PM

 Extension Fields   14-5

over K, then B is a linearly independent subset of L over K and hence linearly

independent over F also (since F ⊆ K) and therefore | B |  [L : F]. Therefore,
[L : K]  [L : F] and hence [L : K] is finite.
Conversely suppose that [L : K] 5 m and [K : F] 5 n, where m and n are posi-
tive integers. Let {x1, …, xm} be a basis of L over K and {y1, …, yn} be a basis
of K over F. Now, let

B 5 {xi yj; 1  i  m and 1  j  n}.

Then, B is a subset of L and | B | 5 mn (since xi’s are independent over K and

yj  K, xi yj 5 xr ys ⇒ i 5 r and j 5 s). We prove that B is a basis of L over F.
To prove the linear independency of B over F, let aij  F for 1  i  m and
1  j  n such that

∑ a x y  0.
1im
ij i j

1 jn

m
For each 1  i  m, let bi  ∑ aij y j  K .
j1

Then, bi  K, since aij  F ⊆ K and yj  K. Now,

m m  n 
∑ b x  ∑ ∑ a y j  xi  0.

i i ij
i1 i1 j1

Since x1, x2, …, xm are linearly independent over K and bi  K, it follows that
bi 5 0 for each 1  i  m. Again, for each 1  i  m,

∑a j1
ij y j  bi  0

and, since {y1, y2, …, yn} is linearly independent over F and aij  F, it follows
that

aij 5 0 for each 1  i  m and 1  j  n.

Thus, B is linearly independent over F.

Next, we prove that B generates L over F. Let x  L. Since xi’s generate L
over K, we get that

x 5 k1x1 1 k2x2 11 kmxm for some k1, …, km  K.

Q001-Algebra-111001_CH 14.indd 5 9/19/2011 4:51:10 PM

 14-6   Algebra – Abstract and Modern

Again, since yj’s generate K over F, we get that

n
K i = ∑ aij y j , for each 1 ≤ i ≤ m
j =1

where aij  F. Therefore, we have

m m  n 
x  ∑ ki xi  ∑ ∑ aij y j  xi  ∑ aij xi y j .
i1 i1 
 j1  1im
1 jn

Therefore, B generates L over F and hence [L : F]is finite. Also,

[L : F] 5 | B | 5 mn 5 [L : K] [K : F].

Corollary 14.1.1. If F ⊆ K ⊆ L are fields and [L : F] is finite, then [L : K] and

[K : F] are divisors of [L : F].

Corollary 14.1.2. Let F ⊆ K ⊆ L be fields such that [L : F] is a prime number.

Then, K 5 F or K 5 L.

Proof: Let [L : F] 5 p. Then, [K : F] is a divisor of the prime p and hence

[K : F] 5 1 or p. If [K : F] 5 1, then K 5 F. If [K : F] 5 p, then [L : K] 5 1
(since [L : K] [K : F] 5 [L : F] 5 p) and hence K 5 L.

Corollary 14.1.3. Let F1 ⊆ F2 ⊆ … ⊆ Fn be fields such that [Fi : Fi21] is finite

for all 1  i  n. Then, [Fn : F1] is finite.

Proof: Apply induction on n.

Worked Exercise 14.1.1. Let F1 ⊆ F2 ⊆ … ⊆ Fn be fields. Then, prove that

[Fn : F1] is finite if and only if [Fi : Fj] is finite for all i . j and, in this case,

n
[ Fn : F1 ]  ∏[ Fi : Fi1 ].
i2

Answer: Suppose that [Fn : F1] is finite. Then, for each 1 , i  n, [Fi : F1]
is finite and [Fi : Fj] is finite for all j , i. Conversely, suppose that [Fi : Fj] is
finite for all i . j. Then, [Fi : Fi21] is finite for all 1 , i  n and, in particular,
[Fn : F1] is finite. Also, in this case,

Q001-Algebra-111001_CH 14.indd 6 9/19/2011 4:51:10 PM

 Extension Fields   14-7

[Fn : F1] 5 [Fn : Fn21] [Fn21 : F1]

5 [Fn : Fn21] [Fn21 : Fn22] [Fn22 : F1]
n
5  [ Fi : Fi1 ] (by induction on n)
i2

Worked Exercise 14.1.2. Let p be a prime number and

Q( p )  {a  b p : a and b ∈ Q}.

Then, prove that Q( p ) is a field extension of Q and is of degree 2 over Q.

Proof: First, let us recall that p is a real number which is not rational. For
any a, b, c and d  Q, we have

( a  b p )  (c  d p )  ( a  c)  (b  d ) p

( a  b p )(c  d p )  ( ac  bdp)  ( ad  bc) p

1 ab p a  b 
and   2  2  p ∈ Q( p ).
ab p
2
a  pb 2
a  pb  a  pb 2 
2

Therefore, Q( p ) is a subfield of R, clearly Q ⊆ Q( p ) . b

Therefore, Q( p ) is a field extension of Q. Also, since {1, p} is a basis
of Q( p ) over Q, it follows that [ Q( p ) : Q] 5 2. That is, Q( p ) is of
degree 2 over Q.

Worked Exercise 14.1.3. Let K be the smallest subfield of R containing

Q ∪ { 2, 3}. Then, prove that K is a finite extension of Q and determine
the degree of K over Q.

Answer: Let F  Q( 2)  {a  b 2 : a and b ∈ Q} by Worked Exercise

14.1.2, F is an extension of Q and [F : Q] 5 2.
Now, observe that 3  F; for, if 3 5 a 1 b 2 with a and b  Q, then a  0
(otherwise 3 5 2b2) and b  0 and hence ab  0 and therefore 3 5 (a 1
b 2)2 5 a2 1 2b2 1 2ab 2 which is not true. Now, consider

F( 3) 5 {a 1 b 3 : a and b  F}.

Then, it can be easily proved that F( 3) is an extension of F and {1, 3} is

a basis of F( 3) over F and hence [F ( 3) : F ]  2. Now, we have Q ⊆ F ⊆
F( 3) and hence

[F( 3) : Q] 5 [F( 3) : F] [F : Q] 5 2 ? 2 5 4.

Q001-Algebra-111001_CH 14.indd 7 9/19/2011 4:51:13 PM

 14-8   Algebra – Abstract and Modern

It can be verified that F( 3) is the smallest subfield of R containing Q ∪ { 2,

3}, that is, K 5 F( 3). Thus, K is a finite extension of Q and is of degree
4 over Q.

EXERCISE 14(a)

1. Let K be the smallest subfield of R containing Q ∪ { 2, 3}. Then, find a basis

of K over Q.

2. Consider two distinct prime numbers p and q and let K be the field given in
Exercise 1 above with p and q in place of 2 and 3, respectively. Then, prove that
K is a finite extension of Q and find the degree and basis of K over Q.

3. Let p1, p2, p3, … be the sequence of all prime numbers. Define Fn recursively as
follows:

F0 5 Q and Fn 5 Fn21( pn ) 5 {a 1 b pn : a and b  Fn21}.

Then, prove that, for each n  Z1, Fn is a finite extension of Fn21 and is of degree
2 over Fn21.

4. Deduce from Exercise 3 above that R is an infinite extension of Q.

5. Determine each of the following.

(i) [C : R]
(ii) [C : Q]
(iii) [R : Q( 2)]
(iv) [Q( 2, 3, 5) : Q], where Q( 2, 3, 5 ) is the smallest subfield
of R containing Q ∪ { 2, 3, 5 }.
(v) [Q( p1 , p2 , …, pn ) : Q] for any distinct primes p , p , …, p .
1 2 n

6. Is Q a field extension of Zp, for any prime p?

7. Construct a field extension of Z3 with exactly 9 elements.

8. If F ⊆ E are fields, prove that char(F) 5 char(E).

14.2 ALGEBRAIC EXTENSIONS

Let us recall that a polynomial f (x) over a field F is called irreducible over F if
f (x) cannot be expressed as a product of two nonconstant polynomials over F.
If E is a field extension of a field F and f (x)  F[x] such that f (a) 5 0 for some
a  E, then a is a called a root of f (x) in E. Note that any polynomial of degree
n over a field F can have at most n roots in any field extension E of F.

Q001-Algebra-111001_CH 14.indd 8 9/19/2011 4:51:15 PM

 Extension Fields   14-9

Theorem 14.2.1. Let F be any field and p(x)  F[x] be irreducible over F.
Then, there exists a field extension E of F such that E contains a root of p(x).

Proof: Since p(x) is given to be irreducible, the principal ideal ,p(x). gen-
erated by p(x) in the ring F[x] is a maximal ideal and hence the quotient ring
F[x]/,p(x). is a field. Define s : F → F[x]/,p(x). by

s(a) 5 a 1 ,p(x). for any a  F.

Then, clearly s is an embedding of F in F[x]/,p(x).. Put E 5 F[x]/,p(x)..

Then, E is a field extension of F. If p(x) 5 a0 1 a1x 1 a2x2 1…1 anxn, n .
0, and ai  F, then x 1 ,p(x). is a root of p(x) in E, since

n
p( x  p( x ) )  ∑ ai ( xp( x ) )i
i0
n
 ∑ ai ( x i p( x ) )
i0

 n 
 ∑ ai ( x i )p( x )
 i=0 
 p( x )p( x )  0 in E .

Thus, E is field extension of F containing a root of p(x), namely x 1

,p(x)..

Corollary 14.2.1. Let F be a field and f (x) be a nonconstant polynomial over

F. Then, F has a field extension E containing a root of f (x).

Proof: Since f (x) is a nonconstant polynomial over F, f (x) is a nonunit

in F[x]. Since F[x] is a unique factorization domain, it follows that there
exists an irreducible polynomial p(x) in F[x] such that p(x) divides f (x). By
Theorem 14.2.1, there exists a field extension E of F such that E contains
a root of p(x).

Let f (x) 5 p(x) g(x). If a  E is a root of p(x), then

f (a) 5 p(a) g(a) 5 0 g(a) 5 0

and therefore a is a root of f (x) in E.

Theorem 14.2.2. Let f (x) be a nonconstant polynomial over a filed F. Then,

F has a field extension E containing all the roots of f (x).

Q001-Algebra-111001_CH 14.indd 9 9/19/2011 4:51:15 PM

 14-10   Algebra – Abstract and Modern

Proof: We prove this by using induction on the degree of f (x). Let deg(f (x)) 5
n. If n 5 1, then f (x) 5 a 1 bx, a and 0  b  F and 2ab21 is the only root of
f (x) and hence F itself is the extension F containing all the roots of f (x). Let n > 1
and assume the theorem for all nonconstant polynomials of degree less then n.
By the above Corollary 14.2.1, there exists a field extension K of F such that K
contains a root a of f (x) in K. Now, x 2 a divides f (x) in K[x] and hence

f (x) 5 (x 2 a) g(x) for some g(x)  K[x].

Now, deg(g(x)) 5 n 2 1 > 0 and, by the induction hypothesis, there exists a

field extension E of K such that E contains all the roots of g(x). Any root of
f (x) other than a must be a root of g(x). Any root of f (x) other than a must be
a root of g(x). Therefore, E contains all the roots of f (x) and E is a field exten-
sion of F, since F ⊆ K ⊆ E.

Corollary 14.2.2. Let f1(x), f2(x), …, fm(x) be nonconstant polynomials over a

field F, then F has a field extension E containing all the roots of fi(x) for all
1  i  m.

Proof: Consider f (x) 5 f1(x), f2(x), …, fm(x) and use Theorem 14.2.2 above.

Definition 14.2.1. Let F ⊆ K be fields and a  K. Then, a is said to be

algebraic over F if a is a root of a nonzero polynomial in F[x]; that is, if
f (a) 5 0 for some 0  f (x)  F[x]. If a is not algebraic over F, then a is
called transcendental over F.

Example 14.2.1
1. For any field F, every element a of F is algebraic over F, since x 2 a 
F[x] and a is a root of x 2 a.
2. 2 is algebraic over Q, since x2 2 2  Q[x] and 2 is a root of x2 2 2.
3. The complex number i is algebraic over Q since i is a root of 1 1 x2
 Q[x].
4. It is known that the real numbers e and  are transcendental over Q. The
proofs of these are beyond the scope of this book.
Let us recall that a nonzero polynomial over a field is called monic if its
leading coefficient is the unity element of the field.

Theorem 14.2.3. Let F ⊆ K be field and a  K be algebraic over F. Then, there

exists a unique monic irreducible polynomial p(x) over F such that p(a) 5 0.
Also, for any f (x)  F[x],

f (a) 5 0 if and only if p(x) divides f (x) in F[x].

Q001-Algebra-111001_CH 14.indd 10 9/19/2011 4:51:15 PM

 Extension Fields   14-11

Proof: Consider the set

I 5 {f (x)  F[x] : f (a) 5 0}.

It can be easily verified that I is an ideal of the ring F[x]. Since F[x] is a prin-
cipal ideal domain, there exists p(x)  I such that

I 5 ,p(x)> 5 {p(x) g(x) : g(x)  F[x]}.

Without loss of generality, we can assume that p(x) is monic; for, if p(x) 5
a0 1 a1x 1…1 anxn, an  0 and g ( x )  an 1a0  an 1a1 x … an 1an1 x n1  x n ,
then p(x) 5 ang(x) and hence p(x) and g(x) are associates in F[x] so that
,p(x)> 5 ,g(x)>. Thus, we can assume that p(x) is a monic polynomial in
F[x] such that

,p(x)> 5 {f (x)  F[x] : f (a) 5 0} 5 I.

Since a is algebraic over F, there exists a nonconstant polynomial f (x) in F[x]

such that f (x) 5 0 and, in this case, f (x) 5 p(x)g(x) for some g(x)  F[x].
Therefore, p(x) is also a nonconstant polynomial. Further,

deg(p(x))  deg(f (x)) for all 0  f (x)  I  (*)

From this, it follows that p(x) is irreducible; for, let p(x) 5 g(x)h(x) for some
nonconstant g(x) and h(x)  F[x]. Then, 0 5 p(a) 5 g(a)h(a) and hence g(a)
5 0 or h(a) 5 0, so that g(x) or h(x)  I. But, since deg(p(x)) 5 deg(g(x))
1 deg(h(x)) and both deg(g(x)) and deg(h(x)) are positive, it follows that
deg(g(x)) , deg(p(x)) and deg(h(x)) , deg(p(x)) which is a contradiction to
(*). Thus, p(x) is an irreducible monic polynomial in F[x] such that p(a) 5 0.
To prove the uniqueness of p(x), let

p(x) 5 a0 1 a1x 1…1 an21xn21 1 xn

      and q(x) 5 b0 1 b1x 1…1 bm21xm21 1 xm

be two irreducible monic polynomials in F[x] such that p(a) 5 0 5 q(a).

Then, q(x)  I and hence q(x) 5 p(x)g(x) for some g(x)  F[x]. The irreduc-
ibility of q(x) implies that g(x) is a unit (that is, g(x)  F) and hence m 5
deg(q(x)) 5 deg(p(x)) 5 n. Now, p(a) 2 q(a) 5 0 and hence p(x) 2 q(x)  I.
This implies, by (*), that p(x) 2 q(x) 5 0. Thus, p(x) 5 q(x).
The last assertion in the theorem follows from the fact that

,p(x)> 5 {f (x)  F[x] : f (a) 5 0}. b

Q001-Algebra-111001_CH 14.indd 11 9/19/2011 4:51:15 PM

 14-12   Algebra – Abstract and Modern

Definition 14.2.2. If F ⊆ K are fields, a  K and a is algebraic over F, then

the unique irreducible monic polynomial in F[x], for which a is a root, is
called the minimal polynomial of a over F.

Example 14.2.2
1. We have Q ⊆ R and 2  R. x2 2 2 is the minimal polynomial of 2
over Q.
2. Let v be a root of 1 1 x 1 x2 1…1 xp21 in C, where p is a given prime
number. Then, using the Eisenstein criterion, we have proved that 1 1 x
1 x2 1…1 xp21 is irreducible over Q and hence it is the minimal poly-
nomial of v over Q.
Let F ⊆ K be fields and a  K. Then, the smallest subfield of K contain-
ing F and a will be demoted by F(a). Note that F(a) is the intersection of all
subfields of K containing F ∪ {a} and it can be easily verified that

F(a) 5 {f (a) g(a)21 : f (x) and g(x)  F[x] and g(a)  0}.

The following provides an elegant description of the elements of F(a),

when a is algebraic over F.

Theorem 14.2.4. Let F ⊆ K be fields and a  K. Then, a is algebraic over

F if and only if F(a) is a finite extension of F and, in this case, [F(a) : F] 5
deg(p(x)), where p(x) is the minimal polynomial of a over F. Also, if

p(x) 5 a0 1 a1x 1…1 an21xn21 1 xn,

     then F(a) 5 {b0 1 b1a 1…1 bn21an21 : bi  F}.

Proof: The theorem is trivial if a  F, since, 2a 1 x is the minimal polyno-

mial of a over F. Therefore, we can assume that a  F.
First suppose that F(a) is a finite extension of F and [F(a) : F] 5 m. Then, 1,
a, a2, a3, …, am are all elements of F(a) and these are m 1 1 in number. Since
dimF(F(a)) 5 m, 1, a, a2, …, am must be linearly dependent. Therefore, there
exists a0, a1, a2, …, am in F, not all zero, such that

a01 1 a1a 1a2a2 1…1 aman 5 0.

Therefore, a is a root of the nonzero polynomial a0 1 a1x 1…1amxm in F[x].

Thus, a is algebraic over F. Conversely, suppose that a is algebraic over F
and let

p(x) 5 a0 1 a1x 1…1 an21xn21 1 xn

Q001-Algebra-111001_CH 14.indd 12 9/19/2011 4:51:16 PM

 Extension Fields   14-13

be the minimal polynomial of a over F. Then,

a0 1 a1a 1…1 an21an21 1 an 5 p(a) 5 0

and hence each of an, an+1, an+2, … can be expressed in the form b0 1 b1a 1…1
bn21an21 with bi  F and hence so is f (a) for any f (x)  F[x]. Now, put

E 5 {b0 1b1a 1…1 bn21an21 : bi  F}.

Then, clearly E is a subring of K containing F and a. Define s : F[x] → E by

s(f (x)) 5 f (a). Then, s is an epimorphism and

f (x)  ker s ⇔ f (a) 5 0

⇔ p(x) divides f (x) (by Theorem 14.2.3)
⇔ f (x)  ,p(x)>

and therefore, ker s 5 ,p(x)>. By the Fundamental Theorem of Homomor-

phisms,

F [ x]  E.
p(x )

Now, since p(x) is irreducible in F[x], ,p(x)> is a maximal ideal of F[x]

and hence F [ x ]/  p(x )  is a field. Therefore, E is also a field. Thus, E is a
subfield of K containing F and a. It is clear that any subfield of K containing
F and a must contain E. Thus, E 5 F(a). Also, {1, a, a2, …, an21} is linearly
independent over F (since a is not a root of any polynomial of degree less
than n) and it generates F(a). Therefore, {1, a, …, an21} is a basis of F(a)
over F and hence

[F(a) : F] 5 n 5 deg(p(x)). b

Definition 14.2.3. Let K be a field extension of F. K is said to be an algebraic

extension of F if every element of K is algebraic over F.

Theorem 14.2.5. Let F be any field. Then, any finite extension of F is an

algebraic extension of F.

Proof: Let E be a finite extension of F and [E : F] 5 n. Then, for any a  E,

we have F ⊆ F(a) ⊆ E and hence [F(a) : F] [E : F(a)] 5 [E : F] 5 n, so that
[F(a) : F] is finite. By Theorem 14.2.4, a is algebraic over F for each a  E.
Thus, E is an algebraic extension of F. b

Q001-Algebra-111001_CH 14.indd 13 9/19/2011 4:51:16 PM

 14-14   Algebra – Abstract and Modern

The converse of the above theorem is not true. That is, an algebraic
extension of a field F need not be a finite extension of F. For, consider the
following example.

Example 14.2.3. It is well known that the set of prime numbers is a countably
infinite set and hence we can express this set in a sequential form. Let p1, p2, …,
pn, … be all the distinct primes. For each n  0, En be the subfield of R defined
recursively by

E0 5 Q and E0  En1 ( pn ) for all n > 0.

Clearly En  Q( p1 , p2 , …, pn ) which is the smallest subfield of R con-

taining Q ∪{ p1 , p2 , …, pn }. We first prove that pn1  En for all
n  0, by using induction on n. Since p1 is irrational, p1  Q  E0 . Let
n > 0 assume that pn  En1 . If pn1  En , then

pn1  a  b pn for some a and b  En21

and hence pn1  a 2  pn b 2  2ab pn which implies that pn  1 ( pn1  a 2  pn b 2

2ab
pn  1 ( pn1  a 2  pn b 2 )  En1 (note that a and b are nonzero elements of En21),
2ab
which is a contradiction. Thus, pn1 ∉ En for all n and hence

[En+1 : En] 5 2 for all n.

Now, Q 5 E0 ⊂ E1 ⊂ E2 ⊂ … ⊂ En ⊂ … is a strictly ascending chain of

subfields of R.
∞
Let E  ∪ En . Then, E  Q( p1 , p2 , …) which is the smallest subfield
n0
of R containing Q and the square roots of all the prime numbers. Now, by
Theorem 14.1.1, we have

[En : Q] 5 2n for all n  0

and, since each En is a subfield of E, it follows that [E : Q] is infinite; that is,

E is an infinite extension of Q. However, we observe that E is an algebraic
extension of Q. For, if a  E, then a  En for some n and hence Q(a) ⊆ En
so that [Q(a) : Q] is a divisor of [En : Q] 5 2n and therefore Q(a) is a finite
extension of Q. By Theorem 14.2.4, a is algebraic over Q. Thus, E is an
algebraic extension of Q.
We have proved in Theorem 14.2.2 that, for any nonconstant polynomial
f (x) over a field F, there exists a field extension E of F containing all the roots

Q001-Algebra-111001_CH 14.indd 14 9/19/2011 4:51:18 PM

 Extension Fields   14-15

of f (x). In the following, we prove that this extension E can be chosen to be a

finite extension containing all the roots of a given f (x).

Theorem 14.2.6 (Kronecker’s Theorem). Let f (x) be a nonconstant polyno-

mial over a field F and deg(f (x)) 5 n. Then there is an extension E of F such
that [E : F]  n! and E contains all the roots of f (x); that is, f (x) can be fac-
tored over E as

f (x) 5 a0(x 2 b1) (x 2 b2)  (x 2 bn), bi  E and a0  F.

Proof: We have f (x)  F[x] and deg(f (x)) 5 n  1. Let p(x) be an irreducible
polynomial R over F such that p(x) divides f (x). By Theorem 14.2.1, there
exists a field extension K of F such that K contains a root of p(x). Let a  K
be a root of p(x) and hence of f (x). By Theorem 14.2.4,

[F(a) : F] 5 deg(p(x))  deg(f (x)) 5 n.

Let F(a) 5 E1. We can write

f (x) 5 (x 2 a) f1(x) for some f1(x)  E1[x].

Now deg(f1(x)) 5 n 2 1. Continuing this process, we get an extension E of E1,

such that [E : E1]  (n 2 1)! and E1 contains all the roots of f1(x). Therefore,
E is an extension F such that

[E : F] 5 [E : E1] [E1 : F]  (n 2 1)! n 5 n!

and E contains all the roots of f (x). b

Definition 14.2.4. Let F ⊆ E be fields. Then, E is said to be finitely generated

over F if there exists a1, a2, …, an  E such that E 5 F (a1, a2, …, an), where

F(a1, a2, …, ai) 5 F(a1, a2, …, ai21)(ai)

for each 1 , i  n.
Note that any finite extension of F is finitely generated over F; but a finitely
generated extension may not be a finite extension. For, if a is a transcendental
number, then Q(a) is finitely generated over Q and it is not a finite extension
of Q. However, we have the following theorem.

Theorem 14.2.7. Let E 5 F (a1, a2, …, an) be a finitely generated extension

of F and a1, a2, …, an be algebraic over F. Then, E is a finite extension of F
and hence an algebraic extension of F.

Q001-Algebra-111001_CH 14.indd 15 9/19/2011 4:51:18 PM

 14-16   Algebra – Abstract and Modern

Proof: We prove this by using induction of n. If n 5 1, then the theorem

­follows from Theorem 14.2.4. Assume the theorem for n 2 1. That is,

[F(a1, a2, …, an21) : F] is finite.

Since an is algebraic over F, it is algebraic over F(a1, a2, …, an21) also and
hence

[F(a1, a2, …, an21)(an) : F(a1, a2, …, an21)] is finite.

Therefore, [F(a1, …, an) : F] is finite, by Theorem 14.1.1. Thus, E is a

finite extension of F and hence, by Theorem 14.2.5, E is an algebraic
extension of F. b

Theorem 14.2.8. Let E be an extension of F and

K 5 {a  E : a is algebraic over F}.

Then, K is a subfield of E and K is an algebraic extension of F.

Proof: Let a and b  K. Then, by Theorem 14.2.7, F(a, b) is an algebraic

extension of F and, since a 6 b, ab and a21 (if a  0)  F(a, b), it follows
that these a 6 b, ab and a21 (if a  0) are all algebraic over F and hence
these are elements of K. Thus, K is a subfield of F and is clearly an algebraic
extension of F. b

Theorem 14.2.9. Let F ⊆ K ⊆ L be fields. Let a  L be algebraic over K and

K an algebraic extension of F. Then, a is algebraic over F.

Proof: Since a  L is algebraic over K, there exists a nonzero polynomial

f (x) 5 a0 1 a1x 1…1 anxn  K[x] for which a is a root. Now,

f (x)  F(a0, a1, …, an)[x]

where each ai is algebraic over F (since ai  K). Put

E 5 F(a0, a1, …, an).

By Theorem 14.2.7, E is a finite extension of F. Also, since f (x)  E[x] and

f (a) 5 0, a is algebraic over E and hence [E(a) : E] is finite. Now,

[E(a) : F] 5 [E(a) : E] [E : F] , ∞.

Q001-Algebra-111001_CH 14.indd 16 9/19/2011 4:51:18 PM

 Extension Fields   14-17

Since F(a) ⊆ E(a), [F(a) : F]  [E(a) : F] , ∞. Thus, F(a) is a finite exten-

sion of F and hence a is algebraic over F. b

Corollary 14.2.3. Let F ⊆ K ⊆ L be fields, L be algebraic over K and K be

algebraic over F. Then, L is algebraic over F.

Definition 14.2.5. Let F ⊆ E be fields. A homomorphism s : E → E is said to

be an F-homomorphism if s(a) 5 a for all a  F. We denote this by s/F 5 Id.

Theorem 14.2.10. Let K 5 F(a1, a2, …, an). Then, any F-homomorphism

s : K → K is completely determined by s(a1), s(a2), …, s(an).

Proof: Let s and t be F-homomorphism of K into K such that s(ai) 5 t(ai)

for all 1  i  n. Put

L 5 {a  K : s(a) 5 t(a)}.

Then L is a subfield of K containing F and a1, a2, …, an. Therefore, F(a1, a2,
…, an) ⊆ L ⊆ K 5 F(a1, a2, …, an) and hence L 5 K. Thus, s(a) 5 t(a) for
all a  K; that is, s 5 t. b

Theorem 14.2.11. Let F ⊆ K be fields, p(x) a monic irreducible polynomial

over F and a and b  K be roots of p(x). Then, there exists an isomorphism s :
F(a) → F(b) such that s(a) 5 b and s(s) 5 s for all s  F.

Proof: As in the proof of Theorem 14.2.4, there exists an isomorphism

s1: F [ x ] → F ( a)
p( x )

defined by s1(f (x) 1 ,p(x)>) 5 f (a). Similarly, there exists an isomorphism

s 2 : F [ x] → F ( b)
p( x ) 

defined by s2(f (x) 1 ,p(x)>) 5 f (b). Now, put

s  s 2 ο s1 1 : F ( a) → F (b).

Then, s is an isomorphism, s(a) 5 s2(x 1 ,p(x)>) 5 b and, for any s  F,

s(s) 5 s2(s 1 ,p(x)>) 5 s. b

Q001-Algebra-111001_CH 14.indd 17 9/19/2011 4:51:18 PM

 14-18   Algebra – Abstract and Modern

Theorem 14.2.12. Let E be an algebraic extension of F and s : E → E be an

embedding such that s | F 5 Id. Then, s is an automorphism of E.

Proof: Let 0  a  E. Then, a is algebraic over F. Let

p(x) 5 a0 1 a1x 1…1 an21xn21 1 xn  F[x]

be the minimal polynomial of p(x) over F. Suppose that a 5 b1, b2, …, bm  E

be all the roots of p(x) in E. Then, by Theorem 14.2.7, F(b1, b2, …, bm) is a
finite extension of F.
Since each bi is a root of p(x), we have

0  p(bi )  a0  a1bi  an1bin1  bin .

By applying s both sides, we get that

0 5 a0 1 a1s(bi) 11 an21s(bi)n21 1 s(bi)n

and hence s(b1), …, s(bm) are also roots of p(x) in E.

Since s is an injection, s(b1), s(b2), …, s(bm) must be the same as
b1, b2, …, bm in some order. Now, Let

E9 5 F(b1, b2, …, bm).

Then, s(E9) 5 s(F(b1, b2, …, bm))  F(s(b1), s(b2), …, s(bm))

5 F(b1, b2, …, bm) 5 E9.

Therefore, [s(E9) : F] 5 [E9 : F], since s(E9)  E9. b

But, since s(E9) ⊆ E9, it follows that s(E9) 5 E9. In particular, since a 5 b1
 E9, there exists b  E9 such that s(b) 5 a. Thus, s is a surjection also.
­Therefore, s is an isomorphism of E onto E. Thus, s is an automorphism of E.

Worked Exercise 14.2.1. Prove that the field C of complex numbers is an

algebraic extension of the field R of real numbers.

Answer: Clearly C is a field extension of R. Also, any a  R is algebraic

over R. Since i is a root of 1 1 x2  R[x], i is algebraic over R. Thus, any a 1
bi  C is algebraic over R, by Theorem 14.2.8. Thus, C is an algebraic
­extension of R.

Worked Exercise 14.2.2. Prove that an algebraic extension E of F is finitely

generated over F if and only if E is a finite extension of F.

Q001-Algebra-111001_CH 14.indd 18 9/19/2011 4:51:19 PM

 Extension Fields   14-19

Answer: Let E be an algebraic extension of F. Suppose that E is finitely gen-

erated over F. Then, E 5 F(a1, a2, …, an) for some a1, a2, …, an  E. Since E
is an algebraic extension of F, each ai is algebraic over F. Now, by Theorem
14.2.7, E is a finite extension of F. Conversely suppose that E is a finite exten-
sion of F. Let [E : F] 5 n and { a1, a2, …, an} be a basis of E over F. Then,
F(a1, a2, …, an) 5 E and hence E is finitely generated over F.

EXERCISE 14(b)

1. Prove that the polynomial p(x) 5 x2 2 x 2 1  Z3[x] is irreducible over Z3 and

that there is a field extension E of Z3 with exactly nine elements and containing
all the roots of p(x).

2. Find the smallest extension of Q having a root of x4 2 2  Q[x].

3. Determine the degrees of the following:

(i) Q( 3) over Q
(ii) Q( 5) over Q
(iii) Q( 3, 5) over Q
(iv) Q( 3, 5) over Q
4. Find the degree of R over Q.

5. Let F ⊆ K be fields and a and b  K be algebraic over F such that [F(a) : F] 5 m

and [F(b) : F] 5 n, where m and n are relatively prime. Then, prove that [F(a, b) : F]
5 mn.

6. Determine the minimal polynomials over Q of the following:

(i) 5
(ii) 2 5
(iii) 5 3 2
(iv) 1 2
(v) 2 3 3
(vi) 6
7. Prove that there is a  R such that Q( 2, 5)  Q( a)
8. Let F ⊆ E be fields such that [E : F] is a prime number. Prove that there are no
fields property between F and E.

9. Let D be an integral domain and F be a field contained in D as a subring such

that [D : F] is finite. Then prove that D is a field.

10. Let F ⊆ E be fields and a  E be algebraic over F. If the minimal polynomial of

a over F is of odd degree, prove that F(a) 5 F(a2).

Q001-Algebra-111001_CH 14.indd 19 9/19/2011 4:51:20 PM

 14-20   Algebra – Abstract and Modern

14.3 ALGEBRAICALLY CLOSED FIELDS

A field F is called algebraically closed if an element a, in any extension of F,
is algebraic over F, then a  F. In other words, no element outside F is alge-
braic over F. In this section, we discuss various properties of algebraically
closed fields. Let us begin with a formal definition in the following.

Definition 14.3.1. A field K is said to be algebraically closed if it has no proper

algebraic extensions; that is, if K ⊆ L and L is algebraic over K, then K 5 L.
The following provides two characterizations of algebraically closed fields.

Theorem 14.3.1. The following are equivalent to each other for any field K.
1. K is algebraically closed.
2. Any nonconstant polynomial over K can be factored completely into
linear factors in K[x].
3. If f (x) is a nonconstant polynomial over K, then all the roots of f (x)
belong to K.
4. Every nonconstant polynomial over K has atleast one root in K.

Proof: Recall that an element a (in any extension of K) is a root of F(x) 

K[x] if and only if x 2 a is a factor of f (x) and therefore (2) ⇔ (3) is clear.
(1) ⇒ (2): Let f (x) be a nonconstant polynomial over K. Then, by Theorem
14.2.6, there exists a finite extension E of K such that E contains all the roots
of f (x). Since any finite extension is algebraic, E is an algebraic extension
of K and hence E 5 K. Thus, K contains all the roots of f (x) and f (x) can be
factored completely into linear factors in K[x].
(3) ⇒ (4) is trivial.
(4) ⇒ (1): Let L be an algebraic extension of K and let a  L. Then, a is alge-
braic over K. Let p(x) be the minimal polynomial of a over K. Then, p(x)  K[x]
and deg(p(x))  1. Then, by (4), p(x) has a root, say b, in K. Then, x 2 b
divides p(x) in K[x] and hence p(x) 5 x 2 b (since p(x) is irreducible monic
polynomial in K[x]). Now a 2 b 5 p(a) 5 0 and hence a 5 b  K.
Thus, L ⊆ K and L 5 K. Thus, K is algebraically closed. b

Corollary 14.3.1. A field K is algebraically closed if and only if every irre-

ducible polynomial in K[x] is of degree 1.

Definition 14.3.2. An algebraic extension E of a field F is called an alge-

braic closure of F if E is algebraically closed.

Example 14.3.1. The field C of complex numbers is an algebraic closure of

the field R of real numbers.

Q001-Algebra-111001_CH 14.indd 20 9/19/2011 4:51:20 PM

 Extension Fields   14-21

We prove later any two algebraic closures of a field F are isomorphic under
an isomorphism that keeps each element of F fixed. The following theorem is
in the direction of proving the existence of an algebraic closure of any field.
For any field F, we know that for any given indeterminates x1, x2, …, xn,
the set F[x1, x2, …, xn] of all polynomials in x1, x2, …, xn with coefficients in F
is an integral domain and we identify xixj with xjxi; that is, the indeterminates
are commuting with each other. For any set S 5 {xa}a∆ of commuting inde-
terminates, we define

F [S ]  ∪
T ⊆S
F [T ].
T is finite

Then, it can be easily verified that F[S] is an integral domain.

Theorem 14.3.2. Any field has an algebraically closed extension.

Proof: Let F be a field. Let us first construct an extension K1 of F in which

every nonconstant polynomial has a root. For each nonconstant polynomial f 5
f (x)  F[x], we correspond an indeterminate xf and let

S 5 {xf : f 5 f (x)  F[x] and deg(f (x))  1}.

Consider the polynomial ring F[S], which is an integral domain. Let A be the
ideal in F[S] generated by all polynomials f (xf) of positive degree in F[S].
We claim that A is a proper ideal of F[S]. Suppose, if possible that A 5 F[S].
Then, 1  A and therefore

1  g1 f1 ( x f1 )  g 2 f 2 ( x f2 )  g x f x ( x fn ),

where g1, g2, …, gn  F[S]. Note that g1, g2, …, gn will involve only a finite
number of variable (indeterminates). Write x fi  xi for each f i  F[x]. After
reindexing, we can assume that x f1  x1 ,, x fn  xn and the variables occur-
ring in all the gi, 1  i  n, are in the set {x1, x2, …, xn, …, xm}. Therefore,
we have
n
1  ∑ gi ( x1 ,…, xm ) f i ( xi ) (*)
i1 

Now, let E be an extension of F in which each of the polynomials f1, f2, …, fn has a
root and let ai be a root of fi in E, for each 1  i  n. If we substitute xi 5 ai for 1
 i  n and xn+1 5  5 xm 5 0 in (*), we get that 1 5 0, which is absurd.

Q001-Algebra-111001_CH 14.indd 21 9/19/2011 4:51:21 PM

 14-22   Algebra – Abstract and Modern

Thus, A is a proper ideal of F[S] and hence, using the Zorn’s lemma, we get a
maximal ideal M of F[S] containing A. Then, there is an embedding a  a 1
M of F into F [S ]/ M and hence F [S ]/ M can be regarded as a field exten-
sion of F. Also, each nonconstant polynomial f 5 f (x)  F[x] has a root in
F [S ]/ M . Thus, we have constructed a field K1 ( F [S ] M ) which is an exten-
sion of F and in which every nonconstant polynomial in F[x] has a root.
Now, inductively, we can form a chain of fields

F 5 K0 ⊂ K1 ⊂ K2 ⊂ 

such that any nonconstant polynomial over Kn has a root in Kn+1 for all
n  0. ∞
Put K  ∪ K n .
n0
Then, K is a field extension of F. If

g(x) 5 b0 1 b1x 11 bmxm, bm  0, m > 0

is a polynomial over K, then there exists n such that b0, b1, …, bm  Kn and
therefore g(x)  Kn[x] so that g(x) has a root in Kn+1 ⊆ K. Thus, F has an alge-
braically closed extension. b
The following will be useful in proving the uniqueness (up to isomor-
phism) of algebraic closure of a given field.

Theorem 14.3.3. Let F be a field and s : F → L be an embedding of F into an

algebraically closed field L. Let E 5 F(a) be an algebraic extension of F. Then, s
can be extended to an embedding h : E → L and the number of such extensions
is equal to the number of distinct roots of the minimal polynomial of a over F.

Proof: Let p(x) 5 a0 1 a1x 11 an21xn21 1 xn be the minimal polynomial

of a over F and write

ps(x) 5 s(a0) 1 s(a1)x 11 s(an21)xn21 1xn

Then, ps(x)  L[x]. Since L is algebraically closed, ps(x) has all the roots in
L. Let p  L be a root of ps(x). By Theorem 14.2.4,

E 5 F(a) 5 {b0 1 b1a 11 bn21an21 : bi  F}.

Define hp : E → L by

hp(b0 1 b1a 11 bn21an21) 5 s(b0) 1 s(b1)p 11 s(bn21)pn21.

Since p(x) is the polynomial of least degree for which a is a root, it follows
that any element of E ( 5 F(a)) can be uniquely expressed as b0 1 b1a 11

Q001-Algebra-111001_CH 14.indd 22 9/19/2011 4:51:22 PM

 Extension Fields   14-23

bn21an21 and hence hp is well defined. It is a routine verification to prove that hp

is an embedding of F(a) into L and that it is an extension of s; that is, hp(s) 5
s(s) for all s  F. Further, p  hp is a bijective correspondence between the
distinct roots of ps(x) and the extensions of s to E. Therefore, the number of
extensions of s to E is equal to the number of distinct roots of ps(x) in L. b
The above theorem is generalised in the following theorem which is proved
by using the Zorn’s lemma.

Theorem 14.3.4. Let E be an algebraic extension of F and s : F → L be an

embedding of F into an algebraically closed field F. Then, s can be extended
to an embedding h : E → L.

Proof: Consider the set

P 5 {(K, u) : F ⊆ K ⊆ E, u : K → L is an embedding and u/F 5 s}.

Then, P is not empty, since (F, s)  P. Define a binary relation  on P by

(K1, u1)  (K2, u2) ⇔ K1 ⊆ K2 and u2/K1 5 u1.

Then, it can be easily verified that  is a partial order on P. We prove that the
partially ordered set (P, ) satisfies the hypothesis of the Zorn’s lemma. Let
{(Ka, ua)}a∆ be a chain of elements in P. Put

K  ∪ K a and define u : K → L by u(a) 5 ua(a) if a  Ka.

a∈∆

If a  Ka ∩ Kb, then ua(a) 5 ub(a), since either

(Ka, ua)  (Kb, ub) or (Kb, ub)  (Ka, ua).

Therefore, u is a well-defined mapping on K and, clearly K is a subfield of E

containing F. It is easy to verify that u : K → L is an embedding and, clearly
(Ka, ua)  (K, u) for all a  ∆. Therefore, every chain in (P, ) has an upper
bound in P and hence, by the Zorn’s lemma, there is a maximal member, say
(M, h), in (P, ). We prove that M 5 E. Suppose, if possible, that M  E.
Then, choose a  E such that a  M. Since E is an algebraic extension of F,
a is algebraic over F and hence over M. By Theorem 14.3.3, there exists an
h
embedding h9 : M(a) → L such that M  h. But then (M, h) , (M(a), h9)
 P, which is a contradiction to the maximality of (M, h). Thus, M 5 E and
h : E → L is an embedding and is an extension of s. b

Q001-Algebra-111001_CH 14.indd 23 9/19/2011 4:51:22 PM

 14-24   Algebra – Abstract and Modern

Theorem 14.3.5. Every field has an algebraic closure.

Proof: Let F be a field. By Theorem 14.3.2, there exists an extension E of F

which is algebraically closed. Let
K 5 {a  E : a is algebraic over F}.
Then, K is a subfield of E and is an algebraic extension of F. Now, we prove
that K is algebraically closed also. Let f (x)  K[x] be a nonconstant polyno-
mial and a be a root of f (x). Since f (x)  E[x] and E is algebraically closed, we
get that a  E. Also, since a is a root of f (x)  K[x], a is algebraic over K.
Now, K is algebraic over F and hence, by Theorem 14.2.9, a is algebraic over
F. Therefore, a  K. That is, every root of f (x)  K[x] is in K. Thus, K is an
algebraic closure of F. b
The following theorem proves the uniqueness (up to isomorphism) of the
algebraic closure of a given field.
Theorem 14.3.6. Let F be a field and K and K9 be algebraic closures of F.
Then, K is isomorphic to K9 under an isomorphism which is identity on F.
Proof: We can treat F as a subfield of K and K9. Let s : F→ K9 be the inclu-
sion map; that is, s(a) 5 a for all a  F. Since K is an algebraic extension of
F and K9 is algebraically closed, we get (from Theorem 14.3.4) an embedding
s* : K → K9 which is an extension of s. Now
K  s*(K) ⊆ K9
and K9 is an algebraic extension of F and hence an algebraic extension of
s*(K) and therefore s*(K) 5 K9. Thus, s* : K → K9 is an isomorphism such
that s*(a) 5 s(a) 5 a for all a  F. b
Definition 14.3.3. By Theorems 14.3.5 and 14.3.6, any field F has a unique
(up to isomorphism) algebraic closure and, henceforth, we denote the alge-
braic closure of F by F .

Definition 14.3.4. Let F be a field of f (x)  F[x] be of degree n  1. Then, a

field extension E of F is called a splitting field of f (x) over F if the following
are satisfied:
1. f (x) factors into linear factors in E[x]; that is, there exists a1, a2, …, an  E
such that
f (x) 5 c(x – a1) (x – a2)  (x – an).
2. E 5 F(a1, a2, …, an); that is, E is generated by a1, a2, …, an over F.
For any nonconstant polynomial f (x) over a given field F, by Theorem
14.2.6, there exists a field extension K of F which contains all the roots of f (x)

Q001-Algebra-111001_CH 14.indd 24 9/19/2011 4:51:22 PM

 Extension Fields   14-25

and [K : F]  n!, where n 5 deg(f (x)). If E is the intersection of all subfields

of K containing F and the roots a1, a2, …, an of f (x), then E 5 F(a1, a2, …, an)
which becomes the splitting field of f (x) and [E : F]  n!. Also, E is a finite
extension and hence an algebraic extension of F. The following theorem says
that any two splitting fields of a given polynomial over a given field F are
isomorphic under an isomorphism which is the identity on F.

Theorem 14.3.7. Let f (x) be a nonconstant polynomial over a field F and let
E and K be two splitting fields of f (x) over F. Then, there exists an isomor-
phism s : E → K which is identity on F.

Proof: Let K and F be the algebraic closures of K and F, respectively. Then,

F ⊆ K ⊆ K , K is algebraic over K and K is algebraic over F. Therefore, by
­Corollary 14.2.3, K is algebraic over F and hence K 5 F . Let l : F → F 5 K
be the inclusion map. Suppose that

f (x) 5 a0 1 a1x 1…1 anxn  F[x], n . 0,

and b1, b2, …, bn  E be all the roots of f (x) (not necessarily distinct). Then,
E 5 F(b1, b2, …, bn) is an algebraic extension of F. By Theorem 14.3.4, there
exists an embedding l* : E → K such that l*/F 5 Id. Put

f l*(x) 5 l*(a0) 1 l*(a1)x 11 l*(an)xn.

Now, l*(b1), l*(b2), …, l*(bn) are the roots of f l*(x) in K , and if b1, b2, …,
bn  K are the roots of f (x), then

{b1, b2, …, bn} 5 {l*(b1), l*(b2), …, l*(bn)},

since K ⊆ K . Also,

K 5 F(b1, b2, …, bn) 5 F(l*(b1), l*(b2), …, l*(bn))

5 l*(F(b1, b2, …, bn)) 5 l*(E)

Thus, l* is an isomorphism of E onto K such that l*|F 5 Id. b

Worked Exercise 14.3.1. Find the splitting field of x3 – 2 over the field Q of
rational numbers and its degree over Q.

Answer: Let f (x) 5 x3 – 2  Q[x]. By the Eisenstein’s criterion, f (x) is irre-

1
ducible over Q. In fact, it is the minimal polynomial of 2 3. Therefore ,

Q[ x ] 1 1
 Q(2 3 ) and [Q(2 3 ) : Q]  3.
 f ( x)

Q001-Algebra-111001_CH 14.indd 25 9/19/2011 4:51:25 PM

 14-26   Algebra – Abstract and Modern

1
However, Q(2 3 ) is not the splitting field of f (x), since
1 1 2
f (x) 5 x3 2 2 5 (x 2 2 3 ) (x2 1 2 3 x 1 2 3 )

and hence f (x) has two complex roots, say w and w. Let
1 2 1
p(x) 5 x2 1 2 3 x 1 2 3
 Q(2 3 )[x].
1
Then, p(x) is irreducible over Q(2 3 ) and hence

1
Q(2 3 ) [ x ] 1
 Q (2 3 ) (w )  Q (2 3 , w )
1

p(x )

1 1
and [Q(2 3 , w) : Q(2 3 )] 5 2, which is the degree of p(x).
1 1
Further w  Q(2 3 , w). Thus, Q(2 3 , w) is the splitting field of x3 – 2
over Q and

1 1 1 1
[Q(2 3 , w ) : Q]  [Q(2 3 , w ) : Q (2 3 )][Q(2 3 ) : Q]
 3.2  6.

Therefore, the degree of the splitting field of x3 2 2 over Q is 6.

Worked Exercise 14.3.2. Let f (x) 5 x2 1 3 and g(x) 5 x2 1 x 1 1 be poly-

nomials over Q. Prove that their splitting fields are equal and find its degree
over Q.

Answer: 3i and  3i are the roots of f (x) and hence Q ( 3i) is the splitting
1 3 i 1 3
field of f (x). Further, are the roots of g(x). Put w  .
2 2
Then, 3 5 2w 1 1 and hence 3  Q(w). Therefore, Q( 3 ) ⊆
Q(w). Also,

1 1
w   3  Q ( 3)
2 2

and hence Q(w) ⊆ Q( 3). Thus, Q(w) 5 Q( 3 ) is the splitting field of f (x)
and g(x). Also, f (x) and g(x) are both irreducible over Q and [Q(w) : Q] 5 2.

Q001-Algebra-111001_CH 14.indd 26 9/19/2011 4:51:27 PM

 Extension Fields   14-27

EXERCISE 14(c)

1. Let K be an algebraically closed field and F be a subfield of K. Then prove that

the algebraic closure F of F in K is also algebraically closed.

2. If F1 ⊂ F2 ⊂ … is a chain of fields such that Fn+1 is a field extension of Fn for all

∞
n > 0, then prove that ∪ Fn is a field extension of each Fn.
n1

3. Construct splitting fields of the following polynomials over the field Q of ratio-
nal numbers.
(i) x4 1 1
(ii) x3 2 1
(iii) x6 2 1

4. Prove that no finite field is algebraically closed.

5. Construct a splitting field for x3 1 x 1 1 over the field Z2 and list all its elements.

6. In the proof of Theorem 14.3.4, prove that K  ∪ K a is a subfield of E and that

a
the map u : K → L is an embedding.

7. Prove that the degree of a splitting field of a polynomial of degree n (> 0) over a
field F is almost n!

8. Let p(x)  F[x] be an irreducible polynomial over F. If p(x) has one root in
a splitting field E of a polynomial f (x)  F[x], then prove that p(x) has all its
roots in E.

9. Let E be any field extension of Q. Prove that the polynomial x3 2 3x 1 1 is either

irreducible or splits into linear factors over E.

10. For any prime p, find the splitting field of xp 2 1 over Q and its degree over Q.

14.4 DERIVATIVES AND MULTIPLE ROOTS

In this section, we introduce the notion of the derivative of a polynomial as
the usual derivative of a function in calculus and relate this to the multiplicity
of a polynomial. Let us recall that a is a root of a polynomial if and only if
x 2 a is a divisor of f (x). If (x 2 a)n divides f (x) for x > 1, then a is called a
multiple root of f (x).
The properties of derivatives which are familiar from the calculus are not
necessarily valid here. For example, f 9(x) 5 0 does not imply that f (x) is
a constant. Consider the polynomial f (x) 5 x3 over the field Z3 of integers
modulo 3. Then, f 9(x) 5 3x2 5 0. However, the ordinary rules for operating
with derivatives remain the same. Let us formally define the derivative of a
polynomial in the following definition.

Q001-Algebra-111001_CH 14.indd 27 9/19/2011 4:51:27 PM

 14-28   Algebra – Abstract and Modern

Definition 14.4.1. Let f (x) 5 a0 1 a1x 1 a2x2 1…1 anxn be a polynomial

over a field F. Then, the derivative of f (x) is defined as

f 9(x) 5 a1 1 2a2x 1 3a3x2 11 nanxn21.

In particular, the derivative of any constant polynomial is defined to be 0.

Example 14.4.1
1. Let f (x) 5 2 1 3x 1 x2 1 4x3 1 2x4  Q[x]. Then,
f 9(x) 5 3 1 2x 1 12x2 1 8x3

2. Let f (x) 5 1 1 x 1 x2 1 x3 1 x4  Z2[x], where Z2 is the field of integers

modulo 2, Then,
f 9(x) 5 1 1 x2

since 2  0, 3  1 and 4  0 in Z2.

3. Let f (x) 5 2 1 3x 1 4x2 1 2x3 1 3x4 1 2x5  Z5[x].
Then, f 9(x) 5 3 1 3x 1 x2 1 2x3
since 8  3, 6  1, 12  2 and 10  0 in Z5.
The following are routine verifications.

Theorem 14.4.1. Let F be any field, f (x) and g(x)  F[x] and a  F. Then
the following holds.
1. (f (x) 1 g(x))9 5 f 9(x) 1 g9(x)
2. (a f (x))9 5 a f 9(x)
3. (f (x)g(x))9 5 f 9 (x)g(x) 1 f (x) g9(x)
4. If char(F) 5 0 and deg(f (x)) 5 n > 0, then deg(f 9(x)) 5 n 2 1
5. If char(F) 5 p and f (x) 5 xp, then f 9(x) 5 0.

Definition 14.4.2. Let F be any field and f (x)  F[x]. Let E be any extension
of F and a  E. For any positive integer m, if (x 2 a)m divides f (x) in E[x] and
(x 2 a)m+1 does not divide f (x), then a is called a root of f (x) of multiplicity m
and m is called the multiplicity of the root a. A root of multiplicity 1 is called
a simple root and a root of multiplicity m > 1 is called a multiple root.
Earlier in Ring Theory, we have proved that a polynomial degree n over F
can have at most n roots in any extension of F. For this counting purpose, we
shall count a root a as m roots if a is a root of multiplicity m and not as one
root. In the following, we obtain a necessary and sufficient condition in terms
of the derivative of f (x) for a root of f (x) to be a multiple root.

Q001-Algebra-111001_CH 14.indd 28 9/19/2011 4:51:28 PM

 Extension Fields   14-29

Theorem 14.4.2. Let f (x)  F[x] be a nonconstant polynomial and a be an

element in any field extension E of F. Then, a is a multiple root of f (x) if and
only if f (a) 5 0 5 f 9(a); that is, a is a root of both f 9(x) and f 9(x).

Proof: Suppose that a is a multiple root of f (x) of multiplicity m > 1. Then,

(x 2 a)m divides f (x) in E[x] and hence

f (x) 5 (x 2 a)mg(x) for some g(x)  E[x].

Now, f 9 (x) 5 m(x 2 a)m21g(x) 1 (x 2 a)mg9(x), m 2 1 > 0

and therefore f 9(a) 5 0 5 f (a); that is, a is a common root of f (x) and f 9(x).

Conversely suppose that f 9(a) 5 0 5 f (a). Then,

f (x) 5 (x 2 a) p(x)

for some p(x)  E[x], and therefore

f 9(x) 5 p(x) 1 (x 2 a) p9(x)

and hence 0 5 f 9(a) 5 p(a) 1 (a 2 a) p9(x) 5 p(a). Therefore, a is a root of

p(x) so that p(x) 5 (x 2 a) h(x) for some h(x)  E[x]. Now,

f (x) 5 (x 2 a) p(x) 5 (x 2 a)2 h(x).

Therefore, (x 2 a)m divides f (x) for some m > 1. Thus, a is a multiple

root of f (x). b

Corollary 14.4.1. Let f (x)  F[x]. Then, f (x) has only simple roots in any
extension E of F if and only if the g.c.d.{f (x), f 9(x)} is a unit in E[x].

Corollary 14.4.2. Let f (x) 5 xn 2 1  F[x] where n > 0. Suppose that

char(F) 5 0 or char(F) 5 p where p does not divide n. Then, the roots of
f (x) are all distinct.

Proof: We have f 9(x) 5 nxn21 and hence, if a is a multiple root of f (x), we

have 0 5 f 9(a) 5 nan21, so that a 5 0 (since n > 0) which is a contradiction
(since 0 is not a root of xn21). Thus, f (x) has no multiple roots; that is, the
roots of f (x) are all distinct. b

Theorem 14.4.3. Let f (x) be an irreducible polynomial over F. Then, f (x) has
a multiple root in some field extension of F if and only if f 9(x) 5 0.

Q001-Algebra-111001_CH 14.indd 29 9/19/2011 4:51:28 PM

 14-30   Algebra – Abstract and Modern

Proof: Suppose that f 9(x) 5 0. If E is a field extension of F and a  E is a

root of f (x), then f (a) 5 0 5 f 9(a) and hence, by Theorem 14.4.2, a is a mul-
tiple root of f (x) in E.
Conversely suppose that f (x) has a multiple root a in a field extension E of F.
Then, by Theorem 14.4.2, a is a root of f 9(x). Suppose, if possible, that f 9(x)  0.
Let p(x) be the minimal polynomial of a over F. Then, since f (a) 5 0 5 f 9(a), we
get that p(x) divides f (x) and f 9(x). But, since f (x) is irreducible, if follows that
f (x) is an associate of p(x) and hence f (x) divides f 9(x), which is a contradiction,
since deg(f 9(x)) , deg(f (x)).
Thus, f 9(x) 5 0. b

Theorem 14.4.4. The following holds for any irreducible polynomial f (x)
over a field F.
1. If char(F) 5 0, then f (x) has no multiple roots.
2. When char(F) 5 p  0, f (x) has a multiple root if and only if f (x) 5
g(xp) for some g(x)  F[x].

Proof:
1. Let char(F) 5 0 and f (x) 5 a0 1 a1x 11 anxn.
Since f (x) is irreducible, deg(f (x)) 5 n > 0. Then,
f 9(x) 5 0 ⇒ a1 1 2a2x 11 nanxn21 5 0
⇒ a1 5 a2 5  5 an 5 0 (since char(F) 5 0)
⇒ f (x) 5 a0, which is not true.
Therefore f 9(x)  0 and hence, by the above theorem, f (x) has no mul-
tiple roots.
2. Let char(F) 5 p  0 and f (x) 5 a0 1 a1x 11 anxn with an  0 and n >
0. Then, f 9(x) 5 a1 1 2a2x 11 nanxn21. Suppose that f (x) has a multiple
root. Then, by Theorem 14.4.3, f 9(x) 5 0 and hence a1 5 2a2 5 3a3 5 
5 nan 5 0. Since char(F) 5 p, it follows that, for each 1  K  n, either
aK 5 0 or p divides K (that is, K 5 K1p for some integer K1 > 0).
Thus, we have
f (x) 5 b0 1 b1xp 1 b2x2p 11bmxmp 5 g(xp)
for some positive integer m, where g(x) 5 b0 1 b1x 11 bmxm.
­Conversely suppose that f (x) 5 g(xp) 5 b0 1 b1xp 11 bmxmp. Then,
f (x )  pb1 x p1  pmbm x mp1  0 (since char(F) 5 p) and hence,
again by Theorem 14.4.3, f (x) has a multiple root. b

Theorem 14.4.5. All the roots of an irreducible polynomial over a field F

have the same multiplicity.

Q001-Algebra-111001_CH 14.indd 30 9/19/2011 4:51:28 PM

 Extension Fields   14-31

Proof: Let p(x) be an irreducible polynomial over F. Let F be the algebraic

closure of F and let a and b be roots p(x) in F with multiplicities m and m9,
respectively.
Then, we know that,

F [x ]  F (a) and F [x ]  F (b)

p(x ) p(x )

under the isomorphisms s: F [ x ]  p ( x ) → F ( a) and t : F [ x ]  p ( x )  F (b) and

given by

s(f (x) 1 ,p(x)>) 5 f (a)

        and t(f (x) 1 ,p(x)>) 5 f (b).

Now, put h 5 t o s21. Then, h : F(a) → F(b) is an isomorphism and is

given by

h(a0 1a1a 11 anan) 5 a0 1 a1b 11 anbn.

Then, h can be extended to an isomorphism h* : F → F (b)  F .

Let a : F [x ] → F [x ] be the ring homomorphism induced by h*. Then, a is
given by

a(a0 1 a1x 11 asxs) 5 h*(a0) 1 h*(a1)x 11 h*(as)xs.

Note that a( p(x ))  p( x ). Since a(x 2 a)m 5 (x 2 b)m, we get that (x 2 b)m is a
factor of p(x) and hence m  m9. By interchanging the roles of a and b, we get
that m9  m. Thus, m 5 m9 and hence a and b are of same multiplicities. b

Worked Exercise 14.4.1. Prove that a polynomial f (x)  F[x] has a multiple
root if and only if f (x) and f 9(x) have a nonconstant common factor.

Answer: Suppose that f (x) has a multiple root a in an extension E of F. Then,

f (x) 5 (x 2 a)mg(x), m > 1 and g(x)  E[x]

Therefore, f 9(x) 5 m(x 2 a)m21g(x) 1 (x 2 a)mg9(x)

     5 (x 2 a)m21 (mg(x) 1 (x 2 a)g9(x)), m 2 1 > 0

and hence (x 2 a)m21 is a nonconstant common factor of f (x) and f 9(x).

Q001-Algebra-111001_CH 14.indd 31 9/19/2011 4:51:29 PM

 14-32   Algebra – Abstract and Modern

Conversely suppose that f (x) and f 9(x) have a nonconstant common factor,
say p(x) with deg(p(x)) > 0.
Suppose, if possible, that all the roots of f (x) are distinct.
Let a1, a2, …, an be all the distinct roots of f (x) in any extension of F. Then,
n
f (x )  a∏ (x  ai ), for some a ∈ F .
i1

n  
      and f (x )  a∑ ∏ (x  a j ).
i1 
 ji 
Therefore, for each 1  i  n, f ( ai )   (ai  a j )  0 and hence no root of
j ≠i
f (x) is a root of f 9(x). Thus, f (x) and f 9(x) have no nonconstant common factor,
which is a contradiction to the hypothesis. Thus, f (x) has a multiple root.

Worked Exercise 14.4.2. Let F be a finite field and f (x) be an irreducible

polynomial over F. Then prove that f (x) has no multiple roots.

Answer: We will be proving later in the next section that |F| 5 pn, where p is
prime, char(F) 5 p and n  Z+ and a  ap is an automorphism of the field F.
Suppose, if possible, that f (x) has a multiple root. Then, by Theorem 14.4.4 (2),

f (x) 5 g(xp) 5 a0 1a1xp 11 anxnp, ai  F.

Since a  ap is an automorphism of F, we can choose b0, b1, …, bn  F such

that ai 5 bip for each 1  i  n.
Now,

f (x )  b0p  b1p x p  b2p x 2 p  bnp x np

5 (b0 1 b1x 1 b2x2 11 bnxn)p

which implies that f (x) is reducible over F, which is a contradiction to the

hypothesis. Thus, f (x) has no multiple roots.

EXERCISE 14(d)

1. Prove Corollary 14.4.1.

n
2. If f (x )   gi (x ), then prove that
i1
n  
f (x )  ∑ ∏ g j (x ) gi ( x )

i1  ji 

3. Let f (x)  F[x] and deg(f (x)) 5 n > 0. If char(F) 5 p and f 9(x) 5 0, then prove
that p divides n and f (x) has at most np distinct roots.

Q001-Algebra-111001_CH 14.indd 32 9/19/2011 4:51:30 PM

 Extension Fields   14-33

4. Let a be a root of f (x)  F[x]. Prove that a is a root of multiplicity m > 1 if

and only if f (r)(a) 5 0 for each 1  r , m and f m(a)  0, where f (r)(a) is the r th
derivative of f (x) at x 5 a. The rth derivative of f (x) is inductively defined by
f (r)(x) 5 (f (r21)(x))9 and f (0)(x) 5 f (x).
5. Let F be a field, a and b  F, f (x)  F[x] and deg(f (x)) 5 n. Prove that
n
1 (r )
f (a  b)  ∑ f ( a )b r .
r0 r!

6. Let F be a field of characteristic 3 and E 5 F[x], the field of quotients of the

integral domain F[x]. Prove that the polynomial y3 2 x  E[y] is irreducible over
E and has multiple roots.

14.5 FINITE FIELDS

Recall that a field having no proper subfields is called a prime field and that
any field E is an extension of a prime field, which is precisely the intersection
of all subfields of E. Also recall that Zp, where p is a prime, and Q are the
only (up to isomorphism) prime fields. If E is a field of characteristic zero,
then Q is the prime subfield of E and, if char(E) 5 p, then Zp is the prime
subfield of E.
In this section, we mainly discuss about finite fields which have necessar-
ily Zp as their prime subfields, where p is the characteristic of the given field.
Let us begin with the following definition.

Definition 14.5.1. Let F be a field of characteristic p, where p is a prime.

Then the intersection of all subfields of F is called the prime subfield of F and
is denoted by Fp. Note that Fp has no proper subfields and hence Fp is a prime
field. Also Fp  Zp, the field of integers modulo p.
The following theorem determines the number of elements in any given
finite field.

Theorem 14.5.1. Let F be a finite field of characteristic p. Then, |F| 5 pn for

some positive integer n.

Proof: It is well known that the characteristic of a finite field must be a

prime number. Therefore, p is a prime number. Consider the prime sub-
field Fp of F. Then, Fp  Zp and hence |Fp| 5 p. Also, F becomes a vector
space over Fp. Since F is finite, F is a finite dimensional vector space over
Fp and hence F  Fpn , where n 5 [F : Fp], the dimension of F over Fp. Thus,
|F |  |Fpn |  |Fp |n  p n , n  0.  b

Q001-Algebra-111001_CH 14.indd 33 9/19/2011 4:51:31 PM

 14-34   Algebra – Abstract and Modern

We have proved above that the number of elements in any finite field is of
the form pn, for some prime p and n  Z1. On the other hand, we prove in
the next two results below that, for any prime p and a positive integer n, there
exists unique (up to isomorphism) finite field with exactly pn elements. First,
we take up the uniqueness.

Theorem 14.5.2. Let F be a finite field with pn elements, where p is a prime

andn n is a positive integer. Then F is the splitting field of the polynomial
x p  x  Z p [ x ].

Proof: First recall that the prime subfield Fp is isomorphic to Zp and hence
Zp can be considered as the prime subfield of F. Let F * 5 F – {0}. Then, F *
pn 1
is a group under multiplicationn
and |F | 5 p – 1. Therefore, a
* n
 1 for all
a  F *. This implies that a p  na for all a  F. Therefore, every element of
F is a root of the polynomial x p  x, which can have at most pn roots in any
pn
extension of F. Thus, F is precisely n
equal to the set of all roots of x  x and
hence F is the splitting field of x p  x, over Zp. b
Since any two splitting of a polynomial are isomorphic, the following is an
immediate consequence of the above theorem.

Corollary 14.5.1. If E and F are finite fields and |E| 5 pn 5 |F|, where p is a
prime and n is a positive integer, then E  F.

Theorem 14.5.3. Let p be a prime and n be a positive integer. Then there

exists a field F with exactly pn elements.
n
Proof: Consider f ( x )  x p  x ∈ Z p [x ]. Let K be an extension of Zp con-
taining all the roots of f (x). Then,
n
f ( x )  p n x p 1 1 1  0

and hence all the roots of f (x) are distinct. We have to prove that the roots of
f(x) form a subfield of K. If a and b are roots of f (x), then
n n n
( a  b) p  a p  b p  a  b (since p divides pnc r )
n
(−a) p a (if p 5 2, note that a 5 –a)
n n n
( ab) p  a p b p  ab
n n
   and ( a1 ) p  ( a p )1  a1, if a  0.

Thus, the set E of all roots of f (x) forms a field and, since the roots are distinct,
|E| 5 deg(f (x)) 5 pn. Note that E is the splitting field of f (x) over Zp. b

Q001-Algebra-111001_CH 14.indd 34 9/19/2011 4:51:32 PM

 Extension Fields   14-35

Corollary 14.5.2. For any prime p and for any positive integer n, there exists
unique (upn to isomorphism) field E such that |E| 5 pn and this is the splitting
field of x p  x over Zp.

Definition 14.5.2. Let p be a prime and n be a positive integer. Then the

field with pn elements is called the Galois field of order pn and is denoted by
GF(pn).
We have proved in Theorem 14.5.1 that any finite field is a GF(pn) for some
prime p and n  Z1. In the following, we prove that any finite field F has an
extension of any given finite degree over F.

Theorem 14.5.4. Let F 5 GF(pn) and m be a positive integer. Then there

exists a field extensions E of F such that [E : F] 5 m and any two such exten-
sions are isomorphic.

Proof: Consider the algebraic closure F of F and the polynomial

mn
f ( x )  x p  x  F [ x ].
n
Since the multiplicative group F * 5 F – {0} is of order pn –1, we have a p 1  1
for all a  F * Also, since

(pn–1)(pn(m21) 1 pn(m22) 11 pn 1 1) 5 pnm–1,

nm
1
we get that a p  1 for all a  F * and hence

nm
a p  a for all a  F.

Therefore, each element of F is a root of f (x) and, as in the proof of Theorem

14.5.3, the pmn roots of f (x) are distinct and form a field E. Now, we have

Zp  Fp ⊂ F ⊂ E ⊂ F

where [F : Fp] 5 n and [E : Fp] 5 mn and thus [E : F] 5 m. b

The following is an important result regarding the multiplicative group of
nonzero elements of a finite field.

Theorem 14.5.5. Let F be a field and F * 5 F – {0}, the group of nonzero

elements of F under the multiplication. Then, F * is a cyclic group if and only
if F is a finite field.

Q001-Algebra-111001_CH 14.indd 35 9/19/2011 4:51:33 PM

 14-36   Algebra – Abstract and Modern

Proof: Suppose that F * is a cyclic group and let

F * 5 ,a. 5 {an : n  Z}.

That is, F * is the cyclic group generated by a. Now, if char(F) 5 0, then

21  1 and 21  F * 5 ,a. and hence 21 5 an so that 1 5 (21)2 5 a2n
for some n  0, which implies that a is of finite order in the group F * and
hence F * and F are finite.
Therefore, we can suppose that char(F) 5 p, a prime. Since F * 5 ,a., we get
that Fp(a) 5 F. If 1 1 a 5 0, then a2 5 1 and hence O(a) is finite in the group
F * 5 ,a., so that F * and F are finite. Suppose that 1 1 a  0. Then 1 1 a 5
ar for some r  Z and hence a is a root of the polynomial xr – x – 1  Fp[x]. This
implies that a is algebraic over Fp and hence [Fp(a) : Fp] and [Fp(a) : F] are
finite. Therefore, [F : Fp] is finite. Thus, F is finite. Conversely suppose that F
is a finite field. We know that F * is abelian group. Let

F * 5 {a1, a2, …, am}.

Then, since F * is an abelian group, there exists a  F * such that

O(a) 5 l.c.m.{O(a1), O(a2), …, O(am)}.

If O(a) 5 r, then clearly r  m 5 |F *|. Also, ari 5 1 for all 1 # i # m and

hence a1, a2, …, am are the roots of the polynomial xr – 1 which has at most
r roots in any extension of F. Therefore, m # r. Thus, m 5 r 5 O(a) so that
,a. 5 F *, which implies that F * is a cyclic group. b

Corollary 14.5.3. Any finite extension E of a finite field F is a simple exten-

sion; that is, E 5 F(a) for some a  E.

Proof: Let F be a finite field and E be a finite extension of F. Let [E : F] 5 n.

Then E is a n dimensional vector space over F and hence |E| 5 |F|n, so that E
is a finite field. By the above theorem, E* is a cyclic group generated by a  E.
Now, E is the smallest subfield of E containing F and a. Thus, E 5 F(a). b

Corollary 14.5.4. Let F be a finite field and n be a positive integer. Then,

there exists an irreducible polynomial of degree n over F.

Proof: By Theorem 14.5.4, there exists an extension E of F such that [E : F]

5 n. Then, by Corollary 14.5.3, E 5 F(a) for some a  E. Since E is a finite
extension of F, a is algebraic over F. Let p(x) be irreducible over F and

deg(p(x)) 5 [F(a) : F] 5 [E : F] 5 n. b

Q001-Algebra-111001_CH 14.indd 36 9/19/2011 4:51:33 PM

 Extension Fields   14-37

For any field F, let us recall that the set Aut(F) of all automorphisms of F is
a group under the composition of mappings. In the following, we prove that
Aut(F) is a cyclic group for any finite field F.

Theorem 14.5.6. Let F be a finite field of characteristic p and |F| 5 pn. Then,
Aut(F) is a cyclic group of order n.

Proof: Define f : F → F by f (a) 5 ap for any a  F. By using the fact that p

is the characteristic of F, it can be easily verified that f is a homomorphism.
Also, for any a and b  F,

ap 5 bp ⇒ (a2b)p 5 0 ⇒ a2b 5 0 ⇒ a 5 b.

Therefore, f is an injection of F into F. Since F is finite, it follows that f is

a surjection also. Thus, f is an automorphism of F. We prove that O(f) 5 n
n
and |Aut(F)| 5 n. Since |F *| 5 pn – 1 , we get that a p 1  1 for all a  F *
and hence
n
f n (a)  a p  a for all a  F.

Therefore, fn 5 Id and hence O(f) # n. On the other hand, suppose fd 5

pd
Id for some d . 0. Then, a  ad, for all a  F and hence every element of
F is a root of the polynomial x p  x. Therefore, pn # pd and hence n # d.
Thus, O(f) 5 n. Now, we have Fp # F # F and by Corollary 14.5.3, F 5
Fp(a) where F * 5 ,a.. Let p(x) be the minimal polynomial of a over F.
Then,

[Fp(a) : Fp] 5 deg(p(x)) 5 n.

Further,
n
by Theorem 14.5.2, F is equal to the set of all roots of the polynomial
x p  x ∈ Fp [x ]. If s : Fp(a) → F is an embedding, then s(a)  F 5 Fs(a)
n
(since s(a) is a root of x p  x), which implies that s(F) ⊆ F. Since F is finite
and s is an injection, we get that s(F) 5 F.
By Theorem 14.3.3, we know that the number of extensions s : Fp(a) → F
is equal to the degree of the irreducible polynomial satisfied by a. Therefore,
Aut(F) contains precisely n elements. Since O(f) 5 n, f is a generator of
Aut(F). Thus, Aut(F) is a cyclic group of order n. b

Worked Exercise 14.5.1. Let F be a finite field and |F| 5 pn, p a prime and
n  Z1. For each divisor m of n, prove that F has exactly one subfield E of
F such that |E| 5 pm.

Q001-Algebra-111001_CH 14.indd 37 9/19/2011 4:51:34 PM

 14-38   Algebra – Abstract and Modern

Answer: Let us recall from group theory that any cyclic group of order n has
a unique subgroup of order d for each divisor d of n. Let m be a divisor of n.
Consider the cyclic group F * 5 F – {0} of order pn – 1. Then, pm – 1 divides
pn – 1; for, if md 5 n, then

pn – 1 5 (pm – 1) (pm(d21) 1 pm(d22) 1…1 pm 1 1).

Therefore, there exists a unique subgroup G of F * of order pm – 1. Then, for

m m
all a m G, a p 1  1 and hence a p  a for all a  G ∪ {0}. Since the roots
of x p  x form a field, it follows that H ∪ {0} is the unique subfield of F of
order pm.

EXERCISE 14(e)

1. Construct fields with 4, 8, 9 and 16 elements.

2. Find a generator for the multiplicative groups of nonzero elements of a field with
8 elements.

3. Prove that a finite extension E of a field F is a simple extension of F if and only

if there are only a finite number of subfields of E containing F.

4. Let F be a field such that |F| 5 4. Then find irreducible polynomials over F of
degree 2, 3 and 4.
n
5. Prove that for n  3, the polynomial x 2  x 1 is irreducible over the field Z2
of integers modulo 2.

6. Let F be a finite field. Prove that any element of F can be expressed as the sum
of two squares.

7. Let a and b be two elements of a finite field F. Then prove that there exist ele-
ments a and b in F such that a 1 aa2 1 bb2 5 0.

8. Let F be a finite field and char(F) 5 p. Prove that each element a of F has a
p
unique pth root a in F.

Q001-Algebra-111001_CH 14.indd 38 9/19/2011 4:51:35 PM

 15
Galois Theory
15.1 Separable and Normal Extensions
15.2 Automorphism Groups and Fixed Fields
15.3 Fundamental Theorem of Galois Theory

Galois theory of fields is one of the most elegant theories in Abstract Algebra
and it is an excellent combination of group theory and the theory of algebraic
field extensions. In general, Galois theory and, in particular, the fundamental
theorem of Galois theory has several applications to the theory of equations
and geometry. Although we are not making a detailed study of Galois theory,
we shall discuss its fundamental concepts and give certain simple applica-
tions like proving the fundamental theorem of algebra and the nonconstruc-
tability of certain geometric figures using straight-edge and compass alone.
In this chapter, we introduce the concepts of separable extensions and normal
extensions and prove certain important properties of these. Also, we prove the
celebrated theorem ‘the fundamental theorem of Galois theory’ which estab-
lishes a one-to-one correspondence between the subfields of a splitting field
E of a separable polynomial in F[x] and the set of subgroups of the group of
F-automorphisms of E. This one-to-one correspondence transforms certain
problems of subfields of fields into more simpler and amenable problems
about the subgroups of groups. Certain applications of Galois theory are dis-
cussed in the next chapter.

15.1 SEPARABLE AND NORMAL EXTENSIONS

In this section, we discuss certain special extensions of a field, namely sepa-
rable extensions and simple extensions and prove certain important proper-
ties of these. First, we introduce the notion of a separable polynomial in the
following definition.

Q001-Algebra-111001_CH 15.indd 1 9/16/2011 10:11:55 AM

 15-2   Algebra – Abstract and Modern

Definition 15.1.1. Let F be any field. An irreducible polynomial in F[x] is called

a separable polynomial if all its roots are distinct (that is, if it has no multiple
roots). Any polynomial f (x)  F[x] is called separable if all the irreducible factors
of f (x) are separable. A polynomial that is not separable is called i­nseparable.

Example 15.1.1
1. Clearly, any polynomial of degree one is separable.
2. x2 1 1  R[x] is an irreducible polynomial and its roots i and 2i are
distinct and therefore x2 1 1 is a separable polynomial.
3. The polynomial 1 1 x 1 x2  Q[x] is separable, since its roots are
(1)  ( 3 )i and it is an irreducible polynomial.
2 2
4. The polynomial 1 1 2x 1 x2 is separable, since its irreducible factors are
1 1 x alone. In fact, for any a  F, (x 2 a)n is a separable polynomial.

Definition 15.1.2. Let E be a field extension of a field F and a  E. Then, a

is said to be separable over F if a is algebraic over F and the minimal poly-
nomial of a over F is separable. E is said to be a separable extension of F if
every element of E is separable over F.
Clearly, every separable extension of F is an algebraic extension of F.
However, not every algebraic extension is separable. If char(F) 5 0, then
by Theorem 14.4.4, every algebraic extension of F is a separable extension.
Also, by Worked Exercise 14.4.2, every algebraic extension of a finite field F
is a separable extension. In the following example, we give an example of an
algebraic extension which is not separable.

Example 15.1.2. Let K be the field of quotients of Z3[x] and consider

f (y) 5 y3 2 x  K[y].

We prove that f (y) is irreducible over K and has multiple roots. If f (y) is
reducible over K, there must exist a root (g ( x ) h( x ))( h( x ) ≠ 0) in K which
implies that

(g ( x))
3

x
(h( x))
3

and hence 3deg(h(x)) 1 1 5 3deg(g(x)) which is absurd. Therefore, f (y) is

irreducible over K. Also, if b1 and b2 are roots of f (y) in its splitting field, then

b13  x  b23

Q001-Algebra-111001_CH 15.indd 2 9/16/2011 10:11:56 AM

 Galois Theory   15-3

which implies that (b1  b2 )3  b13  b23  0 (note that char(K) 5 3) and hence
b1 2 b2 5 0. This proves that f (y) has only one root whose multiplicity is
three. If a is a root of f (y) in any extension of K, then K(a) is an algebraic
extension of K and K(a) is not a separable extension.

Definition 15.1.3. A field F is called a perfect field if each of its algebraic

extension is separable.
Any field of characteristic zero is a perfect field and likewise, any finite
field is also perfect. The field K given in Example 15.1.2 is not perfect.
Let us recall that an extension E of F is called simple if E 5 F(a) for some
a  E, where a is algebraic over F. By Corollary 14.5.4, every finite exten-
sion of a finite field is simple. This result is extended to separable extensions
of any field in the following theorem.

Theorem 15.1.1. Any finite separable extensions of any field F is a simple

extension of F.

Proof: Let F be any field and E be a finite separable extension of F. By

­Corollary 14.5.4, we may assume that F is an infinite field. We first prove
that, for any a and b  E, there exists c  E such that F(a, b) 5 c and then,
by using induction on n, we can prove that F(a1, a2, …, an) 5 F(c) for some c
 E. Now, let a and b  E. Since E is a separable extension, it is an algebraic
extension of F. In particular, a and b are algebraic over F. Let f (x) and g(x)
be the minimal polynomials of a and b, respectively, over F. Let a 5 a1, a2,
…, an be the roots of f (x) in its splitting field and b 5 b1, b2, …, bm be the
roots of g(x) in its splitting field. By hypothesis, ai’s are distinct and bi’s are
distinct. Put

ai  a
aij = for 1  i  n and 1 < j  m.
b bj
Since F is infinite, we can choose 0  a  F such that a  aij for all 1 
i  n and 1 < j  m. Then,

ai  a
a≠ and hence ab 2 abj  ai 2 a
b bj
for all 1  i  n and 1 < j  m. Take c 5 a 1 ab.
Then, clearly F(c) ⊆ F(a, b) (since a  F).
Now, put h(x) 5 f (c 2 ax)  F(c)[x].
Then, h(b) 5 f (c 2 ab) 5 f (a) 5 0.
Also, h(bj) 5 f (c 2 abj)  0 for all j > 1, since c 2 abj  ai for all
i and a1, a2, …, an are the roots of f (x). Therefore, x 2 b is the only common

Q001-Algebra-111001_CH 15.indd 3 9/16/2011 10:11:56 AM

 15-4   Algebra – Abstract and Modern

factor of g(x) and h(x) in F(c)[x]. Now, b is algebraic over F and hence over
F(c). If p(x)  F(c)[x] is the minimal polynomial of b over F(c), then p(x)
divides g(x). Also, p(x) divides h(x) and hence p(x) divides x 2 b which
implies that p(x) 5 x 2 b. Therefore, b  F(c) and hence ab  F(c), since
a  F. Now, a 5 c 2 ab  F(c). Thus, a and b  F(c) and hence F(a, b)
⊆ F(c). Thus, F(a, b) 5 F(c). If E 5 F(a1, a2, …, an), then by using induc-
tion on n, we can prove that E 5 F(c) for some c  E. Thus, E is a simple
extension of F. b
Since any finite extension of a field is an algebraic extension and since any
algebraic extension of a field of characteristic zero is a separable extension,
the following is an immediate consequence of the above theorem.

Corollary 15.1.1. Every finite extension of a field of characteristic zero is a

simple extension.

Theorem 15.1.2. A finite extension E of a field F is a simple extension of F if

and only if there are only finitely many intermediate fields between F and E.

Proof: Let E be a finite extension of a field F. Suppose that E is a simple

extension of F and a  E such that E 5 F(a). Let p(x) be the minimal poly-
nomial of a over F. Let K be a subfield of E containing F. Now, a is algebraic
over K and let

q(x) 5 a0 1 a1x 1 a2x2 1  1 an21xn21 1 xn, ai  K

be the minimal polynomial of a over K. Then, clearly q(x) divides p(x) in K[x]
and [K(a) : K] 5 n. Consider K9 5 F(a0, a1, …, an21). Then,

q(x)  K9[x], K9 ⊆ K and [K9(a) : K9] 5 n

and therefore we have

E 5 F(a) ⊆ K9(a) ⊆ K(a) ⊆ E(a) 5 E.

Therefore, K9(a) 5 K(a). Also,

[K(a) : K] 5 [K9(a) : K9] 5 n and K9 ⊆ K

and hence we get that K 5 K9. Thus, we have proved that, for any intermediate
field K between F and E,

K 5 F(a0, a1, …, an21),

Q001-Algebra-111001_CH 15.indd 4 9/16/2011 10:11:57 AM

 Galois Theory   15-5

where q(x) 5 a0 1 a1x 1  1 an21xn21 1 xn is a divisor of p(x) in K[x]. Since

the number of monic polynomials dividing p(x) is finite, it follows that the
number of intermediate fields between F and E is finite.
Conversely suppose that the number of intermediate fields between F and E is
finite. If F is finite field, then E is finite and hence E 2 {0} is a cyclic group
and, if E 2 {0} 5 <a>, then F(a) 5 E and hence E is a simple extension of F.
Now, suppose that F is infinite. For each a  E, we have F ⊆ F(a) ⊆ E. Let

A 5 {[F(a) : F] : a  E}.

Since [E : F] is finite and [F(a) : F] is a positive divisor of [E : F], we get

that A is a finite set of positive integers and hence A has the largest member.
Choose b  E such that [F(b) : F] is the largest in A; that is,

[F(a) : F]  [F(b) : F] for all a  E.

We prove that F(b) 5 E. Suppose, if possible, that F (b)  E . Choose c  E

such that c  F(b). Now, since the number of intermediate fields between
F and E is finite, we get that

{F(cd 1 b) : d  F} is finite.

Since F is infinite, we can choose d1  d2  F such that F(cd1 1 b) 5

F(cd2 1 b). Putz 5 cd1 1 b. Then, z  F(z) and

cd2 1 b  F(cd2 1 b) 5 F(cd1 1 b) 5 F(z)

and therefore c(d1 2 d2)  F(z). Since d1 2 d2  0  F, it follows that c  F(z)

and therefore

b 5 z 2 cd1  F(z) and F(b) ⊆ F(z).

Also, c  F(z) and c  F(b). Therefore, F (b)  F ( z )

which implies that [F(z) : F(b)] > 1. Hence

[F(z) : F] 5 [F(z) : F(b)][F(b) : F] > [F(b) : F]

which is a contradiction to the largest property of [F(b) : F]. Thus, we have

E 5 F(b) and E is a simple extension of F. b
Recall that an element a in an algebraic extension E of a field F is called
separable if its minimal polynomial over F is separable. Now, we have the
following theorem.

Q001-Algebra-111001_CH 15.indd 5 9/16/2011 10:11:57 AM

 15-6   Algebra – Abstract and Modern

Theorem 15.1.3. Let F ⊆ E be fields and a  E be algebraic over F. Then, a

is separable over F if and only if F(a) is a separable extension of F.

Proof: If F(a) is a separable extension of F, then every element of F(a), in

particular, a is separable over F. Conversely suppose that a is separable over
F and b  F(a). We prove that b is separable over F. We have F ⊆ F(b) ⊆
F(a). Let K be an algebraically closed field and s : F → K be an embedding.
Let p(x) be the minimal polynomial of b over F that has m distinct roots. By
Theorem 14.3.3, there are m distinct extensions of s to F(b). Let s1, s2, …,
sm be such extensions of s to F(b). Let q(x) be the minimal polynomial of a
over F(b) and suppose that q(x) has n distinct roots. Then, again by Theorem
14.3.3, for each si, 1  i  m, there are exactly n extensions sij, 1  j  n
to F(a). Then, clearly the embeddings sij, 1  i  m and 1  j  n, are the
only embeddings of F(a) into K which extend s : F → K. Now, let r(x) be the
minimal polynomial of a over F. Then,

1. [F(a) : F] 5 degree of r(x)

5 The number of distinct roots of r(x) (since a is separable
over F)
5 The number of extensions of s to F(a).
Also, since a is separable over F, we get that a is separable over F(b) and
hence we have
2. [F(a) : F(b)] 5 degree of q(x)
5 The number of distinct roots of q(x)
5 The number of extensions of each si to F(a)
5 n.
Also, we have
3. [F(b) : F] 5 degree of p(x) and
4. The number of distinct roots of p(x) 5 The number of extensions of s to
F(b) 5 m.
Now, from (1) to (4), we get that
n ? deg(p(x)) 5 [F(a) : F(b)][F(b) : F]
5 [F(a) : F] 5 mn
and hence m 5 deg(p(x)) 5 The number of distinct roots of p(x).

Therefore, p(x) is a separable polynomial. Thus, b is separable over F.

Therefore, every element of F(a) is separable over F. Thus, F(a) is a sepa-
rable extension of F. b

Q001-Algebra-111001_CH 15.indd 6 9/16/2011 10:11:57 AM

 Galois Theory   15-7

Next, we introduce the notion of a normal extension which plays an

important role in the applications of the Galois theory and discuss certain
properties of normal extensions.

Definition 15.1.4. An algebraic extension E of a field F is said to be a normal

extension of F if each irreducible polynomial f (x) over F, having a root in E,
splits into linear factors over E.
Note that if E is a normal extension of F and a  E, then all the roots of
the minimal polynomial of a over F belong to E.

Example 15.1.3
1. The field C of complex numbers is a normal extension of the field R of
real numbers and [C : R] 5 2.
2. R is not a normal extension of the field Q of rational numbers; for, x3 2
2  Q[x] is irreducible over Q and has a root 3 2 in R, but it does not
split into linear factors in R, since it has complex roots.
Definition 15.1.5. Let {fa(x)}a be a family of nonconstant polynomials
over a field F. An extension E of F is called a splitting field of {fa(x)}a if
every fa(x) splits into linear factors in E(x) and E is generated over F by all the
roots of the polynomials fa(x), a  .
If {f1(x), f2(x), …, fn(x)} is a finite family of polynomials over F, then the
splitting field of the family {f1(x), …, fn(x)} over F is same as the splitting
n
field of the simple polynomial f ( x )   f i ( x ) over F. On the same lines of
i1
the proof of Theorem 14.3.7, we can prove that any two splitting fields of a
family {fa(x)}a of polynomials over F are isomorphic under an isomor-
phism that keeps each element of F fixed. The following gives equivalent
statements for an extension to be normal.

Theorem 15.1.4. Let E be an algebraic extension of a field F such that E is

a subfield of the algebraic closure F of F. Then, the following are equivalent
to each other:
1. E is the splitting field of a family of polynomials in F[x].
2. Every embedding s : E → F for which s(a) 5 a for all a  F maps
E onto E.
3. E is a normal extension of F.

Proof: (1) ⇒ (2): Let {fa(x)}a be a family of polynomials in F[x] and E be

the splitting filed of {fa(x)}a. If a is a root of some fa(x), a  , then, for
any embedding s : E → F such that s(b) 5 b for all b  F, s(a) is a root
of fa(x). Since E is generated by all the roots of all the polynomials fa(x), it

Q001-Algebra-111001_CH 15.indd 7 9/16/2011 10:11:58 AM

 15-8   Algebra – Abstract and Modern

follows that s(d) 5 d for all d  E. Now, by Theorem 14.2.12, s is an auto-

morphism of E and hence s maps E onto E.
(2) ⇒ (3): Let p(x) be an irreducible polynomial over F that has a root a  E.
Let b  F be another root of p(x). We prove that b  E. Since a and b are
roots of the same irreducible polynomial p(x), we get F-isomorphisms.

F [ x]
s1 : → F ( a)
< p( x )>
F [ x]
and s2 : → F ( b)
          p( x )

given by s1(f (x) 1 < p(x)>) 5 f (a) and s2(f (x) 1 <p(x)>) 5 f (b).
Put s  s 2 ο s1 1; F(a) → F(b). Then, s is an isomorphism such that
s(a) 5 b and s(c) 5 c for all c  F. Then, by Theorem 14.3.4, s can be
extended to an embedding s* : E → F . Now, by (2), s* maps E onto E; that
is, s* is an automorphism of E. Therefore,

b 5 s(a) 5 s*(a)  E, since a  E.

Therefore, E contains all the roots of p(x). Thus, E is a normal exten-

sion of F.
(3) ⇒ (1): Suppose that E is a normal extension of F. For each a  E, let
pa(x) be the minimal polynomial of a over F. Since E is a normal extension
of F, pa(x) splits into linear factors in E, for each a  E. Therefore, it is
immediate that E is a splitting field of the family {pa(x)}aE of polynomials
over F. b

Theorem 15.1.5. A finite extension E of F is a normal extension of F if and

only if E is the splitting field of a polynomial over F.

Proof: Let E be a finite extension of F. Then, E 5 F(a1, a2, …, an), where

each ai  E and ai is algebraic over F. For each 1  i  n, let pi(x) be the
minimal polynomial of ai over F. Now, suppose that E is a normal extension
of F. Since ai  E, we get that every root of pi(x) belongs to E. Put

p(x) 5 p1(x) p2(x)  pn(x).

Then, E 5 F(a1, a2, …, an) becomes the splitting field of f (x).

Conversely suppose that E is the splitting field of a polynomial f (x) 
F[x]. If a1, a2, …, an are all the roots of f (x), then E 5 F(a1, a2, …, an). If
s : E → F is an embedding such that s/F 5 Id, then s(ai) is also a root of

Q001-Algebra-111001_CH 15.indd 8 9/16/2011 10:11:59 AM

 Galois Theory   15-9

f (x) and hence s(ai) 5 aj for some j which implies that s(ai)  E for all
1  i  n. Therefore, s(E) ⊆ E and s is an automorphism of E. By Theorem
15.1.4, E is a normal extension of F. b

Worked Exercise 15.1.1. Let F ⊆ E be fields such that [E : F]52. Then prove
that E is a normal extension of F.

Answer: Since [E : F] > 1, F  E and hence we can choose a  E such that

a  F. Then, a is algebraic over F and let p(x) be the minimal polynomial of
a over F. Then, F  F ( a) ⊆ E and hence

[E : F(a)][F(a) : F] 5 [E : F] 5 2.

Also, since F  F ( a), [F(a) : F] > 1 and hence it follows that [F(a) : F] 5 2 and
[E : F(a)] 5 1. Therefore, F(a) 5 E and

2 5 [F(a) : F] 5 degree of p(x).

Since deg(p(x)) 5 2 and since E has a root of p(x), it follows that the other
root of p(x) must also be in E. Thus, E is the splitting field of p(x)  F[x] and
hence E is a normal extension of F.

Worked Exercise 15.1.2. Let F ⊆ E ⊆ K be fields such that K is a finite sepa-

rable extension of E and E is a finite separable extension of F. Then prove
that K is a finite separable extension of F.

Answer: By Theorem 15.1.1, E 5 F(a) and K 5 E(b) for some a  E and

b  K. Let c  F(a, b) such that c  F(a). Then, we have

F ⊂ F(c) ⊂ F(a, c)
          and F ⊂ F(a) ⊂ F(a, c).

Also, F(a) is a finite separable extension of F and c is a separable element

over F(a). We prove that c is separable over F. Let f (x), g(x), p(x) and q(x) be
the minimal polynomials of a over F, c over F(a), c over F and a over F(c),
respectively, and let deg(f (x)) 5 m, deg(g(x)) 5 n, deg(p(x)) 5 s and deg(q(x))
5 t. Let s : F → L be an embedding of F into an algebraically closed field L.
Since a is separable over F, there are exactly m extensions s1, s2, …, sm of
s to F(a). Also, since c is separable over F(a), there are exactly n extensions
si1, si2, …, sin of each si to F(a, c). Now, there are exactly mn extensions of s :
F → L to sij : F(a, c) → L, 1  i  m, 1  j  n. Similarly, by considering
extensions of s : F → L to F(a, c) via F(c), we get that there are exactly st

Q001-Algebra-111001_CH 15.indd 9 9/16/2011 10:11:59 AM

 15-10   Algebra – Abstract and Modern

extensions of s to F(a, c). Therefore, mn 5 st. Now, suppose, if possible, that

c is not separable over F. Then, the number of extensions of s to F(c) is less
than s (5degree of the minimal polynomial of c over F), which implies that
the number of extensions of s to F(a, c) is less than st 5 mn, a contradiction.
Thus, c is separable over F. Thus, K is a finite separable extension of F.

EXERCISE 15(a)

1. Find a  3  5 such that Q( 3, 5)  Q( a).

2. Ifz is the complex number such that z  1 and z3 5 1, then find a such that
Q( 2, z )  Q( a).
3. Prove that any extension of Q is separable.

4. Prove that any finite extension of a finite field is separable.

5. Let K be a field of characteristic p  0. Then prove that K is perfect if and only

if Kp 5 K.

6. Prove that Q(21/4) is not a normal extension of Q.

7. If E is an extension of a field F, then prove that the set of all elements in E which
are separable over F forms a subfield of E containing F.

8. Let F ⊆ K ⊆ L be fields such that L is a finite normal extension of F. Then prove

that L is a finite normal extension of K.

9. Prove that a field F of characteristic p  0 is perfect if and only if the mapping

a  ap is an automorphism of F.

15.2 AUTOMORPHISM GROUPS AND FIXED FIELDS

Recall that any finite separable extension E of a field F is simple and hence
E 5 F(a) for some a  E. Throughout this section, we confine ourselves to
finite separable extensions and their groups of automorphisms. First recall
that, for any field E, the set Aut(E) of all automorphisms of E forms a group
under the composition of mappings.

Definition 15.2.1. Let F be a field and E be an extension of F. An automor-

phism s of E is called an F-automorphism if s fixes all elements of F; that is,
s (a) 5 a for all a  F. Let

E
G    {s  Aut( E ) : s ( a)  a for all a  F}.
 F 

Q001-Algebra-111001_CH 15.indd 10 9/16/2011 10:12:00 AM

 Galois Theory   15-11

Then, G ( E / F ) becomes a group under the composition of mappings and is

called the group of F-automorphisms of E. Note that G ( E / F ) is a subgroup
of Aut(E).

Theorem 15.2.1. Let E be a finite separable extension of a field F. Then,

E
G    [ E : F ]
 F 

Proof: Recall that any finite separable extension E of F is a simple extension

of F. Therefore, E 5 F(a) for some a  E. Let p(x) be the minimal polyno-
mial of a over F and deg(p(x)) 5 n. Then,

[E : F] 5 [F(a) : F] 5 deg(p(x)) 5 n.

By Theorem 14.3.3, we get that

E
G    n  [ E : F ]. b
 F 


Definition 15.2.2. Let E be any field and H be a subgroup of Aut(E). Then,

the set

EH 5 {a  E : s (a) 5 a for all s  H}

is called the fixed field of H.

It can be easily verified that EH is a subfield of E for any subgroup H of
Aut(E). If E is a field extension of F and H is a subgroup of G ( E / F ), then

F ⊆ EH ⊆ E.

As a simple example, consider H  G (C / R). Then, CH 5 R.

The following result is an important tool in proving the main result of this
section.

Theorem 15.2.2 (Dedikind Theorem). Let F and E be fields and s1, s2, …, sn
be distinct embeddings of F into E. Let a1, a2, …, an  E such that

a1s1(a) 1 a2s2(a) 1  1 ansn(a) 5 0 for all a  F.

Then, a1 5 a2 5  5 an 5 0 (This can also be expressed by saying that dis-

tinct embeddings of F into E are linearly independent over E.)

Q001-Algebra-111001_CH 15.indd 11 9/16/2011 10:12:01 AM

 15-12   Algebra – Abstract and Modern

Proof: Suppose, if possible, that there exist a1, a2, …, an  E, not all zero,
such that

a1s1(a) 1 a2s2(a) 1  1 ansn(a) 5 0 for all a  E.

Then, we can find such a relation having as few nonzero coefficients as pos-
sible. On renumbering, we can assume that this relation is

b1s1(a) 1 b2s2(a) 1  1 bmsm(a) 5 0 for all a  E. (1)

If m 5 1, then b1s1(a) 5 0 for all a  E and, in particular, b1 5 b11 5 b1s1(1)

5 0. Therefore, we can assume that m > 1. Now, s1  sm and hence there
exists an element c  E such that s1(c)  sm(c). The Equation (1) holds for
all a  E and, in particular, for ca for all a  E. Therefore,

b1s1(ca) 1 b2s2(ca) 1  1 bmsm(ca) 5 0

    and hence b1s1(c)s1(a) 1  1 bmsm(c)sm(a) 5 0 (2)

for any a  E. Multiplying (1) by s1(c) and subtracting the result from (2),
we get that

b2(s2(c) 2 s1(c))s2(a) 1  1 bm(sm(c) 2 s1(c))sm(a) 5 0

for all a  E. This is a contradiction to the choice of Equation (1), since

bm(sm(c) 2 s1(c))  0. b
Now, we prove the following which is the main theorem in this section.

Theorem 15.2.3. Let E be any field and H be a finite subgroup of Aut(E).

Then, E is a finite extension of EH and

[E : EH] 5 |H|.

Proof: We have EH 5 {a  E : s(a) 5 a for all s  H}. Then, clearly EH is a

subfield of E and hence E is an extension of EH. Let |H| 5 n <  and H 5 {e 5
g1, g2, …, gn}. Suppose, if possible, that n < [E : EH]. Then, there exist elements
a1, a2, …, an11 in E such that {a1, a2, …, an11} is linearly independent over EH.
Consider the system of n homogeneous linear equations

gj(a1)x1 1 gj(a2)x2 1  1 gj(an11)xn11 5 0, 1 j  n

Q001-Algebra-111001_CH 15.indd 12 9/16/2011 10:12:01 AM

 Galois Theory   15-13

in n 1 1 unknowns x1, x2, …, xn11. This system should have a nontrivial

solution. Therefore, there exists elements y1, y2, …, yn11 in E, not all zero,
such that

gj(a1)y1 1  1 gj(an11)yn115 0 (1)

for all 1  j  n. We can choose these y1, …, yn11 such that as few of them as
possible are nonzero and renumber them such that

yi  0 for i 5 1, …, r and yi 5 0 for r < i  n 1 1.

Then, Equation (2) becomes

gj(a1)y1 1  1 gj(ar)yr 5 0 (2)

for all 1  j  n. Now, let g  H and operate on (2) with g. Then, we get the
system of equations

g(gj(a1))g(y1) 1  1 g(gj(ar))g(yr) 5 0 (3)

Since H 5 {gg1, gg2, …, ggn} 5 {g1, g2, …, gn}, (3) is equivalent to the sys-
tem of equations

gj(a1)g(y1) 1  1 gj(ar)g(yr) 5 0 (4)

By multiplying (2) by g(y1) and (4) by y1 and by subtracting, we get

gj(a2)(y2g(y1) 2 g(y2)y1) 1  1 gj(ar)(yrg(y1) 2 g(yr)y1) 5 0 (5)

for all 1  j  n. This is a system of equations like (2), but with fewer terms,
which becomes a contradiction to our assumption, unless

yjg(y1) 2 g(yi)y1 5 0 for all 2  i  r.

If this happens, then g ( yi y11 )  yi y11 for all g  H and hence yi y11  EH
for all 2  i  r. Therefore, there exist z1, z2, …, zr  EH such that y1zi 5 yi
−1
for 1  i  r (take z1 5 1 and zi = yi y1 for 2  i  r). Then, Equation (2)
with j 5 1 becomes

g1(a1)y1z1 1  1 g1(ar)y1zr 5 0

Q001-Algebra-111001_CH 15.indd 13 9/16/2011 10:12:01 AM

 15-14   Algebra – Abstract and Modern

which implies that g1(a1)z1 1  1 g1(ar)zr 5 0 (since y1  0). Now, since zi

 EH, we get that

g1(a1z1 1  1 arzr) 5 0

and hence a1z1 1  1 arzr 5 0, since g1 is an embedding. Now, since {a1, …, ar}
are linearly independent over EH, it follows that

z1 5 z2 5  = zr 5 0

and hence y1 5 y2 5  5 yr 5 0, which is a contradiction.

Thus, we must have [E : EH]  n <  and therefore E is a finite extension of
EH and [E : EH]  n.
On the other hand, suppose, if possible, that [E : EH] < n. Let [E : EH] 5 m
and {a1, a2, …, am} be a basis of E over EH. Consider the system of m homo-
geneous linear equations

g1(aj)x1 1  1 gn(aj)xn 5 0, 1  j  m,

in n unknowns x1, …, xn. Since m < n, this system has a nontrivial solution.
Therefore, there exist y1, y2, …, yn  E, not all zero, such that

g1(aj)y1 1  1 gn(aj)yn 5 0 (6)

for all 1  j  m. Since {a1, a2, …, am} is a basis of E over EH, any element
a  E can be uniquely written as a 5 s1a1 1  1 smam with si  EH, and
hence

n n  m 
∑ g (a) y  ∑ g ∑ s a  y
i1
i i
i1
i
j1
j j i

m n
 ∑∑ s j gi ( a j ) yi
j1 i1

m  n 
 ∑ s j ∑ gi ( a j ) yi 
j1

 i1 
 0 (by (6))

Therefore, we have y1, y2, …, yn  E, not all zero, and

y1g1(a) 1 y2g2(a) 1  1 yngn(a) 5 0 for all a  E

Q001-Algebra-111001_CH 15.indd 14 9/16/2011 10:12:02 AM

 Galois Theory   15-15

which is a contradiction to the fact that g1, g2, …, gn are distinct embeddings
of E into E (refer Dedikind’s Theorem 15.2.2). Therefore, m < n is impossible
so that [E : EH]  n. Thus,

[E : EH] 5 n 5 |H|. b

Theorem 15.2.4. Let E be a finite separable extension of a field F and let H

be a subgroup of G ( E / F ). Then,

 E   E 
G    H and [E : EH] 5 G  
 E H   E H 

Proof: If s  H, then s(a) 5 a for all a  EH and hence s  G ( E / EH ).

Therefore, H is a subgroup of G ( E / EH ). By Theorem 15.2.3, |H| 5 [E : EH].
Also, by Theorem 15.2.1,

 E 
| H |  G    [ E : E H ]  | H |
 E H 

and hence G ( E / EH )  H , since H is a subgroup of G ( E / EH ) and

| H |  | G ( E / EH ) | . Also, we have

 E 
[ E : E H ]  G   . b
 E H 


Theorem 15.2.5. A finite separable extension E of a field F is a normal exten-

sion of F if and only if the fixed field of G ( E / F ) is F.

Proof: Let E be a finite separable extension of a field F, Then, by Theorem

15.1.1, E is a simple extension of F and hence E 5 F(a) for some a  E.
Let p(x) be the minimal polynomial of a over F and deg(p(x)) 5 n. Then,
we have

[E : F] 5 [F(a) : F] 5 n.

Let E0 be the fixed field of G ( E / F ); that is,

E0 5 {s  E : s(s) 5 s for all s(s) 5 s for all s  G ( E / F )},

Q001-Algebra-111001_CH 15.indd 15 9/16/2011 10:12:03 AM

 15-16   Algebra – Abstract and Modern

Then, F ⊆ E0 ⊆ E and, by Theorem 15.2.3, we have

E
[ E : E0 ]  G   .
 F 

Now, we prove that E is a normal extension of F if and only if E0 5 F.

First, suppose that E0 5 F. Then, G ( E / F )  [ E : E0 ]  [ E : F ]  n. Let
G ( E / F )  {e  s1 , s 2 , …, s n }, where e is the identity automorphism,
­consider the polynomial

n
f ( x )  ∏ ( x  s i ( a)).
i1

Each s i  G ( E / F ) induces a natural homomorphism s *i : E[ x ] → E[ x ]

given by

s *i ( a0  a1 x  ar x r )  s i ( a0 ) s i ( a1 ) x s i ( ar ) x r .
n
Now, s *i f ( x )   (x s i (s j ( a))).
j =1
Since sis1, sis2, …, sisn are distinct members of G ( E / F ) and |G(E/F)| 5 n,
we get that G ( E / F ) = {s i s1 , s i s 2 , …, s i s n } and hence s *i ( f ( x ))  f ( x )
for all 1 i  n. By expanding f (x), we have

f (x) 5 xn 2 c1xn21 1 c2xn22 1  1 (21)ncn, ci  E

and, from s *i f ( x )  f ( x ), we have

si(cj) 5 cj for all 1 i, j  n.

Therefore, cj  E0, the fixed field of G ( E / F ) and, by hypothesis cj  F for

all 1  j  n. This implies that f (x)  F[x]. Further, all the roots of f (x) lie
in E 5 F(a) and a is one of the roots of f (x). Thus, E is the splitting field of
f (x)  F[x] and hence E is a normal extension of F.
Conversely, suppose that E is a normal extension of F. By Theorem 14.3.3,
the number of extensions of the inclusion map F → F to the embedding
F ( a) → F is equal to the number of distinct roots of p(x). Since E is a sepa-
rable extension of F and a  E, a is a separable element of E and hence its
minimal polynomial p(x) over F has distinct roots. Therefore, the number
of distinct roots of p(x) is n. Also, since E 5 F(a) is a normal extension of
F, any embedding s : F ( a) → F should map F(a) onto F(a). Further, any

Q001-Algebra-111001_CH 15.indd 16 9/16/2011 10:12:05 AM

 Galois Theory   15-17

member of G ( E / F ) is an extension of the inclusion map F → F and hence

we have

E
G    The number of distinct roots of p( x )  n.
 F 

Therefore, [E : F] 5 n 5 | G ( E / F ) | 5 [E : E0]. Since F ⊆ E0 ⊆ E, it

follows that [E0 : F] 5 1 and hence E0 5 F. Thus, F is the fixed field of
G ( E / F ). b

Corollary 15.2.1. A finite separable extension E of a field F is a normal

extension of F if and only if [ E : F ]  | G ( E / F ) | .

Proof: Let E be a finite separable extension of a field F. Suppose that E is

a normal extension of F. Then, by Theorem 15.2.5, F is the fixed field of
G ( E / F ). Then, by Theorem 15.2.4,

E
[ E : F ]  G   .
 F 

Conversely suppose that [ E : F ] = |G ( E / F ) | . Let E0 be the fixed field of

G ( E / F ). Then, [ E : E0 ]  | G ( E / F ) | and hence [E : F] 5 [E : E0]. Since
F ⊆ E0 ⊆ E, it follows that [E0 : F] 5 1 so that F 5 E0. Again by Theorem
15.2.5, E is a normal extension of F. b

Worked Exercise 15.2.1. Let f (x)  F[x] has r distinct roots in its splitting
field E over F. Then prove that G ( E / F ) is isomorphic to a subgroup of the
symmetric group Sr of degree r.

Answer: Let a1, a2, …, ar be all the distinct roots of f (x) in its splitting field
E over F. For any s  G ( E / F ), s(ai) is again a root of f (x) in E. Also,
s(ai)  s(aj) for ai  aj. Thus, s(a1), s(a2), …, s(ar) is a permutation of a1,
a2, …, ar and let us denote this permutation by s. Therefore, s  Sr for
each s  G ( E / F ) . Define  : G ( E / F ) → Sr by (s) 5 s. For any s and
  G ( E / F ),

(s o )(ai) 5 (s o )(ai) 5 s((ai)) 5 ((s) o ())(ai)

for all 1  i  r and hence (s o ) 5 (s) o (). Therefore,  is a homo-

morphism of groups. Also, for any s  G ( E / F ),

Q001-Algebra-111001_CH 15.indd 17 9/16/2011 10:12:07 AM

 15-18   Algebra – Abstract and Modern

(s) 5 Id ⇒ s(ai) 5 ai for all 1  i  r

⇒ s 5 Id, since E 5 F(a1, …, ar)

Thus, s : G ( E / F ) → Sr is a monomorphism of groups and s is an isomor-

phism of G ( E / F ) onto s (G ( E / F )), which is a subgroup of the symmetric
group Sr.

Worked Exercise 15.2.2. Let E  Q ( v, 3 2), where v3 5 1 and v  1 (that

is, v is a cube root of unity in C) and let H be the subgroup of G ( E / Q) given
by H5{Id, s} where s : E → E is defined by s(a) 5 a for all a  Q, s(v) 5
v2 and s ( 3 2)  3 2 v 2 . Then, find the fixed field EH.

Answer: For simplicity, let 3 2  c. Then, c is a real number such that c3 5

2. We are given that E 5 Q(v, c) and H 5 {Id, s}, where s is defined by
(s / Q)  Id, s(v) 5 v2 and s(c) 5 cv2. First note that Q ⊂ Q(c) ⊂ Q(v, c),
{1, c, c2} is a basis of Q(c) over Q and {1, v} is a basis of (v, c) over Q(c)
therefore the basis of E over Q is {1, c, c2, v, vc, vc2}. Consider any a  E.

Then, a 5 r0 1 r1c 1 r2c2 1 r3v 1 r4cv 1 r5c2v, with ri  Q, and

s(a) 5 r0 1 r1cv2 1 r2c2v 1 r3v2 1 r4cv 1 r5c2
5 r0 1 r1c(21 2 v) 1 r2c2v 1 r3(21 2v) 1 r4cv 1 r5c2
(since 1 1 v 1 v250)
5 (r0 2 r3) 1 (2 r1)c 1 r5c2 1 (2 r3)v 1 (2 r1 1 r4)cv 1
r2c2v.

 (a) 5 a ⇒ r0 2 r3 5 r0, 2 r15 r1, r5 5 r2, 2 r3 5 r3, 2 r1 1 r4

Therefore, s
5 r4 and r2 5 r5
⇒ r3 5 0, r1 5 0 and r2 5 r5
⇒ a 5 r0 1 r2c2 1 r4cv 1 r2c2v
⇒ a 5 r0 1 r4cv 1 r2c2(1 1 v)
⇒ a 5 r0 1 r4cv 2 r2(cv)2  Q(cv).

Therefore, EH  Q(cv). On the other hand, if a  Q(cv), then clearly

s(a) 5 a and hence a  EH. Thus, the fixed field EH of H is equal to
Q(c v)  Q( 3 2 v). Note that (cv) 2  c 2 v2  3 4 v2 and (c2v2)2 5 2cv and
therefore EH  Q(cv)  Q(c 2 v 2 ) = Q( 3 4 v 2 ).

EXERCISE 15(b)

1. Let f (x) 5 x4 2 2  Q[x] and E be the splitting field of f (x) over Q. Prove that
G ( E / Q) is isomorphic to the group of symmetries of a square.

Q001-Algebra-111001_CH 15.indd 18 9/16/2011 10:12:08 AM

 Galois Theory   15-19

2. Prove that G (C / R) is a cyclic group of order 2.

3. Let F be a field of characteristic 2 and x2 2 a be an irreducible polynomial over

F. Prove that G ( E / F ) is a group of order 2, where E is the splitting field of
x2 2 a over F.

4. Let v  1 be a cube root of unity and E  Q( v, 3 2) . Let s be the automorphism

of E defined by s(r) 5 r for all r  Q, s(v) 5 v2 and s ( 3 2)  3 2 v. Then
prove that {Id, s} is a subgroup of G ( E / Q) and find its fixed field.

5. Let E be the splitting field of x4 2 x2 1 1 over the field of rationals Q. Then,

determine the group G ( E / Q).

6. Let a  1 and a5 5 1. Then prove that Q(a) is a normal extension of Q and that
G (Q( a) Q) is isomorphic to Z4, the group of integers modulo 4.

15.3 FUNDAMENTAL THEOREM OF GALOIS THEORY

In this section, we prove the much awaited main theorem of the present chapter,
which is popularly known as the fundamental theorem of Galois theory. Before
going to prove this theorem, let us introduce the following terminology.

Definition 15.3.1. Any finite normal and separable extension of a field F is

called a Galois extension of F.

Definition 15.3.2. Let E be the splitting field of a polynomial f (x)  F[x]

over F. Then, group G ( E / F ) of F-automorphisms of E is called the Galois
group of f (x) over F.

Theorem 15.3.1 (Fundamental Theorem of Galois Theory). Let E be a Galois

extension of F,  be the class of all subfields of E containing F and  be the
class of all subgroups of G ( E / F ). Then, the following holds for any K  .
1. G ( E / K ) is a subgroup of G ( E / F ) and K  G ( E / K ) is a bijection of
 onto  such that, for any K1 and K2  .
K1 ⊆ K 2 ⇔ G ( E / K 2 ) is a subgroup of G (E K1 ).
2. K is a normal extension of F if and only if G (E K ) is a normal sub-
group of G (E F ) .
3. If K is a normal extension of F, then
E
G  
 F  K 
 G  .
 E   F 
G  
 K 

Q001-Algebra-111001_CH 15.indd 19 9/16/2011 10:12:10 AM

 15-20   Algebra – Abstract and Modern

Proof:
1. Define  :  →  and u :  →  by

( K )  G (E K ) and u (H) 5 EH, the fixed field of H

for any K   and H  . Clearly, G (E K ) is a subgroup of G (E F )
for any field K such that F ⊆ K ⊆ E, and EH is an intermediate field
between F and E for any subgroup H of G (E F ) . Therefore,  and u
are well-defined mappings. We prove that these mappings  and u are
inverses to each other. If K  , then F ⊆ K ⊆ E and E is a normal exten-
sion of K and hence, by Theorem 15.2.5, K is the fixed field of G (E K ) ;
that is,
  E 
(u  )( K ) uG    K .
  K 

This is true for any K   and hence u ο  is the identity map of . On

the other hand, let H  . Then, H is a subgroup of G (E F ) and, by
Theorem 15.2.4,
 E 
H  G    (  u)( H ).
 E H 

This implies that  ο u is the identity map of . Therefore,  is a bijec-

tion and 21 5 u. Also, for any K1 and K2  ,

E E
K1 ⊆ K 2 ⇒ G   ⊆ G   ⇒( K 2 ) ⊆ ( K1 )
 K 2   K1 

E E
and G   ⊆ G   ⇒ E  E  ⊆ E  E  ⇒ K1 ⊆ K 2.
 K 2   K1  G  
 
 K1 
G  
 
 K2 

2. First we prove that, for any F ⊆ K ⊆ E, K is a normal extension of F

if and only if s(K) 5 K for any s ∈ G (E F ) . Let K  . Then, F ⊆
K ⊆ E. Suppose that K is a normal extension of F and s ∈ G (E F ) .
Then, s K : K → E is an embedding. Since E ⊆ F , s K is an embed-
ding of K into F . Then, (s K )( K )  K and hence s(K) 5 K. Con-
versely, suppose that s(K) 5 K for any s ∈ G (E F ) . Let s * : K → F
be an embedding such that (s * F ) Id. Then, s* can be extended to
 : E → F . Since E is a normal extension of F, we have λ(E) 5 E and
hence  ∈ G (E F ) and λ(K) 5 K. Since  is an extension of s* to E,

Q001-Algebra-111001_CH 15.indd 20 9/16/2011 10:12:12 AM

 Galois Theory   15-21

it follows that s*(K) 5 K. Thus, K is a normal extension of F. Now,

we prove the assertion in (2). Suppose that K is a normal extension of
F. Clearly, G (E K ) is a subgroup of G (E F ) . Let s ∈ G (E F ) and
t ∈ G (E K ) . By the above observation, s(K) 5 K. For any a  K, s(a)
 K and hence t(s(a)) 5 s(a), so that (s21 ο t ο s)(a) 5 a. This implies
that s1  t  s ∈ G (E K ) and hence G (E K ) is a normal subgroup of
G (E F ) .
  Conversely suppose that G (E K ) is a normal subgroup of G (E F ).
By the above observation, it is enough to prove that s(K) 5 K for all
s ∈ G (E F ) . Let s ∈ G (E F ) and a  K. For any t ∈ G (E K ) , we
have s1ts ∈ G (E K ) and hence
(s21ts)(a) 5 a; that is, t(s(a)) 5 s(a).
This implies that s ( a) ∈ EG(E K )  K . Therefore, we have proved that
s(a)  K for all a  K and hence s(K) 5 K. Thus, K is a normal exten-
sion of F.
3. Let K be a normal extension of F. For any s ∈ G (E F ) , let s * (s K ).
Then, by the above observation s * ∈ G (K F ) (since s(K) 5 K). Now,
define a mapping f : G (E F ) → G (K F ) by
s
f (s )  s * 
K

It can be easily verified that F is a homomorphism of groups. Also,

 E s 
 E
ker f  s  G   :  Id
  G  .
  F  K 

  K 

By the fundamental theorem of homomorphism of groups, we have

E
G  
 F  K 
 Im f ⊆ G  .
E 
F
G  
 K 

Also, we have [E : F] 5 [E : K][K : F], E is a normal extension of F, E is

a normal extension of K and K is a normal extension of F and therefore

E  E  K 
G    [ E : F ]  [ E : K ][ K : F ]  G   G  
 F   K   F 

Q001-Algebra-111001_CH 15.indd 21 9/16/2011 10:12:15 AM

 15-22   Algebra – Abstract and Modern

Now,

E
G  
 F  K 
Im f   G  
 E   F 
G  
 K 

and hence Im f  G (K F ) . Thus, we have

E
G  
 F  K 
 G  .
 E   F  b
G  
 K 


Worked Exercise 15.3.1. Prove that the Galois group of x3 2 2  Q[x] is the
group of symmetries of the triangle.

Answer: Let E be the splitting field of x3 2 2 over Q. We have seen earlier

that E  Q (21/ 3 ,v) , where v is the root of the irreducible polynomial x2 1 x
1 1 in Q(21/ 3 ) and hence [E : Q] 5 6. Notice that

x 3  2  ( x  21/ 3 )( x v21/ 3 )( x v2 21/ 3 ).

Since E is a normal extension of Q and [E : Q] 5 6, there are six automor-

phisms of E and these are determined by the manner in which they transform
the roots of x3 2 2. The root 21/ 3 can have only three images, namely 21/ 3 ,
21/ 3 v and 21/ 3 v2 and the root v can have only two images, namely v and
v2. There are six possible combinations, since there are exactly six automor-
phisms. The Galois group G (E Q) of x3 2 2 over Q is given by following
table, where

E
G    {Id, s , s 2 , t , st , s 2 t}
 Q 

Id s s2 t st s2t
21/ 3 21/ 3 v21/ 3 v 2 21/ 3 21/ 3 v21/ 3 v 2 21/ 3
v v v v v2 v2 v2

Q001-Algebra-111001_CH 15.indd 22 9/16/2011 10:12:16 AM

 Galois Theory   15-23

This group G (E Q) is isomorphic to the group of symmetries of the

triangle.

Worked Exercise 15.3.2. Prove that the Galois group of x4 2 2  Q[x] is the
group of symmetries of a square (that is, the octic group).

Answer: First observe that x4 2 2 is irreducible over Q and we can factorise

x4 2 2 by x4 2 2 5 (x 2 21/4)(x 1 21/4)(x 1 i21/4)(x 2 i21/4), where is the square
root of 21 in the field C of complex numbers. Therefore, E 5 Q(21/4, i) is the
splitting field of x4 2 2 over Q. Also, [E : Q] 5 8 since [E : Q(21/4)] 5 2 and
[Q(21/4) : Q] 5 4.
Since E is a normal separable extension of Q, it follows that

E
G    [ E : Q]  8.
 Q 

Therefore, the Galois group G (E Q) is a group of order 8. Note that {1, 21/4,
21/2, 23/4, i, i21/4, i21/2, i23/4} is a basis of E over Q. Now, if b  E, then

b 5 a0 1 a121/4 1 a221/2 1 a323/4 1 a4i 1 a5i21/4 1 a6i21/2 1 a723/4i

for some ai  Q, 0  i  7 and hence, for any s  G (E Q),

s(b) 5 a0 1 a1s(21/4) 1 a2s(21/2) 1 a3s(23/4) 1 a4s(i)

1 a5s(i 21/4) 1 a6s(i 21/2) 1 a7s(i 23/4)
5 a0 1 a1s(21/4) 1 a2s(21/4)2 1 a3s(21/4)3 1 a4s(i)
1 a5s(i)s(21/4) 1 a6s(i)s(21/4)2 1 a7s(i)s(21/4)3.

Therefore, s is completely determined by s(21/4) and s(i). Since s(21/4) 5

21/4 or 2 21/4 or i21/4 or 2 i21/4; since s(i) 5 i or 2 i and since | G ( E / Q) |  8,
it follows that G (E Q) {s1 , s 2 , s 3 , s 4 , s 5 , s 6 , s 7 , s 8} where si’s are
given by the following table

s1 s2 s3 s4 s5 s6 s7 s8
i i i i i i i i i
1/4 1/4 1/4 1/4
2 2 i2 2 i 21/4 21/4 i21/4
21/4
i 21/4

For convenience, let us write

a1 = 21/4, a2 = i21/4, a3 =221/4 and a4 = 2i21/4.

Q001-Algebra-111001_CH 15.indd 23 9/16/2011 10:12:17 AM

 15-24   Algebra – Abstract and Modern

As demonstrated in the diagram below, the elements of G (E Q) permute the

roots a1, a2, a3 and a4 of x4 2 2.
d2
l2 l1
a2

a3 a1 d1

a4

s1 : rotation by 0°
s2 : rotation by 90°
s3 : rotation by 180°
s4 : rotation by 270°
s5 : rotation by d1
s6 : rotation by l1
s7 : rotation by d2
s8 : rotation by l2
These are precisely all the symmetries of the square whose vertices are a1, a2,
a3 and a4. Thus, the Galois group G (E Q) of the polynomial x42 2 over Q
is isomorphic with the group of symmetries of a square and hence G (E Q)
is the octic group.

Worked Exercise 15.3.3. Prove that the Galois group of x5 2 1  Q[x] is a

cyclic group of order 4.

Answer: Let f (x) 5 x5 2 1  Q[x]. Then,

f (x)5(x 2 1)(x4 1 x3 1 x2 1 x 1 1)

If v is a root of 1 1 x 1 x2 1 x3 1 x4, then 1, v, v2, v3, v4 are all the roots of

f (x) and hence Q(v) is the splitting field of f (x) over Q and [Q(v) : Q] 5 4 and
hence G (Q( v) Q)  4. If s is an automorphism in G (Q( v) Q) such that

s(1) 5 1 and s(v) 5 v2

then s2(v) 5 v4, s3(v)5 v3 and s4(v) 5 v and hence s4 5 Id and s

generates G (Q( v) Q). This implies that G (( v) Q) is a cyclic group of
order 4.

Q001-Algebra-111001_CH 15.indd 24 9/16/2011 10:12:18 AM

 Galois Theory   15-25

EXERCISE 15(c)

1. In each of the following, find all the subgroups of the Galois group of f (x) and
the corresponding fixed fields.
(i) f (x) 5 x3 2 2  Q[x]
(ii) f (x) 5 x4 1 1  Q[x]
(iii) f (x) 5 x4 2 2  Q[x]
(iv) f (x) 5 x2 2 a  F[x], where F is a field of characteristic  2.

2. Prove that the Galois group of x4 1 1  Q[x] is the Klein four-group and is
isomorphic to Z23 Z2.

3. If E  Q( 3, 5), then find the Galois group G (E Q).

4. Let a be a real number such that Q(a) is a normal extension of Q for which [Q(a)
: Q] 5 2m, where m ≥ 0. Prove that there are fields E0 5 Q ⊂ E1 ⊂ E 2 ⊂ … ⊆
Em 5 Q(a) such that [Ei : Ei21] 5 2 for each 1  i  m.

5. If K is the splitting field of x4 2 3x2 1 4 over Q, then find the Galois group
G (K Q).

6. Let a  cos (2 3) i sin (2 3). Then, find the Galois group G (( a) Q) and
all its subgroups and the corresponding fixed fields.

Q001-Algebra-111001_CH 15.indd 25 9/16/2011 10:12:19 AM

 This page is intentionally left blank.

Q001-Algebra-111001_CH 15.indd 26 9/16/2011 10:12:19 AM

 16
Selected Applications of
Galois Theory
16.1 Fundamental Theorem of Algebra
16.2 Cyclic Extensions
16.3 Solvable Groups
16.4 Polynomials Solvable by Radicals
16.5 Constructions by Ruler and Compass

In this chapter, we discuss certain important applications of Galois theory

to classical problems. The first is the fundamental theorem of algebra which
states that any polynomial over the field of complex numbers C has all the
roots in C which is equivalent to saying that any polynomial over C can be
factored completely into linear factors over C. We prove this by using various
techniques of Galois theory. Also we discuss problems of finding solutions of
polynomial equations (that is, finding roots of polynomials) by radicals; that
is, expressing the roots of a polynomial in terms of the coefficients using the
field operations and the operations of taking square roots, cube roots and so on.
We are familiar with the fact that a quadratic polynomial f (x) 5 ax2 1 bx 1
c (a  0) with real coefficients has b b24ac as its roots in C. During
2a
the 17th century, similar formulae were found for cubic and biquadratic equa-
tions. Later, Abel proved that such formulae were not possible for polynomial
of degrees  5. But the general problems of finding a way of deciding whether
a given polynomial could be solvable by radicals was not completed by Abel.
Galois and Liouville gave necessary and sufficient conditions for the solvabil-
ity by radicals for polynomials of degrees  5 which linked with the group
properties of the symmetric group Sn. Finally, we discuss certain impossibilities

Q001-Algebra-111001_CH 16.indd 1 9/16/2011 10:11:12 AM

 16-2   Algebra – Abstract and Modern

of geometric constructions by using ruler and compass only. For example, we

prove that an angle cannot be trisected by using ruler and compass only.

16.1 FUNDAMENTAL THEOREM OF ALGEBRA

It is well known from elementary analysis that the intermediate value theo-
rem gives us a root of a polynomial f (x) over R if there are real numbers a
and b such that f (a) , 0 , f (b). In this section, we prove that any noncon-
stant polynomial over the field C of complex numbers has a root in C. This
result is known as the fundamental theorem of algebra. Even though there
are several proofs of this fundamental result, most of them use the tech-
niques of topology or real analysis or complex analysis. The proof we are
offering here is more elegant and use the techniques of Galois theory. Before
taking up the proof of the main theorem, let us have a brief p­ reparation.

Theorem 16.1.1. The field C has no extension of degree 2.

Proof: Let K be a field extension of C such that [K : C] 5 2. Then, there

exists a  K such that K 5 C(a). Then, p(x) be the minimal polynomial of a
over C. Then, degree of p(x) must be two. Let

p(x) 5 a 1 2bx 1 x2 where a and b  C.

Now,

( )( )
p(x )  x  b  b 2  a x  b  b 2  a ,

where b  b 2  a and b  b 2  a  C. This is a contradiction, since p(x)

is irreducible over C. b

Theorem 16.1.2. Let f (x)  R[x] be of odd degree. Then, f (x) has a real root.

Proof: Without loss of generality, we can assume that f (x) is a monic poly-
nomial. Suppose

f (x) 5 a0 1 a1x 1  1 an21xn21 1 xn, n is odd,

where ai  R. Put s 5 1 1 |a0| 1 |a1| 1  1 |an21|. Then, |ai| # s 2 1 for all

0 # i # n 2 1. Therefore,

|a0 1 a1s 1  1 an21sn21| # (s 2 1)(1 1 s 1  1 sn21)

5 sn21 , sn

Q001-Algebra-111001_CH 16.indd 2 9/16/2011 10:11:13 AM

 Selected Applications of Galois Theory   16-3

and hence f (s) . 0. Also

f (2s) 5 a0 2 a1s 1 a2s2 1 … 1 (21)n21an21sn21 1 (2s)n

5 a0 2 a1s 1 a2s2 2  2 sn, since n is odd
# sn 2 1 2 sn 5 21 , 0.

Therefore, f (2s) , 0 , f (s) and s  R. By the intermediate value ­theorem

in analysis, there exists a real number a such f (a) 5 0. Thus, a is a root of
f (x) in R. b
For any f (x) 5 a0 1 a1x 1  1 anxn  C[x], let us write
f (x ) a0  a1 x  an x n , where ai is the complex conjugate of ai. It is
a straight forward verification to prove that f (x )f (x )  R [ x ] for any f (x) 
2
C[x]. Recall that aa  a ∈ R for any a  C.

Theorem 16.1.3 (Fundamental Theorem of Algebra). Every nonconstant

polynomial f (x) over C completely factors into linear factors in C[x]; that is,
any nonconstant polynomial in C[x] is a product of polynomials of degree
one in C[x].

Proof: Let f (x) 5 a0 1 a1x 1  1 anxn, n . 0 and an  0 be a polynomial

in C[x]. Put

g (x )  (x 2 1) f (x )f (x )
 ( x 2 1)( a0  a1 x  an x n )( a0  a1 x  an x n ).

Then, g(x)  R[x]. Let E be the splitting field of g(x) over R. Then, R ⊆ C
⊆ E, since i is a root of g(x). We prove that E 5 C. First observe that [C : R]
5 2 and is a divisor of [E : R]. Therefore, [E : R] is an even positive integer.
Suppose that

[E : R] 5 2mq, where m and q  Z1 and q is odd.

Let G be the Galois group G(E/R). Since E is a normal extension of R, it

follows that

|G| 5 |G(E/R)| 5 [E : R] 5 2mq.

By the Sylow Theorem I in group theory, there exists a subgroup H of G such

that |H| 5 2m (H is a 2-Sylow subgroup of G). Let EH be the fixed field of H.
Then, R ⊆ EH ⊆ E and

2mq 5 [E : R] 5 [E : EH][EH : R] 5 |H|[EH : R] 5 2m[EH : R]

Q001-Algebra-111001_CH 16.indd 3 9/16/2011 10:11:13 AM

 16-4   Algebra – Abstract and Modern

and hence [EH : R] 5 q. Also, since EH is a finite separable extension of R, EH

is a simple extension of R and hence EH 5 R(b) for some b  R. Let

q(x) 5 b0 1 b1x 1  1 bqxq  R[x]

be the minimal polynomial of b over R (note that deg(q(x)) 5 [R(b) : R] 5

[EH : R] 5 q). By Theorem 16.1.2, q(x) has a real root and hence q(r) 5 0
for some r  R. This implies that x 2 r is a factor of q(x) in R[x]. Being the
minimal polynomial of b over R, q(x) must be irreducible. All these imply
that q 5 1 and [EH : R] 5 1 and hence EH 5 R. Therefore, [E : R] 5 2m. Now,
we prove that m 5 1. Suppose, if possible, that m . 1. Then, [E : C] 5 2m21
and hence |G(E/C)| 5 2m21. Again, by the Sylow Theorem I, G(E/C) has a
subgroup S of order 2m22. If ES is the fixed field of S, then

[E : ES] 5 |S| 5 2m22.

Therefore, [ES : C] 5 2, since [E : C] 5 2m21. This is a contradiction to ­Theorem

16.1.1. Thus, m 5 1 and [E : R] 5 2. Since R ⊆ C ⊆ E and [C : R] 5 2, it
follows that [E : C] 5 1 and hence E 5 C. Thus, C is the splitting of g(x) In
­particular, C contains all the roots of g(x) and hence of f (x). Thus, f (x) com-
pletely factors into linear factors in C[x]. b

Corollary 16.1.1. C is an algebraically closed field.

Worked Exercise 16.1.1. For any f (x)  C[x], prove that f ( x )  R[ x ].

Answer: Let f (x)  C[x] be of degree n. We use induction on n. If n 5 0, then

2
f (x) 5 a  C and f (x ) f (x )  aa  a  R. Let n . 0 and assume that the
result is true for all polynomials g(x) of degree less than n in C[x]. Let

f (x) 5 a0 1 a1x 1  1 anxn 5 g(x) 1 anxn

      and f ( x )  g ( x )  an x n ,

where g(x) 5 a0 1 a1x 1  1 an21xn21  C[x]. Now,

f ( x ) f ( x )  ( g ( x )  an x n )( g ( x )  an x n )
 g ( x ) g ( x )  ( an g ( x )  an g ( x )) x n  an an x 2n
n1
 g ( x ) g ( x )  ∑ ( an ai  an ai ) x ni  an an x 2n
i0

Q001-Algebra-111001_CH 16.indd 4 9/16/2011 10:11:14 AM

 Selected Applications of Galois Theory   16-5

since ab  ab  ab  ab  R , g ( x ) g ( x )  R[ x ] and an an  R and , it fol-

lows that f ( x ) f ( x )  R[ x ].

16.2 CYCLIC EXTENSIONS

In this section, we discuss a special type of Galois extensions, namely cyclic
extensions which are Galois extensions whose corresponding Galois group
is a cyclic group (that is, generated by a single element). Before these, let us
first have the following definition.

Definition 16.2.1. Let F be a field and n be a positive integer. An element a

in an extension E of F is called a primitive nth root of unity in E if an 5 1 and
am  1 for any 0 , m , n.

Example 16.2.1  2 p i 

( ) ( )

 n 
1. For any n  Z , e 1
 cos 2np  isin 2np is primitive nth root of
unity in C.
2. Let F be a field of characteristic zero. Then, for any positive integer n,
xn 2 1 has n distinct roots in its splitting field and form a cyclic group
under multiplication. If a is a primitive nth root of unity, then am, where 0
, m , n and (m, n) 5 1, is also a primitive nth root of unity and we know
that the number of such m is f(n).
1  i 3
3. is a primitive cube root of unity.
2
( ) ( )
4. cos 65p isin 65p is a primitive 5th root of unity.

Theorem 16.2.1. Let F be a field of characteristic zero and E 5 F(a), where

a is a primitive nth root of unity. Then, G(E/F) is isomorphic to a subgroup
of Z*n , the group of all multiplicatively invertible elements in Zn and hence
G(E/F) is abelian.
Proof: Since a is a primitive nth root of unity, E 5 F(a) becomes the split-
ting field of xn 2 1  F[x]. Therefore, E is a finite normal extension of F and
G(E/F) is a Galois group.
Any s  G(E/F) is completely determined by its value s(a) and s(a) is
also an nth root of unity and hence s(a) 5 ai for some i , n and (i, n) 5 1.
Therefore, the correspondence s  i, where s(a) 5 ai, is a homomorphism
from G(E/F) into Z*n . Also, s o  5  o s, for s and   G(E/F). Also the
map s  i, is one-to-one. Thus, G(E/F) is an abelian group and is isomorphic
to a subgroup of Z*n . b
In general, G(E/F) need not be a cyclic group. For example, the Galois group
of x8 2 1 is isomorphic to the Klein 4-group and hence G(E/F) is not cyclic.

Q001-Algebra-111001_CH 16.indd 5 9/16/2011 10:11:15 AM

 16-6   Algebra – Abstract and Modern

Definition 16.2.2. Let E be Galois extension of a field F. Then, E is called a

cyclic extension of F if G(E/F) is a cyclic group (that is, generated by a single
element).

Examples 16.2.2
1. Recall that any finite extension of a finite field is separable. If E is the
splitting field of a polynomial f (x) over a finite field, then E is a Galois
extension of F and, by Worked Exercise 14.5.1, G(E/F) is a cyclic group
and hence E is a cyclic extension of F.
2. Let p be a prime and a be a primitive pth root of unity. Then, E 5 Q(a) is
the splitting field of xp 2 1  Q[x] and E is a cyclic extension of Q.

Theorem 16.2.2. Let F be a field and suppose that F contains a primitive nth
root of unity. Then, E is a finite cyclic extension of F of degree n if and only it
E is the splitting field of an irreducible polynomial xn 2 b  F[x].

Proof: Suppose that E is a finite cyclic extension of F of degree n; that is,

[E : F] 5 n and G(E/F) is a cyclic group of order n. Let s be a generator of
G(E/F). Then, Id, s, s2, …, sn21 are linearly independent over F. Let v be a
nth root of unity in F. Then, v  0 and

Id 1 v21s 1 v22s2 1  1 v2n11s n21

is a nonzero endomorphism of E as a vector space over F. Therefore, there

exists a  E such that

c 5 Id(a) 1 v21s(a) 1 v22s2(a) 1  1 v2n11s n21(a)  0

Now, s(c) 5 s(a) 1 v21s2(a) 1 v22s3(a) 1  1 v2n11sn(a)

5 v(v21s(a) 1 v22s2(a) 1  1v2n11s n21(a) 1 v2nsn(a))
5 v(v21s(a) 1 v22s2(a) 1  1 v2n11sn21(a) 1 a)
5 vc,

since v2n 5 1 and sn 5 Id. Recursively, we can prove that

sr(c) 5 vrc for all 1 # r # n 2 1.

Therefore, sr(cn) 5 (vrc)n 5 cn for all 1 # r # n 2 1.

This implies that cn is in the fixed field of G(E/F). But F is the fixed field
of G(E/F) and hence cn  F. Put b 5 cn  F. Then, xn 2 b  F[x] and

Q001-Algebra-111001_CH 16.indd 6 9/16/2011 10:11:15 AM

 Selected Applications of Galois Theory   16-7

c, cv, cv2, …, cvn–1 are the roots of xn 2 b. However sr(c) 5 cvr implies
that c, cv, cv2, …, cvn21 are also roots of the minimal polynomial f (x) of c
over F. Therefore, xn 2 b divides f (x). Since c is a root of xn 2 b, it follows
that f (x) divides xn 2 b and hence f (x) 5 xn 2 b. Since f (x) is the minimal
polynomial of c over F, f (x) is irreducible over F; that is, xn 2 b is irreducible
over F. Also,

[F(c) : F] 5 n 5 [E : F] and F(c) ⊆ E

and hence F(c) 5 E. Therefore, E is the splitting field of the irreducible poly-
nomial xn 2 b  F[x] over F.
Conversely suppose that xn 2 b  F[x] is an irreducible polynomial over F
and E is its splitting field over F. Let c  E be a root of xn 2 b; that is, b 5
cn. Then, clearly c, cv, cv2, …, cvn21 are the n distinct roots of xn 2 b, where
v  F is a primitive nth root of unity. Therefore, xn 2 b is a separable irreduc-
ible polynomial and hence E 5 F(c) is a Galois extension of F. For each s 
G(E/s), let the set As be defined by

As 5 {r  Z : s(c) 5 vrc}.

Then, As is nonempty, since s(c) is also a root of xn 2 b. Also, for any

r  As,

As 5 r 1 nZ,

since vrc 5 vsc if and only if r  s(mod n). Further, for any s and   G(E/F),

s(c) 5 vrc and (c) 5 vsc ⇒ (s)(c) 5 s(vsc) 5 vss(c) 5 vr1sc.

Therefore, As 5 As 1 A, where the sum on the right side is interpreted as
the binary operation on the additive group Z/nZ (Zn) of integers modulo
n. Finally, if As 5 nZ, the zero in Z/nZ, then s(c) 5 c. Therefore, s is the
identity on E (since E 5 F(c) and s/F 5 Id). Consequently, s  As is an
isomomorphism of G(E/F) onto a subgroup of Z/nZ. Also,

[F(c) : F] 5 The degree of the minimal polynomial of c over F

5 The degree of xn 2 b 5 n.

since E 5 F(c) is a finite separable and normal extension of F, it follows that

|G(E/F)| 5 [E : F] 5 n and hence G(E/F)  Z/nZ, so that G(E/F) is a cyclic
group of order n. Thus, E is a finite cyclic extension of F degree n. b

Q001-Algebra-111001_CH 16.indd 7 9/16/2011 10:11:16 AM

 16-8   Algebra – Abstract and Modern

16.3 Solvable Groups

Before we take up certain applications of Galois theory to the theory of equa-
tions and geometric constructions in the next two sections, we first introduce
the concept of a solvable group and discuss some of their properties.

Definition 16.3.1. Let G be a group. A sequence

{e} 5 G0 ⊆ G1 ⊆ … ⊆ Gn 5 G
of subgroups of G is called a normal series of the group G if each Gi is a
normal subgroup of Gi11 for 0 # i , n.
Note that, in the above, Gi may not be a normal subgroup of G; it is a normal
subgroup of Gi11. Also, {e} is always a normal subgroup of any subgroup of G
and therefore, for being a normal series, we can simply prescribe that Gi is nor-
mal in Gi11 for 0 , i , n and we can talk about the quotient groups Gi11/Gi.

Definition 16.3.2. A group G is said to be a solvable group if there exists a

normal series
{e} 5 G0 ⊆ G1 ⊆ G2 ⊆ … ⊆ Gn 5 G

such that the quotient Gi11/Gi is an abelian group for 0 # r , n; and, in this
case the series {e} 5 G0 ⊆ G1 ⊆ … ⊆ Gn 5 G is called a solvable series.

Examples 16.3.1
1. Any abelian group G is a solvable group, since {e} 5 G0 ⊂ G1 5 G is a
solvable series.
2. Let S3 be the symmetric group of degree 3 and let H 5 {e, (1 2 3), (3
2 1)}, where (1 2 3) is the 3-cycle mapping 1 → 2, 2 → 3, and 3 → 1.
Then, H is a subgroup of S3. Now,
{e} ⊂ H ⊂ S3
is a solvable series in S3, since H is an abelian normal subgroup and S3/H
is a group of order 2 and hence abelian. Therefore, S3 is a solvable group.
Recall from group theory that for any elements a and b of a group G, the
element aba21b21 is called a commutator in G and the subgroup G9 generated
by the set of all commutators in G is called the derived subgroup of G. For
any positive integer n, we define the nth derived subgroup of G, denoted by
G(n), is defined recursively as follows:

G(1) 5 G9 and G(n) 5 (G(n21))9 for n . 1.

Clearly, G is abelian if and only if G9 5 {e}.

Q001-Algebra-111001_CH 16.indd 8 9/16/2011 10:11:16 AM

 Selected Applications of Galois Theory   16-9

Theorem 16.3.1. A group G is solvable if and only if G(n) 5 {e} for some
positive integer n.

Proof: Suppose that G is solvable and

{e} 5 G0 ⊆ G1 ⊆ … ⊆ Gn 5 G

is a solvable series in G. For each i, Gi11/Gi is an abelian group and hence its
derived subgroup is trivial. This implies that

Gi1 ⊆ Gi for all 0 # i , n.

In particular, G Gn ⊆ Gn1 and hence

G (2) ⊆ Gn1 ⊆ Gn2 .

By induction, we can prove that G(i) ⊆ Gn21 and hence G(n) ⊆ G0 5 {e}. Thus,
G(n) 5 {e}.
Conversely suppose that G(n) 5 {e} for some n . 0. Then,

{e} 5 G(n) ⊆ G(n21) ⊆ … ⊆ G(1) ⊆ G

is a solvable series, since H/H, is abelian for any group H. Thus, G is a solv-
able group. b

Theorem 16.3.2. Any subgroup and any quotient group of a solvable group
is solvable.

Proof: Let G be a solvable group. Then, there exists a positive integer n such
that G(n) 5 {e}. If H is any subgroup of G, then H(n) ⊆ G(n) 5 {e} and hence
H(n) 5 {e} so that H is solvable. Let G/N be a quotient group of G, where N is
a normal subgroup of G. Now,

(G/N)(n) 5 G(n) 1 N 5 {N}, since G(n) 5 {e},

and hence G/N is solvable. b

By the fundamental theorem of homomorphisms, any homomorphic
image of a group G is isomorphic to a quotient group of G and hence we have
the following corollary.

Corollary 16.3.1. Any homomorphic image of a solvable group is solvable.

The following is a converse of the above theorem.

Q001-Algebra-111001_CH 16.indd 9 9/16/2011 10:11:16 AM

 16-10   Algebra – Abstract and Modern

Theorem 16.3.3. Let N be a normal subgroup of a group G such that both N

and G/N are solvable groups. Then, G is a solvable group.

Proof: Recall that any subgroup of G/N is of the form H/N, where H is a
subgroup of G containing N. Now, since G/N is a solvable group, there exist
a sequence of subgroups

N ⊆ H1 ⊆ H2 ⊆ … ⊆ Hn 5 G

such that {N} ⊆ H1/N ⊆ H2/N ⊆ … ⊆ Hn/N 5 G/N is a solvable series in G/N.
In particular, each Hi is a normal subgroup of Hi11 (since Hi/N is a normal
subgroup of Hi11/N). Also,

(Hi11/N)/(Hi/N)  Hi11/Hi

and hence Hi11/Hi is an abelian group. Since N is also solvable, there exists
solvable series

{e} 5 N0 ⊆ N1 ⊆ … ⊆ Nm 5 N

in N. Now, the series

{e} 5 N0 ⊆ N1 ⊆ … ⊆ Nm ⊆ H1 ⊆ H2 ⊆ … ⊆ Hn 5 G

is a solvable series in G. Thus, G is a solvable group. b

Theorem 16.3.4. A finite group G is solvable if and only if there exists a

sequence {e} 5 G0 ⊆ G1 ⊆ … ⊆ Gn 5 G of subgroups in G such that, for
each 0 # i , n, Gi is a normal subgroup of Gi11 and Gi11/Gi is a cyclic group
of prime order.

Proof: First observe that, if B is a proper normal subgroup of a finite group

A such that A/B is abelian, then there exists a maximal normal subgroup N
of A containing B and then A/N is a simple abelian group and hence of prime
order. Now, suppose that G is a solvable group and {e} 5 G0 ⊆ G1 ⊆ … ⊆
Gn 5 G be a solvable series in G. By the above observation, we can construct
a sequence of subgroups.

Gi  H i ,0 ⊆ H i ,1 ⊆  ⊆ H i , mi  Gi1

such that Hi, j is a normal subgroup of Hi, j11 and Hi, j11/Hi, j is a group of prime
order. This is true for each 0 # i , n and hence, by clubbing all these sequences
we get a required sequence of subgroups in G. The converse is trivial. b

Q001-Algebra-111001_CH 16.indd 10 9/16/2011 10:11:16 AM

 Selected Applications of Galois Theory   16-11

Worked Exercise 16.3.1. Prove that the symmetric group S4 of degree 4 is a

solvable group.

Answer: Let A4 be the alternating group of degree 4 (that is, the subgroup of
even permutations in S4). Then, A4 is a normal subgroup of S4. Put

a 5 (1 2) o (3 4), b 5 (1 3) o (2 4) and c 5 (1 4) o (2 3)

and H 5 {Id, a, b, c}. Then, H is a normal subgroup of A4 and H is an abelian

group (since ab 5 c 5 ba, etc.). Now, {Id} ⊂ H ⊂ A4 ⊂ S4 is a solvable series
in S4. Thus, S4 is a solvable group.

Worked Exercise 16.3.2. Prove that the dihedral group Dn is solvable for any
positive integer n . 1.

Answer: Let s be the cycle (1 2 3 … n), Then,

H 5 {Id, s, s2, …, sn21}

is a cyclic group of order n and of index 2 (since Dn is of order 2n). ­Therefore,

H is a normal subgroup of Dn and Dn/H is of order 2 and hence an abelian
group. Therefore, {Id} ⊂ H ⊂ Dn is a solvable series in Dn. Thus, Dn is a
solvable group.

Worked Exercise 16.3.3. Prove that the symmetric group Sn is not solvable
for any n . 4.

Answer: Let n . 4. It is known that the derived subgroup of Sn is An, the

alternating group of degree n; that is, Sn  An . Also, since An is simple and An
is a normal subgroup of An, it follows that An  {e} or An  An. But An  {e}
and hence An  An . Now, Sn(2)  An  An and

Sn(r )  An for all r  2.

Thus, by Theorem 16.3.1, Sn is not solvable.

16.4 POLYNOMIALS SOLVABLE BY RADICALS

In this section, we use the techniques of Galois theory to find necessary and
sufficient conditions for a polynomial over field to be solvable by radicals.
First we need some preparation. Throughout this section, we assume that all
fields are of characteristic zero.

Q001-Algebra-111001_CH 16.indd 11 9/16/2011 10:11:17 AM

 16-12   Algebra – Abstract and Modern

Definition 16.4.1. Let F be a field and K be an extension of F. Then, K is said

to be a radical extension of F if there exists a sequence of fields

F 5 F0 ⊆ F1 ⊆ … ⊆ Fn 5 K

such that, for each 0 , i # n, Fi 5 Fi21(ai) for some ai  Fi with the property
that airi  Fi1 for some ri  1.
Observe that, if Fi 5 Fi21(ai) and airi  Fi1 then ai is a root of the polynomial
x ri  airi  Fi1[ x ] and hence Fi is a simple algebraic extension of Fi21 and
therefore, [Fi : Fi 21] is finite. Also, since

[K : F] 5 [K : Fn21][Fn21 : Fn22] … [F1 : F],

K itself is a finite extension of F and

K 5 F(a1, a2, …, an).

Definition 16.4.2. A nonconstant polynomial f (x) over a field F is said to

be solvable by radicals if the splitting field L of f (x) over F is contained in a
radical extension of F.

Example 16.4.1. Let f (x) 5 a 1 bx 1 x2 be a monic quadratic polyno-

2
mial over the field Q of rational numbers. Its roots are a1  b b 4a and a2  b b
2 2
2 2
 b b 4a and a2  b b 4a .
2 2
Let L 5 Q(v), where v  b 2  4a . Then, v2 5 b2 2 4a  Q and there-
fore L is a radical extension of Q. Also, L itself is the splitting field of f (x)
over Q. Therefore, f (x) is solvable by radicals.
It is known that polynomial of degree 3 and 4 are always solvable by radi-
cals. In this section, we prove that not all polynomials of degree greater than
four are solvable by radicals.

Theorem 16.4.1. Let E be a radical extension of a field F and

F 5 E0 ⊆ E1 ⊆ E2 ⊆ … ⊆ Er 5 E

be an ascending sequence of intermediate fields between F and E. Then, there

exists a radical extension E9 of F and an ascending sequence of intermediate
fields

F  E0 ⊆ E1 ⊆ E2 ⊆  ⊆ Es  E

Q001-Algebra-111001_CH 16.indd 12 9/16/2011 10:11:18 AM

 Selected Applications of Galois Theory   16-13

between F and E9 satisfying the following:

1. E  Er ⊆ Es  E
2. E9 is a normal extension of F.
3. For each 1 # i # s, Ei is a splitting field of a polynomial of the form
x mi  bi  Ei1[ x ].

Proof: We can assume that Ei 5 Ei21(ai) where ai is a root of x ni − ci  Ei−1[ x ]

for each 1 # i # r. Put n 5 n1n2 … nr. Let v be a primitive nth root of unity.
Consider the sequence

F 5 E0 ⊂ E0(v) 5 F(v) ⊂ E1(v).

Clearly, E1(v) is a radical extension of F. Since F(v) is a splitting field of

xn 2 1  F[x], F(v) is a normal extension of F. Therefore, F is the fixed field
of G(F(v)/F) and hence the polynomial

f1 (x ) ∏
 F ( v ) 
( x n1 s (c1 ))
s G 
 F 

is in F[x]. Here, we have f1 ( x )  ( x n1  c1 )t, where t 5 |G(F(v)/F)|, since

c1  F and hence s(c1) 5 c1 for all s  G(F(v)/F). Next, let g1(x) 5 (xn 2 1)
f1(x). Then, g1(x)  F[x]. Let K be the splitting field of g1(x) over F, so that K
is a normal extension of F. Clearly, a1  K, v  K and E1 ⊆ K. Further, it is
clear that there is a finite ascending sequence of intermediate fields between
F and K such that each field is a splitting field of a polynomial of the form
xm 2 b over the preceding field.
Now, we construct a field L such that K and E2 are subfields of L and L is a
normal extension of F. For this, we consider the polynomial

g 2 ( x )  g1 ( x ) f 2 ( x ), where f 2 ( x )  ∏ ( x n2 s (c2 )).

K 
s G 
F 

Since K is a normal extension of F, f2(x)  F(x) and hence g2(x)  F[x]. Take
L to be the splitting field of g2(x) over F. Then, L contains a2 and K and hence
E2 5 E1(a2) ⊆ L. Therefore, L is a normal extension of F containing E2. Also,
because of the nature of the polynomial g2(x), it is clear that there exists a
finite ascending sequence of intermediate fields between K and L such that
any member of the sequence is a splitting field of a polynomial of the form

Q001-Algebra-111001_CH 16.indd 13 9/16/2011 10:11:19 AM

 16-14   Algebra – Abstract and Modern

xm 2 b over the preceding member. Continuing like this, we can construct a

radical extension E9 of F satisfying the required properties. b
Before going to the main theorem, we prove the following two important
results which are useful in proving the main theorem.

Theorem 16.4.2. Let n be a positive integer and F be a field containing all the
nth roots of unity. Let E be the splitting field of the polynomial xn 2 a  F[x].
Then, the Galois group G(E/F) is abelian.

Proof: If b is any root of xn 2 a, then b, bc, bc2, …, bcn21 are the roots of xn
2 a, where c is a primitive nth root of unity and, also E 5 F(b). For any s1 and
s2  G(E/F), s1(b) 5 bci and s2(b) 5 bc j for some i and j and hence

(s1 o s2)(b) 5 s1(bcj) 5 bcjci 5 bci1j 5 (s2 o s1)(b).

This implies that s1 o s2 5 s2 o s1 for any s1 and s2  G(E/F). Thus, G(E/F)

is an abelian group. b

Theorem 16.4.3. Let n be a positive integer and F be a field. If E is the split-

ting field of the polynomial xn 2 a  F[x], then G(E/F) is a solvable group.

Proof: Let E be the splitting field of the polynomial xn 2 a  F[x]. If F con-

tains a primitive nth root of unity, then by Theorem 16.4.2, G(E/F) is abelian
and hence solvable. Suppose that F does not contain any primitive nth root
of unity. Let c be a primitive nth root of unity and b  E be a root of xn 2 a.
Then, cb is a root of xn 2 a and hence cb  E. Also, c 5 b21(bc)  E and
therefore

F ⊆ F(c) ⊆ E and {e} ⊆ G(E/F(c)) ⊆ G(E/F).

We prove that the later is a solvable series in G(E/F), so that G(E/F) is a solv-
able group. Since F(c) is the splitting field of the polynomial xn 2 1  F[x], we
get that F(c) is a normal extension of F. By the fundamental theorem of Galois
theory, G(E/F(c)) is a normal subgroup of G(E/F). Also, since F(c) contains the
primitive nth root of unity c and since E is the splitting field of the polynomial xn
2 a  F(c)[x], we get from Theorem 16.4.2 that F(E/F(c)) is an abelian group,
Further, by the fundamental theorem of Galois theory, we have

E
G  
 F   F (c) 
 G   Z*n ,
 E   F 
G  
 F (c) 

Q001-Algebra-111001_CH 16.indd 14 9/16/2011 10:11:19 AM

 Selected Applications of Galois Theory   16-15

where Z*n is the multiplicative group {r : 0 , r , n and (r, n) 5 1}, which

is abelian. Therefore,

{e} ⊆ G(E/F(c)) ⊆ G(E/F)

is a solvable series in G(E/F). Thus, G(E/F) is a solvable group. b

Note that a polynomial f (x) over a field F is solvable by radicals if we can
extract every root of f (x) by using a finite sequence of operations of addition,
subtraction, multiplication, division and taking nth roots, starting with ele-
ments of the field F. Now, we prove the following main theorem.

Theorem 16.4.4. Let f (x) be a polynomial over a field F of characteristic zero

and E be the splitting field of f (x) over F. Then, f (x) is solvable by radicals
over F if and only if the Galois group G(E/F) is a solvable group.

Proof: Suppose that f (x) is solvable by radicals over F. That is, the splitting
field E of f (x) over F is contained in a radical extension of F. Hence we can
find a sequence of fields

F 5 F0 ⊆ F1 5 F0(v1) ⊆ F2 5 F1(v2) ⊆ … ⊆ Fm 5 Fm21(vm)

such that viri  Fi1 for some integers ri  1 and E ⊆ Fm. By Theorem 16.4.1,
we can suppose that Fm is a normal extension of F and Fi is the splitting field
of x ri viri  Fi1[ x ]. Now,

{e} 5 G(Fm/Fm) ⊆ G(Fm/Fm21) ⊆ … ⊆ G(Fm/F)

and Fi is a normal extension of Fi21 (since it is the splitting field of a poly-

nomial). By the fundamental theorem of Galois theory, we have

G(Fm/Fi21)/G(Fm/Fi)  G(Fi/Fi21).

Also, by Theorem 16.4.3, G(Fi/Fi21) is a solvable group. Now, put Hi 5

G(Fm/Fi), we have

{e} 5 Hm ⊆ Hm21 ⊆ … ⊆ H0 5 G(Fm/F),

where Hi21/Hi is a solvable group. Now, Hm (5{e}) and Hm21/Hm are solvable
and hence, by Theorem 16.3.3, Hm21 is a solvable group. Also, since Hm22/
Hm21 is solvable, so is Hm22. Continuing this process, we get that H0 is a solv-
able group. That is, G(Fm/F) is a solvable group. Since E ⊆ Fm and

G(Fm/F)/G(Fm/E)  G(E/F),

Q001-Algebra-111001_CH 16.indd 15 9/16/2011 10:11:20 AM

 16-16   Algebra – Abstract and Modern

we get that G(E/F) is a solvable group.

Conversely suppose that G(E/F) is a solvable group, since the characteristic
of F is zero, E is a normal separable extension and hence

[E : F] 5 |G(E/F)| 5 n, say.

We divide the proof into two parts (1) and (2) as given below.
1. First we assume that F contains a primitive nth root of unity. Then, F
contains primitive mth roots of unity for all positive integers m which
divide n. since G(E/F) is a finite solvable group, there exists a sequence
of subgroups
{e} 5 G ⊆ G ⊆ G ⊆ … ⊆ G 5 G(E/F)
0 1 2 r

of G(E/F) such that Gi is a normal subgroup of Gi11 and Gi11/Gi is a

cyclic group. Let Fi be the fixed field of Gi. Then, we have the sequence
E5F ⊇F ⊇F ⊇…⊇F 5F 0 1 2 r

of intermediate fields between F and E. Then, by the fundamental theo-

rem of Galois theory, we have that G(E/Fi) 5 Gi and Gr21 5 G(E/Fr21) is
a normal subgroup of G(E/F) which implies that Fr21 is a normal exten-
sion of F. Now, E can be regarded as the splitting field of f (x) over Fr21
and hence E is a finite normal extension of Fr21. Since Gr22 is a normal
subgroup of Gr21, Fr22 is a normal extension of Fr21. Continuing this
process, we get that Fi21 is a normal extension of Fi. By the fundamental
theorem of Galois theory, we also have
G(Fi21/Fi)  G(E/Fi)/G(E/Fi21) 5 Gi/Gi21
which is cyclic. Therefore, Fi21 is a cyclic extension of Fi. Then, by
Theorem 16.2.2, Fi21 is the splitting field of an irreducible polynomial
x ni  bi belonging to Fi[x] and Fi21 5 Fi (ai), where aini  bi  Fi. Then,
we have
E 5 F (ar, ar21, …, a2, a1) 5 F(a1, a2, …, ar),
nr ni
a  Fr  F and ai  Fi 5 F (ar, ar21, …, ai) for all 1 # i , r. Thus,
r
f (x) is solvable by radicals.
2. Now, we take up the general case and drop the assumption that F con-
tains a primitive nth root of unity. Note that the polynomial xn 2 1  E
[x] has roots in E. Let v be a primitive nth root of unity in E. Then, E(v)
is the splitting field of f (x) regarded as a polynomial over F(v). If s is an
F(v)-automorphism of E(v), s leaves the coefficients of f (x) unaltered.
Since E is a normal extension of F.
s/E  G(E/F) for all s  G(E(v)/F(v)).

Q001-Algebra-111001_CH 16.indd 16 9/16/2011 10:11:20 AM

 Selected Applications of Galois Theory   16-17

Also, the map s  s/E is an monomorphism of the group G(E(v)/F(v))

into the group G(E/F), since any subgroup of a solvable group is solv-
able and since G(E/F) is a solvable group, it follows that G ( E ( v) / F ( v))
is solvable. By part (1) above, E(v) is a radical extension of F(v) and
hence E(v) is a radical extension of F. Since the splitting field E of f (x)
is contained in the radical extension E(v) of F, it follows that f (x) is
solvable by radicals over F.
Since the symmetric group Sn is not solvable (by Worked Exercise 16.3.3)
for any n . 4, the following is an immediate consequence of the above
theorem.

Corollary 16.4.1. Let f (x)  F[x] such that the Galois group of f (x) is
isomorphic to Sn for some n . 4. Then, f (x) is not solvable by radicals
over F.
Recall from Worked Exercise 15.2.1 that the Galois group of a polynomial
f (x)  F[x] having r distinct roots is isomorphic to a subgroup of Sr which is
the group of permutations of the r distinct roots of f (x). Before we give some
more applications of Theorem 16.4.4, let us have the following definition.

Definition 16.4.3. Let Sn be the symmetric group of degree n. Then, a sub-

group H of Sn is called transitive permutation group if, for any i, j  {1, 2, …, n},
there exists s  H such that s(i) 5 j.

Theorem 16.4.5. Suppose that a polynomial f (x)  F[x] has no multiple

roots. Then, f (x) is irreducible over F if and only if the Galois group of f (x) is
isomorphic to a transitive permutation group.

Proof: Let E be a splitting field of f (x) over F and a1, a2, …, an be the roots
of f (x) in E. Let G be the Galois group of f (x) over F. Then, G  G ( E / F ).
For each s  G, s(a1), s(a2), …, s(an) are roots of f (x) in E and hence a per-
mutation of a1, a2, a3, …, an. As in Worked Exercise 15.2.1, we can consider
G as a subgroup of Sn.
Now, suppose that G is a transitive permutation group. Let p(x) be the mini-
mal polynomial of a1 over F. For each root ai, there exist s  G such that
s(a1) 5 ai (since G is transitive). Then,

p(ai) 5 p(s(a1)) 5 s(p(a1)) 5 s(0) 5 0.

Therefore, each ai is a root of p(x). Since p(x) divides f (x), it follows that
f (x) 5 cp(x) for some c  F. Since p(x) is irreducible over F, f (x) is also
irreducible over F.

Q001-Algebra-111001_CH 16.indd 17 9/16/2011 10:11:21 AM

 16-18   Algebra – Abstract and Modern

Conversely suppose that f (x) is irreducible over F. Then, ,f (x)., the ideal
generated by f (x) in F(x), is a maximal ideal of F(x) and hence F [ x ]  f (x ) 
is a field and

F (ai )  F [ x]< f (x) > for each 1 # i # n.

If ai : F [ x ]  f (x )  → F(ai) is the isomorphism given by ai (g(x) 1 ,f (x).) 5

g(ai), then a j  ai 1 is an isomorphism of F(ai) onto F(aj) sending ai to aj
and fixing the elements of F. Since E is a normal extension of F, a j ο ai 1
can be extended to an F-automorphism s of E. Then, s  G(E/F) 5 G
and s(ai) 5 aj. Thus, G is a transitive permutation group.

Theorem 16.4.6. Let p be a prime number and f (x) be a monic irreducible

polynomial over Q of degree p. Suppose that f (x) has exactly two nonreal roots
in C. Then, the Galois group of f (x) is isomorphic to the symmetric group Sp.

Proof: Let E be the splitting field of f (x) over Q. Then, Q ⊆ E ⊆ C. By

­Theorem 16.4.5, G(E/Q) is isomorphic to a transitive permutation group H,
which is a subgroup of Sp. Let a1, a2, …, ap be the roots of f (x) among which
a1 and a2 are nonreal. Since the coefficients in f (x) are rational numbers, it
follows that a complex number a is a root of f (x) if and only if its conjugate
a is also a root of f (x). Therefore, a1 and a2 must be conjugates to each other;
that is a1  a2 and a2  a1. Consider the embedding s;E → Q defined by
s (z )  z , where Q is the algebraic closure of Q. Since E is a normal exten-
sion of Q, s maps E onto E and hence s  G(E/Q). s takes a1 to a2 and a2 to a1
and s(ar) 5 ar for all 2 , r # p. This implies that H contains the transposition
(a1 a2). Also, since [Q(a1) : Q] 5 p and [E : Q] 5 |G(E/Q)| 5 |H|, it follows
that p divides the order of the group and hence, by the Cauchy’s theorem in
group theory, H has an element of order p. That is, there exist e    H such
that p 5 e. Since p is prime,  must be a p-cycle. Thus, H contains a p-cycle
and a transposition. By the exercise given below in Worked Exercise 16.4.1,
H 5 Sp. Thus, G(E/Q)  Sp.

Worked Exercise 16.4.1. Let H be a transitive permutation group in Sn con-

taining an n-cycle and a transposition. Then prove that H 5 Sn.

Answer: Since any permutation is a product of transpositions, it is enough if

we prove that H contains all transpositions in Sn. Without loss of generality,
we can assume that s = (1 2 3 … n)  H and a 5 (1 2)  H. Now, it can be
verified that

Q001-Algebra-111001_CH 16.indd 18 9/16/2011 10:11:22 AM

 Selected Applications of Galois Theory   16-19

(i 1 1 i 1 2) 5 si ο a ο si  H for all 0 # i # n 2 2.

Therefore, (i i 1 1)  H for all 1 # i , n. Also,

(i i 1 1) ο (i 1 1 i 1 2) ο (i i 1 1) 5 (i i 1 2)
   and (i i 1 2) ο (i 1 2 i 1 3) ο (i i 1 2) 5 (i i 1 3) and so on.

Therefore, (i j)  H for all 1 # i , j # n. Thus, H contains all transpositions

in Sn and hence H 5 Sn.

Worked Exercise 16.4.2. Prove that the polynomial x5 2 9x 1 3  Q[x] is

not solvable by radicals over Q.

Answer: By the Eisenstein’s criterion for the irreducibility of polynomial

over Q, it follows that f (x) 5 x5 2 9x 1 3 is irreducible over Q. Since f (0) 5
3 . 25 5 f (1), we get a real root of f (x) in the interval (0, 1) (by the inter-
mediate value theorem in analysis). In the same way, we get a real root in the
interval (1, 2) also. Further, by the Descarte’s rule of signs, we have

the number of positive real roots # the number of

               changes of signs in f (x) (52)
and the number of negative real roots # the number of
              changes of signs in f (2x) (51).

Also 0 is not a root of f (x). Thus, there are at most three real roots of
f (x). Also, the roots occur in conjugate pairs. Thus, all these imply that f (x)
has exactly two nonreal roots in C. By Theorem 16.4.6, the Galois group of
f (x) over Q is isomorphic with S5, which is not a solvable group. Thus, by
­Theorem 16.4.4, f (x) is not solvable by radicals over Q.

Worked Exercise 16.4.3. Let f (x)  F[x] be an irreducible polynomial over

a field F. If f (x) has a root in a radical extension of F, then prove that f (x) is
solvable by radicals over F.

Answer: Let E be a radical extension of F containing a root of f (x). By

Theorem 16.4.1, there exists a radical extension E9 of F such that E ⊆ E9
and E9 is a normal extension of F. Since f (x) is irreducible over F and has a
root in E, it follows that f (x) has a root in E9. Further, since E9 is a normal
extension of F, E9 contains a splitting field of f (x). Thus, f (x) is solvable by
radicals over F.

Q001-Algebra-111001_CH 16.indd 19 9/16/2011 10:11:22 AM

 16-20   Algebra – Abstract and Modern

16.5 CONSTRUCTIONS BY RULER AND COMPASS

There are several problems, open for several years, in Euclidean geometry
and some of these are the following:
1. Can we construct by ruler and compass a square having the same area as
that of a given square?
2. Can we construct by ruler and compass a cube having twice the volume
of a given cube?
3. Can an angle be trisected using ruler and compass?
4. Can we construct a regular polygon having n sides using ruler and com-
pass?

Mathematicians proved that all of the above impossible using the techniques
of Galois theory. In this section, we present proofs of the impossibilities of the
above.
Before we take up the proofs, let us be clear that a ruler (or straight edge)
is an instrument through which we can draw a line segment joining two given
points in the Euclidean space and that a compass is an instrument by which we
can draw a circle with a given point as centre and passing through another given
point. Also, let us understand that ‘construction by using ruler and compass’
means ‘construction by using ruler and compass only in finite number of steps’.
Let us imagine that we are given a line segment which we shall define to
be one unit in length.

Definition 16.5.1. A real number a is said to be constructible if one can con-

struct a line segment of length |a| in a finite number of steps from the given
line segment of unit length by using ruler and compass alone.
Recall that by using ruler and compass, it is possible to draw a perpen-
dicular to a given straight line at the given point on the line and to draw a line
passing through a given point and parallel to a given line.

Theorem 16.5.1. If a and b are constructible real numbers, then so are a1b,
a 2 b, ab and a/b if b ≠ 0.

Proof: Let a and b be constructible real numbers. We can suppose that a and b
are not zero. Then, there are line segments of lengths |a| and |b| available to us.
Now, extend the line segment of length |a|, and lay off on the extension the length
b with compass. This constructs a line segment of length a 1 b. Similarly, we
construct a 2 b. Notice that, clearly 2a and 2b are constructible.
Draw a line OA of length |a| and extend it. Draw a line through O not
containing A. Suppose OB is of length |b|. Take the point P such that OP is

Q001-Algebra-111001_CH 16.indd 20 9/16/2011 10:11:22 AM

 Selected Applications of Galois Theory   16-21

of unit length, by using the compass. Draw a line parallel to PA and passing
through the point B which cuts the line OA at Q. By the property of similar
triangles, we get that

B
P

O A Q
1 b

a OQ

and hence |OQ| 5 |a||b| 5 |ab|. Thus, ab is constructible. Next, let |OA| 5 |a|
and draw a line OB of length |b| through O not containing A. Suppose P is a
point on OB such that OP is of unit length. Draw BA and a line parallel to BA
passing through P which cuts the line OA at Q, say. Then, again by a property
|a|
of similar triangles, we have | OQ |  a . Thus, a/b is constructible. b
|b| b
B

O Q A

Corollary 16.5.1. The set of all constructible real members forms a subfield
F of the field R of real numbers.
Since Q is the prime subfield of R, it follows that the field F of con-
structible real numbers is a field extension of Q. Now, in the two-dimensional
Euclidean plane, we can locate any point (q1, q2) whose both coordinates are
rational ­numbers.

Definition 16.5.2. A point (a, b) is said to be a constructible point if both a and

b are constructible real numbers. A line is said to be constructible if it passes
through two distinct constructible points. A circle is said to be constructible if
its centre is a constructible point and it passes through another constructible
point (or equivalently, its radius is a constructible real number).

Q001-Algebra-111001_CH 16.indd 21 9/16/2011 10:11:23 AM

 16-22   Algebra – Abstract and Modern

Theorem 16.5.2. Let a  R be constructible. Then, there exists a subfield K

of R containing a such that [K : Q] 5 2n for some nonnegative integer n.

Proof: First notice that starting from Q 3 Q, any further point in the plane
which can be located by using ruler and compass can be found in one of the
following three ways:
1. As an intersection of two constructible lines (that is, lines passing
through two given points having rational coordinates).
2. As an intersection of a constructible line and a constructible circle.
3. As an intersection of two constructible circles.
Equations of lines and circles of the type mentioned in (1), (2) and (3) above
are of the form.

ax 1 by 1 c 5 0
     and x2 1 y2 1 dx 1 ey 1 f 5 0,

where a, b, c, d, e and f are all rational numbers. It is clear that, for case
(1) above, a simultaneous solution of two linear equations with rational
coefficients can only lead to rational values of x and y, which give us a
new constructible point. For case (2) above, upon substitution in a qua-
dratic equation using linear equation and when solved by the quadratic
formula, gives solutions involving square roots of numbers which are pos-
sibly not square in Q. In case (3), the intersection of two circles, whose
equations are

x2 1 y2 1 d1x 1 e1y 1 f 1 5 0
        and x2 1 y2 1 d2x 1 e2y 1 f 2 5 0,

is same as the intersection of the first circle and the line whose equation is

(d1 2 d2)x 1 (e1 2 e2)y 1 (f 1 2 f 2) 5 0

which reduces to case (2). In any case, the new constructed numbers lie in a
field Q( a ) for some 0 , a  Q. If E is the smallest field containing those
real numbers constructed so far, the above argument shows that the next new
number constructed, lie in a field E ( a ) for some a  H. Therefore, starting
from Q, we get

Q1 5 the set of constructible numbers from Q in the above way.

In a similar way, we get Q2 from Q1 and so on. In this way, we get an ascend-
ing chain of subfields of R

Q001-Algebra-111001_CH 16.indd 22 9/16/2011 10:11:23 AM

 Selected Applications of Galois Theory   16-23

Q 5 Q0 ⊆ Q1  Q2 ⊆ …

It follows from all these that, if a real number a is constructible from Q, then
there exists an ascending chain of subfields of R

Q 5 Q0 ⊆ Q1 ⊆ Q2 ⊆ … ⊆ Qm

satisfying the following:

1. a  Qm
2
2. Qi 5 Qi21(ai), where ai  Qi–1 for 1 # i # m
3. [Qi, Qi–1] # 2.
Thus, it follows that [Qm : Q] 5 2n for some 0 # n # m and Qm is the required
subfield K of R containing a. b

Corollary 16.5.2. If a  R is a constructible number, then

[Q(a) : Q] 5 2r for some 0 # r  Z.

In fact, the converse of the above result is also true. But before this, we first
prove the following theorem.

Theorem 16.5.3. If a . 0 is constructible, then so is a.

Proof: Let OA be a line segment of length a and find a point P on

extended OA so that OP is of unit length. Find the midpoint of PA and
draw a semicircle with PA as diameter. Draw a perpendicular at O inter-
secting the semicircle at Q. Now, the triangles OPQ and OQA are similar
( OPQ OQA and OAQ  PQO) and hence we have

OQ OQ OP 1
   .
a OA OQ OQ

Therefore, OQ is of length a . Thus, a is constructible. b

P O A

Q001-Algebra-111001_CH 16.indd 23 9/16/2011 10:11:24 AM

 16-24   Algebra – Abstract and Modern

Corollary 16.5.3. Let K be the subset of R consisting of numbers construct-

ible from Q. Then, K is a subfield of R containing Q and square roots of all
nonnegative numbers in K.

Theorem 16.5.4. Let a  R such that [Q(a) : Q] 5 2r for some 0 # r  Z.

Then, a is constructible.

Proof: Using the fundamental theorem of Galois theory, we can get a1, a2,
…, ar such that

Q ⊆ Q(a1) ⊆ Q(a1, a2) ⊆ … ⊆ Q(a1, a2, …, ar) 5 Q(a),

where [Q(a1, …, ai) : Q(a1, …, ai21)] 5 2 and a1, a2, …, ar are constructible.
Therefore, a is constructible. b

Definition 16.5.3. An angle u is said to be constructible by ruler and com-

pass if the point (cos u, sin u) is constructible from Q 3 Q.
Now, we prove the impossibilities of certain geometric constructions men-
tioned in the beginning of this section.

Theorem 16.5.5. Doubling the cube is impossible; that is, when a cube is
given, it is not always possible to construct with ruler and compass alone the
side of a cube which has double the volume of the given cube.

Proof: Without loss of generality, we can assume that the side of the given
cube is of unit length and hence its volume is 1. Let the side of the cube to be
constructed, if possible, be x. Then, x3 5 2 (double the volume of the given
cube) and hence x3 2 2 5 0. Therefore, we have to construct the number 21/3.
But x3 2 2 is an irreducible polynomial over Q and hence [Q(21/3) : Q] 5 3
which is not of the form 2n. Thus, by Corollary 16.5.2, 21/3 is not constructible
by ruler and compass alone. b

Theorem 16.5.6. Squaring a circle is impossible; that is, when a circle is

given, it is not always possible to construct with ruler and compass alone a
square having area equal to the area of the given circle.

Proof: Without loss of generality, we can assume that the given circle has
radius 1 and hence its area is p. We need to construct a square of side p .
But p is not algebraic over Q (the proof of this is not given in this book) and
hence p is not algebraic over Q. Therefore, [Q( p ) : Q] is not finite. Thus,
p is not constructible. b

Q001-Algebra-111001_CH 16.indd 24 9/16/2011 10:11:25 AM

 Selected Applications of Galois Theory   16-25

Theorem 16.5.7. Trisecting the angle is impossible; that is, there exists an
angle which cannot be trisected with ruler and compass alone.

Proof: The adjoining figure indicates that an angle u can be constructed if

and only if a segment of length |cos u| can be constructed. Now, 60° is a con-
structible angle and we shall show that it cannot be trisected. If 60° can be
trisected, then the number cos 20° is constructible from Q. Let a 5 cos 20°.
Since

cos 3u 5 4 cos3u 2 3 cos u for any u,

we get that cos 60 5 4 cos320 2 3 cos 20 and hence

1
8a3 2 6a 2 1 5 0 (since cos 60 5 ).
2
C

�
O B

Now, the polynomial 8x3 2 6x 2 1 is irreducible over Q and has a root a and
therefore

[Q(a) : Q] 5 3  2r for any r  Z.

This implies that a is not constructible. Thus, 20° is not constructible and
hence 60° cannot be trisected. b

Theorem 16.5.8. It is impossible to construct a regular 7-gon by ruler and

compass alone.

Proof: Suppose, if possible, that a regular polygon of 7 sides (regular 7-gon)

is constructible. Then, the angle (p / 7) is constructible. Let a 5 2 cos (2p / 7).
Then, a is a constructible number. Put u  (2p / 7). Then,

sin 4u 5 2sin 3u

and hence 8 cos3u 1 4 cos2u 2 4 cos u 2 1 5 0. Therefore,

a3 1 a2 2 2a 2 1 5 0.

Q001-Algebra-111001_CH 16.indd 25 9/16/2011 10:11:25 AM

 16-26   Algebra – Abstract and Modern

Then, a is a root of the polynomial

x3 1 x2 2 2x 2 1

which is irreducible over Q. Therefore,

[Q(a) : Q] 5 3  2r for any r  Z,

which is a contradiction to our assumption that a regular 7-gon is constructible.

Thus, a regular 7-gon is not constructible. b

EXERCISES 16

1. Prove that the identity map is the only automorphism of the field R of real num-
ber and hence the Galois group of R/Q is trivial.

2. Prove that the Galois group of C/R is a cyclic group of order 2.

3. Find the Galois group of the splitting field of the polynomial x4 2 x2 1 1 over Q.

4. Prove that the symmetric group S5 is not solvable.

5. Prove that the polynomial x7 2 10x5 1 15x 15 is not solvable by radicals over Q.

6. Prove that x3 1 x2 2 2x 2 1 is irreducible over Q.

7. Prove that it is impossible to construct a regular 9-gon using ruler and compass
alone.

8. Prove that a regular 17-gon is constructible with ruler and compass alone.

Q001-Algebra-111001_CH 16.indd 26 9/16/2011 10:11:26 AM

 Answers/Hints to Selected
Even-Numbered Exercises
CHAPTER 1
EXERCISE 1(A)

2. (i) {0, 1, 2, 3, 4, 5}
(ii) {(21, 0) ∪ (0, 1)} ∩ Q
(iii) {, {a}, {c}, {a, c}}
(iv) {012, 013, …, 019, 023, …, 029, …, 089, 123, 124, …, 129,
134, …, 139, …}
(v) {(0, 0)}
(vi) , the empty set.

4. X 5 {1, 2, 3}, P(X) 5 {, {1}, {2}, {3}, {1, 2}, {2, 3}, {1, 3}, X}
6. X 5 {1, 2, 3, …, 100}
A 5 {1, 4, 9, 16, 25, 36, 49, 64, 81, 100} and B 5 {1, 3, 5, …, 99}
Ci 5 {i, i 1 1, i 1 2, i 1 3, i 1 4} for each 1  i  96
(i) A ∩ B 5 {1, 9, 25, 49, 81}
(ii) A ∪ B ∪ C2 5 {1, 2, 3, 4, 5, 6, 7, 9, 11, 13, 15, 16, 17, …}
96

(iii) i∪
1
Ci  X and X ∩ A  A
25

(iv) B ∩ ( i∪20 Ci )  {21, 23, 25, 27, 29}

(v) X 2 (A ∪ B) 5 {2, 6, 8, 10, 12, 14, 18, 20, 22, 24, 26, 28, …}
90

(vi) X  ( i∪
6
Ci )  {1, 2, 3, 4, 5, 95, 96, 97, 98, 99, 100}
(vii) 
(viii) A 2 B 5 {4, 16, 36, 64, 100}

8. Straight forward verifications.

Q001-Algebra-111001_Answers.indd 1 9/16/2011 10:00:48 AM

 A-2   Algebra – Abstract and Modern

10. (i) A  B 5 (A 2 B) ∪ (B 2 A) 5 (B 2 A) ∪ (A 2 B) 5 B  A
(ii) (A  B)  C 5 [((A 2 B) ∪ (B 2 A)) 2 C] ∪ [C 2 ((A 2 B) ∪
(B 2 A))]
5…
5 (A ∩ B ∩ C) ∪ ((A 2 B) 2 C) ∪ ((B 2 C) 2 A) ∪
((C 2 A) 2 B)
The other equality is by symmetry.
12. (i) True; since A ⊆ X ∩ Y ⇔ A ⊆ X and A ⊆ Y.
(ii) False; for, if X 5 {1, 2} and Y 5 {3, 4}, then the set A 5 {2, 3}
belongs P(X ∪ Y), but A is in neither P(X) nor P(Y).
(iii) False; for the empty set   P(X 2 Y), but   P(X) 2 P(Y).
(iv) True; since X  P(X) 5 P(Y) ⇒ X ⊆ Y.

EXERCISE 1(B)

2. (i) Clearly, A 5 B ⇒ A 2 C 5 B 2 C.
But A 2 C 5 B 2 C  A 5 B; for, if A and B are any subsets of
C, then A 2 C 5  5 B 2 C.
(ii) A ∩ C 5 B ∩ C  A 5 B; for, if C is any subset of A ∩ B, then A
∩ C 5 C 5 B ∩ C.
(iii) A ∪ C 5 B ∪ C  A 5 B; for, if A and B are any subsets of C, then
A ∪ C 5 C 5 B ∪ C.
(iv) (A 2 B) 3 (C 2 D) ⊆ (A 3 C) 2 (B 3 D). The other inclusion
may not be true; for, take A 5 {1, 2}, B 5 {3, 4}, C 5 {2, 3}
and D 5 {2, 5}. Then, (1, 2)  (A 3 C) 2 (B 3 D) and (1, 2) 
(A 2 B) 3 (C 2 D), since 2  C 2 D.
(v) (A 3 C) ∪ (B 3 D) ⊆ (A ∪ B) 3 (C ∪ D). The other inclusion
may not be true; for, in (iv) above, (2, 5)  (A ∪ B) 3 (C ∪ D), but
(2, 5)  (A 3 C) ∪ (B 3 D).
4. For each a  A, there are exactly m b’s in B such that (a, b)  A 3 B.
Since A is n elements, A 3 B has nm elements. Also, the number of rela-
tions from A to B is 2nm.
6. The number of functions from A into B is mn.
8. If f (a) 5 f (b), then a 5 b (since f |{a, b} is an injection). Therefore, f is
an injection if f |Z is an injection for any subset Z of X. The converse is
clear. This statement is not valid for surjections. The map f : R → R1 ∪ {0}

Q001-Algebra-111001_Answers.indd 2 9/16/2011 10:00:48 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-3

given by f (a) 5 |a| is a surjection, but the restriction of f to (0, 1) is not

a surjection.
10. Yes; the number of such functions g is equal to the number of functions
from X 2 A into Y.
12. f −1[  2, 8)  (  8, 8);
f21(2, 0] 5 {0}; f 21(21, 1) 5 (21, 1);
f −1 (Z)  { n : 0  n  Z}.
14. (1) If g : B → A such that g o f 5 IA, then, for any x, y  A,
f (x) 5 f (y) ⇒ g(f (x)) 5 g(f (y)) ⇒ x 5 y
and hence f is an injection, conversely suppose that f is an injection.
Choose a0  A and define g : B → A by
a if b  f (a), a  A
g (b)  
a0 if b  f (A).

Then, g is a function and g o f 5 IA.

(2) If f o h 5 IB, then, for any b  B, f (h(b)) 5 b and h(b)  A and hence
f is a surjection. Conversely, suppose that f : A → B is a surjection.
For each b  B, f21({b}) is a nonempty subset of A and choose an
element ab  f 21({b}) for each b  B. Then, h : B → A defined by
h(b) 5 ab is a function and f (h(b)) 5 b for all b  B.
16. Straight forward verification.
18. Define f : Z → R by f (n) 5 |n| for all n  Z. If B is the interval (0, 1) in
R, then f 21(B) 5  and f (f 21(B)) 5 f () 5  , B.
20. (i) a  A ⇒ f (a)  f (A) ⇒ a  f 21(f (A)).
Therefore, A ⊆ f 21(f (A)) and hence f (A) ⊆ f (f 21(f (A))).
f (f 21(B)) ⊆ B for all B ⊆ Y
and hence f (f 21(f (A))) ⊆ f (A). Thus, f (f 21(f (A))) 5 f (A).
(ii) Proof is similar to (i).
(iii) Clear.
22. For any a and b in X,
f (a) 5 f (b) ⇒ (g o f )(a) 5 (g o f )(b) ⇒ a 5 b

Q001-Algebra-111001_Answers.indd 3 9/16/2011 10:00:48 AM

 A-4   Algebra – Abstract and Modern

(since g o f is an injection). Therefore, f is an injection. If f is a surjection

also, then f is a bijection and, for any x and y in Y, choose a and b  X
such that f (a) 5 x and f (b) 5 y. Now,
g(x) 5 g(y) ⇒ g(f (a)) ⇒ g(f (b)) ⇒ a 5 b ⇒ x 5 y.
Thus, g is an injection.
24. (i) n!
(ii) 0
(iii) mCn  n!
26. (1) ⇒ (2): Clearly, A ⊆ f21(f (A)). If x  f21(f (A)), then f (x)  f (A) and
hence f (x) 5 f (a) for some a  A. Since f is an injection x 5 a  A.
Thus, f21(f (A)) 5 A.
(2) ⇒ (3): Clearly, f (A1 ∩ A2) ⊆ f (A1) ∩ f (A2).
x  f (A1) ∩ f (A2) ⇒ x 5 f (a1) 5 f (a2) for some ai  Ai
⇒ a1  f 21(f ({a2})) 5 {a2} (by (2))
⇒ a1 5 a2  A1 ∩ A2 and x 5 f (a1)  f (A1 ∩ A2)
Thus, f (A1) ∩ f (A2) 5 f (A1 ∩ A2).
(3) ⇒ (1): For any a and b  X,
f (a) 5 f (b) ⇒ f (a)  f ({a}) ∩ f ({b}) 5 f ({a} ∩ {b})
⇒ a 5 b (otherwise, {a} ∩ {b} 5 )
28. Consider ( f1 ο f 2 ο  ο f n ) ο ( f n1 ο f n11 ο  ο f11 ).

30. (i) ⇒ (ii): If f has a left inverse, then f is an injection (by Ex. 14 of
­Exercise 1(B)). Suppose that f (X)  Y. Choose x1  x2  X and define
 x if y  f ( x ), x  X
g1, g2 : y → x by g1 ( y )  
 x1 if y  f ( X )
 x if y = f (x ), x  X
      and g 2 (y )   .
 x2 if y  f (X )
Then, g1  g2 (since Y 2 f (X)  ) and g1 o f 5 IX 5 g2 o f, which is a
contradiction to the uniqueness of the left inverse of f.
(ii) ⇒ (i) and (ii) ⇒ (iii) are clear.
(iii) ⇒ (ii): If f has a right inverse, then f is a surjection. (by Ex.
14 of Exercise 1(B)). Suppose a and b  X such that f (a) 5
f (b) 5 y0, say. For each y  y0 in Y, choose xy  X such that f (xy) 5 y.
Define h1, h2 : Y → X by

Q001-Algebra-111001_Answers.indd 4 9/16/2011 10:00:49 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-5

 x y if y  y0  x y if y  y0
h1 (y )   and h2 (y )   .
 a if y  y0  b if y  y0
 

Then, f o h1 5 IY 5 f o h2 and hence, by (iii), h1 5 h2 so that a 5 b. Thus,

f is an injection also.
32. If a, b and c are three distinct elements of X, define f and g : X → X by
a if x  b a if x  c
 
g (x )  b if x  a and f (x )  c if x  a
 
 x otherwise  x otherwise

Then, (g o f )(a) 5 g(c) 5 c  b 5 f (b) 5 (f o g)(a) and hence g o f 

f o g.
34. Define g and h : P(X) → P(P(X)) by
g(A) 5 P(A) and h(A) 5 {{a} : a  A} for any A  P(X).
Then, (f o g)(A) 5 A 5 (f o h)(A); that is, g and h are distinct right
inverses of f (note that   g(A) and   h(A)).
36. (i) Follows from 10 of Exercise 1(B).
(ii) Suppose that  is a bijection and A  X. Let y1  y2  Y. Choose
x0  X 2 A and define f and g : X → Y by
 y1 if x  A
f (x) 5 y1 for all x  X and g (x )   .
 y2 if x  A

Then, f (x0) 5 y1  y2 5 g(x0). But (f ) 5 (g) and f  g. Therefore, 
is a bijection implies that A 5 X.
38. This follows from Ex. 37.

EXERCISE 1(C)

2. Let X 5 {a, b, c} and

R 5 {(a, b), (b, a)}
S 5 {(a, a), (b, b), (c, c), (a, b), (b, c)}
T 5 {(a, b), (b, c), (a, c)}.
Then, R is symmetric, but neither reflexive on X nor transitive. S is
reflexive, but neither transitive nor symmetric. T is transitive, but neither
symmetric nor reflexive.

Q001-Algebra-111001_Answers.indd 5 9/16/2011 10:00:50 AM

 A-6   Algebra – Abstract and Modern

4. Easy verification.

6. Since a 1 b 5 b 1 a, ((a, b), (a, b))  R and hence R is reflexive on X.

a 1 d 5 b 1 c ⇒ c 1 b 5 d 1 a. Therefore, R is symmetric.
((a, b), (c, d))  R and ((c, d), (e, f ))  R
⇒ a 1 d 5 b 1 c and c 1 f 5 d 1 e
⇒ a 1 d 1 f 5 b1 c 1 f 5 b 1 d 1 e
⇒ a 1 f 5 b 1 e ⇒ ((a, b), (e, f ))  R

Therefore, R is transitive. Thus, R is an equivalence relation on X.

8. (i) (x, y)  R ⇒ (a  R(x) ⇔ (a, x)  R ⇔ (a, y)  R ⇔ a  R(y)
and hence R(x) 5 R(y))
R(x) 5 R(y) ⇒ x  R(y) ⇒ (x, y)  R
(ii) R(x) ∩ R(y)   ⇔ (a, x)  R and (a, y)  R for some a  X
⇔ (x, y)  R
(iii) is trivial.

10. Straight forward verifications. In Exercise 5 above, X/R is bijective with

the set Q of rational numbers.

12. Suppose that the given conditions are satisfied.

(a, b)  R ⇒ (a, b) and (b, b)  R ⇒ (b, a)  R.
Therefore, R is symmetric. Also,
(a, b)  R and (b, c)  R ⇒ (c, a)  R ⇒ (a, c)  R.
Thus, R is transitive also. The converse is trivial.

14. {Q 1 r : r  R}
(i)
{Z 1 r : r  Q}
(ii)
(iii)
R is not reflexive and hence not an equivalence relation.
(iv){{(a, b)  R2 : a2 1 b2 5 r} : 0  r  R}.
(v)is not equivalence relation (it is not symmetric).
(vi)is not reflexive, since (1, 1)  R
(vii)is not reflexive, since (2¼, 2¼)  R
(viii)is not transitive.
(ix){{A  B : B  F} : A  P(X)}, where F is the class of all finite
subsets of X.
(x) {rQ1 : r  R*}

Q001-Algebra-111001_Answers.indd 6 9/16/2011 10:00:50 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-7

EXERCISE 1(D)

2. (i) If |X| 5 |Z| and h : X → Z is a bijection, then f o h21 is an injection

of Z into Y. By Theorem 1.4.5, |Y| 5 |Z|.
(ii) If |X| 5 |Y| and f (X) ⊆ A ⊆ Y, then
|X| 5 |f (X)|  |A|  |Y|5 |X|
and hence |A| 5 |Y|.
(iii) X ⊆ A ⊆ Y ⇒ |X|  |A|  |Y|
4. If g : Z1 → X is a bijection, then f o g : Z1 → Y is a surjection. By
­Theorem 1.4.5, Y is at most countable.
6. Follows from Corollary 1.4.8.
8. |X| < |P(X)| < |P(P(X))|
10. |P(X)| 5 |2X| 5 |2Y| 5 |P(Y)|, where 2 5 {0, 1}.
12. # 5 {C : C is a circle with radius for 1/n some n  Z1}.
14. Define f : Z1 → Z1 2 In by f (a) 5 a 1 n. Then, f is a bijection.
16. Let Pn be the set of polynomials over Q of degree < n. Then, Pn  Qn and
the set P of all polynomials over Q is the unions of all Pn, n  Z1. |Pn| 5
|Qn| 5 |Q| 5 |Z1| and | P | | ∪  Pn |  | Z | .
nZ

18. If A and T are the sets of algebraic numbers and transcendental numbers,
respectively, then A ∪ T 5 R. If T is countable, than A ∪ T (5 R) is
countable, since A is countable.
20. The given set is bijective with Q 3 Q  Z1.

CHAPTER 2
EXERCISE 2(A)

2. Apply second principle of induction (Theorem 2.1.4) to prove that 2n 1

1 < 2n21 for all 5  n  Z1.
4. This is trivial when n 5 1 or r 5 n. Use Theorem 2.1.3 and the identity
n  n  n 1
    for all 1  r < n.
 r  r 1  r 1
6. If p1, p2, …, pn are all the primes, then, by Theorem 2.1.6, there exists
a prime p (5 pi) dividing p1p2 … pn11, so that p divides 1, which is
an absurd.

Q001-Algebra-111001_Answers.indd 7 9/16/2011 10:00:50 AM

 A-8   Algebra – Abstract and Modern

8. Let A be the set of all positive integers which cannot be expressed in

the required form. If A is nonempty, then A has smallest member, say a.
Since a cannot be prime and a > 1 (note that 1  A), a 5 bc, where 1 < b
< a and 1 < c < a. By the smallest property of a, b and c  A and hence
a  A, a contradiction. Therefore, A is empty.
10. Use Corollary 2.1.3 and Ex. 8 above.

12. Let a  p1r1 p2r2  ptrt and b  p1s1 p2s2  ptst , where pi’s are distinct primes
and ri  0, si  0.

By interchanging pi’s if necessary, we can assume that ri  si for 1  i 

k and si  ri for k < i < t.

Then, l.c.m. {a, b}  p1s1  pksk pkrk11  ptrt

and g.c.d. {a, b} = p1r1  pkrk pksk++11  ptst .

t
Now, ab   piri +si  l.c.m. {a, b}  g.c.d. {a, b}.
i1

14. Suppose that n > 1 and n is not a prime. Let p be the least prime dividing
n. Then, pm 5 n for some m > 1. Any prime dividing m should divide n
and hence p  m. Now, p2  pm 5 n and hence p  n .

16. Follows from the fact that a positive integer n divides a and b if and only
if n divides b and d (since a 5 bc 1 d and a 2 bc 5 d).

18. Use Theorem 2.1.8.

20. 1273 5 969 1 304

969 5 3  304 1 57
304 5 5  57 1 19
57 5 3 ? 19.
Therefore, g.c.d. {969, 1273} 5 19 5 304 2 5  57
5 304 2 5  (969 2 3  304)
5 16  304 2 5  969
5 16(1273 2 969) 2 5  969
5 16  1273 2 21  969; x 5 221
and y 5 16

22. Use Theorem 2.1.8.

24. As in Ex. 12 above,

Q001-Algebra-111001_Answers.indd 8 9/16/2011 10:00:51 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-9

t
a
c  ∏ piri si
(a, b) ik1
k
b
       and d  ∏ pisi ri .
(a, b) i1

Since p1, …, pk, …, pt are all distinct primes, (c, d) 5 1.

EXERCISE 2(B)

2. (i) {x : 3x  5 (mod 7)} (since 5 divides 15, 25 and 35)

5 {x : x  5 ? 5 (mod 7)} (since 5 ? 3  1 (mod 7))
5 {x : x  4 (mod 7)} 5 4 1 7Z
(ii) 4 1 7Z
(iii) The empty set, by Theorem 2.2.6.
(iv) 7 1 8Z
(v) 12 1 17Z
(vi) Empty set, by Theorem 2.2.6.

4. Use Theorem 2.2.8 for (i) and (ii) and, for others, use Theorem 2.2.9
(note that m7 5 5, m11 5 10, m17 5 12, m19 5 2, m23 5 7 and m29 5 3).
6. (i) For any i  j  {0, 1, …, n21}, i  j (mod n) and, since ai  i 1
nZ, ai  i (mod n) and hence ai  aj (mod n).
(ii) Clear.
(iii) For any a  Z, choose i such that a 5 qn 1 i, 0  i  n 2 1 and
then a  i  ai (mod n) and hence a  ai (mod n).
8. Let m, m 1 1, …, m 1 (n 2 1) be n consecutive integers, where m  Z.
Choose 0  i  n 2 1 such that m  i (mod n). Then, m 1 1  i 1 1
(mod n), …, m 1 k  i 1 k (mod n) for all k. This implies that {m, m 1
1, …, m 1 (n 2 1)} is a transversal for congruence modulo n.
10. If n 5 a1a2 … ar and m 5 as(1)as(2) … as(r), where s is any rearrangement
r r
of {1, 2, …, r}, then 3 divides n ⇔ 3 divides  ai   as (i ) ⇔ 3 divides
i1 i1
m.
12. 3 divides 12x 527846531 ⇔ 3 divides 1 1 2 1 x 1 5 1 2 1 7 18 1 4
1 6 15 1 3 11
⇔ 3 divides 44 1 x
⇔ x  {1, 4, 7}.

Q001-Algebra-111001_Answers.indd 9 9/16/2011 10:00:51 AM

 A-10   Algebra – Abstract and Modern

Similarly, 9 divides the number ⇔ 9 divides 44 1 x

⇔ x 5 1.
14. (i) a  b (mod m, mod n) ⇔ m and n divide a 2 b
⇔ mn divides a 2 b (since (m, n) 5 1)
(ii) c  a, d  a (mod n) ⇒ c  d (mod n)
Similarly, c  b (mod m) and d  b (mod m) ⇒ c  d (mod m).
Therefore, by (i), c  d (mod mn).

EXERCISE 2(C)

2. Let a, b  Q, a < b and A 5 {r  Q : a < r < b}.

Define f : A → Q1 by f (r )  bra
r
and g : Q1 → A by g ( s)  a1sb
s .
Then, f ο g  I Q+ and g o f 5 IA and hence f is a bijection. Choose c 
Q such that a < c < b. Then, (a, b) 5 (a, c) ∪ c ∪ (c, b)  Q2 ∪ {0} ∪
Q1 5 Q (since (a, c)  Q1  Q2 and (c, b)  Q1).
4. Imitate the proof of Ex. 2 above to prove that (a, b)  R  (c, d) and
(a, b)Q  Q  (c, d)Q.
6. Let  > 0 be given. Choose n0  Z1 such that (2 / )  n0 . Then, for any
n, m  n0, we have
1 1 1 1 1 1  
−        .
n m n m n0 n0 2 2

Thus, {1/n} is a Cauchy sequence.

8. Clearly, ~ is symmetric and reflexive on CS(Q). If |an 2 bn| → 0 and
|bn 2 cn| → 0, then
0  |an2 cn|  |an 2 bn| 1 |bn 2 cn| → 0.
Therefore, ~ is transitive also.
10. Let r  R. For each n  Z1, choose an  Q such that r  (1/ 2n)  an  r  (1/ 2n).
r  (1/ 2n)  an  r  (1/ 2n). Then, an → r and {an}  CS(Q). If {bn}  CS(Q) such that
bn → r, then
|an 2 bn|  |an 2 r| 1 |r 2 bn| → 0
and hence {an} ~ {bn}.

Q001-Algebra-111001_Answers.indd 10 9/16/2011 10:00:52 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-11

12. {a }  {b} ⇒ {a} ~ {b} ⇒| a  b |  0 ⇒ a  b.

14. | ( an  bn )  ( an  bn ) |  | an  an |  | bn  bn | → 0

| an bn  an bn |  | an (bn  bn )  ( an  an )bn |

 | an || bn  bn |  | an  an | | bn | → 0,

since we can find real numbers K and K9 such that |an|  K and | bn |  K 
for all n.

EXERCISE 2(D)

2. If a < b, then b < a is not possible. The number of pairs (a, b) with a <
n( n1)
b can be atmost n 2n  2 . Therefore, the number of partial orders
2

n ( n1)
on X can be at most 2 2 .
4. For any a, b  X, the set {a, b} has least element, that is, a  b or b  a.
6. Let each (Xi, ) be well ordered. Consider
A⊆X5X 3X 3…3X. 1 2 n

Put A1 5 {x1  X1 : (x1, x2, …, xn )  A for some xi  Xi, i  2}.

Then,   A1 ⊆ X1 and A1 has a least element, say a1. Next, put
A2 5 {x2  X2 : (a1, x2, x3, …, xn)  A for some xi  Xi, i  3}

Then,   A2 ⊆ X2 and A2 has a least element, say a2.

Again, consider A3 5 {x3  X3 : (a1, a2, x3, x4, …, xn)  A for some xi 
Xi, i  4}. Continue this procedure to obtain a1, a2, a3, …, an. Then,
(a1, a2, …, an) will be the least element in A.

EXERCISE 2(E)

2. Let A be an m 3 n matrix and B be an r 3 s matrix. AB is defined if and

only if n 5 r, and BA is defined if and only if s 5 m. Therefore, AB and
BA are defined if and only if B is an n 3 m matrix (or Bt is an m 3 n
matrix).
4. Let A, B and C be m 3 n, r 3 s and t 3 u matrices, respectively.
A(BC) is defined ⇔ BC is defined and A(BC) is defined
⇔ s 5 t and n 5 r
⇔ AB and (AB)C are defined.

Q001-Algebra-111001_Answers.indd 11 9/16/2011 10:00:53 AM

 A-12   Algebra – Abstract and Modern

Let A 5 (aij), B 5 (bij) and C 5 (cij) and suppose that A(BC) (and hence
(AB)C) is defined. For any 1  i  m and 1  j  u,

n  n 
the ijth entry in A( BC )  ∑ aip ∑ bpq cqj 
p1  q1 
n n
 ∑∑ aip (bpq cqj )
q1 p1

n  n 
 ∑ ∑ aip bpq  cqj
q1 
 p1 

5 the ijth entry in (AB)C.

Thus, A(BC) 5 (AB)C. Similarly, others can be proved.

6. Let A 5 (aij) and Sa 5 (sij). Then, sij 5 a or 0 according as i 5 j or i  j.
An ijth entry in SaA is
n

∑s a
r1
ir rj  aaij , for any 1  i, j  n.

Therefore, SaA 5 aA and similarly, ASa 5 aA.

8. (i) Let A 5 (aij) and B 5 (bij). Then, AB and (AB)t are m 3 r and r 3
m matrices, respectively. For any 1  i  r and 1  j  m,

ijth entry in (AB)t 5 jith entry in AB

n
 ∑ a js bsi
s1
n
 ∑ bis asj
s1

5 ijth entry in BtAt,

where bis and asj are the isth and sjth entries in Bt and At, respectively.
(ii) If B1 and B2 are such that AB1 5 I and B2A 5 I, then B1 5 IB1 5
(B2A)B1 5 B2(AB1) 5 B2I 5 B2.
(iii) (AB)(B21A21) 5 A(BB21)A21 5 AIA21 5 AA21 5 I and, similarly
(B21A21)(AB) 5 I.
(iv) If AB 5 I 5 BA, then AtBt 5 (BA)t 5 It 5 I 5 BtAt and hence
(At)21 5 Bt 5 (A21)t.
10. Straight-forward verification.

Q001-Algebra-111001_Answers.indd 12 9/16/2011 10:00:54 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-13

12. If A 5 (aij) is a skew-symmetric matrix, then A 5 2At and hence aii 5

2aii, so that aii 5 0.
14. Let C 5 1 (A 1 At) and D 5 1 (A 2 At). Then, C is a symmetric matrix
2 2
and D is a skew-symmetric matrix. Also, A 5 C 1 D.

EXERCISE 2(F)

2. Use induction on the order of the matrix (if A is an upper triangular

matrix, then det A 5 a11A11).
4. Subtract x1 times the (r 2 1)th column from the rth column to obtain
1 0 0  0 
 
1 x  x x2 ( x2  x1 )  x2 ( x2  x1 )
n2

det A  det  2 1

     
 
1 xn  x1 xn ( xn 1)  xn ( xn  x1 )
n2

and now expand with respect to the 1st row, factor out xj 2 x1 and apply
induction on n.
6. A2 5 A ⇒ (det A)2 5 det A ⇒ det A 5 0 or det A 5 1. If A is nonsingular,
then det A  0.
8. Am 5 On3n ⇒ (det A)m 5 0 ⇒ det A 5 0.
10. Follows from Theorems 2.6.2 and 2.6.8.
12. By Ex. 11, det A 5 2det A and hence det A 5 0.
14. (i) det(AB) 5 det A  det B 5 det B  det A 5 det(BA).
(ii) det(A  At) 5 det A  det At 5 det A  det A.
(iii) det(ABA21) 5 det A  det B  det A21
5 det A  det B  (det A)21 5 det B.

CHAPTER 3
EXERCISE 3(A)

2. In Exercise 1(A), Z4 5 {0, 1, 2, 3} and, in Exercise 1(F ), S 5 {1, i,

21, 2i}
0 is the identity 1 is the identity
1 14 1 5 2 i ? i 5 21
1 14 1 14 1 5 3 i ? i ? i 5 2i

Q001-Algebra-111001_Answers.indd 13 9/16/2011 10:00:54 AM

 A-14   Algebra – Abstract and Modern

Both the tables are same, if we rename 14 by ?, 0 by 1, 1 by i, 2 by 21,

and 3 by 2i.
4. The table in (3) is given by

* a b c d e f
a d e a c a b
b e f d c f d
c a d c a d c
d c c a e d b
e a f d d b d
f b d c b d a

(b*(d * a)) * (c*(b * a)) 5 (b * c) * (c * e) 5 d * d 5 e

((a * b) * c) * d 5 (e * c) * d 5 d * d 5 e
(a * b) * (c * d) 5 e * a 5 a

6. Let S3 be the set of bijections of {1, 2, 3} onto itself and o be the com-
position of mappings in S3. Then, o is associative, but not commutative;
for consider f and g defined by f (1) 5 1, f (2) 5 3, f (3) 5 2 and g(1) 5
2, g(2) 5 1 and g(3) 5 3. Then, (f o g)(1) 5 3 and (g o f )(1) 5 2 and
hence f o g  g o f.
8. No, the statement is false. Consider the operation * on {a, b} given by
the table

* a b
a b a
b a a

Here, (a * a) * b 5 b * b 5 a
  and a * (a * b) 5 a * a 5 b.

Therefore, * is not associative. But it is commutative, since a * b 5 b * a.

10. Yes, the statement is true.
It is trivial when S has exactly one element. To prove the converse, define
a * b 5 b for all a and b  S. If * is commutative, then a 5 b for all a,
b  S and hence | S | 5 1.

Q001-Algebra-111001_Answers.indd 14 9/16/2011 10:00:54 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-15

2
12. The number of binary operations on an n-element set is n2n . The number
n +n
of commutative binary operations on an n-element is n 2 .
∴ The number of noncommutative binary operations on an n-element
2 n2 +n
set is nn  n 2
and, on a 3-element set, it is 39 2 36 5 36 3 26.
14. By definition of 1, e 1 x 5 x 5 x 1 e for all x  S9.

EXERCISE 3(B)

2. All elements in Z10

(i)
(ii)
1, 3, 7, 9
(iii)
1, 5, 7, 11, 13, 17, 19, 23, 25, 29, 31, 35
(iv)1, 3
(v)All bijections
(vi)All elements in P(X)
(vii)All elements in R1
(viii)1 only.
1 0 1 0 1 0  1 0 
4. Let e   , a   , b    and c   . Then, e
0 1  0 1  0 1 0 1
is the identity in G 5 {e, a, b, c}, a2 5 e 5 b2 5 c2, ab 5 c 5 ba, ac 5
b 5 ca, bc 5 a 5 cb.
6. x * x 5 x 5 e * x ⇒ x 5 e.
8. The map e : X → G defined by e ( x ) = e for all x  X is the identity ele-
ment in GX and, for any f  GX, the map 2f : X → G, defined by (2f )(x)
5 2f (x) for all x  X, is the inverse of f in GX.
10. Zn 5 {0, 1, 2, …, n 2 1} and 0 has no inverse in (Zn, n) and hence (Zn, n)
is not a group.
12. For any 0 < m < p, (m, p) 5 1 and hence there exist a and b  Z such that
am 1 bp 5 1. If we choose r  Z1 such that a 5 qp 1 r, 0 < r < p, then
rm 1 qp 1 bp 5 1 and hence rm  1 (mod p), so that r is the inverse of
m in (Gp, p).
14. Yes, take p 5 5 in Ex. 12 above.
16. (1) ⇒ (3): If n is not a prime, then n 5 ab for some 0 < a < b and 0 < b
< n and hence a n b 5 0, a contradiction to (1).

Q001-Algebra-111001_Answers.indd 15 9/16/2011 10:00:55 AM

 A-16   Algebra – Abstract and Modern

(3) ⇒ (2): Follows from Ex. 12 above.

(2) ⇒ (4): (2) ⇒ (1) ⇒ (3) ⇒ (4)
(4) ⇒ (1): 0 < a < n and 0 < b < n ⇒ (a, n) 5 1 5 (b, n)
⇒ (ab, n) 5 15 a n b  0
18. This is an n-element group and is equal to {e, a, a2, …, an21}, where a is
the rotation by angle 2np about the origin.
20. 0 is the identity and, for any r  Q, −r is the inverse of r in Q/Z.

EXERCISE 3(C)

2. Let S be any finite set with more than one element and define x * y 5 y
for all x and y  S. Then, (S, *) is a finite semigroup which satisfies the
left cancellation law and does not satisfy the right cancellation law.
4. Imitate the proof of Theorem 3.3.2.
6. Let (S, ) be a finite semigroup and a  S. Then, {an : n  Z1} ⊆ S and
hence there exist m < n in Z1 such that am 5 an. Put a 5 n 2 m  Z1.
Then,

an 5 am1a 5 am
a2m 5 am  am 5 am  am1a 5 a2m1a
By induction on k, akm 5 akm1a for all k  Z1. Also, akm12a 5 akm1a  aa 5
akm  aa 5 akm1a 5 akm.
By induction on r,

akm1ra 5 akm for all k and r  Z1.

In particular, aam1ma 5 aam; that is, (aam)2 5 aam.

8. Let G be a finite group and |G| 5 2m. For any a  G, let Aa 5 {a, a21}.
Then, |Aa|  2 for all a  G. Also, Aa ∩ Ab 5  or Aa 5 Ab for each
pair of all elements a and b in G. {Aa : a  G} is a partition of G and
hence

2m 5 |G| 5 ∑|Aa|.

Since Ae 5 {e}, there must be atleast one more a  e such that |Aa| 5 1;
that is, a 5 a21 or a2 5 e and a  e.

Q001-Algebra-111001_Answers.indd 16 9/16/2011 10:00:56 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-17

10. For each a  G, {an : n  Z} is a subset of G and hence finite, so that,

there exist m < n in Z such that am 5 an or an2m 5 e and n 2 m  Z1.
Therefore, for each a  G, there exist a positive integer na such that
a na  e. If n is the product of all na, a  G, then an 5 e for all a  G.

12. See Ex. 8 of Exercise 3(B).

14. If |G| 5 1, then G 5 {e}. If |G| 5 2, then G 5 {e, a}, where a2 5 e. Let
|G| 5 3. If e  a  G, then a2  e (otherwise, a2 5 e implies that there
exists b  G 2 {e, a} such that ab, b, e, a are distinct elements of G)
and hence {e, a, a2} 5 G. Let |G| 5 4. If a2 5 e for all a  G, then a 5
a21 for all a  G and hence ab 5 (ab)21 5 b21a21 5 ba for all a and b 
G. Therefore, we can assume that there exists a  G such that a2  e.
If a3 5 e, then there exists b  G 2 {e, a, a2} such that ba is an
element of G other than e, a, a2 and b so that |G| > 4. Thus, a3  e
and hence G 5 {e, a, a2, a3}, which is abelian. Next, let |G| 5 5. We
can prove that, for any e  a  G, G 5 {e, a, a2, a3, a4} which is
abelian.

EXERCISE 3(D)

2.

11 1 2 3 4 5 6 7 8 9 10

1 1 2 3 4 5 6 7 8 9 10

2 2 4 6 8 10 1 3 5 7 9

3 3 6 9 1 4 7 10 2 5 8

4 4 8 1 5 9 2 6 10 3 7

5 5 10 4 9 3 8 2 7 1 6

6 6 1 7 2 8 3 9 4 10 5

7 7 3 10 6 2 9 5 1 8 4

8 8 5 2 10 7 4 1 9 6 3

9 9 7 5 3 1 10 8 6 4 2

10 10 9 8 7 6 5 4 3 2 1

Q001-Algebra-111001_Answers.indd 17 9/16/2011 10:00:56 AM

 A-18   Algebra – Abstract and Modern

4.

14 0 1 2 3

12 0 1 2 0 0 1 2 3

12 0 1 0 0 1 2 1 1 2 3 0

0 0 1 1 1 2 0 2 2 3 0 1

1 1 0 2 2 0 1 3 3 0 1 2

15 0 1 2 3 4

0 0 1 2 3 4
* e a b c
1 1 2 3 4 0
e e a b c
2 2 3 4 0 1
a a e c b
3 3 4 0 1 2
b b c e a
c c b a e 4 4 0 1 2 3

6. The bijections on a 3-element set form a nonabelian group with 6 ele-

ments. Any group with less than 6 elements is abelian (see Ex. 14 of
Exercise 3(C)).
8. Zn, the group of integers modulo n.
1 0
10. Put A 5 a, B 5 b, AB 5 c and    e. Then, a2 5 b2 5 c2 5 2e,
0 1
BA 5 2c, a4 5 b4 5 c4 5 e and Q8 5 {AnBm : n, m  Z} 5 {e, 2e, a,

2a, b, 2b, c, 2c}.

CHAPTER 4
EXERCISE 4(A)

2. Recall that, for any n 3 n matrices A and B, |AB| 5 |A||B|.

(i) No
(ii) Yes
(iii) Yes
(iv) Yes

Q001-Algebra-111001_Answers.indd 18 9/16/2011 10:00:56 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-19

(v) Yes
(vi) Yes
(vii) No
(viii) Yes.

4. If G 5 <a>, then G 5 <a21 > also and therefore a 5 a21 so that a2 5 e

and | G |  2.
6. Zn 5 <d> ⇔ ad  1 (mod n) for some a  Z
⇔ 1 5 ad 1 bn for some a and b  Z
⇔ (d, n) 5 1.
8. x, y  Ca ⇒ ax 5 xa and ay 5 ya
⇒ x21a 5 ax21 and axy 5 xay 5 xya
⇒ x21 and xy  Ca
10. Z (G )  ∩ Ca
aG

12. Any subgroup of Zn is of the form <d> for some divisor d of n.

14. x  axa21 is a bijection of H onto aHa21.
16. Follows from Theorem 4.1.1.

EXERCISE 4(B)

2. First note that any cyclic group is at most countable.

(i) No, since R 2 {0} is uncountable.
(ii) (P(X), 1) is cyclic ⇔uP(X)u  2 (since A 1 A 5 )
⇔ |X|  1
(iii) No, since Q8 is not abelian and a cyclic group is always abelian.
(iv) No, if Q1 5 <a>, then 1a  a n for some n  Z and hence a 5 1,
so that Q1 5 <1> 5 {1}, which is absurd.
(v) No, since R1 is uncountable.
(vi) Yes, since cos 2np  i sin 2np generates.
4. (i) <7> 5 Z18, since (7, 18) 5 1
(ii) In Z20, <5> 5 {0, 5, 10, 15}
(iii) In Z12, <3> 5 {0, 3, 6, 9}
(iv) In Z16, <3> 5 Z16, since (3, 16) 5 1

Q001-Algebra-111001_Answers.indd 19 9/16/2011 10:00:57 AM

 A-20   Algebra – Abstract and Modern

(v) In C 2 {0}, <2> 5 {2n : n  Z}

n
(vi) In (R , ),  2  {2 2 : n  Z}
(vii) In (R, ),  2  {n 2 : n  Z}
(viii) In C {0},  2  {(i 2) n : n  Z}
(ix) In any group G, <e> 5 {e}.
(x) In (3Z, 1), <12> 5 {12n : n  Z}.
6. Not necessary.
8. Let G be an abelian group, a and b  G and O(a)  m  p1r1  pkrk pk+
rk+1 ru
1  pu
s1 sk sk+1 su
and O(b)  n  p1  pk pk+1  pu , where p1, …, pu are distinct primes
and ri and si  Z such that 0  ri  si for 1 i  k and 0  si  ri for k
< i  u. Then,

l.c.m.{m, n}  p1s1  pksk pkrk11  puru .

Put m1  p1r1  pkrk and n1  pk+

sk+1 su
1  pu .

Let c  a m1 and d  b n1 . Then,

O( a)
O(c)   pkrk11  puru
m1
O(b) s1
and O(d )  p1  pksk .
n1
     
O(c) and O(d) are relatively prime and therefore, by Ex. 7, G has a cyclic
subgroup of order O(c) ? O(d) 5 l.c.m.{m, n}.
10. x  <a> ∩ <b> ⇒ x 5 ar 5 bs for some 0  r < O(a) and 0  s < O(b)
O(a) O(b)
⇒  O(x )  (by Theorem 4.2.6)
(O(a), r ) (O(b), s)
⇒ O(x) divides O(a) and O(b)
⇒ O(x) 5 1 ⇒ x 5 e.
12. 16 20 4 5 4, 16 20 a 5 a for all a  G 5 {4, 8, 12, 16}. Therefore, 16 is
the identity in (G, 20). This is a cyclic group generated by 8 and also by
12 (12 is the inverse of 8).
14. No.
16. Let G 5 <a>, H 5 <b>, O(a) 5 m and O(b) 5 n. Suppose that (m, n) 5
1. Then, am1bn 5 1 for some a and b in Z. Let (x, y)  G 3 H. Then,
x 5 ar and y 5 bs for some 0  r < m and 0  s < n. Put t 5 brn 1 asm.

Q001-Algebra-111001_Answers.indd 20 9/16/2011 10:00:58 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-21

Then,

     at2r 5 at2r(am1bn) 5 am(as2ar) 5 e

        and bt2s 5 bt2s(am1bn) 5 bn(br2bs) 5 e
     and hence (x, y) 5 (ar, bs) 5 (at, bt) 5 (a, b)t.

Thus, (a, b) is a generator of G 3 H and hence G 3 H is a cyclic group.

Conversely, suppose that G 3 H is cyclic and G 3 H 5 <(c, d)>. Then,
O(c, d) 5 |G 3 H| 5 mn. If r 5 l.c.m.{m, n}, then (c, d)r 5 (cr, dr) 5 (e,
e) and hence O(c, d) divides r; that is, mn divides r. This is possible only
when r 5 mn or, equivalently (m, n) 5 1.

18. <an> ∩ <am> 5 <ar>, where r  l.c.m.{m, n} modulo 24.

20. {0}, <2>, <3>, <4>, <6>, <8>, <12> and Z24 are all the subgroups of Z24.

{0}, <2>, <3>, <5>, <6>, <10>, <15> and Z30 are all the subgroups
of Z30.

22. {0}, <5>, <25>, <125> and Z625 are all the subgroups of Z625.

24. Any infinite cyclic group has infinitely many subgroups. Therefore, if G
has finitely many subgroups, then <a> is finite for any a  G and, since
G  ∪ < a >, G is finite.
a∈G

26. O(a) 5 O(a21) for any a  G. Also, if O(a) and O(b) divide n, then
O(ab) divides l.c.m. {O(a), O(b)} and hence O(ab) divides n.

28. If O(a) 5 m, then 12 < m  24 and hence O(a) 5 24 and a is a generator

of G.

30. O(a) 5 O(a21) and O(ab) is a divisor of l.c.m.{O(a), O(b)}.

EXERCISE 4(C)

2. Define f : [0, 1) → R (5 {x 1 Z : x  R}) by

Z
f (r) 5 r 1 Z.
If 0  r  s < 1 and f (r) 5 f (s), then s 2 r  Z and hence s 2 r 5 0.
Therefore, f is an injection. For any x  R, let [x] be the largest integer
 x and put r 5 x 2 [x]. Then, 0  r <1 and
f (r) 5 r 1 Z 5 x 2 [x] 1 Z 5 x 1 Z (since [x]  Z).
Thus, f is a surjection also.

Q001-Algebra-111001_Answers.indd 21 9/16/2011 10:00:58 AM

 A-22   Algebra – Abstract and Modern

4. Ha21 5 Hb21 ⇔ a21b  H ⇔ aH 5 bH.

6. If a, b and c are the transposions (1 2), (2 3) and (3 1), respectively,
then
aH 5 bH 5 cH 5 {a, af, af21} 5 Ha 5 Hb 5 Hc
and eH 5 fH 5 f 21H 5 H 5 He 5 Hf 5 Hf 21
8. f (x) 5 x 5 g(x) ⇒ f21(x) 5 x and f (g(x)) 5 f (x) 5 x
Therefore, Ax is a subgroup of S(X). Since the identity e  Ax, y, Ax, y is not
a subgroup of S(X).
10. xyA 5 xzA ⇒ (xy)21 xzA ⇒ y21z  A ⇒ yA 5 zA
Ayx 5 Azx ⇒ (yx) (zx)21 A ⇒ yz21  A ⇒ Ay 5 Az
12. xA is a subgroup ⇒ e  xA ⇒ x  A ⇒ xA 5 A.
14. If A and B are subgroups of G and x and y  G such that xA 5 yB, then
x  xA 5 yB and hence xB 5 yB so that xA 5 xB and A 5 B.

EXERCISE 4(D)

2. A ∩ B is a subgroup of A and hence |A ∩ B| is a divisor of |A|. If |A| is

prime and A ∩ B  {e}, then |A ∩ B| 5|A| and hence A ∩ B 5 A so that
A ⊆ B.
4. iG(A) 5 A ⇔ xA 5 yA for all x and y  G 2 A ⇔ x21y  A
6. For any x  G, x(A ∩ B) 5 xA ∩ xB. If x1A, …, xnA are all the distinct
left cosets of A in G and y1B, …, ymB are all the distinct left cosets of B
in G, then {xiA ∩ yjB : xiA ∩ yjB  } are all the distinct left cosets of A
∩ B in G and hence iG(A ∩ B)  n  m 5 iG(A)  iG(B).
8. AB is a subgroup of G and A and B are subgroups of AB. For any x 5 ab,
a  A, b  B,
x(A ∩ B) 5 xA ∩ xB 5 bA ∩ aB.
If G 5 AB in Ex. 6 above, then xiA ∩ yjB   for all i and j and hence
iAB(A ∩ B) 5 iAB(A)  iAB(B).
10. O(b) 5 1 or 31.
12. By Ex. 8 in Exercise 3(C), there is a  e in G such that a2 5 e. Let A 5
{e, a}. Then, iG(A) 5 n. If b  a and O(b) 5 2, then b  A and 2 divides
iG(A) 5 n, which is a contradiction, since n is odd.
14. The group of bijections on a 3-element set is a nonabelian group of order
6, see Ex. 14 in Exercise 3(C).

Q001-Algebra-111001_Answers.indd 22 9/16/2011 10:00:59 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-23

EXERCISE 4(E)

2. b  B and x  A ∩ B ⇒ bxb21  B ∩ A (since A is normal in G).

4. Let X 5 {1, 2, 3} and f : X → X be defined by f (1) 5 2, f (2) 5 3 and
f (3) 5 1. Then, {e}, {e, f, f21} and S(X) are the only normal subgroups
of S(X).
6. For any x  G and a  A, xaA 5 xA and hence Axa 5 Ax so that
xax21  A.
8. Let G 5 D4, the group of symmetries a square (see Example 3.2.8). Let s
be the clock-wise rotation about the centre of the square through an angle
p and d be the reflection about the diagonal D . Let A 5 {e, d} and B 5
2 1
{e, d, s2, s2d}. Then, B is a normal subgroup of G (since iG(B) 5 2), A is
a normal subgroup of B and A is not normal in G (since s21as  A).
10. Let | A | 5 m and iG ( A)  n  G . Since (m, n) 5 1, there exist integers
A
a and b such that am 1bn 5 1. Clearly xm 5 e for all x  A. Also,
xm 5 e ⇒ xA 5 (xA)am1bn 5 xmaA  (xA)nb 5 A ⇒ x  A.
12. a  A and x  C(A) ⇒ xax21 5 a  A.
14. C(A) ⊆ N(A). Equality may not hold.
16. For any a  A and x  G, xa  xa  A and hence xax  Aa21 5 A. So
that xax21 5 (xax)(x21)2  AA 5 A.
18. Straight forward verification.
20. Let f  Z(S(X)) and x  X. If f (x)  x, then choose y  X, such that y
 x and y  f (x) (since |X|  3). Define g : X → X by g(x) 5 y, g(y) 5
x and g(z) 5 z for all z  X 2 {x, y}. Then, (f o g)(x) 5 f (y) and (g o f )
(x) 5 f (x)  f (y) and hence f o g  g o f.
22. ( x1 , x2 )(e1 , y2 )( x1 , x2 )1  ( x1e1 x11 , x2 y2 x2 1 )  (e1 , x2 y2 x2 1 )  {e1} G2 .

EXERCISE 4(F )
|G|
2. A 5 {e, a3, a6, a9, a12}. G   3. Let A1 5 a5A and A2 5 a7A. Then,
G  { A, A , A }. A | A|
A 1 2

 A A1 A2
A A A1 A2
A1 A1 A2 A
A2 A2 A A1

Q001-Algebra-111001_Answers.indd 23 9/16/2011 10:00:59 AM

 A-24   Algebra – Abstract and Modern

4. See Ex. 16 in Exercise 4(E). (xA)2 5 x2A 5 A for all x  G. Therefore,

G/A is abelian.
6. (i) is clear, by Lagrange’s theorem.
(ii) (xA)n 5 A and hence xn  A.
(iii) If O(a) 5 m, then (aA)m 5 amA 5 A and hence O(aA) divides m.
8. If A 5 <x1, …, xn> and G  y1 A, …, ym A , then G 5 <{xi yj : 1  i  n
A
and 1  j  m}.
10. Let X 5 {a, b, c}. Then, {e}, N 5 {e, (a b c), (c b a)} and S(X) are all the
normal subgroups of S(X). S(X)/N has only two members N and N(a, b).

CHAPTER 5
EXERCISE 5(A)

2. If f (ab21) 5 f (a)f (b)21 for all a, b  G, then f (e) 5 f (aa21) 5 f (a)f (a)21
5 e9, f (b21) 5 f (eb21) 5 f (e)f (b)21 5 e9f (b)21 5 f (b)21 and f (ab) 5
f (a(b21)21) 5 f (a)(f (b)21)21 5 f (a)f (b) for all a and b  G. The converse
is clear.
4. G is abelian ⇔ ab 5 ba for all a and b  G
⇔ (ab)21 5 b21a21 5 a21b21 for all a, b  G
⇔ f (ab) 5 f (a)f (b)
6. For any homomorphism f : Z → Z2, f (n) 5 nf (1) for all n  Z and
hence
f (1) 5 0 ⇒ f (n) 5 0 for all n  Z
and f (1) 5 1⇒ f (n) 5 0 for all even n and
         f (n) 5 1 for all odd n
(since 1 1 1 5 0 in Z2)
8. Let f : Z → Z be a nontrivial endomorphism. Then, f (n) 5 nf (1) for all
n  Z and hence f (1)  0 so that
f (n) 5 f (m) ⇒ nf (1) 5 mf (1) ⇒ n 5 m.

10. If f : Z → R is a homomorphism and f (1) 5 a, then f (n) 5 nf (1) 5 na

5 fa(n) for all n  Z and hence f 5 fa.
12. If O(a) 5 n, then f (a)n 5 f (an) 5 f (e) 5 e9 and hence O(f (a)) is finite
and is a divisor of n (by Theorem 4.2.4 (3)).

Q001-Algebra-111001_Answers.indd 24 9/16/2011 10:01:00 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-25

14. If |G| 5 p, a prime, then |ker f | divides |G| 5 p and hence |ker f | 5 1 or
p, so that ker f 5 {e} or G.
16. f is a homomorphism ⇒ aabb 5 f (1 1 1, 1 1 1) 5 f (1, 1)f (1, 1) 5 abab
⇒ ab 5 ba
f is a homomorphism ⇔ ab 5 ba.
18. Use induction on n to prove that f (n 1 m) 5 f (n)f (m) (that is, an1m 5
anbm) for all n, m  Z. ker f 5 {0} if a is of infinite order and ker f 5 nZ
if O(a) 5 n.
20. If f : Q8 → Z2 is a homomorphism such that f (i) 5 0 and f (j) 5 1, then
f (a) can be evaluated for any a  Q8.
22. Follows from the fact that ab > 0 ⇔ (a > 0 and b > 0) or (a < 0 and b <
0) for any a, b  R 2 {0}. Also, for any x, y  G, xy 5 1 ⇔ x 5 1 5 y
or x 5 21 5 y.
24. If f : Z8 → Z24 is defined by f (a) 5 3a for all a  Z8, then f is a mono-
morphism.
26. If f is an isomorphism of (Z, 1) onto (Z, 1), then f (n) 5 nf (1) for all
n  Z and f (1) 5 1 or 21 (otherwise f is not surjective) and hence f is
either identity map or the inverse map.
28. (f 1 g)(x) 5 f (x) 1 g(x) for all x. The trivial homomorphism is the
identity in Hom(G, H). Also, if f  Hom(G, H), then 2f  Hom(G, H)
and 2f is the inverse of f.
30. g o f is an injection ⇒ f is an injection.
g o f is a surjection ⇒ g is a surjection.

EXERCISE 5(B)

2. In Ex. 1, define f : G → G/N 3 G/M by f (x) 5 (xN, xM). Then, ker f 5

N ∩ M and hence, by the Fundamental Theorem of Homomorphisms,
G/N ∩ M  f (G) which is a subgroup of G/N 3 G/M. If G 5 Z, N 5
nZ and M 5 mZ, then N ∩ M 5 kZ, where k 5 l.c.m.{n, m}. Recall
that Zk  Z/kZ, Zn  Z/nZ and Zm  Z/mZ.
4. If Z27 is a homomorphic image of Z72, then Z27 is isomorphic to a quo-
tient of Z72 and hence there must be a subgroup A of Z72 whose index is
27. This is impossible, since 27 is not a divisor of 72.
6. If f : G → G9 is an epimorphism, then G/ker f  G9
and hence |G9| 5 |G/ker f | 5 iG(ker f ), which is a divisor of |G|.

Q001-Algebra-111001_Answers.indd 25 9/16/2011 10:01:00 AM

 A-26   Algebra – Abstract and Modern

8. (i) ⇒ There is an epimorphism f : Zn → Zm

⇒ Zn/k  Zm for some subgroup k of Zn
⇒ iZn ( k )  m ⇒ m divides | Z n |  n ⇒ (b)

(ii) ⇒ n 5 md for some d  Z1

⇒ The map f : Zm → Zn defined by f (a) 5 ad is a monomorphism
⇒ f is an isomorphism of Zm onto f (Zm) ⊆ Zn ⇒ (c)

(iii) ⇒ There is a monomorphism f : Zm → Zn

⇒ O(1) 5 m in Zm and O(f (1)) 5 m in Zn
⇒ m divides |Zn| 5 n ⇒ (b)

(iv) ⇒ md 5 n for some d  Z1

⇒ The map f : Zn → Zm, defined by f (a) 5 r, where a 5 qm 1 r,
0  r < m, is an epimorphism ⇒ (a)

10. Follows from Ex. 8 above.

EXERCISE 5(C)

2. Let f : G → G9 be a homomorphism of groups. In Ex. 1 above, take

G1 5 G/ker f, G2 5 G9, f2 5 f and f1 : G → G1 be defined by f1(a) 5
a(ker f ). Then, there exists a homomorphism g : G1 → G2 such that
g o f1 5 f2; that is, g o f1 5 f and g is a monomorphism and f1 is an
­epimorphism.

4. Define a(A) 5 f21(A) for any subgroup A of G9 and b(B) 5 f (B) for any
subgroup B of G containing ker f. Then, (a o b)(B) 5 B and (b o a)(A) 5
A and hence a is a bijection of the set of subgroup of G9 onto the set of
subgroups of G containing ker f.

6. Refer Example 3.2.8. Let a be the clock-wise rotation about the centre
of the square through an angle p/2 and b be the reflection about the
diagonal D1. Then, O(a) 5 4, O(b) 5 2 and aba 5 b. The group G8 is
of order 8 and G8 5 {e, a, a2, a3, b, ab, a2b, a3b}. It can be verified that
N1 5 {e}, N2 5 {e, a, a2, a3}, N3 5 {e, a2, b, a2b}, N4 5 {e, a2, ab, ab,
a3b}, N5 5 {e, a2} and N6 5 G8 are all the normal subgroups of G8 and
hence G8/Ni, 1  i  6, are all (up to isomorphism) the homomorphic
images of G8. Also, Z(G8) 5 {e, a2} 5 N5 and N2, N3, N4, N5 and G8 are
all subgroups of G8 containing Z(G8) and hence N2/N5, N3/N5, N4/N5, N5/
N5, and G8/N5 are all the subgroups of G8/Z(G8).

Q001-Algebra-111001_Answers.indd 26 9/16/2011 10:01:01 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-27

8. (i) x  f (A) ∩ f (B) ⇒ f (a) 5 x 5 f (b) for some a  A and b  B

⇒ ab21  ker f ⊆ A ∩ B ⇒ a 5 (ab21)b  B
⇒ x 5 f (a)  f (A ∩ B)
(ii) x  A ker f ⇒ x 5 ab, a  A, b  ker f ⇒ f (x) 5 f (ab) 5
                f (a)  f (A)
⇒ x  f21( f (A)) ⇒ f (x) 5 f (a), a1  A
   ⇒ a121x  ker f ⇒ x 5 a1(a121x)  A ker f.
(iii) Define g : G → G9/[G9, G9] by g(x) 5 f (x)[G9, G9] since f is an
epimorphism, so is g and ker g 5 [G, G]. Therefore, G/[G, G] 
G9/[G9, G9].
10. (i) If x1A, …, xnA are all the distinct left cosets of A in G1, then verify that
f (x1)f (A), …, f (xn)f (A) are all the distinct left cosets of f (A) in G9.
(ii) If A 5 f21(A9) and f (x1)A9, …, f (xn)A9 are all the distinct left cosets of
A9 in G9, then x1A, …, xnA are all the distinct left cosets of A in G.
12. If |G| 5 p and e  a  G, then O(a) divides p and O(a) 5 p and hence
G 5 <a>  Zp.
14. A1/A1 ∩ N  A1N/N 5 A2N/N  A2/A2 ∩ N.

EXERCISE 5(D)

2. 2Z is a proper subgroup of (Z, 1) and f : Z → 2Z, defined by f (a) 5 2a,

is an isomorphism.
4. Ta(A) ⊆ A, where Ta is the inner automorphism of G corresponding to
any a  G.
6. If x and a  G and f (a) 5 a for all f  Aut(G), then f (xax21) 5 f (x)f (a)
f (x)21 5 Tf (x)(a) 5 a 5 Tx(a) 5 xax21 for all f  Aut(G).
8. For Ex. 7, x  x21f (x) is an injection of G into G; G being finite G 5
{y21f (y) : y  G}. In addition, suppose that f 2 5 IG. Then, for any a  G,
a21f (a) 5 f 2(a21f (a)) 5 f (f (a21)a) 5 f ((a21f (a))21)
and therefore, by Ex. 7, f (x) 5 x21 for all x  G. Now,
xy 5 (y21x21) 21 5 f (y21x21) 5 f (y21)f (x21)f (x21) 5 yx,
                 for any x, y  G.

10. f (aba21b21)5 f (a)f (b)f (a)21f (b)21

12. If f : G → G9 is a homomorphism and G 5 <a>, then f (G) 5 <f (a)>.

Q001-Algebra-111001_Answers.indd 27 9/16/2011 10:01:01 AM

 A-28   Algebra – Abstract and Modern

14. For any 0 < i < p, fi : Zp → Zp defined by f (a) 5 ai, is an automorphism of

Zp and Aut(Zp) 5 {f i : 0 < i < p}, which is isomorphic to the multiplica-
tive groups Zp 2{0} modulo p.
16. (m, n) 5 1 ⇔ am 1 bn 5 1 for some a, b  Z
⇔ 1  am (mod n)
If (m, n) 5 1 and am 5 bm, then aam 5 bam and hence a 5 b, so that fm is
automorphism. Conversely, if fm  Aut(G), then <a> 5 G 5 <fm(a)> 5
<am> and hence a 5 aam for some a  Z so that 1  am (mod n).
18. Aut(Z2 3 Z2)  S3
20. Aut(Z2 3 Z3)  Aut(Z6)  Z2
Aut(Z2) 5 {e}, Aut(Z3)  Z2

CHAPTER 6
EXERCISE 6(A)

2. (i) Yes, since Zn is finite.

(ii) No, since f : Z → Z defined by f (a) 5 2a is an injection, but not
a surjection.
(iii) Yes, since Zn is finite.
(iv) Yes, since |X| 5 n → |S(X)| 5 n! and hence S(X) is finite.
(v) Yes, since |X| 5 n → |P(X)| 5 2n
(vi) Yes, refer the proof of Theorem 6.1.1.
4. |X| 5 n < ` ⇒ |S(X)| 5 2n
If X is infinite and Y ⊆ X ⇒ S(Y) ⊆ S(X) and hence 2n < |S(X)| for all n
 Z1, so that S(X) is infinite.
6. Refer proof of Theorem 6.1.2.
8. Use Theorem 6.1.3.
a 1 if a  
10. Define f : R → R by f ( a)   .
a if a  
EXERCISE 6(B)

2. O(a) divides |Sn| 5 n! for any a  Sn

4. O(a) 5 8 5 O(b) and O(c) 5 4

Q001-Algebra-111001_Answers.indd 28 9/16/2011 10:01:01 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-29

6. No, otherwise there is an epimorphism a : S8 → S12 and hence |S8| $ |S12|,

which is an absurd.

1 2 3 4 5 6 7 8
8. a b a1   
7 4 6 5 3 1 8 2
1 2 3 4 5 6 7 8
b c b1   
5 6 7 3 2 1 8 4
1 2 3 4 5 6 7 8
c a c1   
4 1 8 7 2 5 3 6

 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 
10. e, ,  ,  ,  ,
 2 1 4 3 4 3 2 1 3 4 1 2 2 3 1 4 
 
 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 
S4    ,  ,  ,   ,
 3 1 2 4 2 4 3 1 4 1 3 2 3 2 4 1 
 
 1 2 3 4 1 2 3 4 1 2 3 4 
  , 
 , 
 
 
 4 2 1 3 1 3 4 2 1 4 2 3 

EXERCISE 6(C)

2. (i) Yes, (1 4 2 6 5 3 8 7)
(ii) No
(iii) Yes, (1 2 3 4 5 6 7 8)
(iv) No
4. (i) 12
(ii) 2
(iii) 6
(iv) 7.
6. Let |f | 5 Supp f and f and g be disjoint.

i  |f | ∪ |g| ⇒ f (i) 5 i 5 g(i) ⇒ f (g(i)) 5 i ⇒ i  | fg|.

i  |f | ⇒ f (i)  i ⇒ g(i) 5 i ⇒ f (g(i)) 5 f (i)  i ⇒ i  |fg|.
i  |g| ⇒ g(i)  i ⇒ f (g(i))  f (i) 5 i ⇒ i  |fg|.

8. f (i) 5 i ⇒ f m(i) 5 i for all m  Z

10. Supp f 5  ⇔ i  Supp f for any i  In ⇔ f (i) 5 i for all i.

Q001-Algebra-111001_Answers.indd 29 9/16/2011 10:01:02 AM

 A-30   Algebra – Abstract and Modern

12. Follows from the observation that i  Supp f if and only if f (i)  Supp
f and that f is a 3-cycle if and only if Supp f 5 {i, f (i), f2(i)} for some
i. This cannot be extended for any r-cycle. For example, if f 5 (1 2) o
(3 4), then Supp f 5 {1, 2, 3, 4}, which is a 4-element set, but f is not a
4-cycle.
14. (g f g21)n 5 g f n g21 and hence f n 5 e ⇔ (g f g21)n 5 e for any n  Z1.
16. If a 5 (i1 i2 i3 … i2m11), then

a2 5 (i1 i3 i5 … i2m11 i2 i4 … i2m)

17. No; if a 5 (i1 i2 i3 … i2m), then
a2 5 (i1 i3 … i2m21) o (i2 i4 … i2m).
O( a)
18. If a is an r-cycle and 1  s < r, then O(a) 5 r and O( a s )   r
s O( a) r ( s, O( a)) ( s, r )
O( a )   and hence, if (s, r) 5 1, then O(a ) 5 r.
s
( s, O( a)) ( s, r )
20. If a 5 (i1 i2 … ir) and b 5 (j1 j2 … js), then f o a o f21 5 (f (i1) f (i2) … f (ir))
and f o b o f21 5 (f (j1) f (j2) … f (js)). Also, f (it) 5 f (ju) ⇔ it 5 ju.

22. Partitions of 4 are {1, 1, 1, 1}, {1, 1, 2}, {2, 2}, {1, 3}, {4} and of 5 are
{1, 1, 1, 1, 1}, {1, 1, 1, 2}, {1, 2, 2} {1, 1, 3} {2, 3}, {1, 4}, {5}.

24. Follows from Ex. 20 and 23 above.

26. Use Ex. 25 and the facts that (2 3) (1 2) (2 3) 5 (1 3), (3 4) (1 3) (3 4) 5

(1 4), … to prove that sn is generated by (1 2), (2 3), …, (n21, n). If a 5
(1 2 3 … n) and b 5 (1 2), then ak b a2k 5 (k k 1 1) for all 1  k  n 2 1.
Therefore, Sn is generated a and b.

28. See Ex. 20 above.

30. Let |G| 5 6. If G is abelian, then G  Z6. Suppose that G is not abelian.
Then, O(a) 5 3 for some a  G (O(a) 5 2 for all a ⇒ G is abelian and
O(a) 5 6 for some a ⇒ G  Z6). Choose b  G 2 {e, a, a2}. Then,
we can check that b2  a and b2  a2. Also, b2  {b, ab, a2b}. Since e,
a, a2, b, ab, a2b are all distinct elements, we have G 5 {e, a, a2, b, ab,
a2b}. Therefore, b2 5 e and O(b) 5 2. Since <a> is of index 2, <a> is
normal in G and hence b a b21 5 e, a or a2. If b a b21 5 e, then a 5 e
and, if b a b21 5 a, then ba 5 ab and G is abelian. Therefore, b a b21 5
a2. Thus, G 5 <a, b>, a3 5 e 5 b2, b a b 5 a2. Also, if f 5 (1 2 3) and
g 5 (1 2), then f3 5 e 5 g2 and g f g 5 f2 in S3. If a : S3 → G is defined
by, a(f ) 5 a, a(g) 5 b, a(f2) 5 a2, a(e) 5 e, a(fg) 5 ab and a(f2g) 5
a2b, then a is an isomorphism.

Q001-Algebra-111001_Answers.indd 30 9/16/2011 10:01:02 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-31

EXERCISE 6(D)

2. (i) f 5 (1 3 2 4) o (6 7 8 9) and CI(f ) 5 4 1 4 2 2 5 6

(ii) f 5 (1 3 6 9 5) o (2 4 7 8) and CI(f ) 5 5 1 4 2 2 5 7
(iii) f 5 (2 8 3 9 5 4) o (6 7) and CI(f ) 5 6 1 2 2 2 5 6
(iv) f 5 (2 3 6 8 9 4 5 7) and CI(f ) 5 8 2 1 5 7
4. If f  a1  a2 ο  ο as, then f 1  as 1 ο  ο a2 1 ο a11 and length of
ai is equal to that of ai−1. Also, ai’s are pair-wise disjoint if and only ai−1’s
are so. Therefore, CI(f ) 5 O(a1) 1 … 1 O(as) 2 s 5 CI(f21).
6. If f 5 a1 o a2 o … o as, then g o f o g21 5 (g o a1 o g21) o (g o a2 o g21)
o … o (g o as o g21)
ai is an r-cycle ⇔ g o ai o g21 is an r-cycle
ai’s are disjoint ⇔ g o ai o g21’s are disjoint.
s s
Therefore, CI( f )   O( ai )  s   O( g ο ai ο g1 )  s  CI( g ο f ο g1 ).
ai ο g )  s  CI( g ο f ο g1 ).
1 i1 i1

8. Let f 5 a1 o a2 o … o as and ai’s be disjoint cycles.

s s
Then, CI( f )   O( ai )  s   (O( ai ) 1). Since O(ai) 2 1  0,
i1 i1

CI(f ) 5 1 ⇔ O(ai) 2 1 5 1 for some i and O(aj) 2 1 5 0 for all j  i

⇔ O(ai) 5 2 and O(aj) 5 1 for all j  i
⇔ f 5 a1 and a1 is a transposition.
10. All transpositions in S6; that is, (i j), 1  i < j  6.
12. A1 5 A2 5 {e}, A3 5 {e, a, a2}, where a 5 (1 2 3). Therefore, An is abe-
lian for n < 4. Also, A4 is not abelian (since (1 2 3) o (2 3 4) 5 (1 2) o (3
4) and (2 3 4) o (1 2 3) 5 (1 3) o (2 4)) and hence An is not abelian for
all n  4 (since A4 is a subgroup of An).
14. |Sn| 5 n! and |An| 5 1 n! and hence |Sn 2 An| 5 |An|.
2
16. If f 5 (1 2 3 4 5) o (6 7 8), then O(f ) 5 5 ? 3 5 15.
18. If f 5 a1 o … o ar, ai’s are transpositions, then f2 5 a1 o … o ar o a1
o … o ar.
20. If An  H  Sn, then |H| divides n! and 1 n! divides |H| and 1 n! < |H| <
2 2
n! and hence 1 > n  2 which is an absurd. Since n! is an integer.
|H | |H |
22. Let A be a subgroup of Sn and f  A be odd. Then, An ∪ An f 5 Sn and An
∩ An f 5 , since An is of index 2. From this we get that A is the disjoint

Q001-Algebra-111001_Answers.indd 31 9/16/2011 10:01:03 AM

 A-32   Algebra – Abstract and Modern

union of (An ∩ A) and (An f ∩ A). Also, g  gf is a bijection of An ∩ A

onto An f ∩ A. Therefore, |A| 5 2(An ∩ A), so that |A| is even.
24. If O(f ) 5 5 in S6, then f cannot be a product of any two or more disjoint
cycles; for, if f 5 f1 o f2 and f1 and f2 are disjoint cycles of length > 1, then
O(f ) 5 l.c.m. of |f1| and |f2| 5 2 or 6.
26. This is an example to prove that the converse of the Langrange’s Theo-
rem is false.
28. In S4, C((2 4 1)) 5 {e, (2 4 1)}
and C((1 2) o (3 4)) 5 {e, (1 2) o (3 4), (1 3) o (2 4), (1 4) o (2 3)}.
30. O(1 2 3 4) 5 4 and 3 does not divide 4.
32. 16, since 21 is the least positive integer n such that f 21 5 e.
34. All 7-cycles only.
36. m 5 3 or 7 or 9.
38. An has only two cosets An and An f, where f is any odd permutation and An
f is the set of all odd permutations.
5
40. C3  2  20 .
42. 0, since any 4-cycle is add.
44. Let f 5 f1 o f2 o … o fr, where fi’s are disjoint cycles. Then, O(f ) 5 l.c.m.
{|f1|, …, |fr|}. If O(f ) is odd, then | fi| is odd and hence fi is even for all i.
Therefore, f is an even permutation.
46. Since |Dn| 5 2n and O(a) 5 n, <a> is of index 2 and hence normal and
| Dn / < a >|  2.
48. If N is a normal subgroup of Sn, then N ∩ An is a normal subgroup of An
and, since An is simple, N ∩ An 5 {e} or An and therefore An ⊆ N so that
N 5 An or Sn. Since N is proper, N 5 An.

CHAPTER 7
EXERCISE 7(A)

2. Yes, u is effective.
4. ker u 5 {a  G : axa21H 5 xH for all x  G}, which is not necessarily
{e} and hence u is not necessarily effective.

Q001-Algebra-111001_Answers.indd 32 9/16/2011 10:01:04 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-33

6. Let X 5 G/H. Then, |X| 5 n. Define u : G 3 X → X by u(a, xH) 5

axa21H. Then, u is an action of G on X and hence fu : G → S(X) (5 Sn)
is a homomorphism whose kernel is ∩ xHx1  H . Therefore, G/H 
xG
f(G) ⊆ Sn.
8. Each element of G maps edges (or faces or vertices or diagonals) onto
edges (or faces or vertices or diagonals, respectively).
10. Straight forward verification.
12. If u is the given action of G on X, then ker u is a normal subgroup of G
contained in H.
|G|
14. Let H be a subgroup of G and  p. Let X 5 {xH : x  G}. Then, |X|
|H |
5 p and G acts on X by left translation; that is, u(a, xH) 5 axH. Then,
ker u is a normal subgroup of G and ker u 5 ∩ xHx−1 ⊆ H ; G / ker u
xG
is isomorphic to a subgroup of Sp and hence |G/ker u| is a divisor of p!. If
q is a prime dividing |G/ker u| then q divides p! and hence q 5 p. Also,
|G|
p2 does not divide p! and hence |G/ker u| 5 p 5 . Therefore, |ker u|
|H |
5 |H|. Since ker u ⊆ H, we get that H 5 ker u, which is normal in G.
16. If |G| 5 2695 and H is a subgroup of 539 in G, then H is of index 5 and
5 is the least prime dividing |G|. By Ex. 14 above, H is normal in G.
18. If |G| is odd and 3 divides |G|, then 3 is the least prime dividing |G|. Now
use Ex. 14 above.

EXERCISE 7(B)

2. See answer for Ex. 6 in Exercise 7(A).

4. Its kernel is ∩ xHx1 .
xG

6. Let (G) be the set of all subgroups of G and let G act on (G) by
conjugation (that is, u(a, K) 5 aKa21). Then, the orbit of H is the set of
conjugates of H in G and the stabilizer of H is the normalizer of H in G.
Now use Corollary 7.2.1.
8. We can assume that the action is nontrivial. By Theorem 7.2.7, St(x) is a
maximal subgroup of G for each x  X. If N ⊆ St(x) for all x  X, then the
induced action of N on X is trivial. Suppose that N  St(x) for some x  X.
Then, since N is normal NSt(x) is a subgroup containing St(x) properly and
hence NSt(x) 5 G. Now, for any y  X, there exists g  G such that gx 5
y (since the action of G on X is primitive, it is transitive by Corollary 7.2.2).

Q001-Algebra-111001_Answers.indd 33 9/16/2011 10:01:04 AM

 A-34   Algebra – Abstract and Modern

Since g 5 G 5 NSt(x), g 5 ab for some a  N and b  St(x). Now, y 5

gx 5 abx 5 ax, a  N. Thus, the induced action of N on G is trivial.
10. Let the action u of G on X is doubly transitive and R be an equivalence
relation on X which is compatible with u. If R  X, then there exist x
 y  X such that (x, y)  R and now, for any z  X, there exists a  G
such that ax 5 x and ay 5 z and hence (x, z)  R. This proves that R 5
X 3 X. Thus, the action is primitive. The converse is not in general. For,
consider the following example.
Let N be a proper nontrivial normal subgroup of a group G such that N
is a maximal subgroup of G (for example, take G 5 Z and N 5 pZ for
some prime p). Let u be the action of G on G/N by left translation. Then,
u is primitive (since St(xN) 5 N for any x  G); but u is not doubly tran-
sitive (choose e  x  N and y  G 2 N. Then, there is no a  G such
that aeN 5 xN and aeN 5 yN).

EXERCISE 7(C)

2. If C(x) 5 {axa21 : a  G}, then |C(x)| 5 1 ⇔ axa21 5 x for all a  G

⇔ x  Z(G).
4. Let e, a 5 (1 2 3), b 5 (1 3 2), c 5 (2 3), d 5 (3 1) and s 5 (1 2) be all
the elements of S3 (see Worked Exercise 6.2.1). C(e) 5 {e}, C(a) 5 {a,
b} and C(s) 5 {s, c, d} are all the conjugate classes in S3 and Cent(a) 5
{e, a, b} and Cent(s) 5 {e, s} and, by the class equation 7.3.2,
6 5 |S3| 5 i(Cent(a)) 1 i(Cent(s)) 1 1 5 2 1 3 1 1.
Note that Z(S3) 5 {e}.
6. For any a  G, |C(a)| 5 iG(St(a)) and hence |C(a)| divides |G|.
8. Let iG(H) 5 n and X 5 {xH : x  H}. Then, G acts on X and f : G →
S(X), defined by f (a)(xH) 5 axH, is a homomorphism whose Kernel is
K  ∩ xHx1 . Then, K is a normal subgroup of G, K ⊆ H and G/K  a
xG
subgroup of S(X). Since S(X) is finite, |G/K| 5 iG(K) is finite.
10. Let |N| 5 n and |G/N| 5 m and (n, m) 5 1. Choose integers a and b such
that an 1 bm 5 1. For any a  G,
O(a) divides n ⇒ an 5 e and O(aN) divides m
⇒ aan 5 e and am  N
⇒ a 5 aan1bm 5 e(am)b  N.
12. |N| 5 3, N  Z(G). Choose x  N such that x  Z(G). Let G act on N
by conjugation. Then, the orbit C(x) 5 {axa21 : a  G} is a subset of N,

Q001-Algebra-111001_Answers.indd 34 9/16/2011 10:01:05 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-35

since N is normal and x  N. Since x  Z(G), |C(x)| > 1. Since e  C(x)

(otherwise x 5 e  Z(G)), it follows that 1 < |C(x)| < |N| 5 3 and hence
|C(x)| 5 2 and hence the stabiliser St(x) is of index 2.
14. As in Worked Exercise 7.3.1, let Z6 act on the set X of all possible neck-
laces. We have |X| 5 26. Let us compute Xa 5 {x  X : ax 5 x} for all a
 Z6 5 {0, 1, 2, 3, 4, 5}. Clearly, |X0| 5 |X| 5 26. Let b1, b2, b3, b4, b5, b6
be the beads. The action of 1 transforms the beads b1, b2, …, b6, in this
order, to b2, b3, …, b6, b1. In order that these are same necklaces, b1, b2,
…, b6 must be add of same colour and hence |X1| 5 2, since we are given
with two colours. The action of 2 will transform b1 b2 b3 b4 b5 b6 to b3 b4
b5 b6 b1 b2; for these to be same necklaces, b1, b3, b5 must be of one colour
and b2, b4, b6 must also be of one colour. Thus, |X2| 5 2 ? 2 5 4. Similarly,
we can see that |X3| 5 8, |X4| 5 4, |X5| 5 2 (note that X a  X a1for any a
 Z6).
Therefore, by Theorem 7.3.7, the number of different necklaces is
1  | X |  1 (26  2  4  8  4  2)
|Z p| aZ6 a 6
514.
16. If G acts on P(G) by conjugation, then C(S) is the orbit of S and N(S) is
the stabiliser of S.
18. Use Ex. 14 in Exercise 7(A).
20. By Worked Exercise 7.3.1, the number of different necklaces is
5 111 5
(5 111)  (510 10)  4,338,925.
11 11

EXERCISE 7(D)

2. (i) <3> 5 {0, 3, 6, 9, 12, 15, 18, 21} is the only Sylow 2-subgroups.
<8> 5 {0, 8, 16} is the only Sylow 3-subgroup {0} is the only
Sylow p-subgroup for prime p  2, 3.
(ii) Refer Worked Exercise 6.2.1. {e, a, b} is the only Sylow 3-sub-
group and {e, c}, {e, d} and {e, s} are all the Sylow 2-subgroups
{e} is the Sylow p-subgroup for any prime p  2, 3.
(iii) Refer Worked Exercise 6.4.2. <a1>, <b1>, <c1> and <d1> are the
only Sylow 3-subgroups. {e, p, q, r} is the only Sylow 2-subgroup.
4. Let |G| 5 56. Then, n7 5 1 1 7k should divide |G| 5 23 ? 7 and hence n7
should divide 23, so that n7 5 1 or 8. Any subgroup of order 7 is cyclic
and hence generated by each of the six nonidentity elements. If n7 5
8, then there are 6 ? 8 5 48 elements of order 7 and the remaining 8

Q001-Algebra-111001_Answers.indd 35 9/16/2011 10:01:05 AM

 A-36   Algebra – Abstract and Modern

elements should form a unique Sylow 2-subgroup, which turns out to

be normal and hence G is not simple. If n7 5 1, then the unique Sylow
7-subgroup is normal and hence G is not simple.
6. Follows from the fact that, if A is a Sylow p-subgroup, then xAx21 is also
a Sylow p-subgroup for any p.
8. Let G be a group of order pr ? n. Then, np 5 1 1 pk for some k  0
and np should divide the index of any Sylow p-subgroup. Therefore, np
should divide n which is impossible, unless k 5 0 (since n < p). There-
fore, np 5 1 and hence there exists unique Sylow p-subgroup which is
normal in G.
10. Use Ex. 9 above and Ex. 14 of Exercise 7(A).
12. Let |G| 5 225 5 32  52. Then, n5 5 1 1 5k should divide 9 and hence
n5 5 1 so that there is a unique Sylow 5-subgroup which is normal in G.
14. 323 5 19 ? 17 and use Worked Exercise 7.4.2.
16. (i) Routine verification.
(ii) If a 5 (i1 i2 … ir) and b 5 (j1 j2 … jr) are r-cycles, define f  Sn by
f (ik) 5 jk for 1  k  r and f (i) 5 i for all i  ik. Then, f o a o f21
5 b (by (i)).
(iii) Use (i) and (ii) and Theorem 6.3.5.
(iv) use (iii).
18. Let n > 2, f 5 (1 2) and g 5 (2 3). Then, f o g  g o f.
20. Follows from Theorem 7.4.2 and the fact that |G| 5 |N|  |G/N|.
22. If |G| 5 pn, then np 5 1 1 pk should divide n, which is possible only
when np 5 1.
24. Z7[x], the additive group of polynomials over Z7.
26. Let S be a normal (and hence unique) Sylow p-subgroup of G. Then, f (S)
( S/S ∩ ker f ) is a p-subgroup and hence f (S) is contained in a Sylow
p-subgroup. Thus, f (S) ⊆ S.
28. If |G| is square-free, then |G| 5 p1 p2 … pr, where p1, …, pr are distinct
primes. If Ai is the Sylow pi-subgroup of G, then Ai ∩ Aj 5 {e} for i  j
and G 5 A1A2 … Ar. Now, G  A13 … 3 Ar  Z p1  Z pr  Z|G|.
30. 30 5 2 ? 3 ? 5, A2 5 {0, 15}, A3 5 {0, 10, 20} and A5 5 {0, 6, 12, 18,
24} are respectively Sylow 2, 3 and 5-subgroups of Z30. For primes p 
2, 3, 5, {0} is the Sylow p-subgroup.

Q001-Algebra-111001_Answers.indd 36 9/16/2011 10:01:06 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-37

CHAPTER 8
EXERCISE 8(A)

2. Since there is no element of order 4 in Z2 3 Z2, Z4 is not isomorphic

with Z2 3 Z2.
4. Use Corollary 8.1.2.
6. A 3 B  AB 5 Z10 (Use Ex. 4 above)
8. No, otherwise Z 3 Z  Z and hence any nonzero subgroup of Z 3 Z is
of finite index (since, it is so in Z). Note that Z 3 {0} is of infinite index
in Z 3 Z.
10. Use Corollary 8.1.4 and the facts that Z/mZ  Zm and mZ 5
( p1r1 ) ∩ ( p2r2 ) ∩  ∩ ( pnrn ).
12. By Corollary 6.4.6, the group of symmetries of a square is indecom-
posable. Since nZ ∩ mZ 5 [n, m]Z, where [n, m] is the l.c.m. of n
and m, the intersection of any two nonzero subgroups of Z is not zero.
Therefore, Z is indecomposable. If A is a nonzero subgroup of (Q,
1), then A ∩ Z is a nonzero subgroup of Z and hence A ∩ B  {0}
for any nonzero subgroups A and B of (Q, 1). Therefore, (Q, 1) is
indecomposable.
14. Since |A ∩ B| is a common divisor of |A| and |B|, we get that A ∩ B 5
{e}. Also, |AB| 5 |A|  |B| 5 |G| and hence AB 5 G. Now, use Corollary
8.1.4.
16. If S3 is decomposable, then S3  A 3 B, where A and B are subgroups of
order 2 and 3, respectively, and hence S3  Z2 3 Z3  Z6, which is not
true.
18. Straight forward verification.
20. (i) The map f : G1 3 G2 → G2 3 G1, defined by f (a, b) 5 (b, a) is an
isomorphism.
(ii) f : (G1 3 G2) 3 G3 → G1 3 (G2 3 G3), defined by f ((a1, a2), a3) 5
(a1, (a2, a3)) is an isomorphism.
(iii) If f : G1 → G2 is an isomorphism, then g : G1 3 G3 → G2 3 G3,
defined by g(a, b) 5 (f (a), b) is an isomorphism.
(iv) G1 3 {e} and {e} 3 G2 are subgroups of G1 3 G2 and are isomor-
phic to G1 and G2, respectively.
(v) Z 3 Z is not cyclic and Z is cyclic.

Q001-Algebra-111001_Answers.indd 37 9/16/2011 10:01:06 AM

 A-38   Algebra – Abstract and Modern

EXERCISE 8(B)

2. Use Ex. 16 of Exercise 4(B).

4. Follows from a m  e ⇔ aim  e for all i ⇔ O(ai) divides m for all 1 
i  n.
6. Let a 5 (a1, …, an)  G1 3 … 3 Gn. Then, a  Z(G) ⇔ aibi 5 biai for
all bi  Gi and for all 1  i  n ⇔ ai  Z(Gi) for all i.
8. (x1, x2)(a1, a2)(x1, x2)21 5 (x1a1x121, x2a2x221)  N1 3 N2 if ai  Ni. Define
f : G1 3 G2 → G/N1 3 G/N2 by f (x1, x2) 5 (x1N1, x2N2). Then, f is an epi-
morphism and ker f 5 N1 3 N2.
10. If G is an infinite cyclic group, then G  Z and, since Z is not a p-group,
G is so.
12. Use Ex. 4 above.
14. Use Ex. 2 above.
16. Use Ex. 14 above and Ex. 12 of Exercise 8(A).
18. Since G 5 A1A2 … Ar, where Ai 5 <ai>, we can use Theorem 8.1.1.
20. Theorem 8.3.1 can be used.

EXERCISE 8(C)

2. Since P( X )  Z52  Z 2  Z 2  Z 2  Z 2  Z 2 , the invariants of P(X) are

2, 2, 2, 2, 2.
4. G  Z4 3 Z4 3 Z4 3 Z4.

CHAPTER 9
EXERCISE 9(A)

2. (i) 14
(ii) 3
4 4
(iii)  
4 4
(iv) 78
(v) 0
(vi) {2}.

Q001-Algebra-111001_Answers.indd 38 9/16/2011 10:01:07 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-39

4. (i) ⇒ (ii) for any x  R, x 5 x 1 0 5 x 1 (0  0) 5 (x 1 0)  (x 1 0) 5

x2 and x 5 x 1 (x  0) 5 (x 1 x)  (x 1 0) 5 x2 1 x2 5 x 1 x and hence
x 5 0.
(ii) ⇒ (iii) ⇒ (i) are trivial. Also (ii) ⇔ (iv) is clear.
6. This follows from px 5 0 for all x  Zp.
8. Straight forward verification. Note that 21 is the new additive identity
and 0 is the new multiplicative identity and that 2a 2 1 2 1 is the addi-
tive inverse of a.
10. a 5 0 ⇒ ab 5 0 5 ba.
a  0 ⇒ ab 5 ba, since a(ba) 5 (ab)a

12. ab 5 0 ⇒ ba 5 (ba)n 5 b(ab)n21a5 b o a 5 0.

14. The ring of 2 3 2 matrices over Z2.

EXERCISE 9(B)

2. If a is a unit, then ab 5 0 ⇒ b 5 a21(ab) 5 a21(0) 5 0.

4. a is a unit ⇒ a 5 a2a21 5 aa21 5 1.
a is nilpotent ⇒ an 5 0, n  Z1 ⇒ a 5 an 5 0.
a1 ar
6. Let n  p1 … pr , where pi’s are distinct primes and ai  Z1. Then, C(n)
5 p1 … pn. Any 0  x < n is a nilpotent if and only if C(n) divides x and
hence x is of the form C(n)a, 0  a  n . Thus, the number of nilpo-
C (n)
tents in Zn is n/C(n).
8. Straight forward verification.
10. (ab)(ab) 5 a(ba)b 5 a(ab)b 5 a2b2 5 ab and hence ab is an idempotent.
a 1 b need not be an idempotent. For example, the unity 1 is an idempo-
tent and 1 1 1 is not in Z3.
12. Clear.
14. If n is the least positive integer such that an 5 0, then n > 1 (since a  0),
an21  0 and a(an21) 5 0 and hence a is a zero divisor.
16. We can suppose that a  0. If an 5 0 and b 5 2a, then bn 5 0 for some
n > 1 and (1 1 a)(1 1 b 1 … 1 bn21) 5 (1 2 b)(1 1 b 1 … 1 bn21) 5 1
and hence 1 1 a is a unit.

Q001-Algebra-111001_Answers.indd 39 9/16/2011 10:01:07 AM

 A-40   Algebra – Abstract and Modern

18. Consider (axa 2 ax)2 5 axa  axa 1 ax  ax 2 axa  ax 2 ax  axa 5

axaxa 1 axax 2 axax 2 axaxa 5 0 and hence axa 2 ax is a nilpotent,
so that axa 2 ax 5 0; i.e., axa 5 ax. Similarly, axa 5 xa.

EXERCISE 9(C)

2. (a 1 b)3r 5 a3r 1 b3r (since 3 divides 3rCs for any 0 < s < 3r).
4. If R is a finite ring, then O(a) is finite in the group (R, 1) and, if m is the
l.c.m. {O(a) : a  R} then ma 5 0 for all a  R and hence char(R) is
finite.
6. Z5.
8. (a 1 b)p 5 ap 1 bp 5 a 1 b and (a  b)p 5 apbp 5 ab. Also, a  A ⇒ a
5 ap ⇒ a(1 2 ap21) 5 0 ⇒ a 5 0 or ap21 5 1 ⇒ a 5 0 or a is a unit.
10. Direct verification.
12. Let ab 5 1 and R 5 {a1, a2, …, an}. Consider Ra 5 {a1a, a2a, …, ana}
⊆ R. For any i and j,
aia 5 aja ⇒ (ai 2 aj)a 5 0 ⇒ ai 2 aj 5 (ai 2 aj)ab 5 0
Therefore, Ra and R have the same number of elements and hence R 5
Ra. In particular, 1 5 ca for some c  R. Now, b 5 1b 5 (ca)b 5 c(ab) 5
c and hence ba 5 1.

EXERCISE 9(D)

2. Direct verification.
4. S is a subring of (Zn, 1n, n) ⇔ S is a subgroup of (Zn, 1n) ⇔ S 5 <m>
for some divisor m of n.
6. Let A 5 {x1x2 … xn : n > 0 and x1, x2, …, xn  X}.
Then, (X) 5 {a1 1 a2 1 … 1 an : ai or 2ai  A}.
8. Let X be an infinite set and R be the set of all finite subsets of X. Then,
(R, 1, ∩) is a ring without unity. For any finite subset A of X, P(A) is a
subring of R and A is the unity in P(A). (Here, 1 is the symmetric differ-
ence of sets.)
10. S ∪ T is a subring ⇒ S ∪ T is a subgroup of (R, 1) ⇒ S ⊆ T or T ⊆ S.

Q001-Algebra-111001_Answers.indd 40 9/16/2011 10:01:07 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-41

12. Follows from Theorem 4.1.2.

14. Consider (a 1 b)2 1 (a 1 b)  c(R). Therefore, ab 1 ba  c(R). a(ab 1

ba) 5 (ab 1 ba)a and hence a2b 5 ba2. Since (a2 1 a)b 5 b(a2 1 a), we
get ab 5 ba.

EXERCISE 9(E)

2. (i) Yes; since Zn is finite, any injection of Zn into Zn is a bijection.

(ii) No, there is no homomorphism of Z2 into Z3.
(iii) Yes; since Zn will be isomorphic to a subgroup of Zm and we can
use Lagrange’s Theorem.
(iv) Yes; if f : Z/mZ → Z/nZ is defined by f (a 1 mZ) 5 a 1 nZ, then
f is a well-defined epimorphism.
(v) Yes, if R is not trivial.
(viii) Yes, namely the zero map and the identity map.

EXERCISE 9(F )

2. The ring of all 2 3 2 matrices over Z3.

 0 0 1  1 1 0 1 0
4. No; consider    .
0 1 1  1 0 0  2 1
8. 0 is the only nilpotent and 0 and 1 are the only idempotents.

10. Direct verification.

12. The set of all scalar matrices; that is, matrices A 5 (aij) such that aij 5 0
for all i  j and aii 5 ajj for all i and j.

14. R

16. If A 5 (aij) is such that BA 5 B 5 AB for all B  Mn(R), then prove that
aij 5 1 or 0 according as i 5 j or i  j.

EXERCISE 9(G)

2. (i) R 3 R is not integral domain and hence not a field, since (1, 0) 
(0, 1) 5 (0, 0)
(ii) Qn is not an integral domain, not a field.

Q001-Algebra-111001_Answers.indd 41 9/16/2011 10:01:08 AM

 A-42   Algebra – Abstract and Modern

(iii) Z[ 2] is an integral domain, but not a field.

(iv) Z[i] is an ID, but not a field.
(v) Z5[i] is a field.
(vi) Z4[i] is not an ID.
(vii) Z3[i] is a field.
(viii) Z2[i] is a field.
(ix) Z5 3 Z3 is not an ID.
(x) Z19 is a field.
4. R is a field and Boolean ring ⇒ R is an ID and a2 5 a ⇒ a(a 2 1) 5 0
⇒ a 5 0 or a 5 1 for all a  R ⇒ R 5 {0, 1}, converse is clear.
6. Direct verification.
8. Zn[i] is finite.
12. No.
14. (R 2 {0}, ) is a finite semigroup satisfying the cancellation laws. By
Theorem 3.3.5, (R 2 {0}, ) is a group. Therefore, R is a field.
16. If F is a finite field, then F is an ID and char(F )  0. Therefore, by
­Theorem 9.7.4, char(F ) is prime.
18. Imitate Ex. 14 above.
20. Let n be the least positive integer such that na 5 0. Then, (n 2 1)a  0
and (nb)  (n 2 1)a 5 (n 2 1)b  na 5 0 and hence nb 5 0 for all b  R.
Thus, char(R) > 0.

CHAPTER 10
EXERCISE 10(A)

2. (i) Z is not an ideal of Q, since 1  1  Z.

2
(ii) Q is not an ideal of R, 2  1  Q.
(iii) No, Z is a subring of Q, but not an ideal.
(iv) True.
(v) True.
(vi) True, {0} is an ideal in any ring R.
(vii) No, if R is finite and |R| 5 n, then there are at most 2n ideals in R.

Q001-Algebra-111001_Answers.indd 42 9/16/2011 10:01:08 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-43

(viii) Z4 has exactly three ideals, namely {0}, {0, 2} and Z4.
4. If R is a division ring, I is a nonzero ideal of R and 0  a  I, then 1 5
aa21  I and hence I 5 R, so that R has only two ideals. The converse is
not true; for, M2(R), the set of 2 3 2 matrices forms a ring with exactly
two ideals, but it is not a division ring.

6. (i) x  (I 1 J)* ⇔ x(a 1 b) 5 0 for all a  I, b  J.

⇔ xa 5 0 5 xb for all a  I, b  J
⇔ x  I* ∩ J*
⇔ x  (I ∪ J)*.
(ii) Clear.
(iii) Since I ∩ J ⊆ I and J, I* ⊆ (I ∩ J)* and J* ⊆ (I ∩ J)* and hence
I* 1 J* ⊆ (I ∩ J)*.
8. Let I be an ideal and p  I, where p is a prime. Then, pZ ⊆ I. If pZ  I,
then there exists a  I such that a  pZ; that is, p does not divide a and
hence (a, p) 5 1. Then, there exist integers a and b such that 1 5 aa 1
bp  I and hence I 5 Z.
10. (i) r(I ∩ J) ⊆ r(I) and r(J)
a  r(I) ∩ r(J) ⇒ an  I and am  J, for some n, m  Z1
⇒ an1m  I ∩ J, n 1 m  Z1
⇒ a  r (I ∩ J).
(ii) I ⊆ J and a  r(I) ⇒ an  I ⊆ J ⇒ a  r(J).
(iii) r(I) ⊆ r(I 1 J) and r(J) ⊆ r(I 1 J), since I and J ⊆ I 1 J.
Therefore, r(I) 1 r(J) ⊆ r(I 1 J).
(iv) I 1 J ⊆ r(I) 1 r(J). Hence, r(I 1 J) ⊆ r(r(I) 1 r(J)).
a  r(r(I) 1 r(J)) ⇒ an 5 x 1 y, n  Z1, x  r(I) and y  r(J)
⇒ an 5 x 1 y, xr  I, ys  J
⇒ an(r 1 s) 5 (x 1 y)r1s  I 1 J
⇒ a  r(I 1 J).
12. (i) 24Z
(ii) 3Z
(iii) 3Z
(iv) 6Z
(v) {0}
(vi) 2Z

Q001-Algebra-111001_Answers.indd 43 9/16/2011 10:01:08 AM

 A-44   Algebra – Abstract and Modern

22. Let I be an ideal in Mn(R) and J be the set of all 11th (first row and first
column) entries in members of I. Then prove that I 5 Mn(J), by estab-
lishing that, if A 5 (aij) and Eij’s are the matrix units, then

E1r (∑ ais Eis ) E j1  arj E11

If R is simple, then J 5 {0} or R and hence I 5 {0} or Mn(R). Thus,

Mn(R) is a simple ring.

EXERCISE 10(B)

2. (i) R has unity ⇒ R/I has unity. The converse fails.

(ii) True.
(iii) True.
(iv) False; for Z is an ID and Z/6Z is not an ID.
(v) False.
(vi) False; for R/R 5 {0}.
(vii) True; for R/{0} 5 R.
4. Suppose that the multiplication of additive cosets of S are well defined.
Then, for any a  S and r  R, a 1 S 5 0 1 S and hence ar 1 S 5 0r
1 S 5 S which implies that ar  S. Similarly, ra  S. Thus, S is an ideal
of R. Converse is trivial.
6.
Idempotents Nilpotents Units
Z/I; I 5 6Z I, 1 1 I, 3 1 I, I 1 1 I, 5 1 I
41I
Z/I; I 5 8Z I, 1 1 I I, 2 1 I, 4 1 1 I, 3 1 I, 7 1
1 I, 6 1 I I, 5 1I
Z/I; I 5 7Z I, 1 1 I I a 1 I, 1  a  6.
8. (i) ⇔ (ii): n is prime ⇔ (a, n) 5 1 for all 1  a < n ⇔ ab  1 (mod n)
for some b, for each 1  a < n ⇔ a 1 nZ is a unit for all 1  a < n.
(ii) ⇔ (iii) is clear, since Z/nZ is finite.
10. nZ, n  0, are the only ideals of Z.
12. If I 5 2Z in Z, then (0 1 I) (0 1 I) 5 4Z  2Z 5 0 1 I
14. If (n, m) 5 1, then nZ ∩ mZ 5 nmZ and nZ 1 mZ 5 Z and hence Zn 
Z/nZ 5 nZ 1 mZ/nZ  mZ/nZ ∩ mZ 5 mZ/nmZ.

Q001-Algebra-111001_Answers.indd 44 9/16/2011 10:01:09 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-45

16. The ideals of Z/nZ are of the form dZ/nZ, where d is a divisor of n and
the homomorphic images of Z/nZ are of the form Z/dZ, where d is a
divisor of n.

EXERCISE 10(C)

2. Direct verification.

EXERCISE 10(D)

2. Let P  ∩ Pa . Then, clearly P is a proper ideal of R. If a  P and b 

a
P, then there exist a and b   such that a  Pa and b  Pb. Choose g
  for which Pg ⊆ Pa and Pg ⊆ Pb. Then, a and b  Pg and hence ab
 Pg, so that ab  P.
4. Use Ex. 2 and 3 above.
6. Clearly prime ⇒ primary. 4Z is primary and not prime in Z.
8. Note that 1  I ⇔ 1  I . Since I is proper, I is a proper ideal of R
and I ⊆ I . Also, ab  I ⇒ ( ab) n  I for some m > 0 ⇒ anbn  I
⇒ an  I or b n  I ⇒ a  I or b  I . Thus, I is a prime ideal.
Also, if I ⊆ P and P is a prime ideal, then I ⊆ P  P.
10. ab  0 (mod pn) ⇒ pn divides ab ⇒ p divides a or b ⇒ pn divides an or
bn ⇒ an  0(mod n) or bn  0 (mod n).
12. Let R be regular and I be an ideal of R. If a  I , then an  I and choose
b  R such aba 5 a. Now, abab 5 ab and ab 5 (ab)n 5 anbn  I and
hence a 5 (ab)a  I. Therefore, I  I . For the converse, if a  R,
then aRa is an ideal and a3  aRa and hence a  aRa  aRa.
14. (i) ⇒ (ii): Let P be the unique prime ideal of R. Then, P 5 N, the set of
nilpotents in R. If a is a nonunit in R, use the Zorn’s lemma to get a prime
ideal (5 p) containing a and hence a  N.
(ii) ⇒ (iii): ab 5 0 and bn  0 for all n > 0 ⇒ b is not nilpotent ⇒ b is
a unit (by (ii)) ⇒ a 5 (ab)b21 5 0. Thus, {0} is primary. If a  0 is a
nonunit, then an 5 0 (by (ii)) for some n > 1. If n is least such that an 5
0, then an21 ? a 5 0 and an21  0 and hence a is a zero divisor.
(iii) ⇒ (i): N 5 radical of {0}. If {0} is primary, then N is a prime ideal.
If P is any prime ideal, then N ⊆ P. Also, 0  a  P ⇒ a is a nonunit
⇒ a is a zero divisor (by (iii)) ⇒ ab 5 0 for some b  0 ⇒ an 5 0 for

Q001-Algebra-111001_Answers.indd 45 9/16/2011 10:01:10 AM

 A-46   Algebra – Abstract and Modern

some n ⇒ a  N. Therefore, P ⊆ N and P 5 N. Thus, N is the only

prime ideal of R.

16. If m > 1 and m2 divides n, then m2r 5 n for some r $ 1. Now, for any a
 R, (mra)2 5 m2r2a 5 n(ra) 5 0 and hence mra  N(R) 5 {0}, so that
mra 5 0, which is a contradiction to the least property of n 5 char(R).

EXERCISE 10(E)

2. 9Z
4. Use Ex. 3 and the fact that a nonzero ideal of Z is prime if and only if it
is maximal.
6. Let X 5 {x1, x2, …, xn} and M be a maximal ideal in (P(x), 1, ∩). Then,
n
 {xi }  X  M and hence {xi}  M for some i and, since {xi} ∩ {xj} 5
i1

  M, {xj}  M for all j  i. Let

Mi 5 {A ⊆ X : xi  A}.
Then, Mi is a maximal ideal of P(X) and M ⊆ Mi so that M 5 Mi.
8. Let x  X and define a : #(X, R) → R by a(f ) 5 f (x) for any f  #(X,
R). a is an epimorphism of rings and kera 5 Mx. Therefore, #(X, R)/
Mx  R. Since R is a field, so is #(X, R)/Mx. Thus, Mx is a maximal
ideal.
10. Consider the map f : Z[i] → Z2 defined by f (a 1 bi) 5 0 if a 2 b is even
and, 5 1 if a 2 b is odd. Prove that f is an epimorphism of rings and
ker f 5 I. Therefore, Z[i]/I  Z2 and hence I is a maximal ideal and the
quotient has exactly 2 elements.
12. I is not a prime ideal and |Z[i]/I] 5 25.
14. Use the fact that R/ker f  S and that I  f (I) is a one-to-one correspon-
dence between the set of ideals of R containing ker f and the set of ideals
of S. Also, I ⊆ J ⇒ f (I) ⊆ f (J).
16. R/P becomes a finite commutative ring without zero divisors and hence
R/P is a field and P is a maximal ideal.
18. M/Mn is the only prime ideal of R/Mn; for, if Q is a prime ideal of R/Mn,
then Q 5 P/Mn for some prime ideal P of R containing Mn and hence M,
so that P 5 M.

Q001-Algebra-111001_Answers.indd 46 9/16/2011 10:01:11 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-47

20. (i) For any 0  a  R, 1 2 ra is a nonunit ⇔ 12 ra  M for some

max ideal M ⇔ a  M, for some M ⇔ a  J(R) ⇔ J(R) 5{0} ⇔
R is semisimple.
(ii) If 0  a  R, then aba 5 a for some b  R and hence a(1 2 ba)
5 0, so that 1 2 ba is a nonunit.
(iii) Let J(R) be the Jacobson radical of R. Then,
a  J(R) ⇒ a 1 I  M/I for all maximal ideals M of R
⇒ a 1 I  J(R/I) ⇒ a 1 I 5 I ⇒ a  I.

EXERCISE 10(F)

2. Since char(Zn) 5 n 5 char(R), n(a, r) 5 (0, 0) for all (a, r)  Zn 3 R and

n is the least such positive integer. Therefore, char(Zn 3 R) 5 n.
4. The field of quotients of R is Q itself, since any nonzero m
n in Q can be
( )
1
expressed as  m n and m n
and  R.
2 2 2 2
6. If F is a field, consider the homomorphism f : Z → F defined by f (n) 5
n  1, where 1 is the unity in F. Then, ker f 5 {0} or pZ for some prime
p (if p 5 char(F )). Then, Z or Zp (5 Z/pZ) is isomorphic to a subring of
F and hence Q or Zp is isomorphic to a subfield of F.
8. (ii) Q
(iii) Z79
10. If |F | 5 9, then char(F ) 5 3 and Z3 is the prime subfield of F.
12. Let R be an integral domain and f : R → R an automorphism. Define f *:
F → F by f * a 
b ( )
f (a)
f (b)
a
for any b in the field F of quotients of F.
It can be checked that f * is well defined and is an automorphism of F.
Also, f */R 5 f.
14. By Ex. 13, R can be embedded in an integral domain S and hence R can
be embedded in the field of quotients of S.
16. Refer the construction of field of quotients.
18. For any prime ideal P of R disjoint with S, S 1 P  { as : a  P , s  S} is
a prime ideal of S21R and any prime ideal of S21R is of this form. It can
be checked that P  S21P is a one-to-one correspondence between the
prime ideals of R disjoint with S and the prime ideals of S21R.

Q001-Algebra-111001_Answers.indd 47 9/16/2011 10:01:11 AM

 A-48   Algebra – Abstract and Modern

CHAPTER 11
EXERCISE 11(A)

2. (i) True, since Z2 is an ID.

(ii) False, since x has no inverse in R[x], for any ring R.
(iii) False, since Z6 is not an ID.
(iv) False, R[x] is always infinite.
(v) False, since 1 and 21 are the only units in Z[x].
(vi) True, since 22 5 0 in Z4 and hence in Z4[x].
(vii) False
(viii) False.
4. Direct verification.
6. Let f [ R[x]. Then, f is a unit ⇒ fg 5 1, g [ R[x] ⇒ deg f 1 deg g 5 0
⇒ deg f 5 0 5 deg g ⇒ f is a nonzero constant polynomial.
8. 26 2 1 5 63.
10. I 5 xR[x].
12. First observe that, if an 5 0, then (1 2 a)(1 1 a 1 …1 an21) 5 1 and
hence 1 2 a is a unit. Therefore, if u is a unit and a is a nilpotent in any
commutative ring R, then u 2 a 5 u(1 2 u21a) is a unit. Prove the result
by induction on the degree of f (x). The result is trivial if deg f (x) 5 0. Let
deg f (x) 5 n . 0 and assume the result for all polynomials of degree less
than n. Let f (x) 5 a0 1 a1x 11anxn. Suppose that f (x) is a unit in R[x].
Then, f (x)g(x) 5 1 for some g(x) 5 b0 1 b1x 11 bmxm [ R[x]. Then,
we have (1) anbm 5 0, (2) an21bm 1 anbm21 5 0, (3) an22bm 1 an21bm21 1
anbm22 5 0, and, in general  ai b j  0 for k . 0 and a0b0 51. Thus, a0
i jk
is a unit and a021 5 b0. By multiplying (2) with an, we get an2 bm1  0.
Also, by multiplying (3) with a2n, we get a3nbm22 5 0 and, in general
ainbm2i11 5 0. In particular, anm11b0 5 0 and hence anm11 5 0. Therefore,
an is nilpotent and so is anxn and hence f (x) 2 anxn is a unit and is of
degree less than n. By induction hypothesis, we get that an21, an22, …,
a1 are nilpotents and a0 is a unit in R. Converse is clear from our first
observation.
14. Clearly, I[x] is a proper ideal of R[x]. Suppose that I is a prime ideal of R.
Let f (x) 5 a0 1 a1x 1…1 anxn and g(x) 5 b0 1 b1x1…1bmxm [ R[x] 2
I[x]. Let i be the least such that ai  I and j is the least such that bj  I.

Q001-Algebra-111001_Answers.indd 48 9/16/2011 10:01:12 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-49

Then, aibj  I, ar [ I, bs [ I for all r , i and s , j. Then, the coefficient

of xi1j in f (x)g(x) is not in I and hence f (x)g(x)  I[x]. Thus, I[x] is a
prime ideal of R[x]. Converse is clear.
16. Define a : R[x] → R by a(f (x)) 5 the constant term in f (x). Then, a is an
epimorphism of rings and its kernel is <x>. Thus, R[x]/<x> > R.

EXERCISE 11(B)

2. (i) 2
(ii) 1
(iii) 0
(iv) 4
(v) 4
4. The Kernel of fa is an ideal of F[x] and any nonzero ideal of F[x] is infinite.
6. (1 1 4x)(1 1 4x) 5 1 in Z8[x].
8. f (a) 5 g(a) ⇔ a is a root of the polynomial f (x) 2 g(x).
10. Use Fermat’s Theorem 4.4.7.

EXERCISE 11(C)

2. Let P be a prime ideal and I be an ideal of R such that P ⊆ I. Choose a

and b [ R such that P 5 aR and I 5 bR. Then, a 5 bc for some c [ R
and hence bc [ P, so that b [ P or c [ P. If b [ P, then I 5 P. If c [
P, then c 5 ad and hence a 5 bad.
4. Let I 5 {f (x) [ F[x] : a is a root of f (x)}. We can assume that I  {0}. If
g(x) is a nonzero polynomial of least degree in I, then we can prove that
I is the ideal generated by g(x) in F[x] (use division algorithm).
6. Every element of Z3 is a root of x3 2 x and hence of (x3 2 x)f (x) for any
f (x) [ Z3[x].
8. By Euler’s Theorem 4.4.6, ap21 ; 1 (mod p) for all 1 # a # p 2 1 and
hence each 1 # a # p 2 1 is a root of x p21 2 1 [ Zp[x]. Also, each 1 # a
# p 2 1 is a root of f (x) 5 (x 2 1)(x 2 2) … (x 2 (p 2 1)) 2 (x p21 2 1).
Therefore, f (x) has p 2 1 roots and its degree is p 2 2. Therefore, f (x)
must be the zero polynomial.
10. Z[x]/<x> (  Z) is an integral domain, but not a field and hence <x> is
a prime ideal but not a maximal ideal in Z[x].

Q001-Algebra-111001_Answers.indd 49 9/16/2011 10:01:12 AM

 A-50   Algebra – Abstract and Modern

12. The evaluation map fi : R[x] → C, defined by fi(g(x)) 5 g(i), is an

­epimorphism and its kernel is <1 1 x2>.
14. Let f (x) 5 a0 1 a1x 1…1 an21xn21 1 xn, ai [ Z and a  rs , where r and
s [ Z, (r, s) 5 1 and s . 0. Suppose that f (a) 5 0. Then, a0 1 a1 (rs )1
… 1 a (r )n1 (r )n  0 and hence sna 1 sn21a r 11 sa rn21 1
n21 s s 0 1 n21

rn 5 0. If s . 1 and p is a prime dividing s, then p should divide rn and

hence p divides r which is a contradiction to the assumption that (r, s) 5 1.
Therefore, s 5 1 and a [ Z.

EXERCISE 11(D)

2. Follows from Theorem 11.4.2.

4. 1 1 x2 is irreducible over Z3 and hence Z3[x]/<1 1 x2> is a field and has
exactly 9 elements.
6. a is a unit in F[x] and hence f (x) and af (x) are associates.
8. 1 1 2x2 1 x4 has no root in Z3, but it is reducible in Z3[x], since 1 1 2x2
1 x4 5 (1 1 x2)(1 1 x2).
10. Follows from the fact that R is an ID if and only if R[x] is an ID.
12. Consider the epimorphism f (x, y)  f (x, 2 x) of F[x, y] onto F[x]. Its
Kernel is <x 1 y>. Therefore, F[x, y]/<x 1 y>  F[x]. Similarly, F[x,
y]/<x 1 y>  F[y].
14. For each n [ Z1, (1 1 2xn)(1 1 2xn) 5 1 in Z4[x] and hence 1 1 2xn is
a unit in Z4[x] for all n [ Z1. Also, (2xn)2 5 0 in Z4[x] and hence 2xn is
a nilpotent in Z4[x] for all n [ Z1.

CHAPTER 12
EXERCISE 12(A)

2. (i) 1, 21
(ii) 1, 21, i, 2i
(iii) 61, 6 (1 2), 6 (1 2)
(iv) 61
(v) R 2 {0}
(vi) 1, 2, 3, 4

Q001-Algebra-111001_Answers.indd 50 9/16/2011 10:01:13 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-51

4. ( 2  3 )( 2  3 )  1 and ( 2  3 ) 3  3 2 3
6. 1 1 2x 1 3x2, 2 1 4x 2 x2, 3 2 x 1 2x2, 23 1 x 1 5x2, 22 1 3x 1 x2,
21 22x 2 3x2
8. Use induction on n.
10. Let a 5 bu, where u is a unit. Then, a 5 bc ⇒ bu 5 bc ⇒ u 5 c.

EXERCISE 12(B)

2. Refer Worked Exercise 11.3.1.

4. R is a PID and S is a homomorphic image of R, then S > R/I for some ideal
I of R and any ideal of R/I is of the form <x>/I for some x [ R such that I ⊆
<x> and then x 1 I generates <x>/I. Therefore, R/I and hence S is a PID.
6. If p is an odd prime, then the ideal 2,  p generated by 2 and
 p in Z[  p ] is not a principal ideal.
8. Let I be a nonzero ideal of Z[ 2] and A 5 {a2 1 2b2 : 0  a 1
b 2  I }. Choose x5 a 1 b 2  I such that a2 1 2b2 is least in A.
Then, I is the ideal generated by x in Z[ 2].
10. If I is an ideal of S21R and J 5 {a [ R : a [ I}, then, J is an
1
ideal of R and hence J 5 <a0> 5 a0R for some a0 [ J. Now,
a0
1 ( 1 1
a
 I as  I ⇒ a  as  s  I ⇒ a  J ⇒ a  a0 r , r  R and as  0  rs .
1 )
12. See Ex. 4 above and consider R/{0} 5 R.
14. M is a maximal ideal of a PID if and only if M 5 <p> for some irreduc-
ible element. Also, <p> 5 <q> ⇔ p and q are associates.
16. R is not an integral domain and hence not PID.
18. Let P be a nonzero ideal of R. Suppose that P is primary. Then, the radi-
cal r(P) is prime and hence maximal ideal in R, so that r(P) 5 <p> for
some prime element p. If p [ P, then P 5 <p>. Suppose p  P. Choose
least n such that pn [ P. Then prove that P 5 <pn>. Conversely, suppose
that P 5 <pn>, where p is a prime in R and n [ Z1. Let a and b [ R
such that ab [ P and a  P. Then, pn|ab and pna. Let m be the largest
integer such that pm|a. Then, pmc 5 a for some c [ R such that pc. Also,
m , n and ab 5 pm(cb), since pm11|ab, it follows that p|b and bn [ <pn>
5 P. Thus, P is a primary ideal.

Q001-Algebra-111001_Answers.indd 51 9/16/2011 10:01:14 AM

 A-52   Algebra – Abstract and Modern

EXERCISE 12(C)

2. Let a1, a2, …, an be elements of R. If some ai 5 0, then 0 is the unique

l.c.m. of a1, a2, …, an. Suppose that ai  0 for each i, we can write

a1  p1r1 p2r2  ptrt and a2  p1s1 p2s2  ptst

for some distinct

t
prime elements p1, p2, …, pt and nonnegative integers
max{ri , si }
 p
ri, si. Then, i1 i is the l.c.m. of {a1 a2}. We can extend this to a1,
a2, …, an using induction.
4. Follows from the observations that, for any s [ S, 1s is a unit in S21R and
a
hence as is prime (irreducible) if and only if 1 is prime (irr) in S21R.
Also, if a is prime in R, then 1a is prime in S21R.
6. Any nonzero nonunit a can be written as a product p1p2 … pn, where pi’s
are primes and hence a , pn. Let I be a nonzero proper ideal in R. Then,
pn [ I for some n [ Z1. Let n be the least such that pn [ I. Then, <pn>
⊆ I. Also, 0  a [ I ⇒ a is a nonunit ⇒ a 5 pm, n  m ⇒ a 5 pm. pm2n
[ <pn>. Thus, I 5 <pn>.
8. Note that <a> ⊆ <b> ⇔ b divides a. Also, if a  p1r1 p2r2  pnrn , where
pi’s are distinct primes and ri’s are nonnegative integers, then there are at
most finitely many principal ideals containing <a>. Thus, there cannot
be a strictly increasing infinite sequence of principal ideals.
10. These follow from the fact that any two nonzero elements a and b can
be written as a  p1r1 p2r2  pnrn , and b  p1s1  pnsn , where pi’s are distinct
primes and ri and si are nonnegative integers and that
n
(a, b) = ∏ pi min{r , s } i i

i1
n
and [a, b ] ∏ pi max{r ,s }
i i

i1

     

EXERCISE 12(D)

2. There are no nonzero nonunits in any field.

4. Use Theorem 12.4.4 and Corollary 12.4.5 or the fact that f (x)
of degree  3 is reducible over a field F if and only if f (x) has a
root in F.

Q001-Algebra-111001_Answers.indd 52 9/16/2011 10:01:15 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-53

(i) f (x) 5 15 2 9x2 1 6x3 1 2x4 is primitive, 3 is a prime, 3|15, 3|9,

3|6, 32 and 3215. Therefore, by Corollary 12.4.5, f (x) is irreduc-
ible over Z.
(ii) There are no roots of 3 1 2x2 1 x3 in Q and hence it is irreducible.
(iii) 4 1 2x 1 x3 has no roots in Z5 and hence it is irreducible.
(iv) Let f (x) 5 1 1 x2 1 x5 [ Z2[x]. Suppose that f (x) is reducible,
since f (x) has no root in Z2, f (x) has no linear factors and hence
there exists an irreducible factor of degree 2 of f (x). 1 1 x 1 x2
is the only quadratic irreducible polynomial over Z2. Therefore,
1 1 x 1 x2 should divide f (x). If 1 1 x2 1 x5 5 (1 1 x 1 x2)(a
1 bx 1 cx2 1 dx3), then by comparing the coefficients, we get
that a 5 1 5 b 5 c 5 d, which is a contradiction. Thus, f (x) is
irreducible over Z2.
(v) 9 2 x3 has no root in Z31 and hence 9 2 x3 is irreducible over the
field Z31.
(vi) If f (x) 5 1 1 x3 1 x6, then f (x 1 1) 5 3 1 9x 1 18x2 1 21x3 1
15x4 1 6x5 1 x6, which is irreducible over Q (by the Eisenstein’s
criterion) and hence f (x) is irreducible over Q.
(vii) By the Eisenstein’s criterion, 5 1 10x 1 15x3 1 2x5 is irreducible
over Q and Z (it is primitive).
(viii) 2 1 2x 1 x4 is irreducible over Z and Q.
(ix) Since 4 is a root of 9 2 x3 in Z11, 9 2 x3 is reducible over Z11.
(x) This is irreducible over Q, by the Eisenstein’s criterion.
6. Use Eisenstein’s criterion.
8. x2, 1 1 x2, 1 1 x 1 x2 and x 1 x2 are the only polynomials of degree
2 over Z2 and among these 1 1 x 1 x2 is the only irreducible
­polynomial.
10. a 1 bx 1 x2, a, b [ Zp, are all the monic polynomials of degree 2 in
Zp[x] and these are p2 in number. Among these the reducible polynomi-
als are of the form (x 2 a)(x 2 b), where a and b [ Zp. Since (x 2 a)
(x 2 b) and (x 2 b)(x 2 a) are the same polynomials, the number of
p( p1)
reducible polynomials is pC 2  p( 2  p). Thus, the number of
irreducible monic polynomials of degree 2 over Zp is

 p ( p 1)  p ( p 1)
p 2   p  .
 2  2

Q001-Algebra-111001_Answers.indd 53 9/16/2011 10:01:15 AM

 A-54   Algebra – Abstract and Modern

EXERCISE 12(E)

2. Z[x] is not a PID and hence not an Euclidean domain.

4. R[x] is an Euclidean domain ⇒ R[x] is a PID ⇒ R is a field.
6. g(2a) 5 g((21)a) 5 g(21)g(a) 5 g(a) (by Theorem 12.5.1 (2)).
8. For any a  b 3  Z( 3), define g ( a  b 3)  |a 2  3b 2 | note that
g ( a  b 3)  0 if and only if ( a  b 3)  0. Verify that g is a gauge
function for Z[ 3] with respect to which Z( 3) is an Euclidean
domain. Use the technique of the proof of Worked Exercise 12.5.3.

EXERCISE 12(F)

2. 2 5 (1 1 i)(1 2 i) and 2 divides neither 1 1 i nor 1 2 i

17 5 (4 1 i)(4 2 i) and 17 divides neither 4 1 i nor 4 2 i.
4. This is true in any Euclidean domain and Z[i] is a Euclidean domain.
6. Let g(a) be least in A 5 {g(b) : b is a nonzero nonunit in R}. Let x [ R.
Then, x 5 qa 1 r for some q and r [ R such that r 5 0 or g(r) < g(a).
If r 5 0, then a divides x. If r  0, then, by the least property of g(a),
r is a unit and x 2 r 5 qa and hence a divides x 1 (2 r) and 2r is a
unit in R.
8. g.c.d.{23 1 11i, 8 2 i} 5 2 1 i (use the Euclidean algorithm).
10. x 5 1 and y 5 23i.

CHAPTER 13
EXERCISE 13(A)

2. (i), (ii), (iii), (iv) are R-submodules of R4 and others are not.
4. Z is a faithful Z-module and, for any nonzero proper ideal I of a
ring R with unity, R/I is an R-module, which is not faithful, since
Ann(R/I) 5 I.
6. Follows from P ⊆ P 1 Q and P ∩ Q ⊆Q.
8. Let M 5 Z2 3 Z2. Then, M is a Z-module. Let N 5 {(0, 0), (1, 1)},
P 5 Z2 3 {0} and Q 5 {0} 3 Z2. Then, N, P, Q are Z-submodules of M.

Q001-Algebra-111001_Answers.indd 54 9/16/2011 10:01:16 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-55

Here, we have N ∩ (P 1 Q) 5 N ∩ M 5 N, N ∩ P 5 {(0, 0)} and N ∩ Q 5

{(0, 0)} and hence (N ∩ P) 1 (N ∩ Q) 5 {(0, 0)}  N ∩ (P 1 Q).
10. If 0  x [ N0, then <x> is a nonzero R-submodule and hence N0 ⊆
<x> ⊆ N0, so that N0 5 <x>.
12. Clearly, uN is an R-congruence on M. If u is any R-congruence on M and
N 5 u(0), then N is an R-submodule of M and uN 5 u.

EXERCISE 13(B)

4. If f and g are R-endomorphisms of M, it can be verified that f 1 g and

f o g are also R-endomorphisms of M.
6. Direct verification.
8. If f and g [ M, then f 1 g [ M and af [ M for all a [ R. Also, (f 1 g)9
5 f 9 1 g9 and (af )95 (af )9 and hence f  f 9 is an R-homomorphism of
M into RR.
10. If f : M → N is an R-isomorphism and x1, …, xn [ M, then M 5 <x1, …,
xn> ⇔ N 5 < f (x1), …, f (xn)>.
12. For any a [ Q, define fa : Q → Q by fa(r) 5 ar. Then, fa [ EndZ(Q) and
a  fa is a ring isomorphism of Q onto EndZ(Q). (If f [ EndZ(Q) and
n ) f ( n)  an
f (1) 5 a. Then, verify that f (n) 5 an for all n [ Z and m f (m
and hence f 5 fa.)
14. Define f : M → M/A 3 M/B by f (x) 5 (x 1 A, x 1 B). Then, f is an
R-homomorphism and ker f 5 A ∩ B. Also, let (x 1 A, y 1 B) [ M/A
3 M/B, x, y [ M. Then, x 2 y [ M 5 A 1 B and hence x 2 y 5 a 1 b
for some a [ A and b [ B. Put z 5 x 2 a 5 y 1 b. Then, z 2 x [ A and
z 2 y [ B and hence f (z) 5 (z1 A, z1 B) 5 (x 1 A, y 1 B). Thus, f is an
epimorphism. By Theorem 13.2.4, M/A ∩ B > M/A 3 M/B.
16. Clearly, uf is an R-congruence and uf (0) 5 ker f. For any s 5 f (z),
z [ M,

{x [ M : f (x) 5 s 5 f (z)} 5 z 1 ker f.

EXERCISE 13(C)

2. No in (i) and (iv).

Yes in (ii) and (iii).

Q001-Algebra-111001_Answers.indd 55 9/16/2011 10:01:17 AM

 A-56   Algebra – Abstract and Modern

4. For each i [ I, let Ni 5 {a [ RI : |a| ⊆ {i}}. Then, each Ni is an

R-submodule of R(I). Define f i : Ni → R by f i(a) 5 a(i). Then, f i is an
R-isomorphism and hence Ni > R, as R-modules. Also, g : R( I ) → ⊕ N i ,
iI
defined by g (a)(i )  f i1 (a(i )), is an R-isomorphism.
6. Let N   M i be the direct product and Pi : N → Mi be the ith projec-
iI
tion. Then, each Pi is an R-epimorphism. By Theorem 13.3.1 (2), there
exists a R-homomorphism P : N → M such that pi o P 5 Pi. Since Pi is
surjective, so is pi. Thus, pi is an R-epimorphism.
pi
M
Mi

�
�i

8. Let N  {a   M i : |a| is finite}.Then, N is an R-module ­(R-submodule

iI
of  M i ). Define gi : Mi → N by gi(xi)(j) 5 xi or 0 according as
iI
j 5 i or j  i. Then, gi is an R-monomorphism. By Definition 13.3.3
(2), there exists homomorphism g : M → N such that g o f i 5 gi.
Since gi is an injection, f i is also an injection. Thus, each f i is a
R-monomorphism.
fi M
Mi

g
gi

10. Let each fi be an R-isomorphism. Then, for any x, y   M i ,

i

f (x) 5 f (y) ⇒ f (x)(i) 5 f (y)(i) for all i [ I

⇒ f i(x(i)) 5 fi(y(i))
       ⇒ x(i) 5 y(i) for all i [ I ⇒ x 5 y.

Thus, f is an injection. Also, for any y   N i , choose xi [ Mi such that

i
fi(xi) 5 y(i) for each i [ I. Then, f (a) 5 y, where a   M i is defined by
i
a(i) 5 xi. Thus, f is an R-isomorphism.

Q001-Algebra-111001_Answers.indd 56 9/16/2011 10:01:18 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-57

EXERCISE 13(D)

2. M is precisely the external direct sum ⊕ M x where Mx is the ­R-module

x X
R for each x [ X. Since R is simple as an R-module (by Example 13.4.1
(1)), so is each Mx. Therefore, M is a completely reducible R-module.
M is simple ⇔ M > R ⇔ X 5 {x} (if x  y [ X, then Mx can be treated
as a nonzero proper submodule of M).
4. Since N is proper, M/N  {0}. By Theorem 13.4.3, M = N ⊕ ( ⊕ M a ),
I
where {Ma}a[I is a nonempty (since M  N) family of simple R-
modules. Then, M / N  ⊕ M a and hence M/N is completely
aI
­irreducible.

6. If M  ⊕ M a and M a  ⊕ M ai , for each a [ , where each M ai is

a iI a
simple, then M  ⊕ M ai and hence M is completely reducible.
a , iI a

EXERCISE 13(E)

2. Any element of M can be written as r1x1 1  1 rnxn, x1, …, xn [ B and

r1, …, rn [ R. Also, since B is linearly independent, this expression is
unique.

4. If Bi is a basis for Mi, then prove that B1 3 B2 3  3 Bn is a basis for

M1 3 M2 3  3 Mn.

6. Let I be an ideal of Z. Then, I 5 nZ for some n  0 and {n} is a basis

for I as a Z-module.

8. Let M be an R-module. Let N  ⊕ Rx , where Rx is the R-module R.

x M
Then, N is a free R-module and define f : N → M by f (a)   a( x ) x.
x M
Then, f is an R-epimorphism and hence M is a homomorphic image of
the free R-module N.

10. If f : M → N is an R-isomorphism and B ⊆ M, then B is a basis for M ⇔

f (B) is a basis for N.

12. If B is a basis and B  A ⊆ M, then choose y [ A such that y  B. Then,

y 5 r1x1 1  1 rnxn for some xi [ B. Then, r1x1 1  1 rnxn 1 (21)y 5 0
and hence A is linearly dependent.
14. Let B be a basis for M. For each x [ B, choose yx [ M such that f (x) 5
yx 1 N. Any y [ F can be uniquely expressed as y 5 r1x1 1  1 rnxn,

Q001-Algebra-111001_Answers.indd 57 9/16/2011 10:01:19 AM

 A-58   Algebra – Abstract and Modern

where xi [ B and ri [ R and now define g ( y )  r1 y x1   rn y xn . Then,

g : F → M is a homomorphism and
w n  
g ( y )  M  ∑ ri ( y xi  M )  ∑ ri f ( xi )  f ∑ ri xi   f ( y ).
i1 i1

 i 

EXERCISE 13(F )

2. Let e4 5 (0, 0, 0, 1, 0) and e5 5 (0, 0, 0, 0, 1). Then, {e1, e2, e3, e4, e5} is
a basis of F5.

4. The matrices of D with respective to B and C are respectively

 0  0
0 0 0 0
  0 0 0 0

1 0 0 0 0  1 0 0 0 0
   
0 2 0 0 0 and 1 2 0 0 0.
   
0 0 3 0 0  1 1 3 0 0
   
0 0 0 4 0 1 1 1 4 0

The matrices of transformations from B to C and C to B are respectively

 1 0 0 0 0  0
  1 0 0 0

1 1 0 0 0 1 1 0 0 0
   
 0 1 1 0 0 and 1 1 1 0 0.
   
 0 0 1 1 0 1 1 1 1 0
   
 0 0 0 1 1  1 1 1 1 1 

CHAPTER 14
EXERCISE 14(A)

2. [K : Q] 5 4, since [K : Q( p ) ] 5 2 and [ Q( p ) : Q] 5 2

4. [Fn : Q] 5 [Fn : Fn21][Fn21 : Fn22] … [F1 : Q] 5 2n

and [R : Q] > [Fn : Q] 5 2n for all n.
6. No, since char(Zp) 5 p and char(Q) 5 0
8. For any field F, char(F ) 5 O(1) in the group (F, 1).

Q001-Algebra-111001_Answers.indd 58 9/16/2011 10:01:20 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-59

EXERCISE 14(B)

2. {a0 1 a1r 1 a2r2 1 a3r3 : a0, a1, a2, a3 [ Q}, where r4 2 2 5 0

4. Infinite (see Ex. 4 in Exercise 14(A)).
6. (i) x2 2 5
(ii) 23 2 10x 1 x2
(iii) 7 2 10x 1 x2
(iv) 21 1 2x2 1 x4
(v) 625 2 58x2 1 x4
(vi) x2 2 6

8. F ⊆ K ⊆ E ⇒ [E : K][K : F] 5 (E : F ) 5 p, a prime
⇒ [E : K] 5 1 or [K : F] 5 1
       ⇒ K 5 E or K 5 F.

10. Since a2 [ F (a), we have F ⊆ F (a2) ⊆ F (a) and hence n 5 [F (a2) : F]

# [F (a) : F] 5 m, say. Let f (x) 5 a0 1 a1x 1 … 1anxn5 be the minimal
polynomial of a2 over F. Then, a0 1 a1a2 1 … 1 ana2n 5 0 and hence
a is a root of the polynomial g(x) 5 a0 1 a1x2 1 … 1anx2n. This implies
that m divides 2n since m is odd, m divides n and hence m # n. Thus,
[F (a2) : F] 5 n 5 m 5 [F (a) : F].
Also since F (a2) ⊆ F (a), it follows that F (a2) 5 F (a).

EXERCISE 14(C)
∞
2. Let F  ∪ Fn . a and b [ F ⇒ a [ Fn and b [ Fm for some n and m ⇒
n1
a and b [ Fn or Fm, according as m # n or n # m ⇒ a 6 b and a ? b [
Fn ⇒ a 6 b and a ? b [ F. Thus, F is a field and Fn is a subfield of F for
each n.

4. If K is algebraically closed, then, for any a [ K, xn 2 a has all the roots

in K (by Theorem 14.3.1) and hence n , |K| for all n [ Z1. Therefore, K
is infinite.

6. Use the argument given in Ex. 2 above and the fact that, for any a and b
[ , Fa ⊆ Fb or Fb ⊆ Fa.

8. Let a be a root of p(x) in E. Let b be another root of p(x) in some exten-

sion K of F. Then, F[x]/<p(x)>  F (a) and F[x]/<p(x)>  F (b) and

Q001-Algebra-111001_Answers.indd 59 9/16/2011 10:01:20 AM

 A-60   Algebra – Abstract and Modern

hence F (a)  F (b). Therefore, we can treat f (x) as a polynomial over

F (a) as well as over F (b). Also, E is a splitting field of f (x), treated as a
polynomial over F (a). If L is a splitting of f (x), treated as a polynomial
over F (b), then by the uniqueness of the splitting field, E  L. Since
f (x), as a polynomial over F, splits in E, we get that L ⊆ E(b). Since
L contains all the roots of f (x), in particular, b [ L, we have E(b) ⊆ L.
Therefore, E(b) 5 L. Also,
[E(b) : F] 5 [L : F] 5 [L : F (b)][F (b) : F]
     5 [E : F (a)][F (a) : F] 5 [E : F]

and, since E ⊆ E(b), we have E(b) 5 E and, in particular, b [ E. Thus, E

contains all the roots of p(x).
10. Consider f (x) 5 xp 2 1 5 (x 2 1)(xp21 1 xp22 1 … 1 x 1 1). The splitting
field of f (x) [ Q[x] is same as the splitting field of g(x) 5 1 1 x 1  1
xp21 [ Q[x]. If a is a primitive pth root of unity, that is, a 5 e2p/p, and E is
the splitting field of f (x), then
E 5 Q(1, a, a2, …, ap21) 5 Q(a).

Since g(x) is irreducible in Q, we have [E : Q] 5 p 2 1.

EXERCISE 14(D)

2. Use induction on n.
4. Use induction on m.
6. Use Theorem 14.4.4(2).

EXERCISE 14(E)

2. F 5 GF (23). F * 5 F 2 {0} is a cyclic group of order 7 and hence any

1  a [ F * generates F *.
4. 1 1 x 1 x2 is an irreducible polynomial over Z2 and hence Z2[x]/<1 1
x 1 x2> is a field with four elements, say F. Then, F 5 {0, 1, a, 1 1 a},
where 1 1 a 1 a2 5 0. Then, x2 1 ax 1 1, x3 1 x2 1 ax 1 (1 1 a) and
x4 1 ax3 1 ax2 1 (1 1 a) x 1 1 are irreducible polynomials of degrees
2, 3 and 4, respectively over F.
6. Let F be a finite field. Then, by Theorem 14.5.1, |F | 5 pn, where n [ Z1
and p 5 char(F ). If p 5 2, then the map a : F → F defined by a(a) 5 a2
is an injection and hence a bijection, so that any x [ F can be expressed
as x 5 a2 5 a2 1 02 for some a [ F. Next, suppose that p . 2; that is,

Q001-Algebra-111001_Answers.indd 60 9/16/2011 10:01:21 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-61

p is an odd prime. Let x [ F and S 5 {x 2 a2 : a [ F} and T 5 {b2 : b

pn1
[ F}. Then, S and T are subsets of F, |S| 5 (pn 2 1)/2 1 1 5 and
pn1 2
|T| 5 (note that a  2a for any a [ F, since char(F )  2). Since
2
|S ∪ T| # pn and |S ∪ T| 5 |S| 1 |T| 2 |S ∩ T|, it follows that S ∩ T is
nonempty and hence there exists x 2 a2 5 b2 [ T ∩ S; that is, x 5 a2 1
b2 for some a and b [ F.
8. Since char(F) 5 p, the map a : F → F defined by a(x) 5 xp is an
epimorphism of the ring F (recall that x|F | 5 x for all x [ F and |F | 5
pn). Since F is finite and a : F → F is a surjection, we get that a is an
injection and hence, for each a [ F, there exists unique x [ F such
that a 5 a(x) 5 x p.

CHAPTER 15
EXERCISE 15(A)

2. We have 1 1 z 1 z2 5 0. Consider a 5 2 1 z. Then, a2 5 2 1 2 2 z 2

(1 1 z) 5 1 1 (2 2 2 1)z [ Q(a) and hence (2 2 2 1)z [ Q(a).
Therefore, (2 2 2 1)3 5 ((2 2 2 1)z)3 [ Q(a); that is, 8 2 1 2 6 2
(2 2 2 1) [ Q(a) and hence 2 [ Q(a). Since ( 2 2 1)z [ Q(a), it
follows that z [ Q(a). Thus, Q( 2 , z) ⊆ Q(a). Clearly, a [ Q( 2 , z)
and hence Q(a) ⊆ ( 2 , z).
4. Let K be a finite extension of a finite field F and |F | 5 pn, where char(F )
5 p 5 char(K). Then, K is also finite and |K| 5m pm for some m [ Z1.
We have K 5 F (a) for some a [ K. Hence, a p  a . Therefore, a is a
separable element and hence F (a) (5K) is a separable extension of F.
6. x4 2 2 is an irreducible polynomial over Q and 21/4 is a root of x4 2 2
belonging to Q(21/4); but 21/4i, which is also a root of x4 2 2 does not
belong to Q(21/4). Therefore, Q(21/4) is not a normal extension of Q.
8. Use Theorem 15.1.5.

EXERCISE 15(B)

2. Clearly the identity map and the map c : C → C, defined by c(a 1

ib) 5 a 2 ib, are in G(C/R). Also, if f [ G(C/R), then f (i)2 5 f (i2) 5
f (21) 5 1 and hence f (i) 5 6i so that, for each a 1 ib [ C, f (a 1
ib) 5 a 1 ib or a 2 ib. Therefore, f 5 Id, or c. Thus, G(C/R) is a two
element group and cyclic.

Q001-Algebra-111001_Answers.indd 61 9/16/2011 10:01:23 AM

 A-62   Algebra – Abstract and Modern

4. Let 3 2  u. Then, u3 5 2 and u [ R. We are given that s is the auto-

morphism of E 5 Q(v, u) such that s(r) 5 r for all r [ Q, s(v) 5 v2
and s(u) 5 uv. Clearly, s o s 5 Id and hence H 5 {Id, s} is a subgroup
of G(E/Q). Let EH be the fixed field of H. Since s (uv2) 5 s(u)s(v)2 5
uv2 and hence uv2 [ EH, Q(uv2) ⊆ EH. On the other hand, let a [ EH.
That is, s(a) 5 a. Observe that {1, v} is a basis of Q(u, v) over Q(u)
and that {1, u, u2} is a basis of Q(u) over Q. Therefore, {1, u, u2, v, vu,
vu2} is a basis of Q(u, v)(5 E) over Q and hence we can write
a 5 r0 1 r1u 1 r2u2 1 r3v 1 r4vu 1 r5vu2
for some ri [ Q. Now,
a 5 s(a) 5 r0 1 r1uv 1 r2u2v2 1 r3v2 1 r4v3u 1 r5v4u2
5 r0 1 r1uv 1 r2u2(21 2 v)1 r3(21 2 v) 1 r4u 1 r5vu2
5 (r0 2 r3) 1 r4u 2 r2u2 2 r3v 1 r1vu 1 (2r2 1 r5)vu2
From this, it follows that r0 5 r0 2 r3, r1 5 r4, r2 5 2r2, r3 5 2r3, r4 5
r1, and r5 5 2r2 1 r5 and hence r3 5 0 5 r2, r1 5 r4 and r0 and r5 are
arbitrary. Therefore,
a 5 r0 1 r1(u 1 vu) 1 r5vu2 5 r0 2 r1v2u 1 r5(v2u)2
Thus, a [ Q(v2u). Therefore, EH 5 Q(v2u).
6. Since 0 5 a5 2 1 5 (a 2 1)(1 1 a 1 a2 1 a3 1 a4) and a  1, it follows that
a is a root of the polynomial f (x) 5 1 1 x 1 x2 1 x3 1 x4 which is irreduc-
ible over Q. Therefore, [Q(a) : Q] 5 4. The roots of x5 2 1 are 1, a, a2, a3
and a4 which are in Q(a). Therefore, Q(a) is the splitting field of x5 2 1 over
Q. Thus, Q[a] is a normal extension of Q and the Galois group G(Q(a)/Q)
is of order [Q(a) : Q] 5 4. Observe that {1, a, a2, a3} is a basis of Q(a) over
Q and hence any element of Q(a) can be uniquely expressed as
r0 1 r1a 1 r2a2 1 r3a3, ri [ Q
and the four elements of G(Q(a)/Q) are Id, f, g and h, where f (a) 5 a2,
g(a) 5 a3 and h(a) 5 a4, which form a cyclic group with f (or g) as a
generator. Thus, G(Q(a)/Q)  Z4.

EXERCISE 15(C)

2. It is known that the Klein four-group is the group of order 4 in which

each element is of order # 2 and is isomorphic to Z2 3 Z2. We prove
that the Galois group of x4 1 1 [ Q[x] is of order 4 and f 2 5 Id for
each f in this group. Let F 5 Q(a), where a 5 eip/4. Since a, a3, a5
and a7 are the roots of x4 1 1, it follows that F is the splitting field of

Q001-Algebra-111001_Answers.indd 62 9/16/2011 10:01:23 AM

 Answers/Hints to Selected Even-Numbered Exercises   A-63

x4 1 1 over Q. Also, [F : Q] 5 4, since x4 1 1 is irreducible over Q.

char(Q) 5 0 and hence F is a normal separable extension of Q. There-
fore, G(F/Q) is of order [F : Q] 5 4. Note that any element b of F can
be expressed as b 5 r0 1 r1a 1 r2a2 1 r3a3 with ri [ Q, and any f [
G(F/Q) is determined by its value at a; since f (a) must be a root of x4
1 1, the four elements of G(F/Q) are Id, f, g and h, where f (a) 5 a3,
g(a) 5 a5 and h(a) 5 a7. Also, note that f2(a) 5 f (f (a)) 5 f (a3) 5 f (a)3
5 a9 5 a, h2(a) 5 a49 5 a and g2(a) 5 a25 5 a and hence f2 5 Id 5 g2
5 h2. Thus, G(F/Q)  Z2 3 Z2.
4. Let Em 5 Q(a), G 5 The Galois group G(Em/Q). Then, |G| 5 [Em : Q]
5 2m. Choose subgroup Hm21 of order 2m21 in G (Use Sylow Theorem I).
Then, Hm21 is of index 2 in G and hence normal in G. Let Em21 be fixed
field of Hm21. Then, Em is a normal extension of Em21. Also, Q ⊆ Em21 ⊆
Em, [Em : Em21] 5 2 and

E 
G  m 
E   Q 
G  m1  =  2m1
 Q   Em 
G  
 E 
m1

Now, choose a subgroup Hm22 of order 2m22 in G(Em21/Q) and let Em22
be the fixed field of Hm22. This process can be continued to construct the
required fields Q 5 E0 ⊂ E1 ⊂ … ⊂ Em21 ⊂ Em 5 Q(a).
6. Note that the only nonidentity automorphism in G(Q(a)/Q) is the f
for which f (a) 5 a2 (and hence f (a2) 5 a, since a3 5 1). Therefore,
G(Q(a)/Q) is the two element group, which has only two subgroups,
namely {Id} and the whole group and the corresponding fixed fields are
Q(a) and Q, respectively.

CHAPTER 16
EXERCISE 16

2. If f [ G(C/R), then f (i)2 5 f (i2) 5 f (21) 5 21 and hence f (i) 5 i or

2 i. Therefore, G(C/R) 5 {Id, f}, where f (a 1 bi) 5 a 2 bi for any
a 1 bi [ C.
4. Refer Worked Exercise 16.3.3.
6. If f (x) 5 x3 1 x2 2 2x 2 1, then f (x 12) 5 x3 1 7x2 1 14x 1 7. Now,
use the Eisenstein’s criterion.

Q001-Algebra-111001_Answers.indd 63 9/16/2011 10:01:24 AM

 This page is intentionally left blank.

Q001-Algebra-111001_Answers.indd 64 9/16/2011 10:01:24 AM

 Index

A D
algebraic extensions, 14-8–14-19 Dedikind theorem, 15-11–15-12
algebraically closed fields, determinants, 2-43–2-54
14-20–14-26 dihedral group, 6-23–6-36
alternating group, 6-23–6-36 direct products, 8-1–8-10
automorphism groups, 15-10–15-18 disjoint set, 1-7
fixed fields and, 15-10–15-18 division algorithm, 4-13–4-14,
automorphisms, 5-29–5-36 11-15–11-23

B E
binary systems, 3-3–3-16 endomorphisms, 5-29
Burnside’s theorem, 7-23–7-24 equivalence relations, 1-21-1-25
partitions and, 1-21–1-25
C Euclidean division algorithm, 12-36
Euclidean domain, 12-36–12-43
canonical homomorphism, 5-5,
Euler’s theorem, 4-34
10-23
Euler–Totient function, 4-19
cardinality of sets, 1-27–1-35
even permutation, 2-44
Cartesian product, 1-11
extension fields, 14-3–14-38
Cauchy’s Theorem, 7-28–7-29
Cayley’s theorem, 6-1–6-6
Chinese remainder theorem, F
10-29–10-33 Fermat’s theorem, 4-34–4-35
choice function, 2-33 fields, 2-25
class equation, 7-21–7-22 integral domains and, 9-43–9-50
class of sets, 1-6 extensions of, 14-3–14-8
compass, 16-20 polynomials and, 11-25–11-29
completely reducible modules, finite fields, 14-33–14-38
13-31–13-41 finite groups, 3-45–3-50
congruence modulo, 2-14–2-21 finite set, 1-28
coordinate-wise ordering, 2-31 finitely generated abelian groups,
cycles, 6-11–6-20 8-10–8-27
cyclic extensions, 16-5–16-7 invariants of, 8-29–8-32
cyclic groups, 4-12–4-22 fundamental structure theorem for,
cyclotomic polynomial, 12-33 8-16–8-17

Q001-Algebra-111001_Index.indd 1 9/21/2011 4:51:17 PM

 I-2   Index

first principle of induction, 2-2–2-3 L

fundamental theorem of algebra, Lagrange’s theorem, 4-30–4-37
16-2–16-5 lexicographic ordering, 2-31
fundamental theorem of arithmetic,
2-5–2-6
M
fundamental theorem of functions, 1-25
matrices, 2-34–2-41
matrix unit, 2-40
G maximal ideals, 10-43–10-54
Galois Theory, 15-1–15-26 modules, 13-2–13-8
applications of, 16-1–16-26 completely reducible modules,
fundamental theorem of, 13-31–13-41
15-19–15-24 direct products and sums for,
group tables, 3-45–3-50 13-18–13-29
groups of small order, 8-32–8-36 homomorphism and quotients of,
groups, 3-16-3-29 13-10–13-16
homomorphisms of, 5-1–5-34 simple reducible modules,
elementary properties of, 3-32–3-43 13-31–13-41
multiple roots, 14-27–14-32
H
homomorphism N
fundamental theorem of, 5-16–5-22 nilpotent, 9-17
normal extensions, 15-1–15-10
normal subgroups, 4-39–4-43
I
ideals, 10-1–10-17
O
identity homomorphism, 5-3
identity matrix, 2-39 orbits, 7-8–7-17
indexed class, 1-6 ordering, 2-30–2-34
infinite set, 1-28
integers, 2-1–2.10 P
integral domains, 12-1–12-50 partial ordering, 2-30
applications to number theory, partially ordered set, 2-30
12-44–12-49 perfect field, 15-3
factorization in, 12-1–12-50 polynomial rings, 11-1–11-36
divisibility in, 12-2–12-9 prime ideals, 10-34–10-41
principal ideal domains and, principal ideal domains,
12-10–12-16 12-10–12-16
unique factorization domains and, principle of well-ordering,
12-18–12-24 2-33–2-34
irreducible polynomials, 11-31–11-35 proper subset, 1-6
isomorphism theorem, 5-23
Q
K quotient groups, 4-45–4-50
Kronecker’s theorem, 14-15 quotient rings, 10-20–10-27

Q001-Algebra-111001_Index.indd 2 9/21/2011 4:51:17 PM

 Index   I-3

R solvable groups, 16-8–16-11

radical extension, 16-12 stabilizers, 7-8–7-17
rational numbers, 2-23–2-24 subgroups, 4-2–4-10
recursion theorem, 2-9–2-10 cosets of, 4-24–4-28
rings, 9-3–9-52 submodules, 13-2–13-8
homomorphisms of, 9-29–9-34 subrings, 9-25–9-29
polynomial tings and, 11-1–11-36 subsets, 1-3–1-9
special types of, 9-35–9-42 Sylow Theorem II, 7-33–7-34
embeddings of, 10-56–10-65 Sylow Theorem III, 7-34–7-35
examples and elementary properties symmetric difference, 1-9
of, 9-3–9-14
special elements in, 9-16–9-20 T
characteristics of, 9-22–9-24 third isomorphism theorem, 5-26–5-27
subrings and, 9-25–9-29 totally ordered set, 2-31
rings of polynomials, 11-1–11-13 trivial homomorphism, 5-3
ruler, 16-20
U
S unique factorization domains,
second isomorphism theorem, 12-18–12-24
5-23–5-24
second principle of induction, 2-4 V
separable extensions, 15-1–15-10
vector spaces, 13-42–13-51
set intersection, 1-7
set union, 1-7
sets, 1-3–1-9 Z
simple reducible modules, 13-31–13-41 Zorn’s lemma, 2-32–2-33

Q001-Algebra-111001_Index.indd 3 9/21/2011 4:51:17 PM