Basic Abstract

Algebra

Otto F. G. Schilling
Purdue University

W. Stephen Piper
Purdue University
and the Center for Naval Analyses
an affiliate of the University of Rochester

Allyn and Bacon, inc.

Boston
© Copyright 1975 by Allyn and Bacon, Inc.
470 Atlantic Avenue, Boston. All rights reserved.
Printed in the United States of America.
No part of the material protected by this copy-
right notice may be reproduced or utilized in any
form or by any means, electronic or mechanical,
including photocopying, recording, or by any
informational storage and retrieval system,
without written permission from the
copyright owner.

Library of Congress Cataloging in Publication Data

Schilling, Otto Franz Georg, 1911—1973.

Basic abstract algebra.

Bibliography: p.
I. Algebra, Abstract. I. Piper, William Stephen
1940- joint author. 11. Title.
QA162.S34 512'.02 74-4547
 Contents

Preface ix

Chapter 1 • Notation and introductory Concepts

§1.1 Set Notation I
§1.2 Equivalence Relations and Classes 6

Chapter 2 • Arithmetic of Integers 10

§2.1 Algebraic Properties of Integers 10
§2.2 Analytic Properties of Integers and induction 14
§2.3 The Division Algorithm 21
§2.4 Ideals in Z 23
§2.5 Divisibility 25
§2.6 Prime Numbers 31
§2.7 Unique Factorization 33
§2.8 Congruences Modulo m 37
§2.9 Addition and Multiplication of Cosets Modulo in 40
§2.10 Definitions and Examples 43
§2.11 Simultaneous Systems of Congruences 48
§2.12 Two Topics in Number Theory 52

Chapter 3 • Introduction to Ring Theory 55

§3.1 Basic Elements of Ring Theory 56
§3.2 Ring Homomorphisms 62
§3.3 Direct Sums of Rings 71
§3.4 Residue Class Rings of Integers 74
§3.5 Direct Sum Decompositions of Zm 77
§3.6 Integral Domains and Fields 83

Chapter 4 • Aspects of Linear Algebra 94

§4.1 Vector Spaces 95
§4.2 Linear Independence and Bases 101
§4.3 Transformations of Vector Spaces 105
§4.4 Matrices and Linear Transformations 110
§4.5 Determinants 121

Chapter 5• Polynomials and Polynomial Rings 125

§5.1 Polynomial Rings 126
 Contents

§5.2 Divisibility and Factorization of 130

§5.3 Residue Class Rings for Polynomials 136
§5.4 Residue Class Fields of irreducible Polynomials 142
§5.5 Roots of Polynomials 146
§5.6 The Interpolation Formula of Lagrange 150
§5.7 Polynomial Functions 152
§5.8 Primitive and Irreducible Polynomials 154
§5.9 Characteristic Polynomials of Matrices 157

Chapter 6 e Group Theory 162

§6.1 Elements of Group Theory 163
§6.2 Subgroups and Orders of Elements 171
§6.3 Coset Decompositions and the Theorem of Lagrange 177
§6.4 Normal Subgroups and Factor Groups 180
§6.5 Group Homomorphisms 184
§6.6 Cyclic Groups 191
§6.7 Groups of Permutations 194
§6.8 The Isomorphism Theorems of Group Theory 201
§6.9 Automorphisms, Center, Commutator Group 205
§6.10 Direct Product 212
§6.11 Homomorphisms of Abelian Groups 217

Chapter 7 • Selected Topics In Group Theory 221

§7.1 Finitely Generated Abelian Groups 221
§7.2 Characters of Finite Abelian Groups 230
§7.3 Bijections of Sets 235
§7.4 The Class Equation and Normalizers 237
§7.5 The Elementary Theorems on Sylow Subgroups 241
§7.6 Composition Series and the Jordan-H older Theorem 248
§7.7 Groups with Operators 255
§7.8 Modules 258

Chapter 8 • Field Theory 266

§8.1 Algebraic Elements 266
§8.2 Finite Fields 271
§8.3 The Theorem of the Primitive Element 275
§8.4 Equivalence of Fields 277
§8.5 Counting of Isomorphisms and Separability 280
§8.6 Prelude to Galois Theory 284
§8.7 The Fundamental Theorem of Galois Theory 288
§8.8 Consequences of the Fundamental Theorem 292
§8.9 Algebraic Closure 299

Chapter 9 • Selected Topics in Field Theory 304

§9.1 Cyclotomic Fields 305
§9.2 Equations Solvable by Radicals 310
§9.3 Constructions with Ruler and Compass 314
§9.4 Trace and Norm 319
§9.5 Theorem of the Normal Basis 325
 Contents vil

§9.6 Hubert's Theorem 90 and Noether's Equations 328

§9.7 Kummer or Radical Extensions 334
§9.8 Unique Factoriza(ion Domains and
Elementary Symmetric Functions 341
§9.9 The Fundamental Theorem of Algebra 348
§9.10 Finite Division Rings 351
§9.11 Simple Transcendental Extensions 354
§9.12 Perfect Fields 358

Bibliography 363
Index of Mathematicians 367
Index of Notation 373
Index of Mathematical Terms 379
 Preface

The Science of Pure Mathematics, in its modern develop-

ments, may claim to be the most original creation of the human
spirit.
Alfred North Whitehead
Science and the Modern World

Perspective
Why study abstract algebra? What good is it? The answer to such
questions lies, of course, in the questing human intellect that admits no
restraints upon its probes. nor limits to its adventures. Utility marches after
discovery. The formal systems we are about to study have played important
roles in the physical sciences and are of increasing importance in the
biological and social sciences. While the algebra may be abstract, it none-
theless provides the setting or language for consideration of concrete physical
and social problems.
Our approach to abstract algebra involves thinking of abstract as a
verb form, not as an adjective. The development of algebraic structures as
abstractions of the properties of integers and polynomials is shown. Thus
familiar properties, observed in examples, are given a more general axiomatic
setting. The intent is to involve the student in the evolution of algebraic
concepts as a participant, rather than as a spectator having to assimilate
lists of axiomatic descriptions. Besides the pedagogical merit of this
approach, there is the mathematical merit of more closely indicating methods
of mathematical research and advancement. Thus our proofs are designed
to show details of arguments that later facilitate solution of exercises.

Content
This book is intended as a text for the first undergraduate and
introductory graduate courses in abstract algebra. It has been so used at
Purdue University. The order of topics and discussion derives from class-
ix
x Preface

room experience. Students first encountering abstract algebra commonly

have difficulty working with cosets of subgroups, factor groups, and residue
class rings. These difficulties center on insufficient understanding of equival-
ence relations and classes. We address this topic in Section 1.2 and subse-
quently in the text refer to this initial discussion. Similarly, mathematical
induction is carefully presented in Section 2.2 with subsequent references.
Early chapters stress algebraic concepts and techniques so that the student
may understand algebraic structures and gain facility in handling algebraic
arguments.
Chapter 2 provides a careful introduction of the algebraic structure
and congruence relations of integers, culminating in concrete problems in
residue class rings of integers (Section 2.9) and the solution of simultaneous
systems of congruences (Section 2.11). Thus, the student is provided with
examples of rings and with the basic number theory necessary for questions
in the theory of groups involving orders of elements, cyclic groups, and
concepts like direct products.
We provide in Section 2.4 explicit proof of the equivalence of the
Well-Ordering Principle and two statements of the Principle of Induction.
As these equivalence proofs are often omitted from undergraduate texts,
some classes may want to accept parts of this section without proof. Later
we continue to state inductive arguments in detailed steps so that students
may gain an appreciation of their proper use, rather than skirt the issue by
blithely concluding arguments "by induction."
Rings are introduced before groups, contrary to a common sets—
groups—rings—fields approach to abstract algebra. The reason is pedagogical;
students are already familiar with integers and polynomials. Ready reference
to integers and polynomials, where the axiomatic algebraic properties and
their logical consequences are introduced, provides the student with examples
that develop and test the concepts of ideals, nilpotence, idempotence,
factorization, and congruence.
After discussing the algebraic structure of the integers, we abstract
that structure to rings. Chapter 3 begins with the basic elements of ring
theory (Sections 3.1 and 3.2): axiomatic description, subrings, ring homo-
morphisms, and ideals. Sections 3.4 and 3.5 consist of an optional discussion
of orthogonal idempotents and ideals in residue class rings of integers and
of the external and internal direct sum structure of such rings, the concept
of direct sum having been introduced in Section 3.3. These sections may be
omitted without loss of continuity by instructors who prefer to follow Section
3.2 immediately with the discussion of integral domains, fields, and poly-
nomial rings (Sections 3.6, 5.1, and 5.2). Here extensive sets of exercises
provide ample choice of material for further development.
The basics of linear algebra are presented in Chapter 4. This chapter
is intended as a refresher for students who have had a prior course in linear
algebra or as a concise source of the linear facts used in field theory
(Chapters 8 and 9) and in some of the examples of groups (Chapter 6).
 Preface ii

Chapter 5 is devoted to polynomials. The first four sections continue

the ring theory of Chapter 3 with discussion of polynomial rings and residue
class rings of polynomials, including Kronecker's construction of roots of
irreducible polynomials. The remaining sections introduce properties of
polynomials that will be useful in subsequent field theory study (especially
roots of polynomials in Chapter 8 and other topics in Chapter 9).
Groups (Chapter 6) are not introduced until students have had
sufficient experience in congruence theory to handle arguments involving
orders of elements. The first section provides an extensive set of examples
of groups from different branches of algebra. The student is urged to use
these examples to verify the details of group-theoretic arguments while
carrying out explicit computations with the elements of particular groups.
Appropriate explicit examples of normal subgroups, quotient groups,
homomorphisms, and cyclic and permutation groups appear throughout
the chapter. The theorem that the image of a group homomorphism is
isomorphic to its domain modulo the kernel is given special attention together
with its immediate consequences (Section 6.8). Chapter 6 concludes with
three optional sections designed to give the student more experience with
group theory and to prepare him for subsequent topics.
For honors sections or classes seeking greater emphasis on group
theory, Chapter 7 includes all the topics common to an advanced under-
graduate or first graduate course: the structure of finitely generated abelian
groups, group characters, the Sylow theory (Sections 7.3 through 7.5),
composition series, and groups with operators and modules. These topics
are presented independent of one another, so that they may be studied in
any order. Most topics were chosen not only for their inherent algebraic
interest but also for later reference in Chapter 9, Selected Topics in Field
Theory.
Chapter 8 builds to the Galois theory. We emphasize the concept of
prolongation of mappings, since it prepares the student for more advanced
work in cohomological algebra and its applications. While both separable
and inseparable extensions are treated, one may also conveniently limit the
discussion to fields of characteristic zero.
Chapter 9 offers a broad selection of field-theoretic topics. Sections
9.2 and 9.3 on solvability by radicals and ruler and compass constructions
provide answers to classical questions based on the Galois theory. Conse-
quences of the Galois theory include the inseparable case because of its
importance in modern applications of the subject, especially for abelian
function fields and algebraic groups. The proof of the algebraic closure of
the complex numbers follows the ideas of Lagrange and Gauss.
Introductions to individual sections serve as previews, stating
briefly the principal directions of the section and often the relationship
of the topics covered with those in other sections. They may involve terms
about to be, but not yet, defined. These introductions should be helpful to a
student re-reading the book or reviewing his course work.
 Preface

As one learns best by doing, we have provided over 1400 exercises

by which the student may test developing algebraic skills.

Course Selections
A variety of courses can be designed from the material offered,
each tailored to the objectives of the students and instructors. We suggest
the following for consideration as content of a typical

One-semester undergraduate Chapters 1—2

course, with no linear algebra Chapter 3, less Sections 3.4—3.5
prerequisite Chapter 4, Sections 4.1—4.3
Chapter 5, Sections 5.1—5.4
Chapter 6, Sections 6.1—6.6

One-semester undergraduate Same as above with Chapter 4

course, linear algebra omitted, and additional topics
prerequisite added from Chapters 5—6

Honors undergraduate course Chapters 1—6

First graduate course Review Chapter 2

Chapters 3 and 5
Selection of topics from Chapter 7
Chapter 8, perhaps less Section
8.2, and only emphasizing the
separable case
Second semester of above courses Continuation into Chapters 6—7,
with work from Chapter 4 as
needed
Topics from Chapter 9 for gradu-
ate courses.

Appreciation
No undertaking of the nature of this text can be the sole work of
two individuals. Our manuscript has evolved over many semesters of trial.
It owes its present existence and form to the provocative probes of colleagues
and students. We wish to thank especially Michael Drazin for his helpful
criticism, April Kihlstrom for her valuable comments, Bonnie Schnitta for
editorial assistance, and Betty Lewis and Judy Snyder for their patience in
typing and retyping the manuscript. We are also grateful to the several
reviewers for their assessments, notably Thomas R. Berger of Trinity
College, Connecticut, Richard L. Faber of Boston College, David Hertzig
 Preface xiii

of the University of Miami, Eugene F. Krause of Arizona State University,

Joel Roberts of the University of Minnesota, and C. R. B. Wright of the
University of Oregon.
Otto F. G. SchiHing
W. Stephen Piper
West Lafayette, Indiana
June 1973

Otto Schilling died in June 1973 just as our manuscript was entering
the production process. In reviewing the editing and composition phases of this
project I have tried to be faithful to the writing done it was a joint project,
and 1 have been guided by the senior author's insistence on mathematical
quality.
His Purdue University colleagues wrote of him: "Otto was a
dedicated teacher. He tried to awaken enthusiasm for mathematics by
absolute integrity and strict adherence to rigorous standards." This book,
which he never saw in final form, caps the end of his long career in algebra.
Otto Schilling's interests developed at Gottingen, where he was writing a
thesis on algebraic number fields under Professor Emnzy Noether when she
was compelled to leave her native Germany. Upon her recommendation,
Otto Schilling completed his thesis under Professor 1-lelmut Hasse at Marburg.
He then left Germany (in 1934) for a year at Trinity College, Cambridge
and a long teaching career in the United States. His last twelve years were at
Purdue University, after twenty-two years at tile University of chicago.

W. Stephen Piper
Falls Church, Virginia
June 1974
 1

Notation and Introductory

Concepts

A basic concept in algebra is that of equivalence relation, together

with the related equivalence, congruence, or residue classes, also called cosets.
After defining needed basic set theoretic notation [*1.1], we provide the
definition and some examples of equivalence relations, so that the reader
may focus on the concept itself before encountering residue class rings of
integers modulo an integer in residue class rings of polynomials
and cosets of ideals Exercise 31 and of groups
Although subsequently we reflect common usage by referring to
equivalence classes by the various names of residue or congruence classes
or cosets, the underlying concept is always that of the defining equivalence
relation. Especially important is the fact that two equivalence classes defined
by an equivalence relation are either equal or disjoint. Chapter 1 is brief, by
design, so that the foundation is quickly laid for heuristic examination of the
integers in Chapter 2.

§1.1 Set Notation

Throughout algebra, in fact throughout mathematics, one constantly
refers to sets or collections of elements. In general we shall denote sets by
capital letters A, B, C, ... S. ... and their elements by lower case letters
.
 Notation and Introductory Concepts chapter 1

a,b,c, etc. If a belongs to the set or collection A, we write a e A, and if a does

not belong to A, we write a A.
Membership in a set S is often denoted either by explicitly listing the
elements of the set or by stating a rule for membership, as follows:

A = {l,2,3,4,8,9} asetof6integers
or S = {integers n such that n is even}.

The latter has the following common abbreviated form:

S = {integers n : n is even}.
Mathematicians have evolved certain standard symbols and termin-
ology to aid in stating concepts. For instance, we shall make frequent of
the following symbols and set designations.

"implies," or "is a sufficient condition that"

"if and only if," "implies and is implied by," "is a necessary and
sufficient condition," "is equivalent to," or "means the same
as"
e "is an element of" or "belongs to"
N set of positive integers (also called the set of natural numbers)
Z set of integers
Q set of rational numbers
R set of real numbers
C set of complex numbers

The standard abbreviation for "there exists" is 3, and for "for all" is V,
although we shall not need to use them in this text.
The set S mentioned above can also be expressed as

S= {n€Z:niseven}.
The set with no elements belonging to it is called the empty or null
set, designated by 0. A set A is said to be a subset of a second set B if
(*)
(Implicitly the sign means that this holds for all a e A.) Translated into
English words this says that if a is an element of A, then it is also an element
of B. Alternatively, we can say that every element in A belongs to B. The
relationship in line (*) is denoted by
A c B.

Although many authors do not distinguish between A B and A B, we

shall reserve A c B for the situation when A is strictly or properly contained
§1.1 Set Notation 3

in B, meaning that the set B contains A and also elements not belonging to A.
We say then that A is a proper subset of B. Thus,
A
B B, and there isan elementbEBsuch
For every set S,
ØcS and ScS.
The subsets 0 and Sofa set S are commonly called trivial subsets; any other
subset is called a nontrivial subset.
Many proofs involve proving that two sets Consist of the same
elements; that is, that they are equal. The customary technique for proving
the equality of two sets A and B is to prove that each is contained in the other;
that is, to show that
a A aE B (and hence A c B),
and conversely that
b B beA (and hence B A).

For sets A and B. we make the following definitions.

Difference: B\A = {b B: b A}.

Union: AuB={x:x€AorxeB}.
(The "or" used here is inclusive and means that x belongs to A or B, or to
both.)
Intersection: A B= (x:xeA andxeB}.
Symmetric difference:
A B = (A t.s B)\(A B)

Observe that for two arbitrary sets A and B, we have

and ØcB\AcBcAuB.
The preceding definitions are commonly illustrated by Venn diagrams,
named after the English logician John Venn (1834—1923). For given sets

0E115
 Notation and Introductory Concepts chapter 1

we represent the following sets by the shaded portions of the diagrams:

Ar'1B

AAB

The definitions of union and intersection extend to a finite collection

{A1, of sets as follows.
n
Union: U A1 = {x:x EA1,forSo,flei= l,...,n).
i= I

Intersection: flA1 = {x:x€A1,foralli= l,...,n}.

With only a notational change, we can treat infinite collections of sets

indexed by an arbitrary index set 1. The statement "1 is an index set
for a collection of sets E" is used in an intuitive sense, meaning simply that
to each element i e I there corresponds a set A1 in the collection E. Also,
to each set in there corresponds at least one i E I. The set of natural num-
bers is a commonly used infinite index set, but any set (finite or infinite) can
serve as an index set.

REMARK. Use of an index set in referring to the union and intersection of sets in
a collection E is convenient, but can be avoided by using

se1
and fls
SeE

to denote the union and intersection, respectively, of all sets in the collection Z.

We shall call sets A and B disjoint if their intersection is empty,

that is, if A and B have no elements in common. We say that the sets in a
collection are mutually disjoint ifS1 m = 0 for any distinct i,je I.
Given sets A and B, we often consider the set A x B of ordered pairs
of elements in A and B. This set, called the cartesian product of A and B(after
the French philosopher and mathematician René Descartes, 1596—1650), is
§1.1 Set Notation 5

defined to be the set

Ax B = {(a,b):aeA, bEB}.
The pairs (a, b) are ordered in the sense that the first element (component)
listed belongs to the first set A and the second element to the second set B.
Ordered pairs (a,b), (a',b') are defined to.be equal if and only if a = a' and
b = h'. The usual cartesian coordinates in the euclidean plane are ordered
pairs of real numbers; the plane is simply the cartesian product R2 = R x R
of the set of real numbers with itself (together with a distance function).
In our subsequent study of rings and of groups and
7.1] we shall extend the concept of the cartesian product to n sets (or rings,
or groups) A ,, . . .,

A, x A2 x x = i= 1,...,n},
the set of ordered n-tuples of elements a, E A., I = I, ...,n.

Exercises

1. For arbitrary sets A and B, verify the following statements.

a. Au(B\A)—AuB
b. Ari(B\A)=Ø
c.
2. Let S = {1,2,3,4}. Write out explicitly all subsets of S. How many
subsets are there? How many are proper subsets?
3. If a set has elements, how many subsets are there? How many proper
subsets?
4. For sets A, B, C verify the following statements.
a.

A (A B) (A C)
A C) = C)
h.
I.
5. For sets A, B, C verify that
a.
b. (AxB)U(AxC)=Ax(BUC)
c. (AxB)\(AXC)=Ax(B\C)
d. (AxC) = A
xB A B
a one-one correspondence between (A x B) x C and
Ax(BxC).
 Notation and Introductory Concepts cimpter I

6. a. For subsets A, B, C of a given set 5, verify that union and inter-

section satisfy the following statements:
(I) Au(BUC)=(AUB)L)C
(ii)
Theseare statements of the associativily of the union and intersection
of sets. Associativity, as an algebraic concept, is discussed in
3.1, and 6.1, among others.
b. State why, despite Exercise 5(f), the cartesian product of sets is not
associative.

§1.2 Equivalence Relations and Classes

In this section we introduce the concept of "relatedness" of elements
in a set S. For example, in the set of all living persons we are familiar with
many examples of relations or relatedness. We say "a is a child of b" or
"a is a sister of b" to indicate a (family) relationship. Mathematically,
consider a set S, and let stand abstractly for a relation between two
elements of S, such that for any pair a, b in S, either a b (read "a is in
relation to b") or a b (read "a is not in relation to b").
Three common descriptive properties of relations on a set S are
the following.
Reflexive property: s s for all s S.

Symmetric property: s S' s' s for all s,s' S.

Transitive property: s s', S' SI' s s" for all s,s',s" S.

Our interest in relations is restricted to ordering relations (to be

discussed in §2.2) and equivalence relations. A relation is called an
equivalence relation, denoted if it is reflexive, symmetric, and transitive.
Several distinct equivalence relations may be defined on a particular
set S. For instance, consider the fourth example below with differing values
of m.

Examples of Sets with Equivalence Relations

1. In R, equality of real numbers is an equivalence relation.

2. For a set of triangles S. similarity ot triangles is an equivalence relation on S.
3. In R, real numbers r, r' can be defined to be equivalent if r — r' a Z.
4. In Z, for a fixed nonzero integer tn, integers a, a' can be defined to be equivalent
if their difference a — a' is an integral multiple of ni.

Once an equivalence relation has been defined on a particular

set 5, we consider subsets of equivalent elements of S. For s a S, the set
[sJ = {s' a 5: s' s}
§1.2 Equivalence Relations and Classes 7

is called the equivalence class or coset of s with respect to the relation

(At present we shall use the term equivalence class; in the later study of
rings and groups the term coset is common.) The
class of an element s S consists of all elements in S which are -equivalent
to S.

In the fourth example above,

[3] = (a e Z : a 3 is an integral multiple of nz}.
In the specific case in which m 6,

[3] = {3+6k:kEZJ,
[6] = (6k : k e Z} = [0].

If is an equivalence relation on a set S, then every element s e S

belongs to some class, namely [sJ, because of the reflexive
property of an equivalence relation.
We now turn to an important fact concerning equivalence classes.

Proposition. For a giveii equivalence relation on a set S. equivalence

classes [s], [a] are either equal or disjoint; that is,
[s] = [a] or {s) [a] = 0.
Proof Suppose that [s] [a] 0. We then show that [s] = [a] by first
considering b e {s] n [a]. By definition of the equivalence classes,

By symmetry and transitivity of the relation we have then a s. Since

*

[a] = {a' : a' a}. again by transitivity we have a' s for all a' e [a].
Hence [a] [s]. Similarly, [s] [a], and thus [s] = [a], as was to be
shown.

As a corollary. whose proof is left as an exercise, we have

[s] = [a] s a.
A partition of a set S is a collection of mutually disjoint subsets
of S whose union isS, i.e.,
S. S1 = 0 unless S. =

S:= US1.
1€,

For example, the equivalence classes corresponding to an equivalence

relation on a set S constitute a partition of S.
Conversely, given any partition {S,},ei of S, indexed by some
(finite or infinite) set I, we can define an equivalence relation on S as follows.
8 Notation and Introductory Concepts chapter 1

Define two elements s,s' E S to be equivalent if they belong to the same

subset in the partition of S. Then for each s E S, [s] = S., where I is such
that s e Thus each equivalence relation on S defines a partition of S,
and each partition gives rise to an equivalence relation.
Given an equivalence relation on S. we shall need to distinguish
carefully between elements s of a set S and equivalence classes of elements
of S. We have
s e [s] S.

Note that [sJ is not an of S; it is a subset of S. However, [s] is an

element of the set Z of equivalence classes in S. Thus [s] will be considered
both as an element and as a set. We must ascertain how it is to be considered
from the context. In the same way, we should distinguish between the element
s0 E S and the subset (s0} of S consisting only of the single element s0.

Exercises

1. Verify that the third and fourth examples of sets with equivalence relations
satisfy the properties of an equivalence relation.
2. With reference to Example 3, show that every real number,• is equivalent
tosomer',Or'< 1.
3. a. In Example 4 let m be 7 and write out all the equivalence classes.
b. Repeat part (a) with m 10.
4. Show by example that a given set might have several equivalence relations
defined on it, and that in general different relations will give rise to
different equivalence classes.
5. How many distinct equivalence relations can be defined on a set of three
elements?
6. In a set S with equivalence relation prove that [s] = [a] s a.
7. Prove that the equivalence relations in Examples 3 and 4 satisfy the
following statement:
x y, x' y+y,.
8. Prove that the equivalence relation in Example 4, but not that in Example
3, satisfies the following statement:
x y, x' V xx'
9. Let be an equivalence relation on the finite set S. If each of the m
equivalence classes with respect to has n distinct elements, how many
elements are there in the set S? Why?
10. Consider arbitrary sets S. T and a mapping 1: S T, defined for all
s e S. Define s' to be -equivalent to s if f(s) = f(s') T. Prove that is
an equivalence relation on S.
§1.2 Equivalence Relations and Classes

11. a. With the naappingf and relation defined in Exercise 10, iff satisfies
the surjective property that for each Ic T the set
{suS:f(s)=:} 0,
show that there is a one-to-one correspondence between -equiva-
lence classes in S and elements of T.
b. Show that every equivalence relation on S determines a surjective
map f to the set of -equivalence classes of elements of S.
12. Show that the set of all lines in the cartesian planeR2 which are parallel
to a given line constitute a partition of R2.
 2

Arithmetic of Integers

The integers Z have properties of addition and multiplication well

known to all students. We review these properties, both algebraic and
analytic to include the equivalent Principles of Well-Ordering and
induction, in preparation for abstracting them to sets more general than the
integers. The basic algebraic properties of interest are addition and multi-
plication; the analytic property of importance here is that of ordering.
Algebraic concepts of the division algorithm ideals
prime element unique factorization and congruence relations
are introduced within the context of the integers to provide rationale
and experimental material for general contexts Chapters 3 and 5].
Unique factorization domains are studied in detail in §9.8. Ring-theoretic
properties of residue classes of integers provide explicit examples of
finite rings. We emphasize the Chinese Remainder Theorem I] because
of its significance in the direct sum decomposition of Zm presented in
Chapter 3. it is also important in the theory of algebraic numbers and in
algebraic geometry. The Euler q-function and Fermat's Little Theorem are
the two number theory topics presented in §2.12.

§2.1 Algebraic Properties of Integers

Familiarity with the ordinary laws for addition and multiplication
of integers a, b Z, the comparison of integers (a> b, and equivalently
b < a), and the properties of absolute value lal is assumed. For the sake of

I0
§2.1 Mgebraic Properties of Integers 11

completeness and ease of reference, the following summaries of these axioms

are given.
I. Equality. The basic properties of equality (an equivalence
relation) are reflexivity, symmetry, and transitivity [cf. §1.2].
11. Addition. For every pair of integers a,b there exists a unique
integer s called the sum of a,b arid denoted by s = a+b. This operation of
addition obeys the following axioms for all a, b, c e Z.
(I) Uniqueness of the sum: a+b = a* + b*.
=
(We also say that the sum is He/I-defined.)
(ii) Associativity: a+(b+c) = (a+b)+c.
(iii) Commutativity: a+b = b+a.
(iv) Identity (existence of an identity): There exists an integer 0, called
zero, such that 0+a = a for all a e Z.
(v) Inverse: For each a e Z, there exists an integer a' such that a+a' = 0.
1Ff. Multiplication. For every pair of integers a, b there exists a
unique integer p called the product of a, b and denoted by p = a b, or briefly,
p = ab. This operation of multiplication obeys the following axioms for all
a,h,ce Z:
(i') Uniqueness of the product: : ab = a*b*.
= }
(We also say that the product is iiell-defined.)
(ii') Associativity: a(bc) = (ab)c.
(iii') Commutativity: ab = ba.
(iv') Identity (existence of an identity): There exists an integer I, such that
I •a = a, for all a e Z.
(v * )
.
Cancellation:
ac=bc a = b.
The general associative laws for addition and multiplication of more
than three integers are proved in the discussion on groups These
general laws and the proof of identities like (a+b)+(c+d) = a+[(b+c)+d]
are consequences of the Principle of Induction
IV. Distributive Axiom. A connection between addition and
multiplication of arbitrary integers a, b, c is provided by the distributive
axiom of multiplication over addition: a(b + c) = ab + ac.
Note the absence of an axiom (v'), "inverse," for multiplication
analogous to the inverse axiom (v) of addition. Experience in working with
integers shows that ax I has a solution x in Z if and only if a = + 1 or — I.
The equation 2x = I has no integral solution.
We conclude this section with a series of consequences of the addition
and multiplication axioms for integers; some of the proofs are left as exercises.
Property 1. The additive identity or zero element is unique. That is,
a-i-z=a forallaeZ='z=O.
 Arithmetic of Integers chapter 2

Let a = 0. Then 0+: = 0, and by symmetry of the equivalence

relation =, 0 = 0+:. Since 0+z = z, by the properties of 0, we have, by
transitivity,
0= 0+: = z, andbysymmetryz =0.

Property 2. Furthermore, a = a + z for a E Z implies z = 0.

Adding to each side a' Z, such that a'+a = 0 (i.e., an additive

inverse of a), we have
0= a' +a= a' + (a+z)
= (a'+a) + z = 0+ z = z.

Note that this result is stronger than Property I because of the weaker
hypothesis.

Property 3. The cancellation law for addition ho(ds; that is,

a+c= b+ c a = b.

Let c' Z be an additive inverse of c. Adding c' to the right of each

side of the equality, we have
a= a +0= a + (c+c') = (a+c) + c'
= (b+c) + c' = b + (c+c') = b + 0 = b,
making use of the properties of the inverse of an element, of the 0 element,
and of associativity.
Because of commutativity, we also have the left cancellation law
c+a= c +b a = b.

Property 4. The additive inverse a' of an element a Z is unique.

The following proof of uniqueness of the additive inverse is typical

of many uniqueness proofs in mathematics. Suppose there exist two integers
a' and a" satisfying the properties of the additive inverse; that is,
a + a" = 0, a + a' = 0.

By transitivity of the equality relation

a + a" = a + a'.
and by the law of cancellation (Property 3)
= a'.
Hence the two inverses are equal.
§2.1 Algebraic Properties of Integers 13

COMMENT. It is common usage to let —a denote the unique additive of

a c Z. Then b+(—a) is written more simply as b—a. Observe that the inverse or
negative of a, namely — a, is described as a solution of an equation involving elements
of Z:
a + x = 0.
No appeal whatsoever is made to the ordering properties of integers.

Property 5. Every equation a+x = b. for a,b e Z, has a unique solution

x€Z.
Adding —a to each side, using associativity and the properties of 0,
we have x = —a+b. The uniqueness of the solution follows from the can-
cellation law in the same way as the uniqueness of the additive inverse was
proved.
Some authors cite the solvability property as an axiom in place of
the existence of the inverse. The solvability and inverse axioms are equivalent,
in the sense that in conjunction with associativity and identity, each implies
the other.

Property6. ForanyaeZ,a0=0.
Since a = a + 0, we have, using the uniqueness of multiplication,
aa = a(a+O) = aa + by distributivity.
Therefore = 0 by Property 2.

Property 7. The multiplicative identity is unique; that is,

am = a for all a Z m = 1.

Property 8. Furthermore, am = a for some nonzero a e Z implies m = I.

Property 9. The rule of signs for multiplication states that for all a, b e Z,
(—a)(—b) = ab.

This rule is not dependent upon the ordering properties of the

integers. It will now be proved using only the preceding algebraic properties.
(—a)(-—b) = (—a)(—b) + 0 property of the zero element
= (— a) ( — b) + a .0 Property 6
= (—a)(-—h) + a(b—b) property of inverse
= (—a)(—b) + [a(—b)+ab] distributivity, commutativity
= [(— a) + a] ( — b) + associativity, distributivity
= 0(—b) + ab property of inverse
= 0 + ab Property 6
= ab property of the zero element.
Similarly, we can prove (—a)b = —(ab) = a(—b).
14 Arithmetic of Integers chapter 2

Property 10: The cancellation law (v*) under III (Multiplication) above is
equivalent to the statement that the product of two nonzero integers is not
zero.

Assuming the cancellation law we must show that

ab = 0, a 0 b = 0.

Since ab = 0 = by cancellation, b= 0. To prove the converse, note that

ac = be (a—b)c = 0.
Then c 0 implies a—b = 0, or a = b.
The integers are an example of an algebraic structure called a ring.
While rings are formally introduced in §3.1, we note now that a ring R is a
set of elements together with two rules for composing elements (commonly
called addition and multiplication) which satisfy axioms 11(i) through (v);
111(1'), (ii'), and (iv'): and IV.
If the multiplication in a ring R satisfies the commutative axiom
1lI(iii') and the cancellation law we call the ring Ran integral domain
[see §3.6] because it abstracts all of the algebraic properties of the integers.
In what follows we shall refer to Z as the ring of integers.

§2.2 Analytic Properties of Integers and Induction

The previous section emphasized addition and multiplication of
integers and their governing laws. We now discuss briefly the basic rules for
integers dealing with their "size" and "positiveness." These rules are generally
referred to as the ordering properties of the integers. Our introduction of
ordering is followed by two equivalent, fundamental principles: the Principle
of Well-Ordering and the Principle of Induction (two versions of the latter
are presented). The axioms of induction and well-ordering are essential in
many mathematical proofs. Which of the two equivalent principles is used
in a particular proof is a matter of convenience or ease of reference, so usage
varies.
A word of warning is appropriate. The set of integers has rules of
addition, multiplication, and ordering. These properties alone do not, however,
define or categorize the integers, because other very different sets can have
these same properties.
Some readers might wish to skip over the proofs of the equivalence
of induction and well-ordering, as these are not commonly presented in
undergraduate courses. The proofs are not so difficult as they are formal or
technical.
The set of integers Z contains as a subset the natural numbers
N= {n Z : n > 0),
§2.2 Analytic Properties of Integers and Induction 15

where the symbol > is read "greater than." The set N has the following
properties:
Additive closure: x,y e N x+y N.

Multiplicative closure: x,y e N xy E N.

The Law of Trichotomy: For xe Z precisely one of the three statements
xeN, x=0, —x€N
is true.
A simple formal consequence of these facts about N is that a2 N
for all nonzero integers a. To prove this assertion note that if a 0, then
either a E N or —a e N. In the first case a2 N in accordance with (2) above;
in the second the rule of signs I, Property 91 implies that
a2 = (—a)(—a) EN, since(—a) EN.
Asaspecialcase, I= 1.1 EN.
Next an ordering of the integers is defined by writing
a<b (equivalently, b > a) ifb — a e N
(i.e., if a is strictly less than b). This ordering or relation among pairs of
integers satisfies:
I. a < b and b < e a<c (transitivity).
2. forallc€Z.
3. foralineN.
4.
5. and c<d=.a+c<b+d.
Proofs of these properties are left to the reader.
The "less than" relation or ordering of integers is transitive, but not
symmetric or reflexive, and hence is not an equivalence relation. A similar
ordering of integers is "less than or equal to," denoted by , which is
transitive and reflexive, but not symmetric.
Furthermore there is associated to each a E Z its absolute value
lal e N u {O}, defined as follows:
(i) If a N, then lal = a.
(ii) If a = 0, then laP = 101 0.
(iii) If —a e N, then al = —a.
Immediate consequences of the definition are that for all a, b Z
the following relations hold:

—laP a laP, Pal = I—al, labI = alibi.

16 ArIthmetic of Integers chapter 2

The important triangle inequality

also holds. it is demonstrated by considering the following relations:

and

Adding inequalities, we have

—(!aI+lbI) a+b <IaI + IbI.
Hence, Ia+bI + Ibi.

REMARK. The algebraic and analytic properties of the integers described so far
are shared by other systems of numbers; for example, the rational numbers Q, the
real numbers R, and real numbers of the form a+b where a,b e Z [see §3.6].
We now consider an ordering property that distinguishes Z from these other systems
of numbers.

A set S issaid to be well-ordered if every nonempty subset S' has a

least element That is, s for all s E S'.

Principle of Well-Ordering. The positive integers N form a well-ordered set.

In fact, more generally for any fixed k e Z the subset
S=
of Z is well-ordered. Applications of the Well-Ordering Principle often
involve nonnegative integers, not just the positive ones.

We accept the Principle of Well-Ordering as an axiom. After indi-

cating some of its simple, though remarkable consequences, we show that it
is equivalent to the Principle of Induction.

Proposition 1. Between 0 and I there are no integers.

Proof Recall that 0 and I are defined to be the additive and multiplicative
identities, respectively, and that we proved, using the rule of signs, that I > 0.
The proof that no integer ni satisfies 0 < m < 1 will be indirect. Let
S= (m€N:0<rn<l}
and suppose S 0. Since S N, and by assumption is nonempty, the
Well-Ordering Principle ensures the existence of a least integer a, such that
0 < a < I. Since N is closed under multiplication, a2 E N, and further
0< a <I 0< a2 < l•a = a.

Therefore, a2 e S. hut a2 <a, contradicting the choice of a as the least

element in S. Hence S must be empty.
§2.2 Analytic Properties of Integers and Induction 17

Thus Proposition 1 states that I is the least element in N. Note the

following corollary.

Corollary. There are no integers between z and z + I, for each z Z.

Proposition 2. If a set S c N contains I and if se S implies s+ I e S, then

S=N.
Proof Again we choose an indirect argument. Let
SC = {n E N n S}.
Our objective is to show that SC is the empty set, and thus that S = N.
Suppose that S' 0. As a subset of N, SC has a least element a
by the Well-Ordering Principle. The positive integer a cannot be I, since
I e S. Therefore a> I, and a— I e N. Because a is the least element in Sc,
a— 1 e S. Then by hypothesis a (a— I)+ I belongs to S. and so a eS Sc.
This intersection is empty by the definition of SC, so that a S SC. We
have arrived at a logical contradiction and must conclude that cannot
be nonempty. That is, S must equal N.
Proposition 2 implies the following customary principle.

Principle of Induction. Assume that with each positive integer a there is

associated a statement (proposition) P(n) that is either true or false. Then,
P(n) is true for all positive integers a provided the following conditions hold:
(i) P(l) is true.
(ii) The truth ofF(s), for any given se N, implies the truth of P(s+ 1).
For the proof note that the set S of positive integers s for which
P(s) is true satisfies the hypotheses of Proposition 2.
An alternate form of the Principle of Induction is used frequently
in proofs.

Principle of Induction (Alternate Form). Assume, as before, that with each

positive integer n there is associated a proposition P(n). Then, P(n) is true
for all positive integers a provided the following conditions hold:
(i) P(l) is true.
(ii) For each rn E N, the truth of P(s), for all positive s < tn, implies
the truth of
Proof The alternate form of induction can be derived directly from the
Well-Ordering Principle or from the original form of the Principle of Induction.
We choose the latter course, leaving the former as an exercise. If P(n), for each
n e N, is a proposition for which P(l) is true and, if for each me N, the
truth of F(s), for all positive s < m, implies that of P(rn), then we wish to
conclude that P(n) is true for all a e N.
18 Arithmetic of Integers chapter 2

Let Q(s) be the proposition that P(k) is true for all Ic <s. First
note that Q(l) is true, and then note that the truth of Q(s) implies that of
Q(s+ I), for all s N, because the truth of P(k) for all k s was assumed
to imply the truth of P(s+ 1), i.e., of P(k) for all k s+ I. Applying the
original form of induction to Q(s), we conclude that Q(n) is true for all
n E N. But then P(n) is true for all n e N, which establishes the alternate
form of induction.
Conversely, if P(n), for each n E N, is a proposition for which P(l)
is true and if, for any se N, the truth of P(s) implies that of P(s+ 1), we can
conclude, using the alternate form of induction, that P(n) is true for all
n E N. The hypotheses of the original form certainly imply those of the
alternate; i.e., if the truth of P(s) implies the truth of P(s+ I), then, afortiori,
the truth of P(k), k s, implies the truth of P(s+ I). Thus the validity of
the alternate form of induction implies that of the original. Consequently the
two forms are equivalent, and we shall not distinguish between them in
what follows.

Proposition 3. The Principle of Induction implies the Well-Ordering

Principle.

To prove this assertion we first use induction to show that I is the

smallest positive integer or, in other words, that there are no integers between
O and 1. The previous proof of this fact (Proposition I) utilized the Well-
Ordering Principle. Let
T = {t N: t l}.
Obviously I E T, and if n T, then n+ I, since it is greater than ii, is also
greater than I and an element of T. The Principle of Induction then implies
that T = N. Thus, n I for all n e N. which means that 1 is the least
element of N.
We prove next that an arbitrary nonempty subset S of N has a least
element. For each n E N, let P(n) be the proposition that every set of positive
integers containing an integer less than or equal to n has a least element.
Clearly P(l) is true since I is the least positive integer according to the
preliminary observation.
Now suppose n is an integer for which P(n) is true, and hence that
any subset Tc N containing an integer less than or equal to n possesses
a least element. Consider next a set N which contains an element less
than or equal to n + I. There are two possibilities to be considered. If Tk
contains no integer less than n+ I, then certainly n+ 1 is the least element
of If does contain an integer less than n + I, then it also must contain
an integer less than or equal to n. Since P(n) was assumed to be true, has
a least element. Thus, the truth of P(n) implies that of P(n+ I). The hypotheses
of the Principle of Induction are satisfied, so we conclude that P(n) is true
foralinEN.
§2.2 Analytic Properties of Integers and Induction 19

It is worth repeating that application of the Principle of Induction

to the proof of a given proposition involves verifying that the first and second
hypotheses are satisfied, and then using the conclusion of the Principle of
Induction to state that the given proposition is true for all positive integers n.
The first hypothesis is that P(n) is true for some initial value of n e N. It is
convenient to take this initial value to be I, although in some particular situ-
ations it may be greater than I. (Of course, when the induction starts at s,
the result is that the statement is true for all integers n s.) The second
hypothesis is essentially that truth (assumed) for previous value(s) implies
truth for the next integer value.
The exercises following this section develop several mathematical
formulas with the assistance of the Principle of Induction. In proving the
case for n = s+ I, we cannot simply substitute into a general formula, but
rather must usually express the statement for n = s+ I in terms of that for
n S.
As an example of the use of induction in proving formulas, consider
the Binomial Theorem which states that for arbitrary integers a, b and for
n>O,

where
= (n—i)!(i)! =
an integer [see Exercise 10], and n!=n(n—l)(n—2)•"3.2.l. Note that
= eN = 1. We verify that
fs+l\ / s \ fs\
and then show by induction on n that c = ('fle Z. Later it will be observed
that the formula for holds for elements a and b in an arbitrary com-
mutative ring. Then c•a"1b1 will be interpreted as the c4-fold sum of
For thoroughness of discussion we cite an additional fundamental
property of integers. The Archimedean Principle (attributed to Archimedes
of Syracuse, 287—2 12 B.c.), used primarily in analysis, says that for integers
a and b, satisfying 0 <a < b, there exists an integer t such that b at.

Exercises

1. Without reference to the original form of the Principle of Induction, use

the Well-Ordering Principle to prove the alternate form.
2. Prove that n+m n for all positive integers n,m N. To start, take any
m e N and prove
N.
20 ArIthmetic of Integers chapter 2

3. Prove the following formulas.

iz(n+l) n(n+I)(2n+l)
a. b.
1=1 2 6

= = n(n+l)(6n3±9n2+n—I)
c. d.

= n(2n— I)(2n+ 1)
e. =
n+l
1. E (21_l)2
4=1 3

" I
4. a. Prove that i2 < 2—— for ii 2.
i=I fl

I,
b. Prove that — 3 — — for n 1.
i=O1! ii

5. Prove that x2 > 0 if and only if x 0.

6. For integers a,b,, prove the following general distributive law:

a(bj + + = ab1 + ... + aba.

7. Given collections of sets A, and where I i n and 1 m,

prove the following.

a. A, (C) = 13 (A, B), for fixed i.

A1) (C) = (3 (3 (4,

1=1 i=Ij=1
8. Verify that
In+l\ / ii \ in
1=1 1+1
I
i / \i—I/ \i
where
\i/I =
I
i!(n—i)!

0! is defined to be 1, and ii! =fli.

9. Prove the Binomial Theorem, for integers a, b:

= i=o
E \i/

The Binomial Theorem for integral powers has several independent origins.
The Precious Mirror of the Four Elements (1303) by Chu Shih-chieh (fi. 1280—
1303) begins with a diagram of the binomial coefficients through n = 8. In
his Algebra the Persian poet and mathematician Omar Khayyam (C. 1050—
1121/3) claimed to have found a rule for writing the sixth and higher powers
of a binomial. The binomial theorem in array form for the coefficients appears
 The Division Algorithm 21

in the work of al-Kashi (died c. 1436). The array of binomial coefficients for
integral powers, illustrated in Figure 2.1 [cf. Exercise 8], is known in the
Western world as Pascal's triangle (after Rlaise Pascal, 1623—1662) although
it was printed on the title page of the 1527 algebra Rechnung by Peter Apian
(1495—I 552).

/\l
/l\3\/3\/
etc.
Figure 2.1

The corresponding theorem for nonintegral powers was discovered in

1664 or 1665 by Isaac Newton (1643—1727) and published in Algebra (1685)
by John Wallis (1616—1703) with credit to Newton.

10. Prove that for all

it, the binomial coefficients (7) in Exercise 9 are
integers. (The result of Exercise 8 is helpful.)
11. Prove that = 0 for rational numbers x1 if and only if Xj 0,
I I n.
12. For integers a,b prove that a(—b) = (—a)b. Give reasons for each
step in your proof.
13. Prove that the multiplicative identity 1 E Z is unique.
14. Prove that an = a for some nonzero a e Z implies that n = I.
IS. Prove the Archimedean Principle for integers.
16. Let f,g be continuous real-valued functions defined on the interval
(— 1, 1] c R, whose iith derivatives exist for all n e N. Find a formula
for the nth derivative of the product fy and (by induction) prove its
validity. (This result is due to Gottfried Wilhelm Leibniz, 1646—1716.)
17. Prove that a3 b3 a < b, for a,b e Z.
18. Prove the corollary to Proposition 1.

§2.3 The Division Algorithm

The following division algorithm for integers, often called the
Euclidean Algorithm, is fundamental in algebraic number theory and group
theory. Furthermore, generalizations of the division algorithm to rings other
than the integers are significant in commutative algebra. (See, for example,
§5.1 on polynomials.) An analogous algorithm is not valid in arbitrary rings.
22 Arithmetic of Integers chapter 2

Division Algorithm (for Integers). For given a. b e Z, b 0, there exist

unique integers q and r, such that
(*) a=bq+r. whereor<Ibl.
The proof will be in two parts: first the existence of q and r is shown,
and second their uniqueness is demonstrated.
Case I. Assume initially that a is nonnegative and that b is positive. Then
apply the Principle of lnduction to a for fixed b. Thus P(a) is the statement
that there exist unique q,rE Z such that equation (*) hoLds. We use the
following steps.
(I) For a = 0, take q = r = 0.
(ii) For a = I, if b = I, take q = I, r = 0;
if b> 1, take q=0, r= I.
(iii) The general inductive step consists in showing that
a— I = bq1 + r1, where 0 r1 <b,
implies a= bq + r, where 0 r < b.
Now, a = (a—l)+l =(bq1+r1)+l = bq1 +(r1+l).
If r1 + I = b, let q q1 + 1, r = 0. Otherwise, let q = r = r1 + 1. The
second hypothesis of the Principle of Induction is satisfied [i.e.. the truth of
P(a— I) implies the truth of P(a)], and we conclude that for all nonnegative a,
arbitrary b > 0, there exist integers q,r such that a bq+r.
Case 2. For the case a 0 and b <0, we have — b > 0 and a = (— b)q* + r,
0 r < —b = therefore, let q = _q* to obtain
a = + r = bq + r, 0 r <Ibj.
Case 3. Finally, for the case a <0, we have — a> 0, and by the preceding
argument
—a = bq' + r', where 0 r' <Ibi.
Hence a = b(—q') + (—,').
Then if r' = 0, a = b(—q')
= hq, where q = —q';
and if r' >0,
a b> 0
s=—l ifb<0
= bq + r, where q = —q' — r = bJ —r', and 0 r < hi.
To prove the uniqueness of q and r, assume a = bq+r = bq*+r*
where 0 r, r* <Ibi. Then b(q_q*) = r*_.r, whence =Jr*_rl.
Now observe that the assumptions on r* and r imply Ir*_rI < IbI. Also,
is a nonnegative integer. If 0 (and hence at least 1),
§2.4 Ideals in Z 23

but Therefore, and

An alternate proof of the existence portion of the Division Algorithm

follows from the Well-Ordering Principle. Assume b > 0. and let S be the
set of nonnegative integers defined by
S= {a—bt:IEZ, a—ht0}.
If aO, then aES for ,=0. If a<O, then
ha e S. Thus, in either case S is a nonempty set. Application of the Well-
Ordering Principle implies that S has a least element r 0, and r a—bq,
for some q e Z. Thus,
a = bq + r.
It remains to show that r < b = hi. Suppose to the contrary that r b;
then r = b+d, where 0 d < r, and
a — bq = b + d.
Then a—b(q+l)=d
is an element of S strictly less than contradicting the choice of r as the
least element in S. Hence, r must be strictly less than b, and the existence
proof is complete for positive b.
The result for b < 0 follows as in Case 2 of the previous proof.

§2.4 Ideals in Z
In and 2.3 of this chapter we considered properties of integers
or pairs of integers. We now begin a more thorough examination of the
algebraic structure of the integers by introducing the concept of an ideal.
In §3.1 we shalt consider ideals in more general rings, again as a tool for analysis
of algebraic structure.
An ideal in the ring of integers Z is a nonempty subset A satisfying
the following axiomatic properties:
(i)
(ii)

Example 1. The (trivial) subsets and Z are ideals in Z, and are often referred
to as the trivial ideals in Z.

Example 2. For each ni e Z, the set A = {kni : k e is an ideal in Z. More

specifically, form = 6, the set A = {..., — 12, —6,0,6,12. ...}of integral multiples of
6 is an ideal.
The reader is urged to check that A is indeed an ideal in Z. and to write
out elements of other ideals (i.e., for other values of in).
24 Arithmetic of Integers chapter 2

An ideal A is said to be a principal ideal if there exists some element

ill E A such that
A {kni : k E Z}.
In other words, there is a single element in, such that every element in A is
a multiple of in. Such an ideal is commonly denoted (m) and is referred to
as the ideal generated by rn. The generator of a principal ideal is not necessarily
unique. In Z, if,n generates the ideal A = (ni). then so does —in.

Theorem. Every ideal A in Z is principal.

Proof For a given ideal A, we have A = {0} = (0) or {0} c A Z. The
last case is the one to be examined further. For a nonzero element k of A
either k > 0 or —k = (— l)k >0. In either event, the ideal A contains a
positive element. Hence the set
P= {a A : a> 0)
is nonempty, and by the Well-Ordering Principle, there is a least element
in P. We claim that A = (m). Certainly (rn) = {km : k e Z) c A, since A
is an ideal. Now to prove conversely that A c (rn), consider an arbitrary
a E A. By the Division Algorithm,
a = rnq+r, where 0 r <in.
To show that r = 0, observe that mq A, and hence
r= a — inq A,
by the properties of an ideal. We conclude that r = 0, since r > 0 leads to a
contradiction, as m was chosen to be the least element in A which is positive.
Thus A c (m) and hence A = (nl), a principal ideal.

The ideal generated by gil'en integers a1, ..., ..., denoted

(a1, is defined to be

(a1, = : r1 z}

i.e., the set of all linear combinations of the elements a1 with integral co-
efficients. By the preceding theorem the ideal (a1, . . ., is principal, that is,

(a1, m an integral combination of the

elements a.:

k1a1, e Z.
=
Of particular interest will be the case of an ideal generated by two
integers a, b. Let d denote the positive generator of the ideal (a, b), assuming
§2.5 DivIsibility 25

that one of the integers a, b is not zero. Then

(d) = (a, b),
and there are integers s. t such that
d = sa + ib.
We leave as an exercise the proof of the following proposition.

Proposition. The set-theoretic intersection of two ideals (a), (h) in Z is an

ideal.

Exercises

I. a. Verify that (he set of integral multiples of 7 is an ideal in Z.

b. Verify that the set (ni) of multiples of ni e Z is an ideal.
2. Find a (single) generator of the ideal in Z generated by the following
integers.
a. 2 and 3 b. 8 and 24.
3. Find the positive generator of the smallest ideal in Z containing the
following ideals.
a. (4) and (18) b. (6) and (35).
4. If (a) and (b) are two ideals in Z, prove that their intersection (a) (b)
is again an ideal in Z.
5. Find the positive generator of the ideal in Z that is the intersection of the
following ideals.
a. (2) and (3) b. (4) and (18)
c. (8) and (78) d. (6) and (35).
6. Find q and r as in the Division Algorithm when:
a. a=7,b=12 b. a=47,b=4
c. a= 182,b=—3 d. a=—189,b=--17.
7. If a, b,c, d, x, y are integers, and ad—bc = 1, show that the ideals generated
by x, y and by ax + by, cx+dy are equal.

§2.5 Divisibility
In this section we consider the divisibility or factorization of integers.
Its most significant aspect is the existence of the greatest common divisor
of two integers. Proof of the existence and uniqueness follows from the
Division Algorithm of §2.3 and the theorem in §2.4 that all ideals in Z are
principal. Our discussion of factorization of integers culminates in the
statement of unique factorization in §2.7.
An integer h 0 is said to divide an integer a if there exists an integer
q such that a = bq. The notational shorthand is hi a, read "b divides a" or
26 Arithmetic of Integers chapter 2

"b is a divisor (factor) of a." Similarly bi'a means there is no XE Z such

that a = bx,'and is read "b does not divide a."
Note that any integer divides zero, but zero is not a divisor of any
integer. Also
bla a G (b), the ideal in Z generated by b.

Divisibility has the following properties.

I. Fora3&O,ala.
2. Forb,c nonzero, cib and bja cIa.
3. For a,b nonzero, alb and bla b = ±a.
The greatest common divisor (abbreviated GCD) d of integers a
is defined as follows:
1. lfa=h=O,thend=O.
2. If either a or b 9& 0, then
(i) d is a common divisor of a and b.
(ii) dis divisible by any other common divisor of a and b.
(iii) d>O.

Theorem. Any integers a,b have a unique GCD.

Proof In the singular case a = b = 0, we define d to be 0. Thus the questions
of existence and uniqueness are considered only when either a orb is different
from zero.
(i) Let A = (a,b) = {ax+by : x,y e Z) be the ideal consisting of all
linear integral combinations of a and b. Then, by the theorem of §2.4,
A = (m), where m = sa+ ib with 5,, E Z, and in> 0. This integer in is
a common divisor of a and b. Because a = a• I and b =
a b=
for some q1,q2
q2nz
(ii) If eja, elb, then a = es1 and b = es2 with s1,s2 E Z. Hence
m = es1s+es2 1 = e(s1s+s2 1), which shows that elm.
(iii) Finally in > 0 by the normalization of the generator of A. If $
were another GCD, then and mlö (by definition of the GCD), and so
= in for 5 > 0. Hence in is the unique GCD of a and b.

It is customary to denote the GCD of integers a and b by

d = (a,b).
Since in this expression (a, b) is interpreted as an integer, there should be no
confusion with the ideal generated by a and b, also denoted (a, h), introduced
in §2.4. In ideal-theoretic notation the ideals (d) and (a, b) are equal.
From part (I) of the proof, note the useful fact that the greatest
common divisor m = d = (a, b) of integers a, b can be expressed as a linear
§2.5 Divisibility 27

combination of a, b. That is, given a, b Z. there exist s, t e Z, such that

(a,b) = d = sa + lb.
The existence of such s and I will be used repeatedly in the subsequent
developments of the arithmetic of integers as well as that of polynomials of
one indeterminate [cf. §5.23.
If (a, b) = 1, then a and b are said to be relatively prime to each other.
Thus, for relatively prime integers a, b we can find integers s, t such that
sa+tb= I.

Proposition. For integers a, b, and m,

(a, m) = 1, (b, m) = I (ab, m) = 1.

The proof is left as an exercise.

The GCD d = (a1, of more than two integers, aj, not
all zero, is defined as follows:
(i) dia,,
(ii) If ela,, I I n, then eld.
(iii) d> 0.
The GCD d is the unique positive generator of the ideal A (a1,
Furthermore, d can be found inductively by setting
d1 = (a1,a2), d2 = (d1,a3), ..., d1 = (d1_1,a1+1),
d= =
(Verify this statement.)
So far we have proved that two or more integers must have a unique
GCD. An algorithm attributed to Euclid of Alexandria (C. 330-275 B.C.)
provides for explicit computation of these GCDS. The following procedure
with use of the Principle of Induction also provides an alternate proof of the
existence of the GCD.

The Euclidean Algorithm. For given integers a and b, b 0, write the

following using the Division Algorithm repeatedly:
a=q0b+r1. 0ri<IbI,
b = q1 r1 + r2, 0 r2 < r1,
r1 = q2r2 + r3, 0 r3 < r2,
4= 0 < Tk+l'

= i.e.. r,,_1 = 0.

Let n be the smallest positive integer such that the remainder upon
28 ArIthmetic of Integers chapter 2

division of by is zero. Since 0 rk+2 < < <2< r1 <IbI,

such an integer n must exist and can be found in a finite number (at most IbI)
of steps. Then r, = (a, b).
Certainly, starting with the last equation of the above series of
equations, we have j
- hence
where = +
Ultimately, on the one hand, and hence On the other hand,
dia and dlb imply that r1 = a—q0b is divisible by d. Since r2 = b—q1 r1,
the remainder r2 is divisible by d. Continuing in this fashion, we determine
that d I
Since > 0, necessarily d =
This algorithm can be used to determine integers s and i such that
d = (a, b) = sa + ib. From 2 = -' - + and = -2 _2+
we obtain
= —

= —

= +
Continuing this process, in the end we eliminate the remainders r1
in the expression for = d. Collecting terms yields d as a linear combination
of a and b.
Explicit examples will help clarify the preceding argument.

Example 1. For integers 38 and 122,

(38,122) = 2 = —16•38+ 5•122.
First, 122 = 338 + 8,
38 = 4.8+6,
8= 1•6+2,
6 = 3.2 + 0.

Consequently, by reversing the process, we find

2 = 8— 1.6
= 8— = — l•38 + 5•8
= —1.38 + 5•(122—3•38) = 5.122 — 16•38.

Example 2. For integers 38 and 119,

(38,l19)= I =
First, 119 = + 5,

38 = 7•5 + 3,
5=1.3+2,
3 = 1•2+ 1.
 Divisibility 29

Then, 1=3—2
= 3— (5—1•3) = 2•3 —5
= 2•(38—7•5)— 5 = 2.3g—
= 2•38 — 15.(119—3.38) = 47•38— 15•119.

We note for future use the following application of the Euclidean

Algorithm:
Let n be a fixed integer greater than I; then every positive integer a
has a unique representation
a= w0 + n + w2 n2 + + n3, where 0
called the n-adic expansion of a.
Many students will have encountered n-adic expansions in earlier
mathematics courses concerned with bases of number systems as, for
example, when considering numbers to the base 2, 7, 8, 12, 16, etc., rather
than to the commonly used base 10. Most common examples are in computer
applications of numbers to the base 2 (the binary number system), base 8
(octal), or base 16 (hexadecimal).

Example 3. The n-adic expansion of 38 is:

For,: = 2: 38 = 1.2 + 1.22 ÷ 12g.
Forn=8: 38=6+48.
Forn=16: 38.6+216.
Example 4. The n-adic expansion of 57 is:
Forn=3:
57 = 9 +

Example 5. The decimal or lO-adic expansion of 412 is

412 = 2+1.10+4.102.

The proofs of existence and uniqueness of the n-adic expansion ca

be accomplished simultaneously by induction. If a = 1, let w0 = and 1

= 0 for j> 0. Now, to verify the second hypothesis of the Principle of

Induction, assume that for all positive integers a < k such a unique rep-
resentation exists. To prove that k has a unique n-adic expansion, write,
using the Division Algorithm,
k = qn + r, 0 r <a, where q, r are unique.
Since k, a, and r are nonnegative integers, so is q, and in fact, q < k. If q = 0,
then k (= r < n) has a unique n-adic expansion, with = k and to, = 0 for
I> 0. If q> 0. then by the induction hypothesis,
q= a0 + a1n + a2n2 + + as_ins_I, o <n.
 Arithmetic of Integers chapter 2

Substituting this unique expression for q into k = qn + r, we have

a0n + + a2n3 + + +r
= + w2 n2 + w5n',
letting co0 = r and w. = as.. for 1> 0. Thus, existence of a unique n-adic
expansion for all a < k implies existence of a unique n-adic expansion for k.
Applying induction, we conclude, for all positive integers a, the existence of
a unique n-adic expansion.
Such expansions have become extremely important in the arith-
inetization of the theory of algebraic functions of one variable.

NOTE. The n-adic expansion of a negative integer need not terminate after a finite
number of steps. This remark will be developed, with an example, in §2.9 following
discussion of residue classes of integers.

Exercises

1. Find the greatest common divisor of each of the following pairs of

integers and express it as a linear combination of the two numbers with
integral coefficients.
a. —132,630 44,359
b.
c. 8273, 4565 d. 3472, 812.
2. For arbitrary integers a,b, c prove that the following greatest common
divisors are equal:
((a, c), b) = (a, b, c).
3. For integers a, b.c whose greatest common divisor is I, prove that
(a+b,b,c) = 1.
4. Adapt the proof of the uniqueness of q and r in the Division Algorithm
to prove that the coefficients w1, 0 < i s, in the n-adic expansion of a
are unique.
5. Find a (single) generator of the ideal in Z generated by the following
pairs of integers.
a. 12 and 56 b. 8 and 78 c. 34 and 672
d. —35 and 565 e. 48 and —1024.
6. Write out the n-adic expansion of each of the following integers for the
given value of n.
a. 7,n=2,3 b. 132,n=5
c. 14,n=2 d. l999,n=2,8,16
e. 1,000,000, n = 12 1. 153, n = 7.
7. Show that if d = (a, b), the integers s, I in the expression d = sa + lb are
not necessarily unique. Describe the different pairs s, t.
 Prime Numbers 31

S. a. Show that if (a,b) = I, then integers s, I such that 1 = sa+tb also

satisfy I = (s, b) = (s, I) = (a, t).
b. Show that if (a,b) d, then integers s,t such that d= sa+tb also
satisfy (5,1) Id. Need d = (s, t)?
c. For all s,tc Z show that (a,b,as+bI) = (a,b).
9. Show that if mlab and mlac, where (b,c) = 1, then rnla.
10. Prove that for integers a, b, and rn,
(a,rn) = 1 and (b,,n) = 1 (ab,m) = I.
11. In Q, for a fixed nonzero integer m, we can define rational numbers
r = a/b, r' = a'/b' to be equivalent if ab' — ba' is an integral multiple of
ni, for a,b,a',b' E Z; where b,b' 0 and (a,b) I, (a',b') = 1. Prove
that this is an equivalence relation [cf. Example 4, §1.2].
12. Prove for all n a N that if the GCD of integers a1, .. is 1, then there
exist h1 a Z such that

h1a, + h2a2 + I.

13. Prove for positive integers a, b, that if a b, then a b.

14. a. If (a,4) = 2 and (b,4) = 2, prove that (a+b,4) = 4.
b. For all naN, prove that 4r(n2+2).
a. For all n aN, prove that 61(n3 — n).
15. In 1202 Leonardo of Pisa (c. 1180—1250), better known as Fibonacci,
published a treatise, Liber abaci, on algebraic methods which contains
the following problem:
How many pairs of rabbits will be produced in a year, beginning
with a single pair, if in every month each pair bears a new pair which
becomes productive from the second month on?
Sequences of real numbers (or integers) whose terms are
defined recursively by = for n 3, where a1,a2 are
arbitrary, are called Fibonacci sequences. The above problem gives rise
to the particular Fibonacci sequence for which a1 = a2 = I. Prove that
the terms of this sequence satisfy
= 1 for all n eN.

§2.6 Prime Numbers

Integers a,b were defined in the preceding section to be relatively
prime if their greatest common divisor (a, b) = I. Now we consider properties
of prime or irreducible integers, leading up to the theorem on unique
factorization of integers which is of special significance not only for
the arithmetic of integers but also ultimately in the theories of algebraic
numbers and functions [see §9.8].
Any nonzero integer n has the trivial divisors ± I, ±n. A nonzero
integer p. different from ± 1, is said to be prime or a prime number if its only
divisors are ± 1, ±p, that is, if it has no nontrivial divisors.
32 Arithmetic of Integers chapter 2

Many, but not all, authors consider only positive prime numbers. However
it is more convenient for the subsequent development of irreducible elements in rings
and 9.8] to consider —2, —3, —5, etc., as prime numbers as well as 2, 3, 5.
Obviously p is prime if and only if —p is also prime.

Proposition 1. There are infinitely many positive prime numbers.

Proof. First of all, note that there exists at least one positive prime, namely
I+I= 2. Suppose to the contrary that the set P of all positive primes is
finite, say P = {p, Let it be 1 plus the product of al/n positive primes:
n—p1p2...pn+l.
The assertion is that it is a prime not in P, which would contradict the
assumption that the finite set P contains all positive primes. First, observe
that i = 1, ...,n. We now show that it has no proper divisors. Let D
be the set of positive proper divisors of it. If D # 0. by the Well-Ordering
Principle, D has a least element s. To show that s is prime, observe that
sin, als, a>I ajn.
Therefore a e D. Since s is the least positive divisor of it, a must equal s.
In other words, s has no proper divisors. But s I I n; therefore s
is a prime not in P. To avoid contradiction, we must have D = 0, which
means that it has no proper divisors, and hence is prime. Since it >
i= i n, it is a prime not in P. Thus P cannot then be the set of all positive
primes.

Proposition 2. For a prime number p.

plab Pla or pIb.

This is Theorem 2 in Book VII of Euclid's Elements. For the proof,
suppose pl'a. (If pla, nothing has to be proved.) Then (p,a)= I; thus
px+ay = I with x,y€ Z. Hence p(bx)+(ab)y = b, and using plab or
ab = pz for some integer z, we find b = p(hx+zy),
Integers, other than 0 and ± I, which are not prime are called
composite. Note that Proposition 2 characterizes prime numbers in the sense
that
mlab or mlb
if and only if m is prime. If in is prime, this is simply Proposition 2. A com-
posite m can be written in = qr for proper factors q, r, where I <lmi,
1 < Imi. Hence, miq, In subsequent applications in ring theory
it is convenient to use the above characterization of prime number, rather
than our original definition that m is prime if and only if it has no proper
factors.
§2.7 Unique Factorization 33

Proposition 3. If (c, a) = I and e ab, then c b.

Proof Proceed as in the proof for a prime p in Proposition 2, starting with
= cx+ay. Then
b = bcx + bay
= bex + ezy since dab
c(bx+zy),
which states that Cl b.

Proposition 4. If (a, e) = I, a mi, and elm, then ac m.

Proof We have tn = ad and elm; hence by Proposition 3, c d, or d = cc'.
Therefore m = acc* or acim.

§2.7 Unique Factorization

While the following statement is sometimes referred to as the
Fundamental Theorem of Arithmetic, we use the more descriptive name.

Unique Factorization Theorem. For each integer a, one of the following

statements is true:
1. a=O.
2. a=±l.
3. a = epixt •i'" where £ = + I, if a 2, and s = — I, if a —2,
and where the exponents ; are greater than zero and the p1 are distinct
positive primes. Furthermore, the integers p and ; in statement 3 are
uniquely determined by a.
Proof We prove the theorem in two parts: existence and uniqueness. For
the existence proof we use an indirect argument based on the Well-Ordering
Principle Suppose that the theorem is false, and let S be the set of
positive integers a> I that cannot be expressed as the product of primes.
If the theorem is false, then S must be a nonempty set, and so contain a
smallest integer m. This integer m cannot be a prime, as then it would trivially
be a product of primes. Therefore, m is composite with a factorization
m = be, where 1 <1,, c < rn. Since m is the smallest integer in S, neither b
nor c can belong to S: hence each is the product of primes, and therefore
m = be is the product of primes and does not belong to S. This logical con-
tradiction shows that S must be the empty set, and so the existence proof is
complete.
For the uniqueness portion of the proof, suppose that an integer a
can be written two ways as the product of primes:
a=elpl•ph,
34 Arithmetic of Integers chapter 2

where and c2 are ± 1. and q1 are not necessarily distinct

positive primes. Then =&2=+l if a>O, and =s2 =—l if a<O.
Next a implies since a = q1 (q2
J
Ifp1 q1. Proposition
2 of §2.6 implies that Pi (q2 ... Repeating this argument, if necessary,
it follows that Pi equals (at least) one of the primes q1, Relabelling
these primes so that q1 = Pi we obtain, after cancellation (by axiom llI(v*)
of §2.1), P2 Ph = q2 '1k. Now repeat this argument for P2'
Ultimately I is obtained on one side after successive cancellations, hence I
must also appear on the other side. This concludes the proof of the Unique
Factorization Theorem.

REMARK. In expressing an integer a as the product of primes, we commonly

collect equal prime factors and write
a =± I p,°•,

where ..., are nonnegative integers and the Pi, I I, .. ., .r, are distinct primes.

Unique factorization provides a convenient means to determine the

GCD of integers a, b. Let
(*) a= . .. b= ...
where; = 0 if p, 4 a, and similarly for Then

(a,b) = p1Yu where yj = min(a,,f31).

Jn addition to the GCD, we are interested in the least common multiple
(abbreviated LCM) of two nonzero integers. For nonzero integers a, b, the
LCM m is defined as follows:

(i) m is a multiple of both a and b.

(ii) ni divides any integer n that is a multiple of both a and b.
(iii) m > 0.
It is customary to denote the LCM of a and bby [a,b]. For each two nonzero
integers, there exists a unique positive least common multiple. The proof is
left as an exercise.
If a and b have the prime factorizations in equation (*) above, then
{a,bJ = Pi6' where = max(;,flj.
We have already observed that the ideal in Z generated by integers
a and h is the same as the ideal generated by their greatest common divisor
Now, we conclude with two propositions concerning the LCM. The
proofs are left as exercises,

Proposition I. Consider integers a,b. Then in Z, the ideals ([a,b]) and

(a) (b) are equal.
§2.7 Unique Factorization 35

Proposition 2. The (.CD and LCM for arbitrary nonzero integers a and b
satisfy
labi = (a,h)[a,b].

Exercises

1. Prove that [a, —h] [a,b].

2. Repeat Exercises I and 5, §2.5, by comparing the unique prime fac-
torizations of each of the integers involved.
3. Find the least common multiple of each of the following pairs of integers.
[Compare Exercise 1, §2.5 and Exercise 5, §2.4.]
a. — 132, 630 b. 8273, 4565
c. 44, 359 d. 3472, 812
e. 2.3 f. 4,18
g. 8, 78 h. 6, 35.
4. Verify for each of the pal rs of integers in Exercise 3, and prove in general,
that
[a,b](a,b) = IabL
for nonzero integers a,h.
5. The following statements can be proved using the prime factorization
of the integers a and b.
a. Prove that ([a, b]) = (a) (b). Thus the set-theoretic intersection
(a) (b) of ideals is sometimes relerred to as the least common
multiple of the ideals (a), (b).
b. Given any ideal J Z for which (a) J and (b) c 1, prove that I
also contains the ideal (a, b) generated by a and b.
C. Prove that (a) + (b) = ((a, b)), where the sum of ideals (a), (b) is
defined to be
(a)+(b) = {ra+sbEZ},
i.e., the set of sums of elements of (a) and of (b). The sum of ideals
is also called the greatest common divisor of the ideals.
6. a. For integers a,b, prove that (a) (b) 2 (ab).
b. Give examples of ideals (a), (b) for which the inclusion in part (a) is
an equality, and examp'es of ideals for which it is not.
7. Find the generator of the following ideals.
a. (8) (5) b. (8) (12)
c. (8) + (5) d. (8) + (12)
e. (36) (27) f. (36) + (27).
S. For a prime number p prove that the equation x" = p has no solution
of the form a/b, a,b E Z, for any,:> 1.
9. For nonzero integcrs a, b use the Well-Ordering Principle to prove the
existence of a unique positive least common multiple.
36 Arithmetic of Integers chapter 2

10. Prove that a positive integer of the form

a 0 and 0 a,b,c 9. must be divisible by 7, 11, and 13.

11. Suppose that the integers a and bare relatively prime. Let axo—byo =
with x0, e Z. Prove that ax — by = I if and only if x = xo + rb and
y=y0+Ia, fEZ.
12. Prove that a prime integer p divides the binomial coefficients ('),
0 <i <p [see Exercise 8, §2.2].
13. Prove that the integer 2 is a prime. Recall that there are no integers x or
y such that 0 < x < I <y < 2, by Proposition I and its corollary, §2.2.
14. For each prime number p, define the function on the integers Z as
follows:
v,,(O) =

v9(a) = 0, if pta,

= but pa*ira.
in other words, is the greatest integer such that pa a. Now given
an arbitrary prime p, prove the following for a, b E Z.
a.
b. = +
c. =
d. v,([a,b]) =
15. Extending the function in Exercise 14 to rational numbers, define
= a for a e Q if a = pa(e/d) where c, dare relatively prime integers,
such that pi'c, Prove the first two parts of Exercise 14 for every
pair of rational numbers a and b.
16. Take a e Q, as defined in Exercise 15, and define the p-norm of a:

= = 0.

Now show the following.

a. < + IbIs
b.
c. =
17. Take aL, as defined in Exercise 16 for a EQ. Prove that the classical
Cauchy condition for convergence of infinite sums na,, a, e Q, can
be replaced by = 0.
1$. a. Consider arbitrary integers u, v and a prime p. Prove that if
then plu orpi v.
b. Extend the result of part (a) by induction to prove that if
p ..., then u, for some 1, 1 I n.

19. a. Prove that a necessary and sufficient condition that the equation
ax + by = c, a,b,cc Z,

have integral solutions x, y is that the OCD d = (a, b) divide c.

§2.8 Congruences Modulo m 37

b. Show further that if there exists a solution x0, then there will be
infinitely many solutions of the form
x=-x0+(b/d)t, y—y0—(a/d)t, teZ.
c. Construct explicit numerical examples of equations of the type cited
in part (a). Give multiple solutions, as in part (b) for those which
have solutions.
d. Suppose that a1 are integers. Prove that the Diophantine
equation (after the Greek mathematician Diophantos, c. 250 A.D.)

has integral solutions if and only it' the GCD d of

divides b.
20. Let n = flh p," be the factorization of the positive integer n with
distinct positive primes Denote by ö(n) the number of distinct positive
divisors of n. Prove

ô(n) =

First, prove 5(inn) = if(m,n) = I for positive

21. a. Use the Well-Ordering Principle to prove that every integer n> 1
is divisible by some positive prime p. (Note that this problem implies
the existence of at least one positive prime number if we admit the
existence of integers greater than I.)
b. Using part (a), prove the existence of infinitely many positive primes.
(This is the type of proof given by Euclid in his Elements, Book IX,
Theorem 20.)
22. For integers a, b, c, d prove the following statements involving the oco
and LCM.
a. [(a,b),c} = ([a,c],[b,c])

b. (ab,cd) = (a,c)(b,d)i—,
Ia d\Ic
ii—, — b
\(a,c) (b,d)/ \(a,c) (b,d)
23. For a prime number p and integers a,b answer the following questions.
a. If(a,p2) = p and (b,p3) =p2, then what is (pa+b,p5)?
b. !f(a,b)=p, what is(a2,b5)?

§2.8 Congruences Modulo m

In this section we turn to the concept of congruence, which is all
pervading in mathematics, although it is presented now only for integers.
The underlying idea is to replace Strict equality by a "loosened-up" concept.
Rather than discussing the equality of integers a. b, we consider them now
to be equivalent (or congruent) if they differ, not by zero, but by a multiple
of a fixed integer rn. Mathematicians, perhaps even Pierre de Fermat
(1601—1665), made mistakes in algebraic number theory because of a lack of
understanding of congruence prior to the systematic study of congruences by
Carl Friedrich Gauss (1777—1855).
 Arithmetic of Integers chapter 2

Application of the congruence concept permits formulation of

meaningful results on the structure and explicit description of algebraic
entities as, for example, in algebraic number theory [see §3.4], in the solution
of polynomial equations [*5.3]. and in group theory [*6.4].
Let in be a fixed integer, in 0. ± 1. An integer b is said to be
congruent to an integer a modulo in; in symbols
b a modulo in or b a (mod in),
if rn (b — a), that is, h — a = inq or b a + inq for some q E Z.
With reference to the properties of divisibility [*2.5], we can easily
verify the following properties of congruence modulo m.

Property 1. Congruence modufo in is an equivalence relation [see §1.2.

Example 4]. We have:
Reflexivity: a a (mod in).
Symmetry: a b (modm) b a (mod,n).
Transitivity: a b (modni) and b c (modm) a c (modni).

Property 2. a b (mod,n) implies a+c b+c (mod in) for all c e Z.

Property 3. some CE Z, then

a b (mod (mod in), for all c Z.

But note that ac be (mod m) for some c e Z does not necessarily imply a b
(mod in). That is, a multiplicative cancellation law for integers modulo in does not
hold (see *2.1, axiom HI(v*)J.
For example, 8•9 6.9 (mod 18), but 8 6 (mod 18) because 8—6 is not
a multiple of 18.

Property 5. However, ac be (mod in), where (c, in) = 1, implies a b

(mod in).
Proof. Note that c: + my = I. Consequently cz 1 (mod m), and
a a(c:) = (ac): (bc)z = b(c:) 1, (modm).
An integer a e Z is called a prime residue modulo m if (a, in) = I.
In particular, if p is a prime number, then 1,2, ...,p— I, and more generally
i+kp, I i<p, for any k e Z are prime residues modulop.
The coset (also called residue or congruence class) [a] modulo in
of a e Z is the subset of Z given by
[a] = (a' Z : a' a (modm)}.
We conclude this section with some comments on these cosets. In
the next section we define an arithmetic for them, that is, rules for addition
and multiplication of cosets.
§2.8 Congruences Modulo m 39

Since a a (mod in), the coset [a] is nonempty for all a e Z. Note
that [a] contains a unique r e Z such that 0 r Write, by the Division
Algorithm
a = 'nq+r. 0r
and observer a (mod,,,) [cf. Exercise 2, §1.2].
Thus a' e [a] means that a and a' have the same remainder r upon
division by in; each a' [a] can be written as a' = sin+r for some SE Z.
To say that [a] [bJ means that the residue classes [a] and [b],
considered as subsets of Z, have no element in common. Since congruence
modulo in is an equivalence relation, the corresponding equivalence classes
(i.e., the congruence classes modulo in) form a partition of Z. By the
proposition we then have either [a] [b] or [a] [b] = 0.
We denote the collection of distinct residue classes in Z modulo m
by either Z/(m) or Zm. and refer to an element a' c [a] as a representath'e
of the coset [a]. Note that there are mi "elements" in Zm.

Exercises

1. Write out explicitly the congruence classes in Z modulo in for

a. m=2 b. m=3
c. m=8 d. m=l0.
2. Describe the elements of Zm for
a. m=4 b. m=5.
3. Find the least positive representative of each of the following congruence
classes.
a. 5 modulo 11 b. 52 modulo 11
c. 58 modulo 11 d. 628 modulo 8
e. 17 modulo 5 f. 641 modulo 17
g. modulo 23 h. 145 modulo 15.
4. In the text Zm was defined for m 0, ± 1. Now extend the definition of
residue classes of integers modulo in to include the cases in = 0, ± 1.
a. Describe the residue classes of Z/(l) = Z/(— I). How many are
there?
b. What are the equivalence classes of Z for congruence modulo 0?
S. Prove that the congruence ab ac (mod an:) for a 0 implies b c
(Compare Property 5.)
6. ha b (mod in) and (a,m) = I, prove that (b,m) 1.
7. Prove fora prime p that (a,p)>1
8. Let p be an odd prime. Prove that
a b(modpk), k >1,
implies (modp" 1)•
 Arithmetic of Integers chapter 2

9. a. Let p be an odd prime and suppose that

cEl+gp(modp2) withgaZ.
Prove that
1+ (modp"41) for all k 1.

b. Prove that c I +22g (mod implies

forp2.
10. Prove by induction on a that for integers x and y,
(x + y)P' xI' + (mod p)
+ (modp).
11. Suppose that p is an odd prime. Prove that the congruences x2 a
ti> I, where (a,p) = I, have integral solutions x if and only
if the congruence x2 a (modp) has an integral solution. Hint: The
proof is by induction on n. For the inductive step try to find x,,+ i of
the form = where a
12. Generalizing Exercise II, prove that for an odd prime p,
xm a ii > I,
is solvable if and only if xm a (modp) is solvable, where (,,i, p) =
(a,p)= 1.
13. Prove that there are infinitely many positive prime numbers of the form
4n+ 3. Hint: Suppose there are only finitely many-—say P = 3,
—and consider m = +3.
.

§2.9 Addition and Multiplication of Cosets Modulo m

We now define rules of addition and multiplication for cosets
[a], [b] modulo rn, or, briefly, for elements of Zm for fixed integer distinct
from Oand ±1.
Define a suni [a] + [b] in Zm by taking a' e [a] and b' [bJ as
arbitrary representatives and setting [a] + [bl = [a' + b'J; i.e., the sum coset
is the coset modulo m of the sum a' + b' (in Z) of the representatives a', b'.
Similarly, define a product [a]. [b], usually denoted simply [a] [b],
ifl Zm to be the coset [a'b'] modulo m of the product (in Z) of the representatives
a', b'.
In order that these definitions give rise to cosets Ua]+[b], [a].[b]
uniquely determined by the given cosets [a] and [h], it is necessary to show that
a different choice of the representatives a" of [a] and b" of [b] will lead to the
same sum and product cosets. That is, we must prove that
[a'-i-h'] [a"+h"] and [a'h'3 = [a"b"].
For a,a',a" [a] and b,b',b" [b] we can write
a' a" a (modm) and b' b" b (modm),
§2.9 AdditIon and Multiplication of Cosets Modulo m 41

according to the definition of cosets. Now adding congruences we have

a' + b' a" + b" a + b (modm),
and multiplying, a'b' a"b" ab (modm).
Thus [a'+b'] = [a"+b"] and [a'b'] = [a"b"], as asserted.
We observe that the elements [a] of Zm satisfy the same arithmetic
axioms of addition, multiplication (with the important exception of the
cancellation law), and distributivity as given for elements a of Z in §2.1.
These axioms are:
1. The sum [a]+[b] = [a+b] of cosets satisfies the following properties.
(I) Associativity: [a] + ([b] + [c]) = ([a] + [b])+ [c].
(ii) Commutativity: [a] + [b] [b] + [a].
(lii) Identity: There exists a zero coset or additive identity [0] =
{krn : k E Z} such that [0]-i-[aJ = [a] for alt [a] e Zm.
(iv) Inverse: For each [a] Zm, there exists an element [a'] e Zm
such that [a]+[a'] = [0].
II. The product [a3[b] = [ab] of cosets satisfies the following properties.
(i) Associativity: [a]([b][c])= ([a][b])[c].
(ii) Commutativity: [a][b] = [b][a].
(iii) Identity: There exists an element (the class of 1 modulo in) which
plays the role of multiplicative identity. That is, [l][a] =
[a][I] = [a].
III. The sum and product of cosets satisfy the distributivity property:
[a]([b]+[c]) = [a] [b]+[a] [c].
We call Zm the residue class ring of integers modulo m. Such rings
were first constructed by Adrien-Marie Legendre (1752—1833). Note care-
fully that in general the cancellation law does not hold in Zm. For example,
let m = 18, then
[8][9] = [6][9J but [8] [6].
Proofs of the preceding statements are obtained by appealing to the
corresponding facts valid in Z. For example, consider the distributive law.
Let a' E [a], b' E fb] and c' e [c] be any representatives of the given cosets.
Then, in Z, a'(b'+c') = a'b'+a'c'. Consequently [a'(b'+c')] = [a'b'+a'c']
for the corresponding cosets. Hence, using the definitions of sum and
product in Z,,, we have
[a'(b' + c')] = [a] [h' + c'] = [a] ([b'] + [c']) = [a] ([b] + [c]),
and similarly,
[a'b'+a'e'] = [a'b'] + [a'e'] = [a][b] + [a][c].
To illustrate the importance of the choice of coset representatives, we
consider, as at the end of §2.5, the n-adic expansion of any a e N for a fixed n N.
Such an expansion need not be finite if a fixed system of representatives for the n
 Arithmetic of Integers chapter 2

residue classes modulo n is used. For example let ii = 5 and take 0, 1, 2, 3, and 4 to
be she representatives of the cosets of Z5. Then — 1 4 (mod 5), where — I =
and Consequently,
—l
4.50 + +

4.50+4.51+4.52+_l.53, etc.

Thus — I = an infinite sum which does not make sense for the ordinary
constructions of real and complex analysis with the customary absolute value I

as in §2.2. However, if we define al5 to be where a = aaQ and

(b, c) = I for b, c e Z, then the infinite sum turns out to be convergent. (This dis-
cussion of I is a special case of the p-norm defined in Exercise 16, §2.7. See also
Exercise 17, §2.7.)

Exercises

I. For ii = a0 + lOa, + + prove that

k

a, 0(mod3).
i=O

The following problem furnishes the basis for the bookkeeper's rule of
casling out nines, which is used for checking addition and multiplication.
2. a. Let s be the sum of the decimal digits of an integer a. Show that
a 0 (mod 9) if s 0 (mod 9).
b. Prove that the remainder of a sum upon division by 9 is equal to the
sum (reduced modulo 9) of the remainders of the addends upon
division by 9.
c. Prove that the remainder of a product upon division by 9 is equal to
the product (reduced modulo 9) of the remainders upon division
by 9.
d. Prove that if an integer in is obtained from ii by permuting the digits
of n, then in ii (mod 9).
"It is often said that. . . the out of nines' is a Hindu invention, but
it appears that the Greeks knew earlier of this property, without using it
extensively, and that the method came into common use only with the Arabs
of the eleventh century."t
3. Find all powers of the following.
a. [6],3inZ13 b. 12]1oinZ1o
c. in d. [6127 Ifl Z27.

4. For any a e Z, prove that precisely one of the congruences

holds modulo 8.

From Carl Boyer, History of Mathematics, p. 241, © 1968 by John Wiley & Sons, Inc.
 Definitions and Examples 43

5. Prove that no integer a satisfies the congruence a2 17 (mod 100).

6. Prove that the congruence ax b (mod m) has a solution if (a, m) 1.

7. Prove the associative and commutative laws of addition and multi-

plication for the arithmetic of cosets. Note that the proofs are essentially
carried out in Z and transfered to Zm.
8. Match each nonzero element in Z1 with its multiplicative inverse.
9. Repeat Exercise 8 for Z7 and Z17.
10. What elements in Z6 and Z12 have multiplicative inverses?

§2.10 Definitions and Examples

In this section we consider the residue class rings Zm for several
values of m, explicitly giving the addition and multiplication tables for
ni = 2, 3, 4, and 6. We also introduce ring-theoretic terminology for elements
in Zm with particular properties: prime residue or unit, zero divisor, nilpotent,
and idempotent. These special elements and their properties are examined in
specific instances in the rings Zm so as to provide an elementary setting (where
all computations can be executed easily) for the introduction of concepts
necessary for the later study of rings, groups, and algebraic number theory,
algebraic geometry, etc. These definitions, which carry over to general ring
theory and polynomial rings [Chapter 5], provide an opportunity to
gain experience in computing with elements of the residue class rings Zm.
The final topic of this section is the solvability theory for congru-
ences (modulo in) of integers, which is equivalent to solving linear equations
in Zm. Section 2.11 treats the general theory for solving simultaneous
congruences.
In the following examples note that Zm always satisfies the algebraic
axioms 11(i) through (v) and III(i') through (iv') but that the can-
cellation law (y*) is valid only in specific instances.
We call [a] a prime residue class modulo in if some (and hence any)
representative a' is a prime residue modulo m. that is, if (a', m) = 1. If m is
prime, [a] is a prime residue class for all a, I a < m. Also, for example,
[10] is a prime residue mod 21.
A nontrivial divisor of zero in 4, is a residue class [a] # [0] for
which there exists a nonzero residue class [b] E Zm satisfying [a] [h] = [0].
For example, in Z18, if [a] = [3], then [b] = [12] and [b] = [6] satisfy
[aJ[b] = [0].
A nilpotent element in Zm is a residue class [a] satisfying
[a]h = [0]. in other words. mid',
for some positive integer h. (The smallest positive h having this property is
called the index or exponent of nilpotency of [a].) Thus, the coset [6] [0]
is a nilpotent element (of index 2) in Z18 because [612 [36] = [0]. The
zero class [Oils trivially nhlpotent.
44 ArithmetIc of Integers chapter 2

Note that divisors of zero are not necessarily nilpotent elements.

For example. [4] e isadivisorofzero—-[4][27J = [4][9] = [36] = [0]—
but it is not a nilpotent element because no power [4]" equals [01, that is,
4?!
22?! is not divisible by 18 for any positive integer h. Similarly. [3] is a
divisor of zero, but is not nilpotent.
We now illustrate the preceding concepts in examples of addition
and multiplication tables for Zm. For the sake of clarity in the tables all residue
classes [a] modulo in are denoted by overlining, viz., a.

Example I. For rn = 2.

Addition Multiplication
0 1 5 1

0 1 0 1

Note the absence of

nontrivial zero divisors.

Example 2. For in = 3.

Addition Multiplication
0 1 2 '0 1 2

5,0 2 0 0
is
1

r 1 2 o r 2
2 2 0 1 2 1

Note the absence of

nontrivial zero divisors.

Example 3. For nz = 4.

Addition (vi ultiplicat ion

0123 123
1 2 3 ojo O S

T 2 3

0 2 2

3 1

Note that the nontrivial zero

divisor is nilpotent.
*2.10 Definitions and Examples 45

Example 4. For m = 6.

Addition Multiplication
1 2 3 4 3 0 1 2 3 4 3

01 2 3 4 5

1012343
2024024
2i234501
3345012 3030303
4430123
5501234 4042042
5054321
Note that no nontrivial zero
divisor is nilpotent.

The following three propositions summarize and generalize the

observations made in these examples.

Proposition 1. ln 4,. a coset [a] is a nontrivial zero divisor if and only if

(a,m)> 1.

Proof: First, if [a] is a zero divisor, then there exists a coset [bJ [0],
such that m ab. If (a, m) = 1, then by Proposition 3. §2.6, rn b, and [h] [0],
a contradiction. Hence (a, m) I. Now conversely, if (a, tn) = d> I we may
write
m=qd, 0<q<,n, and a=ds, 0<s<a.
Then, ms = qdc = qa, and so [q][a] = [0]. Since 0< q < rn, and so
[qJ # 0. Therefore [a] is a zero divisor.

Proposition 2. If m with distinct primes Ii 2, then Zm has

no nonzero nilpotent elements.
ProoJ Any nilpotent element is a zero divisor. For any nontrivial zero
divisor [a] in Zm, one of the prime factors Pt Ph. say must not divide
a. Hence for any se N, which means that [a] cannot be nilpotent,
as asserted.

Proposition 3. If rn = p5. 2. then every divisor of zero in 4 is nilpotent.

Proof. Clearly, if [a] is a zero divisor, it is necessary that pa a, for some /1,
I fi < because there exists a coset [b) [0] such that p2 I ab. Pick s so
that sfl and observe that p2 a5 or equivalently, = [0].
An idempotent (element) of Zm is a coset [a] for which
[a]2 = [a2] = [a];
46 Arithmetic of Integers chapter 2

or, in other words, a solution of the equation —x = [0]. For example, if

36, then [28]2 = [28]. In Z18, [10] is an idempotent element which is
also a divisor of zero: [10][9] = [0]. We refer to [0],[l] as trivia! ide,;z-
patents Zm for any
A unit of Zm is a coset [a] for which there exists a (necessarily unique)
coset [b] such that [a][b] = [I]. We then speak of {b] as the multiplicative
inverse of [a]. While we write [b] = [a] ',we have no illusions that the coset
[a] -' e contains proper fractions. As we show next, the units [a] of Zm
are precisely the prime residue classes modulo m.
To wit, if (a,m)= 1, then as+mI = I for some s, 1€ Z as in §2.5.
Hence as I = [I]. Conversely, if [a] is a
(mod rn), or equivalently [a][s]
unit, there exists [b) such that [a][b]=[l]. Then ab+mq= I for some
q E Z hence (a, m) = I.
This argument generalizes as follows to the solution of equations
[a][x] = [b] in Zm.

Proposition 4. The congruence ax b (mod m) is solvable if and only if

d= (a,m) divides b. The solutions, if any, differ by multiples of q = mid.
Proof First, we prove the existence of a solution assuming that dib.
Write d= sa+tni and b = qd. Then

b = qd = qsa + qim

and x = qs satisfies the congruence.

Second, assume that the congruence is solvable. Then b = ax+mw
for some w Z. Hence dib, because dja and dI.'n.
Finally, suppose that x" is another solution of the congruence
ax b (mod m). Then a(x— x*) 0 (mod m). Consequently, writing a = dv
and m=dq since we have (moddq), and
v(x_x*) 0 (modq). Therefore (modq) as asserted, because
(v. q) = 1. Furthermore, if is a solution of ax b (mod m), then so is any
x' x* (mod q).
Proposition 4 provides a useful tool for the study of cyclic groups
in §6.6.

Exercises

1. Find all zero divisors in the residue class ring Zm for

a. m=20 b. ,,i=16
C. m=7 d. m=24.
2. Find all nilpotents in each of the rings in Exercise 1.
§2.10 Definitions and Examples 47

3. Give necessary and sufficient conditions on in for each of the following

statements.
a.No proper zero divisor of Zm is nilpotent.
b.All proper zero divisors of Zm are nilpotent.
4. Find all x in the residue class ring Zm such that x2 = [1], the multiplica-
tive identity in Zm, for
a. in = 3 b. in 8 c. in 6
d. in = 4 e. m 10 f. in = 12.

5. For each of the following congruences describe all integral solutions.

a. 13x 9 (mod 19) b. 4x — I (mod 21)
c. 2x 17 (mod 27) d. 3x 36 (mod.54)
e. 3x 54 (mod36) f. 12x —I (mod35).
6. Show that any elements [a], [b] a which satisfy
[a] + [b] = [1] and [a][b] = [0]
are idempotent. Further, show that for each nontrivial idempotent
[a] Zm there exists an idempotent [b] such that the above identities
are satisfied.
7. Show that the congruences
a+b I (modp2) and ab 0 (modp2)
have only the solutions a 1, b 0. or a 0, b I. Conclude that
Z,,. has no trivial idempotents for a > I. (The discussion at the end of
§5.2 indicates that has no nontrivial idempotents.)
8. With reference to Exercises 6 and 7, write out explicitly all idempotents
in the residue class ring Zm for
a. in = 20 b. in = 70 in = 75
c.
d. m = 16 e. in 24 in = 30.
f.
9. Prove that [a] a Z,,, is a nontrivial idempotent if and only if there exist
in1,ns2 such that
(i) in = in1 in2, (in1, I,
(ii) a = sin2, where sin2 I (modin1).
10. Prove that [a] Z,, is either a unit or a zero divisor.
It. Prove that [a] a Z,,, is nilpotent if and only if every prime divisor of in
divides a.
12. a. Prove that has nontrivial nilpotents if and only if in for
some n> I.
b. Describe the set N of all nilpotents in the residue class ring Zm.
c. Prove that N is a principal ideal and describe a generator.
13. Prove that if [a] is idempotent, then [I —a] is also idempotent.
14. In the residue class ring Z,,. show that
a. (a] is a Unit if(a, pa) 1.
b. [a] is nilpotent if (a, p°) I.
15. a. Ifx2+x+ I 0(modm), prove that {x] is not a zero divisor in Zn,.
b. lf[x] is nilpotent in Zn,, prove that [1 +x+x2] is a unit in Zm.
c. Prove that x2+x+l has no integral solution.
d. Find all integral solutions of x2+x+ I 0
48 Arithmetic of Integers chapter 2

§2.11 Simultaneous Systems of Congruences

In §2.9 we listed the additive and multiplicative properties of the
congruence classes [a] E Then, Proposition 4 in the preceding §2.10 gave
necessary and sufficient conditions for the solution of the equation
[a][x] [b]
Zm, or equivalently, of the congruence
ax b (modm).
In this section we take up necessary conditions and techniques for solving
systems of congruence equations simultaneously and prove the so-called
Chinese Remainder Theorem. This theorem on the solution of compatible
systems of congruences has extremely important generalizations in the theories
of algebraic numbers and of algebraic functions of one and more variables.
With the Fundamental Theorem of Arithmetic (unique factorization of inte-
gers) it is one of the foundations of class field theory, a currently active area
of mathematical research that involves abelian extensions (discussed further
in Chapter 9). We shall make use of the theorem in and 5.3.
The general method for solving simultaneous congruences is best
explained after examining two examples.

Example I. Consider two congruences,

y 18 (mod 7) and y 3 (mod 12).
Any solution y of the first congruence must be of the form
y= 18 + 7k.
Substituting this general solution of the first congruence into the second, we obtain
(*)

or 7k 15 9 (mod 12).
Since (7, 12) = 1, we can write I 7s+ 121. In fact,
= _5.7 + 3.12.
Now multiplying by 9, we have
—45-7 = +9—27.12
from which we see that k = —45 3 (mod 12) satisfies congruence (*). Substituting
k =3 into y = 18+7k, we obtain y = 39 as a particular solution of the double
congruence. The general solution is
y = 18+7(3+12q) =
Example 2. Consider the system of congruences
2x —3 (modl),
4x 1 (mod 9),
x 5 (mod 13).
§2.11 Simultaneous Systems of Congruences 49

Following the procedure of Example I, we solve for x satisfying the first

congruence, obtaining
x 2 + 7r.
Putting this value for x into the second congruence, we have
4(2+7r) I (mod9),
8 + 28, 1 (mod 9),
r —7 (mod9),
r 2 (mod 9).
The general solution of congruence (**) is then
r= 2 + 9s
and of the first and second congruences of the original system is
x= 2 + 7(2+9s) = 16 + 7•9s.
We repeat the process, rewriting the third congruence as
16 + 63s 5 (mod 13)
or (***) I Is 2 (mod 13).
Writing I = (11,13) as I = 6.11—5.13, and multiplying by 2, we obtain
12.11 2 (mod 13).
The general solution of congruence (***) is then
s= 12 + 131,
and of the set of three congruences is
x 16+7•9(12+131) = 772-f
Note that in specific cases some congruences, such as 2x —3
(mod 7), can be solved at sight, x = 2, without resort to the formal techniques
of solution, which involve expressing I = (2,7) as a combination of 2 and 7.
With these two examples in mind we consider the Chinese Remainder
Theorem proper, an existence and uniqueness statement for the solution of a
general system of congruences. The name is derived from the fact that the
Chinese mathematician Sun-Tsu found solutions of x 2 (mod 3), x 3
(mod 5), and x 2 (mod 7) in the first century AD. and subsequent Chinese
mathematicians have considered similar problems.

The Chinese Remainder Theorem. Let in1 ni,, be n pairwise relatively

prime integers, each greater than I (that is, = 1, if I i,j n)
and let ...,a,, be n arbitrary integers. Then the system of n congruences
x a1 (mod,ii1)

x aj (mod in,)

x a,, (mod in,,)

 Arithmetic of Integers chapter 2

has an integral solution x Z that is uniquely determined modulo in =

In other words. if x (modm).
The proof is by induction on the number of congruences n in the
system.
The Case n = 2. Consider two congruences
y a1 (modm1) and y a2 (modm2)
for arbitrary but relatively prime moduli m1 and in2. There exist integers s
and t such that
(*) m1s+m21=l.
A solution y of the first congruence must have the form y = a1 + km1. Hence
we must have
a1 + km1 a2 (mod,n2)
or kin1 a2 — ai (modm2).
Multiplying this congruence by s,
km1s (mod,n2),
and making use of equation (*) to write in1 s = I —m2 t, we find that
k s(a2—a1) (modm2),
The general solution of this congruence is s(a2—a1)+qm2, and hence
y = a, + [s(a2 — a1) + qm2] in1 is the general solution of the pair of congruences.
The General Case. To verify the second hypothesis of the Principle of
Induction, §2.2, assume that z is a solution of the system of n—I congruences
z a1 (modm,)

z (mod
and consider the pair of congruences
x z (mod,,,1 m2 in,,... and x a,, (modm,,).
Since (m, rn,,.. ,,nz,,) = I, this pair of congruences has a solution x
by the proof for the case n = 2. This element x Z satisfies the original
system of congruences, because x z (modm1 m,,_ implies that x—z
is divisible by or (modm1), I j<n.
Hence by the assumption
on z,
x (modm1), 1j < n.
In addition x satisfies xa,, (modm,,), so that x is a solution of the entire
system. By the Principle of Induction such systems of n congruences are
solvable for all n.
§2.11 Simultaneous Systems of Congtuences 51

Finally, suppose that x" is another solution of the given system,

namely.
x a, I j ?1•
Then x— 0 (mod ni'), or in, (x — x*). Thus
,nI(x_x*)
by Proposition 4, §2.6, so that x (mod in).

The Chinese Remainder Theorem also yields the existence of a

solution of the system of congruences
a.x=b,(modm1),
I and, for (rn1,in,) = 1, since this system can be
transformed into the system in the theorem by multiplying the ith congruence
by s,, where as1 I (modm1). The existence of such s1 is assured by the
hypothesis that (a1, = I.

Exercises

1. Find all x e Z which satisfy the following simultaneous congruences.

a. 2x 8 (mod22) b. 3x 9 (mod II)
x 4 (mod5) 2x 8 (mod23)
x —I (mod9)
c. x 2 (mod 5) d. x I (mod 4)
x 3 (mod?) 3x I (mod5)
x 5 (mod 11) x 9 (mod 17).
2. In parts (a) and (b) find an integer x that satisfies the simultaneous
congruences.
a.
x
x (mod
x
x
2 (mod 3)
x (mod
x (mod
x 7 upon 1

division by 2, 3, 4, 5, and 6. Solved by lbn al-Haitam, c. 1000.

d. Find an integer x that is a multiple of 7 and has remainders 1, 2, 3, 4,
and 5 upon division by 2, 3, 4, 5, and 6, respectively. Solved by
Fibonacci in 1202.
3. a. Find all x, y c Z which satisfy the pair of congruences
3x + 4y 2 (mod I 1), 5x + 2)' 3 (mod II).
b. Repeat part (a) replacing ii by 9.
52 Arithmetic of Integers chapter 2

4. Is the pair of congruences

2(modl), 5x+2y_ 3(mod7)
solvable? Justify your answer.
5. Find conditions on the integers u and V so that the congruences
x + 2y + 3z u (modp), 4x + 5y + 6z v (modp)
have solutions for all primes p.
6. Prove that the system of congruences
x a, (mod n;1), I I n,
is solvable if and only if for all I i,j n.

§2.12 Two Topics in Number Theory

We conclude this chapter on integers by giving Euler's formula for
the number of units in Zm and Fermat's Little Theorem: for a prime p any
integer is congruent, modulo p. to its pth power.
For positive the number ,p(in), called the Euler v-function of rn
after Switzerland's great mathematician Leonhard Euler (1707—1783), is
defined to be the number of integers between 1 and m which are relatively
prime torn. In §3.5 we prove the following formula for p(rn).

Euler q-function. If ni has the prime factorization

= p12t .
where the p, are distinct positive primes, then

co(m)=nI(l_..i_)...(l (pa_I).

In particular, for a prime p.

q,(p) = p—I and q(p3) = p2_p3.••I•

To verify the latter, observe that every pth integer between I and p is divisible
by p, a total of p8/p =pX integers. Thus are relatively prime to
as asserted.

We noted in §2.10 that [a] e Z,, is a unit if and only if [a] is a prime
residue, or equivalently, if and only if (a,rn)= 1. Euler's then
expresses the number of units in Zm as a function of m.
We make several further observations about the set of all units
Of It is closed with respect to multiplication of cosets; that is, if
[a],[b]eUm, then [a]{b]EUm. Furthermore [aJ[b]=[b]{a], and every
[a] e has a multiplicative inverse. In other words, Urn is an example of a
finite corn?nutative group with q(m) elements, a concept discussed in
Chapter 6.
§2.12 Two Topics In Number Theory 53

The ring Z,, has q(p) =p— I units and has in common with the real
numbers R. the complex numbers C, and the rational numbers Q the property
that every nonzero element has a multiplicative inverse. Rings with such
properties are called fields [see §3.6 for further details].
En addition to ([a] [b]V = [a]"[b]", we prove for [a], [b] E Z,, that
= [a)" + [b]".
From the Binomial Theorem we have

([a]+[b])" =

where =I
fp\I = p!
Z.
\iJ (p—i)!(i!)
Here denotes the product added to itself c. times
in Z,,. Observe that for I because (p—i)!(i!) divides p! and
so must divide the factor (p—I)! of p! [see Proposition 3, §2.6, and §2.7].
Now set [t,J = For I i<p, c.{1,) = [c,tj =[O] e Z,,, since
Consequently
([a] + [bfl" = + [0] + + [0] + [hJ".
Next, by a recursive argument, we can obtain the following theorem.

Fermat's Little Theorem. For all [a] e Z,,,

[a]" = [a],
which is equivalent, for all a e Z, to
a" a (modp)
or, if a 0 (modp), to
I (modp).

The proof of this result, stated by Fermat in 1640, was published by

Euler in 1736. We leave the proof to the reader with the observation that
[a]" = [a] [a+ 1]" = ([a]+[l])"
= [a]" + [i]"
= [a] + [I] = [a+ I].

Exercises

1. Prove Fermat's Little Theorem.

2. Show that the Euler function ço(n) gives the number of distinct proper
fractions between 0 and I whose denominators equal n.
54 Arithmetic of Integers chapter 2

3. a. Prove that p(2n:) if and only if in is odd.

Prove that
b. 0 (mod 2) for all ii> 2.
4. a. Prove that 4(?n2) =
b. Prove that q(mn)q'(d) = dq,(rn)q,(n), where d— (m,n) is the GCO of
m and ii.
5. For which of the integers 7, 9, 13, IS, 17, 24, 33, and 36 is the Euler
p-function a power of 2? (We shall find in §9.3 that = 2' if and
only if a regular n-gon can be constructed by ruler and compass.)
6. Suppose 2h+ I isa prime number p. Prove that
a. The residue class [2] c has multiplicative order 2h; that is,
[2]2h [I]
b. Ii is a power of 2.
c. p—I
This problem of number theory is important in the discussion of the
generalization of Gauss' construction of the regular polygon of 17 sides
7. For an odd integer n show that Z, has an even number of prime residue
classes (units).
8. Provide an alternate proof of 9,(pa) = by using the p-adic
expansion w0-t-w1 p+ 0< w1 <p, of a, I a <pa, as in
I if and only if I
§2.5. Note that (a, p) = <p.
 3

Introduction to Ring
Theory

Chapter 2 reviewed the associative, commutative, and distributive

properties for the addition and multiplication of integers. We referred to the
integers as a ring and defined residue class rings of integers. With these
serving as explicit examples, we now abstract certain of the algebraic proper-
ties of the integers to more general sets and operations that will constitute
rings, defined in the abstract.
Section 3.1 provides the basic definitions of ring theory. These
definitions, each illustrated by one or more examples, must be mastered,
almost like an alphabet. Sections 3.2 and 3.3 introduce from a ring-theoretic
view the fundamental algebraic concepts of homomorphism and direct sum.
Homomorphisms are discussed in general, with numerous specific examples
drawn from the integers and residue class rings of integers. The discussion
of direct sums in §3.3 is general, with specific direct sum considerations for
the ring Zm worked out in full in §3.5. Later the concepts of homomorphism
and direct sum arise in other contexts: for vector spaces in Chapter 4 and
for groups in Chapter 6.
Sections 3.4 and 3.5 have a twofold purpose. First, the complete
description of the algebraic-arithmetic structure of the residue class rings
4,, such as determination of orthogonal idempotents. subrings, and units, is

55
 Introduction to Ring Theory chapter 3

given. Second, techniques for analyzing ring structures are established. They
are general enough to apply to rings of polynomials and ultimately
provide the model for generalization to advanced topics in number theory
and function theory.
A natural continuation of the ring-theoretic development of §3.1 is
the examination in §3.6 of rings with either of two additional properties:
a multiplicative law of cancellation or existence of a multiplicative inverse
for each nonzero element. Here we abstract properties of the integers and
the rationals. The student is asked to examine prime, primary, and maximal
ideals in the exercises for two reasons: first, they provide further experience
in abstracting properties of Z and Zm to more general rings; and second,
significant theorems involving these special ideals can be proved only if
restrictive assumptions, such as the chain conditions, are imposed on the
rings.
The definitions, as presented and as typically applied in a first course
in algebra, do not lead to important theorems, but not for any lack of sig-
nificant ring and ideal theory. Such topics are customarily reserved for
graduate level study. Thus we provide a basic introduction to the subject
with emphasis on those aspects to be used later, first in a detailed analysis
of residue class of integers and field theory [Chapter 8]. In Chapters
8 and 9 the proofs of the existence of algebraically closed fields and the
Fundamental Theorem of Algebra utilize ideal-theoretic arguments.

§3.1 Basic Elements of Ring Theory

At this point in our discussion of the residue class ring Zm it is
helpful to have at hand some general definitions and facts from the theory of
rings. More importantly, we now abstract many of the algebraic properties
of the integers to define a particular algebraic structure called a ring [cf. §2.1,
axioms H—IV, and §2.9].
A ring R is a set of elements together with two laws of composition
(commonly called addition and multiplication) that associate to each ordered
pair (a, b) of elements in R a unique third element (denoted a + b for addition,
ab for multiplication) in R,subject to the following axioms:

Addition. Addition is associative and commutative, there exists an

additive identity element (commonly denoted by 0), and each element
r R has an additive inverse r' such that r + r' 0.
Multiplication. Multiplication is associative, and there exists a multi-
plicative identity element (commonly denoted by e or I) such that
R.
Distrihutivity. Multiplication is distributive over addition:
r(a+b) ra + rb,
(a-1 b)r = ar + br for all a,b,r e R.
 Basic Elements of Ring Theory 57

If furthermore ab = ba for all elements a, b E R, then the ring R is called a

commutative ring.
In §2.1 we encountered the ring of integers Z, and in §2.9 the residue
class rings of integers Zm. These constitute our basic examples of rings;
other examples are given in exercises. Students familiar with matrices and
linear transformations (topics treated in §4.4) will recognize that the set of
nxn matrices (n> I) with either integral or rational coefficients is a non-
commutative ring with the usual rules of matrix addition and multiplication.
Equivalently, the set of all endomorphisms (linear transformations) of an
n-dimensional vector space over a field constitutes a noncommutarive ring
[see the end and Proposition 3, §4.4].
In this book all rings are assumed to be commutative unless explicitly
stated to the contrary.
Because the defining algebraic axioms of a ring are abstracted from
the corresponding axioms for integers, their logical consequences carry over
to arbitrary rings R. We adapt from §2.1 the following properties, valid for
any (commutative or noncommutative) ring R.
Property I. The additive identity is unique.
Property 2. Furthermore, a = a+r, for some a e R, implies r = 0.
Property 3. The cancellation law for addition holds.
Property 4. The additive inverse a', commonly written —a, of an element
a E R is unique.
Property 5. Every equation a + x = b, for a, b e R, has a unique solution
x R.
Property 6. For any a R, a•0 = 0.

Property 7. The multiplicative identity is unique.

Property 9. The rule of signs for multiplication holds; that is, (— a) ( — b) = ab
for all a, b e R.
Note that Properties 8 and 10 involve the multiplicative law
of cancellation, valid in the ring of integers, but not valid for rings in general
(as was seen in §2.9 in the case of Z18). Therefore they are not included in
the present list. Rings in which these two additional properties hold are called
integral domains and will be discussed in §3.6.
Furthermore, the general associative and distributive laws for the
addition and multiplication of n elements hold since the proofs require the
Principle of Induction on the integral subscripts in the enumeration of the
elements (and not on the ring elements themselves). Thus for example,

for a, b1,.. , E R. We defer the proof of the general associative law to §6.1.
 Introduction to Ring Theory chapter 3

A nonempty subset T of a ring R is called a subring of R, if

(i) Tis closed under addition and additive inverse, as determined in R
(i.e.. for all a,b e T. a+h and —a are elements in T; these imply
that 0 e T).
(ii) T is closed under the given law of multiplication of R (a, b e T
ab e T), and T contains a multiplicative identity e1 such that
era = ae1 = a for all a e T.
Note that it is not postulated that the multiplicative identity er of T be the
same element as the multiplicative identity e in R.

Example 1. In Z20 the subset S — of cosets modulo 20 satisfies

the definition of a subring. Here [5] is the multiplicative identity of S, while [I] is
the multiplicative identity of Z20. (The additive identity of a subring is always that
of the ring itself.)

Example 2. The subset {[0], [10)) of Z20 is not a subring. It has no multiplicative
identity, but does satisfy the other properties of a subring.

REMARK I. In particular note that a subring is itself a ring; that is, its elements
satisfy the axiomatic properties of a ring.

REMARK 2. We use the preceding definition of subring because the ring-theoretic

interpretation of systems of simultaneous congruences and its generalizations in
the theory of algebraic numbers and algebraic geometry give rise to mappings of
rings R into rings S for which the image of the multiplicative identity of R is not
the multiplicative identity of S. (Some authors require that the subring have the same
multiplicative identity as the ring. Others, especially those concerned with non-
commutative structures, do not require a ring o have a multiplicative identity.)

It is common usage to refer to a subring Sc R as a nontrivial subring

of R ifS is neither R itself nor the zero subring {0}. Later, similar conventions
will apply to ideals, fields, groups, and other algebraic structures.
The following paragraphs are replete with definitions, which essen-
tially are agreements on terminology. While now they may seem a burden,
their importance can be appreciated later on. The concept of an ideal has
become extremely important in algebra since the work of Ernst Eduard
Kummer (1810—1893), Leopold Kronecker (1823—1891), and J. W. R.
Dedekind (1831—1916).
A nonempty subset A of a (commutative) ring R is called an ideal
in R if
(i) A is closed under addition
(ii) A is closed under multiplication by elements re R; that is, if
a e A and r e R, then rae A.
For noncommutative rings R it is necessary to distinguish between the fol-
lowing types of ideals:
§3.1 Basic Elements of Ring Theory 59

Left ideals. a e A, r R ra e A.
Right ideals. a e A, r R ar E A.
Two-sided ideals. a e A, r c R ar a A, ra a A.
For commutative rings, ra = ar implies that all ideals are two-sided.
We have already encountered ideals in the ring of integers Z In
any commutative ring R, a given element a generates an ideal (a) = {ra: r E R},
also denoted Ra. More generally, if S= {s1, ...,Sm} is any finite subset of R,
the ideal generated by S is the subset of all elements of R of the form
r1 s, +rmsm. where r. a R, I <i rn. Such an ideal is commonly denoted
bY(Si,...,Sm)OFRS1+"+RSm.
There is a considerable arithmetic of ideals paralleling and general-
izing that of the integers. For ideals A and B in a (commutative) ring R, we
define terms as follows.
1. Product:

AB = : a a B, for arbitrary n EN);

that is, the set of all finite sums of products of elements in A and B.
2. Sum:
A+B= {a+b:aaA,baB},
also called the greatest common divisor of A and B, denoted (A, B).
3. Intersection:
A rB = {raR: reA and ra B}
is simply the set-theoretic intersection. We also call A n B the least
common multiple of the ideals A, B, denoted [A, B].
4. Power: The power A" of an ideal. k a N, is the k-fold product
of A with itself. It consists of all finite sums of k-fold products a1
of elements of A.
By finite induction we can extend the concepts of product, sum, and inter-
section to any finite number of ideals.
Recalling the properties of GCD and LCM for integers and
2.7], we note the significance of the terms greatest common divisor and least
common multiple for ideals in a given ring in Proposition 1.

Proposition 1. Ideals A, B, C in a ring R satisfy the following properties.

(I) GCD properties:

A c (A,B).
B (A,B).
A c C,B c C (A,B) c C.
 Introduction to Ring Theory chapter 3

(ii) LCM properties:

[A,B] c A.
[A,B] c B.

Cc A,C c B Cc [A,B].
These properties are all immediate consequences of the definitions.
Less immediate is a second proposition.

Proposition 2. For ideals A, B, C in a ring R,

(i) [A,B](A,B) c AB.
(ii) (A,B) = (A,C) = R (A,BC) = R.
(iii) (A,B) = R A n B = AB.
Proof of (i) [cf. §2.7, Exercise 4].
[A,B](A,B) = (An B)(A-f-B)
= (A n B)A + (A n B)B c BA + AB = AB.
Equality does not always hold in general rings.
Proof of (ii).
forsomeaeA,beB.
Similarly
(A,C)= e=a'+c, forsomea'eA, eeC,
and therefore
e = e2 = aa' + ac + ba' + be.
EA eBC

Hence every element r e R is the sum of

r(aa'+ac+ba') e A and rbc E BC,

which means that R = A+ BC = (A, BC).
Proof of (iii). First, note that ABc A n B, since ABc A and ABc B.
To prove the reverse inclusion, consider x E A n B. Since A + B = R, we
can write e = a+b for some ac A, be B. Hence
x = ye = xa + xb = ax + xb e AB,
as each summand on the right-hand side lies in AB. Therefore A n Bc AB,
and A n B= AD, as asserted.

As we did in the particular rings Zm in §2.10, we consider here

elements in a ring with specific properties. An element a in a ring R is:
§3.1 Basic Elements of Ring Theory 61

A zero divisor if a 0 and if there exists b 0 e R such that ab = 0.

Nilpotent if a" = 0 for some n N.
Idempotent if a2 a.
A unit if there exists some b E R such that ab = e, the multiplicative
identity of R.

Exercises

1. a. With reference to Exercise 4(a)—(e) in §1.1 verify that the set of

subsets of a fixed set S constitutes a ring where intersection of sets
is taken as the multiplicative law of composition and symmetric
difference as the additive law.
b. Show that, despite Exercise 4(f) in §1.1, the set of subsets of S with
the laws of intersection and union does not Constitute a ring.
2. Suppose that R is a ring such that a2 = a for all its elements a. Prove
that a+a = 0 for all ae Rand that ab ba for all a,b e R.
3. Let a, b be elements of a commutative ring R with multiplicative identity e.
a. Prove that (—e)a = —a and —(a—b) = b—a.
b. For all m, n e N prove that (ab)" = a"b", e" = e, and amn =
4. a. Let x be a nilpotent element in the ring R. Prove that e + x and e — x
units in R.
are
b. If a is an idempotent element in R, prove that e—a is an idempotent
and a zero divisor.
5. An element x e R is called unipotent if e — x is nilpotent. Prove that the
product xy of unipotent elements is again unipotent. Prove also that
every unipotent element of R is a unit of R whose inverse is unipotent.
6. Find the unipotent elements in Z,, for
a. ii = 28 b. a = 18 c. n 36.
7. Suppose that a and b are nilpotent elements in the commutative ring R.
Prove that
a. a+b is nilpotent.
b. The nilpotent elements constitute an ideal in R.

The ideal of Exercise 7(b) is sometimes referred to as the radical of the ring,
denoted either Rad(O) or Rad R [cf. Exercise 8, below].

8. Let A be an ideal in the commutative ring R. Define the radical of A to

be the set
Rad A = {r E R r" a A, for some a a N}.

a. Prove that RadA is an ideal in R.

b. Prove that Rad(RadA) = RadA.
62 Introduction to Ring Theory chapter 3

9. a. Describe all ideals A in Z for which Rad A = A [cf. Proposition 2,

b. Describe RadA for A = (100) C for A = (80) C Z,

JO. Determine the radicals of the following rings [Cf. Exercise 7(b)] where
p,q are distinct primes. Distinguish the cases ii = i and is> I,
a. 4, b.
11. Using the definitions of product and sum of ideals, prove the following
associative, commutative, and distributive laws for ideals A, B, C in a
ring R.
(i) (AB)C A(BC)
(ii) AB BA
(iii) (,4+B)C = AC+ BC.
12. Show that the set-theoretic union of ideals A, B in a ring R need not be
an ideal.
13. a. Consider the ideal A = (a1 generated by the elements a1, ...,
a (commutative) ring R. Describe the elements of the ideal A2 in
terms of those in A.
b. Prove that if A = A2, then there exists an idempotent b E A such
that A = Rb = {rb: r R}. In other words, prove that A = A2
implies A is a principal ideal generated by an idempotent element b.
c. IfS {s1, ..., S.,,} is a finite subset of R, prove that the ideal generated
by S is the GCD of the principal ideals

S, = R = {s1 r: r E R}, I i in.

14. Define laws of composition # and * on N by

a#b (a,b), the

a * b = [a,b], the 1CM.

With these laws of composition is N a ring?

§3.2 Ring Homomorphisms

We turn now to mappings (functions) from one ring to another which
are compatible with the ring structure. The root word is commonly
used in higher level algebra courses to denote a mapping from one algebraic
structure (such as a ring) to another of the same kind. The prefix homo-
means "structure-preserving" or "compatible": other prefixes, to be intro-
duced shortly, denote homomorphisms with special additional properties.
in later chapters we shall encounter homomorphisms of other algebraic
structures such as vector spaces and groups. Vector space homomorphisms
are customarily called linear transformations or linear operators.
Consider rings R and S. each with its own laws of composition.
A (ring) homomorphism q' of R into S is a single-valued mapping

p: R S
§3.2 Ring Homomorphisms 63

such that for all a, b e R,

(I) (p(a+b) = q(a) + ip(b)
I I
addition as addition as
defined in R defined in S

q(ab) = q(a)q(b)
I I multiplication
multiplication
as defined in R as defined tn S

The subset of S
= {seS:s=.q(a), for someae R)
= (q(a):aeR}
is a subring of S, called the image of R by ço. The image is also denoted

An important but simple consequence of property (i) of a homo-

morphism q: R —* S is that c (OR) = where °R are the additive identities
of the rings R, S respectively. Since
= = +
by property (i), the uniqueness of the additive identity Os of S implies that
co(OR) = Os [Property 2, §3.1].
The multiplicative property (ii) of such a homomorphism implies
that the image of the multiplicative identity eR of R is the multiplicative
identity of the subring p(R) c S. The defining equations of a ring homo-
morphism do not imply that p(eR) is necessarily the multiplicative identity
of the ring S. (See Example 6 below.) We gather these facts in the following
proposition.

Proposition 1. If q: R —' S is a ring homomorphism, then

(i) 4'(OR) =
(ii) (p(eR) =

One application of Proposition I is to show that some mappings

from one ring to another cannot be homomorphisms. (See Example 11
below.)
Before considering further properties and special types of homo-
morphisms, we construct a few examples of homomorphisms, and of
mappings between rings which are not homomorphisms. When several moduli
m for the residue class rings Zm are involved, subscripts on the cosets can
clarify which ring a coset belongs to.

Example 1. The ring contains the subring (which happens also to be an ideal)
S = {(O]20,[5]2o,(1O]20,[15}2o}.
 Introduction to Ring Theory chapter 3

We can define a homomorphism Z4 by specifying that = [114. To

complete the definition of we compute:
49(110)20) = ço([5J20+[5)20)
= c([5]2o) + 9'(t5120)
= [1)4 + [1J4 = [2J4,
4'U15]2o) = 49([5]2o) + co([10)20)
= 9'([512o) + 4'([I0]20)
= (1)4 + [2)4 = [314,
49([0]20) = 4'([20]20) = 49([l0]20+[l0]20)
4'([10]20) + 4'([10]20)
= [2)4 + (2]4 = [4]4 = [014.
To be assured that the mapping 4', as defined, is indeed a homomorphism, we must
verify that respects the multiplicative structures of S and Z4. That is, that
fp([a]20 [b]20) rp([a]20)4'([b]20) for any [a)20, [b]20 E S. For example,

4'([15]2o[lO]2o) = 4'([150)20)
= [3]4[2]4
= 49((l5]20)q.'((l0]2o).
The complete verification of the multiplicative property (ii) for follows easily from
the fact that the elements of S are integral multiples of [5)20. Since general elements
[a]20, [b]20 e Scan be expressed as [5m)20, [Sn)20,
4'([5fl1)20 [511)20) ç([25inn]20 = qi([5mn]20)
= by property (I)
= = [mn]4
= [ni]4 [n]4
= ço([5m]20)ço([5n}20).

Example 2. A similar example is the mapping çb of the subring S' {[0]20, [4]2o,
(8)20, (12120, [16)20) C z20 to Z5 given by = [l]5 and more generally
n-[l]5

That b?', so defined, is a homomorphism is left as an exercise. The proof is analogous

to the discussion in Example 1.

it may seem an obvious point, but in considering whether a rule of

correspondence 1: R S between two rings is a homomorphism, we must
first insure that it is a mapping as, for instance, in the next
examples.

Example 3. We cannol define a mapping! from Z3, Z by setting J([aJ3) = a.

This is not a well-defined function, because [1313-, = [50)3, but 13 50.

Example 4. Similarly, setting qi([1]4) = cannot define a homomorphism

from Z4 to Z5, because [1)4 = [5)4 and, if were a homomorphism,
4'([5)4) = 4'(5-{114) 5•4'([114)
= = 4'([l]4).
§3.2 Ring Homomorphisms 65

We shall find later that the only homomorphism from Z4 to Z5 is trivial in

the sense that each element of Z4 must be mapped to [015.

Examples 3 and 4 also illustrate the fact that for a homomorphism

p: R S. the original relations in R must be carried into relations valid in S.
For instance, if a+b+c=O in R. then must equal 0 inS.

Example 5. There are homomorphisms

a: Z4 Z20, Z20 Z4,

given by a([aJ4) = [5a120 and 9"([a]20) [a]4.

These mappings are independent of the choices of representatives of cosets, since
a([a+4r]4) [5(a+4r)]20 = [5a]20 a([aJ4),

[a+20s]4 = [a]4

Combining Examples 1 and 5, we have the following sequence of homomorphisms:

a
S —, Z20 —+ Z4,

where S=

Example 6. For the homomorphism a: Z4 —# Z20 in Example 5, note that

a([l]4) = (5]20 [1120, illustrating that a homomorphism from one ring to another
need not map the multiplicative identity of the first to the multiplicative identity of
the second.

Example 7. We make two final observations about the homomorphisms

a: lmo' = a(Z4) = ([5120, [10120,
Z4 of Example 5. First,
[15120. [0120) is a subring of Z20, whose multiplicative identity is [5120 = cr([l]4).
Second, the subset
{[a]20 e Z20 : = [014) = {[a]20 e Z20 a 0 (mod 4))
:

[4120, [8)20, [12120, [16]20)

is an ideal in Z20. (In fact it is a subring with multiplicative identity [16]20).

Propositions 2 and 3 helow state the general case of the properties

illustrated here: the image of a homomorphism q: S is a subring of S.
and the set of elements in R mapped to the additive identity of S is an ideal in R.
The following agreements on terminology pertain to special types of
ring homomorphisms
q: R S.

A homomorphism is said to be onto (alternatively, surjective or

an epimorphism) if p(R) = S; in other words, if for each se S there exists
some r E R for which p(r) = s.
66 Introduction to Ring Theory chapter 3

Example 8. The homomorphisms ,p, and of Examples 1, 2, and 5 above are

onto. For instance,
= [014. [2)4,

4"([lJzo) = [1)4, q'([3)20) = [3]4;

hence there is some element in Z20 which is mapped by to any given element in Z4.

A homomorphism q' is said to be one-one (alternatively, injective or

a monomorphism) if implies r=r'. An equivalent description
of a one-one homomorphism is that implies that r=OR since
(p(r') implies that p(r—r') = os•

Example 9. The homomorphisms and a of Examples 1, 2, and 5 are one-one.

For example, consider [x)4 e Z4 such that a((x)4) = [0)20. First, ci([x]4) =
= [x.5]20; further [x5]20 = [0)20 if and only if which means that
[x)4 = (O]4. Hence a is one-one.

A homomorphism q, is said to be one-one, onto (alternatively,

bijective or an isomorphism) if it is both injective (one-one) and surjective
(onto).

Example 10. Only the homomorphisms and ifr in Examples 1, 2, and 5 are
isomorphisms.

Example 11. To illustrate that not all one-one, surjective mappings of rings need
be isomorphisms, consider the function T: Z Z defined by T(n) = n+ 1. Since
T(0) 0, T is not a homomorphism by Proposition I. Therefore T cannot be an
isomorphism.
Similarly the one-one mapping T': Z Z, defined by T'(n) = 2tz, is not a
homomorphism, even though T'(O) = 0. The image T'(Z) has no multiplicative
identity, and so is not a subring; hence T' cannot be a ring homomorphism. More
explicitly, observe, for example, that 16 = T'(S) = T'(2•4) 48 = 32.

If p: R—'S is an isomorphism. then R and S are said to be

isomorphic rings: this fact is denoted
R S.

For given s e S. the set (which may be empty)

= {reR:ço(r)=s}
is called the complete inverse image of s S with respect to the homomorphism
q,. Note that '(s) 0
for all s S is equivalent to saying is onto, and
that is one-one if and only if - i(s) has at most one element for each
se S. For an isomorphism q the inverse image 1(i) consists of preeise/t
one element for each s S.
Proposition 2 concerns ideals and inverse images.
§3.2 Ring Homomorphisms 67

Proposition 2. If At is an ideal in the image çp(R) of a homomorphism

qi: R—+ S, the complete inverse image A = is an ideal in R.

Proof. Consider a. b A and r a R. Then

q,(a+b) = + ç(b) a At,
(since each belong to At, so does their sum, because At is an ideal)
p(ra) = a At,
(since belongs to At, so does the product .o(r)ço(a), because At is an ideal)

and hence a+b and ra are elements in A. We conclude that A is an ideal in R.

The kernel of the homomorphism q' is

kerq = = {reR:q(r)=O5}.
Notethat ço is injective if and only if kerq = For an integer tn, the
mapping of a a Z to its residue class [a] in Zm is a homomorphism of Z
onto Zm. whose kernel is the ideal (m).

Proposition 3. If R—.S is a ring homomorphism, then kerq' is a (two-

sided) ideal in R (even if R is noncommutative).

We present two proofs. First, since is an ideal in S, then the

complete inverse image = kerq is an ideal in R by Proposition 2.
The second proof is by direct verification. For any a, b a ker p and any r E R,
q(a+b) = q(a) + q(b) = os + =
q,(ar) = q(a)q(r) = = Os,
q,(ra) = q(r)q(a) = = Os.

In the special case that S= R, a homomorphism p: R—R is called

an endomorphism. More commonly, an isomorphism q: R will be
called an automorphism.

Example 12. The mapping Z20 Z20 given by [a]20 —+ [5a]20 is an endo-
morphism whose kernel {[4n]20 : n a Z} is isomorphic to

Example 13. The mapping of the ring of complex numbers C given by

a+ bi a— bi, where a, b a R, is an automorphism of C.

To complete our listing of terminology frequently used to describe

homomorphisms, we introduce the term embedding to denote an injective
homomorphism q: S for which (p(eR) = e5. Not all injective homo-
morphisms are embeddings (consider Example 5 in which a: is
injective, and a([1J4) = {5]20 [1120). Nor are all embeddings isomor-
phisms. For example, the identity map of the integers Z into the rational
68 Introduction to Ring Theory chapter 3

numbers Q is an embedding, but not an isomorphism, as the map is not

surjective.
If = Os, we call çø a trivial homomorphism because every
element of R must then be mapped to the additive identity of S; p(R) =
S.
We conclude this section by outlining a fundamental property of
ring homomorphisms for which we need first an equivalence concept for
elements in a ring. The reader is asked to develop the details in Exercise 3.
For an ideal A in a commutative ring R, introduce a congruence
(or equivalence) relation in R by defining ab (mod A) if and only if b—a e A
for a,b in R. Abstracting the construction of Zm = Z/(m) given in §2.8 and
§2.9, we can construct a residue class ring R/A of R modulo A. In other
words, the set of residue or equivalence classes of elements of R modulo
the ideal A can be given a ring structure analogous to the ring structure of
Zm. (The ring RIA is also called the quotient ring of R by A.)
By mapping each element of R to its residue (equivalence) class
modulo the ideal A, we obtain a surjective homomorphism 'TA from the
ring R onto the residue class ring R/A. Further, the kernel of ltA is the ideal A.
The homomorphism ltA commonly is called the natural or canonical projection
of R onto the quotient ring R/A.
The following theorem completes this discussion. We omit the proof,
preferring to provide it in a group-theoretic context in §6.8 after the reader
has gained more abstract algebraic experience. The proof is not especially
difficult, but since it is similar to the analogue in group theory, we present
it only once.

Theorem. If q: R—+S is a surjective ring homomorphism with kernel A,

there exists an isomorphism A of the residue class ring R/A onto S. such that
= 1. o

Exercises

1. Consider subrings
S= ([Os, [5],[10], [15]),
T = ([0], [4], [8], [12],
of Z20, and define the following ring homomorphisms:
S Z4 by = [114,
Z5 by =
a: Z4 Z20 by c([l]4) = [5120,
t: Z5 Z20 by =
p': Z20 Z4 by = [a]4,
u": by = [a]5.
§3.2 Ring Homomorphisms

a. Prove that 0', and are surjective.

b. Prove that a, and r are injective.
c. Show that and are isornorphisms.
d. Determine explicitly the kernel of each of these six homomorphisms.
e. Determine the complete inverse image of each element in Z4 and
Z20 with respect to the appropriate maps 9,, a, and
f. Determine im cx and Im r.
2. a.Prove that the mapping Z — 4, defined by 9,(a) = [a],,, is a ring
homomorphism.
b. What is the kernel of this mapping?
c. Is the mapping 42—' Z36 defined by tl'([a]12) = [a]36 for
0 a < 12 a homomorphism? Why?
3. Let A be an ideal in the commutative ring R.
a. Show that the congruence relation a b (mod A) if and only if
b—a E A for a,b c R is an equivalence relation on R.
b. Generalize the construction of Z/(m) in §*2.8 and 2.9 to the residue
class ring RIA. That is, give the set R/A of residue (equivalence)
classes of R modulo A the structure of a ring.
c. Show that the mapping ir4: R R/A, which maps each element of
R to its residue class modulo A, is a surjective homomorphism.
d. Show that kernA = A.
4. Prove that if R S is a surjective ring homomorphism, then
S Rfkerç. (No assumption of commutativity is needed.)
5. Consider a homomorphism R S.
a. If is surjective, prove that (D(eR) = e5.
b. Prove that 9,(eR) is an idempotent in S.
c. If co(eR) isdifferent from Os and es, prove that S must contain non-
trivial zero divisors. Hint: Consider e5 —
6. a. Verify that the mapping Z,, given by 9,([a]) = [ai is a ring
homomorphism [see Fermat's Little Theorem, §2.12]. Is an
isomorphism?
b. Is the mapping Z Z given by 9,(a) = a homomorphism?
7. An ideal A in a commutative ring R is termed dense if rA = (ra : a A} =
(0) for any r e R implies r = 0.
a. Prove that if A is dense, then so is any ideal B A.
b. Prove that if the ideals A,B are dense in R, then so are AB and
A C'i B.
8. In analysis we frequently encounter the ring of continuous (or dif-
ferentiable) real-valued functions defined on the real line R, where laws
of composition are defined for f,g e F by
(J+g)(x) =J(x)+g(x), (Jg)(x) =f(x)g(x)
for alIxeR.
a. Verify that F so defined is a (commutative) ring.
b. Verify for any nonempty subset S of R that .9' = {fe F :f(x) = 0
for all xc is an ideal in F.
c. Show that F has nontrivial zero divisors.
70 Introduction to Ring Theory chapter 3

9. Consider a left ideal J in a noncornrnutative ring R. The set A A(J) =

e R: rj 0 for all j E J} is called the annihilator of J.
a. Prove that A(J) is a two-sided ideal in R.
b. With reference to Exercise 8(b), what is the annihilator of ..V when
S is the Set of integers Z?
10. Use induction to prove that the sum of ideals A1 is an ideal in
R for any ii N. The ideal

= EA1 = 1=

also denoted B = A11), is called the greatest common divisor of the

ideals A1.
11. For elements a1, e R, I I in, I j in, prove the distributivity
property that
jm n \ m+n

\i=I / \j=i ,/ h=I \i+j=h

12. Let R be a ring with laws of addition and multiplication denoted by +
and juxtaposition, respectively. Let e denote the multiplicative identity.
Now define new laws of addition # and multiplication * as follows:
a#b=a+b+e, a*b=a+b+ab.
a. Prove that R is a ring with respect to the composition laws # and *.
b. Prove that

given by = a—e, for ae R, is an isomorphism,

c. Determine the inverse of the isomorphism
13. Prove that if R S is a ring homomorphism and if J is an ideal in S,
then '(J) is an ideal in R. It is not required that I im [Cf.
Proposition 2.]
14. The ring 3 of quaternions (due to William Rowan Hamilton, 1805—1865)
can be described as the set of all formal combinations {a+bi+cj+dk:
a,b,c,de RI, i.e., a four-dimensional vector space over R with basis
{l, i,j, k} [cf. and 4.2]. Addition is defined component-wise. Mul-
tiplication of two quaternions is given by taking I as the multiplicative
identity in requiring that all real numbers r commute with I, j, and k,
and extending by distributivity the following products of the basic
elements:
= J2 = k2 = —1,
ij=k, jk=i, ki=j,
ji=—k, kj=—i, ik=—j.
Observe that 3 is a noncommutative ring.
a. Define the conjugate 4* of A = a0+a1i.l-a2j+a3k E 3 to be
= ao—a1i—a2j—a3k.
Prove that AA* is a positive real number for .4 0.
b. Show that every nonzero element of 3 is a unit.
§3.3 Direct Sums of Rings 71

In 1878 Georg Frobenius (1849—1917) proved that the quaternions are the only
vector space (of any dimension) over R with a noncommutative associative
product for which every nonzero element is a unit. The complex numbers are
a two-dimensional vector space over R, but with a commutative product.

Ic. The ring in Exercise 14 is called the ring of (real) quaternions. Define
the ring of rational quaternions by replacing real numbers in Exercise 14
with rational numbers.
a. Prove that for each A 0 in this new system AA* is a positive
rational number.
b. Show that every nonzero element is a unit.
16. In Exercise 15, replace the rational numbers by the set of complex
numbers of the form a+b where a,b eQ. Assume that
i,j,k and that commutes with i,j,k. Prove that the
resulting ring contains nontrivial zero divisors.
17. An ideal I in a ring R is said to be a prime ideal if ab c J implies either
a I
or b E J. Suppose is a surjective homomorphism of the ring R
onto the ring S, and P is a prime ideal in S. Prove that the inverse image
9i1(P) is a prime ideal in R.
IS. Consider an ideal A contained in Rad R [cf. Exercise 8, §3.!]. Assume
that x eR satisfies x2 x (mod A), i.e., x is idempotent modulo A.
Prove the existence of an idempotent y e R such that y x (mod A).
Hint: Try y = x+t(I —2x) with ze R, requiring that y2 = y, and use
the /órrnal expansion z = —(I +4zY 112], letting z = x2—x.

§3.3 Direct Sums of Rings

continue our general discussion of rings with consideration of
We
two related concepts: the external direct sum of rings and the internal direct
sum of ideals. Section 3.5 provides examples in Zm of the direct sum dis-
cussion of the present section. Thus most of the exercises on direct sums are
deferred until then.
The external direct sum is a tool to construct new rings from given
ones, while internal direct sums are important for the analysis and decompo-
sition of a given ring into algebraically "simpler" rings. The significance of
these concepts will be explored within the context of residue class rings of
integers in §3.5 and of polynomials in §5.3. In §6.10 we extend the concept
of direct sum to groups. It should be noted too that these concepts and their
applications in Chapter 7 are prototypes of important aspects of the theory
of algebraic numbers and functions.
Let R1 = {a1,b,,...), ..., = ...} be rings with respective
multiplicative identities e1. ..., n> 1. Consider the set

of all n-tuplcs a = (a1. ..., a,,), b (b1 h,,) e = (e1, . .., e,,) of elements
in R1,...,R,,. Define (a1,.. ,a,,)=(b1,...,b,,) if and only 1 in.
72 IntroductIon to Ring Theory chapter 3

Furthermore, define rules of addition and multiplication by

a+b =
ab =
Then R is, as is easily checked, a commutative ring with e for its multiplicative
identity. The ring R is called the (external) direct sum of the component rings
R1, ..., For each i, I i n, the map
a1 (0 0,a1,0, ...,0)
I
ith component

is an ifijective homomorphism R1—. R. Furthermore, the multiplicative

identity e of R satisfies

In particular, e, and is a subring (and also an ideal) in R.

A ring R is said to be the (internal) direct sum R = A1 ®
of ideals A1, if every element r E R can be expressed uniquely as the
sum of elements a I i n. (If R is a noncommutative ring we require
that the ideals A, be two-sided.)

Proposition 1. if R = A ... is the (internal) direct sum of ideals,

then the ideals A, are subrings of R, I I n.
Proof An ideal is a subring if and only if it has a multiplicative identity.
Write the multiplicative identity e of R as e = e1 +••• the unique sum
of elements e• e A,. We claim ea, a. for all a e A.. and hence that e, is the
multiplicative identity in A.. The proof rests on the fact that a,e1 0 for
any a,€A,, where Suppose to the contrary that for some dis-
tinct land). Then we would have
a, = a,e = a,e1+ ae, + + a,ej+ +
I I
= a,e1 + + + +0+ +
I I
EA,

which are two distinct expressions for a1 as the sum of elements in the ideals
Ak, contradicting the hypothesis that the sum A1 ... is direct.
Therefore we have
a, = ae = a,(e,+• = a,e,,
for all a, A4. Hence e4 is the multiplicative identity in A,, which is then a
subring as asserted.
§3.3 Direct Sums of Rings 73

We note further that e.e1 = e• and

e elements will be discussed in detail
in §3.4 for the ring Zm.
We have the following consequence of this proof.

Corollary. If a1 and aj E where I then = 0.

Since ej is the multiplicative identity of we can write

= = — 0.

The terms external and internal refer to the fact that the external direct sum
R of rings R1 is outside of ("external to") the rings ..., whereas
expressing R as the internal direct sum of ideals A1, ..., emphasizes the structure
within R. The following proposition relates these two concepts.

Proposition 2. If R= and q: R.—lR is the mapping of R1 to

the ith component of R, then R=p1(R1)® Conversely, if

The proof is not difficult, and is given in the case of R = in §3.5.

The general proof is left as an exercise in §35.

Exercises

I. Show that the (external) direct sum Z2 + Z2 is not isomorphic to Z4.

2. a. Prove that R = Z2 + Z3 and S = Z3 -I- Z2 are isomorphic rings.
b. Let R1, R2 be two distinct rings, and define R = R1 + R2 and
S= -i- Prove that R and S are isomorphic rings and that
R S.
3. Prove that Z6 is isomorphic to the ring of Exercise 2(a).
4. In Z36 consider the ideals A and B generated by (28] and (9], respectively.
a. Verify that [28] and (9] are orthogonal idempotents in Z.36, whose
sum is [1].
b. Prove that Z36 A ® B.
c. Verify that A and B are also subrings of Z36.
5. a. Write Z20 as the (internal) direct sum of two ideals, which are also
subrings of Z20.
b. Prove that the ideals of part (a) are principal with idempotent
generators.
6. Let G Z2+Z24 Z4, and write
co(([a]2, [b]2, [c14)) = ([12, [c]2, [a+ b]4),
vQhere a, b, c e Z. Does co describe a homomorphism from G to G?
74 IntroductIon to Ring Theory chapter 3

7. An idempotent u in a ring R, in which e-f e 0, is termed irreducible if

there exist no nonzero idempotents v, w e R such that u v + w and
vw 0. Prove that two distinct irreducible idempotents u and u1 of R
satisfy 0 (i.e., they are orthogonal).
8. Prove that R = Ru Rv if u, v are orthogonal idempotents and e = u + v,
where Ru = fru: r c R} denotes the ideal generated by u a R.

§3.4 Residue Class Rings of Integers

With the general ring-theoretic results of 1 through 3.3 available
we undertake a systematic examination of the residue class rings of integers
Zm, introduced in §2.9. First we construct systems of orthogonal idempotents
in Zm and prove that every ideal in Zm is a principal ideal. In the next section
we express 1,,, as the internal direct sum of ideals generated by these orthogonal
idempotents and as the external direct sum of certain other residue class
rings of integers.
In this section and the next we consider the residue class ring Zm
for an arbitrary, but fixed modulus in; in #0, ± I. Assume throughout that
m has the factorization
= ... en,,, where (me, = 1, for I # J.
(It is implicitly assumed that the modulus in has at least two distinct prime
divisors.)
For each i, I I < n, consider the system of congruences
0 (modm1)

I (mod in1)

0 (mod
or briefly, (mod in), I j n,
where = 0 for j I, and = 1. (We refer to as the Kronecker delta,
after Leopold Kronecker.) Each of then systems of congruences has a solution
e5 which is unique modulo in by the Chinese Remainder Theorem of §2.11.
Furthermore,
e2 e. 1j n.
Consequently e12 e1 (modm) [cf. Proposition 4, §2.6]. In terms of elements
Of 4,, have [e1j2 [ej and [e4) # [0], [e1] # [I]; thus each coset [e,]
we
is a nontrivial idempotent in the residue class ring
Moreover,
9, (mod I i,j, h n,
and so for h 1,

9, e,, 0 (mod in), 1 .1 ii.

§3.4 Residue Class Rings of Integers 75

Therefore, whenever Ii
e,eh 0 (modnz),

and in
[e'] [eh] = [0].
In other words, [e1] and [eh] are orthogonal idempotents in Zm. Finally,
for I

e1+ + e1+ + =

and therefore
e1 + + e, + + I (mod in).
Equivalently in Zm,
{e1] + + [e1] + .. + [en] = [1].
Thus we have proved the following proposition.

Proposition I. To the product decomposition rn = in1 of the modulus

in, there corresponds an additive decomposition [1] [e1] of the identity
[1] of the residue class ring Zm into mutually orthogonal idempotents [e1].

In §2.4, we demonstrated that any ideal in the ring of integers Z is

principal. We now prove the corresponding result for the ring Zm.

Proposition 2. Any ideal A* c Zm is principal.

Proof. The ideals ([0]) and ([I]) = Zm are the trivial cases as for Z. Let
be any nontrivial ideal in Zm and consider the set
A = {a e Z: [a] e A*}.
Then A, the inverse image of A* for the coset mapping a-÷[a] E Zm, is an
ideal of Z by Proposition 2, §3.2.
Since every ideal in Z is principal [Theorem, §2.4], we can write
A = (h) for some h E Z. Therefore [h] E A* and ([h]) c
Conversely, for
any element [a] in A*, each b e [a] belongs to A. Thus b = hx for some
x e Z, and
[a] = [b] = {/s][x] e ([h]).
in other words, A* ([h]), the ideal generated by the residue class [h].
Consequently A* = ([h]) is a principal ideal. as asserted.
Note that for any ideal A* Zm, the corresponding ideal A in Z
contains (m). in fact this observation is half of the proof of the following
proposition.
76 Introduction to Ring Theory chapter 3

Proposition 3. There is a one-one correspondence between ideals in Zm and

those in Z which contain (rn).

To complete the proof, consider any ideal A Z, containing (m).

The surjective homomorphism it: defined by ir(a)=[a], maps A
to an ideal n(A) as is easily verified. Further, for ideals A, B, both
containing (m),
ir(A) = ir(B) A = B.

Example. We conclude this section with an illustration of the general discussion of

the residue class rings Zm. Consider Z20. The modulus m = 20 has the factorization
20 = 22.5. Solving for orthogonal idempotents, we have
e1 I (mod 4) and e1 0 (mod 5);
hence e1 = I + 4s 0 (mod5).
A particular solution is e1 = 5. Furthermore, the congruences
e2 0 (mod4) and e2 I (mod 5)
imply e2 = I + 5u 0 (mod 4);
a particular solution is e2 = 16. Direct computation then yields
e1e2 = 5.16 0 (mod20),
e12 = 25 5 e1 (mod 20),
= 256 16 e2 (mod 20).
Furthermore, e1 + e2 = 21 I (mod 20).
These congruences are reflected as equalities for the cosets in Z20 as follows:
[e1][e2] 0, [e1] + (e2] = [1],
[e]2 = [e1], [e2]2 = [e2].
The following observation concerning this illustration is a prelude to the
next section. The multiples [0], [5], [JO], [IS) of [e1) = [5] constitute an ideal (and
subring) S = ([5)) in Z20, isomorphic to Z4. The isomorphism is given by
9'([512o) = [114. Similarly, T= ([4)) is an ideal (and subring) in Z20, isomorphic to
Z5. The isomorphism is given by = [1)5 [cf. §3.2].

Exercises

I. Find all ideals in the residue class ring Zm for

a. m=28 b. m=21 c. m=l5.
2. Find orthogonal idempotents in Z. for
a. ,n=36 I,. ,n=40 c. ,n—30
d. m = 35 e. m = 58 1. m = 120.
§3.5 Direct Sum Decompositions of Z, 77

3. a. Find in Z28 two distinct idempotents whose sum is the multiplicative

identity.
b. Exhibit in Z28 a subring isomorphic to Z7.
c. Show that every nonzero element of the subring in part (b) has a
multiplicative inverse.
4. Repeat Exercise 3 with 28 replaced by 21.
5. Repeat Exercise 3(a) with 28 replaced by 15.
6. a. Let A be an ideal in a ring R and R—' S a ring homomorphism.
Prove that çD(A) is an ideal in S if is surjective.
b. Show by example that need not be an ideal in S if is not
surjective.
7. Verify that if A is an ideal in Z, then ir(A) is an ideal ifl Zm, where
ii: Z —' is the projection onto cosets modulo in, it(a) = (aim.

§3.5 Direct Sum Decompositions of Z1,

This section continues the study of a general residue class ring Zm
as an internal and external direct sum. It entails arguments on homomorphisms
which can be considered as models for similar structure problems in general
ideal theory and in group theory [see 7.1, and 7.8].
As in §3.4 we write as the product m = m1 m2 of n 2
relatively prime factors. Setting
9= rn/nh = in I I inn,

we observe that the GCD (g1, . .,g,,) = 1. Thus

[cf. Exercise 12, §2.5].Since every 1, is divisible by rn, we have

= 9i h1 + . + h, + - -. + g• (mod rn.)
and 94h1 0 (mod for everyj I.

Consequently, g.h4 (mod for each fixed I and all j, 1 i, j n. In

other words, g, h, is a solution of the system of congruences x1 (mod
1 whose solution e, is unique modulo in according to the Chinese
Remainder Theorem Ii]. Thus
g4h, e, (modrn) or = [e1] in Zm.

Next let B' = ([e1]) denote the principal ideal generated in Zm by

the ith orthogonal idempotent [e1] Of Zm. That is.
= [91h1] Zm = {[gek] [z] : [z] E Zm}.
The following proposition states properties of the ideals
1 fn.
78 Introduction to Ring Theory chapter 3

Proposition 1. The ideals 87, I <I n. satisfy

(i) B72 = B.*.
(ii) B7B' = ([0]). for i k.
(iii) B7 B," = ([0]). for i k.

Detailed proofs are left as exercises. Note that since

B7 has a multiplicative identity,
namely [e1]. Hence [a][e,][e1] = [a][e1], as [e1] is idempotent, which
proves (i). Since the ideal B7 has a multiplicative identity, Proposition 1(i)
implies that it is a subring of Zm.
Utilizing the additive decomposition of [I] e Z,,, obtained in §3.4,
we have for [a] Zm,
[a] = [a][l] =
= + +
[a] [e1] of [a] lies in B7; symbolically.

In other words, is the sum of the ideals B7, I I< n.

Proposition 2. For B7 defined as above, the sum Zr, = B' is

direct.

That is, for any [a] Zm the components {b1] e B7, I i n. in

the expression
(a) [al = [b1J + ... + [bj
are uniquely determined by the coset [a] [see §3.3]. To establish the uniqueness
of representation of [a] in equation (*) we must show for any other represen
tation
[a] = [c1] + + [en], [c1]
that [b1] = [c1]. I I n. Since
[a] — [a] = [0] = [b1—c1] + +
we have [c2—b2] + + = [b1 —c1] e B,
or equivalently b1 — 0 (modm2
Also, cj — 0 (mod in1 ni,, rn,= for all j distinct from I. This
congruence implies c,— b, 0 (mod in,). Consequently

— =

so that (b, — c,) E (in,) (in, But

(in,) inn) (in1 rn2 inn) = (in),
 Direct Sum Decompositions of Zm 79

because of the assumption that (in,,mj) = I for Thus [b1 —c1] [0]
and [c1] = [b1]. Examining in turn the dillerences on the left side,
we find that [ei] = [by], which completes the proof.

Notational ('onvention. In order to simplify notation throughout the

remainder of this section, we denote the residue class rings Z/(,n,) = Zmj by
Z., and their elements [aim, by We continue to use unmodified brackets
to denote the elements [a] = [aIm in Zrn.

Proposition 3. There exists, for each i. I i n, an isomorphism (p. of the

residue class ring Z, with the subring = ([e,]) of Z,,.
Proof. To define an isomorphism
Zrn,

consider first a given residue class e Z. and any integer a in [a],. To

this element a e Z we associate its coset [a] E Zm and define
= [ajJ[eJ e B.*.

(1) We first observe that the image (p,([aJ,) does not depend upon
the choice of the representative a e [a],. For any other representative
a' e [a],, we have a' = a+km, for some k e Z. Hence
[a'] [e,] = [a + kin,] [e,] = [a] [e,] + [k] [m, g, h,]
[a] [es] + [k] [inh,] = [a] [e,] + [k] [0]
= [a][e,] =
since e1 g,h, (modm) and gm, = m. Thus the map 'p, is well-defined on
the residue classes [a],.
(ii) Next, ça, is a homomorphism of Z, into Zm. Let a,b be
representatives of [a],. fb],. respectively. Then
[a],+ [b], = [a+b],
I
sum in Z,
I
sum in Z

and (p,([a],+[b],) = [a+b][e,] the product of classes in Z5,

= ([a]+[b])[ej
= [a] [e,] + [b] [e,]
= (p,([a1,) + (p,([bi,).
Furthermore, [a],[b], = [ab],
I
product in Z,
I
product in Z.

and (p1([a],[b,]) = [ab][e,] the product of classes in Zm

=
= [a] [e,] {b] [e,J since [e,] is an idempotent
=
80 Introduction to Ring Theory chapter 3

(iii) The homomorphism q, is surjective. Consider [x] E pick

XE [x], and take the residue class modulo of x, i.e., [x]1. By the
definition ofq',. = [x]; thus is surjective.
(iv) Finally, to show that p. is an isomorphism, it remains only
to verify that kerq, in Z. is Suppose that q,([x]1) = [0]. Then any
x' e satisfies [0] = [x'] [eJ = [X'91 h1] in Zm. Hence ,n x'g1 h. and
m1 x'h4 since = From

= 9i h1 + ... + g h, + ... +
we note that (m,,h1)= I. and therefore ,n1Ix'; that is. [x']1=[0]1, as was
to be shown.

NOTE. Since [e,] is an idempotent ifl Zm,

= {e1][e,] = [e,].
Thus, the identity of is mapped by not to the identity of Zm, but to the identity
of the subring Z,,.

We now show that Zm is isomorphic to an external direct sum. As a

first step we define a mapping

by selecting a representative a of [a] and setting

A([aJ) = ([a]1, ...,[a]1 [a]5) e Z1 -j-

The element A ([a]) does not depend on the choice of the representative a.
For arbitrary a' E [a], we have a' = a + qm with q E Z and hence
[a'], = [a+qm], = [a]1 + [qin]1 = [a]1,
since m 0 (mod in1), I I ii.
Proposition 4. The map A defined above is an isomorphism:

Proof For b E [h],

[a+h]5)
'F

Sum in Z_
I
sum in Z

= ([a]1+[b] .., [aJ5+[h]5)

= ([a]1,.... + ([h]1 [h]5)

I
sum in R

= A([a]) + A([b]) by the definition of A.

Similarly, A([a][h]) = A([a]).A({h]), and so A is a homomorphism.
 Direct Sum Decompositions of Z,, 81

Furthermore, A is injective. Suppose there are given residue classes

[a].[h]eZm such that A([a])=A([b]). Then [a],=[bJ,, or
(mod ,,,,). for all i. I i n. Consequently a—b 0 (modnz) by Proposition
4, §2.6. or [a] = [I,].
Lastly. A is surjective. Given ([a1]1 [a,,],,) e R. let a be a solution
of the n congruences x a- (modm1), I i n [see §2.11). The element
a e Z satisfies [a], = [a1]1. and so A([a]) ([a1]1, ..., [aj,,). Consequently
A is an isomorphism of Z,,, onto the external direct sum R as asserted.

Proposition 4 yields a proof of the formula for the Euler q-function.

The function is defined to be the number of positive integers less than m
that are relatively prime to m, or equivalently, the number of units in the
residue class ring Zm. In particular, as stated in §2.12, for

= fl n,1, where ni, =

and where the p, are distinct positive primes, the Euler is given by
n /
q(rn) = p(m1) = fj ——
I

For the proof we use the fact (proven below) that there is a one-one
correspondence between the units U,, of Z,,, and the cartesian product
U1 x .. U,, of units U. in the rings Z,. I I ii. [See §2.12.] Since U, has
q (ni,) elements, Urn has q (m) = fl7 q (m1) elements. Writing m, =
recall from §2.12 that (p(m1) — Hence we have the formula

(,o(rn) =

In establishing the one-one correspondence we emphasize on the

one hand the principle of kcali:aiion (passing to the examination of powers
of primes), and on the other the Chinese Remainder Theorem, as the means
to pass from "local" results (the structure of the units U1 c Z.) to "global"
results (the structure of the units Urn c Zm). For [a] e Urn, we know that
(a,m)= I, and "by localization" that (a,n11)= 1 since ,n,Ini. Hence
[a],E U.. I and so
A([a]) = ([a]1 [a],,) e x x U,,.
Conversely, each set ([a1] i,.. , [a,,],,) of n units [a1]1 Z,. I i n,
determines a unique unit a E Zm. The system of congruences

x a1 (mod m1)
has a unique solution a modulo in by the Chinese Remainder Theorem
 lntroductioa to Ring Theory chapter 3

(since the integers are pairwise relatively prime). Further (a, in) = 1, since
(a1,in1) = I and a1 imply that (a,i;:1) = I. We conclude then that
a E Urn. and note that
= [a],,) = ([a1] [a,,],,).
Thus maps Urn onto the cartesian product of units U1 x x U,,, which
completes the proof.

Exercises

1. Express Z,. as the direct sum of ideals B7 for

a. in = 36 b. in = 35 c. in 30
d. in = 120 e. ni = 28 f. in = 15.
2. Show that the mapping Zm Z,,, given by co([a]) = [a] [e1] + [a] [e2),
where in in1 2) and [e1], [e2J, ..., [e,,] are as defined in this
in.. (n
section, is an endomorphism. What is the multiplicative identity of im 9'?
3. Describe explicit isomorphisms which show the following:
a. b.
d.
4. Prove directly that Z5 is isomorphic to a quotient ring of Z20. [See the
example at the end of §3.4 and Exercise 3, §3.2.]
5. Following the notational convention of this section, let
Z1 .•• + Z,,. For each I, I s I ii, prove that Z, is isomorphic to a
residue class ring of Zm.
6. Prove the properties of ar cited in Proposition 1.
7. To generalize the discussion of the direct sum decomposition of Zm, let
R be a commutative ring containing ideals A1, ...,A,, such that for all
I I 1,] ii, the ideal (A1, generated by A, is equal o R.
a. Describe R1(A1 A..) as an internal direct sum.
b. Prove that R/(A1 A2 - A,,) is isomorphic to the (external) direct sum
R/A1 + R/A2 -i- 4- RIA,,.
c. Note that (A1 A,,) = R and prove that for any choice of k1 E N,
I I n,
(A1k A,,k.) = R
where A1k, is the k1th of the ideal A1.
8. Consider a (commutative) ring R in which there are idempotents
e1 e,,(forsomen> 1) such that
(i) e= e1 +e,,, e the identity element in R:
(ii) 0 for i i,j ii.
1

a. Prove that (here exist ideals A1 A,, in R such that

b. Probe that ((0), for I I i,j <ii.

c. Prove that A1 A1 and that the ideal A1 is a subring, 1 <1 ii.
§3.6 Integral Domains and Fields 83

9. With reference to Exercise 4(b), §3.1, prove that if a ring R contains a

nontrivial idempotent, then it contains two ideals A,,A2 such that
R = A1 ® A2.
10. Suppose that a commutative ring R can be written as the sum of ideals

Prove that this sum is direct if and only if for all i, 2 i n,

A, (A1+•••+A,,.1) = (0).
(This proof of directness is used later in a similar argument for vector
spaces and groups
11. a. Consider a divisor dofni. Prove that there is a unique ideal A Zm
which has d elements.
b. Prove that A has p(n) distinct elements which generate it, where
nd = in.
c. Conclude that the Euler satisfies
= in,
dIm

where the sum is taken over all positive divisors d of in.

The argument in Exercise 7(a) and (b) is used again in our discussion of cyclic
groups [ci. Propositions 2 and 3, §6.61.

12. Prove the statement in Exercise 11(c) by induction on the number of

distinct prime factors of,,,, i.e., by induction on r, where rn = p,at ...
p, for
a. For the case r = I, show that

qi(rn) = q(l) + + + +
b. For the inductive step, consider in = p°,,, (n, p) = I, and show that
+
dim din din din din

the sums being taken over all divisors 1 of in and n, as indicated.

13. Prove Proposition 2 of §3.3.

§3.6 Integral Domains and Fields

In §2.1 we listed algebraic properties of the integers. All these
properties, except the cancellation law for multiplication, were included in
defining the concept of a commutative ring in §3. I. A set of elements which
has al/the algebraic properties of the integers is called an integral domain.
An integral domain D is a commutative ring (with at least two ele-
ments) in which the cancellation law that is, if ac=bc for a,hE D
and nonzero c D, then a = b.
84 Introduction to Ring Theory chapter 3

It is customary to require that an integral domain have at least two elements,

as in subsequent arguments it is awkward to consider the zero ring R = as an
integral domain. This requirement is equivalent to stipulating that the multiplicative
and additive identities are distinct.

In a ring the validity of the multiplicative cancellation law and the

absence of zero divisors are equivalent properties, an equivalence we shall
use often later on. We have already proved this equivalence for the ring of
integers [Property 10, §2.1]; the proof for general rings is similar. If a is a
proper (i.e., nonzero) divisor of zero, there is a nonzero element c D such
that ac=0=0c. Then by the cancellation law a=0, a contradiction.
Conversely, letting ac = be with e 0, we obtain (a — b) c = 0. If there are
no proper divisors of zero, then necessarily a—b = 0.

Examples of Integral Domains

I. The integers Z.
2. The rational, real, and complex numbers.
3. The residue class ring Z,,, p a prime integer [see §2.10].
4. The so-called gaussian integers

Z[i] = :a,bEZ},
named for Carl Friedrich Gauss. (Test what conditions on the coefficients
aj,bj,j = 1,2, are required for the product
(a1+b1
of two gaussian integers to be zero.)
5. More generally, systems of real (or complex if rn < 0) numbers of the form

6.
7. The set of all polynomials p(x) with integral (also rational, real, or complex)
coefficients. (Polynomials will be formally introduced in §5.1.)

Since the axiomatic properties ofan integral domain are an abstraction

of the algebraic properties of integers, Properties I through 10 in §2.1
generalize to integral domains, as do their proofs [cf. §3.1].
Note that nonzero elements a in an integral domain do not neces-
sarily have multiplicative inverses in the domain. For example, in Z no
integer times 2 is equal to 1.

Also, in Z { [2] there is no element u + z' with u, v e Z such that

(3+ [2)(u+v [2) = 1. We have, for the real number (3-1-
— I
=317—(1/7)JT.
§3.6 Integral Domains and Fields 85

That is, the real number (3 + belongs to Q( but not to the given integral
domain with coefficients a,b in Z.
An alternate proof follows from the fact that (3+ =
implies
3,, + 2v = 1 and 3v + ii = 0.

This pair of linear equations has a unique rational, but no integral, solution.

An integral domain F, containing at least two elements, in which

every nonzero element (every element different from the additive identity)
has a multiplicative inverse, is called a field.
The integral domains in Examples 2, 3, and 6 above are fields. The
trivial ring (0) is not considered a field.
A field can also be described as a commutative ring F(with at least
two elements) in which each nonzero element has a multiplicative inverse.
The equivalence of the two definitions is immediate since an element r e F
having a multiplicative inverse r cannot be a zero divisor. If r e F were
to be a divisor of zero, there would exist a nonzero element r' E F such that
rr' = 0.

Multiplying by we reach the contradiction r' = 0. Therefore, as asserted,

r is not a zero divisor.

Proposition 1. A finite integral domain D is necessarily a field.

Proof Let a,, ..., be the distinct elements of D. For any a 0 the n
products are distinct, because if
a= the cancellation law. Since I D is one of the products aak,
the element ak is the multiplicative inverse of a. Thus, each nonzero element
has a multiplicative inverse.
It is known—from experience—that every nonzero integer has an
inverse in the ring (field) of rational numbers Q. We now show that for every
integral domain D we can construct a field Q(D), called the fie'd of quotients
of D, that contains an isomorphic image of D. [See §3.2 for the definition
of ring isomorphism.] The proof of the existence of the quotient field Q(D)
involves ordered pairs (a, b) of elements of D, where b = 0. Such pairs are
simply an abstraction of the rationals, written a/b, for integers a and b 0.

Theorem. For a given integral domain D there exists a field Q(D), and an
embedding 1: .D—4Q(D).
Prcof. (I) Consider the totality of pairs of elements in D, such as (a, b),
(a*,b*), (e,d), (c.*,d*), in which the second components h,b*,d,d*, ..., are
dqfferentfron, zero. In this set of pairs introduce the relation
(a,h) (a*,b*) if ab* = ba* in D
 Introduction to Ring Theory chapter 3

an equivalence relation in the sense .2. Specifically,

(a. b) (a, b) because ab = ha,

(a,b) (c,d) (c,d) (a,b)

because ad = be cb = do,
(a,b) (c,d) and (c,d) (e,f) (a,b) (e,f)
because ad = be and ef = de imply
adf= bcf= bde and af= be.
Since is an equivalence relation, each pair (a,b) belongs to one
and only one equivalence class [a,b]. Let Q(D) denote this set of equivalence
classes of pairs of elements in D.
(ii) In Q(D) we introduce addition and multiplication as follows:
[a,b] + [e,d:1 = [a*d*+c*b*,b*d*]
[a, b] [c, d] = [a*c*, b*d*]
where (0*, b*) E [a, b] and (c*, d*) e [c, d]. These definitions are a direct
carry-over of the rules for sum and product of fractions a/b E Q.
It appears at first glance that these sums and products depend on
the choice of the representatives (a*, b*) and (c*, d*) of the classes [a, bJ
and [c,d], i.e., that they may not be single-valued functions of the classes.
Thus it must be proved that these sums and products are independent of the
choice of representatives.
To show that the sum is well-defined we verify for
and (c, d) (c*, d*) that
(ad4 be,bd) (a*d*+b*e*,b*d*),
and for the product that
(ac,bd) (a*c*,b*d*).
In other words we must verify that
(ad+b)b*d* = (a*d*+b*e*)bd,
(ac)(b*d*) = (bd)(a*c*),
using ab* = a*b and cd* = c*d.
These and subsequent verifications are carried ou, in the integral
domain D. We find
(ad+bc)b*d* = adb*d* + bcb*d* = (a*d*+b*c*)bd,
and (ac)(b*d*) = a*bcd* = (bd)(a*c*)
by the distributive, associative, and commutative laws in D. Thus the defi-
nitions of addition and multiplication of equivalence classes in terms of class
representatives are indeed independent of the choice of representatives.
§3.6 integral Domains and Fields 87

(iii) The associative, commutative, and distributive laws are easily

verified for the sum and product of the elements of Q(D), i.e., the equivalence
classes of pairs of elements in D. The element z = [0, 1] satisfies z + [a, bJ =
[a,bJ (z is then the additive identity of Q(D)), and e = [I, I] satisfies
e[a,b] = [a,b] for all [a,b] e Q(D) (e is then the multiplicative identity of
Q(D)). Each element [a,h] E Q(D) has an additive inverse [—a,b]. This
completes the verification of the additive properties of a field for Q(D).
For [a,b] :—that is, 0—we have [b,a] e Q(D), and therefore
[a,bJ{b,aJ = [ab,ah] = [1, 1] = e. Thus Q(D) is a field. In particular the
elements [a, I], a 0, in D have the multiplicative inverses [l,aJ in Q(D).
(iv) Next we define a mapping 1.: D—. Q(D) by 2(a) = [a, 1], which
we show to be an injective homomorphism. First, for all a, b e
1(a+b) = [a+b, 1] = [a, 1] + [b, I] 2(a) + 2(b)
and 1(ah) = [ab,l] = [a, lJ[h, I] = 2(a)2(b).
Second, 2(a) = [a, I] = [0, I] =: implies that a is zero. Hence A is injective.
Finally, as an immediate consequence of its definition, A is an
embedding since 2(1) = [I. I] = the multiplicative identity in Q(D).

should observe that constant use was made of the assumption

We
that D is an integral domain, specifically that the "denominators" bd and
b*d* of the sums and products are different from 0 in D. It is customary to
identify (i.e., equate) the images 2(a) = [a, I] of the elements a in 0 with
the elements in D. With this understanding we say that D is embedded in
its field of quotients Q(0). In particular Q(Z) is identified with the field Q
of all "fractions" of elementary (down-to-earth) algebra.
As with subrings particular subsets of a field are called
subfields. A nonempty subset K of a field F is called a subfield of F, if
(i) K is closed under addition and additive inverse, i.e., for all
a,b e K, we have a+b and —a E K.
(ii) K is closed under multiplication and multiplicative inverse (for
all nonzero elements), i.e.. for all a, b E K, the product ab E K,
and for a 0. the inverse a -' of a in F belongs to K.
Alternatively, a subfield of a field Fis a subring that Contains the multiplicative
inverse of each of its elements different from the additive identity. Again,
{0} is not considered a subfield. Of course, a subfield F' of a field F is itself
a field.

The rational numbers Q are a subfield of the real numbers R. Both Q

and R are subfields of the complex numbers C.

We now turn to what amounts to a proof of the uniqueness of the

field of quotients Q(D) of a given integral domain 0.
 Introduction to Ring Theory chapter 3

Proposition 2. If D is an integral domain contained in a field F. then F

contains an isomorphic image of the field of quotients Q(D).
Proof. First observe that the multiplicative identities of Fand D are equal,
as F has no zero divisors. Then let a/b denote the quotient of elements
a,beD, in F). and define a mapping

ço([a,b]) =

from Q(D) into F. Using the definitions of sum and product in Q(D) and the
rules for sums and products of elements a/b and c/din F, we note immediately
that q is a single-valued homomorphism on [a, b]. Observe here that if
(a*,b*) E [a,b], then a*/b* = a/b in F. Furthermore,

(p([a,b] + [e,d]) = +

since [a, b] + [e, d] = [ad+ bc, bd] has the image (ad+ bc)/(bd) equal to
a/b + c/d in F. Also

p([a,b][e,d])
=

and ([a, b] = if [a, b] z.

Finally, ço is a one-one mapping because p([a,bJ)=ço([c,d])

means a/b = c/d, and thus ad= be or [a,bJ = [c,d]. Therefore the set of
images 4(Q(D)) is a subfield K of F. This subfield K is contained in every
subfield L of F containing D; it is uniquely determined by D as the intersect ion
of all subfields L of F containing D.

To conclude this section we consider the prime field of a field F.

The properties of this smallest subfield contained in a field Frefiect significant
algebraic and arithmetic distinctions between fields, as we shall note in later
sections. The prime field P of a field F is defined to be the intersection of all
subfields L of F. Thus P is a subfield of F which contains no subfields.

Proposition 3.The prime field P of F is isomorphic either to the field of

rational numbers Q or to the residue class field Z,,, where the modulus p is
uniquely determined by F.
Said another way, any field F has a subfield isomorphic to Q or to
Z,,. In the first case the field Fis said to have characteristic 0 (or co) denoted
charF= 0; in the second case F is said to have characteristic p. denoted
charF=p.
Proof. Let e denote the multiplicative identity of F and A the additive
subset of F generated bye; that is,
A = (0) u {(—n).e:neN),
§3.6 Integral Domains and Fields 89

where n times, and (—n).e=n'(—e). The associative law

of addition in F implies that (r.e)+(s.e) = (r+s).e; and the distributive
law for multiplication in F implies that (r. e)(s. e) = (rs) e, for any r, s e Z.
Consequently A is an integral domain, since it is a subset of the field F and
hence has no zero divisors. We show next that A is isomorphic either to Z
or to Z,, for some prime p.
Define a mapping it: Z— F by setting it(l) = e, and more generally
ir(z) = ze a homomorphism from Z onto the subring
A of F follows immediately from its definition and the description of the
elements in A. The ideal
kent = {zeZ: ,r(z)=z.e=O€A}
= (m) for some ni e Z,
as every ideal in Z is principal. We now show that the integer in can only be
either 0 or a prime number p. (The ideal (ni) = (1) = Z is excluded because
ir(l) = e, and hence I kent.)
Case I. If (rn) = 0, then Z is isomorphic to A. Hence Q(A) Q, and the
prime field P of F is isomorphic to Q. In this case the characteristic of F
(and of the subfield Q(A)) is zero.
Case 2. if (m) 0, then m must be a prime p. for otherwise m = m1 m2
where 1 <m1, m2 <ni implies that ir(rn) = ir(m,)ir(m2) = 0. But ir(m1)
and ,t(m2) are distinct from 0 since ni1,m2 do not belong to (m)=kerir.
Thus ir(m,),ir(m2) would be zero divisors, but a field has no nontrivial
zero divisors.
Now define a mapping p: A by p ([a]) = a represents
the equivalence class [a]. Since p.e = 0, we have (a+rp).e = for all
r Z; hence the definition of p is independent of the choice of a e [a]. That p
is indeed an isomorphism follows immediately from its definition. In this
case, since Z,, is a field, A is itself a field, the prime field of F. Hence
char F= p.

NOTE. In a field ofcharacteristicp the mapping a = a monomorphism

of the field into (not necessarily onto) itself. We have

= = + (u').(aPlb) +

+(").(abPI)+bP
= + b" = cop(a) +
since p divides the binomial coefficients (i), 1 i <p. For the product, the
equation = holds trivially. The kernel of being an ideal in the
field F, is zero, for = 0 fora 0, then qp(e) = ')= 1) = 0,
a contradiction.
 Introduction to Ring Theory chapter 3

If, in particular, F is a finite field, then {a" : a e F} = F, and is called

the Frobeniu9 autoinorphism of F. Furthermore, if F= Z,,, then [a]" = (a] for all
cosets [a] of Z9 [Cf. §2.12]. Finite fields will be considered in more detail in §8.2.

Exercises

1. Let a, b, c, d be elements of an integral domain 0. Giving reasons for each

step of your argument, verify the equalities
(a—b) — (c—d) (a+d) — (b+c),
(a—b)(c—d) (ac+bd) — (ad+bc),
where a—I, is defined o be the sum a+(—b) and —b is the unique
solution of the equation x + b 0.
2. a. In an integral domain 0, prove that if au b" and a' = where
(u,v) = 1, then a = b.
b. Construct an explicit example for some elements a,b in some
integral domain D to show that part (a) is false when (u, v) I.
3. Prove that R = {a + b .J7: a, b e Z} is an integral domain.
4. Find the field of quotients of the integral domain R of Exercise 3. Show
that it is isomorphic to the field = x,yeQ}.
5. Find all units in the rings:
a. b.
c. d.
e. Q[.f:Th], m 1,3 and not a square
f.
6. Prove that the ring has infinitely many units. (In fact, the group
of units of this ring is isomorphic to the (external) direct sum of the
additive groups Z2 and Z. See and 6.10.)
7. Let Fbe a finite field of characteristic p. Prove that the mapping given by
= is an isomorphism of F onto itself. For which x e F does
pp(x) = x? Do these elements form a subfield of F? Why?
8. In §3.1 we did not require that the multiplicative identity e' of a subring
R' in a ring R be the same as the multiplicative identity e of R. Prove that
if R is an integral domain then e = e'.
9. Let R—' S be a ring homomorphism such that {Os}. IfS is an
integral domain, prove that 4'(eR) = es, where eM,es are the multi-
plicative identities of R and S, respectively.
10. Let R be a commutative ring in which the only ideals are (0) and R.
Prove that R must be a field.

In §3.2, we defined for an ideal A in a commutative ring R the residue class

ring (also called quotient or difference ring) R/A of equivalence classes of
elements of R modulo A. Exercises 11-17 develop related properties of the
ideal A and of the quotient ring RIA. Note carefully thc distinction between a
quotient ring and the ring (field) of quotients.
§3.6 integral Domains and Fields 91

II. Prove that R/A is an integral domain if and only if for any a, b e R, for
which ab E A, either a A or b e A. (Such an ideal A is called a prime
ideal.)
12. If R is the ring of integers and A = (m), mE Z, for what in is Z/(m) = Z,
an integral domain?
13. Prove that R/A has only nilpotent zero divisors if and only if for any
a,be R, for which abe A, either a e A or E A, for some n EN. (Such
an ideal A is called a primary ideal.)
14. As in Exercise 12, for what in does 4, have only nonzero nilpotent
elements for zero divisors? In other words, describe all primary ideals
in Z. [cf. Proposition 3, §2.10].
15. Prove that R/A is a field if and only if no proper ideal of R properly
contains A; that is, any ideal A' for which A c A' R must be R itself.
(Such an ideal A is called a maximal idea!.)
16. Find all maximal ideals in Z.
17. Conclude from Exercises II and 15 that in any ring R every maximal
ideal is prime, and from Exercise 16 that in Z every prime ideal is maximal.
18. Let be a ring homomorphism from a ring R into a field F. Prove that
is a prime ideal in R.
19. For a given maximal ideal M in a ring R, prove that the residue class
rings ii c N, each contain a unique maximal ideal.
20. Let F be the ring of real-valued continuous functions defined on the
interval [— 1, 1] C R [cf. Exercise 8, §3.2). Prove that for a fixed x0 in
this interval the set .5' (fe F :f(x0) = is a maximal ideal.
21. a. In §3.3 we defined the concept of the (external) direct sum R of rings
R1 Observe that the Cartesian product

R2 = R + R = {(x, y) x, y C R},
is a ring, but not a field.
b. Prove by examples that the (external) direct sum of integral domains
is not an integral domain.
22. a. Prove that R/A has no nonzero nilpotent elements if and only if
A = Rad A. (Such an ideal A is called a semiprime ideal.)
b. Describe all semiprime ideals in Z.
c. Prove that a prime ideal in a ring is also a semiprime ideal.
23. a. Using the notation in the construction of the field of rational numbers
Q as the quotient field of the integral domain Z, define [a,b) > 0
if a'b' > 0, where (a', b') represents the class [a, bi. Prove that an
ordering [see §2.2] of Q is defined and that the customary rules for
the absolute value hold.
Prove that this ordering of Q extends the customary ordering of
b.
Z(b > and that the Archimedian Principle
holds in Q.
24. Prove that any ring isomorphism Z Z must be the identity map.
25. Repeat Exercise 24 for Q —' Q. However, exhibit an isomorphism of
the complex numbers C, which is not the identity map.
92 Introduction to Ring Theory chapter 3

26. Determine the smallest integral domain containing 1/3 and — 1/21 in
the field Q.
27. For a prime number p. define
= (a/b eQ a,be Z, (a,b)= I, (b,p) = l}.
Prove the following statements.
a. is a ring.
b. Every ideal A c is principal and is equal to
= :c
where a = v9(x), as defined in Exercise 14, §2.7.
c. The residue class ring is isomorphic to Z,,.
d. fl,, R,, = Z. where the intersection is taken over all prime integers p.
28. With reference to Exercise 27, prove that
R let N denote the ideal of nilpotent elements
[cf. Exercise 7, §3.1]. Prove that the residue class ring RIN has no nonzero
nilpotent elements.
30. Find a field F and two nonzero elements x, y e F such that x2 +y2 = 0.
31. Let be a (ring) homomorphism of a field F to a ring R such that
0. Prove that ker and hence that isa monomorphism of F
into (not necessarily onto) R.
32. Consider N with the customary properties of addition and ordering.
a. Construct the integral domain of integers analogous to the con-
struction of the field of rational numbers Q from Z. Thus let [a,b]
be the set of all pairs of numbers u,i' in N such that a+v = u+b
for a,b in N.
b. Show that an equivalence relation is defined. Then define sum and
product for these equivalence classes. What role does the class [1, 1]
play?
c. What is the multiplicative unit in Z thus constructed?
d. Is the map a—p [a+ 1,1] an embedding of N into Z?
33. Consider N with the customary properties of addition and ordering.
Prove that the ordering for N can be extended to an ordering of Z,
constructed as in Exercise 32, by defining (a,bJ > [1, I] to mean a > b.
34. Associate to each element a e N a symbol — a; also introduce a symbol
0, defining 0+x = x+0 x for all xc S.
a. Fora,beN,define
(I) a+(—b)=(--b)+a = c, ifa b+e with eeN.
(ii) (—a)+b= b+(—a)= —c€ 5, if 0;
and (—a)+b b+(—a) = 0, ifa = b.
(iii) (—a)+(—b) = —(a+b).
Verify that the set S has all the customary properties of addition; in
particular, that there exists in San element t such that s+ t = t+s = 0
for every element s e S.
b. Next define a product in S as follows:
(iv) Os=sO=Oforallse5.
(v) (—a)b = b(—a) = —(at,) and (—a)(—b) = ab for all a,b eN.
Now prove that S is an integral domain and that S is isomorphic as a
ring to the ring Z, constructed in Exercise 32.
§3.6 Integral Domains and Fields 93

35. An integral domain D is said to be well-ordered if there is an ordering

relation < on D such that every nonempty set of positive (>0) elements
in D contains a least element. Prove that any well-ordered integral
domain D is isomorphic to the ring of integers (see §2.2].
36. II F is a field, prove that the field of quotients Q(F) is isomorphic to F.
37. In an integral domain D, consider a subset S such that 0 S and
imply that ss' E S. Let
s, s' E S be the subset of the field of quotients
F= Q(S) that consists of the elements a/s. ae D and sc S. Prove the
following statements.
a. If A is an ideal in D, then the set
A•D5 = {a/s:anA,sES}
is an ideal in
b. For ideals A, B in 0,
(A+B)D5 = +
(A (BD5).
c. 1ff is any ideal in then D).D5 = J.
d. For any idealA 0, we have A
cD 0 F consists of all
elements a e A for which there exists an s S such that sa e A.
38. With reference to Exercise 37, let .4' and B' be ideals in D5. Prove the
following statements.
a.

(A' D)
e. 0) = (RadA') D.
39. Let P be a prime ideal in a ring Rand let
Prove that the extended ideal PR,. is a maximal ideal in R,. and that all
nonunits of R,. form an ideal. Furthermore, prove that PR,. contains all
proper ideals of R,..
40. With the notation of Exercise 39, prove that the extended ideal
AR,. R,. if and only if A c P.
41. Let 0 be the set of all functions from N to Z, and forf,g €D define

(f+g)(n) =f(n)+g(n)

(f*g)(n) =
din dj
thesum being taken over all divisors d of n. Prove that /) with these
laws of composition is an integral domain.
 4

Aspects of Linear Algebra

In this chapter we pause in our development of abstract algebra to

consider the linear algebra topics of vector space, linear independence, basis,
matrix, and determinant. The material here exemplifies some of the algebraic
concepts already encountered and is necessary for subsequent considerations
of polynomials, groups, and field extensions. Linear algebra is often presented
as a subject separate from abstract algebra, but our intent is to indicate their
interrelations. This chapter is by no means a complete presentation of linear
algebra.
Building on our study of the arithmetic properties of integers I

and then of fields we define an algebraic structure called a vector space.

This concept generalizes to that of a module in §7.7, and some of the aspects
of vector spaces will carry over directly to modules. Vector spaces are used
in §8.1 in the treatment of field extensions leading to the Galois theory.
Then in §9.4, in preparation for the results of Hilbert and Noether on certain
field extensions, we need to consider linear transformations, their represen-
tation by matrices, and determinants.
Some students will already have studied vector spaces and matrices
and can proceed directly to Chapter 5. Others may wish to peruse this chapter,
reviewing the principal statements, and then subsequently use these sections
for reference. However this chapter is intended primarily for those students
with no previous course in linear algebra. While not a substitute for a full
course in linear algebra, it offers a comprehensive introduction.

94
 Vector Spaces 95

§4.1 Vector Spaces

From analytic geometry and calculus we recall the concept of
"vector," usually considered then as a pair of real numbers (a1, a2) or as a
triple (a1,a2,a3). In either case a vector was geometrically described as a
line segment with both direction and magnitude. By geometric argument,
usually by means of a parallelogram of forces, the following algebraic
properties of vectors A = (a1,a2,a3) and B = (b1,b2,b3) were demonstrated:
A ± B = (a1±h1,a2±b2,a3±b3),
and for r e R,

r.A = rA = (ra1,ra2,ra3).
Hence 0. A = (0,0,0), the zero vector, often denoted simply by 0, and
I = A.
But the importance of vectors is not limited to two- or three-
dimensional euclidean space. In fact their value is that they provide a con-
venient means of expressing many interacting aspects of problems in fields
such as physics, linear programming, statistics, and other areas of math-
ematical applications, as well as in advanced algebra. We begin with an axio-
matic description of a vector space together with its associated field of scalars.
The elements of a vector space are called vectors.
A vector space (or linear space) V over a field F is a set of elements
together with two laws of composition (commonly called vector addition and
multiplication by scalars). The law of vector addition associates to each
ordered pair (A, B) in the cartesian product V x V a unique element
A + B V, and the law of multiplication by scalars associates to each
ordered pair (r,A) in Fx V a unique element rA V, subject to the following
axiomatic properties:
Addition. Vector addition is associative and commutative: there exists
an additive identity element (denoted by 0): and each element A V
has an additive inverse B, such that A + B = 0. (This inverse B is
designated — A.)
Multiplication. Multiplication by scalars (that is, multiplication of
elements in V by those in F) is associative,
r(r'A) = (rr')A for A V and r,r' e F;
and if I is the multiplicative identity in F. then I A = A for all A e V.
Distributivity. Multiplication by scalars and vector addition satisfy
the two distributive laws for all A, Be Vand r,r' e F:
r(A+B) = rA + rB,
(r+r')A = rA + r'A.
Because the defining algebraic axioms of a vector space over a field
resemble those of the integers [*2.1] and of a ring [*3.1]. we can easily derive
the following logical consequences.
 Aspects of Linear Algebra chapter 4

Property 1. The additive identity (called the zero vector 0) is unique.

Property 2. Furthermore, A = A + B for some B e V implies B =0.
Property 3. The cancellation law for addition holds.
Property 4. The additive inverse — A of an element Ae V is unique.
Property 5. Every equation A + X = B. for A, B e V. has a unique solution
XEV.
Property 6. For any r F, r.O = 0.
Property 7. The identity for multiplication by scalars is unique. That is,
rA=A. forallAeV='r=leF.
In fact, the next property provides a stronger statement.
PropertyS.
Property 9. For any r E F. and any A e V, the additive inverse —(rA) of
rA is (—r)A.

Most commonly in geometry and analysis we consider vector spaces

V over the field of real numbers R, and also over C and Q. The examples
below are of vector spaces over these fields.

Example 1. Let V be the set of ordered n-tuples (a1 of real numbers c

I i n, where addition of vectors and multiplication by scalars are defined
component-wise:
(a1 + (b1 = (a1 +b1,
r(a,, = (ra1,
We should check that these laws of composition satisfy the axiomatic properties in
the definition of a vector space. (The proof is the same as for pairs of real numbers,
i.e., vectors in the euclidean or cartesian plane.) Thus, V is a vector space over R. It
is commonly denoted by

Example 2. To obtain a vector space over Q or C, Example 1 is modified to

require that the components a1, I I n, of the n-tuple (a1, belong to Q or
C, respectively.

Example 3. Let F denote the set of all real-valued functions defined on the real
number line R. That is, F is the set of all functions 1: R —* R. Define laws of com-
position as follows forJ,g c F, and r R, for all x- R:
J+g is the function which maps x toJ(x)+g(x);
rf is the function which maps x to rf(x).
Again, after verifying that these laws of composition satisfy the axiomatic properties
in the definition of a vector space, we conclude that F is a vector space over R.
§4.1 Vector Spaces 97

We shall encounter further examples of vector spaces later: vector

spaces of matrices in §4.4, vector spaces of polynomials in §5.1, and algebraic
field extensions in §8.1.
In speaking of a vector space V we must also have in mind the
associated field of scalars F. Thus a vector space involves two sets: the set of
vectors V and the set of scalars F. To make this association explicit we denote
a vector space V over a field F by V/F. We use
R of n-tuples of real numbers, and more generally
F of n-tuples of elements of F.
While we shall usually refer to the general case of an arbitrary field
of scalars F, it suffices for our present purposes to think of F as the field R.
A subset W of a vector space V/F is called a snbspace of V if
(i) W is closed under vector addition and additive inverse, as
determined in V; that is,
A+BEW
A,B E W
—A e W.
(Hence 0 e W.)
(ii) W is closed under multiplication by scalars; i.e.,
AeW,
Note that a subspace W of a vector space V/F is itself a vector space over F.

For example, to verify that

W = ((a, b, c) R3 : a = 3c} = ((3c, b, c) : b, CE R}
is a subspace of R3, we note that
(3c,b,c) + (3c',b',c') = (3c+3c',b+b',c+c') = (3(c+c'),b+b',e+c') W,
and —(3C,b,c) (—3c, —b, —e) = (3(—c), —b, —c) e W,
and finally for any e R
r(3c,b,c) = (r(3c),rb,rc) = (3(rc),rb,rC) W.

As a second example consider the set W of 3-tuples (x, y, z) E R3 whose

components satisfy the system of linear homogeneous equations

3x— y+ z=O,
x + 2y — 5z = 0.
The set W is a vector space over R, since if x, y, z and x', y', f are solutions, then
so are y+y', z+z', and rx,ry,rz, for all r ER.

By a linear combination of the elements of a finite subset S =

of a vector space V/F. we mean an element in Vof the form
 Aspects of Linear Algebra chapter 4

where r1, ..., F. Such a combination is called nontrivial if at least one of

the coefficients r1, I i n, is distinct from zero. The set W of all linear
combinations of elements in S with coefficients in F is a subspace of V,
called the subspace generated or spanned by S. We sometimes refer to W as
the span of S, denoted W = Span(S).

Proposition 1. For a finite set S = {A, A,j V/F, let

W
W is subspace of V that contains S.

By "smallest" we mean here that no proper subset of W both is a

subspace of V and contains S. The proof is left as an exercise.

As an example, consider the vector space R' of 4-tuples of real numbers

and the subset
S = ((0,0,0, l),(4,0, 1, 1),(i,0, 1,1)).
Then W= {a(0,0,0,l)+b(4,0,l,l)+c(I,0,1,1):a,b,ccR}
= ((4b+c,0,b+c,a+b+c):a,b,ceR}
= ((u,0,v,w):u,v,weR}. (Why?)
Similarly in R3, the span of S = ((1,0, 1), (0,0,3), (2,0, 5)) is
W= {a(l,0,l)+b(0,0,3)+c(2,0,5):a,b,cER}
= + 2c, 0, a + 3b + Sc) : a, b,c e R}
= {(u,0,v) : u,ve R}. (Why?)

Now given subspaces W1 and W2 of a vector space V/F, we can define

two additional subspaces:
the set-theoretic intersection W1 W2;
the sum W1 + W2, where
W1 + W2 = {AeV:A=A1+A2, A.E W,, i=l,2}.
The definitions of intersection and sum of two subspaces easily
extend to any finite number of subspaces Wi,...,Wm in V. A sum U=
W1+•+W,, of m2 subspaces W1cV, is called an (internal)
I

direct sum, denoted

U can be expressed uniquely as a sum u = + + Wm 0t'

elements w4 e W1.

NOTE. It is not required that the subspaces be distinct.

§4.1 Vector Spaces 99

Proposition 2. Consider subspaces W1, ..., W,,, of V. Then the sum U =

Wi+"+Wm is direct if and only if for all i,

Proof If for some I the intersection W1 (W1 + + contains a

nonzero vector u, then this vector u E W1 +•• + = U has two distinct
expressions,
u= + w2 + + +0+0+ +0
11
EW1 EW2
1

u=0 +0
I
WI

as the sum of elements of W,, I j m, so that U cannot be a direct sum.

Conversely, suppose that for all I, 2 i < m,
= {0}.
To prove that the sum IL! is direct consider a vector u E U which has two
distinct expressions as the sum of elements in I

UWI+"+Wm,
(*)

Then (w1 —W1) + + (WmW,,) = 0.

Let I be the greatest subscript for which ii's — 0. Since we assume the
expressions (*) are distinct, this subscript i is at least 2. Now,

—
• _i
—
• I
i
T Ijtj_
and — w1 is a nonzero vector in W1 + + - Hence

114€

But as this intersection is the 0 subspace. ;,

— = 0, a contradiction. Thus
since u cannot have two distinct expressions (*), the sum U is direct, as
asserted.

A similar argument proves the analogous statement for the direct

sum of rings and the direct product of groups
We say that a vector space V/F is finitely generated if there is some
finite set of vectors S= {A1, ...,A,,j such that every vector A e V can be
written in the form
A = r1A1+ +
that is, V= Span(S). The coefficients r1 rm E F are not required to be
unique.
100 Aspects of Linear Algebra chapter 4

For example. R2 is a finitely generated vector space since every vector

(a, b) a R2 can be written as a linear combination of the vectors ((0, l),(l, l),(1, 2)}:

(a,b) = (b—a)(0,l)+ a(l, 1) + O.(l,2).

Butalso,
Alternatively we can conclude that R2 is finitely generated by observing that every
(a,b) R2 can be written as a linear combination of the vectors ((0, I),(l, l)}.

Exercises

1. Verify that the axiomatic properties are satisfied for the sets in Examples
I and 3.
2. Prove Proposition 1.
3. In euclidean three-space R3, describe the subspaces containing the
following sets of vectors.
a. ((0,0, 1),(0, I, l),(0,4,2)}
b. ((I, I, l),(2, I, l)}

C. ((4,2, l)}
d. ((1,2, l),(l,3, l),(l,0,0)}
((1,2, I),(l,3, l),(1,0,O),(0,5,0)}.
e.
4. Verify, for subspaces W1, W2 of a vector space V, that W1 ri W2 and
W1 + W2 are again subspaces of V.
5. In R3 demonstrate that the set-theoretic union of subspaces need not be
a subspace.
6. Prove Property 7.
7. Prove that the set of triples (x, y, z) e R3, satisfying the homogeneous
system of linear equations

3x— y+ z=0
+ 2y — 5z = 0
is a vector space over R.
8. Determine whether the following subsets of R3 are subspaces.
a. W1
b. W2 = {(x, y,z): 3x—4y—5z 0)
c. W3={(6a—7b,c—4b,a+b—c):a,b,CER}
d. W4 = {(x,y,z):x,y€R}
e. = {(x,y,z):xy}.
9. Determine the intersection of the subspaces W1 and W2 in Exercise 8.
10. Prove that the span of a finite subset S= (A1,..., Am} contained in a
vector space V/F is the intersection of all subspaces U V which
contain S.
11. Consider subspaces W1,W2 V and vectors a,be V. Define u+W1 =
{a+ w: w W1}. Prove that a + W1 b± U'2 if and only if U'1 W2 and
a—be W2.
§4.2 Linear Independence and Bases 101

§4.2 Linear Independence and Bases

In defining a finitely generated vector space {at the end we
made no statement concerning the number of elements in a generating set.
In fact we showed by example that a given vector space could have several
finite generating sets with differing numbers of elements. By introducing the
fundamental concept of linear independence we are able to specify minimal
generating sets (called bases) for a given finitely generated vector space and
thus to define the dimension of such a vector space. Subsequently in our
studies of polynomials and field extensions we shall encounter the similar
concepts of algebraic independence and integral dependence.
A finite subset S = {A1, ..., Am} c V/F is said to be linearly dependent
if there exist m scalars r1 rn,, not all of which are zero, such that

If no such scalars exist we call the subset S linearly independent.

The terms "linear dependence" and "linear independence" may be
better understood when restated as in the following proposition.

Proposition 1. A finite set of vectors S= {A1, ...,Am}, in 2, is linearly

dependent if and only if at least one of them can be expressed as a linear
combination of the remaining ones.
In this phraseology a finite set of at least two vectors is linearly
independent if and only if none of the vectors can be written as a linear
combination of the others. Also note that {O} is a linearly dependent set,
while {A} is linearly independent if A 0.
Proof Suppose S is a linearly dependent set. By definition there exist
scalars r1, E F, not all zero, such that
riAi+...+rmAm=0.
For some subscript i, r. 0, and
—r1A1 = r1A1 + +r1_1A1_1 + +r,,A,,,.
Since — is a nonzero element of F, it has a multiplicative inverse, and
= + ... +s11A1_1 + +S,,,A,,,,

Conversely, if one of S. say A1. is a linear combination of the others,

i.e., if
A. = r1A1 + +r1_1A1_1 + +r,,,A,,,,

then
r1A1 0.
We thus have a nontrivial linear combination of the elements of S equal to
zero. Hence S is linearly dependent, which completes the proof.
 Aspects of Linear Algebra chapter 4

Three further remarks remain. First, any finite set of vectors con-
taining the zero vector is linearly dependent. Second, any finite set of vectors
is either linearly independent or linearly dependent. Third, the most common
means of proving the linear independence of a set of vectors {A1, ...,Am} c V
is to consider a general linear combination

and then to prove that each of the coefficients r1, ..., r,,, must be zero.

For example, to show that {(0, 0, 1), (0,2, I), (1, —3, 1)) is linearly inde-
pendent in R3, consider real numbers a,b,c, for which
a(0,0,l)+b(0,2,l)+c(1,—3,l) = (0,0,0).
This one vector equation translates into three scalar equations:

a+b+c=0, 2b—3c=0, c=0.

Since c = 0, so must b, and hence a. Therefore the vectors are linearly independent
as asserted—no nontrivial linear combination of them can equal the zero vector.
To show that the set of vectors ((0, l),(l, l),(l,2)} in R2 is linearly
dependent, we can determine real numbers a, b, c, not all zero, such that
a(0, I) + b(l, 1) + c(l,2) = (0,0).
That is, we find a, b, c, not all zero, satisfying
b+c=O, a+b+2c=0.
One solution is a = 3, b = 3. c —3. An alternative proof of dependence would be
to write one of the vectors as a linear combination of the two remaining ones. For
instance,

(1,2) = (0,1) + (1, I).

The concepts of linear independence and span of a set are combined

in defining a basis of a finitely generated vector space V/F. A finite subset S of
a nonzero space V/F is called a basis of V if
(i) S is a linearly independent set;
(ii) the span of S equals i.e., each v e V can be expressed as some
linear combination of elements in S.

The fundamental properties of a basis are stated in the following two

theorems.

Theorem 1. Every finitely generated vector space V/F has a basis.

Proof By hypothesis, V/F has a finite generating set S" = {A1 Am}. If
is a linearly independent set, then it is a basis, and nothing is to be proved.
If is a linearly dependent set, can write by Proposition I

A1 = + +s1_1A1_1 +
§4.2 Linear Independence and Bases 103

for some I, 1 i in, and some s, . ,...,SmE F. The span of

S' = (A , ..., A._ Am)

equals = V. If S' is a linearly independent set, it is a basis for V/F.

Otherwise repeat the above process, deleting a vector dependent upon
those remaining, and in at most rn— I steps obtain a linearly independent
set S whose span is V. Hence S is a basis, as desired.

For the second theorem we need the following lemma.

Steinitz Exchange Lemma. If vectors B, are linearly independent in

a vector space V/F generated by vectors A,, ..., Am, then ii m.
Proof. We have already noted that no linearly independent set can contain
the zero vector. Hence B. 0, 1 i n. Since the vectors A,, ..., Am generate
V, the vector B, can be written as a linear combination
B1 = r1 A, + ... + r,nAm.
Since B, 0, at least one of the coefficients r1, ..., r,, is nonzero, say r.1.
Then

= !(B, —r,A, A11_ A11÷

from which we conclude that the set

{Ai,...,Aii_i,Bi,Aii+i,...,Am}
generates V. Now write B2 as a linear combination of these vectors:
B2 = rA1+ ... + r11_1A,,_,+ r1+ A11÷,+ ... + + Sj B1.
A.t least one of the coefficients r is nonzero, as B,, B2 are linearly independent.
Continuing in this fashion we ultimately replace n vectors A.1 As,, by
the linearly independent vectors B,, ..., to obtain a new set
{B, L) [(A , Am}\(Aji A1,,}]

which generates V. Because the B. are linearly independent, at the jth step
the linear combination

where .•• = r'., = 0. has at least one nonzero coefficient r' of A1. We
conclude then that nm, as we cannot exhaust {A, Am} before com-
pleting the n exchanges.
The preceding result is due to Ernst Steinitz (187 1—1928). who was one of
the originators of what today is called "abstract algebra." His paper, A!gebraische
Theorie der Körper (Jour.f.d. r.u.a. Math., 137 (1910), 167—309), presents the founda-
tions of this branch of mathematics. Its introduction clearly demonstrates that
104 Aspects of Linear Algebra chapter 4

Steinitz was motivated by critical analysis and insight, and not by abstraction
per Se.

Theorem 2. Any bases of a finitely generated vector space V/F have the
same number of elements.
Proof. If and are two bases of V/F, then as both
generate V and are linearly independent, we apply the Steinitz Exchange
Lemma twice to obtain n in and in n.
Thus any finitely generated vector space V/F has a basis, and any
two bases have the same number of elements. In other words, the number of
elements in a basis for a given finitely generated vector space V/F is unique,
even if V has many bases. This unique number is called the dimension of V/F,
denoted dime V. or simply dim V. if the scalar field F is understood.
We conclude this section with a characteristic property of bases.

Proposition 2. If S= (A1 is a basis for V/F, then each vector in V

can be expressed unique!)' as a linear combination of elements in S.
Proof Consider an element v e V having two expressions as linear com-
binations of elements of S:
v = r1A1 + ... +

Subtracting, we obtain
(r1 —s1)A1 + ... + = 0.

Since the are linearly independent (by hypothesis), the coefficients

— = 0, I i n, and thus the two expressions (*) are not distinct.
As a corollary, we obtain the converse to Proposition 2. If S is a
spanning, linearly dependent set in V/F, then some v e V has at least two
distinct expressions as linear combinations of elements of S. [See the example
at the end of §4.1.1

Exercises

1. if W is a subspace of a finitely generated vector space, prove that W is

finitely generated (by elements in W).
2. If V is an n-dimensional vector space, prove that V can be written as the
direct sum of n one-dimensional subspaces.
3. II W is a subspace of a finite dimensional vector space V. prove that
dim U" dim V, where equality holds only if W V.
§4.3 Transformations of Vector Spaces 105

4. Consider a nontrivial subspace W of a finitely generated vector space V.

a. Prove there exists a subspace W' such that V equals the direct sum
w'.
b. Is the subspace W' unique? Why?
5. In R2, let 14" be the subspace generated by the vector (1,5). Find W' as
in Exercise 4.
6. Consider a finitely generated vector space V and a finite subset S.
a. if S is a linearly dependent subset of V and generates V. prove that
S contains a basis of V.
b. is a linearly independent subset of V. prove that there exists a
If S
basis S' of V such that S 5'.
c. If S is a basis for V. prove that no proper subset of S generates V.
and that no subset S' of V which properly contains S is linearly
independent.
7. Determine all of the linearly independent subsets of the following sets of
vectors.
a. RI, 1),(2, l),(5, l),(4,3)) c R2
b. ((I, I, l),(5,O, l),(4, — I, — l),(O,2, l)} R3.
8. Show that any vector space V/F has at least two distinct bases if
dim V 2, or if dim V = I and F has more than two elements.
9. State whether the following sets are linearly dependent or independent.
a. {(l,2),(l,3),(l,4)} C
b. {(i,2,l),(1,3,2),(l,4,3fl C R3
c. {(l,2, l),(l, —, 1),(3, —-5,3)}
d. ((5, —3, l),(—2,4,O),(l2, — IO,2fl R3.

10. The set of vectors ((1, — l,O),(2, 1, l),(O,3, l),(O, — 1,1)) spans R3. Find
a subset which is a basis.

11. Prove the following standard result concerning vector space dimensions.
If Sand Tare finite dimensional subspaces of a vector space V. prove that

dim(S+T)+dim(Sr'tT) = dimS+dimT.

§4.3 Transformations of Vector Spaces

Section 4.1 offered several examples of vector spaces, including R".
We now consider isomorphisms of vector spaces and prove that any
n-dimensional vector space over R is isomorphic to R". In fact, for any field
F, every n-dimensional vector space V/F is isomorphic to P.
More generally we consider mappings from one vector space U to
another V, over the same field F, which respect the linear space structure.
Such mappings are called linear transformations, although we might also
think of them as homomorphisms of vector spaces. Specifically, a linear
transformation (Or operator) is a mapping p: U V. such that, for all
4.8€ U and all e F.

(I) q(4+B) =
(ii) (p(rA) = rp(A).
 Aspects of Linear Algebra chapter 4

Vector spaces U/F and V/F are said to be isomorphic if there exists
a linear transformation q: U-+V, such that
(iii) 4(A) = 0 A = 0,
(iv) V={p(A):AEU}.
in other words, a vector space isomorphism is injective, surjective,
and compatible with the operations of vector addition and multiplication by
scalars [properties (i) and (ii) above].
The properties of linearity (I) and (ii) in the definition are most
significant in that, for a linear transformation q: U—' V, once the images
for a basis of U have been specified, the image
is determined for each A e U:
A€ A= r1A1 + +rmAm
with unique scalars r1 rm e F, and so
p(A) = r1(p(A1) + ... + rm(p(Am).

This concept is illustrated by the mapping ço: R2 —' R2, given by

= (3,4), q'((O, I)) = (—3,5). Hence
= + bq,((O, 1))
= a(3, 4) ÷ b( —3, 5) = (3 (a — b), 4a + 5b).

Theorem. if V/F is an a-dimensional vector space. then there exists at least

one isomorphism V—' F".
Proof By Theorem 1, §4.2, V/F has a basis {A1 Now for A e V
there exist unique scalars r1 such that
Define a mapping (p: V—p F" by setting
= = (r1
Certainly the mapping depends upon the choice of basis for V/F. But once
a basis has been chosen, the mapping is well-defined, because the expression
of vectors in V in terms of a given basis is unique. It remains to verify that
is an isomorphism.
Consider also B s1 A where s1, ...,; e F, and ce F.
Then
q'(A + B) = q'(r1A1+ + r,, A,, +s1A1+ + s,, A,,)
=
= (r,+s1
= fr1, ...,r,,) + s,,) = 4)(A) + 4)(B).
=
= q(cr141+
= (er1, .., CT,,)
= c(r =
p(A) = (p(r1A1+••+r,,A,) = (r1 ,..,r,,)
=0 r1 = = r,, = 0 A = 0.
 Transformations of Vector Spaces 107

For any (r1, e P, consider A = r1 A1 + Then (p(A) =

..., re), and hence q is surjective. This completes the proof that the
(r1,
mapping p, defined above, is an isomorphism.

Replacing F by R in the theorem yields the following corollary.

Corollary 1. Any n-dimensional vector space over R is isomorphic to

Corollary 2. Any two finite dimensional vector spaces over F are isomorphic
if and only if they have equal dimensions.

We conclude this section by considering the set of all linear

transformations from one vector space U/F to another V/F. This set, denoted
HomF(U, V), is itself a vector space over F with the following laws of com-
position. [As usual, two linear transformations are said to be equal if
4(A)= for all A eU.] For e HomF(U,V) and reF we define
the sun; ço + cu and the product by a sea/ar rp to be the mappings in HomF (U, V)
given for a/I A e U by
(p+ç&)(A) = q(A) + cuI(A),
(rq)(A) = r4(A).

We must verify first that and rp are indeed linear transform-

ations from U to V, and then that, with these laws of composition, the set
V) is a vector space over F. We verify that e
and leave the remaining proofs as exercises {cf. the axioms for a vector
space, §4.1]. Consider for A,BE U, reP,
(q+i/i)(A+B) = 4(A+B)+
= Q(A) + co(B) + 14i(A) +
= (p+i/í)(A) + (p+i/i)(B)
and
(p+I/i)(rA) = q(rA) + cu'(rA)
= rço(A) + rI/,(A)
=
hence q + e Home (U, V). (Justify each step of the preceding computations.)

Proposition. For vector spaces U/F and V/F of respective dimensions n and
in, V) has dimension nm.
Proof Let {A1, ..., and {B1 BJ be bases for U and V, respectively.
We prove the proposition by displaying a basis of nm elements for HomF(U, V).
Foreachi, I I define
U V
by =
pU(Ah) = 0, for h 1.
 Aspects of Linear Algebra chapter 4

First we verify that the mappings generate HomF(U, V). To this

end consider any linear transformation q: U— V. Since q' is determined by
its action on the chosen basis elements {A1 we can write for each i,
I i < n,
q(A1) = I j in.

But then = I i n, I I in.

as claimed, since for each basis vector I I n,

= = I i ii. I I in.
i,j j
To ascertain that the transformations are linearly independent,
consider a linear combination
= = 0, I I

Since is the zero linear transformation, we have = 0, for each basis

vector That is, for each k, I k a,
0= I I a, I m,

= 1 j 'H•

But the vectors forming a basis of V. are linearly independent. Therefore

any linear combination of them that equals zero must have zero coefficients;
that is, = 0, 1 j in, for each k. Thus the mappings 1 I n,
I j m, are linearly independent and so constitute a basis of rnn elements
of HomF(U, J').

When U = V we refer to V), also denoted EndF(V), as the

ring of endomorphisms of the vector space V/F. Here we can introduce a ring
structure because we can compose two mappings q., i/i E EndF(V), as well as
add them. The composition operation is defined by

o tfr)(A) = for all A e V.

It is left to the reader to verify that EndF(V) with the operations + and o
a ring (noncommutative, if dim V 2).

Exercises

I. If the vector space V/F has dimension greater than 1, prove there exist
at least two distinct isomorphisms V —' P.
2. Prose that for each nonzero e F, the mapping given by = cA is
an isomorphism of V/F with itself.
§4.3 Transformations of Vector Spaces

3. For n> I, prove that each of the following mappings of F" to itself is an

isomorphism.
a. r(a, a,,...,aj = (a,,...,aj,...,a,
(i.e., the ith and jth components are interchanged) for given
i,i, I
b. a,, ...,aj, = (a, a,, ...,aj+cag,
(i.e., c times the ith component is added to the jth) for c e F and
given i,j, I I.
C. pc(ai, ..., a,, ..., = (a,, Ca,, ...,

(i.e., the ith component is multiplied by c C F) for c 0 and given I.

4. Given vector spaces U/F, V/F, and W/F and linear transformations
(9: U—' Vand V—' W, prove that the composite ifr 04' defined by
0 4')(A) = for all A e U,
is a linear transformation from U to W.
5. Given a HomF(U, V) and r a F, prove that re,, defined by (r4')(A) =
r4'(A) for all A a U, belongs to HomF(U, V).
6. Verify that V) with the composition laws of vector addition
and multiplication by scalars, as given in this section, is a vector space
over F. That is, verify the following.
a. + = (L' + (0 b. r((0+ = rço +
c. +). = ço + d. (r+r')4' = rca + r'(0
e. r(r'ço) = f. =
7. Give an explicit basis for
a. HomR(R2,R3) b. Hom0(Q3,Q3).
8. Show by an example that the product law of composition of mappings
for is noncommutative when V = R2.
9. a. Prove that with sum and product as defined in the text, is
a (noncommutative) ring if dim V 2.
b. Describe the units in EndF(V).
10. If a linear transformation R2 —' R3 satisfies = (4,1,2) and
I)) = (4, — 1,0), find the image of the following.
a. (1,0) b. (2,1) c. (4,4)
d. (3,4) e. (—1,—I) f. (2,3).
11. If a mapping R2 —' R2 maps (1,0) to (4,8), (1,1) to (—5,3), and
(0, 1) to (0, 1), can it be a linear transformation?
12. As for rings, define the kernel of a linear transformation U—' V
to be the subset of vectors A U for which = 0. Prove that
is a subspace of U.
13. a. If U —' V is a linear transformation defined on an n-dimensional
vector space U with a k-dimensional kernel (k <n), define the
quotient space and find its dimension [cf. the discussion of
in §2.9).
b. Prove that = (Be V: B= (0(A) for some Ac U} is a subspace
of V. Find its dimension.
c. Prove that and U/kerç,, are isomorphic vector spaces.
 Aspects of Unear Algebra chapter 4

14. Considering F/F as a (one-dimensional) vector space over itself, prove

that F) is isomorphic to U. Choose bases for U and F, and
describe one for HomF(U,F). (The vector space Hom,-(U, F) is called
the dual space of U, commonly denoted its elements are called linear
functionals.)
15. LetU—. V be a linear transformation which is an isomorphism.
Define the inverse mapping from V to U.
a.
b. Show that the inverse is a linear transformation.
16. Consider an idempotent transformation that is, =
Such a transformation is commonly called a projection. Prove that
V= 1m47.
17 Projections (idempotents) e End,(V) are said to be orthogonal

a. Let ..., be orthogonal idempotents such that + + =

the identity map of V onto itself. Prove that there exist subspaces U1
of V such that V= U1 Ui,. Hint: Consider =
b. conversely that if V is the (internal) direct sum of subspaces
U1, ..., there exists a system of orthogonal idempotents
e V) such that + + =

§4.4 Matrices and Linear Transformations

Matrices provide a convenient shorthand notation for describing
linear transformations of finite dimensional vector spaces. Indeed this is how
they arose (in 1858) in the work of Arthur Cayley (1821—1895) at Cambridge,
culminating in his development of matrix algebras. While Cayley had con-
sidered matrices in the abstract, in 1925 Werner Heisenberg recognized the
noncommutative product of matrices as necessary for his significant develop-
ments in quantum mechanics. Matrices are used to deal computationally
with problems involving linear transformations. Besides applications in
linear algebra, they arise in the theories of differential and integral equations,
in group theory, and in linear programming as means of representing
systems of linear equations.
Here our concern is with the simple properties of matrices and the
representation by matrices of the linear transformations introduced in §4.3,
which will be used in §9.4 in the study of algebraic field extensions. In §4.5
we shall discuss briefly the related subject of determinants.
A matrix with coefficients (also entries or components) in a field F
is an ordered rectangular array
a11
f
L amt

of elements c F. I 1 n. We call the array A an mxn matrix

§4.4 Matrices and Linear Transfonnations 111

because it has in rows and n columns. In abbreviated form it is denoted

A = [a3, I i in, I j n,
where is the entry in the ith row and jth column of the array.
The n-tuple of elements in A is commonly referred
to as the ith row of the matrix A. Analogously, we shall refer to the m-tuple
ami) as thejth column.
Two in x n matrices are said to be equal if their zjth coefficients
are equal for all i,j.
The set Fm.n of in x n matrices with entries in F is a vector space
over F with laws of composition given as follows.

(i) Vector addition. The sum of matrices A = and B = is

defined to be the matrix A + B having + for its ijth coefficient.
(ii) Mulliplica: ion by scalars. The product of A = [au] by r e F is
defined to be the matrix r•A = rA whose coefficient is rag.

Direct computations verify that the axiomatic properties of a vector space

1] are satisfied. The zero matrix (additive identity) is simply the m x n
array of zeros.

For example, for real matrices (matrices with components in R) we have

1 1 31
1+1
Ii —211=112 1
I
8 J [5 —IJ [1 7

and 5.
I 3] 1 5 15

—4 8J —20 40

Special matrices of significance are those which for fixed i,j have
all their entries 0, except for a I in the ith row, jth column (the ijth position),
where 1 denotes the multiplicative identity of F. These matrices will be denoted
I I rn, 1n. For any matrix A = E Fmn, we can write

A = I I in, I n.

Thus the inn matrices generate the vector space Fmn. But further they
are linearly independent, since any linear combination

=0 (the zero matrix)

i.j
implies = 0 for all i,j. 1-lence the inn matrices constitute a basis for
Fmn. Collecting these results, we have proved the following proposition.

Proposition 1. The set of in x n matrices with entries in F is an

inn-dimensional vector space over F.
 _

Aspects of Linear Algebra chapter 4

Example 1. The space of matrices R23 has a basis

[1001 [0101 Fool
E 11 —
[000]
I I E 2 =
[000] i E,3
[000
[0001 10001 [000
1 0 oj i oJ 0 1

a11 a12 a13

and [ = EaijE4j, i 1,2, / = 1,2,3.
L a21 a22 a23 ]

Given an in x n matrix A, we define its transpose 'A to be the n x in

matrix obtained from A by making the ith row of A the ith column of 'A.
If A = [au], then the zjth entry in 'A = [ba] is as,. As an example, for
a11 a21
a11 a12 a13
A= = a12 a22
a21 a22 a23
a23

Furthermore, the mapping given by for A

is a (vector space) isomorphism. Note that '('A) = A.
A second matrix operation useful in the representation theory of
linear transformations by matrices is that of matrix multiplication. To define
the product of matrices A, B we require that the number of columns of A be
equal to the number of rows of B. Thus we shall multiply inxn by nxp
matrices.
The product of matrices
I 1

B=[bjk],
is defined to be the m x p matrix
AB=C=[c,,j,
where e,,, = That is, the ikth coefficient of the product is the sum
of ordered products of elements in the ith row of A with those in the kth
column of B.

This notationally somewhat complex rule may be seen more clearly in an

example:
1 2 3

f I 0 5 —4 0 1 C12 C13
[3 8 —1 sJ 2 —l —I Lc2i C22 C23

o i 8

_[ II 1 30]
— t—31 12 58 j
§4.4 Matrices and Linear Transformations 113

where = 1.1 +0(—4) + + 4.0 = II,

1 •2 + 0•0 + I) + 4.1 = 1,
= 1.3+0.1 + = 30,
= 31 + 8(—4) + (— l)2+ 5.0 = —31, etc.

The product of matrices provides a single-valued mapping from the

cartesian product Fm.nXFn.p tO Fm,p, satisfying the following properties for
A,B&Fmn, EEFp,q, and reF.
(I) (A+B)C=AC+BC
I I
sum in F,., sum in F,.,

(ii) A(C+D)=AC+AD
'1 1
sum in F,, sum in F,.,

(iii) (rA) C = A (rC)

(iv) (AC)E=A(CE)
We leave aside verifications of these properties, which are straightforward,
although at times computationally tedious, consequences of the definition
of product.
Note that this product is noncommutative. Indeed if A and B are
and nxp matrices, respectively, where ,n then BA is not even
defined. If both A and B are nxn matrices, then AB need not equal BA,
as is seen in the simple example
00 10 00

but
[{
The set F,,,, of n x ii matrices with entries in F has the matrix
I,, = the Kronecker delta—for its multiplicative identity element.
We turn now to the interrelation between matrices and linear trans-
formations. Consider vector spaces U/F and V/F with respective dimensions
n and rn. In the proposition of §4.3 we proved that dim HomF(U, V) = mn
and in Proposition I above that dim Fmn = inn. Then by Corollary 2, §4.3,
these two vector spaces over F are isomorphic. Thus we have proved the fol-
lowing proposition.

Proposition 2. With vector spaces U, V as described above,

HomF(U,V)
To define such an isomorphism explicitly, first choose bases
d= {A1 A,,) and .B,,,} of U and V respectively. In terms of
 Aspects of Linear Algebra chapter 4

these bases a linear transformation q: U V is described by

p(A,)= a11B1 + +amiBm

p(A3) = + + amjBm =

çp(A,,) = + +amnBm =

Thus, when A

p(A) = =

= rj au) = s1 i i m, i j n,

where the coefficients are given by the matrix product

s1 a11 r1 r1

= =rMQ.

Sm ami

Alternative'y, using the transpose of one-column matrices to save space,

we have

(**)
Now to the linear transformation q associate the in x ii matrix

a11

0ntl a,,,,,

the transpose of the array of coefficients in (*) above. We speak of M4, =

as the matrix representation of the linear transformation iiith respect to the
given bases of U and V. With a different choice of bases we get a different
matrix representation of and a different isomorphism Hom1(U, V) Fm,,. —

Example 2. To illustrate this representation, consider vector spaces U = V R2.

Choose {(l,O),(O, 1)} as a basis for U and {(l,2),(3,4)} as a basis for V. Define a
linear transformation q: U—' V by
= 4(1,2) — 3(3,4),
= 2(l,2)+ 5(3,4)
§4.4 Matrices and Linear Transformations 115

and to q associate the matrix

In representing a linear transformation V—' V. we would use the same

choice of basis for V when considered as the domain as for Vconsidered as the range.
To emphasize the importance of the basis in the representation M0, replace
the previous basis of U by {(l, l),(2, —3)} and compute the representing matrix N0.
First,
I)) = + 1))
= [4(l,2)—3(3,4)] + 12(l,2)+5(3,4)]
= 6(1,2) + 2(3,4),
—3)) = 2(4(l,2)—3(3,4)] — 3[2(l,2)+5(3,4))
= 2(l,2)—2l(3,4).
Thus, N0 M,.
= [

If further, W) is represented by the p x in matrix

= IbM] with respect to bases of V and = {C1,. ç} of W, then
we show that o E W) is represented with respect to the bases
of U. W, respectively, by the p x n matrix
=
For the proof we examine the following diagram:
U
cW
U
and relate the effect of the maps q', cli to the
on a vector A = rj
bases of U,Vj4', respectively. Therefore, expressing the linear
transformation V—p W by

i,l' s7 =

and writing the analogue of'(**) for we have for the components
'[s1 s,,,] =
= ....Sm].

Hence by the associativity of' the matrix product.

= rn])
= rn].
Since, for each vector A E U, the image q)(A) = t/i(p(A)) has
coordinates t1 i,, relative to the basis "C of W, we can conclude that
= Ms,. (Be certain to note however that Mq, is the matrix rep-
resentation of p with respect to the bases d and that is that of cl'
 Aspects of Linear Algebra chapter 4

with respect to the bases and and that is that of with

respect to the bases d and
Now let LJ=V=W and Then the equation
Mq, implies, in addition to the (vector space) isomorphism of Proposition
2, another isomorphism.

Proposition 3. For an n-dimensional vector space V/F the rings EndF(V)

and are isomorphic.

As in Example 2, let U equal R2 with basis (A1 = (1,0), A2 = (0, and

V equal R2 with basis (B1 (1,2), B2 = (3,4)}. Considering as before the mapping
(1 V represented by
4 2
M0=
_—3 5

we compute for A = r1 A1 + r2 42 e U the image ço(A) = s1 B1 +$2

[sill 4 21[ri][ 4r1+2r2

Fszjl—3 5ff nj [—3r1+5r2
or = (4r1+2r2)B1 +(—3r1+5r2)B2.
And if the transformation V —, V is represented by

ii,
then = + t2 C2 has coefficients given by

F" LF 1 -1ff 4 2ff

I — [—1 I J[—3 5ff
11 r1 7r1 — 3r2
— F 1 — I
— F—7 3 iF r2 ] F—7n, + 3,'2

or = (lr, — 3r2) B1 + (— + 3r2) B2.

The relationship, given below in Proposition 4, between matrices

M, and representing a given linear transformation q: V—. V with respect

in §5.9. To this end we now examine the relationship between the components
of a vector A e V relative to two bases d = and = (B1, ...,
of V.
The basis vectors B1 can be expressed in terms of A1, ..., as

and the basis vectors in terms of B1 as

A.
§4.4 Matrices and Linear Transformations 117

Set P [p1k] and Q = Then, since

B1 = = = I j, k n,
we have 'fri PJi = èik,

which is to say that

QP = I,.
Similarly, PQ =
the products PQ, QP of n x n matrices P. Q are the n x ii identity
if
matrix we call P the inverse of Q (denoted p'), and conversely. Matrices
in which have multiplicative inverses (i.e., the units of the ring are
called nonsingular.
Now let
A

= + •.. + U,,B,,

u1 A1) + +

= PJIUI) =
j=1 1=1 j=1

Thus r1 = u1, and setting P = we have

=

With these preparations we are ready to establish the following

relationship between two matrices that represent a linear transformation ço
with respect to different bases.

Proposition 4. Let M9, and Nq, be n x matrices representing a linear

transformation p: V—' V with respect to different bases of V/F. Then
there exists a nonsingular n x n matrix P such that = P'
Proof Expressing A Vas A = A1 + = u1 + relative
to the bases d and we obtain, as in the discussion after Proposition 2,
the following equations for the components of the image q(A):
t{S = re],
'[r1, = N9,t[u us].
Since
t[1. = P.t[u1 un],
 Aspects of Linear Algebra chapter 4

and similarly
= Q•'[s1 ...., .c,,] Pi.t[si,..., sj,
we have

Mq,.'{r1 u,j
and
= ;] =
= un].

This last equality is valid for all choices of u1 i.e., for all A e V. and
therefore N, = P - 'M, P, as was to be shown.

Example 3. Again we return to our earlier example to explicate the computations:

B1 = (1,2) = ÷
B2 = (3,4) = 341 + 4A2,
= (1,0) —2B1 + 82,

42 = (0,1) = —

and p{Pu1 P121=1 1 3

P221 4

q21 q22 j I

Further, R2 R2 defined by = (2a+b, —a+4b) is represented with

respect to the basis d = (41,42) by
2 I
M0=
—I

and with respect to the basis B1, by

—2 2
N0 = P'MP = f
I 4j
1
2 4

15
12

Exercises

1. a. Prove that the map F,,rn given by = 1,4 is a (rector

space) isomorphism.
b. Prove that the map F,,,, F,,,, given by ço(A) = ',4 is not a ring
isomorphism for is 2.
§4.4 Matrices and Linear Transformations 119

2. Evaluate the following matrix products.

1 —1 0
[1 2 0
1

4 0

1 8
1 0
2 0J 001 25
3. Prove associativity of the matrix product. That is, for matrices A, B, C
of respective sizes in x ii, nx p, p x q, prove that (AB)C = A(BC).
4. Prove distributivity of matrix multiplication over matrix addition. That
is, for matrices A, B, C of respective sizes in x n, ii x p, n x p. prove that
A(B+C) = AB+AC.
5. Evaluate the following matrix products

2 4114 811—I
ii 3
a.
—l 5
I

6ff 2 —4

b. I
2 41114
Ill 8 +
—1 3

5 I 6 2 —4

11 2 4 14 —8 i\ 1—1 3
C. (I
1

1+1 III
1 11 —6 2 —4

6. Determine the representing matrix M0 with respect to the usual

(canonical) basis {(l,0,0),(0, 1,0),(0,0, of R3 for each of the linear
transformations R3 —+ R3 described as follows.
a. y, z)) (x, ),0)
b. p((x, y, z)) (x, 2y, — z)
C. ço((x,y,z)) = (x+y+z,y+z,z)
d. p((x, y,:)) = (5x, Sr 5z)
e. 9,((x,y,z)) = (O,y,O)
f. y,z)) = (0,x+ z,0)
g. y, z)) = (z, y, x).
7. Determine the matrix representing the composition of the following
mappings in Exercise 6.
a. The mapping in (a) followed by that in (f)
b. The mapping in (f) followed by that in (a)
c. The mapping in (f) followed by that in (f)
d. The mapping in (c) followed by that in (a)
e. The mapping in (g) followed by that in (c).
120 Aspects of Linear Algebra chapter 4

8. Determine the representing matrix for each of the transformations of

R3 given in Exercise 6 with respect to the basis {( 1, 1,2), (2, 1,0), (2,0,
of K3.
9. Determine the representing matrix MQ with respect to the basis
((1,0), (0, 1)) of R2 for the linear transformation R2 —' R2, where is
rotation through the following angles in radians.
a. ir/6 b. ,z/4
c. 3ir/2 d. 2ir13.
10. Repeat Exercise 9, expressing the matrices with respect to the basis
{(1,0),(l,l)J ofR2.
11. En Exercise 9 use matrices to verify that four successive rotations through
ir/6 radians equal one through 2n13 radians.
12. The canonical (customary) basis of is (E, = I I n}..
In terms of this basis, give the matrix representation of each of the linear
transformations —' described as follows.
a. Forgiven I,],

I k k I,].
b. Forgiven I, I i ii,and nonzero cEF,
to(Ek)=Ek,
c. For given 1,1, 1 i,j it, and CE F,
= + cE1,

d. Prove that the linear transformations in parts (a), (b), and (c) are
isomorphisms.
13. Prove that an n x it matrix A is nonsingular if and only if its rows (or
columns) are linearly independent. 1-lint: Consider the dimension of the
image of the linear transformation determined by A.
14. In the ring of 2 x 2 real matrices prove that

0
a. { is an idempotent zero divisor.
10 ii 1

b.
1211 and
[1 —11 are zero divisors.
4 2 j 2 2j
[2 2] is a unit.
C.
[4 2
1

d.
[001 and
18 —81 are nilpotent elements.
[ 0j 8 —8 j

15. Let R be the set of 2 x 2 matrices with integral components.

a. Prove that R is a noncommutative ring.
b. Describe all units in R.
§4.5 Determinants 121

c. Give examples of matrices which are not units and are not zero
divisors in R.
d. Describe the nontrivial two-sided ideals in R.
16. Let R be the noncommutative ring of 2 x 2 matrices with rational
components.
a. Describe all units in R.
b. Prove that if a matrix n C R is not a unit, then it is a zero divisor.
c. Show that R has nontrivial one-sided ideals, but has no nontrivial
two-sided ideals.
17. a. Describe all units in the noncommutative ring R of 3 x 3 matrices
with rational components.
b. Give examples of idempotent and nilpotent elements in R.
18. a. Give an example of a noncommutative ring in which the sum of two
nonzero nilpotent elements is a unit [cf. Exercise 7(a), §3.1].
b. In the ring of 2 x 2 matrices with coefficients in Z2, prove that the
identity matrix cannot be written as the sum of two nilpotent
matrices. Verify, however, that the following matrices are nilpotent
and sum to the identity:
[00
[ii]'
11 1.j [0 '1
[ooj' [i 0

§4.5 Determinants
We conclude our present discussion of linear algebra with a sum-
mary of the properties of the determinant function, which maps the ring of
n x n matrices F,,, into its field of coefficients F. In and 9.4 we utilize
determinants to define the characteristic polynomial ofa linear transformation.
The earliest use in the Western world of what we call determinants
apparently was by Leibniz in a 1693 letter to G. F. A. de L'Hospital
(1661—1704). Later Gabriel Cramer (1704—1752) and Cohn Maclaurin
(1698—1746) independently developed the theory of determinants related to
the solution of systems of two, three, and four linear equations. Maclaurin's
Treatise of Algebra appeared posthumously in 1748, and Cramer published
his widely known rule for such solutions in 1750. Analogous results were
known to the Japanese mathematician Seki Kowa (or Seki Takakusa,
1642—1708). One of the modern axiomatic treatments of determinants is due
to Karl Theodor Wilhelm Weierstrass (1815—1897), with earlier discussions
by Augustin-Louis Cauchy (1789—1857), William Rowan Hamilton, Pierre-
Simon de Laplace (1749- 1827), and Alexandre Theophile Vandermonde
(1735—1796).
Consider an #zxn matrix A = [au] = [A'. e where
denotes the jth column of A. The determinant of A. denoted
det(A) = IAI = det([A'
is the element of F, determined from A in accordance with the following
axioms.
 Aspects of Linear Algebra chapter 4

1. Linearity with respect to the columns

A =

then det(A)+det(B) = det(C).

(ii) For all c e F,
det([A' cA', ...,A"]) = det([A', ...,A"]).
II. If two adjacent columns are equal, for somej, I j<n, then
det(A)=O.
Ill. For the n x n identity matrix = I.
The determinant function is uniquely determined by the preceding
axiomatic properties from which the following consequences can be derived.
Proofs of these properties are omitted: some require consideration of
permutations of n symbols, a concept to be introduced in §6.7.

Property 1. If the matrix B is obtained from the matrix

A = [A' A' A"]

by interchanging the ith and jth columns, then det(B) = —det(A).

Property 2. If any two columns of' the matrix A are proportional (i.e., if
one is a scalar multiple of the other), then det(A)=O.

Property 3. if the matrix B is obtained from the matrix

A = [A' A' A' A"]

by adding c times the ith column to the) th column, i then det(A) = det(B).

Using properties of permutations we can show another

property.

Property 4. The det([A', ...,A"J) is given explicitly in terms of its n2

entries aU by
det(A) 2

The sum is taken over all of the n! permutations it of {l n}, and

sgn(ir) = ± denotes the sign (parity) of the permutation [cf. Exercise 6,
1

§6.7].
§4.5 Determinants 123

The formulas for determinants of 2 x 2 and 3 x 3 matrices are easily

stated:
a,, a,2
= a11a22 — a,2a21,
a2, a22

a,, a12 a,3

= a,1a22a33 + a12a23a3, + — a,3a22a31
a2, a22 a23
—a23a32a,, —a33a21a,2.
a3, a32 a33

For larger values of n, the following formula, called the Laplace

expansion by minors, reduces the determinant of an n x n matrix A to a linear
combination of the determinants of n (n— I) x (n—I) matrices.

Property 5. For any fixed i, I I n.

det(A) 'as, + + (—
where the (n—l)x(n—l) matrix
A = [au] by deleting its ith row andjth column. The matrix is called the
ijth minor of A.
We complete our list of common properties of determinants with
five additional facts.

Property 6. det(A) = det('A).

Property 7. det(AB) = det(A) det(B) = det(BA).

Property 8. If A is a nonsingular matrix (i.e., if it has a multiplicative

inverse A'), then det(A ')= [det(A)]'.
In Exercise 13, §4.4, we noted that a matrix is nonsingular if and
only if its columns are linearly independent. From this consideration we can
derive Property 9.

Property 9. A matrix is nonsingular if and only if its determinant is nonzero.

The following statement on homogeneous systems of linear equations

is a consequence of Property 9. The system of n equations in n unknowns
x1

= 0, I i n
1

has a nontrivial solution (i.e., not all x1 = 0) if and only if = 0.

Property 10. =
124 Aspects of Linear Algebra chapter 4

The Laplace expansion of a determinant in Property 5 involved the

(ii— l)x(n— 1) minors of a matrix A. The proof of the Cayley-Hamilton
Theorem in §5.9 utilizes the (classical) adjoint matrix Adj(A) of A, also
defined in terms of minors:
Adj(A) where ; =
The component is commonly called the ijth cofactor of A.
Finally. A and its adjoint satisfy
(*) A•Adj(A) = (det(A))I,1.
Adjoints provide a convenient means of computing the inverse of a given
(nonsingular) n x n matrix A, especially those for smail n.
 5

Polynomials and
Polynomial Rings

While the reader no doubt has encountered polynomials with

integer, rational, real, or complex coefficients earlier in his mathematical
studies, he now will see them treated constructively where the coefficients
lie in an arbitrary field. Emphasis is placed upon arithmetic properties
analogous to those in the domain of integers, again an example of the
extension of familiar concepts. in and 5.2. we parallel the previous
discussion of the ring of integers and the residue class rings thereof. Special
attention is given to the congruence relations used to develop the theory of
formal derivatives and the study of multiple roots Thus in part we
prepare for the theory of inseparable field extensions and Chapter 8].
Kronecker's construction of the roots of polynomial equations in
§5.4 will be applied to the proof of the existence of splitting fields. This is
necessary preparation for the study of field extensions and the Galois theory
in Chapter 8. Gauss' Lemma on Primitive Polynomials and Eisenstein's
Irreducibility Criterion are discussed in §5.8. Aspects of the Jatter will be
used to prove that the field of complex numbers admits only linear irreducible
polynomials—the Fundamental Theorem of Algebra
This chapter ends with an examination of very special polynomials:
the characteristic polynomials of a matrix or, equivalently, of a linear
125
 Polynomials and Polynomial Rings chapter 5

transformatjon. These facts are needed for the study of field extensions in
Chapter 9.

§5.1 Polynomial Rings

Polynomials are introduced in the calculus as polynomial functions
i = a0 + a1i + + R

for all t e R and given a1 R, 0 i n, for some n 0. For reasons to

become clearer later, we present a different approach, equivalent for the field
R but applicable to fields F different from R, e.g., Z,,. Polynomial functions
will be treated in §5.7.
We consider infinite vectors or tuples (a0,a1 a1, ...), /3 =
(b0,b1,...,b4,...), and y=(c0,c1,...,c1,...) with components a.,b1,e1 in a
field F, such that a, b., and are zero for all i larger than some positive
integer N, depending on /3, and y. (That is, all but a finite number of the
components a1, and c1 are zero.) As for finite n-tuples, it is agreed that
= /3 if and only if a1 b1 for all subscripts I. We define
= (a0-I-b0,a1+b1,...) and

r F. A quick check shows that a vector space of infinite dimension over F

is obtained (since no finite set generates the entire space).
Next, recall the distributive law of multiplication for polynomial
functions

ait')( =
a.bj)fh. for every teR,
i=O h=0 i+j=h
where 0 I n and 0 m. As an explicit example,
(6+3t+12)(4+12—t3) = 24+ 12t+ lO,2_313_2,4_,5.
Using this distributive law as a model, we define the product to be the
infinite vector (d0, d1, ..., ...) whose components dh are given by the
formulas

j+j=h
for h 0, 1,2 Note that the summation involved in defining each of the
sums dh is finite, and that only a finite number of the dh are distinct from zero.
Simple juggling of indices and summations verifies the following
algebraic properties for infinite vectors z /3. y:
(1) 2/3=/31
(ii) = (2/3))'
(iii) cx(/3+y) = + cy
 Polynomial Rings 127

(iv) (I,0,...)x =
(v) (0, ..., 0, 1 , 0, .. .) (0, ..., 0, 1,0, ...) = (0, . . ., 0, 1,0, ...),
I I I
ith component jth component (i+j)th component

where the enumeration of components begins with zero.

Property (i) follows from the commutativity of the elements
F, because Similarly, property (iii) follows
from the distributivity of multiplication over addition in F, =
and the commutativity of addition in F.
Property (ii) requires more careful attention to the summation
notation. The qth component is

a1( =
i+h17 j+k=h i+h=q j+k.h

= 54h,1 I a. b3 CL = a,
k
(

which is also the component of Note the use of the associative

and distributive properties of elements in F Property (iv) is a simple con-
sequence of the definition of the sums C/h in the product of (1,0, ...,0, ...)
and
In other words, the infinite vectors ... form a set obeying the
same rules of algebra as the usual polynomial functions (the case for F= R).
For this reason and for the sake of notational simplicity we identify (i.e.,
equate) the infinite tuple (a0. 0, ...) with the element a0 e F; the tuple
(0,0 1,0,...), where I is in the (1+ l)st position, with x' (or etc.); and
(a0,a1, ...) with a0+a1 x+ = The resulting set
F[x] (or F[tJ, etc.) is called the ring of polynomials in the indeterminate x
(or t, etc.) with coefficients in the field F. Coefficient fields common in our
subsequent discussion are Q, R, C, or p a prime number.
The principal properties of polynomial rings are summarized below.
Consider two polynomials f(x)= and g(x) = in the
indeterminate x with coefficients a,, b, in F. Here some or all of the coefficients
a1,b1 may be 0. Note that terms like Ox" = 0. in a positive integer, can be
added to any polynomial without changing it. The elements a F are
identified with the polynomials a+Ox+ +Ox&. They are called constant
polynomials, or more briefly. "constants." The polynomial (0,0, ...) is called
the zero polynomial.

Property 1 (Equality of Pulynomials). Polynomials /(x) and g(x) are equal

ifand only ifa1=b1, forall

Property 2 (Addition of Polynomials)

f(x) + g(x) = (a0+b0) + +h1)x + + (ak+bk)x".

 Polynomials and Polynomial Rings chapter 5

Property 3 (Product of Polynomials)

f(x)g(.v) = + (/1X + + + +

where dh = aobh + a1 hh_ + + ahhO 0 h 2k.

=
As indicated in its construction, F[x] is a co,nmuzative ring.

The degree. denoted degf(x), of a nonzero polynomial f(x) is the

greatest integer n for which 0 in the expression of as .1(x) =
a0+a1 x+ +akxA. For convenience, in stating the properties of the degree
of a polynomial we assign to the element 0 e Fc F[x] the symbol — as
degree. It is agreed that
=
—co+k = foreveryintegerk
—cc <k.
If degf(x) = n, then is called the leading coefficient. A monic polynomial
is one whose leading coefficient is I.

Property 4. The degree of polynomials satisfies the following conditions.

(i) deg[f(x)g(x)] = degf(x)+degg(x).
(ii)

(iii) deg[f(x)+y(x)] max(degf(x), degg(x)); i.e., the larger of

degf(x) and degg(x).

Proposition. The polynomial ring Fix] is an integral domain.

Proof To verify that F[x] has no proper zero divisors (or equivalently,
that the multiplicative cancellation law holds, cf. §3.6), suppose that for
f(x) = and g(x) = where 0, 0, we had
f(x)g(x) = 0. Then each of the coefficients of powers of x in the product
would be zero. In particular, anhm = 0, and since the coefficients lie in the
field F, then either a contradiction. Thus, F[x] has no
proper zero divisors.
More generally, we can consider rings of polynomials with co-
efficients in an integral domain D. Nowhere in the preceding discussion was
the existence of a multiplicative inverse used. In particular, Z[x] is a ring,
and by the proof of the above proposition, an integral domain.
The Construction of the quotient field ofan integral domain [Theorem,
§3.6] can be applied to F[x]. The quotient field Q(F[x]) = F(x) is called
the field of rational functions of the indeterminate x with coefficients in F.
 Polynomial Rings 129

its elements are usually identified with the "quotients" a(x)/b(x) of poly-
nomials a(x),b(x) in F[x], with b(x) 00.
Having defined in detail polynomials in one indeterminate x, we
conclude this section with the definition of the ring of polynomials F[x1, ..., xj
in several indeterminates 1,...,x,,. There are two common (and equivalent)
definitions. First, for all i e N, inductively define F[x1 to be the
ring of polynomials F[x1 ,][x1÷1] in the indeterminate with
coefficients in the ring Fix1 x1]. A second definition views F[x1
as the set of finite sums of monomials of the form where a€ F
and ..., I,, are nonnegative integers, together with the customary algebraic
rules of commutativity, associativity, etc.
Polynomials in two indeterminates occur in some exercises, in the
discussion of formal derivatives in §5.5, and in §9.8.

Exercises

1. Prove that (he distributive law holds in F[x).

2. Prove that the ideal generated by x in Z [x) is prime, but not maximal.
3. a. Prove that the ideal generated by x in Q[x] is maximal (and hence
prime).
b. Prove that the ideal generated by x2+2 in Q[x] is a prime and
maximal ideal.
4. Prove that not all ideals in Z[xJ are principal.
5. Prove that F{x, y] contains ideals which are not principal.
6. Let F be a field. Consider the set consisting of all formal power
series a0+a,x+ a, e F. Prove that is an integral
domain if sum and product of power series are defined in the same manner
as in the polynomial ring F[x]. Find all units of and prove that the
elements of the quotient field of have the form xm (a0 + a1 x +
with a0 0.
7. In the polynomial ring F[x, suppose thatg(x, y) = g0(x) + g1(x)y +
+y (x)y" such that go(0) = g(0,O) 0 and g,(0) = ?g(x, y)k?y 0 for
(x, .v) = (0,0). Prove that there exists a unique power series y 1(x) =
a1 xi- for which g(x, y) = 0 in
F be fields and let a1 K. Consider the set A of poly-
nomials f(x1 ,...,x ) in the polynomial ring F[x1, ...,x,J for which

a. Prove that A is an ideal in F[x1,...,

b. Prove that A is a prime ideal in F[x,
9. For an arbitrary ring R, prove that R[xJ is an integral domain if and
only if R is an integral domain.
10. Prove that for each a e F the mapping F[x] F given by
V1a(f(X)) 1(a) is a ring homomorphism.
11. Prove Properties 4(i) and 4(iii) for the degree of polynomials.
130 Polynomials and Polynomial Rings chapter 5

§5.2 Divisibifity and Factorization of Polynomials

Most of the proofs of the following statements concerning the
divisibility and factorization of polynomials are modeled after the proofs
for the corresponding statements for integers. As a general rule, we replace
the absolute value a for 1(x) e F[x]. This
holds especially for applications of the Principle of Induction. Note that the
proof of the division algorithm for polynomials requires that the coefficients
be in a field, not just an integral domain.

Proposition 1. Given a polynomial a(x) and a nonzero polynomial b(x) in

F[x], there exist unique polynomials q(x) and r(x) in F[x] such that
a(x) = q(x)b(x) + r(x),
where either r(x) = O.or 0 degr(x) < degb(x).
Proof [See the Division Algorithm, §2.3.] If dega(x) < degb(x), then set
q(x) = 0 and r(x) = a(x). Otherwise, write
a(x) = a0+a1x+
b(x) = +bmX", bm 0.

If n=m and then the degree of is less than m.

Set q(x) = anbm ',and r(x) equal to the preceding difference.
More generally, if n> m, induction on the degree n of a(x) is made.
Assume the existence statement for polynomials of degree less than n. The
degree of c(x)= is less than n; thus by the induction
hypothesis there exist polynomials (x) and r(x) for which c(x) =
q1 (x)b(x)+r(x), where degr(x) <degb(x) or r(x) = 0. Consequently,
a(x) = q1(x)b(x) + r(x) + (anbm I)XhI_m!,(X)

= [q1(x) + + r(x)
= q(x) b(x) + r(x),
where q(x)= and r(x) = 0 or degr(x) <degb(x). The
Principle of Induction implies now that the existence statement holds for all
a(x) and any given polynomial nonzero b(x).
The uniqueness of q(x) and r(x) is proved as follows. Suppose
a(x) = q(x)b(x) + r(x)
= q*(x)b(x) + r*(x).
Then, = r*(x) — r(x).
Since deg{r(x)—r(x)) <degb(x). properties of the degree imply that
q(x)_q*(x) must be 0. Hence also r*(x)_r(x)=0.
§5.2 Divisibility and Factorization of Polynomials 131

The striking analogy between the division algorithm for polynomials

and that for integers is the basis for further parallel definitions, statements,
and theorems. The subsequent results are obtained by replacing arguments
involving the absolute value used for integers by the degree of polynomials.
(For example, compare the statement of the Division Algorithm with
Proposition I above, or the theorem with Proposition 2 below.)

Proposition 2. All ideals A in F[x] are principal; if A {O}, then A = (a(x)),

where a(x) is a monic polynomial.
Proof. The zero ideal (0) is principal, and its generator is the zero poly-
nomial. Now for A (0), the Set

S = {degf(x) : f(x) 0, f(x) e A}

is a nonempty set of nonnegative integers. By the Well-Ordering Principle S
has a least element k, and so A contains a polynomial c(x) of degree k.
Multiply c(x) by the multiplicative inverse of its leading coefficient, thereby
obtaining a monic polynomial a(x) of degree k. also belonging to A.
To prove that a(x) is the desired generator of A, write
f(x) = q(x)a(x) + r(x)
for arbitrary f(x)EA. Hence r(x)=f(v)—q(x)a(x)e A. and degr(x)<
dega(x). Since no nonzero polynomial in A has degree less than k, we conclude
that r(x) = 0. Therefore a(x)If(x). as claimed.
If dega(x) = 0, then a(x) = I (the monic constant polynomial), and
so (a(x)) = (I) = F[x]. Similarly, for all nonzero ce F, (c) = F[x].
As in an arbitrary ring I], elements u F[x] for which there
exist polynomials z' such that uv= I are called units of the ring F{x].
Using 0 = degl = deg(uz') = degu+degv, we note that the unils of F[x] are
the nonzero constants of F.
A polynomial g(x) is called a divisor of f(x), if f(x) =g(x)h(x) with
h(x) e F[x]; for this we use the notation g(x) I •f(x) [cf. §2.5].
The greatest common divisor d(x) of two polynomials a(x) and b(x),
denoted d(x) = (a(x),b(x)), is defined as follows:
1. If a(x) = b(x) = 0, then d(x) = 0.
2. If either a(x) or h(x) 0. then
(i) d(x) is to be a common divisor of a(x) and b(x),
(ii) any other common divisor e(.v) of a(x) and b(x) is to divide
d(x),
(iii) d(x) is to be monic.
The existence and uniqueness proofs of the greatest common divisor are
modeled after the corresponding proofs for integers. The ideal A, consisting
132 Polynomials and Polynomial Rings chapter 5

of all linear combinations a(x)f(x)+b(x)g(x) in F[x]. has a urnque monic

generator d(x), by Proposition 2 above, which can be shown to be the desired
greatest common divisor [cf. §2.5].

Proposition 3. If d(x) = (a(x),b(x)), there exist polynomials s( x), 1(x) E F[x]

such ihat d(x) = s(x)a(x)+1(x)b(x).

Polynomials a(x) and h(x) are said to be relatiiely prime if

A = (a(x), b(x)) = I. In other words, polynomials a(x), b(x) are relatively
prime if and only if there exist polynomials s(x), 1(x) e F[x] such that
a(x)s(x)+b(x)t(x)= 1.
A polynomial p(x) of F(x) is called an irreducible or prime poly-
nomial, if it has only trivial divisors; i.e., if f(x)Ip(x), then either f(x)=c,
a nonzero constant, orf(x) = cp(x). [Compare the definition of prime number
in §2.6.]

Theorem I (Unique Factorizalion ofPolynomials). Each nonzero polynomial

f(x) can be written uniquely (apart from arrangement of the factors) as a
product
f(x) = oP1

where 0 is the coefficient of the highest nonzero power of f(x) and the poly-
nomials p.(x) are monic irreducible polynomials whose multiplicities cc, are
positive integers. (p(x)° = I as for integers.)

Because the proof is formally the same as that for the Fundamental
Theorem of Arithmetic in §2.7, we do not repeat it here. Note that the
properties of prime integers carry over to irreducible polynomials. In
particular, compare Proposition 2, §2.6, with the following.

Proposition 4. A polynomial p(x) e F[x] is irreducible if and only if

or

We illustrate now the determination of the greatest common

divisors of two polynomials. Note the similarity of this procedure to that
in §2.5 for finding the GCD (a, 1') of integers a, b.

Example. Consider the polynomials

a(x) = x4 — x2 + x — 3, b(x) = x2 — x + 1.

To show that (a(x),b(x)) = 1, and that

x—3\ f—x3-i-2x2-i-4x—2
=
§5.2 Divisibility and Factorizatlon of Polynomials 133

we carry out the long division of polynomials.

x2 + x — 1
x2— x+ I — x2+x—3
x4—X3+ x2
x3 — 2x2 + x

x3— x2+x
—x2 —3
x2+x—1
—x—2
Thus a(x) = x4—x2+x—3 = (x2+x— l)b(x)—(x+2). Dividing b(x)= x2—x+ I
by —x—2, we find
x2—x+1 =(—x+3)(—x—2)+7.
The remainder 7 is not the greatest common divisor, since it is not a monic poly-
nomial. However, since it is constant, it indicates that (a(x),b(x)) = 1.
Reversing the process, we obtain
7 = b(x)—(—x+3)(--x—2)
= b(x)—(—x+3)[a(x)—(x2+x--l)b(x)J
=(x—3)a(x)+[l+(x—3)(--x2—x+l)]b(x)
= (x — 3)a(x) + (—x3 + 2x2 + 4x — 2)b(x).
Dividing this last equation by 7 yields the desired expression for the I.

In §2.6, we proved that there exist infinitely many prime integers.

An analogous proof yields the corresponding result for irreducible (monic)
polynomials. More simply, for infinite fields, the polynomials x—a, for
a e F. are irreducible, monic, and infinite in number.
We conclude this section with a discussion of the number of roots of
a polynomial equation. Consider the ring D[x] of polynomials in one in-
determinate with coefficients in an integral domain D. (In subsequent
applications of this theorem we shall use a field F in place of the domain Li.)
An element a D is called a zero or root of the polynomial

p(x) = e D[x] if p(a) = = 0.

Theorem 2. If p(x) D[x] has degree ii, then p(x) has at most n roots
mD.

The proof is by induction on ii, using the following technical lemma.

Lemma. For a polynomial p(.v) c D[x], a is a root of p(v) (i.e.. p(a) 0)

if and only if in t[x].
134 Polynomials and Polynomial Rings chapter 5

Proof of Lemma. By the Division Algorithm for Polynomials (Proposition I

above), there exist unique polynomials q(x) and r(x) such that
p(x) = q(x)(x—a) + r(x)
where r(x) is a constant polynomial. Since the map D[x] —+ D given by
f(x)—'f(a), a€ D, isa ring homomorphism [cf. Exercise 10, §5.1],
p(a) = q(a)(a—a) + r(a) = r(a).
Thus p(a) 0 if and only if the constant polynomial r(x) = r(a) is the zero
polynomial: that is, if and only if(.v—a)jp(x).
Proof of Theorem. Observe that a linear polynomial equation
Cl X + C0 =
has at most one root in D. (The existence of a root depends upon whether
c0 in D.) Suppose, as the induction hypothesis, that every polynomial of
degree n— I has at most n— I roots in D. Consider an arbitrary polynomial
p(x) of degree n. If p(x) has no roots in D, the statement is satisfied, as
n > 0. Therefore consider a polynomial p(x) having at least one root in D,
say a. Then by the lemma, (x—a)(p(x). In particular,
p(x) = (x—a)q(x),
where the degree of q(x) is n— I. For any root c of p(x),
0 = p(c) = (c—a)q(c).
Since c—a and q(e) are elements of the integral domain D, either c—a 0
or q(c) = 0. By the induction hypothesis, q(x) has at most n— I roots. Hence,
p(x) has at most I +(n— I) = n roots.
The necessity of the hypothesis that the coefficients lie in an integral domain
is evidenced by the following example:
— [4) = [0] with coefficients in Z15
has roots [21, [7], (8], (13]. Examples for other polynomials over rings Z,,, m not
a prime, are easy to construct. The idempotents in Z. are the roots of the polynomial
equation
x2 — x = [0] with coefficients in Z,,.

Exercises

1. Prove that the ideal in Q [x] generated by x2 + 2 is maximal.

2. Find the greatest common divisor of polynomials p(x),q(x) in Q[x],
and express it as a linear combination of p(x) and q(x) with coefficients
in Q[x}, where:
a. p(x) = x3 ÷ 7x — 3 b. p(x) = x2 + x + I
q(x)=x1+5 q(x)=x44x3+x2+x+1
c. p(x) = x4 — 3x3 + 3x —
q(x) = x3 — 5x + 7.
§5.2 Divisibility and Factorization of Polynomials 135

3. Find the greatest common divisor of x6 — I and x3 — in F[x], F an

1

arbitrary field. Generalize this special result' for arbitrary exponents

m, n of x.

4. Find a polynomial q(x) EQ [xl for which

(x2+l)q(x) I (modx3+l).
5. Prove that the polynomial x4 + 2x2 + 2 a Q [x] is irreducible.
6. Let a and b be nonzero relatively prime polynomials in Q[x]. Prove that
the rational function 1/ab can be written as u/a+v/b for some poly-
nomials u,v EQ[x].
7. Suppose that 1(x) F[x], F a field. Prove that
1(x) = (x—a)q0(x)+f(a), for every aeF
where qa(x) C F[xJ. Thus, give an alternate proof that f(a) = 0 if and
only if(x—a)If(x).
8. Prove the Rational Root Theorem: If f(x) = a,x1 is a polynomial
with integral coefficients, prove that any rational root r/s, (r, s) = I, of
J(x) must be such that na0 and sIan.
9. Prove the Integral Root Theorem: Any rational root a of a monic poly-
nomial 1(x) Z [xl must be an integral divisor of the constant term
of f(x).
10. Consider polynomials f(x),g(x) a F[xJ whose degrees are at most n.
Prove that f(x) = g(x) if 1(a) = g(a) for n+ I distinct elements a E F.
11. Let a0,a,, be n+ I distinct elements of the field. Show that the
polynomial

f(x) = b1( (x_aJ)(af_aJ)_I)

i=O j=O, J#i
satisfies f(a,) b1, 0 i n, where the elements b0,b1, are in F
(cf. the Lagrange Interpolation Formula, §5.6].
12. In Z,,[x], p a prime number, consider the polynomials

a(x) = x3 + [7],,x — [3],,,

b(x) = x2 + [5],,.
a. For which p are a(x),b(x) relatively prime?
b. For which p is the greatest common divisor d(x) (a(x), b(x)) the
residue class of
q(x) = (x3+7x—3,x2+5)EQ[x]
taken modulo p?
13. Let a(x)= x3+[l]5x+[3]5 and b(x)= x24-[4]5 in Z5[x]. Find their
greatest common divisor and express it as a linear combination of a(x)
and b(x) with coefficients in Zs[x].
14. Prove that x3+[3]3x+[2]5 is irreducible in Z5[x].
15. Prove that x3 + [3]4x+ is reducible in Z4[x], Find all of its zeros
in Z4[x].
 Polynomials and Polynomial Rings cbapter 5

16. Determine whether the polynomial x3 — e Z11 [x] is irreducible.

17. Determine all roots of the following:
a. x2—[I]inZ12
b. x2+E1] in Z10.
18. Prove that the polynomial x3 — x + {2]3 is irreducible in Z3 [x].
19. Find all irreducible polynomials of degree less than 3 in Z3 [x].
20. Find all x e Z12 satisfying the equation
— {2]12x2 — [3]12x = [0J12.

Express the problem in terms of congruences of integers.

21. Consider x2 + 7x+ 8 e Z[x]. Find all primes p such that:

a. is reducible in
b. is irreducible in
22. Prove that the congruence x3 —9 0 (mod 31) has no integral solution.
23. Prove that the equation x2 = [1115 has precisely 4 zeros in Z15.
24. a. Find all polynomials z E Q[x] such that
(x2+1)z x (modx4—x+ 1).
b. Prove that this congruence is not solvable if the coefficients are taken
in F=
c. Prove that this congruence is solvable if the coefficients are taken in
F= 5.
25. Write out the proof of Theorem I.
26. Following the proof of Proposition 1, §2.6, prove that there exist
many irreducible polynomials.

§5.3 Residue Class Rings for Polynomials

We continue the discussion of analogues to the arithmetic of integers.

In what follows let in = ni(x) be a fixed element (of positive degree)
in the polynomial ring R = Ffx]. It will play the same role in the following
discussion that the integer in did in §2.8 and subsequent sections. For
notational simplicity we frequently F[xJ, etc.
If a, bare polynomials in R, then
/, a(modn,), in 0,

is to mean m (b — a), or b — a e (in), where (in) = mR is the ideal in R

generated by m. Also, (in) # R because degin I. Thus congruence here
has the same meaning as in §2.8.

in were to equal 0, then congruence modulo in would be equality in R.

If
If in or another nonzero constant, then all polynomials are congruent to each
= I
other. Thus, to avoid such singular cases, we assume once and for all that the
modulus in is a polynomial of positive degree.
§5.3 Residue Class Rings for Polynomials 137

All general statements of §2.8 carry over to the polynomial ring R.

We observe especially the following analogue of Property 5, §2.8.

Proposition 1. For polynomials a, b, c, in e R,

ac he (modm), (c,m) = I — a h (mod in).

Furthermore, if a is a prime residue modulo in, then so are the

polynomials a' = a+qm, q R. That is, (a'.iii) = I for all a' a (mod in).
If p is an irreducible polynomial, then all nonzero polynomials (including
the nonzero constants) of degree less than deg p are prime residues modulo p.
Thus, if deg.f< deg p. there exists a polynomial g such thatfg I (modp).
For a fixed modulus in the residue class modulo in (or coset modulo
in), denoted [a], of a polynomial a E R consists of all polynomials a' satisfying
a' a (modrn). That is, [a] a consequence of the
Division Algorithm for Polynomials, each residue class [a] contains a unique
polynomial r R which either is equal to 0, in which case [a] = [0] = rnR =
(in), or has degree less than degm. Note especially that the residue classes,
determined by the elements c, d of F, are distinct if and only if c and dare.
The results for integers in §2.8 through §2.10 hold inutatis niutandis
for polynomials. In particular. Proposition 2 is a special case of the result
in Exercise 3, §3.2. Its proof, which we omit, is an exact analogue of the
argument in §2.9 that Zm satisfies the ring axioms.

Proposition 2. The residue classes of elements of R = F[x] modulo in

constitute a ring = R/(rn).

Note that if in is a reducible polynomial then has zero divisors,

and hence is not an integral domain. We shall prove in §5.4 that Rm is a field
when m is an irreducible polynomial. By Exercise 15, §3.6, this is equivalent
to saying that (ni) is a maximal ideal in R for an irreducible polynomial m.
Let us consider the following examples of arithmetic operations in
Re,, or equivalently of the solution of congruences modulo m.

Example 1. Determination of the polynomialf(x) of smallest degree in x satisfying

the congruence
(x2—x+
x x — 3) I and
= 4(x_3)(x4_x3+x.._3)+4(_x3+2x2+4X_2)(x2_x+l)
from the example of §5.2. Then setting y +(--x3+2x2+4x—2) and taking all
congruences modulo x'—x2+x--3, we have

and y.(x2.—3).
138 Polynomials and Polynomial Rings chapter 5

Thus 1(x) [(—x3+2x2+4x—2)17J(x2—3)

Next, to reduce f(x) modulo x4 x2 + x — 3 we carry out the division

— x+ 2
x4—x2+x—3 12x+ 6
—x5 + 3x
2x4 + 6x3 — 7x2 — 15x + 6
2x4 —2x2+ 2x— 6
6x3 — 5x2 — 17x + 12.
So,

—x5 2x4 + 7x3 — 8x2 — 12x + 6

= (—x+2)(x4—x2+x—3) ÷ 6x3 — 5x2 — 17x+ 12.

Therefore, 1(x) [(6x3 — 5x2 — I 7x + 12)/7] (mod x4 — x2 + x —3),

wherethe cubic polynomial on the right side of this congruence is the polynomial of
minimal degree in x which solves the given congruence.

Example 2. Following the discussion in §3.4, we now determine orthogonal idem-

potents in the residue class ring Rm, where R Q[x] and in x2(x+ l)(x— 1).
The first idempotent is a solution e1 of the simultaneous system of con-
gruences
y I (modx2),
(*) y 0 (modx+ I),
y O(modx—l).
To determine e1, note first that the general solution of the last two congruences is
y = a(x±l)(x— 1) = x(x2— l), a eQ[x].
(Recall that (x+ l)(x— 1) divides y since (x+ (x— l)Iy, and (x+ Lx— 1) = I.)
Now it remains only to determine a eQ [x] for which
a(x2— 1) 1 (modx2) or —a I (modx2).
A particular solution is a = — I. Therefore e1 = — l(x2— I) = I —x2 solves the
system (*) and is the first of our desired set of orthogonal idempotents.
Next, we determine e2 from the system of congruences
w 0 (modx2),
w I (modx+ I),
w 0 (modx— I).
Again note that the general solution of the first and third congruences is w =
11x2(x—l)forfleQ[x]. Thus we need only find a polynomial flsatisfying
flx2(x— I) I (modx+ I).
 Residue Class Rings for Polynomials 139

Noting that x — I (modx+ I), we rewrite this congruence as

fl(—2) I (modx+l),
which has the solution fi = —4 e Q[x]. Hence e2 = — 4x2(x— 1).
The third idempotent e3 is a solution of the system of congruences

z 0 (modx2),
z 0 (modx+1),
z I (modx— I).
As before, we need only solve for y such that
yx2(x+l) I (modx—1).
Since x I (mod x— I), this congruence becomes
I (modx— 1),

which has the solution y = 4. Thus e3 = 4x2 (x + I).

As in §3.4 we easily verify that these polynomials ea,e2,e3 are orthogonal
idempotents in Rm because for I j
[e,J [ej] = [0] and [eg) [e1] = [e,J

and that [I] = [e1J + [e2] + [e3}

= [I—x2] — 4[x2(x— 1)14-

As for the ring Zm, the residue class ring Rm, = 1T17=1 1,
can be written as a direct sum of subrings (ideals) B7, isomorphic to
[see §3.5].
We have seen earlier [Exercise 3(c), §3.2] that the canonical or natural
projection mapping
2: R Rm, 2(f) = [f],
of each polynomial fe F[x] = R to its equivalence class [f] modulo the
given polynomial m is a surjective ring homomorphism. We consider now
the restriction of the mapping 2 defined on F[x] to the subset Fc F[x].
That is, we limit our attention to the effect of I on F, rather than on all of R.

Proposition 3. The restriction of the projection 2: Rm to F is a mono-

morphism of F into Rm.
Proof For c,de F, we have
A(c+d) = 2(c) + 1(d)
I
sum in F
I
sum in R,
and
2(c.d) = 2(c).2(d).
I I
product in F product in R,.,
 Polynomials and Polynomial Rings chapter 5

Further, 2 is a monomorphisni of F since 2(c) = [0] in Rm means

that c e (in). Because c a constant, and is divisible by the polynomial m
is
(of positive degree). it must necessarily equal 0. Therefore, ker2 equals the
zero element 0 of F. (An alternate argument for the injectivity is that ker2 is
an ideal in F. But since a field has only trivial ideals, kerA F implies
ker2 = (0).]
Thus identifying F with its image 2(F) under 2, we speak of F as a
subfield of Rm. In particular, if degrn = I, then
2(F) = {2(c):ceF} = Rm.

An observation important in the general theory of fields [see

Chapter 8] is that Rm can be considered as a vector space over F. To this end
we define the sum of the "vectors" [a] and [b] to be the sum [a+b] of the
cosets in Multiplication by scalars ce Fis defined by
c[a] = [c][a] = [ca].
The customary axioms for a vector space, §4.1, are easily verified.
Since each coset [a] E Rm can be represented uniquely by a poly-
nomial which either is 0 or has degree less than n = degni, it follows that the
cosets [1], [x]. [x2], ..., constitute a basis of Rm over its field of scalars
F [see §4.2]. For, if there were a proper dependence relation
= [0]
(that is, if not all coefficients were 0), then the residue class

would equal [0]. This would imply that

a0 + a1 x + + .V' E (in),
contradicting the fact that m is the monic polynomial of least degree it in (iii).
Hence dim1 R,,, = degm.

Exercises

1. Describe the cosets of F[x] modulo (x3 + I) when

a. F=Q b. F=Z2.
2. Prove that x2—2eZ[x] is an irreducible polynomial, but that
x2— [2] c Z2 [xJ is reducible. Hence conclude that if a polynomial is
irreducible in Z[x] it need not be irreducible in p a prime.
3. Prove that the congruence fg I (modp) always has a solution g for

1, p F[x], when p is an irreducible polynomial not dividing /.

4. a. Find nontrivial idempotent elements in the residue class ring
Q[x]/(x2(x2+ I)).
b. Show that Q[x)/(x(x+ l)(x2 + 3)) has no proper nilpotent elements.
 Residue Class Rings for Polynomials 141
§5.3

1)2)
5. a. Write the multiplicative unit of the ring Q[x]/(x(x+ l)(x— as
a sum of three distinct idempotent elements.
b. Write Q[x]f(ni) as the internal direct sum of three ideals and the
external direct sum of three rings when rn = x2(x+l)(x—l)EQ[x)
[cf. §3.5].
6. Find a polynomial ye Q[x] of degree less than 4 for which (x+ 1)y
x3+l (modx3+3x—l).
7. Find a polynomial ye Z3[x] for which (x2+(l]3)y [l]3 (modx3+
x+[l]3).
S. Find inverses of the residue classes of the polynomials x+ I and x2 + 3
in the residue class ring Q [x]/(x5 — 1).
9. Let F = Q(V7) where is a solution of the equation x3 = 7 in C.
Write (1 + (I — and (I + -' as polynomials in 1,
with coefficients in Q. [See §3.6, Example 6 for the definition of

10. Prove that the polynomial x2—2 is irreducible in and

11. Let F be a subfield of a field K, and the ring homomorphism of

F[x] K given by =f(a) for some fixed a e K. Prove that
either is 0 or is generated by a monic irreducible polynomial
ma(x) C Fix].
12. Prove Proposition 2.
13. Consider polynomials 1(x), ni(x) e F[x], where F is an arbitrary field
and deg'n I. Prove that there exist unique polynomials h0,h1 h, e
F[x] for which
a. degh, < degrn, 0 i <
b. 1=
(Cf. the n-adic expansion for integers at the end of §2.5.) The expression
in (b) is called the m-adic expansion off. Especially important is the case
where nz(x) is an irreducible polynomial p(x).
14. Determine the p-adic expansion of a polynomial f(x) e Q [x] for the
following polynomials p(x) and /(x).
a. p(x) = x + 2 b. p(x) = x2 + 2
1(x) = x3 + I f(x) = x4 + x.
15. Consider the polynomial p = p(x) = x— a in F[x]. Prove that the cosets
[1(x)] of are of the form

[wo+wt(x—a)+ +w,_1(x—aY1] =
i=O
where the coefficients w, e F are unique.

These expressions are the polynomial analogue of the n-adic expansion

introduced in §2.5. This type of analogy with truncated power series expansions
of holomorphic functions of a complex variable led Kurt Hensel (1861—1941)
to his theory of p-adic numbers, a theory which provides a powerful tool for
recent developments in the theory of algebraic numbers and algebraic func-
tions. Isaac Newton used such expansions in his studies of algebraic plane
curves.
142 Polynomials and Polynomial Rings chapter 5

16. Consider an irreducible polynomial p in F[x]. Suppose that a/if is a

proper fraction in F(x), where a E F[xJ. Prove that there exist unique
polynomials a0, . .., such that

where degas < degp, 0 I n.

17. Consider in = x2—[5)11 E Z11[x].
a. Prove that in is reducible.
b. Show that the residue class ring Z11 [x]/(m) is the (internal) direct
sum of two fields isomorphic to Z1

§5.4 Residue Class Fields of Irreducible Polynomials

We turn now to a topic of great significance in the theory of equations
and fields [see Chapter 8]. Recall that in §2.12 we found that the residue
class ring p a prime integer, is a field; that is, every nonzero residue class
has a multiplicative inverse. The initial discussion in this section is directed
toward a similar question. Specifically, if R is the ring of polynomials in x
over a field F and p is an irreducible polynomial, then the residue class ring
is a field. Moreover, Proposition 2 shows that, roughly speaking, for an
irreducible polynomial p E R, the residue class field contains a zero of p.

Proposition I. If p is an irreducible polynomial in R = F[x]. then is a

field.
Proof To prove that each [a] [0] e has a multiplicative inverse, pick
in [a] the representative a' having degree less than degp, as in §5.3. Then
(a', p) = 1, and a'b+pq = I for some polynomials b and q. Consequently
a'b I (modp), or
[a'][b] = [a][h] = [I],
as asserted. Thus is a field. It contains the field which is isomorphic
to F. (Here). is the natural projection of R onto given by ).(a) = [a] for
alt ae R. as in §5.3.)

Proposition 2 (Kronecker's Construction). If p e R = F[x] is an irreducible

polynomial, then the field contains a root of p.

The construction of roots of irreducible polynomials is due to

Leopold Kronecker. As in Proposition 3, §5.3, where we identified the field F
with A(F) we now associate to the polynomial p(x) R = F[x] a
unique polynomial p(t) e where t is another indeterminate over
We shall prove that the residue class (or coset) [x] e R,, is a zero of the
polynomial p(t) e First, if
p(x) = + + a1x + a0,
§5.4 Residue Class Fields of Irreducible Polynomials 143

definep(t) to be the polynomial

p(t) = +a11+a0.

Considering a1 as an element of Rn—that is. identifying ).(a1) = [a1] E with

a1 e R— we may view p(t) as an element of Second,

2(p(x)) = [p(x)J = + + a1 [x] + a0 = [0]

(Again we write a1 for 2(a1) = [a1].) Hence p(t) has a root in
namely [x]. Thus given an irreducible polynomial p R = F[x]. we have
constructed a field R in which p has a root.

Kronecker's construction is admittedly rather subtle. The following

examples should serve to illustrate the argument.

Example 1. Let Q and p = x2—a where a is not the square of a rational

number. Then is a field and dim0 = 2. Also note that [x}2 = [a] for
the residue class [x] E In other words, [xl satisfies the quadratic equation
=

,2_ [a] (l—[x])(I+[x])

inthe polynomial ring

Specifically, let a 2. Then p = x2 — 2 is an irreducible polynomial in
R = Q[x]. The map
2: Q[x] —, Q[x]/(x2—2) =
takes

f + x2"1 + + a1 x + a0 a Q[x]
to its equivalence class

1(f) = [I] = + [x2]' ' [x] -I- -I- a1 [x] + a0,

where 2(a,) = [a1] has been identified with a1 Q. In we have [x]2 = [2] = 2
since x2 2 (mod x1 —2). Hence

[/] = + '[x] + + a1[x] + a0,

and so can be expressed as b[x] + c with coefficients b.c aQ. Furthermore, extending
A to a mapping from R[i] Q[x,t] to by defining 2(t) = t, we obtain (again
using the fact that [x)2 = [2] in
)((2_2) = t2 — [x]2 — (t—[xI)(I+[x])
= (t—

where we conveniently write [x] as ., 2.

In §3.6 we noted that the set

(a4 b a,bEQ}
 Polynomials and Polynomial Rings chapter 5

is a field; it can be described as the field obtained by adjoining the element

Q. By this, we mean that Q
is the a subfield of F for any field F such that
QCF, .,i!eF.
The significant point to be made here is that the field is isomorphic
to the residue class field of Q[x) modulo a particular irreducible (over Q) poly-
nomial, namely x2 —2. The isomorphism

is given by
a + b[x] e R9.
This same polynomial, x2 —2, while irreducible over Q, factors over as
follows:
x2 — 2 (x+ 2)(x—

Example 2. As a second, but very similar example, let F = R and p be the irre-
ducible polynomial p = x2 + 1. Then the residue class field of F[x] modulo p is
isomorphic to C, which is often defined as
C = {a+bi:a.bcR},
where i2 = — 1, or i — The isomorphism
C
is given by
p(a+bi) = [a) + [b]lx] e R,.
As an historical note, it is interesting that Augustin-Louis Cauchy often viewed the
complex numbers algebraically as equivalence classes of real polynomials modulo
x2+l.

Example 3. Consider F = Z3 and m = x2 ÷ [2)3 x + [2)3. Then in is an irreducible

polynomial in F, because no residue class [a] of Z3, a e Z, satisfies [a]2 + [2] [a] +
[2) = [0]. Consequently the residue class ring Rm is a field. Since dimF Rm 2, we
see that Rm is a field of 32 = 9 elements.

We are now ready to prove the theorem that any polynomial in

F[xJ splits into linear factors in some sufficiently large field. This theorem
lays the foundations for the general theory of equations [see the Galois
theory in and 8.7]. The theorem of §5.2 states that a polynomial .1(x)
of degree n in F[x] has at most n roots in F. The following significant
theorem yields the existence of a field in which 1(x) has precisely a (not
necessarily distinct) roots.

Theorem. Consider F[x]. There exists a field fl and an

injective homomorphism p: F—÷f such that ;if(x) = e
§5.4 Residue Class Fields of Irreducible Polynomials 145

factors completely. That is,

= E Q,

where px = an indeterminate over

Proof: We proceed by induction on n, noting that there is nothing to prove
when n = I. As the induction hypothesis, suppose that for any polynomial
g(x) of degree rn — I with coefficients in any field K F there exists a field Q
and an injective homomorphism p': with the properties in the state-
ment of the theorem.
Now consider a polynomial f(x) e F[x] of degree in. If f(x) factors
completely in F[xJ, then nothing has to be proved, and we can take = F.
Therefore assume that f(x) has at least one nonlinear irreducible (over F)
factor (x). Applying the Kronecker construction of Proposition 2, we
obtain a root x off(x) in the field K1 = F[.vJ/(f1(x)), where = [x), the coset
of x with respect to the ideal (f1 (x)).
As in §5.3 let 2 denote the projection mapping to the congruence
classes: for notational simplicity we write if for 2(F) and l.a for 2(a), etc.
Sincef1 (ce) = 0 in K and K F, the image 2f(x) = E K[y],
where 1x = y, an indeterminate over K, has the factor (y —
)f(x) = (y—cc)g(y) e Kfy].
The polynomial g(y) has degree en— 1. Therefore by the induction
hypothesis there exists a field Q and an injective homomorphism ji': K—'Q
such that p'g(y) factors completely in
To complete the proof define /1: F—' by p = 4u' o 2. Then, with
= /I')'.
= p' 2(1(x))

= p' =

1=1

where = and the I1< rn. are the roots in of' g(y) E K[y].
it is customary to identify the isomorphic image pF of F with the
given field F. and to identify the indeterminate with x. Then the theorem
states that there exists a field Q F such thatf(x) factors completely in fl[x).
The Fundamental Theorem of Algebra, to be discussed in §9.9, says
that any polynomial with complex (and hence real, rational, or integral)
coefficients splits in the field of complex numbers C. But C will not be a useful
splitting field for our subsequent considerations as it has infinite dimension
(as a vector space) over Q, whereas the splitting field Q just constructed for
/(x) Q[x] is a finite dimensional vector space over Q. Furthermore the
theorem is applicable for fields of nonzero characteristic, while the
Fundamental Theorem of Algebra certainly is not.
146 Polynomials and Polynomial Rings chapter 5

Exercises

1. Consider I Z1
a. Prove that the polynomial x2 — [5] n F[x) is irreducible.
b. Furthermore, show that the residue class ring F[x]/(x2 — [5) 3)is a
field K having 132 = 169 elements.
2. Prove that the polynomial x2 + [1)7 E Z7 [x} is irreducible. Let t denote
the residue class of x in Z, [x]/(x2 + (1)7).
a. Express the elements [1]7/(t+[l]7) and ([2]7+[3)7t)([4]7+[6]71)
as polynomials of degree less than 2 in i with coefilcients in Z7.
b. Show that Z, [x]/(x2 + [1],) is a field of 49 elements.
3. Let R Q[x, y). Prove that the residue class ring R/A, where A =
(x2—2,y2+3), is a field.
4. Construct a splitting field (i.e., a field in which each polynomial splits into
a product of linear factors) of each of the polynomials f(x) E Q [x].
a. f(x) = —3 b. 1(x) = x3 — 2
c. f(x) = (x2—3)(x3—2) d. 1(x) x4+ 1
e. 1(x) x(x3 +7) 1. J(x) = (x3 + 7)3 (x4 + 1).
5. Determine the dimension of each of the splitting fields in Exercise 4
considered as a vector space over Q.
6. Prove the theorem of this section by successively constructing fields in
which irreducible factors of j'(x) split off at least one root. 1-lini: Apply
Kronecker's construction to an irreducible factor f1(x) of f(x) to obtain
a field K1 in which 1(x) has a root Then consider an irreducible factor,
if any, off(x1) E K1 [xii, where x1 is an indeterminate over K1.
7. Verify that the mapping p in Example I is an isomorphism.
8. a. Show that the ideal A = (y2 — x, x2 — v) REx] is not a prime ideal.
Find prime and maximal ideals in R[x} that contain A.
b. Repeat part (a) with A (r2—x— I,y—l—3x).
9. Prove that(i2—x,y—4x-f 3) isa maximal ideal in R[x].

§5.5 Roots of Polynomials

This section discusses the existence and properties of roots of poly-
nomials over a field in preparation for the field theory in Chapter 8. The formal
derivative of such polynomials, which involves congruences of polynomials,
is needed for this discussion.
Let F{x] be the polynomial ring in the indeterminate x with
coefficients in the field F. Further let t be an indeterminate over the ring
R = F[x]. The typical element of R[r] is

where R, 0 I n.
Such an element can be rewritten as
0 i n, 0 in
i,j
§5.5 Roots of 147

with coefficients ag F, here

= 1i1(x). 0 I n,
and in = Conversely, a polynomial in two indeterminates

= =

with k,(x) = can be considered as an element of R[r].

Now for (x) = e R consider f(x+ r) = v)'.
This element of REt] can be expanded, using the Binomial Theorem, in the
form
f(x+r) = f(x) + + T2f2(x) + +
where the polynomialsj(x) are uniquely determined by 1(x). Next consider
in R[rJ the principal ideal
(r2) tr2/z(x, r) : Ji(x, r) R[r]}.
Then f(.v+ t) fix) + rf1 (x) (mod r2).
The uniquely determined polynomial (x) will be seen to have the formal
properties of a derivative and henceforth will be denoted byf'(x). We call
f'(x) the formal derivative of .1(x).
To verify the (formal) properties of the derivative consider the fol-
lowing congruence for another polynomial g(x):
g(x+t) g(x) +g'(x)t (modx2).
Then, modulo r2,
(f+g)(x+r) (f+g)(x) + [(f+g)(x)]'t
1(x) + g(x) + [f(x)+g(x)]'r
and (f+g)(x+ t) = f(x+ r) + g(x + t)
[.f(x)+f'(x)r] + [g(x)+g'(x)rJ
[f(x) +g(x)] + {f'(x) +g'(x)] r.
Consequently [f(x)+g(x)]' z_—f'(x)+g'(x) since the coefficient of r in the
congruence is uniquely determined. Next. modulo r2,
(fg)(x+r) (fg)(x) + [(fg)(x)]'t
f(x)g(x) + [f(x)g(x)]'r
and (fg)(x+r) =f(x+r)g(x+r)
+f'(x) r] [g(x) +g'(x) tJ
f(x)g(x) + r.
Consequently [f(x)g (x)]' =j'(x)g(x) +f(x)g'(x).
Finally, c' = 0 for ce F, since e c+O.r (mod r2). Also (x")' =
n•x"', since according to the Binomial
Theorem. Here, as at the end denotes the n-fold sum of
 Polynomials and Polynomial Rings chapter 5

NOTE. If F has prime characteristic p and iii 0 (modp), then the derivative of
the nonconstant polynomial x" is 0.

Next we use the (formal) derivative as a tool to examine polynomials

for multiple roots. (Recall from Theorem 2 of §5.2 that a polynomial of
degree n has at most n distinct zeros in F.)
A zero c in a field K F of a polynomial f(x) E F[x] is said to be
Ix-fold, or to have multiplicity if
1(x) 0
(mod (x— C)2+ I) in K[x].
This is equivalent to saying that f(x) = (x—c)1g(x) for some g(x) E K[x]
relatively prime to x—c in K[x].

Proposition I. If c is an x-fold zero of f(x). then c is a zero of multiplicity

at least I of the derivativef'(x).
Proof Letf(x)=(x—crg(x). Then
•f'(x) = +
(x— (x— c)g'(x)J;
i.e., f'(x).
Note that (x—crlf(x) can happen if F has prime characteristic p.
Consider F= Z,, and f(x)= ae F. Then f'(x)=px"' =0, thus
since a = with CE Z, according to §2.12; i.e., (xe—a) =
(x —

PropositIon 2. An irreducible polynomial f(x) in F[x] can have multiple

roots in a field K containing F only if f'(x) = 0.
Proof. From the previous discussion if e K is a multiple zero of
f(x) e F[x], then ['(c) = 0. Therefore x—c divides the GCD (f(x),f'(x))
computed in F[x]. But since f(x) is assumed to be irreducible and since
deg.f' <degf, we have (f(x), f'(x)) I un/essf'(x) is the zero polynomial.
Consequently the assumption that c e K is a multiple zero of f(x) implies
necessarily thatf'(x) is the zero polynomial.

Now suppose an irreducible polynomialf(x) = a, x has multiple

zeros. Then its derivative

(*) f'(x) i•a1 x'

=
is the zero polynomial; hence each coefficient must be 0. Thus,
char F = 0, we must have a, = 0 for i> 0. This fact implies that there are no
nonconstant irreducible polynomialsf(x) in F[x] with multiple roots in any
field K F.
§5.5 Roots of Polynomials 149

However, if char F—p >0 [see §3.6], then the derivative of any
polynomial F[x], whose only nonzero coefficients are a1,
(modp), is zero. If an irreducible polynomialf(x) E F[x] has multiple roots,
its derivative, given in equation (*) above, must be identically zero; that is
f(x) must be of the form

f(x) = where a1 0 only if ph.

Such a polynomial can be expressed as a polynomial e or

g(y) e F[yj, where y = x". Specifically,

g(x") = =

Furthermore, g(y) e F[y] c F[x] is an irreducible polynomial in y

if f(x) is irreducible. It may happen that g(y) also has multiple zeros. If so,
repeating the preceding arguments we find

g(j') = V", ip = rn;

hence, with z = yP,

f(x) = g(.v") = =

Ultimately there exists an integer e I. called the exponent of inseparability,

such that

1(x) = h(.rpe) =
1=0
n.
with h(u) = d,u1 F[u] F[x], u=
1=0
such that h(u) is irreducible in F[u] and does not have multiple roots in
any field F. Such irreducible polynomials 1(x) are said to be inseparable
of reduced degree a" = a/pC; we call pe the degree of inseparability.
This definition of reduced degree n" does not imply that (a"', p) = 1.
(We shall consider inseparable polynomials further in §8.5.) A nonconstant
irreducible polynomial g(x) is said to be separable if g'(x) is not identically
zero.

Proposition 3. Suppose that F has characteristic p > 0 and that 1(x) is an

irreducible polynomial with multiple zeros. Then all zeros have the same
multiplicity pe, e I.
Proof. By Proposition 2, f(x) = = Iz(y), where y = and the
irreducible polynomial I'(y) FEy] has zeros with multiplicity I. Let
h(y) = in some field A F and f(x) = The
150 Polynomials and Polynomial Rings chapter 5

polynomials factor completely in some field according to the

theorem so that = 0 in K. Consequently
— = =

and in K[.v].

§5.6 The Interpolation Formula of Lagrange

In §5.2 we noted that a polynomial of degree ii with coefficients in an
integral domain D had at most n roots in D. Now conversely, given n distinct
elements a1 in D. we can find a polynomial ,n(x) E D[x] of degree n,
whose roots are precisely the given elements a-, I i n; namely.
,n(x) = (x—a1)(x—a2)..
More generally, we have the formula of Joseph Louis Lagrange (1736—1813
given below. While we shall give a straightforward proof, verification that
Lagrange's formula follows from the Chinese Remainder Theorem of §2.11
is left as an exercise in §5.7.

Lagrange's Interpolation Formula. Let a1 be distinct elements and

h1, any iz+1 elements of the field F. Then there exists a polynomial
m(x) e F[x] of degree at most n such that m(a1) = h..
Proof Consider /z,(x) = Then

hI(ak)=O ifk#i
#0 ifk=i.
Now
,,+1
i;i(x) = b-b1 (.'v) (a1)) —
1=1

has degree at most n since degh1(v) = n. Clearly in(a1) = b., I I n+ 1.

Alternatively we can set h(x) = (x—a1); then m(x) has the
expression
n+1
ni(x) = fi b,h(x) [/z'(a1)

For example, to determine the polynomial of least degree with rational

coefficients whose graph contains the points (I, 1), (2,4), (5, —5) in the cartesian
plane, we consider
h1(x) = (x—2)(x—5), h1(1) 4,

h2(x) = (x—l)(x—5), h2(2) = —3,

h3(x) = (x— l)(x—2), h3(5) = 12.

 The Interpolation Formu)a of Lagrange 151

Thus
,n(x) =

—4x2 + '?x 5.

A modification of Lagrange's interpolation Formula yields a means

of constructing idempotents in certain residue class rings.

Proposition. Let h(x) = (x—aj, where the a are distinct elements of

a field F. Any polynomial k (x) e F[x] of degree less than n satisfies
Ik(a1flF /z(x)
k(v)
=
J'roof. Set

(x — a1), where j 1,
=

and q(x) k(a1)I,1(x)(h1(a1))

=
Since q(a1) = k(a,), I i n, the polynomial q(x)—k(x) of degree n—i has
n zeros. Therefore it must be the zero polynomial, i.e., q(x) = k(x), which
proves k (x) has the asserted form.
Fork(x)= I, we have

/z(x) (x — at)] —' = g.(x).

=
so that 0 (modh(x)) for i j,
and g(x)•l
g12(x) + g1(x)g1(x)
j*i

g.2(x) (modh(v)).
The residue classes of the polynomials modulo h(x) form a system of
orthogonal idempotents. whose sum is c':i. in the residue class ring
Ffx]/(/i(x)) [cf. and 5.3]:
[g,(x)J2 =
[g1(x)][g1(x)] = [0] if I j,
and [g1(x)] + + = [I).
This modified argument will be used to prove the Theorem of the
Normal Basis for algebraic field extensions in §9.5.
152 Polynomials and Polynomial Rings chapter 5

§5.7 PoJynomial Functions

We now relate the formal polynomials (infinite tuples) with which
we have been working to the generalization of the polynomial functions of
analysis.
Consider f(x) e F[x] and c F. The mapping F[x] —' F, given
by is a homomorphism. and is the principal ideal
(x—c). For the proof we have only to note that

f'(c) = and g(e) =

are added and multiplied as elements of F according to the same for,na/

rules used for sum and product in F[x].

Example. For
1(x) = a,x' and g(x) =
j=O
n+Ifl/
wehave f(x)g(4=
I
n+mf / it \ Im
and = a1b1)c' = ( a1e')
v0 \i+jv / / \j0
f(c)g(c) = pc(g(x)).

f
A polynomial function in one variable over F is a single-valued
function from F into F whose graph r1 in the cartesian product Fx F is given
by
= {(c,f(c)):ceF},
with f(c) = E7= 0a1 = f(x) e Ftx].
We define the sum and product of polynomia.l functions (and g as
follows:
f+g by = {(c,f(c)+g(c)):cEF}
and
fg by = {(c,f(c)g(c)):ceF}.
Using the preceding remarks on the computation off(c)+g(c) and J(c)9(c)
for each c e F, we can show that all polynomial functions form a ring D(F),
which is an integral domain.
Note that each polynomial f(x) = 7. 0a1 x e F[x] determines a
unique polynomial function f whose graph is

= {(c,J(c)): c F, where f(c) =

The mapping f(x)—+ —_f is a homomorphism of the ring F[x] onto

the ring of polynomial functions 1(F). The kernel of q, is therefore the set of
polynomials k(x) such that p(k(x)) = 0, i.e., Uk = {(c,0) : for all ce F).
§5.7 Polynomial Functions 153

Proposition I. If F is an infinite field, then and F[x] are isomorphic

integral domains.
Proof We have already noted that the above mapping q: is
a surjective homomorphism. It remains to prove that kerq = (0). Consider
an arbitrary polynomialj(x) in kerq; the graph r1 is the set {(c,O): ce F}.
In other wordsf(e) = 0 for all ce F. But since F has an infinite number of
elements and a nonzero polynomial has only a finite number of zeros, 1(x)
must be the zero polynomial. Hence = (0), as asserted.
If F is a finite field, then the collection of all distinct mappings
(functions) from F to F is finite. Accordingly the ring of polynomial
functions must be a finite set, while the ring F[x] of polynomials has an
infinite number of elements. Since q: maps an infinite set to
a finite one, kerqi {0}. (The explicit nature of kerq is considered in
Exercise 7, §8.2.)

Proposition 2 (Wilson's Theorem). For every prime number p,

(p—I)! —I (modp).

In §2.12 we proved that the p—I nonzero cosets [a] in the field Z,,
satisfy =[IJ. Thus they are the p—I distinct zeros of in
Z,, [x]. Consequently
—[I] =
or
x —+ [0], we obtain

— [I]
or —I E(—l)•..(—p+l)(modp),
— l (— I)! (rnodp).
Hence for odd p, (p—I)! —I (modp): for p = 2, then 1! — I I (mod 2).

This theorem, due to John Wilson (1741—1793), was published in 1770 by

his teacher, Edward Waring (1734—1793).

Exercises

1. Prove that the ring R of all polynomial functions of one variable with
coefficients inis not isomorphic to (Show that there exists an
epimorphism of onto R.)
2. Prove that the ring R in Exercise I is isomorphic to the residue class ring
Z,, — x).
 Polynomials and Polynomial Rings chapter 5

3. Consider a prime number p of the form p = 4,, + I, n> 0. Prove that the
congruence x2 — (modp) has a solution. (Hint: Investigate x =
1

[(p— 1)121!.)
4. Find a polynomial p(x)€Q(x] for which:
a. p(l) = 2, p(2) = 4, p(3) = 8, p(4) = 16
b. p(3) = 5, p(O) = 0, p(4) =
c. p(8) = —3, p(—3) = 8.
5. Repeat Exercise 4 when the coefficient field is Z,; that is, replace 2 by
[2] etc.
6. Let f(x) be a polynomial function defined on Z,,. Show that there exists
a polynomial function g(x) of degree less than p such that 1(a) = g(a)
for all aE Z,,.
7. Derive Lagrange's Interpolation Formula of §5.6 from the Chinese
Remainder Theorem, §2.11. Hint: Since the elements at of the statement
of Lagrange's formula are distinct, the Chinese Remainder Theorem,
applied to elements of F[x], states that the n congruences m(x)
(mod x — a1) can be solved simultaneously.

§5.8 Primitive and Irreducible Polynomials

We consider now' polynomials with integral coefficients considered
as elements in both Z[vJ and Q[x]. The discussion begins with primitive
polynomials and the important Lemma of Gauss.
A polynomial E Z[x] is called a primitive polynomial if
the greatest common divisor (c0, c1 is 1.

Lemma of Gauss. The product of primitive polynomials is primitive.

Proof. Consider primitive polynomials

u (x) and r (x) x',

= =
and let
m+n
u(x)t'(x) = cix", with Ch =
h0
Suppose now that u(x)L'(x) is not primitive. Then there exists a prime p
such that Ch_=O (modp), By assumption, not all of the
coefficients of u(x) and v(x)are divisible byp. Consequently there is a smallest
subscript I such that p a1 but a smallest subscript j
such that but Ok <f. Hence, by the formula for Ch with
h 1+], the product
=
is divisible by p. This contradicts the fact that both a1 and are not divisible
byp. Thus the coefficients of u(x)v(x) must have GCD I.
The following theorem makes use of the fact that for any
f(x) Q[x] there exists a rational number such that rxf(x) is a primitive
 Primitive and Irreducible Polynomials 155

polynomial in Z[x]. For the proof simply writef(v) e Q[x] of degree n as

f(x) (s1/1,) x', with Si, 1, Z, (s1, — 1.

=
Let c be the least common multiple of Then

cf(x) = e Z[x].

Now let d be the GCD of the integers csjt1, 0 I n. Then the coefficients of
that is. the elements d'(es1Jtj, have GCD I. Thus
is a primitive polynomial in Z[x].

Theorem 1. If a primitive nonconstant polynomial f(x) in Zfx] is

reducible in Q[x], then it is also reducible in Z[x].
Proof. It must be proved that a factorization f(x) =g(x)h(x) with non-
constant polynomials g(x),h(x) in Q[x] leads to a factorization f(x) =
g*(x)h*(x) with polynomials g*(x),h*(x) in Z[x]. By the preceding argu-
ment there exist rational numbers and $ such that the polynomials =
and h*(x) = 13h(x) belong to Z[x] and are primitive.
Next f(x) = g(x)h(x) implies that cxflf(x) = e Z[x].
Since f(x) is primitive the product afi must be an integer. The Lemma of
Gauss states that is primitive. Hence ccfl is a unit of 1, i.e.,
= ± I. Consequently
f(x) = g*(x)h*(x), where g*(v) =

Corollary. The preceding results remain valid if Z is replaced by a polynomial

ring F[t], F a field. The field of rational numbers Q is then replaced by the
quotient field Q(F[t]), i.e., the field F(i) of all rational functions a(t)/b(t)
with a(t),b(t),&O in F[t] [see §5.1]. The units ±1 of Z are replaced by
the units of F[t], i.e., the nonzero constants of F.

NOTE. The factors g*(x) and h*(x) are unique to within units of Z and F[t],
respectively.

A principal application of Theorem 1 is to show that irreducibility

of a polynomial in Z[xJ implies its irreducibility in the larger ring Q[x],
a result due to Ferdinand Gotthold Eisenstein (1823—1852). The following
theorem gives an important test for irreducibility.

Theorem 2 (Eisenstein's Criterion). A polynomial f(x) = of degree

n I in Z[x] is irreducible in Q[x] if there exists a prime p Z such that
0 (modp)
forOi.<n
a0 0 (mod p2 ).
 Polynomials and Polynomial Rings chapter 5

Proof We assume without loss of generality that /'(x) is primitive. By

Theorem I it suffices to show that a factorizationf(x)=g(x)h(x) in Z[x],
where g(x) and h(x) are not constants, leads to a contradiction. Let
and I:(x) = + + C0,

with integral coefficients c0. Note that 0. Moreover because p

does not divide it does not divide c,,. Since p divides a0=b0c0,
but p2 does not divide a0, one, but not both of b0, c0 must be divisible by p.
Suppose that p divides c0 but not b0.
Next, there is a coefficient c of
= (p + (p — — + + +
farthest to the right that is not divisible by p. Since degh(x) < n, necessarily
i n. Also, i> 0. since we assumed that p I co. Now, p cannot divide
(*) = b0c, + b1c1..., + + b_,c, + b.c0
because p does not divide but it does divide all other terms on the
right-hand side of equation (*) by the choice of I. Since the only coefficient
a of f(x) not divisible by p is and yet for some i< n if a proper
factorization is possible, we have reached a contradiction. Hence f(x) must
be irreducible, as asserted.

Corollary. Theorem 2 remains valid if Z is replaced by F[t] and Q by F(z)

and the primes p e Z are replaced by irreducible polynomials p(t) F[t]
for an arbitrary field F.

Example I. Let p be a prime; then x"—p is irreducible in Q[x].

Example 2. Suppose that F is a field of characteristic p> 0. Let K be the field of

rational functions F(t); then the polynomial f—t is irreducible in K[x].

Exercises

1. Prove is irreducible in Q[x]

if there exists a prime p such that [1(x)] = [ao]f+ + + [an],
[a,] e 0 i n, is irreducible in
2. Prove that the following polynomials are irreducible in Q [x].
a. 3x3—4x2-i-2x—2
b. x4—2x+2
c. x4+3x2+3.
3. Verify that the polynomial in Exercise 2(c) is congruent to
(x—1)(x-i-l)(x2—3) modulo7.
4. Assume that x — a, a e Q, is a divisor of a monic polynomial f(x) e Z [x].
Prove that a a Z.
§5.9 Characteristic Polynomials of Matrices 157

5. Prove that the polynomial is irreducible in Q[xj for all

p> 5. What about p = 2,3?
6. Prove that 1(x) = x5+7x2+ I E Z[x] is irreducible in Q[x]. (Hint:
Reduce Z and 1(x) modulo 2 and examine for reducibility in Z2 [x].)
7. If 1(x) is a monic polynomial in Z[x] for which f(l) = p and if
f(x+ 1) = xs+pg(x) for someg(x) a Z[x], prove thatf(x) is irreducible.
8. a. Prove that 1(x) = + ... + x+ I a Z[x], p a prime, is irreducible
in Z[x]. [Hint: Examine f(x+ I).] Such polynomials are called
cyclotomic [see §9.1].
Let 1(x) = (xe'—
b. — I), p a prime. Noting that f(x) =
yPl + + I, where y = prove that f(x) is irreducible
in Q[x]. [Hint: Examine f(x+ 1).)
9. Prove that the following polynomials are irreducible in F= Z3(t).
a. x3+t3x2—t b. x3+tx—t5.
10. Write out Theorem 2 and its proof with the replacements cited in the
corollary.
11. Prove that
j I,
1(x) = a1 +p a1 x"' E Z [x],
1=0

where 0 j < n, p a prime, has at least one irreducible divisor (in Z [x]),
whose degree is at least n—j, provided that 0 (modp) and at least
one a1 0 (modp), where 0 I
12. Let p be a prime, and suppose that the polynomial 1(x) a Z [xl has a
factorization f(x) go(x)h0(x) (modp), where (go(x),ho(x)) E I
(modp). Prove that f(x) gk(x)hk(x) (modpk), where gk(x),hk(x),
taken modulo p. equal ,q0(x),h0(x), respectively. [Hint: Use induction
on k, letting g*(x) = go(x)+pu1(x)+ etc.)

§5.9 Characteristic Polynomials of Matrices

This section concludes our discussion of polynomials and continues
the study of linear algebra from Chapter 4. We develop the properties of the
so-called characteristic polynomial of a linear transformation of an
n-dimensional vector space V/F to itself, culminating in the celebrated
Cayley-Hamilton Theorem.
Although we shall not do so here, a rich interplay can be developed
between the factorization into powers of irreducible factors of the character-
istic polynomial of a matrix A and the direct sum decomposition of a vector
space into certain subspaces invariant under the linear transformation
described by A. Here a subspace Wc V is called invariant under ço if
ip(W) c W. (See, for example, F. R. Gantmacher, The Theory of Matrices,
Vol. 1.) Of primary importance for work in §9.4 are the concepts of trace
and determinant of a linear transformation q, both of which can be derived
from the characteristic polynomial of any matrix representing q.
158 PolynomIals and Polynomial Rings chapter 5

Consider an ii x matrix A = [an] e

ii {cF. §4.4]. The characteristic
polynomial XA (x) e F[v] of the matrix A is defined to be
x4(x) =
where is the ii x n identity matrix.

NOTE. Here we need to generalize the concept of determinant from §4.5 slightly to
include determinants of matrices with coefficients in a ring. All the properties of
determinants cited in §4.5, except Property 8 which involves inverses, remain valid
if the field of coefficients F is replaced by a commutative ring, such as F[x]. [Note
that the inverse of an n x n matrix B (for example, x A) with polynomial coef-
ficients will in general be a matrix with coefficients in the field F(x) of rational
functions in x over F.]

Using the Laplace expansion for determinants (Property 5, §4.5),

we express the characteristic polynomial in terms of the powers of x as
follows:

(x) = au) x" -' + ... + (— I)" det (A).

—

Example 1. Consider an arbitrary 2 x 2 matrix A = [a11J. Then,

XA(X) det(x.12—A)
x—
det[
—a21 x—a22

— (a11+a22)x + a11a22 — 012021.

I 4 5
Example 2. The characteristic polynomial of A = 0 —1 0 is
4 8 3

x—l —4 —5
XA(x)=det 0 x+l 0
—4 —8 x—3
Expanding by minors of the first column, we obtain

xA(x)=(x—l)
x+l 0
—4
—4 —5
—8 x--3 x+I 0
= (x—l)(x+l)(x—3)—20(x+l)
= x3 — 3x2 — 21x — 17.

The sum of the diagonal elements a., of the matrix A, which

is —l times the coefficient of the term of the characteristic polynomial.
is called the trace T(A) of the matrix A. Proposition I states simple properties
of the trace.
§5.9 Characteristic Polynomials of Matrices 159

Proposition 1. For ii x n matrices A, Be F,,,, and e e F, the trace function

satisfies the following properties:
(i) T(A+B) T(A)+ T(B).
(ii) T(cA) = cT(A).
(iii) T(cI,,) = ne.
(ii) T(AB) = T(BA).
The proofs follow immediately from the definition. Considering F
as a one-dimensional vector space over itself, we obtain a second proposition.

Proposition 2. The trace function T is a linear transformation in

HomF(Fflfl, F).

Suppose now that PE F,,,, is a nonsingular matrix. That is,

det(P) 0 and P ' exists. We then have the following proposition.

Proposition 3. For matrices A, C e F,,,, satisfying C = P - 'AP,

x4(x) = Xc(x).
Proof By the definition of the characteristic polynomial and the properties
of determinants,
Xc(X) = det(x.!,,—P'AP)
= det(x.I,,—C)
= =

= det(!,,) det(x.l,,—A) = det(x.I,,—A) = XA(X).

Since both the trace and determinant of the matrices A and

C = P - 1AP occur as coefficients in the characteristic polynomial, Proposition
3 has the following corollary.

Corollary. T(P'AP) =
det(P'AP) = det(A).

Now consider a linear transformation p of an n-dimensional vector

space V/F. Consider further two bases {A1, .., A,,} and {B1 of V/F
and the corresponding matrix representations M, and N4, of as in §4.4.
Proposition 4, §4.4, states that for some nonsingular n x n matrix P
N4, = P'M4,P.
By Proposition 3 the two matrices Mc,, and N4, representing the linear trans-
formation have the same characteristic polynomials. Thus, to q E
we can associate the uniquely determined polynomial
x4,(x) = E F[x],
 Polynomials and Polynomial Rings chapter 5

called the characteristic polynomial of p. (It is the characteristic polynomial

of any matrix A representing (p.)
Further, the trace T((p) of(p End4,(V) is defined to be — I times the
coefficient of x" ' of and det((p) to be (— times the constant term
of where n = dim V. In other words, T(q') and det(q,) are the trace
and determinant, respectively, of some (and hence any) matrix representing (p.
Both are uniquely determined by (p.
Proposition 4 is a consequence of the properties of matrix represen-
tations of linear transformations (especially Proposition 3, §4.4) and of the
trace and determinant of matrices.

Proposition 4. For c e F and E EndF(V), where n = dim V.

= T((p) +
T(cp) =
and det(q o = det((p)
det(c(p) = det(q).

We conclude this section with a proof of the Cayley-Hamilton

Theorem. For linear transformations this theorem is expressed as follows.

Theorem. Let be the characteristic polynomial of the linear trans-

formation Then in
= (pfl_ + = 0,

where is the identity map on V and 0 denotes the zero map.

Since x,(x) is defined by means of any representing matrix A = M,,

it suffices to prove the classical statement of the Cayley-Hamilton Theorem:
any matrix is a root of its characteristic polynomial.

Cayley-Hamilton Theorem. Let (x) be the characteristic polynomial of

then x n matrix A. Then in
= + = 0.

Proof. The elements of the matrix are linear and constant

functions of x; thus the elements of B=Adj(x.11,—A) are polynomials of
degree less than n in x. [See §4.5 for the definition of the adjoint of a matrix.]
Therefore B has the expression:

(s) B= where B E 0 i < n.

From equation (*) we have the identity
(ss) =
§5.9 Characteristic Polynomials of Matrices 161

For simplicity let

x4(x) = x" +
a E F. 0 i < n. Then using the expression (*) for B, we
can rewrite equation (**):
n—I in—I n—I
—AB0+
i=O i=l

where a,, = I. Comparing coefficients of like powers of x yields

—AB0 = a0!,,,

B0 — AB1 = a1 I,,,

(***)
B,,2 — AB,,_1 =
= 1,,.
Multiplying the equalities in (***) on the left by !,,.A respect-
ively, and adding we obtain
0 = a0 I,, + a1 A + + a,, ' + =
This observation completes the proof.

Exercises

1. Prove that a matrix A, whose coefficients are polynomials over a field F,

has an inverse in F[x} if and only if det(A) is a nonzero constant.
2. Determine the characteristic polynomials of the following matrices.

1121 b.
l —l
0
0

40 j
1

1 1
1

2 —1 01 —l
c. 0 0 21 d. —I 2 0.
—2 1 lJ 8 5 0
3. Verify that the matrix in Exercise 2(a) satisfies its characteristic polynomial.
4. Determine the trace and the determinant of each of the matrices in
Exercise 2.
5. Verify that the matrix of Example 2 satisfies its characteristic polynomial.
 6

Group Theory

A fundamental topic in abstract algebra is that of groups. Our

presentation intersperses many examples with the axiomatic discussion.
Thus an extensive list of examples follows the definition of a group
The basic properties of groups, subgroups, and factor groups I through
6.4] are presented with illustrations. In §6.5 worked-out examples of homo-
morphisms of groups show that the concept of homomorphism is a more
precise tool in studying groups than is that of normal subgroup. (Distinct
homomorphisms may define the same normal subgroup, but distinct normal
subgroups must always correspond to distinct homomorphisms.) Nevertheless
we first use normal subgroups to define the projection homomorphisms onto
the corresponding factor groups. This approach parallels our initial emphasis
on extensive study of residue class rings of integers before introducing the
concept of homomorphism of commutative rings.
Throughout this chapter our aim is that the reader learn to use
explicit computations, no matter how tedious they may appear at first glance,
to isolate the essential meaning of the basic definitions and elementary
theorems of group theory. Such work is necessary so the reader can progress
beyond a superficial absorption of the basic concepts of group theory. Many
of the problems can be solved by "the method of ingenuity" as opposed to
"the method of infinite drudgery." The latter method need not be feared,
as it often will suggest the better approaches and aid understanding.

162
 Elements of Group Theory 163

Special properties of two families of groups—cyclic and

permutation an introduction of structure-preserving map-
pings (homomorphisms) of groups in §6.5; homomorphism properties are
examined in detail in §*6.8 and 6.9. The chapter concludes with an optional
discussion of homomorphisms of direct products of abelian groups, presented
both theoretically and in terms of explicit determination of homomorphisms
defined on the direct product of cyclic groups. The direct products of groups
are prerequisite for the Fundamental Theorem of Finitely
Generated Abelian Groups to be discussed in §7.1, the construction of new
groups, and the analysis of given groups. The student has the choice of fol-
lowing this chapter with additional topics in group theory [Chapter 7] or
with the study of field theory [Chapter 8], which applies the group theory
already studied in developing the Galois theory and the theory for solving
polynomial equations by radicals

§6.1 Elements of Group Theory

A group G is a nonempty set of elements G = {a, b, c, ... } together
with a law of composition or binary operation (that is, a single-valued
mapping), associating to each ordered pair (a, b) in the cartesian product
G x G = {(a, b) : a, b G} a unique element ab = d G, which has the
following properties:

Associativity. a(bc) = (ab) c, for all a. b. c e G.

Identity (existence of an identity). There exists an element e E G such
that eg=ge=g, for allgeG.
Inverse. For each 9€ G, there exists a g' e G, denoted such that
gg'
A group G is said to be commutative or abehan (after the Norwegian
mathematician Niels Henrik Abel, 1802—I 829) if ab = ba for all a, b G.
This is a special property of some groups, and so must be treated separately.
Comparison of the defining properties of groups with those of rings
is invited. By proofs essentially the same as those for integers
and rings we obtain the uniqueness of the identity element and of the inverse
ofeachgeG.
Proposition 1. The identity or Unit element e in a group G is unique. For
eachge G, its inverseg' is unique.
Proof As is customary in uniqueness proofs, suppose that both e,e' have
the properties of an identity element. Then
e= ee', as g = ge', for all g:
e'. as eg = g, for all g.
164 Group Theory chapter 6

g*g = e and gg' = e, then

Similarly, if for a given 9. g'g
9* = g'e = g'.
Proposition 2. Equations of the form
xa=b, a;=b
for arbitrary a, b e G have unique solutions x, y e G.

In each case the proof is an immediate consequence of the existence

of the (unique) inverse of a.
Corollary. The law of cancellation holds in a group. That is, for elements
a,h, ce G,
ab = eb a = c.
REMARK. Authors of algebra texts differ in the listing of axiomatic properties for
the definition of a group. The most common variation is to require the "solvability"
property of Proposition 2 in place of the Identity and Inverse axioms. The two sets
of axioms are seen to be equivalent. Proposition 2 shows that Identity and Inverse
imply Solvability. Solution of ax = a, ya a yields the identity e, and solution of
ax = e, ya = e yields the inverse of a.
Systems of axiomatic properties which may not be so obviously equivalent
can be given. For example, if G has a left identity (eg = g. for all g e G) and each
g e G has a left inverse g' (g'y = e), then we can prove ge = g, and gg' = e. That is,
e is the identity element and g' the inverse of g, as given in our axiomatic description
of a group. No advantage in group theory, theoretic or practical, accrues to such
sets of axiomatic properties.

The general associative law, which states that for any n> 3 elements
e G all products of the n elements, keeping of course their given
order, are equal, is not a group postulate. Rather it is a consequence of the
associativity postulate that 91 (92 93) (g1 g2)g3 for any three elements
E G. The emphasis in the general associative law is on the term
"any n elements g, G." Thus the Principle of induction in one of its forms
is required for the proof.
Define successively to be Thus
= 9i 92. p1 = = 92)93. etc. Now the general
associative law will be proved, if we can demonstrate that every combination
of g, g,, put together in product form, keeping fixed the order of the
elements g,, equals fl7.,
The proof requires induction on n. For n= 1,2 nothing is to be
proved. For n 3. we have simply the given associative law. Assume now
that for all products of Ii. I /i <,z elements, the arrangement of parentheses
in products is immaterial. This means, to be more explicit in a special case,
)2)[t33 = [.v1 (..l2 = ..., for I: = 5.
Thus assume that any ordered product of' I, < elements 9h (a/ways
in this order) equals
 Elements of Group Theory 165

Consequently an arbitrarily bracketed product of n elements

will, by the induction hypothesis, have the form

ci =
where I s <n. The induction hypothesis implies that

g = (g1"g5)flg5+4

Fin—s—I \ 1
= (cii g5)I [1 cis+i)cinj by the defInition of 11
L i=1

I n—s—i 1
=
L
H
1=1 J

Using the induction hypothesis again, we have

n—i—I n—i
= i=I g1,
4=1

'n—I n
so that ci = (n gi),qn = cii.

Numerous examples of different kinds of groups conclude this

section. Some are geometrically derived and others are abstractly presented.
On a first reading, the student should not attempt to master all of these
examples. While studying later sections on group theory, the reader is urged
to return to this section to review appropriate examples and to work out the
details of group-theoretic arguments in terms of explicit computations with
elements of particular groups.

Examples of Groups
1. a. The elements of any ring R with respect to the additive law of composition

b. The elements of the particular rings Z, Q, R, C, and Zm with addition as

the law of composition.
2. a. The units of any ring R with respect to the multiplicative law of composition.
b. The nonzero elements of Q, R, and C, denoted Q*, R*, and C, respectively,
with multiplication as the law of composition.
c. The set of prime residue classes [a] of Zm with multiplication as the law of
composition. In particular, the p— I nonzero cosets of Z,,, p a prime
number.
 Group Theory chapter 6

3. a. The set Tof all complex numbers with absolute value = I, with mul-
tiplication as the law of composition. Such complex numbers can be
expressed in the form cosq,+isinp, — I.

b. The set of all complex numbers of the form

\12,r \ 0 r < in,
cosl—rl+ islnl—rl,
\nr I /
for a given nonzero integer in with multiplication as the law of composition.
4. a. The set of all nonsingular ii x ii matrices ith coefficients in Q, R, C, or Z,,
(p a prime number) with matrix multiplication for the law of composition.
b. The set of all a x n matrices with determinant ± I, with coefficients and
the law of composition as given in part (a).
c. The set of all 2 x 2 matrices with determinant I, integral coefficients, and
matrix multiplication as the law of composition.
For the particular matrices

01]
I
II T=
—10
0 I
,
w=rs=l 0
[--I
V
--I
0 S" =
T4 = W3 =
note that
01] 1 , and that
0 I

for every a Z. Thus the element S = T 13 has infinite order (defined in

§6.2) even though it is in the product of elements T and W of finite order.
Furthermore, (TS)3 = I.
d. The set of matrices
ab
C (I

of the group in part (c) above for which a a' ± I (mod a) and b c 0
(modn), a> I.
e. The set of all 3 x 3 matrices
1 ii
0 I i*'

001
with coefficients in Z, Z,,,, Q, R, or C, with matrix multiplication as the
law of composition.
5. The Cartesian product A x R' of two rings R, R' with the rule of addition
defined by
(a,a') + (b,b') = (a+b,a'+b') for all a,b€ A and a',b' c R'.
a. In particular, we may choose Z, Q, R. and C as the rings R and R'.
b. The Cartesian product Z,, X Zn, with

(('tim, + ([cJ,,,, [d],,) =

Note that this group has nsa elements.
§6.1 Elements of Group Theory 167

c. The set of pairs (a, fi), where a belongs to Urn, the multiplicative group of
prime residues modulo in, and fi belongs to 4
considered as an additive
group, with the law of composition defined by
(a,fl) * (y,ô) =
6. We now introduce "abstract" groups defined formally in terms of generators
and relations.
a. Let G be the set {am: in a Z} of a//integral powers of the symbol a, which
we call the (multiplicative) group generated by a. The element a is then
called a generator of G. The group operation is denoted by multiplication
with the customary law of exponents = etc., and a° is the
multiplicative identity.
b. Let G be the set {am : in a Z} of all integral powers of the symbol a, but
now define a', to represent the same element of the group if i j (mod n)
for some fixed it a N. Thus, the distinct elements of G are {a°,a', ...,a" '};
any other power of a represents the same element in 6 as one of these.

The groups in Examples 6(a) and (b) are called cyclic groups. They are
introduced again in §6.5 and studied in detail in §6.6. They are generated by a single
element a.
In Examples 6(c) through 6(i), we consider "abstract" groups with two
generators a,b. Let S be the set of all finite products of the symbols a,b,a ',b L,
and e, where a° b° = e, the multiplicative identity, and a'a' = etc. The
identifications of symbols in the set S. given in Examples 6(d) through (i), are
sometimes referred to as defining relations.

C. The set S is a group with an infinite number of elements.

d. From the set of symbols S we can obtain a group by making the following
definitions:
(i) ab and ba are equal (that is, they represent the same element of the
group).
(ii) For any fixed positive prime p and positive integer n, and a*bm
are the same element of the group if i k and I m
(modp).
Thus, we equate ar'' and with the identity element e = a° b°. For
example, when it p = 3, the distinct elements of the group are
e, a, a2, a3, a4, a5, a6, a7, a8,

b, ab, a2b, a3h, a4b, a5b, a6b, a'b, a8b,

b2, ab2, a2b2, a3b2, a4b2, a5b2, a6b2, a7b2, a8b2.

Any product of powers of a, b can be reduced to one of these 27 products

by the identifications cited above.
e. From the set of symbols S. we can obtain a group by making the following
definitions, for any fixed integer is 3.
(i) ba and
a' are same group element
and are the same group element.
 Group Theory chapter 6

When ii 3, this group is called the quaternion group; its distinct elements
are
a, a2 = b2, a3, a4 = e; b, ba, ba2, ba3.

For example. ab = a2ba b2ba = bb2a = ba3. The multiplication table

(that is, the array of all products of paIrs of distinct elements in a group)
for the quaternion group follows.

e a a2 a3 b ba ba2 ba3

e e a a2 a3 b ba ba2 ba3
a a a2 a3 e ba3 b ba ba2

a2 a2 a3 e a ba2 ba3 b ba
a3 a3 e a a2 ba ba2 ba3 b

b b ba ba2 ba3 a2 a3 e a

ba ba ba2 ba3 b a a2 a3 e

ba2 ba2 ba3 b ba e a a2 a3

ba3 ba3 b ba ba2 a3 e a a2

When n > 3, such groups are called generalized quaternion groups.

f. From the set of symbols S, we can obtain a group by making the following
definitions for any fixed integer ii
(I) ba and a 'b are the same group element.
(ii) a' and are the same group element if i j (mod ii).
(iii) bk and btm are the same group element if k in (mod 2).
Such groups, denoted are called dihedral; a geometric description of
them is given in Example 7 below. The group F', is noncommutative, and
has 2n (distinct) elements.
g. From the set of symbols S. we can obtain a group by making the following
definitions, for any fixed integer ii 4.
(i) ba and a' are the same group element.
(ii) a' and are the same group element if i j (mod 2' ').
(iii) and btm are the same group element if k in (mod 2).
h. From the set of symbols S. we can obtain a group by making the following
definitions, for any fixed integer n 4.
(i) ba and a2' - 'b are the same group element.
(ii) a' and are the same group element if i j (mod 2"-').
(iii) and b"' are the same group element if k in (mod 2).
These groups are sometimes called semidihedral.
From the set of symbols S, we can obtain a group by making the following
definitions, for any odd prime p.
(i) and are the same group element if I k (modp2) andj m
(modp).
 Elements of Group Theory 169

(ii) ba and a1 +P/, are the same group element.

See Example 2, §6.4, and Example 6, §6.9, for further discussion of this
noncOmmutatiVe group.

j. In a fashion analogous to the definition of the set of symbols Sand Example

6(c), we can define "abstract" groups.with any finite number of generators.

7. The following examples of groups are geometric descriptions of the dihedral

groups in Example 6(f). Here let a be rotation of a regular n-gon (an n-sided
polygon, all sides nonintersecting, of equal length, and inscribed in a circle)
through 2r/n radians, n 3, and let b be reflection about an arbitrary fixed axis
of symmetry.
a. The set f3 of maps (rotations and reflections) of an equilateral triangle to
itself is a group, with 6 elements, with composition of mappings for its
group operation.
b. The set r4 of maps (rotations and reflections) of a square to itself is called
the group of the square. Composition is the group operation; 1'4 has 8
elements.
c. More generally, the rotations and reflections of a regular n-gon constitute
a group of 2n elements.
8. For a prime p the set of all residue classes of q = p", whose representatives
a satisfy a I (modp) where the law of composition is multiplication in
9. The vectors in any vector space with the law of vector addition [cf. §4.1].
10. The set L of all linear transformations of a finite dimensional vector space V
to itself with zero kernel under composition of transformations [cf. §4.3].

There are several origins of the study of groups. Lagrange in the

eighteenth century was concerned with the behavior of permutations of roots
of polynomial equations. Hermann Weyl in the conclusion of his treatise
Symmetry draws the lesson that "Whenever you have to do with a structure-
endowed entity E try to determine its group of automorphisms, the group
of those element-wise transformations which leave all structural relations
undisturbed." Geometrically-minded persons, following Leibniz, would
speak of the group of symmetries, where Weyl uses the term "automorphisms."
Newton and Hermann von Helmhlotz (1821—1894) preferred describing the
structure of a space by the notion of congruence. wherein two parts of a
space are defined to be congruent if they can be occupied by the same rigid
body in two of its positions. (See H. Weyl, Mathematische Analyse des
Raumproblems.) Such concepts underlie the mathematical analysis of
symmetry in ornamental art (highly developed by the Arabians, Egyptians,
and ancient Greeks) and in nature in crystallography, cellular arrangements,
configurations of florets in flowers, and the like. "Symmetry is a vast subject,
significant in art and nature. Mathematics lies at its root, and it would be
Herman Weyl, Symmetry, p. 144. © 1952 by Princeton University Press, Princeton, N.J.
170 Group Theory chapter 6

hard to find a better one on which to demonstrate the working of the

mathematical intellect.'
Example 7 involves groups of symmetries of plane figures. In an
analogous fashion we can consider groups of the five regular (Platonic)
solids in 3-space.

Exercises

1. Let . ii an arbitrary positive integer, be elements of the group

G. Prove that
=
(Make use of the Principle of induction.)
2. In the definition of a group G replace the postulates about the identity
e and the existence of an inverse of an element a by the following:
(i) There exists an element e a G such that e,q = p for all p a G.
(ii) For every g G there exists an element G such that 9*9 = e.
Prove that such a system G is a group in the sense of the definition in
the text.
3. Suppose that G is a given nonempty set of elements on which an asso-
ciative product is defined. Assume that all equations xa = b and ay = b,
a and b in G, have solutions x and y in G. Prove that G is a group.
4. Define the powers of a group element p a G inductively: = e, the
identity element of G, p2 = g•p, p3 = g.gl Let g' denote the
inverse of p. and define p ', n a N, to be (p I
Prove the laws of
exponents:
= and (g')' =
for all u,v,r,sE Z.
5. Let G be the set of all real numbers a,b, ... whose absolute values are
less than 1. Define a law of composition * by the formula
a*b = (a+b)(l+abY1
where a+b and ab denote the customary sum and product of the real
numbers a and b.
a. Prove that G, with * for the law of composition, is a commutative
group.
b. What is the identity element of G and the inverse of a?
6. Let G be the set of functions t, 1, 1 ', (I — t) l, (t— 1)1_I, t(t— 1)1
1

defined on R.
a. Prove that G is a group with respect to the usual composition of
functions. That is, forf,ge G, (fogfls) =f(g(t)).
b. Which function is the identity element?
c. Find the inverse of each of the six functions.

Ibid., p. 145.
§6.2 Subgroups and Orders of Elements 171

7. a. Enumerate the elements of the group G = {ambn :ni, ii c Z} of all

formal products of symbols a,b subject to the identifications
a4 = b2 = e, bab ' = a3 [cf. Example 6(f)].
Construct the multiplication table of this group.
b.
c.Identify this group with the group of the square r4 in Example 7(b).
8. What can be said about the group in Example 6(h) when n = 3?
9. Verify that a vector space V is an abelian group with vector addition for
the law of composition.
10. Let Af(Zm) be the set of transformations
U —, + fi
where u is an indeterminate over Zm, is a unit in Z,,, and flE Zn,.
Prove that Af(Zm), called the affine group over Zn,, is a nonabelian.
group with the usual composition of mappings for its multiplicative law
of operation.
11. Consider the set of all formal products of the symbols a, b, c, a', b1,
c1, and e, where we identify a2 with aa,ea and ae with and
with e, etc. For an odd positive prime p, assume that
ac=ca, bc=cb, c=aba1b1.
Prove that the resulting set is a group with p3 elements.

§6.2 Subgroups and Orders of Elements

We now discuss some elementary concepts of group theory and their
properties. In terms of the symmetry of geometrical objects, two concepts
which arise immediately and are abstracted below are subgroups of sym-
metries, leaving fixed some parts of the object, and the order (sometimes
called period) of a geometric transformation.
A subgroup H of G, denoted by H G, is a nonempty subset of G
such that for elements x, y in H both xy and x' belong to H. Equivalently,
H is a subgroup if 'e H for all x, y E H. In other words, a subgroup is a
subset which is itself a group with respect to the same law of group composition.
A subgroup H of G is termed nontrivial if {e} H G. This definition is
similar to that of a nontrivial subring in §3.1.
While we use the inclusion symbol c for both subgroups and
subsets, the meaning should be clear from the context.
The right coset of a with respect to H is the subset of G
Ha= {xa:xeH}.
The left coset of a with respect to H is the subset of G
aH= (ax:xeH}.
In particular, the left coset hH of hE H equals H; similarly, Hh = H. In fact,
al-I = H a e H.
172 Group Theory chapter 6

As we customarily write groups multiplicatively (that is, use multi-

plication (or juxtaposition) to designate the group law of composition), we
denote the cosets multiplicatively: aH and Ha. For additively-written groups
we usually denote the cosets a+H and H+a. An element a' of a coset aH
(or a+ H) is called a representative of the coset aH. In this case a' = a/i for
some Ii e H. Consequently,
all = a'h'H = a'(h'H) = a'!!,
since H is a subgroup. Similarly Ha = Ha' if a' = /z'a for some h' e H.
The set of right cosets of a given subgroup H c G constitutes a
partition of G [see §1.2]. This means in particular that two cosets Ha,Hh
are either equal or disjoint. To prove it anew, note that g e Ha Hb
implies
g= ha = h'b. for some /z,h' H.

Consequently, a = 'h'h e Hb and hence Ha ç Hb. Similarly h = (h') 'ha E

Ha and hence Hb c Ha. Therefore we can define an equivalence or congru-
ence relation as follows:
aE b (mod!!) ab' E H
Ha = Hh
a e Yb.
There is a one-one correspondence between the elements h H and the
elements of the coset Ha for each a E G. Since cancellation holds in a group,
h1 a = a = h,.
Similarly, the set of left cosets of H constitutes a partition of G
(although not in general the same as the partition by right cosets). There is
again a one-one correspondence between the elements of H and of all.

As an explicit illustration of these concepts, consider Examples 6(f) (for

n 3) and 7(a), the group of the triangle r3, in §6.1. Describe the elements of r3 as

= the identity map,

L3, r clockwise rotation through 120°,
r2 clockwise rotation through 240°,
= reflection in the axis L,,
fi = reflection in the axis L2,
= reflection in the axis L3.

L2

The multiplication table for r3 follows.

 Subgroups and Orders of Elements 173

r r2 fi y

r r2 £ )' P
r2 e r fi y

$ y e r r2
r
fi y

fir r2 e

Check that the only subgroups of r3 are R = {c,r,r2}, S = {e,fl},

and {c, y}. Interpreted geometrically, R is the subgroup of symmetries of the triangle
which preserve the cyclical order of the vertices, whereas S and the other two sub-
groups leave fixed one vertex each.
The left cosets of R are
R= = {e,r,r2}, = {a,fl,y}.
Note that = Re and aR = The left cosets of S are
S = eS = {e, a}, rS = {r, y}, r2S = {r2,fl}.
The right cosets of S are
S = Sc = Sr = {r,fl}, Sr2 = fr2, y}.
Schematically, we represent the subgroups of in a so-called lattice
diagram ordered by inclusion.

{e,

{e)

In a group G the order of g e G, denoted 0(g), is defined to be

(i) the least element in the set {n e N gfl = e};
(ii) if the set in (i) is empty.

In the first case, g is said to be of finite order, and in the second, of infinite
order. The order of an element g e G has several general properties.

Property 1. 0(e) = I.
Property 2. g e EG o(g) > I.
Property 3. o(g) = for all a e G.

Property 4. o(g) =
Property 5. o(g) = = e hln.
174 Group Theory chapter 6

The proof of Property 5 follows from the Division Algorithm

Writing n = qh + r, where 0 r < h, we have
e =
= 9qh+r = (gh)qgr = (Ag' =

Since, by definition h is the least positive integer k such that g" = e, r < Ii
is necessarily zero. But that is to say hi n.

Proposition 1. For any element g of finite order in G. and any integer s,

o(g&) I o(g).

Proof. Consider g of finite order n. For any integer s,

(gS)fl = = (g")3 = = e.
By Property 5 we conclude that n = o(g).
More particularly, we have the following proposition.

Proposition 2. For g e G of finite order and any integer s, 0 <s < o(g),
gS . o(g)
(s, o(g))

Proof Let k = o(g5), n = o(g), and d= (s,n). Then write n = mdand s id.
We wish to prove that k =m, or in other notation, n=kd=o(gs)(s,o(g)).
This we do by showing that k Im and in k.
First, (gs)m = = (g"Y = e; so k I in. Second e = = guiC implies
that n jsk. Now write sk = vii. Substituting s = id, n = md, we obtain
tdk = vmd or 1k = i'm. Since (m, t) = I by the definition of d as the GCD
(s. mm), we conclude that iii k.

A corollary of Proposition 2 is that

o(ys) = 0(g) (s,o(g)) = I.

Proposition 3. For g e G of order rnn, where (ni. n) I, there exist unique

elements a, b e G such that g = ab = ba and o(a) = in, o(b) = n.

We leave the proof as an exercise, with appropriate hints. It is also

a consequence of Proposition 2, §6.6, or the Fundamental Theorem of
Finitely Generated Abelian Groups, §7.1.

REMARK. Suppose that the elements a, b of a group G commute and that they
have relatively prime orders m,n. Then o(ab) = ppm.
This remark is in general not true for noncommuting elements. For
example, the group of the triangle F3 is generated by the elements r and a of orders
3 and 2, respectively. However, ra = y, which is of order 2. The product (ra) a =
= r of elements ra and a, each of order 2, has order 3.
More generally, the dihedral group with 2n elements (given in Example
§6.2 Subgroups and Orders of Elements 175

6(f), §6.1) is nonabelian for ii 3. The element ab satisfies

(ab)2 = abab = = = e;
it is of order 2 and is the product of elements of orders n and 2. The product of ab
and b, each of order 2, has order n.

The set-theoretical intersection flAEA HA = If of a collection of

subgroups of a group G, indexed by 2 e A for some set A, is called the
intersection of the subgroups
The intersection H is a subgroup of G, for if x, y e H, then
x. y E 11A and hence xy e HA for all 2 e A. Thus xy e H. Furthermore
e for all 1, and so E H.
Now let S be an arbitrary nonempty collection of elements in a
group G, and T be the collection of all finite products d1 dk, where either
or Clearly
also and e=dd'eT for d€S. Hence
Sc Tc G, and T is a subgroup of G. The group T is called the subgroup
of G generated by the set S c G. (This is the group-theoretic analogue of the
vector space concept of span.)

Proposition 4. H denote the intersection flpHu of all subgroups

Let
c G which contain a subset S c G, and T the group generated by S.
Then H=T.
Proof First, Sc Tc G, and hence T is one of the groups Therefore,
on the one hand, c T. On the other hand, for each p. Sc 11u implies
that contains all finite products d1 ... dk, where either d e S.
Hence Tc for all p. so T c H
H. K of two subgroups If and K of a group G is the sub-
group of G generated by the set of elements H u K. Thus H K consists of
all finite products /i1k1 ... h1k1 ... of elements h1 H and e K,
I for all yeN.
For the sake of emphasis, we state the following proposition.

Proposition 5. The product H K of two subgroups If, K of a group G is

the intersection of all subgroups L of G such that L If and L K.

Corollary. The product H. K of two subgroups of a group G is the least

subgroup of G containing both H and K.

"Least" is used here in the sense that no proper subgroup of H. K

contains both Hand K.
The definition of the product of two subgroups of a given group G
extends to the product of any finite number of subgroups H1, .., with
.

being the least subgroup of G containing all groups H.,

i= l,...,s. The product subgroup must not be confused with the
176 Group Theory chapter 6

complex (set) of producis HK= (/zk : he H. k e K}. For example, in the

dihedral group r, of 2n elements with generators a, b, for which a" = b2 = e,
for subgroups H {e. ab) and K = {e, b}, then
ll•K= U,,,

I-IK= {e,ab,b,abh=a).
The latter is not a subgroup of f,.
From the definition it is evident that the product K• H denotes the
same subgroup as H• K.

Exercises

I. a. Prove the first assertion in the remark.

b. Let a and b be elements of a group G with respective orders u and
v. Prove that the order of ab divides the least common multiple
[u, vI if ab ba e. Note that the example of U3 in the remark
provides a counterexample when ab ba.
c. With reference to part (b), provide an example to show that the
order of ab need not equal [u, v].
2. Prove that if every element different from the identity has order 2 in a
group G, then G is abelian.
3. Find all subgroups of the additive group of integers Z.
4. Consider a subgroup H of a group G.
a. ProvethataH=H=HaifandonlyifaeH.
b. Prove that there is a one-one correspondence between elements of
H and those of the coset oH for any given a e G.
5. a. Prove that the relation defined as b(rnodH) if and only if
ab' c H for a subgroup H of the group G, is an equivalence relation
in the sense of *1.2. Do the same for the relation a b (mod H) if
and only ifb'aeH.
b. Show that in the case of the group U3 the two relations in part (a)
are different (i.e., define different equivalence classes) for an
appropriately chosen subgroup H.
6. Let H be a subgroup of a group G. For each g e G, prove that
{ghg1 :heH}
is a subgroup of G. It is called a conjugate subgroup of H in G.
7. Given an abelian group G, prove that G, = {g G : o(g) < co} is a
subgroup called the torsion subgroup.
8. Given a group G with subgroup H. show that the set
{g e G : gil = Hg), called the normalizer of H in G, is a subgroup of G.
9. With reference to Exercise 8, prove more generally that the normalizer
{ge G :gS= Sg)ofany set Sin agroup G isasubgroupofG.
 Coset Decompositions and the Theorem of Lagrange 177

10. Let C={aeG:ag=ga, for aligeG); it is called the center of G.

a. Verify that ('is a subgroup of G.
b. What is the normalizer of C in G?
11. a. Prove Property 3. b. Prove Property 4.
12. Prove Proposition 5.
13. Consider finite subgroups S and T of a group G. Let ST denote the
complex (SI : s S, t c T}. Prove that

Card(ST) = Card(S) Card(T)fCard(S C'i T),

where the cardinality Card(U) denotes the number of distinct elements
in a finite subset U of G.
14. Verify that the intersection of two subgroups of a group G is again a
subgroup of G.
15. a. Prove that if a finite subset S of a group G is closed with respect to
the group operation (i.e., if a,b S= ab ES) and contains the
identity element of G, then it is a subgroup.
b. Give a counterexample in the case of an infinite subset.
16. Prove Proposition 3.
a. Show that o(gfl) ,,,, 0(9m) =
b. Let a = g", /3 = gm, and write I = en + urn. Show that o(&') = 0(2) = fll
and o(fl") = o(fi) = n.
c. Set a = b = /3", and prove that a, b are unique. Hint: Suppose
a,b are not unique and raise ab = a'b' to the (mii)th power. We
must show that b b', but this can be accom-
plished by noting that mu = I — en.
17. Generalize Proposition 3 to the case where o(g) = m1 ins, the product
of s pairwise relatively prime factors.

§6.3 Coset Decompositions and the Theorem of Lagrange

We return now to the cosets of the previous section, using them as a
means to enumerate the elements in a finite group.
The order of a group G, denoted by either o(G) or [G: e], is defined
to be
(I) n, if G has n elements,
(ii) if G has an infinite number of elements.
The example of the group of the triangle shows that the order of a
nonabelian group generated by two or more elements of finite order may contain
factors relatively prime to the orders of the generators, in ['3 the elements a and ra
of order 2 are generators, but 31 [F'3 e] = 6.

Let G be a group (either finite or infinite) and H a subgroup (also

finite or infinite). We can express
G = gH
geG
178 Group Theory chapter 6

or more particularly,
6=
1€ I

where the H. i e 1. are disjoint left cosets. Similarly, we can write

6=U
j€J
where the j e J. are disjoint right cosets. Such a representation of a
group 6, called decomposition into left (right) cosets modulo H. is utilized
in proving the well-known theorem discovered by Lagrange in his study of
the behavior of roots of polynomial equations.
Theorem of Lagrange. If G is a finite group with 0(G) = [G : e] elements,
then for any subgroup H, 0(11)10(G).
Proof. In §6.2 we noted that the elements of H are in one-one correspondence
with the elements of any coset Ha, and that 6 = Ha1 u u Ha,, where
for Consequently
which completes the proof.
Let r = [G: H], denote the number of distinct right cosets of H
in 6. If G = b1 H u u h( H (decomposition of G into left cosets modulo If,
for then o(G)=o(H)+.•.+o(H)=/o(H), where
(= [6 : 11], is the number of distinct left cosets of H in G. Thus r = 7.
Thecommon number r = ( of right and left cosets is called the
index of H in G, denoted by [6: H].
The Theorem of Lagrange has the equivalent statement that if II
is a subgroup of a finite group G, then
[G:e] = [G:H]{H:e].
Lagrange gave this theorem in 1770—71, but the first complete proof is due
to Pietro Abbati (1768—1842) some thirty years later. Note;though, thai. the term
"group" was not introduced (by Galois) until 1830.

Corollary 1. The order o(g) of an element q in the finite group G divides

the order of G.
For the proof note that the subgroup H = {g': I e Z}. generated byg,
contains o(g) distinct elements.
Corollary 2. If H, K are subgroups of a finite group 6 and K I-f, then
[G: K] = [G: H][H: K].
Proof. Write, using the Theorem of Lagrange three times,
[G : K][K : = [G : e] = [G : H][II : e]
= [G: H][H: K][K: e].
Canceling [K: e], we have
[G: K] = [G: H][H: K].
§6.3 Coset Decompositions and the Theorem of Lagrange 179

Exercises

1. Prove that any finite group G of even order contains an element of

order 2. (This result is generalized in §7.5 to the statement that a prime
p divides o(G) if and only if G contains an element of order p.)
2. a. If the set S of Exercise 9, §6.2, consists of the single element
prove that the index of the subgroup N0({g01) in a group G equals
the number of elements g in G conjugate to g0.

An element y is said to be conjugate to g0 if there exists an element a u G

such that g = ag0a'. An element g0 is said to be self-conjugate if every con-
jugate ag0 a -'is equal tog0 itself. Thus, the center C of G is simply the set of
self-conjugate elements [cf. Exercise 10, §6.2].

b. Conclude that the number of conjugates of an element in a finite

group G divides o(G).
3. Let a, b, ... be elements of a group G. Define [a, b] to be the
commutator aba - 'b - l• Prove that [ab, c] = {a [b, e] a - } [a, c] and
[a,bc] =
4. Using the notation of Exercise 3, prove by induction on ,i that =
where q = (ii— l)n/2 for all n c N, provided that the com-
mutator [a,b] lies in the center of the group G; that is, provided
g[a,b] = [a,b}g, for all G.
5. Prove that a group G with p elements, pa prime number, must be abelian.
6. The minimal exponent of a group G is the least integer in e N such that
= e for ally E G. If G isa finite abelian group with elements a,, ...,
prove that the minimal exponent of G is the LCM of o(a1)
7. Noting that the prime residue classes in Z,,, constitute a multiplicative
subgroup Urn, prove that
I (mod,,,)
for all x e N for which (x, in) = I. This is known as Euler's Theorem,
and, in the special case when in is a prime, as Fermat's Little Theorem
[cf. §2.12]. (Here denotes the Euler v-function.)
8. If (a,,n) = I, prove that x satisfies the congruence ax b (modni) if
and only if x = ' + kin, for some k e Z.
9. Consider a one-dimensional vector subspace U of R2.
a. Verify that U is an additive abelian subgroup of R2 considered as a
group.
b. Describe the cosets of U.
10. Describe geometrically the cosets in R3 of a one-dimensional vector
subspace U and of a two-dimensional vector subspace V.
11. If[G: H], < and [G: K], < that [G:
H K is the intersection of the subgroups ii and K. (This result is due
to Henri Poincaré, 1854-1912.)
 Group Theory chapter 6

12. Prove that if ye G has finite order, then H = {g' : Ie has o(y) distinct
elements.
13. a. Referring to Exercise 6, §6.2, prove for all g e G that
[G:H]1 < [G:I1],.
b. Prove that = o(H), when o(H) < x.

§6.4 Normal Subgroups and Factor Groups

A class of subgroups of particular interest are the so-called normal
subgroups, for using them we can define groups of cosets called factor or
quotient groups, analogous to the residue classes modulo an ideal in ring
theory.
A subgroup 11 of a group G is called normal if Ha = all for every
a e G, or equivalently, if aHa' = H for every a e G. Note that aHa' = H
means (aha' :hell}=H.
Propositions I through 3 are immediate consequences of this
definition.

Proposition 1. Jf H and K are normal subgroups of a group G, then H K

is a normal subgroup.
Proof First recall from Exercise 14, §6.2, that H n K is a subgroup.
Consider xe H K; then axa ' lies in II and in K. Hence axa e i-I K,
for all a e G.

Proposition 2. If N is a normal subgroup and H is any subgroup of G, then

the product NH = n E N. hE H}. In fact,
NH= NH= HN.
Proof First for n1 e N. H, i = 1,2, the product n1 n2 can be
written as n'h' with n' e N, h' E H:
n1h1n2h2 = = n1n'2h1h2
by the normality of N and the fact that H and N, being subgroups, are closed
with respect to multiplication. Induction on s implies that n1h1n2/i2
can be written as n'/z'. Thus any element in N. H can be expressed in the form
n'h' e NJ-!, and N. H NH. Since Nil N. H, we have
N NH and HN are equal.
A product hn is equal to a'!i for some n' e N. We simply observe that
= /zn(/i 'Ii) = t)/ and set n' =hnh '. Similarly n/i = h(/s 'n/i) =
n" e N. Consequently, when either N or H is a normal subgroup, we shall
commonly write NH for the product subgroup N. H.

Proposition 3. If N and H are both normal subgroups of a group G, then

NH is a normal subgroup of G.
§6.4 Normal Subgroups and Factor Groups 181

Proof. For any g

Nil =
= g1NHg.
since gNy = N and gHy '= H, as both N and H are normal subgroups.
Hence NH is a normal subgroup.

REMARK I. Note carefully that the normality of N in G means gN = Ng, for all
p e G. It does not mean ng = gn. Rather, given g, we have ng = pit' for some it' N;
the elements n,n' e N are in general not equal. For example, in the group of the
triangle F3 R is a normal subgroup, but r€ €r although aR = Ra.

If N is a normal subgroup of a group G, then in a natural way the

set of cosets {gN : g e G} can be given the structure of a group. This new
group, called the factor or quotient group of G modulo N and denoted GIN,
is analogous to the residue class rings considered in and 2.10.
The product of two cosets aN and bN is defined to be the coset of
a'b', namely (a'b') N, where a', b' represent aN, bN, respectively. As was the
case for addition in Zm, we must prove that this rule for composition of two
cosets is independent of the choice of representatives a' E aN, b' bt"I
[cf. §2.9]. In particular. to prove
a'N = aN
a'h'N = ahN,
b'N = bN
we note that
a' e aN. b' e bN a'b' e abN,
since a' = an, b' = bn' imply
a'b' = anbn' = ab(b 'nb)n' = abn"n' abA'.

Therefore a'b'N ahN. A symmetric argument yields abN a'b'N, which

completes the proof.
The identity or unit element in GIN is
eN= (g€G:geN},
since (eN)(gN)=(eg)N=gN=Ng=(gN)(eN), for all peG. Finally, the
inverse of 9N is p - 'N, the coset of p '. Associativity of multiplication of
cosets gN is a consequence of the associativity of multiplication in G. We
conclude that the set GIN of cosets of the normal subgroup N is itself a group.
Note that in the pToof that multiplication of cosets is independent
of the choice of representatives of the cosets, essential use was made of the
fact that N was normal in G. The set of cosets of a nonnormal subgroup H
cannot be given a corresponding structure of a group. In the case of the group
of the triangle can of course define a rule of multiplication on
182 Group Theory chapter 6

the three cosets of the subgroup S = {e. but no such rule of multiplication
will eoriespondto the product in r3. The subgroup S is not a normal subgroup
of r3.

REMARK 2. A trivial consequence of the definition of normality is that every

subgroup H of a group G is a normal subgroup. Thus for a subgroup
H G, G commutative, we can always form the quotient group Gill.

Example I. Consider the abelian group

G = : ,n,,z E
where a!, = ba and o(a) = 4, o(b) = 12.
The element a2b3 has order 4 and generates the subgroup H =
{e,a2b3,b6,a2b9}. Since o(G) = 48, the factor group G/H has order 12. To prove
that GIN is generated by the coset abH, it suffices o prove that the order of abil
in GIN is 12. Observe that

Consequently, o(abH) is not I, 2, 3, 4, or 6; thus it must be 12 since it divides

12 = o(G/H).
Similarly, the element a2b E G generates a subgroup K of order 12,
K = {e, a2b, b2, a2b3, b4, a2b5, b6, a2b', b8, a2b°, & 0 a2b' 't,
and G/K is a group of order 4; each element is a power of aK (and of a3K).

As a second example, consider a nonabelian group of order p3.

p a prime. (When p = 2, the group in Example 2 is the dihedral group IT4;
cf. Example 6(f), §6.1.)

Example 2. Let
G= a Z}
where = e, bab' =
To show that the subgroup
N = {e,
a normal subgroup of G, observe that
= = = a"
implies I = am(bnakhlb_n)a_m = at"
for all ni,n and k.
The factor group GIN hasp2 elements and is abelian because bab =
a" a N. Thus
= eN
or =
and for any In, fl E Z,
(bN)"(aNr'.
(In fact, any group of order p2. p a prime, must be abelian as shown in §7.4,
Exercise 4.)
Furthermore, GIN consists of the p2 distinct cosets 0< fi <p.
 Normal Subgroups and Factor Groups 183

Exercises

1. With reference to Exercise S of §6.2, prove that if H is an arbitrary

subgroup of a group G, then H is a normal subgroup of N6(H).
2. Prove that if a subgroup H of G is a normal subgroup of K, and
K G. then K =
3. Prove that the center C of any group G is a normal subgroup.
4. Prove that if the subgroup H of G has index 2, then H is a normal
subgroup.
5. Consider subgroups H and K of the group G and let HK be the totality
of all products hk with h e H and k e K. Prove the following statements.
a. HK is a subgroup if and only if HK = KH.
b. HK is a subgroup if one of the groups H and K isa normal subgroup.
c. HK is a normal subgroup of G if both H and K are normal subgroups.
6. Find the center C of the dihedral group r7 [see Examples 6(f) and 7(c),
§6.11 of order 14 [cf. Exercise 10, §6.2].
7. Let F4 denote the dihedral group of order 8 [see Example 7(b), §6.1]
commonly called the group of the square.
a. Find all subgroups of F4.
b. Determine which subgroups are normal.
c. Exhibit the multiplication tables of the corresponding factor groups.
d. Construct the lattice diagram of subgroups, as in §6.2.
8. a. Consider the integers Z and the rational numbers Q as additive
groups. Show that every element of the factor group Q/Z has finite
order.
b. Is the corresponding statement true for R/Z, where the set of real
numbers R is considered as an additive group?
9. For a group G with a subgroup H and a normal subgroup N, prove
that H N is a normal subgroup of H.
10. Given a group G with subgroups M, M0, N, and N0, where M0 is a
normal subgroup of M, and N0 is a normal subgroup of N, prove
that MO(M n N0) is a normal subgroup of M0(M N).
11. Prove that multiplication in the quotient group GIN is associative, where
N is a normal subgroup of a group G.
12. Given a group G with subgroup H, show that the centralizer CG(H) of
H in G is a normal subgroup of the normalizer NG(H) of H in G, where
C0(J-I)={aeG:aha' =hfora//heH}.
13. Show that the two equivalence relations defined in Exercise 5(a) of §6.2
are the same if and only if H is a normal subgroup of the group G.
14. Let = (a1,a2) and fl= (b1,b2) be two independent vectors in R2, i.e.,
pairs of real numbers such that a1 b2 — b1 02 0. Define a group structure
on R2 by componentwise addition [see Examples of *6.1].
a. Prove that S = m,n e Z} is a subgroup of R2. Since R2 is
abelian, S is a normal subgroup.
b. Describe geometrically the quotient group R2/S.
184 Group Theory chapter 6

15. In the Cartesian plane R2, consider any two nonparallel translations
ti, r. Let r be the group generated by o- and r.
a. Describe the set of points
U = {(r,r') cR2 : y((O,O)) = (r,r') for some ['I,
called the orbit of (0,0) under r.
b.Prove that U is a normal subgroup of R2. Such groups are special
cases of transformation groups which are of particular interest in
both topology and quantum mechanics.
16. If H G is the only subgroup of ordert, in a group G, prove that H is
a normal subgroup of C. Hint. See Exercise 6, §6.2.
17. Let U be a subspace of a finite dimensional vector space V over a field F.
a. Show that U can be viewed as a normal subgroup of V.
b. Define on the set of (additive) cosets of (I the structure of a vector
space over F Icf. Exercise 9, §6.3]. We call this the quotient space of
V with respect to U, denoted V/U.
c. Prove that dim(Vf U) = dim V—dim U.
19. With reference to Example 2, prove that the cosets a°b"N, 0 a, /S <p
are distinct. Hint: Consider a°b°N and use the commutativity
of GIN to write
N =
=
19. Consider the group G of 2 x 2 nonsingular matrices
[a b
[c d
with coefficients in a field F [cf. Example 4(a), §6.1]. Prove that the set
of matrices

N__ha 0
:aeF
U 0 a
is a normal subgroup. The factor group GIN is called the projective linear
group PL(2,F).
20. Carry out the explicit steps in the induction argument in the proof of
Proposition 2.

§6.5 Group Homomorphisms

We now turn to one of the most important concepts of group theory,
that of homomorphism. A homomorphism is a mapping of one group into
another so that the group operations are preserved. To demonstrate the
significance of such mappings we present various detailed examples which
show in part that the theory of congruences is an essential tool for the study
of finite groups. The examples also show that the abstract definitions are not
§6.5 Group Homomorplisms 185

vacuous. Further discussion of homomorphisms will be found in and

Suppose that there are given two groups G and G' with respective
identity elements e and e'. A single-valued function q' from G into G'
forgeG
is called a (group) homomorphism of G intoG', if
=
I
product in G
I
product in G'

Proposition 1. For any group homomorphism,

q(e) = e' and =
In the second equation, the exponent — I on the left indicates the
inverse of g in G; the exponent — I on the right indicates the inverse of
in G'.
Proof To prove these assertions observe that q(g) = =
hence by the uniqueness of the identity e' of G', q(e) = e'. Furthermore,
implies that by the uniqueness
of the inverse in G'.
For a normal subgroup N of a group G the homomorphism
G —, GIN is called the canonical homomorphism (or natural projection)
of G onto the factor group GIN. Specifically, for g e
= •qN

That is a homomorphism follows, for g. g' in G, from

= (gy')N = (gN)(g'N) =
For a (group) homomorphism p: G —+ G' the complete inverse image
'(e') = e G: = e'} of the identity e' of G' is called the kernel
kerq of The image of a homomorphism .p is defined to be the subset
(actually a subgroup) of G':
= 4(G) = (q(g)eG':geG}.

Proposition 2. The subset kerq is a normal subgroup of G.

That kerq, is a subgroup is left as an exercise. To prove its normality

consider x kerq and a E G. Then
tp(axa') = q(a)q(x)q(a1) ço(a)e'[cp(a)]' =
i.e., axa1 kerq.
186 Group Theory chepter 6

Proposition 3. If the element a e G has finite order h = 0(a). then

divides 0(a).

We have a" = e. and hence = p(ah) p(e) = e'. Then by

Property 5, §6.2. o(p(a)) divides Ii. as asserted.
As was the case with ring homomorphisms. group homomorphisms
with special properties have particular names. We review the most important
special properties below. For others the reader is referred to the analogous
definitions in §3.2. A homomorphism q': G—'G' from one group G to
another G' is said to be
onto (alternatively, surjective or an epimorphism) if for each g' E
there exists at least G such that q'(g)=g;
one-one (alternatively, injective or a monomorphism) if (p(g) =
implies g =
an isomorphism (alternatively, bijective) jilt is both onto and one-one;
that is, if (g') consists of one and only one element g E G, for each
g' e G';
an automorphism if it is an isomorphism and (1' = G.
If there exists an isomorphism G-+ G'. we say that G and G' are isomorphic
groups, denoted G G'.
To facilitate our discussion of examples of group homomorphisms
we introduce cyclic groups. A group G is said to be cyclic if there exists some
E G such that every g G is a power of The element g* is called a
generator of G; it is not necessarily unique. For convenience let
C, {c' 0 i < ,'} =
denote a multiplicatively-written cyclic group of,i elements with I = as the
identity element. The exponents I of r are counted modulo n, i.e.,
C
= etc.

Example 1. Consider the two abelian groups C9 and

G =
where ab = ba and a3 = b6 = e. (In other words, the integers rn and ii are to be
counted modulo 3 and 6, respectively.) Define a homomorphism G -÷ C9 by
=
(Verify that this is indeed a homomorphism.)
To describe ker g' explicitly we need to determine all exponents in,,, for
which = I. This amounts to solving the congruence
3(2m+n) 0 (mod 9),
or equivalently,
2i,, +n 0 (mod 3).
(Why?)
§6.5 Group Ifomomorphisms 1K7

Solutions of the
Stated Congruence Corresponding
Elements of ker

o
o 3 b3

I ab
4 ab4

2 2 a2b2
2 5 1
a2b5

Note that each element of kerQ, is a power of ab: hence K = is a cyclic sub-
group of order 6 with generator ab.
The cosets

K = {e,ab,a2b2,b3,ab4,a2b5} = e,
aK = {a,a2b,b2,ab3,a2b4,b5} = a,
a2K = (a2 ,b,ab2 ,a2b3,b4,ab5} = a2
of K form a cyclic group G/K of three elements with the generator a (also a2).
To describe = C9 observe that c' if and only if there
exist ni, n e Z such that
3(2m+n) I (mod 9).

This congruence is satisfied by any solution to the following congruences.

For in = 0: 3k', j (mod 9), which has a solution if and only if 311. That
is, 1=0,3,6.
For m = 1: 3,, 1—6 (mod 9), which has a solution if and only if
3 I (1— 6) or equivalently 3 i.
For in = 2: 3i, 1-.- 12 (mod 9), which has a solution if and only if
31(1— 12) or equivalently 3j1.

Therefore, lmço = The mapping is described in detail by

a—+c6 b—.c3
ab I a2b —÷ c6 b2

ab2 e3 a2b2 —* I b3 —' I

ab3 —+ c6 a2b3 (3 b4

ab4 -+ I a2b4 c6 b5 —, c6.

ab5-4c3

Our next example shows that distinct homomorphisms may have

the same kernel and image.
 Group Theory chapter 6

Example 2. Let C27 and be cyclic groups of orders 27 and 18 with generators
rand y, respectively. Define distinct homomorphisms and tl' from C2 into C28 by
= 0 k < 27,
= y4k 0 k <27.
To show that = observe that c* c ker4' implies that
2k 0 (mod 18) or k 0 (mod9).
Similarly, E implies also that k 0 (mod 9). Thus, = =
{c0,cO,cISJ.
Furthermore, the images coincide because = y2 and
generate the same subgroup
H=
of C18. This fact is most easily seen by observing that y4 = (y2)2 lies in the sub-
group generated by y2. and conversely that y2 = (y4)5, since exponents are taken
modulo 18, lies in that generated by y4.

We conclude this section with two propositions concerning the image

of a subgroup H c G under a homomorphism q: G G'.

Proposition 4. If G —+ G' is a group homomorphism and H is a subgroup

of G, then ço(H) is a subgroup of G'.
Proof. To prove that H' = is a subgroup it suffices to show that
a'b', a'1 belong to H' for all a',b' e H'. Corresponding to a',b' in H',
there are elements a,b E H such that q,(a) = a', p(b) = b'. By the homo-
morphism property of p, p(ab) = a?,' and q(a ')= a''; hence a'b' E H
and e H'.
Now suppose that H is a normal subgroup of G and that G —,
is surjeciive. To prove that H' = çp(H) is a normal subgroup of G', consider
an arbitrary element a' E G', and (by the surjectivity of q) a corresponding
a E G such that qi(a) = a'. Because H is normal in G, we have aH = Ha, and
so by the homomorphism property of
= q,(H)4(a) and a'H' =
thus H' is a normal subgroup of G'. This proves Proposition 5.

Proposition 5. if H is a normal subgroup of G, and the homomorphism

G— G' is surjective, then q,(H) is a normal subgroup of G'.

A converse to Propositions 4 and 5 is given in Exercise 15. Show by

an example that Proposition 5 need not hold i14 is not surjective.
In §6.8 we examine the images of subgroups (normal subgroups) in
the particular case that is the canonical projection G —+ G/K, where K
is a normal subgroup of G.
 Group Homomorphisms 189

Exercises

1. Verify the following if G G' is a homomorphism defined on a

group G.
a. is a subgroup of G.
b. is a subgroup of G'.
2. Let n be an odd integer. Define, for real numbers x and y,
x*y =
where the real :ith root is taken.
a. Prove that the operation * defines a group law on the set of real
numbers R.
b. Prove that this group is isomorphic to the additive group of real
numbers.
c. Why does this construction of * fail to produce a group if n is even
and the ,,th root is taken to be nonnegative?
3. Let Z be the additive group of integers, and C the field of complex
numbers. Suppose that w is a primitive eighth root of unity (meaning that
every eighth root of I is a power of u). Let be the mapping of Z into
C given by
= no Z.
a. Find the image and the kernel of
b. Find all homomorphisms from Z to the subgroup of C generated
by w.
4. Consider the abelian group G = in,,, e Z}, where a' = b' e
and ab = ba, and the subgroup H generated by ab2.
Determine explicitly the cosets of G modulo H.
a.
Set up the multiplication table for Gill.
b.
5. Consider the abelian group G = in,.': e Z}, where a8 = b2 = e
and ab ba, and the cyclic group C, with generator c.
a. Show that p: G—. C, given by = is not a homo-
morphism.
b. Show that the only homomorphism tl': G —* C, is trivial, i.e.,
= {cO}
6. Let G be the multiplicative group of quaterizions, consisting of the 8
elements ± 1, ±1, ±1, ±k, which satisfy
,2 k' = —I,
=i2 —i (—Ui,
(I = —ii = k, jk = —kj = i, ki = —ik =
1)2 I.
This is a special case (ii = 3) of the generalized quaternion groups in
Example 6(e), §6.1.
a. Find all subgroups of G. Prove that they all are normal.
b. Find a subgroup N of G such that GIN is isomorphic to Klein's
Four-group V(named for Felix Klein, 1849—I 925) of order 4 described
by the 4 elements where a,b are taken modulo 2 and =
= 1, t2 = I. Exhibit such an isomorphism.
190 Group Theory chapter 6

c.. How many such isomorphisms onto V are there?

d. Find all homomorphisms of G into the group T of all complex
numbers of absolute value I.
e. Find the kernel of each such homomorphism in part (d).
7. Conclude from Exercise 6(a) and from Exercise 7, §6.4, that the quater-
nion and dihedral groups of order 8 are not isomorphic.
8. Find all group homomorphisms of the following additive groups.
a. Z12 Z5 b. Z12 Z6 c. Z6 Z28.
9. Let G be a group, Define for fixed a e G different from the identity the
map L of G into itself by the equation L0(g) = ag for all g e G.
a. Prove that is a one-one mapping, but not a homomorphism of G.
b. Prove that the inverse mapping (L0y' exists and is equal to La-'.
10. Define a mapping .o on a group G by Prove that is an
automorphism of G if and only if G is abeliari.
11. a. Let G be an abelian group. Prove that the mapping g —+ g', for fixed
s a N, is a homomorphism of G into itself.
b. If o(G) is finite and relatively prime to s, prove that the above
mapping is an isomorphism.
12. Give an example of a nonabelian group, using a suitable s, for which
the statement of Exercise 11(a) is false.
13. Find all homomorphisms of the dihedral group F'4 of Exercise 7, §6.4,
into the multiplicative group C4 of all nonzero complex numbers. Find
the kernel of each homomorphism.
14. Let R be the integral domain defined in Exercise 3, §3.6. Now let S =
(a+b a,b a Z} be a similarly defined integral domain. Prove that
Rand S are isomorphic if considered as additive groups; prove that they
are not isomorphic, however, if considered as rings.
15. a. Let G G' be a group homomorphism, and H' be a subgroup in
G'. Prove that the complete inverse image H = is a sub-
group in G. Note that it is not required that H'
b. If H' is normal in G', prove that H is normal in G.
16. Consider groups G, G', and G". if G —p G' and 6' 6" are homo-
morphisms, prove that the map o G—. described by o ç(g)
is a homomorphism.
17. Suppose that a: 6 -. r and r: G —+ L are group homomorphisms, and a
is surjective. Prove that the following statements are equivalent.
a. There exists a homomorphism f L such that r = o i.e.,
r(g) = for all g a G.
b. The kernels satisfy ker a kerr. Schematically,
C
§6.6 Cyclic Groups 191

18. Show by an example that Proposition 5 is not true in general if the

homomorphism is not required to be surjective.
19. With reference to Example 2, find distinct integers in,n (different from
27. 18) and distinct homomorphisms cu from Cm to C,,, such that
ker = ker cue. and Im = Im

§6.6 Cyclic Groups

We undertake in this section a systematic study of cyclic groups,
which we shall write multiplicatively.

Theorem. A cyclic group G is either isomorphic to the additive group Z or

to the additive group of the residue class ring Zm for some integer in.
Proof Consider the following mapping A of Z onto G, where p is a given
generator of G:
A(n) = ii e Z.
Since the mapping). is a homomorphism
of Z onto G. Moreover kerA is actually an ideal in the ring Z because
A ( — n) = p "= (g" ) = e for n ker A. Thus ker). = (0) or ker A = (in) with
in > 0 [cf. §2.4]. In the first case A is an isomorphism.
In the second case, let in be the smallest positive integer such that
= A(rn) e. That is, in is the order of p. As an additive group Zm is gener-
ated by the coset [I] = {l +srn : se Z}. and +sm =g'. The in distinct
elements g0 =9a+snt 0 a <in, are thus in one-one correspondence with the
cosets [a]. The definition of the sum in Zm then implies that the mapping
[a] is an isomorphism of the additive group Zm onto G.

Proposition 1. Every subgroup H of a cyclic group G = <p> is itself cyclic.

lfo(G)=n, then H= for some k, such that kin.
Proof, if H is the trivial subgroup (e). it is generated by p0 = e.
More generally, since G is cyclic and generated by g, the subgroup H
is a collection of powers of p. Let
S= {,neN:gmEH}.
Because it is a nonempty set of positive integers, S has a least element k
(by the Well-Ordering Principle, §2.2). To show that generates K. consider
an arbitrary element gS E H. By the Division Algorithm, §2.3,
s=qk+r, 0r<k.
Then = gs(,q&)-q H,
pX,
because and hence (gk)_q belong to H. But since k is the least positive
integer such that k s, and = that is,
=11.
If 0(G) = n. the previous argument indicates that k because
= e E 11 and k s for all s for which p3 e H.
192 Group Theory chapter 6

Conversely every divisor rn of n 0(G) determines a subgroup

{gmi: ie Z} of G whose index is in. Also o(gm) = n/rn: thus o(II) n/rn.
H
These observations prove the following proposition.

Proposition 2. If in divides the order n of a finite cyclic group G, then G

has a subgroup II of index in (and of order n/rn).

That such a subgroup H is unique is left as an exercise.

Proposition 3. A cyclic group G of order n has q(n) distinct generators.

Proof First is a generator of G if and only if o(g') = n. By Proposition 2,
§6.2, = o(g) = ii if and only if (s. o(g)) = I. By definition of the Euler
(p-function is the number of integers s, I s<n. relatively
prime to n.

ProposItion 4. The homomorphic image (7 = (G) of a cyclic group G

of order n is a cyclic group whose order divides n.

generates G, then clearly

If generates since i,ts(çf) =
for a homomorphism iii. By Proposition 3, §6.5,
divides n = o(g) = o(G).
In particular if H is a subgroup of a cyclic group G, then the factor
group G/H is cyclic since it is the image of G under the canonical projection
n11: introduced in §6.5.

Proposition 5. Let G be an abelian group of order n. If there exist for each

divisor in of n at most rn elements whose order divides in, then G is cyclic.
Proof Let n have the prime factorization

ii pi8•.
=
Set in1 By hypothesis, for each i, I i s, at most n = n/p1 < n
elements a in G satisfy = e. Thus at least one element x1 e G satisfies
e. Let y1 = where n = in, Then o(y1)Iin, since y1mi = x," = e.
Moreover, because = e, we have o(y1) = Finally, z =
Yi Yi y5 has order n by the remark in §6.2.

Exercises

1. Find all subgroups of the cyclic group of order in generated by c

when
a. ,n=8 b. rn—35 C. ,n=3l5.
§6.6 Cyclic Groups 193

2. Describe the elements of the factor group C1 8/H, where C18 is the cyclic
group of order 18 generated by c and H is the subgroup generated by c6.
3. For how many distinct groups G, up to isomorphism, can a surjective
homomorphism C54 6 be determined, where C54 is the cyclic group
of order 54? Note that 6 is to equal Im = 4,(C54).
4. Prove that the subgroup H in Proposition 2 is unique.
5. Define a homomorphism C16 -+ C.16, where the cyclic groups C16, C96
have respective generators c, by 4,(em) = 2rn
a. Find the kernel of b. Find the image of 4,.
c. Find all homomorphisms of C16 onto the subgroup G of C96
generated by y'2 Note that if A: G -. 6 is an isomorphism, then
A o 4, is a surjective homomorphism of C16 onto G described by
A o to(c) = A(to(c)) = A(y12).
6. Consider the cyclic group C16 with generator c, and the abelian group
6 = {ambn : in,n E Z} where b2 = e, ab = ba, also of order 16.
Define mappings and çt' from G to C16 by
= C6m and

a. Verify that and are homomorphisms.

b. Find the kernels of 4, and ci'.
c. Find the images of and
7. Suppose that N is a subgroup of the center of a group 6. Prove that G
is abelian if the factor group GIN is cyclic.
8. Prove that the quaternion group [Exercise 6, §6.5] of 8 elements possesses
precisely three cyclic subgroups of index 2 and exactly one subgroup of
order 2.
9. Prove that any group 01 order 4 is isomorphic to one of two abelian
groups. (As a first step, exhibit two nonisomorphic groups of order 4.)
10. Let G be a noncyclic abelian group of order p°. p a prime, and H a cyclic
subgroup of G of order p8 which is not contained in any other cyclic
subgroup. Observe that there exists a y such that for all g e G, gP' H,
but for some g e 6, gP'' H. Prove that every 1: e H which is a
power of an element in G is also a power of an element in H.
11. a. Consider the cyclic group C3,, with generator c, and the homo-
morphism C36 C36 described by 4,(c) = c', for some fixed
s e N. Find the kernel of
b. Generalize the preceding to the case of C,, C,,, described by
= c', for some fixed s N, where c generates Cm. Find the
kernel of
12. a. Verify that the set of units Urn in Z,, for ni = 5, 6, 8, and IS con-
stitute a group [cf. Example 2(c), §6.1].
b. Is Urn a cyclic group for m = 5, 6, 8, or 15? (A general statement of
the structure of U,,, is given in Exercise II, §6.10.)
13. Let p be a prime number and suppose that (a, p) = 1. Prove that the
congruence a (modp) has d solutions modulo p if
(mod p) and has no solution if - 1)/a (modp), where d = (n, p — I).
14. Use Proposition 2 to prove Proposition 3, §6.2.
 Group Theory chapter 6

§6.7 Groups of Permutations

Let S be an arbitrary nonempty set and denote by Z(S) the
collection of all permutations (one-one set mappings or bi-
jections) of the set S onto itself. The set s(S) is made into a group by defining
a product of elements. The product a c fi, here called compositions of two
mappings ct, fi is defined to be the biject ion
(a fl)(s) = x[fl(s)] for all s e S.

The group axioms are readily verified for s(S) with this rule
for composing bijections. Note that two mappings a, a' on a set S are defined
to be equal if a(s) = a'(s) for all s e S.

Associativity. Since for all s e S

(ao{fl8y])(S) = rz[(fJoy)(s)]
= a[fl(y(s))] = [a o fJ](y(s))
= ([aofl]oy)(s),
the composition of mappings is associative, i.e., a (fi v) = (a 0 fi) 0 y.

Identity. The map defined by £(s) = s for all s e S satisfies the properties
of a group identity
I = I = E 02
since for all s e S
(aoc)(s) = a(s) = (soa)(s).

Inverse. An inverse a - of the bijection a is defined by

= if 1(5") — S.

Since a is surjective, given s e 5, we can find an s" such that cc(s*) = s. The
element s"' is unique because a is injective. The bijection so defined
satisfies
= = a(s*) = s, for all SE S.
I
a '(s) = for all
s a a a, that is, the group-theoretic inverse of a.

In particular, if S= {l,2 n} is a finite set, is

called the group of permutations of n symbols, or the symmetric group of n
elements. The group Z,, has n! elements as follows. For I E Z,,. there are
n possible images of I, precisely n — I possible images of 2 (since a (2) a (I)),
and precisely n—2 possible images of 3, etc. An inductive argument shows
that there are n! distinct bijections of that is, [E,, :1] =n!
 Groups of Permutations 195

Notation. There are two standard methods of describing a permutation

If
oh) = a, e { I. ...,n}.
then we write

[ i 2 •.. i •..

Lat a2 a,

This symbol or array describes the map taking each element i into the
element a1 below it.

If it is understood that a given permutation a !,, is defined on a set of n

symbols, we often omit from the array notation symbols left unaltered by a. For
instance, the permutation
[12345
[2 5 3 4 1

is abbreviated
[125
[251
The alternative cycle notation involves writing after each element
its image, thereby forming a chain or cycle, as follows:

A given permutation may involve several cycles

[1, a(l), a (u( 1)), .. .] [a, a(a), a(a(a)), .

The cycle notation, while awkward to describe in general, is, like stenography,
convenient to use in specific cases.
In multiplying cycles, we read from right to left between cycles and
from left to right within a cycle. That is, we write
[15][12] = [125] and [1345][243][135] = [1253].

This procedure is simply a convention. Some authors multiply from left to

right; thus the student must be aware of the different conventions for multiplication
of cycles.
For example, the permutation of mapping I 3, 2 4, 3 5, 4 — 2,
and 5 —+ I is denoted either

[12345
[3452 1
or [1351 [241. Schematically, this permutation can be displayed as

l—.3--'5--41 and 2—'4—'2

196 Group Theory chapter 6

or, literally in terms of cycles,

4
/
1

2 4

In the cycle notation, juxtaposition of two (or more) permutations

denotes their product. The one to the right is performed first. In the array
notation, the composition ta = t a of a, e given by c(i) = a and
= b., is

Ia1 2 ... it-i

II
since t o a(i) = r(a(i)) = b.. Hereafter, omitting the o, we shall simply
denote composition of mappings by juxtaposition (see Example 1, below).
A transposition i is a permutation which interchanges two elements,
leaving all others fixed. In other words for a transposition r on in.
t(h) = k, r(k) = Ii, r(i) = i
for all i different from any given h, k e A transposition t is denoted either
by

11
2 ... k ... It •.. niLkh
or by
[hk][l]."[nJ.
The I-cycles [I] are often omitted.
A cycle y of length r (r n) is a permutation such that for some r
elements ...,i, in
= j2, y(i2) = i3,...,y(i,_1) = I,, y(i,) =
and y(h) = I, for all h ...,i,}. Schematically, a cycle can be displayed as

'2
 Groups of Permutations 197

Thus a transposition is a cycle of length 2 or a 2-cycle:

'I

A cycle of length r. called an r-cycle,

[11,12, "''r] = [i,,ii.i2

= etc.

has order r as an element in the group [see §6.2]. We obtain the inverse
of a cycle y = [11,12. ...,ir] by writing the same elements in reverse order:
= [I,, ...,I2,!i]. In array notation the inverse of a permutation is obtained
by reading up from the second line to the first.

Example 1. The product of permutations

23451
513 4j i and r=I[12345
5243
ar= I [123451
is
[24 5 3 I

because ar(1) = = c(1) = 2,

ar(2) = = a(S) = 4,
cr(3) = a[t(3)] c(2) = 5,
ar(4) = c[T(4)] = c(4) = 3,

ar(5) = c[r(5)J = c(3) = 1.

In addition,

c1='[3 I. I 2
I 4
3 4
5
5

Of particular importance in many applications of permutation

groups are disjoint cycles, or cycles with no common symbols. We call two
cycles o• = [ii. i,] and t = [J1,Ji .151 disjoint if the set-theoretic
intersection
i,.} {J1,i2 f3} = 0.
Proposition 1. Every permutation of can be expressed as the (com-
mutative) product of disjoint cycles.

The proof depends upon the following lemma.

Lemma. For any j a and aa there exists a least integer k. such that
ak(j)jandOk<,z
 Group Theory chapter 6

If = E, then k = 0. Otherwise, consider the n + I elements

j = a°(j),a(j),a2(j)....,
These cannot be distinct, as has only n elements. Let Ii be the least integer
such that
ah(j) = cr"(j)
for some k where 0 Ii <k n. The proof of the lemma is completed by
showing that h must equal 0. To this end, note that since a is a permutation
= = qh(j) =
implies that Now because of the minima! choice of h,
we conclude that h— 1 is negative, that is, Ii must be 0.
Proof of Proposition 1. The identity map E is the product of I-cycles. There-
fore consider a permutation a and the cycle
= [o(I),a2(l) a3'(J)— I],
where the least positive integer k such that ?(l)= I as in the lemma,
is
0 n. Ifs1 = n. then = a. and the proof of Proposition I is complete.
Ifs1 <n, pick 12 E I), .., (I) = 11, and consider the cycle
.

= {a(j2),c2(J2),..., =12]'
where s2 is the least integer k such that ak(j2) =12 Then
{o(l),a2(l),...}
{o(j2),a2(J2). ...} = 0.
Repeat this process, if necessary, with
u
The process cannot be repeated indefinitely as there are only n symbols in
Consequently, the permutation

where since the symbols involved in the permutations

form disjoint sets.
For our analysis of the structure of I,, it is useful to note that an
r-cycle y = [i1, i,] can be expressed as the product of transpositions:
[i1,i2,...,i,] =
=
Then we obtain the following corollary to Proposition 1.

Corollary. Every permutation of can be expressed as the product of

transpositions.

Example 2. The permutation

[1 2 3 4 51
5 3 1 24]
 Groups of Permutations 199

is itsella cycle, written [15423] in cycle notation, since a(l) = 5, a(S) = 4, a(4) 2,
etc.

Example 3. The permutation

11=
1 2 3 4 51 EE5
2 5 I 4 3

is expressed as the product [1253] [4] of disjoint cycles since o(l) = 2, a(2) = 5,
c(5) = 3, and a(3) = I. Also, c(4) = 4.

Example 4. The product

[I 2
1 2 3

of cycles is written as a single cycle

Fl 2 3
[3 1 42
in cycle notation, [132][4)[l][2][34] = [2134]; and as the product of trans-
transpositions, [2134] = [243 [23] [211.

Example 5. The permutation

0=1 11234567891
[4 5 6 2 4 3 8 9 7]
can be written as the product
I 2 4 3 6 7 8
[ = [1425] [361 (7891
[4 5 2 1 ff6 3J[8
[
9 7] 1

of disjoint cycles, and as the product

[15][12] [14] [36] [79] [78]
of transpositions.

The following proposition is used in discussing the solvability of

groups and of polynomial equations by algebraic means

Proposition 2. If n 5, every 3-cycle p = [r, k, i] can be written in the form

arcr called the commutator of a and r. for some 3-cycles a and r.

Proof Given p = [r. k. I]. consider a = [i.j, k] and r = [k, r, s], where j
and .c are arbitrary, but distinct from r, k, and i, and j s. Then, =
[k,j,i] and = [s,r,k]. Consequently,
= [i.j,kl[k,r,s][k,j.jJ[s,r.k] = [r,k,iJ =
as asserted.
 Group Theory chapter 6

Exercises

1. Compute in the symmetric group the products y

where = (143)(2)(5)(6), fi (13)(26)(4)(5), and = (1546)(2)(3).
2. Find the permutation for which y, where a,fl, y are as given in
Exercise 1.
3. Let
[12 34561 I and $=i
1123456
14 2 1 3 5 6] 12 3 5 4 1 6

Express a product of cycles. If the result is a cycle, find its order

as an element of Z6.
4. Prove that two disjoint cycles commute.
5. Prove that the order of a permutation a considered as an element in the
group i.,, is the least common multiple of the lengths of its disjoint cycles.
6. Consider distinct indeterminates x1 v,, ii> 1, over the field Q and
define the polynomial
=
I
fl
Si<jfl

For a define
i, ..., =
I
II X6(,)),

which polynomial is equal to sgn(a)A(x1 where sgn(a) = ± I.

a. Prove that the mapping sgn(c) is a homomorphism from E, to
the multiplicative group ± Il of two elements.
b. Prove that the mapping sgn(c) is surjective.
c. We define to be the kernel of this mapping, and call it the
alternating subgroup of Prove that [).,, = 2. :

We shall call elements of even permutations, and speak of as the group

of even permutations of ii symbols. A permutation is called odd if it is not even.

7. Determine whether a transposition is an odd or an even permutation.

8. a. Prove that a permutation is even if and only if it can be expressed as
the product of an even number of transpositions.
b. For which values of ni is an tn-cycle an even permutation?
9. Write the permutation
12345678
42517386
a product of disjoint cycles. What is the order of a in the group
as
10. Prove that a finite group G of order ii is isomorphic to a subgroup of the
symmetric group Z,,. (This statement is commonly known as Cayley's
Theorem.) Hint: Prove that g —' aq, for fixed a E G, is a permutation of
the elements of G.
§6.8 The Isomorphism Theorems of Group Theory 201

11. Note that ci = (121131 and r [123] generate the group

a. Verify that a2 = = e and cra
b. Prove that the mapping 13 —, defined by = ci', is a
homomorphism.
c. Determine the kernel and image of the homomorphism in part (b).
d. Prove that is isomorphic to the dihedral group of the triangle U3
(see §6.2].
12. Using induction, prove that
= [i,,i2)[i2,i3)
13. Construct a nontrivial homomorphism from to 13. Hint: Examine the
subgroup
ff1] (2] [3] [4], [12] [34], [13] [24], (14] [23]},

which is isomorphic to Klein's Four-group.

14. Let G be the subgroup of 1s generated by = [1234) [5876] and
fi = [1537] [2648]. Prove that G is isomorphic to the quaternion group of
8 elements.
15. a. Prove that the dihedral group [Example 7, §6.1] is isomorphic to
subgroup of 1,,.
a
b. Show that 1'4 is not isomorphic to 14.

§6.8 The Isomorphism Theorems of Group Theory

The isomorphism theorems furnish most important tools for more
advanced results in group theory, such as the Jordan-Holder Theorem in
§7.6, and in the theory of rings and modules The isomorphism
theorems permit transfer of a problem which might be cumbersome in one
group to another group where the solution may be easier.

Theorem 1 (The Isomorphism Theorem). For a homomorphism p of G into

G' with kernel K, there exists an isomorphism 2. of the factor group G/K onto
the image G', such that q' = A o It, where it is the canonical projection
of G onto G/K.

In other words, there exists an isomorphism A, such that the diagram

G
4

IA

G/K

commutes, meaning =A it.

202 Group Theory chapter 6

Proof: (1) Define a mapping 2: G/K—'G' by 2(aK) = q(a). First we must

verify that 2 is well-defined. i.e., is independent of the choice of a representing
the coset aK. To this end, suppose that a and b each represent the coset
aK= bK. Then a 'he Kand q(a 'h) = e', the identity of G', or p(a)= q(h).
(ii) it is easy to check that 2 is a group homomorphism:

2(aK.hK) = 2(ahK) = p(ab)

= = 1(aK)2(bK).
(iii) If A(aK) = q(a) = e', then a K and hence aK = K. Thus the
mapping 2 is injective.
(iv) To verify the surjectivity of 2 consider any a' E By
definition there exists an element a E G such that q(a) = a'. Then 2(aK) =
q(a) = a'; hence 2 is surjective .Thus 2 is an isomorphism of G/K onto
as asserted. That q' = 2 o is a consequence of the definition of 1.

NOTE. Examples show that there may be many isomorphisms of G/K with
Such isomorphisms can be constructed by following the isomorphism of the
theorem by a nontrivial isomorphism a of the image onto itself, i.e.,
A(gK) = e
For example, suppose is isomorphic to the additive group 4,, ,n> 2.
The mapping [c] {a](c) = a([c]), [c] E Zm, where [a] is a prime residue modulo
m, is an automorphism of Zm. Thus a oA: is an isomorphism.

A simple relationship exists between the subgroups of G and those

of its factor group modulo a normal subgroup K via homomorphic images.
Suppose that it = itK is the canonical homomorphism of G o,,to the factor
group G* = GIK, where I( is a normal subgroup of G. If H is a subgroup of G,
then lr(H) isa subgroup of G* by Proposition 4, §6.5. Further, if/lisa normal
subgroup of G, then by Proposition 5, §6.5, ir(H) is a normal subgroup of G*.
Conversely, if H* is a subgroup (normal subgroup) of G*, then the
complete inverse image H = 1(11*) of is, the set of all y e G
for which it(g) a subgroup (normal subgroup) of G satisfying
ir(H)=H* and
This is a special case of the proofs requested in Exercise 15. §6.5. In summary
we have the group-theoretic analogue of Proposition 3, §3.4.

Proposition. There is a one-one correspondence between the subgroups

(normal subgroups) of G containing K and the subgroups (normal subgroups)
of G/K. The correspondence is given by taking the projection mapping it
on the subgroups H K and the inverse images of the subgroups of G/K.

Theorem 2. If G G' and G' —, G" are surjective group homomor-

phisms, then

G'/JV' G/N,
 The Isomorphism Theorems of Group Theory 203

where N' = ken/i and N= Schematically,

G .G"
UI UI

= N ,- N'= ken/i.
Proof By Exercise 16. §6.5, the composite cli go = is a homomorphism
of G onto G". Further
N= =
=
=
Now by the Isomorphism Theorem (for the surjective mappings and i/ip)

G'JN' and G" GIN,

and hence G'/N' GIN, as asserted.

Corollary 1. If G is a finite group, then [G: N] = [G' : N'].

Now suppose a group G has normal subgroups N and H, H N.

Consider the surjective homomorphisms go it11 and tfr:
*
G

where = gN. Then, since ken/i = N/H. we have a second corollary.

Corollary 2. (G/H)/(N/H).

Theorem 3. If N is a normal subgroup of G and if H is a subgroup of G,

then
NH/N H/(H N).
Proof: First, the sets of cosets involved in the theorem are well-defined
quotient groups. Propositions I and 2, §6.4, state that NH is a group (which
contains N as a normal subgroup) and that H N is a normal subgroup
of H.
Next the mapping ).: H—' HN/N, given by
2(h) = hN e HN/N for /i H,
is surjective, the elements of I/N being the products hn, h H, and di e N
[see Proposition 2, §6.4]. Furthermore, 2 is a homomorphism, A(hh')
(hh')N = (hN)(h'N) = ).(h)2(h'), for h,h' H. Also the kernel of A is
N H. since 2(h) = hN = N implies that h N, and thus Ii e N n H. There-
fore, by the Theorem, H/(H n N) HN/N.
CoroHary. If go: G—' G' is a homomorphism with kernel K, then for any
 Group Theory chapter 6

normal subgroup H of finite index in G.

[G:H] =
Proof As in Theorem 2.
G 4(G)
and G* Since KH = we have
p(G)/ço(H) G/KH.
and hence KH] = [ço(G): p(H)]. Applied to the normal subgroup H
and the subgroup K, Theorem 3 states that [K: H K] = [KH: H].
Consequently, as asserted,

Example. Consider the ideals H and K in Z generated by positive integers h and k,

respectively, as additive subgroups of Z. Then the I.CM [1., k] = in generates H K
and the GCD (ii, k) = d generates H + K. (Because the group operation in Z is written
additively the sum of subgroups replaces the product in the preceding discussion.)
The proof that hk = dm [cf. Exercise 4, §2.7] is now a simple consequence
of Theorem 3. (This is truly the hard way to prove that uk = din!) By Corollary 2,
§6.3,
k = [Z:K] = [Z:H+K][H+K:K) = d[H+K:K],
m [Z: = [Z:HJ[H:HrK) =
As a consequence of Theorem 3, however, [H+ K: K] = [H: H n K]. Thus
kdt orklz=n,d.

Exercises

1. Restate the Isomorphism Theorem (Theorem I) for a ring homomorphism

R —' R'. Prove that there exists a ring isomorphism ).: R/A —' ç(R)
where A is the ideal ker4, (see the theorem, §3.2).
2. a. Do Exercise 27(c) of §3.6 by defining a ring homomorphism
Z -+ Prove that is surjective and conclude, using the
isomorphism Theorem for rings (Exercise 1, above), that
Zf(p)
b. Repeat part (a) with reference to Exercise 28 of §3.6.
3. With reference to the discussion of §3.5, in place of

consider the map

§6.9 Automorphisms, Center, Commutator Group 205

defined by = ([a],, ..., [aL, ..., Prove that kerX = (m) and
that X is a surjective ring homomorphism. Using the Esomorphism
Theorem, conclude there exists an isomorphism

REMARK. A principal use of the Isomorphism Theorem for groups and

rings, as well as analogous results for other algebraic structures, is to avoid
having to define maps on sets of cosets in terms of representative elements,
and then necessarily having to prove that such maps are well-defined, i.e.,
that their definition is independent of the choice of representatives of the
cosets. To prove that such a is well-defined is essentially to repeat the
proof of Theorem 1.

4. If H isa normal subgroup of a group G and is contained in a subgroup K

of G, prove that NG,,,(K/H) = NG(K)/H [cf. Exercise 8, §6.2].
5. Consider an abelian group G in which every element different from the
identity has infinite order, and let H be a subgroup with finite index in G.
Define G" and H" to be the subgroups of powers of elements in G and
H, respectively. Prove that [G: II] = H"]. Hint: G" is the image of
G for the homomorphism G G given by Q(g) = g".
6. Consider a finite dimensional vector space V over a field Fas an additive
group. IfS and Tare subgroups (i.e., subspaces), use Theorem 3 to prove
that
dim1(S-f-T) dim,S+ dimFT— T)
[cf. Exercise ii, §4.2, and Exercise 17, §6.4].
7. Give an alternate proof of Theorem 3 by utilizing the canonical pro-
jection ir:G—'G/N. Noting that HN— ir'(n(H)), apply the Isomor-
phism Theorem to obtain HN/ N ,r(H). Conclude the argument by
observing that ir(H) HJH N.

§6.9 Automorphisms, Center, Commutator Group

In this section we study automorphisms. a special class of group
homomorphisms, and the concepts of center and commutator subgroup.
The center is the largest subgroup of a group G whose elements commute
with all g e G. while the commutator subgroup is the smallest normal sub-
group of G whose corresponding factor group is commutative. In investigating
these concepts within the context of dihedral groups we again emphasize
computations involving group elements.
A permutation (Or one-one onto mapping) of the elements of a
group G is called an automorphism is a/so a homomorphism of G into itself.
We denote the collection of automorphisnis by Aut(G). It is a subgroup of
the group of all permutations of G considered as a set [see §6.7].
Denote by a the identity map on G. which is the identity element iii the groups
Aut(G) and
206 Group Theory chapter 6

For a finite group G, the order of Aut(G) can be appreciably smaller

than that For example. if G = ç. the cyclic group with n elements.
then
[Aut(G) a] = < ii < ii! = : a]

if n 3, where q(n) is the Euler q-function. This statement is verified in

detail in Example I.

Example 1. For the multiplicative cyclic group C of order u with generator c,

the multiplicative group of units in the residue class ring Z,,.

Since a homomorphism defined on a cyclic group is completely deter-
mined by the definition ço(c), it is important to recall from Proposition 3, §6.6, that
has ço(n) distinct generators where (s, tO = I. For any isomorphism e
must generate because o(a(c)) = 0(c) = n. Therefore, = c5, for some
s, (s,n) = I. Conversely, the mapping C—' C,1 given by A(c) = Cs, for any s
relatively prime to ii is an automorphism. Thus, the distinct automorphisms of C,,
are in one-one correspondence with the integers s, I s n, (s, n) = 1.
It is convenient to describe the elements in Aut(C,,) as where =
and I s n, (s,n) = I. The mapping
a: Aut(C,,) U,, C Z,,

given by = [s] is a homomorphism, since

o = = [ss')
= [s][s'] =
Verification of injectivity and surjectivity is left to the reader.

Automorphisms of a group G of the form

g xg.v ' for all g E G and any fixed x E G
are called inner automorphisms. We denote the set of inner autornorphisms of
agroupGbyl(G).
Using a Venn diagram to represent the inclusions and Hom(G,G)
the set of homomorphisms from a group G to itself, we have the following
diagram.

Hom(G, G)
§6.9 Automorphisms, Center, Commutator Group 207

NOTE. Normal subgroups, defined in §6.4, are also called invariant subgroups
because a subgroup H is normal if and only if = x 6 G.
In more technical terminology, we describe H as invariant (mapped to itself) under
all inner automorphisms of G. In the more recent literature, greatly influenced by
the French School, "invariant" is often replaced by "stable."

Proposition 1. The subset 1(G) of inner automorphisms of G is a normal

subgroup of Aut(G).
Proof ForalIgEG
= = x(ygy')x'
= (xy)g(xy)' =
the composite of two inner automorphisms is again an inner auto-
Thus,
morphism. Also for any e 1(G), note that ox ° = = a. Hence
= (as)' eI(G), and 1(G) is a subgroup of Aut(G).
Finally, for any a e Aut(G),
(a ox ° a')(g) = '(g))) = a(xcC'(g)x')
= a(x)g(a(x))' =
for all g e G, which proves the normality of 1(G) in Aut(G).

For the group homomorphism c1: G—'I(G) given by ox.

= {x e G: = = g. for all g 6 G}

called the center of G. Thus the Isomorphism Theorem implies

1(G). This fact is summarized in the following proposition.

Proposition 2. Let C be the center of a group G. Then 1(G) (i/C.

For elements x. y in G, the product xyx is called the com-

mutator of x and y. The group G' generated by all elements in G of the form
for x, y e G. is called the commutator subgroup of G. The term
generated means that G' is the group consisting of all finite products of all
commutators of elements in G. Note that the product of two such finite
products is a finite product of commutators, and that the inverse of a com-
mutator is itself a commutator.
The group G' is also often called the first derived group, based on
classical considerations from the Lie theory of "continuous groups," in
which a chain of derived groups is developed (Sophus Lie, 1842—1899).

Proposition 3. A normal subgroup N of G contains G1 if and only if GIN

isabelian.
208 Group Theory chapter 6

The proof is given by the following equivalences:

GIN is abelian (aN)(bN) = (bN)(aN) for all a,b G

abN = baN
=N
eN for all a,!, e G

The last equivalence is a consequence of the definition of G as the group

generated by the set of all commutators.
The commutator subgroup G' is normal in G because for all
x, y, g e G,
a9(.vyx 'y ') = g(xyx ')g
=
and any element in G' is the product of commutators. Thus setting N = G'
in Proposition 3 yields the following corollary.

Corollary. The factor group G/G' is abelian.

Since Gt is a normal subgroup of G and is contained in any normal

subgroup Nc G for which GIN is abelian, we can describe (or define) G'
as the "smallest" normal subgroup of G whose corresponding factor group
is abelian.

Example 2. The automorphism group Aut(V) of Klein's Four-group V is iso-

morphic to the symmetric group of three symbols {cf. §6.7]. The group V consists
of the four elements
(a, b, ab = ba, a2 = b2 = e}.
Any p E Aut(V) can be described by
p (a) =
p (b) = aa21b022,

where c Z2, since the exponents of a,!, are counted modulo 2. It is convenient
to associate to p e Aut(V) the 2 x 2 matrix
a11 a12
a21 a22

with coefficients in Z2. Since p is an automorphism, p(a) and 49(b) must be distinct
elements of order 2; thus the matrix A(p) must have linearly independent nonzero
rows. In other words, as in §4.5, A(p) is a nonsingular matrix. Conversely, each
nonsingular 2 x 2 matrix with coefficients in Z2 determines a unique autornorphism
of V. In fact the mapping A: Aut(V)—4 GL(2,Z2), the (general linear) group of
nonsingular 2 x 2 matrices with coefficients in Z2, is an isomorphism.
 Automorphisms, Center, Commutator Group 209

Now to show that GL(2,Z2) is isomorphic to let

(I]
[0]
[1)
[I]
[0]
and r=I1 [1] [I] [0]

[0]
Then a2
]

and ara 12. The generators a and r of GL(2, Z2> are subject to the same relations
as the generators
11231 I and fl=I
1123
I 3j 3 1

of E3. Thus the mapping A': GL(2, Z2) described by A'(a) = and A'(r) = ft
is an isomorphism. Composing the mappings A and A' we obtain an isomorphism
Aut (V)
To illustrate the concepts of center and commutator subgroup while
providing experience with computations involving group elements, we examine
the dihedral groups V. n 3, in detail. The group can be described by

= {a3b' : s, i e Z}
where a" = b2 = e and ba = th {cf. Examples 6(1) and 7 of *6.1].

Example 3. The center of n 3, is

{e} if 2tn, and if 21n.

An element a3bt a belongs to the center C of if and only if
a(a'b1) = (&bt)a,
b(a5b') (a5bt)b,

because the "general" element of r has the form axbY. To describe the center more
explicitly, consider separately the cases t = 0, t = I.
The Case t = 0. in this case a5b a3, and conditions on s alone have to be found
such that

aa5 = a'a (no condition on s ensues)

and ba5 a3b.

The relation ba = a_'b yields

ba2 = (ba)a = (a'b)a = a1a'b = a2b,
and in general, ba3 = a 'b. Therefore for an element as in the center, asb = ba5 =
implies that
a2'=e or nI2s.
If 21',,, then nls, so that a3 = e, in which case no element of the form a3 e can
belong to C. If 2 ii, however, then for s = ,,/2 we have = as e', and thus
a"'2 cC'
 Group Theory chapter 6

The Case I = 1. If there exist elements of the form a3b such that a(a'b) = (a5b)a,
then

a' or a2 = e, contradicting the assumption that n 3.

Hence no element of the form a'b lies in the center of

Example 4. The factor group 17 = of F',, modulo its commutator subgroup

F',,1 is isomorphic, for n 3, to Z2 if 21'n, and to the Klein Four-group if 21n.
The elements of the factor group 17 can be denoted by
[axbY], [ai, (b'], [e] F',,',

where 0 x < n and 0 y < 2, because [an] = (b2) [e]. Proposition 3 asserts
that 17 is abelian. Thus
[b][a) = (at][b] = [b][a]' and [a]2 = [e].

Consequently a2 E F','. We now distinguish two cases.

The Case 21',z. If then
a= ae = aa" = a F,,'.
Because bab1 a E (a>, the subgroup <a> (of order n) of F, generated by a is a
normal subgroup of F,,. Since r,,/<a> has order 2 it is abelian; therefore by Proposi-
tion 3, <a> r,'. But a a F,,' implies that F',,'. Thus, when 2 4n, = <a>,
and F',/F,,' Z2.
The Case 21,,. As before a2 E F',,, and so <a2> F,'. Moreover <a2> is a normal
subgroup of F',, since ha2 = a 2b (in Example 3 we had = a3b) implies that
ba2b' = a (a2>. The factor group F,,/(a2> has order 4 and so is abelian
[cf. Exercise 9, §6.6]. By Proposition 3, F',,', and therefore F',,' = (a2>.
The four elements of the factor group F: F,/F,,' are
e' = Na2>
a'
=
a'b' = ab(a2> a\a2>b.
Furthermore r: is isomorphic to the Klein Four-group, as each element different
from e' has order 2.

Example 5. We shall prove in §7.4 (Proposition 3) that the center of a group of

prime power order is necessarily nontrivial. As a consequence any group of order
p2 is abelian, p a prime.

Example 6. As a final example, consider again the nonabelian group

G = {ambn
where = = e and bab' a' +P, of order p3. where p is an odd prime, as in
Example 2, §6.4. Since
ba2b'
=
§6.9 Automorphisms, Center, Commutator Group 211

we can show by induction that for all 1 p2,

baab_l =
Furthermore,
b2a0b2 = b(baab_I)b_I
= a°" +p)(I +p)
= +p)b—
=
and a second induction argument shows that for all ft. I ft p,
= or b"a° = a1f9)'bfl.
As noted in Example 2, §6.4, the subgroup N generated by a normal
subgroup of G and Gf N is abelian. Hence N 2 G1. However, since & bab
G', the subgroup K&> = N is contained in G'. Thus G1 =
To determine the center C of G, we consider conditions on in, n such that
C. First, we must have
=
=

or a"" + P) = am. Hence p2 I nip or p m. Therefore elements of the center must have
the form In addition,
= =
or = (a'ba)"
= = a"b",
since & e C from equation (s) with m = p, n = 0. Therefore = e, or p n. Hence
b" =e. Thus C = :0 s <p1, the group of order p generated by

Exercises

1. a. Determine (he commutator subgroup F',' of the dihedral group F', of

order 14.
b.Find a subgroup of the multiplicative group of the nonzero
complex numbers which is isomorphic to the factor group F',fF','.
2. Let G be the multiplicative abelian group of order ptm in which each
element g e has order p.
a. Prove that Aut(G) is isomorphic to the group of units (invertible
elements) in the ring of ni x rn matrices with coefficients in the field
Z,,. Hint: Use the fact that G is isomorphic to the standard rn-dimen-
sional vector space V over considered as an additive group [cf.
Example 2].
b. Find the order of Aut(G).
3. Prove that if H is a normal subgroup of order 2 in a group G, then H is
contained in the center of 6.
4. With reference to Examplc 6, prove that the generator a is a power of
the commutator aba 1b
 Group Theory chapter 6

5. Let G be the group of Example 4.

a. Determine all homomorphisms of the factor group Gf C into the
multiplicative group T of all complex numbers of absolute value I.
b. Write b2a2ba in the form bxaY.
6. Let G be a finite group of more than two elements, but not an abelian
group of order 2', for any SE N. Prove that G has at least one auto-
morphism a distinct from the identity.
7. Consider the integers as an additive group. Prove that Aut(Z) is cyclic
of order 2.
8. Let G be the quaternion group of 8 elements [Exercise 6, §6.5].
a. Determine the center C of G.
b. Determine the commutator subgroup Gt.
c. Prove that the automorphism group Aut(G) is isomorphic to
d. Prove that the group of inner automorphisms 1(G) has 4 elements.
e. Is 1(G) cyclic or is it isomorphic to Klein's Four-group?
9. With reference to Example 7(b), §6.1, and Exercise 7, §6.4, consider the
dihedral group V4 of order 8.
a. Find the automorphism Aut(V4).
b. Find the group 1(r4) of inner automorphisms.
c. Prove that o((Aut(r4)) = 8. (In fact, V4.)
d. Prove that o(1(f4)) = 4.
e. Describe the factor group Aut(F4)//(V4).
10. Prove that if x is the only element of order 2 in a group G, then x must
lie in the center of G.
11. Let G be a finite group in which = for all a,b€ G and somc
n e N. Denote by the set : a C G} and by G(fl) the set
C G : = e}. Prove the following:
a. and G(fl) are normal subgroups of G.
b. : e] = [G :
G a normal cyclic subgroup N of
order ni whose quotient group GIN is cyclic, prove that G is generated
by two elements a,b for which am e, = a', and bab =
where the integers r,s satisfy r(s—l) ?—l 0 (mod,,:), Hint:
Let b be a representative of a generator of GIN.
b.Give an example of such a group, for ni, ii> I.
13. a. With reference to Exercise 12, express a power of a.
b. Determine the commutator subgroup in the general case in Exercise
12(a) and for your example in Exercise 12(b).
14. in a group G prove that the commutator group is the intersection of all
subgroups which contain all of the commutators in G.

§6.10 Direct Product

The concept of direct product. introduced by Otto HOlder (1859—
1937), has applications to the construction of new groups from given ones
and to the analysis of a given group as a composite of subgroups. These
applications are the group-theoretic analogues of the concepts of (internal
 Direct Product 213

and external) direct sums (products) encountered in ring theory and

in vector spaces Recall also the explicit examples of direct sums in
the rings Zm ifl §3.5.
When the group operation is additive, or in the case of rings, we speak
of direct sums: when the operation is multiplicative, we speak of direct
products.
The product H K of two subgroups H and K of a group G was
defined at the end of §6.2 to be the least subgroup of G containing both H
and K. Proposition 2, §6.4. asserted that if either H or K is a normal subgroup
of G. then the product H. K equals the set {hk liE H, k e K}. in this section,
:

we consider only normal subgroups of a group G.

A group G is said to be the internal direct product G = N, ® ®
of the normal subgroups N1, I i s, if
(I)

(ii) (N, {e}, I i<s.

An equivalent statement of this definition is that G is an internal direct
product if

(1)

G has a unique expression as a product g =q, with

components I

Many authors use the notation ® for the tensor product, a concept not
considered in this book. Hence no confusion should result in our use of this symbol
to designate the internal direct product.

To prove this equivalence, first assume (ii) and suppose that g has
two expressions as a product of elements of the N1:

with E N..

Then

In addition, being the product of elements in the

element x belongs to N, Hence by (ii), = e or = h3. We
can show successively that = h5_,, ..., g, = h,.
Conversely, suppose that N, N. Then y has two ex-
pressions asa product of elements of I

y = y,

By the uniqueness assumption (ii)*,

It
y1e ••'
cN1+1 EN,
e and

we must have = e,
=e
IeN1
eye
I
e.
IEN,
Ij 1, and
y = e. But then (N, N.) N.4, = {e} for all j, I i <s.
If G=N, ®N2, then g,g2=g2g, for g1eN,, g2eN2. First
and by the
normality of N1 and N2. Consequently g1g2g,'g2' eN, N2 = (e}. An
214 Group Theory chapter 6

inductive argument proves a corresponding result for the product of any

finite number of normal subgroups.

Proposition I. if G = N1 0 0 is the direct product of normal sub-

groups then forgE N1 and E = where I

This is not the same as saying that G is an abelian group.

Consider s groups G, with respective identity elements e. The set
G={(91,...,g3):91EG1} with
(i) = (h1, ..., hj if 91 = h,, 1 i s,
and
(ii) (g1,...,g5)(h1,...,h5) =
is called the external direct product G1 k ... G5 of the groups G.

That the set G is a group is readily checked because the multiplication

is defined componentwise. Associativity follows from the fact that the product
in each of the components G. is associative. Furthermore, (e1, ..., e is
1)
the unit for multiplication, and is the inverse of(g1,

Proposition 2. The (external) direct product G = * ... * of groups

G,, 1 contains subgroups N isomorphic to G4, I and

Proof Define mappings q4: G1—.G by for each 1,

I <I s.The image subgroup N1 = ç1(G1) G is isomorphic to G,, since
kerq1 = {e4}. Further, N1 is a normal subgroup of G because
(h1 ...,h5)

=
e5)

= 41(h191h4') eN4,
where for Ij
s. and eG. By the defInition of equality
in G, equals the product of unique
elements (e1, ..,g, e3) E N4. Therefore G is the (internal) direct product
of the groups N4 as asserted.

Proposition 3. Let G be a cyclic group of order n = n1 where

for1 Then G is the internal direct product of s subgroups G1,
where o(G1) = n•.

This proposition is the group-theoretic analogue of the (internal)

direct sum decomposition of in §3.5. (Recall the theorem in §6.6, which
states Zn.)
To prove Proposition 3 without reference to §3.5. consider a generator
g of G, and set rn4=n/n4, I Then, as in Proposition 2, §6.6, the
 Direct Product 215

element 9mj generates a subgroup <gmi> = G1 of order n1. To show that

G=G, ® ® note first that (rn,,m2 I and that there exist
integers h1 such that I = in, h1 + For an arbitrary element a e G,

a = a' E

hence
To prove that this product is direct, note that xe (G, ... G.)
implies (since xe G1+,) and o(.v)In, n (since x is the product
of elements x. such that 0(x) In1). Consequently x = e, because (n, ...
,)= I and hence o(x)I I. Thus. G equals the internal direct product
G1 ® ®
G is the (internal) direct product G, ® G2,
and that H, and 112 are normal subgroups of G, and G2, respectively. Then
the factor group (G, ® G2)/(H, ® is isomorphic to the (external) direct
product G,/H, * G2/H2.
Proof. Recall from Proposition I that G = G, ® G2 means that g G can
be expressed uniquely as g = 9,92' for g, E G,, g2 e G2, and moreover that
92 = 929,• We use these facts to construct the mapping
tfr: G G,/H, x G2/H2
given by = H,. 92 112) for g = 9,92. The uniqueness of the expression
of g E Gas the product of elements in G, and G2 implies that cu is well-defined.
We now prove that cu is a surjective homomorphism whose kernel is
H,® 112.
First, is a homomorphism. For g = g, 9i' = 9'2 in G,
cli(gg') =
=
=
(g, !1,,g2 112)(g, H,.g'2 H2) =
Ifg =9,92 e then
= (91 H,,92 H2) = (eH,,e112)
implies thatg, E H,,92 112. Hence kercu' H, ® 112. Conversely ç&(h,h2)=
(eH1,eH2), for all H.. 1=1,2, and so ker,I/= 113 ® H2. Next, the
homomorphism is surjective because the general element in G,/H, x G2/H2
is (g, H,,g2H2), the image of 9,92 E G under k/i.
Since is a surjective homomorphism with kercl, = H, ® H2, the
Isomorphism Theorem assures the existence of an isomorphism
G/(H, ® H2) G,/H, * G2/H2.
Explicitly, the map q is given by
(p[g(!l, ® 112)11 = (g,H1,g,H2),
where g=g,g2.
216 Group Theory chapter 6

Exercises

1. Let G be the external direct sum of Z, and Zm, considered as additive

groups, where (rn, n) = 1.
a. Prove that G is isomorphic to Zmn.
b. Find the image of the subgroup of G in the group Zmn.
2. Let G be the external direct sum of the additive groups Z8 and Z6.
a. Determine the addition table of the factor group G/H where H is the
subgroup generated by x = ([418, [3]6), [41s e Z8 and [316 c Z6.
b. Determine the addition table of the factor group GIK where
K (4]8Z8 4- [316Z6.
3. Find all homomorphisms of the direct sum G = Z3 i-Z3 into Z6.
4. Find all homomorphisms of the direct sum G = Z2 -I- Z2 into the direct
sum r z2+z2÷z6.
5.Consider a group G = HN where H is a subgroup of G and N is a normal
subgroup of G such that H N = {eL Prove that the factor group GIN
is isomorphic to H.
6. If G is the (external) direct product of subgroups H, K, and if L is a
subgroup of G containing prove that L = H*(L x K)).
7. Consider a finite abelian group G for which o(g) = p for all g e in G.
Prove that o(G) = pm, for some e N.
8. a. Prove Proposition 1.
ii. Conclude that elements of N1 commute with those in N1 ®

9. For distinct groups G and G', prove that

G * G' C' C,

but that these two (external) direct products are not equal.

10. If a group C is the product NK of normal subgroups N,K and if

M =N K, prove that

G/M = N/M ® KIM.

11. The following resolution of the structure of the multiplicative group of
units (group of prime residue classes) in the residue class ring Z,, is
due to Gauss (1801). Prove each statement.
a. If p is an odd prime, then U,,.. and are cyclic groups of order
(p_I)pml for allmeN.
(i) Prove by induction on rn the existence of an element w of order
p—I in U,,...
(ii) Prove by induction on m the existence of an element z of order
pm_I in u,,..
(iii) Note that w and z generate U,,.., and since (o(w),o(z)) 1,
that o(wz) = (p—
(iv) Observe that Ui,,.. U2 U,,.. U,,...

b. Ifp=2,then
(i) U2 is cyclic of order 2° = 1.
(ii) is cyclic of order 21 2.
 Homomorphisms of Abelian Groups 217

(iii) for in 3, is the direct product of a subgroup of order 2,

generated by [— l)2 and a cyclic subgroup of order 2m_2,
generated by 1512-.
c. If the modulus ii is not of the form ptm, 2pm (p an odd prime), 2, or
22, then is not a cyclic group.
12. Suppose that G is a finite abelian group of minimal exponent n (the least
positive integer such that g" = e for all g e G). Assume n has a proper
factorizationn = mq where (m, q) = 1. Let S, T be the subsets of elements
s,!, respectively, in G such that = = e.
a. Prove that S and 1' are subgroups.
b. Prove that the mapping T—÷ G given by f((s,t)) = st is an
isomorphism.
c. Conclude that G = S® T.

§6.11 Homomorphisms of Abelian Groups

This section develops the group-theoretic analogue of the argument
in §4.3 that the set HomF(U. V) of linear transformations from one vector
space to another (over the same field F) is itself a vector space over F.
Explicit examples show how the definitions lead from detailed questions on
homomorphisms to simple problems in the theory of integers.
Let G and F be two multiplicatively written abe/ian groups. Consider
the totality Hom(G, r) of homomorphisms of G into F. The set Hom(G, F)
is itself an abelian group when the product in Hom(G,F) of two homo-
morphisms is defined to be the mapping G—*F, described for all
ge G by
9 =
Since q(g) and are elements in the abel/an group F, = More
importantly, is again a homomorphism: for all g, Ii e G,
(qnfr)(gh)
= ço(g)ço(h)i/i(g)i/i(h)
= = (pçli)(g).(pi/i)(h).
The homomorphism c: defined by s(g)=e, the identity
element in F, serves as an identity element in Hom(G, F) with respect to the
product just defined, since
(qe)(g) = = p(g)e = p(g) for all g E G.
The proof of the associativity, = of multiplication in
Hom(G, F) is left as an exercise.
Finally, for a given E Hom(G, F) the mapping p ': G -+ F defined
by '(9) = is the inverse of (in the sense of the law of com-
position for Flom(G, F)) because = ç' = e. (Check that is itself
 Group Theory chapter 6

a homomorphism from G to f.) Thus Hom(G, is an abelian group, as

asserted.

Propositions and 2 develop simple properties of the group

1

Hom(G, with respect to (internal and external) direct products.

Proposition I. If G = G1 ® G2, then

Hom(G, ® G2,T) Hom(G,,F) Hom(G2,r).
To begin the proof, define for each e I-Iom(G, the "restriction"
mappings p,(4'): r, for 1= 1,2, by

= EU
for all g1 G1. Then p1(q,) Hom(G1, f') because

=
= =
i= 1,2.
Next, the mapping

•: Hom(G,f) -# Hom(G1,U) Hom(G2,fl

given by b(4)= (P1(4'). P2 (c')) is an isomorphism. First,

= for e Hom(G,U).
since p.(4'4")(g1) = =
=
for all g, e i = 1, 2. Since 4 E ker implies that p.(4') =;, i = 1,2,
where = c, for all g.€ G1, we have =
Finally, to show that is surjective consider ç', e F),
I = 1,2, and define 4'(g) = g expressed
uniquely as Then (pE Hom(G,U) and =(q1,q'2). Thus
'D is an isomorphism, as asserted.

Proposition 2. If F = F1 ® F2, then

Hom(G,f1 ® Hom(G,U1) Hom(G, IT2).

The proof is left as an exercise.

Example I. We use Propositions I and 2 to determine Hom(G, F) when G = C6,

the cyclic group of order 6 with generator g, and U is the abelian group of order 27
with generators a,b of respective orders 3,9.
First, G <92> ® Kg3> and r <a> where (a> denotes the sub-
group generated by a, etc. Hence,
Hom(G,f) Hom(<g2>,(a>) Hom(<g2>,<b>)
Hom((g'>,<a>) Hom(<g3>,<b>).
 Homomorphisms of Abellan Groups 219

Elementary arguments on the orders of elements and their images yield

Hom(<g2>,<a\) has order 3. Simply r=O, 1, or 2.
has order 3. Simply map g2 —. b3', r = 0, 1, or 2. These
are the only possible homomorphic images of g2 since
o(p(g2))lo(g2) = 3. (These are the only elements in
whose order divides 3.)
isadivisorofo(g3)= 2
(by Proposition 3, §6.5) and of o(<a>) = 3 implies
o(ç(g3)) = I for any homomorphism p.
Hom(<g3\<b>) = te), since, as above, for any homomorphism
II = (2,9).
Any group of order 3 is cyclic, and isomorphic to C3. Therefore
Hom(G,r) C3 * C3.
The elements of Hom(G, fl can be described on the generator g c G by

Example 2. We now describe Hom(r,G) for the groups G,r of Example 1. From
Propositions I and 2 we have
I-Iorn(r,G) * Hom(Ka>,<93>)
* Horn <92>) * Horn ((b>, (g3>).
Analogous to the discussion in Exercise 1:
Horn(Ka>,<g2") C3,
Hom(<a>, <'g3') =
C3,

Hom(Kb>,<g3>) =
Thus Hom(r, G) C3 * C3, and its elements can be described in terms of the
generators a,b of r by
= (g2 = +

Exercises

1. Verify that each of the mappings cor.,, 0 r, s < 3 defined in Example I

is a homomorphism.
2. Prove that the multiplicative group Tof all complex numbers of absolute
value 1 is isomorphic to the additive group R/Z of cosets of real numbers
R modulo the additive group of integers Z.
3. Let G be a cyclic group of order n. Prove the following:
a. Horn(G, T), where T is the group of Exercise 2, is a cyclic group of
order n.
b. Hom(G, T) contains (the Euler function) distinct isomorphisms
from G to T.
220 Group Theory chapter 6

4. Consider Z as an additive group. Prove the following.

Hom(Z, T), where T is the group of Exercise 2, is isomorphic to T.
a.
b. The isomorphisms in Hom(Z, T) can be put in one-one correspond-
ence with the irrational numbers modulo 1.
5. a. Verify that the law of composition defined for Hom(G,f) is
associative.
b. Verify that the mapping çot: G-.. r defined by
for e Hom(G, is itself a homomorphism.
6. Determine the following groups.
a. b. Hom(C12,C14)
c. Hom(C2 C4, C14).
7. Prove Proposition 2. Hint: Show that E Hom(G, r1 ® f'2) gives rise to
mappings e and define
'1': Hom(G,f1 ®F2) -. Hom(G,f'1) Hom(G,f'2)
by 'P(q)
8. Extend the statement of Propositions I and 2 to the case of .s factors.
9. a. Determine explicitly, without reference to Propositions I and 2, the
group Hom(G, r) of Example I.
b. Determine the kernel of each of the elements of Hom(G, r).
10. a. Describe Hom(Cm, for (rn,n) = I (cf. Proposition 3, §6.10].
b. Describe Hom(Cm, when (rn,n) = d> 1.
11. For an abelian group G, prove that Hom(G,G) can be considered as a
(not necessarily commutative) ring, called the ring of endomorphisms of
G, when a second law of operation is defined to be composition of
mappings a = for g€ G, e Hom(G, G).
12. With reference to Exercise 11, describe the ring or endomorphisms of
the following:
a. p a prime b. Cq C. C2 C2.
 7

Selected Topics in Group

Theory

No introductory text can develop all the important aspects of group

theory. We have chosen to present here five topics of special importance in
applications. They may be studied in any order. Analysis of the structure of
finitely generated abelian groups is a prerequisite for the study of
algebraic topology and is basic to the group-theoretic classification of abelian
groups. Topologists and quantum physicists, among others, are concerned
with characters (an analogue of linear functionals on finite dimensional vector
spaces) of finite abelian groups Elementary analysis of finite (non-
abelian) groups is based on the theorems of Sylow through 7.5].
Composition series of groups play an important role in the examination
of finite groups and in the analysis of field extensions, where the Galois
theory relates group and field-theoretic arguments. Commutative algebra
and algebraic geometry place major emphasis on modules, which are viewed
in §7.8 as special cases of groups with operators

§7.1 Finitely Generated Ahelian Groups

This section is devoted to the statement and proof of the Fundamental
Theorem of Finitely Generated Abelian Groups; it provides a complete
description (up to isomorphism) of the structure of such groups. For
221
222 Selected Topics in Group Theory chapter 7

notational convenience we consider additively-written abelian groups and

speak of direct sums. it is a simple exercise to rewrite the definition, state-
ments, and proofs muitiplicatively.
An abelian group G is said to be finitely generated if there exists a
finite subset of elements {A1 Am) c G such that every element A G
can be written as the sum
A = C1 .A1 + ... + Cmhlm, C Z, I , Pfl.
The elements A., I I m, are called generators of G, and c1•A1 denotes
the c1-fold sum of A, with itself. (If < 0, then = (— As).)
No claim is made that the integers c, in a given representation of
A e G are uniquely determined by A; this definition parallels the concept of
a finitely generated vector space [cf. §4.1].

For example, in the additively-written Klein Four-group V with distinct

elements a,b,c,O such that a+a b+b = c-i-c = 0, and a+b = c, the element a
has the following two expressions in terms of generators a, b, c:
a = a + O•b + 0.c,
a= + b + c.
Note that any finite abelian group is finitely generated, as its own
elements constitute a finite set of generators.

Fundamental Theorem of Finitely Generated Abelian Groups. Every finitely

generated abelian group G 0 {0} is the (internal) direct sum of N cyclic
subgroups G1, ..., Gr+ i, ..., where

(1) o(G1) = e, 2, I< I r,

(ii) e1

(lii) ..., are isomorphic to Z.

The numbers r, N and the chain of divisors e1, ..., are uniquely determined
by G. However, the groups G1, I I N, are not uniquely determined when
N is greater than 1.

Fundamental Theorem of Finitely Generated Abelian Groups (Alternate Form).

Every finitely generated (nontrivial) abelian group is isomorphic to a direct
sum of cyclic groups

where q. and the primes are not necessarily distinct; the prime powers
q1, I I s, and the number of summands Z are uniquely determined.

We split the proof of the Fundamental Theorem into separate

existence and uniqueness proofs, beginning with the existence proof. Then
we introduce some terminology and two lemmas before giving the uniqueness
 Finitely Generated Abelian Groups 223

proof. Because of their length we number the major sections of each proof.
The proof of the equivalence of the two forms of the Fundamental Theorem
is left as an exercise. On a first reading, some may prefer to accept the
Fundamental Theorem without proof and to gain understanding of its meaning
by working out a number of the exercises.
The Existence Proof
1. Consider the collection of all finite sets of generators of G.
Let n be the in inimal number of elements in the sets of generators. The existence
of such an integer n is a consequence of the Well-Ordering Principle. We
have two cases.
1(a). If n = 1, the group G is generated by a single element A. If
a = 0, then G is an infinite cyclic group isomorphic to Z.
If a•A = 0 for some a 0, there is a least positive integer m such that
A =0, and G is isomorphic to Zm [cf. the theorem in §6.6]. (Note that the
uniqueness statement is valid trivially.)
1(b). If n> 1, the proof will be by induction on n.

2. Consider a!! sets of n generators, X1, .. ., of G. Furthermore,

consider a!! relations

for a!! sets of n generators.

First no relation has a coefficient x that is a unit (namely, ± I).
For, if there were such a coefficient x1 then
X1=—xi—lx1.X'--...-—xi—ixi—'.Xi—I—xi—1xi+1.X'+'—...—xi—lxn.Xn
could be omitted as a generator, because the set {X1, ..., — 11 + ...,
of n— I elements would generate G, contrary to the minimal choice of n.
There either is or is not a set of generators X1, ..., such that
0. I1 + ... + 0. = 0 is the only relation between them.
2(a). In the first case A e G has a unique representation

(If also then

whence = 0.) Thus A is the n-fold direct sum of the infinite cyclic groups
<11> = {z.X1: z Z}, I i n, and the existence proof is finished [see
§6.10 for properties of the internal direct product].
2(b). In the second case all sets of coefficients (x1, (which
are not all zero) for all relations of all sets of n generators contain at least
one coefficient x satisfying xI> 1. Therefore there exists a system of
generators A1,A2, and a relation
(*) a1 + a2•A2 + + = 0,
such that a1 0 and Ia1 I 1x11 for all nonzero coefficients Xj in all possible
sets of relations.
224 Selected Topics in Group Theory chapter 7

3. .To prove that a1 "i' 2 I ii, pick any particular i. By the

Division Algorithm, a, = a1 q.+ r1, 0 <1a11. If r1 0, relation (s) can
be rewritten
(**) a1 + r1.A1 + = 0, where j I.

In this case the nelements A2,...,A1_1.A1÷1 A,, would

also generate G, but they satisfy relation (**). which has a coefficient less than
a1. This contradicts the choice of a1. Therefore. r, =0 and a = a1 q1,
Setting B1 = Al note that a1 = 0. Now let G1 =
<B1> = {z.B1 : z e Z} and G2 = {A2, ..., the group of integral linear
combinations of A2, ..., A,,. We have G = G1 + G2, because A1 * ..., A,,
generate G.
4. Next, to prove that this sum is direct consider Ce G1 G2.
Then

C= . B1 A1.
=

and consequently 0= . A..

—

Using the Division Algorithm, write Y1 =q1a1+z1, where

0 = z1 .B1 y1.A1.
—

By the minimal choice of 1a11 the coefficient must be zero. Hence

C = z1 BI = 0.
5. Since G2 has a set of n— I generators, the induction hypothesis
implies that
G2 = {A2....,A,,}7 = <B2> <Br> <B,+1> $ <B,,>,
whereo(BJ=e,2, and

<Br+i> <B,+2> ... <B,,> Z.

6. It remains to show that a11=e1 divides e2. Suppose that
e2=e1q+s with 0<s<e1. The elements B1+q.B2,B2,B3,...,B,, generate
G, and

This contradicts the minimal choice of a1 as the coefficient of least magnitude

of any relation. Therefore s = 0, and e1 e2. This concludes the existence
proof.
§7.1 Finitely Generated Abelian Groups 225

The uniqueness proof uses the following concepts and two lemmas.
An element in a group G is called a torsion element if it has finite
order. The torsion elements of an abelian group form a subgroup, the so-
called torsion group G is called torsion free if its identity element
is the only torsion element.
A group is called a p-group if its order is a power of a prime p; in
particular, an abelian p-group is called a p-primary group. We also refer
to the minimal exponent m of a finite group G, that is, the least positive
integer for which, in additive notation, = e for all g e G.

Lemma 1. If G, is the torsion subgroup of an abelian group G, then G/G,

is torsion free.

For the proof consider an element (i.e., an additive coset) g+G,

of of finite order m. Since m.(g+G,) = (tn•g)+G,, we have
(m•g)+G, = 0, or m•g G1. This means that m.g has finite order, say r,
and therefore g has order at most rm. Thus g e and g+ G, = 0 G/G,.

Lemma 2. For a p-primary group G with minimal exponent p2.

where the G. are cyclic of order ps', I ;= and where

the ; are uniquely determined by G.

Note that is is not claimed that the groups ci, are unique, only their
orders p2' are.

For instance the Klein Four-group V, referred to in the third paragraph of

this section, can be written
V = <a> <b> =
<b> <c>, but o(b) = o(c) = 2.

Proof of Lemma 2. The existence of cyclic groups ci, is a consequence of

the existence portion of the Fundamental Theorem, already proved, applied
to the given group G. To prove the uniqueness of the exponents ;, we
proceed by induction on the minimal exponent p2 of G.
If = I, then o(G,) = p. and G. Z,,. Because ci is the direct sum
of the groups G., its order 0(G) = o(G,) The number k of cyclic
summands is uniquely determined by the order of ci.
Now assume that 2. As the induction hypothesis suppose that
the lemma is true for p-primary groups with minimal exponent less than p8.
The group pG = {p.g : g e ci) has minimal exponent p2 ',and

where we omit the trivial summands (if any) pG1 PGm. The cyclic group
226 Selected Topics in Group Theory chapter 7

is generated by the p-fold sum of a generator of Since

By the induction hypothesis the integers;—l, for #n.czik.
are unique, and so then are the;, ,n < i k.
The number of summands Gi,...,Gm isomorphic to Z,, is unique,
because the number k—n, of direct summands of pG is unique by the
induction hypothesis and
k k k

o(G) = fl 0(G1) = pk
fl o(pG1) = fl fl = ptm
1m+l
Application of the Principle of Induction completes the proof of Lemma 2.
Returning to the Fundamental Theorem, we now take up the
uniqueness proof.
The Uniqueness Proof Given G = ... $ G, $
G1 ... ® G,,. where
is cyclic of order e11e21...Ie,, and
I
we have to prove the uniqueness of the integers n—r, r, and e1, 1 I r.
Pick generators I

I. To prove the uniqueness of the number n — r of generators of G

of infinite order, let G* = GIG,. By Lemma I, G* is torsion free. Letting B'
be the coset Bk+G, of in G*, r<kn, note that generate
G* and satisfy only the trivial relation B7.1.1 + ... + 0. B,,4' =
1(a). If there were a nontrivial (i.e., not all 0) relation

then we would have

Hence for some nonzero : E Z,

B,+ + ... + B,, = 0.
contradicting the choice of the B, + B,,.
1(b). To complete the proof of the uniqueness of n—r, select any
positive prime p. Since
= <B:>
and pG4' = ...
Proposition 4, §6.10, yields
(4') G4'/pG4' -I- -I- <B:>/<p.B:>.
Because has order p, for r < k n, and the sum in equation (a)
is direct, G4'/pG4' has Now G,, G4'. and hence G*/pG* are uniquely
determined by G (i.e., are independent of the representation of G as the direct
sum of cyclic subgroups ).Therefore we conclude that n—r is uniquely
determined by G, since p"' = o(G*IpG*).
2. To prove the uniqueness of e1 e,. it suffices to consider the
torsion subgroup G,. Let M = e= o(G,). Writing M as the product of
 Finitely Generated Abelian Groups 227

powers of distinct primes,

M =

set = and set = {g e : p1 divides o(g)}, I j S.

2(a). We now show that

Since (n,1, ...,m5) = I, we know {from Exercise 12, §2.5] that there exist
integers h1 such that I = h1 + +h3ni,. Thus for A G,,
A = 1.4 = (h1m1)•A + +
Noting that for all A E the order of divides p', we have that
E Hence

2(b). This sum is direcL The order of any element

I <j s,
divides (being in H1) and divides q = (being in + +
Therefore o(X) divides the GCD (q. = 1, from which we conclude
that o(X) = I and X = 0.
3. Being subgroups of Gk, the groups = H1 Gk, Ij
I k r, are as they are
cyclic. Their respective orders are powers of
subgroups of H1. Let o(HJk) = elk. From the direct sum representations

H1®
we obtain
(**) = H1 G, = H11 ® •.. ®
= n G, = ... ® I1sk'
where some of the summands might be zero.
Since these sums are direct. = o(Gk) = eJk. Thus to prove
the uniqueness of the orders it suffices to prove the uniqueness of the
integers For this we shall use Lemma 2.
4. Each group 1is pd-primary and so, by Lemma 2,
has a representation as the direct sum of cyclic subgroups, whose number and
orders are unique. Since equation (*s) presents 11, as the direct sum of cyclic
subgroups HJk, the orders of these subgroups must be unique. This com-
pletes the uniqueness proof.
The uniqueness proof implies that the number n, determined (in
part l of the existence proof) to be the minimal number of generators of G,
is the same as the integer N in the statement of the theorem.
 Selected Topics in Group Theory chapter 7

REItIARK. If Z is replaced throughout this section by F[x], the absolute value used
in the existence part of the theorem is replaced by the degree of a polynomial, and
order of a group element is replaced by "monic polynomial in(x) of least degree
such that ,n(x).A = the theorem is the foundation for the theory of "rational
similarity of matrices" [see Exercises 20 and 21, §7.8].

Exercises

I. if G is an abelian, noncyclic group of order p2. prove without reference

to the structure theorems that
G -I- Z,,.

2. Prove the equivalence of the two forms of the Fundamental Theorem of

Finitely Generated Abelian Groups.
3. How many nonisomorphic abelian groups are there of each of the
following orders?
a. 35 b. 36 c. 35.36
d. 48 e. 49 f. 48.49
g. 24 h. 63 i. 24.63.
4. If in = in1 (m,,,n2) = I, and if there are nonisomorphic abelian
groups of order in,, i = 1,2, how many nonisomorphic abelian groups
are there of order m?
5. Generalize the result of Exercise 4 to the case where m is the product of
r pairwise relatively prime factors in,, for which there are ii, nonisomor-
phic abelian groups of order in,.
6. For what integers in must all abelian groups of order in be cyclic?
7. Complete the blanks.
a. There are nonisomorphic abelian groups of order 275.
b. An abelian group of order 882 must be isomorphic to one of the
following groups: , , ,or
8.Can you determine how many subgroups of order 8 there are in:
An abelian group of order 48?
a.
An abelian group of order 200?
b.
9. Prove that for an abelian group G:
a. lf a prime p1 o(G), then G has a (cyclic) subgroup of order p.
b. If an arbitrary integer in I o(G), then C has a subgroup of order in.
c. If in part (b) no square divides in, then the subgroup must be cyclic.
d. If in part (b) a square does divide m, then C might have no cyclic
subgroup of order in.
10. For arbitrary in E N describe the number (up to isomorphism) of abelian
groups of order in (in terms of the unique prime power factorization
of in).
II. Exhibit all nonisomorphic abelian groups of order:
a. 60 b. 360 c. 23527.
§7.1 Finitely Generated Abelian Groups 229

12. Let Fbe isomorphic to the external direct sum of,: copies of the additive
abelian group Z. Write Fas the internal direct sum of ,z groups which are
isomorphic to Z. (Such groups are called free abelian groups of rank n.)
13. Let F= {(a,h) : a,b Z} be a free abelian group of rank 2. Suppose that
H is the subgroup of F generated over Z by the elements (2, —4) and
(1,6). Determine the structure of the (additively-written) factor group
F/H in the sense of the Fundamental Theorem.
14. Let A be the free abelian group with generators u and :', i.e., A =
{au+bt' : a,b E Z}. Consider the subgroup B of A generated over Z by
the elements x = 21,— 3v and y = r. Determine the structure of the
factor group A/B.
15. Let F= {(a1,a2,a3) a1,a2,a3 e Z} be a free abelian group of rank 3.
Suppose that H is the subgroup of F generated over Z by the elements
(2, —4,6) and (1,6,7).
a. Determine the structure of the (additively-written) factor group F/H.
b. Describe the group Aut(F) of automorphisms of F.
16. Let A be the free abelian group with generators ii, v, and i.e.,
A = {au + by + cw : a, b,c Z}. Suppose that B is the subgroup generated
over Z by the elements x = 2:i — v + 4w, •v = 4u + 5v + 6w, and z =
3u+ 2v+ 5w.
a.Is the factor group A/B a finite group?
b.Determine A/B as a sum of cyclic groups.
c.Are the sumniands of A as a direct sum uniquely determined?
17. Consider Z as an additive group.
a. Determine all groups G such that there is a surjective homomor-
phism Z— 6.
b. Determine the kernels of all such homomorphisms
18. Prove (by induction on n) that any subgroup of a free abelian group of
rank n must itself be free.
19. a. Prove that there are as many nonisomorphic abelian groups of order
p° as there are of order q°, where p and q are arbitrary positive prime
numbers.
b. Consider an abelian group C of order puqan:, where p and q are
distinct primes and (p, ni) = (q, rn) = L Show that G has exactly one
subgroup of order p° and one of order Must the number of
subgroups of G of order be the same as the number of those of
order qP, 0 <11 <
20. With reference to Exercise 6, §6.9, prove that if C is a finitely generated
abelian group of order greater than 2, then Aut(G) is a nontrivial group.
21. Let G be the direct product of 'ii groups of prime order p. Prove that C
contains
[a—I
J1(pm_a_l)
hafl(pt_j)i—i
-

distinct subgroups of order pd, I a < in.

22. Suppose that G is a finite abelian group. Prove that
g a direct factor of G. i.e.,
C = <g> ® H, where H is some subgroup of C.
230 Selected Topics in Group Theory chapter 7

§7.2 Characters of Finite Abelian Groups

In §6.11 we studied the group of homomorphisms Horn (G, F), where
G, were arbitrary abelian groups. We now consider the special case when F
1'

is the multiplicative group 7' of all complex numbers z, for which Izi = I.
For a finite abelian group G we call Horn (G, T) the group of
characters or dual group G* of G. The map c: G —÷ { I } c 7' is the identity
element of which we refer to as the identity character. Recall from §6.11
that for q, i/i e Hom(G, T) the product qnfr is the homomorphism on G given
by = ço(g)çls(g).
If n is the minimal exponent of G, then for every Horn (G, T):
(q(g))" = = q(e) = I.
Thus the character values q(g), for every e Hom(G, T) and every g e G,
are dth roots of unity for some d I n, the minimal exponent of G. Consequently
the group T can be replaced by an arbitrary cyclic group of order n.

NOTATION. We shall write the nth roots of unity in Tas powers of the generator
l2rr
= cost—l+isinl—
Cit
\nf
The other generators of the cyclic group of nth roots of unity are
12r\I + islnl—1
.
C,tt .= I,
\'l / /
with (f,n) = 1. An nth root of unity C is said to be primitive if C" = 1, but C' 1,
I I < n; in other words, if C is a generator of the cyclic group of nth roots of unity.
Proposition 1. If G is a cyclic group of order m, then so is its dual G*.

Let g be a generator of G, and define m characters Xk' 0 k <rn,

on G by Xk(9) = Cmk. Since Cm generates the group of mth roots of unity, the rn
powers are distinct; hence the characters Xk are distinct elements of G*.
To prove that G* has no other elements, consider an arbitrary character x.
Then is an ,nth root of unity, and so for somej, 0 j < m,
x= the character Xi has order rn,
the dual group G* = {Xk : 0 k <rn, where Xk(9) = Cm"} is cyclic of order m.
Note that U as they are both cyclic of the same order. Such
an isomorphism is not "natural" in the sense that it depends upon the choice
of the generators of the groups U and G*.

Proposition 2. If G is a finite abelian group, then G* is isomorphic to G.

(This isoinorphisin is not natural.)
Proof Writing U = U1 ® ... ® as the direct product of cyclic groups by
the Fundamental Theorem of Finitely Generated Abelian Groups, §7.1, we
§7.2 Characters of Finite Abelian Groups 231

have
= Hom(G,T) =
Hom(G1, T) * * T)

where by induction we extend Proposition 1, §6.1 1, to the case of s factors.

Proposition 3. If g * e in a finite abelian group G, then there exists a

character x G* such that # 1.
Proof. In terms of the direct product expression G = ® G, fo1low-
ing Proposition 2, e in G, then g=a1 where at least one factor
e. For a character x defined in terms of the s generators of the cyclic
factors G by
X(Ok)=l
and =
where is a generator of the group of n1th roots of unity and n. = o(G,),
we have

= = I, since

a natural isomorphism between G and G** = (G*)*,

the double dual of G.
Proof. By Proposition 2, G but now the problem is to obtain
an isomorphism G :
G** which is independent of the choice of generators
of G. To this end we define by defining K(g): G*_.Tas follows:
K(g)[x] = x(g) e T for every x e G*.
Then is a single-valued function on the dual G* with values in T. To
i.(g)
verifythat g e G, note that by the definition
of K(g) and of the product in G*,
"(9)[x1X21 = (x1x2)(g) = x1(g)x2(g) =
It remains to prove that is an isomorphism. Since
K(g1g2)[xJ = x(g1g2) = x(g1)x(g2)
= e T,

K is a homomorphism. Furthermore, K IS injective. If ic(g) = K(h), then

K(9)[x] = ?C(h)[X] for all xe G*, and so x(/') or
x G*. But then by Proposition 3, gh' must be 1, or g =
The surjectivity of G —* G** is a consequence of the facts that K is
a one-one mapping and that G, G** have the same order [see Proposition 2].
232 Selected Topics in Group Theory chapter 7

The definition of K makes no reference to any choice of generating

elements of G. Thus, is called a "natural" isomorphism.
In the remainder of this section we consider properties of particular
subgroups of the dual group G*. Letting L be a subset of G, define
= (XEG*:X(() = I, for all (eL),
called the annihilator of L. We leave as an exercise the verification of the next
proposition.

Proposition 5. For any subset L G, AG(L) isa subgroup of G*.

Proposition 6. For any subgroup H G, we have AG.[AG(H)] = H, where

we identify II with K(H) and K is the isomorphism of Proposition 4.

Proof By definition A6(H) consists of all those XGG* for which 1

for all Ii e H. Therefore K(h)(X) = I for all x e AG(H); consequently,

K(h) AQ.[AG(H)]. Making the identification of G** with G, we have
Hc AG,[AG(H)] G.
Suppose that H is a proper subgroup of the double annihilator,
i.e., there exists an element g**, more precisely K(g), in the latter which does
not lie in H. This assumption leads to a contradiction as follows. Jfg**H H
in the factor group G/H, then there exists a character qi e (G/H)* such that
(p(g**H) by Proposition 3. Next let it be the canonical homomorphism
of G onto G/H, that is, Tc(g)=gH. Being the composition of homomorphisms
the function (i = çø it from 6 to T belongs to G*. Furthermore, *(h) =
= qi(H) implies e A6(H). Consequently by the definition of
and the assumption on I

according to the choice of iii. Hence Hc is false.

Proposition 7. The mapping H— AG(H) establishes a one-one correspond-

ence between the subgroups of G and G* such that the following relations
hold.
(i) II K implies AG(H) c
(ii) AG(HK) = A(;(H) fl
(iii) K) = AG(H)AG(K).

The definition of annihilator implies that if H K.

If for we had AG(H)=AG(K), then using Proposition 6 we would
obtain
H= = AG.[AG(K)] = K,
a contradiction. Furthermore every subgroup of is given as
where H AG.(S*).
§7.2 Characters of Finite Abelian Groups 233

Next, since HK= HK contains both H and K, the annihilator

AG(HK) is contained in both and A6(K). Thus,
AG(II) n AG(K). For an element ço in this intersection, p(h) = 4(k) = I

for every K and

Ii Therefore
k e K. e A0(HK) and AG(HK) =
AG(H) AG(K), which proves (ii). [The product of groups was introduced
in §6.2.}
Finally, since 11 K is contained in both H and K. its annihilator
K) contains both and A6(K); hence
K) AG(H)AG(K).
If this last inclusion were proper, the mapping AG*(S*) G would
imply that
K)] = H n K c AG.[AG(H)AG(K)]
= AG*[AG(H)1 Ao.[AG(K)1 = H K,
according to (i) and (ii). This is a contradiction.

Proposition 8. For a subgroup If of G,

AG(H) (G/H)*,
and G*/AG(H)

Note carefully that 11* is not a subgroup of G*. To prove the first
assertion, describe a mapping a: AG(H)—+ (G/H)* by defining a(p) by
a(Q)[gHJ = p(g) for p A6(H) and g E G.

The function a acts uniquely on the cosets 9H in G/H, because if g' = gh

with h H, then q(g') = q,(gh) = q(g)qi(/:) = (p(g) since (p e AQ(H). It is a
homomorphism because for all cosetsgH:
a((pqi')[gH] = = q(g)q'(g) =
The mapping a is injective. if = = 1 for all g e
then (p = s. Finally a is surjective. Consider e (G/H)* and define
4) = it, where it is the canonical homomorphism of G onto Gill. That is,
(p(g) = (4)*it)(g) = 4)*[fl(g)] = (p*(YH)
Then (p(991) = (p*(gylff) = çQ*(gH.gIH)
= =
and so e Next 4(h) = (p*(hH) = (p*(H) = I for all Ii H, thus
e Furthermore, (p(g) 4)*(gH) for all ,q G implies
that a(p) =
To prove the second assertion, define a mapping r from G* to H*
as the restriction map
= x(I')
234 Selected Topics in Group Theory chapter 7

for all Ii E H. Clearly t(x) e H*, and t is a homomorphism since r(xx')[I,] =

(xx')(/z) = r(X)[h] for all hE H. The kernel of t is
AG(H) since T(X)[h] = = 1for all lie H implies XE AG(H). Conse-
quently, the Isomorphism Theorem states the existence of an injective
homomorphism z*: G*/A6(H)_.* H*. It satisfies

To establish that G*/A0(!l) and H* are isomorphic, it remains to show that

is surjective. Using the preceding isomorphism o, we have
[A0(H): 1] = [(G/H)*: 1j = [GIll: 1]
= [G:HJ [G: l][H: l]'
and [11* : I] = [H: 1],
where for notational convenience both the identity element and identity
subgroup in each group considered are denoted by I. Consequently,
[G*/AG(H): 1] = {G*: l]{A6(H):
= [G: l][AG(ll): I]' = [H: I] = [H*: I].
REMARK. For a character ç, of H where [G: H] = s, there exist precisely s distinct
characters of G whose restrictions to H equal The
existence of at least one such character follows from the preceding index relations;
the precise number s is attained because A6(H) is the kernel of the map r.

Exercises

1. Suppose the abelian group G is the direct product of the cyclic groups
H1 <s1>, H2 = <s2>, and 113 = Ks3>, whose respective orders are 4, 8,
and 6. Find all characters x E G for which:
a. = x(sl 253) 1

b. x(s12s24s3) = X(512S22) = 1.

2. Let G = II, ® 112 ® H3, where the subgroups H, are cyclic with generators
i = 1,2,3, and y12 = Y22 = = e. Determine the annihilator A1,(H)
and the multiplication table of Gf H for:
a. H = H2 ® H3 b. H )'2> c. H
3. Let G = C3 ® C4, where the subgroups C,,, are cyclic of order m with
generators )',,,, in = 3,4. Determine the annihilator AG(H) and multi-
plication table of G/ll for:
a. H = C4 b. !I = C3.
4. Let G = C6 ® C9, where the subgroups Cm are cyclic of order rn with
generators Ym. ,fl = 6,9. Consider the subgroups H1 =
H2 = and H3 =
a. Determine AG(Hl H2) and verify explicitly that it equals A(,(Hj)
AG(H2).
b. Determine AG(H, 113) and verify explicitly that it equals
AQ(Iij)AG(113).
§7.3 Bijectlons of Sets 235

5. Consider an n-dimensional vector space V over a field F. The set

V" = HomF(V, F) of linear maps x: F is called the dual space of V;
the maps x are called linear functionals on V.
a. Prove that V* is an n-dimensional vector space over F.
b. Prove that there exists a natural (i.e., independent of the choice of
bases for V and V) isomorphism between V and V** = (V*)*.
c. Is a nonzero linear functional x: V—p F necessarily surjective?
d. For given XE V*, prove that kerx is a subspace of V and find its
dimension.
e. If U is an rn-dimensional subspace of V. determine a basis of the
annihilator
f. Prove that = U.
g. Prove that for x, x' e 11*, if kerx c kerx' then there exists an element
a C F such that x' ax.
6. Prove Proposition 5.
7. Consider a group G written as the direct product of cyclic factors
G= •.. ® G,. For each 1, 1 I s, let A1 denote the character
defined in the proof of Proposition 3. Prove that G* = KA1> x ... x <A,>.

Finite abelian groups A, B are said to be dually paired if there exists a single-
valued mapping of their Cartesian product
A T,

such that, for all a,a1,a2 E A and b,b1,b2 e B the following relations hold:
(i) (a1a2,b) = (a1,b)(a2,b),
(a,bib2) (a,b1)(a,b2).

(II) If (a,b) = I for all bE B, then a = 1.

If (a,b) = I for all aeA, then b = I.

8. Prove that G, G are dually paired groups.

9. If A, B are dually paired groups, prove that A B* and A* B.

§7.3 Bijections of Sets

This section is the first of two which are preparatory to the Sylow
theory in §7.5. Bijections (permutations) of groups are used to prove the Sylow
theorems, which are analytical tools for discussing both finite groups and
galois field extensions [see §8.6]. In these sections we use the term bijection
interchangeably with permutation to denote a one-one, surjective mapping
of sets. As such it should not be confused with a bijective homomorphism
of groups [cf. §6.5].
The group E(S) of all permutations (bijections) of a set S onto
itself was introduced in §6.7. In this section the maps E E(S) shall be
written to the left of the elements on which they operate; that is. for a S.
is replaced by and means (Note that some authors
write for cz(a).)
 Selected Topics in Group Theory chapter 7

Now let K be a subgroup of i(S) for a given set S. Define a b

(mod K) for a. b in 5, if there exists an element y K for which b = ya. In
this event we say that b is conjugate to a with respect to the group K. This
concept of being conjugate with respect to K is an equivalence relation;
that is, for a, b.c E S the following relations hold:
(i) a a (modK)
(ii) a b (mod K) implies b a (mod K)
(iii) a b (mod K) and b c (mod K) imply a c (mod K).
The equivalence class of a single element s S with respect to K is
called the orbit of s under action by the elements of K. Denoted
= [ys:yeK},
it consists of all elements in S conjugate to s with respect to K.
To prove that the number of elements in S conjugate to s e S (with
respect to K) divides the order of the group K, we introduce the stabilizer
subgroup
Hj( (s) = {y e K: ys = s} K.

Definition (from General Set Theory). The eardinality of a finite set is the
number of elements in the set.

Note that we consider here only.finiie sets.

Proposition. The cardinality of the orbit Orb,,(s) of an element s E S with

respect to the group K E(S) equals the index [K: Hr(s)J of the subgroup
uK(s) in K.

The proof follows from the observation that for a given s E S there
is a one-one correspondence between elements of OrbK (s) and (left) cosets
of H= HK(s) in K:
y1s = = yi'y1s = s H
=
Since (s) is a subgroup of K, we have immediately by Lagrange's
Theorem, §6.3, an important corollary.

Corollary. The cardinality of OrbK (s) divides the order of K.

Some of the arguments in the Sylow theory utilize a general-

ization of the preceding discussion in which the elements s S are replaced
by subsets M S. Analogous to the definition of the stabilizer subgroup
§7.4 The Class Equation and Normalizers 237

(s) c K of an element s S. the stabilizer subgroup in K of a subset

Mc S is
HK(M) =
The orbit of M with respect to K. denoted OrbK(M). is the set (yM K) y e
of subsets of S conjugate to M with respect to K. As in the proposition, the
number of elements in OrbK(M), for any subset Mc S. equals the index
Hk(M)] of the stabilizer subgroup of it'! in K and so divides the order
of K.

Exercises

1. Prove that the conjugacy relation defined with respect to K c E(S) is

an equivalence relation.
2. Prove that the sets HK(s) and HK(M) are subgroups of K.
3. Verify directly from the definition of that the orbits of elements
s,s' c S are either equal or disjoint subsets of S.
4. A group 6 (with identity e) is said to act (or operate) on a set S if to each
element (g,s) in the Cartesian product 6 x S there is associated a unique
s'=.gs€S, and if es=s and (g1g2)s=g1(g2s) for all seS and
gj,g2 eG.
a. Prove that the stabilizer HG(S) is a subgroup of 6.
b. Prove that the relation defined on S by I s if t e OrbG(s) is an
equivalence relation on S.
c. Suppose that G is a finite p-group and that S is a finite set, whose
cardinaliy Card S is relatively prime to p. Prove that there exists at
least one elements S satisfyinggs = s for all g e 6. (Such an element
is called a fixed point of G on S.)
d. Assume again that 6 and S are finite. Prove that for each s c S.
[6: 1] = [H6(s) :1] .Card OrbG(s).
5. Let 6 be a p group that acts on a finite set S. Let F denote the set of fixed
points of S under the action of 6 [see Exercise 4(c)1. Prove that
Card S Card F (modp).

§7.4 The Class Equation and Normalizers

Continuing the preparation for Sylow theory we now study
normalizers and conjugacy relations in groups. Of interest in their own right,
these concepts, originally introduced in the Exercises of §*6.2 and 6.3,
provide the proof (by means of decomposing a group into conjugacy classes)
that the center of any group of prime power order is nontrivial. The class
equation provides a useful means of enumerating elements in this and other
group-theoretic arguments.
 Selected Topics in Group Theory chapter 7

Whereas §7.3 addressed bijections on an arbitrary (finite) set S. we

now specialize to the case where the set is a finite group G. Further, we pick
the subgroup 1(G) c of inner automorphisms [defined in §6.9] for the
group of actions on G; that is, we replace S and K in §7.3 by G and 1(G),
respectively.
Carrying over the definitions and results we have the following
definitions. Subsets S. S' of G are said to be conjugate with respect to 1(G),
or more simply conjugate, if there exists an element x G such that
S' = as(S) = A subset S is called self-conjugate if = S for
all x G.
In particular, we refer to self-conjugate elements (which are the
elements of the center of G) and self-conjugate (or normal) subgroups [cf.
Exercises 6, §6.2, and 2, §6.3]. The conjugacy class
(g) = {g' G : g' conjugate to g}
= = Orb,(G)(g)
of an element g e G is the set of a/I images of g under the group 1(G) of inner
automorphisms.
Since conjugacy is an equivalence relation, the conjugacy classes
(g) provide a partition of G [see §1.2 and Exercise 3. §7.3]. Thus G is the
union of the distinct conjugacy classes:
G =

From now on we consider only finite groups G.

We utilize the partition of a group G into its conjugacy classes to
state the class equation
(*)

of G. Here C is the center of G and 1,,, is the number of elements in the

conjugacy class the sum is taken over all distinct classes having
more than one element. Again we follow the convention introduced in §7.2
that indicates both the identity element and identity subgroup of the
1

group(s) under discussion.

The class equation (*) follows immediately from the facts that the
conjugacy classes constitute a partition of G and that the center C of G is
the set of self-conjugate elements in G [see Exercises 9 and 10, §6.2 and 2,
§6.3).
Next we develop some propositions which are useful in discussing
structural properties of groups.

Proposition 1. The number h of conjugates of g G divides [G: I].

Proof From the proposition of §7.3, the number of conjugates of a
typical element g E G equals the index h = [1(6): H,(G)(g)] of the stabilizer
§7.4 The Class Equation and Normalizers 239

subgroup of {g}in 1(G). Hence hI [1(G): 1]. By Proposition 2, §6.9,

[G:13 = [1(G):l]{C:lj,
and sohl{G: 1).
Proposition 2. The number of subgroups in G conjugate to a given subgroup
Mc G is [1(G): H,(G)(M)] and divides [G: 1].

In the proof of the Sylow theorems, §7.5, we shall use the following
consequence of Proposition 2 above and the discussion at the end

of 1(G) and a nonempty subset (subgroup)

M of G the number of sets (subgroups) in G conjugate to M with respect to T
is [T: and divides [G: 1].

Proposition 3. The center C of a p-group G consists of more than one

element.
Proof. The class equation for G is
ptm = [G: I] = [C: I]

By Proposition 1, hJ[G: l]—pm. Since by convention liv> I, we conclude

for each v. Hence and so must divide [C:
We conclude this section with some observations on stabilizers
(used here and in §7.3) and normalizers (introduced in Exercises 8, §6.2,
and I and 2, §6.4). For a subset G, the stabilizer H,(G)(S) is a subgroup
of the group 1(G) of inner automorphisms of G; the normalizer
NG(S) = {geG:gSg' =S}
is a subgroup of G.

Proposition 4. For any nonempty subset S

[G: NG(S)j = [1(G): H,(Q)(S)J.
Proof Consider the group homomorphism D: G —+ 1(G) given by D(x) =
as in §6.10. The kernel of D equals the center of G and is a subgroup of
the normalizer N6(S). The inverse image of the stabilizer subgroup
Il,(G)(S) c 1(G) consists of all y G satisfying e H,(G)(S). That is,

'YY'[HI(G)(S)] = {yeG:ySy' =S} = NG(S).

The mapping D: H1(6)(S) is a surjective homomorphism with
kernel C, and so by the Isomorphism Theorem in §6.8, 1-11(Q)(S) NQ(S)/C.
Hence
(**) [P16(S): lJ = [H1(6)(S): l][C: I].
Multiplying
[1(G): 1] = [1(G): H,(G)(S)J[11,(G)(S) : 1]
240 Selected Topics in Group Theory chapter 7

by [C: 1], and combining the result

[G: 1] = [1(G): l][C: I],

[G: 1] = [G: : I],

and the equality (**), we obtain
[G: N.(S)] [NQ(S) : I] = [1(G): H,(6)(S)][NQ(S): 1].
Canceling [NG(S): 1] yields the desired equality.
The class equation (*) can now be written in the common form
[G:l] = {C:1J+E[G:N,.],

where N,. denotes the normalizer of a typical, but not self-conjugate,

element E G. This is an immediate corollary to Proposition 4, if we take for
S the subset {g,,} c G.

Exercises

1. Prove that any group can be written as the union of mutually disjoint
conjugacy classes.
2. Find all classes of conjugate elements for the elements g of the
symmetric group E3. Determine the number of elements in each class.
3. Find all conjugacy classes in each of the following groups, and determine
the number of elements in each class.
a. The nonabelian group of order 27 of Example 6(i), §6.1, with p = 3.
b. The group of the square (dihedral group of order 8).
c. The quaternion group [Example 6(e), §6.11.
4. Prove that a group of order p2, p a prime, is necessarily abelian.
5. Complete the blanks.
a. A group of order 215 can have at most conjugacy classes.
b. A group of order 25 must have conjugacy classes.
6. Prove that a subgroup is normal only if it consists of the union of con-
jugacy classes. Is the union of conjugacy classes necessarily a normal
subgroup?
7. For a subgroup H of a finite group G prove that is the product
of all subgroups IC G in which H is a normal subgroup.
S. Prove that the normalizer NG(H) of a proper subgroup H of the p-group
G properly contains H.
9. a. Suppose that G has p2 elements, p a prime. Prove that all proper
normal subgroups of G lie in the center of G.
b. Generalize part (a) to the following: If H is a proper normal subgroup
of order p in a p-group G, prove that H lies in the center of G.
10. Prove the existence of two nonisomorphic nonabelian groups of order
p3, p a prime.
§7.5 The Elementary Theorems on Sylow Subgroups 241

§7.5 The Elementary Theorems on Sylow Subgroups

Sylow theory (named for Ludwig Sylow, 1832—1918) addresses
subgroups of prime power order and thereby the structure of arbitrary finite
(nonabelian) groups. As we do not assume the Fundamental Theorem of
Finitely Generated Abelian Groups there is some repetition of argu-
ment concerning ahelian groups prior to the Sylow theorems.

Theorem 1 Theorem). If a prime number p divides the order of a

finite abelian group A, then A contains an element of order p.

This is a consequence of either the Fundamental Theorem of

Finitely Generated Abelian Groups or the following lemma.

Lemma. The order of a finite abelian group A divides a power of the minimal
exponent of A.

if A is a cyclic group, the proof consists of noting that the exponent

and order of A are equal.
For noncyclic groups the proof proceeds by induction on the order
of A. To begin the inductive argument, note that the cyclic case establishes
the result for a group of order 2. Suppose now that the result is valid for all
groups of order less than n. Consider a group A of order n, and pick in A
an element b diflerent from the identity. Then B = <b> is a proper subgroup,
and the factor group A/B hich exists since A is abelian) has order less than n.
Let e and e* be the minimal exponents of A and A/B, respectively.
By the induction hypothesis, for some s 1,

[A/B :1] I (e*)s.

Since the eth power of any coset aB in A/B is the identity coset, e*Ie.
Therefore

[A: B] = [A/B: lii?;

since o(b) = [B: 1]j e, we obtain

[A :1] = [A: B]{B: =

Applying the Principle of induction, we conclude that for all finite abelian
groups A, the order divides some power of the exponent.

Proof of Cauchy's Theorem. Since p [A : 1] and [A 1] J?, for some

I
:

a e N, the prime p divides e. The minimal exponent e of A is the least common

multiple of all the orders o(a) of elements a e A. Therefore o(a) for some
element a€ A [cf. Exercise 18, §2.7]. Hence o(a)=ph, and consequently
o(a")
242 Selected Topics in Group Theory chapter 7

A p-Sylow subgroup of a finite group (1 is a p-subgroup whose order

ptm is the highest power of the prime p that divides the order of G. Equivalently,
S,, is a p-Sylow subgroup of G if and only if its order is a power of p and its
index in G is relatively prime to p.

Theorem 2. For a given prime p, any group G for which pI[G: I] has at
least one p-Sylow subgroup S,,.
Proof We use induction on the order of G. If [G: 1] =p, the theorem is
trivial. Now assume that the theorem is proved for all groups whose orders
are less than a. Consider an arbitrary group G of order n, and suppose pm n
but in, m I. Two cases will be distinguished.
Case 1. The group G contains a proper subgroup H whose index is relatively
prime to p. Consequently ptm divides [H: 1]. Since [H: I] <[G: I], the
induction hypothesis implies that H has a Sylow subgroup S, of order ptm.
This subgroup S,, is a p-Sylow subgroup of G.
Case 2. The index of every proper subgroup is divisible by p. If G is non-
abe/ian, the class equation [see §7.4] states that
[G:l] =
where C is the center of G and denotes the typical normalizer of a class of
conjugate elements (ar>>. The indices [G: are greater than I (self-
conjugate elements lie in C). Hence and therefore p divides [G: Np].
Consequently [G : 1] implies that [C: I]. If G is abelian, then G = C,
andpl[C: I].
By Cauchy's Theorem the abelian subgroup C contains an element a
of order p. Moreover, the subgroup <a> is a normal subgroup of G because
a C. Applying the induction hypothesis to the factor group = G/<a>,
which has order 's, where (s, p) = I and n = ptm5, we conclude that G*
has a p-Sylow subgroup S' of order pm_i. Let S,, be the complete inverse
image of where it denotes the canonical homomorphism
mapping 9€ G to its coset g<a> in G. The mapping it: is a surjective
homomorphism with kernel <a>. Hence by the Isomorphism Theorem of
§6.8, S Thus
= <a>][<a> : 1] = : l][<a> : I] = = ptm.

Consequently G contains a subgroup of order ptm, as asserted.

As a corollary note that if p I [G: 1] then G contains an element of

order p. In a p-Sylow subgroup consider element a of order p5, s > 0;
has orderp.

Proposition I. Suppose that P is a p-Sylow subgroup and N is a norma/

§7.5 The Elementary Theorems on Sylow Subgroups 243

subgroup of G. Then N P is a p-Sylow subgroup of N, and NP/N is a

p-Sylow subgroup of GIN.
Proof. Consider the following diagram for the subset relations of selected
subgroups of G.

Relatively prime
NP

Power of p Relatively prime

Relatively prime of p

:Nr\ P
Power of p

{lI
Here, "relatively prime" next to the lines from C to NP and NP to P indicates that
the indices [G: NP] and (NP: PJ are relatively prime top. Also, "power of p" next
to the lines from NP to N, P to N P, and N P to {l} indicates that the cor-
responding indices [NP: N). etc., are powers of the prime p.
We first verify these properties of indices. Since [G : F] and p are
relatively prime, the equation
[G:P] [G:NP][NP:P]
implies that NP] and p1[NP: P]. Since [P: I] =pm, the indices
[P: N F], [N n F: I] are powers of p by Lagrange's Theorem.
Next, using Theorem 3, §6.8, with H replaced by P. we conclude
that P/(N n P) is isomorphic to NP/N; hence [NP: N] = [P: N m P3, a
power of p. This equality together with the equalities

[NP : F] [N: N P].

Consequently [N: N P] and p are relatively prime because F].
Hence N P is a p-Sylow subgroup of N since its order is a power of p.
244 Selected Topics in Group Theory chapter 7

Finally, NP/N is a p-Sylow subgroup of GiN because its order

(equal to [G: NP])
(equal to [NP: N]) is a power of p and its index in GIN
is relatively prime to p.

Proposition 2. If Q is a normal p-Sylow subgroup of G, then it is the only

p-Sylow subgroup of G.
Proof Suppose that P is another p-Sylow subgroup of G. Then PQ/Q is a
p-Sylow subgroup of G/Q according to Proposition I. However, since Q is a
p-Sylow subgroup, [G/Q: I] is relatively prime to p: consequently PQ = Q.
Thus Pc Q and therefore Q = P because [Q: I] = [P:I] =pm.

Since any subgroup is normal in its normalizer [see §7.4 and Exercise
1, §6.4], we have the following corollary.

Corollary. A p-Sylow subgroup P of G is the only p-Sylow subgroup of

Theorem 3. The number r of p-Sylow subgroups of G divides the order of G

and satisfies r I (mod p). All p-Sylow subgroups of G are conjugate.
To prove r I (mod p). If there is only one p-Sylow subgroup, the theorem
is trivially true. Thus, consider the case r> I. (Note that nop-Sylow subgroup
is a normal subgroup by Proposition 2.)
Suppose that P=P1.P2, are the distinct p-Sylow subgroups
of G. Borrowing the notation

of inner automorphisms of G determined by elements x e P

The Set K is also a subgroup of the group of bijections of S. because
P. = x - = { xyx - l : y e P. }
isisomorphic to Thus P1 is a p-Sylow subgroup of G and so is an element
of S. In other words, the mappings in K permute the elements of S.
Furthermore, [K: I] is a power of p. a fact we shall need shortly.
To verify this claim, consider the mapping P—* K which maps x e P to
the inner automorphism of G [cf. §6.9]. Since is surjective (by the
definition of K), P order of
a of p.
As in §7.3, we consider the orbits in S with respect to K. Since P is a
group, for all hence {P) cS is an orbit consisting of one
element. We now show that the cardinality of each of the other orbits
(i.e., the number of conjugates of P., I, with respect to K) is a positive
power of p. First, /z1 [K: (P1)] by the corollary of §7.4. where HK (Ps)
 The Elementary Theorems on Sylow Subgroups 245

is the stabilizer subgroup

1IK(Pa) =
Second. since [K: I] is a power of p, we must have = p' for some n1,
0 P71.
Suppose that = 0. Then IIK(Pj) K; that is,

vP, = P, for all X E P.

and consequently. P the normalizer of 1', in G. The corollary to

Proposition 2 states that is the only p-Sylow subgroup of Thus
P1 = P. a contradiction. Therefore n must be greater than zero for each i,

Hence r = I + p'', n > 0, where the summation is taken over the

distinct orbits. Thus r I (mod p), as asserted.
To prove all p-Sylow subgroups are conjugate. Once this fact is established,
it will follow immediately that r divides the order of G. The number of
conjugates in G of P is
[1(G): = [G:
and divides the order of G by Propositions 2 and 4, §7.4. Thus it must be
shown that, for a given (fixed) p-Sylow subgroup P and any other p-Sylow
subgroup Q, an element g E G exists for which a9 P = Q.
Let T1 denote the class of all p-Sylow groups of G conjugate to
P1 = P with respect to 1(G). If there is a p-Sylow group Q of G which does not
lie in T1, then the class T2 of p-Sylow groups of G conjugate to Q with respect
to 1(G) is distinct from T1 since conjugacy is an equivalence relation. Using
the action of K' = 1(G) : y E Q}, where Q is a typical group in T2, on
the p-Sylow subgroups in the class T2, we show, as in the previous part of
the proof, that the number s2 of distinct groups in the conjugacy class T2
satisfies s2 I (modp).
Next let the mappings in K permute the objects of the class T2.
A typical orbit within T2 with respect to K contains = [K:
p-Sylow groups, where I!K(Q) is the set e K: a,, Q Q}. Since HK(Q)
is a subgroup of K, its index is a power of p. Further, k, is a nonzero power
because Q is the only p-Sylow subgroup of NG (Q). Thus s2, the sum of the
over distinct orbits within is divisible byp. In other words, 0 (modp).
This conclusion contradicts the previously found congruence
(modp). Therefore, T2 = 0; there can be no p-Sylow subgroup Q that is
not conjugate to P.

Theorem 4. If G has a subgroup U whose order is a power of a prime p,

then there exists a p-Sylow subgroup P such that U P.
Proof Consider the set S= {P1, ...,Pr) of the r p-Sylow subgroups of 6,
and let U). As in the proof of Theorem 3, Kis a sub-
 ________

246 Selected Topics in Group Theory chapter 7

group of Z(S) and [K: 1] is a power of p. Also, the orbit of P1 in S contains

= [K: HK (P1)] elements. The indices h• are powers of p, but not all of them
can be greater than 1. lest pir. Since I (modp) by Theorem 3,
h. = [K: HK(PI)] = I for some 1 i 1,that is, K= HK(P1).
For this i, =P1 for all ue U. This fact implies that
P1 is a normal subgroup of Now by Theorem 3, §6.8,
U/(U P1).

Since the order of U is a power of p. so are the orders of the factor groups
U/( U P.) and U• P/P1.
Finally, P. cannot be a proper subgroup of U.P1, since
[G : P1] = [G: U.PJ[U.P1: P1]
and is relatively prime to p. Thus, U.P1 = P1, which implies that U is con-
tained in the p-Sylow subgroup P..

Exercises

1. Prove that the exponent of a cyclic group of order n is n.

2. Prove that the exponent of a finite abelian group A is the least common
multiple of the orders of elements of A.
3. if B is a subgroup of the abelian group A, prove that the exponent of
A/B divides the exponent of A.
4. Let G be a finite group and let H be a Sylow subgroup of G, [H: I] = ptm.
Prove that H is the only subgroup of order ptm in NG(H).
5. With H and NG(H) as in Exercise 4, prove that = NG(H).
6. Prove that any group G has a normal subgroup of order 7 if:
a. o(G) = 28 b. o(G) = 42 C. o(G) = 707.
7. Let G be any group of order 105.
a. Show that G has at least one nontrivial normal subgroup.
b. If G has exactly one nontrivial normal subgroup N, what is o(N)?
8. Complete the blanks.
a. A group of order 48 must have either or 2-Sylow
subgroups.
b. A group of order 595 5.7. 17 must have eithei or
7.Sylow subgroups and 5-Sylow subgroups.
c. A group of order 104 must have either or 2-Sylow
subgroups.
d. A group of order 122 must have either or 2-Sylow
subgroups.
e. An abelian group of order 255 = 3.5.17 must have
3-Sylow subgroups.
9. Show that every group of orderp5m, where p > rn> I, p a prime number,
has a nontrivial normal subgroup.
§7.5 The Elementary Theorems on Sylow Subgroups 247

10. Show that any group of order 595 = 5.7.17 must have at least two
nontrivial normal subgroups.
11. How many elements of order 5 are there in a group G of order 80 if G
has no normal subgroup of order 5?
12. Suppose that the order of a group G is the product of two distinct primes
p and q.
a. If q = 2, prove that G has a normal subgroup oforderp [cf. Exercise
4, §6.4].
b. If p> q, prove that every p-Sylow subgroup of G must be normal
in G.
c. If G has normal subgroups with respective orders p and q, prove
that G must be abelian and (with reference to §7.1) cyclic.
d. If p I (modq) and q $ 1 (modp), prove that G must be cyclic.
13. Let H be a normal p-Sylow subgroup of G. Prove that a(II) = H for all
automorphisms of G.
14. Show that any group of order (35)2 must be abelian. Hint: Show that
such a group is the direct product of two normal abelian subgroups.
15. a. Prove that any group of order 245 must be abelian.
b. How many nonisomorphic groups of order 245 are there?
c. Repeat parts (a) and (b) for a group of order 85.
16. Consider primes p and q, where p> q and p 1 (mod q). Prove that, to
within isomorphisms, there exists precisely one nonabelian group of
order pq.
17. Consider a finitc group G in which every Sylow subgroup is normal.
Prove that G is the direct product of its Sylow subgroups.
18. If G is a finite group in which each element different from the identity
has order p, for a fixed prime p, prove that o(G) = pm, or some m e N.
19. Show by an example that if an integer ii divides the order of an abelian
group A it is not necessarily true that A contains an element of order ii
[cf. Cauchy's Theorem (Theorem I )].
20. Prove that every subgroup H of G which contains the normalizer NG(P)
of a p-Sylow group P is equal to its own normalizer, i.e., NG(H) H
[cf. Exercise 7, §7.4].
21. Prove that if H c G has order p' but is not a p-Sylow subgroup, then
H NG(H).
22. A subgroup H G is called maximal if there is no subgroup H' of G
such that H H' G. Prove that every maximal subgroup of a p-group
is normal and has index p.
23. The following steps constitute an alternate proof of Theorem 3. The
method of double coset decomposition was originally used by Cauchy
and later by Frobenius.
a. Given two subgroups H, K of a group G, define the double coset by
Hand KofgcG to be
HgK= {/igk:/zeH,keK}.
Prove that the double cosets by H and K constitute a partition of G
[see §1.2].
 Selected Topics in Group Theory chapter 7

b. Prove that the number of elements in any double coset divides o(G),
when G is a finite group. Hint. Show that the number of right cosets
Hg' in HgK equals [K: (g 'Hg) K) and use the analogous
statement for left coscis.
c. In fact, prove that the number of elements in HgK is hk/d9, where
o(H) = I,, o(K) k, and d9 = o(gHg' K).
d. Decomposing G by two Sylow p-groups P and Q, show that
o(gPg' Q) = o(Q) for some g, and hence, that P and Q are
conjugate subgroups of G.
24. Prove the following propositions concerning groups of prime power
order by arguments similar to those used in proving the Sylow theorems.
Let C have order p", p a prime.
pS,
a. C has at least one normal subgroup of order for each s, 0 <s <rn.
b. Every subgroup of order pmt is normal in G.
c. If p 2, and if G has only one proper subgroup of order for each
s, 0 < s < ni, then G is cyclic.
d. The number of normal subgroups of order p' is congruent to 1

modulo p.
e. The number of subgroups of order p' is congruent to I modulo p.
25. Suppose that S, and S2 are two distinct p-Sylow subgroups of a sub-
group H of the group C. Prove that they are not contained in the same
p-Sylow subgroup of G.
26. Suppose that G is a p-group which has only one normal subgroup of
index p. Prove that G is cyclic. Hint: Use induction on the order of 6.
27. Let G be a nonabelian group of order which contains a cyclic normal
subgroup of index p. Prove that C contains an element a of order
and an element b of order p. which is not a power of a. Furthermore,
prove that a and b generate G, that <a> is a normal subgroup, and that
bab1 = ar, where r 1 and r" I

§7.6 Composition Series and the Jordan-Holder Theorem

More detailed description of the structure of a nonabelian group is
given by the study of composition series, which are sequences of selected
subgroups. The fundamental result in the study of composition series is the
theorem, due to Camille Jordan (1838—1922) and Otto Holder, stating that
any two composition series of a given finite group are "isomorphic." Com-
position series are used in defining the "solvability" of groups, and as we
shall find in and 9.2, this bears on the related question of solvability
of polynomial equations by radicals. A principal application of composition
series is in the concept of the length of such a series, which plays the role
in group theory and algebraic geometry that dimension does in the study
of vector spaces or field extensions [see §8.1].

For a group
Proposition I (The "Butterfly Lemma" due to Hans Zassenhaus).
(i and four subgroups M, M0, N, N0, where M0 is normal in M and N0 is
§7.6 CompositiOn Series and the Jordan-Holder Theorem 249

normal in N,
M0•(M n N)1M0(M n N0) N0.(N n M)/N0.(Nn M0).

M0•(Mr\N)

N0

The proof has several parts. First we should check that the two factor
groups are indeed well-defined. That is, verify that M0•(M n N0) and
N0.(N n M0) are normal subgroups of n N) and N0.(N n M),
respectively [cf. Exercise 10, §6.4].
Next we adopt the following notation:
K=MnN,
H = M0•(M n N0),
D=HrK.
We leave to the reader the proofs of two technical lemmas.

Lemmal. HK= M0.(MnN0).(MnN)z M0.(MnN).

Lemma2. D=.HnK=(M0nN)•(MnN0).
Since H is a normal subgroup of HK= M0.(M n N), we obtain
from Theorem 3, §6.8, an isomorphism K/(H n K), or
N)/M0.(M n N0) (Mn N)/(M0 n N).(M n N0).
The terms in the factor group on the right-hand side remain unaltered if
M and N, M0 and N0 are interchanged; thus also
n M /N0.(Nn M0) (Mn N)/(M0 n N).(M n N0).
Consequently, combining the isomorphisms, we complete the proof of the
Zassenhaus Lemma (Proposition I).
250 Selected Topics in Group Theory chapter 7

We now consider chains or sequences of subgroups of a given group

G and utilize the Zassenhaus Lemma to prove the Jordan-HOlder Theorem.
A chain of subgroups of a group G

is called a normal chain if is normal in G1 ,. I i r. A normal chain

G= G0 = K,0 K,, =

is called a refinement of the normal chain

is normal in for I I

A composition series of a group G is a normal chain

(*)
which admits no proper refinements and has no coincidences = G1.
The number r in the composition series (*) is called the length of the
composition series.
Equivalently, a composition series is a normal chain such that the
factor groups G,_ 1/G1 are simple groups (that is, have no nontrivial normal
subgroups) or such that the normal subgroups of -' are ,naxi,nal normal
proper subgroups of respectively.
Two normal chains

and
are called isomorphic. if r = s and if there is a permutation r of (I, ..., r}
such that G1_ ,/G1 - I/"t(i) for each i. I I r.

Proposition 2. Any two normal chains of a group G have isomorphic

refinements.
ProoJ Consider two normal chains of subgroups of G:

and
Now define two sets of subgroups for I I r, I s:
= Hi). 11,, n G1).
Also set G10 = , and = Observe that
G1. Hi,. = Hi'
§7.6 Composition Series and the Jordan-Holder Theorem 251

Next apply the Zassenhaus Lemma, setting M = G...1, M0 = G., and

N = H1 N0 = Then

Finally, using these isomorphisms. we conclude that

and

are isomorphic refinements of the given normal chains.

Theorem 1 (Jordan-Holder Theorem). If the group G has a composition

series, any two of its composition series are isomorphic.
Proof Any two composition series are normal chains, and hence by
Proposition 2 have isomorphic refinements. As compostion series they have
no proper refinements however, and so must themselves be isomorphic.
Thus for a group having a composition series, the length r is a property
of the group, independent of the choice of composition series. Note that the
theorem requires the existence of at least one composition series. This
condition is satisfied whenever the group G is finite.

Proposition 3. If G has a composition series and N is a normal subgroup

of G, then there exists a composition series of G one of whose terms equals N.

For the proof, apply Proposition 2 on the existence of isomorphic

refinements to the normal chain G = G0 = G2 = {e} and to a
composition series G = = {e}.
An important subject in group theory is that of solvability, a concept
which is defined in terms of normal chains. The term "solvable" derives from
the corresponding concept of solvability of polynomial equations by radicals
for which the reader should refer to the Galois theory and 9.2].
A group G is called solvable if there is a normal chain

such that each factor group 1/G1 is abelian, I I <S. Finite solvable
groups can be defined equivalently as groups G having a normal chain

such that G1_ 1/Gd is cyclic of prime order, I j r. The equivalence of the
definitions follows from the structure of finite abelian groups I].

Example 1. All p-groups G are solvable.

These groups contain, by Proposition 3 in §7.4, a cyclic group L of order
252 Selected Topics in Group Theory chapter 7

p which lies in the center. Then GIL has order Hence we may assume, as an
induction hypothesis, that G* = GIL = ... ... G7,,

where G?_ is cyclic of order p. The inverse images of the groups G? [see
§6.8] form a chain

of subgroups in G such that G,_ 1/G1, Gm_ 11L, and are cyclic of order p. Hence
G is solvable.

Example 2. The dihedral groups f are solvable.

If F is given [as in Examples 6(f) and 7, §6.1] by
= {a'bt : s,t c Z, & = b2 e, bat, = I)
-
then r. <a> {e} is a normal chain with abelian factors, since [f;: 2 and
<a> is cyclic.

Thus, in particular, the group of permutations of 3 elements is solvable.

Example 3. The symmetric group is solvable.

The group of permutations of 4 elements has order 24; the alternating
subgroup 44 of even permutations has order 12 and index 2 [see Exercise 6, §6.7].
Thus A4 is normal in Furthermore A4 has a normal subgroup V, generated by the
even permutations a = [12] [34] and fi = [13] [24]. Since a2 = = = [1] [2] [3] [4],
V is isomorphic to Klein's Four-group.
Before proving that V is normal 44, we observe that ,44 is generated by
a, /1, and y [123] = (l3][l2]. Since a, Ii, ; are all even permutations, 44 contains
the group B generated by them. Moreover, o(B) must be either 6 or 12 since it is
greater than 4 and divides 12 = 0(44). The proof that B = 44 can be accomplished
by exhibiting seven distinct elements in B such as:
= l
= [213], = [423],
= [413], = [412],
3/3 = (432], 3'2afl = [314].
;'afl = (214],
Hence
3.2,
44 {e, a, ;'a, y2a /3, yfi, ..2fl afl, ;afl, y2ajJ L

Since V and 44 are generated by {a, 131 and {a,fl, respectively, the proof that V is
normal in .44 is reduced to verifying that V; = ;'V. Check that
= [12][34][l3][12] = [243] = yfl,
rn = [13][24][13][12] [214] =
afi;' = = ;'/Ja/3 =
We then have the composition series for
V D {i.a} fr}
whose factors are cyclic groups of respective orders 2, 3, 2, 2.
§7.6 CompositIon Series and the Jordan-Holder Theorem 253

To prove that the symmetric group n 5, is not solvable we

need the following lemma.

Lemma. If N is a normal subgroup of a subgroup ftc Z,,, n 5, where

H contains all 3-cycles, and if H/N is abelian, then N itself contains all
3-cycles.
Proo/: Proposition 2 in §6.7 stated that, for n 5. every 3-cycle is the
commutator of two 3-cycles. Since H/N is abelian, N contains the commutator
subgroup of H, by Proposition 3, §6.9. In particular, N contains all 3-cycles,
as asserted.

Theorem 2. The symmetric group s,,, n 5, is not solvable.

Suppose to the contrary that E,, were solvable, i.e., there exists a
composition series
= H1 H_1 H. H5 = {e},
where is normal in Hi.. and H1_1
is abelian. Then using the lemma
for !.,, = H and = N, we conclude that H1 must contain all 3-cycles.
Repeating this argument s times, we conclude that 115 = {e} must contain
all 3-cycles, an absurdity. Thus no such composition series can exist.

Corollary. The alternating group n 5, is not solvable.

If A,, were solvable, then would be too since is a normal sub-

group olE,, and since the quotient En/A,, is an abelian group.
The structure of n 5, is made more explicit by the following
proposition, the proof of which is left as an exercise.

Proposition 4. The alternating group n 5, is simple.

Exercises

1. Find a composition series of the quaternion group G of order 8 [see

Exercise 6, §6.5].
2. Let G be a solvable group. Prove that
Every subgroup of G is solvable.
a.
b. Every homomorphic image of G is solvable.
3. Let H be a subgroup and N a normal subgroup of the group G. Show
that HN is solvable if H and N are solvable.
4. Let N be a normal subgroup of the group G. Assume that both N and
its factor group GIN are solvable. Prove that G is solvable.
254 Selected Topics in Group Theory chapter 7

5. a. With reference to Exercise JO, §6.1, prove that the set H of(affine)
maps r on Z, given by r(u) = au+fl, where a = 1, is a normal sub-
group of the affine group (We call H the subgroup of
translations.)
b. Prove that Af(Z,)JH Z!, the multiplicative group of Z7.
c. Find a composition series of Af(Z,).
d. Is Af(Z7) a solvable group?
e. Find the commutator subgroup of Af(Z7).
f. Find the center C of Af(Z,).
6. Prove that the affine group Af(Zm) is solvable.
7. Prove that any abelian group is solvable.
8. Find all composition series of a cyclic group G of order 28.
9. Consider the following chain of subgroups G where =
G', = [Gu)]i, ..., = [G" fl]I,.•. where G' is the com-
mutator subgroup of G. Prove that
a. is not only a normal subgroup of I, but is also a
normal subgroup of G.
b. G is solvable if and only if = (e) for some positive integer r.
10. Using the higher commutator subgroups (as in Exercise 9) of a
solvable group G, prove the following statements [cf. Exercise 4].
a. if N isa normal subgroup of G, then the factor group GIN is solvable.
b. If H is a subgroup of U, then H is solvable.
11. Prove that every group of prime power order is solvable.
12. Prove Lemmas I and 2.
13. The following steps outline a proof of the simplicity of A,,, ii S. Com-
plete the proof of each step.
a. The symmetric group X,, is generated by the transpositions [12],
[l3],...,[ln].
b. The alternating group A,, is generated by the 3-cycles [123], [124],
[12,,].

In the remaining steps assume H is a normal subgroup of A,, which

contains a permutation a = r1 r2 tj, where the r, are disjoint cycles.
In each case prove that H = A,,.

c. Suppose a = = [132] (i.e., I = 1), and observe that [12nz} =

,n ii. (This generalizes
[m3 I] [132] [I 3m] belongs to H for all in, 3 <
to the statement that if H contains one 3-cycle, then it contains all
3-cycles, or H = A,,.)
d. Suppose t1 = [aj,a2 am] has length in> 3, and observe that for
p= [a1,a2,a3]e = =
H since disjoint cycles commute.
e. Suppose Ti = [a1,a2,a3] and 12 = [a4,a5,a6} and show that H
Contains a permutation a' involving a cycle of length in> 3 by
considering p = [a2,a3,a4J and pc1p'a.
f. Suppose is a 3-cycle and 12 are transpositions. Observe that
c H.
 Groups with 255

g. Suppose the I I are all transpositions with = [a1,a2],

[a3,a4). Let p = [a2,a3,a4] and = (a1,a4,a5]. (Why must
involve an a5?) Show that (pa 'c),i(a '),j 'is a 3-cycle
in H.
h. Conclude that A, is simple.
14. Prove that is the only normal subgroup of ii 5.
15. Exhibit two nonisomorphic groups which have isomorphic composition
series.
16. Consider a group G of order P1 P2 P3, where the p1 are distinct primes.
Prove that if G contains a nontrivial normal subgroup then it is solvable.
17. Suppose that H isa proper subgroup of a p-group G, so that H
H occurs as a subgroup in some com-
position series of G.
18. Prove that a p-group G has a composition series G = G0 ...
G1 ... such that the subgroups
(e}
G a group with a composition series and a normal subgroup N.
Prove that the length of G is equal to the sum of the lengths of N, GIN
[cf. Exercise 17(c), §6.4].

§7.7 Groups with Operators

In this section and the next we generalize two topics previously
considered: vector spaces from §4.1 and bijections of sets from §7.3.
A vector space V/F [cf. §4.1] is an additive abelian group together
with a field F such that each r E F determines a linear transformation
V —' V given by Wr(V) = rr
for all v e V. Thus in the sense of our study of bijections of sets the
nonzero elements of F operate on or permute the elements of V.
In §7.3 we discussed bijections of sets in which a subgroup K of s(S)
acted on elements of a set S. Then in §7.4 we considered the group 1(G) of
inner automorphisms acting on a group G. We now consider the more general
situation of a set acting on a group. Hence we have the term with
operators." As suggested above, vector spaces are special cases of groups
with operators.
Much of the group-theoretic discussion of subgroups. homomor-
phisms, factor groups, composition series, and the like carries over with only
minor modification to the so-called groups with operators and thus to modules
and vector spaces. In this section we address the more general case of groups
with operators, where there are no structure requirements on the set of
operators. Then in §7.8 we shall specialize our discussion to consideration of
modules.
A group G is said to admit (or to have) a set Q as a domain of
operators if each w Q determines a homomorphism of G into itself. We
then refer to G as a group with operators It is convenient to write iog in
place
256 Selected Topics in Group Theory chapter 7

We begin our discussion of groups with operators with three examples.

Example 1. An abelian group A always admits Z as a domain of operators because

each ii e Z determines a homomorphism A —b A defined as follows [cf. Exercise
II, §6.51:
for n 0, = a";
forn<O,
In each case is a homomorphism:
= (ab)" = a"b" =
or ((ab) = (a 'b1)" = (a)"(b 1)_n
=
If the law of composition for A is expressed additively, then for n 0, we
have a = n a, the n-fold sum of a. For n < 0, o,, a = ml (— a). Thus in particular,
Zn, is a group with the operators Z.

Example 2. An arbitrary (not necessarily commutative) ring R, considered as an

abelian group, admits itself as a domain of operators. For each a e R define the
operator R — R by w0r = ar, an additive group homomorphism.
These mappings and the mappings w;: R — R given by r ra, which
make R into a right operator group, play an important role in the study of non-
commutative rings. Of course if R is commutative, then = (Ofl.

Example 3. An arbitrary group G admits the inner automorphisms 1(G) as a

domain of operators. For each g G, the inner automorphism a9, given by
a9x = gxg1 for xe G, is an operator on G [cf. §7.4].

We turn now to adaptation of the concepts of subgroup, homo-

morphism, factor group, etc. to groups with operators. The reader is invited
in §7.8 to note that these are generalizations of corresponding statements
for modules.
A subgroup H of a group G with operator domain 0 is called
e 0.
fl-admissible (or fl-stable) if wh E H for all Ii E H and all

Considering the preceding examples, check that all subgroups of abelian

groups are Z-admissible. The normal subgroups of a group G are precisely the
admissible subgroups with respect to the operator domain = 1(G) [cf. the note
in §6.9]. The subspaces of a vector space V over a field Fare precisely the admissible
subgroups of V (considered as an abelian group) with operator domain 0 =
r e F), where v = rv, for v e V.
As a note of caution, when considering Q and Z as additive (abelian)
groups, observe that Z is a Z-admissible subgroup of Q, but is not a Q-admissible
subgroup of Q.

Suppose that H is a normal fl-admissible subgroup of G. The

factor group Gift admits 0 as a domain of operators, if we define
w(gH) (wg)H
§7.7 Groups with Operators 257

for all w and g G. This defines a homomorphism on G/H because

= w(gg'H) = (w(gg'))H
= (wg)(wg')H = [(wg)H].[(wg')H]
= o.(gH)w(g'H).
If G and F are two groups with respective operator domains
and C(f), a homomorphism q: G F is called an admissible homomorphism
if there exists a single-valued mapping/i of onto such that
p[w(g)] =
where g,g' e G, w and h(w) e
In many applications of admissible homomorphisms Q(G) =
= and It is the identity map h(w) = w; in this case is called an
a-admissible homomorphism.
We now carry over Proposition 2, §6.5, and the Isomorphism
Theorem, §6.8, to groups with operators. The proofs follow those given
earlier with allowance made now for the operators.

Lemma. The kernel K of an admissible homomorphism q: G F is an

admissible normal subgroup of G.
Proof Since K ker is known to be a normal subgroup of G, it remains
only to verify its admissibility with respect to the domain of operators. For
any wEfl(G) and
= h(w)q(a) = h(w)e e

where e is the identity of F. (Since h(w) belongs to Hom(F, F) it maps e

to e.) Thus wa ker q, and K = ker is an admissible subgroup.

Theorem. If q: G F is an admissible homomorphism with kernel K,

there exists an admissible isomorphism p:

G
\ /
/
,,,,//u

G/K

commutes (that is, q = o it), where it is the canonical projection onto the
quotient.
Proof. The Isomorphism Theorem, §6.8, asserts the existence of a group
isomorphism p: Thus it remains only to verify the admissibility
 Selected Topics in Group Theory chapter 7

of it and p. The canonical projection it: G —' G/K is Q-admissible since

it(wg) = (wg)K = w(9K) = wit(g).
The isomorphism ;n G/K—'p(G) is admissible since
p(w(gK)) = p((wg)K) = (p(LtJg)

= h(w)p(g) = /z(w)p(gK).

Corollary 1. The propositions in §7.6 generalize to groups with operators

and operator homomorphisms and isomorphisms.

Corollary 2. The Jordan-Holder Theorem remains valid for groups with

operators if normal subgroups are replaced by admissible normal subgroups.

These corollaries are discussed in more detail in the special case of

modules in §7.8.

§7.8 Modules
We now specialize the discussion of groups with operators in §7.7 to
modules. Modules may also be viewed as generalizations of the concept of
vector spaces V/F, because a module M over a ring R is an (additive) abelian
group M and an operation (or action) of the elements in R on M. Thus this
section also represents a continuation of linear algebra from Chapter 4.
Modules pervade the modern study of commutative algebra, algebraic geom-
etry, and algebraic topology.
Let R be a (commutative) ring with multiplicative identity I. An
(additive) abelian group Mis called an R-left module if there is given an external
law of composition (or mapping)
(r,m) —+ rin e Ill
for (r. in) in the cartesian product R x M such that
r(,n1 = ru:1 + rn:2,
(r1 r2)nz = r1 (r2un),
(r1 +r2),n = r1 in + r2nz,
In = in
with nz,n:1,m2 eM and l,r,r1,r2 eR.
In module-theoretic discussion we refer to the R-left module M
defined above as a unitary module since un in for all in e M. Vector
spaces V/F are examples of such modules, as are the groups in Examples I
and 2 in §7.7.
In an analogous fashion we can define R-right modules M (R acts to
the right on M), and two-sided modules. As a matter of convenience we
§7.8 Modules 259

shall refer to R-modules in what follows rather than to R-left modules. The
corresponding statements for R-right modules are exactly equivalent. When
R is commutative, left and right R-modules can be identified, but in a strict
sense they are not equal. We maintain the distinction because of the carry-
over of modules to the important cases of noncommutative and nonassoci-
ative rings. (If the ring R is noncommutative. the distinction between R-left
and R-right modules is important.)
An R-module homomorphism q: Al Al' of R-modules Al, Al' is a
group homomorphism which preserves (respects) the module structure; that
is. for a, b E M and e R
p(a+b) = + q(b),
(p(ra) = r(p(a).
The following statements from §7.7 need no further proof.

Lemma. The kernel K of an R-module homomorphism Al —' M' is an

R-submodule of M.

Proposition I. The (additively-written) quotient group MIt! is an R-module

for any submodule H c M.

Theorem 1. If q: M—+ M' is an R-module homomorphism with kernel K,

there exists an R-module isomorphism ;n

the canonical projection onto the quotient module M/K.

As asserted in Corollaries 1 and 2, §7.7, the concept of compos-

ition series and the Jordan-Holder Theorem of §7.6 carry over to groups
with operators and hence to modules. In particular, a composition series of
a module M is a chain of submodules which admits no proper refinements.
Therefore we have the following theorem.

Theorem 2 (Jordan-Holder Theore,n). If the module 41 has a composition

series, any two composition series are isomorphic.

An n-dimensional (n 1) vector space V/F, considered as an

F-module, is simple if and only if dime V = I. (Simplicity means in this
 Selected Topics in Group Theory chapter 7

context that V has no nontrivial subspaces.) Thus a composition series of V

in the sense of F-modules is a chain of subspaces

such that the factor spaces V1_1/V, (over F) are simple F-modules, i.e.,
dimF(VI_l/Vj= 1, 1Consequently dimFV is the length r of a com-
position series of V, because the r vectors ;, 0 1< r. form a basis of V/F
where rj1 is an arbitrary nonzero vector in V, 1\V1. Thus we have proved the
following proposition.

Proposition 2. The length of a composition series of a finite dimensional

vector space is equal to its dimension.

As a corollary, for subspaces S. T of a finite dimensional vector

space V/F considered as F-modules we have
dimF(S+T)—dimFS= T),
as in Exercise 11, §4.2. This relation is sometimes referred to as Grassmann's
relation, after Hermann Grassmann (1809—1877). It follows from the isomor-
phism of Theorem 3, §6.8,

and the consequence of the Jordan-Holder Theorem [cf. Exercise 19, §7.6]
that for any subspace U V
= dimFV— dimFU.
Analogous to the discussion of generating sets for vector spaces V/F
and of finitely generated abelian groups, an R-left module M is said to be
finitely generated if there exists a finite set of elements rn1, in M such
that each element ni in M can be written as a sum
in = r1 m1 + +
with elements r1, I I s. in R. It is important to remember that the elements

rj are not required to be uniquely determined relative to the set of generators

in1, 1

The remainder of this section is devoted to the study of R-module

homomorphisms. Given R-modules M, M', and M" and R-module homo-
morphisms and fi: M'—* M", we call the sequence
p
A'! Al' M"
exact if Im — ker fi. For example, each R-submodule N M determines
an exact sequence
0 'M
where is the inclusion map (z(n) = n, considered as an element of Al, for
z

all n e N) and it is the canonical projection onto the quotient.

§7.8 Modules 261

Denote by M) the set of R-module homomorphisms

a: M'—*M. As in §6.11. HomR(M',M) is an additive abelian group. More-
over it becomes an R-module if we define r e R. a 1-IomR(M', M)
to be the mapping ia: M given by ra(m') = r[a(in')]. (Check that ra
is an R-module homomorphism.)

Proposition 3. For R-modules M, M', Al". each a HomR(M', Al) induces

an R-module homomorphism
a*: HomR(M.M")
and an R-module homomorphism
a,1,: HomR(M",M')—+ l-IomR(M,M).

For (3€ HomR(M, M") define a*(j3)e HomR(M', Al") by

fl(a(rn')) e Al".
Since a and /3 are R-module homomorphisms, so is their composition
a*(fl) = /io a. Similarly, define = a(y(rn")) for y€ HomR(M", M'),
and of course a o y is an R-module homomorphism.
As a simple consequence we note the following more general
corollary.

Corollary. Corresponding to any sequence

p
(*) M' M
of R-modules we have the "induced" sequences
p.
HomR(M',N) < HomR(M,N) < HomR(M",N)
(**)
HomR(N,M') >

for any R-module N.Also, tx" /3* = ((3 o 2)*, and = (a o

Leaving this verification and the proof of the fact that if the sequence
(*) is exact then so are the sequences (**) as exercises, we conclude this
discussion with the so-called Short Five Lemma. Its proof is commonly
referred to as an exercise in diagram chasing.

Short Five Lemma. Consider exact sequences of R-modules with R-module

homomorphisms as indicated below

o it!' '>M >0

'1,

o N' N N" 0
such that the diagram commutes, If a and are isomorphisms, then so is /3.
262 Selected Topics in Group Theory chapter 7

Exercises

I. Let a: M' M be an R-module homomorphism of R-modules M', M.

For re R define a mapping ra: M'—' M by (rcx)(m') = r(n(rn')). Prove
that ra is an R-module homomorphism.
2. a. Prove that ,(fl) and a(y), as given in Proposition 3, are R-module
homomorphisms.
b. Verify that the mappings a* and defined in Proposition 3 are
R-module homomorphisms.
3. Consider R-modules M M', N N'. and an R-module homomor-
phism M N such that N'. Construct an R-module homo-
morphism M/M' NIN' [cf. §6.8).
4. a. Verify the statements in the corollary.
b. Prove that if the sequence (*) is exact, then so are the sequences
5. a. Given R-modulcs N', N", define (as in §3.3) the external direct sum
N'+ N" and verify that it is an R-module.
b. Define maps a, fi so that the sequence
p
0 N'+N" —÷N" -O
is exact.
c. Conclude that the sequence

0 HomR(M,N')
6.
Hom,dM,N") 0
is exact for any R-module M.
6. Consider R-modules M, M', N, and N' and R-module homomorphisms
a: M—' M', N—b N', and N.
a. Show that flq,ae HomR(M,N').
b. Thus, show that the mapping f given = flq,a is an R-module
homomorphism from HomR(M ', N) to HomR(M, N'). Schematically,

M 'M'

.+' p4N.
7. In the Short Five Lemma, prove that if a and y are injective, so is j3.
a.
Prove also that if a and y are surjective, so is fi.
b.
Hence prove the Short Five Lemma.
c.
S. Considering the exact sequences
0 >Z2 Z4 #Z2
0 - Z2 Z2 + Z2 Z2 0,

show that in the Short Five Lemma the mapping fi must be given. That
is, there can exist isomorphisms a and y, hut no homomorphism fi such
that the diagram commutes.
§7.8 Modules 263

9. An R-module Al is called irreducible if it has no nontrivial submodules.

Prove the lemma of lssai Schur (1875—1941) that every R-module homo-
morphism M is either an isomorphism or the zero map
(ço(nz) = 0, for all ni e M).
10. The group ring GQ of a multiplicative finite group G = {g1, . over
the field of rational numbers Q is defined to be

GQ=

addition is componentwise and multiplication is distributive over

addition. Prove that

q =-
in
ii i= I
is an idempotent element (that is, q2 = q).
II. Let V be the Klein Four-group and suppose that F = Z2. Consider,
analogous to Exercise 10, the group ring
a. Prove that y = is nilpotent.
b. Find all ideals A for which A2 = {O}.
c. Prove that there is an ideal B V, such that A c B for all ideals
A for which 42 =
12. Suppose that a cyclic group F = y\ generated by y of order n, operates
on the additively-written abelian group A. Let 0 and N be the mappings
of A into A given for acA by
D(a) = — a = (y— ila
/n—I \
and N(a) a+ya+ +y"1a = (
\i=o /
a. Prove that 0 and N belong to l-lom(A,A).
b. The quotient q(A) [kerD: N(A)][kerN: D(A)]1 is called the
Herbrand Quotient (after Jacques Herbrand, 1908—193 1) if both
indices are finite. Suppose that B is a F-stable subgroup of A (i.e.,
yb E B for all b e B), and define an action of F on the factor group
A/B.
c. If any two of the quotients are finite, show that
q(A) = q(B)q(A/B).
d. Prove that q(A) = I if A is finite.
13. Let R be a subring of a (commutative) ring S, and consider subsets M
and N of S which are R-modules.
a. Prove that the set MN of finite sums of elements me,, ni e M, n E N,
is an R-submodule of S.
b. Prove that if M and N are finitely generated R-modules, then so is
MN.
14. State and prove the analogue, for groups with operators, of:
a. Theorem 2, §6.8.
b. Theorem 3, §6.8.
c. Corollary 2 to Theorem 2, §6.8.
264 Selected Topics in Group Theory chapter 7

15. a. let 0 M —* N P be an exact sequence of R-module

homomorphisms. Prove for every R-module A the exactness of the
sequence 0 —* HomR(A, Mi —÷ i-lomR(A, HomR(A, 1').
b. If M —+ N P o is exact, prove that for every R-module
A the sequence
p.
0 > HomR(P,A) > >
is exact.
16. Given an exact sequence of vector spaces over a field F,
'p
0 >W >V >U >0,
prove the exactness of the sequence
0 >U* >V* >W* >0.
17. Given an n-dimensional vector space V/F and a linear transformation
T: V, define the action of the polynomial f(x)=
a0EF[x)on Vby
f(x)v (ahTh+...+aIT+aoI)v
ahP(v)+ + a1 T(v)+ a0v, for v E
a. Prove that V/F thus can be considered as a finitely generated
F[xJ-modu)e.
b. Prove that the set W(v) = {a(x)u : a(x) e F[x]} is a subspace of V.
the so-called cyclic subspace generated by v with respect to T.
c. Verify that {v, T(v), ..., is a generating set of the subspace
W(v).
d. Find a necessary and sufficient condition on b(x) e F[x] such that
W(b(x)v) = W(v).
18. a. Continuing Exercise 17, for v 0, prove that A(v) = {g(x) e F[x]:
= 0} is a nontrivial ideal in F[x]. The monic generator
of A (v) is called the order of v (relative to T). Thus this exercise states
that each v 0 in V is a torsion element.
b. Prove that divides the characteristic polynomial XT(X) of T.

A subspace W is said to be p(x)-primary for an irreducible polynomial

p(x) if p(x)sw = 0 for some s> 0 and all we W.

c. Prove that V can be uniquely expressed (except for arrangement of

summands) as the internal direct sum V= W1 of p1(x)-
primary subspaces W,, I < I h, where the polynomials p,(x) are
the irreducible factors of the characteristic polynomial Xi (x) of T.
d. Finally, prove that each of the p1(x)-primary submodules W1 is a
direct sum I / k(i).
Prove that
h k(i)
1] fl
1=1 i=I
= Xr(X).
[Cf. the Fundamental Theorem, §7.1.]
§7.8 Modules

NOTE. The discussion in these last two problems is a carry-over of the

direct sum decomposition of finitely generated abelian groups. Here the
degree of a polynomial replaces the absolute value of integers. Instead of
positive numbers we use monic polynomials. The dimension of a vector space
replaces the order of an abelian group. In general though the F[xJ-module
structure of the vector space V/F depends upon the choice of the linear
transformation T.
 8

Field Theory

While undergraduate courses commonly do not discuss fields beyond

what we have already covered in Chapter 5, we provide, for honors courses
or a second semester, the basic theory of algebraic field extensions, both
separable and inseparable (although the latter ease can be stripped from the
arguments) so that the interplay of groups and fields might be displayed in
the Fundamental Theorem of the Galois Theory and in subsequent
explicit examples of galois groups. The existence of algebraically closed
fields requires the Kronecker construction of roots of irreducible
polynomials (from Chapter 5. §5.4) and Lemma in the ideal-theoretic
argument given. The proof of the Fundamental Theorem of Algebra (namely,
that the complex numbers constitute an algebraically closed field) requires
elementary real analysis and further algebraic argument; it is given in
Chapter 9, §9.9. The Galois theory, especially field extensions and galois
groups, might be followed by discussion of cyclotomic extensions, solution
of equations by radicals, and ruler and compass constructions, which are all
topics in Chapter 9.

§81 Algebraic Elements

This section begins our study of field theory with consideration of
extensions of a given field F. A field K containing the field F is called an
extension field (or overfield) of F, and F is called a ground field (base field)
of K. This relationship will be denoted K/F.

266
 Algebraic Elements 267

An element a E K is called algebraic over F if there exists a finite

dimensional vector space V(a)/F in K (considered as a vector space over F)
such that all powers d. I 0, of a lie in V(a). An extension K/F is called
algebraic over F if each a e K/F is algebraic over F.

Proposition 1. If a e K is algebraic over F. then a is a zero of a (unique)

monic irreducible polynomial ?fla(X) F[x]. Further, ma(x) divides any
polynomial g(x) of which a is a root.

We call Ifla(X) the minimal polynomial of a over F, and its degree

the degree of a over F.
Proof. If a = 0, then ?na(X) .v. Now consider a # 0 and the ring homo-
morphism 4a: F[x] —, K given by p0(f(x)) =f(a). Set M(a) = kerqa. By
Proposition 2. §5.2, M(a) is a principal ideal, and so, if it is nontrivial, it has
a (unique) monic generator in0(x). Since it generates M(a), rna(v) divides all
polynomials g(x) e M(a).
Let r=dimV(a). Since then the elements (l,a,a2,...,cf} constitute
a linearly dependent subset of V(a), there exist elements b. e 0 i r, not
all zero, for which = 0. Thus a is a zero of/i(x)= E F[x],
and M(a) is a nontrivial ideal.
To prove that the monic generator of M(a) is irreducible,
suppose to the contrary that ,n0(x) has the proper factorization #Oa(X)
g(x)/z(x). Then 0 = nz0(a)=g(a)h(a), a contradiction since g(a),h(a) are
nonzero elements of the field K.
Proposition 2 is the converse to Proposition I.

Proposition 2. If a E K is a root of a polynomial 1(x) E F[x], then a is

algebraic over F.
Proof Consider a polynomial 1(x) = #0. of which a is a root.
Then

an — (b0
-
Multiplying by a, we note that

a" = a" — at? — ... a

— —

fh0

fb0\

is an element of the vector space V(a) = (l,a,.. generated over F

 Field Theory chapter 8

by {l,a,...,a"'}. An inductive argument shows that kO, lies in

V(a). Hence a is algebraic over F.
We use F[a] to denote the image ofthe homomorphism F[xJ —* K
given by q0(f(x)) =f(a). for a e K. The subring F[aJ of K is isomorphic to
the residue class ring F{X]/(flla(X)). (Recall the Isomorphism Theorem for
rings, Exercise 1, §6.8.) Explicitly, an isomorphism q: F[x]/(rna(x))
is described by
= = 1(a) e F{a].

Proposition 3. The polynomial ring F[a] equals its field of quotients F(a)
when a is algebraic over F.

Consider a nonzero element A =f(a)/g(a) in F(a). Because ma(x) jS

irreducible, (g(x), = I; hence there are polynomials k(x) and h(x) in
the ring F[xJ satisfying y(x)k(x)+ma(x)h(x) = I. Therefore in F[a]
= g(a)k(a) + ma(a)h(a) = g(a)k(a).
Consequently A =f(a)k(a) e F[a].
Proposition 3 has the obvious restatement that every rational function
of an algebraic element over a field F is a polynomial function over F. The
dimension of the field F(a), considered as a vector space over F, equals the
degree of the minimal polynomial ma(x). It is denoted dimF F(a) or [a: F].

Theorem 1. If K F, then all elements of K which are algebraic over F

form a subfield L F, the so-called algebraic closure of Fin K.
Proof Consider a, b in the set L of elements of K algebraic over F. Then
a' E V(a) = ..., c K and b' e V(b) = ..., K for all I, j 0.
This means a' = and b-' = with in F.
Hence

= =

with coefficients thus V for all

i, j 0. Since dimF V is finite (at most SI), ab belongs to L. Next

lies in V for all rn e N. Consequently a + b e L. A nonzero element a e L

satisfies a polynomial

with coefficients e, e F, where 0. Hence ')+ I = 0,

which implies that a e L. Thus L is a field. Moreover, every rational
§8.1 Algebraic Elements 269

expression p(a,b)/q(a,b) with polynomials p(a,b) and 0. whose

coefficients lie in F, is an element of L.
An extension K/F is called a finite algebraic extension of the field F
if K, considered as a vector space over F, has finite dimension n = [K: F].
Tkis dimension is also called the degree of K over F.

REMARK. The elements of a finite algebraic extension K/F are algebraic over F.
We can take V(a) = (k,, for all a E K where {k1 is a basis of K over
F. The proof of Theorem I indicates the utility of defining the concept of algebraic
element in terms of a vector space and not just as the root of a polynomial over the
base field.

Theorem2.
Proof. Letki,...,km bea basisofKover Fand basisofEoverK.
Then every element a€ E is a linear combination of the inn products
with coefficients in F. Specifically, a = e, with b1 = E7'..1 where
E F; hence
a=
i,j
Thus dime E = [E: F] is at most inn. To prove that the elements ej are
linearly independent over F consider = 0 for some coefficients
E F. Then

= 0:

by the linear independence of the elements e, over K, = 0. Moreover

= 0 since the elements k. are linearly independent over F. Thus
[E: F]=fE: K][K: F] =nzn.
Corollary. Let K/F be an algebraic, not necessarily finite, extension. If an
element a in an extension field M/K is algebraic over K, then it is necessarily
algebraic over F.

For the proof letflx) = f+b1 x"'+ ... e K[xJ be the minimal
polynomial of a. The field L = F(b1 is a finite algebraic extension of F
since each field F(b1, ...,b1_1)(b1) is a finite algebraic extension of
F(b1, ..., b._ Consequently, Theorem 2 implies that L(a) is a finite extension
of F. Thus a is algebraic over F.

Let Fbe a field contained in a field fl. is a set of elements in fL

the intersection fl of all subfields L,/Fin that contain the set E is denoted
by F(s), and F(I) is said to have been obtained by adjoining the eLements
of I to the field F. Hereafter we shall be concerned mainly with adjunctions
F(a1, ...,am) where the elements a1, ...,a,, are algebraic over F.
 Field Theory chapter 8

in this case the elements of F(E) = F(a, ,..., am) are polynomials
b,1 ,,,,
amim, I .1 rn,
0 11<

with coefficients b51 in F. To verify this fact, note that F(a,) = F[a,]
by Proposition 3 and by induction that
F(a, = F(a,
= F[a,
= F{a, a1].

Now consider f(x) e F[xJ. According to the basic construction of

Kronecker [see §5.4], there exist n = degf(x) algebraic elements a,, ...,a1
in some extension field cl/F such that

•f(x) = fl a.)

in cl/F. A field K/F, obtained by adjoining to F the zeros a,,...,a1 of a

polynomialf(x) of degree n F[x], which lie in some field cl/F, is called a
in

splitting (or decomposition) field of 1(x);

K=F(a1 a1).
The existence of a splitting field for each E F[x] is the subject of the
next proposition. Essentially it restates the theorem of §5.4. The field
K= F(a,, ...,a1). obtained by adjoining to Fthe zeros a, a1, is a splitting
field off(x).
Proposition 4. For each polynomialf(x) e F[x] there exists a field L/Fsuch
thatf(x) is a product of linear factors in L[x].
lff(x) is irreducible in F[x] the zeros of f(x) in K are called conjugates
of each other. In case the polynomial f(x) is separable—that is, /(x)
with g(x) e F{x]—the results of §5.5 imply that 1(x) has degf(x) distinct
roots in any field containing a splitting field of f(x). A zero of such a poly-
nomial is called a separable element over F. An algebraic extension K/F is
called a separable extension over F if each of its elements is separable over F.

Exercises

1. Suppose that a c KJF, a finite algebraic extension of degree n. Prove that

the degree of the irreducible monic polynomial e satisfied
by a divides n.
2. Consider polynomials f(x)= x2+x+l and g(x) = x3—2 in Q[x].
a. Prove that f(x),g(x) are irreducible in Q[x).
Ii. Prove that g(x) is reducible in F[x], where F is the splitting field
of 1(x).
 ________

Finite Fields 271

3. a. Prove that [Q( ': =

b. Exhibit a polynomial f(x) e Z [x] of degree 4 such that
=
4. Prove that =
5. Let K1 and K2 be extensions of the field F in some field L/F. Assume
that [K1 : F] = p1. [K2 : F] Pi for prime numbers P1 and P2 (which
need not be distinct). Prove that either K1 = K2 or K1 K2 = F.
6. Prove that a subring R of the algebraic extension K/F, where R F, is
a field. Is this statement true if K contains nonalgebraic (called
transcendental) elements over F? Give an example to substantiate your
answer.
7. a. Find the degree of over Q.
b. Find the degree of over Q; prove that the field
contains = 1.
c. Prove that the degree of over Q is 6, where rü is a com-
plex number satisfying w2 + + 1 = 0.
8. a. Prove that the polynomial f(x) x3—x+ I is irreducible in Q[x].
b. Now let K = Q(a) where a is a zero of f(x). Find the monic poly-
nomial with coefficients in Q which has b I for a root. Fur-
thermore express 1/(1 —a2) as a polynomial in l,a,a2 with rational
coefficients.
9. Find the minimal polynomial f(x) E Q [xl of 2 ÷ 3
a.
Find the other zero of f(x).
b.
c. Express (2+3 in the form at-b ./1 with a,beQ.
10. Find : Q].

11. Let be an irreducible polynomial over a field F of characteristic

p> 0. Prove that the polynomial xe—a is also irreducible over F.
12. If a1,a2,a3 are the zeros of the polynomial x3—x2+ I e Q[x], find
cubic polynomials 1(x), g(x) c Q [x] such that:
a. f(x) has roots a12, a22, and a32.
b. g(x) has roots I — I/a1, I — I/a2, and I — I/a3.
13. a. Find the minimal_polynomial in Q[x] of + + Vi —
b. Determine

§8.2 Finite Fields

Whereas the previous section involved algebraic extensions of
arbitrary fields, we now limit our consideration to finite extensions of finite
fields. The discussion culminates in Proposition 6. the finite field analogue of
the Fundamental Theorem of Galois Theory in §8.7. Throughout this section
we consider a field F with q elements and prime field P,, Z,,. As a vector
space over P,,, Fhas dimension nz: hence it has pm = q elements.

Proposition 1. The multiplicative group F* of the nonzero elements of F

is a cyclic group of q— I =pm_ I elements.
272 Field Theory chapter 8

Proof By Lagrange's Theorem [*6.3], a F*, and by

Theorem 2, §5.2, the polynomial x4 I has at most d roots. Now apply
Proposition 5, §6.6, to conclude that F* is cyclic. Since of course = 0, all
the elements a in F are zeros of the polynomial P,[x]. This poly-
nomial is separable because its derivative is .— I. Consequently, the elements
of F are precisely the q distinct zeros of — x.

Proposition 2. If w generates the group F*, then ai is a zero of an irreducible

polynomial of degree in with coefficients in P,,.
Proof Since [F: P,,] = in, the m + I elements { I, w, . .., Om} are linearly
dependent over P,,. Thus the minimal polynomial .1(x) P,[x] of co has
degree h m. Any nonzero a F is a power of w (since w generates F*);
furthermore a = 9(w) for some polynomial 9(x) [x] of degree less than h,
since F= P,[x]/f(x) [cf. the discussion preceding Proposition 3,
§8.1]. There are p" such polynomials (including the zero polynomial); hence
F must have elements. But F has ptm elements; therefore m = h = degf(x),
as asserted.
In the note at the end of §3.6 we introduced the Frobenius
autoinorphisin
a a' =
definedon a finite field F of characteristic p. The Frobenius map p is not
simply a group automorphism of F* (that is, (p(ab) = but is an
automorphism of the additive group as well: (a + 6) = (a) + (6). The
Frobenius automorphism has the following three significant properties.

Property 1. 4)(a)=aifandonlyifaeP,cF.
Property 2. The order of is in. That is, (ptm is the identity map on F while
(p5 is different from the identity map, 0 <s < in.

Property 3. The automorphism generates the group G of all automorphisms

of F that leave fixed the elements of P,,. Consequently G is a cyclic group
of order m.

The verification of Property I is left as an exercise.

Ver f/i cation of Property 2. The order of a generator w of F* is ptm — I. Thus
the elements 4)5(w) = are different from w for 0 < s <in and equal to w
for s = in. Consequently q has order in.
Verfficag ion of Property 3. The generator w of F* satisfies the polynomial
f(x) = xm+am_ lXm_I + of Proposition 2. Since has order in,
0 s < ni} is a set of in distinct roots of f(x). But as degf(x) = in,
this set contains all the roots off(x).
§8.2 Finite Fields 273

For any automorphism a e G, the image a(w) must again be a root

of f(x), since
O=c(O)=a(wm+am_iwm_I+...+ao)
= (a(w))m+am_j(a(w))m_I + +a0.
Consequently a(o) = for some s, 0 s <rn. Because the two auto-
morphisms a and agree on the generator w of the cyclic group F*, they
are equal. Thus, every a G is a power of order m.

The third property is stated more generally for extensions of the

field F as follows.

Proposition 3. Let F be a field with q = ptm elements and K/F an extension

of degree n. The group G(K/F) of all automorphisms of K that leave fixed
the elements of F is generated by i,li = q,", where q, is the Frobenius
automorphism.
Proof The multiplicative group K* of K is cyclic of order qfl_ I =ptm"— I
and has a generator w. As in the proof of Proposition 2 we show that w is a
zero of an irreducible polynomial

in F[x] by noting that {l,w af} are linearly independent over F. If

degf(x) = h were to be less than n, then K would have q" elements,
a contradiction.
For elements b E F, pm(b) = 1,. and thus e G(K/F). Setting
= consider the n automorphisms 0 v <n, of K. where denotes
the identity map. The mappings i,tiv belong to G(K/F), i.e., *"(b) = b for all
b e F. Therefore, (w) is also a zero of/tv). since 0 = = =
As the zeros = 0 v <ii, are distinct so are the n
automorphisms in G(K/F).
Now to prove that G(K/F) is a cyclic group, consider any element a.
Then, 0 = a(0) = a(f(w)) since i leaves fixed all he F. Conse-
quently, a(w) is a root of 1(x) and so must equal one of the n roots
But if a(w) = 9Y(w), then a = because w is a generator of K*, Thus,
every a G(K/F) is a power of = as asserted.

Proposition 4. For every prime integer p and every positive integer in there
exists a field F containing q = ptm elements.

The proof follows immediately from the observation that the q zeros
of the polynomial f(x) = x are distinct and form a subfield of
the splitting field L of F(x). The details are left to the reader.

Proposition 5. All fields of ptm elements are isomorphic.

274 Field Theory chapter 8

Proof: Consider two fields F, F, each having q = ptm elements, with respective
prime fields P,,, The prime fields are isomorphic since each is isomorphic
to Z,,. Extend the isomorphism P,, —i p,,, in the usual way, to an isomorphism
A:
As in Proposition 2 let ai be a generator of F*. The minimal poly-
nomialf(x) e P[x] of w has degree m; it divides /z(x) = since h(w) = 0.
In the polynomial )(f(x)) divides A(h(x)) = In fact
since contains all of the roots of Pick
a root a of The rn elements I, w,..., are a basis for F/Pa, and
I, ..., atm - are a basis for F/Fr. The mapping A: F—. F described by
A(w') = &, 0 < i < m, and A(a) = A(a) for a e is the desired isomorphism.

Proposition 6. Consider a field F with q = ptm elements. Let K/F be an

extension of degree n and p be the Frobenius automorphism of K. There is a
one-one correspondence between the subfields L/F of K/F and the subgroups
where

The correspondence is described by associating to a subgroup H

of G the intermediate field L = {a e K: a(a) = a for all a e H } of elements
in K invariant under H; it is called the fixed field of H. Conversely,
to the extension L/F of degree I: associate the subgroup (of order n/h)
H= e G: b, for alt be L} = G(K/L); it is called the galois group
of K/L.
Proof Consider a given subgroup H = (i/i" : 0 v <n/h} of index h in G.
The set {a e K: = a subfield L of K because, for a, b e L,
= = a ± b,
±
= I1/h(a),I,h(b) = ab,

Since i/i"(a) = the elements of L are precisely the qh zeros of the polynomial
and hence L has qh elements, thus [L : F] = Ii. Consequently
associated to the subgroup 11 of order n/h is an intermediate field L/F for
which
[K: L] = [K: F]/[L : F] = = [H:1].
n/h
Finally H is exactly the subgroup, leaving fixed the elements of L. Suppose
to the contrary that a(b) = b for all b e L for some a G\H. Let H be the
product of the subgroups <a> and H. and L be the fixed field associated to H.
By the preceding argument, [K: L] = [II: 1] > [H: I] = [K: LI, and
therefore L c L. But if a(h) = b for all b e L, then every element in L is fixed
under the automorphisms in H and hence L L, a contradiction.
Conversely consider a given intermediate field L/F of degree h, and
let H be the subset e G: = b for all b e L} of the cyclic group
G = G(K/F) of order n. By definition H is the group G(KIL) of Proposition 3
and is generated by (phm = Hence H is a subgroup of G of order = n//i
and of index/i in G.
 The Theorem of the PrImitive Element 275

Exercises

1. Prove that the polynomial x3+x2+(%]2 is irreducible in Z2[x). Let a be

a zero of this polynomial in a splitting field K of x3 + x2 + [112.
a. Express the other zeros /3 and A as polynomials in I, a, a2 with
coefficients in Z2.
b. Prove that [Z2(a): Z2] 3.
c. Show that a is a primitive 7th root of unity.
2. Prove that the polynomial x3+tx2—2 is irreducible in the field of
rational functions Z3(i).
3. Prove that the equation x2 = —[I] has a solution in Z,, if and only if
p = 4m+3, pa prime.
4. Define (u, (x, y) = (iix + [7] vy, uy + rx), where ii, v, x, y, [0], [I], [7] e
Prove that there exists a pair (x. y) such that (ii, v).(x, y) = ([1], [0])
for given (u, v) ([0], [01).
5. a. Let p be an odd prime. Prove that every element of Z, is a sum of two
squares.
b. If F is a finite field, prove that every element in F is a sum of two
squares.
6. Suppose that F= = {O,T, ...,ã,E, ...} is the field of q =pm elements
where a is a primitive (q— l)st root of unity. Let SL(2,F) = U be the
group of 2 x 2 matrices
ãE
with determinant 1. Prove that
a. [U:1] =q(q2—l).
b. The matrices

c
1011 and Th
1101 h< in,
1 =[
generate the group U. Note that

[1 OffT I
Ia ij[s Tj a-i-El
7. Let F be a field of q = ptm elements, p a prime. Suppose that f(x) e F[x],
and define the polynomial function!: F— F by 1(a) = 1(a) F.
a. Prove that the polynomial functions form an integral domain D.
b. Show that f—f is a homomorphism from F[xJ to D, whose kernel
is the principal ideal generated by — x.
8. Let F be the quadratic extension of Prove that all ekments of are
squares of elements of F.
9. Verify Property l. Hint. Consider the polynomial — x F[x].

§8.3 The Theorem of the Primitive Element

An algebraic extension K/F is said to be simple if K = F(A) for some
A e IC. The element A is called a primitive element of K over F.
 Field Theory chapter 8

Theorem of the Primitive Element. If F is an infinite field and L/F an exten-

sion containing elements e1, ..., algebraic over F such that c2, ..., are
separable over F, then there exists a primitive element we K = F(c1, ...,

The proof is by induction on h. First consider elements a = c1 and

b= where b is separable over F, and monic irreducible polynomials f(x)
c2,
and g(x) in FExj such thatf(a) = g(b) = 0. By the theorem there exists
a field M K which contains all the zeros a = a1, ...,a, and b = b1,
(these are distinct!) off(x) and g(x), respectively. We may assume that b F,
for otherwise K = F(a) and the assertion is proved. Hence s 2. Next each
of the equations aI+ubk = a1 has at most one zero in M, I i r,
I <k s. Since this is a finite set of equations and F is an infinite field, there
exists an element cc F, distinct from the solutions = —bk),
such that aL+cbk a1 +cb1 for all i and all k I.
Setting d=a1+cb1 =a+cbeF(a,b) yields f(a)=f(d—cb)=0,
wheref(d—cb) is a polynomial in b with coefficients in F(d)[x]. Let h(x)
be the GCD (g(x),f(d—.cx)) in F(d)[x]. We seek to prove that h(x) = x—b,
from which we may conclude that b c F(d).
Since f(x),g(x) split in M, so dof(d—cx) and h(x). The element b
is a common zero of g(x),f(d—cx) and so (x—b)Ih(x). No higher power
of x — b can divide h (x) however, because b is a simple zero of the (separable)
polynomial g(x). Any other factor of h(x) must be a product of powers of
the polynomials x—bk, k I, as they are divisors of g(x). This would imply
that f(d— = 0 and that d— cbk = a1 + cb, — is one of the zeros
a1, ..., a, of f(x). By the choice of c, this is impossible; that is, f(d— cbk) 0
for k 1. Consequently h(x) x — b and b e F(d). Also, a = d— cb e F(d)
and therefore F(a, b) = F(d).
Finally, using induction, we have
F(cl,c2,...,ch_l) = F(v);
the preceding argument yields

F(c1, c2, ..., Ch_ i) (c,,) = F(l', ch) =

F the validity of the theorem follows from the fact that

K = F(w) where w is a primitive I)st root of unity [see §8.2). A finite field
has no inseparable extensions because every element a algebraic over F such (hat
[F(a): = is a zero of the separable polynomial — x.

Exercises

1. Let F = y) be the field of rational functions of the indeterminates

x, y. Suppose that and are zeros (not belonging to F) of the poly-
nomials t"—x and in F[:).
 Equivalence of Fields 277

a. Prove that [K: F] = p2 where K = ,i).

Show that there exists no element y E K for which K = (In
other words, the Theorem of the Primitive Element does not hold in
this case.)
c. Prove that there exist infinitely many distinct fields M for which
K M F.
2. Prove that a finite separable algebraic extension K/F always can be written
K = F(A) for some A e K.
3. Let K be a finite algebraic extension of the infinite field F. Prove that
K = F(A) if and only if there exist only finitely many fields F L K.
(Hint: Take B e K such that [F(B) : F] is maximal. Make an indirect
proof by examining the fields F(B+dC) for Ce K\F(B), de F.)
4. Let A and B be nonzero algebraic elements of the field K/F such that A
is separable over the field F and e F where charF= p, e ? I. Prove
that:
a. F(A,B)=F(A+B)
b. F(A,B)=F(AB).
5. Find a primitive element for each of the following extensions of Q.
a.
b.
c. where w2+w.i-l = 0.

§8.4 Equivalence of Fields

If A is an isomorphism of fields F and F and if K. K are extensions of
F, F, respectively, we call an isomorphism A: K —' K a prolongation (also an
extension or continuation) of). if the restriction A F of A to F is equal to the
original isomorphism A; that is, if A(a) = ).(a) for all a E F.
Two field extensions K/F, K/Fare said to be equivalent (or isomorphic)
over F if there exists an isomorphism K—' K prolonging the identity map
on F; in other words, an isomorphism q: K—3.K such that ip(a)=a for all
a E F. In particular, an automorphism of K over F is a (field) automorphism
for which q(a) = a for all ac F.
Equivalent fields K/F and K/F contained in a common field fl/F are
called conjugate fields. In particular, elements A e K and B K are said to be
conjugate if there is an isomorphism q of the extensions F(A) and F(B) for
which q,(A) = B and qi(a) = a for all a E F. Note that these definitions extend
the concept of conjugate roots of a polynomial {cf. the end of §8.11.
An isomorphism A of fields F and F extends to a ring isomorphism
of F[x] and F[y], where y is an indeterminate over F. as follows. Define
A(x) = y, and forf(x) = +a0 in F[x], define ).(f(x)) by
1(f(x)) = + + + ).(a,)y + A(a0).
That A is an isomorphism of the polynomial rings is a consequence of its being
an isomorphism of the ground fields. Further, A(f(x)) =fiy) is irreducible
in F[y] if and only iff(x) is irreducible in F[x].
278 Field Theory chapter 8

The next two lemmas on prolongations are significant in the sub-

sequent development of normal and separable extensions and 8.6].

Lemma 1. If A is a root of an irreducible polynomialf(x) in some extension

L/F, and if B is a root of A(f(x)) =J(y) in some extension ElF, then A has a
prolongation A to F(A) such that A(F(A))= F(B) and A(A)= B.
Proof Consider the following isomorphisms:
(1) F[A] —+ F[xJ/(f(x)), where p(g(A)) = {g(x)], the residue
class of g(x) modf(x).
(ii) i/i: F[BJ —* P{y]/(f(y)), where = the residue
class modf(y).
(iii) A': F[y]/(f(y)), where A' is the map on the residue
class ring F[x]/(f(x)) induced by the original isomorphism
A: F[x] F[yJ. Specifically, =
Schematically,
A — —
F(A) = F[AJ—---'-F[B] F(B)

F[x]/(f(x)) F [y]/(f( y)).

Recall that F[A] = F(A) [Proposition 3. §8.1]. Composing these isomorphisms
we obtain the desired prolongation A = 'A'q: F(B) of A to F(A).

Corollary 1. If the isomorphism A is an automorphism of F, and if A and B

are zeros of the irreducible polynomials f(v) and 7(x) (here x and A (x) = y
are identified for notational simplicity) in an extension L/F, then A can be
extended to an isomorphism A: F(A) F(B), such that A(A) = B.

In particular, we have another corollary, as follows.

Corollary 2. II A, B are zeros of an irreducible polynomial f(x) e F[x],

then F(A) and F(B) are equivalent extensions of F.

REMARK. The irreducibility of f(x) is essential in the hypotheses of Lemma I and

its corollaries. For example, [Q( : Q] = 2 and : Q] = 3. I-fence Q( -.12)
and are not isomorphic (i.e., not equivalent) extensions of Q. Yet, both
and are roots ofh(x) =(x2—2)(x3—3)EQ(x].

Lemma 2. Let A be an isomorphism from F to F extended as in Lemma 1

to the polynomial rings F[xJ and F[y]. Consider a (not necessarily
irreducible) polynomial 1(x) E F{x] of degree n with a splitting field
L = F(A1, correspondingly, let L= F(B be a splitting field
of f(y). There exists a prolongation A of A to F(A1, such that
§8.4 Equivalence of Fields 279

A(F(A1, F(B1, ..., where the zeros B, are arranged so that

A(A1)=B1,
Proof. We proceed by induction on the degree n of f(x), starting with A1.
This element is a zero of an irreducible factor g(x) of 1(x) in F[x]. The
corresponding polynomial has a zero in Now
label the Ba's so that B1 is a zero of and apply Lemma 1 to obtain a
prolongation A1 of). to F(A1) such that A1(A = B1 and A1 (F(A1)) = F(B1).

F(41,...,A1)

1has been extended to an isomorphism

L. = F(A1,...,A1) -, =
such that A1(A,) = I j 1, as in Figure 8.1. Next, factorf(x) in L.[x],
obtaining
f(x) = (x — Ai)j h1 + (x) (x),

where 1(x), are irreducible (but not necessarily distinct) poly-

nomials in L.[x]. There is a corresponding factorization in

.1(y) = [

In the splitting fields L and L the polynomials and h,,(y),

1< v s, split into linear factors (x— Ak) and (y— Bk), i < k n. Now let
be a zero of the polynomial 1(x) and relabel the zeros Bk so that
satisfies = 1(x)).
Either which implies that h1÷1(x) is linear and
correspondingly or 1(x) has degree greater than one. In
the first case, set 1
= A.; and in the second apply Lemma 1 to obtain a
prolongation of A. (and hence of).) to such that
= =
Thus, by induction, the lemma is true for all n.
 held Theory chapter 8

Application of Lemma 2 to splitting fields and

F(B1, ..., of a polynomial f(x) e F[v] of degree n proves the following
theorem.

Theorem. Any two splitting fields of a polynomial f(x) E F[x] are iso-
morphic over F.

Exercises

1. Let n an odd positive integer, p a prime number. Prove that the field
be
admits only the identity automorphism over Q.
2. Prove that the only automorphism of Q is the identity map.
3. Prove that the subfield of C is equivalent to two distinct subfields
in C/Q, other than itself.
4. a. Prove that Q(i, .[3)fQ has degree 4 over Q.
b. What are the conjugate fields a(K) (over Q) of X = Q(i, in C?
c. Prove that the mappings a form a group isomorphic to the direct
product of two cyclic groups of order 2. (Observe that oc(K) = K.)
5. Consider A c C. Find
a. IQ(A) QI and [Q(A, Q(
b. The monic irreducible polynomials in Q[x] and which
have A for a zero.
c. The conjugates of A over Q and Q(
6. If &+co+ I = O, prove that the fields and p a prime
number, are isomorphic over Q.

§8.5 Counting of Isomorphisuns and Separability

We examine now the number of fields equivalent to a given field K/F.

Lemma 1. Let K/F and K/F be given extensions, and the isomorphism
A: K—p K, a prolongation of ).: F. If a e K is a root of the (not neces-
sarily irreducible) polynomial f(x) e F[x], then A(a) e K is a root of
2(f(x)) =J(x) E F[x].
Writingf(x) = e F. we have

0
=
and 0 = A(0) = = j'(A(a)).

In particular, if a is an aulomorpizism of K orer F and if a K is a root of

.1(x) e F[x], then a(a) is also a root of From this observation and
Corollary 2, §8.4, we have the following proposition.
 Counting of Isomorphlsms and SeparabilIty 281

Proposition. Suppose K/F is the splitting field of an irreducible polynomial

f(x) E F[x]. For any two roots fi of f(x), there exists an isomorphism of
K/F mapping to fi. Furthermore any isomorphism of K/F permutes the
roots off(x).
Lemma 2. Let a be an algebraic element of K/F and n'1' the reduced degree
of m0(x), the minimal polynomial of a over F. Assume that K/F contains a
splitting field of ma(x). Then F(a) has precisely n" conjugate fields in K/F.
In other words, precisely distinct isomorphisms defined on F(a)
extend the identity map of F. If a is a separable element, the reduced degree
coincides with the degree n of ma(x); otherwise, n = n*jf, where
p = charF. Since Kcontains all of the (distinct) roots a•, 1 i of m0(x),
K/F contains precisely n" conjugate fields F(a1)/F. These n" conjugate fields
may very well coincide; what counts is that the isomorphisms over F are
distinct.
For example, Q( and Q( — are coincident conjugate fields over Q
in C, but f2 and are distinct roots of x2—2eQ[xJ. The automorphisms
given by
=
a,beQ,
are distinct.
Proof of Lemma 2. Let {a a1, a2, .. ., K be the distinct roots of
Pfla(X). By Corollary I, §8.4, there exist isomorphisms
F(a) F(a1), I< I
such that 1,(a) = and 2,1 F is the identity map. Thus at least n" distinct
isomorphisms are defined on F(a). By Lemma 1 above, if). is an isomorphism
(extending the identity map on F) from F(a) to some other extension L/F,
L K, ).(a) must be a root of ma(x): that is, 2(a) = a for some i. Consequently
L 2 F(a1) 2 F. Since both L and F(a1) are isomorphic to F(a) and L 2 F(a1),
they are equal and A = Therefore there are precisely distinct isomor-
phisms defined on F(a) as asserted.
Lemma 2 gives the key to the proof of the theorem below extending
the enumeration of distinct isomorphisms to arbitrary finite algebraic exten-
sions K/F, which in the case of characteristic p > 0 need not be simple. The
precise statement needed in the theorem is the following generalization of
Lemma 2.

Corollary. With notation as in Lemma 2, there are precisely n" distinct

prolongations of an isomorphism A defined on F to isomorphisms defined
on F(a).

Let K/F be a finite extension of degree n with the vector space basis
(If n= 1, then K=F.) There exists a subset {AI,...,Ak) of
 Field Theory chapter 8

such that and

K. = F(A1,...,A1) = F(A1,...,A1_1)(A1)
= Kg_1(Aj), 1 f k,
where the fields have degrees n1> I over K4_ [see §8.1]. Note for
characteristic p> 0 that = n' where is the reduced degree of over
K1_1 with the corresponding monic irreducible polynomial g,(y) K1_ 1[y].
Let f,(x)=g1(x")e K1_1[x] be the minimal polynomial of A [see §5.5].
lfcharF=0, then
The situation in Lemma 2 has the following graphic representation:

'
F(a) F(a1) F(a2)

whereas that in the corollary is depicted as follows:

F(a) F(a1) P(a2) F(a1.)

With this preparation, we have the following theorem.

Theorem. Consider a finite extension K= F(AI, of F, and let Q/F

be any extension containing all zeros of the minimal polynomialsf(x) of
whose reduced degrees are n$, I i k. Then contains precisely
n* = intermediate fields conjugate to K/F.

Proof First, consider an isomorphism defined on K/F (into Define

an isomorphism on K1/F by restriction, I i k; namely, for c4 K1,

=
a K1_1 = a, Thus, an isomorphism
of K/F determines a chain of isomorphisms (a1, ...,a&}.
 Counting of Isomorpblsms and Separability 2$3

If a and r are distinct isomorphisms on K/F, then there exists a

subscript 1, I I k, such that for the giren tower of fields
aIK1_1 = = =
= t, =
That is. a and i restrict to distinct prolongations on K. =
Conversely, if (a1. a2, a chain of isomorphisms on the tower
of fields such that a.1K1_,
the isomorphism a = a
a typical piece
K1

to describe K/F. An isomorphism ). defined on K,_1 has, by the above

corollary, precisely n? distinct prolongations to K.. It follows that the identity
map on F has exactly distinct prolongations to K/F. The total
number of isomorphisms of K/F (from the corollary) is independent of the
tower of fields used in describing the chain of isomorphisms ...,
associated to an isomorphism a. Hence =
flk1 ,,*
The integer n*, denoted also [K: is called the reduced degree of
K/F ot the degree of separability. The quotient = 101 is
called the degree of inseparability, denoted [K: F]1, of K/F.
We conclude this section with a series of seven corollaries derived
from the preceding theorem. Some of the proofs will be left to the reader.

Corollary 1. If the elements A1 of the theorem are separable over the fields
K1_ then K/F has precisely n = [K: F] isomorphisms which restrict to the
identity on F.

A more general statement is given in Corollary 2. There "sufficiently

large" means any field containing all zeros of the polynomialsj(x) =
in the proof of the theorem.

Corollary 2. Each isomorphism A from F to F = A(F) has precisely [K: F]

prolongations to isomorphisms of K into a sufficiently large field n/F.

Corollary 3. If all the elements A1 are separable over K1_ j. I i k, then

K = F(A) where A is a separable element over F; [A : F] = [K: F] n.

The Theorem of the Primitive Element implies that K = F(A).

Since K/F has, by Corollary 1, exactly [K: F] distinct conjugate fields, the
element A must have [K: F] distinct conjugates, the zeros of the minimal
nIA(x) e F[x] of A. If A were inseparable, then in4(x) would have fewer than
[K: F] distinct zeros [see §5.5]. Hence K/F would have fewer than [K; F]
conjugate fields.
 Field Theory chapter 8

Corollary 4. If K = F(A) with A separable over F, then every element B in

F(A) is separable over F.
Proof: Consider the tower of fields F F(B) c F(B)(A) = IC. If B were
inseparable, then F(B) would have fewer than [F(B): F] isomorphisms over
F. Consequently K would have fewer than [K: F(B)][F(B) : F] = [K: F]
isomorphisms over F, contrary to the assumption of the separability of A over
F, since A is separable over F(B).
This corollary together with Corollary 3 implies that a field
K = F(A1 Ak), obtained by adjoining elements A1 separable over F, is a
separable extension. (An extension K/F was defined in §8.1 to be separable if
each a E K\F is separable over F.)

Corollary 5. A finite extension field K/F is separable if and only if K has

precisely [K: F] isomorphisms over F.
Proof. If K/F is separable, then the number of isomorphisms of K is [K: F]
by Corollary I. Conversely, if K/F is inseparable, consider an element
A4 e K/F inseparable over F. Using the notation of the theorem, write
K = F(A1, A2, ..., and note that <n4. Hence the number of iso-
morphisms n n = [K: F], a contradiction.

Corollary 6. If A and B are separable algebraic elements in a field 12/F, every

rational function h(A, B) of A and B with coefficients in Fis separable over F.
For the proof apply Corollary 4 to the extensions
F F(h(A, B)) c F(A, B).
The field F(A, B) is separable over F. Hence the element h(A, B) is separable
over F.

Corollary 7. If the finite extension K/F is separable and the element A is

separable over K, then A is separable over F.

Exercises

1. Prove the corollary to Lemma 2.

2. Prove Corollaries 1 and 2.
3. Prove Corollary 7.

§8.6 Prelude to Galois Theory

With the preliminaries 8.4, and 8.5 complete, we now begin
consideration of Galois theory proper, culminating in the next section with
the statement and proof of the Fundamental Theorem of Galois Theory. In
 Prelude to Galois Theory 285

this section we focus our attention on two special types of algebraic fIeld

extensions.
A field extension K/F is called normal if
(I) K is algebraic over F,
(ii) any irreducible polynomial g(x) e F[x], which has a zero in K,
has all of its zeros in K.

Theorem 1. The splitting field K/F of a (not necessarily irreducible) poly-

nomialf(x) e F[x] is a normal extension of F.
Proof. Let A1, ..., A,, be the distinct zeros of f(x) e F[xJ, and write
K = F(A1, ..., Am). Suppose now that the irreducible polynomial g(x) has a
zero B in K.
Let 11/F be a field containing K and all zeros B, B', ... of g(x). Con-
sider an arbitrary zero B' e 11 of g(x). By Corollary 1, §8.4, there is an iso-
morphism A over F from F(B) to F(B') such that 2(B) = B'.
Lemma 2, §8.4, asserts the existence of a prolongation A of A defined
Ofl F(B)(A1, ..., Am), such that
A(F(B)(Ai,...,Am)) = F(B')(A(Ai),...,A(Am)).
To complete the argument, observe that by hypothesis an element
B€Kc11 is a polynomial B= in the elements with
coefficients in F. Application of A to B yields
B' = A(B) =
q (A(A ..., A K because A maps the set {A ..., A,,}
to itself since A(A,), 1 i m, must be a root of 1(x) (by Lemma 1, §8.5).
Hence B' e K, and thus g(x) splits in K, as asserted.

The definition of normal extension does no: exclude inseparable extensions.

For example, let F = Z9(t) where : is an indeterminate over The polynomial
E F[x] is irreducible. (See Eisenstein's Criterion, §5.8, and use the prime ideal
(t) in The coset A = t)F[x] in K = F[x]/(x"—:) satisfies I = 0.
Hence y"— g = (y— in K[yJ, y an indeterminate over K. Consequently, x"— t is
an irreducible inseparable polynomial whose splitting field K = F(A) is normal
according to Theorem I.
Hereafter only separable polynomials and finite separable extensions
will be considered unless otherwise mentioned.
Recall that an isomorphism a defined on K= F(A) is an automorphisni
of K/F if and only if a(A) K and a F is the identity map. If [K: F] =
there are n or fewer automorphisms of K/F [see Lemma 2, §8.5].
A separable normal extension K/F of finite degree is called a galois
extension. It is now easy to summarize the preceding discussion on splitting
fields of polynomials and separable extensions in proving the following
important theorem.
 Field Theory chapter 8

Theorem 2. A galois extension K/Fof degreen has preciselyn automorphisms

over F. These automorphisms form a group G(K/F) = G, the so-called
galois group of K over F.
Proof The separability of K/F implies that K/F has exactly n = [K: F]
conjugate fields over F by Corollary 5, §8.5. By the Theorem of the Primitive
Element, §8.3, K = F(A) for some A e K. Let rn(x) = fl71(x—A1) e F[x]
be the minimal polynomial of the primitive element A = A By the normality
ofK/F, and so
F(A) coincides with F(A), and the n mappings
a1: F(A) F(A1) are automorphisms of F(A)/F.
The set G = = a2, is a group under composition of
mappings. That is, ct(a) = a(t(a))for all a e K. It follows immediately from
the definition of at that (crt)p = a(rp). Every automorphism a has an inverse
a is an isomorphism of K onto itself. Hence G(K/F) = G is a
group, and the proof is complete.
Historically the automorphisms of a e G were viewed as permutations
of the ordered set of zeros {A1, ...,A1, of f(x). For a(A1) =
where u(i) is uniquely determined by a for I I n, the map
1 •..
... "l=ir(a)
Lc(l)".a(i)".a(n)J
determines a unique element it(a) of the symmetric group E,,, such that
= n(a)x(t). (The mapping it can be shown to be an injective homo-
morphism of G into In the older literature the galois group G(K/F) is
viewed as a permutation group of the zeros of the "defining polynomial
f(x)" of K/F.
We must guard against the misconception that every element of
originates from or gives rise to an automorphism of K/F. A permutation of
E,, does not necessarily give rise to an automorphism of a normal separable
extension, although there are, for each n 2, extensions K of the field of
rational numbers Q whose galois groups are isomorphic to E,,.

Example 1. The splitting field of x3 —2 e Q [x) is the extension K = Q(w, of

of degree 6 over Q, where o2 + w + I = 0. The automorphism group of K/Q is
described as follows by its effect on the generating elements of K:
a(w) = (02
and r(o) = co, r((2) =
so that = = I and = r2. Thus, G(K/Q)

Example 2. To illustrate that an irreducible cubic polynomial does not necessarily

give rise to an extension of degree 6 whose group of automorphisms is isomorphic
to consider
f(x) = x3 — 21x + 28 Q[x].
 ________

Prelude to Galois Theory 287

This cubic polynomial is irreducible in Z[x], and hence also in Q[x] by the Lemma
of Gauss [*5.8]. If C cC isa zero of 1(x), in particular a real zero sincef(x) has odd
degree, the elements
C'

and =
also are real zeros of 1(x), as is easily checked by direct substitution intof(x). Con-
sequently the group of automorphisms of the splitting field Q(C)/Q is cyclic of
order 3.

Exercises

1. a.Determine the degree of [Q(i) : Q}.

b.Is a galois extension? If so, determine its galois group.
2. Consider the finite normal extension K = of Q. Com-
plete the blanks.
a. The degree[K:Q] is
b. The order of the galois group G(K/Q) is
c. The order of G(L/Q) is when L =
d. The order of G(L/Q) is when L =
e. The order of G(L/Q) is when L = f8).
1. The order of G(KIL) is when L =
3. Let K be the field of Example 1, and (0 a root ofx2+x+1.
a. Find a primitive element A of K over Q.
b. Determine the minimal polynomial of A over Q and over Q(a).
c. Find all conjugates of A over Q and over
d. Find the conjugates of = A(l +2w—w2) over Q and over Q(o).
4. Prove that the polynomial 1(x) = x4 + 4x2 +2 is irreducible in Q [x]. Let
be a root of 1(x) in the splitting field K/Q of f(x).
a. Prove that the mapping
3 3
a1& -+ a1fl',
1=0 1=0

with e Q and fi = + is an automorphism of Q(c)/Q.

b. Find the galois group of Q(or)/Q and determine its structure.
5. Find the galois group of Q(i, 1/Ti) over Q where i2 — I.
6. Prove that x' + I E Q [x] is irreducible. Find the splitting field and
determine the galois group and all the subfields of the splitting field.
7. Let K = Q( Prove that K/Q is a galois extension. Deter-
mine the effect of theelements of the galois group G(K/Q) on 1/i
8. Why is the splitting field of a separable polynomial a galois extension?
9. Prove that the galois group of a splitting field of x" — a c Q [x] is solvable,
when a is not a dth power of an element of Q, din.
 Field Theory chapter 8

JO. Let,4beazerooff(x)=x3—3x+l€Q[x].
a. Prove that — 2+A2 is another zero of 1(x), and that K = Q(A) is a
galois extension of Q.
b. What is the third zero of 1(x)?
11. a. Prove that x4—2 is irreducible in Q[x].
b. Let K be the splitting field of x4 —2 in C. Prove that K = Q 1),
where = — I and is a real 4th root of 2.
c. Prove that the galois group G(K/Q) has order 8 and contains a
cyclic subgroup of order 4.
d. Find a set of generators and relations of G(K/Q).

§8.7 The Fundamental Theorem of Galois Theory

The following terminology is useful in providing a succinct formula-
tion of the Fundamental Theorem of Galois Theory.
The fixed field (1(H) of a subgroup II of G G(K/F) is defined to be
= (a e K: c(a) = a for allae H}.
It is a subfield of K since for a, b e
u(a+b) = a(a) + cr(b) = a + b,
a(ab) = a(a)a(b) ab;
furthermore
cT(aa1) = a(a)a(a') = u(l) = I
implies that a(a') = (a(a))' for a 0. As a is an automorphism of K/F,
by definition a(c) c for all ce F, and thus F c 1(H).
Now associate to an intermediate field L/F of the extension K/F
(that is, L is a subfield of K) the subset
f(L) = (a G(K/F): a(l) = I for all I L}

of G = G(KfF>. That U(L) is a subgroup of G is easily verified. Furthermore,

F(L) = G(K/L), the group of automorphisms of K whose restriction to L is
the identity mapping.

Lemma. If L is an intermediate field of the galois extension K/F, then K is

also a galois extension of L.

Let g(x)e L[xJ be an irreducible polynomial which has a zero

a e K. To prove that all other zeros of g(x) lie in K consider the minimal
polynomial ,na(x) e F[x] of a. Since ma(x) belongs to L[x], as well as to
Fix], and since g(x) is the minimal polynomial of a wit/i coefficients in L,
we have that Therefore, each root of g(x) is a root of
Since K/F is a normal extension and a e K. all roots of ,n0(x) belong to K.
Hence all roots of g(x) belong to K, and K/L is a galois extension as asserted.
 _______________J

The Fundamental Theorem of Galois Theory 289

The Fundamental Theorem of the Galois Theory. Let K/F be a galois

extension with the group G(K/F) = G.
(I)If L/Fis a subfield of K, then D[f(L)] = L.
(ii) lfHis a subgroup of G, then f[D(H)] = H.
(iii) [K: L] = [F(L): I] and [II: 1] = [K:
(iv) If L/F and ElF are conjugate subfields (in the sense of §8.4) of
K/F, then F(L) and r(L) are conjugate subgroups of G by an
inner automorphism. Conversely, if H and H are conjugate
subgroups of G, then 1(H) and are conjugate subfields
of K/F.
(v) An intermediate field L/F of K/F is a galois extension if and only
if the corresponding subgroup in G is normal. The galois
group G(L/F) is by restriction of G to L naturally isomorphic
to the factor group G/r(L).
Parts (i) and (ii) of the theorem can be visualized by a diagram:

Tower of Fields Tower of Groups

UI nI

UI flI

= r(L)
UI flu

F" G = G(K/F)

Proof of(i). if Ic L, then a(l) = I, for all a f(L) by the definition of T(L).
Hence L c D[r(L)]. Consider an element B of the fixed field c1[r(L)]. Then

K a galois extension of L(B) by the preceding lemma. By the definition

of we have c(B) = B for all a e r(L), and thus r(L(B)).
Since every automorphism a of K which prolongs the identity on L(B)
certainly prolongs the identity on L, we also have r(L(B)) r(L). Thus,
r(L) = r(L(B)). However, if L(B)/L were to be a proper extension, then we
would have
[T(L(B)): I] [K: L(B)] < [K:L] = [r(L): 1].
The equalities are taken from Theorem 2, §8.6, and the inequality from
Theorem 2, §8.1. This observation contradicts the equality = r(L(B)).
Thus, L(B)/L must be an improper extension; Be L; and = L.
290 Field Theory chapter 8

Proof of (ii). Let M = For A M we have a(A) = A for all a H;

thus H G(K/M) = r(M). Therefore, by Theorem 2, §8.6,
(*) [K: M] = {G(K/M): I] [H: I] = Ii.

By the Theorem of the Primitive Element, §8.3, K = M(B) for some element
BE K. Now let e = a1, a,, be the elements of H and consider the polynomial

where the coefficients e1, ...,e,, are the elementary symmetric functions of
a1(B),...,a,,(B). That is,
e1 = a1(B) + + a,,(B) =

e2 a,(B)o2(B)+ + a,,_i(B)cr,,(B) =
i<j
e3 = a.(B)aJ(B)a,,(B),
i<j<k

eb = a1(B)...a,,(B) =

Then = for I i,j I,.

Consequently M, and thus B = a1(B) is a zero of the poly-
nomial g(x) E M[x] which has degree h. Therefore, [K: M] I, = [H: 1].
This inequality combined with the inequality (*) implies that
H= = G(K/M) =
Proof of (lii). Since r(L) = G(K/L), by the preceding arguments Theorem 2,
§8.6, implies
[f(L): I] = [K:L].
Similarly, [K: D(H)] = [G(K/b(H)): I] = [H: 1].
Proof of(iv). Suppose that L/Fand L/Fare conjugate subfields in the galois
extension K/F. Let A be an isomorphism L L extending the identity mapping
on F, and let A be a prolongation of A to K/F, as in §8.4. Then A e G(K/F)
since eveiy isomorphism defined on K/F is an automorphism of K.
Let H = G(K/L) and ii G(K/L). Then
[H: I] = [K:L] = [K:F][L:F]' =
= [K:LJ = [H: 1],
since L and E are (isomorphic) conjugates over F. Finally, for an arbitrary
element 1 = 2(l) = A(/) of Leach automorphism a e H satisfies (AaA')l =
[(AaA 1)A]/ = A(/) = 1. Thus AHA' c H, and consequently AHA' = H,
since [AHA' :1] = [11:1] = [fl: 1].
§8.7 The Fundamental Theorem of Galois Theory 291

Now conversely, choose a G such that 11 = aHa Writing

as F(A) for some appropriate A e consider M = F(a(A)).
Since A and a(A) are roots of the same irreducible polynomial in F[x], the
fields F(A) and M = F(oiA)) are conjugate extensions of F; hence they have
the same degree over F. We show next that M = Ffr(A)) is the fixed field
t(H) of H. For any H and any rn E M, where rn is a rational function of
cr(,4) with coefficients in F, observe that
= (in) rn

since t = a e H. and hence t(A) = A. Therefore, M 1(H). Finally,

since M and cD(H) have equal degrees over F, as follows from the equalities
[M: F] = [K: F][K: M3' = [K: F][K: F(A)]'
= [K:F][H: = [K:F][R: l]'
=
=
they coincide. Therefore, M=
Proof of(v). Let L/P be a galois extension ofF with the galois group G(L/F).
As before, let H = G(KJL) = r(L). The restrictions a)L of the auto-
morphisms a E G are automorphisms 5 e G(L/F), since L/F is a galois
extension (i.e., every isomorphism defined on L/F is an automorphism of L).
The restriction mapping a a homomorphism p: G(K/F) -+
G(L/F). The kernel of p consists of those automorphisms a of K which
restricted to L are the identity map on L. That is, kerp = r(L), and thus
H is a normal subgroup of G(K/F).
Consider now e G(L/F). The lemmas on prolongations in §8.4
imply that f has an extension t to K such that r L = f. Furthermore,
t G = G(K/F) since f F is the identity map. Thus, the restriction map p
is surjective. By the Jsomorphism Theorem
GIr(L)) p(G) = G(L/F).
Conversely, suppose that H is a normal subgroup of G with the fixed
field L/F; H = G(K/L). As in the proof of(iv), a conjugate field L/FofL/Fis
equal to a(L)/ F for some a E G, and G(K/L) = aG = aHa' =
11 = G(K/L). Part (I) implies that

L = D(G(K/L)) = 1(G(K/L)) = L.
That any conjugate field L of L must coincide with L is to say, however, that
L is a galois extension field over F. Now let L = F(B). The element B is a root
of an irreducible polynomial E F[x]. Any other root gives rise to a
conjugate extension E = F(B), which coincides with L. Thus L contains all
roots off(x) and by Theorem I, is a normal (and hence) galois extension
ofF.
292 Field Theory chapter 8

As a corollary observation to part (v), note that an intermediate field

L is a galois extension over F if and only if L coincides with its conjugates in K.
The Fundamental Theorem of the Galois Theory provides a corre-
spondence between intermediate fields of K/F and subgroups of the auto-
morphism group G(K/F) and between normal, separable extensions and
normal subgroups. This correspondence was previously discussed in
Proposition 6, §8.2, in the special case of finite fields.

Exercises

1. Consider a galois extension K/Q of degree p2q, where p and q are distinct
primes such that q <p and — I).
a. Prove the existence of intermediate fields L and M, such that
[L:Q]=p2,[M:Q]=q.
Prove that these extensions L and M must be galois extensions of Q.
b.
Prove that K must be an abelian extension of Q (i.e., the galois group
c.
G(K/Q) must be abelian).
2. If K/F is a cyclic galois extension (i.e., has a finite cyclic galois group),
prove:
a. Every intermediate extension L is cyclic over F.
b. For each divisor d of [K: F] there exists a unique intermediate
extension L4 of degree d over F.
c. Is statement (b) necessarily true if K/F is not cyclic? Why?
3. Prove that the splitting field of J(x) = x4 + x2 ÷ 1 e Q [x) has degree 4
over Q.
4. Determine the structure of the galois group of the splitting field of
f(x).= x4—5x2+6cQ[xJ.
5. Find a basis of the splitting field of f(x)= x4+x3+x2+x+1 overQ. If
K where satisfies = 0, determine all subfields of K/Q.
6. Find the galois group of f(x) x3 — 3x2 + 5x— 2 e Q[xJ. (Hint: Show
that 1(x) is irreducible by getting rid of the term," let
with a eQ, and examine the resulting equation y3 + by+ CE Z[xl modulo
suitable primes p e Z.)
7. Determine the field of invariant elements in F(x) associated with the group
of automorphisms generated by the map x —+ l/x.
8. Let K = Q(A) where A is a zero of x4 —10. Find the galois group of K/F
and determine all subfields of K.

§8.8 Consequences of the Fundamental Theorem

This section presents several corollaries to the Fundamental Theorem
of the Galois Theory and extends the results of that theorem to the case of
products of fields, analogous to the products of groups encountered in §6.2.
Two significant consequences of the Galois theory are left to the next chapter:
 Consequences of the Fundamental Theorem 293

solvability of polynomial equations by radicals, §9.2, and constructions by

ruler and compass, §9.3. For the reader who wishes to study these con-
sequences the only further prerequisite is §9.1 on cyclotomic field extensions.

Corollary I. The following strict inclusion relations hold for the subfields
L1/F. L2/F of the galois extension K/F and for the subgroups Ii,, H2 of the
galois group G(K/F) =
L1 L2 implies r(L1)
H1 c H2 implies D(H1)
Suppose that r(L1) = r(L2). Then by part (1) of the Fundamental
Theorem, = L1 = tV[F'(L2)] = L2, a contradiction. Similarly, if
= then by part (ii) F[cb(H1)] = = = 112, again
a contradiction.
It is convenient to let L, L2 denote the least subfield of K containing
both and L2; we speak of the field L1 L2 as the product or composite of the
fields L1 and L2. It is obtained by adjoining the elements of L1 and L2 to F
(and hence consists of finite sums of products of elements in L, and L2).
Similarly, the product of the subgroups 112 of G was defined at the end
of G containing both and 112.

Corollary 2. Consider intermediate fields L1, L2 of a galois extension K/F

and subgroups H2 of the galois group G(K/F). Then
(i) r(L1 L2) = f(L1) r(L2),
(ii) •(H1.H2) =
(iii) r(L1 L2) =
(iv) 'D(H1 =
The following diagrams show the inclusion of subfields and the
corresponding subgroups of Corollary 2.

K {E}

UI fli

L1L2 H2

4,
L1 L2 H1

0/
L1 L, I!1112
UI fli

F G
294 Field Theory chapter 8

Proof of(i). Consider a e r(L1 L2), E L1, and '2 L2. Then '2 L1 L2;
in particular, a and
ae a a r(L1) r(L2). Conversely, for r r(L1) F(L2) we
have = and r(l,) = 12 for all elements a L1 and '2 L2. Since the
elements of L1 L2 are, by definition, finite sums of products of elements in L1
and L2, it follows that r f'(L1 L2). Hence F'(L1 L2) = f(L1) F(L2).
Proof of(iO. We use an analogous argument. Recall from §6.2 that elements
of the form 01 02, where a, a H1, i = 1,2, generate H1. H2. For A a
we have a,(A) = = A, since a H1. H2, I = 1,2, for all e H1.
Thus A and Conversely for Be
'1(H2) we have o1(B) = B for all UI a I = 1,2. Therefore,
a2)(B) = B for all 02 H2 and hence for all rE H2, t(B) = B.
Thus, c1(H1) ct'(H2) c .1/2), and c1(H1 = b(H1)
Proof of (iii). The inclusion relations
and

imply, as in Corollary 1, that

L2) 2 f'(L1) and r(L1 L2) 2

hence r(L1 L2) 2 This inclusion cannot be proper because,
if it were, Corollary I and the properties of the operations t1 and F would
imply that
L2 L, = Li)] c=
= = L1 L2,

a contradiction. Hence V(L1 L2) = F(L1).F(L2).

Proof of (iv). Similarly, the inclusions H1 112 c H1, 1 1,2, imply that
H2) (1)(H1). Hence (D(H1 112) 2 (1)(H1)'Z)(H2). This cannot be
a proper inclusion, because
r, 112 = F[(1)(111 r'i 112)] c f[4)(H1)(1)(H2)]
= r[D(111)]
= H1

leads to a contradiction.
Theorem 1 (On Natural Irrationality). Let K/F be a galois extension with the
galois group G = G(K/F). If LI F is an extension such that both K and L are
contained in a common field I'/F, then:
(i) KL a galois extension of L.
is

(ii) The galois group G(KL/L) is isomorphic to the galois group

G(K/(K L)).
§8.8 Consequences of the Fundamental Theorem 295

The inclusion relations between the various fields of this theorem may
be illustrated as follows:

KL

Proof of(i). Since K/F is a galois extension and hence separable, the Theorem
of the Primitive Element implies that K = F(A). Further XL = L(A),
as follows. First, KL L(A), since KL contains Land A. Second, by definition
KL is the intersection of all fields containing both Kand L. Certainly L(A) L,
but also L(A) K = F(A) since L contains F and A.
Part (i) follows from the fact that the zeros of the minimal poly-
nomial f(x) e F[x] of A are polynomials in A with coefficients in F because
K/F is a galois extension. Thus KL is the splitting field of f(x) and hence a
galois extension over F [see §8.6).
Proof of (ii). Consider the restriction mapping p: G(KL/L) —÷ L))
given by p(u) = a K belongs to
G(K/(K ri L)). Note that a restricted to L (and hence to K n L or F) is the
identity map. Moreover since a(A) is a root of f(x), it is a polynomial in A
with coefficients in F. Thus a(K) K, and so a K belongs to G(K/F) and
hence to G(K/(K L)). Since (a I K)(r 1K) = (air K) by definition of
restriction to K, the mapping p is a homomorphism.
Now for a z in G(KL/L), a(B) r(B), for some element BE K\L,
and hence rIK. Thus, the mapping p is one—one. Denote Imp by
G(KL/L) I K. To prove that G(KL/L) I K = G(K/(K L)) it suffices to show
that the fixed field M K of the subgroup G(KL/L) I K G(K/F) is K L.
For bE M, c(b) = (al K)(b) = b and so bE L. Therefore M K L. Con-
versely for c K c = a(c) = (aI K)(c) and therefore K L M.
As a corollary result, we have the following proposition.

Proposition. Let K/F be a galois extension and L/F a finite extension such
Then [K! :F]=[K:F)[L:F].
 Field Theory chapter 8

For the proof note that the galois groups G(KL/L) and
G(K/(K L)) = G(K/F) are isomorphic. Consequently,
[KL : F] = [XL : L][L : F] = [K : F][L F].

As a cautionary note that this result does not hold for arbitrary finite
extensions K/F and L/F, consider the extensions K = and L =
wherew2+co+l=O. Then and (KL:Q]=6, but
[K:Q][L:Q] = 32

Theorem 2. For galois extensions K11 F. i = 1,2, contained in a field Q/F

with respective galois groups
(I) The product field K1 K2 is a galois extension of F.
(ii) The mapping G(KI K2/F) —' G2 given by g(a)=
(°l K1, a K2) is an injective homomorphism.
(iii) The pair (a1,a2)eG1 xG2 determines an automorphism of
K1 K2/F if and only if
a11(K1 rK2) = a21(K1 K2).

(iv) The galois group G = G(K1 K2/F) is isomorphic to the subgroup

H= ((a1,a2)E G1 nK2)= a2((K1
of G1
Proof of (1). Let K1 = F(a1), I = 1,2. Then the composite of the fields K1
and can be expressed as K1 K2 K1(a2). Suppose now that [a2 : K1] =
k F] = [K2 : F]. Then the elements of K1 K2 are of the form
+Yk_1Q2 j< k.
Furthermore,
= + + ... +
with coefficients h1, F. Therefore a E K1 K2 can be written as
a =
j. I
where 0j < k and 0 1< ii,. For an isomorphism a on K1 l(2/F
a(a) =
j,1
belongs to K1 K2 since (a I K1)(a1) K1, i = 1,2, because K1/F is a galois
extension. Thus K1 K2 is a galois extension of F.
Proof of (II). The mapping p is a homomorphism since for b1 E i = 1,2,
[(ar)1K1](b1) = (ar)(b1) = a(r(b1))
= (ajK1)[(tIK1)(b1)}.
 Consequences of the Fundamental Theorem 297

If a I K1 is the identity mapping on I(1/F, i = 1,2, then

a(a) = =a
j,'
for a/la e K1 K2. Hence a is the identity isomorphism of K1 K2/F, and q' is an
injection of G into G2.
Proof oJ'(iii). If isomorphisms a1 e G1 and a2 G2 have different restrictions
to ri K2, there can be no automorphism a in G such that
= aIK2 = a2.
If there were such an automorphism, then we would have the contradiction:
K2) = aI(K1 K2) = cr2I(Ki K2).
The converse, that an element (c1,a2) e G2 determines an
automorphism a of K1 K2/F such that al K1 = a1, i = 1,2, if I(K1 K2) =
a2 K2), requires careful proof, for which we use the Theorem on
Natural Irrationalities. The following diagram illustrates the relations between
the various fields.

Ki K2

K1
K2

p(H1)]
p(H2)
K1 K2

Here denotes the restriction of the galois group of

K1 K2/K2 (a gatois extension by Theorem 1) to the galois group of
K1/(K1 K2). Similarly p(H2) is the restriction of the galois group H2 of
K1 to the galois group of K2/(1C1 K2).
The groups H1 and H2 are normal in G(K1 K2/F) since K1 K2 is a
galois extension of both K2 and KI by the lemma of §8.7. Consequently the
product group H2 is normal in G. By Corollary 2 above the fixed field
D(H1 = K1 K2 and so H1 •H2 = G(KI K2)). Furthermore
H2 = {e} since t1(H1 H2) = K1 K2. Thus the product H1 H2 is
direct; by the properties of the internal direct product
= for e H., 1 1,2
 Field Theory chapter 8

[see §6.10]. Application of the Fundamental Theorem of the Galois Theory

yields
K2)/F) G(K1 K2/F)/H1 ®
Also the automorphisms t in G((K1 K2)/F) are restrictions of auto-
morphisms in G(K1 K2/F) = G. For a1 E H1, i = 1,2, the restrictions
a2) and are equal. By virtue of the isomorphism (*) each prolonga-
tion oft e G((K1 n K2)/F) to K1K2/Fmust be of the form for suitable

Now finally consider mappings a, e G4, I = 1,2, that are equal on

K1 ci K2. As a consequence of the isomorphism (*) (K1 ci K2) =
I (K1 ci K2) has a prolongation a2 to K1 K2. To prove that the auto-
morphisms a2 can be selected so that
= a1 and =
we must solve the following for a1:
(a1 1K1) ° (a2

= o p(a1) o 1 =
where, according to Theorem 1, the restriction mapping p determines an
isomorphism between H1 = G(K1 K2/K2) and G(K1/(K1 ci K2)). Because of
this isomorphism
p(a1) = ° a1,

and thus a1 e is determined. Similarly we can determine a2 e H2, and so

obtain an automorphism a for which
= (a1,u2)e G1 G2.

Proo/'of(iv). Combining parts (ii) and (iii) yields the desired isomorphism.

Exercises

1. Let a be the map f(x)—i.f(1 —x) on the rational function field F(x).
Prove that F(x2 — x) is the fixed field for {e, a), assuming that char F 2.
2. Assume that A and B are elements of the galois extension K/F with
respective conjugates
AA1,Ai,..,,Am and
and that [F(A, B) : F] = mn. Prove the existence of automorphisms
a,,j e G(K/F) such that a1,1(B) = for 1 I m,
I I n.
3. Prove that if K,/Fand K2/F are solvable galois extensions, then so are
K1 K2 and K1 ci K2.
§8.9 Algebraic Closure 299

4. Denote by Fthe field of rational functions of x with coefficients in C, and

let C be a primitive pith root of unity.
a. Prove that c(x) = Cx and r(x) = x' determine automorphisms of
F of respective orders n and 2.
b. Prove that at = ra', and that the fixed field of the dihedral group
generated by a and r is
5. Let 1(x) = x3+a1x2+a2x+a3 be an irreducible cubic in F[x] whose
galois group is isomorphic to Prove that

D (A1—A2)(A1—A3)(42—43)
generates a subfield L F of the splitting field K = F(A1,A2,A3) if
2.
6. If the degree of every finite algebraic extension of F is divisible by p,
prove that [K: F] must be a power of p, p a prime, for any K/F.
7. Let A = .1TT41, where i2 = — I.
a. Determine [Q(A) : Q].
b. Show that C2 — C + i = 0 has a solution C in the least galois extension
over Q that contains Q(A).
c. Prove that is contained in the galois extension of part (b).
8. Determine the galois group of the polynomial x4 + 2 a Q [x] over Q.
9. Prove that 1(x) = x4 + 30x2 + 45 e Q [x] is irreducible. Let A be a zero
off(x), and show that Q(A)/Q is a cyclic extension of degree 4. Describe
the effect of the galois group on A and exhibit the quadratic subfield
K/Q of Q(A)/Q. (Note that A has the form .13).)
10. Suppose that K/F is a galois extension with the intermediate field L/F.
Prove that the galois group G(K/M) of the smallest field M for which
L M K and M/Fis normal equals

§8.9 Algebraic Closure

In §5.4 we proved that each polynomial /(x) E F[x] splits in some
finite algebraic extension of F, and that such extensions depend upon the
polynomial f(x). We now address the more general problem of constructing
an algebraic extension of F in which every polynomial 1(x) a F[xJ splits. The
proof requires transfinite induction or the lemma of Max A. Zorn. We prefer
the latter approach, and begin with some set-theoretic preliminaries.
A set S is termed partially ordered if there is given for it a relation
between certain pairs of elements x,y in S such that the following relations
hold:
(I) xx
(ii)
xy imply
x x need not hold for arbitrarily selected elements
x, y a S; hence the terminology partial order.
300 Theory chapter 8

A subset T of a partially ordered set S is called totally ordered if

either x y or y x for every pair of elements x,y e T. An element s e S
is called an upper bound of a nonempty subset U S if u s for all u U.
A partially ordered set S is termed inductively ordered if every totally ordered
subset Tof S has an upper bound in S. An element n: in a partially ordered set
S is called maximal ii for x e S with m x necessarily rn = x.

Zorn's Lemma. if S is a partially ordered set which is inductively ordered

and nonempty, then S contains at least one maximal element.

We accept this lemma as an axiom. For an interesting discussion of it,

see S. Lang, Algebraic Structures.
Recall that an ideal M in a commutative ring R is maximal if lvi R
and if there is no ideal B for which M B R. The residue class ring RIM
of an ideal Mis a field if and only if Mis maximal in R [see Exercise 15, §3.6].

Lemma 1. if A is a proper ideal in a ring R, then A is contained in a maximal

ideal M of R.
Proof Let T = {C} be a totally ordered subset of ideals C in R such that
C A and I C. Then the set-theoretic union K = Uc€ is an ideal and
an upper bound for T. To verify that K is an ideal consider x,y e K, r R;
then x C', y E C" for some C', C" E T. Since T is totally ordered, we may
assume, without toss of generality, that C' cC"; therefore x+ye
C". Hence K is an ideal, and of course K A. Finally I K, since I C
for all C e 7', and thus K is a proper ideal.
Now, applying Zorn's Lemma, we obtain a maximal element M in
the set S of all proper ideals in R which contain A. Then M R, since I M;
furthermore M is a maximal ideal.
A field F is said to be algebraically closed if every polynomial
f(x)e F[x] has a zero e in F, i.e.,f(x) = (x—c)g(x) with g(x)€ F[x]. That
a field F is algebraically closed means that every polynomial in F[x] can be
written as the product of linear factors in F[x].

Lemma 2. For any field F there exists an algebraically closed field

containing F.
Proof To each f(x) F[x], degf(x) 1, associate a symbol A'1. Then the
set S of symbols A'1 is in one-one correspondence with the polynomials of
degree at least I in F[x]. Next consider the polynomial ring F[fX1}] con-
sisting of all polynomials in any finite number of the indeterminates X1 with
coefficients in F. To each polynomialf(x) e F[x] associate the unique poly-
nomial f(X1) F[{X1}]. The ideal A ( (A'1), ...) generated by f(Xf),
for all f(x) F[x] of degree at least 1, is a proper ideal in F[{X1}], because
if A were the whole ring then I would have to be a linear combination of a
 Algebraic Closure 301

finite number of polynomialsf(X1) with coefficients in F[(X1}]. Letf1, ...,f5

be these polynomials in the indeterminates Xf, then

= g1 e F[(X1}].

Since there are finitely many polynomials g1, each with a finite number of
indeterminates, there are only finitely many indeterminates
x1, ..., x5, + ...,
involved in the relation (*). We have

According to the basic construction of Kronecker, §5.4, there exists a finite

extension K/F in which every polynomial J (X1) has a zero a. Finally let
a1 = 0 for I> s, and consider the homomorphism given by X1 —p a, for
1 Then in K,

1 = g.(a1, ...,a5,0, ...,0)J(a1) = 0,

a contradiction. Hence I A, and A is a proper ideal.

Lemma I implies that A is contained in a maximal ideal M of
F[{X1}]. Thus, the residue class ring F[(X1}]/M is a field K1. This field K1
contains a zero of every polynomial /(x) e F[x] because f(X1) e A M;
specifically, the coset X,. modulo M, denoted in K1 satisfies = 0.
The construction of the field K1 can be repeated recursively so that
the field contains a zero of every nonconstant polynomial in K,,[x:J:

At each step the field K,, is identified with the residue class ring of a polynomial
ring over K,,... 1,constructed like F[{X1}], modulo a maximal ideal, such as M,
in F[{X,)].
Now let U,,K,,. This field is algebraically closed. First, Q is
indeed a field. If a,b eQ. then a,b lie in K,, for some sufficiently large ii. Since
K,, is afield, a±b, ab, and a/b, if b 0, are in K,, c Q. Second, ifg(x)eQ[xJ,
then g(x) Km[x], for some m, since the number of coefficients in g(x) is
finite. By construction, Km+i contains a zero of g(x); hence Q contains a
zero of g(x). In other words, Q[x] does not contain any irreducible poly-
nomials of degree greater than 1. Therefore Q is algebraically closed.

Theorem. Every field F has an algebraic extension K that is algebraically

closed.

Such a field K is called an algebraic closure of F. All algebraic closures

of F are isomorphic over F.
302 Field Theory chapter 8

Proof. As in Lemma 2, let Q be an algebraically closed extension of F.

Define K to be the set of all elements in f which are algebraic over F. By
Theorem 1, §8.1, K is a field. Furthermore it is algebraically closed because it
contains the zeros of any polynomialf(x) e K[x] as follows. The coefficients
of a given polynomial belong to a finite algebraic extension L/F contained in
K. The zeros of J(x) are algebraic over L and therefore over F. Thus they
belong to K, the totality of elements in that are algebraic over F.
To prove that the field K/F is unique to within isomorphisms over F,
we use Zorn's Lemma. But first, if F(a) is a simple extension and J(x) the
minimal polynomial of a. than an embedding). of Finto an algebraically closed
field C has at most n0 degf(x) distinct extensions to F(a) [see and
8.5]. (Once F(a) is mapped by an extension A into the other embeddings of
F(a) are obtained by mapping A(a) to the other zeros of 2(f(x)) e 2(F)
This result generalizes to arbitrary algebraic extensions K/F. Suppose
that 1 embeds F in an algebraically closed field Q which need not be algebraic
over F. To prove that there exists an extension A to K such that A(K)
let S be the set of all pairs (L/F, a) where L c K and a F 2, a an embedding
of L in This is a nonempty set for (F, ).) e S. Next introduce a partial
ordering in S as follows:
(L/F,a) (L'/F,o')
ifLc L' and a'IL=a.
This ordering is inductive. Let T be a nonempty, totally ordered
subset of S with typical elements (N/F, r) and (N'/F, t'). Now let L/F be the
union of all fields N occurring in T. To prove that ElF is a field, consider
a, h e E, where a e N, b e N' for some fields N, N'. Without loss of generality,
we may assume because T is totally ordered that N c N'; hence a, b e N' and
therefore ab, a+b, and a1 (for a 00) belong to N' and therefore toE.
Now to construct an embedding L -+ Q, define
c L, where (N/F, r) T. This definition of 8 is independent of the choice
of field N containing a, since if a e N' and N c N', then 8(a) = T'(a) = t(a)
because t' N = t. To verify the homomorphism properties
&(a+b) = 8(a) + 8(b),
8(ab) = &(a)ê(b)
fora e N, be N', assume Nc N' and then use the facts that a,b e N' and that
'r' is a homomorphism on N'. Therefore, since each t occurring in T is an
embedding we conclude that 8 is also. Thus (L, 8) is an upper bound of T.
Having verified the hypotheses of Zorn's Lemma, we apply it to
obtain the existence of a maximal element {M, A} in S. M c K. Necessarily
M = K, for if there were an element Be K\M, we could extend A to an em-
bedding A0 of M(B) into Q/F, thereby contradicting the maximality of(A'f, A).
Finally assume that K/F and (i/F are two algebraic closures of F.
By the preceding argument, each can he embedded in the other. Consequently
they are isomorphic over F.
 Algebraic Closure 303

REMARK. The separable elements of an algebraic closure K/F form a subfield

KS/F, called a separable algebraic closure of F [ef. §8.5]. Clearly all separable algebraic
closures of F are isomorphic over F.

Exercises

I. Prove that a finite field F cannot be algebraically closed. (Hint: Examine

f(x) = 1 + (x—a,), where F= (a1: I I m). Compare with §8.2.)
2. Suppose that the elements ..., in an extension field K/Fare algebraic
over F. Prove that the ideal
{f(x1, ...,xh) e F[x1, ...,xh] = 01.

is a maximal ideal of Fix1, ...,Xh].

3. Let F be a countable field, i.e., its elements are in a one-one set cor-
respondence with the set of positive integers N. Prove that the polynomial
ring F[x] has at most a countable number of monic irreducible poly-
nomials.
4. Let F= Z2(x,y). Prove thatf(t)= i'2+xt+y€ F[i] remains irreducible
in fl(x, y)[t], where is an algebraic closure of Z2.
5. Let S be a nonempty subset of a group G, and H a subgroup of G for which
H S= 0. Prove that the set of all subgroups K c G such that
K S 0 and that H K contains at least one maximal subgroup
of G.
 9

Selected Topics in Field

Theory

This chapter has a two-fold purpose: first, to develop further the

Galois theory in order to lay some foundation for algebraic number theory
and algebraic geometry 9.4 through 9.7, and 9.11 through 9.123; and
second, to provide the solutions to problems such as "doubling the cube" and
"trisecting an angle" that tantalized mathematicians from the classical Greek
period through the European Renaissance. General solutions for cubic and
quartic equations by Cardano, Ferrari, and Tartaglia in the sixteenth century
gave rise to the query whether equations of the fifth and higher degrees could
be solved by extracting radicals. These questions are treated in
through 9.3.
In §9.9 we present a proof, derived from the work of Lagrange and
Gauss, of the Fundamental Theorem of Algebra. which states that the field
of complex numbers is algebraically closed. This proof is prefaced
by a discussion of unique factorization domains, elementary symmetric
functions, and the concept of integral dependence. Together with the exercises
in 3.6, and 7.8, this provides a background for general ideal theory.
We include E. Witt's proof of one of Wedderburn's fundamental
theorems on finite dimensional algebras: that every finite division ring is
commutative.

304
§9.1 Cyclotomic Fields 305

§9.1 Cyclotomic Fields

Abelian extensions of the rational number field Q obtained by adjunc-
tion of roots of unity are called cyclotomic fields. Fields of this type, specifically
abelian extensions (and their subfIelds) over fields of finite degree over the
rational numbers Q, were significant in Kummer's attempts to prove Fermat's
Last Theorem and are important in the class field theory of algebraic number
fields. In this section we discuss the algebraic structure of cyclotomic fields,
for which the arguments are essentially arithmetic in nature. Analytic aspects
of the theory will be used in §9.10 concerning finite division rings.
Throughout this section we treat only extensions of the field Q of
rational numbers. We denote by a primitive nth root of unity for a fixed
integer n; that is, satisfies = I, but for I rn <n, 1. (Alternatively,
a primitive nth root of unity is a generator of the cyclic group of nth roots of
unity.) If is a primitive nth root of unity, then so is for each i, I i < n,
that is relatively prime toiz [cf. Proposition 2, §6.2]. For given n we define the
cyclotomic polynomial by

= (i.n) I

Theorem. The cyclotomic polynomial n 1, is irreducible in Q[x]

and has coefficients in Z.

Note that the degree of is the Euler q-function q(n), equal to

the number of generators of a cyclic group of order n [cf. Proposition 3,
§6.6]. Granting that is irreducible, we know then that it is the minimal
polynomial of over Q, and hence that Q] = q(n). The following
lemma is useful in the proof of the theorem.

Lemma. Foranyn I,
f—I =
din

where the product is taken over all positive divisors dofn.

Proof Becauseeach nth root of unity is a primitive dth root of unity for
one and only one divisor d of n, and because x" — I has distinct roots (i.e., is a
separable polynomial), each of then linear factors off—I in some splitting
field of f— I occurs precisely once as a factor of some cyclotomic
polynomial bd(x). Recall that n = [cf. Exercise ii,
Proof of the theorem. We show by induction on n that in Q[x] has
coefficients in Z. For n = 1 the statement is trivial, as D,(x) = x— 1. For the
induction hypothesis assume that d less than n; in
particular, that
= dinfl e Z[xJ.
d<n
 Selected Topics in Field Theory chapter 9

The lemma states that = f—i, and so must have integral

coefficients since f— 1 is primitive [cf. Theorem 1, §5.8]. By induction we
conclude that c Z[x] for all integers ii I.
To complete the proof of the theorem we demonstrate that is
the minimal polynomial of over Q and hence is irreducible. Let
f(x) Q [x] be the minimal polynomial of Then f(x) In order to j

prove conversely that consider a prime number p satisfying

(n, p) =1. Then is also a primitive nth root of unity. Let g(x) be its minimal
polynomial. To prove that = g(x), assume to the contrary *hatf(x)
g(x). Then (f(x),g(x)) = I since both polynomials are irreducible. Further,
f(x)((f—1) and

imply that f(x)g(x) I (x"— 1).

Hence f—I —fix)g(x)h(x) where f(x), g(x), and h(x), as divisors of a

primitive polynomial in Z[x], also lie in Z{x]. Since is a root of J(x) and
these polynomials have a common factor k(x) e Z[x]. Let
f(x) = k(x)f1(x) and g(x") = k(x)g1(x).
Because these polynomials have integral coefficients, we may reduce the
coefficients modulo the prime p, obtaining
= and g*(XP) =

in That is, we map Z{x] to Z,fx) homomorphically by taking

coefficients modulo p. with the image denotedf*(x), etc.
Since a (modp) for any a e Z by Fermat's Little Theorem,
§2.12,

(modp),
i.e., g*(xP) (g*(x))P if q*(x) is an irreducible factor of the common factor
k*(x) ofJ*(x),g*(xP), then implies that Thus
f*(x)g*(x) f (x"—
implies that q*(x)2 (x"—
Consequently in some algebraic extension of x" — would have a
multiple root. This is impossible, however, because the derivative of
(f—I) modp is 0 (modp), according to the choice of p [see §5.5].
Hencef(x) = g(x) and is a xero off(x).
This fact implies that all primitive nth roots of unity are zeros of
f(x). Consider c-',
where (j,n) = I, withj = •..p, (some of the prime factors
may coincide). 0 and
= ... = 0.
Therefore and consequently =1(x), the minimal polynomial
of 1.
§9.1 Cyclotomic Fields 307

Proposition. The galois group G(Q(C)/Q) is isomorphic to the multi-

plicative group of prime residue classes modulo n.
Proof. Since the roots off— I are then powers I i n, of a primitive
,zth root of unity it follows that Q(() is a galois extension of degree 43(n)
over Q [cf. §8.6]. For each o e G = and for a fixed primitive nth
root of unity
= 1;i(c), 0 <j(a) < n,
is also a primitive nth root of unity, and hence (j(a), n) = I; that is,
j(a) Now define a mapping 2: G by
1(a) =
We seek to show that 2 is an isomorphism. First,
=
=

from which we conclude that A is a homomorphism. For a kerl we have

j(a) = 1, which means that a is the identity auto-
morphism. Therefore 2 is injective.
To see that I is surjective, consider a prime residue m modulo n, and
define a e G by = cm. This mapping is an automorphisin since cm is a
zero of the cyclotomic polynomial and hence = (In more
detail, certainly and : Q] = Q]. since both have the :

same minimal polynomial Finally, 2(a) = in, which completes the

proof.

An immediate consequence of the proposition is that the galois

group of has order 43(n).

Example. Let n = 7. The field K = Q(O of seventh roots of unity has degree
= 6 over Q, and its galois group G is cyclic of order 6. A generator a of G is
given by a(C) = Then
a2(ç) = a4(C) = C2, a6(C) =
= a5(C) =

In particular, let be the primitive seventh root of unity expressed as a complex

number by
. . 2ir
C = cos— + I Stfl—.
7 7
Its complex conjugate is
2ir . . 2r
c—1 =cos——,sin—,
7 7

and
 _

308 Selected Topics in Field Theory chapter 9

The intermediate extension Q(A) is associated with the subgroup {l,a3} of order 2,
since
= ':6+': ':+C_I.
Therefore [Q(A) : Q] = [G : {l,a3}] = 3. Thus Q(A) is the maximal real subfield of
Q(C)/Q.
The (irreducible) defining equation of A overQ can be determined as follows.
Its roots are A, a(A), and a2(A). Since

a(A) = u(C+':6) = +
and a2(A) = C' + = C' +
the sum of the roots
A+a(A)+a2(A)= —I.
Furthermore
= —2

and A.a(A).o2(A) = I.
Consequently,

x3 + x2 — 2x — 1 = 0
is the minimal polynomial of the primitive element A of the extension Q(A)/Q.
SinceQ(A)/Q is a galois extension, the zeros c(A)and a2(A) are themselves
polynomials in 1, A, and A2 with rational coefficients. Specifically,
c(A) = C3+':2 = a+bA+cA2
= a+b(C+C')+c(C+':')2
= a + bC + + c':2 + 2c + cC2
(a+2c) + bC + c':2 + c':5 + b':6
Since {I,C, ...,':6} isa basis ofQ(C)/Q, we find c = 1, a= —2, and b = 0. Hence
ci(A) = —2 + A2.
Direct computation implies that c2(A) = I — A — A2.

Exercises

1.Find the cyclotomic polynomials overQ for n = 4, 6, 8, 10, and 12.

2. Prove that the cyclotomic polynomial = + + I, p a prime,
is irreducible over Q. (Hint: Let x = y+ and apply Eisenstein's1

Criterion to 1)/(x— I).)

3. Prove that t12m(x) 4)m(X) for odd integers m. (Hint: I
(x"— 1)((_X)m_
§9.1 Cyclotomic Fields

4. For a given integer ,z, let C1, ..., denote the nth roots of unity.
a. forn>l.
b. Prove 117=1C1 =(—ly'''.
c. For a given nth root, I, prove = 0.
5. Let C be a primitive nth root of unity. Find the galois group of
and determine the subfields of Q(C) and their respective galois groups
over Q for n = 6, 8, and 9.
6. Let ( be a primitive ninth root of unity. Find a primitive element for each
subfield K/Q of Q(C) and determine its minimal polynomial.
7. Find the subfield K/Q of degree 4 in the held of thirteenth roots of unity
and describe the effect of its galois group on a suitably selected
primitive element A of K.
8. Suppose that is a primitive 16th root of unity.
a. Determine the galois group of Q(C)/Q by its action on C.
b. Find all subfields of Q(C) and their associated groups.
9. Suppose that n1,n2 I; (n1,n2) = 1; and and 12 are primitive n1st
and n2nd roots of unity. Prove that
a.
b. = where q is a primitive nth root of unity,
fl = fl2.
10. Generalize Exercise 9 to the case where (n1, n2) I. What then can be
said about the fields Q(i1) Ci Q(12) and Q(11)Q(12) as cyclotomic
extensions? [Cf. Theorem 1, §8.8.]
11. Determine the following cyclotomic polynomials:
a. cb9(x)
in Q and in
an a primitive pth root of unity. Prove that
= 1) in Q(C)lxI.
13. Let C be a primitive 24th root of unity. Find [Zs(C): Z5] and determine
the monic polynomial f(x) e Z5 [x] which has C for a zero.
14. Suppose that is a primitive pth root of unity, p an odd prime.
+ 1)12 =
a. Prove that — — A is a primitive (2p)th root of unity.
b. Prove that
15. Prove that A, as given in the example, is a unit in Z[A].
16. Consider a primitive nth root of unity ii 4. Prove that
:Q] = and that R. Determine explicitly
the automorphism group ofQ(C)/Q(C+C ').
17. For n> 5 and a = sin(2zr/n) prove that:
a. [Q(a) :Q] = if(n,8) 2.
b. [Q(a) :Q] = *p(n)/4 if (n,8) = 4.
c. [Q(a) : Q] = ço(n)/2 if (n,8) = 8.
18. Find a subfield K/Q of Q(C) for some nih root of unity whose gatois
group G(K/Q) is:
a. The direct product of two cyclic groups of order 2.
b. The cyclic group of order 4.
c. The direct product of two cyclic groups of order 4.
310 Selected Topics in Field Theory chapter 9

19. Find a primitive element and its minimal polynomial for the extensions
in parts (a) and (b) of Exercise 18.
20. Prove that the derivative of g(x) = V — 1)/(x — 1) Q [x) satisfies
= n/1((— 1) for every nth root of unity C I.
21. For an nth root of unity C different from I prove that I "I(C— I).

§9.2 Equations Solvable by Radicals

ln"olden days" students learned to solve cubic and quartic equations
by means of extracting cube roots and to trisect angles by using tables of
values of trigonometric functions. Thus they were familiar with the results of
the Italian mathematicians Geronimo Cardano (1501—1576), Ludovico Ferrari
(1522—1565), and Nicolo Tartaglia (c. 1499—1557) on solutions of equations of
the form x3 +px = q, p, q e Z, and with the reduction of a quartic to a cubic
equation. For further discussion of such solutions, the reader might refer, for
example, to B. L. van der Waerden, Algebra, Vol. 1.
Toward the end of the Renaissance speculation arose as to whether
there were general methods uniformly applicable to the solution of equations
of degree n 5 with rational coefficients. Would the extraction of successive
nith, nth roots, etc., yield solutions of every polynomial equation? No com-
plete answer to this question was found until the beginning of the nineteenth
century. Niels Abel (at the age of 19) succeeded in completing an earlier
attempt (1799) of Paolo Ruffini (1765—1822) when he proved that "in general"
the quintic equation was not solvable by extracting and forming rational
combinations of a finite number of nth roots (n is variable). His work is
matched by that of Evariste Galois who reduced the solvability problem to
one of group theory. Galois' work is related to some earlier results (c. 1770)
of Lagrange on permutations of roots of polynomial functions. it is inter-
esting to note that both Abel and Galois had their difficulties with what
might be called the "scientific establishment" of their day, which scarcely
designed to consider their work until some decades later.
For historical reasons and for the sake of simplicity we assume
throughout the discussion below that the underlying base field F has
characteristic 0.
A polynomial J(x) E F[x] is said to be solvable by radicals if there
exists a tower of algebraic extensions
(*)
such that:
(i) E contains the splitting field K off(x)
(ii) E1=E1.1(Aj, where A," =a,€E1_1, I
in particular, some (or all) of the elements may be 1; that is, some of the
extensions EJ/E1_ may be cyclotomic. Further, note that the splitting field K
off(x) need not be a step in any tower (*) with property (ii).
§9.2 Equations Solvable by Radicals 311

For example, the cyclic cubic equation

+ x2 — 2x — I=0
[cf. §9.1] has for a root, a primitive seventh root of unity. For no
e N is an element ofQ; i.e., 'is not a pure radical.

For the special case r = I, n1 = n> 1, we note the following lemma.

Lemma 1. If K = F(A), where = a 6 F. and F contains all the iith roots

of unity, then K/F is a cyclic extension whose degree divides n.
Proof. The minimal polynomial g(x) e F[x] of A divides f—a, and its
zeros occur among then distinct zeros 0 i < ,z, off—a, where is
some fixed primitive nth root of unity. Clearly the splitting field of g(x) is K.
Let a,r, ... be elements of the galois group G of K/F. Then o(A) =
for some i(a) 6 N (counted modulo n). Since e F,
(to)(A) = r[c(A)] =
= =
= (ar)(A).
Further = because
A = =
implies that i(a ')_ — i(a) (modn). Hence G is isomorphic to a subgroup of
the additive group of .Therefore G is cyclic and o(G) and [K: F] divide n.
In the following theorem it will be useful to have the equivalent
definition of solvability given in the following lemma.

Lemma 2. A polynomial f(x) e F[x] is solvable by radicals if and only if

there exists a tower of algebraic extensions (*) such that:
(i) E contains the splitting K off(x)
(ii) E4 = (A1), where = a1 e E1_ I I r.
(iii) E is a galois extension of F
(iv) E1 contains all the Nth roots of unity for N = n1.

Certainly these conditions imply those of the preceding definition.

To prove the converse, suppose thatf(x) is solvable by radicals. The definition
assures the existence of a tower of fields (*) with properties(i) and (ii). To each
step of the tower (*) adjoin a primitive Nth root of unity N= n..
Thus
F = E0 g ç ... =
and = A,). This tower has properties (i) and (ii) required in the
definition. For notational simplicity relabel the intermediate extensions to
omit any trivial ones. Use E to denote E contains all
Nth roots of unity (if they are not already in E0). Thus
(**)
312 Selected Topics In Field Theory chapter 9

It remains only to prove that the tower can be chosen so that E'/F
is a galois extension. To this end we use an inductive argument in modifying
the tower (**). First, observe that is a galois extension of F. If F, then
E is obtained by adjoining to F all roots of 1 E F[x]. If F, then
E= E0(A1) and E contains a//roots of x"—a1 e F[x].
For the general inductive step, suppose that the tower, 2 u s,
F=E0cEc....c:E,.1
has been modified so that the field is contained in a tower

of successive radical extensions such that L1 contains all Nth roots of unity,
the extensions L4/L1_1 are cyclic of degree dividing N, 2 i t, and the field
is a galois extension of F.
Recall that = where An"" c Denote by
the (not necessarily distinct) conjugates of =
All these conjugates
I jrn = [L1: F].
since L1/F is a galois extension. The
belong to
polynomial
g(x) =
is invariant under the galois group of Li/F, and hence g(x) e F[x]. Since we
consider now only fields F of characteristic 0, the extension L,/F is separable,
and hence L, F(B) for some primitive element B. Let h(x) be the minimal
polynomial of B in F[x], and the splitting field of g(x)h(x).
Since B€ Lt+m, the splitting field L. Since is a factor
of g(x), it splits in Lt+m, and thus Lr+m. Let denote a zero of
E 1j m, where = We have then a tower of
radical extensions

Au, 2) (An. ..., Au,m)

= Lr+m.
The degree of where = divides n,, and
hence N.
Finally we conclude by induction that the tower (**) can be modified
for all u so that is contained in a galois extension of F, and so that each
step in the tower, after the first, is a cyclic (radical) extension [see Lemma 1].
The first step involving adjunction of Nth roots of unity is abelian, but not
necessarily cyclic. In particular, E is contained in a galois extension L/F.
From the tower (*) we have obtained one with properties (i) through (iv).
We now turn to the crucial theorem concerning solvability.

Theorem. A polynomial f(x) e F[x) is solvable by radicals if and only if

the galois group of its splitting field K/F is solvable.
§9.2 Equations Solvable by Radicals 313

Proof Suppose that f(x) is solvable. By Lemma 2 there exists a tower of

fields

for which L/LI_ is cyclic, 2 I u, and L1/L0 is abelian. The galois group
G(L/F) has then a composition series with cyclic factors; i.e., it is solvable.
Consequently, since K c L the group G(K/F) is solvable, as it is the quotient
group of a solvable group [see and 8.7].
Conversely, suppose that the galois group G of the splitting field K
of the polynomialfix) F[x] is solvable. Then we have a composition series

with cyclic factors. Consider a primitive Nth toot of unity (, where

N= n. and n1 = [G1_ : G.]. Let K. be the intermediate field associated
with the intermediate group G.. Since G._I/G1 is cyclic, so is
by Theorem I, §8.8. That is, is obtained by adjoining an n1th
root of some element a, e Thus, we have a tower
Fc K1(C)c c =
of radical field extensions. By definition then. f(x) is solvable by radicals, as
stated.

The term "solvable group," as in §7.6, is used to describe groups

arising as galois groups of splitting fields of polynomial equations solvable
by radicals. Polynomial equations of degree less than five are always solvable
by radicals. By the remark following Theorem 2. §8.6. the galois group of a
polynomialflx) of degree n is isomorphic to a subgroup (proper or not) of the
symmetric group E,,. Since the permutation groups E2, Z3, and are solvable
[see §7.6], the preceding theorem implies that equations of degree at most 4
are solvable by radicals.
We should note that nonsolvability depends upon the particular
ground field considered; any polynomial equation over a finite field is solvable,
since any finite extension is cyclic. In special cases (cited below) the galois
group of polynomials of degree n is the symmetric group Since the groups
n 5, are not solvable such polynomials of degree n 5 are then
not solvable by radicals.

For example let F = Q. Oskar Perron proved that for given n the galois
group of the polynomial
n—I jn—I \
f(x) x fl (x—p1
i=I
÷( I]
\i=1 1

is the symmetric group where k1 e Z, I I n; where the p,, I n, are

primes such that the v integers 1< v n—2. p1k1,p1 p2k2,...,p1 p2 ...p,.k, are
relatively prime to and are not congruent to each other modulo p, and that
k., 0 (mod l pj) [see 0. Perron, Algebra, Vol. II, Theorem 92}.
314 Selected Topics in Field Theory chapter 9

For each degree n S there exist infinitely many polynomials which

are not solvable by radicals. The above statement provides an explicit con-
struction. Quite different is an existence proof based on a theorem of Hilbert.
if
f(x1,t1, = + + ... +
where are algebraically independent indeterminates over Q, then
there exist infinitely many specializations —+ a1 E Q (that is, homomorphisms
of Q[i1, to Q) such that f(x,a1 Q[x] is irreducible. A more
arithmetic proof depending upon the theory of algebraic functions of one
variable is due to Herman Ludwig Schmid (1908—1956).

Exercise

I. Considerf(x) = x3—7x+7 eQ[x], and set

A and B=
Verify that
at =
a2 =

and a3 =
where a,2 + (0 + I = 0,are the zeros of 1(x) in C. Prove that these three
zeros are real. (This is an example of the casus irreducibiis of Cardano.
The name is derived from the fact that the real roots cannot be expressed
in terms of real radicals.)

§9.3 Constructions with Ruler and Compass

This section provides the solution to three classical problems of
euclidean geometry. These solutions were obtained in the nineteenth century,
more than two millenia after the problems were debated by the geometers of
Plato's Academy. The central question is which coordinates in the cartesian
plane can be constructed with the euclidean tools, ruler and compass. With a
ruler we can draw a line between any two given points, and with a compass
we can draw a circle with any given point as center and circumference passing
through any second point. We can also use a compass to copy an already
determined line segment onto another line.
We assume as given two perpendicular axes in a plane and a line
segment of unit length. Then using a compass to lay off additional line segments
§9.3 Constructions with Ruler and Compass 315

of unit length and to construct perpendiculars, we can locate all points in the
plane with rational components.

Geometric Lemma. If line segments of nonzero lengths a and h are given,

then elementary ruler and compass constructions yield line segments of
lengths a+b, a—h, ab, I/a, a/b, and

For example, given a line segment PQ of length a, we can obtain a

line segment of length I/a by constructing a perpendicular at P (see Figure
9.1). On this perpendicular lay off a line segment PR of unit length, and at R
construct a perpendicular to RQ to intersect at S the line PQ extended.
Noting that the triangles RSP and PQR are similar, we have the proportion-
ality SP/RP = RP/PQ, or length SP = 1/a.

/ // -'S....
'S'S.-.
-S
/ '.5.

//
S..
'.5

/
/
//
x = f/a -l a
S P Q
Figure 9.1

To construct a line segment of length we lay off on the extension

to the line segment PQ a segment UP of unit length as in Figure 9.2. Now
bisect UQ at 0 and draw the semicircle centered at 0 with UQ as diameter. At
P construct a perpendicular to PQ, to meet the semicircle at V. The line
segments UVand VQ are perpendicular, as the angle between them is inscribed
in a semicircle. Thus the triangles PUV and PVQ are similar, from which we
conclude that PU/P V = PV/PQ or, since the lengths PU = I and PQ a,
that the length (P1')2 = a and finally P1' =

Figure 9.2
316 Selected Topics In Field Theory chapter 9

Complete verification of the geometric lemma is left as an exercise.

Once a cartesian coordinate system has been established in the plane, these
facts and the geometric problems of finding finitely often
(a) the intersection of two lines, and
(b) the intersection of lines and circles, or the intersection of two
circles
can be translated via analytic geometry into the algebraic problems of finding
(a*) the solution of a pair of linear equations, and
(b*) the solution of a linear and a quadratic equation, or of two
quadratic equations.
Conversely, the algebraic solutions of problems like (a*) and (b*) can be
constructed geometrically by ruler and compass, if the coefficients of the
equations are constructible.

Corollary 1. The set of real numbers (lengths of line segments) constructible

by ruler and compass is a subfield of R.

Corollary 2. If every element of a field K c R is constructible by ruler and

compass, then so is ever)' element in the extension for all a e K.

Corollary 3. If every element of a field K R is constructible by ruler and

compass, then so is every element in a galois extension L/Kwhere [L: K] = 2".

The proof of Corollary 3 follows from the fact that L = L, contains

a chain of subfields I i i, such that [L4: L,.. = 2 and L0 = K,
since the galois group G(L/K) has a composition series whose factors have
order 2 [see §7.6, especially Exercise 11, and §8.7]. Now L. =
where B12 = b,..., e L._ and by Corollary 2 and an induction argument
B, can be constructed geometrically. Hence so can L.. (In case is a non-
real complex number, then B. has constructible coordinates in the complex
plane; see Exercise 2, §9.9.)
These observations can be restated as follows.

Proposition 1. If the elements of K c C are constructible by ruler and com-

pass, and if L/K is a galois extension of degree 2"', then L = K(A), where A
can be obtained by (a finite number of) ruler and compass constructions.

REMARK. The same result applies to nongalois extensions K/F given by a finite
tower of stepwise quadratic subfields:
F K0 C K, c ... c K,_ C K, C c K,, [K, : K,...,] = 2.

Conversely, suppose that a e L/K is obtained by a finite sequence of

ruler and compass constructions, where L/K is a finite algebraic extension and
every element of K c C can be constructed with ruler and compass. That is,
a is obtained from the elements of K by solution of a succession of linear and
§9.3 Constructions with Ruler and Compass 317

quadratic equations which give rise to a tower of field extensions of degrees I

and 2, respectively. Hence the extension K(a)/K has degree 2", for some
in e N. Thus, we have the following converse to Proposition 1.

Proposition 2. If the elements of K c N can be constructed with ruler and

compass and if a e L/K is obtained from K by a succession of ruler and com-
pass constructions, then [K(a): K] = 2"', for some ?fl e N.

Behind the preceding general statements are three long-standing

geometric problems. the first two of which date from the classical Greek era,
Athens in the age of Pericles.
1. Doubling of the Cube (Delian Problem). Given a cube whose edges
have length one, can a ruler and compass construction determine
the edge x of a cube that has volume 2?
2. Trisection of an Angle. Given an arbitrary angle can a ruler and
compass construction determine cLI3?
3. Construction of Polygons. Which regular polygons of n sides can
be constructed by ruler and
The answers to these problems, given via Galois theory and
Proposition 2 in the nineteenth century, are determined as follows.
Problem 1. Clearly the length x of the edge must satisfy the equation
x3 = 2. Since x3—2 E Q{x] is irreducible, the real number has degree 3
over Q, and so by Proposition 2 is not constructible over Q by ruler and
compass. Consequently the cube cannot be doubled by euclidean construc-
tions. (The oracle of Apollo at Delos is reported (c. 429 Bc) to have required
that a cubical altar be doubled as a condition for ending a plague. However,
the plague which Phoebus Apollo sent to punish the Athenians finally abated.)
Problem 2. Given the angle j3 = the problem is to determine the angle
For the solution we consider the trigonometric identity 4 cos3 —3 cos =
cos It suffices to determine whether cosc can be found by a ruler and
compass construction. Let fJ = 600; then cosfi = 4, and is a zero of the
polynomial p(x) = e Q[x]. Because p(x) has no root in Q, it is
irreducible over Q. Therefore, cos has degree 3 over Q, and so by Proposition
2 is not constructible over Q by ruler and compass. In other words. "general"
angles cannot be trisected.
Obviously this statement does not exclude trisection of certain angles,
e.g., = 900. Note that it was not proved that there exist infinitely
many angles which cannot be trisected. (Such a result is true, but requires the
proof that 4x3—3x—u€ F[u][x], u an indeterminate over F= Q(i1,
1.E R, is irreducible and possesses infinitely many consistent specializations
u-÷veFsuch that 4x3—3x—v is irreducible in F[x].)
Problem 3. By similarity in the euclidean plane it suffices to examine regular
polygons of n = sides, inscribed in a circle of radius I, where the
318 Selected Topics in Field Theory chapter 9

p are distinct odd primes. Using the formula of Abraham de Moivre (1667—
1754),
= cos(nçs) + isin(np),
we examine the primitive nth root of unity cos(2ir/n)+isin(2ir/n), and
the galois extension Q

[Q(C) : Q] = q,(n) =

By Proposition 2, we must have ço(n) = m 1. Since p 2,

each must equal 1, and furthermore each p1 must have the form
cc,

p, = + 2k1,
1 i < r. These primes 1 + 2&1 must further be Fermat primes—
that is, they must have the form I + 22"—named for the pioneer of modern
number theory, Pierre de Fermat [cf. Exercises 5 and 6, §2.12]. Consequently,
a regular polygon of n sides can be constructed by ruler and compass if and
only if n = 28 where p, = I + Whether there are Fermat primes
other than 3, 5, 17, 257, and 65,537 (22'n + I for 0 in 4) is an open question.
The integers 22m+ 1 for 5 16 are not Fermat primes. For example,
in 1732 Euler showed that 225+1 641 .6,7Q0,4l7.
The condition q(n) = 2" for the constructibility of regular polygons
of n sides implies that in the ruler and compass sense used here, it is impossible
to construct, for example, the regular polygons of 7 and 9 sides.
Gauss actually constructed (c. 1796) the regular polygon of 17 sides.
(For the algebraic interpretation of his solution see Exercise 2.) In this con-
nection it should be noted that he was in possession of very considerable
parts of the structure of the cyclotomic fields Q(C) [see §9.1] and especially
their connection with the quadratic law of reciprocity. These results lead
ultimately to Kronecker's theorem that every quadratic field and
more generally every finite abelian extension of Q, is a subfield of some field
for which the complex function ez with period 2,ri is a "generating"
function. The foundation of some of the most exciting research of today—the
arithmetic of abelian functions, diophantine equations, rationality of certain
t functions, and the like—were laid, we venture to say, by these early
investigations.

Exercises

1. For which of the following values of n can a regular n-gon be constructed

with ruler and compass?
a. n=13 b. n=15
c. n=17 d. n=24
e. n=37 f. n=36.
§9.4 Trace and Norm 319

2. a. Find a composition series of the galois group of K = where is

a primitive seventeenth root of unity.
b. Determine the corresponding chain of subfields of K/Q.
3. Repeat Exercise 2, where is a primitive fifteenth root of unity.
4. Given line segments of lengths a and b, find a line segment of length x by
geometric construction such that:
a. x a = a: 1, or equivalently, x = a2.
b. x : a2 = a :1, or equivalently, x = a3.
c. x : a b : I, or equivalently, x = ab.
Thus verify that if 11, ..., 1, are constructible real numbers, then every
element in the field Q(tt 1,) is geometrically constructible.
5. With the hypotheses of Proposition 2, prove that K(a) is contained in a
galois extension 1.', such that EL': K] is a power of 2 [Cf. Lemma 2 and
the theorem of §9.2].
6. Determine whether or not a root (in C) of the following polynomials can
be constructed by ruler and compass.
a. x4 + 3x2 + 25 b. x4 + 5x2 + 5
c. x4—x2+3.
7. Use the result due to Perron at the end of §9.2 to exhibit a quartic poly-
nomial in Q[x], none of whose roots can be constructed by ruler and
compass.
8. Find the galois group of the polynomial x4 —7 aQ [x]. Prove that a
primitive element of its splitting field is constructible by ruler and compass.

§9.4 Trace and Norm

This section is a prelude to the proofs of Hilbert's Theorem 90 and
Noether's Equations in §9.6 and to the discussion in §9.7 of special abelian
extensions of fields. That discussion relates, in turn, to the classical problem
of the solution of equations by radicals, presented in §9.2.
Here we consider two mappings of a finite algebraic extension K/F
into the field F. The first is the trace, a linear functional defined on K with
values in F. The second is the norm, a homomorphism of the multiplicative
group of K into that of F. These functions can be defined on the one hand by
considering K as a (finite dimensional) vector space over F, and on the other
hand by utilizing the Galois theory. For the sake of simplicity of argument in
the proofs and for ease of extending the results to include inseparable exten-
sions, we shall use both definitions interchangeably.
Let K be an extension of degree ii of the field F. For fixed a e K, the
mapping
K K given by 4'afr) = QC, c e K
is a linear transformation of K considered as a vector space over F. For
c1,c2 in Kand dE F,
= afr1+c2) = ac1 +ac2 = +
and = = d(ac) =
320 Selected Topics in Field Theory chapter 9

In fact the set of mappings {q'a : an K} is a subring of the ring of endo-

morphisms EndF(K) of K/F {cf. §4.3]. We embed K in EndF(K) by
definining i: EndF(K) by :(a) = q(a). For all ce K,
= (a+b)c = ac + be = + pb(C)

and = (ah)c = a(bc) = = =

and therefore = (Pa+(Pb, = which states that i is a ring
homomorphism. The injectivity oft follows immediately by considering a
such that = 0. Then (,oa(C) = ac = 0 for all e K, and in particular for
I. Hence a= a.1 = 0, and kers= (Q}.
A matrix representation Ma = [au] of the linear mapping is
obtained, as in §4.4, by selecting a basis (a1, of K/F, and writing
= The characteristic polynomial of the linear transfor-
mation (or equivalently of the matrix M4 determined by was defined
in §5.9 to be
Xa(X) =

= a1j)x"_' + ... + (—
—

where 1,, denotes the n x n identity matrix. While the matrix Ma corresponding
to a linear transformation ç°a depends upon the choice of a basis for K/F, the
characteristic polynomial x0(x) is independent of this choice. Thus, to each
element a e K/F is associated its unique characteristic polynomial Xa(X) of
degree n = [K: F].
We define the trace of an K with respect to F, denoted T(a), or more
precisely TK,F(a), to be the coefficient of in Equivalently, T(a)
is the trace of the matrix Ma, as defined in §5.9, or the sum of the roots of
Xa(4. Similarly, the norm of a K with respect to F, denoted N(a), or more
precisely NK,F(a), is the determinant of the matrix Ma. It is (— times the
constant term Of Xa(X), or the product of the roots of x0(x). Being defined in
terms of the characteristic polynomial, T(a) and N(a) are independent of the
choice of basis for K/F.
The Cayley-Hamilton Theorem of §5.9 implies the following
theorem, since i is an embedding of K in EndF(K).

Theorem. An element a E K is a zero of its characteristic polynomial Xa(X)

over F.

Continuing to consider K/F. an extension of degree n with basis

(a1, suppose now that L/K is an extension of degree nt with basis
{b1, ...,bm) over K. The elements akhJ form a basis of L/F [see §8.1].
Ordering the elements of this basis as follows
§9.4 Trace and Norm 321

we determine for the endomorphism of L/F(cpa is multiplication by a K)

the representing matlix

S(a)=

where the matrix A = determined by relative to the basis

of K/F. appears rn times on the main diagonal.
Consequently the characteristic polynomial of a (oi of the matrix
S(a)) with respect to the basis {a, h1, ...,anbm} of L/Fis the rnth power of the
characteristic polynomial of a (or of the matrix A) with respect to the basis
(a1, .. ., of K/F.
This discussion yields three propositions.

Proposition 1. Consider extensions L/K and K/F ot' respective degrees m and
n. Then, for a E K,
TL/f(a) = rnTK,F(a) and NL,F(a) = [NK,p(a)]m.

ProposItion 2. For a tower of fields KF(a) F, with m = [K: F(a)]

and [K: F] = the characteristic polynomial Xa(X) of a in K/F is the nith
n,
power of the characteristic polynomial Pa(X) of a in the extension F(a)/F.
Proposition 3. The minimal polynomial ma(x) over F of a e K/F equals the
characteristic polynomial of a in F(a)/F.
Proof. Since the characteristic polynomial p0(x) of a in the extension
F(a)/F has a for a root, rna(X)Ipa(X). These polynomials are monic of the
same degree [F(a): F], and so they coincide.
As a consequence of this discussion, for a e K/F each root of
is conjugate to a because is the power of the irreducible characteristic
polynomial ii0(x) of a F(a)/F. In fact, we have the following alternate
definition of the trace and norm. For a K/F, T(a) and N(a) are the sum and
product, respectively, of the conjugates of a, taken with appropriate multi-
pliciiv, in a splitting field of;s0(x).
Proposition 4 states significant properties of the trace and norm,
derived from their definition in terms of a matrix representing the element
a e K [cf. Proposition 4, §5.9]. Two additional properties, significant for work
in arithmetic field theory, are stated for not necessarily separable extensions
in Proposition 5.

Proposition 4. For elements a, b in an extension K/F of degree n:

(i) T(a+b) = T(a) + T(b)
(ii) N(ab) = N(a).N(b)
(iii) T(ca) = cT(a) for c e F
322 Selected Topics in Field Theory chapter 9

(iv) T(c) = 'ic for cc F

(v) N(ca) = ceF
(vi) N(c) =

Proposition 5. Let L K F be pairwise finite extensions. Then, for a e L,

(i) TL,F(a) = Tr,r[TLI,Ja)]
(ii) NL/F(a) = NK,F[NL,K(a)].

One approach to the proof, which necessitates a theorem on deter-

minants using triangular representations, involves representations of elements
in L relative to bases of L/K and K/F. [See, for example, N. Jacobson,
Lectures in Abstract Algebra, Vol. Ill; Theory of Fields and Galois Theory, pp.
66—70.] We prefer a proof based on properties and enumerations of conjuga-
tions (i.e., equivalences of L/F in a sufficiently large field fl/F, see §8.4).
Proof First, we recall some facts related to the fundamental lemma on
prolongations. Denote by K5 the maximal separable subfield of K/F, by L5
the maximal separable subfield of L/K, and by the maximal separable
subfield of L/F. Schematically,

K5

The degrees of these extensions satisfy

F] = [L3: K][K3 : F] or [L: F]3 = [L: K]5[K: F]5
and [L: = [L: L5][K: K3] or [L: F]1 = [L: K]1[K:F]1
as will be shown by enumerating the conjugates of L/K, L/K, and K/F in a
splitting field fl/F of some polynomial g(x)€ F[x]. Let K= F(a1, ...,a3)
and L = ..., a,), and let g.(x) be the minimal polynomial in F[x] of
a,, I <i t. Then setg(x) = g.(x) [see and 8.5].
There are exactly [L: K]5 distinct isomorphisms of L/K in fl/F;
there are also exactly [K: F]5 distinct isomorphisms of K/F in fl/F.
Furthermore, each isomorphism has exactly [L: K]3 distinct prolongations
to an isomorphism of L/F [See the theorem in §8.5]. For each i,
§9.4 Trace and Norm 323

I I [L: K]5, we have the prolongation If is any other prolonga-

tion of to L/F, then the restriction of to K is the identity map on K.
Hence = a1, for some i, and f1 — Thus there are precisely
[L: KJ5[K: F]5 distinct isomorphisms [K: F]5 and1j
I i [L: K]5. Note that this is the number of distinct isomorphisms of
LJF (namely, [L : F]5). (An isomorphism p of L/F induces in K/F by
restriction a unique isomorphisni consequently p = with a unique
determination of after a fixed extension of to L/K is taken.) Thus
[L: F]5 = [L: K]5[K: F]5. Finally,
[L:F] [L:KJ[K:F]
= [L: KJ5[L: K]1[K: F]5[K: F]1
= [L: F]5[L: F]1;
and so [L : F]1 = [L : KJ,[K: F]1.
These results will now be applied to the tower of fields
K 2 K5 2 F(a) 2 F(a)5 2 F.
where K5 is the maximal separable subfield of K/F(a). and F(a)5 is the maximal
separable subfield of F(a)/F. Let
= F] = [F(a): F]5,
u= pe
[F(a) : F(a)5] = [F(a) : F]1,
N0 = [K5: F(a)] = [K: F(a)]5,
= = [K: Kj = [K: E(a)]4.
Each of the n0 distinct isomorphisms of F(a)/F extends one of the n0
isomorphisms r7 of F(a)5/F. Each of the N0 distinct isomorphisms a, of
K/F(a) extends one of the N0 isomorphisms of K5/F(a). Also there are
precisely n0 N0 isomorphisms of K/F given by a.. where is chosen as
above as a fixed prolongation oft1 to K.
According to the discussion
of a e L is
(x_rj(a))M.

Next, by Propositions 2 and 3, the characteristic polynomial is

= Xa(X)
where m = [K: F(a)]. Consequently, since a,(a) = a.

Xa(X) =

= fl

(x — (a))",
324 Selected Topics in Field Theory chapter 9

where the mappings ph, 1 Ii are the = [K: F]5 distinct iso-
morphisms of K/F, and uv = = [K: F]1. Therefore, by Proposition 1,

TK,f(a) = uv Ph(C)
h I

NK,F(a)
=
The statement of Proposition S follows by noting that the [K: F]5
mappings Ph of L/Fare given as tJak, where the ak are the [L: K]5 distinct
isomorphisms of L/K and the are selected prolongations of the [K: F]5
distinct isomorphisms of K/F. Let i = [L : F]1. For a c L we have
T,.,K(a) = [L: ak(a);

hence TX/F(TL,K(a)) = [K: F],.[L :

=
j. k
= = TL,F(a).
h

Similarly,

consequently, = [n
= [fl
U.k J

ri
— 11 —
h

Exercises

1. If p>O, prove that a polynomial F[x} either is

irreducible or factors completely in F{x]. (Hint: Observe that (x + [I])" =
[1] c Z1,, and examine the trace of potential irreducible factors
of
2. let f—ac F[xj, where ihe prime p is distinct from char F. Prove that
a is either irreducible in F[x] or that it factors completely in F[x].
(Hint: Replace trace by norm in Exercise I.)
3. Suppose that F Q(i) where i2 = — 1. Find the trace of o+ a in Q and
the norm of co+a over Q(co), where a e Fand -i-w+ 0. 1
§9.5 Tbeorem of the Normal Basis 325

4. Prove Proposition 1.
5. Let K = Q w) where w2 + w +1 =0. Find:
a. TK/Q(V5+w) b. NK,Q(w).
6. Verify the formulas in Proposition 4.
7. Let A E p a positive prime. Prove that N(A) = 1 if and only
11.4 is a unit of the ring
8. Suppose that A is a zero of the irreducible polynomial 1(x) x3 —2 e
Q[x]. Let B=(l—AY1 inK=Q(A).
a. Find the minimal polynomial in Q[x] of B.
b. What are the norm and trace of B?
c. What are the norm and trace of B considered as an clement of the
splitting field of f(x)?
9. Let K = Q(JZ Find NK,Q( and Tx,Q(
10. Let be a primitive pth root of unity, p a prime. Denote by T and N the
trace and norm of over Q. Prove:
a. <h<p.
b.

11. Let F= Z3. Prove that a zero A of x3+[2]3x+[l]3 is a primitive 28th

root of unity over F. Find:
a. The conjugates of A as polynomials in A with coefficients in F.
b. An element Be F(A) whose norm with respect to F is [l]3.
c. An element Ce F(A) such that B = C/c(C), where is the
Frobenius automorphism of F(A)/F Isee
12. Let K Q(O, where C is a primitive fifth root of unity, and denote by ,
a generating automorphism of G(K/Q).
a. Find an element A e K whose norm with respect to Q is 1, and
determine all B e K such that B/a(B) = A.
b. Find a nonzero element C e K whose trace with respect to Q is zero,
and determine all Dc K such that D— a(D) = C.

§9.5 Theorem of the Normal Basis

This section is devoted to a single theorem significant in the solution
of arithmetic problems in algebraic number theory. Much of the argument
involves congruences and is reminiscent of the proof of the Chinese Remainder
Theorem in §2.11. The Theorem of the Normal Basis asserts that galois exten-
sions have bases whose elements are conjugates of a single element. For the
sake of simplicity and in order to avoid, for example, the representation theory
of linear algebras, we shall assume that our ground field has infinitely many
elements.

Theorem. Let K = F(a) he a galois extension of degree n over an infinite

field F with the galois group G(K/F) = {a1, There exists an element b
in F(a) such that the n conjugates i1(b) of b form a vector space
basis of K over F.
326 Selected Topics in Field Theory chapter 9

Such a basis is called a normal basis because it consists of conjugates

of a single element.
The first step in the proof is to express the unit element as the sum of
orthogonal idempotents in a residue class ring of K[xJ [cf. and 5.3].
First, extend the automorphisms E G from K to K[x] by defining
x. Then let a, = c1(a), where is the identity map, and denote by
f(x) the minimal (over F) polynomial of a. Then in K, f(x) = (x—a1).
By a modification of the Lagrange Interpolation Formula we obtain, as in
§5.6, the equivalent expression for any polynomial /,(x) of degree in < n in
K[x]:
h(x)=
x—a
Note that for 1 i n,
c1(f(x)) = f(x) and =
In particular. for h(x) = I,
f(x) —
f(x)
— —

Settingg(x) =f(x)/[f'(a)(x—a)] and a1(g(x)) E K[x], we obtain

Further since for i

g.(x) = = a.(g(x)) for I j
As in §2.11, the polynomials g.(x) are orthogonal idempotents modulof(x);
that is,
(modJ(x)).
Since = for some k and for fixedj, the trace function
satisfies

= = = 1.

Let D(g1 denote the matrix with components in K[x] and

its transpose. Then
=
= [cç,J, where; =
= [T(g1g,)].
Consequently,
= (detD(g1 I (modf(x))
 Theorem of the Normal Basis 327

because (mod 1(x)) and

= T(01)]
det I (mod 1(x)).
Therefore the discriminant is a nonzero polynomial in F[xJ. Assuming
now that F is an infinite field, we can find an element c e F for which
0 by Theorem 2, §5.2.
Finally the elements
b g(c) = a1(b) and = g1(c), 2 I n,
are linearly independent, because if there were a dependence relation
(b) + + =0
with coefficients x1, in F, then
x1cr4(a1(b))+ =0
for I i ii. The determinant of this homogeneous system of n linear
equations in x equals A(c), different from zero. Consequently the
system of linear equations has only the trivial solution x1 = = =0
[see §4.5, following Property 9]. Thus the n conjugates < I n, of b 1

are linearly independent and so constitute a normal basis of K/F.

Exercises

I. Determine a normal basis for each of the following extensions of Q.

a. b.
c. Q(w) where + w+I 0 d.
e.
2. Suppose that K1 and K2 are dIstinct quadratic extensions of the field F,
2. Assume that the elements A1,A2 of K1,K2, respectively,
generate normal bases over F.
a. Do the conjugates of A1 A2 determine a normal basis of K1 K2 over F?
b. Generalize this fact to the case distinct quadratic extensions.
3. Let K = Q (A) be the cyclic cubic extension where A = C + C6 with a
primitive seventh root of unity Then A3+A2—2A—l = 0.
a. Show that the mapping a defined by a(A) = —2+A2 belongs to the
galois group G(K/Q).
Show that a2(A) = C4+C3.
b.
c. Find N(A).
d. Find all elements Be K for which B/a(B) = A.
4. Verify the following steps used in proving the theorem:
a. detfT(g1g1)} det[T(g1) T(g1)] (modf(x))
b. T(g,) T(g1) ô1, (modf(x))
c. =
d. gj(x)gj(x) (modf(x)).
 Selected Topics in Field Theory chapter 9

§9.6 Hilbert's Theorem 90 and Noether's Equations

The celebrated Theorem 90 of David Hilbert (1862—1943) originally
was proved about 1897 and subsequently was generalized by Amalie Emmy
Noether (1882—1935). It presents some aspects of the cohomology theory
prevalent in algebraic topology, in the modern version of class field theory,
and in portions of the structure theory of groups. The discussion for finite
fields leads ultimately to local class field theory, i.e., the study of abelian
extensions of fields obtained by the completion of number fields and function
fields with respect to special types of absolute values [for example, see
Exercise 16, §2.7].
Suppose that K = F(a) is a separable extension with the minimal
polynomial =f(x) F[x]. In a splitting field L K of f(x),
f(x) = fl
where a = a1,a2 are the distinct conjugates of a. As in §8.5, we also
denote these conjugates by a.(a) = let denote the identity automorphism
of K/F.
The product 2(a — a.) = f'(a) is called the different of the element
a; and = (— l)3N(f'(a)) with s = n(n— 1)/2, where N is the norm
function of K/F, is called the discriminant of a. Since a is a primitive element of
K with respect to F, necessarilyf'(a) 0 and hence 0.

NOTE. If b e K is not a primitive element of K/F, then F(b) K. Since b e K has

only (F(b): F] distinct conjugates [see §8.5], the images I i [K: F],
cannot be distinct. In other words, c e K has [K: F] distinct conjugates in any galois
extension LI F if and only if c is a primitive element of K/F.

Lemma 1. The discriminant =

1
fl
For any isomorphism Ij n,

= fl
Sincef'(a) 0, and aj is an isomorphism, then a3(a) 2 I n. Hence
N(f'(a)) = fl [f'(a)]

ñ 1=2
= j=1

H
= 12
=
i *j
=
I <j
Then = (— =
§9.6 Hubert's Theorem 90 and Noether's Equations 329

Lemma 2. The discriminant i\(a) is equal to the square of the determinant

1 a1 a,2
I a, a,' I]
D = det =
: : : ...
1

Note that in this determinant the superscripts denote powers of the

conjugates a1 of a. The assertion that ']= is a standard
result in the theory of determinants due to Alexandre Theophile Vandermonde
(1735—1796) for whom determinants of this type are named. We present the
proof for completeness of discussion.
In D subtract a1 times the (n— 1)st column from the nth column so
as to obtain
1 a, a,"' 0
a, (a,—a,)a,"'
D det

I an"'
Next subtract a, times the (n—2)nd column from the (n— 1)st column;
consequently
I a1 0 0

I a, ...
D=det
I a,, (a,,—a1)a,,"'
Repeating these operations, we obtain
I 0 0 0
i a,—a1 (a,—a1)a,
D = det

I a,,—a, (a,,—a,)a,, (a,,—a,)a,,"'

Il a, a, n—2
= JI
j'2 I a,,
n—2
'I

Consequently by recursion.

D = JI fl (a,—a1) =
1=! j=i+1 1

and = =
I
330 Selected Topics In Field Theory chapter 9

After these preparations it is quite easy to prove the following

lemma.

Lemma 3. If K = F(a) is a separable extension of F, there exists an element

d K whose trace T(d) is different from 0.
Proof Let {h3: I j n} be a vector space basis of K/F. Since
I • a, a2, . .,
.
a basis of K/F, we may write = I

where the matrix of coefficients C = [Ckj] is nonsingular and the elemenis

C&, belong to F.
Define elements in any galois extension L/F containing K by
= o1(b) for I / n, where the a are the isomorphisms on K/F. Then

= =

and [by) = - '3. C.

Now set
= D2.(detC)2,
where D = fJ, as above. From matrix theory we recall
that det C = det 'C, where 'C is the transpose of C, and that (det C)2 =
det'CdetC. Thus,
5(!i,, =
= det{T(b1b,)],
where the trace equals bk,bLJ [ci. §9.41.
Since D2 0 for the primitive element a (the conjugates of a are
distinct) and since C is nonsingular, sé 0 for each basis h1, of K
over F. Consequently for each such basis there exists a pair of indices i,j such
that the trace of = d is different from zero. The existence oi such an
element d yields the following theorem.

Theorem (Noeulier's Equations). Let K be a galois extension of F of degree n

whose galois group is G = {e, p, a, r, .. }. There are n nonzero elements
.

x,,, ... in K satisfying the n2 equations

=
if and only if there exists a nonzero element y E K for which
y/a(y) = U E G.

Before proving the theorem, we note that the symbolic power notation
used here and subsequently is interpreted as a° = a(a) for a e K, a e G.
Introduced in Hilbert's Zahiberichu, it was used most significantly in modified
form in Philip Furtwangter's proof of the Principal Ideal Theorem in Class
§9.6 Hubert's Theorem 90 and Noether's Equations 331

Field Theory (1930). In particular,

•yQt
= = = (ytY',
tr = a[y/t(y)J a(y)/at(r) =
If one of the elements satisfying = x0, is 0, then all the
t e G, are 0. Therefore, assume in the statement of the theorem that the
elements xa are different from 0. Also Xl•E(XT) = x1 = implies E K1

for the identity automorphism e in G.

Proof. For a e G, y' = implies that
yt _tr = 3.1
=
i.e., =
For the converse, pick a primitive element a e K/F and set

Z1 = (a(a))1.xG, 0 i < n.

At least one of the sums:, is different from zero (i.e., this is a nonhomogeneous
system of n equations), for if it were a homogeneous system, it would have
only the trivial solution = 0 for all a e G, because =
0 for a primitive element a according to Lemma 2.
Finally we compute as follows, letting : = for some I for which
0, and using Noether's equations in the simplification:

zjzd
= PEG peG

=
peG

= xg [ (p(a))i.xpj/[
peG

= XQ,

i.e., x0 = z1 as asserted.

Corollary (Hi/bert's Theorem 90). Let K/F be a cyclic extension of degree n

whose galois group is generated by a. Then N(x) = I for x e K if and only if
x=y for someyGK.
Proof Set x0 = x and define
=
Then N(x) = x.a(x)...
a"'(x) implies that = for I
i,j n [see §9.4]. Note that Xeo = x,.. = I must hold since N(x) = 1. We have
Xc'S(Xa) = x,,, whence x a(x)... a"1 (x) = N(x).
Therefore, by the preceding theorem x = x, = y' for some y K.
332 Selected Topics in Field Theory chapter 9

For the converse note that x = y' implies

N(x) = N(y)N(a(y)Y' = N(y)N(yY' = 1.

Noether's equations have the following additive analogue.

Proposition 1. If n elements xff of a normal extension K/F with galois group

G {a, r, . . } satisfy the n2 equations
.

(*) x0 + =
x,, = y—a(y) withy e K, and conversely.
Proof. The converse assertion is again easy. We have
(y—a(y)) + a[(y—t(y))] = y — a(y) + a(y) — (ar)(y)
=y .— (cir)(y).
Since K/F is a separable extension, Lemma 3 assures the existence of
an element d E K with nonvanishing trace. For

TEG

we have a(y) = [l/T(d)J Then

T(d)
= xc—
T(d)
= —
xE

and consequently

a(y) + x0 = fa(xj+x6].(at)(d).
T(d) r G

Therefore, using the assumption in equation (s), we have

cr(y) + x0 =
T(d) T 6G

= (ar) (d)] =)'

T(d) G

and =y —

Proposition 2 is the additive analogue of Hubert's theorem. Its

proof can be given by changing multiplication to addition, division to sub-
traction, and norm to trace in the proof of the corollary.

Proposition 2. If K/F is a cyclic extension of degree n whose galois group is

generated by a, then T(x) = 0 for xc K if and only if x = y—a(y) for some
y E K.
§9.6 Hubert's Theorem 90 and Noether's Equations 333

We conclude this section with discussion of two consequences of

Hubert's Theorem 90 significant in the study of "local class field theory."
Let K be a cyclic extension of the finite field F of q = ptm elements,
and consider the Frobenius automorphism of §8.2, which leaves fixed
the elements of F. Then the mappings
and
are hoinomorp/nsms of the multiplicative group K* of the field K into itself.
Following standard notational usage, we set
N(K*) = {N(a) : a e K*},
= a e K*};
and correspondingly for the additive group K,
T(K) = {T(a): a e K},
(l—q,)K= {a—q(a):aEK3.
Then
[K*: 1] l].[F*: I]
by the Galois theory, and
[K*: I] = [N(K*): I)
by Hilbert's Theorem 90. Consequently 1] = IN(K*): I] and F* =
:

since F* N (K*).
The additive version of Hubert's Theorem 90 yields a corresponding
result for the additive group of F. By the Galois theory,
[K:0] = [(l—q')K:O]•IF:O];
by Proposition 2,
[K:0] = [T(K):0].t(l—4)K:0].
Hence [T(K) :0] = IF: 0], and consequently. T(K) = T(K).

Exercises

1. Suppose that the field F has q = pfl elements, p an odd prime. Prove that
the equation ax2 + by2 = c, abc 0 in F, has q + 1 solutions if — ab is not
a square in F.
2. Let K = Q(15) and consider A = 9—4 Find:
a. N,(/Q(A).
b. An element K such that A = where a is a generating
automorphism of the galois group of K/Q.
334 Selected Topics in Field Theory chapter 9

3. Let K = Prove that = I and find all elements 8€ K

such that B/a(fl) = 2— where ci generates the galois group of K/Q.
4. Let K = Frnd all elements Bc K such that 5 = B—c(R),
where a generates the galois group of K/Q.
5. Prove Proposition 2.

§9.7 Kuminer or Radical Extensions

These special abelian extensions were first systematically studied by
Ernst Eduard Kummer in his investigations of 1-ermat's Last Theorem.
Kummer extensions are significant for investigations in class field theory
dealing with abelian field extensions and arithmetic properties of fields, and
in algebraic geometry for problems addressing covering varieties and the
resolution of singularities.
Kummer's theory is expressed multiplicatively, involving finite
abelian extensions whose galois groups have exponents relatively prime to the
characteristic of the base field. We present also the additive version of
Kummer theory (c. 1927), due to Emil Artin (1898--l962) and Otto Schreier
(1901—1929), which treats finite abelian extensions whose galois groups have
exponents equal to the characteristic of the base field.
Throughout this section we shall consider abelian extensions of a field
F, where F contains a/I of the nth roots of unity for a given n e N. (That is, F
contains a splitting field off— I.) IfcharF= Owe place no restriction on n,
but if charF = p > 0 we require that (n,p) = I.
Denote by F* the multiplicative group of F and by a primitive nth
root of unity. Since the derivative of x"— I is 0 (because either
char F = 0 or(n,p) = I ; see §5.5), the ,,th roots of unity are distinct. Therefore,
the group = 0 I < n} has order n.

Lemma I. If K/F is a cyclic extension of degree n with the properties

described above, then there exist elements A e Kand a Fsuch that K = F(A)
and A is a zero of the polynomial f—a e F[x].
Proof By Proposition 4(vi) of §9.4 for F, the norm homomorphism
N= satisfies
N(çt) = = 1.
Consequently, for a generator a of the galois group G(K/F) and for some
A e K,
= A/cr(A)
by Hubert's Theorem 90 [corollary, §9.6]. Thus a(A) = hence
= C'.A
Furthermore the equations
= = =
§9.7 Kummer or Radical Extensions 335

imply that = a F. The polynomial f—a has then distinct zeros =

C.A in K. It is irreducible in F[x], because the minimal (irreducible) poly-
nomial of .4 divides x" — a and is satisfied by the n conjugates a'(A) = A
of A. Hence the minimal polynomial has degree ii and so must equal f—a.
Thus finally, K = F(A) because
ii = [K: F] = [K: F(A)][F(A): F] = [K: F(A)].n.
The additive version of Lemma I deals with cyclic extensions K/F of
degree p. the characteristic of F.

Lemma 1'. If K/F is cyclic of degreep, then K = F(A) where A is a zero of an

irreducible polynomial f—x—a, ae F.
Proof. By Proposition 4(iv) of §9.4, for [—1] e F, the trace T= Tk,F
satisfies
T([—l]) =p.[—l] =0.
(Note that we identify the prime field of F with Z,,: cf. §3.6.) The additive
version of 1-lilbert's Theorem 90 [Proposition 2. §9.6] then yields the
existence of A e K/F such that
a(A) = A + [I],
where o is a generator of the galois group G(K/F). Furthermore cr'(A) =
A + [I], where [I] e Z,, c Next
= (a(A))"— a(A)
= = Ar—A.
Thus, Ar—A is invariant under the action of the galois group, and so
a = A"—A F. Consequently A satisfies the polynomial F[x].
The equations
p = [K: F] = [K: F(A)][F(A): F] and [F(A) : F] > 1
imply that K = F(A) and that the minimal polynomial nlA(x) of A over F
has degree p. Therefore
FUA(X) = x — a,

which must be irreducible.

A finite abelian extension K/F is said to have exponent n, if gfl = c for
all in the galois group G(K/F). The exponent is not unique. In fact, if K/F
has exponent n then it has exponent uk, for all k N. The exponent of K/F
is a multiple of the minimal exponent [defined in §*7. I and 7.2] of the galois
group G(K/F).
A finite abelian extension K/F with exponent n, which contains the
uzth roots of unity, is called a Kummer or radical extension. Again we require
336 Selected Topics in Field Theory chapter 9

(n,p) = I if char F p > 0. These extensions, as we shall show below, are

obtained by adjoining to F the nth roots of elements of F; hence the alternate
name, radical extensions.

Lemma 2. Let K/F be a Kummer extension, and denote by (A} the multi-
plicative group
(A} {AcK:AnEF*}.
The factor group {A}/F4 is isomorphic to the dual G* of the galois group
G = G(K/F).
Proof Since AN C F*, for a e G,
= = I,
and therefore A/a(A) is an nth root of unity. Let
= A/a(A) e
and note that for a, t eG

= t(A/a(A))
= t(.fA,a)
and r(A)/(ta)(A) = [A/f4 J[f4
Hence =fA,ra fA,orr e <C>
(Because G is abelian ta = at.) Thus the mapping
q4: G -# given by p4(c)
is a homomorphism of G. Next, the mapping
A: given by A(A) = q'4.
is a (group) homomorphism since for A, Be
(AB)/a(AB) (A/a(A)).(B/a(B)) = =
To show that A is surjective consider x G*. The elements x(°). x(r) of
<C> c satisfy
X(at) =
and so by Noether's equations [theorem. §9.6] there exists an element
C g* for which
C/a(C) = X(a)
for all ac G. Since n is the minimal exponent of G and hence of G*,
(x(c))" = = = I. Thus e F* or Ce {A}, and
x = A(C) = coc.
For all elements D in the kernel of A, is the unit character of G,
§9.7 Kummer or Radical Extensions 337

and consequently D e F*. Therefore the factor group {A}/F* is naturally

isomorphic to the group of characters G* according to the Isomorphism
Theorem It follows that {A)/F* is isomorphic (but not naturally) toG.
in preparation for the next lemma we consider a subgroup {s} of the
multiplicative group F* of F, such that the index {(s} : = <
where F*' denotes the set of nth powers of elements of Choose repre-
sentatives Si of the cosets of {s}/F*', and let K be the splitting field
extension of F obtained by adjoining the roots of the polynomial

Then K is a galois extension and K = F(..., ...) where S,' = and

I /i n. Clearly K does not depend upon the choice of the ?71 representatives
5m , since for ae
F*
= = F(S1).

Writing
{S} = se
we set K = F({S}) = F({s}lm).
Furthermore, since raising elements of a (commutative) group to
their nth powers is a group homomorphism, the corollary to Theorem 3, §6.8,
implies that
(s) [{S} F*] : F*f][<i) : = : F*n] = [{s) :

since the powers of ( are the only elements in K whose nth powers are equal
to 1, and since F*. Now consider the homomorphism
A: {S}/F* =
given by A (SF*) S0nF*n; it is independent of the choice of the representa-
tive S0 of the coset SF*. Because of the index equalities (*) A is an
isomorphism.

Lemma 3. The galois group G of K = F((S}) is abelian and is isomorphic

to the dual group of {S}/F*.
Proof. Consider a e G. Then S/a(S) is an nth root of unity because
e F* implies
S/a(S) e
defines a homomorphism {S)/F* since, for S1, S2 E {S} and
a e F*,
(S1/a(S1)).(S2/a(S2)) = (S1 S2)/a(S1S2)
and a/a(a) = I.
338 Selected Topics in Field Theory chapter 9

The map a is a homomorphism of the galois group G into the

dual group of {S)/F*, because
(S/a (S)) (S/r(S)) = (S/a(s)). a(S/r(S))
= (S/c(S)). (c(S)/(ot) (S))
= S/(ci')(S)
sinceSIr(S) is an ,ath root of unity and lies in F*. Thus =
Furthermore, if = I then Xa(SF*) = S/a(S) = for all 5€ (5).
1

Since K = F({S}), the Galois theory implies that a is the identity map in G.
Thus, a L. is one-one and
[G: 1] [(S}/F*: I].
Now conversely, to prove that {S}/F* is isomorphic to a subgroup
of G, consider, as in the argument for Lemma 2, the mapping for a fixed
Se(S)
4s: G given by q's(°) = S/a(S).

The mapping is a homomorphism and, as in the first part of the proof,

A: {S)/F* G* given by A(S) =
is a homomorphism. Furthermore if (p5 is the unit map, then
S/a(S)=l forallaeG.
Hence Sc F* and SF* is the identity coset in (S }/F*. Thus). is an isomorphism
from {S}/F* into G*. Consequently
[{S}/F*:1J [G*: 1] =[G: 13.
Thus finally
[G: I] = [{S}IF*: 1]
= [{S} : F*] = [{s} : F*It],
and G is naturally isomorphic to the dual group of {S}/F*. Hence, by
duality, G* is naturally isomorphic to {S)/F* (s)/F*M

We use the preceding notation to state the principal theorem of the

Kummer theory. Corollaries 1 and 2 give important consequences of this
theorem.

Theorem 1. The Kummer extensions K of exponent n over a field F are in

one-one correspondence with the subgroups {s} of the multiplicative group
F* such that [{s) F*nI < Further, if e F*,
G = G(K/F) {S}/F* = {A}/F*.
Proof The first statement is a consequence of Lemmas 2 and 3. For the
identity of (A) and (S}, let [{s} : F*f] be finite. Then the galois group
§9.7 Kummer or Radical Extensions 339

G = G(K/F) of K = F({S}) is isomorphic to {S}/F* by Lemma 3.

Lemma 2 implies that {A}!F* G. Hence {A} = {S).

Corollary 1. The galois group G of an abelian extension K/F of exponent n

is determined by the effect of its elements on the group {S} of elements in K*
whose nth powers lie in F.
Proof Let S1 S, be representatives of the cosets of {S}/F* such that the
cosets S.F* form a basis of (S}/F. Let n1 be the order of S1F*_i.e., the
least positive integer k such that = e F*_and note that n n. Further-
more, let be a fixed n.th primitive root of unity. Also = F for
i where K1 = F(S1), 1 i r. Then the field K is a cyclic extension of
degree n, over the product of fields = K1 ... K1_ K.÷1
K = k(S1), is generated by the automorphism a1 for
which a1(S1) = Then K1 is the fixed field for the group H1 =
G,., and the restriction K. generates the galois group of
K1/F.

Corollary 2. Consider multiplicative subgroups {s1} of F containing

such that F*f] < 1=1,2. Let {S1}. i= 1,2, be the set of nth roots
of elements of and K1 = F({S1}) be the corresponding radical extensions
of F.
(i) If {s1} {S2} then G(K1/K2) {s1}/{s2}.
(II) If {s} = {s2}, then K = equals K1 K2.
(iii) If(s) = {s2}, then K = equals K1 K2.

These properties follow from the duality theory of finite abelian

groups and the Galois theory and 8.8]. For the proofs it is useful
to recall that {S} {S1} implies that the annihilator AG({Sj}) of {S4} in the
group G = G(K/F) is a subgroup of G, and that the fixed field of AG({S,}) is
a proper subfield of K = F((S}). Consequently we can demonstrate a corre-
spondence between
(S1) {S2) and F({S1} {S2});
(S1}{S2} and F({S1}(S2}).
The details of the proof are left as an exercise.
These results have an additive version for finite abelian extensions K
of exponent p over a field F of characteristic p > 0. Replace the multiplicative
group of nth roots of unity in the preceding discussion by the additive group
and apply the additive version of Hilbert's Theorem 90.
Set (x) = — x. Then

p(x+y) = (x+y)" — (x+y) = (x"—x) + (y"_y) = + ga(y),

and p(z) = 0 if and only if 2 Z,,, where we identify the prime field of F with
Z,,. Let B be a root of f--x—b e F[x]. Since p(x+[iJ) = ga(x)+[O] =
340 Selected Topics in Field Theory chapter 9

p(x) for [ii e the other roots of f—x—b are B+[i], [i] Here
we write B = p - '(b), whereas in the preceding multiplicative case we wrote
S=
Theorem 2 is obtained by making the following replacements in the
discussion of Lemmas 2 and 3, and in Theorem 1.
I. A Kummer extension K/F of 1'. An abelian extension K/F of
exponent n, where char F = p, exponent p. p = char F
(p,n) =
2. Consider F* as a multiplicative 2'. Consider Fas an additive group
group
3. The multiplicative group {A} of 3'. The additive group {A} of all
all A E K such that A e Ksuch that eF
4. = A —o(A)foro e G =
G(K/F)
5. =fA.a E <i> =fA.,,
6. fA,tv IA, and thus E G* 6'. JA.,,+fA and thus (p4 E
Hom(G,
7. = and thus 7'. 'PA+B(a) = A(o)+q'Ba) and
2: {A} —' G* given by 2(A) = thus 2: {A} —, Hom(G, given
is a homomorphism by 1(A) = (p4 is a homomor-
phism
8. 8'.
9. {s} a subgroup of F* such that 9'. {s} a subgroup of F such that
[(s} : F*n] < co [{s} : < co, where
p(F)= {p(a):aeF}
10. K = F(..., ...) where Se" = 10'. K = F(..., S.+ fh], ...) where
and is a primitive nth root p(S1) = and [h] e
of unity
11. f{S} : F*] = [{s} : 11'. [{S} : F] = [{s} : p(F)]
12. Xc(SF*)_S/0r(S) 12'. XQ(S+F)= S—a(S)
13. [S/a(S)] .[S/z(S)] = S/(ar)(S) 13'. [S—a(S)]+[S-—r(S)] = S—
(at)(S).

Theorem 2. Finite abelian extensions K/F of exponent p. where p is the

characteristic of F, are in one-one correspondence with the subgroups {s}
of the additive group F such that [{s} : p (F)] < co and G = G(K/F)
{S}/F= {A}/F.

REMARK. In the case that K/F has exponent ptm but not exponent p, where
p = char F, we can use the so-called Witt vectors [see N. Jacobson, bc. cit., pp. 124—
139] or an inductive argument due to A. A. Albert (1905—1972) [see Bull. Amer.
Math. Soc. 40(1934), pp. 625—631].
 ______

Unique Factorization Domains and Elementary Symmetric Functions 341

Exercises

1. Suppose that and x"—b are distinct irreducible polynomials of

Q[x] such that [Q(A,B):Q]=p2 for two zeros ,4,Bof
respectively, in an extension L/Q. Prove that [Q(A + B) Q] = p2.
2. Suppose that F is a finite field, char F = p. and x, are indeterminates
over F.
a. is irreducible in F(x)[t]?
b. Show that i"—t—x' is irreducible in F(x)[i].
3. Assume that f(x) = x" — x — a is an irreducible polynomial in the poly-
nomial ring F[xJ, char F = p > 0. For e F such that 1(a) 0, prove
that the polynomial is irreducible.
4. Let be zeros of the irreducible polynomials
c F[x], respectively, where p char F. Prove that =
F(A2) if and only if /12 = [cJA1+b, where [c) 0 in Z" and ha F.
5. If char F = p > 0. prove that — x — a E F[x] either factors completely

or is irreducible.
6. Prove that for an odd prime p the polynomial —ae is irre-
ducible for all n N if a is not a pth power in F.
7. Prove that the galois group of the splitting field K/F of x" — a E F[x] is
isomorphic to a subgroup of (he affine group of 4 if (n, char F) = 1.
(If charF= 0, no restriction is placed on n.)
8. Suppose that the field Fcontains all nih roots of unity where (n, char F) =
I. Prove that F(A) = F(B). with = a, = b, and [F(..4): F] =
implies that b = where c F and (s, n) = I.
9. Prove Corollary 2 to Theorem I in detail.
10. In the parallel argument preceding Theorem 2, verify statement 8' in
detail.
II. Let K = wnere I = 0.
a. Find the least normal extension of KiQ and determine its galois
group G.
b. Determine a composition series of G.
12. Let K = where is a primitive sixth root of unity. Find the
galois group G of the least normal extension of K/Q(i), and determine a
composition series of G.
13. For relatively prime integers ni and n prove that x" — a E F[xl is
irreducible if and only if both xm—a and f—a are irreducible in F[x].

§9.8 Unique Factorization Domains and Elementary Symmetric

Functions
This section, essentially arithmetic in nature, provides the preparation
for Gauss' proof of the Fundamental Theorem of Algebra [see §9.9]. Emphasis
is placed on the concept of' integral dependence. Thus we obtain a proof of the
Fundamental Theorem of Elementary Symmetric Functions somewhat
342 Selected Topics In Field Theory chapter 9

different from the usual one, and at the same time prepare for important
concepts in ring theory: unique factorization domains and integral dependence.
Suppose now that R is an integral domain with Ffor its quotient field
and that K/F is an algebraic (field) extension. Since we consider only elements
in a field K, left and right R-modules in K can be considered to be equal. Thus
in this section we shall not distinguish between left and right modules.
in and 2.7 we considered the factorization of integers and
properties of certain "irreducible" integers, called primes, which could not be
factored nontrivially. Similarly in §5.2 we considered the questions of divisi-
bility and factorization within the integral domain of polynomials over a field.
We now extend the discussion of factorization to more general integral
domains.
A nonzero element a in an integral domain R is called irreducible if
0) it is not a unit of R, and
(ii) whenever a bc with b.c E R, then one of the elements h and c is
a unit of R.
A nonzero element a e R, which is not a unit, is said to have a unique factoriza-
tion into irreducible elements of R, if
(1) a= u p, with a unit u e R and irreducible elements p1.
I I r, in R, and
(ii) if a = v is another such factorization, then r = s and
= with units U1 R, I i r, where is a permutation
of {l,...,r}.
We call an integral domain R a unique factorization domain (commonly
abbreviated UFD) if every nonzero element has a unique factorization into
irreducible elements.
A nonzero element a e R is termed a divisor of b e R, if ac = b with
c e R; symbolically, we write a Jb. An element d e R is called a greatest
common divisor of a and b if d I a and d b, and if every nonzero element g e R
that divides a and b also divides d. Elements a,a1 of a unique factorization
domain R are said to be associates, a aa, if a1 = ca for some unit c of R.
The relation of being associates is an equivalence relation.
Every element a e F = Q( R) can be written as a quotient u/v where u
and v lie in R and have no common irreducible (nonunit) factors. The factor-
ization property implies that, for each irreducible element p
a __p2a* ,&O
with a* F, where neither numerator or denominator of a* has p as a factor.
We now define an order flinction Ofl F = Q (R) for a given irreducible element
p E R [cf. Exercise 14, §2.7] by setting
= e Z.
Note that = for every unit e of R. For a = 0, we set = The
 Unique Factorization Domains and Elementary Symmetric Functions 343

unique factorization of elements implies that, for any pair of nonzero elements
a,bof F,
= +
The order function i', extends to the polynomials
f
= f(x) = + + + a0

in F{x] as follows:
= :0 i n}.
A p-content of the polynomial f is an element where
= z',,(f) and is a unit of R. The content C(f) is defined to be =
fl,, where is any unit of R, and where the product is taken over a set
of representatives of the distinct classes of associated irreducible elements
p e R. (The product is well-defined because all but a finite number of the
are equal to 1.) This definition implies immediately that
C(af) = aC(f)
for every nonzero element a of the field of coefficients F. Consequently
= cf1(x)
where c = C(f) and the content of f1(x) e REx) is 1. More precisely, the
coefficients off1(x) lie in R and their GCD is 1.
The arguments of the proof of Gauss' Lemma generalize to
yield the proof of Lemma 1. In the proof, which we leave to the reader, it
suffices to consider polynomialsf.g for which C(f) = C(g) = and to prove
1

that C(fg) = l—in other words, that = 0 for all irreducible elements
p e R—because in the general case we write 1= af1, g = bg1, where C(f1) =
C(91) = 1, so that C(fg) = abC(f1g1).

Lemma 1. If R is a unique factorization domain with quotient field

F= Q(R), then forf,ge F[x]
C(fg) = C(f) C(g).

An immediate consequence of Lemma I is that if h(x) e R[x] has

the factorization h(x) =f(x)g(x) in F[x], then
/,(x) = C(J)C(g)J1(x)g1(x),
where C(g)C(h) 6 R and J1(x) and g1(x) are polynomials in R[x] with con-
tent I. A generalization of this argument proves the following lemma [cf. the
unique factorization of integers in §2.7 and of polynomials, §5.2].

Lemma 2. If R a unique factorization domain, then so is the polynomial

is
ring R [x]. The irreducible elements of R [x] are either irreducible elements of
R or polynomials of R[x] which are irreducible in F[x], F = Q(R), and have
content I.
344 Selected Topics in Field Theory chapter 9

The preceding lemmas together with induction on the number of

indeterminates yield the proof of the foltowing theorem.

Theorem 1. If x1, ..., are indeterminates over a field F, then the poiy-
nomial ring F[x1 v,,] is a unique factorization domain.

Before introducing the second objective of this section, elementary

symmetric functions, we define the concept of integral dependence. The
following discussion of integral elements and integral closure is parallel to
that in Chapter 8 of algebraic elements and algebraic closure.
An element a K is said to be integral over R, if there exists a finitely
generated R-module Ma K such that
aM0 = {am : in E M0) M0 = (a1, .

We use a series of four lemmas to develop properties of elements in K which

are integral over R.

Lemma 3. The set of elements in K integral over R is a subring of K.

Proof. Consider elements a, b K, integral over R. Express M0 in terms of
generators a1, ..., a5 as
M0=Ra1+••.+Ra5,
and Mb in terms of generators b1, . .,b, as

M
M= + + + + Ra,b,,
I I Since aMc M and bMc M, we have M
and abM M. Thus, a + b and ab are integral over R.
Consequently the set S (a e K: a integral over R} is a subring of
K; it is called the integral closure of R in K.

Lemma 4. An element a E K is integral over R if and only if it is a zero of a

monic polynomial in R[xJ.
Proof First, consider a K, integral over R;

a a zero of the characteristic polynomial

of the matrix [do] [see §5.9].
Conversely, suppose =0. Set Ma =
Then, by induction on/i k, we can show that aha Ma, and
hence that aM0c M0, since
 Unique Factorization Domains and Elementary Symmetric Functions 345

Lemma 5. An element c e K, integral over the integral closure S of R in K,

is integral over R.
Proof. By Lemma 4, the element satisfies a monic polynomial
c
X"+C&_ + •.+c0 in S[x]. Now let M = Mk, the product module
generated by the set of elements {rn1 : rn1 M., I I k}, where M1
is a finitely generated R-module in K for which c1 M. M., I I k. Then
c4M cM; and Mis finitely generated. Finally = M+cM+...+ck_IM is
a finitely generated R-module, and c
A subring of a field F is said to be integrally closed in F if it contains
all elements of F which are integral over it. Lemma 5 states then that the
integral closure S of R in F is itself integrally closed in F.

Lemma 6. A unique factorization domain R is integrally closed in its quotient

field F = Q(R).
Proof Suppose that a/b c F is integral over R, and a/b R with a, .5 R
having no common irreducible factor. Let p be an irreducible element that
divides b. Then a relation of integral dependence of the quotient a/b over R,
(a/b)tm + c,,, 1(a/h)m '+ + c0 = 0,
with coefficients c4 R implies that
atm + - ibaml + ... + = 0.

The assumption that b implies then that atm and consequently that a,
contrary to the assumption on a. Hence a/b cannot be integral over R, and R
is integrally closed.
We now apply Theorem I to polynomial rings. Let F be a field and
R= xe]. Then K = Q(R) = F(x1 ;) is the field of rational
functions of x1, with coefficients in F. Suppose that

o=IFl
L it
is an element of the symmetric group E,, on n elements [see §6.7]. The mappings
for a defined by
xv)]

are distinct automorphisms of K/F. Let L be the fixed field of G = {y0 : crc
as in §8.6. Then K is a galois extension over L of degree n! and K =

The polynomial

g(f) = fl (t—x1) = + (— in K[tJ.

346 Selected Topics in Field Theory chapter 9

where ej(xi, = for Ij n, can be re-

written as

g(t) = J1

for each a e Consequently the coefficients of g(r) remain unaltered by

application of any a e i,,. The polynomials e1, ...,e1, called the elementary
symmetric functions, therefore lie in the fixed field L.
Hence the rational function field E = F(e1, is a subfield of the
fixed field L of the automorphism group G of K. Actually £ equals L. Since
K = E(x1) contains then distinct zeros x1, ..., of the polynomialg() e
it is a galois extension of E, and consequently the galois group G(KJE) is
isomorphic to a subgroup of [see §8.6]. Therefore [K: L] = = [Es: 11
and L E imply L = £ as asserted.
To summarize, we have proved the following lemma.

Lemma 7. The field K = F(x1, is a galois extension of the field

E = F(e1, of the elementary symmetric functions e1, with co-
efficients in F. The galois group of K/E is isomorphic to the symmetric group

Lemma 8. The elementary symmetric functions e1, ...,e,, of the indeter-

minates x1 over F are algebraically independent over F.
Proof We shall use an indirect recursive argument. Suppose that e1, ...,
are algebraically dependent, that is,
(*)

for some nonzero polynomial g(u1, in the polynomial ring

uj in the indeterminates u1, ..., Let f(u1, ..., be a polynomial
of minimal degree h in among all polynomials satisfying relation (*). It can
be written in the form
f =f0(u1, + ...
where necessarily f0(u1, ..., #0, because of the minimal choice of h.
Hence
fl_ti
— + ...
This is a relation in F[x1, . . ., and remains valid if the indeterminate is
replaced by 0 (a homomorphism of F[x1, . . ., to F[x,, ..., since
= fl7. xj. Hence
= 0,
where the terms = e,.0 are seen by inspection to be the elementary sym-
metric functions of x1, That is, the elementary symmetric functions
e, are then algebraically dependent.
§9.8 Unique Factorization Domains and Elementary Symmefric Functions 347

Since e1 = x1 for n = 1, the preceding reduction from n to n— I

implies that e1, are algebraically independent.

We conclude this section with the important concept of symmetric

polynomials; a basic property of such polynomials (Theorem 2) is sometimes
referred to as the Fundamental Theorem of Elementary Symmetric Functions.
A polynomialf=f(x1 F[x1, is symmetric if, for each ce Z,,,

Ya(J) =j:
Theorem 2. A symmetric polynomial f in the indeterminates x1. is
equal to a polynomial /z(e1 e F[e1, ..., es], where the are the
elementary symmetric functions of the I <1 < n.
Proof. Since y0(f) =f for all ac the polynomialf lies in the fixed field
E= Because the elements are the zeros of the poly-
nomial g(t)e F[e1, they are integrally dependent over F[e,, ...,en]
[see the discussion prior to Lemma 7]. Consequentlyf is integrally dependent
on the unique factorization domain F[e,, by Lemma 4. Finally
F[e1, is integrally closed according to Lemma 6 together with Lemma
8 and Theorem I. Hence F[e,, as asserted.

Exercises

1. a. Express as a polynomial in Q[e1,e2,e3], where

e1, e2, and e3 are the elementary symmetric functions of x1, x2, and
x3.
b. Express and as polynomials of the elementary
functions
2. A ring is called a principal ideal ring if all of its ideals are principal. Prove
that such a ring is a unique factorization domain. (Is the converse true?
Examine Q[x, y].)
3. Let AI,A2,A3 be the zeros of the irreducible polynomial x3+px+qE
Q[x]. Express [(A3—A1)(A3—A2)(A2--A1)]2 in terms of p and q.
4. Let K = Q(w), where w2+w+ = 0. 1

a. Prove that the subring R = {a+bmw : a,b e Z} is not

integrally closed in K; m 0, ± I.
b. Find the integral closure of such a ring R in K.
5. Let K = Q( where [K: Q] = 2 and in e Z is a product of distinct
primes. Prove the following statements:
a. Ii m 2 or in 3 (mod 4), the integral closure of Z in K equals

b. If in I (mod 4), the integral closure of Z in K equals R =

(a+5b : a,be Z} where ö (I + Show that
R =
348 Selected Topics in Field Theory chapter 9

6. Let R be an integrally closed integral domain ER Q(R)]. Suppose that

M is a subset of R such that M and s1s2 M for all s1,s2 eM.
Generalize the construction of the quotient field thus defining
M - 'R, and prove that M - 1R is integrally closed.
7. Consider a ring R with subring S having the same multiplicative unit as
R. Suppose that A and P are prime ideals of S and R, respectively, for
which P m S = A. Prove that there exists an embedding of the residue
class ring S/A into RIP. Furthermore, determine whether or not RIP
consists of integral elements over the image S/A if R consists of integral
elements over S.
8. a. Let R be a unique factorization domain with the quotient field

where a0 0, is necessarily an element in R which divides a0 [cf. the

Integral Root Theorem, Exercise 9, §5.2).
b. State and prove an analogue of the Rational Root Theorem, Exercise
8, §5.2.
9. Suppose that y)E F[x, y) satisfies k(x)f(x, y) = g(x, y)h(x, y) for
some k(x) c F[x], and g(x, y),h(x, y) c Fix, y), and that the coefficients
of powers of y in the polynomial p(x, y) have GCD I. Prove that k(x)
divides each of the coefficients of the powers of y in the polynomial
h(x, y).
10. Suppose that the polynomial f + a1 x' - + + e F[xl has the zeros
...,A in some field K 2 F. Let Sk = + A,,". Prove
Newton's formulas:
a.
b. fork>n.
11. Prove that the only units of the integral closure of Z in for
d=—2ord<—3, are ±1.
12. Prove Lemma 1.

§9.9 The Fundamental Theorem of Algebra

The Fundamental Theorem of Algebra states that the field of complex
numbers C = R(i), i2 = — 1, is algebraically closed. An alternate statement
is that any polynomial with complex coefficients has a root in C, and hence can
be written as the product of linear factors in Cfx].
This theorem is attributed by the mathematical historian, David E.
Smith, to Peter Roth (C. 1608) and Albert Girard (1590—1633) in 1629. Francois
Viête (1540—1603) was cognizant o the relations between the zeros of a
polynomial with rational coefficients and its coefficients. In the middle of the
eighteenth century Jean le Rond d'Alembert (1717—1783) made several
attempts, all unsuccessful, to prove Girard's conjecture, and because of his
efforts his name is commonly given to the theorem, especially in France.
While Euler (in 1749) and Lagrange also attempted proofs, the first rigorous
proof of the Fundamental Theorem of Algebra was given by Carl Friedrich
Gauss (1777—1855) in his doctoral dissertation at the University of Helmstedt
§9.9 The Fundamental Theorem of Mgebra 349

in 1798, published a year later. In this paper Gauss also demonstrated the
inadequacy of earlier proofs, including those of Euler and Lagrange.
Complex numbers, which we write a + bi with a, b E R, were some-
times viewed as cosets in R [x]/(x2 + I), for the geometrical description of the
complex plane was new. In 1798 Caspar Wessel (1745—1818) published his
graphical representation of complex numbers in the transactions of the Danish
Academy of Sciences, and Jean Robert Argand (1768—1822) introduced the
complex plane in 1806. Gauss subsequently strove, albeit without success,
for a strictly algebraic proof of the theorem, presenting two new proofs in 1816
and one in 1850.
There are a number of proofs of this theorem, all of them analytical
(or topological) by nature, a quality inherently due to the definition of the
field C as a quadratic extension of the field of real numbers. The proof we
present appears to be due in part to Lagrange. It uses a minimal amount of real
analysis such as the Intermediate Value Theorem whose proof in turn employs
not more than the existence of least upper bounds, a characteristic property
of the real numbers, and a simple inequality.
Gauss reduced the problem of finding = such that =
a+bi to one of finding the intersection of two hyperbolas a and
afl = 1,12. For purely imaginary numbers, a = 0, and the hyperbola
= 0 is degenerate. Thus the problem is one of finding the intersection
of two lines = ± and the hyperbola cx/3 = h/2.
As a preparatory lemma, we note that a real polynomial
1(x) = x" + - + + a0 e R [x]
of odd degtee has at least one real root. First, there exists a positive real
number p such that
f(x)>0 forx>p
and f(x) <0 for x < —p.
We have the following estimate for a> I:

= >0
Furthermore
f(—a) + ..'
<
= <0.
Hence let p Max{l,IaoI+ + I. The Intermediate Value
Theorem, applied to the interval —p x p, implies that f(x) has at least
one real zero.
350 Selected Topics in Field Theory chapter 9

Proof of the Fundamental Theorem of Algebra. It must be shown that every

polynomial g(x) = +.••+e0 with complex coefficients has a
zero in the complex field C. Set
g*(x) = x" + x" -' + +
where is the conjugate complex number I v n. Then
f(x) = g(x)g*(x) = (g(x)g*(x))*
is a polynomial with real coefficients since it equals its own conjugate. There-
fore the complex zeros of f(x) occur as pairs of conjugate numbers. Con-
sequently the proof is reduced to showing that real polynomials split into
linear factors in C[x]. Thus consider a real polynomialf(x) of degree d = 2mq,
where (q, 2) = I. We proceed by induction on m.
If m = 0, then d is odd and hence has a real zero according to the
preliminary discussion. Suppose now that m I. The theory of splitting fields
[see §8.1) implies the existence of an extension K of C such that

f(x) A (x—a1) E K[x].

=
Now pick an element u R, and define d(d+ 1)/2 elements

The coefficients of the polynomial h(x) = are symmetric poly-

nomials of a1 with coefficients in R. Consequently they are polynomials
ad
(with real coefficients) of' the elementary symmetric functions of a1, ...,ad,
according to Theorem 2, §9.8. (Substitute indeterminates for a1, ...,ad, apply
the theorem, and ultimately replace the indeterminates by the elements as.)
Consequently h(x) is a real polynomial of degree d(d+ 1)12 = 'q', where
= q(d+ I) is odd since both q and d+ I are.
Now li(x) has a zero in C by the induction hypothesis because
(q',2) = 1. Consequently one of the elements must be equal to;; denote
this element by
= aI(U) + aJ(U) +
Since R has infinitely many elements and since the number of pairs (i,j) with
I 1 d is finite, there must exist two distinct real numbers u and v for
which 1(u) = i(v) = r andj(u) =j(v) = s. Consequently, both
a, + a5 + ua, a3 and a, + a5 + va, a5
belong to C, as do
v (a, + a5 + ua, a5) and u (a, + a5 + va, a5).
Thus a,+a5 C, and therefore a,a5 e C.
This fact implies that a, and a5 satisfy a quadratic equation
= 0 with coefficients in C. The roots of this equation are

±
 Finite DivIsion Rings 351

where with a,b in R. Then

±sJa+bi = ±
Consequently. a, and as lie in C, and J(x) has zeros a,,a5 in C, as asserted.

Exercises

1. For u1,ii2 C, prove the Triangle Inequality Iui+usI u11+lusI. (Hint.'

Write the u'sin the form u = a+bi with a,bE R.)
2. Let a and b be real numbers. Find real numbers a, fi such that a + bi =
(a+fii)2.
3. Find the conjugates ofA = ./ThE C overQ(i), P = —1.
4. Assume that A is a zero in the splitting field K of = x4 + 6x3 + lOx2 +
3x+IeQ(xJ. Prove that B=A+A2 has degree 2 overQ. Is B a real
number? Find the other zeros of 1(x).

§9.10 Finite Division Rings

One of the famous theorems of Joseph Henry Macglagen Wedderburn
(1882—1948) asserts that any finite ring with a unit for multiplication whose
nonzero elements form a multiplicative group is necessarily a field. The proof
we present is due to E. Witt (see Abhandlungen des Mathematischen Seminars
der Universitäf Hamburg, 1931). It should be noted that although we are
dealing with a strictly algebraic question, the clinching argument depends
upon an analytic statement concerning the zeros of a cyclotomic polynomial.
For a completely algebraic proof see I. N. Herstein, Noncominutative Rings.
A (not necessarily commutative) ring D with unit for multiplication
whose nonzero elements form a multiplicative group D* is called a division
ring.

Lemma. The center F of' a division ring D is a field.

By definition, the center is the set

F= {x e D ax = xa for alla D}.
For x,y e F, and for all ae
a(x+y) = ax + ay = xa + ya = (x+,v)a,
a(xy) = (xa)y = x(ay) = (xy)a,
ax1 =(a"1x)' = ifx 0,
Thus x+y, xy = yx, and x' all belong to F, so that F is a field.
Theorem (Wedderburn). A finite division ring D is a field.
352 Selected Topics in Field Theory chapter 9

Consider D as a vector space over its center F. Since D is finite, the

dimension D = n is finite. The object of the proof is to show that n> I
leads to a contradiction, and thus that D must equal F. First note that if F has
q elements, then D has q" elements. Next, for a e D let
F0 {u e D: au = ua}
denote the normalizer of a [see §7.4]. Since for u, v F0

a(u+v) = au + av = ua + va = (u+v)a
and a(uv) = (ua)v = (uv)a,
U+v, Ut E F0. The fact that D is finite implies that the set {I,u,u2, ...) of
powers of any nonzero element u is finite. Consequently for some Ii, k with
h> k, we have uh = uk. Thus, = 1 so that u"1 E F0 is the inverse
of u. Therefore F0 is a division ring.
The division ring Fa may be considered as a d-dimensional vector
space over F; then F0 has q" elements. Furthermore, 0 can be considered as a
left module over F0, where the module multiplication of x D by u is
the product ux, so that for y D and u, v e F0 the equations
lx = x, u(x+y) ux + uy,
(u+v)x = ux + vx, (uv)x = u(vx)
verify the axioms of a left module [see §7.8]. Just as in the theory of vector
spaces, it follows that D contains a basis x3, ...,x5 over F0 [see §4.2]. The
number of elements of D, namely q", must equal hence n.
Next we apply the class equation to the group D*. First note
that the number of distinct conjugates (with respect to the group of inner
automorphisms of D) of a nonzero element a e D is (q"— I). Then

(*)

where the summation is taken over the distinct conjugacy classes.

Finally, let = fl Z[xJ be the nth cyclotomic poly-
nomial where the elements denote the q'(n) primitive nih roots of unity
[see §9.1]. If d is a proper divisor of n,
rf-l
Since all polynomials in this relation are monic, it follows as before
that
§9.10 Finite Division Rings 353

with g(x) E Zfx]. Furthermore, I) implies for the same reason

that
(x"—l)
with h(x) E Z[x]. Consequently these factorizations, which are identities in
Z [x], yield for the homomorphism Z.[x] —* Z given by x —' q that

and

The class equation (*) implies that

I).

This relation cannot hold if n> I, because for every primitive nth root of
unity (= I I q(n), the inequality I holds, as is shown
below. Thus
(yP(n)
>

contradicts expression (**).

Note that IcosCI=IRe(l. Since the real part of the complex root of
unity (is less than 1,
12 ReCI = < 2,
q2 — > q2 — 2q,

or (q—KI)(q—I(I) > (q—l)2;

hence

Exercises

1. If F is an algebraically closed field and D is a division ring whose center

contains F, such that the elements of D are algebraic elements over F,
prove that D = F.
2. Let R = ([l],i,j,k)4, p 2, where

=
f2 = k2 = —[1], ii = —ii, jk = —jk, ki = —1k.
Prove that R is isomorphic to the ring of 2 x 2 matrices with coefficients
in

3. In Exercise 2, let p = 2. Does R then contain a two-sided ideal A 0 for

which A2 = 0? If so, what is the structure of the residue class ring
R/A?
354 Selected Topics in Field Theory chapter 9

§9.11 Simple Transcendental Extensions

An apparently easy problem in the structure theory of simple trans-
cendental extensions—that is, fields of rational functions, F(x1,
the question whether subfields properly containing the field of coefficients F
are themselves fields of rational functions. A solution for F(x), Fan arbitrary
field, is provided by the theorem of Jacob LUroth (1844—1910). Its, proof, as
will be seen below, depends upon arithmetic properties of polynomials of one
and two indeterminates. If ii = 2, then algebraic-geometric arguments are
needed [see the work of Guido Castelnuovo and Oscar Zariski cited in the
bibliography]. In the cases of n 3, only special results are known, most of
which are negative. A paper of Frederigo Enriques (1871-.-1946) indicates
that a general theorem is far from being realized. Recently Y. Manin and V.
lskovskibh have given a set of counterexamples in the case n = 3. Here is one
of the outstanding challenges of algebra which will certainly require con-
siderable advances in arithmetic algebraic geometry and analysis.
An element in a proper extension ElF is called transcendental, or
(by abuse of language) an indeterminate over F, if it is not algebraic over F.
That is, e E/F is transcendental if it is not a zero of some polynomial
f(X) F[X].
We define the degree of an element q in the quotient field
of a transcendental element, where without loss of generality
we assume that = 1, to be the larger of and

Theorem I. If the element q e does not lie in F, then is transcendental

over F and = n, where n is the degree of
Proof Consider with relatively prime polynomials and
in Then satisfies the equation = 0 with coefficients
in F{t1]. In detail, if
= + a1 + ... and = + b1 + ...,
then (b0j—a0) + + + = 0.
The leading coefficient ij — is not zero since at most one of can be
zero, since has degree n and F. Consequently is algebraic over the field
Hence must be transcendental over the field of coefficients F, for other-
wise would be algebraic over F, contrary to the hypothesis that is trans-
cendental fsee the corollary to Theorem 2, §8.1].
Now let I be an indeterminate over the field F(,1). Then is a zero of
h(X) = g(X),7 —f(X) e
This polynomial has degree n with respect to I according to our initial
assumption and Furthermore, using the Lemma of Gauss, §5.8,
it suffices to prove that h(X) cannot be factored in =
For the indirect proof of irreducibility assume that lz(X) can be
factored in F[q, I]. The polynomial h(X) is linear with respect to and
§9.11 Simple Transcendental Extensions 355

consequently reducibility implies that one of its factors must be a polynomial

k(X)EF[X]. Write h(X) = in Applying the homo-
morphism X] —. F[X] obtained by setting = 0, we find that k(X) 1(X).
Furthermore, k(X) then divides h(X)—f(X) = and therefore
k(X)Ig(X). This contradicts the assumption that are relatively
prime. Hence is a zero of an irreducible polynomial of degree n with co-
efficients in F(ij), and as asserted, = n.
Note the following, immediate consequence.

Corollary. The polynomial F[q,XJ has no factor in

F[X]\F.

Moreover, if q is replaced by its representation it follows

that 9(X)Jg) has no factors in F[XJ\F.

Theorem 2. The group of automorphisms (over F) of a simple transcendental

extension F is isomorphic to the projective linear group PL (2, F).
Proof Suppose that a is an automorphism of Then =
for Hence a is completely determined by
the image = of Let q where = I in
By Theorem I
= ii,
where n = Max
For an automorphism a, the fields and must coincide.
Hence the degrees of the cannot exceed 1. Furthermore
at least one of these degrees must be 1, for otherwise F contrary to our
hypothesis. Consequently
+b
=
+d
where a, b, c, d lie in F, and at least one of the coefficients a, c is distinct from
zero. Since = I,

det[
ía bl

(If the determinant were 0, then q = would tie in F, since the rows of
[a b]
Lc dJ
would be proportional.) Conversely, each nonsingular matrix
[a bl
Lc d]
356 Selected Topics in Field Theory chapter 9

with coefficients in F determines by =q an auto-

morphism of for = I according to Theorem I.
Finally, we note by direct verification that the mapping
ía bl
Lc
of the group of 2 x 2 matrices with coefficients in F is a homomorphism onto
the automorphism group A of The kernel of this map consists of the
diagonal matrices
ía 01
[o a]'
a 0. Hence A is isomorphic to the factor group
(Ia bl ía bl ) I(Ia 01 1

that is, the two-dimensional projective linear group PL(2, F) of the field F.

Theorem 3 (Lüroth). Let the proper extension KIF be a subfield of the

rational function field F(x). Then K is also a rational function field F(y).
Proof For any z e K\F, F(x) is a finite algebraic extension of F(z) by
Theorem I. Hence F(x) is also algebraic over K; thus F(x) = K(x), where
[F(x): K] = n. Now let
f(X) = + ... +
X an indeterminate over K, be the minimal polynomial of x. Note thatf(X)
F[X}, because otherwise a• F and x would be algebraic over F, a contra-
diction. Thus, at least one of the coefficients a. E K\F.
As elements of K c F(x), the coefficients a are rational functions of
x. Then f(X) can be multiplied by an appropriately chosen polynomial in
F[x] so that the resulting product
f*(x,x) = + + ±
in F[x,X] is irreducible and primitive [see §5.8]. The uniqueness of
implies that
= 0 i<
Sincef(X) FIX], at least one of these coefficients, say actually involves
x because f*(x, X) is primitive. Now write y = aj = g(x)/h(x), where
(g(x),h(x)) = 1. The degrees of g(x) and h(x) with respect to x are at most m,
the degree off*(x, X) with respect to x. Since 0, we have

P(x,X) = g(X) — jh(X) = g(X) — 0.

(x)j
 Simple Transcendental Extensions 357

Applying the homomorphism from F[x,X] to F[x), we obtain

P(x, x) = 0, which implies that x satisfies P(x, X) e K[X]. Therefore
P(x,X) 0 (modf(X)) in K[X] by the definition off(X).
Next the Lemma of Gauss relates divisibility in K(x)[XJ to that in
K[xJ[X]. The preceding congruence implies that P(x,X) = q(X)f(X),
where q(X) e K{X]. All the coefficients of the powers of X in this relation are
quotients of polynomials in F[x]. Hence multiplication by a polynomial in
Ftx] results in an equation in F[x,X] of the form
s(x)f(X)q(X) = s(x)P(x,X)
= s(x)fg(X)/i(x)—g(x)h(X)] = f*(x, X) i(x, X).
Since f*(x, X) is a primitive polynomial, s(x) must be a factor of t(x, X);
hence we may write

A(x, X) = g(X)h(x) — g(x)h(X) = f*(x, K)u(x, X),

with u(x, X), A(x, A) F[x, X]. By the definition of A (x, X) we have
m, whereas = m. Hence u(x.X)e FUX].
Furthermore, referring to the corollary to Theorem 1, we note that
u(x, X) E F. Thus we may write, ignoring factors in F,
A(x,X) = g(X)/z(x)—g(x)Js(X) = cf*(x,X),
o ce F. Now observe that = n. Furthermore A(x,X)
—A(X,x) implies that m = n.
Finally, F(x) K F(y)—thus [F(x): F(y)] m by Theorem 1—
and [F(x): K] = n imply that K = F(y), as asserted.

Exercises

1. a. Prove that the six mappings of 1(x) e Q(x), given by f(x)

J(l—x), f(l/x), f(l—l/x), J(l/[l—x}), and f(x/[x—lfl, are auto-
morphisms of Q(x) and form a group. Which group of order 6 is
this?
b. Show that Q(x) is a galois extension of degree 6 over the field of
rational functions Q(r(x)) of
(x2—x+ 1)2
x2(x— 1)2
2. Prove that K Q(x) is a quadratic extension of the fields F1 = Q(x2)
and F2 Q(x(x + ii). Determine the automorphisms of K/F1 and K/F2
in terms of x. Show that K is not a finite algebraic extension of the
intersection F1 F2.
3. Prove that F(x) is a finite algebraic extension of any subfield E that is a
proper extension of F.
358 Selected Topics in Field Theory chapter 9

§9.12 Perfect Fields

in this final section we discuss the so-called perfect fields: fields
which, as far as irreducible polynomials are concerned, have the same
properties as does the field of rational numbers. All algebraic extensions of
such fields are separable [cf. §8.5]; all irreducible polynomials have distinct
roots. The existence of an algebraic extension of a given field in which all
irreducible polynomials are separable will be proved. We also present Robert
Gilmer's argument on the existence of an algebraically closed extension,
which avoids the double induction approach used in §8.9 [see Amer. Math.
Monthly 75(1968), pp. 1101—1102].
A field F is called perfect if every irreducible polynomial in F[x]
is separable.
in §5.5 we proved that every field of characteristic 0 is perfect. Every
finite field is also perfect. For the proof see §8.2 and note that the field of pfl
elements, determined by a zero of an irreducible polynomial of degree n with
coefficients in the prime field Z1,, is a galois extension of Z,, [see §8.6].
An equivalent statement of the definition is given for fields of non-
zero characteristic in the following lemma.

Lemma 1. A field F of nonzero characteristic p is perfect it and only if every

element a e F has a pth root .4 in F; that is, if and only if the polynomial
is reducible in F[x].
Proof Assume F is perfect. The reducibility of f—a in F[x] implies
f—a = g(x)h(x) with a proper factor g(x) E F[x] of degree s <p. In a
splitting field K of F[x], we have x"—a = [see §5.5]. Hence g(x) =
Now (s,p)= 1, so that st+pq= 1 for some t,qeZ. Consequently
(x — A)1 = g(x)'. (x"— F(x). Hence A 1 and therefore (x — =
f—a in F[x].
Next, if some element a e F is not the pth power of an element in F,
then f—a must be irreducible in F[x] by the preceding argument. Thus F
has the inseparable extension F(A), A = F, which contradicts the
assumption that F is perfect. Consequently A e F, and a has a pth root in F.
Conversely, assume that every element of F is a pth root of an
element in F. if there were an irreducible inseparable polynomialf(x) in F[x],
then
f(x) = = yfl + a1

y= where e is the maximal such integer, as in §5.5. Now a1 =

A e F, by assumption. Consequently

((x) =
andf(x) is reducible. Hence e must be 0, which means that 1(x) is separable.
The proof of the existence of an algebraic closure K of F in §8.9
relied upon a double induction argument in the construction of the fields
 Perfect Fields 359

K1. We now indicate that such an argument is unnecessary if simple facts

concerning inseparable extensions and perfect fields are used. However it
should be noted that the proof in §8.9 does not require the concepts of
normality, separability, and the Theorem of the Primitive Element.
Let K1 be the field F[(X1}]/M of §8.9 and F be the totality of all
elements A e K1 which are purely inseparable over F; that is, for which E F
for some u 0. Then F is a field. For e F, the elements (A + B)', (A B)'
belong to F where r =

Lemma 2. The field F/F is perfect.

Proof Consider C e F; then e F by the definition of F. The polynomial
— of F[x] has a zero D in K1 by the construction of K1 in §8-.9.
Hence e F and D E F. Consequently = 0 in F, and thus
= C. Finally, Lemma I implies that P is a perfect field.

Lemma 3. The field K1/F contains an algebraic closure of P.

Proof The proof of Lemma 3 consists of showing that an irreducible poly-
nomial g(x) in P[x] factors completely in K1[xj. Let g(x) have roots
B1,..., and consider a splitting field S = F(B1, ..., of y(x). Since g(x)
is separable by Lemma 2, the Theorem of the Primitive Element yields
the existence of w S, such that S = F(w). Denote by j(x) the minimal
polynomial of w in P [x].
The p'th powers of the coefficients off(x) lie in F for some e 0 by
the definition of the field P. Hence lies in F[x] and has a zero A in
K1 according to the construction of K1. In other words, f(x) e F[x] has a
root A e K1. Since A and w are roots of the same irreducible polynomial
,f(x) F[xJ, there is an isomorphism over P
A: f'(w) F(A), P(A) c
Since the zeros B1 of g(x) are linear combinations (over P) of powers of w,
they correspond to linear combinations 2(B1) (over F) of powers of A. The
elements 2(B1) E P(A) are roots of 1(g(x)) = g(x). In other words, g(x) =
fl7,j(x—A(B1)), and the polynomial g(x) splits completely in P(A)[xJ c
K1 [x]. Consequently the field K1 contains an algebraic closure ofF as asserted.

Proposition. The field K1/F contains an algebraic closure of F.

The proof is obtained by combining Lemmas 2 and 3. As our final

statement we prove the following theorem.

Theorem. For each field F of nonzero characteristic p there exists a field

which is perfect and algebraic over F. Further, none of its proper
subfields containing F is perfect. The field is uniquely determined to
within isomorphisms over F.
 Selected Topics in Field Theory chapter 9

Proof The field P of Lemma 2 is perfect and, by construction, algebraic

over F. Suppose that L were a proper perfect subfield of P with L F. Then
there would exist an element a e P\L for which = c F. Observe that the
equation x" — = 0 has a solution in L, because L is perfect by assumption.
Next = has a root A2 EL for the same reason. Hence, by
induction, = has a root in L. Consequently,
= = a"
implies that (Ae0)" = 0
and that a = 6 L, a contradiction.
Finally let f'1/Fhe another perfect extension satisfying the hypotheses
of the theorem. Suppose that P1 is embedded in an algebraic closure F1/F
and that P is embedded in an algebraic closure F/F. The fields F1 and F are
isomorphic over F [from §8.9]. and this isomorphism maps F1 onto F, as both
fields have the property that some power of each of their elements lies in F.
Q.E.D.

Exercises

I. Prove that F(x) is not a perfect field if charF= p > 0.

2. Determine the smallest perfect field containing
3. Assume that the field F, char F = p> 0, is not perfect. Prove that the
sets
= s = 1,2,...
are distinct fields, where denotes a zero of E F[x}.
4. If the degree of the algebraic closure F over F of the field F is a prime
prove that F is a perfect field.
5. Making the same assumption as in Exercise 4, show that char F it.
6. Construct the smallest perfect algebraic extension of K F(x), called
the perfect closure of K, where char F p > 0. (Hint: Examine the fields
of Exercise 3. Denote by
7. Prove that an algebraic extension K of a perfect field F is perfect, where
charF=p >0.
8. Subfields K1, K2 of a field M are said to be linearly disjoint over a com-
mon subfield F if every finite set of linearly independent (over F) elements
in K1 remain linearly independent when considered over K2 as the field
of coefficients.
Let K/F be a subfield of an algebraic closure fl/F. Prove that
K/F is separable if and only if K and are linearly disjoint.
9. a. Prove that and are linearly disjoint fields over Q(i),
i2 = -1.
b. Are the fields and Q(w linearly disjoint over Q, where
=0?
§9.12 Perfect Fields 361

10. Let a be an element in a field F of characteristic p > 0, which is not a

pth power of an element in F. Prove that the polynomials — a e Fix],

a e N, are irreducible.
11. Prove that the definition in Exercise 8 is symmetric in K1 and K2. Hint:
Show that the definition is equivalent to the statement that if x1, e
K1 are linearly independent over F and if yi, e K2 are linear
.

independent over F, then the products x, I I n, I j m, are

linearly independent over F.

12. Let (a1, be a set of elements in an algebraic extension fl/F, and

suppose L fl. Prove that:
a. [L(a1 am):L] [F(a1 am):F].
b. Equality of the degrees in part (a) holds if and only if the fields L/F
and F(a1, ., a,,)/F are linearly disjoint with respect to F.
13. Prove that an element A in an algebraic extension K/F belongs to F if it
is simultaneously purely inseparable and separable over F.
14. Prove that a field F of nonzero characteristic p is perfect if and only if
P = F.
15. Let i' be a perfect closure of a field F, where char F = p > 0. Show that
either P = F or P does not have finite degree over F.
16. Let Cl be an algebraic closure of a field F, where char F = p> 0. Prove
that an element A E Cl is left unaltered by any automorphism of Cl/F if
and only if 4" n F for some m e N.
17. Prove that an algebraic element A over a field F, charF= p > 0, is

separable over F if and only if F(A) = F(A").

18. Prove that when charF=p>0:
a. If K/F is a separable (algebraic) extension, then F(K') = K.
b. If F(K") = K and [K: F] < then K/F is separable, where
K" = (a": an Kl.
19. Let K/F be a separable algebraic extension of characteristic p> 0. Prove
that {A/: An A) is a basis of K/F if {AA : A e A} is a basis of K/F.
______

Bibliography

The following bibliography lists books of interest in the history of

mathematics related to algebra, additional readings on material covered in the
text, and suggested readings for students wanting to pursue subsequent
algebraic topics.

ALBERT, A.A., Structure of Algebras, Amer. Math. Soc. Coil. Pubi. XXIV, 1939.
ARTIN, E., Galois Theory, Notre Dame Mathematical Lectures, No. 2, University
of NoIre Dame, 1948.
Theory of Algebraic Numbers, Göttingen, Germany, 1959.
- , C.J., THRALL, R.M., Rings with Minimum condition, University
of Michigan Press, Ann Arbor, Mich., 1944.
BELL, E.T., The Development of Mathematics, second edition, McGraw-Hill
Book Co., New York, 1945.
Men of Mathematics, Simon and Schuster, New York, 1937.
BIRKHOFF, G., see MACLANE, S.
BOYER, C., A History of Mathematics, John Wiley & Sons, New York, 1968.
CASTELNUOVO, G., Sulla razionalità delle involuzioni plane, Math. Ann. 44 (1894),
125—155.
CURTIs, C., Linear Algebra: An Introductory Approach, third edition, Allyn and
Bacon, Boston, 1974.
DICKSON, L.E., Modern Elementary Theory of Numbers, The University of
Chicago Press, Chicago, 1939.
ENRIQUES, F., Sopra una involuzione non razionale dello spazio, Atti Accad. naz.
Lincei, Rend. V.s. 21' (1912), 81—83.
EvES, H., Introduction to the History of Mathematics, third edition, Holt, Rinehart
and Winston, New York, 1969.
FULTON, W., Algebraic Curves, W. A. Benjamin, New York, 1969.
GREUB, W.H., Linear Algebra, Springer-Verlag, New York, 1967.
Mullilinear Algebra, Springer-Verlag, New York, 1967.
GROSSWALD, E., Topics from the Theory of Numbers, The Macmillan Co., New
York, 1966. (For a succinct discussion of the Fermat Conjecture, see pp.
159—182.)

363
_______

Bibliograpby

Macmillan Co., New York, 1959.

HALL, M., The Theory of Groups, The
HALL, TORD, Carl Friedrich Gauss: A Bibliography (translated by Albert
Froderberg), MIT Press, Cambridge, Mass., 1970.
HALMOS, P., Naive Set Theory, D. Van Nostrand Co., Princeton, N.J., 1960.
HERSTEIN, I.N., No,zcommulative Rings, Carus Mathematical Monographs, Number
15, The Mathematical Association of America, Washington, D.C., 1968.
HILBERT, D., Collected Works, Vol. 1, Number Theory, Verlag von Julius Springer,
Berlin, 1932.
Ueber die Irreducibilität ganzer rationaler Funk lionen mit ganzzahligen
Coefficienten, Jour. f.d.r.u.a. Math. 110 (1892), 104—129.
HOCHSCNILD, C., The Structure of Lie Groups, Holden-Day, San Francisco, 1965.
HOFFMAN, K.M., and KUNZE, R.A., Linear Algebra, Prentice-Hall, Englewood
Cliffs, NJ., 1971.
JACOBSON, N., Lectures in Abstract Algebra, Vols. 1—3, D. Van Nostrand Co.,
Princeton, N.J., 1951, 1953, 1964.
KLEIN,F., Lectures on the Icosahedron and the Solution of Equations of f/se Fifth
Degree, 1884, second and revised edition, Dover Publications, New York, 1956.
KUNZE, R.A., see HOFFMAN, K.M.
KURO5H, A.G., The Theory of Groups (translated from the Russian and edited by
K.A. Hirsch), second English edition, 2 vols., Chelsea Publishing Co., New
York, 1960.
LANDAU, E., Grundlagen der Analysis, Chelsea Publishing Co., New York, 1946.
LANG, S., Introduction to Algebraic Geometry, Interscience, New York, 1958.
Algebraic Numbers, Addison-Wesley Publishing Co., Reading, Mass., 1964.
Algebra, Addison-Wesley Publishing Co., Reading, Mass., 1965.
Algebraic Number Theory, Addison-Wesley Publishing Co., Reading,
Mass., 1970.
LEDERMANN, W., Introduction to the Theory of Finite Groups, Oliver and Boyd,
London, 1961.
MACLANE, S., and BIRKHOFF, 0., Algebra, The Macmillan Co., New York, 1967.
MAXFIELD, J.E., and MAXFIELD, M.W., Abstract Algebra and Solution by Radicals,
W.B. Saunders Co., Philadelphia, 1971.
NESBITT, i.E., see ARTIN, E.
NEWMAN, J.R., Editor, The World of Mathematics, Vols. 1—4, Simon and Schuster,
New York, 1956-1960.
NoamcoTr, D.C., An Introduction to Homological Algebra, Cambridge University
Press, London, 1960.
O'MEARA, O.T., Introduction to Quadratic Forms, Springer-Verlag, Berlin, 1963.
ORE, 0., Number Theory and its History, first edition, McGraw-HiH Book Co.,
New York, 1948.
PERRON, 0., Algebra, Vol. H, de Gruyter and Co., Berlin, 1927.
RIBENBOIM, P., Algebraic Numbers, John Wiley & Sons, New York, 1972.
ROTMAN, J., The Theory of Groups: An Introduction, second edition, Allyn and
Bacon, Boston, 1973.
 Bibliography 365

SAMUEL, P., Algebraic Theory of Numbers (translated by Allan J. Silberger),

Houghton-Mifflin, Boston, 1970.
SEIDENBERG, A., Elements of the Theory of Algebraic Curves, Addison-Wesley
Publishing Co., Reading. Mass., 1968.
SERRE, J.P., Corps beaux, Publications de l'Institut de Mathématique de
l'Université de Nancago, VIII, Actualités scientifiques et industrielies, 1296,
1962.
STEINITZ, E., Algebraische Theorie der Körper, Jour. f.d.r.u.a. Math. 137 (1910),
167—309.
STRUIK, D.J., Editor, A Source Book in Mathematics, /200—1800, Harvard
University Press, Cambridge, Mass., 1969.
SUN Tsu, Suan-Ching (trans. Arithmetic), Abh.-Gesch. Math. Wiss. 30 (1912), 32.
THRALL, R.M., see ARTIN, E.
VAN DER WAERDEN, B.L., Modern Algebra, revised edition, 2 vols., Frederick Ungar
Publishing Co., New York, 1969.
VANDIVER, H.S., Fermat's last theorem, its history and the nature of the known
results concerning it, Amer. Math. Monthly 53 (1946), 555—578.
WARING, E., Medilationes Algebraicae, Cambridge, England, 1770.
WElL, A., Foundations of Algebraic Geometry, Amer. Math. Soc. Coil. Pubi.
XXIX, 1946.
WEISS, E., Algebraic Number Theory, McGraw-H ill Book Co., New York, 1963.
WEYL., F!., Mathemarische Analyse des Raumproblems, Verlag von Julius Springer,
Berlin, 1923.
Symmetry, Princeton University Press, Princeton, N.J., 1932.
Wrrr, E., (iber die Kommutativitdt endlkher Schiefkorper, Abh. Math. Semin.,
Hamburg, 8 (193!), 413.
ZARISK1,0., On Castelnuovo's criterion of rationality = P2 = 0 of an algebraic
surface, ill. 3. Math. 2 (1958), 303—315.
and SAMUEL, P., Commutative Algebra, Vols. I and ii, D. Van Nostrand
Co., Princeton, N.J., 1959.
ZASSENHAUS, H., On the fundamental theorem of algebra, Amer. Math. Monthly
74 (1967), 485—497.
 Index of
Mathematicians

Abbati, Pietro 178

Abel, Niels Henrik 163, 310
Albert, Adrian A. 340
al-Haitam, Ibn 51(Ex. 2c)
al-Kashi 21
Apian, Peter 21
Archimedes of Syracuse (9
Argand, Jean Robert 349
Artin, Emil 334

Boyer, Carl B. 42n

Brahmegupta 51(Ex. 2a)
Cardano, Geronimo 304, 310, 314
Castelnuovo, Guido 354
Cauchy, Augustin-Louis 121, 144, 241, 247(Ex. 23a)

Cayley, Arthur 110, 124, 157, 160, 200(Ex. 10)

Chu Shih-chieh 20
Cramer, Gabriel 121

d'Alembert, Jean Ic Rond 348

Dedekind, J. W. Richard 58

367
 Index of Mathematicians

de Moivre, Abraham 318

Descartes, René 4
Diophantos 37(Ex. 19d)

Eisenstein, Ferdinand Gotthold 125, 155

Enrigues, Frederigo 354
Euclid of Alexandria 27, 32, 37(Ex. 21b)
Euler, Leonhard 52, 53, 1 79(Ex. 7), 318, 348—349

Fermat, Pierre de 37, 52, 53, 179(Ex. 7), 305, 318,

334
Ferrari, Ludovico 304, 310
Fibonacci, Leonardo 31, 51(Ex. 2d)
Frobenius, Ferdinand Georg 71(Ex. 14), 90, 247(Ex. 23a), 272
Furtwängler, Philip 330

Galois, Evariste 178, 274, 288—292, 310

Gantmacher, F. R. 157
Gauss, Carl Friedrich 37, 54(Ex. 6), 84, 125, 154,
2l6(Ex. 11), 304, 318,. 341,
34 8—349
Gilmer, Robert 358
Girard, Albert 348
Grassmann, Hermann 260

Hamilton, William Rowan 70(Ex. 14), 121, 124, 157, 160

Heisenberg, Werner 110
Helmholtz, Hermann L. F. von 169
Hensel, Kurt 141
Herbrand, Jacques 263(Ex. 12b)
Herstein, Israel N. 351
 Index of Mathematicians 369

Hubert, David 94, 314, 319, 328, 330, 331

HOlder, Otto 212, 248, 251, 259
L'Hospitai, 0. F. A. de 121

lskovskibh, V. 354

Jacobson, Nathan 322, 340

Jordan, Camille 248, 251, 259

Khayyam, Omar 20
Klein, Christian Felix 189(Ex. 6)
Kowa, Seki 121
Kronecker, Leopold 58, 74, 125, 142, 266, 318
Kummer, Ernst Eduard 58, 305, 334, 338

Lagrange, Joseph Louis 150, 169, 178, 304, 310, 348—349

Lang, Serge 300
Laplace, Pierre-Simon de 121, 123
Legendre, Adrien-Marie 41
Leibniz, Gottfried Wilhelm von 21(Ex. 16), 121, 169
Leonardo of Pisa
(see Fibonacci)
Lie, Marius Sophus 207
LUroth, Jacob 354, 356

Maclaurin, Cohn 121

Manin, Y. 354
370 Index of Mathematicians

Newton, Isaac 21, 141, 169

Noether, Emmy 94, 319, 328, 330

Pascal, Blaise 21
Perron, Oskar 313
Plato 170, 314
Poincaré, Jules Henri 179(Ex. 11)

Roth, Peter 348

Ruffini, Paolo 310

Schmid, Herman Ludwig 314

Schreier, Otto 334
Schur, Issai 263(Ex. 9)
Smith, David Eugene 348
Steinitz, Ernst 103—104
Sun-Tsu 49
Sylow, Ludwig 221, 241—246

Takakusa, Seki
(see Kowa, Seki)
Tartaglia, Nicolo 304, 310

Vandermonde, Alexandre
Theophile 121, 329
van der Waerden, B. L. 310
Venn, John 3
Viète, François 348
 Index of MathematIcians 371

Wallis, John 21
Waring, Edward 153
Wedderburn, J. H. M. 304, 351
Weierstrass, Karl 121
Wessel, Caspar 349
Weyl, Hermann 169, 170
Wilson, John 153
Witt, Ernst 304, 340, 351

Yih-Ling 51(Ex. 2b)

Zariski, Oscar 354

Zassenhaus, Hans 248
Zorn, Max August 266, 299, 300
 Index of
Notation

Symbol Meaning Introduced on

page

Set related and logic symbols:

difference of sets 3

fl intersection (of sets) 3, 4

symmetric difference of sets 3

union (of sets) 3, 4

"implies" 2
"is equivalent to" 2
E "is an element of" 2
proper (strict) containment 2
c containment, with possible equality 2
0 null (empty) set 2

Set operational symbols:

denotes a relation 6
denotes an equivalence relation 6
"is not in relation to" 6
isomorphic 66, 186
RxS cartesian product of sets R, S 4
internal direct sum 72
-I- external direct sum 71
internal direct product 213

373
374 Index of Notation

x external direct product 214

denotes composition of mappings 194

Labels for speczfic sets:

C set of complex numbers 2

N set of natural numbers 2

Q set of rational numbers 2

R set of real numbers 2
set of n-tuples of real numbers 5

Urn (multiplicative group of) Units in Zm 52

Z set of integers 2

Zm residue class ring of integers modulo m 39, 41

Miscellaneous symbols:
{ } denotes a set 2
group generated by g 186
(m) ideal generated by m 24
set of conjugates of g 238
ba "b divides a" 25
b4'a "b does not divide a" 26
IK restriction of mapping a to K 277
congruence 38
Ia! absolute value of a e C 15
[sJ equivalence (congruence) class of s 6
(a,b) GCD of a,b e Z (equivalently, ideal in Z
generated by a, b) 24, 26
[a,b] LCMOfa,bEZ 34

order of an element g in a group 173

[<g> e] }

order of the groupG 177

[Ge])
[G:H] index of the subgroup H in the group G 178
[K: F] degree of the extension field K/F 269
[K: F]1 degree of inseparability of K/F 283
[K: degree of separability of K/F 283

Alphabetized notation:
alternating group on n elements 200 (Ex. 6c)
AG(L) annihilator of a subset L in a group G 232
 Index of Notation 375

'A transpose of the matrix A 112

Adj(A) adjoint of the matrix A 124
Af(Zm) affine group over Zm 171 (Ex. 10)
Aut(G) group of automorphisms of the group G 205
C (commonly) center of a group 177 (Ex. 10),
207
centralizer of the subgroup H in the group 6 183 (Ex. 12)
(multiplicative) cyclic group of order n 186
Card(S) cardinality of the set S 177 (Ex. 13)
char F characteristic of the field F 88
degf(x) degree of the polynomialf(x) 128
det(A) = IAI determinant of the matrix A 121
dim V= dim,V dimension of the vector space V over the
field F 104
EndF(V) ring of endomorphisms of the vector space
VoverthefieldF 108
Fmn set of m x n matrices with components in
thefieldF 111
F[x) polynomial ring in the indeterminate x over
the field F 127
F(x) field of rational functions (field ofquotients)
in the indeterminate x over the field F 128
G' derived (or commutator) group of the 207, 254
group 6 (Ex. 9)
6* the dual group of the group G 230
G(K/F) galois group of the field extension K/F 273, 286
GIN quotient (factor) group 181
gH (left) coset of H by g in a group 171
Hg (right) coset of H byg in a group 171
GL(n, R) general linear group 208
GCD greatest common divisor 26
HK (M) stabilizer of the subgroup M in a group with
respect to the subgroup K 237
HK(s) stabilizer of the element s in a group with
respect to the subgroup K 236
Horn (6, r) group of homomorphisms G —+ f
where
6, rare abelian groups 217
HomF(U, V) vector space of linear transformations
376 Index of Notation

'p: U V of vector spaces U, V over the

field F 107

n xn identity matrix 113

1(G) group of inner automorphisms of the

groupG 206

lmq = image of a homomorphism 4) 63, 185

K/F extension Kofthe field F 266

ker 'p kernel of the homomorphism 'p 67, 185

modm modulo the integer rn 38

LCM least common multiple 34

ma(x) minimal polynomial of a e K/F 267

Mq, matrix representing the linear transform-
ation4) 114

N(a) norm of the field element a 320

N(K) = NG(K) normalizer of a subset (subgroup) K in the
groupG 176 (Ex. 8)
OrbG(s) orbit of the element s with respect to the
group G 236, 237
PL(2,F) projective linear group 184 (Ex. 19),
356
Q(D) field of quotients of the integral domain D 85
Radii radical of the ideal A 61 (Ex. 8)
SL(2, F) special linear group 275 (Ex. 6)
sgn (it) sign or signature of the permutation it 122, 200
(Ex.6)
Span(S) vector space generated by the set S 98
T(a) trace of the field element a 320
T(A) trace of the matrix A 158
T('p) trace of the linear transformation 'p 160

tJFD unique factorization domain 342

Alphabetized greek notation:

XA (x) characteristic polynomial of the matrix A 158

characteristic polynomial of the field
element a 320
characteristic polynomial of the linear
transformation 'p 159
 index of Notation 377

Kronecker delta 74
dihedral group of order 2n 168 (Ex. 6f)
group of automorphisms of G which leave
fixed the elements of the intermediate field L 288
fixed field of a subgroup /1 of a group G 288
cyclotomic polynomial of degree p(n) 305
q(rn) (usually) Euler of rn Z 52
(usually) inner automorphism by x 206
symmetric group on n elements 194
s(S) group of permutations of the setS 194
(commonly) primitive nth root of unity 230
w (commonly) a primitive cube root of unity 271 (Ex. 7c)
 Index of
Mathematical Terms

A Affine group, 17l(Ex. 10), 254(Ex.

5a, 6), 341(Ex. 7)
Abelian extension, 292(Ex. Ic), Algebraic closure (see also Closure):
335—337, 339 of fields, 268, 301
galois group of, 340 separable, 303
Abelian groups, 163, 241 Algebraic element, 267—269
annihilator of, 232 characteristic polynomial of, 320,
characters of, 230, 231 321
dual, 230 conjugate, 277, 278, 321, 325—326
dually paired, 235(Ex. 7) degree of, 267
finitely generated, 222 different of, 328
free, 229(Ex. 12, 18) discriminant of, 328
Fund. Th. of Fin. Gen., 222-227 minimal polynomial of, 267, 321
homomorphisms of, 217—219 norm of, 320—324
order of, 241 separable, 270
rank of, 229(Ex. 12) trace of, 320—324
Absolute value, 15 Algebraic extension (see also Field
Addition, II, 127 extension(s)):
Additive inverse: equivalent, 277
for integers, 11—13 example of, 143—144
for rings, 56, 57 finite, 269
for vector spaces, 95 primitive element of, 275
n-adic expansion, 29, 30, 141(Ex. 15) of Q, 286
example of, 29, 42 simple, 275
of integers, 29, 41—42 Algebraically closed, 300
of polynomials, 141(Ex. 13—15) Alternating group, 200(Ex. 6c),
Adjoint matrix, 124 252(Ex. 3)
Adjunction of elements, 269 not solvable, 253
Admissible homomorphism, 257 simple, 253, 254(Ex. l3)
isomorphism theorem of, 257 Annihilator:
kernel of, 257 double, 232
[i-admissible, 257 of an ideal, 70(Ex. 9)
Admissible subgroup, 256 of a subset of a group, 232
normal, 256—257 Archimedean Principle, 19,91(Ex. 23b)

379
380 Index of MathematIcal Terms

Associates, 342 for groups, 164

Associativity, 6 for an integral domain, 83—84
for coset arithmetic, 41 for multiplication, 11, 14,56
generallawof, 11, 57, 164—165 for rings, 57, 83—84
for groups, 163 for Z, Ii
for ideals, 62(Ex. 11) for Zrn, 38, 41
for integers, ii
for matrices, I 19(Ex. 3) C
of permutations, 194
for rings, 56, 57 Canonical projection, 68, 139, 185,
for vector spaces, 95 201
Automorphism(s) (see also Cardinality, 177(Ex. 13), 236
Endomorphism; Isomorphism; Cartesian plane, 5, 9(Ex. 12), 96(Ex. I),
Automorphism group): 314
of C, 67(Ex. 13) Cartesian product, 4, 91 (Ex. 21a)
of Q, 280(Ex. 2) 166(Ex. 5), 258
of cyclic groups, 206(Ex. I) nonassociativity of, 6(Ex. 6)
of a field, 274, 277, 286 Casting out nines, 42(Ex. 2)
Frobenius, 90, 274 Cauchy Convergence Criterion,
of a group, 186, 205, 229(Ex. 20) 36(Ex. 6b)
group of inner, 207, 238—239, 256 Cauchy's Theorem, 241, 247(Ex. 19)
inner, 206 Cayley's Theorem, 200(Ex. 10)
of a ring, 67 Cayley-Hamilton Theorem, 124, 157,
Automorphism group, 205, 21 1(Ex. 2) 160—161, 320
(see also Galois group) Center:
of Z, 212(Ex. 7) of dihedral groups, 209(Ex. 3)
of cyclic groups, 206(Ex. I) of a division ring, 351
of r4, 212(Ex. 9) of a group, 177(Ex. 10), 179(Ex. 2),
of Klein's Four-group, 208(Ex. 2) 183(Ex. 3), 207, 210—211(Ex. 6),
of quaternions, 212(Ex. 8) 239
Centralizer, 183(Ex. 12)
Chains of groups, 250, 254(Ex. 9)
B (see also Composition series)
isomorphic, 250
Basis: normal, 250, 251
normal, 326 refinement of, 250
of K.", 120(Ex. 12) Characteristic (of a field), 88
ofa vector space, 102—104, 111, 112 Characteristic polynomial:
Bijections (see also Isomorphism; of a field element, 320, 321
Permutations): of a linear transformation, 160,
of sets, 235—236 264(Ex. 18)
Bijective mapping (see Isomorphism) of a matrix, 125, 157, 158, 160—161
Binomial coefficients, 20—21(Ex. 8—10), Characters (see Abelian groups)
36 (Ex. 12) Chinese Remainder Theorem, 10,
Binomial Theorem, 19, 20(Ex. 9), 53,147 48—51, 81, 325
Butterfly Lemma, 248 Class equation, 238, 240
Cancellation, Law of: application of, 239, 242, 352—353
for addition, 12, 57 Closure:
for congruences, 38, 39(Ex. 5), 41 additive, 15
 Index of Mathematical Terms 381

algebraic, 268, 301, 359 Conjugate field elements, 277, 278,

integral, 344 321, 325—326
multiplicative, 15 Conjugate field extensions, 277, 278,
perfect, 360(Ex. 6) 281, 289
separable, 303 example of, 281
Cofactor, 124 Conjugate group elements, 179(Ex. 2),
Commutative diagram, 201 238, 239, 281—283
Commutativity: Conjugate roots of a polynomial,
for coset arithmetic, 41 270, 281
for groups, 163 Conjugate subgroups, 176(Ex. 6),
for ideals, 62(Ex. II) 239, 289
for integers, II Constructions with ruler and compass,
for rings, 56, 57 54(Ex. 5, 6), 314—318
for vector spaces, 95 Content, 343
Commutator: p-content, 343
of group elements, 179(Ex. 3), 207 Coset(s), 1, 7 (see also Residue
of permutations, 199 classes)
Commutator subgroup, 207 double, 247(Ex. 23a)
higher, 254(Ex. 9, tO) examples of, 7
Complete inverse image, 66, 67, 202 of integers, 38
Complex of products, 176, 180 left, 171
Complex numbers, 2, 144 mapping of, 62—63, 205
of absolute value 1, 1 66(Ex. 3a) of polynomials, 137
automorphisms of, 67(Ex. 13) product of, 40—41
as a residue ring, 144, 349 representative of, 39, 172
Composite number, 32 right, 171
Composition series, 248, 250—252 of a subgroup, 171
(see also Solvable group) sum of, 40—41
of factor groups, 254(Ex. 10), Coset decomposition, 177—178,
255(Ex. 19) 247—248(Ex. 23)
of groups, 250 Cycle(s):
length of, 250, 255(Ex. 19), 260 commutator of, 199
of modules, 259 r-cycle, 197
of subgroups, 253(Ex. 2a), 255(Ex. disjoint, 197
19) even, 200(Ex. 6)
Congruence(s): examples of, 198—199
general systems of, 48—51 inverse, 197
modulo m, 37—38 length of, 196
of polynomials, 136, 137 notation for, 195
solutions of simultaneous systems, odd, 200(Ex. 6)
48-51, 52(Ex. 6), 138—139 order of, 200(Ex. 5)
solvability of, 40(Ex. II, 12), 43, product of, 195
46, 52(Ex. 6), l40(Ex. 3) transposition, 196, 198
Congruence classes, 1, 38 Cyclic group, 19 1—193, 214, 219(Ex. 3),
Congruence relations, 38, 176(Ex. Sa) 272—273
Conjugacy class, 238 (see also automorphisms of, 206(Ex. 1)
Equivalence class; Equivalence dual group of, 230
relations) generators of, 186, 192
Conjugacy relations, 236 homomorphic image of, 192
382 Index of Mathematical Terms

number of generators, 192 Direct sum:

subgroup of, 191 external, 72
sufficient condition for, 192 of groups, 213
Cyclotomic extension, 305 of ideals, 72
Cyclotomic fields, 305 internal, 72
examples of, 307-308 of rings, 72
galois groups of, 307 of vector spaces, 98, 99, 264(Ex. 18)
Cyclotomic polynomial, 157(Ex. 8a), Discriminant, 327, 328
305, 308(Ex. 2, 3), 352—353 Distributivity:
for coset arithmetic, 41
general law of, 20(Ex. 6, 7), 57,
D
70(Ex. 11), 126
Defining relations (of a group), 167 for ideals, 62(Ex. 11)
Degree: for infinite vectors, 126—127
of an algebraic element, 267 for integers, II
of a field extension, 269 for matrices, 1 19(Ex. 4)
of inseparability, 149, 283 for polynomial functions, 126
of a polynomial, 128 for rings, 56, 57, 70(Ex. II)
reduced, 149, 283 for vector spaces, 95
of separability, 283 Divisibility, 25, 32, 33, 130—132
of a transcendental element, 354 Division Algorithm (see also
Delian Problem, 317 Euclidean Algorithm):
de Moivre's formula, 318 application of, 24, 134, 174, 191
Derivative (see Formal derivative) for integers, 21—23
Derived group (see Commutator for polynomials, 130, 131
subgroup) Division ring, 351
Determinant, 121—122, 158 (see also center of, 351
Norm) finite, 351
Laplace expansion by minors, 123 Divisor, 26, 131, 342 (see also
of a linear transformation, 160 Greatest common divisor)
properties of, 122—123 Divisor of zero (see Zero divisor)
Vandermonde, 329 Domain:
Different (of an element), 328 integral (see Integral domain(s))
Dihedral group, 168(Ex. 61), of operators, 255, 256
169(Ex. 7), 174—175, 252(Ex. 2) principal ideal, 131, 347(Ex. 2)
automorphisms of, 212(Ex. 9) unique factorization, 342
center of, 209(Ex. 3) Double coset, 247(Ex. 23a)
commutator subgroup of, 210(Ex. 4) Doubling the cube, 317
quotient group of, 210(Ex. 4) Dual group, 230—233, 336, 337
solvability of, 252(Ex. 2) annihilators, 232
Dimension: double, 231
of a field extension, 269 Dual space, I lO(Ex. 14), 235(Ex. 5)
of a vector space, 104, 105(Ex. 11),
260
E
Diophantine equation, 37(Ex. l9d)
Direct product: Eisenstein's Criterion, 125, 155—156,
external, 213, 214 308(Ex. 2)
of groups, 212—215 Elementary symmetric functions,
internal, 213 290, 346
 Index of Mathematical Terms 383

algebraically independent, 346—347 Equivalence classes, 1, 7

Theorem of, 347 disjoint, 7
Elements (in a domain): Equivalence of field extensions, 277,
associated, 342 280—283
divisor. 342 Equivalence relation(s), 1, 6—8
integral, 344 examples of, 6, 8(Ex. 10)
irreducible, 342 Euclidean Algorithm, 21, 27 (see also
unit, 46, 61, 85, 13! Division Algorithm)
Elements (in a field): applications of 28—29
algebraic, 267, 361(Ex. 13) Euler 10, 52—54, 81
conjugate, 270, 277, 321, 325—326 applications of, 54(Ex. 5), 192
degree of, 267, 354 properties of, 54(Ex. 3,4),
inseparable, 283 83(Ex. 11,12)
primitive, 275 Euler's Theorem, 179(Ex. 7)
purely inseparable, 359, 361(Ex. 13) Exact sequence, 260, 262(Ex. 5),
separable, 270, 284, 361(Ex. 13) 264(Ex. 15,16)
transcendental, 271(Ex. 6), 354 Expansion by minors, 123
Elements (in a group): Exponent:
commutator, 179(Ex. 3), 199, 207 of a field extension, 335
conjugate, 179(Ex. 2), 238 of a group (see Minimal exponent)
generators, 186, 222 of inseparability, 149
identity, 163 minimal (see Minimal exponent)
orbit of, l84(Ex. 15), 238 of nilpotency, 43
order of, 173, 174, 178, 186 Exponents, Law of, l67(Ex. 6a),
self-conjugate, 179(Ex. 2), 238 170(Ex. 4)
torsion, 225 Extensions (see Field extensions and
Elements (in a ring): Isomorphism)
idempotent, 45, 61
identity, 56, 57
nilpotent, 43, 61 F
orthogonal idempotent, 73, 138—139
(see also Orthogonal Factor group (see Quotient group)
idempotent) Factorizat ion:
unipotent, 61(Ex. 5) of integers, 33
Unit, 46, 61 (see also Unit(s)) in integral domains, 342—344
zero divisor, 43, 45, 46(Ex. 10), 61,84 of polynomials, 132
Elements (in a set): unique, 33, 132, 342
equivalence classes of, 7 Fermat's Last Theorem, 334
least, 16 Fermat's Little Theorem, 10, 52—53,
maximal, 300 69(Ex. 6), 179(Ex. 7)
upper bound, 300 Fermat primes, 318
Embedding, 67, 85 Fibonacci sequences, 31(Ex. 15)
Endomorphism(s), 67, 160 Field(s), 85—90, 153 (see also Field
ringof,57, 108, 116, 220(Ex. Ii) extension(s); Subfield)
Epimorphism (see Surjective mapping) algebraically closed, 300, 353(Ex. 1),
Equality: 358
of mappings, 194 characteristic of, 88
of polynomials, 127 composite of, 293
of sets, 3 conjugate, 277, 289
384 Index of Mathematical Terms

cyclotomic, 305 Formal derivative, 146—148

examples of, 84—85, 142 Formal power series, ring of,
finite, 271—274, 303(Ex. 1), 313 129(Ex. 6)
fixed, 274, 288 Frobenius automorphism, 90,
ground, 266 272—274, 333
intermediate, 274, 282, 288 Fundamental Theorem of:
intersection of, 293 Algebra, 56, 125, 145, 266, 341, 348—
perfect, 358, 359, 36l(Ex. 14) 351
prime, 88—89 Arithmetic, 33, 48
product of, 293, 296 Elementary Symmetric Functions,
quotient, 85—88, 128 341, 347
of rational functions, l28--129 Finitely Generated Abelian Groups,
separable, 270, 284 163, 222—227, 264(Ex. 18d)
separable closure, 303 Galois Theory, 266, 271, 284,
splitting, 145, 270, 280, 285 289—291, 292
Field extension(s), 266
abelian, 292(Ex. Ic), 335—337, 339
algebraic, 143—144, 267, 269, G
277(Ex. 2), 301, 311—312, 318
algebraic closure, 268, 301 Galois extensions, 285, 330
conjugate, 277, 278, 289 abelian, 292(Ex. Ic), 335
cyclic, 292(Ex. 2), 311, 331, 332, cyclic, 292(Ex. 2), 331, 332, 334,
334, 335 335
cyclotomic, 305 Galois groups, 286, 346, 355
degree of, 269 of abelian extensions, 335—337, 339
equivalent, 277, 278, 289 example of, 286—287, 307—308
exponent of, 335 for finite fields, 272—274
galois, 285, 289, 296, 311, 330, 346 solvable, 312—313
inseparable, 285 Galois Theory, 288—292
Kummer, 335 for finite fields, 274
normal, 285, 289, 332 Gauss, Lemma of, 125, 343
quadratic, 316-318 Gaussian integers, 84(Ex. 4)
radical, 335 General associative law, 11, 57,
separable, 270, 277(Ex. 2), 284, 164—165
285, 330 General distributive law, 20(Ex. 6,7),
simple, 275 57, 70(Ex. Ii), 126
splitting, 145, 270, 280, 281, 285 General linear group, 208—209(Ex. 2)
transcendental, 355 Generalized quaternion group,
Field homomorphism, 92(Ex. 31) l67—168(Ex. 6e)
Finite fields, 27 1—274 Generator:
examples of, 144 of a group, 167, 186, 222
Frobenius automorphisms of, 272 of an ideal, 24, 59
generator of, 272 Grassmann's relation, 105(Ex. 11),
multiplicative group of, 271—272 205(Ex. 6), 260
prime field of, 88 Greatest common divisors:
subfield, of, 274 computational examples, 28—29
units of, 271 of ideals, 35(Ex. Sc), 59, 70(Ex. 10)
Fixed field, 274 of integers, 26, 27, 35
Fixed point, 237(Ex. 4c) in integral domains, 342
 Index of MathematIcal Terms 385

of polynomials, 131 Sylow (see Sylow subgroups)

uniqueness of, 27 symmetric, 194 (see also Symmetric
Group(s), 162—234 (see also Abelian groups)
Affine group; Groups, of symmetries (see Symmetric
examples of; Homo- groups; Symmetries)
morphism(s); Subgroup(s)) torsion, 225
abelian, 163 torsion free, 225
affine, 171(Ex. 10) Groups, examples of, 52, 165—169,
of automorphisms, 205-206 182, 186—188, 210—211 (see a/so
center of, 177(Ex. 10) (see also Automorphism group)
Center) affine group, 171(Ex. 10) (see also
of characters, 230 Affine group)
commutative, 163 alternating, 200(Ex. 6), 252—253,
commutator, 207 254(Ex. 13)
cyclic, 167(Ex. 6a,b), 186, 191—192 cyclic, 167(Ex. 6a,b), 188 (see also
defining properties of, 163-164 Cyclic group)
defining relations of, 167—168 dihedral, 168(Ex. 6f), 169(Ex. 7)
derived, 207 (see also Commutator (see also Dihedral group)
group) general linear, 208—209(Ex. 2)
direct product of, 2 13—214 generalized quaternion,
double dual, 231 167—168(Ex. 6e)
dual, 230—233 group of the square, l69(Ex. 7b)
dually paired, 235(Ex. 7) group of the triangle, I 69(Ex. 7a),
exponent of (see Minimal exponent) 172—174, 201(Ex. 11)
factor, 181 Klein's Four-group, 189(Ex. 6b)
finite, 177 (see also Klein's Four-group)
Fund. Tb. of Finitely Generated of matrices, 166(Ex. 4)
Abelian, 222—227 of order p2. 240(Ex. 4)
galois, 274 (see also Galois group) of order ptm, 248(Ex. 24,26,27),
generated by elements, 167—169 251—252, 254(Ex. 11),
(Ex. 6a—i) 255(Ex. 18)
generator(s) of, 167(Ex. 6a), 186, oforderpq,247(Ex. 12, 16)
222 of permutations, 194
of homomorphisms, 217—218 projective linear, 184(Ex. 19), 355,
identity in, 163—164 356
inverse in, 163, 164, 170(Ex. 1) quaternion, 167—168(Ex. 6e),
isomorphism theorem of, 201-202 189(Ex. 6) (see a/so
law of exponents for, l67(Ex. 6a), Quaternion group)
170(Ex. 4) semidihedral, 168(Ex. 6h)
minimal exponent of, 179(Ex. 6), special linear, 275(Ex. 6)
217(Ex. 12), 225, 335 symmetric, 194, 313 (see also
order of, 177, 241 Symmetric groups)
p-group, 225, 239, 248(Ex. 24) of translations, 254(Ex. 5a)
p-primary group, 225 Groups with operators, 255—258
quotient, 181, 182, 202, 203, 207, (see also Modules)
210, 233 admissible subgroups, 256
simple, 250 examples of, 256, 263(Ex. 12)
solvability property of, 164 Group action, 237(Ex. 4)
solvable, 251, 313 Group characters, 230
386 Index of Mathematical Terms

Group homomorphism (see intersection of, 25, 34, 59

Homomorphism(s)) inverse image of, 67
Group ring, 263(Ex. 10) LCM of, 35(Ex. Sa), 59, 60
left, 59
maximal, 91(Ex. 15)
H of nilpotents, 61(Ex. 7)
of polynomials, 131
Herbrand Quotient, 263(Ex. 12b) power of, 59
Hubert's Theorem Ninety, 328, 331- primary, 9l(Ex. 12)
333, 335, 339 prime, 71(Ex. 17), 91(Ex. 11)
Homomorphism(s). 62—71. 184—191, principal, 24, 75, 131
259 (see also Linear trans- product, 59
formations) proper, 300
of abelian groups, 186—187(Ex. I), radical of, 61
217—219 right, 59
admissible, 257 in a ring, 58
automorphism, 67, 186 semiprime, 91(Ex. 22a)
bijective, 66, 186 sum of, 35(Ex. Sc), 59
canonical, 68, 185 trivial, 23
embedding, 67 two-sided, 59
endomorphism, 67, 221(Ex. 11) Idempotent elements:
epimorphism, 65, 186 irreducible, 74(Ex. 7)
examples of, 64—65, 186—188 orthogonal, 73, 74(Ex. 7), 75
of groups, 185 in a ring, 61
image of, 63, 185 Ifl Zn,, 45—46, 47(Ex. 9)
injective, 66, 186 Idempotent linear transformations,
inverse image of, 66, 67 llO(Ex. 16,17)
isomorphism, 66, 186 Identity element:
kernel of, 67, 185, 257, 259 for coset arithmetic, 41
of modules, 259 for groups, 163—164
monomorphism, 66, 186 for groups of permutations, 194
one-one, 66, 186 for integers, 11—13
onto, 65, 186 for rings, 56, 57, 63
of rings, 62 for vector spaces, 95, 96
surjective, 65, 186 Image (of a mapping), 63, 65, 185,
trivial, 68 187, 188, 192
inverse, 66
of a subgroup, 188
I Indeterminate element (see
Transcendental element)
Ideal(s), 23, 58, 67 Index:
annihilator of, 70(Ex. 9) of nilpoency, 43
dense, 69(Ex. 7) of a subgroup, 178
direct sum of, 72, 73, 82(Ex. 7,8), Index set, 4
83(Ex. 10) Induction, Principle of, 14, 17—19
extended, 93(Ex. 39) alternate form, 17
generators of, 24, 59 application of, Ii, 19, 22, 164—165
GCD of, 35(Ex. Sc), 59, 70(Ex. 10) infinite vectors, 126, 132
of integers, 23—25, 76 Injective mapping, 66, 186
 index of Mathematical Terms 387

Inner automorphism, 206, 207, for rings, 56, 57

238—239, 256 for vector spaces,95, 96
inseparability: Inverse permutation, 197
degree of, 149, 283 Irreducibility Criterion, 155—156
exponent of, 149 Irreducible element of a domain, 342
inseparable: Irreducible polynomial, 132, 148,
extension, 285 149, 155
polynomial, 149 examples of, 305, 308(Ex. 2)
Integers, 2, 14, 92(Ex. 32) (see also Isomorphic chains of groups, 250
Ring of integers) Isomorphic fields, 273, 277, 280
absolute value of, 15 Isomorphic groups, (86
algebraic properties, 10-44, 55 Isomorphic rings, 66
analytic properties, 14—19 isomorphic vector spaces, 106
arithmetic of, Ii Isomorphism:
axioms for, II admissible, 257
composite, 32 extension of, 277
factorization of, 33 of fields, 277, 278, 280—284
GCD of, 26, 27 of groups, 186
ideals in, 23—25 prolongation of, 277, 281, 283
LCM of, 34 restriction of, 277
ordering of, 14—15, 92(Ex. 33—35) of rings, 66—68
prime, 31—33, 37(Ex. 21), 153 Theorem of, 201—202
residue classes of, 38, 41 (see also of vector spaces, 105-106
Residue class rings of integers) Isomorphism Theorem:
solvability in, 13 applications of, 203—205
well-ordering of, 16 for groups, 201—202
integral closure (of a domain), 344, for groups with operators, 257—258
345 for rings, 68, 204(Ex. I)
Integral dependence, 341, 342 for vector spaces, 109(Ex. 13)
Integral domain(s), 14, 57, 83—88
direct sum of, 91(Ex. 21)
divisor in, 25—26, 342 J
examples of, 84, 128, 153
finite, 85 Jordan-Holder Theorem, 201, 251,
GCD in, 26 258, 259
irreducible element in, 31, 342
unique factorization in, 33, 342
well-ordered, 93(Ex. 35) K
Integral elements, 344, 345
Integral Root Theorem, 135(Ex. 9), Kernel (of a homomorphism):
348(Ex. 8a) group, 185, 257
intermediate field, 274, 282, 288 module, 259
Intermediate Value Theorem, 349 ring, 67
Inverse: Klein's Four-group, I 89(Ex. 6b),
for addition, 11—13, 56, 57 201(Ex. 13), 222
for coset arithmetic, 41 automorphisms of, 208(Ex. 2)
for groups, 163, 164 Kronecker, Theorem of, 318
for integers, 11 Kronecker's construction, 125,
for multiplication, 11, 46 142—143
388 Index of Mathematical Terms

Kronecker delta, 74 Matrices, 110—1 14

Kummer, Theorem of. 338 adjoint, 124
Kummer extensions, 334—34 I characteristic polynomial of, 125,
Kummer theory, 334 158
cofactor of, 124
determinant of, 121—124, 158
groups of, 166(Ex. 4)
L inverse of, 117
minor, 123
Lagrange, Theorem of, 178 multiplication of, by scalars, ill
Lagrange interpolation Formula, nonsingular, 117, 120(Ex. 13), 123,
135(Ex. ii), 150—151, 326 159, 161(Ex. 1), 166(Ex. 4a)
Laplace expansion by minors, 123 product of, 112
Lattice diagram, 173 ring of, 57
Leading coefficient, 128 trace of, 158—159
Least common multiple: transpose, 112
of ideals, 35(Ex. 5a), 59, 60 vector addition of, Ill
of integers, 34—35 vector space of, 111—114, 116—117
Linear algebra, 94 Matrix representation:
Linear combinations, 97—98 examples of, 114—115, 118
Linear dependence, 101, 102 of field elements, 3 19—321
Linear functionals, I I0(Ex. 14), of linear transformations, 114, 117,
235(Ex. 5) 159
Linear independence, 101, 102 Maximal element, 300
Linear space, 95 Maximal ideal, 91(Ex. 15), 300
Linear transformations, 105—108, Maximal subgroup, 247(Ex. 22)
169(Ex. 10), 264(Ex, 17) Minimal exponent:
characteristic polynomial of, 160, of a field extension, 335
264(Ex. 18) of a group element, 179(Ex. 6),
composition of, 108, 115 217(Ex. 12), 225, 246(Ex. 2),
determinant of, 160 335
examples of, 106, 109(Ex. 3), 118 Minimal polynomial, 141(Ex. II),
idempotent, I l0(Ex. 16) 267, 321
isomorphism, 105—106 Module(s), 94, 221, 258—261,
kernel of, 109(Ex. 12) 264(Ex. 17,18), 342
matrix representation of, 114, 117, composition series of, 259—260
118, 159 direct sum of, 262(Ex. 5)
orthogonal, 1l0(Ex. 17) finitely generated, 260
ring of, 57, 108 homomorphism of, 259
set of, 107 irreducible, 263(Ex. 9)
trace of, 160 quotient, 259
Luroth's Theorem, 354, 356 R-left, 258
R-module, 259
R-right, 258
sequences of, 261
M unitary, 258
Monic polynomial, 128
Mathematical induction (see Monomorphism (see Injective
Induction, Principle of) mapping)
 Index of Mathematical Terms 389

Multiple roots, 149 Normalizer, 176(Ex. 8,9), 183(Ex. 1,2),

definition of, 148 239, 240
of an irreducible polynomial, 148 Null set, 2, 3
Multiplication:
cancellation, law of, Il, 14, 57
of cosets, 40—41 0
of integers, 11, 13, 14
of natural numbers, 15 One-one mapping (see Injective
rule of signs for, 13, 57 mapping)
by scalars, 95 Onto mapping (see Surjective
Multiplicative identity: mapping)
for groups, 163 Operator domain, 255—256
for integers, 11, 13 Oracle of Apollo, 317
for rings, 56, 57 Orbit, 236, 237
for scalar multiplication, 95 of an element, 184(Ex. ISa), 238
Multiplicative inverse (see Inverse) Order:
Mutually disjoint sets, 4 of an element, 166(Ex. 4c), 173,
174, 264(Ex. 18)
of a group, 177, 241
properties of, 173—174, 178
N Order function:
for polynomials, 343
Natural Irrationality, Theorem on, for rational numbers, 36(Ex. 14—16)
294—295 for ring elements, 342
Natural numbers, 2, 15, 16 Ordered pairs, 4
Natural projection (see Canonical Ordered i,-tuples, 4, 96
projection) Ordering, 9l(Ex. 23)
Newton's formulas, 348(Ex. 10) inductive, 300
Nilpotency: Law of Trichotomy, 15
exponent of, 43 partial, 299
index of, 43 total, 300
Nilpotent elements: well-ordering, 16, 93(Ex. 35)
in a ring, 61 Orthogonal idempotents, 73,
in Zn,, 43—45, 47(Ex. 12) 74(Ex. 7), 75, llO(Ex. 17)
nontrivial, 45, 47(Ex. l2a) examples of, 76, 138—139
Noether's Equations, 330, 332
Norm (see also Determinant):
of a field element, 320-322 P
p-norm, 36(Ex. 16)
Normal basis. 326 Partition(s), 7, 8
Theorem of, 151, 325—326 examples of, 9(Ex. 11,12)
Normal chains: Pascal's Triangle, 2l(Ex. 9)
isomorphic, 250 Perfect closure, 360
refinement of, 250 Permutations, 194—199
Normal field extensions, 285 array notation for, 195
Normal subgroups, 180, 182, 207, commutator of, 199
214, 289 cycle notation for, 195
example of, 182 cycles, 196, 197
direct product of, 213—214 even, 200(Ex. 6)
390 Index of Mathematical Terms

examples of, 195, 197—199 ring of (see Polynomial rings)

group of, 194, 286 root of (see Root of a polynomial)
inverse of, 197 separable, 149, 270
odd, 200(Ex. 6) in several indeterminates, 129
order of, 200(Ex. 5) symmetric, 347
signature (sgn) of, 200(Ex. 6) unique factorization of, 132
transposition, 196 zero, 127
Permutation groups (see Symmetric zero of (see Root of a polynomial)
groups) Polynomial equations:
p-groups, 225, 247(Ex. 22), 248(Ex. 26) cubic, 310, 313, 314(Ex. 1)
center of, 239 of degree greater than or equal to 5,
solvability of, 25 1—252 310, 313
structure of, 248(Ex. 24) quartic, 310, 313
Platonic solids, 170 quintic, 310, 313
Polygons: solvable by radicals, 310—314
constructible, 54(Ex. 5), 317—318 Polynomial functions, 126, 152—153,
regular, 54(Ex. 5,6), 169(Ex. 7) 275(Ex. 7)
Polynomials, I 26—I 55 (see also Polynomial rings, 127—129, 155, 228,
Characteristic polynomial; 268
Polynomial equations) factorization in, 130—132, 343, 344
m-adic expansion of, 141(Ex. 13) ideals in, 131
congruence of, 136, 137 properties of, 129(Ex. 9), 153
constant, 127 in several indeterminates, 129
content of, 343 units in, 140(Ex. 3)
p-content of, 343 Primary ideal, 91(Ex. 13)
cyclotomic, 157(Ex. 8a), 305, Prime field, 88
308(Ex. 2) Prime ideal, 71(Ex. 17), 91(Ex. 11)
degree of, 128 Prime numbers, 3 1—33, 37(Ex. 21),
division algorithm for, 130, 131 153
factorization of, 130—132, 155, 156 Prime residue, 38
GCD of, 131 Prime residue classes, 43, 165(Ex. 2c)
inseparable, 149 Primitive element, 275, 276, 277(Ex. 3)
irreducible, 132, 148, 149, 155, Theorem of, 276, 276—277(Ex. 1)
305, 308(Ex. 2) Primitive polynomials, 154—155
irreducible cubic, 286, 314(Ex. 1) Primitive root of unity, I 89(Ex. 3),
irreducibility criterion for, 155—156 230
leading coefficient of, 128 Principal ideal, 24, 75, 131
of linear transformations, Principal ideal ring, 131, 347(Ex. 2)
264(Ex. 17) Projection, II0(Ex. 16)
minimal, 141(Ex. 11), 267 canonical, 68, 139, 185, 201
monic, 128 onto quotient structure, 68, 139,
order function of, 343 185, 201
prime, 132 orthogonal, I lO(Ex. (7)
primitive, 154—155 Projective linear group, 184(Ex. 19),
properties of, 127—128 355, 356
reduced degree of, 149 Prolongation of an isomorphism,
reducible, 155 277--279, 28 1—283
relatively prime, 132 Proper subset, 3
residue class ring of, 136- 137 Purely inseparable element, 359
 Index of Mathematical Terms 391

ordering, 91(Ex. 23)

partial ordering, 299
Quadratic Law of Reciprocity, 318 reflexive, 6
Quaternions, 353(Ex. 2) (see also symmetric, 6
Quaternion group) total ordering, 300
generalized group of, transitive, 6
167—168(Ex. 6e) Relatively prime integers, 27
ring of, 70(Ex. 14) Relatively prime polynomials, 132
Quaternion group, 167—168(Ex. 6e), Representative of a coset, 39
189(Ex. 6), 193(Ex. 8) Residue classes, 1 (see also Coset)
automorphisms of, 212(Ex. 8) of integers, 38, 41, 137
Quotient field, 85—88, 128 of polynomials, 137
Quotient group, 181, 202, 203, 207 prime, 38
dual group of, 233 Residue class rings, 68, 69(Ex. 3),
example of, 182, 210 82(Ex. 7) (see also Residue
Quotient ring, 68 class rings of integers and of
of polynomials, 128—129 polynomials)
Quotient space, 109(Ex. 13a) Residue class ring of integers, 41—46,
55, 74—77, 139
examples of, 44—45
as an external direct sum, 78—81
R ideals in, 76
idempotents in, 45—46, 47(Ex. 9),
Radical: 75—76
of an ideal, 61(Ex. 8) as an internal direct sum, 78, 79
of a ring, 61(Ex. 7) multiplicative inverses in, 46
Radical extension, 335 nilpotent elements in, 43—45,
Rational functions, field of, 162 47(Ex. 12)
Rational numbers, 2 units in, 46, 52
automorphisms of, 280(Ex. 2) zero divisors in, 43—45
embedding of integers, 85—87 88
geometric constructions of, 315 Residue class ring of polynomials, 137
p-norm of, 36(Ex. 16) examples of, 138—139, 143—144
as a prime field, 88 of irreducible polynomials, 142—144
Rational Root Theorem, 135(Ex. 8), units in, l40(Ex 3)
348(Ex. 8) as a vector space, 140
Real numbers, 2 Restriction of a mapping, 139, 277
geometric constructions of, Ring(s), 14, 56—73 (see also Residue
315—316 class rings; Ring of integers;
Reduced degree (of a polynomial), Subring)
283 axiomatic properties, 56
Refinement, 250 cancellation law, 57, 84
Regular polygon, 54(Ex. 5,6), commutative, 57
169(Ex. 7) of continuous functions, 69(Ex. 8),
constructibility of, 317—318 91(Ex. 20)
Relation: defining properties, 56—57
defining, 167—168 direct sum of, 73, 82(Ex. 7)
equivalence, 6—8 of endomorphisms, 108, 220(Ex. 11)
inductive ordering, 300 examples of, 57, 6l(Ex. Ia)
 index of Mathematical Terms

of formal power series, 129(Ex. 6) Self-conjugate subgroups, 289, 291

group, 263(Ex. 10) (see also Normal subgroups)
ideals in (see Ideal(s)) Semidihedral group, 168(Ex. 6h)
identity in, 56, 57, 63 Semiprime ideal, 91(Ex. 22a)
isomorphism theorem for, 68 Separable algebraic closure, 303
of matrices, 57 Separable element, 270, 284
noncommutative, 57 Separable extension, 270, 277(Ex. 2),
of polynomials, 127, 129 284, 285, 330
principal ideal, 131, 347(Ex. 2) Separable polynomial, 149, 270
of quaternions, 70(Ex. 14) Sequences:
quotient, 68, 91(Ex. II, 13, IS) exact, 260, 262(Ex. 5), 264(Ex. 15, 16)
radical of, 61(Ex. 7) induced, 261
of sets, 6I(Ex. Ia) of modules, 261
subring (see Subring) of module homomorphisms, 261
units of, 61, 131, 165(Ex. 2) Set(s) (see also Subset):
Ring homomorphism, 62 cardinality of, 177(Ex. 13), 236
examples of, 64—65, 67 Cartesian product of, 4
kernel of, 67, 141(Ex. 11) difference of, 3
multiplicative identity under, 63, 65 disjoint, 4
Ring of integers, 14, 57 empty, 2
construction from natural numbers, equal, 3
92(Ex. 32—35) index, 4
ideals in, 23—25, 63—64 inductively ordered, 300
isomorphisms of, 91(Ex. 24) intersection of, 3, 4
Ring of polynomials (see Polynomial maximal element of, 300
rings) null, 2, 3
Root of a polynomial, 133, 267, 320 partially ordered, 299
a-fold, 148 partition of, 7
conjugate, 270, 281 ring of, 61(Ex. Ia)
constructible, 316—318 symmetric difference of, 3
of an irreducible polynomial, totally ordered, 300
148—149 union, 3, 4
multiple, 148 upper bound of, 300
multiplicity of, 148, 149 well-ordered, 16, 93(Ex. 35)
number of, 133—134, 144-145 Set notation, 2
primitive, 305 Short Five Lemma, 26 1—262
Roots of unity, 309(Ex. 4) Simple extension, 275, 355
primitive, 305 Simple group, 250
Rule of signs, 13, 57 Simple vector space, 259
Ruler and compass constructions, Simultaneous systems of congruences:
54(Ex. 5,6), 314—318 examples of, 48—51, 74, 138—139
solvability conditions, 52(Ex. 6)
Solvability axiom, 13
S
Solvability of congruences, 43, 46,
Schur's Lemma, 263(Ex. 9) 52(Ex. 6)
Self-conjugate elements, 179(Ex. 2), Solvable group, 248, 251—253, 312
238 Solvable polynomial, 310—314
Self-conjugate fields (see Galois Span of a set, 98, I00(Ex. 10), 102,
extensions) 175
 Index of Mathematical Terms 393

Special linear group, 275(Ex. 6) Subspace, vector, 97

Splitting field, 145, 270, 280, 285 cyclic, 264(Ex. 17)
construction of, 144-145 direct sum of, 98—99, 264(Ex. 18)
galois group of, 312 generators of, 98
Stabilizer, 236, 237, 239, 240 intersection of, 98
Stable subgroup, 256, 263(Ex. 12) invariant, 157
Steinitz Exchange Lemma, 103—104 p(x)-primary, 264(Ex. 18)
Subfield, 87 sum of, 98
conjugate, 281, 289 union of, 100(Ex. 5)
linearly disjoint, 360(Ex. 8) Surjective mapping, 65, 186
prime, 88 Sylow subgroups, 242
Subgroup(s), 171, 202 (see also conjugate, 244, 245
Sylow subgroups) normal, 244
admissible, 256 normalizer of, 246(Ex. 4,5)
alternating, 200(Ex. 6) number of, 244
annihilator of, 232 Sylow Theorems, 242(Th. 2),
center, 177(Ex. 10), 207 (see also 244(Th. 3), 247(Ex. 23)
Center) Symmetric groups, 194, 286, 313,
centralizer, 183(Ex. 12) 345-346
commutator, 207 alternating subgroup, 200(Ex. 6),
conjugate, 176(Ex. 6), 239, 289 253
of a cyclic group, 191 not solvable, 253
dual group of, 233 solvable, 252
first derived group, 207 Symmetric polynomial, 347
generators of, 175 Symmetries, 169—171
higher commutator, 254(Ex. 9,10)
index of, 178
intersection of, 175, 180, 232 T
invariant, 207
maximal, 247(Ex. 22) Torsion element, 225
nontrivial, 171 Torsion free group, 225
normal, 180, 182, 207, 214 Torsion group, 225
normalizer, 176(Ex. 8) Torsion subgroup, 176(Ex. 7)
order of, 178 Trace:
p-Sylow, 242 of a field element, 320—322, 330
product of, 175, 180, 232 of a linear transformation, 160
stable, 207, 256, 263(Ex. 12) of a matrix, 158—159
stabilizer, 236, 237 Trace function, 158—159 (see also
Sylow, 242 Trace)
torsion, 176(Ex. 7), 225 Transcendental element, 27l(Ex. 6),
Subring(s), 58, 62 354
examples, 58 degree of, 354
identity element in, 58 Transcendental extensions, 355
integrally closed, 345 Transitive relation, 6
nontrivial, 58 Transpose of a matrix, 112
Subset: Transposition, 196, 198
nontrivial, 3 Triangle inequality, 16, 351(Ex. I)
proper, 3 Trichotomy, Law of, 15
trivial, 3 Trisection of an angle, 317
394 Index of Mathematical Terms

U order of an element, 264 (Ex. 18)

polynomial action on, 264 (Ex. 17,
Unipotent element, 61(Ex. 5) 18)
Unique factorization, 342 quotient space, 109(Ex. 13)
of integers, 33—34 ring of endomorphisms, 108, 116
of polynomials, 132 simple, 259
Theorem of, 33 span, 98, 102
UFD, 342—344 sum of, 98
Unit(s): Vector spaces, examples of, 96
group of units in Zm, 52, of complex numbers, 71
165(Ex. 26), 193(Ex. 12), finite fields, 271
206(Ex. I), 216—217(Ex. 11), of linear transformations, 107,
307 113—115
in a ring, 61, 131, 165(Ex. 2a) of matrices, 111—117
in a ring of polynomials, 131 of polynomials, 126
in Zrn, 46, 47(Ex. 10) of quaternions, 70(Ex. 14)
Unitary module, 258 96(Ex. 1), 100
Upper bound, 300 of real-valued functions, 96(Ex. 3)
Venn diagrams, 3—4, 206

V
w
Vectors, 95
as a group, 169(Ex. 9) Wedderburn's Theorem, 304, 351—353
linearly dependent, 101 Well-ordered integral domain,
linearly independent, 101 93(Ex. 35)
Vector addition, 95 Well-ordered set, 16
Vector spaces, 95—108, 256, 267, 269 Well-Ordering, Principal of, 14, 16,
(see also Vector spaces, 18
examples of; Subspaces) application of, 16—17, 23, 24, 33,
bases of, 102—104, 111—112 191
Cartesian n-space, 96(Ex. 1) Wilson's Theorem, 153
dimension of, 104, 105(Ex. II) Wilt vectors, 340
direct sum of, 98, 99, 264(Ex. 18)
dual, I lO(Ex. 14), 235(Ex. 5)
finite dimensional, 104, 107 z
finitely generated, 99, 102
generators, 99, 101 Zassenhaus Lemma, 248
intersection of, 98 Zero (see also Root):
isomorphic, 106 in Z, Il
linear functionals on, I 10(Ex. 14), of a polynomial, 133, 148
235(Ex. 5) Zero divisor, 43, 45, 47(Ex. 10), 61, 84
linearly dependent subset, 101 Zero vector, 95
linearly independent subset, 101 Zorn's Lemma, 300