This action might not be possible to undo. Are you sure you want to continue?
Third Edition
lo N„ Henstein
Late Professor of Mathematics University of Chicago
JOHN W I L E Y & SONS, INC.
N E W Y O R K • CHICHESTER • W E I N H E I M • BRISBANE • SINGAPORE • T O R O N T O
Cover Photograph: Charnley Residence, entryway. Photo by © Nick Merrick/HedrichBlessing This book was typeset in 10/12 Times Ten Roman by University Graphics, Inc. The paper in this book was manufactured by a mill whose forest management programs include sustained yield harvesting of it timberlands. Sustained yield harvesting principles ensure that the number of trees cut each year does not exceed the amount of new growth. Copyright © 1999 by John Wiley & Sons, Inc. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by means, electronic, mechanical, photocopying recording, scanning or otherwise, except as permitted under Sections 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher or authorization through payment of the appropriate percopy fee to the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923, (508) 7508400, fax (508) 7504470. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc. 605 Third Avenue, New York, NY 101580012, (212) 8506008, Email: PERMREQ@WILEY.COM. To order books or for customer service call l800CALLWILEY(2255945). Library of Congress Cataloging in Publication Data Herstein, I. N. Abstract algebra / I.N. Herstein. — 3rd ed. p. cm. Includes index. ISBN 0471368792 1. Algebra, Abstract. I. Title QA162.H47 1995 9521470 512'.02^dc20 CIP Printed in the United States of America. 10 9 8 7 6 5. 4 3 2
To Biska
Preface
ix
U
Things Familiar and Less Familiar
1 2 3 4 5 6 7
1
A Few Preliminary R e m a r k s 1 Set Theory 3 Mappings 8 A(S) (The Set of 11 Mappings of S onto Itself) T h e Integers 21 Mathematical Induction 29 Complex N u m b e r s 32
16
J£M G r o u p s
1 2 3 4 5 6 7 8 9 10 11
40
Definitions and E x a m p l e s of G r o u p s 40 Some Simple R e m a r k s 48 Subgroups 51 Lagrange's T h e o r e m 56 66 H o m o m o r p h i s m s and N o r m a l Subgroups Factor G r o u p s 77 The H o m o m o r p h i s m T h e o r e m s 84 Cauchy's T h e o r e m 88 Direct Products 92 Finite Abelian G r o u p s (Optional) 96 101 Coniugacy and Svlow's T h e o r e m (Optional)
Contents
The Symmetric G r o u p
1 2 3
108
Preliminaries 108 Cycle Decomposition 111 O d d and Even Permutations
119
Ring T h e o r y
1 2 3 4 5 6 7
125
Definitions and Examples 125 Some Simple Results 137 Ideals, H o m o m o r p h i s m s , and Quotient Rings 139 Maximal Ideals 148 Polynomial Rings 151 Polynomials over the Rationals 166 Field of Quotients of an Integral D o m a i n 172
Fields
1 2 3 4 5 6
176
Examples of Fields 176 A Brief Excursion into Vector Spaces Field Extensions 191 Finite Extensions 198 Constructibility 201 R o o t s of Polynomials 207
180
S p e c i a l Topics ( O p t i o n a l )
1 2 3 4 5 6 7
215
T h e Simplicity of A„ 215 Finite Fields I 221 Finite Fields II: Existence 224 Finite Fields III: Uniqueness 227 Cyclotomic Polynomials 229 Liouville's Criterion 236 T h e Irrationality of IT 239 243
Index
PREFACE TO THE THIRD EDITION
W h e n we were asked to p r e p a r e the third edition of this book, it was our consensus that it should not be altered in any significant way, and that Herstein's informal style should be preserved. W e feel that one of the b o o k ' s virtues is the fact that it covers a big chunk of abstract algebra in a condensed and interesting way. A t the same time, without trivializing the subject, it remains accessible to most undergraduates. W e have, however, corrected m i n o r errors, straightened out inconsistencies, clarified and expanded some proofs, and added a few examples. To resolve the many typographical problems of the second edition, Prentice Hall has had the b o o k completely retypeset—making it easier a n d m o r e pleasurable to read. It has b e e n pointed out t o us that some instructors would find it useful to have the Symmetric G r o u p S„ and the cycle notation available in C h a p t e r 2, in order to provide m o r e examples of groups. R a t h e r t h a n alter the arrangement of the contents, thereby disturbing the original balance, we suggest an alternate route through the material, which addresses this concern. After Section 2.5, one could spend an hour discussing permutations and their cycle decomposition (Sections 3.1 and 3.2), leaving the proofs until later. T h e students might then go over several past examples of finite groups and explicitly set u p isomorphisms with subgroups of S„. This exercise would be motivated by Cayley's theorem, q u o t e d in Section 2.5. At the same time, it would have the beneficial result of making the students more comfortable with the concept of an isomorphism. T h e instructor could then weave in the various subgroups of the Symmetric G r o u p s S„ as examples throughout the remain
x
P r e f a c e to T h i r d E d i t i o n
Ch. 6
der of Chapter 2. If desired, one could even introduce Sections 3.1 and 3.2 after Section 2.3 or 2.4. T w o changes in the format have been m a d e since the first edition. First, a Symbol List has b e e n included to facilitate keeping track of terminology. Second, a few problems have b e e n m a r k e d with an asterisk (*). T h e s e serve as a vehicle to introduce concepts and simple arguments that relate in some important way to the discussion. As such, they should be read carefully. Finally, we take this opportunity to thank the m a n y individuals whose collective efforts have helped to improve this edition. W e t h a n k the reviewers: Kwangil K o h from N o r t h Carolina State University, D o n a l d P a s s m a n from the University of Wisconsin, and R o b e r t Zinc from P u r d u e University. And, of course, we thank George Lobell and Elaine W e t t e r a u , and others at Prentice Hall who have been most helpful. B a r b a r a Cortzen David J. Winter
In the last halfcentury or so abstract algebra has b e c o m e increasingly important not only in mathematics itself, but also in a variety of other disciplines. For instance, the importance of the results and concepts of abstract algebra play an ever m o r e important role in physics, chemistry, and computer science, to cite a few such outside fields. In mathematics itself abstract algebra plays a dual role: that of a unifying link between disparate parts of mathematics and that of a research subject with a highly active life of its own. It has been a fertile and rewarding research area both in the last 100 years a n d at t h e present m o m e n t . Some of the great accomplishments of our twentiethcentury mathematics have b e e n precisely in this area. Exciting results have been proved in group theory, commutative and noncommutative ring theory, Lie algebras, Jordan algebras, combinatorics, and a host of other parts of what is k n o w n as abstract algebra. A subject that was once regarded as esoteric has become considered as fairly downtoearth for a large cross section of scholars. T h e p u r p o s e of this b o o k is twofold. F o r those readers w h o either want to go o n to do research in mathematics or in some allied fields that use algebraic notions and methods, this b o o k should serve as an introduction—and, we stress, only as an introduction—to this fascinating subject. F o r interested readers who want to learn what is going on in an engaging part of m o d e r n mathematics, this book could serve that purpose, as well as provide t h e m with some highly usable tools to apply in the areas in which they are interested. T h e choice of subject matter has b e e n m a d e with the objective of introducing readers to some of the fundamental algebraic systems that are b o t h in
xii
P r e f a c e t o First E d i t i o n
teresting and of wide use. Moreover, in each of these systems the aim has b e e n to arrive at some significant results. T h e r e is little purpose served in studying some abstract object without seeing some nontrivial consequences of the study. W e h o p e that we have achieved the goal of presenting interesting, applicable, and significant results in each of the systems we have chosen to discuss. A s the r e a d e r will soon see, there are many exercises in the book. T h e y are often divided into three categories: easier, middlelevel, and h a r d e r (with an occasional very h a r d ) . T h e purpose of these problems is to allow students to test their assimilation of the material, to challenge their mathematical ingenuity, to p r e p a r e the ground for material that is yet to come, and to be a m e a n s of developing mathematical insight, intuition, and techniques. R e a d e r s should not b e c o m e discouraged if they do not manage to solve all the p r o b lems. T h e intent of many of the problems is that they b e tried—even if not solved—for the pleasure (and frustration) of the reader. Some of the p r o b lems a p p e a r several times in the book. Trying to do the problems is undoubtedly the best way of going about learning the subject. We have strived to present the material in the language a n d tone of a classroom lecture. Thus the presentation is somewhat chatty; we h o p e that this will put the readers at their ease. A n attempt is m a d e to give many and revealing examples of the various concepts discussed. Some of these examples are carried forward to be examples of other p h e n o m e n a that come u p . T h e y are often referred to as the discussion progresses. W e feel that the book is selfcontained, except in o n e section—the second last one of the b o o k — w h e r e we m a k e implicit use of the fact that a polynomial over the complex field has complex roots (that is the celebrated Fundamental Theorem of Algebra due to Gauss), and in the last section where we m a k e use of a little of the calculus. W e are grateful to many people for their comments and suggestions on earlier drafts of the book. M a n y of the changes they suggested have b e e n incorporated and should improve the readability of the book. W e should like to express our special thanks to Professor Martin Isaacs for his highly useful comments. W e are also grateful to Fred Flowers for his usual superb j o b of typing the manuscript, and to Mr. Gary W. Ostedt of the Macmillan C o m p a n y for his enthusiasm for the project and for bringing it to publication. With this we wish all the readers a happy voyage on the mathematical journey they are about to u n d e r t a k e into this delightful and beautiful realm of abstract algebra. I.N.H.
aES a£S SCT,TDS S=T
0
AUB AC)B {s E 5  s satisfies P] A  B A' (a, b) AX B
U
f:S> T
m
i: S —» S, i
s
r (A)
l
f°g, fg
A(S)
s„
n\ Z 0(s)
a is an element of the set S, 3 a is not an element of the set S, 3 S is a subset of the set T, 3 The sets 5 and T a r e equal (have the same elements), 4 T h e empty set, 4 T h e union of the sets A and B, 4 T h e intersection of the sets A and B, 4 T h e subset of elements of S satisfying P, 4 The difference of the sets A and fi, 4 T h e c o m p l e m e n t of A, 5 O r d e r e d pair consisting of a, b (see also below), 5 T h e Cartesian product of A and B, 5 T h e set of real n u m b e r s , 8 Function from the set S to the set T, 8 Image of the element s under the function/, 8 T h e identity function on S, 9 Inverse image of t u n d e r f, 10 Inverse image of a subset A of T under / : S —» T, 10 Composition or product of functions / a n d g , 11,18 Set of 11 mappings from a set S to S, 16 Symmetric group of degree n, 16,109 n factorial, 17 Set of integers, 21 Orbit of s relative to mapping /, 21
xiv
S y m b o l List
N m n m n (a, b) C i,—i z = a + bi z = a  bi IIz z ?• (cos 9 + / sin 6) 9„ Q E G  C(a) (a) Z(G) a ~ b a = b mod n a = b(n) [a] cl(a) o(a) i {H) Z„ U„ <p(n) Hb aH G ~ G' cp(G) K e r cp N <i G GIN AB G X G X . . . X G„ fab ... c \ii v ... \v (a, b,... , c) A„ a + ct^' + a j + a k det x
n G 1 2 N / 0 2 3
Set of positive integers, 21 m divides n, 22 m does not divide n, 22 Greatest c o m m o n divisor of a, b (see also above), 23 Set of complex numbers, 32 Square roots of —1, 32 Complex n u m b e r z with real part a and imaginary part b, 32 Conjugate of complex n u m b e r z = a + bi, 32 Inverse of the complex n u m b e r z, 33 Absolute value of complex n u m b e r z, 34 Polar form of a complex n u m b e r , 35 Primitive nth root of unity, 36, 42 Set of rational numbers, 42 G r o u p of nth roots of unity, 42 O r d e r of the group G, 42 Centralizer of a in G, 53,102 Cyclic group generated by a, 53 Center of group G, 53 a is equivalent to b in a specified sense, 57 a is congruent to b modulo n (long form), 57 a is congruent to b m o d u l o n (short form), 57 Class of all b equivalent to a, 58 Conjugacy class of a, 58,101 O r d e r of element a of a group, 60 Index ofHin G, 59 Set of integers m o d n, 61 G r o u p of invertible elements of Z„, 62 E u l e r cp function (phi function), 62 Right coset of subgroup H, 58 Left coset of subgroup H, 64 G r o u p G is isomorphic to group G ' , 68 Image of h o m o m o r p h i s m , 70 Kernel of the h o m o m o r p h i s m s 9 , 70,140 T is a n o r m a l subgroup of G, 72 V Quotient of a group G by a subgroup N, 78 Product of subsets A, B of a group, 79 Direct product of G G , . . . , G„, 93
h 2
Permutation sending a to 11, b to v,..., Cycle sending a to b,. .. , c to a, 111 Alternating group of degree n, 121,215 Quaternion, 131 D e t e r m i n a t e of the 2 X 2 matrix x, 136
c to w, 110
S y m b o l List
xv
H{F) R®$ (a) F[x] deg p(x) g(x)\f(x) R[x] F(x) v G V av a v + . . . + a„v (v v , ...,v„) V®W d i m (V) U+W [K:F] F[a] F{a) E(K)
1 1 u 2 F
n
4>„M
Ring of quaternions over F, 136 Direct sum of rings R, S, 146 Ideal generated by a in a commutative ring, 145 Polynomial ring over the field F, 152 D e g r e e of polynomial p(x), 153 Ideal generated by g(x) in a polynomial ring, 157 Polynomial g(x) divides f(x), 157 Polynomial ring over ring R, 163 Field of rational functions in x over F, 177 Vector v in a vector space V, 180 Scalar a times vector v, 180 Linear combination of vectors v ,.. . , v„, 181 Subspace spanned by v , v , • • •, v„, 181 Direct sum of vector spaces V, W, 181 D i m e n s i o n of V over F, 186 Sum of subspaces U, W of V, 190 D e g r e e of K over F, 191 Ring generated by a over F, 196 Field extension obtained by adjoining to a to F, 196 Field of algebraic elements of K over F, 198 F o r m a l derivative of p o l y n o m i a l / ( x ) , 227 nth cyclotomic polynomial, 230
x t 2
ABSTRACT ALGEBRA
TH1M
'M/llMlt
f
AMIUAi
1. A FEW P R E L I M I N A R Y R E M A R K S
F o r many readers this b o o k will b e their first contact with abstract m a t h e matics. The subject to be discussed is usually called "abstract algebra," but the difficulties that the reader m a y encounter are not so much due to t h e "alg e b r a " part as they are to the "abstract" part. On seeing some area of abstract mathematics for the first time, be it in analysis, topology, or whatnot, there seems to be a common reaction for the novice. This can best be described by a feeling of being adrift, of not having something solid to hang on to. This is not too surprising, for while many of the ideas are fundamentally quite simple, they are subtle and seem to elude one's grasp the first time around. O n e way to mitigate this feeling of limbo, or asking oneself "What is the point of all this?," is to take the concept at hand and see what it says in particular cases. In other words, the best road to good understanding of the notions introduced is to look at examples. This is true in all of mathematics, but it is particularly true for the subject matter of abstract algebra. Can one, with a few strokes, quickly describe the essence, purpose, and background for the material we shall study? Let's give it a try. W e start with some collection of objects S and endow this collection with an algebraic structure by assuming that we can combine, in one or several ways (usually two), elements of this set S to obtain, once more, elements of this set S. These ways of combining elements of S we call operations o n S.
2
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
T h e n we try to condition or regulate the n a t u r e of S by imposing certain rules on how these operations b e h a v e on S. These rules are usually called the axioms defining the particular structure on 5. These axioms are for us to define, but the choice m a d e comes, historically in mathematics, from noticing that there are many concrete mathematical systems that satisfy these rules or axioms. W e shall study some of the basic axiomatic algebraic systems in this b o o k , namely groups, rings, and fields. Of course, one could try many sets of axioms to define new structures. W h a t would we require of such a structure? Certainly we would want that the axioms be consistent, that is, that we should not be led to some nonsensical contradiction computing within the framework of the allowable things the axioms permit us to do. B u t that is not enough. W e can easily set up such algebraic structures by imposing a set of rules on a set S that lead to a p a t h o logical or weird system. F u r t h e r m o r e , there may be very few examples of something obeying the rules we have laid down. Time has shown that certain structures defined by " a x i o m s " play an imp o r t a n t role in mathematics (and other areas as well) and that certain others are of no interest. T h e ones we mentioned earlier, namely groups, rings, and fields, have stood the test of time. A word about the use of "axioms." In everyday language "axiom" m e a n s a selfevident truth. But we are n o t using everyday language; we are dealing with mathematics. A n axiom is not a universal t r u t h — b u t one of several rules spelling out a given mathematical structure. T h e axiom is true in the system we are studying because we have forced it to be true by h y p o t h e sis. It is a license, in the particular structure, to do certain things. W e return to something we said earlier about the reaction that m a n y students have on their first encounter with this kind of algebra, namely a lack of feeling that the material is something they can get their teeth into. D o not b e discouraged if the initial exposure leaves you in a bit of a fog. Stick with it, try to understand what a given concept says, and most importantly, look at particular, concrete examples of the concept u n d e r discussion.
PROBLEMS
1. Let S b e a set having an operation * which assigns an element a * b of S for any a, b G S. Let us assume that the following two rules hold: 1. If a, b are any objects in S, then a * b = a. 2. If a, b are any objects in S, then a * b = b * a. Show that S can have at most one object.
Sec. 2
Set T h e o r y
3
2. Let S b e the set of all integers 0, ± 1 , ± 2 , . . . , ±n,.... For a, b in S define * by a*b = a — b. Verify the following: (a) a * b j= b * a unless a = b. (b) (a * b) * c + a * (b * c) in general. U n d e r what conditions on a, b, c is (a * b) * c = a * (6 * c)? (c) T h e integer 0 has the property t h a t a * 0 = A for every a in 5. (d) F o r a in S, a * a = 0. 3. Let 5 consist of the two objects • and A . W e define the operation * on S by subjecting • and A to the following conditions: 1. n * A = A = A * Q 2. • * • = • . 3. A * A = • . Verify by explicit calculation that if a, b, c are any elements of S (i.e., a, b and c can be any of • or A ) , then: (a) a * b is in S. (b) (a * b) * c = a * (b * c). (c) a * b = b * a. (d) T h e r e is a particular a in S such that a * b = b * a = b for all b in S. . (e) Given b in 5, then b * b = a, w h e r e a is the particular element in P a r t (d).
2 . SET THEORY
With the changes in the mathematics curriculum in the schools in the U n i t e d States, many college students have h a d some exposure to set theory. This introduction to set theory in the schools usually includes the elementary n o tions and operations with sets. Going on the assumption that many readers will have some acquaintance with set theory, we shall give a rapid survey of those parts of set theory that we shall n e e d in what follows. First, however, we need some notation. T o avoid the endless repetition of certain phrases, we introduce a shorthand for these phrases. Let S b e a collection of objects; the objects of S we call the elements of S. To d e n o t e that a given element, a, is an element of S, we write a G S—this is read "a is an element of 5 . " To denote the contrary, namely that an object a is not an element of S, we write a S. So, for instance, if S denotes the set of all positive integers 1, 2, 3 , . . . , « , . . . , then 165 E S, whereas  1 3 £ S. W e often want to know or prove that given two sets S and T, one of these is a part of the other. W e say that S is a subset of T, which we write S C T (read " 5 is contained in 7"") if every element of S is an element of T.
4
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
In terms of the notation we now have: S C T if s G S implies that s G T. W e can also denote this by writing T D S, read "T contains S." (This does not exclude the possibility that S = T, that is, that S and T have exactly the same elements.) Thus, if T is the set of all positive integers and S is t h e set of all positive even integers, t h e n S C T, and 5 is a subset of T. In the definition given above, S D S for any set S; that is, S is always a subset of itself. W e shall frequently need to show that two sets S and T, defined perhaps in distinct ways, are equal, that is, they consist of the same set of elements. T h e usual strategy for proving this is to show that b o t h S C T and T C S. For instance, if S is the set of all positive integers having 6 as a factor and T is the set of all positive integers having both 2 and 3 as factors, then S = T. (Prove!) T h e need also arises for a very peculiar set, namely one having n o elements. This set is called the null or empty set and is d e n o t e d by 0 . It has t h e property that it is a subset of any set S. Let A, B be subsets of a given set S. W e now introduce m e t h o d s of constructing other subsets of S from A and B. T h e first of these is the union of A and B, written A U B, which is defined: A U B is that subset of S consisting of those elements of S that are elements of A or are elements of B. T h e " o r " we have just used is somewhat different in meaning from the ordinary usage of the word. H e r e we m e a n that an element c is in A U B if it is in A, or is in B, or is in both. T h e " o r " is n o t m e a n t to exclude the possibility that b o t h things are true. Consequently, for instance, A U A = A.
If A = {1, 2, 3} and B = {2, 4, 6,10}, then A U B = {1, 2, 3, 4, 6,10}.
W e now proceed to our second way of constructing new sets from old. Again let A and B be subsets of a set S; by the intersection of A and B, written A D B, we shall m e a n the subset of S consisting of those elements that are both in A and in B. T h u s , in the example above, A n B = {2}. It should be clear from the definitions involved that A n B C A and A n B C B. Particular examples of intersections that hold universally are: A D A = A, A n S = A, A D 0 = 0 . This is an o p p o r t u n e m o m e n t to introduce a notational device that will be used time after time. Given a set S, we shall often b e called on t o describe the subset A of S, whose elements satisfy a certain p r o p e r t y P. W e shall write this as A = {s G 5  s satisfies P). F o r instance, if A, B are subsets of S, then Al)B = [s(ES\s&A or s<=B] while AnB = {sGS\sE.A a n d ^ G B). A l t h o u g h the notions of union and intersection of subsets of S have b e e n defined for two subsets, it is clear h o w one can define the union and intersection of any n u m b e r of subsets. W e now introduce a third operation we can perform on sets, the difference of two sets. If A , B are subsets of 5, we define A  B = {a G A\a $L B\.
Sec. 2
Set Theory
5
So if A is the set of all positive integers and fl is t h e set of all even integers, then A — B is the set of all positive odd integers. In the particular case w h e n A is a subset of 5, the difference S — A is called the complement of A in S and is written A'. W e represent these three operations pictorially. If A is @ and B is © , then 1. A U B
=mWM
is the shaded area.
2. ^ H B = ( A M a l i s the s h a d e d area.
3. A — B = Kg,4it ) b ) is t h e s h a d e d area.
4. B  A ={
A [ KBVA is the s h a d e d area.
N o t e the relation a m o n g the t h r e e operations, namely the equality A U B = (A n B) U (A  B) U (B  A). A s an illustration of how one goes about proving the equality of sets constructed by such settheoretic constructions, we prove this latter alleged equality. W e first show that (A fl B) U (A  B) U (fl  A) C A U fl; this part is easy for, by definition, A D B C A, A  B C A, and B  A C B, h e n c e {A
n
fl) U {A  B) U
(5 
A) C A U A U
5
= A U B.
N o w for the o t h e r direction, namely that A U B C (A D B) U (A  5) U (5  A ) . Given u E A U 5, if « £ 4 and » G 5, t h e n u G i f l B , so it is certainly in (A n B) U (A  fl) U (fl  A ) . O n the other hand, if u G A but w ^ fl, then, by the very definition of A — fl, u G A — 5, so again it is certainly in (A n fl) U {A  fl) U (fl  A ) . Finally, if u G fl but u £ A, t h e n m G fl  A , so again it is in (A Pi fl) U (A  fl) U (fl  A ) . W e have thus covered all t h e possibilities and have shown that A U fl C (A Pi fl) U (A  fl) U (fl — A). Having the two opposite containing relations of A U fl and (A n fl) U (A  fl) U (fl  A), we obtain the desired equality of these two sets. W e close this brief review of set theory with yet another construction we can carry out on sets. This is the Cartesian product defined for the two sets A, B by A X fl = [(a, b) \ a G A, b G fl}, where we declare the o r d e r e d pair (a, b) to b e equal to the o r d e r e d pair ( a b{) if and only if a = a and b = b\. H e r e , too, we n e e d not restrict ourselves to two sets; for instance, we
l 5 x
6
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
can define, for sets A, B, C, their Cartesian product as the set of o r d e r e d triples (a, £>, c ) , where a G A, b G B, c G C and where equality of two ordered triples is defined componentwise.
PROBLEMS
Easier Problems 1. Describe the following sets verbally. (a) S = (Mercury, Venus, Earth, . . . , Pluto). (b) S = {Alabama, Alaska, . . . , Wyoming}. 2. Describe the following sets verbally. (a) S = {2, 4, 6, 8, . . . } . (b) S = {2, 4, 8,16, 32, . . . } . (e) S= { 1 , 4 , 9 , 1 6 , 2 5 , 3 6 , . . . } . 3. If A is the set of all residents of the United States, B the set of all Canadian citizens, and C the set of all w o m e n in the world, describe the sets Ar\BnC,AB,AC,CA verbally. 4. If A = {1, 4, 7, a} and B = {3, 4, 9, 11} and you have b e e n told that A n B = {4, 9}, what must a b e ? S.lfACB 6. If AC and B C C, prove that A C C. B, prove that i U C C S U C f o r any set C. = BUA and AnB = BC\ A.
7. Show that AUB
8. Prove that (A  B) U (B  A) = (A U B)  (A n 5 ) . W h a t does this look like pictorially? 9. Prove that A (1 (B U C) = (A D B) U (A D C). 10. Prove that A U (B f) C) = {A U 5 ) n (A U C). 11. Write down all the subsets of S = {1, 2, 3, 4}. MiddleLevel Problems *12. If C is a subset of S, let C d e n o t e the complement of C in S. Prove the De Morgan Rules for subsets A, B of S, namely: (a) (A n B)' = A' U 5 ' . (b) (A U B)' =A'n B'. * 13. Let S b e a set. F o r any two subsets of S we define A + B = (A  B) U (B  A) and A • B = A H B.
Sec. 2
Set Theory
7
Prove that: (a) A + B = B + A. (b) A + 0 = A. ( c ) A • A = A. (d) A + A = 0. (e) A + (B + C) = {A + B) + C. (f) If A + B = A + C, then B = C. ( g ) A • (fl + C) = A • B + A • C. *14. If C is a finite set, let m(C) are finite sets, prove that m(A d e n o t e the n u m b e r of elements in C. If A, B
U fl) = m(A)
+ m(B)
 m(A
n fl).
... ,
15. F o r t h r e e finite sets A, fl, C find a formula for m ( A U fl U C). (Hint: First consider D = fl U C and use the result of P r o b l e m 14.) 16. Take a shot at finding m{A A„.
x
U A U • • • U 4 „ ) for n finite sets A A ,
2 U 2
17. Use the result of Problem 14 to show that if 80% of all Americans have gone to high school and 7 0 % of all A m e r i c a n s r e a d a daily newspaper, t h e n at least 5 0 % of Americans have b o t h gone to high school and r e a d a daily newspaper. 18. A public opinion poll shows that 9 3 % of t h e population agreed with the government on the first decision, 84% on the second, and 74% on the third, for t h r e e decisions m a d e by the government. A t least what percentage of the population agreed with the government on all three decisions? (Hint: Use the results of P r o b l e m 15.) 19. In his b o o k A Tangled Tale, Lewis Carroll p r o p o s e d the following riddle about a group of disabled veterans: "Say t h a t 70% have lost an eye, 7 5 % an ear, 80% an arm, 8 5 % a leg. W h a t percentage, at least, must have lost all f o u r ? " Solve Lewis Carroll's problem. *20. Show, for finite sets A, fl, that m(A 21. If S is a set having five (a) H o w many subsets (b) H o w many subsets (c) H o w m a n y subsets Harder P r o b l e m s 22. (a) Show that a set having n elements has 2" subsets. (b) If 0 < m < n, how many subsets are there that have exactly m elements? X fl) = m(A)m(B). elements: does S have? having four elements does S have? having two elements does S have?
8
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
3.
MAPPINGS
O n e of the truly universal concepts that runs through almost every phase of mathematics is that of a function or mapping from one set to another. O n e could safely say that t h e r e is n o part of mathematics where the notion does not arise or play a central role. T h e definition of a function from one set to another can be given in a formal way in terms of a subset of the Cartesian product of these sets. Instead, here, we shall give an informal and admittedly nonrigorous definition of a mapping (function) from one set to another. Let S, T b e sets; a function or mapping f from S to T is a rule that assigns to each element s G S a unique element t G T. Let's explain a little more thoroughly what this means. If s is a given element of S, t h e n t h e r e is only one element t in T that is associated to s by the mapping. A s s varies over S, t varies over T (in a m a n n e r depending on s). N o t e that by the definition given, the following is not a mapping. Let S b e the set of all p e o p l e in the world and T the set of all countries in the world. Let / b e t h e rule that assigns to every person his or her country of citizenship. Then / is not a m a p ping from S to T. Why not? Because there are people in the world that enjoy a dual citizenship; for such people there would not be a unique country of citizenship. Thus, if Mary Jones is b o t h an English and French citizen, / would not m a k e sense, as a mapping, when applied to Mary Jones. O n the other hand, the rule / : (R —» R, where R is the set of real n u m b e r s , defined by f(a) = a for a G R , is a perfectly good function from R. to R. It should b e noted t h a t / (  2 ) = (  2 ) = 4 = / ( 2 ) , a n d / (  a ) = f{a) for all a G R. W e denote that / is a mapping from S to T by / : S —» T and for the f G T mentioned above we write t = f(s); we call t the image of s u n d e r / . T h e concept is hardly a new one for any of us. Since grade school we have constantly encountered mappings and functions, often in the form of formulas. But mappings need not b e restricted to sets of n u m b e r s . A s we see below, they can occur in any area.
1 2
Examples 1. Let 5 = {all m e n w h o have ever lived} and T = {all w o m e n w h o h a v e ever lived}. Define f:S^>T by f(s) = m o t h e r of s. Therefore, / ( J o h n F. Kennedy) = Rose Kennedy, and according to our definition, R o s e K e n n e d y is t h e image u n d e r / of J o h n F. Kennedy. 2. Let S = {all legally employed citizens of the United States} and T = {positive integers}. Define, for s £ S , f(s) by f(s) = Social Security N u m b e r of s. (For the purpose of this text, let us assume that all legally employed citizens of the United States have a Social Security N u m b e r . ) T h e n / defines a m a p ping from §. to T.
Sec.
3
Mappings
9
3. Let S be the set of all objects for sale in a grocery store and let T = {all real numbers}. Define / : 5 — T by f(s) = price of s. This defines a mapping > from S to T. 4. Let S be the set of all integers and let T = S. Define f:S—> Thy f(m) = 2m for any integer m. Thus the image of 6 u n d e r this mapping, / ( 6 ) , is given b y / ( 6 ) = 2 • 6 = 12, while that of  3 , / (  3 ) , is given by / (  3 ) = 2 (  3 ) =  6 . If s s G S are in S and f(s ) = f(s ), what can you say about s and s 7
u 2 x 2 1 2
5. Let S = T b e the set of all real n u m b e r s ; define / : S  » T by / ( s ) = ^ . D o e s every element of T come u p as an image of some s G SI If not, h o w would you describe the set of all images \f(s) \s G 5}? W h e n is f(s{) = 6. Let 5 = T b e the set of all real n u m b e r s ; d e f i n e / : S — T b y f(s) = s . This > is a function from 5 to T. W h a t can you say about \f(s) \ s E S}? W h e n is 7. Let T b e any nonempty set and let S = T X T, the Cartesian product of T with itself. D e f i n e / : T X T T b y f(t , t ) = t . This mapping from T X T to T is called the projection of I X T o n t o its first component.
l 2 L 3
2
8. Let S b e the set of all positive integers and let T be the set of all positive rational n u m b e r s . Define / : S X 5 > T by / ( m , «) = m/n. This defines a mapping from S X 5 to T. N o t e that / ( l , 2) =  while / ( 3 , 6) = f = § = / ( l , 2), although ( 1 , 2) ^ (3, 6). Describe the subset of 5 X S consisting of those (a, b) such t h a t / ( a , b) = . T h e mappings to b e defined in E x a m p l e s 9 and 10 are mappings that occur for any nonempty sets and play a special role. 9. Let 5, T b e n o n e m p t y sets, and let r b e a fixed element of T. Define f:S—> T by f(s) = 1 for every s G S; / is called a constant function from S to T.
0 0
10. Let S be any nonempty set and define i:S—*Sby i(s) = s for every s G S. W e call this function of S to itself the identity function (or identity mapping) on S. W e may, at times, denote it by i (and later in the book, by e).
s
Now that we have the notion of a mapping we n e e d some way of identifying when two mappings from one set to another are equal. This is n o t G o d given; it is for us to decide h o w t o declare f = g where f:S>T and g : S » T. W h a t is m o r e natural than to define this equality via t h e actions of / and g on the elements of S? M o r e precisely, we declare that / = g if and only if f(s) = g(s) for every s G S. If S is the set of all real n u m b e r s and / is defined on S by f(s) = s + 2s + 1, while g is defined on S by g(s) = (s + l ) , our definition of the equality o f / a n d g is merely a statement of the familiar identity (s + l ) = s + 2s + 1.
z 2 2 z
10
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch, 1
Having m a d e the definition of equality of two mappings, we now want to single out certain types of mappings by the way they behave. Definition. T h e mapping / : S — T is onto or surjective if every t G T » is the image u n d e r / of some s G S; that is, if and only if, given t G T, there exists an s G 5 such that t = f(s). In the examples we gave earlier, in E x a m p l e 1 the mapping is n o t onto, since not every woman that ever lived was the m o t h e r of a m a l e child. Similarly, in Example 2 the mapping is not onto, for not every positive integer is the Social Security N u m b e r of some U.S. citizen. The mapping in E x a m p l e 4 fails to b e onto because not every integer is even; and in Example 5, again, the mapping is not onto, for the n u m b e r —1, for instance, is not the square of any real number. However, the mapping in Example 6 is onto because every real n u m b e r has a unique real cube root. The reader can decide w h e t h e r or not the given mappings are onto in the other examples. If we define f(S) = {f(s) G T \ s G S), another way of saying that the mapping / : 5 — T is onto is by saying that f(S) = T. > A n o t h e r specific type of mapping plays an important and particular role in what follows. Definition. A m a p p i n g / : .V —• 7' is said to be onetoone (written 11) or injective if for s, + s in S, /(i'j) + f(s ) in T. Equivalently, / is 11 if /( i) f ( 2 ) implies that s = s .
2 2 5 = s t 2
In other words, a mapping is 11 if it takes distinct objects into distinct images. In the examples of mappings we gave earlier, the mapping of Example 1 is not 11, since two brothers would have the same mother. H o w e v e r in Example 2 the mapping is 11 because distinct U.S. citizens have distinct Social Security numbers (provided that there is no goofup in Washington, which is unlikely). The reader should check if the various other examples of mappings are 11. Given a mapping / : S — T and a subset A C T, we may w a n t to look at > B = {s G 5 \f(s) G A}; we use the notation f~\A) for this set B, and call F~ (A) the inverse image of A under f. Of particular interest is f~ {t), the inverse image of the subset {f} of T consisting of the element t G T alone. If the inverse image of \t\ consists of only one element, say s G 5, we could try to define f~ (t) by defining / ( 0 = s. A s we note below, this need not be a mapping from T t o S, but is so if / i s 11 and onto. We shall use the same n o tation f~ in cases of b o t h subsets and elements. This f~ does not in general define a mapping from T to S for several reasons. First, if / i s not onto, then
X l l _ 1 l l
Sec.
3
Mappings
 1
11
there is some t in T which is not the image of any element s, so / ( f ) = 0 . Second, if / is not 11, then for some t E T there are at least two distinct s ¥= s in S such that f(s^) = t = f(s ). So f~ (t) is not a unique element of S—something we require in our definition of mapping. However, if / i s both 11 and onto T, t h e n / indeed defines a mapping of T o n t o S. (Verify!) This brings us to a very important class of mappings.
l 1 2 2  1
Definition. T h e mapping / : S —* 7' is said to be a 11 or bijection if / i s both 11 and onto.
correspondence
N o w that we have the notion of a mapping and have singled out various types of mappings, we might very well ask: " G o o d and well, but what can we do with t h e m ? " A s we shall see in a m o m e n t , we can introduce an operation of combining mappings in certain circumstances. Consider the situation g : S —» T and / : T ^> U. Given an element s E S, then g sends it into the element g(s) in T; so g(s) is ripe for being acted on by / Thus we get an element f(g(s)) E U. W e claim that this procedure p r o vides us with a mapping from 5 to U. (Verify!) W e define this m o r e formally in the Definition, if g : S > 7' and f:T^>U, then the composition (or product), d e n o t e d by f°g, is the mapping f°g:S^> U defined by (f°g)(s) = f(g(s)) for every s G S . N o t e that to compose the two mappings / and g—that is, for f°g to have any sense—the terminal set, T, for the mapping g must be the initial set for the mapping /. One special time when we can always compose any two mappings is when S = T = U, that is, when we map S into itself. Although special, this case is of the utmost importance. W e verify a few properties of this composition of mappings. L e m m a 1.3.1. (f g)°
a
If /;: .S'
T, g : T
U. and / : (/ ^ V, then
=
ItProof. H o w shall we go about proving this lemma? To verify that two
mappings are equal, we merely must check that they do the same thing to every element. Note first of all that b o t h f° (g ° h) and ( / ° g)° h define m a p pings from S to V, so it m a k e s sense to speak about their possible equality. O u r task, then, is to show that for every s E S, (f°(g°h))(s) = ((/° g) ° )( ) W e apply the definition of composition to see that
n s
( / o (go h))( )=
S
/((go /,)(,))=
f(g(h(s))).
12
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
Unraveling {(f g)°h)(s) we do indeed see that
0
=
(f°g)(Ks))=f(g(h(s))),
(/»(r#) =
((f°g)°hm
for every s E S. Consequently, by definition, f° (g ° h) = (f° g)° /?. • (The symbol • will always indicate that the proof has been completed.) This equality is described by saying that mappings, under composition, satisfy the associative law. Because of the equality involved t h e r e is really n o n e e d for parentheses, so we write f° (g h) as f° g ° h.
0
Lemma 1.3.2. is also 11.
T a n d f : T> U are both 11, t h e n / ° g : 5  » U
Proof. Let us suppose that (f°g)(si) = ( / ° g ) ( s ) ; thus, by definition, f(g(*i)) = f(g( 2))S i n c e / i s 11, we get from this that g( ) = g(s ); however, g is also 11, thus s = s follows. Since (f°g)(si) = (f°g)(s ) forces S\ = s , the m a p p i n g / g is 11. •
2
s
Sl
2
x
2
2
0
2
W e leave the proof of t h e next R e m a r k to the reader. Remark. If g : S also onto. T and / : T> U are both onto, then f°g : S > U is
A n immediate consequence of combining the R e m a r k and L e m m a 1.3.2 is to obtain L e m m a 1.3.3. If g : S Uis also a bijection. T and f:T^ U are both bijections, then
If / i s a 11 correspondence of S o n t o T, then the "object" f~ :T^S defined earlier can easily b e shown to be a 11 mapping of T o n t o S. In this case it is called the inverse of /. In this situation we have Lemma 1.3.4. I f / : S —>• T is a bijection, t h e n / ° / ~ ' = / and / ^ ° / = i , where i and i are the identity mappings of S and T, respectively.
r s s r
x
Proof. W e verify o n e of these. If / G T, then {f°r ){t) = f{f~\t)). But what is / ( 0 ? By definition, / ( f ) is that element s G S such that
_ 1 _ 1 Q
l
Sec.
3
 1
Mappings
13
t = / ( s ) . So / ( /  ' ( f ) ) = / O o ) = f. I n other words, ( / ° / ) ( f ) = r for every t £ 7 ; h e n c e / ° / = / , the identity mapping on T. •
0 _ 1 r
W e leave t h e last result of this section for the reader to prove. L e m m a 1.3.5. I f / : S —> T and i is the identity mapping of T o n t o itself and i is that of S onto itself, then i °f = / a n d / i = f.
T 0 s T s
PROBLEMS
Easier Problems 1. For the given sets S, T determine if a mapping / : S —» T is clearly and unambiguously defined; if not, say why not. (a) S = set of all women, T = set of all men, f(s) = husband of s. (b) S = set of positive integers, T = S, f(s) = s  1. (c) S = set of positive integers, T = set of nonnegative integers, f(s) = s  1. (d) S = set of nonnegative integers, T = S, f(s) = s  1. (e) 5 = set of all integers, T = 5, / ( s ) = 5 — 1. (f) 5 = set of all real n u m b e r s , T = S, f(s) = V s . (g) 5 = set of all positive real n u m b e r s , T = S, f(s) = 'Vs. 2. In those parts of Problem 1 where / does define a function, determine if it is 11, onto, or both. *3. If / is a 11 mapping of S o n t o T, prove t h a t / o n t o S. *4. I f / i s a 11 mapping of S onto T, p r o v e t h a t / *6. If / : S > T is onto and g:T^U prove that g = h. *7. If g : S > r, then g = :>S T, and i f / : T U is 11, show that if / ° g = f°h,
 1 _ 1
is a 11 mapping of T
° / = /y. g°f=h°f,
5. Give a proof of the R e m a r k after L e m m a 1.3.2. and h:T—>U are such that
8. L e t 5 be t h e set of all integers a n d T = {1, —1); / : S > T is defined by / ( s ) = 1 if 5 is even, / ( j ) =  1 if 5 is odd. (a) D o e s this define a function from S to 7 7 (b) Show that / ( $ ! + j ) = f(si)f(s ). W h a t does this say about the integers? (c) Is f(s s ) = f{si)f(s ) also true?
2 2 1 2 2
14
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
9. Let S be the set of all real numbers. Define f:S^> g:S>Sbyg(s)=s + l. (a) F i n d / ° g . (b) Find (c) I s / ° g = g o / ?
S by f(s)
= s , and
2
10. Let S be the set of all real n u m b e r s and for a,b G S, w h e r e a ¥= 0; define (a) Show that f„ °f _ = /„ „ for some real u, v. Give explicit values for u, v in terms of a, b, c, and d. (b) I s / °/,, = f , °/„, always? (c) Find all f such that / ° / = / ° f ,, • (d) Show that /7J, exists and find its form.
b c d f l ; 6 rf c d 6 a>b a > i u u c b
11. Let S be the set of all positive integers. Define / : S > 5 by / ( l ) = 2, / ( 2 ) = 3, / ( 3 ) = 1, and f(s) = s for any other s G 5. Show that f°f°f = i . W h a t i s / " in this case?
1 s
MiddleLevel Problems 12. Let S be the set of nonnegative rational numbers, that is, S = {m/n  m, n nonnegative integers, n + 0}, and let Tbe the set of all integers. (a) D o e s f:S—> T defined by f(m/n) = 2"'3" define a legitimate function from S to r? (b) If not, how could you modify the definition of / so as to get a legitim a t e function? 13. Let S be the set of all positive integers of the form 2"'3", where m > 0, n > 0, and let T be the set of all rational n u m b e r s . Define / : 5 T by /(2'"3") = m/n. Prove that / defines a function from S to T. (On what properties of the integers does this d e p e n d ? ) 14. Let f:S> S, where S is the set of all integers, b e defined by f(s) = as + b, where a, b are integers. Find the necessary and sufficient conditions on a, b in order t h a t / / = i .
0 s
15. Find a l l / o f the form given in P r o b l e m 14 such t h a t / ° / ° / = 16. I f / i s a 11 mapping of S onto itself, show that ( /
_ 1
i.
s
)
_ 1
= /.
17. If S is a finite set having m > 0 elements, how m a n y mappings are there of S into itself? 18. In Problem 17, how many 11 mappings are there of S into itself? 19. Let S b e the set of all real n u m b e r s , and define / : S —» S by f(s) = s + as + b, where a, b are fixed real numbers. Prove that for n o values at a, b can / be onto or 11.
2
Sec.
3
Mappings
15
20. Let S be the set of all positive real n u m b e r s . F o r positive reals a, c and nonnegative reals b, d, is it ever possible that the mapping / : S S defined by f(s) = (as + b)/(cs + d) satisfies f°f= z ? Find all such a, b, c, d that do the trick.
s
21. L e t S be the set of all rational n u m b e r s and let f : S —> S be defined by f (s) = as + b, where a # 0, b are rational numbers. Find all f of this form satisfying f f„ = /„, ° / , for every / .
a b ab C(i a c> d ib b c d fl b
22. L e t S be the set of all integers and a, b, c rational numbers. Define f:S—>S by f(s) = as + bs + c. Find necessary and sufficient conditions on a, b, c, so t h a t / d e f i n e s a mapping on S [Note: a, b, c n e e d not b e integers; for example, f(s) = %s(s + 1) = ^s + §s does always give us an integer for integral s.}
2 2
Harder Problems 23. L e t S be the set of all integers of t h e form 2"'3", m > 0, n > 0, and let T be the set of all positive integers. Show that there is a 11 correspondence of S onto T. 24. Prove that there is a 11 correspondence of the set of all positive integers onto the set of all positive rational n u m b e r s . 25. L e t S be the set of all real n u m b e r s and T the set of all positive reals. Find a 11 mapping / of S o n t o T such t h a t / I s , + j ' ) = / ( s ) / ( s ) for all s s G S.
2 ] 2 u 2
26. F o r the / i n Problem 25,
find/
1
explicitly.
27. If / g are mappings of S into S and f° g is a constant function, then (a) W h a t can you say about / if g is onto? (b) W h a t can you say about g if / i s 11? 28. If S is a finite set and / is a mapping of S onto itself, show that / must be 11. 29. If S is a finite set and / is a 11 mapping of S into itself, show that / must be surjective. 30. If S is a finite set a n d / i s a 11 mapping of 5, show that for some integer n > 0, . / ° / ° / ° • • • ° / = ^. n times 31. If 5 has 7 « elements in Problem 30, find an n > 0 (in terms of m) that works simultaneously for all 11 mappings of S into itself.
16
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
4 . A(S)
(THE SET O F 11 M A P P I N G S OF S O N T O ITSELF)
W e focus our attention in this section on particularly nice mappings of a nonempty set, S, into itself. Namely, we shall consider the set, A(S), of all 11 mappings of S onto itself. Although most of the concern in the b o o k will b e in the case in which S is a finite set, we do not restrict ourselves to that situation here. W h e n S has a finite n u m b e r of elements, say n, then A (S) has a special n a m e . It is called the symmetric group of degree n and is often d e n o t e d by S„. Its elements are called permutations of S. If we are interested in the structure of S„, it really does not m a t t e r much what our underlying set S is. So, you can think of S as being the set {1, . . . , n). Chapter 3 will be devoted t o a study, in some depth, of S„. I n the investigation of finite groups, S plays a central role. T h e r e are many properties of the set A (S) on which we could concentrate. W e have chosen to develop those aspects here which will motivate the notion of a group and which will give the reader some experience, and feeling for, working in a grouptheoretic framework. G r o u p s will be discussed in Chapter 2. W e begin with a result that is really a compendium of some of the results obtained in Section 3.
n
L e m m a 1.4.1. (a) fgGA
A (5) satisfies the following: (S). h = f° (g° h). that f° i = =
(S) implies that f°gEA implies that (f°g)°
(b) /, g, h G A(S)
(c) T h e r e exists an e l e m e n t — t h e identity mapping i—such z o / = / for every / G A(S). (d) Given / G A(S), 8°f= ithere exists a g G A(S) (g =
such that f°g
Proof All these things were d o n e in Section 3, either in the text m a t e rial or in the problems. W e leave it to the reader to find the relevant part of Section 3 that will verify each of the statements (a) through (d). • W e should now like to k n o w how many elements t h e r e are in A(S) when S is a finite set having n elements. To do so, we first m a k e a slight digression. Suppose that you can d o a certain thing in r different ways and a second independent thing in s different ways. In how many distinct ways can you do both things together? T h e best way of finding out is to picture this in
Sec. 4
A{S)
( T h e S e t of 11 M a p p i n g s o f S O n t o Itself)
17
a concrete context. Suppose that there a r e ;• highways running from Chicago to Detroit and s highways running from Detroit to A n n Arbor. In how m a n y ways can y o u go first to Detroit, then to A n n A r b o r ? Clearly, for every road you t a k e from Chicago to D e t r o i t y o u have s ways of continuing on to A n n A r b o r . Y o u can start your trip from Chicago in r distinct ways, hence you can complete it in s + s + s + • • • + s = rs r times different ways. It is fairly clear that we can extend this from doing two independent things t o doing in independent ones, for an integer m > 2. If w e can d o the first things in \ \ distinct ways, t h e second in r ways, . . . , the m t h in r,„ distinct ways, then we can do all these together in i\r .. . r,„ different ways. Let's recall something m a n y of us have already seen:
2 2
Definition. If n is a positive integer, then nl (read "n factorial") fined by nl = 1 • 2 • 3 • • • n. L e m m a 1.4.2. If S h a s n elements, then A (S) has nl elements.
1 2
is de
Proof. Let / G A(S), w h e r e S = \x , x , . . . , x„}. H o w many choices d o e s / h a v e as a place to send x l Clearly n, for we can send x u n d e r / t o any element of S. B u t now / is not free to send x anywhere, for since / is 11, we must have f(xA ¥= f(x ). So we can send x anywhere except onto f(xi). H e n c e / can send x into n — 1 different images. Continuing this way, w e see that / can send x, into n — (i — 1) different images. H e n c e t h e n u m b e r of s u c h / ' s is n(n ~ l)(n  2) • • • 1 = nl •
x l 2 2 2 2
Example T h e n u m b e r n\ gets very large quickly. T o be able to see the picture in its entirety, we look at t h e special case n = 3, where nl is still quite small. Consider A(S) = S , where S consists of the three elements x x ,x . We list all the elements of 5 , writing out each mapping explicitly by what it does to each of x x , x .
3 h 2 3 3 u 2 3
1. 2. 3.
i: Xi g:x±^
—» X\,
x —> x , x —> x .
2 2 3 3
f:x\—>x ,x ^>x ,x ^X\.
2 2 i 2l
x ,x ~+
2 2 u 2 3 3
x x ^>
u 3 2
x.
3
4. g ° / : * i >x x *x ,x >x .
(Verify!)
18
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
5. f° g : A " !  > a  , a :
3
2
x , A '  > a c j . (Verify!)
2 3
6.
/ " /
: a c
1
^ a c
3
, a c
2
^ a c
1
, a 
3
 > a :
2
.
(Verify!) only does rule and we
Since we have listed h e r e six different elements of S , and S has six elements, we have a complete list of all t h e elements of S . W h a t this list tell us? T o begin with, we n o t e that f°g^g°fso one familiar of the kind of arithmetic we have b e e n used to is violated. Since g G S g G S , we must have g°g also in S . W h a t is it? If we calculate g°g, easily get g ° g = i. Similarly, we get
3 3 3 3 3 3
(f°g)°(f°g) N o t e also that f°(f°f) r e a d e r to show that g°f=
1
= ' =
1
(g°f)°(g°.f)= f°f. Finally, we leave it to t h e
= i, hence f f^ °g.
It is a little cumbersome to write this p r o d u c t in A (S) using t h e °. From now on we shall drop it and write f° g merely as fg. Also, we shall start using t h e shorthand of exponents, to avoid expressions like f°f°f°°f. We define, for / G A(S), f° = i, f = f°f = ff, and so on. F o r negative exponents —n we define by f~" = ( / ) " , w h e r e n is a positive integer. T h e usual rules of exponents prevail, namely / ' / • ' = f and (f') = W e leave these as exercises—somewhat tedious ones at t h a t — f o r the reader.
2 _ 1 r+s s
Example D o not j u m p to conclusions that all familiar properties of exponents go over. F o r instance, in t h e example of the / , g G S defined above, we claim that (fg) ^ f g . T o see this, we n o t e that
3 2 2 2
fg '•
2 t u 2 2 2
X
l ~^ 3'
2 3
X
X
2
X
2> 3
X
~* L)
2 2
X
so that (fg) : x —> x x ~> x , x —> x , that is, (fg) hand, f / and g = i, hence f g = f /, whence (fg)
3 2 2 2
= i. On t h e other ^ fg in this case.
2 2
However, some other familiar properties do go over. F o r instance, if / , g, h are in A (S) and fg = fh, then g = h. W h y ? Because, from fg = fh we have f~\fg) = f\fh); therefore, g = ig = (f^ftg = / ^ ( / g ) = r\fh) = (f f)h ih = h. Similarly, gf = hf implies that g = h. So we can cancel an element in such an equation provided that we do not change sides. In S3 our /, g satisfy gf = f~ g, but since / f~ we cannot cancel the g h e r e .
1 = 1 l
PROBLEMS
Recall t h a t / g stands for f°g will be a n o n e m p t y set.
and, also, w h a t / ' " means. S, without subscripts,
Sec. 4
A(S)
( T h e S e t o f 11 M a p p i n g s of S O n t o Itself)
19
Easier P r o b l e m s 1. If s ¥= s are in S, show that there is a n / G A(S)
{ 2
such that f(s )
{
= s.
2
2. Ifsi G S, let / / = { / G A (5) (a) / G //. (b) I f / . g G W . t h e n / i , ' G //. (c) I f / G # , t h e n / G # .
_ 1 2
= s,}. Show that:
3. Suppose that s\ + s are in S and f(s{) = s , w h e r e / G A (5). Then if H is as in P r o b l e m 2 and K = {g G A ( 5 )  g ( s ) = s } , show that:
2 2 2
(a) If g&K, then ftgfGH. (b) If h G //, then there is some g G K such that /z = 4. If/, g, h G. A(S), show that ( / about ( r ' g / y ?
_ 1
f~ gf. W h a t can you say
1
g / ) ( / ~ V ) = f~ (gh)f.
x
5. If /, g G A (5) a n d / g = g / show that: (a) (b> {fgffg . ( / g r w  v
2 1
=
6. Push the result of P r o b l e m 5, for t h e same / and g, to show that (fg)'" f"'g"' for all integers m. *7. Verify the rules of exponents, namely ff / G A (S) and positive integers r, s. 8. If / g G A (5) and (fg)
2 3 4 2 2 2 s
= / ' ' ' and (f'Y
+
= / " for
= / g , p r o v e t h a t / g = gf.
4
9. If 5 = {x^ x , x , x } , l e t / , g G S be defined by
and g : Calculate: (a)/ ,/ ,/ . (b) g , g . (c)/g. (d) g / (e) (fg)\ (f) r .*
3 1 1 2 3 2 3 4
Xj
 » x , x > x
2 2
1 ;
x
3
x , x > x .
3 4 4
(gf) .
3
10. I f / G S , show t h a t /
6
= z.
4
11. Can you find a positive integer m such that /"' = i for all f G S ?
20
T h i n g s F a m i l i a r a n d Less
Familiar
Ch. 1
MiddleLevel Problems * 12. If / G S„, show that there is some positive integer k, depending o n / , such that/''" = i. (Hint: Consider the positive powers o f / . ) * 1 3 . Show that there is a positive integer / such that / ' = i for all f G S„. 14. If m < n, show that there is a 11 mapping F: S„, — S„ such that F(fg) > F(f)F(g) for a l l / g G 5 .
m
=
15. If 5 has three or more elements, show that we can find / g G / I (5) such that/g # g / . 16. Let S be an infinite set and let M C A(S) be the set of all elements / G A (S) such that f(s) # s for at most a finite n u m b e r of s G S. Prove that: ( a ) / g G M implies t h a t / g G M. (b) / G Mimplies t h a t / ' G M .

17. For the situation in P r o b l e m 16, show, if / G A (5), that / £
G
_
1
M/
=
^}
m
u
s
t
equal Af.
18. Let S T and consider the subset [7(7) = { / G A (5)  / ( f ) G T f o r every t G T}. Show that: (a) / G U(T). (b) / g G U(T) implies t h a t / g G U(T). 19. If the 5 in P r o b l e m 18 has n elements and T has m elements, h o w many elements are there in U(T)7 Show that there is a mapping F: U(T) S,„ such that F(fg) = F(f)F(g) f o r / g G U(T) and F i s o n t o c 20. If /7? < «, can i in P r o b l e m 19 ever be 11? If so, when?
7
21. In S„ show that the mapping / defined by
[i.e., /(*;) = A .  if i < n, f(x„) = A^] can b e written as / = g j g • • • g„_! w h e r e each g G 5„ interchanges exactly two elements of S = {x , . . . , A'„), leaving the other elements fixed in S.
/+]
2
f
x
Harder Problems 22. I f / G S„, show t h a t / = h\h '' ' *23. Call an element in S a transposition ing the others fixed. Show that any sitions. (This sharpens the result of
2 n 3
for some E S„ such that /r = /. if it interchanges two elements, leavelement in S is a product of transpoProblem 22.)
n
2
24. If n is at least 3, show that for some / in S„, / cannot be expressed in the f o r m / = g for any g in S„.
Sec. 5
3
T h e Integers
21
25. If / G S„ is such that / it i b u t / = j , show that we can n u m b e r the elements of S in such a way that / ( . t j ) = .v , f(x ) = v , f(x ) = X ) , / ( x ) = x ,f(x ) = x , f(x ) = x , ... ,f(x ) = x ,f(x ) = x , f(x ) =x for some k, and, for all the other x, G 5, f(x) = x .
2 2 3 3 4 5 5 6 6 4 3k+1 3k+2 3k+2 3 k + 3 3k+3 3 k + 1 t
26. View a fixed shuffle of a deck of 52 cards as a 11 mapping of the deck o n t o itself. Show that repeating this fixed shuffle a finite (positive) number of times will bring the deck back t o its original order. *27. If / G A(S), call, for s G S, the orbit of s (relative to / ) the set 0(s) = { f (s)  all integers / } . Show that if s, t G S, then either 0(s) n 0(t) = 0 or 0(s) = 0(t).
j
28. If 5 = [xi, x ,..., x ] a n d / G 5 is defined by/(x,) = x , if i = 1, 2, . . . , 11 a n d / ( x ) = x , find the orbits of all t h e elements of 5 (relative to / ) .
2 12 1 2 + 1 n 1
29. If / G A (S) satisfies /
3
= ;', show that the orbit of any element of S has is an integer p>l such that p cannot b e facp
one or three elements. *30. Recall that a prime number tored as a product of smaller positive integers. If / G A(S) satisfies f ~ i,
what can you say about the size of the orbits of the elements of S relative to / ? W h a t property of the prime numbers are you using to get your answer? 3 1 . P r o v e that if S has m o r e than two elements, then the only elements / „ in A(S) such t h a t / / = ff
0 0
for a l l / G A(S) m u s t satisfy f
0
= i.
*32. W e say that g G A(S) commutes with / G A(S) if fg = gf. Find all the elements in A(S) that c o m m u t e with f:S > S defined by /(*,) = x ,f(x ) = x , a n d / ( s ) = s if s # x x .
2 2 x u 2
33. In 5„ show that t h e only elements commuting with / defined by f(x) = x i if / < n, f{x,) = X i , are the powers off, namely i = f°, f f ,.. . ,/""'.
2 j+
34. F o r / G A(S), let C ( / ) = [g G A(S) \fg = gf). Prove that: (a) g, h G C ( / ) implies that gh G C(.f). (b) g G C(f) implies that g G C ( / j . (c) C ( / ) is not empty.
 1
5. THE INTEGERS
T h e mathematical set most familiar to everybody is that of the positive integers 1 , 2 , . . . , which we shall often call Equally familiar is t h e set, Z, of all integers—positive, negative, and zero. Because of this acquaintance with Z, we shall give here a rather sketchy survey of the properties of Z that we shall use often in t h e ensuing material. Most of these properties are well k n o w n to all of us; a few are less well known. T h e basic assumption we m a k e about the set of integers is the
22
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
WellOrdering Principle. has a smallest member.
A n y n o n e m p t y set of nonnegative integers
M o r e formally, what this principle states is that given a n o n e m p t y set V of nonnegative integers, there is an element v G V such that v < v for every v G V. This principle will serve as the foundation for our ensuing discussion of t h e integers. T h e first application we m a k e of it is to show something we all k n o w and have taken for granted, namely that we can divide o n e integer by another to get a r e main der that is smaller. This is known as Euclid's Algorithm. W e give it a m o r e formal statement a n d a proof based on wellordering.
Q 0
Theorem 1.5.1 (Euclid's A l g o r i t h m ) . If m and n are integers with n > 0, then there exist integers q and r, with 0 < r < such that m = qn + r. Proof. Let W be the set of m  tn, where t runs t h r o u g h all the integers, i.e., W = [m — tn 11 G Z}. N o t e that W contains some nonnegative integers, for if t is large enough and negative, then m — tn > 0. L e t V = {v G W  v ^ 0}; by t h e wellordering principle V has a smallest element, r. Since r G V, r £ 0, a n d r = m  qn for some q (for that is t h e form of all elements in W D V). W e claim that r < n. If not, r = m  qn > n, hence m  (q + > 0. B u t this puts m  (q + l)n in V, yet m  (q + l ) n < r, contradicting the minimal n a t u r e of r in V. With this, Euclid's A l g o r i t h m is proved. • Euclid's Algorithm will have a host of consequences for us, especially about the notion of divisibility. Since we are speaking about t h e integers, be it understood that all letters used in this section will be integers. This will save a lot of repetition of certain phrases. Definition. Given integers m 0 a n d n we say that m divides n, written as m  n, if n = cm for some integer c. Thus, for instance, 2 [ 14, (  7 )  14, 4  (  1 6 ) . If m  n, we call m a divisor or factor of n, and n a multiple of m. T o indicate that m is not a divisor of n, w e write 7771n; so, for instance, 3 f 5. T h e basic elementary properties of divisibility are laid out in Lemma 1.5.2. (a) 1 T h e following are true:
I
n for all n.
(b) If 777 * 0, then m \ 0.
Sec.
5
T h e Integers
23
(c) If m  n and n  q, then m \ q. (d) If m  « and m  q, then m  (un + vq) for all it, v. (e) If m 11, then m = 1 or m = — 1 . (f) If m  n and «  m, then 777 = ±n. Proof. T h e proofs of all these parts are easy, following immediately from the definition of m \ n. W e leave all but Part (d) as exercises but prove Part (d) h e r e to give the flavor of h o w such proofs go. So suppose that m  n and m \ q. T h e n n = cm and q = dm for some c and d. Therefore, un + vq — u(cm) + v(dm) = (uc + vd)m. Thus, from the definition, m \ (un + vq). • Having the concept of a divisor of an integer, we now want to introduce that of the greatest common divisor of two (or m o r e ) integers. Simply enough, this should be the largest possible integer that is a divisor of b o t h integers in question. However, we want to avoid using the size of an integer— for reasons that may become clear m u c h later when we talk about rings. So we m a k e the definition in what may s e e m as a strange way. Definition. Given a, b (not b o t h 0), then their greatest common sor c is defined by: (a) c>0. divi
(b) c
I
a and c \ b. and d \ b, then d \ c.
(c) lid\a
W e write this c as c — (a, b). In other words, the greatest common divisor of a and b is the positive n u m b e r c which divides a and b and is divisible by every d which divides a and b. Defining something does not g u a r a n t e e its existence. So it is incumbent on us to prove that (a, b) exists, and is, in fact, unique. The proof actually shows more, namely that (a, b) is a nice combination of a and b. This combination is not unique; for instance, (24,9) = 3 = 3  9 + (  1 ) 2 4 = (  5 ) 9 + 2  2 4 . T h e o r e m 1.5.3. If a, b are not b o t h 0, then their greatest common divisor c = (a, b) exists, is unique, and, moreover, c = rn^a + n b for some suitable mo and n .
0 n
24
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
Proof. Since not b o t h a and fa are 0, the set A = {ma + nb \ m, n G Z) has nonzero elements. If .v G A and x < 0, then —x is also in A and —x > 0, for if x = m a + n b, then —x = (—m )a + (—n^fa, so is in A T h u s has positive elements; hence, by the wellordering principle t h e r e is a smallest positive element, c, in A. Since c G / i , by the form of the elements of A we k n o w that c = m a + n b for some / 7 J f ) , We claim that c is our required greatest c o m m o n divisor. First note that if rf  « and d \ fa, then d \ (m a + n b) by Part (d) of L e m m a 1.5.2, that is, d  c. So, to verify that c is our desired element, we need only show that c \ a and c  b. By Euclid's Algorithm, a = qc + ;•, where 0 < /• < c, that is, « = q(m a + /? /3) + r. Therefore, ;• = —gn b + (1 — qm )a. So r is in A. B u t r < c and is in A, so by the choice of c, r cannot be positive. H e n c e r = 0; in other words, a = qc and so c  a. Similarly, c  b. For the uniqueness of c, if t > 0 also satisfied r ] a, 11 /j> and d  f for all d such that d  a and d  fa, we would have t \ c and c 11. By Part (f) of L e m m a 1.5.2 we get that t = c (since both are positive). •
x { { 0 Q Q 0 Q 0 a 0
Let's look at an explicit example, namely a = 24, fa = 9. By direct examination we know that (24, 9) = 3; note that 3 = 3  9 + (  1 ) 2 4 . W h a t is (  2 4 , 9)? H o w is this done for positive numbers a and fa which may be quite large? If fa > A, interchange a and fa so that a > fa > 0. Then we can find (a, fa) by 1. observing that (a, fa) = (fa, r) w h e r e a ~ qb + r with 0 < r < fa (Why?); 2. finding (fa, r), which now is easier since one of the n u m b e r s is smaller than before. So, for example, we have (100, 28) ( 28, 16) ( 16,12) This gives us (100, 28) (12, 4) = 4.
a
(28,16) (16, 12) (12, 4)
since 100 since since 28 16
3 (28) + 16 1 (16) + 12 1 (12) + 4
It is possible to find the actual values of m 4 = m
0
and n such that
0
100 + n
0
28
Sec. 5
The Integers
25
by going backwards through the calculations m a d e to find 4: Since 16 = 1 (12) + 4, 4 = 12= 16 + ( 28 + ( 1) 12 1) 16
Since 28 = 1 (16) + 12, Since 100 = 3 (28) + 16, B u t then
16 = 100 + ( 3) 28
4 = 16 + (  1 ) 12 = 16 + (  1 ) ( 2 8 + (  1 ) 16) = (  1 ) 28 + (2) 16 = (™1) 28 + (2)(100 + (  3 ) 28) = (2) 100 + (  7 ) 28 so that m = 2 and « = —7. This shows how Euclid's Algorithm can be used to compute (a, b) for any positive integers a and b. W e shall include some exercises at the end of this section on other properties of (a, b). W e come to the very important
0 n
Definition.
W e say that a and b are relatively prime if (a, b) = 1.
So the integers a and b are relatively p r i m e if they have no nontrivial c o m m o n factor. A n immediate corollary to T h e o r e m 1.5.3 is T h e o r e m 1.5.4. The integers a and b are relatively prime if and only if 1 = ma + nb for suitable integers in and n. T h e o r e m 1.5.4 has an immediate consequence T h e o r e m 1.5.5. If a and b are relatively prime and a  be, then a \ c.
Proof. By T h e o r e m 1.5.4, ma + nb = 1 for some in and n, hence (ma + nb)c = c, that is, mac + nbc = c. By assumption, a \ be and by observation a  mac, hence a  (mac + nbc) and so a \ c. • Corollary. If b and c are both relatively prime t o a, then be is also relatively prime to a. Proof. W e pick up the proof of T h e o r e m 1.5.5 at mac + nbc = c. If d = (a, be), then d  a and d\ be, hence d  (mac + nbc) = c. Since d \ a and d \ c
26
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
and (a, c) = 1, we get that d = 1. Since 1 = d = (a, be), we have that be is relatively prime to a. • W e now single out an ultraimportant class of positive integers, which we m e t before in P r o b l e m 30, Section 4. Definition. A prime number, or a prime, is an integer p > 1, such that for any integer a either p \ a or p is relatively prime to a. This definition coincides with the usual one, namely that we cannot factor p nontrivially. For if p is a prime as defined above and p = ab w h e r e 1 < a < p, then (a,p) = a (Why?) and p does not divide a since p > a. It follows that a = 1, so p = b. O n the other hand, if p is a prime in the sense that it cannot b e factored nontrivially, and if a is an integer not relatively prime to p, then (a, b) is not 1 and it divides a and p. But then (a, b) equals p, by our hypothesis, so p divides a. A n o t h e r result coming out of T h e o r e m 1.5.5 is T h e o r e m 1.5.6. If p is a prime and p\(a a some i with 1 < ; < ; ? .
l u 2
• • • a,), then p \ a for
f
Proof. If p  a there is nothing to prove. Suppose that p\a ; then p and ffj are relatively prime. B u t p \ a (a • • • a„), hence by T h e o r e m 1.5.5, p  a • • • a . R e p e a t the argument just given on a , and continue. •
{ l 2 2 n 2
T h e primes play a very special role in the set of integers larger than 1 in that every integer n > 1 is either a prime or is the product of primes. W e shall show this in the next t h e o r e m . In the t h e o r e m after the next we shall show that there is a uniqueness about the way n > 1 factors into prime factors. T h e proofs of both these results lean heavily on the wellordering principle. T h e o r e m 1.5.7. primes. If n > 1, t h e n either n is a prime or n is the product of
Proof. Suppose that the t h e o r e m is false. Then there must be an intger m > 1 for which the t h e o r e m fails. Therefore, the set M for which the t h e o r e m fails is nonempty, so, by the wellordering principle, M has a least element m. Clearly, since m G M, m cannot be a prime, thus m = ab, where 1 < a < m and 1 < b < m. Because a < m and b < m and m is the least element in M, we cannot have a G M or b G M. Since a ^ M, b £ M, by the definition of M the t h e o r e m must b e true for both a and b. T h u s a and b are
Sec.
5
T h e Integers
27
primes or the product of primes; from m = ab we get that m is a product of primes. This puts m outside of M, contradicting that m G M. This proves the t h e o r e m . • W e asserted above that there is a certain uniqueness about the decomposition of an integer into primes. W e m a k e this precise now. T o avoid trivialities of the kind 6 = 2 • 3 = 3 • 2 (so, in a sense, 6 has two factorizations into the primes 2 and 3), we shall state the t h e o r e m in a particular way. T h e o r e m 1.5.8. Given n > 1, then there is one and only one way to write n in the form n = p ^p • • • p , where p < p < • • • < p are primes and the exponents a , a • • •, a are all positive.
a c ak k 2 x 2 k x 2> k
Proof. W e start as we did above by assuming that the t h e o r e m is false, so there is a least integer m > 1 for which it is false. This m must have two distinct factorizations as m = pVpT ' ' ' Pt <7i 72 ' ' ' where P \ P 2 ' ' ' P k > q \ 2 ' ' ' e are primes and where t h e exponents a , ... , a and b , .. . , b are all positive. Since p \ pi • • • p = q\ • • • q ', by T h e o r e m 1.5.6 p \ qf> for some /; hence, again by T h e o r e m 1.5.6, p  q hence p = cp. By the same t o k e n q = p for some /; thus p ^ /?• = q < q, = p . This gives us that p = q . N o w since m/p < m, ni/p has the unique factorization property. But m/p = p" ^ p  • • • p" = Pi <?2 ' ' ' q e d since m/p can b e factored in one and only one way in this form, we easily get k = £, p = q , .. . , p = q , a — 1 = b — 1, a = b , ... , a = b . So we see that the primes and their exponents arising in the factorization of m are unique. This contradicts the lack of such uniqueness for m, and so proves the t h e o r e m . •
k = 1( 2 < < < < a < < a 1 ak k l b x k x ( x e x x h x x x x x x v x x 1 1 k x 2 k 1 1 2 h c a n 1 2 2 k k x x 2 2 k k
W h a t these last two t h e o r e m s tell us is that we can build u p the integers from the primes in a very precise and welldefined manner. O n e would expect from this that there should b e m a n y — t h a t is, an infinity—of primes. This old result goes back to Euclid; in fact, the argument we shall give is due to Euclid. T h e o r e m 1.5.9. T h e r e is an infinite n u m b e r of primes.
Proof. If the result were false, we could e n u m e r a t e all the primes in p , p , ... ,p . Consider the integer q = 1 + p p • • • p . Since q > p, for every i = 1, 2, . . . , k, q cannot be a prime. Since p \ q, for we get a remainder of 1 on dividing q by p , q is n o t divisible by any of p ,.. . , p . So q is not a prime n o r is it divisible by any prime. This violates T h e o r e m 1.5.7, thereby proving the t h e o r e m . •
x 2 k x 2 k s t x k
28
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
Results much sharper than T h e o r e m 1.5.9 exist about h o w m a n y primes there are u p to a given point. T h e famous prime n u m b e r t h e o r e m states that for large n the n u m b e r of primes less than or equal to n is " m o r e or less" nl\og n, where this " m o r e or less" is precisely described. T h e r e are many open questions about the p r i m e numbers.
e
PROBLEMS
Easier Problems 1. Find (A, b) and express (a, b) as ma + /ib for: (a) (b) (c) (d) (116,  8 4 ) . (85,65). (72, 26). (72, 25).
2. Prove all the parts of L e m m a 1.5.2, except part (d). 3. Show that (ma, mb) = m(a, b) if m > 0. 4. Show that if a \ m and b  m and (a, b) = 1, then (ab)  m. 5. Factor the following into primes. (a) 36. (b) 120. (c) 720. (d) 5040. 6. If 777 = p'l • • • p" and n = p\ • • • p , where p . , . , p are distinct primes and a±, .. . , a are nonnegative and b , . . . , b are nonnegative, express (m, n) as p^ • • • p'jj by describing the c's in terms of the a's and /j's.
1 k x bk k c u k k 1 k 1 1
* 7. Define the least common multiple of positive integers m and n t o be the smallest positive integer v such that both 7771 v and 771 v. (a) Show that v = mn/(m, 71). (b) In terms of the factorization of m and n given in P r o b l e m 6, what is u? 8. Find the least c o m m o n multiple of the pairs given in P r o b l e m 1. 9. If 777, 77 > 0 are two integers, show that we can find integers u, v with — 7 7 / 2 < 7; < 7 7 / 2 such that 777 = 7777. + v. 10. T o check that a given integer 77 > 1 is a prime, prove that it is e n o u g h to show that 77 is not divisible by any prime p with p < V77.
Sec.
6
Mathematical Induction
29
11. Check if the following are prime. (a) 301. (b) 1001. (c) 473. 12. Starting with 2, 3, 5, 7 , . . . , construct the positive integers 1 + 2  3 , 1 + 2  3  5 , 1 + 2 • 3 • 5 • 7 , . . . . D o you always get a prime number this way? MiddleLevel P r o b l e m s 13. If p is an odd prime, show that p is of the form: (a) An + 1 or An + 3 for some n. (b) 6n + 1 or 6n + 5 for some n. 14. A d a p t the proof of T h e o r e m 1.5.9 to prove: (a) T h e r e is an infinite n u m b e r of primes of the form An + 3. (b) T h e r e is an infinite n u m b e r of primes of the form 6n + 5. 15. Show that n o integer u = An + 3 can be written as u = a + b , where a, b are integers. 16. If T is an infinite subset of N, the set of all positive integers, show that there is a 11 mapping of T onto N. 17. If p is a prime, prove that one cannot find nonzero integers a and b such that a = pb . (This shows that V p is irrational.)
2 2 2 2
6. M A T H E M A T I C A L I N D U C T I O N
If we look back at Section 5, we see that at several places—for instance, in the proof of T h e o r e m 1.5.6—we say "argue as above and continue." This is not very satisfactory as a m e a n s of nailing down an argument. W h a t is cleatis that we n e e d some technique of avoiding such phrases when we want to prove a proposition about all the positive integers. This is provided for us by the Principle of Mathematical Induction; in fact, this will be the usual m e t h o d that we shall use for proving t h e o r e m s about all the positive integers. T h e o r e m 1.6.1. such that: (a) P(l) is true. (b) If P(k) happens to b e true for some integer k S: 1, then P(k + 1) is also true. T h e n P(n) is true for all n > 1. Let P(n) b e a statement about the positive integers
30
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
Proof. Actually, t h e arguments given in proving T h e o r e m s 1.5.7 a n d 1.5.8 are a prototype of the argument we give here. Suppose that t h e t h e o r e m is false; then, by wellordering, there is a least integer m S: 1 for which P(m) is not true. Since P(l) is true, m # 1, hence m > 1. N o w 1 < m  1 < m, so by t h e choice of m, P(m  1) must b e valid. B u t then by the inductive hypothesis [Part (b)] w e must have that P(m) is true. This contradicts that P(m) is n o t true. Thus t h e r e can b e n o integer for which P is not true, and so t h e t h e o r e m is proved. • W e illustrate h o w t o use induction with some r a t h e r diverse examples. Examples 1. Suppose that n tennis balls are p u t in a straight line, touching each other. T h e n we claim that these balls m a k e n. — 1 contacts. Proof. If n = 2, t h e m a t t e r is clear. If for k balls we have k — 1 contacts, t h e n adding o n e ball (on a line) adds o n e contact. So k + 1 balls would have k contacts. So if P(n) is what is stated above about t h e tennis balls, we see that if P(k) h a p p e n s to b e true, t h e n so is P(k + 1). Thus, by t h e t h e o rem, P(ri) is true for all n > 1. • 2. If p is a prime a n d p \ a a
x 2
• • • a„, t h e n p \ a, for some 1 < i < n.
Proof. Let P(ri) b e t h e statement in E x a m p l e 2. T h e n P ( l ) is true, for if p  a , it certainly divides a for some 1 < < 1. Suppose we k n o w that P(k) is true, and that p  a a • • • a a . Thus, by T h e o r e m 1.5.6, since p \ (a a • • • a )a either p \ Uk+i ( desired conclusion) or p  a i • • • a . In this second possibility, since P(k) is t r u e we have that p\a for some 1 < i s k. Combining b o t h possibilities, w e get that p  cij for some 1 < / < / < : + 1. So P a r t (b) of T h e o r e m 1.6.1 holds; hence P(n) is true for all n > 1. • 3. F o r n > 1, 1 + 2 + • • • + « = \n(n + 1).
x f x 2 k k + x x 2 k k+1 a k f
Proof. If P(n) is t h e proposition that 1 + 2 + • • • + n = \ n(n + 1), then P(l) is certainly true, for 1 = \ (1 + 1). If P(k) should b e true, this means that 1 + 2 + • • • + k = \k(k T h e question is: Is P(k + 1).
+ 1) then also true, that is, is 1 + 2 + • • • + k +
(k + 1) =  ( / c + l ) ( ( / c + 1) + 1)? N o w 1 + 2 + • • • + k + (k + 1) =
(1 + 2 + • • • k) + (k + 1) = \k(k + 1) + (k + 1), since P(k) is valid. B u t Jfc(fc + 1) + (/c + 1) =  ( / c ( / c + 1) + 2(k + 1)) = + \){k + 2), which assures us that P(k + 1) is true. Thus the proposition 1 + 2 + • • • + n = \n(n + ' T ) is true for all n > 1. •
Sec. 6
Mathematical Induction
31
W e must emphasize one point here: Mathematical induction is not a m e t h o d for finding results about integers; it is a m e a n s of verifying a result. W e could, by other means, find t h e formula given above for 1 + 2 + • • • + n. Part (b) of T h e o r e m 1.6.1 is usually called the induction step. In t h e problems we shall give some other versions of the principle of induction.
PROBLEMS
Easier P r o b l e m s 1. Prove that l + 2 + 3 + • • • + n = \n(n 2. Prove that I + 2 + • • • + « = \n (n
3 3 3 2 2 2 2 2
+ l)(2n + 1) by induction.
2
+ l ) by induction. — 1) subsets having ex
3. Prove that a set having n s 2 elements has \n(n actly two elements.
4. Prove that a set having /? > 3 elements has n(n — l)(n — 2)/3! subsets having exactly three elements. 5. If n > 4 a n d S is a set having n elements, guess (from Problems 3 a n d 4) how many subsets having exactly 4 elements there are in S. Then verify your guess using mathematical induction. *6. Complete t h e proof of T h e o r e m 1.5.6, replacing t h e last sentence by an induction argument. 7. If a tion. 1, prove that 1 + a + a + • • • + a" = (a"
2 +1
 1)1 (a  1) by induc
8. By induction, show that
J 
1223
+ J  + . . . +
1
n(n + 1)
=
2L_
n. + 1 "
*9. Suppose that P(n) is a proposition about positive integers /( such that P(n ) is valid, and if P(k) is true, so must P(k + 1) be. W h a t can you say
0
about P(n)l Prove your statement. *10. L e t P(n) b e a proposition about integers n such that P(l) is true and such that if P(j) is true for all positive integers / < k, then P(k) is true. Prove that P(n) is true for all positive integers n. MiddleLevel P r o b l e m s 11. Give an example of a proposition t h a t is not true for any positive integer, yet for which the induction step [Part (b) of T h e o r e m 1.6.1] holds. 12. Prove by induction that a set having n elements has exactly 2 " subsets.
32
T h i n g s F a m i l i a r a n d Less F a m i l i a r
3
Ch. 1
13. Prove by induction on n that n — n is always divisible by 3. 14. Using induction on n, generalize the result in Problem 13 to: If p is a prime number, then n — n is always divisible by p. (Hint: T h e binomial theorem.)
p
15. Prove by induction that for a set having n elements the n u m b e r of 11 mappings of this set o n t o itself is nl.
7. C O M P L E X
NUMBERS
W e all know something about the integers, rational n u m b e r s , a n d real numbers—indeed, this assumption has b e e n m a d e for some of the text material and many of the problems have referred to these n u m b e r s . Unfortunately, the complex numbers and their properties are much less known to presentday college students. A t one time the complex numbers were a part of the high school curriculum and the early college one. This is n o longer the case. So we shall do a rapid development of this very important mathematical set. T h e set of complex numbers, C , is the set of all a + bi, w h e r e a, b are real and where we declare: 1. a + bi = c + cli, for a, b, c, d real, if and only if a = c and b = d. 2. (a + bi) ± (c + di) = (a ± c) + (b ± d)i. 3. (a + bi)(c + di) = (ac  bd) + (ad + bc)i. This last property—multiplication—can best b e r e m e m b e r e d by using i =  1 and multiplying out formally with this relation in mind. For the complex n u m b e r z = a + bi, a is called the real part of z and b the imaginary part of z. If a is 0, we call z purely imaginary. We shall write 0 + Oi as 0 and a + Oi as a. N o t e that z + 0 = z, zl = z for any complex n u m b e r z. Given z = a + bi, t h e r e is a complex n u m b e r r e l a t e d to z, which w e write as z, defined by z = a  bi. This complex n u m b e r , z, is called t h e complex conjugate of z. T a k i n g t h e complex conjugate gives us a m a p p i n g of C o n t o itself. W e claim
2
L e m m a 1.7.1. (a) \J)=z.
If z, w G C , then:
(b) (z + w) = z + w. (c) (zw) = zw (d) zz is real and nonnegative and is, in fact, positive if z i= 0.
Sec. 7
Complex Numbers
33
(e) z + z is twice the real part of z. (f) z ~ z is twice the imaginary part of z times i. Proof. Most of the parts of this lemma are straightforward and merely involve using the definition of complex conjugate. W e do verify Parts (c) and (d). Suppose that z = a + bi,w = c + di, where a, b, c, d are real. So zw = (ac  bd) + (ad + bc)i, hence (zw) = (ac — bd) + (ad + bc)i = (ac — bd) — (ad + bc)i. O n the other hand, z = a  bi and w = c  di, hence, by the definition of the product in C, z w = (ac  bd) — (ad + be) i. Comparing this with the result that we obtained for (zw), we see that indeed (zw) = z. w. This verifies Part (c). W e go next to the proof of Part (d). Suppose that z = a + bi # 0; then z = a — bi and zz = a + b . Since a, b are real and not both 0, a + b is real and positive, as asserted in Part (d). •
2 2 2 2
T h e proof of Part (d) of L e m m a 1.7.1 shows that if z = a + bi' + 0, then zz = a + b # 0 a n d z ( l / ( V + b )) = 1, so
2 2 2
acts like the inverse 1/z of z. This allows us to carry out division in C, staying in C while doing so. W e now list a few properties of C. L e m m a 1.7.2. C behaves u n d e r its sum and product according to the following: If u, v, w G C, then (a) u + v = v + it. (b) (u + v) + w = u + (v + w). (c) uv = vu. (d) (uv)w = u(vw).
l l
(e) u # 0 implies that u~ = Vu exists in C such that uu~
= 1. •
Proof. W e leave the proofs of these various parts to the reader.
These properties of C m a k e of C what we shall call a field, which we shall study in much greater depth later in the book. W h a t the lemma says is that we are allowed to calculate in C m o r e or less as we did with real n u m bers. However, C has a much richer structure than the set of real numbers.
34
T h i n g s F a m i l i a r a n d Less Familiar
Ch. 1
W e now introduce a "size" function on C . Definition. If z = a + bi G C , then t h e absolute value of z, written as \z\, is defined by \z\ = V z i " = V o + b .
2 2
W e shall see, in a few m o m e n t s , what this last definition m e a n s geometrically. In the m e a n t i m e we prove Lemma 1.7.3. If u, v G C , then \uv = \u \ \v\.
Proof. By definition, \u \ = \Zial a n d u  = V u u . N o w \uv\ = V(uv)(uv)  V(MU) (i< v) = \/(uu)(vv) = \/uu\/vxJ (by Part (c) of L e m m a 1.7.1) (by L e m m a 1.7.2) = \u\ \v\. •
A n o t h e r way of verifying this l e m m a is to write u = a + bi, v = c + di, u v = (ac — bd) + (ad + bc)i and t o n o t e t h e identity (ac  bd)
2
+ (ad + be)
2
= (a + b )(c
2
2
2
+
d ).
2
N o t e several small points about conjugates. If z G C , t h e n z is real if and only if z = z, and z is purely imaginary if and only if z = — z. If z, w G C , then (zw + zw) = zw + zw = zw + zw,
so zw + zw is real. W e want to get an upper b o u n d for \zw + zw\; this will come up in the proof of T h e o r e m 1.7.5 below. But first we must digress for a m o m e n t to obtain a statement about quadratic expressions. Lemma 1.7.4. L e t a, b, c b e real, with a > 0. If act + ba + c > 0 for every real a, then b — 4ac £ 0.
2 2
Proof. Consider the quadratic expression for a = b/2a. W e get a(b/2a) + b(bl2a) + c > 0. Simplifying this, we obtain that (4ac b )/4a > 0, and since a > 0, we end u p with 4ac  b > 0, and so b  4ac < 0. •
2 2 2 2
W e use this result immediately to prove the important Theorem 1.7.5 (Triangle Inequality). For z,w E.C, \z + w \ <\z\ + \w\.
Proof. If z = 0, t h e r e is nothing to prove, so we m a y assume that z j= 0; thus zz > 0. Now, for a real,
Sec. 7
Complex Numbers
35
0 < \az + w\
2
= (az + w) (az + w) = (az + w)(az = crzz + ctfzw + zw) + ww.
+ w)
If a = zz > 0, b = zw + zw, c = wW, then L e m m a 1.7.4 tells us that b  4ac = (zw + zw)  4(zz)(ww) < 0, hence (zw + zw) < 4(zz)(ww) = 4  z   w  . Therefore, zw + zw < 2\z\ \w\. F o r a = 1 above,
2 2 2 2 2
z + w j = zz + ww + zW + zw =  z i +  w  + zW + zw < z +
2
2
2
2
M +
2
2\z\ \w\
2 2
from the result above. In other words, z + w \ s (z  +  w  ) ; taking square roots we get the desired result, z + w \ < z +  w . • Why is this result called the triangle inequality? T h e reason will b e clear once we view the complex numbers geometrically. Represent the complex n u m b e r z = a + bi as the point having coordinates (a, b) in the xy plane.
T h e distance r of this point from the origin is V a + b , in other words, z. T h e angle 6 is called the argument of z and, as we see, tan 6 = b/a. Also, a = r cos 9, b = r sin 0; therefore, z = a + bi = r ( c o s 0 + i sin 0). This representation of z is called its polar form. Given z = a + bi, w = c + di, then their sum is z + w = (a + c) + (b + d)i. Geometrically, we h a v e the picture:
2
2
(a+c , b + d)
36
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
T h e statement \z + w\ ^ \z\ + \ w\ merely reflects the fact that in a triangle one side is of smaller length than the sum of the lengths of the other two sides; thus the t e r m triangle inequality. The complex n u m b e r s that come u p in the polar form cos 9 + i sin 9 are very interesting n u m b e r s i n d e e d . Specifically, cos 9 + i sin Q\ =
Vcos
2
9 + sin 9
2
= VI =
1,
so they give us many complex n u m b e r s of absolute value 1. In truth they give us all the complex n u m b e r s of absolute value 1; to see this just go back and look at t h e polar form of such a number. Let's recall two basic identities from trigonometry, cos(# + i/>) = cos 9 cos \jj — sin 8 sin i/> and sin(0 + i/>) = sin 9 cos i/> + cos 9 sin if/. T h e r e fore, if z = /'(cos 9 + i sin 9) and w = s(cos < + i sin i/>), t h e n / > zw = /(cos 9 + i sin 9) • s(cos <p + i sin i/>) = ra(cos 9 cos i/> — sin 0 sin i/>) + z rs(sin 0 cos i/> + cos 0 sin i/y) = « [ c o s ( 0 +(/>) + / s i n ( 0 + i//)]. Thus, in multiplying two complex numbers, the argument of the product is the sum of the arguments of the factors. This has another very interesting consequence. T h e o r e m 1.7.6 ( D e Moivre's T h e o r e m ) . [r(cos 9 + i sin 9)]" = r"[cos(nd) + i sin(;i0)]. F o r any integer n > 1,
Proof. W e proceed by induction on n. If n = 1, the statement is obviously true. Assume then that for some k, [r(cos 9 + isin9)] = r [cos k9 + isink9]. Thus
k k
[/(cos 8 + i sin 9)]
k + 1
= [r(cos 0 + i sin 0)] • r(cos 6 + i sin 0) = r (cos kd + i sin kO) • /(cos 0 + / sin 9) = r [cos(k
k+1 fc
/c
+ 1)9 + isin(k
+ 1)9]
by the result of the p a r a g r a p h above. This completes the induction step; hence the result is true for all integers n s 1. • In the problems we shall see that D e Moivre's T h e o r e m is true for all integers m; in fact, it is true even if m is rational. Consider the following special case: 9„ = cos — + i sin — , where n > 1 is an integer.
Sec. 7
Complex Numbers
37
By D e Moivre's T h e o r e m ,
= cos + / sin 1. So 6", = 1; you can verify that 6'" + 1 if 0 < m < n. This property of 6„ makes it one of the primitive nth roots of unity, which will b e encountered in P r o b lem 26.
27 7
27 = 7
PROBLEMS
Easier P r o b l e m s 1. Multiply. (a) (6  7i)(8 + i). (b) (  +  0 (   § 0 (c) (6 + 7i)(8  02. Express z in the form z (a) z = 6 + 8/. (b) 2 = 6  8 / .
1  1  1
= a + bi for:
( C ) Z
=
V2
+
V=2 _
1
L
*3. Show that (?)
= (z" ).
_ 1
1
4. Find (cos 6 + i sin 0 ) . 5. Verify parts a, b, e, fof L e m m a 1.7.1. *6. Show that z is real if and only if z = z, and is purely imaginary if and only if z =  z. 7. Verify the commutative law of multiplication zw = wz in C . 8. Show that for z * 0,  z  = 9. Find: (a) 6  4i.
_ 1
V\z\
(b)
(c)
I+fi—j=
H
7=1 .
V2 V2
10. Show that z = z.
38
T h i n g s F a m i l i a r a n d Less F a m i l i a r
Ch. 1
11. Find the polar form for
..
i\
V2
1 .
(b) z = Ai. 6 , 6 . (c) Z = —7= H 7= I. ,a\ 13 ^ 39 .
2
12. Prove that ( c o s (  0 ) + z sin(f 0)) = cos 0 + i sin 0. 13. By direct multiplication show that (  +  V 3 z') = — 1. MiddleLev el P r o b l e m s 14. Show that (cos 0 + i sin 9)"' = cos(m8) + i sin(m0) for all integers m.
3
15. Show that (cos 9 + z'sin 6)'' = cos(r0) + isin(/'0) for all rational n u m bers r. 16. If z £ C and « > 1 is any positive integer, show that there are n distinct complex numbers w such that z = w". 17. Find the necessary and sufficient condition on k such that:
(
cos
{^r)
2Trk\
\ n
+1
cos
{^rj) , . . / 7ic 2 rf \ + z sm
\ n
sin
= 1
and
¥=1
if 0 < m < n.
18. Viewing the xy plane as the set of all complex n u m b e r s x + iy, show that multiplication by i induces a 90° rotation of the xy plane in a counterclockwise direction. 19. In P r o b l e m 18, interpret geometrically what multiplication by the complex n u m b e r a + bi does to the xy plane. *20. Prove that \z + w\
2
+ \z  w\ = 2(\z\
2
2
+
\w\ ).
2
21. Consider the set A = [a + bi  a, b £ Z}. Prove that there is a 11 correspondence of A onto N. (A is called the set of Gaussian integers.) 22. If a is a (complex) root of the polynomial x" + axx"" ^
t 1
• • • + a^x
n 1
+ a„,
where the a are real, show that a must also b e a root, [r is a root of a polynomial p(x) if p(r) = 0.]
Sec. 7
Complex Numbers
39
Harder P r o b l e m s 23. Find the necessary and sufficient conditions on z and w in order that z + w\ = z + \w\. 24. Find the necessary and sufficient conditions on z i , • • •, z zi + • • • + z \ =  z i  + • • • + \z \.
k k k k
in order that
*25. T h e complex n u m b e r 9 is said to have order n > 1 if 0" = 1 and 0"' =h 1 for 0 < m < n. Show that if 9 has order « and 9 = 1, where /c > 0, t h e n « [ A:. *26. Find all complex n u m b e r s 0 having order «. (These are the primitive roots of unity.) nth
1. DEFINITIONS A N D E X A M P L E S OF GROUPS
We have seen in Section 4 of C h a p t e r 1 that given any n o n e m p t y set, the set A(S) of all 11 mappings of S onto itself is not just a set alone, but has a far richer texture. T h e possibility of c o m b i n i n g two e l e m e n t s of A(S) to get yet a n o t h e r e l e m e n t of A(S) endows A(S) with an algebraic structure. W e recall how this was done: If / , g G A (S), then we c o m b i n e t h e m to form the mapping fg defined by (fg)(s) = f(g(s)) for every s G S. W e called fg the product of / a n d g, and verified t h a t / g G A(S), and that this p r o d u c t obeyed certain rules. F r o m the myriad of possibilities we s o m e h o w selected four particular rules that govern t h e behavior of A (S) relative to this product. These four rules were 1. Closure, namely if / g G A(S), then fg G A(S). W e say that A(S) = (fg)h. all/G A(S). is
closed u n d e r this product. 2. Associativity, 3. Existence i G A(S) 4. Existence that is, given / g, h G A(S), then f(gh)
of a unit element, of inverses, xnA{S)
namely, there exists a particular element there exists an element, = i.
(the identity mapping) such that/*' = if = f'for that is, given / G A(S) such that ff~
x
denoted by f~\
= f~ f
x
T o justify or motivate why these four specific attributes of A (S) w e r e singled out, in contradistinction to some other set of properties, is not easy to
Sec. 1
D e f i n i t i o n s a n d E x a m p l e s of G r o u p s
41
do. In fact, in the history of the subject it took quite some time to recognize that these four properties played the key role. W e have the advantage of historical hindsight, and with this hindsight we choose t h e m not only to study A(S), but also as the chief guidelines for abstracting to a much wider context. Although we saw that the four properties above enabled us to calculate concretely in A(S), there were some differences with the kind of calculations we are used to. If S has three or m o r e elements, we saw in Problem 15, Chapter 1, Section 4 that it is possible f o r / , g G A(S) to have fg # gf. H o w ever, this did not present us with insurmountable difficulties. Without any further polemics we go to the Definition. A nonempty set G is said to b e a group if in G there is defined an operation * such that: (a) a, b G G implies that a * b E G. (We describe this by saying that G is closed u n d e r •*.) (b) Given a, b, c G G, then a * (b * c) = (a * b) * c. (This is described by saying that the associative law holds in G.) (c) T h e r e exists a special element e G G such that a * e = e * a — a for all a G G (e is called the identity or unit element of G ) . (d) F o r every a G G there exists an element b G G such that a * b = b * a = e. (We write this element b as a~ and call it the inverse of a in G.)
l
These four defining postulates (called the group axioms) for a group were, after all, p a t t e r n e d after those that hold in A (S). So it is not surprising that A (S) is a group relative to the operation "composition of mappings." T h e operation * in G is usually called t h e product, but k e e p in m i n d that this has nothing to do with product as we k n o w it for the integers, ration a l , reals, or complexes. In fact, as we shall see below, in many familiar examples of groups that come from n u m b e r s , what we call the product in these groups is actually the addition of numbers. However, a general group need have no relation whatsoever to a set of numbers. W e reiterate: A group is n o more, n o less, than a n o n e m p t y set with an operation * satisfying the four group axioms. Before starting to look into the n a t u r e of groups, we look at some examples. Examples of Groups 1. Let Z b e the set of all integers and let * b e the ordinary addition, + , in Z. That Z is closed and associative u n d e r * are basic properties of the integers. What serves as the unit element, e, of Z u n d e r *? Clearly, since a = a * e =
42
Groups
Ch. 2
a + e, we have e = 0, and 0 is the required identity e l e m e n t u n d e r addition. W h a t about a " ? H e r e too, since e = 0 = a * a = a + a , the a " in this instance is  a , and clearly a * (  a ) = a + (—a) = 0.1  1  1 1
2. Let < be the set of all rational n u m b e r s and let the o p e r a t i o n * o n Q b e Q the ordinary addition of rational n u m b e r s . A s above, < > is easily shown to b e G a group under *. N o t e that I C Q and both Z and Q are groups u n d e r t h e same operation *. 3. Let Q' be the set of all nonzero rational n u m b e r s and let the operation * on Q' be the ordinary multiplication of rational n u m b e r s . By the familiar properties of t h e rational n u m b e r s we see that <Q' forms a group relative to *. 4. Let R be the set of all positive real n u m b e r s and let the operation * on R b e the ordinary product of real n u m b e r s . Again it is easy to check that R is a group under *.
+ + +
5. Let E„ be the set of d'„, i = 0 , 1 , 2 , . . . , « — 1, where 0„ is the complex number 0„ = cos(277/«) + /' sin(27r/n). Let 0* * B' = 6* , the ordinary product of the powers of 0„ as complex numbers. By D e Moivre's T h e o r e m we saw that 0" = 1. W e leave it to the reader to verify that E„ is a group under *. The elements of E„ are called the n'th roots of unity. The picture below illustrates the group E , whose elements are represented by the dots on the unit circle in the complex plane.
+i n h
N o t e one striking difference b e t w e e n the E x a m p l e s 1 t o 4 and E x a m p l e 5; t h e first four have an infinite n u m b e r of elements, whereas E„ has a finite number, n, of elements. Definition. A group G is said to be a finite group if it has a finite n u m b e r of elements. The n u m b e r of elements in G is called the order of G and is denoted by \ G\.
Sec. 1
Definitions a n d Examples of Groups
43
Thus E above is a finite group, and \E \ = n. All the examples presented above satisfy t h e additional property that a * b = b * a for any pair of elements. This n e e d not b e true in a group. Just witness t h e case of A (S), w h e r e S h a s three o r m o r e elements; there w e saw that we could find fgE.A (S) such that fg ¥= gf This p r o m p t s us t o single o u t as special those groups of G in which a * b = b * a for all a, b G G.
n n
Definition. a, b £ G.
A g r o u p G is said t o b e abelian
if a * b = b * a for all
The word abelian derives from the name of the great Norwegian mathematician Niels Henrik Abel (18021829), one of the greatest scientists Norway has ever produced. A group that is n o t abelian is called nonabelian, a not t o o surprising choice of n a m e . We now give examples of some nonabelian groups. Of course, t h e A(S) afford us an infinite family of such. But w e present a few other examples in which we can compute quite readily. 6 Let 1R b e t h e set of all real numbers, and let G be t h e set of all mappings . T„ : U —> U defined by T (r) = ar + b for any real n u m b e r r, where a, b are real n u m b e r s and a 0 Thus, for instance, T _ is such that T _ ( / ) = . 5r 5 r ._ (7r) = 5tt ' 6. T h e T are 5 6 pings of U onto itself, and we let T * T b e t h e product of two of these mappings. So
b ab 
6; r _(14) = • 1  6 = 64, 4
ab
5
6
5
6
5
6
a>b
11 m a p 
cd
(T ,„*
a
r , )(r) =
c rf
r , (r (r))
a 6 Cirf
= «rc,rf(r) + b = a(cr + d) + b T,^ (r).
ad+b
= (nc)r + (ad + b) = So we have t h e formula T ,b* T
a B cd
= T
acad
+ b
.
() 1
This result shows us that T * T is in G—for it satisfies t h e membership requirement for belonging to G—so G is closed u n d e r *. Since w e are talking about t h e product of mappings (i.e., t h e composition of mappings), * is associative. T h e element T = i is t h e identity mapping of R onto itself. Finally, what is T ? Can we find real numbers x # 0 a n d y, such that
cd i0  1 f c
44
Groups
Ch. 2
T ,b * T
a
xy
— T
x<y
*T
llb
—T ?
l 0 iQ
G o back to (1) above; we thus want T = T , that is, ax = 1, ay + ft = 0. R e m e m b e r now that a + 0, so if we put x = a and y =  a f t , the required relations are satisfied. O n e verifies immediately that
a x a y + b  1  1
a,h* a a b So G is indeed a group.
l
l
,
~ a^a^b * „,b ~ M , 0 •
1 1
W h a t is T * T l According to the formula given in (1), w h e r e we replace a by c, c by a, b by d, d by ft, we get
c d a b
T ,a* T„, ~ T
c b cd ih f l i cd
c i i c h + d
.
(2)
T h u s T * T„ = if r * T and only if ftc + d = ad + ft. This fails to b e true, for instance, if a — 1, ft = 1, c = 2, d = 3. So G is nonabelian. 7. L e t H C G, where G is t h e group in Example 6, and H is defined b y H [T ,b G  a is rational, ft any real}. W e leave it t o the r e a d e r to verify that H is a group u n d e r the operation * defined on G. H is nonabelian.
= G a
8. Let K C H C G, where 77, G are as above and K = [T G G  ft any real}. T h e reader should check that K is a group relative t o the operation * of G, and that K is, however, abelian.
x b
9. Let S b e the p l a n e , that is, S = {(x, y)  x, y real} and consider f,gEA (S) defined b y f(x, y) = (x, y) and g(x, y) = (—y, x); / is the reflection a b o u t the yaxis and g is the rotation through 90° in a counterclockwise direction about the origin. W e then define G = [fg' \ i = 0, 1 ; = 0, 1, 2, 3}, and let * in G be the product of elements in A(S). Clearly, f = g = identity mapping;
2 4
(f*g)(x,y) and (g*f)(x,y)
= (fg)(x,y)=f(g(x,y))=f(y,x)
= (y,x)
= g(f(x,y))
= g(x,y)
= (  y , x).
 1
So g * / = £ / * g. W e leave it t o the r e a d e r to verify that g * / = / * g and G is a nonabelian group of order 8. This group is called the dihedral group of order 8. [Try t o find a formula for (fg') * ( / ' g ' ) = fg that expresses a, ft in terms of i, j , s, and f\]
b
10. Let S b e as in E x a m p l e 9 a n d / t h e mapping in E x a m p l e 9. Let /i > 2 and let A be the rotation of the plane about the origin t h r o u g h a n angle of Ittju in t h e counterclockwise direction. W e then define G = [f h'\k = 0, 1; j = 0, 1, 2 , — 1} and define the product * in G via t h e usual product of mappings. O n e can verify that f = h" = identity mapping, and fh = ft /.
k 2 1
Sec. 1
Definitions a n d E x a m p l e s of Groups
45
T h e s e relations allow us t o show (with some effort) that G is a nonabelian g r o u p of o r d e r 2n. G is called the dihedral group of o r d e r 2n. 11. Let G = {/£ A(S) \f{s) # s for only a finite n u m b e r of s £ S}, where we suppose that S is an infinite set. W e claim that G is a group under the product * in A(S). T h e associativity holds automatically in G, since it already holds in A(S). Also, i £ G, since i(s) = s for all s £ S. So we must show that G is closed u n d e r the product and if / £ G, t h e n / " £ G. W e first dispose of the closure. Suppose that / g £ G; then / ( s ) = s except, say, for s , s , . . . , s a n d g (s) = s e x c e p t for s[, s , • • • , s,' . T h e n (fg)( ) = f(g( )) = u s , . . . , s,„ s{,. . .,s,'„ (and possibly even for some of these). So fg moves only a finite n u m b e r of elements of S, so fg £ G. Finally, if f(s) = s for all s other t h a n s , s , . . . , s,„ then f~ (f(s)) — f'^s), but f^s) = f\f{s)) = (r\f)(s) = i(s) = s. So w e obtain that f^ (s) = s for all 5 except s . .., s . Thus / £ G and G satisfies all the group axioms, hence G is a group.
1 x 2 n 2 n s s s f o r a 1 1 5 o t h e r t h a n s 2 1 x 2 1 _ 1 u n
12. Let G be the set of all mappings T , where T is the rotation of a given circle about its center through an angle 0 in the clockwise direction. I n G define * by the composition of mappings. Since, as is readily verified, T * = T , G is closed under *. The other group axioms check o u t easily. Note that T „ = T = the identity mapping, and Tg = T_ = T _ . G is an abelian group.
B e # e+ljl 1 2 Q e 2lT g
A s we did for A (S) we introduce the shorthand notation a" for a *a *a•••*a n times and define a~" = (a" )", for n a positive integer, and a" = e. T h e usual rules of exponents then hold, that is, (a'")" = a'"" and a'" * a" = a"' " for any integers m and n. N o t e that with this notation, if G is the group of integers u n d e r + , t h e n a" is really na. Having seen the 12 examples of groups above, the reader might get t h e impression that all, or almost all, sets with some operation * form groups. This is far from true. W e n o w give some examples of nongroups. I n each case we check the four group axioms a n d see which of these fail to hold.
+ 1
Nonexamples 1. Let G b e the set of all integers, a n d let * b e t h e ordinary product of integers in G. Since a * b = ab, for a, b £ G we clearly have that G is closed a n d associative relative t o *. F u r t h e r m o r e , the n u m b e r 1 serves as t h e unit ele
46
Groups
Ch. 2
ment, since a * 1 = al = a = la = 1 * a for every a G G. So we are threefourths of the way to proving that G is a group. All we n e e d is inverses for the elements of G, relative to *, to lie in G. But this just isn't so. Clearly, we cannot find an integer b such that 0 * 5 = 0/3 = 1, since Ob = 0 for all b. B u t even other integers fail to have inverses in G. For instance, we cannot find an integer b such that 3 * b = 1 (for this would require that b = §, and § is not an integer). 2. Let G b e the set of all n o n z e r o real n u m b e r s and define, for a, b G G, a * b = a b; thus 4 * 5 = 4 ( 5 ) = 80. Which of the group axioms hold in G u n d e r this operation * and which fail to hold? Certainly, G is closed u n d e r *. Is * associative? If so, (a * b) * c = a * (b * c), that is, (a * b) c = a (b * c), and so (a b) c = a (b c), which boils down to a = 1, which holds only for a = ± 1 . So, in general, the associative law does not hold in G relative t o *. W e similarly can verify that G does not have a unit element. T h u s even to discuss inverses relative to * would not m a k e sense.
2 2 2 2 2 2 2 2 2
3. Let G b e the set of all positive integers, under * w h e r e a * b = ab, the ordinary product of integers. T h e n one can easily verify that G fails to be a group only because it fails to have inverses for some (in fact, most) of its elements relative to *. W e shall find some other nonexamples of groups in the exercises.
P R O B L E M S
Easier Problems 1. D e t e r m i n e if the following sets G with the operation indicated form a group. If not, point out which of the group axioms fail. (a) G = set of all integers, a * b = a  b. (b) G = set of all integers, a*b = a + b + ab. (c) G = set of nonnegative integers, a * b = a + b. (d) G = set of all rational n u m b e r s — 1, a * b = a + b + ab. (e) G = set of all rational n u m b e r s with denominator divisible by 5 (written so that n u m e r a t o r and denominator are relatively p r i m e ) , a * b = a + b. (f) G a set having m o r e t h a n one element, a * b = a for all a, b G G. 2. In the group G defined in Example 6, show that the set H = {T b any real} forms a group under the * of G. 3. Verify that E x a m p l e 7 is indeed an example of a group.
a h
j a = ±1,
Sec. 1
Definitions a n d E x a m p l e s of G r o u p s
47
4. Prove that K defined in E x a m p l e 8 is an abelian group. 5. I n E x a m p l e 9, prove that g * f = f * g " , and that G is a group, is nonabelian, and is of order 8. 6. L e t G a n d H b e as in Examples 6 a n d 7, respectively. Show that if T E G, then T„_ *V* T~\ G H if V G 77.
llh h 1
7. D o Problem 6 with 7/ replaced by the group K of E x a m p l e 8. 8. If G is an abelian group, prove that {a *fa)"= a" * b" for all integers n. 9. If G is a group in which a = e for all a E G, show that G is abelian. 10. If G is the group in E x a m p l e 6, find all T„ E G such that T Ti * T , for A// real x.
b iX aJ ab 2
*T
Ux
=
11. In E x a m p l e 10, for n = 3 find a formula that expresses {fh') / " / 7 . Show that G is a nonabelian group of order 6.
6
* {fh') as
12. D o Problem 11 for n = 4. 13. Show that any group of order 4 or less is abelian. 14. If G is any group and a, b, c G G, show that if A *fa= a * c , then fa = c, and iffa* a = c * a, then fa = c. 15. Express (a *fa)" in terms of A " and fa . 16. Using the result of P r o b l e m 15, prove that a group G in which a = A " for every a G G must b e abelian. 17. In any group G, prove that ( A " )
l 1  1 1 1 1 1
= a for all A G G.
* 18. If G is a finite group of even order, show that there must b e a n element a + e such that a = a~ . (Hint: Try to use the result of Problem 17.) 19. In S , show that there are four elements x satisfying x = e and three elements y satisfying y = e.
3 3 2
20. Find all the elements in S such that x = e.
4
4
MiddleLevel Problems 21. Show that a group of order 5 must b e abelian. 22. Show that the set defined in E x a m p l e 10 is a group, is nonabelian, and has order 2n. D o this b y finding the formula for {fh') * {fh') in t h e form fh .
b
23. I n t h e group G of Example 6, find all elements U G G such that U * T ,b = T ,b * U for every T
a a a<b
E G.
24. If G is the dihedral group of order 2n as defined in Example 10, prove that: ( a ) If 77 is odd and A G G is such that a *fa=fa* a for allfaG G, then a = e. ( b ) If 77 is even, show that t h e r e is an a G G, a + e, such that a * b =fa* a for allfaG G.
48
Groups
Ch. 2
(c) If 77 is e v e n , find all t h e e l e m e n t s a £ G such t h a t a * b = b * a for all b G G. 25. If G is any group, show that: (a) e is unique (i.e., i f / G G also acts as a unit element for G, then / = e). (b) Given a £ G, then a G G is unique.
1
*26. If G is a finite group, prove that, given a G G, there is a positive integer «, depending on a, such that a" = e. *27. In P r o b l e m 26, show t h a t t h e r e is a n i n t e g e r m > 0 such t h a t a'" = e for all a G G. H a r d e r Problems 28. Let G be a set with an operation * such that: 1. G is closed u n d e r *. 2. * is associative. 3. T h e r e exists an element e G G such that e * x = x for all x G G. 4. Given x G G, there exists a y G G such that y * x = e. Prove that G is a group. (Thus you must show that x * e = x a n d x * y = e for e, y as above.) 29. Let G be a /imYe n o n e m p t y set with an operation * such that: 1. G is closed u n d e r *. 2. * is associative. 3. Given a, b,c G G with a * b = a * c, then b = c. 4. Given a, i>, c, G G with b * a = c * a, then b = c. Prove that G must b e a group u n d e r *. 30. Give an example to show that the result of P r o b l e m 29 can b e false if G is an infinite set. 31. Let G be the group of all nonzero real numbers u n d e r the operation * which is the ordinary multiplication of real n u m b e r s , and let H b e the group of all real n u m b e r s under the operation #, which is the addition of real n u m b e r s . (a) Show that there is a mapping F: G * H of G o n t o H which satisfies F(a * b) = F(a)#F(b) for all a, b G G [i.e., F ( a 6 ) = F{a) + F(b)]. (b) Show that n o such mapping /"can b e 11.
2. SOME SIMPLE REMARKS
In this short section we show that certain formal properties which follow from the group axioms hold in any group. A s a matter of fact, most of these results have already occurred as problems at the end of the preceding section.
Sec. 2
S o m e Simple Remarks
49
It is a little clumsy to k e e p writing t h e * for the product in G, and from now on we shall write the product a * b simply as ab for all a, b G G. T h e first such formal results we prove are contained in L e m m a 2.2.1. If G is a group, then: unique.
1
(a) Its identity element is
(b) Every a G G has a unique inverse A " G G. (c) If a G G, ( a ) "
 1 1
= a.
1
(d) F o r a,b£G,
(A/3)" =
b'^'K
Proof. W e start with Part (a). W h a t is expected of us to carry out t h e proof? W e must show that if e, / G G and af = fa = a for all a G G a n d ae = ea = a for all a G G, then e = f. This is very easy, for then e = ef a n d /" = e/; hence e = ef = f as required. Instead of proving P a r t (b), we shall p r o v e a stronger result (listed below as L e m m a 2.2.2), which will have Part (b) as a n immediate consequence. W e claim that in a group G if ab = ac, then b = c; that is, we can cancel a given element from the same side of an equation. T o see this, we have, for a G G, an element u G G such that ua = e. T h u s from ab = ac w e have u(ab) = u{ac),
so, by t h e associative law, (ua)b = (ua)c, that is, eb = ec. H e n c e b = eb = ec = c, and o u r result is established. A similar argument shows that if ba = ca, then b = c. However, we cannot conclude from ab = ca that b = c; in any abelian group, yes, but in general, n o . N o w to get Part (b) as a n implication of t h e cancellation result. Suppose that b,c G G act as inverses for a; then ab = e = ac, so by cancellation b = c and we see that the inverse of a is unique. W e shall always write it as a . T o see Part (c), n o t e that by definition A ^ A ) " = e; but A " a = e, so by cancellation in ^ " ' ( a " ) " = e = a~ a we get that ( a ) " = a.
 1  1 1 1 1 1 l  1 1
Finally, for Part (d) we calculate
( A / > ) ( / 3 A ) = ((A/3)/3" )A" = (A(/3/3~ )A"
= (AC)A
1
_1
_1
1
1
(associative law) (again t h e associative law)
1
1
1
= aa~
= e.
1
Similarly, (b~ a~ )(ab)
1
1
= e. H e n c e , by definition, (ab)'
= b~ aT .
l
x
•
50
Groups
Ch. 2
W e promised to list a piece of the argument given above as a separate lemma. W e k e e p this promise and write L e m m a 2.2.2. In any group G and a, b, c G G, we h a v e :
(a) If ab = ac, then b = c. (b) If ba = ca, then b =c. Before leaving these results, note that if G is the g r o u p of real n u m b e r s u n d e r + , then P a r t (c) of L e m m a 2.2.1 translates into the familiar — (—a) = a. There is only a scant bit of mathematics in this section; accordingly, we give only a few problems. N o indication is given as to the difficulty of these.
PROBLEMS
1. Suppose that G is a set closed u n d e r an associative operation such that 1. given a,y G G, t h e r e is an x G G such that ax = y, and 2. given a, w G G, t h e r e is a u G G such that ua = w. Show that G is a group. *2. If G is a finite set closed under an associative operation such that ax = ay forces ,Y = y and ua = wa forces u = w, for every a, x, y, it, w G G, prove that G is a group. (This is a repeat of a problem given earlier. It will be used in the body of the text later.) 3. If G is a group in which (ab) = a'b' for three consecutive integers i, prove that G is abelian. 4. Show that the result of P r o b l e m 3 would not always be true if the w o r d " t h r e e " were replaced by " t w o . " In other words, show that t h e r e is a group G and consecutive n u m b e r s i, i + 1 such that G is not abelian but does have the p r o p e r t y that (ab)' = a'b' and (a/3)' = a' b' for all a, b in G.
+ 1 + l + l 1
5. Let G be a group in which (ab) Show that G is abelian.
3
= ab
3
3
and (ab)
5
= a 6 for all a, b G G.
3 5
6. Let G be a group in which (ab)" = a"b" for s o m e fixed integer n > 1 for all a, b G G. For all a, b G G, prove that: (a) (ab)"= b" a" . ( b ) a"b" = b"~ a".
1 [ ] 1 }
(c) (afor ** )"*" ) = e.
[Hint for P a r t (c): N o t e that (aba' ) '
1 1
1
1
1
= ab a"
r
x
for all integers;.]
Sec. 3
Subgroups
51
3. SUBGROUPS
I n order for us to find out m o r e about t h e m a k e u p of a given group G, it m a y be too much of a task to tackle all of G headon. It might be desirable to focus our attention on appropriate pieces of G, which are smaller, over which we have some control, and are such that the information gathered about them can be used to get relevant information and insight about G itself. T h e question t h e n becomes: W h a t should serve as suitable pieces for this kind of dissection of G ? Clearly, whatever we choose as such pieces, we want t h e m to reflect the fact that G is a group, not merely any old set. A group is distinguished from an ordinary set by the fact that it is endowed with a wellbehaved operation. It is thus natural to d e m a n d that such pieces above behave reasonably with respect to the operation of G. O n c e this is granted, we are led almost immediately to the concept of a subgroup of a group. Definition. A n o n e m p t y subset, / / . of a group G is called a of G if, relative to the product in G, / / i t s e l f forms a group. subgroup
W e stress the phrase "relative to t h e product in G." Take, for instance, the subset A = {1,  1 } in Z, the set of integers. U n d e r the multiplication of integers, A is a group. B u t A is not a subgroup of Z viewed as a group with respect to + . Every group G automatically has two obvious subgroups, namely G itself and the subgroup consisting of the identity element, e, alone. These two subgroups we call trivial subgroups. O u r interest will b e in the remaining ones, the proper subgroups of G. Before proceeding to a closer look at the general character of subgroups, we want to look at some specific subgroups of some particular, explicit groups. Some of the groups we consider are those we introduced as examples in Section 1; we maintain the numbering given there for them. In some of these examples we shall verify that certain specified subsets are indeed subgroups. W e would strongly r e c o m m e n d that the reader carry out such a verification in lots of the others and try to find other examples for themselves. In trying to verify w h e t h e r or not a given subset of a group is a subgroup, we are spared checking one of the axioms defining a group, namely the associative law. Since the associative law holds universally in a group G, given any subset A of G and any t h r e e elements of A, then the associative law certainly holds for them. So we must check, for a given subset A of G, whether A is closed under the operation of G, whether e is in A, and finally, given a EL A, whether a " is also in A .
1
52
Groups
Ch. 2
Note that we can save one m o r e calculation. Suppose that A C G is nonempty and that given a, b G A, then ab G A. Suppose further that given a G A, then a G A Then we assert that e G A. For pick a G A; then a G A by supposition, hence a a G A, again by supposition. Since a a = e, we have that e E. A. T h u s a is a subgroup of G. In other words,
 1  1 _ 1  1
L e m m a 2.3.1. A nonempty subset A C G is a subgroup of G if and only if A is closed with respect to the operation of G and, given a G A, then a G A.
 1
W e now consider some examples. Examples 1. Let G be the group Z of integers u n d e r + and let 77 be the set of even integers. We claim that 77 is a subgroup of Z. Why? Is 77 closed, that is, given a, b G H, is a + b G 77? In other words, if a, b are even integers, is a + b an even integer? T h e answer is yes, so 77 is certainly closed under + . N o w to the inverse. Since the operation in Z is + , the inverse of a G Z relative to this operation is —a. If a G 77, that is, if a is even, then a is also even, hence —a G 77. In short, 77 is a subgroup of Z u n d e r + . 2. Let G once again b e the group Z of integers u n d e r + . In E x a m p l e 1, 77, the set of even integers, can b e described in another way: namely 77 consists of all multiples of 2. T h e r e is nothing particular in E x a m p l e 1 that m a k e s use of 2 itself. Let m > 1 b e any integer and let 77„, consist of all multiples of m in Z. W e leave it to the reader to verify that H,„ is a subgroup of Z u n d e r + . 3. Let S be any { / G A(S)\f(a) /, g G 77(a), then fg G 77(a). Also, n o n e m p t y set and let G = A(S). If a G S, let H(a) = = a}. W e claim that 77(a) is a subgroup of G. For if ( / g ) ( a ) = / ( g ( a ) ) = / ( a ) = a, since / ( a ) = g(a) = a. T h u s i f / G 77(a), t h e n / ( a ) = a, so t h a t / ( / ( a ) ) = / ( a ) . B u t
 1  1 T h u s s i n c e a 1 w e n a v e
r\f(a))
= = »(«) = «• > = J" (/(«)) = that / G 77(a). Moreover, 77 is nonempty. (Why?) Consequently, 77(a) is a subgroup of G.
 1
4. Let G be as in E x a m p l e 6 of Section 1, and 77 as in E x a m p l e 7. T h e n 77 is a subgroup of G (see P r o b l e m 3 in Section 1). 5. Let G be as in E x a m p l e 6, 77 as in E x a m p l e 7, and K as in E x a m p l e 8 in Section 1. T h e n K C 77 C G and 7C is a subgroup of b o t h 77 and of G. 6. Let C b e the n o n z e r o complex n u m b e r s as a group u n d e r the multiplication of complex n u m b e r s . Let V = {a G C  \a \ is rational}. T h e n F i s a subgroup of C . F o r if  a  and \ b\ are rational, then \ ab\ =  a  \b\ is rational, so
Sec. 3
Subgroups
1
53
ab G V; also, \ a group of C .
 = 1/ \a  is rational, hence a
1
G V. Therefore, V is a sub
7. Let C and K b e as above and let U = (A G C  a = cos 0 + / sin 0, 0 any real). If fl = cos 6 + 7 sin 6 and = cos i/> + 2 sin I/A we saw in Chapter 1 that ab = c o s ( 0 + \}t) + i sin(0 + (/)), so that A6 G £/, and that A " = cos 0  / sin 0 c o s (  0 ) + i s i n (  e ) G U. Also, \a\ = 1, since \a\ = V c o s 0 + s i n 0 = 1. Therefore, U C V C C and (/ is a subgroup both of V and of C .
1 2 2
8. Let C , £/, V be as above, and let n > 1 be an integer. Let 0„ = COS(2T77H) + i sin(277//7), and let E„ = {1, 0,„ 0?„ .. . , dT'}. Since e;; = 1 (as we saw by D e Moivre's T h e o r e m ) , it is easily checked that E„ is a subgroup of U, V, and C , and is of order n. 9. Let G be any group and let a G G. T h e set A = {fl'  / any integer) is a subgroup of G. For, by the rules of exponents, if a' G A and a' G A , then a'a' = A ' , SO is in A. Also, ( A ' ) " = A " ' , so ( f l ' ) G A. This makes A into a subgroup of G. A is the cyclic subgroup of G generated by a in the following sense.
+ ; 1  1
Definition. T h e cyclic subgroup integer). It is denoted (A).
of G generated by A is a set {a' j / any
N o t e that if e is the identity element of G, then (e) = je). In Example 8, the group B is the cyclic group (0„) of C generated by 0„. 10. Let G be any group; for a G G let C ( A ) = [g G G  ga = ag}. We claim that C ( A ) is a subgroup of G. First, the closure of C ( A ) . If g, h G C(a), t h e n g A = A g and ha = a/i, thus (gh)a = g(ha) = g(ah) = (ga)h = (ag)h = a(gh) (by the r e p e a t e d use of the associative law), hence gh G C(a). Also, if g G C ( a ) , t h e n from ga = ag we h a v e g~ (ga)g~ = g (ag)g~\ which simplifies to ag~ = g~ a\ whence g~ G C(a). So, C(a) is thereby a subgroup of G. These particular subgroups C(a) will come up later for us and they are given a special n a m e . W e call C(a) the centralizer of a in G. If in a group ab = fea, we say that a and b commute. Thus C ( A ) is the set of all elements in G that commute with a.
1 1 1 ! l l
11. Let G b e any group and let Z ( G ) = {z G G  z.r = xz for all x G G ) . W e leave it to the r e a d e r to verify that Z ( G ) is a subgroup of G. It is called the center of G.
54
Groups
1
Ch. 2
12. Let G be any group and 77 a subgroup of G. For a G G, let a' Ha = {a~ ha  h G 77}. W e assert that a~ Ha is a subgroup of G. If x = a / i a and y = a" h a where h , h G 77, then xy = (a~ h a)(a~ h a) = cr (h h )a (associative law), and since 77 is a subgroup of G, /?i/? £ 77. Therefore, a~\h h )a G a~ Ha, which says that xy G a~ Ha. T h u s a~ Ha is closed. Also, if x = a / z a G a~ Ha, then, as is easily verified, x = (a /™)" = a~' h~ a G a~ Ha. Therefore, a~ Ha is a subgroup of G.
[ ]  1 t l l l l 2 x 2 l 2 i 2 2 l l l l 2
 1
x
_ 1
 1
1
l
1
l
l
A n even dozen seems to be about the right n u m b e r of examples, so we go on to other things. L e m m a 2.3.1 points out for us what we need in order that a given subset of a group be a subgroup. In an important special case we can m a k e a considerable saving in checking w h e t h e r a given subset 77 is a subgroup of G. This is the case in which 77 is finite. L e m m a 2.3.2. Suppose that G is a group and 77 a n o n e m p t y finite subset of G closed u n d e r the product in G. Then 77 is a subgroup of G. Proof. By L e m m a 2.3.1 we must show that a G 77 implies a' G 77. If a = e, t h e n a~ = e a n d we a r e d o n e . S u p p o s e t h e n t h a t a e ; consider t h e e l e m e n t s a, a , . . . , a " , w h e r e n =  7 7  , the o r d e r of 77. Here we have written down n + 1 elements, all of them in H since H is closed, and 77 has only n distinct elements. H o w can this b e ? Only if s o m e t w o of the elem e n t s listed are equal; p u t a n o t h e r way, only if a' = a' for s o m e 1 < i < ; < ; ! + 1. B u t then, by the cancellation p r o p e r t y in groups, a'~ ' = e. Since ;  / > 1, a'~' G 77, h e n c e e G 77. H o w e v e r , j  i  1 > 0, so 'a'~ G 77 and aa''^ = a'~ ' = e, w h e n c e a = a' '" G 77. This p r o v e s the lemma. •
1 2 + 1 1 1
1 1
1
A n immediate, but nevertheless important, corollary to L e m m a 2.3.2 is the Corollary. If G is a finite group and 77 a n o n e m p t y subset of G closed under multiplication, then 77 is a subgroup of G.
PROBLEMS
Easier P r o b l e m s 1. If A, B are subgroups of G, show that A n B is a subgroup of G. 2. W h a t is the cyclic subgroup of Z generated by  1 u n d e r + ? 3. Let.S be the symmetric group of degree 3. Find all the subgroups of S .
3 3
4. Verify that Z ( G ) , the center of G, is a subgroup of G. (See E x a m p l e 11.)
Sec. 3
Subgroups
55
5. If C ( A ) is the centralizer of a in G (Example 1 0 ) , prove that Z(G)
=
n
fleG
C(a).
3
6. Show that a E Z ( G ) if and only if C(a) = G. 7 . In S , find C ( A ) for each A G 5 .
3
8. If G is an abelian group and if H = {a G G  a subgroup of G.
2
= e}, show that H is a
9. Give an example of a nonabelian group for which the H in Problem 8 is not a subgroup. 10. If G is an abelian group and n > 1 an integer, let A„ = {a" \ a G G } . Prove that A,, is a subgroup of G. * 11. If G is an abelian group and H = [a G G  a " = e for some 77 (a) > 1 depending on a } , prove that His a subgroup of G.
(a)
W e say that a group G is cyclic if there exists an a G G such that every x G G is a p o w e r of A, t h a t is, x = A for some /'. In o t h e r words, G is cyclic if G = (A) for some a G G, in which case we say that A is a generator for G. *12. Prove that a cyclic group is abelian.
7
1 3 . If G is cyclic, show that every subgroup of G is cyclic. 14. If G has n o p r o p e r subgroups, prove that G is cyclic. 15. If G is a group and H a n o n e m p t y subset of G such that, given A, b G H, then a/3" G / / , prove that His a subgroup of G.
1
MiddleLevel P r o b l e m s * 16. If G has n o p r o p e r subgroups, prove that G is cyclic of order p, w h e r e p is a prime number. (This sharpens t h e result of P r o b l e m 1 4 . ) 1 7 . If G is a group and a, x G G, prove that C(x~ ax) amples 1 0 and 1 2 for the definitions of C(b) and
1
= x~ C{a)x. ofx~ C(a)x.]
1
x
[See ExC
18. If S is a nonempty set a n d X C S, show that T(X) X) is a subgroup of A ( 5 ) if X is finite.
= {fGA(S)\f(X)
19. If A, B are subgroups of an abelian group G, let AB = {ab  a G A, b G B } . Prove that is a subgroup of G. 20. Give an example of a group G and two subgroups A, B of G such that is nor a subgroup of G. 21. If A, B are subgroups of G such that b~ Ab A / 3 is a subgroup of G.
l
C A for all b E B, show that
* 22. If A and B are finite subgroups, of orders m and n, respectively, of the abelian group G, prove that AB is a subgroup of order inn if m and n are relatively prime.
56
Groups
Ch. 2
23. What is the order of AB in Problem 22 if m and n are not relatively prime? 24. If 77 is a subgroup of G, let N = C\ x~ Hx. of G such that y~ 'TVy = TV for every y G G.
xSG 1
Prove that TV is a subgroup
H a r d e r Problems 25. Let 5, X, 7 ( X ) be as in P r o b l e m 18 (but X n o longer finite). Give an example of a set S and an infinite subset X such that T(X) is not a subgroup of A (S). * 26. Let G be a group, 77 a subgroup of G. Let Hx = [hx \ h G 77}. Show that, given a,b E. G, t h e n 77A = 77/3 or 77a D 77/3 = 0. *27. If in Problem 26 77 is a finite subgroup of G, prove that Ha and 77/3 have the same n u m b e r of elements. W h a t is this n u m b e r ? 28. Let M , TV be subgroups of G such that x~ Mx for all x G G. * 29. If M is a subgroup of G such that . Y M X C M for all x G G, prove that actually M Y = M. 30. If M , TV are such that . v " ' M r = M and ,Y TVY = TV for all x G G, a n d if M n TV = (e), prove that inn = « m for any m G M. n G TV. (Hint: Consider the element m~ n~ nm.)
1 l l  1 l
C M and x~ Nx
l
C TV for C MTV
all x G G. Prove that MTV is a subgroup of G and that x~\MN)x
4. L A G R A N G E ' S THEOREM
We are about to derive the first real grouptheoretic result of importance. Although its proof is relatively easy, this t h e o r e m is like the A  B  C ' s for finite groups and has interesting implications in n u m b e r theory. A s a m a t t e r of fact, those of you who solved Problems 26 and 27 of Section 3 have all the necessary ingredients to effect a proof of the result. T h e t h e o r e m simply states that in a finite group the order of a subgroup divides the order of the group. To smooth the argument of this theorem—which is due to L a g r a n g e — and for use many times later, we m a k e a short detour into the realm of set theory. Just as the concept of "function" runs throughout most phases of m a t h ematics, so also does the concept of "relation." A relation is a statement aRb about the e l e m e n t s a, b G S. If S is t h e set of integers, a = b is a r e l a t i o n on S. Similarly, a < b is a relation on S, as is a < b.
Sec.
4
Lagrange's T h e o r e m
57
Definition. A relation — on a set S is called an equivalence for all a, b, c G S, it satisfies: (a) a ~ a (reflexivity). (symmetry). (transitivity).
relation if,
(b) a — b implies that b ~ a
(c) a ~ b, b ~ c implies that a ~ c
Of course, equality, = , is an equivalence relation, so the general notion of equivalence relation is a generalization of that of equality. In a sense, an equivalence relation measures equality with regard to some attribute. This vague r e m a r k may become clearer after we see some examples. Examples 1. L e t S be all the items for sale in a grocery store; we declare a ~ b, for a, b G S, if the price of a equals that of b. Clearly, the defining rules of an equivalence relation hold for this ~ . N o t e that in measuring this "generalized equality" on S we ignore all properties of the elements of S other than their price. So a ~ b if they are equal as far as the attribute of price is concerned. 2. Let S b e the integers and n > 1 a fixed integer. W e define a ~ b for a, b G S if n  (a  b). W e verify that this is an equivalence relation. Since n  0 and 0 = a — a, we have a ~ a. Because n \ (a  b) implies that n\(b — a), we have that a ~ b implies that b ~ a. Finally, if a ~ b and b ~ c, t h e n 771 (A — b) and 771 (b — c); hence n \ ((a — b) + (b — c)), that is, n \ (a — c). Therefore, a ~ c. This relation on the integers is of great importance in n u m b e r theory and is called congruence modulo n; when a ~ b, we write this as a = b m o d 77 [or, sometimes, as a = b(n)], which is r e a d "a congruent to b m o d 7 7 . " We'll b e running into it very often from now on. A s we shall see, this is a special case of a much wider p h e n o m e n o n in groups. 3. We generalize Example 2. Let G b e a group and H a subgroup of G. For A, b G G, define a ~ b if ab~ G H. Since e G H and e = aa~\ we have that a ~ fl. Also, if A/J" G H, then since His a subgroup of G, (ab~ )~ G H. B u t (ab~ y = (b~ y a~ = ba' , so ba~ G H, hence b ~ a. This tells us that a ~ b implies that b ~ a. Finally, if a ~ b and b ~ c, then ab~ G H and / 3 c G H. But (ab~ )(bc~ ) = ac~ , whence a c " G H and therefore a ~ c. W e have shown the transitivity of ~ , thus ~ is an equivalence relation on G. N o t e that if G = Z, the group of integers u n d e r + , and H is the subgroup consisting of all multiples of n, for 77 > 1 a fixed integer, t h e n A/3" G H
l 1 ] l l l x 1 l 1 x } _) l l [ 1 1
58
Groups
Ch. 2
translates into a = b(n). So congruence m o d n is a very special case of the equivalence we have defined in E x a m p l e 3. It is this equivalence relation that we shall use in proving Lagrange's theorem. 4. Let G be any group. For a, b G G we declare that a ~ b if t h e r e exists an x G G such that b = x~ ax. W e claim that this defines an equivalence relation on G. First, a ~ a for a = e~ ae. Second, if a ~ b, then b = x~ ax, hence a = (x^ y b(x' ), so that b ~ a. Finally, if a ~ b, b ~ c, then b = x~ ax, c = y~ by for some x, y G G. Thus c = y~ (x~ ax)y = (xyy a(xy), and so a ~ c. W e have established that this defines an equivalence relation on G.
1 l 1 1 1 i 1 x l l l
This relation, too, plays an important role in group theory and is given the special n a m e conjugacy. W h e n a ~ b we say that "a and b are conjugate in G." N o t e that if G is abelian, then a ~ b if and only \fa = b. W e could go on and on to give n u m e r o u s interesting examples of equivalence relations, but this would sidetrack us from our main goal in this section. There will be no lack of examples in the problems at the end of this section. W e go on with our discussion and m a k e the Definition. If ~ is an equivalence relation on S, then [a], the class of a, is defined by [a] = [b G S \ b ~ a}. Let us see what the class of a is in the two examples, E x a m p l e s 3 and 4, just given. In E x a m p l e 3, a ~ b if ab~ G 77, that is, if ab~ = h, for some h G 77. Thus a ~ b implies that a = hb. O n the other hand, if a = kb w h e r e k G 77, then ab' = = k G 77, so a ~ b if and only if a G Hb = [hb  h G 77}. Therefore, [b] = Hb. T h e set Hb is called a right coset of H m G. W e ran into such in P r o b lem 26 of Section 3. N o t e that b G Hb, since b = eb and e G 77 (also because b G [b] = Hb). Right cosets, and left h a n d e d counterparts of t h e m called left cosets, play important roles in what follows. In Example 4, we defined a ~ b if b = x~ ax for some x G G. Thus [a] = [x~ ax  x G G } . W e shall d e n o t e [a] in this case as cl(a) and call it the conjugacy class of a in G. If G is abelian, then cl(fl) consists of a alone. In fact, if a G Z ( G ) , the center of G, then cl(a) consists merely of a. T h e notion of conjugacy and its properties will crop u p again often, especially in Section 11. W e shall examine the class of an element a in E x a m p l e 2 later in this chapter. The important influence that an equivalence relation has on a set is to b r e a k it u p and partition it into nice disjoint pieces.
l l 1 l l
Sec. 4
Lagrange's T h e o r e m
59
Theorem 2.4.1. If  is an equivalence relation on S, then S = U[«], where this union runs over o n e element from each class, and w h e r e [a] # [fa] implies that [a] Pi [b] = 0 . T h a t is, ~ partitions S into equivalence classes. Proof. Since a £ [a], we have U [ a ] = S. T h e proof of t h e second assertion is also quite easy. W e show that if [a] # [b], then [a] Pi [b] = 0 , or, what is equivalent to this, if [a] Pi [b] 0 , then [a] = [fa]. Suppose, then, that [a] Pi [fa] 0 ; let c G [a] Pi [fa]. By definition of class, c ~ a since c G [a] and c ~ fa since c G [fa]. Therefore, a ~ c by symmetry of ~ , a n d so, since a ~ c a n d c ~ fa, we have a ~ fa. Thus a G [fa]; if x G [ a ] , then x ~ A, fl ~ fa gives us that x ~ fa, h e n c e x G [fa]. T h u s [a] C [fa]. T h e a r g u m e n t is obviously symmetric in a a n d fa, so we have [fa] C [a], whence [a] = [fa], a n d our assertion a b o v e is proved. T h e t h e o r e m is now completely proved. •
n E 5
W e n o w can prove a famous result of Lagrange. Theorem 2.4.2 (Lagrange's Theorem). If G is a finite group and H is a subgroup of G, then t h e order of H divides t h e order of G. Proof. L e t us look back at E x a m p l e 3, w h e r e we established that t h e relation a ~ fa if flfa"' G H is an equivalence relation a n d that [a] = Ha = \ha\hEH).
x k
Let k be the n u m b e r of distinct classes—call t h e m Ha , ... , Ha . By T h e o rem 2.4.1, G = Ha U Ha U • • • U Ha and we know that Ha Pi Ha = 0 if i * j. W e assert that any Ha, has \H\ = order of / / n u m b e r of elements. M a p H > Hai by sending h — fafl,. W e claim that this m a p is 11, for if ha = » /7'fl,,then by cancellation in G we would have h = h'\ thus the m a p is 11. It is definitely o n t o by t h e very definition of Ha . So H a n d Ha h a v e the s a m e n u m b e r ,  H \, of elements. Since G = Ha U • • • U Ha a n d the Ha a r e disjoint and each Ha; h a s 1771 elements, we have that \G\ = k\H\. T h u s  7 /  divides \G\ and L a grange's T h e o r e m is proved. •
x 2 k } t t { t { k t
Although Lagrange sounds like a French name, J. L. Lagrange (17361813) was actually Italian, having been born and brought up in Turin. He spent most of his life, however, in France. Lagrange was a great mathematician who made fundamental contributions to all the areas of mathematics of his day. If G is finite, t h e n u m b e r of right cosets of 77 in G, namely  G  /1771, is called t h e index of 77 in G a n d is written as i (77).
G
60
Groups
Ch. 2
Recall that a group G is said to be cyclic if t h e r e is an element a G G such that every element in G is a p o w e r of a. Theorem 2.43. A group G of prime order is cyclic.
Proof. If 77 is a subgroup of G then, by invoking Lagrange's T h e o r e m , \H\ divides  G  = p, p a prime, so  H\ = 1 or p. So if 77 + (e), then 77 = G. If fl G G, a e, then the powers of a form a subgroup («) of G different from (e). So this subgroup is all of G. This says that any x G G is of the form x = Hence, G is cyclic by the definition of cyclic group. • If G is finite and a G G, we saw earlier in the proof of L e m m a 2.3.2 that »(«) = f some n(a) > 1, depending on a. W e m a k e the
e or
a
Definition. If G is finite, then the order of a, written o ( a ) , is the Zea.vf positive integer m such that a'" = <?. Suppose that a G G has order m. Consider the set A = [e, a, a ,..., a'" }; we claim that A is a subgroup of G (since a'" = e) and that t h e m elements listed in A are distinct. W e leave the verification of these claims to the reader. Thus \ A\ = m. — o(a). Since \A \  \G\, we have T h e o r e m 2.4.4. If G is finite and a G G. then o(a)  \G\.
2  1
If a G G, where G is finite, we have, by T h e o r e m 2.4.4,  G  = k • o(a). Thus
fl
lG
=
a
kM«)
= ( °W)* =
f l
e
* =
e
.
W e have proved the T h e o r e m 2.4.5. a G G. If G is a finite group of order n, t h e n a" = e for all
W h e n we apply this last result to certain special groups arising in n u m ber theory, we shall obtain some classical n u m b e r  t h e o r e t i c results due to F e r m a t and Euler. Let Z be the integers and let n > 1 be a fixed integer. W e go back to E x a m p l e 2 of equivalence relations, where we defined a = b m o d n (a congruent to b m o d n) if n \ (a — b). T h e class of a, [a], consists of all a + nk, where k runs through all the integers. We call it the congruence class of a.
Sec.
4
Lagrange's T h e o r e m
61
By Euclid's Algorithm, given any integer b, b = qn + r, where 0 < r < n, thus [fe] = [/•]. So the n classes [0], [1], . . . , [ « — 1] give us all the congruence classes. W e leave it to the r e a d e r to verify that they are distinct. \ Let Z„ = {[0], [1], . . . , [ «  1]}. W e shall introduce two operations, + and • in Z „ . U n d e r + Z„ will form an abelian group; u n d e r • Z„ will not form a group, but a certain piece of it will b e c o m e a group. H o w to define [a] + [b]l W h a t is m o r e natural than to define [a] + [b] = [a + b]. But t h e r e is a fly in the ointment. Is tjhis operation + in Z„ welldefined? W h a t does that m e a n ? W e can represent [a] by many a's—for instance, if n = 3, [1] = [4] = [—2] = • • • , yet we are using a particular a to define the addition. W h a t we must show is that if [a] = [a'] and [b] = [b'], then [a + b] = [a' + b'], for then we will have [a] + [b] = [a + b] = [a' + b'] = [a'] + [b']. Suppose that [a] = [a']; t h e n n \ (a — a ' ) . Also from [b] = [/>'], n.  (b  b'), hence n \ ((a  a') + (b  b')) = ((a + b)  (a' + b')). T h e r e fore, a + b = (a' + b') m o d n, and so [a + b] = [a' + b']. So we now have a welldefined addition in Z „ . T h e element [0] acts as the identity element and [a] acts as — [fl], the inverse of [a]. W e leave it to the reader to check out that Z„ is a group u n d e r + . It is a cyclic group of order n generated by [1]. W e summarize this all as T h e o r e m 2.4.6. [« + b]. Z„ forms a cyclic group u n d e r the addition [a] + [b] =
Having disposed of the addition in Z„, we turn to the introduction of a multiplication. Again, what is m o r e natural than defining [a][b] = [ab]?
So, for instance, if n = 9, [2][7] = [14] = [5], and [3][6] = [18] = [0]. U n d e r this multiplication—we leave the fact that it is welldefined to t h e r e a d e r — Z„ does not form a group. Since [0][a] = [0] for all a, and the unit element under multiplication is [1], [0] cannot have a multiplicative inverse. O k a y , why not try the nonzero elements [fl] + [0] as a candidate for a group u n d e r this product? H e r e again it is no go if n is not a prime. For instance, if n = 6, then [2] + [0], [3] * [0], yet [2] [3] = [6] = [0], so the nonzero elements d o not, in general, give us a group. So we ask: Can we find an appropriate piece of Z„ that will form a
62
Groups
Ch. 2
group u n d e r multiplication? Yes! Let U„ = [[a] £ Z„  (a, n) = 1), noting that (a, n) = 1 if and only if (b, n) = 1 for [a] = [b]. By the Corollary to T h e o r e m 1.5.5, if (a, n) = 1 and (b, n) = 1, then (ab, n) = 1. So [a][b] = [ab] yields that if [a], [b] G U„, then [aft] G £/„ and U„ is closed. Associativity is easily checked, following from the associativity of the integers u n d e r multiplication. T h e identity element is easy to find, namely [1]. Multiplication is commutative in U„. N o t e that if [a][b] = [a][c] where [a] G U„, t h e n we have [ab] = [ac], and so [ab — ac] = [0]. This says that n \a(b — c) = ab — ac; but a is relatively prime to n. By T h e o r e m 1.5.5 one must have that n \ (b  c), and so [b] = [c]. In other words, we have the cancellation property in U„. By P r o b lem 2 of Section 2, U„ is a group. W h a t is the order of U„l By the definition of U„, \ U„ \ = n u m b e r of integers 1 < m < n such that (m, n) = 1. This n u m b e r comes u p often and we give it a name. Definition. T h e Eider cpfunction, (p(n), is defined by ^ ( 1 ) = 1 and, for n > 1, <p(n) = the n u m b e r of positive integers m with 1 < m < n such that (m, n) = 1. Thus \ U„ \ = (p(n). If n = p, a prime, we have <p(p) = p — 1. W e see that <p(8) = 4 for only 1, 3, 5, 7 are less t h a n 8 and positive and relatively prime to 8. W e try another one, <p(15). T h e n u m b e r s 1 s m < 15 relatively prime to 15 are 1, 2, 4, 7, 8 , 1 1 , 1 3 , 1 4 , so <p(15) = 8. Let us look at some examples of U„. 1. U = {[1], [3], [5], [7]}. N o t e that [3][5] = [15] = [7], [5] = [25] = [1]. In fact, U is a group of order 4 in which a = e for every a G C / .
s 2 s 8 2
2. L/ = {[1], [2], [4], [7], [8], [11], [13], [14]}. N o t e that [11][13] = [143] = [ 8 ] , [ 2 ] = [1], and so on.
15 4
T h e reader should verify that a = e = [1] for every a G
4 1 g
U.
15 2
3. U = {[1], [2], [4], [5], [7], [8]}. N o t e that [2] = [2], [2] = [4], [if = [8], [2 ] = [16] = [7], [2] = [32] = [5]; also [2] = [2][2] = [2][5] = [10] = [1]. So the powers of [2] give us every element in U . Thus U is a cyclic group of order 6. What other elements in U generate f7 ?
4 5 6 5 9 g 9 9
In parallel to T h e o r e m 2.4.6 we have T h e o r e m 2.4.7. U„ forms an abelian group, u n d e r the [a][b] = [ab], of order <p(n), where (p(ri) is the Euler ^function. product
Sec. 4
Lagrange's Theorem
63
A n immediate consequence of T h e o r e m s 2.4.7 a n d 2.4.5 is a famous r e sult in n u m b e r theory. Theorem 2.4.8 (Euler). = I J
m o c
If a is a n integer relatively prime to n, then
p(n)
a
<p(n)
Proof. U„ forms a group of order <p(ri), so by T h e o r e m 2.4.5, a' = e for all a <EU„. This translates into [a' ] = [aY = [1], which in turn translates into n  (a " — 1) for every integer a relatively prime to p. In other words, a' " = 1 m o d n. •
pin) (n) v( ) Fl )
A special case, where n = p is a prime, is d u e to Fermat. Corollary (Fermat). Up is a prime and p \ a, then a~ F o r any integer b,b
p _ 1 p p l
= 1 m o d p.
= b m o d p.
1 p l p
Proof. Since <p(p) = p — 1, if (fl, p ) = 1, w e have, by T h e o r e m 2.4.8, that a = l ( p ) , hence a • a ~ = a (p), so that fl = a (p). If p\b, t h e n b^O(p) and ft" = 0 ( p ) , so that b = b (p). •
p
Leonard Euler (17071785) was probably the greatest scientist that Switzerland has produced. He was the most prolific of all mathematicians ever. Pierre Fermat (16011665) was a great number theorist. Fermat's Last Theorem—which was in fact first proved in 1994 by Andrew Wiles—states that the equation a" + b" = c" (a, b, c, n being integers) has only the trivial solution where tf = 0orfe = 0 o r c = 0 i f « > 2 . O n e final cautionary word about Lagrange's T h e o r e m . Its converse in general is not true. That is, if G is a finite group of order n. then it need n o t be true that for every divisor m of n there is a subgroup of G of order m. A group with this property is very special indeed, and its structure can be spelled out quite well and precisely.
PROBLEMS
Easier Problems 1. Verify that the relation ~ is a n equivalence relation on the set S given. (a) (b) (c) (d) S S S S = = = = U reals, a ~ b if a  b is rational. C, t h e complex n u m b e r s , a~b\f\a \ = \b\. straight lines in the plane, a ~ b if a, b are parallel. set of all people, a ~ b if they have t h e same color eyes.
64
Groups
Ch. 2
2. T h e relation ~ on the real n u m b e r s IR defined by a ~ b if b o t h a > b and b > a is not an equivalence relation. Why not? W h a t properties of an equivalence relation does it satisfy? 3. Let ~ be a relation on a set S that satisfies (1) a ~ b implies that b ~ a and (2) a ~ b and b ~ c implies that a ~ c. These seem to imply that a ~ a. For ifa~b, then by (1), b ~ a, so a ~ b, b ~ a, so by (2), a ~ a. If this argument is correct, then the relation ~ must be an equivalence relation. Problem 2 shows that this is not so. What is wrong with the argum e n t we have given? 4. Let S be a set, {S } n o n e m p t y subsets such that S = U S and S fl Sp = 0 if a ¥= /3. Define an equivalence relation on S in such a way that the 5 are precisely all the equivalence classes.
a a a a
a
* S. Let G be a group and H a subgroup of G. Define, for a, b £ G, a ~ b if «" '6 £ i / . Prove that this defines an equivalence relation on G, and show that [A] = aH = [ah\h £ H). T h e sets A / / are called left cosets of in G. 6. If G is S and 77 = [i, f}, where f:S S is defined by / ( x ) = 2  f( i) i , f( i) 3 , list all the right cosets of H in G and list all the left cosets of H in G.
3 : x x = x x = x
7. In Problem 6, is every right coset of H in G also a left coset of H in G ? 8. If every right coset of H in G is a left coset of H in G, prove that aHa^ = Hfor all a £ G.
1
9. In Z , write down all the cosets of the subgroup H = {[0], [4], [8], [12]}. (Since the operation in Z„ is + , write your coset as [a] + H. W e d o n ' t need to distinguish between right cosets and left cosets, since Z„ is abelian u n d e r +.)
1 6
10. In Problem 9, what is the index of H in Z ? (Recall that we defined the index i (H) as the n u m b e r of right cosets in G.)
1 6 G
11. F o r any finite group G, show that there are as many distinct left cosets of H in G as there are right cosets of H in G. 12. If aH and bH are distinct left cosets of H in G, are Ha and Hb distinct right cosets of H in G ? Prove that this is true or give a counterexample. 13. Find the orders of all the elements of U .
iS
Is U Is U
18
cyclic? cyclic?
2
14. Find the orders of all the elements of U .
20
20
* 15. If p is a prime, show that the only solutions of x 1 m o d p or x = —1 m o d p. *16. If G is a finite abelian group and a ,..., that x = « i f l • • • fl„ must satisfy x = e.
± 2 2
= 1 m o d p are x =
a„ are all its elements, show
17. If G is of odd order, what can you say about the x in Problem 16?
Sec. 4
Lagrange's T h e o r e m
65
18. Using the results of Problems 15 a n d 16, prove that if p is an odd prime n u m b e r , then (p  1)! =  1 m o d p. (This is known as Wilson's Theorem.) It is, of course, also true if p = 2. 19. Find all the distinct conjugacy classes of S.
3
20. In the group G of Example 6 of Section 1, find the conjugacy class of the element T . Describe it in terms of a and b.
ab
21. Let G be the dihedral group of order 8 (see E x a m p l e 9, Section 1). Find the conjugacy classes in G. 22. Verify Euler's T h e o r e m for n = 14 and a = 3, and for n = 14 and a = 5. 23. In U , show that there is an element a such that [a] = [  1 ] , that is, an integer a such that a =  1 m o d 4 1 .
4i 2 2
24. If p is a prime n u m b e r of the form An + 3, show that we cannot solve x = 1 m o d p [Hint: Use F e r m a t ' s T h e o r e m that a '
1 2
'' = 1 m o d p if p \ a.]
25. Show that the nonzero elements in Z„ form a group under the product [a][b] = [ab] if and only if n is a prime. MiddleLevel Problems 26. Let G be a group, H a subgroup of G, and let S be the set of all distinct right cosets of H in G, T the set of all left cosets of H in G. Prove that there is a 11 mapping of 5 onto T. (Note: T h e obvious m a p that comes to mind, which sends Ha into aH, is not the right one. See Problems 5 and 12.) 27. If all = bll forces Ha = Hb in G, show that n/Zrv" = H for every A £ G. 28. If G is a cyclic group of order n, show that there are <p(n) generators for G. Give their form explicitly. 29. If in a group G, aba' gers /•.
5 s 1 1
= b', show that a'ba~
1 2
r
= b' for all positive inte
r
30. If in G a = e and aba'
= b , find o(b) if 6 # e.
*31. If o («) = 777 and a = e, prove that m  s. 32. Let G be a finite group, H a subgroup of G. Let f(a) be the least positive 777 such that a'" £ H. Prove t h a t / ( a )  o(a). 33. If z ¥= f E. A(S) is such that / = i, p a prime, and if for some s £ S, f'(s) = s for some 1 < / < p, show f h a t / ( s ) = 5 . 34. If / £ A (S) has order p, p a prime, show that for every s £ S the orbit of s u n d e r / has one or p elements. [Recall: T h e orbit of s under / is {/'"(*) I / a n y integer).]
p
66
Groups
Ch. 2
35. If / G A (S) has order p, p a prime, and S is a finite set having n elements, where (n,p) = 1, show that for some s G S , f(s) = s. Harder Problems 36. If a > 1 is an integer, show that n \ <p(a"  1), where <p is the E u l e r (^function. [Hint: Consider the integers mod(fl"  1).] 37. In a cyclic group of order n, show that for each positive integer m that divides n (including m = 1 and m = ri) there are <p(m) elements of order m. 38. Using the result of P r o b l e m 37, show that n = 2  „ < p ( m ) .
m
39. Let G be a finite abelian group of order n for which the n u m b e r of solutions of x'" = e is at most m for any m dividing n. Prove that G must b e cyclic. [Hint: Let i//(m) be the n u m b e r of elements in G of order m. Show that i/)(ra) ^ <p(m) and use P r o b l e m 38.] 40. Using the result of P r o b l e m 39, show that U , if p is a prime, is cyclic. (This is a famous result in n u m b e r theory; it asserts the existence of a primitive root m o d p.)
p
41. Using the result of Problem 40, show that if p is a prime of the form p = 4n + 1, then we can solve x = — 1 m o d p (with x an integer).
2
42. Using Wilson's T h e o r e m (see Problem 18), show that if p is a prime of the form p = 4/7 + 1 and if
then y 41.)
2
==  1 m o d p. (This gives another proof of the result in Problem
43. Let G b e an abelian group of order n, and a ,... , a its elements. L e t x = a a • • ' « „ • Show that: (a) If G has exactly one element b # e such that /3 = e, then x = fo. (b) If G has m o r e than one element b e such that fo = e, then x = e. (c) If n is odd, then x = e (see Problem 16).
1 n i 2 2 2
5. H O M O M O R P H I S M S A N D N O R M A L S U B G R O U P S
In a certain sense the subject of group theory is built u p out of three basic concepts: that of a h o m o m o r p h i s m , that of a normal subgroup, and that of the factor or quotient group of a group by a normal subgroup. W e discuss the first two of these in this section, and the third in Section 6. Without further ado we introduce the first of these.
Sec. 5
Homomorphisms and Normal Subgroups
67
Definition. Let G, G' b e two groups; t h e n the mapping <p : G —» G' is a homomorphism if <p(ab) = cp(a)cp(b) for all a,b £ G. (Note: This cp has nothing to d o with the E u l e r <pfunction.) In this definition the product on t h e left side—in <p(ab)—is that of G, while the product cp(a)cp(b) is that of G'. A short description of a h o m o m o r phism is that it preserves the operation of G. We do not insist that cp be onto; if it is, we'll say that it is. Before working out some facts about homomorphisms, we present some examples. Examples 1. L e t G be the group of all positive reals u n d e r the multiplication of reals, and G ' t h e group of all reals u n d e r addition. L e t cp: G — G' b e defined by » <p(x) = l o g x for x £ G. Since l o g ( x y ) = l o g x + l o g y , we have <p(xy) = <p(x) + <p(y), so cp is a h o m o m o r p h i s m . It also happens to be o n t o and 11.
1 0 10 1 0 1 0
2. Let G be an abelian group and let cp: G —> G be defined by cp (a) = a . Since cp(ab) = (ab) = a b = (p(a)cp(b), cp is a h o m o m o r p h i s m of G into itself. It need n o t be onto; the r e a d e r should check that in U (see Section 4) a = e for all a £ U , so cp(G) = (e).
2 2 2 s 2 8
2
3. T h e example of U above suggests t h e socalled trivial homomorphism. Let G be any group and G ' any other; define <p(x) = e', the unit element of G ' , for all x £ G. Trivially, cp is a h o m o m o r p h i s m of G into G ' . It certainly is n o t a very interesting one. A n o t h e r homomorphism always present is the identity mapping, i, of any group G into itself. Since i(x) = x for all x £ G, clearly i(xy) = xy = i(x)i(y). T h e m a p i is 11 and onto, but, again, is not too interesting as a homomorphism.
s
4. L e t G b e the group of integers u n d e r + and G' = {1,  1 } , t h e subgroup of the reals under multiplication. Define cp(m) = 1 if m is even, cp(m) =  1 if m is odd. T h e statement that cp is a h o m o m o r p h i s m is merely a restatement of: even + even = even, even + odd = odd, and odd + odd = even. 5. Let G be the group of all nonzero complex numbers under multiplication and let G' be the group of positive reals under multiplication. L e t cp: G —> G' be defined by cp (a) = \a\; t h e n cp (ab) = \ab\ = \ a\ \ b\ = cp (a) cp (b), so cp is a h o m o m o r p h i s m of G into G ' . In fact, cp is onto. 6. L e t G be the group in E x a m p l e 6 of Section 1, and G' t h e group of nonzero reals u n d e r multiplication. Define cp: G —> G' by cp(T ) = a. T h a t
ab
68
Groups
Ch. 2
cp is a h o m o m o r p h i s m follows from t h e product rule in G, namely, T T T
ab
1
cd
=
ac, ad + b •
1. L e t G = Z b e t h e group of integers u n d e r + and let G' = Z„. Define <p : G — Z„ by <p(m) = [/»]. Since t h e addition in Z„ is defined by [in] + [/*] = > [m + r], we see that <p(m + r) = <p(m) + <p(r), so cp is indeed a h o m o m o r phism of Z onto Z„. 8. T h e following general construction gives rise to a wellknown t h e o r e m . Let G be any group, and let A(G) be the set of all 11 mappings of G onto itself—here we are viewing G merely as a set, forgetting about its multiplication. Define T : G — G by T (x) = ax for every i £ G , W/raf K the product, T T , of > T and T as mappings on G ? Well,
a a a b a b
(T T ){x)
a h
= T CT x)
a b
= T (bx)
a
= a(bx) = (ab)x =
a b ab
T (x)
ah
(we used t h e associative law). So we see that T T = T . Define t h e mapping tp: G —> A(G) by <p(a) = T , for a E G. T h e p r o d uct rule for t h e T's translates into q>(ab) = T = T„T = cp(a)cp(b), so < is a p h o m o m o r p h i s m of G into A(G). W e claim that cp is 11. Suppose that cp(a) = cp(/3), that is, T = T . Therefore, a = T (e) = T (e) = b, so cp is indeed 11. It is n o t onto in general—for instance, if G has order n > 2, t h e n A(G) has order n\, and since nl > n, cp doesn't have a ghost of a chance of being onto. It is easy to verify that the image of cp, cp(G) = [T  a G G } , is a subgroup ofA(G). T h e fact that cp is 11 suggests that p e r h a p s 11 h o m o m o r p h i s m s should play a special role. W e single t h e m o u t in t h e following definition.
a ab b a b a b a
Definition. T h e h o m o m o r p h i s m cp: G — G ' is called a » monomorphism if < is 11. A m o n o m o r p h i s m that is o n t o is called an isomorphism. A n p isomorphism from G t o G itself is called an automorphism. O n e m o r e definition. Definition. T w o groups G a n d G ' are said to be isomorphic if there is an isomorphism of G onto G'. W e shall d e n o t e that G a n d G ' are isomorphic by writing G — G'. This definition seems t o b e asymmetric, but, in point of fact, it is not. For if there is an isomorphism of G onto G ' , there is o n e of G' o n t o G (see P r o b l e m 2). W e shall discuss m o r e thoroughly later what it m e a n s for t w o groups t o be isomorphic. B u t now we summarize what we did in E x a m p l e 8.
Sec, 5
Homomorphisms and Normal Subgroups
69
Theorem 2.5.1 (Cayley's Theorem). E v e r y group G is isomorphic to some subgroup of A (S), for an appropriate S. T h e appropriate S we used was G itself. But there may be better choices. W e shall see some in the problems to follow. W h e n G is finite, we can take the set S in T h e o r e m 2.5.1 to be finite, in which case A (S) is S„ and its elements are permutations. In this case, Cayley's T h e o r e m is usually stated as: A finite group can be represented as a group of permutations. (Arthur Cayley (18211895) was an English mathematician who worked in matrix theory, invariant theory, and many other parts of algebra.) This is a good place to discuss the importance of "isomorphism." L e t cp be an isomorphism of G o n t o G'. W e can view G' as a relabeling of G, using the label cp(x) for the element x. Is this labeling consistent with the structure of G as a group? T h a t is, if x is labeled cp(x), y labeled <p(y), what is xy labeled as? Since cp(x)cp(y) = cp(xy), we see that xy is labeled as cp(x)cp(y), so this renaming of the elements is consistent with the product in G. So two groups that are isomorphic—although they need not be equal—in a certain sense, as described above, are equal. Often, it is desirable to b e able to identify a given group as isomorphic to some concrete group that we know. W e go on with m o r e examples. 9. Let G be any group, a £ G fixed in the discussion. Define cp: G —> G by cp(x) = a~ xa for all x £ G. W e claim that cp is an isomorphism of G onto itself. First,
l
cp(xy) = a~ (xy)a
x
= a~ xa
l
• a~ ya
x
= cp(x)cp(y),
so cp is at least a h o m o m o r p h i s m of G into itself. It is 11 for if cp (x) = cp (y), then a~ xa = a ~ ya, so by cancellation in G we get x = y. Finally, cp is o n t o , for x = fl" (fl.vfl" ')A = cp (flxfl" ) for any x £ G. H e r e cp is called the inner automorphism of G induced by a. The n o t i o n of automorphism and some of its properties will come u p in the problems. O n e final example:
l l 1 l
10. Let G b e the group of reals u n d e r + and let G' be the group of all nonzero complex numbers u n d e r multiplication. Define cp : G —> G' by cp(x) = c o s x + i s i n x . W e saw that ( c o s x + / s i n x ) ( c o s y + / s i n y ) = cos(x + y) + i sin(x + y ) ,
70
Groups
Ch. 2
h e n c e cp (x)cp(y) = cp(x + y) and < is a homomorphism of G into G ' . cp is n o t p f1 because, for instance, <p(0) = <p(277) = 1, nor is cp onto. Now that we have a few examples in hand, we start a little investigation of homomorphisms. W e begin with L e m m a 2.5.2. If cp is a h o m o m o r p h i s m of G into G ' , then: (a) cp(e) = e', the unit element of G ' . (b) cp{  )
a 1
= cp{aY for all a G G.
v
P r o o / Since x = xe, cp(x) = cp(xe) = cp(x)<p(e); by cancellation in G' we get cp{e) = e'. Also, cpf/zfl" ) = cp(e) = e', hence e' = cp(aa~ ) = cp(a)cp ( a ) , which proves that <p(fl ) = cp ( A ) . •
1 l  1 _1
 1
Definition.
T h e ('mage of cp, <p(G), is cp(G) = (cp(a)  a G G } .
W e leave to the reader the proof of L e m m a 2.5.3. If is a h o m o m o r p h i s m of G into G ' , then the image of cp is a subgroup of G ' . W e singled out certain homomorphisms and called them m o n o m o r phisms. Their property was that they were 11. W e want to m e a s u r e how far a given h o m o m o r p h i s m is from being a monomorphism. This p r o m p t s the Definition. If cp is a h o m o m o r p h i s m of G into G ' , t h e n the kernel of cp, Ker cp, is defined by Ker cp = \a G G  cp(a) = e'}. Kercp measures the lack of 11' ness at one point e'. W e claim that this lack is rather uniform. W h a t is W = {x G G\cp(x) = w'} for a given w' G G'l We show that if cp(x) = w' for some x G G, t h e n W = [kx\k G Kercpj = (Ker<p).\\ Clearly, if k G Ker cp and = w ' , then cp(kx) = cp(k)cp(x) = e'cp(x) = w', so /cx G W. Also, if <p(x) = <p(y) = w ' , t h e n <p(x) = ip(y), hence ^ ( y ) ^ ^  ) " = e'\ but < ( x ) " = cp (x ~ ') by p L e m m a 2.5.2, so e ' = cp(y)<p(x)~ = cp(y)cp(x~ ) = c p ( y x ) , whence yx~ G K e r cp a n d so y G ( K e r cp)x. Thus the inverse image of any element w ' in cp ( G ) G G ' is the set (Ker cp)x, where x is A«y element in G such that cp(x) = w'. W e state this as
1 1 1 l _ 1 1
L e m m a 2.5.4. If w' G G' [y G G  cp(y) = w'} = (Ker cp)x.
is of the form
cp(x)
=
w',
then
Sec. 5
Homomorphisms and Normal Subgroups
71
W e n o w shall study some basic properties of t h e kernels of h o m o m o r phisms. T h e o r e m 2.5.5. If cp is a h o m o m o r p h i s m of G into G', t h e n (a) K e r cp is a subgroup of G. (b) Given a G G, a ^ f K e r cp)a C K e r cp. Proof. A l t h o u g h this is so i m p o r t a n t , its proof is easy. If a, b G K e r cp, then cp(a) = cp(b) = e', hence cp(ab) = cp(a)cp(b) = e', whence ab G K e r cp, so K e r cp is closed u n d e r product. A l s o cp (a) = e' implies that cp (a~ ) = <p(a)~ = e', a n d so a^ G K e r cp. Therefore, K e r cp is a subgroup of G. If k G K e r cp and a G G, then <p(/c) = e'. Consequently, cp(a~ ka) = cp (a~ )cp(k)cp(a) = cp{a~ )e'cp{a) = ^ ( a ) ^ ( a ) = cp(cT a) = <p(e) = e'. This tells us that a~ ka G K e r 9 , h e n c e a ( K e r cp)a G K e r <p. T h e t h e o r e m is n o w completely proved. •
J l 1 1 l x _ 1 l l _ 1
Corollary. If cp is a h o m o m o r p h i s m m o n o m o r p h i s m if and only if K e r cp = (e).
of G into G',
t h e n 9 is a
Proof. This result is really a corollary t o L e m m a 2.5.4. W e leave t h e few details to t h e reader. • Property (b) of K e r cp in T h e o r e m 2.5.5 is an interesting a n d basic o n e for a subgroup to enjoy. W e r a n into this property in t h e text material a n d problems earlier on several occasions. W e use it t o define the ultraimportant class of subgroups of a group. Definition. T h e subgroup N of G is said t o b e a normal subgroup G if a~ Na C N for every a G G.
x
of
Of course, Ker cp, for any h o m o m o r p h i s m , is a normal subgroup of G. A s we shall see in the next section, every n o r m a l subgroup of G is the kernel of some appropriate h o m o m o r p h i s m of G into a n appropriate group G'. So in a certain sense t h e notions of h o m o m o r p h i s m a n d n o r m a l subgroups will b e shown t o b e equivalent. A l t h o u g h w e defined a n o r m a l s u b g r o u p via a~ Na C TV, we actually have a~ Na = N. F o r if a~ Na C Nfor all a G G, then N = a(a~ Na)a~ C aNa' = ( / r ' ^ T ' / V i i T C TV. So N = aNa' for every a G G. T r a n s p o s i n g , we h a v e Na = aN; t h a t is, every left coset of T in G is a right coset of V N in G. O n the o t h e r hand, if every left coset of 7V in G is a right coset, then t h e left coset aN, which contains a, must b e equal to the right coset containing a,
1 1 1 1 1 1 1 1
72
Groups
l
Ch. 2
namely Na. Thus, aN = Na and N = a~ Na for all a G G, which is to say that T is normal in G. V W e write "TV is a normal subgroup of G " by the abbreviated symbol N<G. N o t e that a~ Na = N does not m e a n that a~ na = n for every n G TV. N o — m e r e l y that the set of all a' na is the same as the set of all n. W e have proved
l l y
T h e o r e m 2.5.6. N < G if and only if every left coset of TV in G is a right coset of T in G. V Before going any further, we pause to look at some examples of kernels of homomorphisms and normal subgroups. If G is abelian, then every subgroup of G is normal, for a~ xa = x for every a, x G G. T h e converse of this is not true. Nonabelian groups exist in which every subgroup is normal. See if you can find such an example of order 8. Such nonabelian groups are called Hamiltonian, after the Irish m a t h e m a t i cian W. R. Hamilton (18051865). T h e desired group of order 8 can b e found in t h e quaternions of Hamilton, which we introduce in Chapter 4, Section 1. In E x a m p l e 1, <p(x) = l o g x , a n d K e r cp = [x  l o g x = 0} = {1}. In E x a m p l e 2, where G is abelian, a n d cp(x) = x ,
l H ) 1 0 2
K e r cp = [x G G  x
2
= e}
T h e kernel of the trivial h o m o m o r p h i s m of E x a m p l e 3 is all of G. In E x a m ple 4, K e r cp is t h e set of all even integers. In E x a m p l e 5, Ker cp = {a G C  \a  = 1), which can b e identified, from the polar form of a complex n u m b e r , as Ker cp = ( c o s x + i s i n x  x real}. In E x a m p l e 6, K e r cp = {T j, G G  b real}. In E x a m p l e 7, K e r cp is the set of all multiples of n. In E x amples 8 and 9, t h e kernels consists of e alone, for t h e maps are m o n o m o r phisms. In E x a m p l e 10, w e see that K e r cp = \2irm \ m any integer}. Of course, all the kernels above are n o r m a l subgroups of their respective groups. W e should look at some n o r m a l subgroups, intrinsically in G itself, without recourse to t h e kernels of h o m o m o r p h i s m . W e go back to the examples of Section 1.
x
1. In E x a m p l e 7, H = {T„ G G  a rational}. If T G G, we leave it to t h e r e a d e r to check that T~], HT C H a n d so H < G.
b x y xy
2. I n E x a m p l e 9 t h e subgroup [i, g, g , g } < G. H e r e t o o w e leave t h e checking to the reader. 3. In E x a m p l e 10 the subgroup H = {i, h, h ,..., we also leave to the reader.
2
2
3
h"~ } is normal in G. This
1
Sec. 5
Homomorphisms and Normal Subgroups
73
4. If G is any group, Z(G), the center of G, is a normal subgroup of G (see E x a m p l e 11 of Section 3). 5. If G = S , G has the elements i, f, g, g , fg, and gf, where f(x ) = x , f(x ) = x , , f(x ) = x and g(x ) = x , g(x ) = x , g(x ) = X j . W e claim that the subgroup /V = [i, g, g } <i S . A s we saw earlier (or can c o m p u t e
3 1 2 2 3 3 x 2 2 3 3 2 3 2
now), fgr = g = g ,fg r
2 3
1
1
2
2
l
= g, (fg) gifgr = fggg'r = M~ =
1
1
l
g , and so on. So N <\ S follows. T h e material in this section has b e e n a r a t h e r rich diet. It m a y not s e e m so, b u t the ideas presented, although simple, are quite subtle. W e recomm e n d that the r e a d e r digest the concepts and results thoroughly before going on. O n e way of seeing h o w complete this digestion is, is to t a k e a stab at many of the almost infinite list of problems that follow. The material of the next section is even a richer diet, and even h a r d e r to digest. Avoid a m a t h e matical stomachache later by assimilating this section well.
PROBLEMS
Easier Problems 1. D e t e r m i n e in each of the parts if the given mapping is a h o m o m o r p h i s m . If so, identify its kernel and w h e t h e r or not the mapping is 11 or onto. (a) G = Z u n d e r + , G' = Z „ , cp(a) = [a] for a G Z . (b) G group, cp: G —> G defined by cp(a) = a for a e G. (c) G abelian group, cp: G » G defined by cp (a) = cC for a G G. (d) G group of all nonzero real n u m b e r s under multiplication, G' = {1, —1}, cp(r) = 1 if r is positive, cp(r) =  1 if r is negative. (e) G an abelian group, n > 1 a,fixed integer, and cp: G » G defined by cp(a) = a" for a G G.
 1 l
2. Recall groups (a) G (b) G (c) G i
x x
that G — G' m e a n s that G is isomorphic to G ' . Prove that for all Gx, G , G : = Gj. = G implies that G = G . = G , G = G implies that G ! — G .
2 3 2 2 t 2 2 3 3 _ 1
3. L e t G b e any group and A ( G ) the set of all 11 mappings of G, as a set, o n t o itself. Define L : G — G by L ( x ) = x a . Prove that: » (a) / „ G / 1 ( G ) . (b) L L = L . (c) T h e mapping </>: G —> A(G) defined by ip(a) = L is a m o n o m o r phism of G into A ( G ) .
a a a b f l h a
74
Groups
Ch. 2
4. In Problem 3 prove that for all a, b G G, T L fined as in Example 8.
a
b
= LT,
b a
w h e r e T is dea
5. In Problem 4, show that if V G A(G) is such that T ,V = VT for all a G G, then V = L for some b G G. (Hint: Acting o n e G G, find out what b should be.)
t a b
6. Prove that if cp : G » G' is a h o m o m o r p h i s m , then cp ( G ) , the image of G, is a subgroup of G ' . 7. Show that <p: G — G'', where < is a h o m o m o r p h i s m , is a m o n o m o r p h i s m > p if and only if Ker cp = (e). 8. Find an isomorphism of G, the group of all real n u m b e r s u n d e r + , onto G', the group of all positive real n u m b e r s under multiplication. 9. Verify that if G is the group in E x a m p l e 6 of Section 1, and 77 = {T„ G G  a rational), then 77 O G.
b
10. Verify that in E x a m p l e 9 of Section 1, the set 77 = {;', g, g , g } is a normal subgroup of G, the dihedral group of order 8. 11. Verify that in E x a m p l e 10 of Section 1, the subgroup H= is normal in G. 12. Prove that if Z ( G ) is the center of G, then Z ( G ) <1 G. 13. If G is a finite abelian group of order n and cp: G — G is defined by > (r?) = o'" for all A G G, find the necessary and sufficient condition that cp be an isomorphism of G o n t o itself. 14. If G is abelian and cp: G — G ' is a h o m o m o r p h i s m of G onto G ' , p r o v e > that G ' is abelian. 15. If G is any group, T <3 G, and cp: G > G' a h o m o m o r p h i s m of G onto V G', prove that the image, cp(N), of T is a normal subgroup of G'. V 16. If T < G and M < G and MTV = [ran  m G M, n G TV), p r o v e that M/V is V a subgroup of G and that MN O G. 17. If M < G, N < G, prove that M n T < G. V 18. If 77 is any subgroup of G and T = D V
l a e G
2
3
[i,h,h ,
2
...,h"~ }
[
a~ Ha, prove that N <\ G.
l
19. If 77 is a subgroup of G, let /V(/7) be defined by the relation N(H) = [fl G G  a~ Ha = 77). Prove that: (a) 7V(77) is a subgroup of G and 7V(77) D 77. (b) 77 < N(H). (c) If 7C is a subgroup of G such that 77 < K, then K C 7V(77). [So 7V(77) is the largest subgroup of G in which 77 is normal.] 20. If M< G, T < G, and M n T = (e), show that for ra G M, n G TV, ran = nra.. V V
Sec. 5
Homomorphisms and Normal Subgroups
75
21. Let S b e any set having m o r e than two elements and A(S) the set of all 11 mappings of S onto itself. If s G S, we define H(s) = { / G A(S)  f(s) = s}. Prove that H(s) cannot be a normal subgroup of A(S). 22. Let G = S ,
3
the symmetric group of degree 3 and let H = [i, / } , w h e r e
2
f(.X )
{
=
X
2
, f(x )
=
X j , f(x )
3
=
x
3
.
(a) Write down all the left cosets of If in G. (b) Write down all the right cosets of H in G. (c) Is every left coset of H a right coset of HI 23. L e t G be a group such that all subgroups of G are normal in G. If a, b G G, prove that ba = a'b for some;'. 24. If G G are two groups, let G = Gj X G , the Cartesian product of G G [i.e., G is the set of all ordered pairs (a, b) where a G G b G G ] . Define a product in G by ( a b )(a , b ) = (a a , b b ). (a) Prove that G is a group. (b) Show that there is a m o n o m o r p h i s m <pj of G into G such t h a t cp (G ) <1 G, given by cp (aA = 6 2 ) . where e is the identity element of G . (c) Find the similar m o n o m o r p h i s m cp of G into G. (d) Using the mappings <p cp of Parts (b) and (c), prove that <P\(G )(p (G ) = G and <Pi(Gj) D <p (G ) is the identity element of G. (e) Prove that G x G — G X G .
l 5 2 2 l 5 2 l 5 2 1 ; t 2 2 t 2 x 2 x 1 1 x 2 2 2 2 u 2 x 2 2 2 2 x 2 2 x
25. Let G be a group and let W = G X G as defined in Problem 24. Prove that: (a) T h e mapping cp: G defined by cp (a) = (a, a) is a m o n o m o r p h i s m of G into W. (b) T h e image <p(G) in W [i.e., {(a, A)  a G G}] is n o r m a l in W if and only if G is abelian. MiddleLevel P r o b l e m s *26. If G is a group and a G G, define cr : G > G by cr„(<?) g • We saw in Example 9 of this section that a is an isomorphism of G o n t o itself, so cr„ G A(G), the group of all 11 mappings of G (as a set) o n t o itself. D e fine iff: G —> A (G) by ijj(a) = a for all a G G. Prove that: (a) i/> is a h o m o m o r p h i s m of G into A (G). (b) Ker <// = Z ( G ) , the center of G.
= fl
a
a
l
a
a
27. If 9 is an automorphism of G and T O G, prove that 6(N) < G. V 28. L e t 9, ip b e automorphisms of G, and let 9ip b e the product of 9 and (// as mappings on G. Prove that 0i/y is an automorphism of G, and that 0 " is an automorphism of G, so that the set of all automorphisms of G is itself a group.
1
76
Groups
Ch. 2
* 2 9 . A subgroup 7 of a group Wis called characteristic if <p(T) C 7 for all automorphisms, <p, of W. Prove that: (a) M characteristic in G implies that M <\ G. (b) M, T characteristic in G implies that MTV is characteristic in G. V (c) A n o r m a l subgroup of a group need not be characteristic. (This is quite hard; you must find an example of a group G a n d a noncharacteristic normal subgroup.) 30. Suppose that \G\ = pm, w h e r e p\m and p is a prime. If 77 is a n o r m a l subgroup of o r d e r p in G, prove that 77 is characteristic. 31. Suppose that G is an abelian group of o r d e r p " m w h e r e p \ m is a prime. If 77 is a subgroup of G of order p", prove that 77 is a characteristic subgroup of G. 32. D o P r o b l e m 31 even if G is n o t abelian if you h a p p e n t o k n o w that for some reason or other 77 O G. 33. Suppose that T < G and M C T is a characteristic subgroup of TV. P r o v e V V that M < G. (It is not true that if M < T and T < G, t h e n M m u s t b e V V normal in G. See P r o b l e m 50.) 34. Let G b e a group, s&(G) t h e group of all automorphisms of G. (See P r o b lem 28.) L e t 7(G) = [a \ a G G}, where cr is as defined in P r o b l e m 26. Prove that 7(G) < si(G).
a a
35. Show that Z(G), t h e center of G, is a characteristic subgroup of G. 36. If T < G and 77 is a subgroup of G, show that 77 n T < 77. V V Harder Problems 37. If G is a nonabelian group of o r d e r 6, prove that G — S .
3
38. L e t G b e a group a n d 77 a subgroup of G. L e t S = { 77A  a £ G} b e t h e set of all right cosets of 77 in G. Define, for b G G, T : S  > 5 by 7 (77fl) = 77fl/3" . (a) Prove that T T = 7 for all b, c G G [therefore the mapping <> G — A (5) defined b y i/>(/3) = 7 is a h o m o m o r p h i s m ] . /: » (b) Describe K e r i/>, t h e kernel of i/>: G — A ( S ) . > (c) Show that K e r i/> is t h e largest n o r m a l subgroup of G lying in 77 [largest in t h e sense t h a t if T < G a n d T C 77, then T C K e r i/>]. V V V 39. U s e the result of P r o b l e m 38 to r e d o P r o b l e m 37. Recall that if 77 is a subgroup of G, then t h e index of 77 in G, / ( 7 7 ) , is the n u m b e r of distinct right cosets of 77 a n d G (if this n u m b e r is finite).
b 6 1 b c 6 c A G
40. If G is a finite group, 77 a subgroup of G such that n f i (H)l w h e r e n  G j„. prove that t h e r e is a n o r m a l subgroup T + (e) of G contained in 77. V
G
Sec, 6
Factor G r o u p s
77
4 1 . Suppose that you know that a group G of order 21 contains an element a of order 7. Prove that A = ( a ) , the subgroup generated by a, is normal in G. (Hint: Use the result of Problem 40.) 42. Suppose that you know that a group G of order 36 has a subgroup H or order 9. Prove that either H <3 G or t h e r e exists a subgroup N <1 G, NC H, and /V = 3. 43. Prove that a group of order 9 must be abelian. 44. Prove that a group of order p , p a prime, has a normal subgroup of order p. 45. Using the result of P r o b l e m 44, prove that a group of order p , prime, must be abelian.
3 5 2 2
p a
46. Let G be a group of order 15; show that there is an element a ¥= e in G such that a = e and an element ft e such that ft = e. 47. In P r o b l e m 46, show that b o t h subgroups A [e, 6, ft , ft , ft } are normal in G.
2 3 4
= \e, a, a } and Z? =
2
48. F r o m t h e result of P r o b l e m 47, show that any group of order 15 is cyclic. Very H a r d P r o b l e m s 49. Let G be a group, H a subgroup of G such that i (H) is finite. Prove that t h e r e is a subgroup N d H,N <\ G such that i (TV) is finite.
G c
50. Construct a group G such that G has a n o r m a l subgroup TV, and N h a s a normal subgroup M (i.e., N < G, M < TV), yet M is not n o r m a l in G. 51. Let G be a finite group, cp an a u t o m o r p h i s m of G such that <p is the identity automorphism of G. Suppose that cp(x) = x implies that x = e. Prove that G is abelian and cp(a) = a~ for all a G G.
x 2
52. Let G be a finite group and cp an a u t o m o r p h i s m of G such that cp(x) — x~ for /7?ore than threefourths of t h e elements of G. Prove that <p(y) = y~ for a// y G G, and so G is abelian.
l l
6. FACTOR GROUPS Let G b e a group and T a n o r m a l subgroup of G. In proving Lagrange's T h e V o r e m we used, for an arbitrary subgroup H, the equivalence relation a ~ ft if ab~ G H. Let's try this out when T is n o r m a l and see if we can say a little V more than one could say for just any old subgroup. So, let a ~ ft if aft" G N and let [a] = {x G G  x ~ a}. As we saw earlier, [a] = Na, the right coset of N in G containing a. Recall that in looking at Z„ we defined for it an operation + via [a] + [ft] = [a + ft]. W h y
1 1
78
Groups
Ch. 2
not try something similar for an arbitrary group G and a n o r m a l s u b g r o u p T of G? V So let M = {[A]  a G G ] , w h e r e [a] = [x G G  G TV} = TVA. W e define a product in M via [«][&] = [ab]. W e shall soon show that M is a group u n d e r this product. But first and foremost we must show that this product in M is welldefined. In other words, we must show that if [a] = [a'] and [b] = [b'], then [ab] = [a'b'], for this would show that [a][b] = [ab] = [a'b'] = [«'][£>']; equivalently, that this product of classes does not depend on the particular representatives we use for the classes. Therefore let us suppose that [a] = [a ] and [b] = [£>']. F r o m the definition of our equivalence we have that a' = na, where n G TV. Similarly, b' = mb, where m G TV. T h u s a'b' = namb = n(ama~ )ab; since T < G, V anuC is in TV, so n(ama~ ) is also in TV. So if we let n = n(ama~ ), then rt] G TV and a'b' = n ab. B u t this tells us that a'b' G Nab, so t h a t a'b' ~ ab, from which we have that [a'b ] = [ab], the exact thing we required to ensure that our p r o d u c t in M was welldefined.
l 1 l l x x 1 1
T h u s M is now e n d o w e d with a welldefined p r o d u c t [fl][6] = [ab]. W e now verify t h e g r o u p axioms for M. Closure we h a v e from t h e very definition of this p r o d u c t . If [a], [b], a n d [c] are in M, t h e n [A]([/3][C]) = [fl][/3c] = [a(be)] = [(ab)c] (since t h e p r o d u c t in G is associative) = [A/J][C] = ( [ f l ] [ 6 ] ) [ c ] . T h e r e f o r e , t h e associative law has b e e n established for t h e p r o d u c t in M. W h a t a b o u t a unit e l e m e n t ? W h y not try t h e obvious choice, namely [e]l W e immediately see that [a][e] = [ae] = [a] a n d [e][a] = [ea] = [a], so [e] d o e s act as the unit e l e m e n t for M. Finally, w h a t a b o u t inverses? H e r e , too, t h e obvious choice is the correct o n e . If a G G, t h e n [ A ] [ A ] = [ f l i j T ] = [e], h e n c e [ f l ' ] acts as t h e inverse of [a] relative to t h e p r o d u c t we h a v e defined in M.
 1 1

W e want to give M a n a m e , and better still, a symbol that indicates its dependence on G and TV. The symbol we use for M is GIN (read " G over T or V G mod TV") and GIN is called the factor group or quotient group of G by TV. What we have shown is the very important T h e o r e m 2.6.1. If T < G and V GIN = {[A]  A G G ) = {TVfl  A G G ] , then G/TVis a group relative to the operation [a][b] = [ab]. O n e observation must immediately b e m a d e , namely T h e o r e m 2.6.2. If T O G, then there is a h o m o m o r p h i s m < of G o n t o V / > GIN such that Ker i/>, t h e kernel of if, is TV.
Sec. 6
Factor G r o u p s
79
Proof. T h e most n a t u r a l m a p p i n g from G to GIN is t h e one that does the trick. Define i/>: G » GIN b y i/y (a) = [a]. O u r p r o d u c t as defined in GIN m a k e s of f a h o m o m o r p h i s m , for ip(ab) = [ab] = [a][b] = \fj(a)4>(b). Since every e l e m e n t X G GIN is of t h e form X = [b] = i//(/3)for some b G G, i/> is o n t o . Finally, what is the k e r n e l , K e r ifi, of i/>? By definition, K e r i/> = [a G G  i/> (a) = 7?}, w h e r e £ is t h e unit e l e m e n t of G/7V. B u t what is El N o t h i n g o t h e r t h a n E = [e] = TVe = TV, a n d a G K e r i/> if a n d only if £ = TV = i/> (a) = TVA. B u t TV<7 = TV tells us t h a t a = ea E. Na = N, so we see t h a t K e r <p C N. T h a t TV C K e r i/>—which is e a s y — w e leave t o the reader. So K e r i/> = TV. • T h e o r e m 2.6.2 substantiates the r e m a r k we m a d e in the preceding section that every normal subgroup T of G is the kernel of some h o m o m o r V phism of G onto some group. T h e "some h o m o m o r p h i s m " is t h e i// defined above and the "some g r o u p " is GIN. This construction of the factor group G by T is possibly the single most V important construction in group theory. In other algebraic systems we shall have analogous constructions, as we shall see later. O n e might ask: W h e r e in this whole affair did the normality of T in G V enter? Why not d o the same thing for any subgroup H of G? So let's try and see what h a p p e n s . As before, we define W= {[a]flGG} =
1
[Ha\aEG]
where the equivalence a ~ b is defined by ab' G H. W e try to introduce a product in W as we did for GIN by defining [a][b] = [ab]. Is this product well defined? If h G H, then [hb] = [b], so for the product to be well defined, we would n e e d that [a][b] = [a][hb], that is, [ab] = [ahb]. This gives us that Hab — Hahb, and so Ha = Hah; this implies that H = Haha" , whence aha~ G H. That is, for all a G G and all h G H, aha^ must be in H; in other words, H must be normal in G. So we see that in order for the product defined in W to be welldefined, H must be a normal subgroup of G. W e view this m a t t e r of the quotient group in a slightly different way. If A, B are subsets of G, let AB = [ab \ a G A, b G B}. If H is a subgroup of G, then HH C H is another way of saying that H is closed under the product of G. Let GIN = {TVa  a G G] b e the set of all right cosets of the normal subgroup TV in G. Using the product of subsets of G as defined above, w h a t is (Na)(Nb)7 By definition, (Na)(Nb) consists of all elements of the form (na)(mb), where n, m G TV, and so
1 l 1
(na)(mb)
= (nama~ ){ab)
l
= n^ab,
80
Groups
1
Ch. 2
where n = noma is in TV, since /Vis normal. Thus (Na)(Nb) other hand, if n G N, then
l
C Nab. O n the
n{ab) = (na)(eb)
G
(Na)(Nb),
so that Nab C (Na)(Nb). In short, we have shown that the p r o d u c t — a s subsets of G—of Na and Nb is given by the formula (Na)(Nb) = Nab. All the other group axioms for GIN, as defined here, are now readily verified from this product formula. A n o t h e r way of seeing that (Na)(Nb) = Nab is to n o t e that by the normality of TV, aN = Na, hence (Na)(Nb) = N(aN)b = N(Na)b = NNab = Nab, since NN = T (because T is a subgroup of G). V V H o w e v e r we view GIN—as equivalence classes or as a set of certain subsets of G — w e do get a group whose structure is intimately tied to that of G, via the natural h o m o m o r p h i s m \poiG o n t o GIN. W e shall see very soon how we combine induction and the structure of GIN to get information about G. W h e n G is a finite group and TV < G, then the n u m b e r of right cosets of T in G, i (N), is given—as the proof of Lagrange's T h e o r e m s h o w e d — b y V 'G (n) =  G/TV. But this is the order of GIN, which is the set of all the right cosets of TV in G. Thus  G/TV =  G  /1 T . W e state this m o r e formally as V
G
Theorem 2.63. G/TV.
If G is a finite group and T < G, then G/TV = V
A s an application of what we have been talking about h e r e , we shall prove a special case of a t h e o r e m that we shall prove in its full generality later. T h e proof we give—for the abelian case—is not a particularly good one, but it illustrates quite clearly a general technique, that of pulling back information about GIN to get information about G itself. The theorem we are about to prove is due to the great French mathematician A. L. Cauchy (17891857), whose most basic contributions were in complex variable theory. T h e o r e m 2.6.4 (Cauchy). If G is a finite abelian group of order  G  and p is a prime that divides \ G\, then G has an element of order p. Proof. Before getting involved with the proof, we point out to the reader that the t h e o r e m is t r u e for any finite group. W e shall p r o v e it in the general case later, with a proof that will be much m o r e beautiful t h a n the one we are about to give for the special, abelian case. W e proceed by induction on  G . What does this mean precisely? W e shall
Sec.
6
Factor G r o u p s
81
assume the theorem to be true for all abelian groups of order less than G and show that this forces the theorem to be true for G. If  G  = 1, there is no such p and the theorem is vacuously true. So we have a starting point for our induction. Suppose that there is a subgroup (e) # N =h G. Since 7V < \ G\, if p divides  N , by our induction hypothesis there would b e an element of o r d e r p in N, hence in G, and we would b e done. So we may suppose that p   7 V  . Since G is abelian, every subgroup is normal, so we can form GIN. Because p divides \G\ and p\\N\, and because \GIN\ =  G  /  7 V  , we have that p divides \GIN\. T h e group GIN is abelian, since G is (Prove!) and since N * (e), \N\ > 1, so \G/N\ = \G\I\N\ <  G  . Thus, again by induction, there exists an element in GIN of order p. In other words, there exists an a G G such that [a] = [e], but [a] + [e]. This translates to a E N, a <£ N. So if m = \N\, then (a ) = e. So (a'") = e. If we could show that b = a'" e, then b would b e the required element of order p in G. But if a'" = e, t h e n [a]'" = [e], and since [a] has order p,p \ m (see Problem 31 of Section 4). But, by a s s u m p t i o n , p \ m = \N\. So we are done ifG has a nontrivialsubgroup.
p p p m p
Hut if G has n o nontrivial subgroups, it must b e cyclic of prime order. (See Problem 16 of Section 3, which you should b e able to handle m o r e easily now.) W h a t is this "prime o r d e r " ? Because p divides  G  , we must h a v e j G  = p. But then any element a ¥= e G G satisfies a = e and is of order p. This completes the induction, and so proves the theorem. •
p
W e shall have other applications of this kind of grouptheoretic argument in the problems. T h e notion of a factor group is a very subtle one, and of the greatest importance in the subject. T h e formation of a n e w set from a n old one by using as elements of this new set subsets of the old one is strange to the n e o phyte seeing this kind of construction for the first time. So it is worthwhile looking at this whole m a t t e r from a variety of points of view. W e consider GIN from a n o t h e r angle now. W h a t are we doing when we form G/7V? Sure, we are looking at equivalences classes defined via TV. Let's look at it another way. W h a t we are doing is identifying two elements in G if they satisfy the relation ab^ G N. I n a sense we are blotting out N. So although GIN is not a subgroup of G, we can look at it as G, with T blotted out, and two elements as equal if they are V equal "up to TV." For instance, in forming Z/N, w h e r e Z is the group of integers and TV is the set of all multiples of 5 in Z, what we are doing is identifying 1 with 6, 11, 16, —4, —9, and so on, and we are identifying all multiples of 5 with 0. T h e nice thing about all this is that this identification jibes with addition in Z when we go over to Z/N.
1
82
Groups
Ch. 2
Let's look at a few examples from this point of view. 1. L e t G = {T \a * 0, b real) (Example 6 of Section 1). L e t T = V [T \b real) C G; we saw that TV <1 G, so it makes sense to talk about G/N. N o w T and T are in the same left coset of T in G, so in G/N we are getV ting an element by identifying T with T . T h e latter element just depends on a. Moreover, the T multiply according t o T T =T and if we identify T with T , T with T , then their product, which is T , is identified with T . So in GIN multiplication is like that of the group of nonzero real numbers u n d e r multiplication, and in some sense (which will b e m a d e m o r e precise in the next section) GIN can b e identified with this group of real numbers.
aJ) lb ab a0 ab aQ aJ> ub cd a c a d +b a ai0 cd c0 nc a d + b ac<0
2. Let G be the group of real n u m b e r s u n d e r + and let Z b e the g r o u p of integers under + . Since G is abelian, Z < G, and so we can talk about G/Z. W h a t does G/Z really look like? In forming GIT, we are identifying any two real numbers that differ by an integer. So 0 is identified with —1, —2, —3, . . . and 1, 2, 3, . . . ;  is identified with §, §, —§, —§, . . . . E v e r y real n u m b e r a thus has a mate, a, where 0 < a < 1. So, in G/Z, the whole real line has b e e n compressed into the unit interval [0, 1]. B u t a little more is true, for we have also identified the end points of this unit interval. So we are bending the unit interval around so that its two end points touch and b e c o m e one. W h a t d o we get this way? A circle, of course! So G / Z is like a circle, in a sense that can be m a d e precise, and this circle is a group with an appropriate product. 3. L e t G b e t h e group of n o n z e r o complex n u m b e r s a n d let T = V [a£ G  \a  = 1} which is the unit circle i n the complex plane. T h e n TV is a subgroup of G and is n o r m a l since G is abelian. I n going t o GIN we are declaring that any complex n u m b e r of absolute value 1 will b e identified with the real number 1. Now any a £ G, in its polar form, can b e written as a = r(cos 8 + i sin 8), where r = \a\, and [cos 8 + i sin 0 = 1. I n identifying cos 8 + i sin 8 with 1, we are identifying a with r. So in passing t o GIN every element is being identified with a positive real number, and this identification jibes with the products in G and in the group of positive real numbers, since \ab\ = \a \ \b\. So GIN is in a very real sense (no pun intended) the group of positive real numbers under multiplication.
PROBLEMS
1. If G is the group of all nonzero real n u m b e r s under multiplication and T V is the subgroup of all positive real numbers, write out GIN by exhibiting the cosets of TV in G, and construct the multiplication in GIN.
Sec. 6
Factor G r o u p s
83
2. If G is the group of nonzero real n u m b e r s u n d e r multiplication and /V = {1, —1}, show how you can "identify" GIN as the group of all positive real numbers u n d e r multiplication. W h a t are the cosets of /Vin Gl 3. If G is a group and TV <l G, show that if M is a subgroup of GIN and M = {a G G\Na G M], then M is a subgroup of G, and M D J V . 4. If M in Problem 3 is normal in GIN, show that the A/ defined is normal in G. 5. In P r o b l e m 3, show that MIN must equal M. 6. Arguing as in the E x a m p l e 2, where we identified G/Z as a circle, w h e r e G is the group of reals u n d e r + and Z integers, consider the following: let G = [(a, b)\a,b real}, where + in G is defined by (a, b) + (c, d) = (a + c, b + d) (so G is the plane), a n d let N = {(a, b) G G  a, b are integers}. Show that GIN can b e identified as a torus (donut), and so we can define a product on the d o n u t so that it becomes a group. H e r e , you may think of a torus as the Cartesian product of two circles. 7. If G is a cyclic group and TV is a subgroup of G, show that GIN is a cyclic group. 8. If G is an abelian group and TV is a subgroup of G, show that GIN is an abelian group. 9. D o Problems 7 and 8 by observing that GIN is a homomorphic image of G. 10. Let G be an abelian group of order p'^P? '"' PV > where p p , .. . ,p are distinct prime n u m b e r s . Show that G has subgroups S S ,... , S of orders p'(\ p ,.. ., p ", respectively. (Hint: U s e Cauchy's T h e o r e m and pass to a factor group.) This result, which actually holds for all finite groups, is a famous result in group theory known as Sylow's Theorem. We prove it in Section 11.
ly z k lt 2 k 2 a 2 k !
11. If G is a group and Z ( G ) the center of G, show that if G / Z ( G ) is cyclic, then G is abelian. 12. If G is a group and T <] G is such that GIN is abelian, prove that V aba' b~ G N for all a,b G G.
1 L
13. If G is a group and N <1 G is such that aba^b'
1
GT V
for all a, b £. G, prove that G/7V is abelian. 14. If G is an abelian group of order P\p  • • p , where p distinct primes, prove that G is cyclic. (See Problem 15.)
2 k ly
p ,...
2
,p
k
are
84
Groups
Ch. 2
15. If G is an abelian group and if G has an element of order m and o n e of order n, where m and n are relatively prime, prove that G has an element of order mn. 16. Let G b e an abelian group of order p"m, where p is a prime a n d Let P = [a G G \ a = e for some k depending on a}. Prove that: (a) P is a subgroup of G. (b) GIP has n o elements of order p. (c)  P  = p".
pk
p\m.
17. Let G be an abelian group of order mn, where m a n d n are relatively prime. Let M = { « E G a"' = e). Prove that: (a) M is a subgroup of G. (b) G / M has no element, x, other t h a n t h e identity element, such that x'" = unit element of GIM. 18. Let G be an abelian group (possibly infinite) a n d let t h e set T = {« G G  = e, m > 1 depending on A}. Prove that: (a) T is a subgroup of G. (b) GIT has no e l e m e n t — o t h e r than its identity element—of finite order.
7. T H E H O M O M O R P H I S M T H E O R E M S
Let G be a group and cp a h o m o m o r p h i s m of G onto G'. If K is t h e kernel of cp, then if is a normal subgroup of G, hence we can form G/K. It is fairly natural to expect that there should be a very close relationship b e t w e e n G' and GIK. T h e First Homomorphism Theorem, which we are about to prove, spells out this relationship in exact detail. But first let's look back at some of the examples of factor groups in Section 6 to see explicitly what t h e relationship m e n t i o n e d above might be. 1. Let G = [T  A ¥= 0, b real} a n d let G' be the group of n o n z e r o reals u n d e r multiplication. F r o m t h e product rule of these T's, namely T .hT = T ±, we determined that t h e mapping cp: G —> G' defined by cp(T ) = a is a h o m o m o r p h i s m of G onto G ' with kernel K = {T  b real). O n t h e other hand, in E x a m p l e 1 of Section 6 w e saw t h a t GIK = [KT \a + Q real}. Since
a b a c d acad b aJj t h a0
(*T . )(tfT,, ) =
fl 0 0
KT ,
ax 0
the mapping of GIK onto G ' , which sends each KT„ o n t o a, is readily seen to be an isomorphism of GIK onto G ' . Therefore, GIK — G'.
n
2. In Example 3, G was the group of nonzero complex numbers under multiplication and G' the group of all positive real numbers under multiplication.
Sec. 7
The Homomorphism Theorems
85
Let cp:G —>• G' defined by cp(a) = \a\ for a G G. Then, since \ab\ = ff   b, < is a h o m o m o r p h i s m of G o n t o G ' (can you see why it is o n t o ? ) . p Thus the kernel K of cp is precisely K = {a G G  \a = 1}. But we have already seen that if fl  = 1, t h e n a is of the form cos 0 + z' sin 0. So the set K = {cos 0 + z sin 0  0 < 0 < 2rr\. If A is any complex number, then a = /'(cos 0 + z sin 0), where r = fl , is t h e polar form of a. Thus Ka = /(/•(cos 0 + z sin 0) = /((cos 0 + /' sin 0)r = i ( / \ since /((cos 0 + /' sin0) = / ( because cos 0 + z sin 0 £ K. So GIK, whose elements are the cosets Ka, from this discussion, has all its elements of t h e form Kr, where r > 0. T h e mapping of GIK onto G ' defined by sending Kr onto ;• then defines an isomorphism of GIK onto G'. So, here, too, GIK G'. With this little experience behind us we are ready to m a k e the j u m p t h e whole way, namely, to T h e o r e m 2.7.1 (First H o m o m o r p h i s m T h e o r e m ) . Let cp be a h o m o morphism of G onto G' with kernel K. T h e n G ' — GIK, the isomorphism b e tween these being effected by t h e m a p
i//:
GIK — G ' »
defined by ip(Ka) = <p(a). Proof. T h e best way to show that G//C and G ' are isomorphic is to exhibit explicitly an isomorphism of GIK onto G ' . T h e statement of the t h e o rem suggests what such an isomorphism might b e . So define i/>: GIK — G' by *jj{Ka) = cp(a) for a G G. As usual, our first > task is to show that i/> is well defined, that is, to show that if Ka = Kb, then ip(Ka) = ij/(Kb). This boils down to showing that if Ka = Kb, then cp(a) = cp(b). But if Ka = Kb, then a = kb for some /c G / ( , hence cp(a) = cp(kb) = cp(k)cp(b). Since A: G / ( , the kernel of cp, then <p(/c) = e', the identity element of G', so we get cp(a) = 9(6). This shows that t h e mapping i/>is well defined. Because < is onto G ' , given .T G G', then ,v = cp(a) for some a G G, p thus x = p ( a ) = ip(Ka). This shows that i/> maps G/Z( onto G ' . Is < 11? Suppose that </>(#«) = </>(^): then <p(a) = ip(Ka) = / > cp(b). Therefore, e' = cp{a)cp{b)~ = ( p ^ f i r ) = cpiab' ). Because ab~ is thus in the kernel of cp—which is K—we have ab~ G / ( . This implies that Ka = Kb. In this way i/> is seen to b e 11. Finally, is i/> a h o m o m o r p h i s m ? W e check: tp((Ka)(Kb)) = 4i(Kab) = cp(ab) = cp{a)cp{b) = ip(Ka)ij/(Kb), using that cp is a h o m o m o r p h i s m and that (Ka)(Kb) = Kab. Consequently, < is a h o m o m o r p h i s m of GIK onto G ' , V / and T h e o r e m 2.7.1 is proved. •
i/z(/<:6) =
x
1
1
]
l
86
Groups
Ch. 2
Having talked about the First H o m o m o r p h i s m T h e o r e m suggests that there are others. T h e next result, however, is an extension of t h e First H o m o m o r p h i s m T h e o r e m , and is traditionally called the Correspondence Theorem. In the context of the t h e o r e m above, it exhibits a 11 correspondence between subgroups of G and those subgroups of G that contain K.
1
Theorem 2.7.2 (Correspondence Theorem). Let the map <p: G G ' be a homomorphism of G onto G' with kernel K. If H' is a subgroup of G' and if H = [a £ Gj<p(a) £ / / ' } , then H is a subgroup of G, 77 D K, and HIK = 7 7 . Finally, if H' <\G\ H<G. then
Proof. We first verify that the 77 above is a subgroup of G. It is not empty, since e G 77. If a, b G 77, t h e n <p(a), <p(£>) G 77', hence cp (ab) = cp(a)cp(b) G 77', since 77' is a subgroup of G ' ; this puts ab in 77, so 77 is closed. Further, if a G 77, then cp(a) G 77', hence cp(a~ ) = cp(a)~ is in 77', again since 77' is a subgroup of G ' , whence a~ G 77. Therefore, 77 is a subgroup of G. Because cp(K) = {e'\ C 77', w h e r e e' is the unit element of G ' , we h a v e that K C 77. Since K <\ G and 7C C 77, it follows that 7C < 77. T h e mapping <p restricted to 77 defines a h o m o m o r p h i s m of 77 onto 77' with k e r n e l K. By the First H o m o m o r p h i s m T h e o r e m we get HIK — 77'. Finally, if 77' < G' and if a G G, then cp( y H'cp(a) C 77', so cp(  )H'cp(a) C 77'. This tells us that cp(  Ha) C 77', so a 7 7 a C 77. This proves the normality of 77 in G. •
l l l 1 a 1 1 1 a a
It is worth noting that if K is any normal subgroup of G, and cp is the natural h o m o m o r p h i s m of G o n t o G/X, then the t h e o r e m gives us a 11 correspondence between all subgroups 77' of G/K and those subgroups of G that contain K. Moreover, this correspondence preserves normality in the sense that 77' is normal in G/K if and only if 77 is normal in G. (See P r o b l e m 7, as well as the last conclusion of the theorem.) We now state the Second Homomorphism Theorem, leaving its proof to the r e a d e r in P r o b l e m 5.
Theorem 2.7.3 (Second Homomorphism Theorem). Let / / be a subgroup of a group G and T a normal subgroup of G. T h e n 77TV = V {hn  h G 77, n G TV} is a subgroup of G, 77 Pi T is a n o r m a l subgroup of 77, V and 77/(77 fl TV) — (HN)/N.
Sec.
7
The Homomorphism Theorems
87
Finally, we go on t o the Third Homomorphism Theorem, which tells us a little m o r e about the relationship b e t w e e n T a n d N' when N' <3 G'. V T h e o r e m 2.7.4 (Third H o m o m o r p h i s m T h e o r e m ) . If the m a p cp : G —» G' is a homomorphism of G o n t o G' with kernel / ( t h e n , if TV' <1 G' and N = [a E G\ cp(a) E TV'}, we conclude t h a t G//V — G7/V'. Equivalently, GIN=*(G/K)/(NIK). Proof. Define the mapping I/>: G  » G'lN' by I/)(a) = N' cp(a) for every a E G. Since <p is onto G ' and every element of G'lN' is a coset of the form 7VV, and x' = <p(x) for some x E G, w e see that I/>maps G o n t o G'lN'. F u r t h e r m o r e , ifi is a h o m o m o r p h i s m of G onto G'lN', for tp(ab) = N'q>(ab) = N'q>(a)cp(b) = (N'<p(a))(N'(p(b)) = ip(a)ip(b), since TV' < G ' . W h a t is the kernel, Af, of I/>? If a G M , t h e n if/(a) is the unit element of G'lN', that is, tf/(a) = N'. O n t h e other hand, by the definition of ip, tp(a) = N'cp(a). Because N'cp(a) = N' w e must have cp(a) E TV'; but this puts a in N, by the very definition of N. Thus MEN. T h a t T C Af is easy and is left to V the reader. Therefore, M = N, so ip is a h o m o m o r p h i s m of G onto G'lN' with kernel TV, whence, by the First H o m o m o r p h i s m T h e o r e m , GIN — G'lN'. Finally, again by T h e o r e m s 2.7.1 a n d 2.7.2, G ' = GIK, N' = M/C, which leads us t o G/7V  G'lN' = (G/K)/(N/K). • This last equality is highly suggestive; w e are sort of "canceling o u t " the K in t h e n u m e r a t o r and denominator.
PROBLEMS
1. Show that M D N in t h e proof of T h e o r e m 2.7.3. 2. Let G be the group of all realvalued functions on the unit interval [0, 1], where we define, for f g E G, addition by ( / + g)(x) = f(x) + g(x) for every x E [0, 1]. If N = {f E G \ f(\) = 0}, prove that G/7V = real numbers u n d e r + . 3 . Let G b e the group of nonzero real n u m b e r s under multiplication a n d let N = {1,  1 } . Prove that G/7V — positive real n u m b e r s under multiplication. 4. If Gi, G are two groups a n d G = G X G = {(a, b) \ a E G b E G ), where we define (a, b)(c, d) = (ac, bd), show that: (a) N = {(a, e )  a E G ), w h e r e e is the unit element of G , is a normal subgroup of G. (b) 7Y= G i . (c) G/N=*G .
2 t 2 u 2 2 x 2 2 2
88
Groups
Ch. 2
5. Let G b e a group, H a subgroup of G, and N < G. L e t the set fflV = [h <= H,n <E N\. Prove that: (a) H H N < H. (b) fflV is a subgroup of G. (c) T C HN and TV < 77/V. V (d) (HN)/N = H/(H n TV). *6. If G is a group and N < G, show that if a G G has finite order o ( a ) , t h e n Na in G/7V has finite order m, where m \ o(a). (Prove this by using t h e h o m o m o r p h i s m of G onto GIN.) 7. If ( is a h o m o m o r p h i s m of G onto G' and N < G, show that 9 ( T < G ' . p A)
8. C A U C H Y ' S THEOREM
In T h e o r e m 2.6.4—Cauchy's T h e o r e m — w e proved that if a p r i m e p divides the order of a finite abelian group G, then G contains an element of order p. W e did point out there that Cauchy's T h e o r e m is true even if the group is not abelian. W e shall give a very neat proof of this here; this proof is due to McKay. W e return for a m o m e n t to set theory, doing something t h a t w e m e n tioned in the problems in Section 4. Let S be a set, / £ A(S), and define a relation on S as follows: s ~ t if t = f'(s) for some integer /' (i can b e positive, negative, or zero). W e leave it to the r e a d e r as a problem that this does indeed define an equivalence relation on S. T h e equivalence class of s, [s], is called the orbit of s u n d e r / . So S is the disjoint union of the orbits of its elements. W h e n / is of order p,p a prime, we can say something about the size of the orbits u n d e r / ; those of t h e readers w h o solved P r o b l e m 34 of Section 4 already know the result. W e prove it here t o p u t it on the record officially. [If f (s) = s, of course f (s) = s for every integer t. (Prove!)]
k tk
L e m m a 2.8.1. If / £ A(S) is of order p, p a prime, then the orbit of any element of S u n d e r / h a s 1 or p elements. Proof. Let s £ S; iif(s) = s, then the orbit of s u n d e r / c o n s i s t s merely of s itself, so has one element. Suppose then that f(s) ± s. Consider the elements s, f(s), f (s), . . . , f ' (s); we claim that these p elements are distinct and constitute the orbit of s u n d e r /. If not, then f'(s) = f'(s) for some 0 < i < ; ' < / ?  1, which gives us that f'~'(s) = s. L e t m = j  i; t h e n 0 < m < p  1 a n d / ' " ( s ) = s. B u t / " ( s ) = s and s i n c e p \ m , ap + bm = 1 for some integers a and b. Thus f(s) = f {s) = f {f {s)) = f (s) = s,
2 p 1 ap+bm ap bm ap
Sec. 8
p
Cauchy's T h e o r e m
89
since f'"(s) = f (s) = s. This contradicts that f(s) ¥= s. Thus t h e orbit of s u n d e r / c o n s i s t s of s, f(s), f (s),... ,f ~ (s), so as p elements. •
2 p l
W e now give McKay's proof of Cauchy's T h e o r e m . T h e o r e m 2.8.2 (Cauchy). If/; is a prime and p divides t h e order of G, then G contains an element of order p. Proof. If p = 2, the result amounts to Problem 18 in Section 1. Assume that p + 2. Let S be the set of all orderedptuples (a , a , • • •, a \, a ), where a a , • • •, a G and where a a • • • a \a = e. W e claim that S has n?~ elements where n = \G\. Why? W e can choose a ,..., a _ arbitrarily in G, and by putting a = (a a • • • a^~\ theptuple (a , a ,..., a ^ , a ) then satisfies
x 2 p p a r e m x h 2 p x 2 p p x p x p x 2 x 2 p x p
aa
L
2
• • •a_a
p x p
p
= aa
x 1
2
• • • a _ (a a
p l x
2
• • • cip^y
1
= e,
so is in S. Thus S has n ~ elements. Note that if a a • • • a _ a = e, then a a a • • • a _ = e (for iixy = e in a group, then yx = e). So the mapping / : S ^ S defined by f(a ,..., a) = (a , a , a , . .. , is in A(S). N o t e t h a t / =h /, the identity m a p on S, and
x 2 p x p p x 2 p x x p p x 2
= i, so / is of order p. If the orbit of s u n d e r / has one element, then f(s) = s. O n the other hand, if f(s) ¥= s, we know that the orbit of s u n d e r / consists precisely of p distinct elements; this we have by L e m m a 2.8.1. Now when is f(s) + si W e claim that f(s) + s if and only if when s = (a , a , .. . , a ), t h e n for some i + j , a i + cij. (We leave this to the reader.) So f(s) = s if and only if 5 = (a, a,... , a) for some a G G.
x 2 p
that f
Let m b e the n u m b e r of s G S such that f(s) = s; since for s = (e, e,. . ., e), f(s) = s, we know that m > 1. O n the other hand, if f{s) # s, the orbit of s consists of p elements, and these orbits are disjoint, for they are equivalence classes. If t h e r e are k such orbits where f(s) # s, we get that p~ = + / p f have accounted this way for every element of S.
l n m C ; or w e
But p\nby assumption and p  (kp). So we must have p \ m, since m = ii ~  kp. Because m + 0 and p \ m, we get that m > 1. But this says that there is an s = (a, a... ., a) + (e, e , . . . , e) in S; from the definition of S this implies that a = e. Since a + e, a is the required element of order p. •
p x p
N o t e that the proof tells us that the n u m b e r of solutions in G of x = e is a positive multiple of p. W e strongly urge the r e a d e r who feels uncomfortable with the proof just given to carry out its details for p = 3. In this case the action of / o n S becomes clear and our assertions about this action can b e checked explicitly.
p
90
Groups
Ch. 2
Cauchy's T h e o r e m has m a n y consequences. W e shall present o n e of these, in which we determine completely the nature of certain groups of order pq, where p a n d q are distinct primes. O t h e r consequences will b e found in the problem set to follow, a n d in later material on groups. L e m m a 2.8.3. L e t G be a group of order pq, where p, q are primes and p > q. If a E G is of o r d e r p a n d A is the subgroup of G generated by a, then A <\G. Proof. W e claim that A is the only subgroup of G of order p. Suppose that B is another subgroup of order p. Consider the set AB = [xy  x G A, y G B\; we claim that AB has p distinct elements. F o r suppose that xy = uv where x, u G A, y, v G B; then u~ x = vy~K B u t u~ x G A, vy~ G B, and since u~ x = t r y , we have u~ x G A fl B. Since B + A . a n d A Pi B is a subgroup of A a n d A is of prime order, we are forced t o conclude that A fl B = (e) and so iC x = e, that is, u = x. Similarly, v = y. Thus the n u m b e r of distinct elements in AB is p . B u t all these elements are in G, which has only pq < p elements (since p > q). With this contradiction we see that B = A and A is the only subgroup of order/? in G. B u t if x G G, B = x ~ A x is a subgroup of G of order p, in consequence of which we conclude that x ~ A x = A; hence A<\G.O
2 l 1 l l 1 l i 2 2 L ]
x~ ax
l
Corollary. If G, a are as in L e m m a 2.8.3 and if x G G, then = fl', where 0 < i < p, for some i (depending on x).
l
Proof. Since e + a G A and x ~ ' A x = A, x~ ax G A . B u t every element of A is of the form a', 0 < i < /J, and x 'AX e. In consquence, x ^ a x = A', where 0 < i< p. •

W e n o w prove a result of a different flavor. L e m m a 2.8.4. If a G G is of order m and 6 G (7 is of o r d e r n, w h e r e m and n are relatively p r i m e and ab = /3A, then c = ab is of order Proof. Suppose that A is the subgroup generated by a and B that generated by b. Because  A  = m a n d B  = n and (m, «) = 1, w e get A fl B = (e), which follows from Lagrange's T h e o r e m , for A D 5 1  n and \ A D B\ \ m. Suppose that c' = e, w h e r e i > 0; thus (ab)' = e. Since A/3 = ba, e = (A&)' = a'b'; this tells us that a' = b ~ ' G A n 5 = (e). So A'' = e, whence in  z', and />' = e, whence n \ i. Because (m, n) = 1 and m a n d n both divide i, inn divides i. So / > mn. Since (ab)'"" = a"'"b""' = e, we see that m/z is t h e smallest positive integer i such that (ab)' = e. This says that ab is of o r d e r mra, as claimed in the lemma. •
Sec.
8
Cauchy's T h e o r e m
91
Before considering the m o r e general case of groups of order pq, let's look at a special case, namely, a group G of order 15. By Cauchy's T h e o r e m , G has elements b of order 3 and a of order 5. By the Corollary to L e m m a 2.8.3, b" ab = a', where 0 < i < 5. T h u s
x
b' ab
2
2
= b'\b~ ab)b
3 3 3
l
= b^db
3
= (b^ab)'
3
= (V)' =
a'
2
nnd similarly, b~ ab = a' . But b = e, so we get a' = a, whence a' ~ = e. Since a is of order 5, 5 must divide i  1, that is, i = 1(5). However, by Fermat's T h e o r e m (Corollary to T h e o r e m 2.4.8), i = 1(5). These two equations for /' tell us that / = 1(5), so, since 0 < i < 5, i — 1. In short, b" ab = a' = a, which m e a n s t h a t ab = ba. Since a is of o r d e r 5 and b of order 3 , by L e m m a 2.8.4, c = ab is of order 15. This m e a n s that t h e 15 powers e = c°, c, c ,. .., ; are distinct, so must sweep out all of G. In a word, G must be cyclic. T h e argument given for 15 could h a v e b e e n m a d e shorter, but the form n which we did it is the exact prototype for the proof of the m o r e general
3 3 4 l 2 1 4
3
4
Theorem 2.8.5. Let G b e a group of order pq, where p, q are primes m d p > q. li q \p — 1, then G must be cyclic. Proof. By Cauchy's T h e o r e m , G has an element a of order p a n d an element b of order q. By the Corollary to L e m m a 2.8.3, b~ ab = a' for s o m e i with 0 < i < p. Thus b'ab = a '' for all r > 0 (Prove!), and so b~«ab<< = a'". But b = e; therefore, a'' = a and so a ^ = e. Because a is of order p, we conclude that p\i — 1, which is to say, i = l(p). However, by F e r m a t ' s Theorem, z = l(p). Since q\p — 1, we conclude that i = l(p), and since ) < i < p, i = 1 follows. Therefore, b~ ab = a' = a, hence ab = ba. By L.emma 2.8.4, c = ab has order pq, so the powers of c sweep out all of G. Thus G is cyclic, and the t h e o r e m is proved. •
l r 1 q 1 1 1 q q p _ 1 l
PROBLEMS
MiddleLevel Problems 1. In the proof of T h e o r e m 2.8.2, show that if some two entries in s = (ai, a , ... , a ) are different, then f(s) s, and the orbit of s u n d e r / has p elements.
2 p
2. Prove that a group of order 35 is cyclic. 3 . Using the result of P r o b l e m 40 of Section 5, give another proof of L e m m a 2.8.3. (Hint: U s e for H a subgroup of order p.) 4. Construct a nonabelian group of order 21. (Hint: Assume that a
3
= e,
92
1
Groups
l
Ch. 2
b = e and find some / such that a ba = a' j= a, which is consistent with the relations a = b = e.)
3 1
5. Let G b e a group of o r d e r p " m , w h e r e p is prime a n d p \ m . Suppose that G has a normal subgroup P of order p". Prove that 9(P) = P for every automorphism 0 of G. 6. Let G be a finite group with subgroups A, B such that \A \ > V  G j and  5  > V j G ) . Prove that / l n 6 # (e). 7. If G is a group with subgroups A, B of orders in, n, respectively, where m and n are relatively prime, prove that the subset of G, AB = [ab ) a G A, b E B), has distinct elements. 8. Prove that a group of order 99 has a nontrivial n o r m a l subgroup. 9. Prove that a group of order 42 has a nontrivial n o r m a l subgroup. 10. F r o m the result of P r o b l e m 9, prove that a group of o r d e r 42 has a normal subgroup of order 2 1 . H a r d e r Problems 11. If G is a group and A, B finite subgroups of G, prove that t h e set AB = {ab  a G A, b £ B} has (\A\ \B\)I\A f l B\ distinct elements. 12. Prove that any two nonabelian groups of order 21 are isomorphic. (See Problem 4.) Very Hard Problems 13. Using the fact that any group of order 9 is abelian, prove that any group of order 99 is abelian. 14. Let p > q b e two primes such that q\p  1. Prove that t h e r e exists a nonabelian group of o r d e r pq. (Hint: Use the result of P r o b l e m 40 of Section 4, namely that U is cyclic if p is a prime, and the idea n e e d e d to do Problem 4 above.)
p
15. Prove that if p > q are two primes such that q\p abelian groups of order pq are isomorphic.
— 1, then any two non
9. DIRECT P R O D U C T S
In several of the problems and examples that a p p e a r e d earlier, we went through the following construction: If G , , G are two groups, t h e n G = Gi X,G is the set of all ordered pairs (a, b), where a G G and b G G and
2 2 x 2
Sec. 9
Direct P r o d u c t s
93
where the product was defined componentwise via (a , b )(a , b) = (a a , b b ), the products in each c o m p o n e n t being carried out in the respective groups G and G . W e should like to formalize this procedure here.
x l 2 2 x 2 x 2 x 2
Definition. If G , G ,. . . , G„ are n groups, then their (external) direct product G X G X G X • • • X G„ is the set of all ordered ntuples ( a , , a ,. . ., a„) where a , G G,, for i = 1, 2,.. . , n, and where the product in G] X G X • • • X G„ is defined componentwise, that is,
x 2 x 2 3 2 2
(«!, a ,.
2
.., a )(b ,
n 1 x 2
b ,...,
2
b„) = (a b ,
x x
a b ,a„b„).
2 2
That G = G X G X • • • X G„ is a group is immediate, with (e , e ,.. ., e„) as its unit element, w h e r e e is the unit element of G and where (a , a , • • • > „ ) = ( « f \ « 2 > • • • , fl,r')' G is merely the Cartesian product of the groups G , G , . . . , G„ with a product defined in G by componentwise multiplication. We call it external, since the groups G , G ,. . ., G„ are any groups, with n o relation necessarily holding among them. Consider the subsets G, C G , X G X • • • X G„ = G, w h e r e
x 2 t h a 1
_ 1
x
2
x
2
x
2
2
G, =
;
. . ., e
M
, a,, e
i + 1
, . . . , e„) \ a G G,};
t
in other words, G consists of all /ituples w h e r e in the iih component any element of G, can occur and where every other c o m p o n e n t is the identity element. Clearly, G, is a group and is isomorphic to G, by the isomorphism TTJ\ Gj —> G defined by 7r (e , e ,. . . , a . . ., e ) = a,. F u r t h e r m o r e , not only is G, a subgroup of G but G, <J G. (Prove!) Given any element a = (a , a ,. . . , a„) G G, t h e n
t / 1 2 h n x 2
a = (a,, e ,..
2
. , e )(e ,
n x
a
2 )
e , . . . , e„) • • • ( e e , . . . , e,,.,, a„);
3 1; 2
2
that is, every a G G can b e written as a = a j a • • • a „ , where each a, G G,. Moreover, a can be written in this way in a unique manner, that is, if a = a d • • • a„ = b b • • • b , where the a, G G, and b, G G,, then a = & ! , . . . ,a = b . So G is built u p from certain normal subgroups, the Gj, as G = G G • • • G„ in such a way that every element a G G has a unique representation in the form a = a j a , • • • a„ with a, G G,. This motivates the following
x 2 L 2 n 2 n n ± 2
Definition. T h e group G is said to b e the (internal) direct product of its normal subgroups N N ,. . . , N„ if every a G G has a unique representation in the form a = a a • • • a„, w h e r e each a, G TV, for = 1, 2 , . . . , n.
u 2 x 2
F r o m what we have discussed above we have the
94
Groups
Ch. 2
Lemma 2.9.1. If G = G , X G X • • • X G„ is t h e external direct product of Gi, G , • • •, G „ , then G is the internal direct product of the normal subgroups G G , . . ., G„ defined above.
2 2 l 5 2
We want t o go in the other direction, namely t o prove that if G is the internal direct product of its normal subgroups N , N ,..., N„, t h e n G is isomorphic to N X TV, X • • • X A ,,. T o d o so, we first get some preliminary results. T h e result we a r e about t o prove h a s already occurred as P r o b l e m 20, Section 5. F o r the sake of completeness we prove it h e r e .
x 2 7 x
Lemma 2.9.2. L e t G b e a group, M, N normal subgroups of G such that M fl N = (e). Then, given ra E M and n G N, mn = nm. Proof. Consider t h e element a = mnm~ n~ . Viewing a as b r a c k e t e d one way, a = (mnm~ )n~ ; then, since N <I G a n d n G N, mnm~ G N, so a = {mnm~ )n~ is also in N. N o w bracket a in the other way, a = ni(nm~ n~ ). Since M <3 G a n d m ' G M , we have nm~ n~ G M a n d so fl = m(nm~ rC ) G M . T h u s a G M, n N = (e), which is t o say, mmn~ nT = e. This gives us that mn = n m , as required. •
l l l l l 1 1 1 l l x v x x l x
If G is the internal direct product of the normal subgroups N N ,... , N„, w e claim that A , PI Nj = (e) for i + j . F o r suppose that T A E A, Pi A^; then A = e • e • • • eae • • • e, where the a occurs in the z'th place. This gives us one representation of A in G = N N • • • N . O n the other hand, a = e e  e a e e, where the a occurs in the y'th place, so a has the second representation as an element of A^A/^ • • • N„. By the uniqueness of the representation, we get fl = e, and so A, Pi Nj = (e). Perhaps things would b e clearer if w e d o it for n = 2. So suppose that A ) <l G, N <I G, a n d every element a G G has a unique representation as a — A j • A , where a G A/j, fl G Af . Suppose that a E. N C\ N ; then A = A • e is a representation of a = a • a with a = a G N , a = e G A^. H o w ever A = e • a, so A = b • b , where b = e E. N , b = a E N . By t h e uniqueness of t h e representation w e must have a = b , that is, a = e. So N PI N = (e). T h e argument given above for A ,,..., N is t h e same argument as that given for n = 2, but perhaps is less transparent. A t any rate we have proved
h 2 7 1 2 n 7 7 2 2 x 2 2 x 2 x 2 x x 2 x 2 x x 2 2 { x x 2 7 n
Lemma 2.9.3. If G is t h e internal direct product of its normal subgroups N , N ,..., N„, then, for / ;', A , Pi Nj = (e). T x 2
Corollary. If G is as in L e m m a 2.9.3, then if / # / a n d A, £ A , a n d cij GW, we have A;fly = A A,.
;
7
Sec.
9
Direct Products
;
95
Proof. By L e m m a 2.9.3, iV fl TV} = (e) for i /. Since the TV's are normal in G, by L e m m a 2.9.2 we have that any element in N commutes with any element in TV,, that is, a a = aja, for a G N a  G TV,. •
t l j t h ;
With these preliminaries out of the way we can now prove T h e o r e m 2.9.4. Let G be a group with normal subgroups N , N ,. • •, N„. T h e n the mapping ip(a , a , • •., a„) = a a • • • a„ is an isomorphism from N X N X • • • X N„ (external direct product) onto G if and only if G is the internal direct product of N , N ,. . . , N„.
x 2 x 2 x 2 x 2 x 2
Proof. Suppose G is an internal direct product of N ,. . ., N„. Since every element a in G has a representation a = a a • • • a„, with the a, £ N we have that the mapping ip is onto. W e assert that it is also 11. F o r if tjj((a a , . . . , a,,)) = !/>((&!, b ,.., b,,)), then by the definition of a a • • • a = b b • • • b„. By the uniqueness of the representation of an element in this form we deduce that a = b a = b , ... , a„ = b„. Hence < is 11. / > All that remains is to show that i/> is a h o m o m o r p h i s m . So, consider
x x 2 h u 2 2 x 2 n x 2 x h 2 2
(//((«!, a ,...,
2
«„)(&!, b ,...,
2
b„)) = ip((a b ,
L x
a b ,a„b„))
2 2
=
faA)(flA)
x L 2 2
•'"
n n
( «b,)
a
= abab
x l5 h t
• • • ab.
Since b G 7V it commutes with a b for i > 1 by the Corollary to L e m m a 2.9.3. So we can pull the b across all the elements to the right of it to get a ba b • • • a„b„ = a a b a b • • • a„b„b . Now repeat this procedure with b , and so on, to get that a b a b • • • a„b„ = (a a • • • a )(b b • • • b ). Thus
x 1 l 2 2 1 2 2 3 3 1 2 y 1 2 2 1 2 n 1 2 n
^((fli,
a ,...,
2
a )(b ,
n t
b ,...,
2
b„)) = a b a b
x x 2
2
• • • a„b • • • a )(b b
n x 2
n
= {a a
x
2
• • • b„)
x 2
= <K(fli, « . • • • . a,))4i{{b b ,...,
2
b,,)).
In other words, i// is a h o m o m o r p h i s m . O n the other hand, suppose that ii/ is an isomorphism. T h e n the conclusion that G is the internal direct p r o d u c t of N , N ,..., 7V„ easily follows from the fact that »> is onto and 11. / With this the proof of T h e o r e m 2.9.4 is complete. •
x 2
Corollary. Let G be a group with normal subgroups N , N . T h e n G is the internal direct product of N and 7V if and only if G = N N and N n N = (e).
x 2 x 2 X 2 x 2
96
Groups
Ch. 2
Proof. This follows easily from the fact that ifj\N XN —> given by i/f(flj, i) \ i> is isomorphism if and only if N\N N nN = (<?)• •
i 2 a = a a a n l 2
2
G, which is = G and
In view of the result of T h e o r e m 2.9.4 and its corollary, we drop the adjectives "internal" and "external" and merely speak about the "direct product." W h e n notation G = N X N is used it should be clear from context whether it stands for the internal or external direct product. T h e objective is often to show that a given group is the direct product of certain normal subgroups. If o n e can do this, the structure of the group can b e completely d e t e r m i n e d if we h a p p e n to k n o w those of the n o r m a l subgroups.
x 2
P R O B L E M S
1. If Gi and G are groups, prove that G
2 2
x
X G —G
2
2
X
G.
1
2. If Gi and G are cyclic groiips of orders m and n, respectively, prove that G X G is cyclic if and only if m and n are relatively prime.
x 2
3. Let G be a group, A = G X G. In A let T = {(g, g)\g<= (a) Prove that T ~ G. (b) Prove that T<AH and only if G is abelian.
k k
G).
4. Let G be an abelian group of order p'fp'p • • • p' , w h e r e p , p ,..., p are distinct primes and m > 0, m > 0 , . . . , m > 0. By P r o b l e m 10 of Section 6, for each i, G has a subgroup T , of order p"' . Show that G = P j X P X • • • X P .
x 2 x 2 k 3 1 2 k
k
5. Let G be a finite group, N , N ,.. ., N normal subgroups of G such that G = N N • •• N and  G  = \N \ \N \ • • • \N \. Prove that G is the direct product of N , N ,. . . , N .
x 2 k t 2 k L 2 k x 2 k
6. Let G be a group, N , N ,. 1. G = N N • • • N .
x 2 L 2 k
.., N normal subgroups of G such that:
k
2. F o r each i, Nj n (N N
1
2
• • • N y  i ^ + i • • • N ) = (e).
k lt
Prove that G is the direct product of N
N ,...
2
, N.
k
10. FINITE A B E L I A N G R O U P S (OPTIONAL)
W e have just finished discussing the idea of t h e direct p r o d u c t of groups. If we w e r e to leave that topic at the point where we e n d e d , it might s e e m like a nice little construction, but so what? To give some m o r e substance to it,
S e c . 10
Finite A b e l i a n Groups (Optional)
97
we should prove at least one t h e o r e m which says that a group satisfying a certain condition is the direct product of some particularly easy groups. Fortunately, such a class of groups exists, the finite abelian groups. W h a t we shall prove is that any finite abelian group is the direct product of cyclic groups. This reduces most questions a b o u t finite abelian groups to questions about cyclic groups, a reduction that often allows us to get complete answers to these questions. T h e results on the structure of finite abelian groups are really special cases of some wider and deeper theorems. T o consider these would be going too far afield, especially since the story for finite abelian groups is so important in its own right. T h e t h e o r e m we shall p r o v e is called the Fundamental Theorem on Finite Abelian Groups, and rightfully so. Before getting down to the actual details of the proof, we should like to give a quick sketch of how we shall go about proving the theorem. O u r first step will be to reduce t h e problem from any finite abelian group to one whose order is p", w h e r e p is a prime. This step will be fairly easy to carry out, and since the group will have order involving just one prime, the details of the proof will not be cluttered with elements whose orders are somewhat complicated. So we shall focus on groups of order p". Let G be an abelian group of order p". W e want to show that there exist cyclic subgroups of G, namely Ai, A , • • • ,A , such that every element x £ G can be written as x = b b b , w h e r e each b £ A in a unique way. Otherwise put, since each Aj is cyclic and generated by a say, we want to show that x = a'" a  • • • a"' , where the elements a'"' are unique.
2 k y 2 k { h h l k 2
A difficulty appears right away, for there is not just one choice for these elements a , ... , a . For instance, if G is the abelian group of order 4 with elements e, a, b, ab, where a = b = e and ab = ba, then we can see t h a t if A, B, C are the cyclic subgroups generated by a, b, and ab, respectively, t h e n G = AxB=AXC=BxC.So there is a lack of uniqueness in the choice of the «,. H o w to get around this?
x k 2 2
W h a t we need is a mechanism for picking a and which, when applied after we have picked a , will allow us to pick a , and so on. W h a t should this mechanism be? Our control on the elements of G lies only in specifying their orders. It is the order of the element—when properly used—that will give us the means to prove the theorem.
x x 2
Suppose that G = A X A X • • • X A , where \G\ = p" and the A's have been numbered, so that \Aj\ = p"> and «j s n > • • • > n , and each A; is cyclic generated by a . If this were so and x = fl"' • • • a' , then
x 2 k 2 k 1 k t k
x"
p
1
= (fl'// • • • < * ) ' " " = ci'{'"""a' " "
2
1
2P
1
• • • a' " "
k
kP
1
98
Groups
iP p
Ch. 2
because n ~ n p"'\ p"\ so since every a"' "' = e, thus x "' = e. In other words, a should be an element of G whose order is as large as it can possibly be. Fine, we can now pick a . W h a t do we do for a ? If G = GIA , then to get the first element n e e d e d to represent G as a direct p r o d u c t of cyclic groups, we should pick an element in G whose order is maximal. W h a t does this translate into in G itself? W e want an element a such that a requires as high a power as possible to fall into A . So that will be the road to t h e selection of the second element. However, if we pick an element a with this property, it may not do the trick; we may have to adapt it so that it will. T h e doing of all this is the technical part of the argument and does go through. Then one repeats it appropriately t o find an element a , and so on.
x h x x 2 X 2 2 x 2 3
This is the p r o c e d u r e we shall be going through to prove the t h e o r e m . But to smooth out these successive choices of a , a ,..., we shall use an induction argument and some subsidiary preliminary results. With this sketch as guide we h o p e the proof of the t h e o r e m will m a k e sense to the reader. O n e should not confuse the basic idea in the proof— which is quite reasonable—with the technical details, which m a y cloud the issue. So we now begin to fill in the details of the sketch of the proof that we outlined above.
x 2
L e m m a 2.10.1. Let G be a finite abelian group of order mn, where m and n are relatively prime. If M = {x G G  x' = e] and N = {x G G  x" = e], then G = M X N. Moreover, if neither m nor n is 1, then M i= (e) and N + (e).
n
Proof. T h e sets M and N d e n n e d in the assertion above are quickly seen to b e subgroups of G. Moreover, if m + 1, then by Cauchy's T h e o r e m ( T h e o r e m 2.6.4) we readily obtain M (e), and similarly if n 1, that N + (e). F u r t h e r m o r e , since M n N is a subgroup of b o t h M and N, by Lagrange's T h e o r e m , \ M D N \ divides \M\ = m and jV = n. Because m and n are relatively prime, we obtain \M fl N\ = 1, hence M C\ N = (e). To finish the proof, we n e e d to show that G = MN and G = M X N. Since m and n are relatively prime, there exist integers r and s such that rm + sn = 1. If a G G, then a = a = a " = a "a "; since (a*") " = a""" = e, we h a v e that a " G M. Similarly, a'" G T . T h u s a = a "a " is in V MN. In this way G = MN. It n o w follows from Corollary to T h e o r e m 2.9.4 that G = M X T . • V
1 s +rm s n 1 s 1 s n
A n immediate consequence is the Corollary. Let G be a finite abelian group and let p b e a prime such that p divides  G  . T h e n G = P X T for some subgroups P and T, w h e r e \P\ v. p"\ m > 0, and  T\ is not divisible by p.
S e c . 10
pS
Finite A b e l i a n G r o u p s ( O p t i o n a l )
99
Proof. Let P = [x G G\ x = e for some s) and let the subset T = [x G G  x' = e for r relatively prime to/?}. By L e m m a 2.10.1, G = P X T and / (e) Since every element in P has order a power of p, \P\ is not divisible by any other prime (by Cauchy's T h e o r e m ) , so \P\ = p'" for some m. It is easy to see that p j  T\ by making use of Lagrange's Theorem. T h u s we really have that P is not merely some subgroup of G but is what is called a pSylow subgroup of G. (See Section 11). •
J
W e now come to the key step in the proof of the theorem we seek. The proof is a little difficult, but once we have this result the rest will be 3asy. T h e o r e m 2.10.2. Let G be an abelian group of order p", p a prime, and let a G G have maximal order of all the elements in G. Then G = A X Q, where A is the cyclic subgroup g e n e r a t e d by a. Proof. W e proceed by induction on n. If n = 1, then \ G\ = p and G is already a cyclic group generated by any a j= e in G. W e suppose the t h e o r e m to b e true for all m < n. W e first show that the theorem is correct if there exists an element b G G such that b £ A = (a) and b = e. Let B = (b), the subgroup of G generated by b; thus A n B = (e) (see Problem 1). Let G = GIB\ by assumption B = (e), hence \G\ < \G\. In G, what is = the order of d = 5 a ? We claim that o(a) = o(a). To begin with, we know that o(a)  o(a) (see Problem 6 of Section 2.7). O n the other hand, a " e, so fl ' G 5 . Since a G A , we see that fl" ' G A f l B = (e), whence <.(«) _ tells us that o(a) \ o(a). H e n c e o(a) = o(d). Since a is an element of maximal o r d e r in G, by the induction we k n o w that G = ( f l ) X T for some subgroup T of G. By the Correspondence T h e o r e m we also know that T = QIB for some subgroup Q of G. We claim that G is the internal direct product A X Q. T h a t G = AQ is left to the reader. It remains to show that A (~l O = (e). Let a' £ A fl (J. T h e n fl' G <2A/3 = 7", and since (a) (1 T = (e), we have that d' = e. But since o(a) = o(fl), this implies a' = e. Therefore, A f l Q = (e) and we obtain that G = A X Q. Suppose, then, that t h e r e is no element b in G, b not in A , such that b = e. We claim that this forces G = A = (fl), in which case G is a cyclic group. Suppose that G + A and let x G G, x £ A have smallest possible order. Because o(x ) < o(x), we have, by our choice of x, that x G A, h e n c e x = «' for some i. W e claim that p\i. Let o(a) = /r , and n o t e that the maximality
p o ( ) = o( 7) o(7,) ( 7) fl e p p p p !
100
Groups
pS pS p pS l i pS l
Ch. 2
of the order of a implies that x = e. But x = (x ) ~ = (a ) ~ = e. Since o(a) = p' , we have /?  /. T h u s x = a', where p  i. Let y = fl~' ' • x. T h e n y = a~'x = cT'a' = e. Moreover, y £ (a) = A , because x & A. But this puts us back in the situation discussed above, w h e r e t h e r e exists a b 6 G, £ £ A such that b = e;m that > case we saw that the t h e o r e m was correct. So we must have G = («), and G is a cyclic group. This finishes the induction and proves the t h e o r e m . •
5 p /f p p p
We are n o w able to prove the very basic and important Theorem 2.10.3 ( F u n d a m e n t a l Theorem on Finite A b e l i a n G r o u p s ) . A finite abelian group is the direct product of cyclic groups. Proof. Let G b e a finite abelian group and p a prime that divides  G  . By the Corollary to L e m m a 2.10.1, G = P X T, where  P  = p". By T h e o r e m 2.10.2, P = Ai X A X ••• X A , where the A, are cyclic subgroups of P. A r guing by induction on G, we may thus assume that T = T X T X • • • X T , where the T, are cyclic subgroups of T. T h u s
2 k x 2 q
G = (A = A
}
x
X
A A
2
2
X
• • •
X
A)
k
X
(T,
X
T
2
2
X
• • •
X
T)
q
X
X
• • •
X
A
k
x
r,
X
T
X
• • •
X
T.
q
This very important t h e o r e m is n o w proved.
•
We return to abelian groups G of order p". W e now have at h a n d that G = A X A X • • • X A , w h e r e the A are cyclic groups of order p"'. W e can arrange the n u m b e r i n g so that n s n > • • • > n . Also, \G\ = \A X A X • • • X A \ = \A \ \A \ • • • \A , which gives us that
l 2 k t i 2 k 1 2 k X 2 k
hence n = n + n + • • • + n . T h u s the integers ( 1 , 2 : 0 give us a partition of n. It can be shown that these integers H j , n ,.. . , n —which are called the invariants of G — a r e unique. In other words, two abelian groups of order p" are isomorphic if and only if they have the same invariants. G r a n t e d this, it follows that the n u m b e r of nonisomorphic abelian groups of o r d e r p" is equal to the n u m b e r of partitions of n. For example, if /; = 3, it has the following three partitions: 3 = 3, 3 = 2 + 1,3 = 1 + 1 + 1, so there are three nonisomorphic abelian groups of order p (independent of p). T h e groups corresponding to these partitions are a cyclic group of order p , the direct product of a cyclic group of order p by one of order p, and the direct product of t h r e e cyclic groups of order p, respectively.
x 2 k 2 k 3 3 1
S e c . 11
Conjugacy and S y l o w ' s T h e o r e m (Optional)
101
For n = 4 we see the partitions are 4 = 4, 4 = 3 + 1, 4 = 2 + 2, 4 = 2 + 1 + 1 , 4 = 1 + 1 + 1 + 1, which are five in number. Thus there are five nonisomorphic groups of order p . C a n you describe t h e m via the partitions of 4? Given an abelian group of order n  p p ' ' ' p"l\ w h e r e the p are distinct primes and the A, are all positive, then G is the direct product of its s o  c a l l e d p ,  Sylow subgroups (see, e.g., the Corollary to L e m m a 2.10.1). F o r each prime p t h e r e are as many groups of order p" as there are partitions of A;. So the n u m b e r of nonisomorphic abelian groups of order n = p\ • • • p" is f(adf( i) ' ' ' /( /<r)' where f(m) denotes the n u m b e r of partitions of m. Thus we know how many nonisomorphic finite abelian groups there are for any given order.
4 ai a 2 2 t 1 t y k k a fl
F o r instance, how many nonisomorphic abelian groups are there of order 144? Since 144 = 2 3 , and there are five partitions of 4, two partitions of 2, there are 10 nonisomorphic abelian groups of order 144. T h e material treated in this section has b e e n hard, the p a t h somewhat tortuous, and the effort to understand quite intense. To spare the reader too much further agony, we assign only t h r e e problems to this section.
4 2
PROBLEMS
1. Let A be a normal subgroup of a group G, and suppose that b G G is an element of prime order p, and that b £ A. Show that A D (b) = (e). 2. Let G be an abelian group of order p", p a prime, and let a G G have maximal order. Show that x = e for all x £ G.
o ( a )
3. Let G be a finite group, with N <\ G and a G G. Prove that: (a) T h e order of aN in GIN divides the order of a in G, that is, o(aN)  o(fl). (b) If (A) n T = (e), then o(aN) = ola). V
1 1 . C O N J U G A C Y A N D SYLOW'S THEOREM (OPTIONAL)
In discussing equivalence relations in Section 4 we mentioned, as an example of such a relation in a group G, the notion of conjugacy. Recall that the element b in G is said to be conjugate to a G G (or merely, a conjugate of a) if there exists an x G G such that b = x~ ax. W e showed in Section 4 that this defines an equivalence relation on G. T h e equivalence class of a, which we denote by cl(fl), is called the conjugacy class of a.
l
102
Groups
Ch. 2
F o r a finite group an immediate question presents itself: H o w large is cl(fl)? Of course, this depends strongly on t h e element a. F o r instance, if a £ Z(G), the center of G, then ax = xa for all x £ G, hence x~ ax = a; in other words, the conjugacy class of a in this case consists merely of the element a itself. O n the other hand, if cl(a) consists only of t h e element a, then x~ ax = a for all x £ G. This gives us that xa = ax for all x £ G, hence a £ Z ( G ) . So Z ( G ) is characterized as t h e set of those elements a in G whose conjugacy class has only o n e element, a itself.
x l
F o r an abelian group G, since G = Z ( G ) , two elements are conjugate if and only if they are equal. So conjugacy is not an interesting relation for abelian groups; however, for nonabelian groups it is a highly interesting n o tion. Given a £ G, cl(fl) consists of all x~ ax as x runs over G. So to determine which are the distinct conjugates of a, we need to know w h e n two conjugates of a coincide, which is the same as asking: W h e n is x~ ax = y~ ay? In this case, transposing, we obtain a(xy~ ) = (xy~ )a; in other words, xy~ must c o m m u t e with a. This brings us to a concept introduced as E x a m p l e 10 in Section 3, that of the centralizer of a in G. W e repeat something we did there.
x x x x x x
Definition. If a £ G, then C(a), the centralizer by C(a) = [x £ G  xa = ax}.
of a in G, is defined
W h e n C(a) arose in Section 3 we showed that it was a subgroup of G. W e record this now more officially as L e m m a 2.11.1. For a £ G, C{a) is a subgroup of G.
x x
A s we saw above, t h e two conjugates x' ax and y~ ay of a are equal only if x y £ C(a), that is, only if x and y are in the same right coset of Cia) in G. O n the other hand, if x and y are in the same right coset of C{a) in G, then xy~ £ C{a), hence xy~ a = axy~ . This yields that x~ ax = y~. ay. So x and y give rise to the same conjugate of a if and only if x and y are in the same right coset of C(a) in G. Thus there are as many conjugates of a in G as there are right cosets of C(a) in G. This is most interesting when G is a finite group, for in that case the n u m b e r of right cosets of C(a) in G is what we called the index, i ( C ( a ) ) , of C(a) in G, and is equal to  G  /1 C(a) . W e have proved
 1 x l x x x c
T h e o r e m 2.11.2. L e t G be a finite group and a £ G; t h e n t h e n u m b e r of distinct conjugates of a in G equals the index of C(a) in G.
S e c . 11
Conjugacy and S y l o w ' s T h e o r e m (Optional)
103
In other words, the n u m b e r of elements in cl(«) equals i (C(a)) = \G\l\C{a)\. This t h e o r e m , although it was relatively easy to prove, is very important and has many consequences. We shall see a few of these here. O n e such consequence is a kind of b o o k k e e p i n g result. Since conjugacy is an equivalence relation on G, G is the union of t h e disjoint conjugacy classes. M o r e o v e r , by T h e o r e m 2.11.2, we k n o w how many elements there are in each class. Putting all this information together, we get
G
T h e o r e m 2.11.3 ( T h e Class E q u a t i o n ) . G\=^i (C(a))
G
If G is a finite group, then ^
a
=
M
\C(a)\'
a
where the sum runs over one a from each conjugacy class. It is almost a sacred tradition a m o n g mathematicians to give, as the first application of the class equation, a particular t h e o r e m about groups of order p", where p is a prime. N o t wanting t o b e accused of heresy, we follow this tradition and prove the pretty and important T h e o r e m 2.11.4. If G is a group of order p", w h e r e p is a prime, t h e n Z ( G ) , the center of G, is not trivial (i.e., there exists an element a ¥= e in G such that ax = xa for all x G G ) . Proof. W e shall exploit the class equation to carry out the proof. L e t z =  Z ( G )  ; as we pointed out previously, z is then the n u m b e r of elements in G whose conjugacy class has only o n e element. Since e G Z ( G ) , z > 1. F o r any element b outside Z ( G ) , its conjugacy class contains m o r e than one element and C(/3) <  G  . Moreover, since C(/>) divides ]G by Lagrange's t h e o r e m , \C(b)\ = p \ w h e r e 1 < n{b) < n. W e divide the pieces of the class equation into two parts: that coming from the center, and the rest. W e get, this way,
n{b
P" = \G\ = z
+
2) r j j y p * + X
6£Z(G) \^K J\
U
zSy = ^
(b) u{b)<n
2
n(fc)oi
P" 
lb)
n(b)<nP
Clearly, p divides the lefthand side, p", and divides 1 p"~" . T h e net result of this is that p  z, and since z > 1, we have that z is at least p. So since z =  Z ( G )  , there must b e an element a # e in Z ( G ) , which proves the t h e o rem. • This last t h e o r e m has an interesting application, which some readers m a y have seen in solving P r o b l e m 45 of Section 5. This is
104
Groups
2
Ch. 2
T h e o r e m 2.11.5. G is abelian.
If G is a group of order p , where p is a prime, then
Proof. By T h e o r e m 2.11.4, Z ( G ) # (e), so that t h e r e is an element, a, of order p in Z ( G ) . If A = (a), the subgroup generated by a, then A C Z ( G ) , hence A C C(x) for all * G G. Given x G G, x £ A , t h e n C(x) D A and .V G C(x); so  C ( x )  > p, yet C(x) must divide p . T h e n e t result of this is that  C ( x )  = p , so C(x) = G, whence x G Z ( G ) . Since every element of G is in the center of G, G must b e abelian. •
2 2
In the problems to come we shall give many applications of t h e n a t u r e of groups of order p", where p is a prime. The natural attack on virtually all these problems follows the lines of the argument we are about to give. W e choose o n e of a wide possible set of choices t o illustrate this technique.
T h e o r e m 2.11.6. If G is a group of order p", p a prime, t h e n G contains a normal subgroup of order p"~ .
l
Proof. W e p r o c e e d by induction on n. If n = 1, t h e n G is of order p and (e) is the required normal subgroup of o r d e r p ~ = p = 1. Suppose that w e know that for some k every group of order p has a normal subgroup of order p ~ . Let G be of order p ; by T h e o r e m 2.11.4 there exists an element a of order p in Z ( G ) , the center of G. Thus the subgroup A = (a) generated by a is of order p and is n o r m a l in G. Consider T = G/A; r is a group of order \G\I\A\ = p lp = p by T h e o r e m 2.6.3. Since T has order p , we know that T has a n o r m a l subgroup M of order p ~ . Since T is a h o m o m o r p h i c image of G, by t h e Correspondence T h e o r e m ( T h e o r e m 2.7.2) t h e r e is a normal subgroup N in G, N D A, such that N/A = M. B u t then we have
l l a k k l k+1 k+1 k k k 1
pki
k x
= M = \N/A\ =
k
j^j,
that is, p ~ = \N\/p, leading us to \N\ = p . Thus TV is our required normal subgroup in G of order p . This completes the induction and so proves the theorem. •
k
By far the most important application we m a k e of the class equation is the proof of a farreaching t h e o r e m due to Sylow, a Norwegian m a t h e m a t i cian, w h o proved it in 1871. W e already showed this t h e o r e m t o b e true for abelian groups. W e shall now prove it for any finite group. It is impossible t o overstate the importance of Sylow's Theorem in the study of finite groups. W i t h o u t it the subject would n o t get off the ground.
S e c . 11
Conjugacy and S y l o w ' s T h e o r e m (Optional)
105
Theorem 2.11.7 (Sylow's Theorem). Suppose that G is a group of order p"m, where p is a prime and p \ m. T h e n G has a subgroup of order p". Proof. If n = 0, this is trivial. W e therefore assume that n > 1. H e r e , again, we p r o c e e d by induction o n \ G\, assuming the result to b e true for all groups Hsuch that \H\ < \ G\. Suppose that the result is false for G. T h e n , by our induction hypothesis, p" cannot divide \H\ for any subgroup H of G if H # G. In particular, if a <£ Z ( G ) , then C(a) + G, hence p"\\C(a)\. Thus p divides \G\l\G{a)\ = i {C(a))lova£Z(G). Write down the class equation for G following the lines of the argument in T h e o r e m 2.11.4. If z =  Z ( G )  , t h e n z > 1 and
G
p"m
G
=\G\=z+
2
it<£Z(G)
e z ( G ) G
io(C(a)).
But p  i (C(a)) if a Z ( G ) , so 2„ / ( C ( « ) ) . Since p  p " m , we get p ) z. By Cauchy's T h e o r e m there is an element a of order p in Z ( G ) . If A is the subgroup generated by A, then A = p and A < G, since A G Z ( G ) . Consider T = G/A; T =  G  /  A  = p"mlp = p"~ m. Since T <  G  , by our induction hypothesis T has a subgroup M of order p"~ . However, by the Correspondence T h e o r e m t h e r e is a subgroup P of G such that PDA and P/A = M. Therefore,  P  = \M\ \A\ = p"~ p = p" and P is the soughtafter subgroup of G of order p", contradicting our assumption that G had no such subgroup. This completes the induction, and Sylow's Theorem is established. •
l l 1
Actually, Sylow's T h e o r e m consists of three parts, of which we only proved the first. T h e other two are ( a s s u m i n g p " m =  G  , w h e r e p \ m ) : 1. A n y two subgroups of order p " in G are conjugate; that is, if  P  = \Q\ = p" for subgroups P, Q of G, then for some x G G, Q = vides  G  . Since these subgroups of order p" p o p u p all over the place, they are called pSylow subgroups of G. A n abelian group has o n e pSylow subgroup for every p r i m e p dividing its order. This is far from t r u e in the general case. For instance, if G = S , the symmetric group of degree 3, which has o r d e r 6 = 2  3 , t h e r e are three 2Sylow subgroups (of order 2) and one 3Sylow subgroup (or order 3). For those who want to see several proofs of that part of Sylow's T h e o rem which we proved above, and of the other two parts, they might look at the appropriate section of our b o o k Topics in Algebra.
3
x~ Px.
1
2. T h e n u m b e r of subgroups of order p" in G is of the form 1 + kp and di
106
Groups
Ch. 2
PROBLEMS
Easier P r o b l e m s 1. In 5 , t h e symmetric group of degree 3, find all the conjugacy classes, and check the validity of the class equation by determining the orders of t h e centralizers of t h e elements of S .
3 3
2. D o P r o b l e m 1 for G the dihedral group of order 8. 3. If a G G, show that C(x~ ax) 5. If  G  = p
3 2 l
= x~ C(fl)x.
1
4. If cp is an a u t o m o r p h i s m of G, show that C(cp(a)) = cp(C(a)) for a £ G. and  Z ( G )  > p , prove that G is abelian. 6. If P is a /7Sylow subgroup of G and P < G, prove that P is t h e only pSylow subgroup of G. 7. If P < G, P a jDSylow subgroup of G, prove that cp(P) = P for every automorphism cp of G. 8. U s e the class equation to give a proof of Cauchy's T h e o r e m . If TP is a subgroup of G, let N(H) H<G, = {x G G  x ^ / P x = 7/7}. This «oes G N(H), a G if. F o r instance, if then 7V(/7) = G, yet 7 / n e e d not be in the center of G. is a subgroup of G, 77 C 7V(77) and in fact 77 < = x " N(77)x. N(P).
1
/ £ r mea/7 that xa = «x w h e n e v e r x 7>
9. Prove that N(H)
_ x
N(H).
10. Prove that 7V(x Hx)
11. If P is a oSylow subgroup of G, prove that P is a /jSylow subgroup of 7V(P) and is the only oSylow subgroup of that if a~ Pa = P then a G P . 13. Prove that if G is a finite group and 77 is a subgroup of G, then t h e number of distinct subgroups x 7 7 x of G equals cannot be a multiple of p. 15. If T < G, let 73(TV) = [x G G  XA = «x for all a G TV}. Prove that V B (TV) < G. MiddleLevel P r o b l e m s 16. Show that a group of order 36 has a n o r m a l subgroup of o r d e r 3 or 9. (Hint: See Problem 40 of Section 5.) 17. Show that a group of o r d e r 108 has a normal subgroup of o r d e r 9 or 27. 18. If P is a pSylow subgroup of G, show that N(N(P))
1 _ 1 l
12. If P is a pSylow subgroup and a G G is of order p'" for s o m e 777, show
i (N(H)).
G
14. If P is a pSylow subgroup of G, show that the n u m b e r of distinct
x~ Px
l
= 7V(P).
19. If [G\ = p", show that G has a subgroup of o r d e r p " for all 1 < m < n.
Sec.
11
CONJUGACY AND SYLOW'S THEOREM (OPTIONAL)
107
20. If p'" divides G, show that G has a subgroup of order p'". 21. If G = p" and H + G is a subgroup of G, show that N(H) 22. Show that any subgroup of order p"" in
1
J //.
in a group G of order p" is n o r m a l
G.
H a r d e r PROBLEMS 23. Let G b e a group, H a subgroup of G. Define for a, b G G, a ~ b if /3 = /?" ah for some /z G H. Prove that (a) this defines an equivalence relation on G. (B) If [a] is the equivalence class of a, show that if G is a finite group, then [a] has m elements where m is the index of H fl C(a) in H.
l
24. If G is a group, 7f a subgroup of G, define a relation B ~ A for subgroups A, 5 of G by the condition that B = h~ Ah for some h & H. (a) Prove that this defines an equivalence relation on the set of subgroups of G. (B) If G is finite, show that t h e n u m b e r of distinct subgroups equivalent to A equals the index of N(A) n H in H.
l
25. If is a pSylow subgroup of G, let S b e t h e set of all pSylow subgroups of G. For <2 g G S define G^ ~ g if g = a~ Q a with a E. P. Prove, using this relation, that if Q # i , t h e n the n u m b e r of distinct a~ Qa, with a G / \ is a multiple of p.
l 1; 2 2 2 x 3 1
26. Using the result of P r o b l e m 25, show that t h e n u m b e r of pSylow subgroups of G is of the form 1 + kp. (This is t h e third part of Sylow's T h e o rem.) 27. Let P be a pSylow subgroup of G, and Q a n o t h e r one. Suppose that Q + x~ Px for any x G G. L e t S be t h e set of all y~ Qy, as _ runs over G. y For g  , <2 G S define g ~ Q if £> = a ^ g j a , where a G P. (a) Show that this implies that t h e n u m b e r of distinct y~ Qy is a multiple of p. (B) Using the result of P r o b l e m 14, show that the result of Part (a) cannot hold. (C) Prove from this that given any two /?Sylow subgroups P and Q of G, then Q = x~ Px for some x G G. (This is the second part of Sylow's T h e o r e m . )
1 1 2 : 2 2 l 1
28. If II is a subgroup of G of order p'" show that H is contained in some pSylow subgroup of G. 29. If P is a pSylow subgroup of G and a, b G Z(/ ) prove that they are already conjugate in N(P).
J
are conjugate in G,
1.
PRELIMINARIES
Let us recall a t h e o r e m p r o v e d in Chapter 2 for abstract groups. This result, k n o w n as Cayley's Theorem ( T h e o r e m 2.5.1), asserts t h a t any group G is isomorphic to a subgroup of A (S), the set of 11 mappings of t h e set S onto itself, for some suitable S. In fact, in t h e proof we gave we used for S t h e group G itself viewed merely as a set. Historically, groups arose this way first, long before the notion of an abstract group was defined. W e find in the work of Lagrange, Abel, Galois, and others, results on groups of permutations proved in t h e late eighteenth and early nineteenth centuries. Yet it was not until t h e midnineteenth century that Cayley m o r e or less introduced the abstract concept of a group. Since the structure of isomorphic groups is the same, Cayley's T h e o r e m points out a certain universal character for the groups A (S). If we knew t h e structure of all subgroups of A (S) for any set S, we would k n o w the structure of all groups. This is m u c h too much to expect. Nevertheless, o n e could try t o exploit this embedding of an arbitrary group G isomorphically into some A(S). This has t h e advantage of transforming G as an abstract system into something more concrete, namely a set of nice mappings of some set onto itself. W e shall n o t b e concerned with the subgroups of A (S) for an arbitrary set S. If S is infinite, A (S) is a very wild and complicated object. Even if S is finite, the complete n a t u r e of A (S) is virtually impossible to determine.
1flR
Sec.
1
Preliminaries
109
In this chapter we consider only A (S) for S a finite set. Recall that if S has n elements, then we call A (S) the symmetric group of degree n, and denote it by S„. T h e elements of S„ are called permutations; we shall denote them by lowercase G r e e k letters. Since we multiplied two elements <x, T E A(S) by the rule (ar)(s) = (T(T(S)) this will have the effect that when we introduce the appropriate symbols to represent the elements of S„, these symbols, or permutations, will multiply from right to left. If the readers look at some other book on algebra, they should m a k e , s u r e which way the permutations are being multiplied: right to left or left to right. V e r y often, algebraists multiply p e r m u t a tions from left to right. To b e consistent with our definition of the composition of elements in 5„, we do it from right to left. By Cayley's T h e o r e m we know that if G is a finite group of o r d e r n, then G is isomorphic to a subgroup of S„ and 5„ has n\ elements. Speaking loosely, we usually say that G is a subgroup of S„. Since n is so much smaller than n\ for n even modestly large, our group occupies only a tiny little corner in S„. It would be desirable to embed G in an S„ for n as small as possible. For certain classes of finite groups this is achievable in a particularly nice way. Let S be a finite set having n elements; we might as well suppose that S = {x , x ,..., x„}. Given the p e r m u t a t i o n cr E S„ = A(S), then cr(x ) E S for k = 1, 2 , . . . , n, so cr(x ) = x for some i ,l ^ i £ n. Because a is 11, if j ¥= k, then x = cr(x ) i= cr(x ) = x . Therefore, the numbers i ,..., /„ are merely the numbers 1,2,... ,n shuffled about in some order. Clearly, the action of cr on S is d e t e r m i n e d by what cr does to the subscript of x , so the symbol " x " is really excess baggage and, as such, can be discarded. In short, we may assume that S = [1, 2,..., n}. Let's recall what is m e a n t by the product of two elements of A (S). If <r, r E A(S), then we defined err by (OTT){S) = CT(T(S)) for every s €L S. We showed in Section 4 of Chapter 1 that A (S) satisfied four properties that we used later as the model to define the notion of an abstract group. Thus S„, in particular, is a group relative to the product of mappings. O u r first need is some handy way of denoting a permutation, that is, an element a in S„. O n e clear way is to m a k e a table of what a does to each element of S. This might be called the graph of a. W e did this earlier, writing out cr, say cr E S , in the fashion: cr: .Vj — x , x — x , x —> X ] . But this is > > cumbersome and space consuming. W e certainly can make it more compact
{ 2 k k t k k f ; k ik 2 ; 3 2 2 3 3
the second row is the image u n d e r cr of the n u m b e r in the first row directly above it. T h e r e is nothing holy about 3 in all this; it works equally well for any n.
110
The Symmetric Group
Ch. 3
If cr G 5„ a n d cr(l) = r' , <x(2) = i , • •., cr(«) = /„, we use the symbol
x 2
) to represent <x and we write a = ( } ? " ]. N o t e i) \h h • • • h,) that it is not necessary t o write the first row in the usual order 1 2 • • • n; any way we write the first row, as long as we carry t h e z's along accordingly, we still have a. F o r instance, in the example in S cited, 'l 2 3\ /3 1 2\ / 2 1 3^ °" \2 3 l j U 2 3/ \3 2 1
'i
'2 • ' •
n 3
f
?
W
1 If we know a = [ . symbol for <r over to get a 1 2 2 3
=
2 .
•••
w\ . , what is c r ? It is easy, just flip the
1
= ^
=
h


_ _ _ ^ j . (Prove!) In our example
1
" "
l„
3\ _! / 2 3 1\ 1 ' " " 1 2 3
\3
2 3 1 2 2 ^
The •
'
identity n'
n
element—
fl which we shall write as e—is merely e = ( ^
) H o w does the product in S„ translate in terms of these symbols? Since err means: "First apply T and to the result of this apply cr," in forming the product of the symbols for a and T we look at the n u m b e r k in the first row of T and see what n u m b e r i is directly below k in the second row of T. W e then look at the spot i in the first row of cr and see what is directly below it in the second row of cr. This is the image of k u n d e r ar. W e t h e n run t h r o u g h k = 1, 2 , . . . , n and get t h e symbol for CTT. W e just do this visually. W e illustrate this with two permutations 1 2 3 4 5\ , /1 2 3 4 5 2 3 1 5 4 3 4 5 1 2
k k a n d T =
in
1 2 3 4 5 1 5 4 2 3 E v e n the e c o n o m y achieved this way is not enough. After all, the first row is always 1 2 • • • n, so we could dispense with it, and write cr = T h e n <JT 2 l
(} ? " ) as (z'j, i , • • •, i„). This is fine, but in the next section we \h h ' ' ' n) shall find a better and briefer way of representing permutations.
PROBLEMS
1. Find the p r o d u c t s : 1 16 2 4 3 5 4 2 5 1 6\/l 3A2 2 3 3 4 4 5 5 6 6 1
( a )
Sec. 2
Cycle D e c o m p o s i t i o n
111
( b )
/l ^2 (1 ^4
2 1 2 1
3 3 3 3
4 4 4 2
5\ 1 5j\3
1 1
2 2
3 1 2 1
4 4 3 3
5 5 4 4 5\/l 5J\4 2 1 3 3 4 2 5^ 5
/
( c )
5V / 5/ ^2
2. Evaluate all the powers of each p e r m u t a t i o n cr (i.e., find cr* for all k). '1 3 4 5 6 3 4 5 6 1 (a) v
2
1
2
1
(b)
2 2
4
3 3 3 5 1
4 4 4
5 6 5 1
6 5 6 3
7 7
2
(c)
'l ,6
3. Prove that
2 2 •••
n\
1
/z 1
t
2
z
2
• • • z', ••• n
4. Find the o r d e r of each element in P r o b l e m 2. 5. Find the order of the products you o b t a i n e d in P r o b l e m 1.
2. CYCLE D E C O M P O S I T I O N
We continue t h e process of simplifying the notation used to represent a given permutation. I n doing so, we get something m o r e t h a n just a new symbol; we get a device to decompose any p e r m u t a t i o n as a product of particularly nice permutations. Definition. Let i , i , . . . , i be k distinct integers in S = {1, 2,..., zz). The symbol (z'j i • • • i ) will r e p r e s e n t t h e p e r m u t a t i o n cr G S„, where o(i ) = i , cr{i ) = z , . . . , cr(z .) = for / < k, a(i ) = i and CJ(S) = s for any s G 5 if s is different from z i , • • •, i •
x 2 k 2 k L 2 2 3 ; k u l5 2 k
Thus,
in
S ,
7
the ^
permutation
C C a
(1
3 i°
n
5 °f
t
4)
n e
is t h e f°
r m
permutation h
''' 'A)
(325I467)'
(i/ci i
k
^
a
P
e r m u t a t
(*i
2
a kcycle. F o r the special case k = 2, the permutation (z, position. N o t e that if cr = (z\ i\ ii 5 ''' 42)) 1 z'
a n 2
z ) is called a rrawsk
•••
s o o n
i ), then cr is also (i
k
z,
z'
2
•••
i \),
k
d
 (Prove!) For example, 4 1
(1
3
4) = (4
3
5) = (5
3) = (3
5
4
1).
Two cycles, say a /ocycle and an mcycle, are said to be disjoint cycles if
112
The Symmetric
Group
Ch, 3
they have no integer in c o m m o n . W h e n c e (1 3 5) and (4 2 6 7) in S are disjoint cycles. Given two disjoint cycles in S„, we claim that they c o m m u t e . W e leav< t h e proof of this to the reader, with the suggestion that if cr, r are disjoin cycles, the reader should verify that (O"T)(7) = (TCT)(Z) for every i G S = [1,2,... ,n). We state this result as L e m m a 3.2.1. If cr, T G S„ arc disjoint cycles, t h e n ar = rcr.
Let's consider a particular /ocycle a = (1 2 • • • k) in S„. Clearly cr(l) = 2 by the definition given above; h o w is 3 related t o 1? Since cr(2) = 3 we have c r ( l ) = cr(2) = 3. Continuing, we see that cr '(l) = /' + 1 foi j < k  1, while a (1) = 1. In fact, we see that cr = e, w h e r e e is t h e identity element in S . T h e r e are two things to b e concluded from t h e p a r a g r a p h above.
2 ; k k n
1. T h e order of a kcycle, as an element of S„, is k. (Prove!) 2. If cr = (i i • • • i ) is a A>cycle, then t h e orbit of i u n d e r a (see P r o b l e m 27 in Section 4 of Chapter 1) is {i,, i , • • •, i }. So we can see that the /ccycle cr = (/j i • • • i ) is
t 2 k x 2 k 2 k
<r=(i
v
oih)
a (h)
2
•••
er* ^)).
1
Given any p e r m u t a t i o n Tin S„, for / G {1, 2 , . . . , « } , consider the orbh of i u n d e r r. W e have that this orbit is [i, T ( I ) , r (i),. .., T (;')}, where T ( Z ' ) = i and s is t h e smallest positive integer with this p r o p e r t y . Consider the 5cycle (i r(z') T ( Z ) ••• r ( / ) ) ; w e call it t h e cycle of T determined by i. W e take a specific example and find all its cycles. Let
2 s _ 1
S
2
s _ 1
1 3
2 9
3 4 5 6 7 8 4 1 5 6 2 7
9\ 8/
what is t h e cycle of T d e t e r m i n e d by 1? W e claim that it is (1 3 4). Why? T takes 1 into 3, 3 into 4 and 4 into 1, and since T(1) = 3, r ( l ) = T(3) = 4, T ( 1 ) = T(4) = 1. W e can get this visually by weaving through
2 3
with t h e thin p a t h . W h a t is t h e cycle of T d e t e r m i n e d by 2? Weaving through
Sec. 2
Cycle D e c o m p o s i t i o n
113
with the thin path, we see that t h e cycle of T determined by 2 is (2 9 8 7). T h e cycles of T d e t e r m i n e d by 5 and 6 are (5) and (6), respectively, since 5 and 6 are left fixed by T. S O t h e cycles of T are (1 3 4), (2 9 8 7), (5), and (6). Therefore we have that T = (1 3 4)(2 9 8 7) (5)(6), w h e r e we view these cycles—as defined above—as permutations in S because every integer in S = {1, 2 , . . . , 9} appears in o n e and only one cycle, and the image of any i u n d e r r is read off from the cycle in which it appears.
9
T h e r e is nothing special about t h e permutation T above that m a d e the argument we gave go through. T h e s a m e argument would hold for any permutation in S„ for any n. W e leave t h e formal writing down of the proof to the reader. T h e o r e m 3.2.2. cycles. In writing a p e r m u t a t i o n cr as a p r o d u c t of disjoint cycles, we omit all Icycles; that is, we ignore t h e z's such t h a t cr(/) = i. T h u s we write cr = (1 2 3)(4 5)(6)(7) simply as cr = (1 2 3)(4 5). In other words, writing eras a p r o d u c t of /ccycles, with k > 1, we assume that cr leaves fixed any integer not p r e s e n t in any of t h e cycles. T h u s in the group S t h e p e r m u t a tion r = (1 5 6)(2 3 9 8 7) leaves fixed 4, 10, and 11.
n
Every permutation in S„ is the product of disjoint
L e m m a 3.2.3. If T in S„ is a fccycle, then the order of 7 is k; that is, r = e and 7' i= e for 0 < j < k.
k
Consider t h e p e r m u t a t i o n 7 = (1 2)(3 4 5 6)(7 8 9) in S . What is its o r d e r ? Since the disjoint cycles (1 2), (3 4 5 6), (7 8 9) commute, T ' " = (1 2 ) ( 3 4 5 6)'"(7 8 9)"'; in order that r'" = e we need (1 2)'" = e, (3 4 5 6)'" = e, (7 8 9)"' = e. (Prove!) T o have (7 8 9)" = e, we must have 3  m, since (7 8 9) is of order 3; to have (3 4 5 6)'" = e, we must h a v e 4  m, because (3 4 5 6) is of o r d e r 4, and to have (1 2)"' = e, we must have 2  m, because (1 2) is of order 2. This tells us that m must b e divisible by 12. O n the other hand,
9 ,n !
T
1 2
= (1
2) (3
12
4
5
6) (7
12
8
9)
1 2
= e.
So T is of order 12. H e r e , again, the special properties of 7 d o not enter the. picture. W h a t we did for 7 works for any p e r m u t a t i o n . T o formulate this properly, recall that t h e least c o m m o n multiple of m and n is the smallest positive integer v which is divisible by m and by n. (See P r o b l e m 7, Chapter 1, Section 5.) T h e n we have
114
The Symmetric
Group
Ch. 3
T h e o r e m 3.2.4. Let cr E S„ have its cycle decomposition into disjoint cycles of length m , , m , • • •, m . T h e n the order of cr is the least c o m m o r multiple of m m , • • •, ni .
2 k u 2 k
Proof. Let cr = T T • • • r , where t h e T, are disjoint cycles of length m Since the r, are disjoint cycles, T,T, = TJT,; therefore if M is the least common multiple of m m ,. . . , m , then cr' ' = (T T • • • r) = T^T • • •r = e (since rf~ e because T, is of o r d e r m, and m \ M). Therefore, the o r d e r ol cr is at most M. O n the other hand, if cr = e, then T^T • • • r = e. Thb forces each T ' / = e, (prove!) because r, are disjoint p e r m u t a t i o n s , so m \ A since T, is of o r d e r m,. T h u s N is divisible by the least c o m m o n multiple oi m , m ,. . ., m , so M  7Y. Consequently, we see that <J is of o r d e r M as claimed in the theorem. •
X 2
k
t
1
M
M
b
2
k
X
2
k
2
k
t
N
2
k
r
t
l
2
k
N o t e that t h e disjointness of the cycles in the t h e o r e m is imperative F o r instance, (1 2) and (1 3), which are not disjoint, are each of order 2 but their product (1 2)(1 3) = (1 3 2) is of order 3. Let's consider T h e o r e m 3.2.4 in the context of a card shuffle. Suppose that we shuffle a deck of 13 cards in such a way that t h e t o p card is put intc the position of t h e 3rd card, the second in that of the 4 t h , . . . , t h e ith into the i + 2 position, working m o d 13. A s a permutation, cr, of 1, 2 , . . . , 13, the shuffle becomes 1 3 2 4 3 5 4 6 5 7 6 8 7 9 8 10 9 11 10 12 11 13 12 1 13 2
and ois merely the 13cycle (1 3 5 7 9 11 13 2 4 6 8 10 12), so a is of order 13. H o w m a n y times must we r e p e a t this shuffle to get the cards back to their original order? T h e answer is merely the order of cr, that is, 13. So it takes 13 repeats of the shuffle to get the cards back to their original order. Let's give a twist to the shuffle above. Suppose that we shuffle the cards as follows. First t a k e the t o p card and put it into t h e secondtolast place, and then follow it by the shuffle given above. H o w m a n y r e p e a t s are now n e e d e d to get the cards back to their original o r d e r ? T h e first operation is the shuffle given by the p e r m u t a t i o n T = (1 12 11 10 9 8 7 6 5 4 3 2) followed by cr above. So we must c o m p u t e err and find its order. But o r = (1 X(l = (1)(2 3 5 12 3 7 11 4 5 9 10 6 11 9 7 13 8 8 2 7 9 4 6 10 6 5 8 4 11 10 3 12 2) 13), 12)
Sec. 2
Cycle D e c o m p o s i t i o n
115
so is of order 12. So it would take 12 repeats of the shuffle to get back to the original order. Can you find a shuffle of the 13 cards that would require 42 repeats? Or 20 repeats? What shuffle would require the greatest n u m b e r of repeats, and what would this number be? We return to the general discussion. Consider the permutation (1 2 3); we see that ( 1 2 3) = (1 3)(1 2). W e can also see that (1 2 3) = (2 3)(1 3). So two things are evident. First, we can write (1 2 3) as the product of two transpositions, and in at least two distinct ways. Given the &cycle (i i • • • /*), then (i i ••• i) = (/[ ' ' ' 0 i h)> every /tcycle is a product of k — 1 transposi{ 2 i 2 k s o
tions (if k > 1) and this can be d o n e in several ways, so not in a unique way. Because every permutation is the product of disjoint cycles and every cycle is a product of transpositions we h a v e T h e o r e m 3.2.5. tions. This t h e o r e m is really not surprising for it says, after all, nothing m o r e or less than that any p e r m u t a t i o n can b e effected by carrying out a series of interchanges of two objects at a time. We saw that there is a lack of uniqueness in representing a given permutation as a product of transpositions. But, as we shall see in Section 3, some aspects of this decomposition are indeed unique. As a final word of this section, we would like to point out the convenience of cycle notation. W h e n we r e p r e s e n t elements of a permutation group as products of disjoint cycles, many things become transparent—for example, the order of the p e r m u t a t i o n is visible at a glance. T o illustrate this point, we now give a few examples of certain geometric groups, which are in fact permutation groups that have already appeared in Chapter 2 u n d e r different guises.
Examples
Every p e r m u t a t i o n in S„ is the product of transposi
1. Informally, a motion of a geometric figure is a permutation of its vertices that can be realized by a rigid m o t i o n in space. For example, there are eight motions of a square, whose vertices are n u m b e r e d 1, 2, 3, 4 as below. 4 3
1
2
116
The Symmetric Group
Ch. 3
a = (13) is the reflection about the axis of symmetry joining vertices 2 and 4 in the original position, 3 = (1234) is the counterclockwise rotation by 90°, p = (13)(24) is the counterclockwise rotation by 180°, P = (1432) is the counterclockwise rotation by 270°, a(i = (12)(34) is the reflection in the vertical axis of symmetry, a p = (24) is the reflection in the other diagonal axis, ap
3 2 3 2
= (14)(23) is the reflection in the horizontal axis, and, of course
4
a = (3 = (1) is the " m o t i o n " that leaves the vertices unchanged. W e also have the relation (3a = a ( 3 . These motions, or symmetries of a square, form a subgroup of S which is called the octic group, or the dihedral group of o r d e r 8. This group (or, strictly speaking, a group isomorphic to it) was introduced in E x a m p l e 9 of Section 2.1 without mention of permutations.
4 3
2
2. There are only four symmetries of a n o n  s q u a r e rectangle:
4 3
1
2
the reflections in the two axes of symmetry, rotation by 180° and the identity. These motions can be identified with permutations (1), (14)(23), (12)(34), (13)(24), and form a subgroup of the group obtained in E x a m p l e 1. This subgroup is often called Klein's 4group. 3. W e leave it to the reader to verify that the group of all motions of an equilateral triangle is the full symmetric group S .
3
3
1
2
4. T h e motions of a regular hexagon form the dihedral group of order 12, generated by the p e r m u t a t i o n s a = (15)(24), corresponding to a reflection
Sec. 2
Cycle Decomposition
117
about one of t h e axes of symmetry, and p = (123456), corresponding to the counterclockwise rotation by 60°. 5 4
1
2
In general, t h e dihedral group of o r d e r 2n, which was first introduced in Example 10 of Section 2.1, can b e interpreted as the group of symmetries of a regular /7gon (a polygon with n edges of equal length).
PROBLEMS
Easier P r o b l e m s 1. Show that if cr, T are two disjoint cycles, t h e n en = TIT. 2. Find the cycle decomposition and order. (a) 1 3 1 7 1 7 2 1 2 6 2 6 3 4 3 5 3 5 4 2 4 4 4 3 5 7 5 3 5 4 6 6 6 2 6 2 7 9 7 1 7 W 1 2 3 4 5 6 7 l/\2 3 1 5 6 7 4/' 8 8 9^ 5
(b) (c)
3 . Express as t h e product of disjoint cycles and find t h e order. (a) (1 2 3 5 7)(2 4 7 6). (b) (1 2)(1 3)(1 4). (c) (1 2 3 4 5)(1 2 3 4 6)(1 2 3 4 7). (d) (1 2 3)(1 3 2). (e) (1 2 3 ) ( 3 5 7 9)(1 2 3 ) . (f) (1 2 3 4 5) .
_ 1
3
4. Give a complete proof of T h e o r e m 3.2.2. 5. Show that a /ccycle has o r d e r k. 6. Find a shuffle of a deck of 13 cards that requires 42 repeats to r e t u r n the cards to their original order.
118
The Symmetric Group
Ch. 3
7. D o Problem 6 for a shuffle requiring 20 repeats. 8. Express t h e p e r m u t a t i o n s in P r o b l e m 3 as the product of transpositions. 9. Given t h e two transpositions (1 such that cr(l 2)a~ = (1 3).
l
2) a n d (1
3), find a p e r m u t a t i o n cr 2)cr~ = (1 cr(l 2
1
10. Prove that there is n o p e r m u t a t i o n crsuch that c r ( l 11. Prove (4 (1 5 2 that 6). there is a p e r m u t a t i o n cr such that
2 3)cr
_1
3). = =
12. Prove that t h e r e is n o p e r m u t a t i o n 4)(5 6 7).
cr such that c r ( l 2
3)cr~
l
MiddleLevel P r o b l e m s 13. Prove that (1 cycles. 14. Prove that for any p e r m u t a t i o n cr, c n  c r position. 15. Show that if r is a /ccycle, then tion cr.
CTTO^
1
2) cannot b e written as t h e p r o d u c t of disjoint 31
is a transposition if r is a trans
is also a A;cycle, for any p e r m u t a 
16. L e t $ b e an a u t o m o r p h i s m of S . Show that there is an element a G S such that <E>(T) = O~~ TCT for every T G S .
3
1
3
3
17. L e t (1 2) and (1 2 3 ••• n) be in S„. Show that a n y subgroup of S„ that contains both of these must b e all of S„ (so these two p e r m u t a t i o n s generate S„). 18. If T , a n d T are t w o transpositions, show that product of 3cycles (not necessarily disjoint).
2
TT
X
2
can b e expressed as t h e J= e, t h e identity
19. Prove that if T element of S„. 20. If
T,
T
1 ;
T , and T are transpositions, then T T T
2 3
1 2
3
T
2
are distinct transpositions, show that
TT
V
2
is of o r d e r 2 o r 3.
2 1 . If cr, T a r e two p e r m u t a t i o n s that disturb n o c o m m o n element and err = e, prove that cr = T = e. 22. Find an algorithm for finding crTcr" for any p e r m u t a t i o n s cr, T of 5 „ . 23. L e t cr, T be t w o p e r m u t a t i o n s such that they both have decompositions into disjoint cycles of cycles of lengths m , m ,. . . , m . ( W e say that they have similar decompositions into disjoint cycles.) Prove that for some p e r m u t a t i o n p , r = p c r p .
x 2 k  1 1
24. Find t h e conjugacy class in S„ of (1 centralizer of (1 2 • • • n) in S„? 25. D o Problem 24 for a = (1 2) (3 4).
2
• • • n). W h a t is t h e o r d e r of t h e
Sec.
3
Odd
and Even Permutations
119
3 . O D D A N D EVEN PERMUTATIONS
We noticed in Section 2 that although every permutation is the product of transpositions, this decomposition is not unique. We did comment, however, that certain aspects of this kind of decomposition are unique. W e go into this now. Let's consider the special case of S , for here we can see everything explicitly. Let f(x x , x ) = (x  x ) ( x ,  x ) ( x ~ x ) be an expression in the three variables X , , x , x . W e let S act on f(x) = f(x x , x ) as follows. If oG S , t h e n
3 u 2 3 x 2 3 2 3 2 3 3 u 2 3 3
<r*(f(x))
= (Vl) ^(2))(^(l) "
_
^(3))(*(2)  ^(3))C T
3
We consider what cr* does to f(x) for a few of the o's in S . Consider cr = (1 2 ) . T h e n <x(l) = 2, o(2) = 1, and a(3) = 3, so that o*(.f(x)) = (Vt)
= (X
2
~ x )(x^
cr(2)
V)
 x )(x\
a(3) 3
r(2}

x
(r(3)
)

x) ( x  x )
2 2 1
(Xj
 x)
3 2
=  ( x ,  x )(x
 x )(x
3

x )
3
=
/(*)•
So a'* coming from cr = (1 2) changes the sign of f(x). Let's look at t h e action of another element, T = (1 2 3), of S o n / ( x ) . Then
3
T*(f(x))
= (x
2
T ( 1 )
 x
3
T ( 2 )
)(x
T ( l )
 X
3
T ( 3 )
)(X
t ( 2 )

x
T ( 3 )
)
= (x  x ) ( x = (x, X)
2
2
 x )(x
x 3
2
x^
3
(Xj
 x ) (x  x )
so f* coming from T = (1 2 3) leaves / ( x ) unchanged. W h a t about the other permutations in S ; how do they a f f e c t / ( x ) ? Of course, t h e identity element e induces a m a p e* on / ( x ) which does not change / ( x ) at all. What does T , for T above, do t o / ( x ) ? Since r*f(x) = f(x), we immediately see that
3 2
(T ) * (/(x)) = (xv  x v ) ( x > = / ( x ) . (Prove!)
(1) ( 2 )
2
( 1 )
 x> )(xv
( 3 )
( 2 )

x> )
(3)
Now consider err = (1 2)(1 2 3) = (2 3); since T leaves / ( x ) alone and cr changes the sign of f(x), o r m u s t change the sign of f (x). Similarly, (1 3) changes the sign of f(x). We have accounted for the action of every element of S on f(x).
3
120
The Symmetric Group
Ch. 3
Suppose that p E S is a product p = T T • • • r of transpositions T ..., T ; then p acting on / ( x ) will change t h e sign of f(x) k times, since each T, changes the sign of fix). So p*(f(x)) = (~l) f(x). If p = ojcr, • • • o>. where < j . . . , a, a r e transpositions, by t h e same reasoning, p * ( / ( x ) ) = (  1 ) 7 W  Therefore, = (~1)7W. whence (  1 ) ' = (  1 ) * . This tells us that t and k have t h e same parity; that is, if t is odd, t h e n /<; must be odd, and if t is even, then k must b e even. This suggests that although the decomposition of a given p e r m u t a t i o n a as a product of transposition is not unique, the parity of the number of transpositions in such a decomposition of a might be unique. W e strive for this goal now, suggesting t o readers that they carry out the argument that we d o for arbitrary n for the special case n = 4. A s we did above, define f(x) = f(x . . . , x„) to b e
3
X 2
k
U
K
k
1;
h
fix)
= ( x  x ) • • • (*!  x„) {x  x ) • • • (x  x„) • • • (,*„_!  x„)
x 2 2 3 2
= EI ( ,  xj), *where in this product i takes o n all values from 1 t o n — 1 inclusive, and j all those from 2 to n inclusive. If <x E 5„, define cr* on f(x) by cx*(/(x)) = If cr, T E S„, then
ri(^(o
Kj = a*

w

(crr)*(/(x)) =
n (x „
(
m
 x
(!TT)(j)
)
(n ( x
T(/)

x )j
T(/)
= o* ( T *
(x,.  )Jj
X]
= cr*(r*(/(x))) = (cr*r*)(/(x))
So (err)* = cr*T* when applied t o / ( x ) . W h a t does a transposition r d o to f{x)l W e claim that T * ( / ( X ) ) = —fix). T o prove this, assuming that T = (/ where r < j , w e count u p t h e number of (x — x ), with u < v, which get transformed into an (x„ — x ) with a > b. This h a p p e n s for (x„ — x ) if i < u < j , for (x, — x„) if i < v < j , and finally, for (xj — x ). E a c h of these leads to a change of sign on / ( x ) and since there a r e 2(j  z' — 1) + 1 such, that is, an o d d n u m b e r of them, w e get a n odd n u m b e r of changes of sign o n fix) when acted o n by T * . Thus T * ( / ( X ) ) = ~fix). Therefore, o u r claim that r * ( / ( x ) ) = fix) for every transposition r is substantiated.
u v b ; ;
Sec. 3
Odd and Even Permutations
121
If cr is any permutation in S„ and cr — T T • • • r , where T , T , . . . , r are transpositions, then cr* = (T T ' ' ' k)* * * • • • T  as acting on / ( x ) , and since each T * ( / ( X ) ) =  / ( x ) , we see that a*(f(x)) = (~l) f(x). Similarly, if cr = £i£ " " ' £r> where £ £ , . . . , £f transpositions, then (j*(/(x)) = ( — l ) / ( x ) . Comparing these two evaluations of cr*(/(x)), we conclude that (  1 ) * = (  1 ) ' . So these two decompositions of cr as the product of transpositions are of the same parity. Thus any permutation is either the product of an odd number of transpositions or the product of cm even number of transpositions, and no product of an even number of transpositions zan equal a product of an odd number of transpositions.
] 2 k { 2 r = T T
X 2
k
k
a r e
2
1 ;
2
f
This suggests the following Definition. T h e p e r m u t a t i o n cr E S is an odd permutation if cr is the product of an odd n u m b e r of transpositions, and is an even permutation if a is the product of an even n u m b e r of transpositions. What we have proved above is
n
T h e o r e m 3.3.1. A p e r m u t a t i o n in S„ is either an odd or an even permutation, but cannot be both. With T h e o r e m 3.3.1 b e h i n d us we can deduce a n u m b e r of its consequences. Let A„ b e the set of all even permutations; if cr, T G A„, t h e n we immediately have that ar G A„. Since A„ is thus a finite closed subset of t h e (finite) group S„,A„ is a subgroup of S„, by L e m m a 2.3.2. A„ is called the alternating group of degree n. We can show that A„ is a subgroup of S„ in another way. We already saw that A„ is closed under the p r o d u c t of S„, so to know that A„ is a subgroup of S„ we merely need show that cr G 5„ implies that c r G S„. F o r any permutation cxwe claim that crand a~ are of the same parity. Why? Well, if a = T T • • • r , where the r are transpositions, then
1 ]
X 2
k
(
O"
1
=
(7iT
2
• • • T
k
y
l
=
T^T/T ! • • •
1
T
L 2
T
L X
=
T T —\
K K
• ' • T 7, ,
2
since rf = r,. Therefore, we see that the parity of a and a' is (l) , they are of equal parity. This certainly shows that a G A„ forces <r G whence A„ is a subgroup of S„. But it shows a little more, namely that A„ is a normal subgroup of For suppose that cr G A„ and p G 5„. W h a t is the parity of p~ opl By
1 l
1
1
k
so A,
n
S„. the
122
The Symmetric Group
l
Ch. 3
l
above, p and p~ are of the same parity and cr is an even permutation so p~ ap is an even permutation, hence is in A„. Thus A„ is a normal subgroup of S„. We summarize what we have done in T h e o r e m 3.3.2. subgroup of S„. A„, the alternating group of degree n, is a normal
W e look at this in yet another way. F r o m the very definitions involved we have the following simple rules for the product of p e r m u t a t i o n s : 1. T h e product of two even permutations is even. 2. T h e product of two odd permutations is even. 3. T h e product of an even p e r m u t a t i o n by an odd one (or of an odd one by an even one) is odd. If cr is an even p e r m u t a t i o n , let 8 (a) = 1, and if cr is an odd p e r m u t a tion, let 0(cr) = — 1 . T h e foregoing rules about products translate into 8(OT) = 9(O)9(T), SO 9 is a homomorphism of S„ o n t o the group E = {1, —1} of order 2 u n d e r multiplication. W h a t is the kernel, N, of 91 By the very definition of A„ we see that N = A„. So by the First H o m o m o r p h i s m T h e o r e m , E — S„/A„. T h u s 2 = \E\ = \S /A„\ =  5 „  /  A „  , if n > 1. This gives us that
n
Therefore, T h e o r e m 3.3.3. Corollary. permutations. F o r n > 1, A is a normal subgroup of S„ of order §7;!.
n
For n > 1, S„ contains \n\ even p e r m u t a t i o n s and \n\ odd
A final few words about the proof of T h e o r e m 3.3.1 before we close this section. M a n y different proofs of T h e o r e m 3.3.1 are known. Quite frankly, we do not particularly like any of them. Some involve what might be called a "collection process," where one tries to show that e cannot be written as the product of an o d d n u m b e r of transpositions by assuming that it is such a shortest product, and by the appropriate finagling with this product, shortening it to get a contradiction. O t h e r proofs use other devices. T h e proof we gave exploits the gimmick of the function f(x), which, in some sense, is extraneous to the whole affair. However, the proof given is probably the most transparent of t h e m all, which is why we used it. Finally, the group A„, for n > 5, is an extremely interesting group. W e
Sec. 3
Odd and Even Permutations
123
shall show in Chapter 6 that the only normal subgroups of A,„ for ». > 5, are (e) and A„ itself. A group with this property is called a simple group (not to be confused with an easy group). T h e abelian finite simple groups are merely the groups of prime order. T h e A„ for n > 5 provide us with an infinite family of nonabelian finite simple groups. There are other infinite families of finite simple groups. In the last 20 years or so the heroic efforts of algebraists have determined all finite simple groups. T h e determination of these simple groups runs about 10,000 printed pages. Interestingly enough, any nonabelian finite simple group must have even order.
PROBLEMS
Easier P r o b l e m s 1. Find the parity of each permutation.
( a )
/I ^2
2 3 4 5 6 7 8 4 5 1 3 7 8 9 2 3 4 2 3 4 2)(1 2 5 5
9\ 6
(b) (1 (c) (1 (d) (1
6)(7 8 9). 6)(1 2 3 4 5 7). 3)(4 5)(5 6 8)(1 7 9).
2. If cr is a fccycle, show that cr is an odd p e r m u t a t i o n if k is even, and is an even permutation if k is odd. 3. Prove that cr and 4. If m < n, we 1 , 2 , . . . , m,.. Prove t h a t the element of S„,
T (XT,
_ 1
for any a, r G S„, are of the same parity.
can consider S,„ C 5„ by viewing cr G S„ as acting on ., n as it did on 1, 2 , . . . , m a n d cr leaves j > m fixed. parity of a p e r m u t a t i o n in S„,, when viewed this way as an does not change.
5. Suppose you are told that the p e r m u t a t i o n 1 3
9
2 3 4 1 2
5
6 7
7 8
8 9
9\ 6)
in 5 , where the images of 5 and 4 have been lost, is an even permutation. W h a t must the images of 5 a n d 4 b e ? MiddleLevel Problems 6. If n > 3, show that every element in A„ is a product of 3cycles. 7. Show that every element in A„ is a product of ncycles. 8. Find a normal subgroup in A of order 4.
4
124
The Symmetric Group
Ch. 3
H a i d e r Problems (In fact, very h a r d ) 9. If n 2 5 and (e) N C A„ is a normal subgroup of A„, show that 7Y must contain a 3cycle. 10. Using the result of Problem 9, show that if n ' 5, the only normal subgroups of A„ are (e) and A„ itself. (Thus the groups A„ for n. > 5 give us an infinite family of nonabelian finite simple groups.)
1. DEFINITIONS A N D EXAMPLES
So far in our study of abstract algebra, we have b e e n introduced to one kind of abstract system, which plays a central role in the algebra of today. T h a t was the notion of a group. Because a group is an algebraic system with only one operation, and because a group n e e d not satisfy the rule ab = ba, it ran somewhat counter to our prior experience in algebra. W e were used t o systems w h e r e you could b o t h add and multiply elements and where the elements did satisfy the commutative law of multiplication ab = ba. Furthermore, these systems of our acquaintance usually came from sets of numbers—integers, rational, real, and for some, complex. T h e n e x t algebraic object w e shall consider is a ring. I n m a n y ways this system will b e m o r e r e m i n i s c e n t of w h a t we h a d previously k n o w n t h a n w e r e g r o u p s . F o r o n e thing rings will b e e n d o w e d with addition and multiplication, and t h e s e will b e subjected t o m a n y of the familiar rules we all k n o w from arithmetic. O n t h e o t h e r h a n d , rings n e e d not c o m e from o u r u s u a l n u m b e r systems, a n d , in fact, usually have little to d o with t h e s e familiar ones. A l t h o u g h m a n y of t h e formal rules of a r i t h m e t i c hold, m a n y s t r a n g e — o r w h a t m a y s e e m as s t r a n g e — p h e n o m e n a do t a k e place. A s w e p r o c e e d a n d see e x a m p l e s of rings, we shall see s o m e of t h e s e things occur. With this p r e a m b l e over we are ready to begin. Naturally enough, the first thing we should do is to define that which we'll be talking about.
125
126
Ring T h e o r y
Ch. 4
Definition. A n o n e m p t y set R is said to be a ring if in R t h e r e are two operations + and • such that: (a) a, b E R implies that a + b E R. (b) a + b = b + a for a, b E R. (c) (a + b) + c = a + (b + c) for a,b,c<= R. (d) There exists an element 0 E R such that a + 0 = a for every a G R. (e) Given a E R, t h e r e exists ab G R such that a + 6 = 0. (We shall write b as —a.) N o t e that so far all we have said is that R is an abelian group u n d e r + . W e now spell out the rules for the multiplication in R. (f) A, b E R implies that a • b E R. (g) a  (b • c) = (a • b) • c for fl, b,c E R. This is all that we insist on as far as the multiplication by itself is concerned. But the + and • are not allowed to live in solitary splendor. W e interweave t h e m by the two distributive laws (h) « • (b + c) = a • b + a • c and (b + c) • a: = b • a + c • a, for a, b, c E R. These axioms for a ring look familiar. They should be, for the concept of ring was introduced as a generalization of what happens in the integers. Because of Axiom (g), the associative law of multiplication, t h e rings we defined are usually called associative rings. Nonassociative rings do exist, a n d some of these play an important role in mathematics. But they shall not b e our concern here. So whenever we use the word "ring" we shall always m e a n "associative ring." Although Axioms (a) to (h) are familiar, there are certain things they do not say. W e look at some of the familiar rules that are not insisted u p o n for a general ring. First, we do not postulate the existence of an element I E / ? such that A • 1 = 1 • A = A for every a E R. Many of the examples we shall encounter will have such an element, and in that case we say that 7? is' a ring with unit. In all fairness we should point out that many algebraists do d e m a n d that a ring have a unit element. W e d o insist that 1 ^ 0 ; that is, the ring consisting of 0 alone is not a ring with unit.
Sec. 1
Definitions and Examples
127
Second, in our previous experience with things of this sort, whenever a • b = 0 we concluded that a = 0 or b = 0. This n e e d not b e true, in general, in a ring. W h e n it does hold, the ring is kind of nice and is given a special n a m e ; it is called a domain. Third, nothing is said in the axioms for a ring that will imply the commutative law of multiplication a • b = b • a. T h e r e are noncommutative rings where this law does not hold; we shall see some soon. O u r main concern in this chapter will b e with commutative rings, but for many of the early results the commutativity of the ring studied will not b e assumed. A s we mentioned above, some things m a k e certain rings nicer than others, and so b e c o m e worthy of having a special n a m e . W e quickly give a list of definitions for some of these nicer rings. Definition. A commutative ring A' is an integral domain if a • b = 0 in R implies that a = 0 or b = 0. It should be pointed out that some algebra books insist that an integral domain contain a unit element. In reading a n o t h e r book, the reader should check if this is the case there. T h e integers, Z, give us an obvious example of an integral domain. W e shall see other, somewhat less obvious ones. Definition. A ring R with unit is said to b e a division ring if for every a + 0 in R there is an element b G R (usually written as a~ ) such that a • A " = A ' • A = 1.
x
1 
T h e reason for calling such a ring a division ring is quite clear, for we can divide (at least keeping left and right sides in mind). Although noncommutative division rings exist with fair frequency and do play an important role in noncommutative algebra, they are fairly complicated and we shall give only one example of these. This division ring is the great classic one introduced by H a m i l t o n in 1843 and is k n o w n as the ring of quaternions. (See E x a m p l e 13 below.) Finally, we come to perhaps the nicest example of a class of rings, the field. Definition. ring. In other words, a field is a commutative ring in which we can divide freely by nonzero elements. Otherwise put, R is a field if the nonzero elements of R form an abelian group u n d e r •, the product in R. A ring R is said to b e afield if R is a commutative division
128
Ring Theory
Ch, 4
F o r fields we do have some ready examples: the rational n u m b e r s , the real numbers, the complex n u m b e r s . B u t we shall see many m o r e , p e r h a p s less familiar, examples. C h a p t e r 5 will be devoted to the study of fields. W e spend the rest of the time in this section looking at some examples of rings. We shall drop the • for the product and shall write a • b simply as ab. Examples 1. It is obvious which ring we should pick as our first example, namely Z, the ring of integers u n d e r the usual addition and multiplication of integers. N a t u rally enough, Z is an example of an integral domain. 2. T h e second example is equally obvious as a choice. Let Q be the set of all rational numbers. As we all know, Q satisfies all the rules n e e d e d for a field, so 0 is a field. 3. T h e real numbers, IR, also give us an example of a field. 4. T h e complex numbers, C , form a field. N o t e that Q C U C C ; we describe this by saying that Q is a subfield IR (and of C ) and (R is a subfield of C.
6
of
5. Let R = Z , the integers m o d 6, with the addition and the multiplication defined by [a] + [b] = [a + b] and [a][b] = [ab]. N o t e that [0] is the 0 required by our axioms for a ring, and [1] is the unit element of R. Note, however, that Z is not an integral domain, for [2] [3] = [6] = [0], yet [2] * [0] and [3] * [0]. R is a commutative ring with unit.
6
This example suggests the Definition. A n element a # 0 in a ring R is a zerodivisor for some b ¥= 0 in R. in A if ab  0 *
W e should really call what we defined a left zerodivisor; however, since we shall mainly talk about commutative rings, we shall not n e e d any leftright distinction for zerodivisors. N o t e that both [2] and [3] in Z are zerodivisors. A n integral d o m a i n is, of course, a commutative ring without zerodivisors.
6
6. Let R = Z , the ring of integers m o d 5. 7? is, of course, a commutative ring with unit. But it is more; in fact, it is a field. Its nonzero elements are [1], [2], [3], [4] and we note that [2][3] = [6] = [1], and [1] and [4] are their own inverses. So every nonzero element in Z has an inverse in Z . W e generalize this to any prime p.
5 5 5
Sec. 1
Definitions and Examples
129
7. Let Z b e the integers m o d p, w h e r e p is a prime. Again Z is clearly a commutative ring with 1. W e claim that Z is a field. T o see this, note t h a t if [a] ¥= [0], then p\a. Therefore, by F e r m a t ' s T h e o r e m (Corollary to T h e o r e m 2.4.8), a ' = l(p). F o r t h e classes [•] this says that [ A " " ] = [1]. But [a ~ ] = [ A ] ' ' " , so [ a ] ' ' " = [1]; therefore, [a] ~ is the required inverse for [a] in Z , hence Z is a field.
p p p p x
1
p
l
1
1
p
2
p
p
Because Z has only a finite n u m b e r of elements, it is called a finite field. L a t e r we shall construct finite fields different from the Z^'s.
p
8. Let Q b e t h e rational numbers; if a G Q, we can write a = m/n, w h e r e m and n are relatively prime integers. Call this the reduced form for a. Let R be the set of all a G Q in whose r e d u c e d form t h e denominator is odd. U n d e r the usual addition and multiplication in Q the set R forms a ring. It is an integral d o m a i n with unit but is not a field, for \ , t h e n e e d e d inverse of 2, is not in R. Exactly which elements in R do have their inverses in R? 9. Let R b e the set of all « G < > in whose r e d u c e d form the d e n o m i n a t o r is Q not divisible by a fixed p r i m e p. A s in (8), R is a ring under t h e usual addition and multiplication in Q, is an integral d o m a i n but is not a field. W h a t elements of R have their inverses in R7 B o t h E x a m p l e s 8 and 9 are subrings of Q in the following sense. >
Definition. If R is a ring, then a subring of R is a subset S of R which is a ring if the operations ab and a + b a r e just t h e operations of R applied to the elements a, b G S. F o r S to b e a subring, it is necessary and sufficient that S b e n o n e m p t y and that ab, a ± b G S for all a, b G 5. (Prove!) W e give one further commutative example. This one comes from the calculus. 10. Let R be t h e set of all realvalued continuous functions on the closed unit interval [0, 1]. F o r fgELR and x G [0, 1] define ( / + g)(x) = f(x) + g (x), and ( / • g)(x) = f(x)g(x). F r o m t h e results in t h e calculus, f+g and / • g are again continuous functions on [0, 1]. With these operations R is a commutative ring. It is not an integral domain. F o r instance, if f(x) = — x + \ for 0 < x <  and f(x) = 0 for  < x s 1, and if g(x) = 0 for 0 < x < \ and g{x) = 2x — 1 for  < x < 1, then figER and, as is easy to verify,/• g = 0. It does have a unit element, namely the function e defined by e (x) = 1 for all x G [0,1]. W h a t elements of R h a v e their inverses in R?
130
Ring Theory
Ch. 4
We should now like to see some noncommutative examples. T h e s e are not so easy to come by, although noncommutative rings exist in abundance, because we are not assuming any knowledge of linear algebra on the reader's part. T h e easiest and most natural first source of such examples is the set of matrices over a field. So, in our first noncommutative example, we shall really create the 2 X 2 matrices with real entries. 11. Let F b e the field of real n u m b e r s and let R b e the set of all formal square arrays
a b c d
where a, b,c, d are any real n u m b e r s . For such square arrays we define addition in a natural way by defining
b\
dJ
x
+
/«2
\c
2
b\ d)
2
=
(a + « 2 \c + c
x x
2
b + b d + d
x 2 x
2
It is easy to see that R forms an abelian group under this + with ^ ing as the zero element and ^ _ " _ the negative of ^
^
act
^ T o m a k e of
R a ring, we need a multiplication. W e define one in what may seem a highly unnatural way via
a c
b\(r d)\t
s\ _ far + bt uj \cr + dt
as + bu cs + du
It m a y b e a little laborious, but one can check that with these operations R is a noncommutative ring with ^ N o t e that 1 0 while 0 1 so OVl 0\ 0 A 0 0/
=
^
acting as its multiplicative unit element.
oVo
O/ll
0\ 0/
=
/0 lO
0 0
0
\1
0 0/"
i
0
oyo
O/ll
o\
0 1
(o
owi o
0 0 0
Sec. 1
Definitions and Examples
131
N o t e that
1 0
0 and 0 0
0 1 0
0 are zerodivisors; in fact, 0
o o
0 0
so
0 1
0 0
is a nonzero element whose square is the 0 element of R. This R is known as the ring of all 2 x 2 matrices over F, the real field. F o r those unfamiliar with these matrices, and w h o see n o sense in the product defined for them, let's look at how we do compute the product. To get the top left entry in the product AB, we "multiply" the first row of A by the first column of B, where A, B G R. F o r t h e top right entry, it is the first row of A versus the second column of B. T h e b o t t o m left entry comes from the second row of A versus the first column of 73, and finally, the b o t t o m right entry is the second column of A versus the second column of 73. W e illustrate with an example: Let and Then the first row of A is 1, \ and the first column of B is §, TT; we "multiply" these via 1 • § + \ • TT = 7r/2 + §, and so on. So we see that
In the problems we shall have many matrix multiplications, so that the reader can acquire some familiarity with this strange but important example. 12. Let R b e any ring and let
with + and • as defined in E x a m p l e 11. O n e can verify that S is a ring, also, under these operations. It is called the ring of 2 X 2 matrices over R. O u r final example is one of the great classical examples, the real quaternions, introduced by Hamilton (as a noncommutative parallel to the complex n u m b e r s ) . 13. The quaternions. Let F be the field of real n u m b e r s and consider the set of all formal symbols a + a i + a j + ct k, where a , a , a , a G F. Equality and addition of these symbols are easy, via the obvious route
0 x 2 3 0 { 2 3
132
Ring T h e o r y
Ch. 4
a
0
0
+ afi + a j
2 0 x x
+ ak
3 2
= B + Bj + B j + B fc
0 2 3 2 3
if and only if a = B , a = B , a = B a n d a (a
0
= B , and
3 2
+ afi + a j
2
+ a k)
3
+ (B + B.i + B j +
0
B k)
3 3
= (ao + ft) + («i
+
ft)
' +
(«2 +
ft)
j + (a
+
B )k.
3
W e n o w come to the tricky part, the multiplication. W h e n H a m i l t o n discovered it on October 6, 1843, h e cut t h e basic rules of this product o u t with his penknife on B r o u g h a m Bridge in Dublin. T h e product is based on r = j = k = —l,ij = k, jk = i, ki = j and ji = —k, kj = —i, ik = —j. If we go a r o u n d the circle clockwise
2 2
o
the product of any t w o successive ones is t h e next one, a n d going a r o u n d counterclockwise we get the negatives. W e can write out the product n o w of any two quaternions, according to the rules above, declaring by definition that (a
Q
+ a,/ + a j
2
+ a k)(B
3
0
+ BJ + B j +
2 3
B k)
3
= T + Jii o
where
+ J2J
+ y £,
T = "oft  «ift o
J\ Ji =
aB
2
2 3
2
3
aB
3
3
y
3
«oft + "ift + a B = "oft  «ift + <x B + = «oft + «ift « 2 f t +
2 0
aB
2
^
x
aB
3

aB
3
0
It looks horrendous, doesn't it? B u t it's n o t as b a d as all that. W e a r e multiplying out formally using the distributive laws and using t h e product rules for the i, j , k above. If some a , is 0 in x — a + ad + a j + a k, we shall omit it in expressing x; thus 0 + 0/ + 0/ + Ok will be written simply as 0 , 1 + 0; + 0/ + 0k as 1, 0 + 3i + 4j + 0k as 3z + 4/, and so on. A calculation reveals that
0 2 3
(a
0
+ a i + a j + a k)(a
x 2 3
0
 ai  aj x 2
a k)
3
(II)
Sec. 1
Definitions and Examples
133
This has a very important consequence; for suppose that x = a + a i + a j + a k i= 0 (so some # 0). T h e n , since the a's are real, /3 = al + a\ + a\ + a\ + 0. T h e n from (II) we easily get
Q x 2 3
So, if x + 0, t h e n x has an inverse in the quaternions. Thus the form a noncommutative division ring.
quaternions
A l t h o u g h , as we m e n t i o n e d earlier, t h e r e is n o lack of noncommutative division rings, t h e quaternions above (or some piece of them) are often the only n o n c o m m u t a t i v e division rings that even m a n y professional mathematicians have ever seen. W e shall h a v e m a n y p r o b l e m s — s o m e easy and some quite a bit h a r d e r — a b o u t t h e two examples: t h e 2 X 2 matrices a n d the q u a t e r n i o n s . This way t h e r e a d e r will b e able to acquire s o m e skill with playing with n o n c o m m u t a t i v e rings. O n e final c o m m e n t in this section: If y , y , y , y are as in (I), then
0 x 2 3
{a
2
+ al+
a+
2
2
a I ) ( / 3 + [3J + & + f3 )
3
2
2
(III)
To +
yl
+ rl + yl •
This is k n o w n as Lagrange's Identity; it expresses t h e product of two sums of four squares again as a sum of four squares. Its verification will be one of the exercises.
PROBLEMS
Easier P r o b l e m s * 1. Find all the elements in Z inverse) in Z .
2 4
2 4
that are invertible
(i.e., have a multiplicative
2. Show that any field is an integral domain. 3. Show that Z„ is a field if and only if /; is a p r i m e . 4. Verify that E x a m p l e 8 is a ring. Find all its invertible elements. 5. D o P r o b l e m 4 for E x a m p l e 9. 6. In E x a m p l e 11, the 2 x 2 matrices over t h e reals, check t h e associative law of multiplication.
134
Ring T h e o r y
Ch. 4
7. W o r k out the following: 'i \4 2\/i 7A0
r
1
( H )
<» (1 1
<> (o
( d )
c
o)'
dj\0 0\ 01 /l 10 0\/ff 0/\c /3 d
fl \c
8. Find all matrices
, such : s ) —that ( : s $ ° B » ) ( ° s) *• ^ ) that commute with all 2 X 2 matrices.
9. Find all 2 X 2 matrices 
10. Let R be any ring with unit, S the ring of 2 X 2 matrices over R. (See E x ample 12.) (a) Check the associative law of multiplication in S. ( R e m e m b e r : A' n e e d not b e commutative.) (b) Show that T 'a ,, „ ,0
A, b, c, £ i? V is a s u b r i n g o f 5 .
c
(c) Show that ^
^ j has an inverse in T if and only if a and c h a v e
A
6
^j
W
inverses in R. In that case write down ^
explicitly.
11. Let F: C  » C be defined by P ( « + W) = A  oz. Show that: (a) F(xy) = F(x)F(y) for x, y £ C. (b) F(xx) = x . (c) Using Parts (a) and (b), show that
2
( A + b )(c
2
2
2
+ d ) = (ac  bd)
2
2
+ (ad +
be) .
2
[Note: F(x) is merely x.] 12. Verify the identity in Part (c) of Problem 11 directly. 13. Find the following products of q u a t e r n i o n s . (a) (/ + j)(i  j). (b) (1  i + 2/  2k)(l + 2/  4/ + 6k). (c) (2i  3; + 4/c) . (d) i(a + a i + a j + a k) — (a + a i + a j +
2 0 Y 2 3 0 x 2
a k)i.
3
Sec. 1
Definitions and Examples
135
14. Show that the only quaternions commuting with i are of the form a + Bi. 15. Find the quaternions that c o m m u t e with both i and j . 16. Verify that (cv + afi + a j
0 2 2
+ a k)(a
3
0
 af  a j
2

a k)
3
= a , + a\ + a\ +
a\.
17. Verify Lagrange's Identity by a direct calculation.
MiddleLevel Problems 18. In the quaternions, define
a
n
+ ai
x
+ aj
2
+ a k\
3
=
V«5 +
«i
+
oi ,
2
+ «i2
Show that \xy \ = \x\\y \ for any two quaternions x and y. 19. Show that there is an infinite n u m b e r of solutions to x quaternions. = — 1 in the
20. In the quaternions, consider the following set G having eight elements: G = { ± 1 , ±i, ±j, ±k). (a) Prove that G is a group (under multiplication). (b) List all subgroups of G. (c) W h a t is the center of G? (d) Show that G is a nonabelian group all of whose subgroups are normal. 21. Show that a division ring is a domain. 22. Give an example, in the quaternions, of a noncommutative domain that is not a division ring. 23. Define the m a p * in the quaternions by (a
n
+ af
+ a j + a k)*
2 3
= (a
0
— a i — <x j —
t 2
a k).
3
Show that:
( a ) x * * = ( x * ) * = x.
(b) ( x + y)* = x* + y*. (c) x x * = x * x is real and nonnegative.
(d) (xy)* = y*x*.
[Note the reversal of order in Part (d).] 24. Using *, define  x  = V x x * . Show that \xy\ =  x  \y\ for any two quaternions x and y, by using Parts (c) and (d) of Problem 23.
136
Ring T h e o r y
Ch, 4
25. U s e the result of P r o b l e m 24 to prove Lagrange's Identity. In Problems 26 to 30, let R b e the 2 x 2 matrices over the reals. 26. If ^ ^ j G R, show that ^ is invertible in R if and only if ad — be ¥= 0.
In that case find I ° c
K
^ di
Ia 27. Define detl c
v
b d, I = ad 
s
be. F o r ,r, y £ R show that d e t ( x y )
(det x)(det y). 28. Show that {x £ R  det x # 0} forms a group, G, u n d e r matrix multiplication and that N = {x £ R  det x = 1} is a normal subgroup of G. 29. If x £ R is a zerodivisor, show that det x = 0, and, conversely, if x # 0 is such that det x = 0, then x is a zerodivisor in R. 30. In P , show that \ I
11
£  o
6
a
«, /3 real \ is a field.
H a r d e r Problems 31. Let R be the ring of all 2 X 2 matrices over Z , p a prime. Show that if
p
32. Let
be as in Problem 31. Show that for x, y & R, det(xy) = det(x) det(y).
33. Let G be the set of elements x in the ring R of P r o b l e m 31 such that det(x) + 0. (a) Prove that G is a group. (b) Find the order of G. (Quite hard) (c) Find the center of G. (d) Find a pSylow subgroup of G. 34. Let T be the group of matrices A with entries in the field Z such that det A is not equal to 0. Prove that T is isomorphic to S , the symmetric group of degree 3.
2 3
35. For R as in Example 10, show that S = {/ £ R  / is differentiable on (0,1)} is a subring of R which is not an integral domain. If F is a field, let H(F) b e the ring of quaternions over F, that is, the set of all a + ad + a j + a k, where a , a , ct , a £ F and where equality, addition, and multiplication are defined as for the real quaternions.
0 2 3 0 t 2 3
Sec. 2
S o m e S i m p l e Results
137
36. If F = C, the complex n u m b e r s , show that H(C) is not a division ring. 37. In H(C), find an element .v # 0 such that x
7 2
= 0.
2 3
38. Show that H(F) is a division ring if and only if a , + a\ + a\ + a for a i , a , a , a in Z forces a = a = a 3 0.
= a = 2 3 4 0 y 2
= 0
39. If Q is the field of rational n u m b e r s , show that H(Q) 40. Prove that a finite domain is a division ring.
is a division ring.
41. U s e P r o b l e m 40 to show that Z is a field if p is a prime.
p
2. S O M E SIMPLE RESULTS
N o w that we have seen some examples of rings and have had some experience playing a r o u n d with them, it would seem wise to develop some computational rules. These will allow us to avoid annoying trivialities that could beset a calculation we might be making. T h e results we shall prove in this section are n o t very surprising, not too interesting, and certainly not at all exciting. Neither was learning the alphabet, but it was something we had to do before going on to bigger and better things. T h e same holds for the results we are about to prove. Since a ring R is at least an abelian group under + , there are certain things we know from our group theory background, for instance, («) = a,  ( A + b) = (fl) + {~b); if a + b = a + c, t h e n b = c, and so on. W e begin with
L e m m a 4.2.1. (a) flO = Ofl = 0. (b) a(b) (c) (~a)(b) = (a)b = ab.
Let R be any ring and let a. b G R. Then
=
(ab). a.
(d) If 1 G R, then (  l ) a =
Proof. W e do these in turn. (a) Since 0 = 0 + 0, aO = a(0 + 0) = flO + aQ, hence «0 = 0. We have used the left distributive law in this proof. T h e right distributive law gives Ofl = 0. (b) ab + a(b) = a(b + (b)) = flO = 0 from Part (a). Therefore, a(—b) = —(ab). Similarly, (~a)b = (ab). (c) By P a r t (b), (a)(—b) = —((a)b) = ((ab)) = ab, since we are in an abelian group.
138
Ring T h e o r y
Ch.
4
(d) If 1 G R, then (  l ) a + a = (  l ) a + (l)a = (  1 + 1)A = OA = 0. So (  l ) a =  a by the definition of  A. • A n o t h e r computational result.
Lemma 42.2.
In any ring R, (a + bf = a + b + ab + ba for a, h G A'.
2 2 2 2 2
Proof. This is clearly the analog of (a + B) = a + 2a/3 + B in t h e integers, say, but keeping in mind that R may be noncommutative. So, to it. B y the right distributive law (a + b) = (a + b)(a + b) = (a + b)a + (a + b)b = a + ba + ab + b , exactly what was claimed. •
2 2 2
Can you see the noncommutative version of the binomial t h e o r e m ? T r y it for (a + bf. O n e curiosity follows from t h e two distributive laws w h e n R has a unit element. T h e commutative law of addition follows from the rest.
L e m m a 4.2.3. If R is a system with 1 satisfying all the axioms of a ring, except possibly a + b = b + a for a, b G R, then R is a ring. Proof. W e must show that a + b = b + a for fl, b G R. By the right distributive law (A + b)(l + 1) = (a + b)l + ( A + 6)1 = a + b + a + b. O n the other hand, by the left distributive law (A + b)(l + 1) = a (I + 1) + 6 ( 1 + 1) = a + a + b + b. But then a + b + a + b = a + a + b + b; since w e are in a group under + , we can cancel a on t h e left a n d b on the right to obtain b + a = a + 6, as required. R is therefore a ring. • W e close this brief section with a result that is a little nicer. W e say t h a t a ring R is a Boolean ring [after the English mathematician G e o r g e Boole (18151864)] if x = x for every x G i?.
2
We prove a nice result on Boolean rings.
L e m m a 4.2.4.
2
A Boolean ring is commutative.
2 2 2 2 2
Proof. Let x, y G K , a Boolean ring. Thus x = x, y = y, (x + y) = x + y. B u t (x + y) = x + xy + yx + y = x + xy + yx + y, by L e m m a 4.2.2, so we have (x + y) = (x + y) = x + xy + yx + y, from which we have xy + yx = 0. Thus 0 = x(xy + yx) = x y + xyx = xy + xyx, while 0 = (xy + yx)x = xyx + yx = xyx + yx. This gives us xy + xyx = xyx + yx, and so xy  yx. Therefore, R is commutative. •
2 2 2
Sec. 3
I d e a l s , H o m o m o r p h i s m s , a n d Q u o t i e n t Rings
13S
PROBLEMS
1. Let R b e a ring: since R is an abelian group u n d e r + , na has meaning for us for n E Z, a E R. Show that (na)(mb) = (nm)(ab) if n, m are integers and a, b E R. 2. If 7? is an integral domain and ab = ac for a 4. If R is a ring and e G _ is such that e R (ex — exe) = 0 for every x G A.
2 3 2
0, 6, c G R, show that b = c. = <?, show that (xe  exe)
2
3. If A is a finite integral domain, show that R is a field. =
5. Let R be a ring in which x = x for every x G R. Prove that R is commutative. 6. If a = 0 in A, show that ax + xa c o m m u t e s with a. 7. Let R b e a ring in which x = x for every x G R. Prove that R is commutative. 8. If F is a finite field, show that: (a) T h e r e exists a prime p such that pa = 0 for all a E F. (b) If F has o elements, then o = p" for some integer n. (Hint: Cauchy's Theorem) 9. Let p b e an odd prime and let 1 + § + ••• + ll(p  1) = «/£>, where A, b are integers. Show that p \ a. (Hint: A s a runs through (7 , so does «"'.)
p 4 2
10. If p is a prime and /; > 3, show that i f l +  +   + l / ( p  l ) = fl/fo, where a, are integers, then p 1 a. (Hint: Consider 1/a as a runs through L^.)
2 2
,
3. IDEALS, H O M O M O R P H I S M S , A N D QUOTIENT RINGS
In studying groups, it turned out that h o m o m o r p h i s m s , and their k e r n e l s — the normal subgroups—played a central role. T h e r e is no reason to expect that the same thing should not be true for rings. A s a matter of fact, the analogs, in the setting of rings, of h o m o m o r p h i s m and normal subgroup do play a key role. With the background we have acquired about such things in group theory, the parallel development for rings should b e easy and quick. And it will be! Without any further fuss we m a k e the Definition. T h e mapping cp : R > R' of the ring R into the ring A" is a homomorphism if (a) cp(a + b) = cp(a) + cp(b) and (b) cp(ab) = cp(a)cp(b) for all a, b E R.
140
Ring T h e o r y
Ch, 4
Since a ring has two operations, it is only natural and just that we dem a n d that both these operations be preserved u n d e r what we would call a ring homomorphism. Furthermore, Property (a) in the Definition tells us that cp is a homomorphism of R viewed merely as an abelian group under + into R' (also viewed as a group u n d e r its addition). So we can call on, a n d expect, certain results from this fact alone. Just as we saw in C h a p t e r 2, Section 5 for groups, the image of R u n d e r a homomorphism from R to R', is a subring of R', as defined in C h a p t e r 4, Section 1 (Prove!). Let cp: R —» R' be a ring h o m o m o r p h i s m and let K e r cp = [x G R  cp(x) = 0}, the 0 being that of R'. W h a t properties does K e r cp enjoy? Clearly, from group theory Ker cp is an additive subgroup of R. B u t m u c h more is true. If k G K e r cp and r G R, then cp(k) = 0, so cp(kr) = cp(k)cp(r) = 0cp(r) = 0, and similarly, cp(rk) = 0. So K e r cp swallows u p multiplication from the left and the right by arbitrary ring elements. This property of Ker cp is n o w abstracted to define the i m p o r t a n t analog in ring theory of the notion of normal subgroup in group theory. Definition. ideal of R if: Let R be a ring. A n o n e m p t y subset / of R is called an
(a) I is an additive subgroup of R. (b) Given r G R, a G I, then ra G I and ar G L W e shall soon see some examples of h o m o m o r p h i s m s and ideals. B u t first we n o t e that Part (b) in the definition of ideal really has a left a n d a right part. W e could split it and define a set L of R to be a left ideal of R if L is an additive subgroup of R and given r G R, a: G L, then ra G L. So we require only left swallowingup for a left ideal. W e can similarly define right ideals. A n ideal as we defined it is b o t h a left and a right ideal of R. By all rights w e should then call what we called an ideal a twosided ideal of R. Indeed, in working in noncommutative ring theory one uses this name; here, by "ideal" we shall always mean a twosided ideal. Except for some of the problems, we shall not use the notion of onesided ideals in this chapter. Before going on, we record what was d o n e above for Ker cp as L e m m a 4.3.1. ideal of R. If cp: R —> R' is a homomorphism, then K e r cp is an
W e shall soon see that every ideal can b e m a d e the k e r n e l of a h o m o morphism. Shades of what happens for n o r m a l subgroups of groups!
Sec. 3
I d e a l s , H o m o m o r p h i s m s , a n d Q u o t i e n t Rings
141
Finally, let K be an ideal of R. Since K is an additive subgroup of R, the quotient group RIK exists; it is merely the set of all cosets a + K as a runs over R. But R is not just a group; it is, after all, a ring. N o r is K merely an additive subgroup of R; it is m o r e than that, namely an ideal of R. We should be able to put all this together to m a k e of RIK a ring. H o w should we define a product in RIK in a natural way? What d o we want to declare (a + K)(b + K) to b e ? T h e reasonable thing is to define (a + K)(b + K) = ab + K, which we do. A s always, the first thing that comes u p is to show that this p r o d u c t is welldefined. Is it? W e must show that if a + K = a' + K and b + K = b' + K, then {a + K)(b + K) = ab + K = a'b' + K=(a' + K)(b' + K). However, if a + K = a' + K, then a a' G K, so (a — a')b G K, since K is an ideal of R (in fact, so far, since if is a right ideal of R). Because b + K = b' + K, we have b  b' G K, so a'(b  b') G K, since K is an ideal of R (in fact, since if is a left ideal of R). So both (a  a')b = ab  a'b and a'(b  b') = a'b  a'b' are in K. Thus (ab  a'b) + (a'b  a'b') = ab  a'b' G K. But this tells us (just from group theory) that ab + K = a'b' + K, exactly what we needed to have the p r o d u c t welldefined. So RIK is now endowed with a sum and a product. F u r t h e r m o r e , the mapping cp: R —» RIK defined by cp(a) = a + K for a G R is a h o m o m o r phism of R o n t o RIK with kernel K. (Prove!) This tells us right away that RIK is a ring, being the h o m o m o r p h i c image of the ring R. W e summarize all this in
T h e o r e m 4.3.2. Let K b e an ideal of A*. Then the quotient group RIK as an additive group is a ring u n d e r t h e multiplication (a + K)(b + K) = ab + K. Furthermore, the map cp: R * RIK defined by cp(a) = a + K for a G R is a h o m o m o r p h i s m of R onto RIK having K as its kernel. So RIK is a h o m o morphic image of R. Just from grouptheoretic consideration of R as an additive group, we have that if cp is a h o m o m o r p h i s m of R into A", then it is 11 if and only if Ker cp = (0). A s in groups, we define a h o m o m o r p h i s m to be a monomorphism if it is 11. A m o n o m o r p h i s m which is also onto is called an isomorphism. W e define R and A" to be isomorphic if there is an isomorphism of R onto R'. A n isomorphism from a ring R o n t o itself is called an automorphism of R. For example, suppose R is the field C of complex numbers. T h e n the m a p ping from C t o C sending each element of C t o its complex conjugate is an automorphism of C . (Prove!)
142
Ring Theory
Ch. 4
O n e would have to be an awful pessimist to expect that the h o m o m o r phism theorems proved in Section 7 of Chapter 2 fail in this setting. In fact they do hold, with the slightly obvious adaptation n e e d e d to m a k e the proofs go through. We state the h o m o m o r p h i s m theorems without any further ado, leaving the few details n e e d e d to complete the proofs to the reader. Theorem 4.3.3 (First Homomorphism Theorem). Let the mapping cp : R —> R' be a h o m o m o r p h i s m of R onto R' with kernel K. T h e n R' — RIK; in fact, the mapping ip: RIK —» R' defined by if/(a + K) = cp(a) defines an isomorphism of RIK onto R'. Theorem 4.3.4 (Correspondence Theorem). Let the mapping <p: R —> R' b e a h o m o m o r p h i s m of R onto R' with kernel K. If V is an ideal of R \ let I = [a E R [ cp(a) E / ' } . T h e n / i s an ideal of R, I D K and I/K = / ' . This sets up a 11 correspondence between all the ideals of R' and those ideals of R that contain K. Theorem 4.3.5 (Second Homomorphism Theorem). Let A be a subring of a ring R and I an ideal of R. T h e n A + I = {a + i \ a E A, i E 1} is a subring of i?, / is an ideal of A + / , and (A + / ) / / = yl/(^4 n /). T h e o r e m 4.3.6 (Third Homomorphism Theorem). Let the mapping cp : R —> R' be a h o m o m o r p h i s m of /? onto / ? ' with kernel K. If / ' is an ideal oiR' and I = {a E R\ cp(a) E / ' } , t h e n RII = / ? ' / / ' . Equivalently, if K is an ideal of R and / D /C is an ideal of /?, then RII = (RIK)I(IIK). We close this section with an inspection of some of t h e things we have discussed in some examples. Examples 1. A s usual we use Z , the ring of integers, for our first example. Let n > 1 b e a fixed integer and let /„ be the set of all multiples of n; then /,, is an ideal of Z . If Z „ is the integers m o d «, define cp: Z —» Z„ by <p(a) = [«]. A s is easily seen, < is a h o m o m o r p h i s m of Z onto Z„ with kernel / „ . So by T h e o r e m p 4.3.3, Z„ — Z/I„. (This should come as no surprise, for that is how we originally introduced Z „ . ) 2. Let F be a field; what can the ideals of F be? Suppose that / + (0) is an ideal of F; let a + 0 E /. T h e n , since / is an ideal of T , 1 = <7~'a E /; but now, since 1 E /, rl = r E I for every r E F. In short, / = F. So F has only the trivial ideals (0) and F itself.
7
3. Let R be the ring of all rational n u m b e r s having odd denominators in their reduced form. Let / be those elements of R which in reduced form have an
Sec.
3
Ideals, H o m o m o r p h i s m s , and Quotient Rings
143
even n u m e r a t o r ; it is easy to see that / is an ideal of R. Define cp: R  » Z , the integers m o d 2, by cp(a/b) = 0 if a is even (a, b have n o c o m m o n factor) and cp(alb) = 1 if a is odd. W e leave it to the reader to verify that cp is a h o m o morphism of R onto Z with kernel I. T h u s Z = RII. Give the explicit isomorphism of RII onto Z .
2 2 2 2
4. Let R be the ring of all rational n u m b e r s whose denominators (when in reduced form) are not divisible by p,p a fixed prime. Let I be those elements in R whose n u m e r a t o r is divisible by p; / is an ideal of R and RII — 7L , the integers m o d p. (Prove!)
p
5. L e t R b e the ring of all realvalued continuous functions on the closed unit interval where ( / + g)(x) = f(x) + g(x) and (fg)(x) = f(x)g(x) for /, g G R, x G [0, 1]. Let I = { / G R /(f) = 0}. W e claim that / is an ideal of R. Clearly, it is an additive subgroup. F u r t h e r m o r e if / G / and g G R, t h e n /(f) = 0, so (/g)(f) = /(f)g(f) = 0g(f) = 0. T h u s / g G / ; since / is commutative, g / i s also in /. So / is an ideal of R. W h a t is RII1 Given any f G R, then f(x) = (f(x)
 /(I)) + /(f) =
g(x) +
/(I),
where g(x) = f(x)  /(f). Because g (  ) = /(f)  /(j) = 0, g is in / . So g + 1= I. T h u s / + / = ( / (  ) + g) + / = / (  ) + /. Because /(f) is just a real n u m b e r , A// consists of the cosets a + I for a real. W e claim that every real a comes up. F o r if /(f) = B 0, then «/3 /
L
+ / = (a/T + / ) ( / + / ) = ( a r + /)(/(!) + /)
= (a/T
1
1
1
+ 1)08 + / ) =
a/r ^
1
+ I = a + I.
So A / / consists of all a + I, a real. T h u s we have shown directly that RII is isomorphic to the real field. W e now use T h e o r e m 4.3.3 to show that RII = real field. Let cp:R^U be defined by <p(f) = /(I) T h e n it is easy to verify that cp is a h o m o m o r phism, cp is o n t o and K e r cp = { / G R \ /(f) = 0}; in other words, Ker cp = I. So RII = image of cp = IR. 6. Let R be the integral quaternions, in other words, R = {a a k  a , a , a , a G Z } . For a fixed prime p, let
3 Q x 2 3 = 0
+ ai + aj
x 2
+
( o + 4
a
a
+ 27 +
p
A
^1 P i ; f ° i
a
r
=
0> 1, 2, 3).
p
T h e r e a d e r should verify that I is an ideal of R and that RII Problem 36 of Section 1 and the p a r a g r a p h before it).
~ //(Z^) (see
144
Ring T h e o r y
Ch. 4
a, b E IR k R is a subring of the 2 X 2 matrices over the
reals.
L x
'0 b
Let / = •  q { q ]  6 £ IR }•; / is easily seen to be an additive subgroup of R. Is it
an ideal of R? Consider x so it is in /. Similarly,
y \ / 0 b\ 0 x\0 0/
=
(0
xb
\0
0
0 b\x 0 o)\o
y\ x)
=
0 \0
bx\
0 '
y
so it, too, is in /. So / is an ideal of R. W h a t is Rill points of view. G,ve„
, h e
W e approach it from two
(«
l) .
eR
° (o « ) " ( o ")
+
(o o)
s 0 , h a
<
since ^
^
is in I. Thus all the cosets of / in R look like ^ ^j + ij
^
+ /. If we map
this onto a, that is, i<
= a, we can check that ifi is an isomorphism
o n t o the real field. So RII = U.
We see that RII = R another way. Define <p: R > U. by < ^ p claim that < is a h o m o m o r p h i s m . For, given ^ p ^ ^fj, t h e n
^ ) = A. We
a
/3\
= r t
fed
no «
a M O a j /c \0
' no
d ^ / f l cj [
= c > c
+ c
/j + a"
a + c
0
Sec. 3
I d e a l s , H o m o m o r p h i s m s , a n d Q u o t i e n t Rings
145
,
a l l d
/a ((J
b\f c a)\0 a
d\ c) b\ a
=
f ac {0 , (c
+
ad + bc\ ac )' d\\
= <P
,
h e n C C
(a + c
^ 0
0 c
\
0
a
b + d + c
n
M ,
c
d
,
a l l d
, fa
b\(c
d\\
fac
ad + be
*[[o
a
h Jh(o
flC = < p
ac
lo A > \ O cr
^ G Ker 9,
So cp is indeed a homomorphism of R onto IR. What is Ker < ? If ^ p
then
9^
=
fl
'
^ (0
=
^' '
S
n c e
(0
a) ^ ^
6 r
*'
>
T h u s A = 0. F r o m this we see that I = K e r cp. So A / / — image of cp = IR by T h e o r e m 4.3.3.
8. Let R = {I
1
—b
b
«, b G R [ and let C b e the field of complex numbers. a
11
Define tf/: R —» C by <//^ _ ^
^ j = a + 6z. We leave it to the reader to verify that
i/f is an isomorphism of R onto C . So R is isomorphic t o the field of complex numbers.
9. Let R be any commutative ring with 1. If a G R, let (A) = [xa \ x G A). W e claim that (a) is an ideal of R. T o see this, suppose that u, v G (A); thus M = xa, v = ya for some x, y G R, whence u ± v
= XA
± yfl = (x ± y) a G
(A).
Also, if » G (fl) and r G R, then » = XA, hence ru = r(xfl) = (rx)a, so is in (a). Thus (fl) is an ideal of R. N o t e that if R is not commutative, then (a) need not be an ideal; b u t it is certainly a left ideal of R.
146
Ring T h e o r y
Ch. 4
PROBLEMS
Easier Problems 1. If R is a commutative ring a n d a E R, let L(a) = [x E R \ xa = 0}. Prove that L(a) is an ideal of R. 2. If 7? is a commutative ring with 1 and R has n o ideals other t h a n (0) and itself, prove that R is a field. (Hint: Look at Example 9.) * 3 . If <p : R —> R' is a h o m o m o r p h i s m of 7? onto i?' and 7? has a unit element, 1, show that <p(l) is the unit element of R'. 4. If I, J are ideals of R, define 7 + 7 by I + J = [i + j  / E 7 , ; E / } . Prove that 7 + 7 is an ideal of R. 5. If 7 is an ideal of 7? and A is a subring of 7?, show that I D A is an ideal of A 6. If 7, / are ideals of R, show that 7 n 7 is an ideal of R. 7. Give a complete proof of T h e o r e m 4.3.2. 8. Give a complete proof of T h e o r e m 4.3.4. 9. Let cp : R  » 7?' be a h o m o m o r p h i s m of 7? onto 7?' with kernel 7v. If A is a subring of 7?', let A = [a E R \ cp(a) E / I ' } . Show that: (a) A is a subring of 7\, A D 7C. (b) A X = A . (c) If A is a left ideal of R', then ^ is a left ideal of R. 10. Prove T h e o r e m 4.3.6. 11. In Example 3, give the explicit isomorphism of RII o n t o Z .
2
12. In Example 4, show that RII — Z .
p
13. In Example 6, show that R/I
p
= 77(Z ).
p
14. In Example 8, verify that the mapping i/> given is an isomorphism of 7? onto C. 15. If 7, 7 are ideals of 7?, let IJ b e the set of all sums of elements of the form ij, where i E 7, j E 7. Prove that 77 is an ideal of 7?. 16. Show that the ring of 2 X 2 matrices over the reals has nontrivial left ideals (and also nontrivial right ideals). 17. Prove T h e o r e m 4.3.5. If R, S are rings, define the direct sum of R and S, R © S, by R@S where (r, s) = (r
u x
= {(r, s)\
lt
rER,sES] s = s
] (
s ) if and only if r = r
and w h e r e
(r, s) + (t, u) = (/• + t,s + u),
(r, s)(t, u) = (rt, su).
Sec. 3
I d e a l s , H o m o m o r p h i s m s , a n d Q u o t i e n t Rings
147
18. Show that R © S is a ring and that the subrings ((/, 0)  /• E R] a n d ((0, s)  s G S] are ideals of R © S isomorphic to R and S, respectively. 19. I f R a, b, c real \ and I = 0 0 b real j , show that:
(a) R is a ring. (b) / is an ideal of R. (c) RII — F © F, where F is the field of real n u m b e r s . 20. If /, / are ideals of R, let R = RII and R = RIJ. Show that <p:R^ A\ © R defined by <p(r) = (/ + I, r + J) is a h o m o m o r p h i s m of R into R ® R such that K e r <p = I C\ J.
x 2 x 2
2
2 1 . L e t Z j b e the ring of integers m o d 15. Show that Z
5
1 5
—Z © Z .
3 5
MiddleLevel Problems 22. Let Z be the ring of integers and m, « two relatively prime integers, /„, the multiples of m in Z , and /„ the multiples of n in Z . (a) What is /„, n 7„? (b) U s e the result of Problem 20 t o show that there is a onetoone h o m o m o r p h i s m from ZII to Zll © ZII .
mn m n
23. If m, n are relatively prime, prove that Z„„, — Z„, © Z „ . (Hint: Use a counting argument to show that the homomorphism of Problem 22(b) is onto.) *24. Use the result of Problem 23 to prove the Chinese Remainder Theorem, which asserts that if rn and n are relatively prime integers and a, b any integers, we can find an integer x such that .v = a m o d m and x = b m o d n simultaneously. 25. Let R b e the ring of 2 X 2 matrices over the real numbers; suppose t h a t / is an ideal of R. Show that / = (0) or I = R. (Contrast this with the result of Problem 16.) Harder Problems 26. Let R be a ring with 1 and let S be t h e ring of 2 X 2 matrices over R. If / is an ideal of S show that there is an ideal I of R such that I consists of all the 2 X 2 matrices over / . 27. If P i , p , • • • , p„ are distinct odd primes, show that there are exactly 2" solutions of x = x m o d ( p ! • • • p„), w h e r e 0 s x < p • • • p„.
2 l x
28. Suppose that R is a ring whose only left ideals are (0) and R itself. Prove that either R is a division ring or R has p elements, p a prime, and ab = 0 for every a,b G R.
148
Ring T h e o r y
Ch. 4
29. Let R be a ring with 1. A n element a G R is said to have a /e/r inverse if /3A = 1 for some o G R. Show that if the left inverse b of a is unique, then a/3 = 1 (so b is also a right inverse of a).
4. M A X I M A L IDEALS
This will be a section with o n e major theorem. T h e importance of this result will only become fully transparent when we discuss fields in C h a p t e r 5. H o w ever, it is a result that stands on its own two feet. It isn't difficult to prove, but in mathematics the correlation between difficult and important isn't always that high. T h e r e are m a n y difficult results that are of very little interest and of even less importance, and some easy results that are crucial. Of course, there are some results—many, many—which are of incredible difficulty and importance. L e m m a 4.4.1. Let R be a commutative ring with unit whose only ideals are (0) and itself. T h e n R is a field. Proof. Let a + 0 be in R. T h e n (a) = [xa  x G R} is an ideal of R, as we verified in Example 9 in the preceding section. Since a = la G (a), (a) + (0). Thus, by our hypothesis on R, (a) = R. B u t then, by the definition of (a), every element i G R is a multiple xa of a for some x G R. In particular, because 1 G R, 1 = ba for some b G R. This shows that a has the inverse b in R. So R is a field. • In T h e o r e m 4.3.4—the Correspondence T h e o r e m — w e saw that if cp: R —» R' is a h o m o m o r p h i s m of R onto R' with kernel K, then t h e r e is a 11 correspondence between ideals of R' and ideals of R that contain K. Suppose that t h e r e are n o ideals other than K itself and R which contain K. W h a t does this imply about R"> Since (0) in R' corresponds t o K in R, and R' corresponds to all of R, we must conclude that in this case R' has no ideals other than (0) and itself. So if R' is commutative and has a unit element, then, by L e m m a 4.4.1, R' must be a field. This prompts the following definition. Definition. A p r o p e r ideal M o\R is a maximal ideals of R that contain M are M itself and R. ideal of R if the only
The discussion preceding this definition has already almost proved for us
Sec. 4
M a x i m a l Ideals
149
T h e o r e m 4 4 . 2 . Let R be a commutative ring with 1, and let M b e a maximal ideal of R. Then RIM is a field. Proof. T h e r e is a homomorphism of R onto R' = RIM, and since 1 G 7? we have that R' has 1 + Af as its unit element. (See Problem 3, Section 3). Because M is a maximal ideal of R, we saw in t h e discussion above that R' has n o nontrivial ideals. Thus, by L e m m a 4.4.1, R' = RIM is a field. • This t h e o r e m will be our entry into the discussion of fields, for it will enable us to construct particularly desirable fields whenever we shall n e e d them. T h e o r e m 4.4.2 has a converse. This is T h e o r e m 4.4.3. If R is a commutative ring with 1 and M an ideal of R such that RIM is a field, then M is a maximal ideal of R. Proof. W e saw in Example 2 of Section 3 that the only ideals in a field F are (0) a n d F itself. Since RIM is a field, it has only (0) and itself as ideals. B u t then, by t h e correspondence given us by T h e o r e m 4.3.4, there can b e n o ideal of R property between M and R. T h u s M is a maximal ideal of R. • W e give a few examples of maximal ideals in commutative rings. Examples 1. L e t Z b e t h e integers a n d M an ideal of Z . A s an ideal of Z w e certainly have that M is an additive subgroup of Z , so must consist of all multiples of some fixed integer n. Thus since RIM — Z„ and since Z„ is a field if and only if n is a prime, we see that M is a maximal ideal of Z if a n d only if M consists of all the multiples of some prime p. Thus the set of maximal ideals in Z corresponds to t h e set of prime numbers. 2. L e t Z b e the integers, and let R = [a + bi\a, b G Z ) , a subring of C (j =  1 ) . Let M be the set of all a + bi in R, where 3  a a n d 3  b. W e leave it to t h e reader to verify that M is an ideal of R. W e claim that M is a maximal ideal of R. F o r suppose that T D M and V N M is an ideal of R. So there is an element r + si G TV, where 3 does not divide r or 3 does not divide s. Therefore, 3 \ (r + s ). (Prove using congruences m o d 3 !) But t = r + s = (r + si)(r  si), so is in TV, since r + si G T V and TV is an ideal of R. So T has an integer t = r + s not divisible by 3. T h u s V ut + 3v = 1 for some integers u, v; but t G TV, hence ut G T and 3 G M C TV, V so 3L> G TV. Therefore, 1 = ut + 3v G TV. Therefore, (A + £n')l G TV, since T is V an ideal of R, for all a + bi G R. This tells us that TV = R. So the only ideal of R above M is i? itself. Consequently, M is a maximal ideal of R.
1 2 2 2 2 2 2
150
Ring Theory
Ch. 4
By T h e o r e m 4.4.2 we k n o w that RIM is a field. It can b e shown (see Problem 2) that RIM is a field having nine elements. 3. Let R b e as in E x a m p l e 2 and let I = {a + bi \ 5\a and 5/3}. W e assert that I is not a maximal ideal of R. In R we can factor 5 = (2 + i)(2  i). Let M = {x(2 + i )  x 6 R}. M is an ideal of R, and since 5 = (2 + i)(2  i) is in M, we see that / C M . Clearly, / + M for 2 + i G M and is not in / because 5 j 2. So / + M. C a n M = R1 If so, then (2 + /)(« + bi) = 1 for some a, 6. This gives 2a — b = \ and 2/3 + A = 0; these two equations imply that 5a = 2, so a = §, b = — B u t  g Z,   £ Z; the element a + bi = §  g/ is n o / in R.SoM =t R. O n e can show, however, that M is a maximal ideal of R. (See Problem 3.) 4. Let /? = {« + /3V5  A, 6 integers}, which is a subring of the real field u n d e r the sum and product of real n u m b e r s . That R is a ring follows from
(A +
by/2)
+ (c + d\/2)
= (a + c) + (b +
d)V2
and (A + /3V2)(c + « V 2 ) = (ac + 2bd) + (ad + bc)V2.
Let M = [a + /3V2 G R \ 5\a and 5\b}. M is easily seen to b e an ideal of R. W e leave it to the r e a d e r to show that M is a maximal ideal of R and that RIM is a field having 25 elements. 5. Let R b e the ring of all realvalued continuous functions on the closed unit interval [0, 1]. W e showed in E x a m p l e 5 of Section 3 that if M = [f&R\ / (  ) = 0}, then M is an ideal of R and RIM is isomorphic t o the real field. Thus, by T h e o r e m 4.4.3, M is a maximal ideal of R. Of course, if we let M = {/ G R  f(y) = 0}, where y G [0,1], t h e n M is also a maximal ideal. It can be shown that every maximal ideal in R is of the form M for some y G [0,1], but to prove it we would require some results from real variable theory. W h a t this example shows is that the maximal ideals in R correspond to the points of [0, 1].
y y y
PROBLEMS
1. If a, b are integers and 3 \ a or 3 \ b, show that 3 \ (a + b ). 2. Show that in E x a m p l e 2, RIM is a field having nine elements. 3. In E x a m p l e 3, show that M = \x(2 + i) \ x G R] is a maximal ideal of R.
2 2
Sec. 5
Polynomial Rings
5
151
4. In E x a m p l e 3, show that RIM — Z . 5. In E x a m p l e 3, show that RII ~ Z © Z .
5 5
6. In E x a m p l e 4, show that M is a maximal ideal of R. 7. In E x a m p l e 4, show that AVM is a field having 25 elements. 8. Using E x a m p l e 2 as a model, construct a field having 49 elements. We make a short excursion back to congruences mod p, where p is an odd prime. If a is an integer such that p\a and x = a m o d p has a solution x in Z , we say that a is a quadratic residue mod p . Otherwise, a is said to be a quadratic nonresidue mod p .
2
9. Show that (p  1)12 of the n u m b e r s 1, 2, . . . , p  1 are quadratic residues and ( p — l ) / 2 are quadratic nonresidues mod p. [Hint: Show that {x 1 x # 0 G Z \ forms a group of order  l)/2.]
2 p
10. Let m > 0 be in Z , and suppose t h a t m is not a square in Z . Let A = [a + Vm 6  a, £ G Z }. Prove that u n d e r the operations of sum and > product of real n u m b e r s i? is a ring. 11. If p is an odd prime, let us set l = [a + Vm b p\a and p\b), w h e r e a + V m b €z R, the ring in P r o b l e m 10. Show that I is an ideal of R.
p p
12. If m is a quadratic nonresidue 11 is a maximal ideal of R. 13. In P r o b l e m 12 show that RII
P
mod./?, show that the ideal / is a field having p
2
in Problem
elements.
5. P O L Y N O M I A L R I N G S
T h e material that we consider in this section involves the notion of polynomial and the set of all polynomials over a given field. W e h o p e that most readers will have some familiarity with the notion of polynomial from their high school days and will have seen some of the things one does with polynomials: factoring them, looking for their roots, dividing one by another to get a remainder, and so on. T h e emphasis we shall give to the concept and algebraic object k n o w n as a polynomial ring will b e in a quite different direction from that given in high school. B e that as it may, what we shall strive t o do h e r e is to introduce the ring of polynomials over a field and show that this ring is amenable to a careful dissection that reveals its innermost structure. A s we shall see, this ring is very wellbehaved. T h e development should remind us of what was d o n e for the ring of integers in Section 5 of C h a p t e r 1. T h u s we shall run into the analog of Euclid's algorithm, greatest c o m m o n divisor, divisibility, and possibly most important, the appropriate analog of prime number. This will lead to
152
Ring Theory
Ch. 4
unique factorization of a general polynomial into these "prime polynomials," and to the nature of the ideals and the maximal ideals in this new setting. But the polynomial ring enjoys one feature that the ring of integers did not: the notion of a root of a polynomial. T h e study of the n a t u r e of such roots—which will b e done, for the most part, in the next chapter—constitutes a large and important part of the algebraic history of the past. It goes u n d e r the title Theory of Equations, and in its honorable past, a large variety of magnificent results have been obtained in this area. Hopefully, we shall see some of these as our development progresses. With this sketchy outline of what we intend to do out of the way, we now get down to the nittygritty of doing it. Let F be a field. By the ring of polynomials in x over F, which we shall always write as F[x], we m e a n the set of all formal expressions p(x) = a + a x + • • • + a ix"~ + a„x", n s 0, where the a,, the coefficients of the polynomial p(x), are in F. W e sometimes employ the alternative notation: p(x) = a x" + a  [ A ' " + • • • + a„. In F[x] we define equality, sum, a n d product of two polynomials so as to m a k e of F[x] a commutative ring as follows:
l Q x n
 1
0
1. Equality. W e declare p(x) = a + a x + • • • + a„x" and q(x) = b + b x + • • • + b„x" to b e equal if and only if their corresponding coefficients are equal, that is, if and only if a, = 6 for all / > 0. W e combine this definition of equality of polynomials p(x) and q(x) with the convention that if
0 x 0 x ;
q(x) = b + b x + • • • +
0 x
b x"
n
and if b q (x) as
m + :
= • • • = b„ = 0, t h e n we can d r o p the last n  m terms and write
q(x) = b + b +
0 lX
•••
+
bx.
m
m
This convention is observed in the definition of addition that follows, where s is the larger of m and n and we add coefficients a = ••• = a = 0 if n < s or b = • • • = b = 0 if m < s.
n+1 s m + l s
2. Addition. If p(x) = a + a x + • • • + a„x" and q(x) = b + b x + • • • + b,„x'", we declare p(x) + q(x) = c + cpt + • • • + c x\ where for each c, = a + £>,. So we add polynomials by adding their corresponding coefficients. T h e definition of multiplication is a little more complicated. W e define it loosely at first and then more precisely.
Q x 0 x 0 s {
Sec.
5
Polynomial Rings
153
If p(x) = a + a x + • • • + a x " a n d q(x) = b + b x + • • • + b,„x'", we declare p(x)q(x) = c + c x + • • • + c,x', where the Cj are determined by multiplying t h e expression o u t formally, using t h e distributive laws and the rules of exponents x " x " = x " " , a n d collecting terms. M o r e formally,
0 L n
0 x
3. Multiplication.
0
s
+
Cj =
Qjb
Q
+
a _ b
{ x
x
+
• • • +
a bj_
x
x
+
a bj
0
for every
i.
W e illustrate these operations with a simple example, but first a notational device: If some coefficient is 0, we just omit that term. T h u s we write 9 + Ox + 7 x + Ox  I 4 x as 9 + 7 x  14x . Let p(x) = 1 + 3 x , q(x) = 4  5x + 7 x  x . Then p(x) + q(x) = 5  5x + 1 0 x  x while
2 3 4 2 4 2 2 3 2 3
p(x)q(x)
= (1 + 3 x ) ( 4  5x + 7 x 2 3 2
2
2
x )
2 3
3
= 4  5x + 7 x  x + 3 x ( 4  5x + 7 x  x ) = 4  5x + 7 x  x + 1 2 x  15x + 21x  3x = 4  5x + 19x  16x + 2 1 x  3 x . Try this product using the c, as given above. In some sense this definition of F[x] is n o t a definition at all. W e h a v e indulged in some hand waving in it. B u t it will do. W e could employ sequences to formally define F[x] m o r e precisely, but it would merely cloud what to most readers is well known. T h e first r e m a r k that we m a k e — a n d do n o t verify—is that F[x] is a commutative ring. T o go through t h e details of checking the axioms for a commutative ring is a straightforward b u t laborious task. However, it is important t o note L e m m a 4.5.1. F[x] is a commutative ring with unit.
0 2 3 4 5 2 3 2 3 4 5
degree
Definition. If p(x) = a + ape + • • • + a „ x " a n d a „ ¥= 0, then t h e of p(x), denoted by d e g p ( x ) , is n.
So the degree of a polynomial p(x) is the highest power of x that occurs in the expression for p(x) with a nonzero coefficient. Thus deg(x  x + x ) = 4, deg(7x) = 1, deg 7 = 0. (Note that this definition does not assign a degree to 0. It is, however, sometimes convenient to adopt the convention that the d e gree of 0 b e — co, in which case many degree related results will hold in this extended context.) T h e polynomials of degree 0 and t h e polynomial 0 a r e called the constants; thus the set of constants can b e identified with F. T h e degree function on F[x] will play a similar role to that played by
2 4
154
Ring Theory
Ch. 4
the size of the integers in Z, in that it will provide us with a version of E u clid's Algorithm for F[x]. O n e immediate and important property of the degree function is that it behaves well for products.
Lemma 4.5.2. If p(x), deg(p(x)q(x)) = degp(x) +
q(x) are nonzero elements of F[x], degq(x).
then
Proof. Let m = degp(x) and n = deg q(x); thus the polynomial p (x) = a + a x + • • • + a x"', where a ¥= 0, and the polynomial q(x) = 6 + b x + • • • + b„x", where b„ # 0. T h e highest power of x that can occur in p(x)q(x) is x'" ", from our definition of the product. What is the coefficient of x"' "? T h e only way that x"' " can occur is from (a x'")(b„x") = a b x"' . So the coefficient of x"' " inp(x)q(x) is a b„, which is not 0, since a,„ + 0, b„ + 0. Thus deg(p(x)q(x)) = m + n = degp(x) + degq(x), as claimed in the lemma. •
0 t m m 0 x + + + m +n + m n m
O n e also has some information about deg(p(.t) + q(x)). This is Lemma 4.5.3. If p(x), q(x) (E F[x] and p(x) d e g ( p ( x ) + q(x)) < max(degp(x), deg q(x)). + q(x) # 0, then
W e leave the proof of L e m m a 4.5.3 to the reader. It will play n o role in what is to come, whereas L e m m a 4.5.2 will b e important. W e put it in so that the " + " should not feel slighted visavis the product. A n immediate consequence of L e m m a 4.5.2 is Lemma 4.5.4. F[x] is an integral domain.
Proof If p(x) * 0 and q(x) # 0, then degp{x) > 0, deg q(x) > 0, so deg(p(x)q(x)) = degp(x) + deg q(x) ^ 0. Therefore, p(x)q(x) has a degree, so cannot be 0 (which has n o degree assigned t o it). Thus F[x] is an integral domain. • O n e of the things that we w e r e once forced t o learn was to divide o n e polynomial by a n o t h e r . H o w did we do this? T h e process was called long division. W e illustrate with an e x a m p l e h o w this was d o n e , for what we do in the e x a m p l e is the m o d e l of what we shall d o in t h e general case. W e want to divide 2 x + 1 into x  Ix + 1. W e do it schematically as follows:
2 4
Sec. 5
Polynomial Rings
155
(2x
2
+ 1)1^
Ix
+
1
i _4
Ix lx and we interpret this as:
+1
+ \\
lx + 1 = (2x + l)(fx  i) + (7* + f)
and —lx + f is called the remainder in this division.
2 2 2 4
2
2
W h a t exactly did we do? First, where did the \x come from? It came from the fact that when we multiply 2x + 1 by \x we get x , the highest 2 z power occurring in x  lx + So subtracting from x  lx gets rid of the x term and we go on to what is left and repeat the procedure. This " r e p e a t the p r o c e d u r e " suggests induction, and that is h o w we shall carry out the proof. B u t k e e p in mind that all we shall be doing is w h a t we did in the example above. W h a t this gives us is something like Euclid's Algorithm, in the integers. However, here we call it the Division Algorithm.
4
1.
x (2x + 1)
4
+1
4
T h e o r e m 4.5.5 (Division Algorithm). g(x) G F[x], where g(x) ¥= 0, then
Given the polynomials / ( x ) ,
f(x) = q(x)gix) + r(x), where q(x), r(x) G F[x] and r(x) = 0 or deg r(x) < d e g g ( x ) . Proof. W e go by induction on d e g / f x ) . If either / ( x ) = 0 or d e g / ( x ) < d e g g ( x ) , then / ( x ) = 0g(x) + / ( x ) , which satisfies the conclusion of the theorem. So suppose that d e g / ( x ) > d e g g ( x ) ; thus the polynomial / ( x ) = a + a x + • • • + a x"', where a„, # 0 and the polynomial g(x) = b + h x + • • • + b„x", where b„ i= 0 and w h e r e m > n. Consider
0 x m 0 x
'g(x) = ^x"'~"(b
0
+ b,x+
•••
+
b,,x")
156
Ring Theory
Ch. 4
Thus (a,„/b„)x'"~"g(x) has the same degree and same leading coefficient as does f(x), so f(x)  (a,„lb, )x'"~"g(x) = h(x) is such that the relation deg ft (x) < degf(x) holds. Thus, by induction,
l
h(x)
= q (x)g(x)
x
+ r(x),
where
c {x),r(x)
h
GF[x]
and r(x) = 0 or deg r(x) < deg g(x). R e m e m b e r i n g what h(x) is, we get
h(x) so
x"'"g(x)
= (x)g(x)
gi
+
r(x)
If q(x) = (ajb )x theorem. •
n
m
" + q[(x),
we have achieved the form claimed in the
T h e Division Algorithm has one immediate application: It allows us to determine the nature of all the ideals of F[x]. A s we see in the next t h e o r e m , an ideal of F[x] must merely consist of all multiples, by elements of F[x], of some fixed polynomial.
Theorem 4.5.6. If / * (0) is an ideal of F[x], then / = [f(x)g(x)  f(x) G F[x)}; that is, / consists of all multiples of the fixed polynomial g(x) by the elements of F[x]. Proof. To prove the t h e o r e m , we need to p r o d u c e that fixed polynomial g(x). W h e r e are we going to dig it up? T h e one control we have numerically on a given polynomial is its degree. So why not use the degree function as the mechanism for finding g(x)1 Since / (0) there are elements in I having nonnegative degree. So there is a polynomial g(x) # 0 in / of minimal degree; that is, g(x) # 0 is in / and if 0 t(x) G /, then deg t(x) s deg g(x). Thus, by the division algorithm, t(x) = q(x)g(x) + r(x), where r(x) = 0 or deg r(x) < d e g g ( . t ) . But since g(x) G / and / is an ideal of F[x], we have that q(x)g(x) G /. By assumption, t(x) G I, thus t(x)  q(x)g(x) is in /, so r(x) = t(x)  q(x)g(x) is in /. Since g(x) has minimal degree for the elements of J and r(x) G /, deg r(x) cannot be less than degg(.v). So we are left with r(x) = 0. But this says that t(x) = q(x)g(x). So every element in / is a multiple of g(x). On the other hand,
Sec.
5
P o l y n o m i a l Rings
157
since g(x) G / a n d / is an ideal of F[x], f(x)g(x) G / for all f(x) G F[x]. T h e net result of all this is that I = [f(x)g(x)  f(x) G F[x]}. • Definition. A n integral domain R is called a principal ideal domain if every ideal / in R is of the form I = {xa \ x G R} for some a G I. T h e o r e m 4.5.6 can be stated as: F[x] is a principal ideal domain. W e shall write the ideal generated by a given polynomial, g(x), namely {f(x)g(x)\f(x)EF[x]}, (g(x)). T h e proof showed that if / is an ideal of F[x], t h e n / = (g(x)), where g(x) is a polynomial of lowest degree contained in /. But g(x) is n o t unique, for if a + 0 G F, then ag(x) is in / and has t h e same degree as g(x), so I = (ag(x)). T o get some sort of uniqueness in all this, we single out a class of polynomials.
aS
Definition. f(x) highest power is 1.
G F[x] is a monic polynomial
if the coefficient of its
T h u s f(x) is monic means that f(x) = x" + a,,!*" + • • • + a x +
x 1
a.
0
W e leave to the r e a d e r to show that if / is an ideal of F[x], then t h e r e is only one monic polynomial of lowest degree in /. Singling this out as the generator of / does give us a " m o n i c " uniqueness for the generation of /. O u r next step in this parallel development with what happens in the integers is to have the notion of o n e polynomial dividing another. Definition. g(x) divides f(x), F[x]. Suppose f(x), g(x) G F[x], with g(x) + 0. W e say that written as g(x) \f(x), if f(x) = a(x)g(x) for some a(x) G
N o t e that if f(x) + 0 and g(x) \ f(x), then deg g(x) s d e g / ( x ) by L e m m a 4.5.2. Moreover, the ideals (/(*)) and (g(x)) of F[x], generated by f(x) and g(x), respectively, satisfy the containing relation (f(x)) C (g(x)). (Prove!) W e again emphasize the parallelism between Z, the integers, and F[x] by turning to t h e notion of greatest common divisor. In order t o get some sort of uniqueness, w e shall insist that the greatest common divisor always b e a monic polynomial.
158
Ring T h e o r y
Ch. 4
Definition. For any two polynomials / ( x ) and g(x) G F]x] (not both 0), the greatest common divisor of f(x) and g(x) is the monic polynomial d(x) [sometimes denoted (f(x), g(x))], such that: (a) d(x)  f(x) and d(x)  g(x). (b) If H O  / ( x ) and /?(x)  g(x), then /7(x) ] d(x). A l t h o u g h we defined the greatest common divisor of two polynomials, we neither know, as yet, that it exists, n o r what its form m a y b e . W e could define it in another, and equivalent, way as the monic polynomial of highest degree that divides both f(x) and g(x). If we did that, its existence would b e automatic, but we would not know its form. Theorem 4.5.7. Given f(x) and g(x) + 0 in F[x], then their greatest c o m m o n divisor d(x) £ F[x] exists; moreover, d(x) = a(x)f(x) + b(x)g(x) for some a(x), b(x) G F[x\. Proof. Let / be the set of all r ( x ) / ( x ) + s(x)g(x) freely over F[x]. W e claim that I is an ideal of R. For, (r {x)f(x)
x
as r(x),
s(x)
run
+ (x)g(x))
Sl x
+ (r (x)f(x)
2
+ +
2
s (x)g(x))
2
= (r (x)
+ r (x))f(x)
2
+ ( (x)
Si
s (x))g(x),
so is again in I, and for r(x) G F[x], t(x)(r(x)f(x) + s(x)g(x)) = (t(x)r(x))f(x) + (t(x)s(x))g(x),
so it, too, is again in I. Thus / is an ideal of F[x]. Since g(x) ¥= 0, we k n o w that 7 ^ 0 , since both f(x) and g(x) are in I. Since 7 ^ 0 is an ideal of F[x], it is generated by a unique monic polynomial d(x) ( T h e o r e m 4.5.6). Since f(x), g(x) are in 7, they must t h e n b e multiples of d(x) by elements of F[x]. This assures us that d(x) \f(x) and d(x)\g(x). Because d(x) G 7 and 7 is the set of all r(x)f(x) + s(x)g(x), we h a v e that d(x) = a(x)f(x) + b(x)g(x) for some appropriate a(x), b(x) G F[x]. Thus if h(x) \f(x) and h(x) \g(x), t h e n h(x)  (a(x)f(x) + b{x)g(x)) = d(x). So d(x) is the greatest c o m m o n divisor off(x) and g(x). This proves the theorem; the uniqueness of d(x) is g u a r a n t e e d by the d e m a n d that we have m a d e that the greatest c o m m o n divisor be monic. • A n o t h e r way to see the uniqueness of ^ ( x ) is from Lemma 4.5.8. I f(x), t h e n f(x) If / ( x ) ¥= 0, g(x) * 0 are in F[x] and fix) = ag(x), where a G F.  g(x) and
g( )
x
Sec.
5
Polynomial Rings
159
Proof. By t h e mutual divisibility condition on f(x) and g(x) we h a v e , by L e m m a 4.5.2, d e g / ( x ) < d e g g ( x ) < d e g / f x ) , so d e g / ( x ) = d e g g ( x ) . B u t f(x) = a(x)g(x), so degf(x) = d e g a ( x ) + d e g g ( x ) = deg a(x) + d e g / ( x ) ,
in consequence of which deg a(x) = 0, so a(x) = a, an element of F. • W e leave the proof of the uniqueness of t h e greatest common divisor via L e m m a 4.5.8 to t h e reader. Definition. T h e polynomials / ( x ) , g(x) in F[x] are said to b e relatively prime if their greatest common divisor is 1. Although it is merely a very special case of T h e o r e m 4.5.7, to emphasize it a n d to have it to refer to, we state: Theorem 4.5.9. If f(x), g(x) G F[x] are relatively prime, t h e n a(x)f(x) + b(x)g(x) = 1 for some a(x), b(x) G F[x]. Conversely, if a(x)f(x) + b(x)g(x) = 1 for some a(x), b(x) G F[x], t h e n / ( x ) a n d g(x) are relatively prime. Proof. W e leave this "conversely" p a r t to the r e a d e r as exercise. • A s with t h e integers, we have Theorem 4.5.10. <?(*) I f(x)g(x), If q(x) a n d f(x) are relatively p r i m e and if
t h e n q(x) \ g(x). + b(x)q(x) = 1 for some a(x),
Proof. By T h e o r e m 4.5.9 a(x)f(x) b{x) G F[x]. Therefore,
a(x)f (x)g(x) + b(x)q(x)g(x)
= g(x).
(1)
Since q(x) \ b(x)g(x)q(x) a n d q(x) \f(x)g(x) by hypothesis, q(x) divides t h e lefthand side of the relation in (1). T h u s q(x) divides the righthand side of (1), that is, q{x) \ g(x), the desired conclusion. • W e are n o w ready to single out the important class of polynomials that will play the same role as prime objects in F[x] as did the prime numbers in Z. Definition. T h e polynomial p(x) G F[x] is irreducible if p(x) is of positive degree a n d given any p o l y n o m i a l / ( x ) in F[x], then either p(x)  / ( x ) or p(x) is relatively prime t o / ( x ) .
160
Ring Theory
Ch. 4
W e should n o t e h e r e that t h e definition implies that a polynomial p(x) of positive degree is irreducible in F[x] if and only ifp(x) cannot be factored as a product of two polynomials of positive degree. In other words, if p(x) = a(x)b(x), where a(x) a n d b(x) are in F[x], then either a(x) is a constant (that is, an element of F), or b(x) is constant. T h e proof of this fact is very similar to the proof of an analogous observation concerning two equivalent definitions of a prime n u m b e r . First suppose that p(x) (of positive degree) cannot b e factored as a product of two n o n  c o n s t a n t polynomials. Then, given any f(x) in F[x], w e have only two possibilities for (p(x), f(x)), namely 1 or a monic polynomial of the form cp(x), where c is an element of F. Thus (p(x),f(x)) = 1 or p(x)  fix), which shows that p(x) is irreducible. N o w let p(x) be irreducible in F[x] and suppose p(x) = a(x)b(x) for some a(x), b(x) in F[x]. According to the definition, we must have p(x)  a(x) or (p(x), a(x)) = 1. If p(x) \ a(x), then b(x) must be a constant. If, on t h e other hand, p(x) a n d a(x) are relatively prime, then by T h e o r e m 4.5.10, p(x)  b(x), and in this case a(x) must be a constant. This shows that an irreducible polynomial cannot b e factored as a product of two n o n  c o n s t a n t polynomials. N o t e that the irreducibility of a polynomial depends on t h e field F. F o r instance, the polynomial x  2 is irreducible in Q[x], where Q is the field of rational numbers, but x  2 is n o t irreducible in R[x], w h e r e U is t h e field of real numbers, for in U[x]
2 2
x
2
 2=
(x  V2)(x +
V2).
Corollary to T h e o r e m 4.5.10. If p(x) is irreducible in F[x] and p(x)  a (x)a {x) • • • a (x), where a (x),.. ., a (x) are in F[x], t h e n p ( x ) \ a (x) for some
x 2 k t h t
Proof. W e leave t h e proof t o the reader. (See T h e o r e m 1.5.6.) • Aside from its other properties, an irreducible polynomial p (x) in F[x] enjoys the property that (p(x)), t h e ideal generated by p(x) in F[x], is a maximal ideal of F[x]. W e prove this now. T h e o r e m 4.5.11. If p(x) E F[x], then the ideal (p(x)) generated by p(x) in F[x] is a maximal ideal of F[x] if a n d only if p(x) is irreducible in F[x]. Proof. W e first prove that if p(x) is irreducible in F[x], then t h e ideal M = (p(x)) is a maximal ideal of F[x]. F o r , suppose that N is an ideal of F[x], and W D M . B y T h e o r e m 4.5.6,
Sec.
5
Polynomial Rings
161
N = (f(x))
for some
f(x)
G F[x].
Because p(x) G M C N,p{x) = a ( x ) / ( x ) , since every element in T is of this V form. B u t / ? ( x ) is irreducible in F[x], hence a(x) is a constant o r / ( x ) is a constant. If a(x) = a G F, t h e n p ( x ) = af(x), so f(x) = Thus f(x) G M, which says that N G M, hence N = M. O n the other hand, if / ( j c ) = b G F] then 1 = D o G N, since A is an ideal of F[x], thus g(x)l G N for all g(x) G F[x]. This says that N = Ffx]. Therefore, we have shown M to b e a maximal ideal of F[x].
_ 1 7
In the other direction, suppose that M = (p(x)) is a maximal ideal of F[x]. If p ( x ) is not irreducible, t h e n p(x) = a(x)fc(x), where d e g a ( x ) > 1, deg 6(x) > 1. Let A = (a(x)); then, since p ( x ) = a(x)b(x), p(x) G N. T h e r e fore, M C N. Since deg a(x) > 1, A = («(x)) ^ F[x], since every element in (a(x)) has degree at least that of a(x). By the maximality of M we conclude that M = N. But then a(x) G N = M, which tells us that a(x) = f(x)p(x); combined with p(x) = a(x)b(x) = b(x)f(x)p(x), we get that b(x)f(x) = 1. Since deg 1 = 0 < deg b (x) < deg (6 ( x ) / ( x ) ) = deg 1 = 0, we have reached a contradiction. T h u s p ( x ) is irreducible. •
7 7
This t h e o r e m is important because it tells us exactly what the maximal ideals of F[x] are, namely the ideals generated by the irreducible polynomials. If M is a maximal ideal of F[x], F[x]/M is a field, and this field contains F (or m o r e precisely, the field {a + M  a G F}, which is isomorphic to F). This allows us to construct decent fields K D F, the decency of which lies in that p (x) has a root in K. T h e exact statement and explanation of this we p o s t p o n e until Chapter 5. T h e last topic in this direction that we want to discuss is the factorization of a given polynomial as a product of irreducible ones. N o t e that if p(x) = a x" + fljx" + • • • + a„^ x + a„, a i= 0, is irreducible in F[x], t h e n so is a^pix) irreducible in F[x]; however, a p(x) has the advantage of being monic. So we have this monic irreducible polynomial trivially obtainable from p (x) itself. This will allow us to m a k e m o r e precise the uniqueness part of the next theorem.
1 0 L 0 1 ()
T h e o r e m 4.5.12. Let / ( x ) G F[x] b e of positive degree. Then either / ( x ) is irreducible in F[x] or f(x) is the product of irreducible polynomials in F[x]. In fact, then,
f( )
X
=
fl (x)'> (xrA 2
•
• (x)'"\
Pk
k
where a is the leading coefficient of fix), Pi(x),. . ., p (x) are monic and irreducible in F[x], m > 0, . . . , m > 0 and this factorization in this form is unique u p to the order of the p (x).
x k t
162
Ring T h e o r y
Ch. 4
Proof. We first show the first half of the theorem, namely that f(x) is irreducible or the product of irreducibles. T h e proof is exactly the same as that of T h e o r e m 1.5.7, with a slight, obvious adjustment. W e go by induction on d e g / ( x ) . If d e g / f x ) = 1, t h e n f(x) = ax + b with a + 0 and is clearly irreducible in F[x]. So the result is true in this case. Suppose, then, that the t h e o r e m is correct for all a(x) E F[x] such that dega(x) < degf(x). If f(x) is irreducible, then we have nothing to prove. Otherwise, f(x) = a(x)b(x), a(x) and b(x) E F[x] and degfl(x) < d e g / ( x ) and deg b(x) < d e g / ( x ) . By the induction, a(x) [and b(x)] is irreducible or is the product of irreducibles. But t h e n f(x) is the product of irreducible polynomials in F[x]. This completes the induction, and so proves the opening half of the theorem. Now to the uniqueness half. Again we go by induction on d e g / ( x ) . If deg/(.v) = 1, t h e n / f x ) is irreducible and the uniqueness is clear. Suppose the result true for polynomials of degree less t h a n d e g / f x ) . Suppose that f{x)
t
= a (x)>"ip (x)'"
Pl 2 t
2
•• p^x)"*
= (x)"*
aqi
•••
q (x)'\
r h {
where the p (x), q {x) are monic irreducible polynomials and the m n are all positive and a is the leading coefficient of f(x), that is, the coefficient of the highest power term of f(x). Since p (x)\f(x), we have that Pi( ) I • • • q (x)" , so by the corollary to T h e o r e m 4.5.10, p (x) \ q (x) for some i. Since q (x) is monic and irreducible, as is pi(x), we get p (x) = q,(x). W e can suppose (on renumbering) t h a t p { x ) = q (x). T h u s
l x r r x t t x x x
= ap^xy^p.ix)"'
1 1 2
2
••
2
•p (x)"'*
k
= ap (x)"^ q (xr
x
•••
q (x)'\
r
By induction we have unique factorization in the required form for f( )/Pi(x), whose degree is less than d e g / ( x ) . H e n c e we obtain that nil  1 = « i — 1 (so nil iX i i , • • • , m = n, r = k andp (x) = q (x), . . . , p (x) = q (x), on renumbering the g's appropriately. This completes the induction and proves the theorem. •
= w m = n k k 2 2 k k
W e have pointed out how similar the situation is for the integers Z and the polynomial ring F[x]. This suggests that there should be a wider class of rings, of which the two examples Z and F[x] are special cases, for which m u c h of the argumentation works. It w o r k e d for Z and F[x] because we h a d a measure of size in them, either by the size of an integer or the degree of a polynomial. This measure of size was such that it allowed a Euclidtype algorithm to hold.
Sec. 5
Polynomials Rings rings.
163
This leads us to define a class of rings, the Euclidean
Definition. A n integral domain R is a Euclidean ring if there is a function d from the nonzero elements of R to the nonnegative integers that satisfies: (a) F o r a + 0, b * 0 G R, d(a) < (b) Given a r = 0 or < rf(fl). d(ab).
0, b ¥= 0, there exist g and r G R such that b = qa + r, where
T h e interested student should try to see which of the results proved for polynomial rings (and the integers) hold in a general Euclidean ring. A s i d e from a few problems involving Euclidean rings, we shall not go any further with this interesting class of rings. T h e final comment we m a k e h e r e is that what we did for polynomials over a field we could try to do for polynomials over an arbitrary ring. T h a t is, given any ring R (commutative or noncommutative), we could define the polynomial ring R[x] in x over R by defining equality, addition, and multiplication exactly as we did in F[x], for F a field. T h e ring so constructed, R[x], is a very interesting ring, whose structure is tightly interwoven with that of R itself. It would b e too much to expect that all, or even any, of the theorems proved in this section would carry over to R[x] for a general ring R.
PROBLEMS
In the following problems, F will always denote a field. Easier P r o b l e m s 1. If F is a field, show that the only invertible elements in F[x] are the nonzero elements of F. 2. If R is a ring, we introduce the ring R[x] of polynomials in x over R, just as we did F[x]. Defining d e g / ( x ) for f(x) G R[x] as we did in F[x], show that: (a) d e g ( / ( x ) g ( x ) ) < d e g / ( x ) + d e g g ( x ) if / ( x ) g ( x ) ± 0. (b) T h e r e is a commutative ring R such that we can find f(x), g(x) in R[x] with d e g ( / ( x ) g ( x ) ) < deg f(x) + d e g g ( x ) . 3. Find the greatest c o m m o n divisor of the following polynomials over Q, the field of rational numbers. (a) x  6x + 7 a n d x + 4. (b) x  1 and 2 x  4 x + 2.
3 2 7 5
164
Ring T h e o r y
Ch. 4
6 4
(c) 3 x + 1 and x + x + x + 1. (d) x  1 and x  x + x  1.
3 1 4 3
2
4. Prove L e m m a 4.5.3. 5. In Problem 3, let I = [f(x)a(x) + g(x)b(x)}, where / ( x ) , g(x) run over Q[x] and a(x) is t h e first polynomial and b(x) the second o n e in each part of the problem. Find d(x), so that I = (d(x)) for Parts (a), (b), (c), and (d). 6. Ug(x),f(x) E F[x] and g(x) \f(x), show that ( / ( * ) ) C ( g ( x ) ) .
7 . Prove the uniqueness of the greatest c o m m o n divisor of two polynomials in F[x] by using L e m m a 4.5.8. 8. If f(x), g(x) G F[x] are relatively prime and / ( x )  h(x) show t h a t / X x ) g ( x )  h(x). 9. Prove the Corollary to T h e o r e m 4.5.10. 10. Show that the following polynomials are irreducible over the field F indicated. (a) x + 7 over F = real field = U. (b) x  3x + 3 over F = rational field = < > Q. (c) x + x + 1 over F = Z • (d) x + 1 o v e r F = Z . (e) x  9 o v e r F = Z . (f) x + 2 x + 2 over F = Q.
2 3 2 2 2 1 9 3 1 3 4 2
and g(x)  h(x),
11. If p(x) G F[x] is of degree 3 and p{x) = a x + a x + a x + a , show that p ( x ) if irreducible over F if there is n o element r G F such that p(r) = a r + a r + a r + a = 0.
Q x 2 3 3 2 0 x 2 3
3
2
12. If F C K are two fields and / ( x ) , g(x) G Ffx] are relatively prime in F[x], show that they are relatively prime in K[x]. MiddleLevel Problems 13. Let IR be the field of real n u m b e r s and C that of complex n u m b e r s . Show that R[x]/(x + 1) = C. [Hint: If A = (R[x]/(x + 1), let u b e t h e image of x in A; show that every element in A is of the form a + bu, w h e r e A, b G U and u =  1 . ]
2 2 2
14. Let F = Z , the integers m o d 11. (a) Let p(x) = x + 1; show that p(x) is irreducible in F[x] a n d that F[x)/(p(x)) is a field having 121 elements. (b) Let p(x) = x + x + 4 G F[x]; show that p(x) is irreducible in F[x] and that F[x]l(p(x)) is a field having l l elements.
u 2 3 3
Sec. 5
p
P o l y n o m i a l s Rings
165
15. Let F = T b e the field of integers m o d p, where p is a prime, a n d let q(x) G be irreducible of degree n. Show that F[x]/(<7(x)) is a field having at most p" elements. (See P r o b l e m 16 for a m o r e exact statement.) 16. Let F, q(x) be as in Problem 15; show that F[x]/(q(x)) ments.
2 k
has exactly p" ele
17. Let Pi(x), p (x),... , p {x) G F[x] b e distinct irreducible polynomials and let q(x) = p (x)p (x) • • • p (x). Show that
x 2 k
(q(x))
( (x))®
Pl
(p (x))®
2
^( (x)y
Pk
18. Let F b e a finite field. Show that F[x] contains irreducible polynomials of arbitrarily high degree. (Hint: Try to imitate Euclid's proof that t h e r e is an infinity of prime numbers.) 19. Construct a field h a v i n g p elements, for p an odd prime. 20. If R is a Euclidean ring, show that every ideal of R is principal. 21. If A' is a Euclidean ring, show that R has a unit element. 22. If R is the ring of even integers, show that Euclid's algorithm is false in R by exhibiting two even integers for which the algorithm does not hold. Harder Problems 23. Let F = Z and let p(x) = x'  2 and q{x) = x + 2 b e in F\x\. Show that p(x) and q(x) are irreducible in F[x] and that the fields F[x]/(p(x)) and F[x]/(q(x)) are isomorphic.
7 3 2
24. Let Q b e the field of rational n u m b e r s , and let q(x) = x + x + 1 b e in Q(x). If a is a complex n u m b e r such that a + a + 1 = 0, show that the set [a + ba \ a, b G 0} is a field in two ways; the first by showing it to be isomorphic to something you know is a field, the second by showing that if a + ba + 0, then its inverse is of the same form.
2
2
25. If p is a prime, show that q(x) = 1 + x + x Q[x].
2
2
+ •••x~
p
l
is irreducible in
26. Let R b e a commutative ring in which a = 0 only if a = 0. Show that if q(x) G R[x] is a zerodivisor in R[x], then, if q(x) = a x"
0
+ a x"~
x
l
+•••+«„,
0 x
there is an element b ¥= 0 in R such that ba = ba
ba„ = 0.
166
Ring T h e o r y
Ch. 4
27. Let R be a ring and I an ideal of R. If R[x] and I[x] are the polynomial rings in x over A and I, respectively, show that: (a) I[x] is an ideal of R[x]. (b) R[x]II[x] = (/?//)[*]. Very H a r d Problems *28. D o Problem 26 even if the condition " a = 0 only if a = 0" does n o t hold in P . 29. Let R = [a + bi  a, b integers} C C . Let d(a + bi) = a + b . Show that R is a Euclidean ring w h e r e d is its required Euclidean function. (R is known as the ring of Gaussian integers and plays an important role in n u m b e r theory.)
2 2 2
6. P O L Y N O M I A L S OVER THE R A T I O N A L S
In our consideration of the polynomial ring F[x] over a field F, the particular nature of F never entered the picture. All the results hold for arbitrary fields. However, t h e r e are results that exploit the explicit character of certain fields. O n e such field is that of the rational numbers. W e shall present two important theorems for Q[x], the polynomial ring over the rational field Q, These results depend heavily on the fact that we are dealing with rational numbers. T h e first of these, Gauss' Lemma, relates the factorization over the rationals with factorization over the integers. T h e second one, known as the Eisenstein Criterion, gives us a m e t h o d of constructing irreducible polynomials of arbitrary degree, at will, in Q[x]. In this the field Q is highly particular. For instance, there is no easy algorithm for obtaining irreducible polynomials of arbitrary degree n over the field Z of the integers m o d p, p a prime. Even over Z algorithm is nonexistent; it would b e highly useful to have, especially for coding theory. But it just doesn't exist— so far. W e begin our consideration with two easy results.
p s u c n a n 2
L e m m a 4.6.1.
Let f(x) fix)
0
£ Q[x]; then +
a "~i
lX
= ±(a x"
+ ...
0t x
+a„)
n
where u, m, a ,. . ., a are integers and the a a ,..., a have n o c o m m o n factor greater than 1 (i.e., are relatively prime) and (w, m) = 1.
0 n
Sec.
6
Polynomials Over the Rationals
l 0 x n
167
Proof. S i n c e / ( x ) E Q[x], f(x) = q x" + q x"~ + • • • + q , where the q are rational n u m b e r s . So for i = 0, 1, 2 , . . . , n, q = 6,/c,, w h e r e b , c, are integers. Thus
t ( t
clearing of denominators gives us
where the are integers. If w is the greatest common divisor of u , u . . . , u„, then each u, = w<7,, where a , flj,. . . , a„ are relatively prime integers. T h e n
0 l t 0
/ »
= c,cZ
••
c
( o*" +
fl
" i * "

1
+
• •"
+
«»);
canceling out the greatest c o m m o n factor of w and f(x) = ^(a x"
a
CqCj •
• • c„ gives us
+ ••• + «„)
where M, m are relatively prime integers, as is claimed in the lemma. • T h e next result is a result about a particular h o m o m o r p h i c image of R[x] for any ring R. L e m m a 4.6.2. If R is any ring and / an ideal of R, then I[x], the polynomial ring in x over I, is an ideal of R[x]. F u r t h e r m o r e , P[x]/7[x] — (R/I)[x\, the polynomial ring in x over RII. Proof. L e t R = RII; then there is a h o m o m o r p h i s m <p: R—> R, defined by <p(a) = a + I, whose kernel is /. Define 5>: R[x] —> R[x] by: If / ( x ) = a x" +
0
fljx"
1
+•••+«„,
then
<D(/(x)) = <p(«oK +
pCfli)^" + • • • + cp(a„).
0
1
W e leave it to t h e reader to prove that $ is a h o m o m o r p h i s m of R[x] onto R[x]. W h a t is the kernel, Ker <&, of < > I f / ( x ) = a x" + • • • + a„ is in K e r E? then 3>(/(x)) = 0, the 0 element o£R[x]. Since <f>(f(x)) = <p(a )x" + tpiajx"0 u L 1
+ ••• + < p K ) = 0,
we conclude <p(fl ) = 0, <p(a ) = 0, . . . , cp(a„) = 0, by the very definition of what we m e a n by the 0polynomial in a polynomial ring. Thus each a, is in
168
Ring T h e o r y
Ch. 4
the kernel of cp, which h a p p e n s to b e /. Because a ,a ,..., a„ are in /, f(x) = a x" + a x"~ + • • • + a„ is in I[x]. So K e r $ C I[x]. T h a t I[x] C K e r <> is imE mediate from the definition of the mapping H e n c e I[x] = K e r $ . By the First H o m o m o r p h i s m T h e o r e m ( T h e o r e m 4.3.3), the ring I[x] is then a n ideal of R[x] and R[x] — i?[x]/Ker $ = K[x]/J[x]. This proves t h e lemma, r e membering that R = RII. •
Q 1 l Q x
As a very special case of the lemma we have the Corollary. Let Z be t h e ring of integers, p a prime n u m b e r in Z, a n d I = (p), the ideal of Z generated by p . T h e n Z[x]//[x] = Z [x\.
p
Proof. Since Z — Zll, the corollary follows by applying the l e m m a t o r = z . u
p
W e are ready to prove the first of the two major results we seek in this section. Theorem 4.6.3 (Gauss' Lemma). Let f(x) G Z[x] be a monic polynomial and f(x) = a(x)b(x), where a(x) and b(x) are in Q[x]. T h e n f(x) = a (x)b (x), where a (x), b (x) are monic polynomials in Z[x] a n d deg a (x) = deg a(x), d e g b (x) = deg £>(x).
1 1 x x x x
Proof. Suppose f(x) G Z[x] is monic and f(x) = a(x)b(x), w h e r e a(x), b(x) G Q[x], and deg a(x) = s, deg b(x) = r. By L e m m a 4.6.1, we can express each of a(x), b(x) as a product of a rational n u m b e r and a polynomial with integer coefficients. M o r e precisely, a(x) = —(a£x' ' m
v x
+ a{x ^
s
+ •••+»;)=
'
m
x
^aAx), '
v
w h e r e a' , a{,...,
Q
a' are relatively prime integers and
s
b(x) = ^
r/(2
(bo'v" + b{x>
1
+ • • • + /,/) = ^
'''2
^(x),
where b^b'
u
,. . ,b' are relatively prime. Thus
r
/ ( x ) = a(x)b(x)
= ^a (x)b (x)
x x
=
^a (x)b (x),
x x
where v a n d w are relatively prime, by canceling o u t the c o m m o n factor of u ii a n d m m . Therefore, wf(x) = va (x)b (x), a n d / ( x ) , a (x), b (x) are all in Z[x], Of course, we m a y assume with n o loss of generality that the leading coefficients of a (x) and b (x) are positive. If w = 1, then, s i n c e / ( x ) is monic, we get that va' b' = 1 and this leads
x 2 x 2 x x x x x x 0 0
Sec. 6
P o l y n o m i a l s Over the Rationals
169
easily to v = 1, a' = b' = 1 and so f(x) = a (x)b (x), where b o t h a (x) and bi(x) are monic polynomials with integer coefficients. This is precisely the claim of the theorem, since deg a (x) = deg A (A:) and deg b (x) = deg b(x). Suppose then that w ¥= 1; thus there is a prime p such that p  w and, since (i>, vv) = 1, p j f . Also, since the coefficients a' , a\,.. ., a' of A i ( x ) are relatively prime, there is an i such that p\a\; similarly, there is a such that p\b'j. Let 7 = (p) b e the ideal g e n e r a t e d by p in Z; t h e n Z/I — Z and, by the Corollary to L e m m a 4.6.2, Z[x]/7[x] — Z [x], so is a n integral domain. However, since p  w, w, the image of w in Z[x]/7[x], is 0, and since p\v, v the image of v in Z[x]/7[x] is not 0. T h u s 0 / (x) = v d (x)b (x), where v 0 and fl^x) 0, b (x) ¥= 0 b e c a u s e p \ a \ a n d p \ b ' j for the given i, j above. This contradicts that Z[x]/7[x] is an integral domain. So w 1 is not possible, and the theorem is proved. •
0 Q l 1 1 x L 0 s p p 1 1 x
It might b e instructive for the r e a d e r to try to show directly that if x + 6x  7 is the product of two polynomials having rational coefficients, then it is already the product of two monic polynomials with integer coefficients.
3
One should say something about C. F. Gauss (17771855), considered by many to be the greatest mathematician ever. His contributions in number theory, algebra, geometry, and so on, are of gigantic proportions. His contributions in physics and astronomy are also of such a great proportion that he is considered by physicists as one of their greats, and by the astronomers as one of the important astronomers of the past. A s we indicated at the beginning of this section, irreducible polynomials of degree n over a given field F may b e very hard to come by. However, over the rationals, due to the next t h e o r e m , these exist in abundance a n d are very easy to construct. Theorem 4.6.4 (The Eisenstein Criterion). Let f{x) = x" + a x"~ + • • • + a„ b e a nonconstant polynomial with integer coefficients. Suppose that there is some prime p such that p \ a p \ a ,. • •, p \ a , but p f a . T h e n f(x) is irreducible in Q[x].
x 2 u 2 n n l
Proof. Suppose t h a t / ( x ) = u(x)v(x), w h e r e u(x), v(x) are of positive degree and are polynomials in Q[x]. By G a u s s ' L e m m a we may assume that both u(x) and v(x) are monic polynomials with integer coefficients. Let I = ( p ) b e the ideal generated by p in Z, and consider Z[x]//[x], which is an integral domain, since we k n o w by the Corollary to L e m m a 4.6.2 that Z[x]//[x]  ( Z / / ) [ x ] =* Z [x\. T h e image of f(x) = x" + a x ^ + • • • + a„ in Z[x]//[x] is x", since p\a ,... ,p \ a„. So if w(x) is the image of u(x) and U(x)
n p L x
170
Ring Theory
Ch. 4
that oiv(x) in Z[x]/I[x], t h e n * " = 77(X)TJ(X). Since Ti(x)\x", v(x)\x" in Z[x]/I[x], we must have that u(x) = x , v(x) = x " ~ for some 1 £ r < n. But then u(x) = x + pg(x) and v(x) = x"~ + ph(x), where g(x) and h(x) are polynomials with integer coefficients. Since u(x)v(x) = x" + px'hix) + px"~'g(x) + p g(x)h(x), and since 1 < r < n, the constant term of u(x)v(x) is p st, where s is the constant term of g(x) and t the constant term of h(x). Because f(x) = u(x)v(x), their constant terms are equal, hence a„ = p st. Since s and t are integers, we get that p 1 a „ , a contradiction. In this way we see t h a t / ( x ) is irreducible. •
r r
r r 1 2 2
2
W e give some examples of t h e use to which the Eisenstein Criterion can be put. 1. Let f(x) = x " — p , p any prime. T h e n one sees at a glance that f(x) irreducible in Q[x], for the Eisenstein Criterion applies.
5 2
is
2. L e t / ( j c ) = x  4x + 22. Since 2  22, 2  2 2 and 2 divides the o t h e r relevant coefficients of f(x), t h e Eisenstein Criterion tells us that f(x) is irreducible in Q[x]. 3. L e t / f x ) = x  6 x + 1 2 x + 36x  6. W e see t h a t / ( x ) is irreducible in Q[x] by using either 2 or 3 to check the conditions of t h e Eisenstein Criterion. 4. Let / ( x ) = 5 x  lx + 7; f(x) is not monic, but we can modify f(x) slightly to be in a position w h e r e we can apply the Eisenstein Criterion. Let g(x) = 5 / ( x ) = 5 x  7 • 5 x + 7 • 5 = ( 5 x )  175(5x) + 875. If we let y = 5x, then g(x) = h(y) = V ~ 175y + 875. T h e polynomial h(y) is irreducible in Z[y] by using the prime 7 and applying the Eisenstein Criterion. T h e irreducibility of h(y) implies that of g(x), and so that o f / ( x ) , in Q[x]. This suggests a slight generalization of the Eisenstein Criterion to n o n m o n i c polynomials. (See Problem 4.) 5. Let / ( x ) = x + x + x + x + 1; as it stands we cannot, of course, apply the Eisenstein Criterion to / ( x ) . W e pass to a polynomial g(x) closely related to / ( x ) whose irreducibility in Q[x] will ensure that of f{x). Let g(x) = f(x + 1) = (x + l ) + (x + l ) + (x + I ) + (x + 1) + 1 = x + 5x + 1 0 x + lOx + 5. T h e Eisenstein Criterion applies to g{x), using the prime 5; thus g(x) is irreducible in Q[x]. This implies t h a t / ( x ) is irreducible in Q[x]. (See Problem 1.)
4 3 2 4 3 2 4 3 2 4 3 4 4 3 3 4 4 n 4 3
Sec. 6
Polynomials Over the Rationals
171
Gotthold Eisenstein (18231852) in his short life made fundamental contributions in algebra and analysis.
PROBLEMS
1. In E x a m p l e 5, show that because g(x) is irreducible in Q[x], then so is fix)2. Prove t h a t / ( x ) = x
1 2 3
+ 3x + 2 is irreducible in Q[x]. =
3. Show that there is an infinite n u m b e r of integers a such that f(x) x + 1 5 x — 30x + a is irreducible in Q[x]. W h a t a's do you suggest?
1
4. Prove the following generalization of the Eisenstein Criterion. Let f(x) = a x" + flpv" + • • • + a„ have integer coefficients and suppose that t h e r e is a prime p such that
0
p\a ,
a
p\a ,p\a ,.
l 2
. . ,p\a _
n
u
p\a„,
but p 1 a„; then f(x) in Q[x].
2
is irreducible in Q[x]. = x~
p l
5. If p is a prime, show that f(x)
+ x~
p
2
+ • • • + x + 1 is irreducible
6. L e t F b e the field and cp an a u t o m o r p h i s m of F[x] such that cp(a) = a for every a G F. If / ( J C ) G 77[x], p r o v e that f(x) only if g(x) = (p(f(x)) 7 . Let is. b e a field. Define t h e mapping p:F[*]>FM for every f(x) by <p(/(x)) = f(x + 1) is irreducible in F[x] if and
G F[x]. Prove that cp is an a u t o m o r p h i s m of F[x] such that
<p (a) = a for every a G F. 8. Let F b e a field and b + 0 an element of F. Define the mapping cp : F[x] > F[x] by cp(f(x)) = f(bx) for every / ( x ) G F[x]. Prove that cp is an a u t o m o r p h i s m of F[x] such that cp{a) = a for every a G F . 9. Let F b e a field, o ^ 0, c elements of F. Define the mapping cp : F[x] > F[x] by <p(/(x)) = f(bx + c) for e v e r y / ( x ) G F[x]. Prove that cp is an a u t o m o r p h i s m of F[x] such that < (a) = a for every a G F. p 10. Let p b e an automorphism of F[x], w h e r e F i s a field, such that cp(a) = a for every a GF. Prove that if / ( x ) G F[x], t h e n deg cp(f(x)) = d e g / ( x ) . 11. Let cp b e an automorphism of F[x], w h e r e F is a field, such that cp(a) = a for every a G F. Prove t h e r e exist b 0, c in F such that cp(f(x)) = /(/3x + c ) for e v e r y / ( x ) G F[x].
172
Ring Theory
2
Ch. 4
12. Find a nonidentity a u t o m o r p h i s m <p of Q[x] such that cp is t h e identity automorphism of Q[.v]. 13. Show that in P r o b l e m 12 you do not n e e d the assumption cp(a) = a for every a G Q because any a u t o m o r p h i s m of Q[x] automatically satisfies cp(a) = a for every a G Q. 14. Let C be the field of complex n u m b e r s . Given an integer n > 0, exhibit an automorphism cp of C[x] of o r d e r n.
7. FIELD OF Q U O T I E N T S O F A N I N T E G R A L D O M A I N
Given the integral domain Z, the ring of integers, then intimately related to Z is t h e field Q of rational n u m b e r s that consists of all fractions of integers; t h a t is, all quotients m/n, where m, n # 0 are in Z. N o t e that t h e r e is n o unique way of representing f, say, in < > because  = f = (—7)/(14) = • • •. Q In other words, we are identifying § with f, (—7)/( —14), and so on. This suggests that what is really going on in constructing t h e rationals from the integers is some equivalence relation o n some set based on the integers. T h e relation of Q to Z carries over to any integral domain D. Given an integral domain D, we shall construct a field FDD whose elements will b e quotients alb with a, b G D, b # 0. W e go through this construction formally. Let D be an integral domain, a n d let S = {(a, b) \ a, b G D, b ¥= 0}; S is thus the subset of D X D—the Cartesian product of D with itself—in which the second c o m p o n e n t is not allowed to be 0. Think of (a, b) as alb for a m o ment; if so, when would we w a n t to declare that (a, b) = (c, d)? Clearly, we would want this if alb = c/d, which in D itself would b e c o m e ad = be. W i t h this as our motivating guide we define a relation ~ on S by declaring: ( A , b) ~ (c, d) for (A, b), (c, d) in S, if and only if ad = be. W e first assert that this defines an equivalence relation on S. W e go through the t h r e e requirements for an equivalence relation t e r m by term. 1. (fl, b) ~ (a, b), for clearly ab = ba (since D is commutative). So ~ is reflexive. 2. ( A , b) ~ (c, d) implies that (c, d) ~ (a, b), for (a, b) ~ (c, d) m e a n s flfl* = be; for (c, d) ~ {a, b) we n e e d cb = da, but this is t r u e , since cb = be — ad = da. So ~ is symmetric. 3. ( A , b) ~ (c, d), (c, d) ~ (e,f) implies that ad = be, cf = de, so adf = bef = bde; but d ¥= 0 and we are in an integral domain, hence af = be follows. This says that (a, b) ~ (e, / ) . So the relation is transitive.
Sec. 7
Field of Quotients of an Integral Domain
173
W e have shown that ~ defines an equivalence relation on S. Let F b e the set of all the equivalence classes [a, b] of the elements (a, b) G S. F is our required field. T o show that F is a field, we must endow it with an addition and multiplication. First the addition; what should it b e ? Forgetting all the fancy talk about equivalence relation and the like, we really want [a, b] to be alb. If so, what should [a, b] + [c, d] be other t h a n the formally calculated a_ b This motivates us to define [a,b] + [c,d] = [ad + bc,bd]. (1) c _ ad + be ^ d~ bd
N o t e that since b j= 0, d ¥= 0 and D is a domain, then bd j= 0, hence [ad + be, bd] is a legitimate element of F. A s usual we are plagued with having to show that the addition so defined in F is welldefined. In other words, we must show that if [a, b] = [a', b'] and [c, d] = [c', d'], t h e n [a, b] + [c, d] = [a',b'] + [ c \ d']. F r o m (1) we must thus show that [ad + be, bd] = [a'd' + b'c', b'd'], which is to say, (ad + bc)b'd' = bd(a'd' + b'c'). Since [a, b] = [a', b'] and [c, d] = [c', d'], ab' = ba' and cd' = dc'. Therefore, (ad + bc)b'd' = ab'dd' + bb'cd' = ba'dd' + bb' dc' = (a'd' + b'c')bd, as required. Thus " + " is welldefined in F. T h e class [0, b], b 0, acts as the 0 u n d e r " + ," we denote it simply as 0, and the class [a, b] is the negative of [a, b]. T o see that this m a k e s of F an abelian group is easy, but laborious, for all that really needs verification is the associative law. N o w to t h e multiplication in F. Again motivated by thinking of [a, b] as alb, we define [a,b][c,d] = [ac,bd]. (2)
Again since b 0, d ¥= 0, we h a v e bd =fc 0, so the element [ac, bd] is also a legitimate element of F. A s for the " + " we must show that the product so introduced is welldefined; that is, if [a, b] = [a',b'], [c, d] = [c', d'], then [a,b][c,d] = [ac,bd] = [a'c',b'd'\ = [a',b'\c',d'\
W e know that ab' = ba' and cd' = dc', so acb'd' = ab'cd' = ba'dc' = bda'c', which is exactly what we need for [ac, bd] = [a'c', b'd']. Thus the product is welldefined in F. W h a t acts as 1 in Fl W e claim that for any a i= 0, b i= 0 in D, [a, a] = [b, b] (since ab = ba) and [c, d][a, a] = [ca, da] = [c, d], since (ca)d = (da)c.
174
Ring T h e o r y
Ch. 4
So [a, a] acts as 1, and we write it simply as 1 = [a, a] (for all a # 0 in D). Given [a, b] # 0, t h e n a 0, so [£>, o] is in .F; hence, because [a, b][b, a] — [ab, ba] = [ab, ab] = 1, [a, 6] has an inverse in F. All that remains t o show that the nonzero elements of F form an abelian group u n d e r this product is the associative law and commutative law. W e leave these to the reader. T o clinch that F is a field, we n e e d only now show the distributive law. But [ad + be, bd][e,f] = [(ad + bc)e, bdf], so ([a,b] while [a,b][e,f] + [c,d][e,f] bdf ] bce,bdf][ff]
2
+ [c,d])[e,f]
= [ade +
bee,bdf],
= [ae, bf] + [ce, df] = [aedf + bfee, = [(ade + bee) f bdf ] = [ade + bee, bdf],
2
= [ade +
which we have seen is ([a, b] + [c, d])[e, / ] . T h e distributive law is n o w established, so F is a field. Let a + 0 be a fixed element in D and consider [da, a] for any d G D. T h e m a p cp: d [da, a] is a m o n o m o r p h i s m of D into F. It is certainly 11, for if cp(d) = [da, a] = 0, then da = 0, so d = 0, since D is an integral domain. Also, cp(d d ) = [d d a, a] while cp(d )cp(d ) = [d a, a][d a, a] = [d d a , a ] = [d d a, a][a, a] = [d d a, a] = cp(d d ) F u r t h e r m o r e ,
1 2 x 2 x 2 x 2 2 2 x 2 x 2 x 2 1 z
[d a,a]
x
+ [d a,a]
2
= [d a
t x
2
+ ad,
2 2
2
a]
2
= [d a + d a, a][a, a] = [(d
x 2 x x x x
+ d ) a, a]
2 2 x 2
so cp(d + d ) •= [(d + d )a, a] = [d a, a] + [d a, a] = cp(d ) + cp(d ). T h u s cp maps D monomorphically into F. So, D is isomorphic to a subring of F, and we can thus consider D as " e m b e d d e d " in F. W e consider every element [a, b] of F as the fraction alb. Theorem 4.7.1. Let D b e an integral domain. T h e n t h e r e exists a field which consists of all fractions alb, as defined above, of elements in D .
FDD
T h e field F is called the field of quotients of D. W h e n D = 7L, t h e n F is isomorphic to the field Q of rational numbers. Also, if D is the domain of even integers, then F is also the entire field Q. W h a t we did above in constructing the field of quotients of D was a long, formal, wordy, and probably dull way of doing something that is by its
Sec. 7
Field of Quotients of an Integral Domain
175
n a t u r e something very simple. W e really are doing nothing m o r e than forming all formal fractions alb, a, b + 0 in D, w h e r e we add and multiply fractions as usual. However, it is s o m e t i m e s necessary to see something d o n e to its last detail, painful t h o u g h it m a y b e . M o s t of us h a d never seen a really formal a n d precise construction of t h e rationals from the integers. N o w that we h a v e c o n s t r u c t e d F from D in this formal m a n n e r , forget the details and think of F as the set of all fractions of e l e m e n t s of D.
PROBLEMS
1. Prove the associative law of addition in F. 2. Prove the commutative law of addition in F. 3. Prove that the product in F is commutative and associative. 4. If K is any field that contains 73, show that K D F. (So F is the smallest field that contains D.)
5
FIELDS
T h e notion of a ring was unfamiliar territory for most of us; that of a field touches m o r e closely on our experience. While the only ring, other t h a n a field, that we might h a v e seen in our early training was the ring of integers, we had a bit m o r e experience working with rational n u m b e r s , real n u m b e r s , and, some of us, complex n u m b e r s , in solving linear and quadratic equations. T h e ability to divide by nonzero elements gave us a bit of leeway, which we might not have had with the integers, in solving a variety of problems. So at first glance, when we start working with fields we feel that w e are on h o m e ground. A s we p e n e t r a t e d e e p e r into the subject, we start running across new ideas and n e w areas of results. Once again we'll b e in unfamiliar territory, but hopefully, after some exposure to the topic, the notions will become natural to us. Fields play an important role in geometry, in the theory of equations, and in certain very important parts of n u m b e r theory. W e shall touch u p o n each of these aspects as we progress. Unfortunately, because of the technical machinery we would need to develop, we do not go into Galois theory, a very beautiful part of the subject. W e hope that many of the readers will m a k e contact with Galois theory, and beyond, in their subsequent mathematical training.
1 . E X A M P L E S O F FIELDS
Let's recall that a field F is a commutative ring with unit element 1 such that for every nonzero a E F there is an element a' G F such that aa~ = 1. In other words, fields are "something like" the rationals < > But are they really? T h e inQ.
1 l
176
Sec. 1
E x a m p l e s of Fields
177
tegers mod p, Z , where p is a prime, form a field; in Z we have the relation
p p
pi = 1 + 1 + • • • + 1,= 0
(p times) Nothing akin to this happens in Q. T h e r e are even sharper differences a m o n g fields—how polynomials factor over them, special properties of which we'll see in some examples, and so on. W e begin with some familiar examples. Examples 1. Q, the field of rational numbers. 2. IR, the field of real numbers. 3. C, the field of complex numbers. 4. Let F = {a + bi \ a, b G Q} C C. T o see that F is a field is relatively easy. W e only verify that if a + bi 0 is in F, then (a + bi)~ is also in F. B u t what is (fl + biy ? It is merely
] 1
>
(A
2 2
a
2
i
+
b')
~ / 2 i u2\
(A
bi + b )
_1
(verify!),
2 2
,
.
and since a + b i= 0 and is rational, therefore a/(a + b ) and also b/(a + b ) are rational, hence (a + / 3 / ) is indeed in F.
2 2
5. Let F = {A + b\/2 \ a, b G 0 } C R. A g a i n the verification that F i s a field is not too hard. Here, too, we only show the existence of inverses in F f o r the n o n z e r o elements of F. Suppose that a + bVl ^ 0 is in F ; then, since V 2 is irrational, a — 2b + 0. Because
2 2
( A + b V 2 ) ( a  bV2)
= A 
2
2b ,
2 2
2
we see that (a + bV2)(alcV 2 bid) = 1_, where c = a  2b . T h e required inverse for a + &V2 is a/c  V 2 b/c, which is certainly an element of F, since a/c and b/c are rational. 6. Let F be any field and let F[x] be the ring of polynomials in x over F. Since F[x] is an integral domain, it has a field of quotients according to T h e o r e m 4.7.1, which consists of all quotients f(x)/g(x), where f(x) and g(x) are in F[x) and g(x) + 0. This field of quotients of F[x] is denoted by
F(x) and is called the field of rational functions in x over F.
7. Z , the integers modulo the prime p, is a (finite) field.
p
8. In Example 2 in Section 4 of Chapter 4 we saw how to construct a field having nine elements. These eight examples are specific ones. Using the theorems we have proved earlier, we have some general constructions of fields.
178
Fields
Ch. 5
9. If D is any integeral domain, then it has a field of quotients, by T h e o r e m 4.7.1, which consists of all the fractions alb, where a a n d b are in D and b # 0. 10. If R is a commutative ring with unit element 1 and M is a maximal ideal of R, then T h e o r e m 4.4.2 tells us that RIM is a field. This last example, for particular R's, will play an i m p o r t a n t role in what is to follow in this chapter. W e could go on, especially with special cases of E x a m p l e s 9 a n d 10, t o see m o r e examples. B u t the 10 that we did see above show us a certain variety of fields, and we see that it is n o t too hard to run across fields. In Examples 7 and 8 the fields are finite. If F is a finite field having q elements, viewing F merely as an abelian group u n d e r its addition, " + ," we have, by T h e o r e m 2.4.5, that qx = 0 for every x £ F. This is a behavior quite distinct from that which happens in the fields that we are used to, such as the rationals and reals. W e single out this kind of behavior in the Definition. A field /' is said to have (or, to be of) characteristic p # 0 if for some positive integer p, px = 0 for all x £ F, and n o positive integer smaller than p enjoys this property. If a field F is not of characteristic p 0 for any positive integer p, we call it a field of characteristic 0. So Q, U, C are fields of characteristic 0, while Z is of characteristic 3. In the definition given above t h e use of the letter p for the characteristic is highly suggestive, for we h a v e always used p t o d e n o t e a prime. In fact, as we see in the next theorem, this usage of p is consistent.
3
T h e o r e m 5.1.1. number.
T h e characteristic of a field is either 0 or a prime
Proof. If the field F has characteristic 0, there is nothing m o r e t o say. Suppose then that mx = 0 for all x £ F, where m is a positive integer. L e t p be the smallest positive integer such that px = 0 for all x £ F. W e claim that p is a prime. F o r if p = uv, where u > 1 and v > 1 are integers, then in F, (ul)(vl) = (uv)l = 0, w h e r e 1 is the unit element of F. B u t F, being a field, is an integral domain (Problem 1); therefore, ul = 0 or vl = 0. In either case we get that 0 = (ul){x) = ux [or, similarly, 0 = (vl)x = vx] for any x in F. But this contradicts our choice of p as the smallest integer with this property. H e n c e p is a prime. •
Sec. 1
Examples of Fields
179
N o t e that we did not use the full force of the assumption that F was a field. W e only needed that F was an integral domain (with 1). So if we define the characteristic of an integral domain to be 0 or the smallest positive integer p such that px = 0 for all i G F . w e obtain the same result. T h u s the Corollary. If /) is an integral domain, then its characteristic is either 0 or a prime n u m b e r .
P R O B L E M S
1. Show that a field is an integral domain. 2. Prove t h e Corollary even if D does not have a unit element. 3. Given a ring R, let S = R[x] be the ring of polynomials in x over R, and let T = S[y] be the ring of polynomials in y over S. Show that: (a) A n y element f(x, y) in T has the form 2 2 f l x V , where the a are in R. (b) In t e r m s of the form of f(x, y) in T given in Part (a), give the condition for the equality of two elements f(x, y) and g(x, y) in T. (c) In terms of the form for f(x, y) in Part (a), give the formula for f(x, y) + g(x, y), for f{x, y), g(x, y) in T. (d) Give the form for the product of fix, y) and g(x, y) if f(x, y) and gix, y) are in T. ( T i s called the ring of polynomials in two variables over R, and is denoted by R[x, y].)
; / /y
4. If D is an integral domain, show that D[x, y] is an integral domain. 5. If F is a field and D = F[x, y], the field of quotients of D is called the field of rational functions in two variables over F, and is usually d e n o t e d by F(x, y). Give the form of the typical element of F(x, y). 6. Prove that F(x, y) is isomorphic to F(y, x).
p p p
7 . If F is a field of characteristic p # 0, show that (a + b)  a + b for all a,b G F. (Hint: Use the binomial t h e o r e m and the fact that p is a prime.) 8. If F is a field of characteristic p ¥= 0, show that (a + b) where m = p", for all a, b in F and any positive integer n.
p m
= a " + b"\
1
9. Let F b e a field of characteristic p =h 0 and let cp : F —> F be defined by cpia) = a for all a €E F. (a) Show that <p defines a m o n o m o r p h i s m of F into itself. (b) Give an example of a field F w h e r e cp is not onto. (Very hard.) 10. If F is a finite field of characteristic p, show that the mapping cp defined above is onto, hence is an automorphism of F.
180
Fields
Ch. 5
2 . A BRIEF E X C U R S I O N I N T O V E C T O R S P A C E S
T o get into the things we should like to do in field theory, we n e e d some technical equipment that we do not have as yet. This concerns the relation of two fields K D F and what we would like to consider as some m e a s u r e of the size of K compared to that of F. This size is what we shall call the dimension or degree of K over F. However, in these considerations, much less is n e e d e d of K than that it be a field. W e would be remiss if we proved these results only for the special context of two fields K D F because the same ideas, proofs, and spirit hold in a far wider situation. W e need the notion of a vector space over a field F. Aside from the fact that what we d o in vector spaces will be important in our situation of fields, the ideas developed a p p e a r in all parts of mathematics. Students of algebra must see these things at some stage of their training. A n appropriate place is right here. Definition. A vector space V over a field ./•' is an abelian group u n d e r " + " such that for every a E F and every v E V there is an element av E V, and such that: 1. a(v
{
+ v) =
2
aV[
+ av , for a £ F, u , v £ V.
2 t 2
2. (a + B)v = av + Bv, for a, B £ F, v £ V. 3. a(Bv) = (aB)v, for a, BEF.v E V. 4. lv = v for all v E V, where 1 is the unit element of F. In discussing vector spaces—which we will do very briefly—we shall use lowercase Latin letters for elements of V and lowercase G r e e k letters for elements of F. O u r basic concern here will be with only one aspect of the theory of vector spaces: the notion of the dimension of V over F. W e shall develop this n o tion as expeditiously as possible, not necessarily in the best or most elegant way. W e would strongly advise the readers to see the other sides of what is done in vector spaces in other books on algebra or linear algebra (for instance, our books A Primer on Linear Algebra and Matrix Theory and Linear Algebra). Before getting down to some results, we look at some examples. W e leave to the reader the details of verifying, in each case, that the example really is an example of a vector space. Examples 1. Let F b e any field and let V = {(a,, a ,. .., a„)  a, £ F f o r all /} be the set of /?tuples over F, with equality and addition defined componentwise. F o r
2
Sec. 2
A Brief Excursion into Vector Spaces
181
v = (c*i, a , • • •, a„) and B E F, define Bv by Bv = (Ba V is A vector space over F.
2
u
Ba ,...,
2
Ba ).
n
2. Let F b e any field and let V = F[x] b e the ring of polynomials in x over F. Forgetting the product of any arbitrary elements of F[x] but using only that of a polynomial by a constant, for example, we find that B(a
0
+ ax
y
+ • • • + a„x") = Ba
0
+ Ba x
x
+ ••• +
Ba„x".
In this way V becomes a vector space over F. 3. Let V be as in Example 2 and let W = \f(x) E V \ d e g ( / ( x ) ) < «}. T h e n W
is a vector space over F, and W C V is a subspace of V in the following sense. Definition. A subspace of a vector space V is a n o n e m p t y subset W of that a w £ W a n d w + w £ W f o r all a in F a n d w, w ,w E W.
x 2 l 2
Vsuch
N o t e that t h e definition of subspace W of V implies that W is a vector space whose operations are just those of V restricted to the elements of W. 4. Let V be the set of all realvalued differentiable functions o n [0, 1], t h e closed unit interval, with the usual addition and multiplication of a function by a real n u m b e r . T h e n V is a vector space over R. 5. Let W be all t h e realvalued continuous functions on [0,1], again with the usual addition and multiplication of a function by a real n u m b e r . W, too, is a vector space over R, and the V in E x a m p l e 4 is a subspace of W. 6. Let F b e any field, F[x] the ring of polynomials in x over F. Let f(x) b e in F[x] and / = ( / ( * ) ) t h e ideal of F[x] generated by f(x). Let V = F[x]/J, w h e r e we define a(g(x) + J) = ag(x) + J. V is then a vector space over F. 7. Let U be the real field and let V b e the set of all solutions t o the differential equation d yldx + y = 0. V is a vector space over U.
2 2
8. Let Vbe any vector space over a field F, v , v ,..., v„ elements of V. L e t (v , v , . . . , v ) = {a v + a v + • • • + a v \ a , a , . . . , a„ £ F). T h e n (v , v ,..., v ) is a vector space over Fand is a subspace of V. This subspace (v , v ,..., u „ ) is called t h e subspace of V generated or spanned by v ,... , v„ over F; its elements are called linear combinations of v ,.. ., v„. We shall soon have a great deal to say about (v ,v ,..., v„).
x 2 x 2 n x x 2 2 n n x 2 x 2 n x 2 x x x 2
9. Let V and W b e vector spaces over t h e field F and let V © W = {(u, vt>) 11» £ V, w £ W}, with equality and addition defined componentwise,
182
Fields
Ch. 5
and where a(v, w) = (av, aw). T h e n V © W is easily seen to be a vector space over F; it is called the direct sum of V and W. 10. Let K D Fbe two fields, where the addition " + " is that of K and w h e r e av, for a E F and v E K is the product, as elements of the field K. T h e n Conditions 1 and 2 defining a vector space are merely special cases of the distributive laws that hold in K, and Condition 3 is merely a consequence of the associativity of the product in K. Finally, Condition 4 is just the restatement of the fact that 1 is the unit element of K. So K is a vector space over F. In one respect there is a sharp difference among these examples. W e specify this difference by examining some of these examples in turn. 1. In E x a m p l e 1, if
V l
= (1, 0 , . . . , 0), v
2
= ( 0 , 1 , 0 , . . . , 0), . . . , v„ = (0, 0 , . . . , 1),
then every element u in V has a unique representation in the form v = a v + • • • + a„v„, where a <. ., a„ are in F.
x { u
2. In Example 3,ifv = 1, v = x, . . . , v, = x'^ , . . . , v = x", t h e n every v E Vhas a unique representation as v = a v + • • • + a„v , with the a,in F.
1 2 n + l x t n
1
3. In Example 7, every solution of d y/dx y = a cos x + 6 sin x, with a and B real.
x
2
2
+ y = 0 is of the u n i q u e form
4. In E x a m p l e 8, every v £ (v ,..., v„) has a representation—albeit not necessarily unique—as v = a v + • • • + a v„ from the very definition of (v ,... , v„). Uniqueness of this representation depends heavily o n the elements V i , . . . , V .
x x n {
n
5. In the special case of E x a m p l e 10, w h e r e K = C, the field of complex numbers, and F = IR that of the real n u m b e r s , then every v E C is of the unique form v = a + Bi, a, B E U. 6. Consider K = F(x) D F, the field of rational functions in x over F. W e claim—and leave t o the r e a d e r — t h a t we cannot find any finite set of elements in K which spans K over F. This p h e n o m e n o n was also true in some of the other examples we gave of vector spaces. The whole focus of our attention here will be on this notion of a vector space having some finite subset that spans it over the base field. Before starting this discussion, we must first dispose of a list of formal properties that hold in a vector space. You, dear reader, are by now so
Sec. 2
A Brief Excursion into Vector Spaces
183
sophisticated in dealing with these formal, abstract things that we leave the proof of the next lemma to you. L e m m a 5.2.1. If l* is a vector space over the field /', then, for every a E F and every v £ V: (a) aO = 0, where 0 is the zeroelement of V. (b) Ov = 0, w h e r e the first 0 is the zero in F. (c) av = 0 implies that a = 0 or v = 0. (d) (a)v = (av).
In view of this lemma we shall not run into any confusion if we use the symbol 0 both for the zero of F and that of V. W e forget vector spaces for a m o m e n t a n d look at solutions of certain systems of linear equations in fields. T a k e , for example, the two linear h o m o geneous equations with real coefficients, x + x + x = 0 and 3A'! ~ X + X = 0. W e easily see that for any A ' X such that 4x + 2x = 0 and x =  ( A ' I + x ), we get a solution to this system. In fact, there exists an infinity of solutions to this system other than the trivial one x = 0, x = 0, x = 0. If we look at this example and ask ourselves: " W h y is there an infinity of solutions to this system of linear equations?", we quickly come to the conclusion that, because there are m o r e variables than equations, we h a v e room t o maneuver to produce solutions. This is exactly the situation that holds m o r e generally, as we see below.
x 2 3 2 3 1 ; 3 t 3 2 3 x 2 3
Definition. Let /•' be a field; then the ntuple (B . .., B„), where the Bj are in F, and not all of them are 0, is said to b e a nontrivial solution in F to the system of h o m o g e n e o u s linear equations
u
ax
n
a
1
+ ax
X2
2
+ ••• + a + ••• + a
Ul
x x
n
= 0 = 0 =
0
2i i
x
+
a
22 2
x
2ll
n
(*) ax
n x
+ ax
i2
2
+ ••• + ax
in
n
= 0
ax
rl 7
1
+ a,. x
2
2
+ • • • + a x„
rn x 1
= 0
where the a, are all in F, if substituting x t h e equations of (*).
= B , . . . , x„ = B„ satisfies all
F o r such a system (*) we have the following
184
Fields
Ch. 5
T h e o r e m 5.2.2. If n > r, that is, if the n u m b e r of variables (unknowns) exceeds the n u m b e r of equations in (*), then (*) has a nontrivial solution in F. Proof. T h e m e t h o d is that, which some of us learned in high school, of solving simultaneous equations by eliminating one of the unknowns and at the same time cutting the n u m b e r of equations down by one. W e proceed by induction on r, the n u m b e r of equations. If r = 1, the system (*) reduces to a X \ + • • • + a x„ = 0, and n > 1. If all the a = 0, t h e n X j = x = • • • = x„ = 1 is a nontrivial solution to (*). So, o n r e n u m b e r ing, we may assume that a # 0; we then have the solution to (*), which is nontrivial: x = • • • = x„ = 1 a n d x = — ( l / a ) ( a + ' ' ' + \n)Suppose that the result is correct for r = k for some k and suppose that (*) is a system of k + 1 linear h o m o g e n e o u s equations in n > k + 1 variables. A s above, we may assume that some a,  + 0, and, without loss of generality, that a 0. We construct a related system, (**), of k linear h o m o g e n e o u s equations in n. — 1 variables; since n > k + 1, we have that n. — 1 > k, so we can apply induction to this new system (**). H o w do we get this n e w system? W e want to eliminate x among the equations. W e do so by subtracting a /a times the first equation from the ith o n e for each of / = 2, 3 , . . . , k + 1. In doing so, we end up with the new system of k linear h o m o g e n e o u s equations in n — 1 variables:
n ln u 2 n a 2 x n l 2 ; H x n u
Bx
22
2
+ ••• + •••
+ +
B x„
2n
= 0 = 0
Bx
32
2
j8 jr„
3lI
n
Pk + 1 , 2 * 2 + • • ' + Pk+l,nX„ = °,
w h e r e B  a,  a l a for i = 2, 3 , . . . , k + 1 and j = 2, 3 , . . . , n. Since (**) is a system of k linear h o m o g e n e o u s equations in n  1 variables and n  1 > k, by our induction (**) has a nontrivial solution (y ,..., y„) in F. Let y = —(ot y + • • • + a „ y „ ) / a ; we leave it to the r e a d e r to verify that the (y y ,..., y„) so obtained is a required nontrivial solution to (*). This completes the induction and so proves the theorem. •
u ; n l l 2 , x l2 2 1 11 lt 2
With this result established, we are free to use it in our study of vector spaces. W e now return to these spaces. W e repeat, for emphasis, something we defined earlier in Example 8.
Sec. 2
A Brief E x c u r s i o n into V e c t o r S p a c e s
185
Definition. Let V be a vector space over /' and let u v , • • •, v„ b e in V. T h e element u G V is said to be a linear combination of u i> , • • •» « if y = a v + • • • + a„v„ for some a , • • • a in P.
u 2 u 1; 2 x x x n
A s we indicated in E x a m p l e 8, the set (v u , • • . , v„) of all linear combinations of v , v , • • •, u„ is a vector space over F, and being contained in V, is a subspace of K Why is it a vector space? If a v + • • • + a„u„ and Pi i + • • • + B„v„ are two linear combinations of v ,... , v„, then
u 2 x 2 x x v x
(a v
1
1
+ • • • + a„v„) + (B v
x
1
+ • • • + B„v )
n
= («! + B )v
x
x
+ •••
+ (Q„ + A , ) v„
x
by the axioms defining a vector space, and so is in (v ,.. + • • • + a„u„ G . . . , u„), then y(a v
x lt x
., u ). If y G F and
n
+ • • • + a„v„) = ya v
l x n
l
+ •••
+ya„v„,
and is also in (v . .., v„). Thus (v ,. .., v ) is a vector space. A s we called it earlier, it is the subspace of V spanned over F by v . . ., v„. This leads us to the ultraimportant definition.
l t
Definition. The vector space V over F is finite dimensional over F if V = (vi,.. ., v„) for some v .. . , v„ in V, that is, if V is spanned over Fby a finite set of elements.
u
Otherwise, we say that V is infinite dimensional over F if it is not finite dimensional over F. N o t e that although we have defined what is meant b y a finitedimensional vector space, we still have not defined what is meant by its dimension. T h a t will come in due course. Suppose that V is a vector space over F and v ,. . ., v„ in V are such that every element v in ( u . . . , v„) has a unique representation in the form v = a v + • • • + a„v„, where a , • • •, a„ G F. Since
{ 1 ? x x x
0 G (v ...,
1:
v„)
and
0 = 0v
l
+ ••• +
x x
0v ,
n n
by the uniqueness we have assumed we obtain that if a v + • • • + a v„ = 0, then a = a = • • • = a„ = 0. This prompts a second ultraimportant definition.
x 2
Definition. Let V be a vector space over F ; then the elements v ,..., v„ in V are said to be linearly independent over F if a v + • • • + a„v„ = 0, where a ,.. ., a„ are in F, implies that a = a = • • • = a„ = 0.
x x x x 1 2
186
Fields
Ch. 5
If the elements v v in V are not linearly i n d e p e n d e n t over F, then we say that they are linearly dependent over F. For example, if IR is the field of real numbers and V is the set of 3tuples over IR as defined in E x a m ple 1, then (0, 0, 1), (0, 1, 0), and (1, 0, 0) are linearly i n d e p e n d e n t over IR (Prove!) while (1,  2 , 7), ( 0 , 1 , 0), and (1,  3 , 7) are linearly d e p e n d e n t over R, since 1(1,  2 , 7) + (  l ) ( 0 , ' l , 0) + (  1 ) ( 1 ,  3 , 7) = (0, 0, 0) is a nontrivial linear combination of these elements over R, which is the 0vector.
{ n
N o t e that linear i n d e p e n d e n c e depends on the field F. If C D U are the complex and real fields, respectively, then C is a vector space over R, but it is also a vector space over C itself. T h e elements l,i in C are linearly independent over R but are not so over C, since il + (  l ) i = 0 is a nontrivial linear combination of 1, i over C. W e prove Lemma 5.2.3. If V is a vector space over F and v , • . •, v„ in V are linearly independent over F, t h e n every element v E (v ,. . . , v„) has a unique representation as
x x
v = av
x
x
+ ••• +
av
n
n
with a ,.
x
. . , ct„ in F.
1
Proof. Suppose that v £ (v ,. . ., v„) has the two v = a v + • • • + a„v„ = + • • • + B„v„ with the as gives us that (a  B )v + • • • + (a„ ~ B„)v„ = 0; since early independent over F, we conclude that a  B = 0, yielding for us the uniqueness of the representation. •
} x x x x x
representations as and /3's in F. This v ,..., v„ are lin. . . , a„  B = 0,
x n
H o w finite is a finitedimensional vector space? To m e a s u r e this, call a subset v i , . . . , v„ of V a minimal generating set for V over F if V = (v ,..., v„} and n o set of fewer than n elements spans V over F. W e now come to the third vitally important definition.
x
Definition. If V is a finitedimensional vector space over /•'. then the dimension of V over F, written dim (V), is n, the n u m b e r of elements in a minimal generating set for V over F.
F
In the examples given, d i m ( C ) = 2, since 1, i is a minimal generating set for C over R. However, d i m ( C ) = 1. In E x a m p l e 1, d i m ( V ) = n and in E x a m p l e 3, dim (V) = n + 1. In E x a m p l e 7 the dimension of V over F is 2. Finally, if(v ,..., v„) C V, then dim (v ,..., v„) is at most n.
H c F F x F x
W e now prove
Sec. 2
A B r i e f E x c u r s i o n into V e c t o r S p a c e s
187
L e m m a 5.2.4. If V is finite dimensional over F of dimension n a n d if the elements v ,..., v„ of V generate V over F, then v ,. .. ,v„ are linearly i n d e p e n d e n t over* F.
x x
Proof. Suppose that v , . . . , u„ are linearly dependent over is a linear combination a v + • • • + a v = 0, where not all the may suppose, without loss of generality, that a + 0; (l/a )(a v + • • • + a„v ). Given v E. V, because v . .., v„ is set for V over F,
x i 1 n H x x 2 2 n u
F; thus there a are 0. W e then v = a generating
t x
V = ByVy + • • • + B„V„
(a v
2 2 2
+ • • • + a„v )
n
+ Bv
2
2
+ • • • + B„v„;
l 5
thus v ,. .., v„ span V over F, contradicting that the subset u minimal generating set of V over F. • W e now come to yet another important definition.
v ,...,
2
v„ is a
Definition. Let V be a finitedimensional vector space over F; then v v „ is a basis of V over F if the elements v ,...,v„ span V over T and are linearly independent over F.
7 L
By L e m m a 5.2.4 any minimal generating set of V over F is a basis of V over F. Thus, finitedimensional vector spaces have bases. Theorem 5.2.5. Suppose that V is finite dimensional over /•. Then any two bases of V over F have the same n u m b e r of elements, and this number is exactly dim (V).
F
Proof. L e t u . . . , v„ and w ,.. . , w b e two bases of V over F. W e want to show that m = n. Suppose that m > n. Because v ,..., v„ is a basis of V over F, we know that every element in V is a linear combination of the Vj over F. In particular, w ,..., w,„ are each a linear combination of v . . . , v„ over F. Thus we have
1 ( x m x x l t
w'i
= =
auUi a , ^ !
+ +
ai y
2
2
+ +
• • • + • • • +
a ,v„
h
w
2
a
2 2
i;
2
a v„
2ll
w,„ = a v
ml
x
+ a v
m2
2
+ ••• +
a „v„
m
188
Fields
Ch. 5
where the a,  are in F. Consider
;
ftiVi
+
• • • +
B,„w„, = (a^By
+ aB
2i
2
+ •
+ a B )v
ml m
1
+
+ (a B
ln
x
+ aB
2n
2
+
+ oi „B )
m m
v,
T h e system of linear h o m o g e n e o u s equations «i, A + a B
2i 2
+ • • • + a B„,
mi 7
= 0,
i = 1, 2 , . . . , «,
has a nontrivial solution in T by T h e o r e m 5.2.2, since the n u m b e r of variables, nu exceeds the n u m b e r of equations, n. If B ,..., B,„ is such a solution in then, by the above, B w + • • • + B w,„ = 0, yet not all the /3, are 0. This contradicts the linear independence of w . . . , w over A. Therefore, m s 77.. Similarly, 77 < ; 777; hence 777 = 77. T h e t h e o r e m is now proved, since a minimal generating set of V over F is a basis of V over F and the n u m b e r of elements in this minimal generating set is d i m ( V ) , by definition. Therefore, by the above, n = dim (V), completing the proof. •
x x x m 1 ; m F F
A further result, which we shall use in field theory, of a similar n a t u r e to the things we have been doing is T h e o r e m 5.2.6. Let V b e a vector space over F such that d i m ( V ) = 77. If 777 > 77, then any 777 elements of V are linearly d e p e n d e n t over F.
F
Proof. Let w .. ., w,„ G V and let v ,..., here n = dim (V) by T h e o r e m 5.2.5. Therefore,
u x F
v„ be a basis of V over F;
iv 1 =
au
n
x
+ • • • + a v,
Ul
n
. . . ,w„, =
a
m
l
V i
+
• • • +
a „v„.
m u m
T h e proof given in T h e o r e m 5.2.5, that if 777 > 77 we can find B ... , B in F, and not all 0, such that B w + • • • + B w = 0, goes over word for word. B u t this establishes that w ,. .., w,„ are linearly dependent over F. •
x x m m {
W e close this section with a final t h e o r e m of the same flavor as the p r e ceding ones. T h e o r e m 5.2.7. Let V be a vector space over F with dim (V) = 77. T h e n any 77 linearly i n d e p e n d e n t elements of V form a basis of V over F.
F
Sec. 2
A Brief E x c u r s i o n into V e c t o r S p a c e s
189
Proof. W e want to show that if v ,..., v„ G V are linearly i n d e p e n d e n t over F, t h e n they span V over F. Let v G V; t h e n v, v ,. . ., v„ are n + 1 elements, hence, by T h e o r e m 5.2.6, they are linearly d e p e n d e n t over F. T h u s t h e r e exist elements a, a . . . , a„ in F, not all 0, such that av + a ^ j + • • • + a v = 0. T h e element a cannot b e 0, otherwise + • • • + a„v„ = 0, and not all the a, are 0. This would contradict the linear independence of the elements v ,..., v„ over F. Thus a # 0, and so i; = ( — l / a ) ( a i ; + • • • + a„y„) = B Vi + • • • + /3„u„, where B = a /a . Therefore, v ,. . ., v„ span V over F, and thus must form a basis of V over i . •
x x 1 ; n n x 1 1
x
t
i
1
x
7
PROBLEMS
Easier P r o b l e m s 1. D e t e r m i n e if the following elements in V, the vector space of 3tuples over R, are linearly independent over U. (a) ( 1 , 2, 3), (4, 5, 6), (7, 8, 9). (b) (1, 0, 1), (0, 1, 2), (0, 0, 1). (c) ( 1 , 2 , 3), (0, 4, 5), ( 1 , 3 ,  r ) . 2. Find a nontrivial solution in Z equations: x
x 5
of the system of linear h o m o g e n e o u s
+
x
2
+
2
v = 0
3 3
XL
+ 2 x + 3x = 0
2 3
3x + 4x + 2x = 0
x
3. If V is a vector space of dimension n over Z , p a prime, show that V has p" elements.
p
4. Prove all of L e m m a 5.2.1. 5. Let F be a field and V = F[x], the polynomial ring in x over F. Considering V as a vector space over F, prove that V is not finite dimensional over F. 6. If V is a finitedimensional vector space over F and if W is a subspace of V, prove that: (a) W is finite dimensional over F a n d dim (W) < dim (y). (b) If d i m (W) = d i n v ( y ) , t h e n V = W.
F F F
* 7. Define what you feel should b e a vector space h o m o m o r p h i s m ip of V into W, w h e r e V and W are vector spaces over F. W h a t can you say about the kernel, K, of i/> where 7C = {i> G V  i/>(u) = 0}? W h a t should a vector space isomorphism b e ?
7
190
Fields
Ch. 5
8. If V is a vector space over F and W is a subspace of V, define t h e requisite operations in VIW so that V/W becomes a vector space over F. 9. If V is a finitedimensional vector space over F and v ,. . ., v„, in V are linearly independent over F, show we can find w ,... , w in V, where m + r = d i m ( V ) , such that v ,..., v,„, w ,..., w form a basis of V over F.
x x r F x x r
10. If i/>: F — V" is a h o m o m o r p h i s m of V onto V with kernel if, show that » V — VIK (as vector spaces over F). (See Problem 7). 11. Show that if d i m ( y ) = n and W is a subspace of V with d i m ( W ) = m, then d i m ( V / W ) = n  m.
F f F
12. If V is a vector space over F of dimension n, prove that V is isomorphic to the vector space of /ztuples over F (Example 1). (See P r o b l e m 7).
MiddleLevel Problems 13. Let if D F b e two fields; suppose that K, as a vector space over F, has finite dimension n. Show that if a E if, then there exist a , c v j , . . . , a„ in F, not all 0, such that
0
a
0
+ aa + aa
x 2
2
+ • • • + a„a" = 0.
14. Let F be a field, F[x] the polynomial ring in x over F, and fix) 0 in F[x]. Consider V = F [ x ] / / as a vector space over F, w h e r e J is the ideal of F[x] generated by fix). Prove that d i m ( V ) = deg
F
fix).
15. If V and W are two finitedimensional vector spaces over F, prove that V © W is finite dimensional over F and that dim iV © W) = d i m (V) + d i m ( W ) .
F f F
16. Let V be a vector space over F and suppose that U and TV are subspaces of V. Define £/ + W = {u + w  u E [7, w E W). Prove that: (a) U + W is a subspace of V". (b) £7 + W is finite dimensional over F i f both U and W are. (c) £7 n W is a subspace of V. id) £7 + W is a h o m o m o r p h i c image of £7 © W. (e) If £ and W are finite dimensional over F, then 7 dim (£7 + W) = dim (£7) + d i m ( I ¥ )  d i m ( £ 7 n W).
F F F F
Sec. 3
Field Extensions
191
Harder Problems 17. Let K D Fbe two fields such that dim (K) = m. Suppose that Vis a vector space over K. Prove that: (a) V is a vector space over F. (b) If V is finite dimensional over K, t h e n it is finite dimensional over F. (c) If dim (V) = n, then dim (V) = mn [i.e., d i m ( V ) = dimjcCV) dim (K)].
F K F F F
18. Let K D Fbe fields and suppose that Vis a vector space over 7 T such that C dim (V) is finite. If dim (K) is finite, show that dim (V) if finite and determine its value in terms of dim (V) and d i m (K).
F F K F f
19. Let D b e an integral domain with 1, which h a p p e n s to be a finitedimensional vector space over a field F. Prove that D is a field. (Note: Since F\, which we can identify with F, is in D, the ring structure of D and the vector space structure of D over F are in h a r m o n y with each other.) 20. Let V be a vector space over an infinite field F. Show that V cannot be the settheoretic union of a finite number of proper subspaces of V. (Very hard)
3 . FIELD E X T E N S I O N S
O u r attention now turns to a relationship b e t w e e n two fields K and F, w h e r e K D F. W e call K an extension (or extension field) of F, and call F a sub field of T h e operations in F are just those of K restricted to the elements of F. In all that follows in this section it will be understood that F C K. W e say that K is a finite extension of F if, viewed as a vector space over F, d'\m (K) is finite. We shall write d i m (K) as [K: F] and call it the degree of K over F. W e begin our discussion with what is usually the first result one proves in talking about finite extensions.
F f
T h e o r e m 5.3.1. Let L D K D F b e three fields such that both [L : K] and [A^: F] are finite. T h e n L is a finite extension of F and [L : F] = [L:K][K:F]. Proof. W e shall prove that L is a finite extension of F by explicitly exhibiting a finite basis of L over F. In doing so, we shall obtain the stronger result asserted in the theorem, namely that [L : F] = [L : K][K: F]. Suppose that [L : K] = m and [K:F]= n; then L has a basis v , v ,..., v,„ over A^, and K has a basis w , w ,..., w„ over i . W e shall prove that the mn
x 2 7 t 2
192
Fields
Ch.
5
elements vWj, where / = 1, 2 , . . . , m and / = 1, 2 , . . . , n, constitute a basis of L over F. W e begin by showing that, at least, these elements span L over F; this will, of course, show that L is a finite extension of F. L e t a G L; since the elements v , . . . , u„, form a basis of L over K, we have a = k v + • • • + k v„„ where k , k ,. .., k,„ a r e in K. Since w ,..., w„ is a basis of K over F, w e can express each /c, as
x x x
m
x
2
x
=
fnWi
+
fw
i2
2
+ .• • +//„w„,
t
where the are in F. Substituting these expressions for the k in the foregoing expression of a, we obtain a = (/nWi + fw
12 2
+ • • • + f w„)
ln
vy
mn n
+
• • • +
(fmlWi
+ f W
m2
2
+
• • • + f W)
V .
m
Therefore, on unscrambling this sum explicitly, we obtain
a = fyyWyVy + f WV
12 2 1
+
• • • + fijWjVj
+
• • • +
f W„V .
mn m
Thus the mn elements v Wj in L span L over F; therefore, [L : F] is finite and, in fact, [L : F] < m « . To show that [L : F] = mn, w e need only show that t h e mn elements vjWj above are linearly independent over F, for t h e n — t o g e t h e r with the fact that they span L over F—we would have that they form a basis of L over F. By T h e o r e m 5.2.5 w e would have t h e desired result [ L : F] = mn = [L:K][K:F]. Suppose then that for some b in F.we have the relation
t
u
0 = bvw
u 1
1
+ b vw
12 1
2 2
2
+ • • • + b „v w
1 1
n
n
+
+
b^v^y
• • • b „v„,w„.
m
+
• • • + b „v w
+
• • • + b v Wy
mX m
Reassembling this sum, we obtain c v + c v + • • • + c,„v = 0, w h e r e c = b \ i i + • • • + b w„, . . . , c = £>„,iWi + • • • + b w . Since the c, are elements of K and the elements v , . .., v „ in L are linearly i n d e p e n d e n t over K, we obtain c — c = • • • = c = 0. Recalling that q = b w + • • • + b w„, where t h e b a r e in F and where w ,..., w in K are linearly i n d e p e n d e n t over F, w e deduce from the fact that c = c = • • • — c,„ = 0 that every b^ = 0. Thus only the trivial linear combination, with each coefficient 0, of the elements y,iv over F can b e 0. H e n c e the v Wj are linearly i n d e p e n d e n t over F. W e saw above that this was enough to prove the t h e o r e m . •
x x 2 2 m x w
ln
m
mn
n
x
t
2
m
n
x
ln
tj
x
n
x
2
;
t
Sec. 3
Field Extensions
193
T h e reader should compare T h e o r e m 5.3.1 with the slightly more general result in Problem 17 of Section 2. T h e r e a d e r should n o w be able to solve P r o b l e m 17. A s a consequence of the t h e o r e m we have the Corollary. If L D K D F are t h r e e fields such that [L : F] is finite, t h e n [K : F] is finite and divides [L : F]. Proof. Since L D K, K cannot have m o r e linearly independent elements over F than does L. Because, by T h e o r e m 5.2.6, [L : F] is the size of the largest set of linearly i n d e p e n d e n t elements in L over F, w e therefore get that [K: F ] < [L : F ] , so must be finite. Since L is finite dimensional over F and since K contains F, L must b e finite dimensional over K. Thus all the conditions of T h e o r e m 5.3.1 are fulfilled, whence [L : F] = [L : K][K : F]. Consequently, [K : F ] divides [L : F ] , as is asserted in t h e Corollary. • If K is a finite extension of F , we can say quite a bit about the behavior of t h e elements of K visavis F. T h e o r e m 5.3.2. Suppose t h a t K is a finite extension of F of degree n. Then, given any element u in K there exist elements a , a ,.. . , a in F , not all zero, such that a + a ti + • • • + a u" = 0.
0 x n 0 x n
Proof. Since [K: F] = dim (K) = n and t h e elements 1, u, u ,. . . , u" are n + 1 in n u m b e r , by T h e o r e m 5.2.6 they must b e linearly d e p e n d e n t over F. T h u s w e can find a , a . . . , a„ in F , not all 0, such that a + a ii + a u + • • • + a„u" = 0, proving t h e t h e o r e m . •
F Q 1 ; n x 2 2
1
T h e conclusion of t h e t h e o r e m suggests that we single out elements in an extension field that satisfy a nontrivial polynomial. Definition. If K D F are fields, t h e n a G K is said to b e algebraic F if there exists a polynomial p (x) ^ 0 in F[x] such that p (a) = 0.
1 Q
over
By p(a) we shall m e a n the element a a" + ata"^ + • • • + a„ in X, where p(x) = a x" + a x ~ + •••+ a. If K is an extension of F such that every element of K is algebraic over F, w e call K an algebraic extension of F . In these terms T h e o r e m 5.3.2 can be restated as: IfK is a finite extension ofF, then K is an algebraic extension ofF. T h e converse of this is not true; an algebraic extension of F need not be of finite degree over F . C a n you come u p with an example of this situation?
n l 0 x n
194
Fields
Ch. 5
A n element of K that is not algebraic over F is said to b e transcendental over F. Let's see some examples of algebraic elements in a concrete context. Consider C 73 <Q, the complex field as an extension of the rational one. The complex n u m b e r a = 1 + i is algebraic over 0 , since it satisfies a  2a + 2 = 0
2
over Q. Similarly, the real n u m b e r b = V 1 + V l + \/2 is algebraic over Q>, since b = 1 + V T T v l , so ( 6  l) = 1 + V 2 , and therefore ((b — l ) — I ) = 2. Expanding this out, we get a nontrivial polynomial expression in b with rational coefficients, which is 0. Thus b is algebraic over Q. It is possible to construct real n u m b e r s that are t r a n s c e n d e n t a l over Q fairly easily (see Section 6 of C h a p t e r 6). H o w e v e r , it takes s o m e real effort t o establish the transcendence of certain familiar n u m b e r s . T h e two familiar n u m b e r s e and IT can b e shown to b e transcendental over Q. T h a t e is such was p r o v e d by H e r m i t e in 1873; the proof that TT is t r a n s c e n d e n t a l over Q is much h a r d e r and was first carried out by L i n d e m a n n in 1882. W e shall not go into the proof h e r e t h a t any particular n u m b e r is t r a n s c e n d e n tal over Q. H o w e v e r , in Section.,7 of C h a p t e r 6 we shall at least show t h a t 7r is irrational. This m a k e s it a possible candidate for a t r a n s c e n d e n t a l n u m ber of Q, for clearly any rational n u m b e r b is algebraic over < because it Q satisfies the polynomial p (x) = x  b, which has rational coefficients.
2 2
1
2
3
2
Definition. A complex n u m b e r is said to b e an algebraic number is algebraic over <Q.
if it
A s we shall soon see, the algebraic n u m b e r s form a field, which is a subfield of C. W e return to the general development of the theory of fields. W e have seen in T h e o r e m 5.3.2 that if K is a finite extension of F, then every element of K is algebraic over F. W e turn this m a t t e r around by asking: If K is an extension of F and a G K is algebraic over F, can we somehow p r o d u c e a finite extension of F u s i n g al T h e answer is yes. This will come as a consequence of the next t h e o r e m — w h i c h we prove in a context a little m o r e general than what we really need.
T h e o r e m 5.3.3. Let D b e an integral domain with 1 which is a finitedimensional vector space over a field F . T h e n D is a field. Proof. To prove the t h e o r e m , we must produce for a verse, a~ , in D such that aaT = 1.
l 1
0 in D an in
Sec. 3
Field Extensions
195
A s in the proof of T h e o r e m 5.3.2, if dim (D) = n, then 1, a, cr, in D are linearly dependent over F. Thus for some appropriate a , a . . . , a„ in F, not all of which are 0,
F 0 u
a a"
0 T r 1 0 x
+ aid"'
1
+ •••
+ «„ = ().
Let p(x) = B x + B x ~ + • • • + B,. ¥= 0 b e a polynomial in F[x] of lowest degree such that p (a) = 0. W e assert that B + 0. For if B = 0, then
r r
0 = Ba
0
r
+ B cr
x 1
l
+ ••• + + •••
B_a
r x
= (B^a'
+
+ B _ ) a.
r L l a
Since D is an integral domain and a ¥= 0, we conclude that B a'~ + Bya'— + ••• + B 1 = 0, hence q(a) = 0, where q(x) = Bf) ' + Bx~ + • • • + / 3 _ ! in F[x] is of lower degree than p(x), a contradiction. Thus B i= 0, hence B~ is in F a n d
2 x 1 r r 2 x r 1 r
Pr
giving us that —(p a ~ + • • • + / 3 _ i ) / / 3 , which is in D, is the a~ in 73 that we required. This proves t h e t h e o r e m . •
Q r r
r
l
l
Having T h e o r e m 5.3.3 in hand, we want to m a k e use of it. So, h o w do we produce subrings of a field K that contain F and are finite dimensional over Fl Such subrings, as subrings of a field, are automatically integral domains, and would satisfy the hypothesis of T h e o r e m 5.3.3. The means to this end will b e the elements in K that are algebraic over F. But first a definition. Definition. The element a in the extension K of F is said to be algebraic of degree n if t h e r e is a polynomial p (x) in F[x] of degree n such that p (a) = 0, and n o nonzero polynomial of lower degree in F[x] has this p r o p erty. W e m a y assume that the polynomial p (x) in this definition is monic, for we could divide this polynomial by its leading coefficient to obtain a monic polynomial q(x) in F[x], of the same degree as p(x), and such that q(a) = 0. W e henceforth assume that this polynomialp(x) is m o n i c ; we call it the minimal polynomial for a over F. L e m m a 5.3.4. Let a G K b e algebraic over F with minimal polynomial p(x) in F[x]. T h e n p(x) is irreducible in F[x].
196
Fields
Ch. 5
Proof. Suppose that p(x) is n o t irreducible in F[x]; t h e n p(x) = f(x)g(x) where f(x) a n d g ( x ) are in F[x] and each has positive degree. Since 0 = p(a) = / ( a ) g ( a ) , a n d since f(a) a n d g(o) are in the field K, w e conclude that f(a) = 0 or g ( a ) = 0, both of which are impossible, since b o t h f(x) a n d g(x) are of lower degree t h a n / f x ) . T h e r e f o r e , i s irreducible in F[x]. • Let a G K be algebraic of degree n over F and let p (x) G F[.r] b e its minimal polynomial over F. Given f(x) G F[x], t h e n / ( x ) = q(x)p(x) + r(x), where q(x) and r(x) are in F[x] a n d = 0 or deg r(x) < deg/?(x) follows from the division algorithm. T h e r e f o r e , / ( a ) = q(a)p(a) + r(a) = r ( a ) , since p(a) = 0. In short, any polynomial expression in a over F can b e expressed as a polynomial expression in a of degree at most n  1. Let F[a] = {/(a)  f(x) G F[x]). W e claim that F[a] is a subfield of # that contains both F and a, and that [ F [ a ] : F] = n. By the remark m a d e above, F[a] is spanned over F by 1, a, a , . . . , A " , so is finite dimensional over F. Moreover, as is easily verified, F[a] is a subring of K; as a subring of K, F[a] is an integral domain. Thus, by T h e o r e m 5.3.3, F[a] is a field. Since it is spanned over F by 1, a, a , . . . , a " , we have that [F[a] :F] < 72. T o show that [F[a]: F] = n we must merely show that 1, «, a , . . . , a"~ are linearly independent over F. But if cr + a a + • • • + e v ^ a " = 0, with t h e a, in F, t h e n q(a) = 0, where q(x) = a + ape + • • • + a _ x"~ is in F[x]. Since q(x) is of lower degree than p(x), which is the minimal polynomial for a in F[x], we are forced t o conclude that q(x) = 0. This implies that a = a = ••• = « „ _ ! = 0. Therefore, 1, a, a , . . . , a " are linearly independent over F a n d form a basis of F[a] over F. Thus [ F [ a ] : F ] = n. Since F[a] is a field, not merely just a set of polynomial expressions in a, we shall denote F[a] by F ( a ) . N o t e also that if M is any field that contains b o t h F and a, then M contains all polynomial expressions in a over F, hence M D F ( a ) . So F ( a ) is f/ze smallest subfield of K containing both F and a.
2  1 2  1 2 l  1 0 x x Q n x 0 x 2  1
Definition. a to F.
F ( a ) is called the field or extension obtained by
adjoining
W e n o w summarize. T h e o r e m 5.3.5. L e t K D F and suppose that a in K is algebraic over F of degree n. T h e n F ( a ) , the field obtained by adjoining a to F , is a finite extension of F, and [ F ( a ) : F ] = n. Before leaving T h e o r e m 5.3.5, let's look at it in a slightly different way. Let F[x] b e the polynomial ring in x over F , and let M = (p(x)) b e the ideal
Sec. 3
Field Extensions
197
of F[x] g e n e r a t e d by p(x), t h e minimal polynomial for a in K over F. By L e m m a 5.3.4, p(x) is irreducible in F[x]; hence, by T h e o r e m 4.5.11, M is a maximal ideal of F[x]. Therefore, F[x]/(p(x)) is a field by T h e o r e m 4.4.2. Define t h e mapping i/>: F[x] — > by \f>(f(x)) = / ( a ) . T h e mapping i/> is a h o m o m o r p h i s m of F[x] into /C, and t h e image of F[x] in /C is merely 7 ( a ) by the definition of 7 ( a ) . What is the kernel of (/>? It is by definition / = {f{x) G F[x]  «/>(/(x)) = 0}, and since we k n o w i{,(f(x)) = / ( a ) , / = {fix) G 7[x]  f(a) = 0}. Since p ( x ) is in / and p(x) is t h e minimal polynomial for a over F, p (x) is of the lowest possible d e g r e e a m o n g the elements of 7. Thus / = ip(x)) by the proof of T h e o r e m 4.5.6, and so / = M. By t h e First H o m o m o r p h i s m T h e o r e m for rings, F[x]IM — image of F[x] under if/ — Fid), and since F[x]/M is a field, we have that F{a) is a field. W e leave t h e proof, from this point of view, of [ 7 ( a ) : 7 ] = d e g p ( x ) to t h e reader.
PROBLEMS
1. Show that t h e following n u m b e r s in C are algebraic n u m b e r s . (a) V 2 + V 3 . (b) V 7 + #12.
(c) 2 + i V 5 . (d) cos(27r//c) + i sin(2ir/k), k a positive integer. 2. D e t e r m i n e the degrees over < of t h e n u m b e r s given in Parts (a) a n d (c) Q of P r o b l e m 1. 3. W h a t is t h e degree of cos(27r/3) + i sin(27r/3) over Q ? 4. W h a t is t h e degree of COS(2TT/8) + i sin(27r/8) over Q ? 5. If p is a p r i m e number, prove that the degree of cos(27r/p) + / sin(27r/p) over Q is p — 1 and that fix) = 1 + x + x
2
+ ••• + x ^
1
is its minimal polynomial over Q. 6. (For those who have had calculus) Show that 1
e = 1 + +
1
+
. . . +
1+ . . . m
1! is irrational.
2!
7. If a in K is such that a is algebraic over t h e subfield 7 of K, show that a is algebraic over 7 .
2
198
Melds
Ch.
5
8. M F C if and f(a) is algebraic over F, where / ( x ) is of positive degree in F[x] and a G if, prove that a is algebraic over F. 9. In the discussion following T h e o r e m 5.3.5, show that F[x]/M is of degree n = deg TJ>(X) over F, and so [F(a): F ] = n = deg/?(*)• 10. Prove that cos 1° is algebraic over Q. (1° = one degree.) 11. If a E if is transcendental over F, let F(a) = {f(a)/g(a) \ f(x), g(x) + 0 G F[x]}. Show that F(a) is a field and is the smallest subfield of if containing both F and a. 12. If a is as in Problem 11, show that F(a) — F ( x ) , where F ( x ) is the field of rational functions in x over F. 13. Let if be a finite field and F a subfield of if. lf[K:F] elements, show that if has q" elements. = n and F has q
14. Using the result of P r o b l e m 13, show that a finite field has p" elements for some prime p and some positive integer n. 15. Construct two fields if and F such that if is an algebraic extension of F but is not a finite extension of F .
4. FINITE E X T E N S I O N S
W e continue in the vein of the preceding section. Again K D F will always d e n o t e two fields. Let F ( i f ) b e the set of all elements in if that are algebraic over F. Certainly, F C E(K). O u r objective is to prove that E(K) is a field. Once this is d o n e , we'll see a little of how E(K) sits in if. Without further ado we proceed to T h e o r e m 5.4.1. F ( i f ) is a subfield of if.
Proof. What we must show is that if a, b G if are algebraic over F, t h e n a ± b, ab, and alb (if 0) are all algebraic over F. This will assure us that F ( i f ) is a subfield of if. We'll do all of a ± b, ab, and alb in one shot. Let if o = F{a) be the subfield of K obtained by adjoining a t o F. Since a is algebraic over F, say of degree m, then, by T h e o r e m 5.3.5, [K : F ] = m. Since Z is algebraic over F and since if contains F, we certainly have t h a t b ? is algebraic over K . If 6 is algebraic over F of degree n, t h e n it is algebraic over if of degree at most n. T h u s if! = K (b), the subfield of i f obtained by adjoining b to K , is a finite extension of K and [ i f : if ] < n. Thus, by T h e o r e m 5.3.1, [if, : F ] = [if : if ][if : F ] < 777/7; that is, if, is a finite extension of F. A s such, by T h e o r e m 5.3.2, K is an algebraic exten0 0 0 0 0 0 0 : 0 2 0 0 t
Sec. 4
Finite E x t e n s i o n s
199
sion of F, so all its elements are algebraic over F. Since a E K C K and 6 £ K , then all of the elements a ± b, ab, alb are in K , hence are algebraic over F. This is exactly what we wanted. T h e t h e o r e m is proved. •
0 x x x
If we look at the proof a little m o r e carefully, we see that we have actually p r o v e d a little more, namely the Corollary. If a and b in K are algebraic over F of degrees m a n d n, respectively, t h e n a ± b, ab, and alb (if b ¥= 0) are algebraic over F of degree at most mn. A special case, but one worth noting and recording, is the case K = C and F = <Q. In that case we called the algebraic elements in C over Q t h e algebraic numbers. So T h e o r e m 5.4.1 in this case becomes T h e o r e m 5.4.2. T h e algebraic n u m b e r s form a sublield of C.
For all we know at the m o m e n t , t h e set of algebraic n u m b e r s may very well be all of C. This is n o t the case, for transcendental n u m b e r s do exist; we show this to be true in Section 6 of C h a p t e r 6. W e r e t u r n t o a general field K. Its subfield E(K) has a very particular quality, which we prove next. This p r o p e r t y is that any element in K which is algebraic over E(K) must already b e in ELK). In order not to digress in the course of t h e proof we are about to give, we introduce the following notation. If a , a , . . . , a are in K, t h e n F(a ,.. ., a„) will be the field obtained as follows: K = F(a ), K = Ki(a ) = F(a , a ), K = K (a ) = F(a , a , a ), ... , K„ = K _ (a ) = F(a , a ,. .., a„). W e now prove
x 2 n x x x 2 2 x 2 3 2 3 x 2 3 n x n x 2
T h e o r e m 5.4.3.
If u in K is algebraic over E(K),
then u is in
E(K).
Proof. To prove the theorem, all we must d o is show that u is algebraic over F; this will put u in E(K), and we will be done. Since u is algebraic over E(K), t h e r e is a nontrivial p o l y n o m i a l / ( x ) = x" + a x"~ + a x"~ + • • • + a„, where a , a ,. . ., a„ are in E(K), such that / ( « ) = 0. Since a , a ,. .., a„ are in E(K), they are algebraic over F of degrees, say, m , i , • • • > «> respectively. W e claim that [F(a ,. .., a„): F] is at most m m • • • m„. To see this, merely carry out n successive applications of T h e o r e m 5.3.1 to the sequence K , K ,..., K„ of fields defined above. W e leave its proof to the reader. Thus, since u is algebraic over the field K = F(a , a , . . . , a ) [after all, the polynomial satisfied by u is f(x) =
l 2 x 2 t 2 x 2 m W7 x x l 2 x 2 n x 2 n
200
Fields
n l
Ch. 5
x" + a x ~ + ••• + «„, which has all its coefficients in F(a , a ,. • ., a,,)], the field K (it) is a finite extension of K„, and since K is a finite extension of F, we have, again by T h e o r e m 5.3.1, that K„(u) is a finite extension of F. B e cause u G K„(u), we obtain from T h e o r e m 5.3.2 that u is algebraic over F. This puts u in E(K) by the very definition of E(K), thereby proving the t h e o rem. •
x x 2 n n
T h e r e is a famous t h e o r e m d u e to Gauss, often referred t o as the Fundamental Theorem of Algebra, which asserts (in terms of extension) that the only finite extension of C , the field of complex numbers, is C itself. In reality this result is not a purely algebraic one, its validity depending heavily on topological properties of the field of real numbers. Be that as it may, it is an extremely important t h e o r e m in algebra and in many other parts of m a t h e matics. T h e formulation of the F u n d a m e n t a l T h e o r e m of A l g e b r a in terms of the nonexistence of finite extensions of C is a little different from that which is usually given. T h e most frequent form in which this famous result is stated involves the concept of a root of a polynomial, a concept we shall discuss at some length later. In these terms the F u n d a m e n t a l T h e o r e m of A l g e b r a b e comes: A polynomial of positive degree having coefficients in C has at least one root in C . T h e exact m e a n i n g of this statement and its equivalence with the other form of the t h e o r e m stated above will b e c o m e clearer later, after the development of the material on roots. A field L with the property of C described in the paragraphs above is said to be algebraically closed. If we grant that C is algebraically closed (Gauss' T h e o r e m ) , then, by T h e o r e m 5.4.3, we also have The field of algebraic numbers is algebraically closed.
PROBLEMS
1. Show that a = V 2 — V 3 is algebraic over < of degree at most 4 by exQ hibiting a polynomial f(x) of degree 4 over Q such that f(a) = 0. 2. If a and b in K are algebraic over F of degrees m and n, respectively, a n d if m and n are relatively prime, show that [F(a, b): F] = mn. 3. If a G C is such that p (a) = 0, where p(x) = x
5
+ Vix
3
+ V5x
2
+ Vlx
+
VLT,
show that a is algebraic over < of degree at most 80. Q
Sec. 5
Constructibility
201
4. If K D F is such that [K: F] = p, p a prime, show that K = F(a) for every a in K that is not in F. 5. If [K:F] = 2" and T is a subfield of K containing F, show that [T:F]= for some m £ n. 2"'
6. Give an example of two algebraic n u m b e r s a and £ of degrees 2 and 3, re> spectively, such that ab is of degree less than 6 over Q. 7. If/v D F a r e fields and a , . . . , a„ are in show that F(a . F(fl , . . . , a„(„)) for any p e r m u t a t i o n cr of 1, 2 , . . . , «.
x ly tT(1)
. ., a„) equals
5.
CONSTRUCTIBILITY
In ancient Greece, unlike in the other cultures of the time, the Greek m a t h e maticians w e r e interested in mathematics as an abstract discipline r a t h e r than as a pragmatic bag of tricks to do accounts or t o carry out measurements. They developed strong interests and results in n u m b e r theory and, most especially, in geometry. In these areas they posed penetrating questions. T h e questions they asked in g e o m e t r y — t w o of which will m a k e u p the topic treated h e r e — a r e still of interest and substance. The English m a t h e matician G. H. Hardy, in his sad but charming little b o o k A Mathematician's Apology, describes the ancient G r e e k mathematicians as "colleagues from another college." Two of these G r e e k questions will b e our concern in this section. But, as a m a t t e r of fact, the answer to both will emerge as a consequence of the criterion for constructibility, which we will obtain. W e state these questions now and will explain a little later what is entailed in them. QUESTION 1 Can one duplicate a cube using just straightedge and compass? (By duplicating a cube, we m e a n doubling its volume.) QUESTION 2 Can one trisect an arbitrary angle using just straightedge and compass?
Despite the seemingly infinite n u m b e r of angletrisectors that crop up every year, the answer to both questions is " n o . " A s we shall see, it is impossible to trisect 60° using just straightedge and compass. Of course, some angles are trisectable, for instance, 0°, 90°, 145°, 180°,. . . , but most angles (in a very precise meaning of "most") are not.
202
Fields
Ch. 5
Before getting to t h e exact m e a n i n g of the questions themselves, we want to spell out in explicit terms exactly what the rules of the g a m e are. By a straightedge we do not mean a rider—that is, an i n s t r u m e n t for m e a suring arbitrary lengths. No! A straightedge is merely a straight line, with n o quantitative or metric properties attributed to it. W e are given a line s e g m e n t — t o which we assign length 1—and all o t h e r lengths that we get from this must b e obtainable merely employing a straightedge a n d compass. Let us call a n o n n e g a t i v e real n u m b e r , b, a constructible length if, b y a finite n u m b e r of applications of t h e straightedge and c o m p a s s and t h e points of intersection o b t a i n e d b e t w e e n lines and circles so c o n s t r u c t e d , we can construct a line segment of length b, starting out from t h e line segm e n t we have assigned length 1. F r o m our high school geometry we recall some things we can d o in this framework. 1. Whatever length we construct o n one line can be constructed on any other line by use of the compass acting as a transfer agent. 2. We can draw a line parallel to a given line that goes t h r o u g h a given point. 3. W e can construct a length n for any nonnegative integer n.
F r o m these and by using results about the similarity of triangles, we can construct any nonnegative rational length. W e don't do that at this m o m e n t for it will come out as a special case of what we are about to do. We claim the following properties: 1. If a and b are constructible lengths, then so is a + b. For if AB is a length segment of length a and CD is one of length b, we can transfer this line segment CD, by m e a n s of a compass, to obtain the line ABE, w h e r e AB is of length a and BE is of length b. Thus the line segment AE is of length a + b. If b > a, how would you construct b — al 2. If a and 6 are constructible lengths, then so is ab. W e m a y assume that < 7 ^ 0 and b 0, otherwise, the statement is trivial. Consider the following diagram
L
j
B
Sec. 5
Constructibility
203
W h e r e L and L are two distinct lines intersecting at P, and such that PA has length a, PB has length b, and PJ has length 1. Let L be the straight line t h r o u g h ./ and A and L the line parallel to L passing through B. If C is the point of intersection of L and L , we have the diagram
x 2 3 4 3 x 4
All of these constructions can b e carried out by straightedge and compass. F r o m elementary geometry the length of PC is ab. Therefore, ab is constructible. 3. If a and b are constructible and b + 0, t h e n alb is constructible. Consider the diagram
L
L
w h e r e P, A, B, J, L and L are as in Property 2 above. Let L be the line through A and B and let L be the line through / parallel to L . If D is the point of intersection of L and L , then, again by elementary geometry, the length of PD is alb. W e stress again that all the constructions made can be carried out by straightedge and compass. Of course, this shows that the nonnegative rational numbers are constructible lengths, since they are quotients of nonnegative integers, which we know to b e constructible lengths. B u t t h e r e are other constructible lengths, for instance, the irrational n u m b e r V 2 . Because we can construct by straightedge and compass the rightangle triangle
u 2 5 6 s x 6
204
Fields
Ch. 5
with sides AB and BC of length 1, we know, by the Pythagorean T h e o r e m , that AC is of length V 2 . So V 2 is a constructible length. In Properties 1 t o 3 we showed that t h e constructible lengths almost form a field. W h a t is lacking is the negatives. T o get a r o u n d this, we m a k e the Definition. T h e real n u m b e r a is said to be a constructible \a\, the absolute value of a, is a constructible length. number if
A s far as we can say at t h e m o m e n t , any real n u m b e r might b e a constructible one. W e shall soon have a criterion which will tell us that certain real numbers are not constructible. F o r instance, we shall be able to deduce from this criterion that both V 2 and cos 2 0 ° are not constructible. This in turn will allow us t o show that t h e answer t o both Questions 1 a n d 2 is " n o . " But first we state
T h e o r e m 5.5.1. of real numbers.
T h e constructible n u m b e r s form a sublield of the field
Proof. Properties 1 to 3 almost do the trick; we must adapt Property 1 slightly t o allow for negatives. W e leave the few details to the reader. • O u r next goal is to show that a constructible n u m b e r must b e an algebraic n u m b e r — n o t any old algebraic n u m b e r , but one satisfying a r a t h e r stringent condition. N o t e , first, that if a s 0 is a constructible n u m b e r , then so is \fa. Consider the diagram
It is of a semicircle of radius (a + 1)12, center at C, AD is of length a, DB is of length 1, and DE is t h e perpendicular to AB at D, intersecting the circle at E. All this is constructible by straightedge and compass. F r o m elementary geometry we have that DE is of length V a , hence \/~a is constructible. W e n o w head for the necessary condition that a real n u m b e r b e constructible. L e t K be the field of constructible numbers, a n d let K b e a subfield of K. By the plane of K we shall m e a n the set of all points (a, b) in the real Euclidean plane whose coordinates a and b are in K . If (a, b) a n d (c, d)
0 Q 0
Sec. 5
Constructibility
205
are in t h e plane of if , then the straight line joining t h e m has the equation (y  b)/(x  a) = (b  d)/(a  c), so is of the form ux + vy + w = 0, w h e r e u, v, and w are in if . Given two such lines upc + v y + w = 0 and u x + v y + M2 = 0, where u , v , w and u , v , w are all in if , either they are parallel or their point of intersection is a point in if . (Prove!) Given a circle whose radius r is in if and whose center (a, b) is in t h e plane of if , then its equation is (x — a) + (y — b) = r , which we see, on expanding, is of the form x + y + dx + ey + / = 0, where d, e, and / are in if . T o see where this circle intersects a line in the plane of if , ux + vy + w = 0, we solve simultaneously the equations of the line and of the circle. F o r instance, if v # 0, then y = (ux + w)/v; substituting this for y in the equation of the circle x + y + dx + ey + f = 0 leads us to a quadratic equation for the xcoordinate, c, of this intersection point, of the form c + s c + s = 0, with s and s in K . By the quadratic formula, c = ( — s ± Vsf 4s )/2, and if the line and circle intersect in the real plane, t h e n s\  4s > 0. If s = s\ — 4s > 0 and if K = if()(Vi'), then we see that the xcoordinate, c, lies in if j . If Vs G K , then K = K ; otherwise, [K : K ] = 2. Since t h e ycoordinate d = (—uc + w)/v, we have that d is also in K . Thus the intersection point (c, d) lies in the plane of K w h e r e [K : K ] = 1 or 2. The story is similar if v = 0 and u i= 0. Finally, to get the intersection of two circles x + y + dx + ey + f = 0 and x + y + gx + hy + k = 0 in the plane of if , subtracting one of these equations from the other gives us the equation of the line in the plane of if , (d  g)x + (e  h)y + (f  k) = 0. So to find the points of intersection of two circles in the plane of K is the same as finding the points of intersection of a line in the plane of if with a circle in that plane. This is precisely the situation we disposed of above. So if the two circles intersect in the real plane, their points of intersection lie in the plane of an extension of if of degree 1 or 2. To construct a constructible length, a, we start in the plane of <Q, the rat i o n a l ; the straightedge gives us lines in the plane of Q, and the compass circles in the plane of Q. By the above, these intersect at a point in the plane of an extension of degree 1 or 2 of Q. T o get t o a, we go by this procedure from the plane of Q t o that of L say, w h e r e [L :Q] = 1 or 2, then to that of L , w h e r e [L : L ] = 1 or 2, and continue a finite n u m b e r of times. W e get, this way, a finite sequence < = L C L C • • • C L of fields, where each Q [Lj: L,_!] = 1 or 2 and w h e r e a is in L„. By T h e o r e m 5.3.1, [L„ : Q] = [L„ : L ^ j j f L ^ j : L „ _ ] • • • [L : Q] and since each of [ L , : L ^ ] = 1 or 2, we see that [L„ : Q] is a power of 2. Since a G L„, we have that Q(a) is a subfield of L„, hence by the Corollary to Theorem 5.3.1, [Q(a): Q ] must divide a power of 2, hence [Q(a): < > = 2"' for Q] some nonnegative integer m. Equivalently, by T h e o r e m 5.3.5, the minimal
0 0 x x ; 2 2 x x x 2 2 2 0 0 0 2 2 2 0 2 2 0 0 2 2 2 x 2 x 2 0 x 2 2 2 x 0 x 0 x 0 x x x 0 2 2 2 2 0 0 0 0 0 u t 2 2 x 0 x n 2 x
206
Fields
Ch. 5
polynomial for a over Q must have degree a p o w e r of 2. This is a necessary condition that a be constructible. W e have proved the important criterion for constructibility, namely T h e o r e m 5.5.2. In order that the real n u m b e r a be constructible, it is necessary that [Q(a) : < > b e a p o w e r of 2. Equivalently, the minimal polynoQ] mial of a over 0 must have degree a p o w e r of 2. T o duplicate a cube of sides 1, so of volume 1, by straightedge and compass would require us to construct a cube of sides of length 6 whose volu m e would be 2. B u t the volume of this cube would be 6 , so we would have to be able to find a constructible n u m b e r b such that b = 2. Given a real n u m b e r b such that b = 2, then its minimal polynomial over Q is p(x) = x  2, for this polynomial is monic and irreducible over Q (if you want, by the Eisenstein Criterion), and p(b) = 0. Also, as is clear to the eye, p(x) is of degree 3. Since 3 is not a power of 2, by T h e o r e m 5.5.2, there is n o such constructible b. Therefore, the question of the duplication of the cube by straightedge and compass has a negative answer. W e summarize this in
3 3 1 3
T h e o r e m 5.5.3. It is impossible to duplicate a cube of volume 1 by straightedge and compass. W e now have disposed of the classical Question 1, so we t u r n our attention to Question 2, the trisection of an arbitrary angle by straightedge and compass. If we could trisect the particular angle 60°, we would b e able to construct the triangle ABC in the diagram c
where 0 = 20° and AC is of length 1, by straightedge and compass. Since AB is of length cos 20°, we would h a v e that b = cos 20° is a constructible number. W e want to show that b = cos 20° is not a constructible n u m b e r by producing its minimal polynomial over <Q, and showing that this polynomial is of degree 3. T o this end we recall the tripleangle formula from trigonometry, namely that cos 3(f> = 4 c o s <j> — 3 cos <j>. If 6 = cos 20°, then, since cos(3 • 20°) = cos 60° =  , this trigonometric formula becomes Ab  3b = §, and so 8 6  6b  1 = 0. If c = 26, this becomes c  3c  1 = 0. If b is constructible, then so is c. B u t p(c) = 0, where p(x) = x  3x  1, and this
3 3 3 3 3
Sec.
6
R o o t s of P o l y n o m i a l s
207
polynomial is irreducible over ©. (Prove!) So p(x) is t h e minimal polynomial for c over Q. Because p(x) is of degree 3, and 3 is not a power of 2, by T h e o rem 5.5.2 we have that c is not constructible. So we cannot trisect 60° by straightedge and compass. This answers Question 2 in the negative. T h e o r e m 5.5.4. compass. It is impossible to trisect 60° by straightedge and
W e h o p e that this t h e o r e m will dissuade any reader from joining the hordes of angletrisectors. T h e r e are m o r e profitable and pleasanter ways of wasting o n e ' s time. T h e r e is yet another classical p r o b l e m of this kind to which the answer is " n o . " This is the question of squaring the circle. This question asks: Can we construct a square whose area is that of a circle of radius 1 by straightedge and compass? This is equivalent to asking whether x/TT is a constructible number. If this were the case, then since TT = (VTT) , the n u m b e r it would be constructible. But Lindemann proved in 1882 that TT is in fact transcendental, so certainly is not algebraic, and so cannot be constructible. Therefore, the circle of radius 1 cannot b e squared by straightedge and compass.
2
Of course, what we did above does not constitute a proof of the impossibility of squaring the circle, since we have p r e s u p p o s e d Lindemann's result without proving it. T o prove that TT is transcendental would take us t o o far afield. O n e might expect that it would b e easier to prove that TT is not constructible than to prove that it is not algebraic. This does not seem to b e the case. Until now all proofs that TT is not constructible go via the route of exploiting t h e transcendence of TT.
PROBLEMS
1. C o m p l e t e the proof of T h e o r e m 5.5.1. 2. Prove that x
3
 3x  1 is irreducible over Q. a & O does indeed give us V a .
3. Show that the construction given for \/a,
4. Prove that t h e regular h e p t a g o n (sevensided polygon with sides of equal length) is not constructible by straightedge and compass.
6 . R O O T S OF P O L Y N O M I A L S
Let F[x], as usual, be the polynomial ring in x over the field F and let K b e an extension field of F. If a E if and
f(x) = <x
Q
+
ctyX
+
• • • +
a„x'\
208
Fields
Ch. 5
then by f(a) we understand the element
f(a) = a
0
+ ajfl + • • • +
a„a"
in K. This is the usage we have m a d e of this notation t h r o u g h o u t this chapter. W e will now b e interested in those a's in K such that f(a) = 0. Definition. iff (a) = 0. T h e element A £ if is a root of the polynomial f(x) £ F[x]
In what we have d o n e u p until now we have always h a d an extension field K of F given t o us and we considered the elements in K algebraic over F, that is, those elements of K that are roots of n o n z e r o polynomials in F[x]. W e saw that if a £ K is algebraic over F of degree n—that is, if t h e minimal polynomial for a over F is of degree n—then [F(a): F] = n, w h e r e F(a) is t h e subfield of K obtained by adjoining a to F. W h a t we do now is turn t h e p r o b l e m around. W e n o longer will have t h e extension K of F at our disposal. In fact, our principal task will b e t o produce it almost from scratch. W e start with some polynomial f(x) of positive degree in F[x] as our only bit of data; our goal is to construct an extension field K of F in which f(x) will have a root. Once we have this construction of K u n d e r control, we shall elaborate on t h e general t h e m e , t h e r e b y obtaining a series of interesting consequences. Before setting off o n this search for the appropriate K, we must get some information about t h e relation between the roots of a given polynomial and the factorization of that polynomial. L e m m a 5.6.1. If a £ /. is a root of the polynomial f(x) £ F[x] of degree n, where L is an extension field of F, then f(x) factors in L[x] as fix) = (x  a)q(x), where q(x) is of degree n  1 in L[x]. Conversely, if f(x) = (x  a)q(x), w i t h / ( x ) , q(x), and a as above, then a is a root of f(x) in L. Proof. Since F C L, F[x] is contained in L[x]. Because a £ L, x  a is in L[x]; by the Division Algorithm for polynomials, we have f(x) = (x — a)q(x) + r(x), w h e r e q(x) and r(x) are in L[x] and where r(x) = 0 or d e g r ( x ) < deg(x  a) = 1. This yields that r(x) = b, some element of L. Substituting a for x in the relation above, and using t h e fact t h a t / ( G ) = 0, we obtain 0 = (a  a)q(a) + b = 0 + b = b; thus 6 = 0. Since r(x) = b = 0, we have what we wanted, n a m e l y / ( x ) = (x ~ a)q(x). For the statement that dagq(x) = n — 1 we note that since f(x) = (x  a)q(x), then, by L e m m a 4.5.2, n = d e g / ( x ) = deg(.v  a) + degq(x) = 1 + deg q(x). This gives us the required result, deg q(x) = n — 1. T h e converse is completely trivial. •
Sec.
6
Roots of Polynomials
209
O n e immediate consequence of L e m m a 5.6.1 is T h e o r e m 5.6.2. L e t f(x) in F[x] have degree n; then f(x) can have at most n roots in any extension, if, of F. Proof. W e go by induction o n n. If n = 1, then f(x) = ax + b, where a and b are in F and where a ¥= 0. T h u s the only root of f(x) is £>/A, an elem e n t of F. Suppose that the theorem is correct for all polynomials of degree k  1 over any field. Suppose that f(x) in F[x] is of degree k. If / ( x ) has no roots in if, then the theorem is certainly correct. Suppose, then, that a E if is a r o o t of / ( x ) . By L e m m a 5.6.1, / ( x ) = (x  a)q(x), where g(x) is of degree k — 1 in if [x]. A n y root i? in i f o f / ( x ) is either A or is a root of q(x), since 0 = f(b) = (7J> — a)q(b). By induction, o(x) has at most k — 1 roots in if, hence / ( x ) has at most k roots in if. This completes the induction and proves the theorem. • Actually, the proof yields a little m o r e . T o explain this "little m o r e , " we n e e d the notion of the multiplicity of a root. Definition. If if is an extension of /•'. t h e n the element a in K is a root of multiplicity k > 0 of fix), w h e r e f(x) is in F[x], if f(x) = (x  a) q(x) for some q(x) in ifjx] and x  a does not divide q(x) (or, equivalently, where q(a) * 0).
k
T h e same proof as that given for T h e o r e m 5.6.2 yields t h e sharpened version: Let f(x) be a polynomial of degree n in F[x]; then f(x) can have at most n roots in any extension field K of F, counting a root of multiplicity k as k roots. T h e o r e m 5.6.3. L e t f(x) in F[x] be monic of degree n and suppose that if is an extension of F in which / ( x ) has n roots, counting a root of multiplicity k as k roots. If these roots in if are a a , • • •, a,„, each having multiplicity k k , • •., k respectively, t h e n f(x) factors in if[x] as / ( x ) = (xa ) '(xa )^(xa ) .
u 2 u 2 m k k l 2 m
Proof. T h e proof is easy by making use of L e m m a 5.6.1 and of induction on n. W e leave the carrying out of t h e proof t o t h e reader. • Definition. W e say that / ( x ) in F[x) splits into linear factors over (or in) if if f(x) has the factorization in K[x] given in T h e o r e m 5.6.3. T h e r e is a nice application of T h e o r e m 5.6.3 t o finite fields. Let F be a finite field having q elements, a n d let a,, a ,. . ., be t h e nonzero ele2
210
Fields
Ch. 5
ments of F. Since these form a group of order q — 1 under the multiplication in F, by T h e o r e m 2.4.5 (proved ever so long ago), a ~ = 1 for any a ¥= 0 in F. T h u s t h e polynomial x''~  1 in F[x] has q  1 distinct roots in F. By T h e o r e m 5.6.3, the polynomial x ~  1 = (.r  a,)(x — a ) • • • (x — a ^ ). If we also consider 0, then every element a in F satisfies a' = a, so that the polynomial x — x has t h e q elements of F as its distinct roots. B y T h e o r e m 5.6.3 we have
q l ] q l 2 q l 1 q
T h e o r e m 5.6.4. factors in F[x] as
Let F be a finite field having q elements. T h e n x
q
—x
x'i  x = x(x  a )(x
x
«,)•••
(x 
A, _ )>
7 1
where a
u
a ,.
2
.., a  \ are the n o n z e r o elements of F, and
q
x"~'
 1 = (.v  fl,)(x  a ) • • • (x 2
a _ ).
q y
A very special case of this t h e o r e m is that in which F = Z , the integers modulo the prime p. H e r e q = p a n d ay,a ,..., a \ are just 1, 2 , . . . , p  1 in some order.
p 2 p
Corollary.
In Z [ x ] , the polynomial x ~
p
p
1
 1 factors as
 1 = (
X
 l ) ( x _ ) • • • ( *  (p  1)).
2
Try this out for p = 5,7, and 11. A s a corollary t o the corollary, we have a result in n u m b e r theory, known as Wilson's Theorem, which we assigned as Problem 18 in Section 4 of Chapter 2. Corollary. If p is a prime, t h e n (p — 1)1 = —1 m o d p.
Proof. By the Corollary above, x"1
 1 = 0 
1)(*  2) • • • (x  (p  1));
substituting x = 0 in this gives us  1 = (  l ) (  2 )  • • (  ( >  1)) = (  l ) '  l  2  • (p = (  l ) '  ^  1)! in Z . In the integers this translates into "congruent m o d p." Thus
p 1 l
 1)
(  l ) " " ^  1)! =  1 m o d p , and so (p  1)1 = (~l) m o d p . B u t (  1 ) ' ' =  1 moAp; p r o v e d Wilson's T h e o r e m . •
p
1
hence we h a v e
Sec.
6
R o o t s of P o l y n o m i a l s
211
W e change direction to consider the p r o b l e m mentioned at the beginning of this section: given/'(x) G F[x], to construct a finite extension K of F in which / ( x ) has a root. A s we shall see in a m o m e n t , this construction of K will be quite easy when we bring the results about polynomial rings proved in C h a p t e r 4 into play. However, to verify that this construction works will t a k e a bit of work.
T h e o r e m 5.6.5. Let ./•" be a field and fix) a polynomial of positive degree n in F[x]. T h e n there exists a finite extension K of F, with [K : F] < n, in which f(x) has a root. Proof. By T h e o r e m 4.5.12, f(x) is divisible in F[x] by some irreducible polynomial p(x) in F[x]. Since p(x) divides fix), degp(x) < d e g / ( x ) = n, a n d / ( x ) = p(x)q(x) for some polynomial q(x) in F[x]. If 6 is a root of p(x) in some extension field, then b is automatically a root of f(x), since f(b) = p(b)q(b) = Oq(b) = 0. So to prove the t h e o r e m it is enough to find an extension of F in which p (x) has a root. Because p(x) is irreducible in F[x], the ideal M = (p(x)) of F[x] generated by p(x) is a maximal ideal of F[x] by T h e o r e m 4.5.11. Thus by T h e o r e m 4.4.2, K = F[x]/M is a field. W e claim that this is the field that we are seeking. Strictly speaking, K does not contain F; as we now show, however, K does contain a field isomorphic to F. Since every element in M is a multiple in F[x] of p(x), every such nonzero element must have degree at least that of p(x). Therefore, M fl F = (0). Thus the homomorphism i/>: F[x] — A defined > T by <A(g(*)) g( ) + M for every g(x) in F[x], when restricted to F, is 1 — 1 on F. Therefore, the image F of F in K is a field isomorphic to F. W e can identify F, via ifj, with F and so, in this way, we can consider K an extension of F. D e n o t e x + M G K by a, so that ip(x) = a, a G K. W e leave it t o the reader to show, from the fact that ifi is a h o m o m o r p h i s m of F[x] onto A" with kernel M, that i//(g(x)) = g(a) for every g(x) in F[x]. W h a t is i(/(p(x))7 O n the one hand, since p(x) is in iTx], ip(p(x)) = O n the other hand, since p(x) is in M, the kernel of iV i]/(p(x)) = 0. E q u a t i n g these two evaluations of z, i/f(p(x)), we get that p(a) = 0. In other words, the element a = i//(x) in K is a root ofp(x). T o finish the proof, all we need is to show that [K : F] = deg p (x) < n. This came u p earlier, in the alternative proof we gave of T h e o r e m 5.3.5. T h e r e we left this point to b e proved by the reader. We shall be a little m o r e generous h e r e and carry out the proof in detail. Given h(x) in F[x], then, by the Division Algorithm, h(x) = p(x)q(x) + r(x) where q(x) and r(x) are in F[x], and r(x) = 0 or d e g r ( x ) < d e g p ( x ) . Going modulo M, we obtain that
= x
212
Fields
Ch.
5
ijj(h(x)) = 4>(p(x)q(x) + r(x)) = Mp(x)q(x))
+ <A(r(x))
= MPtoMqix)) = H'ix)) = r(a)
+
H'ix))
[since if/(p(x)) = p(a) = 0]. So, since every element in K = F[x]/M is (/)(/? (x)) for some /r(x) in F[x] and <//(h(x)) = r(a), we see t h a t every element of K is of the form r(a), where r(x) is in F[x] a n d deg r(x) < d&gp(x). If deg/?(x) = m, the discussion just m a d e tells us that 1, a, a ,... , a ' " span K over F. Moreover, these elements are linearly independent over F, since a relation of the type a + a a + • • • + a , , ,  ^ ' " = 0 would imply that g(a) = 0 where g(x) = a + a x + • • • + a  \ X ~ is in F[x]. This p u t s g(x) in M, which is impossible since g(x) is of lower degree than p(x), unless g(x) = 0. In other words, we get a contradiction unless a = a\ = • • • = = 0. So the elements 1, a, a , . . . , a"'~ are linearly independent over F . Since they also span K over F, they form a basis of K over F. Consequently,
2  1 Q x  1 0 x
m x m
2
l
0
dim AT = [K: F ] = m = degj? (.v) < n = d e g / ( x ) .
F
T h e theorem is proved.
•
We carry out an iteration of t h e argument used in t h e last proof to prove the important Theorem 5.6.6. Let f(x) G F[x] b e of degree n. Then there exists an extension K of F of degree at most nl over F such that f(x) has n roots, counting multiplicities, in K. Equivalently, f(x) splits into linear factors over K. Proof. W e go by induction on n. If n = 1, then f(x) = a + Bx, where a, B G F and where B ^ 0. T h e only root of f(x) is —a/B, which is in F. Thus K = F and [K : F) = 1. Suppose that t h e result is true for all fields for polynomials of degree k, and suppose t h a t / f x ) G F[x] is of degree k + 1. By T h e o r e m 5.6.5 there exists an extension K of F with [/Cj : F ] < k + 1 in which /'(x) has a r o o t a ^ Thus in K [x], f(x) factors a s / ( x ) = (x  a ) o ( x ) , where q(x) G K [x] is of degree k. By induction there exists a n extension K of K j of degree at most /c! over Zfi over which q(x) splits into linear factors. But t h e n / ( x ) splits into linear factors over K. Since [K: F ] = [K: KJK, : F] < (Ar + l)/c! = (k + 1)!, the induction is completed and the t h e o r e m is proved. •
x x 1 x
W e leave the subject of field extensions at this point. W e are exactly at what might be described as the beginning of Galois theory. Having a n exten
Sec.
6
R o o t s of P o l y n o m i a l s
213
sion K of F of finite degree over which a given polynomial f(x) splits into linear factors, t h e r e exists an extension of least degree enjoying this property. Such an extension is called a splitting field off(x) over F. O n e t h e n proceeds to prove that such a splitting field is unique u p to isomorphism. Once this is in h a n d the Galois theory goes into full swing, studying the relationship b e tween the group of automorphisms of this splitting field and its subfield structure. Eventually, it leads to showing, a m o n g m a n y other things, that t h e r e exist polynomials over the rationals of all degrees 5 or higher whose r o o t s cannot b e expressed nicely in terms of the coefficients of these polynomials. This is a brief and very sketchy description of w h e r e we can go from h e r e in field theory. But there is n o hurry. T h e readers should assimilate t h e material we have presented; this will p u t t h e m in a good position to learn Galois theory if they are so inclined.
PROBLEMS
1. Prove T h e o r e m 5.6.3. 2. If F is a finite field having the q — 1 n o n z e r o elements a , a , •.., prove that a a • • • a ^ = ( — l ) .
x 2 9 x 2 q x 4 3 2
a i,
q
3. Let Q be t h e rational field and let p(x) = x + x + x + x + 1. Show that there is an extension K of < > with [K: Q] = 4 over which p (x) splits Q into linear factors. [Hint: Find the roots ofp(x).] 4. If q(x) = x" + a x ~ + • • • + a„, a„ ¥= 0, is a polynomial with integer coefficients and if the rational n u m b e r r is a root of q (x), prove that ;• is an integer and r\a„.
x n l
5. Show that q(x) = x
3
 Ix + 11 is irreducible over Q.
p
6. If F is a field of characteristic p # 0, show t h a t (a + b) a and b in F.
= a + b for all
m
p
p
1. E x t e n d the result of P r o b l e m 6 by showing that (a + b) w h e r e m = p".
p
= a'" + b'",
8. Let F = Z , p a prime, and consider the polynomial x'" — x in Z [x], where m = p". Let K be a finite extension of Z over which x'" — x splits into linear factors. In K let K be the set of all roots of x'  x. Show t h a t Ko is a field having at most p" elements.
p p n Q
9. In P r o b l e m 8 show that K 14.)
0
has exactly p" elements. (Hint: See Problem
10. Construct an extension field K„ of < such that [K„ : Q] = n, for any Q n > 1.
214
Fields
Ch. 5
11. Define the mapping 8 : F[x] —» F[x] by S(a
a
+ ape + a x
2
2
+ • • • + a„x")
+ • • • +
= a + 2a x
y 2
iaiX "
1
1
+ • • •
+
na,,x"\
Prove that: (a) S(f(x) + g(x)) = 8(f(x)) (b) 8(f(x)g(x)) = f(x)8(g(x))
+ 8{g(x)). + 8(f(x))g(x)
for a l l / O ) and g(x) in F[x}. in F[x] such that
12. If F is of characteristic p j= 0, characterize all f(x) S(f(x)) = 0.
13. Show that if f(x) in F[x] has a root of multiplicity greater than 1 in some extension field of F, then f(x) and 8 ( / ( x ) ) are not relatively prime in F[x]. 14. If F is of characteristic p m = p", are distinct. 0, show that all the roots of x'" — x, where
15. If f(x) in F[x] is irreducible and has a root of multiplicity greater t h a n 1 in some extension of F, show that: (a) F must b e of characteristic p for some prime p. (b) f(x) = g(x ) for some polynomial g(x) in F[x].
p
In this final chapter we treat several u n r e l a t e d topics. One of these comes from group theory, and all the rest from the theory of fields. In handling these special topics, we draw from m a n y of t h e results and ideas developed earlier in the book. A l t h o u g h these topics are somewhat special, each of t h e m has results that are truly i m p o r t a n t in their respective areas. T h e readers who have m a n a g e d t o survive so far should have picked up a certain set of techniques, experience, and algebraic knowhow to be able to follow the material with a certain degree of ease. W e now feel free to treat the various matters at hand in a somewhat sketchier fashion than we have heretofore, leaving a few m o r e details to the r e a d e r to fill in. T h e material we shall handle does not lend itself readily to problems, at least not to problems of a reasonable degree of difficulty. Accordingly, we will assign relatively few exercises. This should come as a relief to those wanting to assimilate the material in this chapter.
1 . T H E S I M P L I C I T Y O F A„
In Chapter 3, where we discussed S„, the symmetric group of degree n, we showed that if n > 2, then S„ has a n o r m a l subgroup A„, which we called the alternating group of degree n, which is a group of order n\l2. In fact, A„ was merely the set of all even p e r m u t a t i o n s in S„.
215
216
Special Topics (Optional)
Ch. 6
In discussing A , we said that A„, for n £ 5, was a simple group, that is, that A„ has n o normal subgroups other than (e) a n d itself. W e p r o m i s e d there that we would prove this fact in Chapter 6. W e now m a k e good on this promise. To m a k e clear what it is that we a r e about to prove, we should p e r h a p s repeat what we said above a n d formally define what is m e a n t by a simple group.
n
Definition. A nonabelian group is said t o b e simple if its only n o r m a l subgroups are (e) and itself. W e impose t h e proviso that G b e nonabelian to exclude t h e trivial examples of cyclic groups of prime order from t h e designation "simple." T h e s e cyclic groups of prime o r d e r have n o nontrivial subgroups at all, so, perforce, they have no p r o p e r normal subgroups. A n abelian group with n o p r o p e r subgroups is easily seen to b e cyclic of prime order. W e begin with t h e very easy L e m m a 6.1.1. If n s 3 a n d T T are two transpositions in S„, then is either a 3cycle or t h e product of two 3cycles.
U
2
TT
X
2
Proof. If Tj = T , then T T = r\ = e a n d e is certainly t h e p r o d u c t of two 3cycles, for instance as e = (123)(132). If r # T , then they either have o n e letter in c o m m o n or n o n e . If they have o n e letter in c o m m o n , we m a y suppose, on a suitable renumbering, that T , = (12) a n d T = (13). B u t t h e n T , T = (12)(13) = (132), which is already a 3cycle. Finally, if T, a n d T have n o letter in common, we may suppose, without loss of generality, that r = (12) a n d T = (34), in which case T I T = (12)(34) = (142)(143), which is indeed t h e product of two 3cycles. T h e l e m m a is n o w proved. •
2 1 2 x 2 2 2 2 x
2 2
A n immediate consequence of L e m m a 6.1.1 is that for /z > 3 the 3cycles generate A„, the alternating group of degree n. T h e o r e m 6.1.2. If a is an even p e r m u t a t i o n in S„, where n > 3, t h e n a is a product of 3cycles. I n other words, t h e 3cycles in S„ generate A .
n
Proof. L e t cr G S„ b e an even permutation. By the definition of the parity of a permutation, a is a product of an even n u m b e r of transpositions. T h u s cr = TJTJ • • • T , ^ [ T ,  • • • T ~ i T is a product of 2m transpositions Ty, T , . . . , T . B y L e m m a 6.1.1, each T O / ^ T ^ is either a 3cycle o r a product
2 2 2 m 2 m 2 2 M
Sec. 1
T h e S i m p l i c i t y of A„
217
of two 3cycles. So we get that a is either a 3cycle or the product of at most 2m 3cycles. This proves the t h e o r e m . • W e now give an algorithm for computing t h e conjugate of any permutation in S„. Let a G S„, and suppose that a(i) = /'. What does TCTT~ look like if T G 5„? Suppose that T ( Z ) = s and T ( / ) = f; then TO~T~ *(s) = TO(T (.V)) = Tir(i) = r(/') = r. In other words, to compute T C T T replace every symbol in a by its image u n d e r T. F o r instance, if a = (123) and T = (143), then, since T(1) = 4, T(2) = 2, T(3) = 1, and T(4) = 3, we see that T C T T " = (421) = (142). Given two /ccycles, say (12 • • • k) a n d (i i • ' • 4)> then they are conjugate in S„ because if r is a permutation that sends 1 into ^ , 2 into i , • • • , k into i , then T(12 • • • / C ) T ~ = (iii , • • •, h) Since every permutation is t h e product of disjoint cycles and conjugation is an automorphism, we get, from the result for /ccycles, that to c o m p u t e T a r ' for any permutation cr, replace every symbol in a by its image u n d e r T. In this way we see that it is extremely easy to c o m p u t e t h e conjugate of any p e r m u t a t i o n . Given two p e r m u t a t i o n s a a n d o~ in S„, t h e n they are conjugate in S„, using t h e observation above, if in their decompositions into products of disjoint cycles they have the same cycle lengths and each cycle length with the same multiplicity. Thus, for instance, (12)(34)(567) and (37)(24)(568) are conjugate in S , but (12)(34)(567) and (37)(568) are not. Recall that by a partition of t h e positive integer n, we mean a decomposition of n as n = n + n + • • • + n , w h e r e 0 s n < n S ' " < n . If cr in S„ is the disjoint product of an «jcycle, an n cycle,..., an 7r cycle, then \ + i + • • • + n = n, and a p e r m u t a t i o n T i s conjugate to cr if and only if T is the disjoint product of cycles in the same way. Therefore, the n u m b e r of conjugacy classes in S„ is equal to t h e n u m b e r of partitions of n. F o r instance, if n = 4, then t h e partitions of 4 are 4 = 4, 4 = 1 + 3, 4 = 1 + 1 + 2, 4 = 1 + 1 + 1 + 1, and 4 = 2 + 2, which are five in n u m ber. Thus S has five conjugacy classes, namely t h e classes of (1234), (123), (12), e, and (12)(34), respectively. W e summarize everything we said above in three distinct statements.
1 _ 1
 1
1
1 2
2
!
k
2
1
x
2
8
x
2
k
1
2
k
2
/c
n
n
k
4
L e m m a 6.1.3. To find nrr ' in S, structure of a by its image u n d e r T. L e m m a 6.1.4. Two elements in S
n
replace every symbol in .the cycle
n
are conjugate if they have similar
decompositions as the product of disjoint cycles. L e m m a 6.1.5. T h e n u m b e r of conjugacy classes in S„ is equal t o t h e n u m b e r of partitions of n.
218
Special Topics (Optional)
Ch. 6
Clearly, from t h e results above, any two 3cycles in S„ a r e conjugate in S„. A 3cycle is an even p e r m u t a t i o n , so is in A„. O n e might w o n d e r if any two 3cycles are actually conjugate in t h e smaller group A„. F o r n "s 5 t h e answer is "yes," and is quite easy t o prove. L e m m a 6.1.6. gate mA„. If n ~ 5, then any two 3cycles in S„ are already conju
Proof. Let CT\ and r/ be two 3cycles in S„; by L e m m a 6.1.4 they a r e conjugate in S„. By renumbering, we m a y assume that a = (123) and a = T ( 1 2 3 ) T for some r G S„. If T is even, then we are d o n e . If r is odd, t h e n p = T(45) is even a n d p ( 1 2 3 ) p " = T ( 4 5 ) ( 1 2 3 ) ( 4 5 )  T ~ = T ( 1 2 3 ) T = cr . Therefore, <Tj and a are conjugate in A„. W e thus see that the lemma is correct. •
2 x 2 _ 1 1 1 1 _ 1 2 2
In S the two 3cycles (123) and (132) are conjugate in 5 b u t a r e n o t conjugate in A , which is a cyclic group of order 3. W e now prove a result that is not only important in g r o u p theory, but also plays a key role in field theory and t h e theory of equations.
3 3 3
T h e o r e m 6.1.7. group of S„ is A„.
If n > 5. t h e n t h e only nontrivial p r o p e r n o r m a l sub
Proof. Suppose that /V is a n o r m a l subgroup of S„ and N is neither (e) nor S„. Let cr # e be in N. Since the center of S„ is just (e) (See P r o b l e m 1) and the transpositions generate S„, t h e r e is a transposition r such that err i= ra. By L e m m a 6.1.4, r = C T T O  " is a transposition, so TT = TCTTO~ ^ e is in N, since cr G N and T U T = r c r r G N because N is normal in S . So N contains an element that is t h e product of two transpositions, namely TT .
1
1
x
X
 1
n
X
If T and Tj have a letter in c o m m o n , then, as we saw in t h e proof of L e m m a 6.1.1, TT is a 3cycle, hence TV contains a 3cycle. By L e m m a 6.1.4 all 3cycles in S„ are conjugate to TT S O must fall in N, by t h e normality of N in S„. Thus t h e subgroup of S„ g e n e r a t e d by t h e 3cycles, which, according t o T h e o r e m 6.1.2, is all of A„, lies in N. N o t e that up to this point we have not used that n > 5. W e may thus assume that T a n d T have n o letter in c o m m o n . Without loss of generality we m a y assume that r = (12) and T = ( 3 4 ) ; therefore, (12)(34) is in N. Since n > 5, (15) is in S„ hence (15)(12)(34)(15)^ = (25)(34) is also in JV; thus (12)(34)(25)(34) = (125) is in N. Thus in this case also, /Vmust contain a 3cycle. T h e argument above then shows that N D A„.
X X X 2 1
W e have shown that in both cases rV must contain A„. Since there a r e n o subgroups strictly b e t w e e n A a n d S a n d N ¥= S„, we obtain t h e desired result that N = A„. •
n n
Sec. 1
T h e S i m p l i c i t y of A
N
219
T h e result is false for n = 4; the subgroup T = {e, (12)(34), (13)(24), (14)(23)} V is a p r o p e r normal subgroup of S and is not A . W e n o w know all the normal subgroups of S„ when n a 5. Can we determine from this all the normal subgroups of A„ for n > 5? T h e answer is "yes"; as we shall soon see, A„is a simple group if n a 5 , The proof we give may strike m a n y as strange, for it hinges on the fact that 60, the order of A , is not a perfect square.
4 4 5
T h e o r e m 6.1.8,
T h e group A
5
5
is a simple group of order 60.
Proof. Suppose that A is not simple; t h e n it has a proper normal subgroup TV whose order is as small as possible. Let the subset T = [a E S \ T V c r C TV}, the normalizer of TV in S . Since TV is n o r m a l in A , we know that T D A . T is a subgroup of S , so if T + A , we would have that T = S . But this would tell us that TV is normal in 5 , which, by T h e o r e m 6.1.7, would imply that N D A , giving us that TV = A , contrary to our supposition that TV is a proper subgroup of A . So we must have T = A . Since (12) is odd, it is not in A , hence is not in T. Therefore, M = (12)7V(12) + TV. Since TV < A , we also have that M <1 A (Prove!), thus both M D TV and M T V = [mn j m G M,n G TV} are n o r m a l in A . (See Problem 9.) Because M + TV we have that M D TV + TV, and since TV is a minimal proper normal subgroup of A , it follows that M f l TV = (e). O n the other hand, ( 1 2 ) M / v ( 1 2 ) = ( 1 2 ) M ( 1 2 )  ( 1 2 ) M 1 2 ) = T V M (since (12)TV(12)" = M and ( 1 2 ) M ( 1 2 ) = TV) = M T V by the normality of M and TV in A . T h e r e fore, the element (12) is in the normalizer of M T V in S , and since MN is normal in A , we get, as we did above, the M T V is normal in S , and so MTV = A by T h e o r e m 6.1.7. Consider what we now have. B o t h M and TV are normal subgroups of A , each of order TV, and M T V = A and M D TV = (<?). W e claim, and leave to the reader, that MN must then have order  T V  . Since M T V = A , we obtain that 60 = yl 1 =  M T V  =  T V  . But this is sheer nonsense, since 60 is not the square of any integer. This establishes T h e o r e m 6.1.8. •
 1 5 5 5 s s 5 5 5 5 5 5 s 1 5 5 5 5 5 _1 1 _ 1 1 _ 1 5 5 5 5 5 5 5 2 5 2 5
To go from the simplicity of A to that of A„ for n s 5 is not too hard. N o t e that t h e argument we gave for A did not depend on 5 until the punch line " 6 0 is n o t the square of any integer." In fact, the reasoning is valid as long as we know that n\!2 is not a perfect square. Thus, for example, if n = 6, then 6!/2 = 360 is not a square, hence A is a simple group. Since we shall n e e d this fact in the subsequent discussion, we record it before going on.
s s 6
220
Special Topics (Optional)
Ch. 6
Corollary to the Proof of T h e o r e m 6.1.8.
A
6
is a simple group.
W e return to the question of whether or not n\l2 is a square. A s a matter of fact, it is not if n > 2. This can b e shown as a consequence of the beautiful t h e o r e m in n u m b e r theory (the socalled B e r t r a n d Postulate), which asserts that for m > 1 there is always a prime between m and 2m. Since we do not have this result at our disposal, we follow another road to show the simplicity of A„ for all n s 5. W e now prove this important t h e o r e m . Theorem 6.1.9. F o r all n > 5 the group A„ is simple.
Proof. By T h e o r e m 6.1.8 we may assume that n > 6. T h e center of A„ for n > 3 is merely (e). (Prove!) Since A„ is generated by the 3cycles, if cr + e is in A„. then, for some 3cycle T, err TCT. Suppose that N # (e) is a n o r m a l subgroup of A„ a n d that cr ^ e is in TV. Thus, for some 3cycle r, err rcr, which is to say, c r r c r T i= e. Because N is normal in A„, the element <TCT~ T~ is in N, hence OTO~ T~ is also in N. Since T is a 3cycle, so must c r r c r also be a 3cycle. Thus N contains the product of two 3cycles, and this product is not e. These two 3cycles involve at most six letters, so can be considered as sitting in A which, since n ^ 6, can be considered e m b e d d e d isomorphically in A„. (Prove!) But then N D A j= (e) is a normal subgroup of A , so by the Corollary above, N C\ A = A . Therefore, Nmust contain a 3cycle, and since all 3cycles are conjugate in A„ ( L e m m a 6.1.6), N must contain all the 3cycles in S„. Since these 3cycles generate A„, we obtain that N is all of A„, thereby proving the theorem. •
_ 1 _ 1
1 [ X 1
1
6
6
6
6
6
T h e r e are many different proofs of T h e o r e m 6.1.9—they usually involve showing that a normal subgroup of A„ must contain a 3cycle—which are shorter and possibly easier t h a n the one we gave. However, we like the bizarre twist in the proof given in that the whole affair boils down to the fact that 60 is not a square. W e r e c o m m e n d to the reader to look at some other proofs of this very important theorem, especially in a book on group theory. T h e A„ provide us with an infinite family of finite simple groups. T h e r e are several other infinite families of finite simple groups and 26 particular ones that do not belong to any infinite family. This determination of all finite simple groups, carried out in the 1960s and 1970s by a large n u m b e r of group theorists, is one of the major achievements of twentiethcentury m a t h ematics.
Sec. 2
Finite Fields I
221
PROBLEMS
* 1. Prove t h a t if n > 2, the center of S„ is (e). * 2 . P r o v e that if n > 3, the center of A
n
is (e).
3 . W h a t can you say about the cycle structure of the product of two 3cycles? 4. If m < /?, show that there is a subgroup of S„ isomorphic to S,„. 5. Show that an abelian group having n o p r o p e r subgroups is cyclic of p r i m e order. 6. H o w m a n y conjugacy classes are t h e r e in S ?
6
7 . If the elements a a ,..., a„ generate the group G and b is a noncentral element of G, prove that ba, # a b for some i.
u 2 {
8. If M < N and N < G, show that aMa~
1
is n o r m a l in TV for every a G G.
9. If M <\ G and N < G, show that MN is a normal subgroup of G. 10. If 77 > 5 is odd, show that the ncycles generate A„. 11. Show that the centralizer of (12 • • • k) in 5 „ has order k(n — k)\ and that (12 • • • k) has n\l(k(n  k)\) conjugates in S .
n
12. In the proof of T h e o r e m 6.1.8, show that \MN\ =
2. FINITE FIELDS I
\N\ .
2
O u r goal in this section and the next two is to get a complete description of all finite fields. W h a t we shall show is that the multiplicative group of nonzero elements of a finite field is a cyclic group. This we do in this section. In the next two, the objectives will be to establish the existence and uniqueness of finite fields having p" elements for any prime p and any positive integer 77. Some of the things we are about to do already came up in the problem sets in group theory and field theory as h a r d problems. The techniques that we use come from group theory and field theory, with a little number theory thrown in. W e recall what the Eider <pfunction is. W e define the Euler ^function by: cp(l) = 1 and, for 77 > 1, cp(n) is the n u m b e r of positive integers less than 77 and relatively prime to 77. W e begin with a result in n u m b e r theory whose proof, however, will exploit group theory. Before doing the general case, we d o an example. Let 77 = 12; then <p(12) = 4, for only 1, 5, 7, and 11 are less than 12 and relatively p r i m e to 12. W e compute <p(d) for all the divisors of 12. We have:
222
Special Topics (Optional)
Ch. 6
(p(l) = 1, ip(2) = 1, (f(3) = 2, <p(4) = 2, 9 ( 6 ) = 2, and cp{\2) = 4. N o t e that the sum of all <p(d) over all the divisors of 12 is 12. This is n o fluke but is a special case of T h e o r e m 6.2.1. divisors d of 77. Proof. Let G be a cyclic g r o u p of o r d e r 77 generated by t h e element a. If d  n, how m a n y elements of G have o r d e r dl lib = a" , t h e n all t h e solutions in G of .t'' = e are t h e powers e, b, b ,. . . , /3 ~' of o. H o w m a n y of these have order dl W e claim, and leave t o t h e reader, that b" h a s o r d e r d if and only if r is relatively p r i m e to d. So t h e n u m b e r of elements of o r d e r d in G, for every divisor d of n, is (p(d). E v e r y element in G has o r d e r some divisor of / I , so if we sum up t h e n u m b e r of elements of order d—namely ip(d)— over all d dividing 77, we account for each element of G once a n d only once. H e n c e 2^>(d) = n if we run over all the divisors d of 77. T h e t h e o r e m is n o w proved. •
ld 2 rf
If n > 1, then 2(p(d) = 77, where this sum runs over all
In a finite cyclic group of o r d e r 77 the n u m b e r of solutions of x = e, t h e unit element of G, is exactly d for every d that divides 77. W e used this fact in t h e proof of T h e o r e m 6.2.1. W e n o w prove a converse t o this, getting thereby a criterion for cyclicity of a finite group. T h e o r e m 6.2.2. L e t G be a finite group of order 77 with t h e p r o p e r t y that for every d that divides 77 t h e r e are at most d solutions of x = em G. T h e n G is a cyclic group. i Proof. Let 41(d) be t h e n u m b e r of elements of G of o r d e r d. By hypothesis, if a 6 G is of o r d e r d, t h e n all t h e solutions of x  e are t h e distinct powers e, a, a ...., a ~ , of which n u m b e r , <p(d) a r e of o r d e r d. So if there is an element of order d in G, then tzV(d) = <p(d). O n the o t h e r hand, if t h e r e is n o element in G of order d, then (/7(d) = 0. So for all d  n w e have that ip(d) s <p(d). H o w e v e r , since every element of G has some o r d e r d that divides T we have that 2i//(d) = n, where this sum runs over all divisors d of J 77. B u t
d d 2 d l
d
77 = 2 ^ ( d ) < Xcp(d) = 77 since each i/V(d) s <p(d). This gives us that Ei/f(d) with (/'(d) < <p(d), forces <//(d) = cp(d) for every particular, i/>(77) = 9 ( 7 7 ) s 1. W h a t does this tell number of elements in G of order n, and since tjj(n) ment 'a in G of order n. Therefore, t h e elements e, = X<p(d), which, t o g e t h e r d that divides n. Thus, in us? After all, ijj(n) is the ' 1 there must be an elea, a ,. . . , a"~ a r e all dis2 l
Sec. 2
Finite Fields I
223
tinct and are n in number, so they must give all of G. Thus G is cyclic with a as generator, proving the theorem. • Is there any situation where we can be sure that the equation x = e has at most d solutions in a given group? Certainly. If K* is the group of nonzero elements of a field u n d e r multiplication, then the polynomial x" ~ 1 has at most n roots in K* by T h e o r e m 5.6.2. So, if G C K* is a finite multiplicative subgroup of K*, then the n u m b e r of solutions of x = 1 in G is at most d for any positive integer d, so certainly for all d that divide the o r d e r of G. By T h e o r e m 6.2.2 G must be a cyclic group. W e have proved
d d
Theorem 6.2.3. If K is a field a n d A'* is the group of nonzero elements of K u n d e r multiplication, then any finite subgroup of K* is cyclic. A very special case of T h e o r e m 6.2.3, but at the m o m e n t the most imp o r t a n t case for us, is Theorem 6.2.4. If A is a finite field, t h e n K* is a cyclic group. T
Proof. K* is a finite subgroup of itself, so, by T h e o r e m 6.2.3, K* is cyclic. • A particular instance of T h e o r e m 6.2.4 is of great importance in n u m ber theory, where it is known as the existence of primitive roots mod p for p a prime. Theorem 6.2.5. If p is a prime, t h e n Z* is a cyclic group.
PROBLEMS
1. If A G G has order d, prove that a' also has order d if and only if r a n d d are relatively prime. 2. Find a cyclic generator (primitive root) for 2 . \ \ . 3. D o Problem 2 for Z f v 4. Construct a field K having nine elements and find a cyclic generator for the group K*. 5. If p is a prime and m = p , t h e n Z is not a field but the elements { [ A ]  ( A , p) = 1} form a group u n d e r the multiplication in Z,„. Prove that this group is cyclic of order p(p  1).
m 2
224
Special Topics (Optional)
Ch. 6
6. D e t e r m i n e all the finite subgroups of C*, where C is t h e field of complex numbers. In the rest of the problems h e r e cp will be the Euler ^function. 7. If p is a prime, show that <p(p") = p"^ (p
1
— 1).
8. If m and n are relatively p r i m e positive integers, prove that <p(mn) = <p(rri)cp(ri). 9. Using t h e result of Problems 7 and 8, find cp(n) in terms of the factorization of n into p r i m e p o w e r factors. 10. Prove that lim cp(n) = °°.
3. FINITE FIELDS II: E X I S T E N C E
Let if be a finite field. Then if must b e of characteristic p, p a prime, and K contains 0, 1, 2 , . . . , p  1, the p multiples of the unit element 1 of if. So if D Z , or, m o r e precisely, if contains a field isomorphic to Z . Since if is a vector space over Z and clearly is of finite dimension over Z , if [if: Z ] = n, then if h a s p " elements. T o see this, let v v , • • •, v„ be a basis of if over Z . T h e n for every distinct choice of ( a a , . . . , a ), where the a, are in Z , the elements
;; p p p p h 2 p
b 2 n
p
a Vi
L
+
av
2
2
+
• • • +
a„v
n
are distinct. Thus, since we can pick ( a , a , . . . , a„) in p" ways, if has p " elements. T h e multiplicative group K* of n o n z e r o elements of if is a group of order p" — 1. So, we have that a ' " = 1, where w = p", for every a in if, h e n c e a" = a. Since this is also obviously true for a = 0, we have that a'" = a for every a in K. Therefore, the polynomial x  x in Z [x] has m = /?" distinct roots in if, namely all the elements of K. Thus x'"  x factors in K[x] as
x 2
 1
1
m
p
x'"  x = (x  a )(x
L 2
 a ) • • • (x 2
a,„),
where a,, a , . . . , a,„ are the elements of if. Everything we just said we already said, in m o r e or less t h e s a m e way, in Section 6 of Chapter 5. Since we wanted these results to b e fresh in t h e r e a d e r ' s mind, we r e p e a t e d this material here. W e summarize what we just did in T h e o r e m 6.3.1. Let if b e a finite field of characteristic p, p a prime. T h e n if contains m = p" elements where n = [if: Z ], a n d the polynomial x'"  x in Z [x] splits into linear factors in K[x] as
p p
Sec, 3
Finite Fields II: Existence
225
x'"  x = (x  )(x
ai
 a ) • • • (x 2
a,„),
where a , a ,...,
x 2
a
m
are the elements of K.
T w o natural questions present themselves: 1. F o r what primes p and what integers n does there exist a field havingp" elements? 2. H o w m a n y nonisomorphic fields are there having p" elements? W e shall answer both questions in this section a n d the next. T h e answers will be 1. F o r any prime p and any positive integer n there exists a finite field having p" elements. 2. T w o finite fields having the same n u m b e r of elements are isomorphic. It is to these two results that we now address ourselves. First, we settle the question of the existence of finite fields. W e begin with a general r e m a r k about irreducible polynomials. L e m m a 6.3.2. Let F be any field and suppose that p(x) is an irreducible polynomial in F[x]. Suppose that q(x) in F[x] is such that in some extension field of F, p (x) and q (x) have a c o m m o n root. T h e n p (x) divides q (x) in F[x]. Proof. Suppose that p(x) does not divide q(x); since p(x) is irreducible in F[x], p(x) and q(x) must therefore b e relatively p r i m e in F[x]. Thus there are polynomials u (x) and v (x) in F[x] such that u(x)p(x) + v(x)q(x) = 1.
Suppose that the element a in some extension K of F is a root of both p (x) and q(x); thus p(a) = q(a) = 0. B u t then 1 = u(a)p(a) + v(a)q(a) = 0, a contradiction. So we get that p(x) divides q(x) in F[x]. • N o t e that we can actually prove a little m o r e , namely Corollary. If / ( x ) and g(x) in F(x) are not relatively prime in K[x], where K is an extension of F, t h e n they are not relatively prime in F[x]. Let F b e a field of characteristic p + 0. W e claim that the polynomial = x'"  x, where m = p", cannot have a multiple root in any extension
fix)
226
Special Topics (Optional)
Ch. 6
field K of F. D o you r e m e m b e r what a multiple root of a polynomial is? W e refresh your memory. If g(x) is in F[x] and if K is an extension field of F, then a in K is a multiple root of g(x) if g(x) = (x  afq(x) for some q(x) in AT[.v]. W e return to the polynomial f(x) = x'"  x above. Since f(x) = x(x'"~  1) and 0 is not a root of x ' " — 1, it is clearly true that 0 is a simple (i.e., not multiple) root o f / ( x ) . Suppose that a G K, K D F, is a root o f / ( x ) ; thus a'" = a. If y = x — a, then
[  1
fiy)
= )'"' ~ y = (
x
~ )'"
a
~ (
x
 a) = x'
n
 a'"  (x 
a)
(since we are in characteristicp = x'"  x (because a'" = a) = So fix) = f(y) = y"  y
1
0 and m = p")
fix).
= (x1
«)"'  (x 
a)
= (x  a)((x
 a ) ' " "  1),
and clearly this is divisible by x  a only to the first power, since x  a does not divide (x  a)" '  1. So a is n o t a multiple root of f(x). W e have proved
1 1
T h e o r e m 6.3.3. If n > 0, t h e n / ( x ) = x'" — x, where m = p", has n o multiple roots in any field of characteristic p. W e should add a word to the proof above to nail down the s t a t e m e n t of T h e o r e m 6.3.3 as we gave it. A n y field of characteristic p == 0 is an extension # of Z , and the polynomial f(x) is in Z [x]. So the argument above, with K any field of characteristic p and F = Z , proves the t h e o r e m in its given form. W e have exactly what we n e e d to prove the important
p p p
T h e o r e m 6.3.4. For any p r i m e p and any positive integer n t h e r e exists a finite field having p" elements. Proof. Consider the polynomial x'"  x in Z [x], w h e r e m = p". By T h e o r e m 5.6.6 there exists a finite extension K of Z such that in K[x] the polynomial x'" — x factors as
p p
x'"  x = (x  flj)(x  a ) • • • (x 2 lt 2 m
a„,),
where a a ,.. ., a are in K. By T h e o r e m 6.3.3, x'"  x does not have any multiple roots in K, hence the elements a a ,..., a are m = p" distinct elements. W e also know that a , a , • • •, a,„ are all the roots of x'"  x in K, since x'" — x is of degree m.
u 2 m ± 2
Sec. 4
1
Finite Fields III: Uniqueness
227
Let A = {a G K ] a" = a}; as we just saw, A has m distinct elements. W e claim that A is a field. If a, b G A, then a"' = a and £>'" = b, hence (a/?)'" = fl"'/3"' = ab, thus A/> G A. Because we a r e in characteristic p # 0 and m = /?", ( A + b)'" = fl'" + fo'" = a + b, h e n c e a + 6 is in A. Since A is a finite subset of a field and is closed with respect to sum and product, A must be a subfield of K. Since A h a s m = p" elements, A is thus the field whose existence was asserted in the statement of the theorem. With this the t h e o r e m is proved. •
PROBLEMS
* 1 . Give the details of the proof of the Corollary to L e m m a 6.3.2. T h e next two problems are a r e p e a t of ones given earlier in the b o o k . 2. If f(x) = a x" + fl,*"" + • • • + a„ is in F[x], let f'(x) b e the formal derivative of fix) defined by the following equation: f'(x) = «a x" + in  l ^ x " + • • • + ( «  i)a x"~'~ + • • • + a„_ . Prove that: (a) ifix) + gix))' =f\x)+ g'{x) (b) ( / ( x ) g ( x ) ) ' = f\x)g(x) + fix)g'ix) for a l l / 0 : ) and g(x) in F[x].
0 _ 1 0  2 i i y 1
* 3 . Prove that f(x) in F[x] has a multiple root in s o m e extension of F if and only if fix) a n d / ' ( x ) are not relatively prime. 4. If fix) = x" — x is in F[x], p r o v e t h a t / ( x ) does not have a multiple root in any extension of F if F is either of characteristic 0 or of characteristic p # 0, where p does not divide « — 1. 5. U s e the result of P r o b l e m 4 to give a n o t h e r proof of T h e o r e m 6.3.3. 6. If F is a field of characteristic p i= 0, construct a polynomial with multiple roots of the form x"  x, where p \ {n — 1). 7. If K is a field having p" elements, show t h a t for every m that divides n there is a subfield of K havingp'" elements.
4 . F I N I T E F I E L D S III: U N I Q U E N E S S
N o w that we know that finite fields exist having p" elements, for any p r i m e p and any positive integer n, we might ask: H o w many finite fields are there with p" elements? For this to m a k e any sense at all, what we a r e really asking is: H o w m a n y distinct nonisomorphic fields are there with p" elements? The answer to this is short and sweet: one. W e shall show here that any two finite fields having the same n u m b e r of elements are isomorphic.
228
Special Topics (Optional)
Ch. 6
L e t K and L b e two finite fields having p" elements. Thus K and L are both vector spaces of dimension n over Z . A s such, K a n d L are isomorphic as vector spaces. O n t h e other hand, K* and L * are b o t h cyclic groups of order p" — 1 by T h e o r e m 6.2.4; hence K* and L* are isomorphic as multiplicative groups. O n e would imagine that o n e could p u t these t w o isomorphisms together to prove that K a n d L are isomorphic as fields. B u t it just isn't so. T h e proof does not t a k e this direction at all. B u t the finiteness of K and L together with these two isomorphisms (of two structures carried by K and L) do suggest that, p e r h a p s , K and L are isomorphic as fields. This is ind e e d the case, as we n o w proceed to show. W e begin with
p
L e m m a 6.4.1. If q(x) in Z [x] q(x)  (x  x), where m = p".
p m
is irreducible of degree n,
then
Proof. By T h e o r e m 4.5.11 the ideal (q(x)) of Z [x] g e n e r a t e d by q(x) is a maximal ideal of Z [x] since q(x) is irreducible in Z [x]. L e t A = Z [x)l(q(x)); by T h e o r e m 4.4.3, A is a field of degree n over Z , hence h a s p " elements. Therefore, u'" = u for every element u in A. Let a = x + (q(x)) be the coset of x in A = Z [x\l(q(x))\ thus q{a) = 0 and q(x) is the minimal polynomial for a over Z . Since a is in A, a " = a, so a is seen as a root of the polynomial x'"  x, where m = p". Thus x  x and q(x) have a common root in A. By L e m m a 6.3.2 we have that q(x) \ (x'"  x). •
p p p p p p 1 p m
W e are now in a position to prove the main result of this section. T h e o r e m 6.4.2. ff K and L are finite fields having the same n u m b e r of elements, then K and L are isomorphic fields. Proof. Suppose that K and L have p" elements. B y T h e o r e m 6.2.4, L* is a cyclic group generated, say, by the element b in L. T h e n , certainly, Z (b)—the field obtained by adjoining b t o Z —is all of L. Since [L : Z ] = n, by T h e o r e m 5.3.2 b is algebraic over Z of degree n, with n = d e g ( g ( x ) ) , where q(x) is the minimal polynomial in Z [x] for b, and is irreducible in Z [x]. T h e mapping i/>: Z [x] » L = Z (/b) defined by i/>(/(x)) = / ( & ) is a h o m o m o r p h i s m of Z [x] onto L with kernel (o(x)), the ideal of Z [x] generated by q(x). So L = Z [ x ] / ( ? ( x ) ) . Because o ( x ) is irreducible in Z [ x ] of degree «, by L e m m a 6.4.1 q(x) must divide x'"  x, where m = /J". However, by L e m m a 6.3.1, t h e polynomial x'" — x factors in K[x) as
p ;J P p p p p p p p p p
x
m
 x = (x  a )(x  « , ) • • • (x  «„,),
x
Sec. 5
Cyclotomic Polynomials
229
w h e r e a , a , • • • , a,„ are all the elements of K. Therefore, q(x) divides (x  a )(x — a ) • • • (x  a,„). By the Corollary t o T h e o r e m 4.5.10, q(x) cann o t be relatively prime to all the .v — a in K[x], hence for some 7, q(x) and x  cij have a c o m m o n factor of d e g r e e at least 1. In short, x  a, must divide q(x) in K[x], so q(x) = (x  cij)h(x) for some h(x) in K[x). T h e r e fore, q(cij) = 0.
x 2 x 2 t
Since q(x) is irreducible in Z [x] and a is a root of q(x), q(x) must be the minimal polynomial for aj in Z [x]. Thus Z (aj) — Z [x]/(q(x)) = L. This tells us, among other things, that we have [Z (aj) :Z ] = n, and since Zp(fly) C K and [K:Z ] = n we conclude that Z (aj) = if. Therefore, K = Zp(fly) — L. T h u s we get the result that we are after, namely, that K and L are isomorphic fields. This proves the theorem. •
p y p p p p p p p
Combining T h e o r e m s 6.3.4 and 6.4.2. we have
T h e o r e m 6.4.3. For any prime p and any positive integer n there exists, up to isomorphism, one and only one field having p" elements.
5. C Y C L O T O M I C P O L Y N O M I A L S
Let C be the field of complex n u m b e r s . A s a consequence of D e Moivre's T h e o r e m the complex n u m b e r B„ = cos 27r/« + i sin Irrln satisfies 6" = 1 and 6'" + 1 if 0 < m < n. W e called 0„ a primitive nth root of unity. The other primitive nth roots of unity are
where (/c, n) = 1 and 1 < /<: < n. Clearly, d satisfies the polynomial x"  1 in Q[x], where Q is the field of rational n u m b e r s . W e want to find the minimal (monic) polynomial for 6„ over Q. W e define a sequence of polynomials inductively. At first glance they might not seem relevant to the question of finding the minimal polynomial for 9„ over Q. It will turn out that these polynomials are highly relevant to that question for, as we shall prove later, the polynomial <j>„(x) that we are about to introduce is a monic polynomial with integer coefficients, is irreducible over Q, and, moreover, c6„(#„) = 0. This will tell us that (p„( ) i * desired monic minimal polynomial for 8 over Q. W e now go about the business of defining these polynomials.
n x s n e n
230
Special Topics (Optional)
Ch. 6
Definition. T h e polynomials <p„(x) are defined inductively by: (a) cp^x) = x  1. (b) If n > 1, then <p,,(x) = (x"  l)/TIcfc (x), where in the product in the denominator d runs over all the divisors of n except for n itself. These polynomials are called the cyclotomic polynomials and c6„(x) is called the nth cyclotomic polynomial.
rf
A t the m o m e n t it is not obvious that the <p„(x) so defined are even polynomials, nor do we, as yet, have a clue as to the n a t u r e of the coefficients of these polynomials 4> (x). All this will come in due time. But first we want to look at some early examples.
n
Examples 1. <p (x) = (x  l)/cMx) = (*  l ) / ( *  1)
2 2 2
= X + = X
2
1.
+ X +
2. <p (x) = (x 3
3
= (x 2
3
l)/(x  1)
4
1. l)(x
2
3. <f3 (x) = (x  l)/(<Mx)c/> (x)) = (x (x  l)/(x  1) = x + 1.
4 4 4 2 2 5 5 9 5 I
4
l)/((x 3
+ 1)) =
1.
4. ( x ) = ( x  l ) / c f e ( x ) = ( x  l ) / ( x  1) = x + x + x x — 1
6
5
+ X +

U
x
)
=
<t> (x)tex)<h(x)
x =
x
X
3
6
 1
2
(x  l)(x + l ) ( x
=
+ X +
+ X +
+
1)
1 , . = x 1
2 z

X
, . 1.
W e notice a few things about the polynomials above: 1. They are all monic polynomials with integer coefficients. 2. The degree of <f>„(x) is <p(n), where <p is the E u l e r ^function, for 1 < n < 6. (Check this out.) 3. Each cfe„(x), for 1 < n ^ 6, is irreducible in Q(x). (Verify!) 4. For 1 < n < 6, 6„ is a root of <p„(x). (Verify!) These few cases give us a hint as to what the general story might b e for all <f>„(x). A hint, yes, but only a hint. To establish these desired properties for c6„ (x) will take some work. To gain some further insight into these polynomials, we consider a par
Sec. 5
Cyclotomic Polynomials
231
ticular case, one in which n — p " \ w h e r e p is a prime. T o avoid c u m b e r s o m e subscripts, we shall d e n o t e </>„(x) by i/> (x), w h e r e n = p ' " . The prime p will be k e p t fixed in the discussion. W e shall obtain explicit formulas for t h e tl/('"\x)'s and d e t e r m i n e their basic properties. H o w e v e r , the m e t h o d w e use will not be applicable to the general case of <p (x) TO study the general situation will require a wider and d e e p e r set of techniques than those n e e d e d
(m) n
for
4i "\x).
(1)
(l
W e n o t e one simple example. If p is a p r i m e , the only divisor of p that is not p itself is 1. F r o m the definition of 4> (x) = i// (x) we have that
p
^\x)
= <f (x) =
p
= x"~ + • • • + X + 1.
l
N o t e that in studying the Eisenstein Criterion w e showed that this polynomial is irreducible in Q(x). W h a t can we say for t h e higher i/> (x)'s? L e m m a 6.5.1.
(
(m)
For all m > 1,
=
1
<l> "'\x) = ^"~i
x
+ "'~'
xP
+
x2
"""
i
+ •••
+
j^ *"'" .
1
1
Proof. W e go by induction on m. If m = 1, we showed above that <// (x) = (x  l ) / ( x  1) = 1 + x + x + • • • + x ~\ so t h e lemma is true in this case. Suppose that >p = (x'"'  l ) / ( x "  1) for all r < m. Consider Since the only p r o p e r divisors of p ' " are 1, p , p , . .., p " ' ~ \ from t h e definition of i//" (x) we have that
1 p (r) p r L 2 0
X
(1)
p
P"'
_
i
^'" ^
}
=
(x  1 ) I A ( X )
0 )
• • • ^"'^(jt)"
By induction, ip (x)
(,)
= (x>'  l ) / ( x
{ l
r
p r _ 1
 1) for r < m, hence
(x  l ) i / / " ( x ) ' ' ' <!> '"~ \x)
yl> — 1 yP = (r
{ X
2
— 1
yP'"'
1
_
1 = xP'"1
i
Uj
~
 • • •±

i
x 
1 x" 
1
x>""~ 
2
1
But t h e n x "'
!//<"'»(X) =
p
 1 •
completing t h e induction and proving t h e lemma. N o t e here that
^'»\x)
=
x P m
_ r
\
=
1 + x '"~*
p
+ ••• +
x " '""~
(
r)
l
232
Special Topics (Optional)
l
Ch. 6
is a monic polynomial with integer coefficients. Its degree is clearly p'" (p — 1), which is indeed <p(p'"). Finally, if 9 is a primitive p"'th root of unity, then 0 " = 1, but d "'~ + 1, hence i\i (0) = 0; so 0 is a root of <// (x). T h e final thing we want to know is: Is <//'"'(x) irreducible over 0 ?
pl p l (m) ffl)
Note that ip \x)
im
= 1 + x "'~'
(1)
p
+ • • • + xO ^'"
1
 1
=
^(x ""' )
1
1
and we know that i/y (x) is irreducible in <G>[x]. W e shall use the Eisenstein Criterion to prove that ip "'\x) is irreducible in <Q[x]. W e digress for a m o m e n t . If f(x) and g(x) are two polynomials with integer coefficients, we d e f i n e / ( x ) = g(x) m o d p if / ( x ) = g(x) + pr(x), where r(x) is a polynomial with integer coefficients. This is equivalent to saying that the corresponding coefficients of / ( x ) and g(x) are congruent m o d p. Expanding (f(x) + g(x)) by the binomial theorem, and using that all the binomial coefficients are divisible by p, since p is a prime, we arrive at (f{x) + g(x)Y  f(x)" + g(x)" mod p. G i v e n / ( x ) = a x" + ciix"~\ + ••• + «„, where the «,• are integers, t h e n , by the above,
( p 0
f(x)
p
= (« x" + a.x"0
1
+ • • • + « „ ) " = fl£x" + of x " >" + • • • +
1
p
<
1
a
p
= G X"
0
p
+
fl!^"" ^
+ • • • + a„ m o d p,
the latter congruence being a consequence of F e r m a t ' s T h e o r e m (the Corollary to T h e o r e m 2.4.8). Since f(x ) = a x" + a "V>' + • • • + a , we obain that
p p 0 1 t n
f(x )^f(xY Iterating what we just did, we arrive at f(x ) for all nonnegative k.
pk
p
modp.
=
f(x) modp
pk
W e return to our ip (x). Since i// (x) = i}/ (x "'~ ) we have, from the discussion above, that i//" (x) — iiV (x '"~') mod p. Therefore,
!) (1) p
(m)
m)
{l)
p
l
2 = <jj (x) '"~ (1) p l(p l)
x + p
.p'
)
m o d p = ijj '"\x
(
+ 1) m o d p.
Sec. 5
Cyclotomic Polynomials
233
This tells us t h a t
i/> "°0
m)
(
+ 1) = '"'~ x
p
l(p l)
+
pr{x),
w h e r e r(x) is a polynomial with integer coefficients. So all the coefficients of i / / ( x + 1), with the exception of its leading coefficient 1, are divisible by p. If we knew for some reason that the constant t e r m of h(x) = (// (x + 1) was not divisible by p , we could apply the Eisenstein Criterion to show that h(x) is irreducible. B u t what is the constant t e r m of h(x) = ip (x + 1)? It is merely n(0) = (//'"'(l), which, from the explicit form of i//"°(x + 1) that we found four paragraphs earlier, is exactly p. Thus h(x) is irreducible in Q[x], that is, (//'"'(x + 1) is irreducible in Q[x]. B u t this immediately implies that t / / 0 ' ) is irreducible in Q[x). Summarizing, we have proved
m) 2 (m) m)
T h e o r e m 6.5.2. For n = p"\ w h e r e p is any prime and m any n o n n e g ative integer, the polynomial cp (x) is irreducible in Q[x].
n
A s we pointed out earlier, this is a very special case of the theorem w e shall soon prove; namely, that <p„(x) is irreducible for all positive integers n. Moreover, the result and proof of T h e o r e m 6.5.2 play no role in the proof of the general proposition that <f>„(x) is irreducible in Q[x]. But because of the result in T h e o r e m 6.5.2 and the explicit form of <p„(x) when n = p'", we d o get a pretty good idea of what should be true in general. W e now proceed t o the discussion of the irreducibility of cp„(x) for general n. T h e o r e m 6.5.3. For every integer h > 1,
m
<p„{x) = (x  6 )where 6 , 9 ,...
(i> (2)
•  fx  0<*<"»),
,
are the ip(n) distinct primitive nth roots of unity.
Proof. W e proceed by induction o n n. If n = 1, then cpi(x) = x  1, and since 1 is the only first root of unity, the result is certainly correct in this case. Suppose that result is true for all m < n. Thus, if d  n and d + n, t h e n , by the induction, 0 ( x ) = (x  0 /) • • • (x  6 f ), where the fl^are the primitive t/th roots of unity. N o w
(l ( d)] d
x"  1 = 0 
Q { x  & ) • • • (x  Q ,
where the £, run over all nth roots of unity. Separating out the primitive n t h roots of unity in this product, we obtain x"
 1 = ( J E 
6» )
(1)
• • • o  ^""M*).
234
Special Topics (Optional)
Ch. 6
where u(x) is t h e product of all t h e other x — thus b y our induction hypothesis v(x) is t h e p r o d u c t of t h e 4> i{x) over all divisors d of n with t h e exception of d = n. Thus, since
c
0
n
W
(x"  1) (x •••(*fl<*"»)i;fr) u(x) u(x) = (x  e )(x  6 ) • • • (x 0 "»),
= m (2) M
we have proved t h e result claimed in t h e t h e o r e m .
n
•
F r o m the form of 4> (x) in T h e o r e m 6.5.3 we immediately see that 4> {x) is a monic polynomial in C[x] of degree cp(n). Knowing this, we prove that, in fact, t h e coefficients of 4>„(x) are integers. Why is this t r u e ? Proceeding by induction on n, we m a y assume this to be t h e case if d \ n a n d d ¥= n. Therefore, if v(x) denotes t h e polynomial used in t h e proof of T h e o r e m 6.5.3, then (x"  l)/v(x) = <f>,,(x) G C[x], hence v(x) \x"  1 in C[x]. B u t , by the longdivision process, dividing t h e monic polynomial v(x) with integer coefficients into x" — 1 leads us t o a monic polynomial with integer coefficients (and n o remainder, since rj(x)  (x"  1) in C[x]). T h u s (x"  l)lv(x) = (j>„(x) is a monic polynomial with integer coefficients. A s we saw, its degree is cp(n). Thus
n
T h e o r e m 6.5.4 F o r every positive integer n the polynomial (f>„( ) is a monic polynomial with integer coefficients of degree cp(n), where cp is t h e Euler <pfunction. Knowing that (fi (x) is a polynomial, we can see that its degree is cp(n) in yet a n o t h e r way. F r o m cp„(x) = (x"  1)1 v(x), using induction on n, d e g ( 0 „ ( x ) ) = n  deg(u(x)) = n — 1cp{d), t h e s u m over all divisors d of n other than d = «, from t h e form of u(x). Invoking t h e result of T h e o r e m 6.2.1, n  1cp(d) = cp(n), where again this sum is over all d \ n, d ¥= n. W e thus obtain that deg(</>„(*)) <P(' )T h e result we are a b o u t to prove is without question o n e of t h e most basic ones about cyclotomic polynomials.
n = 7
x
T h e o r e m 6.5.5. irreducible in Q[x].
F o r every positive integer n the polynomial </>„ (x) is
Proof. Let f(x) in Q[x] b e an irreducible polynomial such that / ( x )  <p„(x). Thus 4>„(x) = f(x)g(x) for some g(x) in Q[x]. By G a u s s ' L e m m a we m a y assume that both / ( x ) a n d g(x) a r e monic polynomials with integer coefficients, thus are in Zjx]. O u r objective is to show that <p„(x) = f(x); if
Sec. 5
Cyclotomic Polynomials
235
this w e r e the case, then, since / ( x ) is irreducible in Q[x], we would have that </>„(x) is irreducible in Q[x]. Since c/>„(x) has n o multiple roots, f(x) and g(x) must b e relatively prime. Let p b e a prime n u m b e r such that p does not divide n. If 6 is a root of fix), it is then a root of <j> (x), h e n c e by T h e o r e m 6.5.3 9 is a primitive nth root of unity. Because p is relatively p r i m e to n, 9 is also a primitive nth root of unity, thus, by T h e o r e m 6.5.3, 9 is a root of <p {x). W e therefore have that 0 = <p (6 ) = f(O )g(d ), from which we deduce that either f(8 ) = 0 or g(9 ) = 0. O u r aim is to show that f(9 ) = 0. Suppose not; then g(9 ) = 0, h e n c e 9 is a root of g(x ). Because 9 is also a root of the irreducible polynomial f(x), by L e m m a 6.3.2 we obtain t h a t / ( A )  g{x ). A s we saw in the course of the proof of T h e o r e m 6.5.2, g(x ) = g(x) m o d p. L e t / b e the ideal in Z g e n e r a t e d by p; by the Corollary to T h e o r e m 4.6.2, Z[x]/7[x] — Z [x], which m e a n s that reducing the coefficients of any polynomial m o d p is a h o m o m o r p h i s m of Z[x] onto Z [x]. Since all the polynomials <p„(x), v(x), fix), and g(x) are in Z[x], if 4>„(x),v(x),f(x),andg(x) are their images in Z [x], all the relations a m o n g t h e m are preserved going m o d p. T h u s we have t h e relations x"  1 = (x)rj(x), (x)_ = f(x)g (x) and f(x) \ g(x ) = g(x)" . Therefore, f(x) and g(x) have a c o m m o n root, A , in some extension K of Z . N o w x" — 1 = (p„ (x)tT(x) = / ( x ) g ( x ) , h e n c e a, as a root of both f(x) and g(x), is a multiple root of x"  1. B u t the formal derivative (x"  1)' of x"  1 is n x " 0, since p does not divide n; therefore, (x" — 1)' is relatively prime to x"  1. By t h e result of P r o b l e m 3 of Section 3 the polynomial x" — 1 cannot have a multiple root. With this contradiction arrived at from the assumption that 8 was not a root of f(x), we conclude that whenever 9 is a r o o t of fix), then so must 9 be o n e , for any prime p that does not divide n. R e p e a t i n g this argument, we arrive at: 9' is a root of / ( x ) for every integer ;• that is relatively p r i m e to n. B u t 9, as a root of fix), is a root of <p (x), so is a primitive nth root of unity. T h u s 9 is also a primitive nth root of unity for every r relatively prime to n. B y running over all r that are relatively prime to n, we pick u p every primitive nth root of unity as some such 9 . Thus all the primitive nth roots of unity are roots of f{x). By T h e o r e m 6.5.3 we see that c6„(x) = fix), hence c/J„(x) is irreducible in Q[x]. •
n P P n p p p p n p p p p p p p p p p p P _ 1 P P n r r
It may strike the r e a d e r as artificial and u n n a t u r a l to have resorted to the passage m o d p to carry out t h e proof of t h e irreducibility of a polynomial with rational coefficients. In fact, it may very well be artificial and unnatural. A s far as we know, n o proof of the irreducibility of c6„ (x) has ever been given
236
Special Topics (Optional)
Ch. 6
staying completely in Q[x] and not going m o d p. It would b e esthetically satisfying to have such a proof. O n t h e other hand, this is not the only instance where a result is p r o v e d by passing to a related subsidiary system. M a n y theorems in n u m b e r t h e o r y — a b o u t the ordinary integers—have proofs that exploit the integers mod p. Because 4>„(x) is a monic polynomial with integer coefficients which is irreducible in Q[x], and since 8„, t h e primitive nth root of unity, is a root of <p„(x), we have T h e o r e m 6.5.6. ep„(x) is the minimal polynomial in Q[x] for the primitive nth roots of unity.
P R O B L E M S
1. Verify that the first six cyclotomic polynomials are irreducible in Q[x] by a direct frontal attack. 2. Write down the explicit forms of: (a) <p (x). (b) cp (x). (c) cp (x).
10 l5 20
3. If (x'"  1 )  (x"  1), prove that m \ n. 4. If « > 1 is an integer and (fl'"  1)  (a"  1), prove that m \ n. 5. If K is a finite extension of Q>, the field of rational n u m b e r s , prove that there is only a finite n u m b e r of roots of unity in K. (Hint: U s e the result of Problem 10 of Section 2, together with T h e o r e m 6.5.6.)
6. LIOUVILLE'S CRITERION
Recall that a complex n u m b e r is said to be algebraic of degree n if it is the root of a polynomial of degree n over Q, the field of rational numbers, and is not the root of any such polynomial of degree lower than n. In the terms used in Chapter 5, an algebraic n u m b e r is a complex number algebraic over Q. A complex n u m b e r that is not algebraic is called transcendental. Some familiar numbers, such as e, TT, e"*, and m a n y others, are k n o w n to b e transcendental. Others, equally familiar, such as e + TT, eir, and TT ', are suspected of being transcendental but, to date, this aspect of their n a t u r e is still open. T h e French mathematician J o s e p h Liouville (18091882) gave a criterion that any algebraic n u m b e r of degree n must satisfy. This criterion gives us a condition that limits the extent to which a real algebraic n u m b e r of degree n .can be approximated by rational numbers. This criterion is of such a
1 (
Sec.
6
Liouville's Criterion
237
nature that we can easily construct real n u m b e r s that violate it for every n > 1. A n y such n u m b e r will then have to b e transcendental. In this way we shall be able to produce transcendental n u m b e r s at will. However, n o n e of the familiar n u m b e r s is such that its transcendence can be proved using Liouville's Criterion. In this section of the book we present this result of Liouville. It is a surprisingly simple and elementary result to prove. This takes nothing away from the result; in our opinion it greatly enhances it. T h e o r e m 6.6.1 (Liouville). Let a b e an algebraic n u m b e r of degree n > 2 (i.e., a is algebraic but not rational). T h e n there exists a positive constant c (which depends only on a) such that for all integers u, v with v > 0,  a — u/v  > civ". Proof. Let A be a root of the polynomial f(x) of degree n in Q[x], where < is the field of rational n u m b e r s . By clearing of denominators in the Q coefficients of f(x), we may assume that f(x) — r x" + r ^ x " + • • • + r„, where all the r, are integers and where r > 0. Since the polynomial fix) is irreducible of degree n it has n distinct roots a = a±, a , . . . , a„ in C, the field of complex numbers. Therefore, f(x) factors over C as /Yx) = r ( x  a)(x ~ a ) • • • (x  a„). Let u, v be integers with v > 0; then
 1 0 0 2 0 2
u\ _ i\ "_ , riu"'
)U
1
r„_ M
T
hence
V
" \i)
f
=
+
''i ""
M
L
u
+
• • • + I'niuv"'
1
+ r„v"
is an integer. Moreover, since f(x) is irreducible in Q[x] of degree n s 2, fix) has no rational roots, so v"f(u/v) is a nonzero integer, whence  v"f(u/v)  > 1. Using t h e factored form o f / ( x ) , we h a v e that
hence u    a v
:
\f(ulv)\ r \(ulv)
{)
 a \ • • • \{ulv)  a„\
2
r v"\(ulv)
0
v"\f(u/v)\  a \ • • • \(u/v) 2
q,\
r v"\(ulv)
0
1  a \  • • \{ulv)  a„\ '
2
238
Special Topics (Optional)
Ch. 6
Let s be the largest of \a\, \a \, • • •, \a„\. W e divide t h e argument according as \u/v\ > 2s or \it/v\ s 2s. If \u/v\ > 2s, then, by the triangle inequality, \a  (u/v)\ > \u/v\ — \a\ > 2s — s = s, and, since v > 1, \a  (u/v)\ > s/v". O n t h e other hand, if \u/v\ ^ 2s, then, again by t h e triangle inequality,  (u/v) \ < a, + «/rj < 5 + 2s = 3s. Therefore,
2
t =
(u\
W
 1 , M 0 2 1 1 0
a
n

W
x 0
~
(3s)"
so that lit > l / O ) " = l / ( 3 " ~ y ) . Going back t o t h e inequality above that \a  (u/v)\ > l/[r v"\a  (u/v)\ ••• \a„  (u/v)\], we have that \a  (u/v)\ 5: l/(r 3"~ s"~ v"). T h e s e n u m b e r s r , 3"'\ s"~ ate d e t e r m i n e d once and for all by a and its minimal polynomial f(x) a n d do not depend on u or v. If we let b = 1 / ( 7  3 " " J " " " ) b > 0 and \a ~ (u/v)\ > b/v". This covers t h e second case, w h e r e \ u/v\ s 2s. ff c is a positive n u m b e r smaller than both b and s, we have from the discussion that \a — u/v\ > c/v" for all integers u, v, w h e r e v > 0, t h e r e b y proving t h e theorem. •
1 1 T H E N 0 ;
Let's see t h e particulars of t h e proof for t h e particular case a = V2. T h e minimal polynomial for a in Q[x] isf(x) = (x  a)(x + a), so a = a a n d —a = a . So if u and v are integers, a n d v > 0, then
x 2

/
(
^
<
2
©
2

"
;
)
^
T
2
©
2

2
)
=
"
S

2
^
0
'
an integer. So  u / ( w / u )  > 1 > llv . T h e ^ above is t h e larger of  V 2  and   V2); that is, s = V5. Also, the above is l / ( 3 " ( V 2 )  ) = 1/(3V2), so if c is any positive n u m b e r less t h a n 1/(3V2), t h e n V2  ulv\ > c/v . W h a t the t h e o r e m says is t h e following: A n y algebraic real n u m b e r has rational n u m b e r s as close as we like to it (this is true for all n u m b e r s ) , b u t if this algebraic real n u m b e r a is of degree n s 2, there are restrictions o n h o w finely we can approximate a by rational numbers. These restrictions a r e t h e ones imposed by Liouville's T h e o r e m . H o w do we use this result to p r o d u c e transcendental n u m b e r s ? All we n e e d d o is to produce a real n u m b e r T, say, such that whatever positive integer n m a y b e , and whatever positive c we choose, we can find a pair of integers u, v, with v > 0 such that  T  u/v\ < c/v". W e can find such a r easily by writing down an infinite decimal involving 0's and l ' s , w h e r e we m a k e t h e 0's spread o u t between t h e l ' s very rapidly. F o r instance, if T = 0.10100100000010 . . . 010 . . . , where t h e 0's between successive l ' s go like ml, t h e n T is a n u m b e r that violates Liouville's Criterion for every n > 0. Thus this n u m b e r r is transcendental.
2 1 / 2 1 2
Sec. 7
The
I r r a t i o n a l i t y of IT
239
W e could, of course, use other wide spreads of O's between the l ' s — m'", (ml) , and so o n — t o produce h o r d e s of transcendental numbers. Also, instead of using just l's, we could use any of the nine nonzero digits to obtain m o r e transcendental numbers. W e leave to t h e r e a d e r the verification t h a t the n u m b e r s of the sort we described d o not satisfy Liouville's Criterion for any positive integer n and any positive c. W e can use the transcendental n u m b e r r and the variants of it we described to prove a famous result due to Cantor. This result says that t h e r e is a 11 correspondence between all the real n u m b e r s and its subset of real transcendental numbers. In other words, in some sense, there are as m a n y transcendental reals as t h e r e are reals. W e give a brief sketch of how we carry it out, leaving the details to the reader. First, it is easy to construct a 11 mapping of the reals onto those reals strictly between 0 and 1 (try to find such a mapping). This is also true for the real transcendental numbers and those of t h e m strictly between 0 and 1. Let the first set be A and the second one B, so B C A . Then, by a theorem in set theory, it suffices to construct a 11 mapping of A into B . Given any n u m b e r in A, we can represent it as an infinite decimal 0.0^2 . • • a„ .. . , where the a fall b e t w e e n 0 and 9. (We now wave our hands a little, being a little bit inaccurate. T h e r e a d e r should try to tighten up the argument.) Define the mapping / from A to B by / ( 0 . a r t • .. a„ . . .) = O.fl 0a 00a 000000fl . . . ; by the Liouville Criterion, except for a small set of a a , • • •, a ,. . . , the n u m b e r s 0.a 0a 00fl 000000a . . . are transcendental. T h e / w e wrote down then provides us with the required mapping. O n e final word about the kind of approximation of algebraic n u m b e r s by rationals expressed in T h e o r e m 6.6.1. T h e r e we have that if a is real algebraic of degree n s 2, then \a — u/v\ > c/v" for some appropriate positive c. If we could decrease the n to \a — u/v\ > civ'" for m < n and some suitable c (depending on a and m), we would get an even sharper result. In 1955 the (then) young English mathematician K. F. R o t h proved the powerful result that effectively we could cut the n down to 2. His exact result is: If a is algebraic of degree n "s 2, then for every real n u m b e r r > 2 there exists a positive constant c, depending on a and r, such that \a — ujv\ > c/v for all but a finite n u m b e r of fractions u/v.
2 t 1 2 1 2 3 4 u 2 n 1 2 3 4 r
7 . T H E I R R A T I O N A L I T Y OF TT
A s we indicated earlier, L i n d e m a n n in 1882 proved that TT is a transcendental number. In particular, from this result of L i n d e m a n n it follows that TT is irrational. W e shall not prove the transcendence of TT here—it would require a
240
Special Topics (Optional)
Ch. 6
rather long detour—but we will, at least, prove that TT is irrational. T h e very nice proof that we give of this fact is due to I. Niven; it a p p e a r e d in his p a p e r " A Simple Proof T h a t TT Is Irrational," which was published in the Bulletin of the American Mathematical Society, vol. 53 (1947), p. 509. T o follow Niven's proof only requires some material from a standard firstyear calculus course. W e begin with L e m m a 6.7.1. If u is a real n u m b e r , then lim u"/n! = 0.
Proof. If u is any real n u m b e r , then e" is a welldefined real n u m b e r and e" = l + u + « / 2 ! + M / 3 ! + • • • + u"ln\ + • • •. The series 1 + u + u /2l + • • • + u"ln\ + • • • converges to e"; since this series converges, its nth t e r m must go to 0. Thus lim u"lnl = 0. •
2 3 2
W e now present Niven's proof of the irrationality of TT. T h e o r e m 6.7.2. TT is an irrational n u m b e r .
Proof. Suppose that TT is rational; then TT = alb, w h e r e a a n d b are positive integers. For every integer n > 0, we introduce a polynomial, whose properties will lead us to the desired conclusion. T h e basic properties of this polynomial will hold for all positive n. T h e strategy of the proof is to m a k e a judicious choice of n at the appropriate stage of the proof. L e t / ( x ) = x"(a  bx)"ln\, where TT = alb. This is a polynomial of degree 2/7 with rational coefficients. Expanding it out, we obtain _ a x"
0
+ a x"^
Y n
TW where a = a", a, = na'^b,
n
+ ••• \
+
ax"
n
2
...,«,=
^ ' " j , a""b',
...
,a„ = (  1 ) " 6 »
are integers. W e denote the ith derivative of f(x) with respect to x by the usual n o tation f '\x), understanding f \x) to m e a n / ( x ) itself. W e first n o t e a symmetry property of f(x), namely, that f(x) = /(TT  x). T o see this, note t h a t / f x ) = (b"ln\)x"(TT  x)", from whose form it is clear t h a t / f x ) = /(TT  x). Since this holds for / ( x ) , it is easy to see, from the chain rule for differentiation, t h a t / ( x ) = (  l ) ' / ( 7 r  x).
( {0 ( ; ) ( / )
Sec. 7
The Irrationality of T T
241
This statement about f(x) and all its derivatives allows us to conclude that for the statements that we make about the nature of all f / ; e / ( 0 ) , there are appropriate statements about all the f('\ir).
w
W e shall be interested in t h e value o f / ( 0 ) , a n d / ( 7 r ) , for all nonnegative i. N o t e that from t h e expanded form of fix) given above we easily obtain t h a t / ( 0 ) is merely /'! times the coefficient of .v' of the polynomial f(x). This immediately implies, since t h e lowest p o w e r of x appearing in f(x) is the nth, t h a t / ( 0 ) = 0 if / < n. For i > n we obtain t h a t / ( 0 ) = ila^Jnl; since i s n, i\ln\ is an integer, and as we pointed out above, a,_„ is also an integer; therefore / ( 0 ) is an integer for all nonnegative integers i. Since f^(ir) = (  l ) ' f ( O ) , we have that / ( 7 r ) is an integer for all nonnegative integers i.
w ( , ) ( ; ) ( , ) (,)
( , )
w
W e introduce an auxiliary function F(x) = f{x) Since f '" {x) g
( ]
 f \x)
2
+ •••
+
il)J "\x).
(2
= 0 ifm = F\x)
> In, we see that = fV\x) + f(x).  f^ix) + ••• + (1)"/ '°M
( 2
= F{x) Therefore, iF\x)
dx
sin x  Fix) cos x)
F"(x) sin x + F'ix) cos x — F\x) (F"(x) cos x + Fix) sin x + Fix)) s i n * = fix) smx.
F r o m this we conclude that I fix) sin x dx = [F'ix) sin x  Fix) cos X]Q = (F'(TT) sin T  Fin) cos TT)  ( F ' ( 0 ) sin 0  7(0) cos 0) T = Fin)
+
F(0).
(,) ( 0
B u t from the form of Fix) above and the fact that all f ( 0 ) a n d / ( 7 r ) are integers, we conclude that F(TT) + F(O) is an integer. Thus JZf( ) dx is an integer. This statement about JQ fix) sin * dx is true for any integer n > 0 whatsoever. W e now w a n t to choose cleverly enough to m a k e sure that the statement " / J fix) sin x dx is an integer" cannot possibly be true.
x s m x
242
Special Topics (Optional)
x x
Ch. 6
W e carry out an estimate o n Sof( ) sin dx. F o r 0 < x < TT the funct i o n / ( . v ) = x"(a  bx)"ln\ ^ ir"a"ln\ (since a > 0), and also 0 < sin.v < 1. T h u s 0 < JZf(x) sin x dx < / S T T V / H ! dx = TT" V/7?!. Let u = TTCI; then, by L e m m a 6.7.1, lirn^ u"ln\ = 0, so if we pick n large enough, we can make sure that u"ln\ < IITT, hence 7 r " V 7 « ! = mi"ln\ < 1. But then Jof(x) sin x dx is t r a p p e d strictly b e t w e e n 0 and 1. But, by what we have shown, Jaf(x) sin .v dx is an integer. Since there is no integer strictly between 0 a n d 1, w e h a v e r e a c h e d a contradiction. T h u s the p r e m i s e t h a t TT is rational was false. T h e r e f o r e , TT is irrational. This c o m p l e t e s t h e proof of the theorem. •
+ +
i
;
Index
245
A(S), 16 Abel, 43 Abelian group, 43 Algebraic element, 193 extension, 193 number, 194 Algebraically closed field, 200 Alternating group, 121, 215 Associative law, 12, 41 Associative ring, 126 Automorphism of a group, 68 inner, 69 Automorphism of a ring, 141 Basis, 187 Bertrand Postulate, 220 Bijection, 11 Boole, 138 Boolean ring, 138 Cantor, 239 Carroll, 7 Cartesian product, 5 Cauchy, 80 Cauchy's Theorem, 80, 89 Cayley, 69 Cayley's Theorem, 69 Center, 53 Centralizer, 53,102 Characteristic subgroup, 76 Characteristic of a field, 178 Chinese Remainder Theorem, 147 Class Equation, 103 Commutative ring, 127 Commuting elements, 53 Commuting mappings, 21 Complement of a set, 5 Complex number, 32 absolute value of, 34 argument of, 35 complex conjugate of, 32 imaginary part of, 32 polar form of, 35 purely imaginary, 32 real part of, 32
Composition of mappings, 11 Congruence, 57 class, 60 Conjugacy, 58 class, 58 Conjugate elements, 58 of a complex number, 32 Constant function, 9 polynomial, 153 Constructible length, 202 number, 204 Correspondence Theorem for groups, 86 for rings, 142 Coset left, 64 right, 58 Cycle of a permutation, 111 Cyclic group, 53, 55, 60 generator of, 53 Cyclotomic polynomial, 230
Degree of an algebraic element, 195 of a field extension, 191 of a polynomial, 153 De Moivre's Theorem, 36 De Morgan Rules, 6 Dihedral group, 45,116,117 Dimension of a vector space, 186 Direct product of groups, 93 Direct sum of rings, 146 of vector spaces, 181, 182 Divides, 22,157 Division Algorithm, 155 Division ring, 127 Divisor, 22 Domain integral, 127 principal ideal, 157 Duplication of a cube, 201
246
Index
Eisenstein Criterion, 169 Element, 3 algebraic, 193 identity, 41 invertible, 133 orbit of, 21,65 transcendental, 194 unit, 41 Empty set. 4 Equality of mappings, 9 of sets, 4 Equivalence class, 58 Equivalence relation, 57 Euclid, 27 Euclidean ring, 163 Euclid's Algorithm, 22 Euler, 63 Euler cpfunction, 62 Euler's Theorem, 63 Extension field, 191 Factor, 22 Factor group, 78 Factorial, 17 Fermat, 63 Fermat's Theorem, 63 Field, 127 algebraically closed, 200 characteristic of, 178 extension, 191 of algebraic numbers, 199, 200 of quotients, 172175 of rational functions, 177,179 splitting, 213 Field extension, 191 algebraic, 193 degree of, 191 finite, 191 Finite abelian groups, 96 Finite dimensional vector space, 185 First Homomorphism Theorem for groups, 85 for rings, 142 Formal derivative, 227 Function, 8 constant, 9 Euler, 62
identity, 9 injective, 10 onetoone, 10 onto, 10 surjective, 10 Fundamental Theorem of Algebra, 200 on Finite Abelian Groups, 100 Galois theory, 212 Gauss, 169 Gauss' Lemma, 168 Gaussian integers, 38,166 Greatest common divisor of integers, 23 of polynomials, 158 Group, 41 abelian, 43 alternating, 121, 215 axioms, 41 center of, 53 cyclic, 53, 55, 60 dihedral, 45,116,117 factor, 78 finite, 42 Hamiltonian, 72 homomorphism, 67 Klein's, 116 nonabelian, 43 octic, 116 order of, 42 quotient, 78 simple, 123, 216 Hamilton, 72,132 Hamiltonian group, 72 Hardy, 201 Hermite, 194 Homomorphism of groups, 67 image of, 70 kernel of, 70 trivial, 67 Homomorphism of rings, 139 kernel of, 140 Homomorphism Theorems for groups, 8487 for rings, 142
Index
247
Ideal, 140 left, 140 maximal, 148 right, 140 trivial, 142 twosided, 140 Identity element, 41 Identity function, 9 Image, 8 inverse, 10 Index of a subgroup, 59 Induction, 29 Induction step, 31 Inductive hypothesis, 30 Injective mapping, 10 Integers, 21 Gaussian, 38,166 relatively prime, 25 Integral domain, 127 Intersection of sets, 4 Invariants of abelian groups, 100 Inverse of a group element, 41 of a mapping, 12 Invertible element, 133 Irreducible polynomial, 159 Isomorphic groups, 68 Isomorphic rings, 141 Isomorphism of groups, 68 of rings, 141 Kernel of a homomorphism for groups, 70 for rings, 140 Klein's 4group, 116 Lagrange, 59 Lagrange's Identity, 133 Lagrange's Theorem, 59 Leading coefficient, 162 Least common multiple, 28 Lindemann, 194 Linear combination, 185 Linear dependence, 186 Linear independence, 185 Liouville, 236 Liouville's Criterion, 236
Mappings, 8 commuting, 21 composition of, 11 identity, 9 injective, 10 onetoone, 10 onto, 10 product of, 11 surjective, 10 Mathematical induction, 29 Matrices real 2 X 2 , 1 3 0 , 1 3 1 2 X 2 over a ring, 131 Maximal ideal, 148 McKay, 89 Minimal generating set, 186 Minimal polynomial, 195 Monomorphism of groups, 68 of rings, 141 Motion of a figure, 115 Multiple, 22 Multiplicity, 209 Niven, 240 Normal subgroup, 71 index of, 59 Null set, 4 Number algebraic, 194 complex, 32 prime, 21, 26 purely imaginary, 32 transcendental, 194, 236 Octic group, 116 Onetoone correspondence, 11 mapping, 10 Orbit, 21, 65 Order of an element, 60 of a group, 42 Partition of a set, 58 of a positive integer, 100
248
Index Roots of unity, 42 primitive, 37, 39, 229 Root of a polynomial, 208 multiplicity of, 209 Roth, 239 Second Homomorphism Theorem for groups, 86 for rings, 142 Sets, 3 Cartesian product of, 5 difference of, 4 equality of, 4 intersection of, 4 union of, 4 Simple group, 123, 216 Splitting field, 213 Squaring of a circle, 207 Subfield, 128,191 Subgroup, 51 characteristic, 76 cyclic, 53 index of, 59 normal, 71 proper, 51 Sylow, 101 trivial, 51 Subring, 129 Subset, 3 Subspace of a Vector space, 181 spanned by elements, 181 Surjective mapping, 10 Sylow, 104 Sylow subgroup, 101,105 Sylow's Theorem, 105 Symmetric group, 16,109 Symmetry, 57 Third Homomorphism Theorem for groups, 87 for rings, 142 Transcendental element, 194 number, 194, 236 Transitivity, 57 Transposition, 20, 111 Triangle Inequality, 34
Permutation, 16,109 even, 121 odd, 121 Polynomial, 152,179 coefficients of, 152 constant, 153 cyclotomic, 230 degree of, 153 irreducible, 159 leading coefficient of, 162 minimal, 195 monic, 157 relatively prime, 159 root of, 208 Polynomial ring, 152 Prime number, 21, 26 Primitive root of unity, 37, 39, 229 Primitive root mod p, 66 Principal ideal domain, 157 Principle of Mathematical Induction, 29 Product of mappings, 11 Projection, 9 Quadratic nonresidue, 151 Quadratic residue, 151 Quaternions, 131,136 Quotient group, 78 Rational functions, 177,179 Reftexivity, 57 Relation, 56 congruence relation, 57 conjugacy relation, 58 equivalence relation, 57 Relatively prime integers, 25 polynomials, 159 Ring, 126 associative, 126 Boolean, 138 commutative, 127 division, 127 Euclidean, 166 homomorphism, 139 noncommutative, 127 of polynomials, 152,179 with unit, 126
Index Trisection of an angle, 201 Union of sets, 4 Unit element of a group, 41 in a ring, 126 Vector space, 180 basis of, 187 dimension of, 186 finite dimensional, 185 infinite dimensional, 185 minimal generating set for, 186 WellOrdering Principle, 22 Wiles, 63 Wilson's Theorem, 65, 210 Zero divisor, 128
249