Classical Modern Algebra

  • Commentary
  • Some pages duplicated, with some variants more readable than others. Same source as http://gen.lib.rus.ec/book/index.php?md5=971ABFAC5E3448BC410BBD269E3D7ED5 but different binarization.
Citation preview

CLASSICAL MODERN ALGEBRA

CLASSICAL MODERN ALGEBRA

SETH WARNER Professor of Mathematics

Duke University

Prentice-Hall, Inc., Englewood Cliffs, New Jersey

© 1971 by Prentice-Hall, Inc.

Englewood Chfls, New Jersey

All rights reserved. No part of this book may be reproduced in any form or by any means without permission in writingfrom the publisher.

Current printing (first digit): 1 2 3 4 5 6 7 8 9 10 13-]36069-8 Library of Congress catalog card number 78-12003 Printed in the United States of America

Prentice-Hall International, Inc., London

Prentice-Hall of Australia, Pty. Ltd., Sydney Prentice-Hall of Canada, Ltd., Toronto Prentice-Hall of India Private Ltd., New Delhi

Prentice-Hall ofJapan, Inc., Tokyo

To three Susan Emilys

PREFACE

To more than a generation of mathematics students, “modern algebra” has come to be identified with the contents of B. L. van der Waerden’s classic text bearing that name. To refer now to subject matter that was already standard in 1930 as “modem” is somewhat anomalous, but it is

difiicult to abandon the terminology we have received from our mathematical forebears. For this reason, “classi ” has been added to the traditional

“modern algebra” in the title of this text not only to indicate the historical place of the subject matter but also to suggest its continuing, decisive role in

all of mathematics. Although this text is intended primarily as an introduction to abstract algebra for advanced undergraduates, it may prove useful to anyone seeking an introduction to the subject. Many instructors will find more material here than can be covered conveniently in an academic year. Consequently, this text may serve a variety of different courses. Instructors may choose, for example, to assume as “known” the natural numbers and integers and therefore to omit sections 11, 12, and 14 of Chapter II; or instructors may

wish to omit sections 25 and 26 of Chapter IV, sections 31 and 32 of Chapter V, all of Chapter VII, or sections 44 and 47 in Chapter VIII.

The contents of this text are mostly drawn from the first eight chapters of the author’s Modem Algebra. At many points the exposition has been simplified and made less general to facilitate the use of the text in the class-

room. Material on finite groups, including the Sylow theorems, has been added.

Readers will find all the linear algebra needed for a reading of Chapters VI—VIII in Chapter V, plus some additional material on duality that is designed to relate concepts of analytic geometry with those of linear algebra. The text does not contain enough material for a thorough course in linear algebra, however. ix

X

PREFACE

The author is convinced that the only way a beginner can really absorb algebra is by solving problems. For this reason a large number of exercises is included, and it is hoped that the text when read is sufliciently clear so that the instructor and his students may devote most of class time to the exercises. They range in difliculty from the routine to the very advanced, but hints are

supplied for those of even modest difliculty so that they will be amenable to many students. Exercises that are not routine are starred. Instructors will perceive that often exercises on the same topic are grouped together; suCh a group of exercises may serve as the basis for papers by students particularly interested in the topic considered. Errors undoubtedly persist, especially in the exercises, and the author will welcome corrections from readers.

Many exercises are based on contributions to The American Mathematical

Monthly, and the author wishes to express his thanks to the many contributors to that journal who will find their contribution (sometimes slightly modified) among the exercises.

SETH WARNER

CONTENTS

9999:“?!91‘

I ALGEBRAIC STRUCTURES

The Language of Set Theory 2 Compositions 12 Unions and Intersections of Sets 20 Neutral Elements and Inverses 26 Composites and Inverses of Functions Isomorphisms of Algebraic Structures Semigroups and Groups 52 Subgroups 57

1

33 41

II THREE BASIC NUMBER SYSTEMS 9. 10. 11. 12. 13. 14. 15. 16. 17.

Orderings 68 Ordered Semigroups 76 The Natural Numbers 79 Finite Sets 100 The Division Algorithm 105 The Integers 110 Rings, Integral Domains, and Fields Equivalence Relations 135 The Rational Field 138

67

123

xii

CONTENTS

155

GROUPS AND RINGS 18. 19. 20. 21. 22.

Normal Subgroups and Quotient Groups Ideals and Quotient Rings 164 Homomorphisms of Groups 173 Homomorphisms of Rings 189 Principal Ideal Domains 195

155

213

FINITE GROUPS 23. 24. 25. 26.

Cyclic Groups

213

Direct Products 220 Extensions of Cyclic Groups by Cyclic Groups Sylow’s Theorems 238

230

VECTOR SPACES 27. 28. 29. 30. 31. 32.

Vector Spaces and Modules

252

252

Subspaces and Bases 260 Linear Transformations 275

Direct Sums and Quotient Spaces Duality Matrices

284

290 303

POLYNOMIALS 310 33. Algebras 34. The Algebra of Polynomials 329 ‘ 35. Substitution 36. Irreducibility Criteria 341 348 37. Adjoining Roots 38. Finite Fields 362

310 319

THE REAL 'AND COMPLEX NUMBER FIELDS 39. 40. 41. 42.

369

Dedekind and Archimedean Ordered Fields 370 The Construction of a Dedekind Ordered Field 380

Isomorphisms of Archimedean Ordered Groups The Field of Complex Numbers 399

388

CONTENTS

VIII

xiii

407

GALOIS THEORY

43. Algebraic Extensions 407 Constructions by Ruler and Compass 414 44. 45. Galois Theory 420 46. Separable and Normal Extensions 432 47. The Euler-Lagrange Theorem 442 48. Roots of Unity 447 49. Permutations Groups 461

50. Solving Polynomials by Radicals

470

APPENDIX A

INDUCED N-ARY OPERATIONS

485

APPENDIX B

COMBINATORIAL ANALYSIS

498

LIST OF SYMBOLS

505

INDEX

509

CLASSICAL MODERN ALGEBRA

CHAPTER I

ALGEBRAIC STRUCTURES

The best way of learning what modern algebra is all about is, of course, to study it. But some preliminary insight may be obtained by comparing it with the mathematics taught in many secondary schools. Beginning algebra is concerned with the operations of addition and multiplication of real or complex numbers and with the related operations of subtraction and division.

Emphasis is on manipulative techniques (such as rearranging parentheses, solving quadratic equations) and on translating problems into the language of algebra whose solutions may then be worked out. Emphasis in modern algebra, in contrast, is on deriving properties of algebraic systems (such as the system of integers, or of real numbers, or of

complex numbers) in a formal, rigorous fashion, rather than on evolving techniques for solving certain kinds of equations. Modern algebra is thus more abstract in nature than elementary algebra. In contrast with elementary algebra, modern algebra studies operations on objects that need not be numbers at all but are assumed to satisfy certain laws. Such operations are already implicit in elementary mathematics. Thus, addition and multiplication of polynomials are best thought of as operations on polynomials considered as objects themselves, rather than on numbers. In the theorem

“the derivative of the sum of two differentiable functions is the sum of their derivatives” of calculus, “sum” refers essentially to an operation not on numbers but on difl‘erentiable functions. In spirit, modern algebra is more like plane geometry than elementary algebra. In plane geometry one deduces properties of plane geometrical figures from a given set of postulates. Similarly, in modern algebra we deduce properties of algebraic systems from given sets of postulates. Some statements proved, such as “— (—a) = a and a - 0 = 0 for any number a,” are so

familiar that it is sometimes hard to realize that they need proof. Others, such as “there are polynomial equations of fifth degree that cannot be solved by 1

2

AmEBRArc STRUCTURES

Chap. I

radicals,” are solutions to problems mathematicians worked on for centuries. In plane geometry, however, only one set of postulates is considered, and' these postulates are applicable only to the plane. In contrast, many sets of postulates are considered in modern algebra, some of which are satisfied by many different algebraic systems. This generality, of course, greatly increases the applicability of the theorems proved. But increased generality is not the only benefit of the abstract approach in algebra. In addition, the abstract approach exposes the essential ideas at work in a theorem and clears away what is merely fortuitous. Consider, for

example, the following two assertions: (1) There are at most three real numbers x satisfying the polynomial equation x3—3x2+2x—1=0,

for the degree of the polynomial is 3. (2) There is at least one real number x between 2 and 3 satisfying

x3—3x’+2x—1=0, for 23—3-23+2-2—1=—10.

The truth of (1) depends only on relatively few properties of the real numbers, and the statement obtained by replacing “rea ” with “rational” or “complex” is also valid. In contrast, the truth of (2) depends on much

deeper properties of the real numbers, and, indeed, the statement obtained by replacing “re ” with “rational” is false. The theorem underlying (1) is actually valid for many algebraic systems besides the real numbers, but for relatively few of such systems is the theorem underlying (2) valid. When we examine these two theorems in their natural abstract setting, we shall see at

once what properties of the real numbers are essential for the truth of (2) but incidental for the truth of (1).

1. The Language of Set Theory

Any serious discussion must employ terms that are presumed understood without definition. Any attempt to define all terms used is futile, since it can

lead only to circular definitions. The terms we shall discuss but not attempt to define. formally are object, equals, set, element, is an element of, function,

ordered pair.

Sec. 1

ADGEBRAIC STRUCTURES

3

The English verb equals, symbolized by =, means for us “is the same as,” or “is identical with.” Thus, if a is an object and if b is an object, the expression a = b

means “a” and “b” are both names for the same object. For example, the sixteenth president of the United States = the author of the Gettysburg address. By a set we mean simply a collection of objects, which may be finite or infinite in number and need not bear any obvious relationship to each other. For variety of expression, the words class, collection, and occasionally space

are sometimes used as synonyms for “set.” An object belonging to a set is called an element of the set or a member of the set. A set E and a set F are considered identical if and only if the elements of E are precisely those of F. The symbol 6 means “is an element 0 ” or “belongs to” or grammatical variants of these expressions, such as “belonging to.” The assertion x is an element of E

is thus symbolized by e.

Similarly, the symbol 9% means “is not an element of.” For example, if P is the set of presidents of the United States before 1900,

then Hayes 6 P, Tilden (f. P, 17 ¢ P. If H is the set of all heads of state in 1863 together with the number 1863, then Lincoln E H, Wilson ¢ H, Victoria E H,

1776 ¢ H, Napoleon III e H, 1863 e H. Five mathematical sets, with which the reader is already acquainted, are so important that we shall introduce symbols for them now. The set of natural numbers consisting of the whole numbers 0, 1, 2, 3, etc., we denote

by N. The set ofintegers consisting of all natural numbers and their negatives, we denote by 2. Thus, —3 is an integer but not a natural number. The set of all rational numbers consisting of all fractions whose numerators are integers and whose denominators are nonzero integers is denoted by Q. Every integer n is also the fraction n/l and so is an element of Q. The set of real numbers is denoted by R and may be pictured as the set of all points on an infinitely long line, e.g, the X-axis of the plane of analytic geometry. Every rational number 1s a real number, but not conversely; for example, \/2 and 11 are real numbers but not rational numbers. We shall say that a real number x is positive if x 2 0 and that x is strictly positive if x > 0.

It is simply for convenience that we deviate in this way from the ordinary

4

Amanmc STRUCTURES

Chap. I

meaning of “positive.” Similarly, a real number x is negative if x s 0, and x is strictly negative if x < 0. With this terminology, zero is both a positive and a negative number and, indeed, is the only number that is both positive

and negative. The set of complex numbers is denoted by C and may be pictured as the set of all points in the plane of analytic geometry, the point (x, y) corresponding to the complex number x + iy. Complex numbers may also be written in

trigonometric form: If 2 is a nonzero complex number and if z = x + iy, then also 2 = r(cos 0 + isin 0) where

r=«/x‘+y”, the length of the line segment joining the origin to (x, y), and where 0 is the angle whose vertex is the origin, whose initial side is the positive half of the X-axis, and whose terminal side passes through (x, y). If

21 = x1 + iy1 = r1(cos 01 + isin 01), z, = x2 + iy, = r,(cos 092 + isin 0,), then Zr + Z: = (x1 + x2) + i()’1 +y2)

and

2122 = (xlxa — yiya) + i(x1ya + x01) = rlr2[cos(01 + 9,) + i sin(01 + 92)]. In this and subsequent chapters our discussion will frequently proceed on two levels. In §11 we shall formally state postulates for addition and the ordering on the natural numbers, prove theorems about them, and go on later to define and derive properties of the integers, the rational numbers, the real numbers, and the complex numbers. On an informal level, however, we shall

constantly use these sets and the familiar algebraic operations on them in illustrations and exercises, since the reader is already acquainted with many of their properties. It is important to keep these levels distinct. In our formal development we cannot, of course, use without justification those properties of N, Z, Q, R, or C that we tacitly assume known on the informal level.

We shall say that a set E is contained in a set F or is a subset of F if every element of E is also an element of F. The symbol 9 means “is contained in”

Sec. 1

ALGEBRAIC STRUCTURES

5

and its grammatical variants, such as “contained in,” so that EE F means E is contained in F.

Similarly, we shall say that E contains F and write E2 F

if F is contained in E; the symbol .3. thus means “contains” and its grammatical variants. Clearly, E S E, and if E E F and F S G, then E S G. Since E and F are identical if they have the same elements,

E=FifandonlyifES FandFS E. For example, N E Z, Z S Q, Q s R, and R s C, or more succinctly,

NEZSQEREC One particularly important set is the empty set or null set, which contains no members at all. It is denoted by 0. One practical advantage in admitting {D as a set is that we may wish to talk about a set without knowing a priori whether it has any members. For any set E, we have 0 s E; indeed, every member of 0 is also a member of E, simply because there are no members at all of (2). A set having just a few elements is usually denoted by putting braces around a list of symbols denoting its elements. No significance is attached to the order in which the symbols of the list are written down, and several symbols in the list may denote the same object. For example, {the hero of Tippecanoe, President Wilson, the paternal grandfather of President Benjamin

Harrison} = {the author of the League of Nations, President William Henry Harrison}. Very often, a set is defined as consisting of all those objects having some

given property. In this case, braces and a colon (some authors use a semicolon or a vertical bar in place of a colon) are often used to denote the set,

as in the following illustrations. For example, {x: x has propertyQ} denotes the set of all objects x such that x has property Q (the colon should

be read “such that”); the set H discussed on page 3 may thus be denoted

by {x: either x was a head of state in 1863, or x = 1863}.

6

ALGEBRAIC smUcrumzs

Chap. I

Often, a set is described as the subset of those elements of some set possessing a certain property. Thus,

{x e E: x has property Q} denotes the set of all objects x that belong to the set E and have property Q.

Many different properties may serve in this way to define the same set. For example, the sets

{2, 3, 5}, {x e N: x is a prime number and x < 7},

{x eN: 2x4 — 21x3 + 72x2 — 91x + 30 = 0} are identical. A function is a rule which associates to each element of a given set, called the domain of the function, one and only one element of a given set, called the codomain of the function. Thus, to specify a function, we must specify two sets, the domain and codomain (they may be the same set) and a rule which

associates to each element of the domain a unique element of the codomain. Iff is a function with domain E and codomain F, for each x e E the unique element of F that is associated to x by f is called the value off at x, or the

image of x under f, and is usually denoted by f(x). The expression “fis a function from E into F” means “fis a function whose domain is E and whose codomain is F” and is symbolized by f:E—>F.

If b is the image under a function f of an element a, then

b =f(a) by the definition of the expression “f a ”; another frequently used way of

expressing this is to write f: a —> b. For example, let us determine all functions from E into F where E = {0, 1}

and F = {4, 5}. Clearly, there are precisely four such functions, which we denote by f, g, h, k, whose rules are given as follows:

f:0—>4,l—->4

h:0—>5,1—>4

g:0—>4,l—>5

k:0—>5,l—>5.

Sec. 1

ALGEBRAIC smUcrUiuas

7

Sometimes, a function may be defined by means of a formula, in which case

the formula and an arrow or equality sign are often used in specifying the function. For example, f: R —> R, where

f:x—->xsinx

means that f is the function with domain R and codomain R that associates to each number x the number x sin x. Similarly, g: R —> R, where

()

—-lifx$2,

x:

g

x‘+lifx>2

means that g is the function with domain R and codomain R that associates —1toxifx sZandx3+ltoxifx>2.

If f and g are functions, then f and g are regarded as identical if they have the same domain, the same codomain, and if to each element of their domain they associate the same element of their codomain. Thus, iff: R —> R

and g: R —> R are given by fix—>\/DF

gix-r IXI, then f and g are the same function, a fact we express, of course, by writing f = g. 0n the other hand, if h is the function with domain R and codomain

the set R+ of positive numbers that is given by h:x—>|x|, then h 72 g, since h and g have difi'erent codomains.

The last of our undefined terms is ordered pair. To every object a and

every object b (which may possibly be identical with a) is associated an object, denoted by (a, b) and called the ordered pair whose first term is a and whose second term is b, in such a way that for any objects a, b, c, d,

(OP)

if(a,b)=(c,d),thena=candb=d.

This concept is familiar from analytic geometry. Indeed, the points of the plane of analytic geometry are precisely those ordered pairs both of whose terms are real numbers. It is possible to give a definition of “ordered pair” that yields (OP) solely in terms of the set-theoretic concepts already introduced (Exercise 1.11). Therefore, it is not really necessary to take “ordered pair” as an undefined term. However, the precise nature of any definition of

AmnnnAIc STRUCTURES

8

Chap. I

“ordered pair” is unimportant; all that matters is that (OP) follow from any

proposed definition.

Using “ordered pair,” we wish to define the ordered triad (a, b, c) of objects a, b, c in such a way that if (a, b, c) = (d, e,f), then a = d, b = e,

and c = f. We accomplish this by defining (a, b, c) to be the ordered pair ((a, b), c). Indeed, if

((0.17), c) = ((d, e),f). then

(a, b) = (d, e) and

c =f by (OP), whence again by (OP), (1 = d, b = e. Similarly, if a, b, c, and d are objects, we define (a, b, c, d) to be ((a, b, c), d), and so forth. We could, of course, have defined (a, b, c) to be (a, (b, c)) and again derived the result that if (a, b, c) = (d, e, f), then a = d, b = e, and c = f.

Thus, we have simply chosen for our definition of (a, b, c) one of several equally plausible possibilities. Definition. If E and F are sets, the cartesian product of E and F is the set

E X F defined by EX F={(x,y):eandyeF}. Similarly, if E, F, and G are sets, the cartesian product of E, F, and G is the

setE X FX Gdefinedby EX FX G={(x,y,z):e,yeF,andzeG}. For example, R X R is the plane of analytic geometry, R X {0} is the X-axis and {0} X R the Y-axis of that plane, R X {1, 2} is the set of points lying on either of two certain lines parallel to the X-axis, and {0, 1} X {0, 2} is the set of corners of a certain rectangle. Similarly, R X R X R is space of

solid analytic geometry, R X R X {0} is the X Y-plane, and R X {0} X R is the XZ-plane.

manure smucrmuzs

Sec. 1

9

The concept of the cartesian product of two sets enables us to define in a natural way the graph of a function. If f is a function from E into F, the graph of f is the subset Gr(f) of

E X F defined by Grtf) = {(x.y) 6E X Fzy =f(x)}-

When is a subset G of E X F the graph of a function from E into F? Suppose that G = GrU'), wheref is a function from E into F. Asf associates to each element of E an element of F,

(Gr 1)

for each x e E, there exists at least one element y e F such that

(x, y) e G, and since to each x e E is associated only one element of F,

(Gr 2)

if (x, y) e G and if (x, z) e G, theny = 2.

Conversely, if G is a subset of E X F satisfying (Gr 1) and (Gr 2), then G is the graph of exactly one function whose domain is E and whose codomain is F, namely, the function that associates to each x 6E the unique [by (Gr 2)] element y of F such that (x, y) e G. If G is a subset of E x F satisfying (Gr 2) [but not necessarily (Gr 1)], then G is the graph of exactly one function whose domain is a subset of E and whose codomain is F, namely, the function f whose domain D is

{x GE: for somey EF, (x, y) e G} that associates to each x e D the unique [by (Gr 2)] element y of F such that

(x. y) E GFor example, the graph G of a function f whose domain is a subset of R and whose codomain is R is a certain subset of R X R. By (Gr 2), every

line parallel to the Y-axis contains at most one point (i.e., ordered pair of real numbers) belonging to G; conversely, every subset of the plane having this property is the graph of exactly one function whose domain is a subset of R and whose codomain is R. Thus, the circle of radius 1 about the origin

is not the graph of a function, but the semicircles G1 and G3 defined by Gl={(x,y)eRxR:x*+y”=landyzO}

Gz={(X,y)eR>0? (c)m>0andn=0?

(b)m=0andn>0? (d)m=0andn=0?

*1.10. Prove that V5 ¢ Q. [Arrive at a contradiction from the assumption that

V5 6 Q by use of the fact that every rational number may be expressed as the quotient of integers at least one of which is odd.]

*1.11. (a) For each object a and each object b, define (a, b) to be {{a}, {a, b}}. Prove that if (a, b) = (c, d), then a = c and b = d. [Consider separately

the two cases {a} = {a, b} and {a} aé {a, b}.] (b) For each object a and each object b, define ]a, b[ to be {{b}, {a, b}}. Prove that if ]a, b[ = ]c, d[, then a = c and b = d. [Use (a).]

2. Compositions

Addition and multiplication, the two basic operations of arithmetic, are examples of the fundamental objects of study in modern algebra: Definition. A composition (or a binary operation) on a set E is a function from E X E into E. The symbols most frequently used for compositions are - and +. We shall at first often use other symbols, however, to lessen the risk of assuming

without justification that a given composition has properties possessed by

addition or multiplication on the set of real numbers. If A is a composition on E, for all x, y e E we denote by

xAy the value of A at (x, y). Just as in elementary calculus attention is limited to certain special classes of functions (e.g., differentiable functions in differential calculus, continuous

functions in integral calculus), so also in algebra we shall consider only

compositions having particularly important properties. If A is a composition on E and if x, y, and z are elements of E, (xAy)Az may very well be difi‘erent

from xA (yAz). Definition. Acomposition A on E is associative if (xAy)Az = xA (yAz) for all x, y, 2 6E.

13

assume STRUCTURES

Sec. 2

If A is an associative composition, we shall write simply xAyAz for

(xAy)Az. It is intuitively clear that if A is associative, all possible groupings of a finite number of elements yield the same element; for example, (xA(yAz))A (uAv) = xA (yA ((zAu)Av)).

We shall formulate precisely and prove this in Appendix A. Although algebraists have intensively studied certain nonassociative compositions, we shall investigate associative compositions only. We shall, however, encounter both commutative and noncommutative compositions:

Definition. Let A be a composition on E. Elements x and y of E commute (or permute) for A if xAy = yAx.

The composition A is commutative if xAy = yAx for all x, y e E.

If E is a finite set of n elements, a composition A on E may be completely described by a table of n rows and n columns. Symbols denoting the n elements of E head 'the n columns and, in the same order, the n rows of the

table. For all a, b e E, the entry in the row headed by a and the column headed by b is the value of A at (a, b). It is easy to tell from its table whether A is commutative; indeed, A is commutative if and only if its table is symmetric with respect to the diagonal joining the upper left and lower right corners. However, usually it is not possible to determine by a quick inspection of its table whether a composition is associative. Example 2.1. Let E be one of the sets Z, Q, R, C. Then addition and

multiplication (the functions (x, y) —> x '+ y and (x, y) —> xy respectively) are both associative commutative compositions on E. Subtraction is also a composition on E, but it is neither associative nor commutative since, for example, 6 — (3 — 2) 9E (6 — 3) — 2 and

2 — 3 75 3 — 2. Example 2.2. Let E consist of the English words “odd” and “even.” We define two compositions GB and 0 on E by the following tables: 6

even

odd

0

even

odd

even

even

odd

even

even

even

odd

odd

even

odd

even

odd

l4

ALGEBRAIC STRUCTURES

Chap. I

These compositions mirror the rules for determining the parity of the sum and product of two integers (e.g., the sum of an even integer and an odd integer is odd, and the product of an even integer and an odd integer is even). It is easy to verify that 6 and O are both associative and commutative. Example 2.3. For each positive integer m, we define Nm to be the set {0, l, . . . , m — l} of the first m natural numbers. Two very important

compositions on Nm, denoted by +,,, and ~,,, and called addition modulo m and multiplication modulo m, are defined as follows: x +,,, y is the remainder after x + y has been divided by m, and x -,,, y is the remainder after xy has been divided by m. In other words, x- +,,, y = x + y — jm where j is the largest integer such that jm S x + y, and x -,,, y = xy — km where k is the largest integer such that km 3 xy. The tables for addition modulo 6 and multiplication modulo 6 are

given below. +.

0

l

2

3

4

5

'.

0

l

2

3

4

5

0

0

l

2

3

4

5

T

0

—0—

0

0

0

0

l

l

2

3

4

5

0

l

0

l

2

3

4

5

2

2

3

4

5

0

l

2

0

2

4

0

2

4

3

3

4

5

0

l

2

3

0

3

0

3

0

3

4

4

5

0

l

2

3

4

0

4

2

0

4

2

5

5

0

l

2

3

4

5

0

5

4

3

2

l

Manipulations with these compositions are in some respects easier than the corresponding manipulations with ordinary addition and

multiplication on Z. For example, to find all x e N. such that (x'sx) +sx+o4=o:

we need only calculate the expression involved for each of the six possible choices of x to conclude that l and 4 are the desired numbers. No analogue of the quadratic formula is needed. The modulo m compositions arise in various ways: In computing time in hours, for example, addition modulo 12 is used (3 hours after

10 o’clock is 1 o’clock, i.e., 3 +13 10 = 1). It is easy to infer the associativity and commutativity of +,,, and

AmEBRAIc STRUCTURES

Sec.2

15

-,,, from the associativity and commutativity of addition and multiplication on Z. Let us prove, for example, that (x 'my) 'mz = x .m (y .1112)“

Let j be the largest integer such that jm s xy, and let p be the largest integer such that pm 3 yz. By definition,

x any = xy —jm, y -m z = yz — pm.

Let k be the largest integer such that km s (xy — jm)z, and let q be the largest integer such that qm g x(yz — pm). Then (jz + k)m S (xy)z and (q + xp)m s x(yz), and by definition

(x 'm y) -... z = (xy - jm)z — km, x -m(y-,,,z) =x(yz—pm) —qm. But jz + k is the largest of those integers i such that im 3 (xy)z; if not, (jz + k + l)m S (xy)z, whence (k + l)m 3 (xy — jm)z, a

contradiction of the definition of k. Similarly, q + xp is the largest of those integers i such that im g x(yz). As (xy)z = x(yz), we have jz+k=q+xp, andthus (x-my)-mz = (xy —jm)z— km = xyz— (jz+k)m =xyz— (q+xp)m=x(yz—pm)—qm

= x 'm (y 3.. Z)Example 2.4. On any set E we define the compositions by xya=y for all x, y e E. Clearly, are associative compositions, but, if E contains more than one element, neither is commutative.

Example 2.5. Let P be a plane geometrical figure. A symmetry ofP, or

a rigid motion ofP into itself, is a motion of P such that the center of

l6

ALGEBRAIC smucrum

Chap. I

P is unmoved during the motion and the figure occupies the same position at the end of the motion that it did at the beginning. Two such motions are regarded as the same motion if they have the same efi'ect on each point of the figure; for example, a counterclockwise rotation of a square through 270° about its center is regarded as the same as a clockwise rotation through 90° about its center. Each

symmetry of P is either a rotation of P in the plane of the figure about its center or a rotation of the figure in space about an axis of symmetry

of the figure. If x and y are symmetries of P, it is intuitively evident that the motion obtained first by performing y and then x is again a symmetry, which we shall denote by x o y (note the order: at o y is the motion obtained by first performing y and then x). Consequently, if x, y, and z are symmetries,

(x°y)°2=x°(y°2), for each is the symmetry obtained by first performing 2, then y, and finally x. Let G be the set of symmetries of a square, and let us number the corners of the square 1, 2, 3, and 4 in a counterclockwise fashion. Then G contains eight members: the counterclockwise rotations r0, r1, r2, and r3 in the plane of the square through 0°, 90°,

180°, and 270° respectively, the rotations h and v in space about the horizontal and vertical axes of the square, and the rotations al1 and d, in space about the diagonal joining the upper left and lower right corners and the diagonal joining the lower left and upper right corners respectively. If two symmetries of the square have the same effect on each corner of the square, they will also have the same efi'ect on each point of the square and so will be the same symmetry. Thus, to determine the table of the composition 0 on G, it suflices to find out

for each x, y e G what happens to each corner of the square if y is performed and then x. For example,

21 34

”12's 43

and 21

so r3 0 v = d,; similarly,

d.

41

41 32

Sec. 2

ALGEBRAIC STRUCTURES

17

and 21d123 34

—->

14

so 0 0 r3 = d1. A partial table for o is given below. 0

r0

r1

r,

r,

h

v

d1

d.

r0

r0

r1

r.

r.

h

v

d,

d,

r1

‘ r1

r,

r,

r8

r8

h

h

v

u

d1

d1

d,

d.

h

h d1

d, h 0

We have just seen that o is not commutative. A formal, direct .verification of the associativity of o is tedious; we shall be able to prove associativity formally more easily in §8.

EXERCISES 2.1. Write out the tables for +2 and -,. Can you make precise the assertion that

these compositions are respectively “just like” the compositions (-B and O of Example 2.2? [In making the assertion precise, use the functionfdefined byf(even) = 0,f(odd) = 1.]

2.2. Let E = {a, b} be a set containing two elements. Write out the tables for all

compositions on E (head the first row and column by “a” and the second row and column by “b”). Collect the tables into classes so that if two distinct tables are in the same class, the entry in each square of one table is not the

same as the entry in the diagonally opposite square ofthe other table. (There are ten classes, six of which contain two tables and four of which contain

only one table.) Determine which tables define associative compositions and which ones define commutative compositions. Are there two tables in the

18

ALGEBRAIC smucnmns

Chap. I

same class, of which one defines an associative (commutative) composition

but the other a nonassociative (noncommutative) composition? Can you

make precise the assertion that the compositions defined by two tables in the same class are “just like” each other? [Rewrite one of the tables by heading its first row and column by “b” and its second row and column by “a”.]

2.3. If E is a finite set of n elements, how many compositions are there on E? Of these, how many are commutative? 2.4. Find all x E N6 such that (a)3-sx=3.

(b)2-6x=5.

(c) (S-ax) +s3=4-

(d)x-6x=1.

(e) x'sx =5-

(0 (X'sX) +s(3'sx) =4-

(g)2-6x=4'6x. (i) x-sx-6x=4~6x.

(h)x-6x-6x=(5'6x) +65. (j) (x +6x+6x) +3(x+sx+.x)=0.

What facts that you remember from elementary algebra concerning the solution of equations no longer hold when addition and multiplication modulo 6 replace ordinary addition and multiplication of real numbers? 2.5. Let m > 2. Prove that (m — 1) -m (m — l) = 1. Infer that there exists p e Nm such that x -,,, x 9e p for all x 6 NM; i.e., there exists an element of

Nm that is not a square for multiplication modulo m. 2.6. Prove that +,,n is associative. 2.7. Prove that +m and -,,, are commutative.

*2.8. Let m 2 0, n > 0. Let + mm be the composition on Nm+fl defined as follows: if x +y and 1, 1 is the neutral element for multiplication modulo

m. The composition 0 defined in Example 2.5 on the set of all symmetries of the square admits re, the symmetry leaving each point of the square fixed, as neutral element. On the set SB(E) of all subsets of E, (D is the neutral element

for U and E is the neutral element for n. If x is a nonzero real number, —x and x‘1 (or l/x) play similar roles for addition and multiplication, for —x added to x yields the neutral element 0 for addition, and x—1 multiplied by x yields the neutral element 1 for multiplication. For this reason, —x and x—1 are called inverses of x for addition and multiplication respectively: Definition. Let A be a composition on E. An element x e E is invertible for A if there is a neutral element e for A and if there exists y e E such that xAy = e = yAx.

An element y satisfying xAy = e = yAx is called an inverse of x for A. Theorem 4.2. If A is an associative composition on E, an element x EE admits at most one inverse for A. Proof: If xAy =’ e = yAx,

xAz = e = zAx, then y = yAe = yA(xAz)

= (yAx)Az = eAz = z.

28

ALGEBRAIC STRUCTURES

Chap. I

Therefore, if A is associative and if x is invertible for A , we may use the

definite article and speak of the inverse of x. The inverse of an element x invertible for an associative composition denoted by a symbol similar to + is denoted by —x, so that by definition,

x+(-X)=0=(—x)+xThe inverse of an element x invertible for an associative composition denoted by a symbol similar to - is often denoted by x—l, so that by definition, xx‘:l = l = x‘lx.

Unless otherwise indicated, if x is invertible for an associative composition

denoted by A, we shall denote its inverse by x*. If E is either Q, R, or C, every nonzero element of E is invertible for multiplicatidn on E; however, 1 and —1 are the only integers invertible for

multiplication on Z. By inspection of the tables, every element of N3 is

invertible for addition modulo 6, but only 1 and 5 are invertible for multiplication modulo 6. Every symmetry of the square is invertible for the composition defined in Example 2.5; for example, r;1 = rs and d;1 = d1. The conclusion of Theorem 4.2 need not hold for nonassociative compositions. For example, if A is the composition on Na defined by the following table, then 0 is the neutral element, and both 1 and 2 are inverses of 1 (and also of 2) for A.

Theorem 4.3. If y is an inverse of x for a composition A on E, then x is an inverse ofy for A . Thus, an inverse of an invertible element is itself invertible.

In particular, if x is invertible for an associative composition A, then x* is invertible, and x** = x.

Proof. The first two assertions follow at once from the equalities xAy = e =yAx.

ALGEBRAIC STRUCTURES

28

Chap. I

Therefore, if A is associative and if x is invertible for A, we may use the

definite article and speak of the inverse of x. The inverse of an element x invertible for an associative composition denoted by a symbol similar to + is denoted by —x, so that by definition,

x+(—x)=0=(—x)+xThe inverse of an element x invertible for an associative composition denoted by a symbol similar to ' is often denoted by x—l, so that by definition, xx‘1 = 1 = x‘lx.

Unless otherwise indicated, if x is invertible for an associative composition

denoted by A, we shall denote its inverse by x*. If E is either Q, R, or C, every nonzero element of E is invertible for multi-

plication on E; however, 1 and —1 are the only integers invertible for multiplication on Z. By inspection of the tables, every element of N6 is invertible for addition modulo 6, but only 1 and 5 are invertible for multiplication modulo 6. Every symmetry of the square is invertible for the composition defined in Example 2.5 ; for example, r1—3l = r3 and d;1 = d1. The conclusion of Theorem 4.2 need not hold for nonassociative compositions. For example, if A is the composition on Na defined by the following table, then 0 is the neutral element, and both 1 and 2 are inverses of l (and also of 2) for A.

Theorem 4.3. If y is an inverse of x for a composition A on E, then x is an inverse ofy for A. Thus, an inverse of an invertible element is itself invertible.

In particular, if x is invertible for an associative composition A, then x* is invertible, and x** = x.

Proof. The first two assertions follow at once from the equalities

xAy = e =yAx.

28

AIGEBRAIC STRUCTURES

Chap. I

Therefore, if A is associative and if x is invertible for A, we may use the definite article and speak of the inverse of x. The inverse of an element x invertible for an associative composition denoted by a symbol similar to + is denoted by —x, so that by definition,

x+(—x)=0=(-x)+xThe inverse of an element x invertible for an associative composition denoted

by a symbol similar to - is often denoted by x‘l, so that by definition, xx—1 = 1 = x‘lx.

Unless otherwise indicated, if x is invertible for an associative composition

denoted by A, we shall denote its inverse by x*. If E is either Q, R, or C, every nonzero element of E is invertible for multiplication on E; however, 1 and —1 are the only integers invertible for

multiplication on Z. By inspection of the tables, every element of N6 is invertible for addition modulo 6, but only 1 and 5 are invertible for multiplication modulo 6. Every symmetry of the square is invertible for the composition defined in Example 2.5; for example, r;1 = r3 and d;1 = d1. The conclusion of Theorem 4.2 need not hold for nonassociative compositions. For example, if A is the composition on Na defined by the following table, then 0 is the neutral element, and both 1 and 2 are inverses of l (and also of 2) for A.

Theorem 4.3. If y is an inverse of x for a composition A on E, then x is an inverse ofy for A. Thus, an inverse of an invertible element is itself invertible. In particular, if x is invertible for an associative composition A, then x* is invertible, and x** = x.

Proof. The first two assertions follow at once from the equalities xAy = e = yAx.

Sec. 4

ALGEBRAIC STRUCTURES

29

In particular, if x is invertible for an associative composition A, then x* is invertible, and the inverse of x*, which is unique by Theorem 4.2, is x; thus,

1:" = x. If + is an associative composition and if x is invertible for +, then by Theorem 4.3, —x is invertible and

—(—x) = x.

Similarly, if - is an associative composition and if x is invertible for -, then by Theorem 4.3, x‘1l is invertible and

(x—1)—1 = x. Theorem 4.4. If x and y are invertible elements for an associative composition A on E, then xAy is invertible for A, and

(xAy)* =y*Ax*. Proof. We have

(xAy)A(y*Ax*) = ((xAy)Ay*)Ax* = (xA(yAy*))Ax* = (e)Ax* = xAx“ = e

and similarly (y*Ax*)A(xAy) = ((y*Ax*)Ax)Ay

= 0*A(x*Ax))Ay

= 0*Ae)Ay = y*Ay = e. Thus, if x and y are invertible for an associative composition +, then so

is x + y, and

—(x + y) = (—y) + (—x)If x and y are invertible for an associative composition ', then so is xy, and (xy)—1 = y—lx—l.

Sec. 4

ALGEBRAIC STRUCTURES

29

In particular, if x is invertible for an associative composition A, then x* is invertible, and the inverse of x*, which is unique by Theorem 4.2, is x; thus,

x” = x.

If + is an associative composition and if x is invertible for +, then by Theorem 4.3, —x is invertible and —(—x) = x. Similarly, if - is an associative composition and if x is invertible for -, then

by Theorem 4.3, x—1 is invertible and

(x_1)_1 = x. Theorem 4.4. If x and y are invertible elements for an associative composition A on E, then xAy is invertible for A, and (xAy)* =y*Ax*. Proof. We have

(xAy)A(y*Ax*) = ((xAy)Ay*)Ax*

= (xA(vAy*))Ax* = (e)Ax* = xAx* = e

and similarly (y*Ax*)A(xAy) = ((y*Ax*)Ax)Ay = (y*A(x*Ax))Ay

= (VAeMy = y*Ay = e. Thus, if x and y are invertible for an associative composition +, then so is x + y, and

—(x + y) = (—J’) + (-95)If x and y are invertible for an associative composition -, then so is xy, and (xy)—1 = y—lx—l.

Sec. 4

ADGEBRAIC STRUCTURES

29

In particular, if x is invertible for an associative composition A, then x* is invertible, and the inverse of x*, which is unique by Theorem 4.2, is x; thus, x** = x. If + is an associative composition and if x is invertible for +, then by Theorem 4.3, —x is invertible and —— (—x) = x.

Similarly, if - is an associative composition and if x is invertible for -, then by Theorem 4.3, x—1 is invertible and

(x—1)—1 = x. Theorem 4.4. If x and y are invertible elements for an associative composition A on E, then xAy is invertible for A, and

(xAy)* = y*Ax*. Proof. We have

(xAy)A(y*Ax*) = ((xAy)Ay*)Ax* = (xA(yAy*))Ax* = (e)Ax* = xAx* = e and similarly

(y*Ax*)A(xAy) = ((y*Ax*)Ax)Ay = (y*A(x*Ax))Ay

= U*Ae)Ay = y*Ay = e. Thus, if x and y are invertible for an associative composition +, then so is x + y, and

—(x + y) = (—y) + (-x)If x and y are invertible for an associative composition ', then so is xy, and (xy)‘1 = y—lx—1_

30

ALGEBRAIC STRUCTURES

Chap. I

Of course, if - is commutative, then we also have (xy)‘1 = x'ly—l, but in the contrary case, (xy ‘1 need not be x—ly‘l. In Example 2.5, for instance,

('1 ° ’0‘1 = df‘ = 012, but '1_1° h—1 = r30 h =d1. Similarly, to undo the result of putting on first a sweater and then a coat, one does not first remove the sweater and then the coat, but rather one removes first the coat and then the sweater.

The conclusion of Theorem 4.4 need not hold for a nonassociative composition. For example, if A is the composition on Na given by the table below, then 0 is the neutral element for A, and each element x admits a unique inverse x“, but (1A1)* ¢ 1*A1*. A 0

Theorem 4.5. Let A be an associative composition on E, and let x, y, and 2 be elements of E.

1° If both x and y commute with 2, then xAy also commutes with 2. 2° If x commutes with y and ify is invertible, then x commutes with y*. 3° If x commutes with y and if both x and y are invertible, then x* commutes with y*. Proof. If xAz = zAx and yAz = s, then (xAy)Az = xA(yAz) = xA(s)

= (xAz)Ay = (zAx)Ay = 2A (xAy).

If xAy = yAx and if y is invertible, then

y*Ax = y*A(xA(yAy*)) = y*A((xAy)Ay*) =J’*A(0’Ax)Ay*) = (y*A(yAx))Ay* = ((y*Ay)Ax)Ay* = xAy*.

Sec. 4

ADGEBRAIC STRUCTURES

31

Finally, if both x and y are invertible and if xAy = yAx, then

x*Ay* = (yAx * = (xAy)* =y*Ax* by Theorem 4.4. Now that we have begun to present some theorems of algebra, the reader

may ask why certain theorems are chosen for presentation in the text while others are, presumably, omitted. This is a perfectly reasonable question, and it will recur frequently. The answer, often, is simply that they are chosen because of their usefulness in later developments—the discussion of axioms for the natural numbers and the construction of the integers, the rationals,

the real and complex numbers, for example, to mention just a few of the basic concerns of algebra—that the beginner in algebra cannot know about at this point. Consequently, the reader should hold these questions in abeyance, since the answers will be easier to give later. Rereading earlier sections after later chapters have been studied is often rewarding, because the reader will then be able to add to his logical understanding of the statements and proofs of theorems an insight into the reasons they were chosen for presentation.

EXERCISES 4.1. Rewrite the statement and proof of Theorem 4.4 if the composition is (a) denoted by + rather than by A, (b) denoted by - rather than by A. 4.2. Rewrite the statement and proof of Theorem 4.5 if the composition is (a) denoted by + rather than by A, (b) denoted by - rather than by A. 4.3. (a) Let M be the set of married men now alive, W the set of married women

now alive. Is the statement “For every w E Wthere exists m E M such that m is the husband of w” equivalent to the statement “There exists m E M such that for every w e W, m is the husband of w” ? (b) Is the statement “For every x E E there exists e e E such that eAx =

x = x Ae” equivalent to the statement “There exists a neutral element for A” 'I (c) Let E be a set containing more than one element. Prove that for every xEE there exists ee E such that e F, g: F—> G, and h: G —> H be three functions. The composite

functions g of and h o g are then defined, as are h o (g of) and (h o g) o f.

Taking apart the meaning of the composite of two functions, we see that both h o (g o f) and (h o g) o f are, roughly speaking, the function obtained by first applying f to each element x of E, then applying g to the resulting element f(x) of F, and finally applying h to the resulting element g(f(x)) of G. Consequently, we may conjecture that (h o g) of and h o (g of) are really, the same function, namely, the function that associates to each x e E the

element h(g(f(x))) of H. The following theorem is a formal verification of this conjecture. Theorem 5.1. Iffis a function from E into F, if g is a function from F into G, and if h is a function from G into H, then

(h°g)°f=h°(g°f)Proof. The domain of both (h o g) of and h o (g o f) is the domain E off, and the codomain of both functions is the codomain H of h. For every x E E,

[(h ° g) °f](x) = (h ° 9006)) = h(g(f(x))) = h((g °f)(x)) = [h ° (8 °f)](x)Hence, (hog)of=ho(gof). Figure 7 is a pictorial representation of Theorem 5.1; all ways indicated of going from E to H are identical. In view of Theorem 5.1 , we shall write simply

h°(g°f) 6

g of

h°g°ff°r(h°g)°fDefinition. For any set E, the identity function on

\

E is the function 1,; from E into E defined by 1E(x) = x

F

for all x e E.

(h0g)°f

The graph of IE is the set of all ordered pairs Figure7

(x, x) where x GE; this set is also called the diagonal subset of E X E.

Iff is a function from E into F, clearly f° 1E =f9

1F0f=f.

34

ADGEBRAIC STRUCTURES

Chap. 1

Letf: E —> F, g: F —> G, and h: G —> H be three functions. The composite functiOns g of and h o g are then defined, as are h o (g of) and (h o g) o f. Taking apart the meaning of the composite of two functions, we see that both h o (g o f) and (h o g) o f are, roughly speaking, the function obtained by first applying f to each element x of E, then applying g to the resulting element f(x) of F, and finally applying h to the resulting element g(f(x)) of G. Consequently, we may conjecture that (h o g) o f and h o (g o f) are really the same function, namely, the function that associates to each x e E the element h(g(f(x))) of H. The following theorem is a formal verification of

this conjecture. Theorem 5.1. Iff is a function from E into F, if g is a function from F into G, and if h is a function from G into H, then

(h°g)°f=h°(g°f)Proof. The domain of both (h o g) of and h o (g of) is the domain E off, and the codomain of both functions is the codomain H of h. For every x E E,

[(11 ° 5’) °f106) = (’1 ° g)(f(X)) = h(g(f(X))) = h((g °f)(x)) = [h ° (3 °f)](x)Hence, (hog)of=ho(gof). Figure 7 is a pictorial representation of Theorem 5.1; all ways indicated of going from E to H are identical. In view of Theorem 5.1, we shall write simply

h°(g°f) g of

h°g°ff°r(h°g)°fDefinition. For any set E, the identity function on E is the function 1E from E into E defined by

\ If

l

g \

F

nW

1E(x) = x for all x EE.

(hog)°f

The graph of IE is the set of all ordered pairs (x, x) where x GE; this set is also called the diagonal subset of E x E. Iff'is a function from E into F, clearly

Figure7

f°1E=fa

1F°f=f-

Sec. 5

ALGEBRAIC STRUCTURES

35

Iff is a function, there may exist elements u, x of its domain such that u 75 x but f(u) =f(x). For example, if f: )5s from R into R, f(2) = f(—2). Functions for which this does not happen, that is, functions f for

which u 75 x implies that f(u) 7E f(x), are sufliciently important to warrant a special name:

Definition. A functionffrom E into Fis an iniection if the following condition holds: (Inj)

For all u, x E E, if u 7'5 x, thenf(u) #flx).

A function f is injective or one-to-one iff is an injection.

An equivalent formulation of (Inj) is the following: For all u, x e E, if f(u) =f(x), then u = x.

Definition. A function f from E into F is a surjection if for every element y in the codomain F off there exists at least one element x in the domain E off such that y = f(x). The function f is surjective or a function onto F if f is a surjection. For example, the function f: x 1—) x2 from R into R is not a surjection,

but the function g: x n—-> x2 from R into the set R+ of positive numbers is a surjection. Definition. A function f is a bijection or a one-to-one onto function if f is both an injection and a surjection. A function if bijective if it is a bijection. A permutation of a set E is a bijection whose domain and codomain are both E. Thus, f: E—> F is bijective if and only if for each y e F there is exactly one element x e E such that y = f(x). Consequently, iff is a bijection from E onto F, we may define a function f‘— from F into E by associating to each

y e F the unique x e E such that f(x) = y. Definition. Let f be a bijection with domain E and codomain F. The inverse off is the function f‘— with domain F and codomain E that associates to each

y e F the unique x e E such that f(x) = y. Thus, iff is a bijection from E onto F, then

y =f(x)

Sec. 5

ALGEBRAIC STRUCTURES

35

Iff is a function, there may exist elements u, x of its domain such that

u 75 x but f(u) =f(x). For example, if f: x 1—) x2 from R into R,f(2) =

f(—2). Functions for which this does not happen, that is, functions f for which u 72 x implies that f(u) 7'5 f(x), are sufficiently important to warrant a special name: Definition. A functionffrom E into F is an injection if the following condition holds:

(Inj)

For all u, x e E, if u 7'5 x, thenf(u) aéf(x).

A function f is injective or one-to-one iff is an injection. An equivalent formulation of (Inj) is the following: For all u, x e E, if

f(u) =f(x), then u = x. Definition. A'function f from E into F is a surjection if for every element y

in the codomain F off there exists at least one element x in the domain E off such that y = f(x). The function f is surjective or a function onto F if

f is a surjection. For example, the function f: x H x2 from R into R is not a surjection, but the function g: x |—) x9 from K into the set R+ of positive numbers is a surjection.

Definition. A function f is a bijection or a one-to-one onto function if f is both an injection and a surjection. A function if bijective if it is a bijection. A permutation of a set E is a bijection whose domain and codomain are both E. Thus, f: E —> F is bijective if and only if for each y e F there is exactly one element x e E such that y = f(x). Consequently, iff is a bijection from E onto F, we may define a function f“ from F into E by associating to each y e F the unique x e E such that f(x) = y.

Definition. Let f be a bijection with domain E and codomain F. The inverse offis the functionf“ with domain F and codomain E that associates to each y e F the unique x e E such that f(x) = y. Thus, iff is a bijection from E onto F, then

y =f(x)

Sec. 5

ALGEBRAIC STRUCTURES

35

If f is a function, there may exist elements u, x of its domain such that

u :15 x but f(u) =f(x). For example, if f: xI—nc2 from R into R, f(2) = f(—2). Functions for which this does not happen, that is, functions f for

which u 7E x implies that f(u) 7e f(x), are sufficiently important to warrant a special name:

Definition. A functionffrom E into Fis an injection if the following condition holds: (Inj)

For all u, x E E, if u yé x, thenf(u) ¢f(x).

A function f is injective or one-to-one iff is an injection. An equivalent formulation of (Inj) is the following: For all u, x e E, if f(u) =f(x), then u = x. Definition. A function f from E into F is a surjection if for every element y in the codomain F off there exists at least one element x in the domain E off such that y = f(x). The function f is surjective or a function onto F if f is a surjection. For example, the function f: x 1—) x” from R into R is not a surjection,

but the function g: x H x2 from R into the set R+ of positive numbers is a surjection. Definition. A function f is a bijection or a one-to-one onto function if f is both an injection and a surjection. A function if bijective if it is a bijection. A permutation of a set E is a bijection whose domain and codomain are both E. Thus, f: E —> F is bijective if and only if for each y e F there is exactly one element x e E such that y = f(x). Consequently, iff is a bijection from E onto F, we may define a function f“ from F into E by associating to each y e F the unique x e E such that f(x) = y.

Definition. Let f be a bijection with domain E and codomain F. The inverse offis the functionf“ with domain F and codomain E that associates to each y e F the unique x e E such that f(x) = y. Thus, iff is a bijection from E onto F, then

y =f(X)

36

manure STRUCTURES

Chap. I

if and only if

f“(1') = xCustomarily, the inverse of a bijection is denoted by f'1 rather than f. That notation, however, may lead to ambiguities in certain contexts, and consequently, we shall avoid denoting the inverse of a bijection f by f“.

The following table gives some familiar examples of bijections and their inverses. Domain f

Codomain f

f(x) =

f“(y) =

x20

y23

2x+3

Ky—a)

13x33

lgyg9

x'

1/?

13x32

iSySl

x"1

y“1

ngn-IZ

Ogygl

sinx

arcsiny

xgo

0g°f is a composition on EE, which we shall denote by 0. By Theorem 5.1, o is associative, and the f+og n? if m < n? [Use Exercise 5.4.]

5.6. If E is a nonempty set and iff is an injection from E into F, then there is a

function g from F into E such that g of = 1E. [Use Exercise 3.8.] 5.7. (a) Iff is the function from N into N defined by f(n) = n + 1,

then f has infinitely many left inverses (Exercise 4.11) for the composition 0

on N". (b) If g is the function from N into N defined by n_

.

ilfn is even, £01):

11—1

2

if n is odd,

then g has infinitely many right inverses (Exercise 4.11) for the composition

0 on N”. 5.8. Let E, F, G, and H be sets. For each subset A of E x F and each subset B

of F X G, the composite of B and A is the set B o A defined by BoA = {(x,z)eE x G: for someyeF, (x,y)eA and (y,z)eB},

and the set A‘— is defined by A‘— = {(y,x)eF x E: (x,y)EA}.

(a) Iff: E —> Fand ifg: F —> G, then Gr(g) o Gr(f) = Gr(gof). (b) If A, B, and C are subsets respectively ofE X F, F x G, and G x H, then (CoB)oA = Co(BoA).

(c) If A and B are subsets respectively of E x Fand F x G, then (B o A)“ = A‘— o B“. (d) If A is a subset of E x F, then A o Gr(lE) = A and Gr(1E) o A '= A.

(e) If A is a subset of E x F, then A o A“ = Gr(1F) if and only if A is the

graph of a surjection whose domain is a subset of E and whose codomain is F. (f) If A is a subset ofE x F, then A o A‘— = Gr(1F) and A“ o A = Gr(1E)

if and only if A is the graph of a bijeetion from E onto F.

40

ALGEBRAIC STRUCTURES

Chap. I

[Obtain the answer first for m = 1, 2, and 3, and then determine a formula

n n for (m) in terms of (m _ 1).] Express your answer as a quotient whose numerator is a factorial and whose denominator is a product of two factorials. 5.5. How many injections are there from a set having m elements into a set having n elements if m = n? ifm > n? ifm < n? [Use Exercise 5.4.]

5.6. If E is a nonempty set and iff is an injection from E into F, then there is a

function g from F into E such that g of = 1E. [Use Exercise 3.8.] 5.7. (a) Iff is the function from N into N defined by f(n) = n + 1,

then f has infinitely many left inverses (Exercise 4.11) for the composition 0

on N". (b) If g is the function from N into N defined by n_

.

Eifniseven,

g(n) = 2

if n is odd,

then g has infinitely many right inverses (Exercise 4.11) for the composition

o on N”. 5.8. Let E, F, G, and H be sets. For each subset A of E x F and each subset B

of F x G, the composite of B and A is the set B o A defined by BoA = {(x,z)eE x G: for someyeF, (x,y)eA and (y,z)eB},

and the set A“ is defined by A“ = {(y,x)eF x E: (x,y)eA}.

(a) Iff: E —> Fand ifg: F —> G, then Gr(g)o Gr(f) = Gr(g of). (b) If A, B, and C are subsets respectively of E x F, F x G, and G x H, then (CoB)oA = Co (BoA).

(c) If A and B are subsets respectively of E x F and F x G, then (B o A)“ = A‘— o B“. ((1) HA is a subset ofE x F, then A o Gr(lE) = A and Gr(lE) o A '= A. (e) If A is a subset of E X F, then A o A“ = Gr(l 17') if and only if A is the

graph of a surjection whose domain is a subset of E and whose codomain is F. (f) If A is a subset of E X F, then A o A“ = Gr(lf) and A“ o A = Gr(lE)

if and only if A is the graph of a bijection from E onto F.

Sec. 6

manure STRUCTURES

41

(g) The composition (B, A) i—) B o A on $(E x E), which we denote by 0, is associative with neutral element Gr(1E). A subset A of E x E is invertible

for o if and only if A is the graph of a permutation of E. If A is invertible for 0, its inverse for o is A“.

6. Isomorphisms of Algebraic Structures

In mathematics, ordered pairs or ordered triads are frequently used formally to create a new entity from two or three component parts. Similarly, ordinary language often suggests the origins of something new in naming it by joining

together in some fashion the names of its component parts, e.g., “bacon, lettuce, and tomato sandwich,” “A.F.L.-C.I.O.,” “State of Rhode Island

and Providence Plantations.” We shall define an algebraic structure eSsentially to be a nonempty set together with one or two compositions on that set: Definition. An algebraic structure with one composition is an ordered pair (E, A) where E is a nonempty set and where A is a composition on E. An algebraic structure with two compositions is an ordered triad (E, A, V) where E is a nonempty set and where A and V are compositions on E. An algebraic structure is simply an algebraic structure with either one or two compositions.

This definition is quite artificial in two respects. First, on the set E we consider only binary operations. In a more general definition, ternary operations (functions from E X E X E into E), unary operations (functions from E into E), and, in general, n-ary operations for any integer n > 0 would be allowed.

Second, we have limited ourselves to at most two compositions. A more general definition would allow any number of compositions, even infinitely many. However, algebraic structures with more than two binary operations are rarely encountered in practice. The only reason we have for imposing these limitations is that it is con-

Venient to do so. The reader is invited to consider a more general definition of algebraic structure and to modify correspondingly the concepts subsequently introduced for algebraic structures. It is customary to use expressions such as “the algebraic structure E under A ," for example, instead of the more formal “the algebraic structure (E, A).”

Two algebraic structures, though distinct, may be “just like” each other. But before we explore this concept, let us consider some nonmathematical examples of situations just like each other. In checker game (1) of Figure 9 white is to move, and in game (2) black

is to move (the initial position of the black checkers is always the lowernumbered squares).

Sec. 6

AIJGEBRAIC STRUCTURES

41

(g) The composition (B, A) i—> B o A on ‘13(E x E), which we denote by 0,

is associative with neutral element Gr(l E). A subset A of E x E is invertible

for o if and, only if A is the graph of a permutation of E. If A is invertible for 0, its inverse for o is A“.

6. Isomorphisms of Algebraic Structures

In mathematics, ordered pairs or ordered triads are frequently used formally to create a new entity from two or three component parts. Similarly, ordinary language often suggests the origins of something new in naming it by joining together in some fashion the names of its component parts, e.g., “bacon, lettuce, and tomato sandwich,” “A.F.L.-C.I.O.,” “State of Rhode Island

and Providence Plantations.” We shall define an algebraic structure eSsentially to be a nonempty set together with one or two compositions on that set: Definition. An algebraic structure with one composition is an ordered pair (E, A) where E is a nonempty set and where A is a composition on E. An algebraic structure with two compositions is an ordered triad (E, A, V) where

E is a nonempty set and where A and V are compositions on E. An algebraic structure is simply an algebraic structure with either one or two compositions.

This definition is quite artificial in two respects. First, on the set E we consider only binary operations. In a more general definition, ternary operations

(functions from E X E X E into E), unary operations (functions from E into E), and, in general, n-ary operations for any integer n > 0 would be allowed. Second, we have limited ourselves to at most two compositions. A more general definition would allow any number of compositions, even infinitely many. However, algebraic structures with more than two binary operations are rarely encountered in practice.

The only reason we have for imposing these limitations is that it is conVenient to do so. The reader is invited to consider a more general definition of algebraic structure and to modify correspondingly the concepts subsequently introduced for algebraic structures. It is customary to use expressions such as “the algebraic structure E under

A ,” for example, instead of the more formal “the algebraic structure (E, A).” Two algebraic structures, though distinct, may be “just like” each other.

But before we explore this concept, let us consider some nonmathematical examples of situations just like each other. In checker game (1) of Figure 9 white is to move, and in game (2) black

is to move (the initial position of the black checkers is always the lowernumbered squares).

Sec. 6

ADGEBRAIC STRUCTURES

41

(g) The composition (B, A) i—> B o A on ENE x E), which we denote by 0, is associative with neutral element Gr(l E). A subset A of E x E is invertible

for o if and only if A is the graph of a permutation of E. If A is invertible for 0, its inverse for o is A“.

6. Isomorphisms of Algebraic Structures

In mathematics, ordered pairs or ordered triads are frequently used formally to create a new entity from two or three component parts. Similarly, ordinary language often suggests the origins of something new in naming it by joining

together in some fashion the names of its component parts, e.g., “bacon, lettuce, and tomato sandwich,” “A.F.L.-C.I.O.,” “State of Rhode Island

and Providence Plantations.” We shall define an algebraic structure e'ssentially to be a nonempty set together with one or two compositions on that set: Definition. An algebraic structure with one composition is an ordered pair (E, A) where E is a nonempty set and where A is a composition on E. An algebraic structure with two compositions is an ordered triad (E, A, V) where E is a nonempty set and where A and V are compositions on E. An algebraic structure is simply an algebraic structure with either one or two compositions. This definition is quite artificial in two respects. First, on the set E we consider only binary operations. In a more general definition, ternary operations

(functions from E X E X E into E), unary operations (functions from E into E), and, in general, n-ary operations for any integer n > 0 would be allowed. Second, we have limited ourselves to at most two compositions. A more

general definition would allow any number of compositions, even infinitely many. However, algebraic structures with more than two binary operations are rarely encountered in practice. The only reason we have for imposing these limitations is that it is con-

Venient to do so. The reader is invited to consider a more general definition of algebraic structure and to modify correspondingly the concepts subsequently introduced for algebraic structures. It is customary to use expressions such as “the algebraic structure E under A ,” for example, instead of the more formal “the algebraic structure (E, A).”

Two algebraic structures, though distinct, may be “just like” each other. But before we explore this concept, let us consider some nonmathematical examples of situations just like each other. In checker game (1) of Figure 9 white is to move, and in game (2) black

is to move (the initial position of the black checkers is always the lowernumbered squares).

42

ALGEBRAIC STRUCTURES

/

1

27//

5

3

4

77/

e///ZJ

fiV/

l2

- V/ V/ //A

28

Zéfi

26

27



Chap. I

1// .27/ / 5%

9y

/

/

(1)

25"

6

3

7/

10/

26

/ .4 9°

11

I2

V/

(2) Figure9

Strategically, situation (1) for white is just like (2) for black. A white man (respectively, white king, black man, black king) occupies square k in (1) if

and only if a black man (respectively, black king, white man, white king) occupies square 33 — k in (2), and the move of a white piece from square k to square m in (1) has the same strategic value as the move of a black piece from square 33 — k to square 33 — m in (2). Though not identical, the two situations strategically have the same form, or to use a word whose Greek roots mean just that, they are “isomorphic” situations. A winning strategy for white in (1), for example, is the following: white moves 31 —> 26, black must jump 23 —> 30, white moves 17 —> 21, black must move 30 —> 26, white double-jumps 21 —> 30 —> 23, black must move 29 ——> 25, white moves 23 —> 26 and 26 —> 22 on his next move and wins. The corresponding strategy in (2) is obtained mechanically by interchanging black and white and replacing k by 33 — k. Thus, black moves 2 —> 7, white must move 10 —> 3, black moves

16 —> 12, etc. Similarly, any situation in checkers has a strategically isomorphic counterpart, obtained by corresponding to a man (king) on square k of the given situation a man (king) of opposite color on square 33 — k. It is easy to describe situations in other games that are strategically just like each other. If two decks of cards are stacked so that the nth card of one deck has the same denomination and suit as the nth card of the other for 1 S n S 52, then corresponding hands dealt from the two decks for any card game whatever will have the same strategic value, even though the backs of cards from different decks have different decorative patterns. For a less trivial example, consider any card game in which the strategic value of a card depends only on its color and denomination and not on its specific suit. If two decks are such that the nth card of one deck is a spade (respectively,

42

ADGBBRAIC STRUCTURES

28

/

2|

A

A

25

M

/

log10 x. That problems involving multiplication of strictly positive numbers can be transformed into problems involving addition by means of the logarithmic function is essentially just a restatement of the fact thatf‘— is an isomorphism of these two algebraic structures. Although (R, +) and (RI, -) are isomorphic, (R, +, -) and (R1, -, +) are not, since R has a neutral element for multiplication but R: has no neutral element for addition. Our final theorem will assist us in making certain constructions in later chapters. Occasionally in algebra we are given an algebraic structure (E, A) and a bijectionffrom E onto a set F, and we wish to “transplant” A from E to F by means off, that is, we wish to construct a composition V on F such

thatfis an isomorphism from (E, A) onto (F, V). This can always be done, and moreover, it can be done in only one way. Theorem 6.3. (Transplanting Theorem) Let (E, A) be an algebraic structure, and let f be a bijection from E onto a set F. There is one and only one composition V on F such thatfis an isomorphism from (E, A) onto (F, V), namely, the composition V defined by

(3)

“V9 =f(f ‘1q“(10)

for all u, v e F.

Proof. To show that there is at most one composition having the desired properties, we shall show that if V is a composition on F such that f is an isomorphism from (E, A) onto (F, V), then V satisfies (3). Indeed, as f is an isomorphism and as f of“ = 1F,

f(fh(“)Af‘-(9)) =f(f“(14))Vf(f '"(v)) = q.

It remains for us to show that the composition V defined by (3) has the desired properties. Let x, y e E, and let u = f(x), v = f(y). Then x = f“(u) and y = f“(0), so

f(my) =f(f"(q“(10) = W0 =f(x)Vf(y). Thus, f is an isomorphism from (E, A) onto (F, V).

50

ADGEBRAIC STRUCTURES

Chap. I

It is natural to call the composition v defined by (3) the transplant of A

underf If F = E and iffis an automorphism of (E, A), then the transplant of A under f is, of course, A itself.

In Example 6.3 we calculated the transplant of a composition on {a, b} under J. The transplant V of multiplication on Z under the bijection f: n HZn from Z onto the set E of even integers is given by

=2-M=

4

Lil"

nvm =f(f(n) rm» =f (g- 12'!) nm.

The transplant V of multiplication under the bijection f: x-n—y 10” from R onto the set R: of strictly positive real numbers is given by

“V0 =f(f“(14) 'f“(19) =f((1°8 “)(108 v)) = 1000310003») = (lologu)logv = “log 9.

If two algebraic structures are isomorphic, it is sometimes easy to find a “natural” isomorphism between them. But if we suspect that two algebraic structures are not isomorphic, how should we go about trying to prove that they are not? If the sets involved are infinite,' it may be impossible to examine all possible bijections from one onto the other, and if they are both finite with a large number of elements, it may be too arduous to show that none of the bijections is an isomorphism. The best procedure is to try to find some property of one structure not possessed by the other that would have to be preserved by any isomorphism. We have already seen, for example, that neither the even integers under multiplication nor the strictly positive real numbers under addition are isomorphic either to the integers or to the real numbers under multiplication, because the latter two algebraic structures have neutral elements while the former two do not. A more diflicult problem

is to prove that the nonzero real numbers R*'under multiplication and the real numbers R under addition are not isomorphic. Both compositions are associative, commutative, possess neutral elements, and all elements of either

set are invertible for the composition in question, so we must look further for some algebraic property that differentiates the two. Such a property is the following: There are two elements of R* that are their own inverses for multiplication, namely, —1 and 1, but there is only one element of R that is its own

Sec. 6

ALGEBRAIC STRUCTURES

51

inverse for addition, namely, 0. Iff were an isomorphism from (R*, -) onto

(R, +), then

0 =f(1) =f((—1)(-1))

= f(—1)+ f(—1)= 2f(-1), whence

f(—1)= 0 =f(1), sof would not be injective, a contradiction.

EXERCISES If E S C, the set of nonzero members of E is denoted by E*, and if E S R, the

set of strictly positive members of E is denoted by E}; 6.1. Given a deck of cards to be used to play a game in which the strategic value of a card depends on its denomination only and not on its suit (e.g., certain forms of tummy), describe (4!)18 decks of cards that strategically have the same form as the given deck. How many of these have the same strategic form as the given deck for a game in which the strategic value of a card depends on its denomination and color, but not on its specific suit?

How many decks “isomorphic” to the given one can you describe for canasta? [Special attention must be given the 3’s.] 6.2. Complete the proof of Theorem 6.1. 6.3. Complete the proof of Theorem 6.2. State and prove assertions similar to those of Theorem 6.2 concerning left neutral elements, right neutral elements (Exercise 4.5), left inverses, right inverses (Exercise 4.11), and the commuta-

tivity of two elements.

6.4. For what real numbers at is the function f: xi-_'>.ax an automorphism of (R, +) ? For what strictly positive real numbers a is the function g: XI—) a” an isomorphism from (R, +) onto (R1, ~)?

*6.5. Show that multiplication modulo 5 is a composition on Ng". Exhibit two isomorphisms from (N4, +4) onto (Ng‘, -5). Show that these two are the

only ones. ”‘66. Let v and A be the compositions on R defined by

x v y = maX{x,y}, x A y = min{x, y}.

Sec. 6

ADGEBRAIC STRUCTURES

51

inverse for addition, namely, 0. Iff were an isomorphism from (R*, -) onto

(R, +), then 0 =f(1) =f((-1)(-1)) = f(—1)+ f(—1)= 2f(—1), whence

f(—1)= 0 =f(1), so f would not be injective, a contradiction.

EXERCISES If E s C, the set of nonzero members of E is denoted by E*, and if E s R, the

set of strictly positive members of E is denoted by El“. 6.1. Given a deck of cards to be used to play a game in which the strategic

value of a card depends on its denomination only and not on its suit (e.g., certain forms of rummy), describe (4!)13 decks of cards that strategically

have the same form as the given deck. How many of these have the same strategic form as the given deck for a game in which the strategic value of a card depends on its denomination and color, but not on its specific suit? How many decks “isomorphic” to the given one can you describe for canasta? [Special attention must be given the 3’s.]

6.2. Complete the proof of Theorem 6.1. 6.3. Complete the proof of Theorem 6.2. State and prove assertions similar to those of Theorem 6.2 concerning left neutral elements, right neutral elements (Exercise 4.5), left inverses, right inverses (Exercise 4.11), and the commuta-

tivity of two elements. 6.4. For what real numbers at is the function f: x|->. aux an automorphism of (R, +) ? For what strictly positive real numbers at is the function g: x H at”

an isomorphism from (R, +) onto (Rj, -)?

*6.5. Show that multiplication modulo _5 is a composition on Ng". Exhibit two isomorphisms from (N4, +4) onto (Nf, -5). Show that these two are the

only ones.

*6.6. Let v and A be the compositions on R defined by x v y = max{x,y}, x A y = min{x, y}.

Ammo STRUCTURES

52

Chap. I

(8.) Exhibit an isomorphism from (R, v) onto (R, A) and one from (R, -) onto (R, -). (b) Prove that (R, v, -) is not isomorphic to (R, A, -).

[If f were an isomorphism, show first that x > f‘—(0) would imply that f(x) < 0, and then use the fact that every square is positive] 6.7. Let Z[\/§] = {m + m5: m, nEZ}. (a) The sum and product of two elements of Z[\/§] belong to Z[V§];

consequently, addition and multiplication are compositions on 2[1/3].

(b) The function fzm + m/giflm — 111/3 is an automorphism of

(Zn/5], +, -). (c) Multiplication is a composition on the set E = {2"8": m, n E Z}. Exhibit an isomorphism from (Z[ 1/3], +) onto (E, -). 6.8. Determine the transplant of (a) multiplication on R under the permutation f: xii—g 1 — x of R;

(b) addition on R: under the permutation f: x973 x” of R:; (c) addition on R: under the bijectionf: x H loglo x from R: onto R. 6.9. Prove that (0*, -) is not isomorphic with (R*, -), (C, +), or (R, +). *6.10. Prove that (Q, +) is not isomorphic with (Q1, -). [Use Exercise 1.10.] *6.11 . (a) Every permutation of E is an automorphism of (E, ). (b) If E contains at least two elements, then (E, F. The range off is the subset {yey =f(x) for somee} of F. Let G be a subset of F that contains the range off. The function obtained by restricting the codomain off to G is the function

x 'I-.>f(x) with domain E and codomain G. Thus, the function obtained by restricting the codomain of a function f

has the same domain and graph asfbut, in general, has a smaller codomain. The function obtained by restricting the codomain of a function to its range is clearly a surjection; the function obtained by restricting the codomain of

an injection to its range is clearly a bijection. Definition. Let f: E —> F, g: D —> G. We shall say that f is an extension of g to a function from E into F, or that g is a restriction offto a function from D into G, ifD S E, G s F, and g(x) =f(x) for all x e D. Thus, if D E E and G E F, then there is a (unique) restriction off to a

function from D into G if and only if f(x) e G for all x e D. Clearly, if g: D —> G is a restriction off: E —> F, then

mw=mmnwxa

Sec. 8

ALGEBRAIC STRUCTURES

59

In particular, let A be a composition on E, and let A s E. If there is a restriction of A to a function from A X A into A, that restriction will be a composition on A; such a restriction exists if and only if xAy GA for all x, y e A.

Definition. Let A be a composition on E. A subset A of E is stable for A, orclosed under A, if xAy e A for all x, y e A.

If A is stable for a composition A on E, we shall denote the restriction of A to A X A by AA when it is necessary to emphasize that it is not the same as the given composition on E; but when no confusion would result, we shall simply drop the subscript and use the same symbol to denote both the given composition and its restriction. The composition AA is called the compo-

sition induced on A by A. Example 8.1. If m is a positive integer, the set of all integral multiples of m is a subset of Z stable for both addition and multiplication on Z. If m and p are positive integers, the set of all integral multiples of m

that are greater than F is another subset of Z stable for both addition and multiplication. Example 8.2. The sets of nonzero integers, of nonzero rationals, of

nonzero real numbers, and of nonzero complex numbers are all stable for multiplication. When is the set N; of all nonzero elements of N,” stable for multiplication modulo m? If m > 1 and if N; is stable, then m is a prime number; otherwise, m = rs where r and s are strictly positive integers less than m, and consequently, r and s belong to N5” but, since rs = m, r -,,, s = 0. Later we shall prove the

converse (cf. Theorems 19.9 and 19.11): If m is prime, then N; is stable for multiplication modulo m. Example 8.3. The sets E and (D are stable for every composition A on E. If e is a neutral element for A, then {e} is stable for A. More

generally, if x is any element of E satisfying xAx = x, then {x} is

stable for A. Example 8.4. Let V, be the composition on N defined by x v y = max{x, y}.

Then (N, v) is a semigroup whose neutral element is zero, and every subset of N is stable for v.

60

ALGEBRAIC STRUCTURES

Chap. I

Example 8.5. It is easy to enumerate all the stable subsets of the group G of symmetries of the square: They are (0, {r0}, {rm r3}, {rm 1:}, {’0’ U}, {’0’ d1}: {’0’ d8}, {70, r1: r2: r3}: {rm 7'2, ha 1’}, {r05 r23 d1 d2}: and

G. Let (E, A) be an algebraic structure, and let A be a subset of E stable

for A. What properties does AA inherit from A ? Surely if A is associative or commutative, so also is A.4- In addition, every element of A cancellable for A is again cancellable for AA. If E has a neutral element e for A, three

possibilities arise: (l) A contains e, in which case e is, of course, the neutral element for AA ; (2) there is no neutral element for AA (for example, if A is

the set of all strictly positive integers, then A is stable for addition, but there is no neutral element for addition on A); (3) A does not contain e, but there is, nevertheless, a neutral element for AA (in Example 8.4, for instance, the

set A of all strictly positive integers is stable for v: and 1 is the neutral element for VA, although 0 is the neutral element for v). The third possibility cannot arise if every element of E is cancellable, for then e is the only element x

satisfying xAx = x: Theorem 8.1. If e is the neutral element for a composition A on E, then e is the only cancellable element x e E satisfying xAx = x. Proof. If x is a cancellable element satisfying xAx = x, then x = e since xAx = x = e. Definition. Let H be a nonempty stable subset of an algebraic structure (E, A). If (H, A H) is a semigroup, we shall say that (H, A H) is a subsemigroup

of (E, A), and if (H, A H) is a group, we shall say that (H, A H) is a subgroup of (E, A). ' If (H, A H) is a subsemigroup (subgroup) of (E, A), the subset H of E is

also called a subsemigroup (subgroup) of (E, A). Depending on the context, therefore, the words “subsemigroup” and “subgroup” may refer either to a set or to an algebraic structure. Each of the nonempty stable subsets of the group of symmetries of the square is, for example, a subgroup.

Theorem 8.2. If a subsemigroup of a group has a neutral element, that element is the neutral element of the group. The assertion follows at once from Theorem 8.1 and the corollary of Theorem 7.1.

A nonempty subset of a semigroup is a subsemigroup if and only if it is stable. However, a nonempty stable subset of a group need not be a subgroup; for example, N is a subsemigroup but not a subgroup of the group

(Z, +)-

60

ALGEBRAIC STRUCTURES

Chap. I

Example 8.5. It is easy to enumerate all the stable subsets of the group G of symmetries of the square: They are (2), {r0}, {rm r2}, {’0’ h}, {’09 U}, {rm d1}: {’0’ d8}, {’09 ’1’ r2, r3}: {’0’ r2: h, 0}: {r01 '2’ d1 d2}: and

G. Let (E, A) be an algebraic structure, and let A be a subset of E stable

for A. What properties does AA inherit from A? Surely if A is associative or commutative, so also is AA. In addition, every element of A cancellable for A is again cancellable for AA. If E has a neutral element e for A, three possibilities arise: (1) A contains e, in which case e is, of course, the neutral

element for AA; (2) there is no neutral element for AA (for example, if A is

the set of all strictly positive integers, then A is stable for addition, but there is no neutral element for addition on A); (3) A does not contain e, but there is, nevertheless, a neutral element for AA (in Example 8.4, for instance, the

set A of all strictly positive integers is stable for v: and 1 is the neutral element for VA, although 0 is the neutral element for v). The third possibility cannot arise if every element of E is cancellable, for then e is the only element x

satisfying xAx = x: Theorem 8.1. If e is the neutral element for a composition A on E, then e is the only cancellable element x e E satisfying xAx = x. Proof. If x is a cancellable element satisfying xAx = x, then x = e since x Ax = x = x Ae.

Definition. Let H be a nonempty stable subset of an algebraic structure (E, A). If (H, A H) is a semigroup, we shall say that (H, A H) is a subsemigroup of (E, A), and if (H, A H) is a group, we shall say that (H, A H) is a subgroup of (E, A). ' If (H, A H) is a subsemigroup (subgroup) of (E, A), the subset H of E is

also called a subsemigroup (subgroup) of (E, A). Depending on the context, therefore, the words “subsemigroup” and “subgroup” may refer either to a set or to an algebraic structure. Each of the nonempty stable subsets of the group of symmetries of the square is, for example, a subgroup. Theorem 8.2. If a subsemigroup of a group has a neutral element, that element is the neutral element of the group. The assertion follows at once from Theorem 8.1 and the corollary of Theorem 7.1.

A nonempty subset of a semigroup is a subsemigroup if and only if it is stable. However, a nonempty stable subset of a group need not be a subgroup; for example, N is a subsemigroup but not a subgroup of the group

(2, +)-

Sec. 8

ALGEBRAIC STRUCTURES

61

We shall frequently apply either criterion 2° or 3° of the following theorem to determine whether a given nonempty subset of a group is a subgroup.

Theorem 8.3. Let H be a nonempty subset of a. group (G, A). The following statements are equivalent:

1° H is a subgroup of G. 2° For all x, y e G, if H contains x and y, then H contains xAy and y*. 3° For all x, y e G, if H contains x and y, then H contains xAy*.

Proof. Statement 1° implies 2°, for by Theorem 8.2, the neutral element for AH is also the neutral element e for A, and consequently, for every y e H, the inverse of y for AH is the unique inverse y* of y for A. Statement 2°

implies 3°, for if x 'and y belong to H, then x and y* belong to H by 2°, whence xAy* e H again by 2°. Finally, 3° implies 1°: There exists an element a in H by hypothesis, so e e H by 3° since e = aAa*. Also ify e H, then y* e H by 3° since y* = eAy*. Moreover, H is stable for A, for if x, y e H, then by what we have just proved, x and y* belong to H, so as

xA(y*)* = xAy, xAy also belongs to H by 3°. Thus, H is stable for A, H contains a neutral

element for AH, every element of H has an inverse in H for AH, and AH is, of course, associative since A is; therefore, (H, A H) is a group.

If- the composition of the group G is denoted by + rather than A, 2° and 3° become respectively: For all x, y e G, if H contains x and y, then H contains x + y and —y. For all x, 'y e G, if H contains x and y, then H contains x + (—y). If the composition of G is denoted by - rather than A, 2-0 and 3° become

respectively:

For all x, y e G, if H contains x and y, then H contains xy and y-l. For all x, y e G, if H contains x and y, then H contains xy‘l. Definition. Let (G, A) be a group. The center of G is the set Z(G) consisting ofall x e G such that xAy =yAx for everyy e G. A group G is abelian, of course, ifand only if Z(G) = G.

Sec. 8

ADGEBRAIC smucrmuas

61

We shall frequently apply either criterion 2° or 3° of the following theorem to determine whether a given nonempty subset of a group is a subgroup. Theorem 8.3. Let H be a nonempty subset of a. group (G, A). The following statements are equivalent:

1° H is a subgroup of G. 2° For all x, y e G, if H contains x and y, then H contains xAy and y*. 3° For all x, y E G, if H contains x and y, then H contains xAy*. Proof. Statement l° implies 2°, for by Theorem 8.2, the neutral element for

A H is also the neutral element e for A, and consequently, for every y e H, the inverse of y for AH is the unique inverse y* of y for A. Statement 2° implies 3°, for if x and y belong to H, then x and y* belong to H by 2°, whence xAy* e H again by 2°. Finally, 3° implies 1°: There exists an element a in H by hypothesis, so e eH by 3° since e = aAa*. Also ify EH, then y* e Hby 3° since y* = eAy*. Moreover, His stable for A, for ifx,y e H, then by what we have just proved, x and y* belong to H, so as xA(y*)* = xAy, xAy also belongs to H by 3°. Thus, H is stable for A, H contains a neutral element for AH, every element of H has an inverse in H for AH, and AH is, of course, associative since A is; therefore, (H, A H) is a group.

If- the composition of the group G is denoted by + rather than A, 2°

and 3° become respectively: For all x, y e G, if H contains x and y, then H contains x + y and —y. For all x, 'y e G, if H contains x and y, then H contains x + (—y). If the composition of G is denoted by - rather than A, 2“ and 3° become

respectively: For all x, y e G, if H contains x and y, then H contains xy and y‘l. For all x, y e G, if H contains x and y, then H contains xy‘l.

Definition. Let (G, A) be a group. The center of G is the set Z(G) consisting of all x e G such that xAy = yAx for every y e G.

A group G is abelian, of course, ifand only if Z(G) = G.

Sec. 8

ALGEBRAIC STRUCTURES

61

We shall frequently apply either criterion 2° or 3° of the following theorem to determine whether a given nonempty subset of a group is a

subgroup.

Theoremp8.3. Let H be a nonempty subset of a. group (G, A). The following statements are equivalent: 1° H is a subgroup of G. 2° For all x, y e G, if H contains x and y, then H contains xAy and y*.

3° For all x, y e G, if H contains x and y, then H contains xAy*. Proof. Statement 1° implies 2°, for by Theorem 8.2, the neutral element for AH is also the neutral element e for A, and consequently, for every y e H,

the inverse of y for AH is the unique inverse y* of y for A. Statement 2° implies 3°, for if x‘and y belong to H, then x and y* belong to H by 2°, whence xAy* e H again by 2°. Finally, 3° implies 1°: There exists an element a in H by hypothesis, so e e H by 3° since e = aAa*. Also ify e H, then y* E H by 3° since y* = eAy*. Moreover, H is stable for A, for if x, y e H, then by what we have just proved, x and y* belong to H, so as xA(y*)* = xAy, xAy also belongs to H by 3°. Thus, H is stable for A, H contains a neutral

element for AH, every element of H has an inverse in H for AH, and A H is, of course, associative since A is; therefore, (H, AH) is a group.

If- the composition of the group G is denoted by + rather than A, 2° and 3° become respectively: For all x, y e G, if H contains x and y, then H contains x + y and —y.

For all x, y e G, if H contains x and y, then H contains x + (—y). If the composition of G is denoted by - rather than A, 2" and 3° become

respectively:

For all x, y e G, if H contains x and y, then H contains xy and y‘l. For all x, y e G, if H contains x and y, then H contains xy“.

Definition. Let (G, A) be a group. The center of G is the set Z(G) consisting

of all x e G such that xAy = yAx for everyy e G. A group G is abelian, of course, if and only if Z(G) = G.

62

ALGEBRAIC STRUCTURES

Chap. I

Theorem 8.4. The center of a group (G, A) is a subgroup. Proof. Clearly the neutral element of G belongs to Z(G). If x, y e Z(G), then

xAy e Z(G) and x* e Z(G) by 1° and 2° of Theorem 4.5. Thus, by Theorem 8.3, Z(G) is a subgroup.

Theorem 8.5. Let (E, A) be a semigroup with neutral element e. The set G

of all invertible elements of E is a subgroup of E. Proof. If x and y are invertible, so is xAy by Theorem 4.4. Thus, G is a stable subset of E. Clearly, e e G, and if x e G, then x* e G by Theorem 4.3. Also AG is clearly associative as A is. Thus, G is a group.

From our discussion in §5, the subgroup of invertible elements of the semigroup (EE, 0) of all functions from E into E is the group (SE of all

permutations of E. Theorem 8.6. The set of all automorphisms of an algebraic structure E is a subgroup of SE. The assertion follows at once from 2° of Theorem 8.3 and Theorem 6.1.

We shall denote by Aut(E) the group of all automorphisms of an algebraic structure E. Definition. A group (5 is a permutation group on E if (5 is a subgroup of the group 6E. Permutation groups were studied before the general definition of a group was formulated. When groups first began to be studied, mathematicians wondered if the concept of a group were really more general than that of a

permutation group, that is, if there existed a group not isomorphic to any permutation group. The answer, furnished by Cayley, was negative; every group is isomorphic to a permutation group. Theorem 8.7.- (Cayley) Let (G, A) be a group, and for each a e G, let L, be

the permutation of G defined by La(x) = a. Then the function L: a 1-39 L,

is an isomorphism from (G, A) onto a permutation group on G. Proof: By Theorem 7.1, L, is indeed a permutation of G. If L, = L,, then a = aAe = L,(e) = Lb(e) = bAe = b.

Sec. 8

ADGEBRAIC smucrumzs

63

Therefore, L is injective. For every x e G,

(La ° Lb)(x) = L..(Lb(x)) = La(b) = aA(b) = (aAb)Ax

=

am“);

Therefore, La ° Lb =

«Ab

for all a, b e G. Thus, the set (5 = {L,,: a e G} is a stable subset of Ga, and L is an isomorphism from G onto (5. Consequently, (5 is a group and, therefore, is a permutation group on G. The isomorphism L of Theorem 8.7 is called the left regular representation of (G, A). Let (1,, a2, . . . , a, be elements of a set E such that a, yé a, whenever i 72 j.

We shall denote by (a1, a2, . . . , a,) the permutation a of E defined as follows: 0(a‘) =a¢+1ifl Si< r,

“(‘19) = a1,

0(x) = x for all x ${a1, . . . , a,}.

Such permutations are called cycles. With this notation, for example,

(1, 2, 4) ° (3, 4, 2) = (1, 2, 3)Certain permutation groups have interesting geometrical models. For example, let G be the group of symmetries of the square. For each u e G, let a“ be the permutation of {1, 2, 3, 4} describing the efl'ect of u on the corners of the square. Thus, 0'1; = (I! 2) ° (3: 4)

since 12 sends corner 1 into corner 2, corner 2 into corner 1, corner 3 into corner 4, and comer4 into corner 3, and since (1 , 2) o (3, 4) is the permutation taking 1 into 2, 2 into 1, 3 into 4, and 4 into 3. Similarly,

0,1 = (l, 2, 3, 4),

0,, = (l, 4) o (2, 3),

or, = (1: 3) ° (2’ 4):

ad; = (1’ 3):

6r, = (l: 4: 3a 2):

ad. = (2: 4):

ALGEBRAIC STRUCTURES

64

Chap. I

and 0,0 is, of course, the identity permutation. It is easy to verify that Gilli—)0“

is an isomorphism from G onto a subgroup (5 of 6‘. Indeed, instead of proving directly that the composition of G is associative, it is easier to prove that a is an isomorphism from G onto (5; since G is, therefore, isomorphic

to a group, we conclude that the composition of G is associative. Similarly,

the group of symmetries of any polygon of n sides is isomorphic to a subgroup of 6”.

EXERCISES 8.1. Which subsets of E are stable for the compositions . on E?

8.2. Let E be a finite set of n elements. For how many compositions on E is every subset of E stable? Of these, how many are commutative?

8.3. If F S E, is EMF) a stable subset of ME) for U? for n? for A (Example

7.4)? For which of those compositions is the set of all finite subsets of E stable? For which is the set of complements of all finite subsets of E stable?

8.4. Show that H = {3,4, 5, 6} is a subgroup of the semigroup (N7, +3.4) (Exercise 2.8). Is the neutral element of H the neutral element of N,? Is

every element of H invertible for the composition +3“, on N,? 8.5. Write out the table for the group of (the ten) symmetries of a regular pentagon. Exhibit an isomorphism from it onto a subgroup of 65. 8.6. Denote each element of 63 other than the identity permutation as a cycle, and write out the table for the composition of the group (53. Exhibit an isomorphism from the group of (the six) symmetries of an equilateral triangle onto 63. Which subgroups of 653 are isomorphic to the group of symmetries of an isosceles nonequilateral triangle? 8.7. Let R* be the set of all nonzero real numbers, and let E = R" x R.

.(a) Let A be the composition on E defined by

(a, b)A(C,d) = (ac, ad + b) for all (a, b), (c, d) e E. Show that (E, A) is a group. (b) For each (a, b) e E let flu, be the function from R into R defined by fa'b(x) = ax + b,

64

ALGEBRAIC STRUCTURES

Chap. I

and (1,0 is, of course, the identity permutation. It is easy to verify that Gilli—>0“

is an isomorphism from G onto a subgroup (5 of (54. Indeed, instead of proving directly that the composition of G is associative, it is easier to prove that a is an isomorphism from G onto 6; since G is, therefore, isomorphic

to a group, we conclude that the composition of G is associative. Similarly, the group of symmetries of any polygon of n sides is isomorphic to a subgroup of G".

EXERCISES 8.1. Which subsets of E are stable for the compositions ' on E? 8.2. Let E be a finite set of n elements. For how many compositions on E is every subset of E stable? Of these, how many are commutative? 8.3. If F s E, is s13(1‘7') a stable subset of ENE) for U? for n? for A (Example

7.4)? For which of those compositions is the set of all finite subsets of E stable? For which is the set of complements of all finite subsets of E stable?

8.4. Show that H = {3,4, 5, 6} is a subgroup of the semigroup (N7, +3.0 (Exercise 2.8). Is the neutral element of H the neutral element of N,? Is every element of H invertible for the composition +a.4 on N-,?

8.5. Write out the table for the group of (the ten) symmetries of a regular pentagon. Exhibit an isomorphism from it onto a subgroup of 65. 8.6. Denote each element of 63 other than the identity permutation as a cycle, and write out the table for the composition of the group 63. Exhibit an isomorphism from the group of (the six) symmetries of an equilateral triangle onto 63. Which subgroups of 63 are isomorphic to the group of symmetries of an isosceles nonequilateral triangle? 8.7. Let R“ be the set of all nonzero real numbers, and let E = R* x R.

(a) Let A be the composition on E defined by (a, b) A(c, d) = (ac, ad + b) for all (a, b), (c, (1)6 E. Show that (E, A) is a group.

(b) For each (a, b) e E let f“, be the function from R into R defined by fa,,,(x) = ax + b,

Sec. 8

ADGEBRAIC STRUCTURES

65

and let

(5 = {f,,_,,: (a, b)eE}. Show that (5 is a permutation group on R isomorphic with the group (E. A) described in (a). Construct a table for each composition on {a, b, c} satisfying the following conditions: b is the neutral element, c is cancellable, a is invertible but not

cancellable, and {a} is not stable. 8.9. Let A be a composition on E, and let A = {ez (xAy)Az = xA(yAz) for ally, zeE}. If A 95 0, then A is a subsemigroup of (E, A).

8.10 If (G, A) is a group and if a e G, then the set of all elements of G commuting

with a is a subgroup. 8.11. Let f be a function from E into E. (a) The function f is left cancellable (Exercise 7.6) for the composition 0 on

EE if and only iff is injective, and f is right cancellable if and only iff is surjective. Conclude that the subsemigroup of cancellable members of E” coincides with the subgroup of its invertible members. (b) The function f is idempotent (Exercise 2.17) for o if and only if the function obtained by restricting both the domain and codomain off to its range F is the identity function. on F. 8.12. If E is a commutative semigroup, is the set of all noncancellable elements of

E a subsemigroup ? ‘8.l3. If H and K are subgroups of a group G neither of which contains the other,

then there exists an element of G belonging neither to H nor to K. ‘8.l4. Let E = {a, b, c} be a set of three elements. Let M, a, ’61, «a: 3’3, and

9 be respectively the set of all compositions A on E such that the group of automorphisms of (E, A) is 6E, {1E, (a, b, c), (a, c, b)}, {1E, (a, b)},

{1E, (a, c)}, {1E, (b, c)}, and {115;} respectively, and let V = «’1 U «2 U 93. (a) Show that .9! contains three members. [Use Exercise 6.ll(c).] Write

out the table for the unfamiliar composition belonging to .21. Show that 9 contains 33 — 3 members and that each of ((1, “a, and «’3 contains 3‘ — 3 members. Conclude that 9 has 19,422 members.

(b) Show that the isomorphic class determined by a composition in a! consists of that composition alone, that the claSS determined by a composition in 9 contains one other composition, which is also from .93, that the

class determined by a composition in ? contains three compositions, one each from 3’1, 1’2, and fl, and that the class determined by a composition in 9 contains five other compositions, all from 9. Conclude that there are

Sec.8

ALGEBRAIC STRUCTURES

65

and let

0; ={fl,',,:(a,b)eE}. Show that 03 is a permutation group on R isomorphic with the group (E, A) described in (a). Construct a table for each composition on {a, b, c} satisfying the following conditions: b is the neutral element, c is cancellable, a is invertible but not

cancellable, and {a} is not stable. 8.9. Let A be a composition on E, and let A = {x 6E: (xAy) Az = xA(yAz) for ally, zeE}. If A ab 0, then A is a subsemigroup of (E, A).

8.10. If (G, A) is a group and if a e G, then the set of all elements of G commuting

with a is a subgroup. 8.11. Letf be a function from E into E. (a) The function f is left cancellable (Exercise 7.6) for the composition 0 on

EE if and only iff is injective, and f is right cancellable if and only iff is surjective. Conclude that the subsemigroup of cancellable members of EE coincides with the subgroup of its invertible members. (b) The function f is idempotent (Exercise 2.17) for o if and only if the

function obtained by restricting both the domain and codomain off to its range F is the identity function. on F. 8.12. If E is a commutative semigroup, is the set of all noncancellable elements of

E a subsemigroup ? III8.13. If H and K are subgroups of a group G neither of which contains the other, then there exists an element of G belonging neither to H nor to K.

*8.14. Let E = {a, b, c} be a set of three elements. Let 4!, Q, g1, V2, *3, and

9 be respectively the set of all compositions A on E such that the group of automorphisms of (E, A) is (5E, {1E, (a, b, c), (a, c, b)}, {1E, (a, b)},

{113, (a, c)}, {1E, (b, c)}, and {1E} respectively, and let if = ((1 U V, U «’3. (a) Show that .2! contains three members. [Use Exercise 6.11(c).] Write

out the table for the unfamiliar composition belonging to .91. Show that 93 contains 33 - 3 members and that each of “6'1, «2, and «’3 contains 34 — 3

members. Conclude that 9 has 19,422 members. (b) Show that the isomorphic class determined by a composition in .9! consists of that composition alone, that the class determined by a composition in 9 contains one other composition, which is also from 5?, that the

class determined by a composition in 3’ contains three compositions, one each from @1, (6’2, and V3, and that the class determined by a composition

in 9 contains five other compositions, all from 9. Conclude that there are

66

ALGEBRAIC STRUCTURES

Chap. I

3,330 isomorphic classes; that is, there are to within isomorphism 3,330

compositions on' E. (c) For how many compositions in each of the six classes is there a neutral element? Conclude that there are to within isomorphism 45 compositions on E admitting a neutral element. (d) How many compositions in each of the six classes are commutative? Conclude that there are to within isomorphism 129 commutative compositions on E.

CHAPTER II

THREE BASIC NUMBER SYSTEMS

In this chapter we shall axiomatize the natural number system and then construct in a formal manner two other familiar number systems, the

integers and the rational numbers. In plane geometry certain of the most self-evident statements about lines, points, angles, and circles are chosen as postulates; here also we shall select as postulates certain statements about the natural numbers that are particularly self-evident. Whether one mathematical statement is more self-evident than another is a psychological, not a mathematical, matter, and there are several alternative sets of postulates one may choose, all approximately equally self-evident. Incidentally, there is no mathematically compelling reason for preferring as postulates selfevident rather than unfamiliar though true statements about the natural numbers; indeed, some mathematicians delight in choosing as postulates for some familiar mathematical system certain true but unfamiliar statements and then deriving as theorems the more self-evident statements. Our discussion of the natural numbers will not present any information that the reader does not already know, but rather it will serve to show how familiar facts about the natural numbers and finite sets may be derived from certain axioms. Consequently, readers who are willing to accept as “known” facts familiar to everyone about the natural numbers and finite sets may omit most of §§11—12, and similarly, readers willing to accept the integers as known

and who are uninterested in their construction from the natural numbers may omit most of §14. (For a more precise statement, see the footnote at the beginning of each of those sections.) In §11 we shall postulate the existence of a set N, a composition + on N,

and an “ordering” g on N that satisfy certain conditions. Since the concept of an ordering is fundamental in all mathematics, we shall first consider it separately.

67

68

THREE BASIC NUMBER SYSTEMS

Chap. II

9. Orderings

The concept of a “relation” as used in everyday discourse is too fundamental

to be defined in simpler terms, although, as we shall shortly see, a very simple definition may be given that is adequate for mathematics. In any event, before we can talk sensibly about a relation on a set E, we must be sure that

for any elements x and y of E, either the relation holds between x and y or else it does not hold between x and y. Seizing upon this fact, we make the following informal definition of a relation: A relation on a set E is a linguistic expression that may contain _-b1anks and . . .-blanks but contains blanks of no other kind, such that for every

x e E and every y e E, if “x” is inserted in every _-blank and “y” in every . . .-blank, the resulting expression is a sentence that is either true or false. The following, for example, are relations on the set of all people now alive: (a) (b) (c) (d) (e) (f)

__ loves . . .. . . . is the husband of _. __ and . . . have the same parents. . . . and __ are first cousins. __ and . . . have at least two grandparents in common. — is no taller than . . . .

(g) _ has the same parents and siblings as . . . . (h) _ is the mother of . . . .

(i) (j) (k) (1) (m) (11)

_ is either an ancestor of . . . or the same person as . . . . _ is either a parent of . . . or the same person as . . . . _ and . . . are descendents of George III. _ is either at least as tall as . . . or at least as heavy as . . . . _ loves cheese. The Sistine Chapel is in Rome.

No claim is made that the informal definition of “relation” given here corresponds exactly with the intuitive concept one may have. For example, it may seem contrary to ordinary usage to call (m) and (n) relations on the set of all people now alive, even though they satisfy the informal definition given.

If R is a relation, we shall say that x bears R to y if the sentence obtained by putting “x” in every _-blank and “y” in every .. .-b1ank is true. If x bears R to y, we shall write x R y, and if not, we shall write x R y. A relation R on E is reflexive if x R x for all x e E, symmetric if for all

x, y e E, x R y implies that y R x, antisymmetric if for all x, y e E, x R y and y R x together imply that x = y, transitive if for all x, y, 2 GE, x R y and y R 2 together imply x R 2. Our human condition is such that relation (a), alas, is neither reflexive, symmetric, nor transitive, but at least it is not antisymmetric. A relation

Sec. 9

THREE BASIC NUMBER SYSTEMS

69

that is both symmetric and antisymmetric is also transitive (Exercise 9.1); consequently, there are at most 24 — 2 possible combinations of the four properties defined that a given relation may possess; that there are, in fact,

14 possibilities is shown by the examples listed above. Thus, (m) is transitive but neither reflexive, symmetric, nor antisymmetric. Since (11) is a false statement, relation (11) is symmetric, antisymmetric, and transitive, but not

reflexive.

To each relation R on E we may associate the subset of E X E consisting of all ordered pairs (x, y) such that x R y; this set is called the truth set of R. Every subset A of E X E is the truth set of at least one relation on E, namely,

the relation (_,...)eA.

Knowing how to determine whether any given ordered pair in E X E belongs to the truth set of a given relation on E is certainly a long advance towards understanding the full meaning of the relation, and it is an epistemological question whether such knowledge should be regarded as the same as complete understanding of the meaning. Happily, the issue may be avoided in mathematics: Just as a set is completely determined by its elements (that is, E and F are identical sets if they have the same elements), so also two relations

on a set of mathematical objects are regarded as the same relation if they have the same truth sets. We may, therefore, regard a relation on E simply as a certain subset of E X E, namely, its truth set. Consequently, we make

the following formal definition: Definition. A relation on E is a subset of E X E. A relation R on E is a reflexive relation on E if (x, x) e R for all x e E; R is a symmetric relation on E if (x, y) e R implies that (y, x) e R for all x, y e E; R is an antisymmetric relation on E if (x, y) e R and (y, x) e R together imply that x = y for all x,

y e E; R is a transitive relation on E if (x, y) e R and (y, 2) 6R together imply that (x, z) e R for all x, y, z e E.

As before, we shall write x R y and say that x bears R to y if (x, y) e R. In discussing relations we shall freely use both formal and informal defini-

tions. The relation

on E is, for example, simply the diagonal subset of E X E.

Definition. A relation R on a set E is an ordering on E if R is reflexive, antisymmetric, and transitive. An ordering R on E is a total ordering if for all x,

yeE, either n orn.

THREE BASIC NUMBER SYSTEMS

70

Chap. 11

On the set of all British monarchs, the relation

M: _ was monarch after or at the same time as . . . is a total ordering, but the relation

_ was president of the United States after or at the same time as . . . on the set of all presidents of the United States is not antisymmetric because of President Cleveland’s nonconsecutive terms. On the set of all persons who have ever lived, the relation

D: _ is a descendant of or the same person as . . . is an ordering that is not total, and on the set of all straight lines in the plane of analytic geometry, the relation

L: _ is parallel to . . . , and if . . . is not parallel to the Yaxis, then _ coincides with or lies below . . . , but if . . . is parallel to the Y-axis, then _ coincides with or lies to

the right of . . . is also an ordering that is not total (a line is considered parallel to itself). If F is a set having more than one element, the relation

9 : __ is contained in . . . on the set ‘B(F) is an ordering that is not total. Definition. Let R be a relation on a set E, and let A be a subset of E. The restriction of R to A, or the relation induced on A by R, is the relation RA

satisfying xRAyifand onlyify

for all x, y e A. If S is a relation on subset A of E, relation R on E is an extension of S if RA = S. Thus, by definition, RA = R n (A x A).

If R is respectively reflexive, symmetric, transitive, or antisymmetric, then RA is also; consequently, if R is an ordering or a total ordering on E, then

Sec. 9

THREE BASIC NUMBER SYSTEMS

71

RA is also on A. To avoid cumbersome notation, we shall use the symbol denoting a given relation on E also to denote the relation it induces on A unless confusion results. Figure 10 shows how an ordering R on a finite set may be represented

diagrammatically. Each small circle represents an element of the set, and the line segments

connecting the circles are so drawn that x R y if and only if either x = y or

(1)

(2)

(3)

Ms (5 )

Figure 10

there is an ascending path of line segments joining the circle representing x to the circle representing y. Thus, diagram (1) is the diagram for the restriction of M to the set of all

British monarchs x such that x M Victoria and Elizabeth II M x. Diagram

(2) is the diagram for the restriction of D to the set of all persons x such that x D Terah and Joseph D x (we assume that Abraham is truthful in Genesis 20:12). Diagram (3) is the diagram for the restriction of L to the set of all lines parallel to and one unit away from either the X-axis or the Y-axis.

Definition. A relational structure is an ordered pair (E, R) where E is a set and R a relation on E. If R is an ordering on E, the relational structure

(E, R) is also called an ordered structure.

Sec. 9

THREE BASIC NUMBER SYSTEMS

71

RA is also on A. To avoid cumbersome notation, we shall use the symbol denoting a given relation on E also to denote the relation it induces on A unless confusion results. Figure 10 shows how an ordering R on a finite set may be represented

diagrammatically. Each small circle represents an element of the set, and the line segments connecting the circles are so drawn that x R y if and only if either x = y or

(1)

(2)

(3)

, (4)

(5)

(6)-

Figure 10

there is an ascending path of line segments joining the circle representing x to the circle representing y. Thus, diagram (1) is the diagram for the restriction of M to the set of all British monarchs x such that x M Victoria and Elizabeth II M x. Diagram (2) is the diagram for the restriction of D to the set of all persons x such that x D Terah and Joseph D x (we assume that Abraham is truthful in Genesis 20:12). Diagram (3) is the diagram for the restriction of L to the set of all lines parallel to and one unit away from either the X-axis or the Y-axis. Definition. A relational structure is an ordered pair (E, R) where E is a set and R a relation on E. If R is an ordering on E, the relational structure (E, R) is also called an ordered structure.

72

THREE BASIC NUMBER SYSTEMS

Chap. II

Corresponding to the concept of an isomorphism between algebraic structures, an isomorphism of two relational structures is a bijection preserving everything in sight; Definition. Let (E, R) and (F, S) be relational structures. An isomorphism from (E, R) onto (F, S) is a bijection f from E onto F such that for all x, y e E,

x Ry if and only iff(x) Sf(y). Relational structures (E, R) and (F, S) are isomorphic if there is an isomorphism from one onto the other. Theorem 9.1. Let (E, R), (F, S), and (G, T) be relational structures. 1° The identity function is an isomorphism from (E, R) onto itself.

2° Iffis a bijection from E onto F, thenf is an isomorphism from (E, R) onto (F, S) if and only if f“ is an isomorphism from (F, S) onto (E, R). 3° Iff is an isomorphism from (E, R) onto (F, S) and if g is an isomorphism from (F, S) onto (G, T), then g of is an isomorphism from (E, R) onto (G, T). Definition. Let R be a relation on E. We define R“ to be the relation on E

satisfying xR"yifand only ifn for all x, y e E.

If R is defined by a linguistic expression involving __-blanks and . . .blanks, R“ is obtained either by replacing every _-blank by a . . .-b1arik and every . . .-blank by a __-blank or by replacing certain crucial words and phases by their opposites. For example, M‘— is the relation _ was monarch before or at the same time as . . . on the set of British monarchs. Clearly, R““ = R

for any relation R.

Theorem 9.2. If R is a reflexive (respectively, symmetric, antisymmetric transitive) relation on E, then R“ is also. If f is an isomorphism from a

Sec. 9

THREE BASIC NUMBER SYSTEMS

73

relational structure (E, R) onto a relational structure (F, S), thenfis also an isomorphism from (E, R") onto (F, S“).

If R is an ordering on a finite set, the diagram for R“ is obtained simply by turning the diagram for R upside down. Symbols similar to s and 2 (e.g., s and 2 , < and k) are most frequently used to denote orderings. If g is an ordering, then s“ is denoted by 2. Similar, if 2 is an ordering, then 2“ is denoted by 3. If g is an ordering on E, for all x, yeE we write x y if x 2 y and x sh y. It is customary to use such' words as “less,” “smaller,” “greater,” and “larger” and the corresponding superlatives when discussing orderings denoted by these symbols. However, the meaning of such words depends not only on the ordering involved but also on the particular symbol chosen to denote it. For example, if R is an

ordering and if x R y, one would say “x is less than or equal to y” if s were chosen to denote R, but in contrast “x is greater than or equal to y” if 2 were chosen. Consequently, the use of such words for orderings not denoted by symbols similar to s or 2 could easily lead to confusion.

An ordering s on E is total if and only if for all x, y e E, either x < y or x = y or x > y, a condition known as the trichotomy law.

Definition. A well~ordering on E is a total ordering S such that every nonempty subset A of E possesses a smallest element, i.e., such that every nonempty subset A of E contains an element a satisfying a g x for all x e A. The familiar ordering g on R is not a well-ordering; indeed, if A =

{x e R: x > 1}, then A has no smallest member, for if a e A, then §(a + l) is a yet smaller element of A. It is intuitively evident that a total ordering on a finite set is a wellordering; we shall formally prove this in §12. Clearly, if s is a well-ordering on E and if A is a subset of E, the induced ordering SA on A is a well-

ordering. Definition. Let (E, g) and (F, f0’»; f is strictly increasing (strictly decreasing) if for all x, y eE, ifx nm on N is called multiplication. In view of the properties characterizing gm, nm may be described as the result of adding m to itself n times. The definition of nm is a special case of the following definition: Definition. Let A be a composition on E, and let a e E. If there is a neutral element e for A, we define A”a by

Na = gaOI)

Sec. 11

THREE BASIC NUMBER SYSTEMS

87

for all n e N, where g, is the unique function from N into E satisfying ga(0) = e~

ga(r + 1) =' ga(r)Aa for all r e N. If there is no neutral element for A , we define A”a by A"a = f,(n)

for all n e N*, where f; is the unique function from N* into E satisfying

fa(1) = a, fa(r + 1) =fiz(r)Aa for all r E N*.

If a composition is denoted by a symbol similar to + instead of A, we write n.a or simply na for (Va, and if it is denoted by a symbol similar to -, we shall write a" for A”a. Thus, by definition, Ala = a,

A”"'1a = (A”a)Aa for all n e N*, and furthermore, if there is a neutral element e for A,

A°a = e. A familiar rule of arithmetic is that multiplication is distributive over addition, that is, that

m(n +p) = mn + mp,

(m+")P=mP+"P for all numbers m, n, and p.

Definition. A composition v on E is distributive over a composition A on

E if xV(yAz) = (xVy)A(s),

(xAy)v'z = (s)A(s), for all x, y, z e E.

Each of the compositions U and n on SB(E), for example, is distributive over the other and over itself by Theorem 3.1. The distributivity of multiplication over addition on N is one consequence of the following theorem.

Sec. 11

THREE BASIC NUMBER SYSTEMS

87

for all n e N, where g, is the unique function from N into E satisfying ga(0) = e-

ga(r + 1) =' ga(r)Aa for all r e N. If there is no neutral element for A , we define A"a by

Na = f«(n) for all n e N*, where fa is the unique function from N* into E satisfying

fa(1) = a,

fa(r + 1) = fa(r)Aa for all r e N*. If a composition is denoted by a symbol similar to + instead of A, we write n.a or simply na for A"a, and if it is denoted by a symbol similar to -, We shall write a" for A”a. Thus, by definition,

Ala = a,

A”+1a = (A"a)Aa for all n e N*, and furthermore, if there is a neutral element e for A,

A°a = e.

A familiar rule of arithmetic is that multiplication is distributive over addition, that is, that m(n +p) = mn + mp, (m+n)p=mp+np

for all numbers m, n, and 17. Definition. A composition v on E is distributive over a composition A on

E if xv(yAz) = (xvy)A(s),

(xAy)v'z = (xvz)A(s), for all x, y, z e E.

Each of the compositions U and n on ‘B(E), for example, is distributive over the other and over itself by Theorem 3.1. The distributivity of multiplication over addition on N is one consequence of the following theorem.

88

THREE BASIC NUMBER SYSTEMS

Chap. 11

Theorem 11.8. Let A be a composition on E, and let a and b be elements of E. If A is associative, then for all n, m E N*, (l)

A”+’"a = (A"a)A(A”‘a).

If A is associative and if aAb = bAa, then for all n, m e N*,

(2)

(Ama)A(A"b) = (A"b)A(A""a ,

(3)

A”(aAb) = (A”a)A(A”b).

If there is a neutral element e for A, then (1), (2), and (3) hold if either n = 0 -orm=0,andforallmeN, (4)

Ame = e.

Proof. We shall assume that A is associative. Let n e N*, and let S be the set of all m e N* for which (1) holds. Then 1 e S, for A"+1a = (A"a)Aa = (A”a)A(A1a) by definition. If m e S, then m + l e S, for An+ 0, then m — 1 EN," and Nm—l = Nm _ {m _ 1}'

Theorem 12.3. If n and m are natural numbers, then N" ~ Nm if and only if n = m. '

Proof. The condition is sufi‘icient by 1° of Theorem 12.1. Necessity: Let S={neN:qbN,,forallmeN,,}. Clearly, 0 GS, for N0 = (2). Let n ES. To prove that n + 1 eS, let

m e N..+1. Ifm = 0, then Nm'qb Nn+1 since N0 = Q) and Nn+l 7e (0. Ifm > 0 and if Nm ~ Nu“, then N,,,,_1 ~N,, by Theorem 12.2 and m — 1 < n by Theorem 11.3, in contradiction to our assumption that n e S. Thus, n + l e S. By induction, therefore, S = N, so Nm 7‘! N” whenever m < n.

Consequently, by 2° of Theorem 12.1, Nm 79M, whenever m 3/: n.

Thus, by Theorem 12.3 and 3° of Theorem 12.1, if E ~ N”, then E 7L Nm for all m 75 n. In view of this fact, we may make the following definition. Definition. A set E is finite if there exists a natural number n such that

E~N,,. If E is finite, the unique natural number n such that E~Nn is called the number of elements in E. A set E is infinite if E is not finite.

We shall also say that E has n elements if E is finite and the number of elements in E is n. By 3° of Theorem 12.1 , to Show that a set has n elements, it suffices to Show that it is equipotent to a set having n elements.

102

THREE BASIC NUMBER SYSTEMS

Chap. 11

Theorem 12.4. If E has n + 1 elements and if a GE, then E — {a} has n elements. The assertion is an immediate consequence of Theorem 12.2. Definition. A subset E of a set F is a proper subset of F if E 9E F. The set F

properly contains E if E is a proper subset of F.

The symbol C means “is a proper subset of,” and 3 means “properly contains.” We, therefore, write E C F or F D E if E is a proper subset of F.

Theorem 12.5. If E is a proper subset of F and if F has n elements, then E is finite and has fewer than n elements. Consequently, every subset of a set having n elements is finite and has at most n elements. Proof. Let S be the set of all natural numbers m such that every proper subset of any set having m elements is finite and has fewer than m elements. Clearly, 0 e S, for Q) is the only set having zero elements, and there are no

proper subsets of (2). Let m e S. To Show that m + 1 e S, let A be a proper subset of a set B having m + 1 elements. Then there exists b e B such that b ¢A, so A E B — {b}, which has m elements by Theorem 12.4. If A = B — {b}, then A has m elements; if A C B — {b}, then A is finite and has fewer than m elements, since m E S; since m < m + 1, therefore, A has fewer than m + 1 elements. Hence, m + 1 e S. By induction, therefore, S = N. In particular, n E S, and the proof is complete.

Theorem 12.6. Let E be a set having n elements. Iff is a surjection from E onto F, then F is finite and has at most n elements, and furthermore, F has

exactly n elements if and only iff is a bijection. Proof. Since E has n elements, there is a surjection from E onto F if and only if there is a surjection from N” onto F. Therefore, we need only consider the case where E = N”. For each y e F the set E, defined by

E. = {x e E=f(x) = y} is not empty asfis surjective. Therefore, by (NO 1) we may define a function g from F into E by n) = the smallest number in E,

for ally e F. In F, then g(y) e E”, whencef(g(y)) = y. Thus,fo g = 1F, so g is injective by Theorem 5.2. The function obtained by restricting the codomain of g to its range E’ is thus a bijection from F onto E'. By Theorem

Sec. 12

THREE BASIC NUMBER SYSTEMS

103

12.5, E’ is finite, and m s n where m is the number of elements in E’. By

Theorem 12.1, F has m elements. Suppose that m = n. Then by Theorem 12.5, E' = E, so g is a bijection from F onto E. Therefore,

f=f°lm =f°g°g" =1F°g" =g"In particular, f is a bijection from E onto F.

Theorem 12.7. Let E and F be finite sets having the same number of elements, and let f be a function from E into F. The following statements are equivalent: 1° f is bijective. 2° f is injective.

3° f is surjective. Proof. By Theorem 12.5, 2° implies 3°, for iff is injective, then the function obtained by restricting the codomain off to its range F’ is a bijection, so the subset F’ of F has the same number of elements as F and consequently is

1“. By Theorem 12.6, 3° implies 1°. Theorem 12.8. The set N of natural numbers is infinite.

Proof. The function s defined by s(n) = n + 1 for all n e N is an injection from N into N that is not a bijection, for

s(n) 2 0 + 1 > 0 for all n e N, and consequently, 0 is not in the range of 3. Therefore, N is

infinite by Theorem 12.7. Theorem 12.9. If (E, g) is a totally ordered structure, then every nonempty finite subset A of E has a greatest and a least member.

Proof. Let S be the set of all n e N* such that every subset of E having n elements has a greatest and a least member. Clearly, I e S. Letn e S. To show that n + 1 e S, let A be a subset of E having n + 1 elements. Then there exists b e A, and A — {b} has n elements by Theorem 12.4. Consequently, A — {b} has a greatest element c and a least element a as n e S. The greater of c and b is then clearly the greatest element of A, and the lesser of a and b is the least element of A. Therefore, n + 1 e S. Consequently, S = N* by induction, and the proof is complete.

104

THREE BASIC NUMBER SYSTEMS

Chap. II

EXERCISES 12.1. If every element of a finite semigroup E is cancellable, then E is a group. [Use Theorem 12.7 and Exercise 7.13.]

12.2. A set containing an infinite set is infinite. 12.3. IfE1~F1andis ~F2,thenE1 x E2 ~F1 x F2. 12.4. IfE~GandifF~H,thenFE~HG. 12.5. If E, F, and G are sets, then (F x G)E ~ FE x GE. 12.6. If E, F, and G are sets, then (GE)? ~ GE x173 *12.7. A set E is denumerable if E is equipotent to N. A set is countable if it is

either finite or denumerable. If E is a nonempty set, the following statements are equivalent:

1° E is countable. 2° There is a surjection from N onto E. 3° There is an injection from E into N.

[To show that 2° implies 3°, let f be a surjection from N onto E, and consider the function g from E into N defined by g0!) = the smallest number in {x e N: f(x) = y}

for all y e E. To show that 3° implies 1°, show first that any infinite subset A of N is denumerable; for this, consider the function h from A into N

defined by

h(x) = the number of elements in A n N, for all x e A.]

12.8. Every subset of a countable set is countable, and every infinite subset of

a denumerable set is denume'rable. *12.9. Let E be a set. There is a proper subset F of E such that E ~ F if and only if E contains a denumerable subset. [To show that the condition is neces-

sary, apply the Principle of Recursive Definition to a bijection s from E onto F.]

12.10. (Cantor’s Theorem) If E is a set, there is no surjection from E onto SB(E). [Iff is a function from E into $(E), Show that {x e E: x ¢f(x)} is not in the range off] Infer that $(N) is an uncountable set.

12.11. If (E, s) and (F, , |(n, 0)] was an isomorphism from (N, +) onto

a stable subset of (N X N)/R. A discussion and construction of the “inverse-

completion” of a commutative semigroup having cancellable elements are

presented in the exercises. We next wish to extend the total ordering on Z to a total ordering on Q compatible with addition so that the product of any two positive numbers is

positive. As we shall see, this can be done in only one way. Definition. Let (A, +, -) be a ring. An ordering S on A is compatible with the ring structure of A if g is compatible with addition and if for all x, y e A, (OR)

ifx 2 0 and ify 2 0, then xy 2 0.

If g is an ordering on A compatible with its ring structure, we shall say that (A, +, -, g) is an ordered ring. An element x of an ordered ring A is positive if x 2 0, and x is strictly positive if x > 0.

The set of all positive elements of an ordered ring A is denoted by A+, and the set of all strictly positive elements of A is denoted by A1. If (A, +, ', S) is an ordered ring and if g is a total ordering, we shall, of course, call (A, +, -, g) a totally ordered ring; if (A, +, -) is a field, we shall call (A, +, -, 3) an ordered field, and if, moreover, s is a total ordering, we shall call (A, +, -, g) a totally ordered field.

It follows at once from Theorem 14.6 that (Z, +, -, g) is a totally ordered ring. If (A, + , -) is any ring, the equality relation on A is an ordering

THREE BASIC NUMBER SYSTEMS

148

Chap. H

compatible with the ring structure of A, and for certain rings this is the only

compatible ordering (Exercise 17.8). The following theorem is a collection of important and easily proved facts about ordered and totally ordered rings. Theorem 17.6. Let x, y, and 2 be elements of an ordered ring A.

1° x 0, and hencex s Oifand onlyif—x 20.

5° Ifx > 0, then n.x > 0for allneN*.

6° Ifx Syandifz 20,thenxz Syzandzx 32y. 7° Ifx syandifzso,thenxz 2yzandzx 22y. [f A is a totally ordered ring, the following also hold:

8° Ifxy>0,theneitherx>0andy>0, orx 0, and x < 0 if and only if x—1 < 0.

Our next theorem shows that an ordering s compatible with the ring structure of a ring A is completely determined by the set of its positive elements. Theorem 17.7. If A is an ordered ring and if P = A+, —P = {—x: x EA+}, then (P l)

P is stable for addition,

(P 2)

P n (—P) = {0},

(P 3)

P is stable for multiplication.

Furthermore, if A is a totally ordered ring, then (P 4)

P U (—P) = A.

Conversely, if A is a ring and if P is a subset of A satisfying (P 1), (P 2), and (P 3), then there is one and only one ordering s on A compatible with the ring structure of A such that P = A+. Furthermore, if (P 4) is also satisfied, then s is a total ordering.

Sec. 17

THREE BASIC NUMBER SYSTEMS

149

Proof. If x 2 0 and if y 2 0, then x+y2x2m

so P is stable for addition. By 4° of Theorem 17.6,

—P={xeA:x g0}. Hence, (P 2) holds, for if x 61’ n (—P), then x 2 0 and x s 0, so x = 0. Clearly, (P 3) is equivalent to (OR). Also, if s is a total ordering, then (P 4) holds since for every x E A, either x 2 0 or x s 0.

Conversely, let P be a subset of a ring A satisfying (P l), (P 2), and (P 3). By 2° of Theorem 17.6, there is at most one ordering on A compatible with the ring structure of A such that P = A+, namely, that satisfying

x gyifandonlyify—xeP. _ so defined has the requisite It remains for us to show that the relation< properties. For all x e A, x s x since x — x 6P by (P2). If x Oby (OR), so x 2 0, and hence, 2 GP; thus, Q g P.

We shall denote again by s the unique total ordering on the field Q of rational numbers that is compatible with its ring structure and induces on Z its total ordering Thus, (Q, +, ', g) is a totally ordered field.

Definition. Let (A, +, -, s) and (A’, +', °’, g’) be ordered rings. An

isomorphism from A onto A’ is a bijection f from A onto A’ that is also an isomorphism from the ordered group (A, +, 3) onto the ordered group (A’, +', S') and an isomorphism from the semigroup (A, -) onto the semigroup (A’, -'). An automorphism of an ordered ring is an isomorphism from itself onto itself. As before, the identity function is an automorphism of an ordered ring,

a bijection f is an isomorphism from an ordered ring A onto an ordered ring A’ if and only iff‘— is an isomorphism from the ordered ring A’ onto A, and the composite of two isomorphisms is again one.

Theorem 17.9. If K and L are totally ordered quotient fields of totally ordered integral domains A and B respectively and if f is an isomorphism from the ordered ring A onto the ordered ring B, then there is one and only one isomorphism g from the ordered field K onto the ordered field L extending f, and moreover,

22 =f(_x)

for all x EA, y e A*.

g(y )

f(y)

Sec. 17

THREE BASIC NUMBER SYSTEMS

151

Proof. By Theorem 17.5 it suffices to prove that if x, u e A and if y, u 6 AL

S

cl:

‘clx

then

@SM

if and only if

f(y) f(v)'

If x/y S u/v, then

xv = £0103 E (w) = uy y

I)

as yv > 0, and conversely if xv S uy, then

y

yU

yU

U

as l/yv > 0. Asf(y)f(v) = f(yv) > 0, similarlyf(x)[f(y) S f(u)[f(v) if and only if f(x)f(v) s f(u)f(y). But as f is an isomorphism from the ordered ring A onto B, xv S uy if and only iff(xv) Sf(uy), that is, if and only if

f (x)f(v) Sf(10f(y)EXERCISES 17.1. (a) If A = {m'+ 2n1/5: m, n e Z}, then A is a subdomain of R, and the

quotient field of A is {r + 31/5: r, s e Q}. (b) If B = {m + 5n\/§:m, 1162}, then B is a subdomain of R. What subfield of R is its quotient field? (c) Prove that the integral domains A and B are not isomorphic. 17.2. Let C = {m + in: m, neZ}, andlet D = {m + int/5: m,neZ}. (a) Show that C and D are subdomains of C. What subfields of C are

their quotient fields? (b) Prove that C and D are not isomorphic integral domains.

(c) Prove that a subdomain of C containing 1' cannot be isomorphic with a subdomain of R.

17.3. If K is a quotient field of an integral domain A, then K is also a quotient field of every subdomain of K containing A. 17.4. Let 6 be the composition on Z x 2* defined by (m, n) 6 (p, q) =

(mg + np, nq). Is (Z x Z*, (-9) a commutative group? 17.5. Complete the proof of Theorem 17.2: (a) Show that addition is commutative and that |(0,1)| is the neutral element for addition.

THREE BASIC NUMBER SYSTEMS

152

Chap. 1]

(b) Show that multiplication is associative, commutative, and distributive over addition and that |(l , 1)] is the identity element for multiplication.

(c) Show that A’ is a subdomain of K'. 17.6. Prove Theorem 17.6.

17.7. (a) Let K be a quotient field of a totally ordered integral domain A that is not a field, and let P = A+. If 5’ is the ordering on K defined by P, then (K, +, -, s’) is an ordered field. Which of 8°—11° of Theorem 17.6 hold

for S'? (b) What part of the assertion obtained by deleting the words “totally” and “total” from the statement of Theorem 17.7 is true?

17.8. If A is a boolean ring (Exercise 15.21), then the equality relation is the only ordering on A compatible with its ring structure. Show by a specific example that if E is a nonempty set, the ordering S on WE) is not compatible with the ring structure of ENE) (Example 15.2). In the remaining exercises, (E, -) is a commutative semigroup that has at least one cancellable element, and C is the set of all cancellable elements of E. An inverseeompletion of E is a semigroup (G, -) that contains (E, -) algebraically such that

1° Every element of C is invertible in (G, ), 2° For each 2 e G there exist x e E, y e C such that z = xy“1.

(We note that we could not hope to find a semigroup (G, -) containing (E, -) algebraically in which a noncancellable element of E had an inverse, for ifan element of E has an inverse in G, it is a cancellable element of G by Theorem 7.1 and a fortiori is a cancellable element of the given semigroup (E, -).) 17.9. Rewrite the definition of an inverse—completion if the composition of E is denoted by + rather than -. 17.10. Prove that C is a stable subset of (E, -). *17.11. Let (G, -) be an inverse-completion of (E, -).

(a) G is a commutative semigroup. (b) Every cancellable element of G is invertible; i.e., G is an inversecompletion of itself. (c) If E possesses a neutral element e, then e is also the neutral element of G (d) The inverse of an invertible element of (E, ) is also its inverse in G. *17.12. Let R be the relation on E x C satisfying (a, b) R (c, d) if and only if ad = cb.

(a) Show that R is an equivalence relation on E x C. (b) Let G’ be the quotient set (E x C)/R, and show that the composition

on G’ defined by K“. b)| |(a, d)| =‘ I016; 174)] for all |(a, b)[, |(c, d)| e G’ is well defined.

Sec. 17

THREE BASIC NUMBER SYSTEMS

153

(c) Let SEC, and let E’ = {|_(fl_,_s)_|:aeE}. Show that E’ is a stable subset of G’, and that f: a H, |(as, s)] is an isomorphism from E onto E’. (d) If t is any other element of C, then [(at, t) = |(as, s)| for all a E E.

(e) Show that G’ is an inverse-completion of E’. (f) Conclude that there is an inverse-completion G of E. [Modify the proofs of Theorems 17.2 and 17.4.] *17.13. Let f be an isomorphism from (E, -) onto a semigroup (F, -), and let G

and H be respectively inverse-completions of E and F. Show that the function g: G —> H defined by

flab—1) = f0!)!"01)“ for all a e E, b e C is well-defined, and that g is an isomorphism from G

onto H extending f [Modify the proof of Theorem 17.5.] Conclude that if G and G’ are inverse-completions of E, there is an isomorphism g from

G onto G’ satisfying g(a) = a for all a e E. 17.14. What subsemigroup of (R, -) is the inverse-completion of (E, -) if

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

E = Z? E is the set of all positive even integers ? E is the set of all positive odd integers ? E is the set of all integers 2 5 ‘2

(e) E = {a + bV22a,bEZ}? (f) E = {a + bfl/E: a, b e N*}? [First show that every strictly positive

number of the form a — b1/i where a, b EN* belongs to the inversecompletion] 17.15. What subsemigroup of (R, +) is the inverse completion of (E, +) if (a) E = Z?

(b) E is the set of negative integers? (c) E is the set of all integers 2 5 ? (d) E is the set of all positive even integers 7 I"17.16. Let (E, +, S) be a totally ordered commutative semigroup all of whose elements are cancellable, and let (G, +) be an inverse-completion of (E, +).

The relation 3’ on G satisfying

x—y s’z—wifandonlyifx+wSy+z for all x, y, z w eE is a well-defined relation, and s’ is the only total ordering on G that is compatible with addition and induces the given ordering s on E. [First show that for all a, b, c 6E, if a + c _ xa from H into Ha is a bijection. Theorem 18.3. (Lagrange) Let (G, -) be a finite group having n' elements, and let H be a subgroup of G having m elements. Then m I n, and both the set G/H of left cosets of Hand the set of right cosets of H have n/m members.

Proof. Let k be the munber of left (right) cosets of H in G. As the left (right) cosets form a partition of G and as each left (right) coset has m elements by Theorem 18.2, the total number n of elements of G is km.

Thus, m | n and k = n/m. The number of elements in a finite group is called the order of the group; an infinite group is said to have infinite order. If H is a subgroup of a finite

group G, the number of left cosets of H in G is called the index of H in G. By Lagrange’s Theorem, order G = (order H)(index H).

Simple though it is, Lagrange’s Theorem is fundamental to the study of finite groups. For example, in determining the subgroups of a finite group G, by virtue of Lagrange’s Theorem we need only seek subgroups whose order divides that of G.

Here is a simple criterion for cosets to be identical. Theorem 18.4. If H is a subgroup of a group (G, -) and if a, b e G, then

aH = bH if and only if b‘1a e H, and Ha = Hb if and only if‘ab‘1 e H. Proof. If aH = bH, then a E bH, so a = bh for some h e H, whence b-la = h e H. Conversely, if b—la e H, then a = b(b‘1a)e bH, so aH n

bH ¢ (0, whence aH = bH by Theorem 18.1. A similar argument establishes the assertion concerning right cosets.

GROUPS AND RINGS

158

Chap. III

Definition. A subgroup H of a group (G, -) is a normal subgroup of G if aH = Ha for every a e G. If H is a normal subgroup of G, we shall call a left coset of H in G

simply a coset of H in G, since it is also a right coset. The importance of normal subgroups arises from the fact that if H is a normal subgroup of G, we may introduce a composition on G/H making it a group in a fairly natural way. Before doing this, we list some conditions equivalent to normality. Theorem 18.5. Let H be a subgroup of a group (G, -). The following statements are equivalent:

1° H is a normal subgroup of G. 2° 3° 4° 5°

aHa‘1 S H for all a e G. a—lHa E HforallaeG. HE aHa—lforallaeG. H S a—lHa for all a e G.

Proofi If H is a normal subgroup and if a e G, then aH = Ha, so clearly (aH)a'1 = (Ha)a—1 = H(aa'1) = H. Thus, 1° implies 2° and 4°. If 2° holds, then in particular for each a e G, a‘1H(a'1)'1 E H, that is, a‘lHa S 'H. Therefore, 2° implies 3°. If 3° holds, then for each a e G, a(a—1Ha)a—1 S aHa‘l, whence H S aHa‘l. Thus, 3°

implies 4°. Similarly, 4° implies 5°, and 5° implies 2°. Therefore, statements 2°—5° are all equivalent. If they hold, then aHa—1 = H for each aEG, whence (aHa‘1)a = Ha, that is, aH = Ha. Consequently, all five statements

are equivalent. Clearly, {e} and G are normal subgroups of G. If G is an abelian group,

every subgroup of G is normal. More generally, every subgroup of G contained in the center of G is clearly a normal subgroup of G. However,

a normal subgroup of G need not be contained in its center, as the following example shows. Example 18.1. Let (G, o) be the group of symmetries of the square (Example 2.5). It is easy to verify that the center of G is {rm r,}. The index of the subgroup H = {r0, r1, r2,ra} is 2 by Lagrange’s Theorem, and consequently, H is a normal subgroup of G not con-

tained in its center by the following theorem: Theorem 18.6. If H is a subgroup of a finite group (G, -) whose index is 2, then H is a normal subgroup of G.

GROUPS AND RINGS

Sec. 18

159

Proof. If aeH, then aH = H = Ha. Let a ¢H. Since there are exactly two left (right) cosets of H in G, one of which is H itself, the remaining left coset is aH and the remaining right coset is Ha; since the left (right) cosets of H form a partition of G, aH ==H°

Ha.

Example 18.2. The subgroup L = {rm h} of the group G of symmetries of the square is not a normal subgroup of G. Indeed, the left cosets of L are {ro, h}, {r1, d2}, {rm 1:}, (rs, d1}, and the right cosets of L are {’02 h}, {71, d1}, {’29 17}, {rss d2}-

Let H be a subgroup of a group (G, -). We should like to define a composition on G/H by (l)

(aH) - (bH) = abH

for all aH, bH e G/H. We cannot simply declare that (1) defines a composition on G/H, for the right side of (1) appears to depend on the particular choice of elements in the two given left cosets of H. For (1) to define a composition, we must show that if aH = a’H and bH = b’H, then abH =

a’b’H. That this is not always the case is shown by Example 18.2, for there rloL= dgoL and rsoL=raoL, but (rlor3)oL=L;é (dgora)oL= v o L. To see what conditions H must satisfy for (1) to define a composition, suppose that for all a, a’, b, b’ eG, if aH = a’H and bH = b'H, then abH = a’b’H. Letting a = xh, a’ = x, b = b’ = x—1 where x e G, h eH,

we must have xhx—lH = H, that is, xhx‘1l e H. Thus, for (1) to define a composition on G/H, we must have xHx-1 s H for all x e G, so H must

be a normal subgroup of G. Actually, the normality of H is suflicient for (l)

to define a composition, and G/H, equipped with this composition, is a group. Theorem 18.7. If H is a normal subgroup of a group (G, ), the composition

on G/H defined by (aH) - (bH) = abH for all aH, bH e G/H is well-defined, and (G/H, °) is a group whose neutral element is H; moreover, the inverse of aH is a‘lH for all a e G.

Proof. Let a, a', b, b’ E G be such that aH = a’H and bH = b’H. By the normality of H, abH = a(bH) = a(b'H) = a(Hb’) = (aH)b’ = (a’H)b’ = a’(Hb') = a'(b'H) = a’b’H.

160

GROUPS AND RINGS

Chap. III

Thus, the composition is well-defined. Multiplication is associative, for if a, b, c e G,

[(aH)(bH)](cH) = (abH)(cH) = (ab)cH, (aH)[(bH)(cH)] = (aHXbcH) = a(190)11As H = eH, where e is the neutral element of G, clearly, H is the neutral

element for multiplication on G/H. Consequently, a—lH is the inverse of aH for each a e G, since (aH)(a—1H) = (1l = H, (a‘lH)(aH) = a‘laH = H. The group G/H is called the quotient group of G defined by H, and its

composition is called the composition induced on G/H by that of G. Example 18.3. In Example 18.1 we saw that H = {r0, r1, r2, r3} is a

normal subgroup of the group G of all symmetries of the square. There is exactly one other coset of H besides H itself, and it may be variously denoted by hH, vH, l, daH (or Hh, etc.), since those cosets must all be identical. If that coset is denoted by hH, for example,

the table for the induced composition on G/H is hH H

H

hH

hH-

hH

H

If the composition of G is denoted by + rather than -, then, of course,

the composition induced on G/H is also denoted by + and is defined by (a+H)'+ (b.+-m= (0+ bl+H

foralla+H,b +HeG/H; also the inverse ofa+His ——a+H. Example 18.4. Let H = {0, 4, 8}. Clearly, H is a normal subgroup of

the group (N12, +12) of integers modulo 12. The cosets of H are

H={0,4,8}=4+H=8+H, 1+H={1,5,9}=5+H=9+H, 2+H={2,6,10}=6+H=10+H, 3+H={3,7,ll}=7+H=11+H.

Sec. 18

GROUPS AND RINGS

161

In constructing the addition table for the composition induced on Nu/H by +12, it makes no difl‘erence, of course, how we choose to

denote the four elements of Nm/H. We might choose to denote them byH, l + H,2 + H, and3 + H,forexample,orby4 + H, 5 + H,

10 + H, and 3 + H. In the former case, the table for the induced composition is

H

1+H

2+H

3+H

H 1+H 2+H

H 1+H 2+H

1+H 2+H 3+H

2+H 3+H H

3+H H 1+H

3+H

3+H .

H

1+H

2+H ’

and in the latter case, the table for the induced composition is 4+H

5+H

10+H

3+H

4+H

4+H

5+H

10+H

3+H

5+H

5+H

10+H

3+H

4+H

10+H

10+H

3+H

4+H

5+H

3+H

3+H

4+H

5+H

10+H

EXERCISES

In these exercises, the composition induced on ME) by a composition A on E is denoted by Asp. 18.1. (a) Let A be a composition on E. If X, Y, and Z are subsets of E, then XA(YUZ)=(XAY)U(XAZ),

XA(YnZ)s(XAY)n(XAZ),

(Y uz)AX = (YAX) u (ZAX), (Y nZ)AX 0. Hence, 1p is indeed a bijection from N onto the set of all ideals of Z.

The following theorem, in conjunction with Theorem 19.8, shows that a quotient ring of Z determined by a nonzero ideal is isomorphic to the ring of integers modulo m for some m > 0 (Examples 2.3, 15.1). Theorem 19.9. Let m be a strictly positive integer. The function (pm from N," into 2](m) defined by

0, and it holds

Sec. 20

GROUPS AND RINGS

177

also if n = 0 by 3° of Theorem 20.5. If m > 0,

f(a‘m) =f((a‘1)'”) =f(a‘1)"‘

= (f(a)‘1)"‘ =f(a -m by 3° of Theorem 20.5 and Theorem 14.7. Of particular importance is the case where f is a homomorphism from a group E into a group F. If the compositions of both groups are denoted additively and if the neutral elements of both are denoted by “0,” then

f(0) = 0,

f(—x) = —f(X), f(n.x) = n.f(x) for all x e E, n e Z; if the compositions of both groups are denoted multipli-

catively and if the neutral elements of both are denoted by “1,” then

f(l) = 1, f(96“) = f(x)‘1, f(36") = f(96)” for all e, neZ. Not all algebraic properties are preserved by epimorphisms. Indeed, the image of a cancellable element under an epimorphism need not be cancellable. For example, 2 is cancellable for multiplication on Z, but 2 + (4) is not cancellable for multiplication on Z4. Theorem 20.6. If f is a homomorphism from (E, A) into (F, V) and if g

is a homomorphism from (F, v) into (G, V), then g of is a homomorphism from (E, A) into (G, V).

The proof is similar to that of 3° of Theorem 6.1. Definition. Let f be a homomorphism from a group (G, -) into a group (G’, -), and let e’ be the neutral element of G’. The kernel off is the subset

kerf of G defined by kerf = f *({e’}). Thus, x e kerf if and only iff(x) = e’.

Theorem 20.7. The kernel K of a homomorphism f from a group (G, -) into a group (G', -) is a normal subgroup of (G, -).

178

GROUPS AND RINGS

Chap. 1]]

Proof: Let e and e’ be the neutral elements of G and G’ respectively. By 3° of Theorem 20.5,f(e) = e’, so e 6K, and thus Kari 0. If x,y 6K and if a e G, then

f00’“) =f(x)f0"1) =f(1‘))"00‘l = e’e"1 = e’, f(axa‘l) = f(a)f(X)f(a‘1) =f(a)e'f (a)‘1 = e’ by 2° of Theorem 20.5, so xy—1 6 K, war1 6 K. Thus, K is a normal subgroup by Theorems 8.3 and 18.5. Theorem 20.8. Let f be a homomorphism from a group (G, -) into a group

(G’, -). If H is a normal subgroup of G that is contained in the kernel K off, then

g: xH:—>f(x)

is a well-defined homomorphism from G/H into G’ satisfying g 0 (pH = f. Proof. If xH = yH, then y'lx e H S K, sof(y‘lx) ; e’, the neutral element of G’, whence

f(x) =f(y(y‘1x)) =f(.v)f0"1x) =f0’)e' =f0')Thus, g is well-defined, and clearly, g o (pH = f. If x, y E G,

g((xH)(yH)) = g(xyH) =f(xy)

= f(x)f(y) = g(xH)g('yH)Hence, g is a homomorphism. Theorem 20.9. (Factor Theorem for Groups) Letf be an epimorphism from a group (G, -) onto a group (G’, -), and let K be the kernel off2 Then g: xK I—>f(x)

is a well-defined isomorphism from G/K onto G’ satisfying g 0 (px = f. Moreover, f is an isomorphism if and only if K = {e}. Proof. By Theorem 20.8, g is a well-defined homomorphism, and clearly

g is surjective, since f is. If f (x) = f (y), then fol—1x) = f (y‘l)f(x) = f (y)‘1f(x) = e’, the neutral element of G’, whence y‘lx e K, and thus xK = yK. Therefore, g is injective and hence is an isomorphism. Iff is an

isomorphism, then {Fir is injective as go «px = f, whence K = {e}. Conversely, if K = {e}, then (pg is clearly an isomorphism, whence f is also.

Sec. 20

GROUPS AND RINGS

179

The Factor Theorem shows that an epimorphism from one group onto another can be “factored” into an isomorphism and the canonical epimorphism from a group onto a quotient group. Example 20.1. The circle group is the group (T, -) where

T= {zeCz |z| = 1}. Let f be the function from R into T defined by f(x) = cosx + isinx.

By a theorem of analysis, fis surjective. Consequently,fis an epimorphism from (R, +) onto (T, -), for

f (30f(y) = (cos x + isin x)(cos y + isin y) = (cos x cosy — sin x sin y) + i(sin x cosy + cos x siny) = cos(x + y) + isin(x + y)

=f(x + y) for all x, y e R. The kernel of f is the subgroup 21rZ of all integral multiples of 211-. Hence, by the Factor Theorem,

gzx + 27a—->cosx + isinx

is an isomorphism from the quotient group (R/21rZ, +) onto (T, -). Example 20.2. Let n be a strictly positive integer, and let 0" be the set of all nonzero complex numbers. The function f defined by

f(2) = 2" is an epimorphism from (C*, -) onto itself. Indeed,

f(W) = (WZ)" = W”Z“ =f(W)f(Z) for all w, z e C*, and f is surjective, for if

w = r(cos at + isin a), then w = f(z) where M

.

-

at

z = M(cos—+ 1s1n—). n

n

GROUPS AND RINGS

180

Chap. III

The kernel off is the set R“ of all complex nth roots of unity, so by the Factor Theorem,

g: ai—a z" is an isomorphism from (C*/R,,, -) onto (0", -). Thus, if n 9e 1, f is an example of a surjective endomorphism that is not an automor-

phism. For a more concrete picture of the quotient group and the associated isomorphism, let us consider the case where n = 3. The set R3 of all complex cube roots of unity has three members, namely, 1, 271'

w=cos——

3 +

. . 211-

ism—, 3

and 417

w2=cos—

3 +

. 471'

ism—. 3

For each 2 e C*, the coset 2R3 is thus the set {2, zco, zw’}, and multiplication on the set C"‘/R3 of all such cosets satisfies {Zn 21a), 21002} ' {22: 22a), 2260”} = {1122: 2122‘”, 2122“”-

The associated isomorphism g from C*/R3 onto 0" takes an equivalence class {2, zw, zw’} into the cube 23 = (zco)3 = (2a)”)3 of any one of its members. Example 20.3. For each complex number 2, we denote the real part of 2 by 9% and the imaginary part of 2 by .fz. Thus, if z = x + iy where x and y are real numbers, then $2 = x. and J2 = y. Clearly, .9? is an endomorphism of (C, +) whose range is R. The kernel of .9? is the subgroup iR of purely imaginary complex numbers, so by the Factor Theorem, g: z + iR-H 922 is an isomorphism from (C/iR, +) onto (R, +). The set at of purely imaginary complex numbers may be described geometrically as the Y-axis of the plane of analytic geometry. For each 2 e C, 2 + iR is simply the set of all the complex numbers having the same real part as 2, i.e., the set of all points of the plane whose abscissa is the real

part of z; consequently, z + iR may be described geometrically as the line through the point 2 parallel to the Y-axis. Thus, C/iR is the

Sec. 20

GROUPS AND RINGS

181

set of all lines parallel to the Y—axis. The sum of two such lines L1 and L2 is the line all of whose points have for abscissa the sum of the

number that is the abscissa of all the points of L1 and the number that is the abscissa of all the points of L2. The associated isomorphism g takes a given line parallel to the Y-axis into the number that is the abscissa of all its points.

Similarly, I“ is an endomorphism of (C, +) whose range is R. The kernel of J is the subgroup R of real numbers, so h:z+R—>.fz is an isomorphism from (C/R, +) onto (R, +). Thus, J is an example

of an endomorphism whose kernel is the same as its range. However, .9? and .f are not endomorphisms of (C, -), since

321-2 = —1 9e 0 = (gel-)2, 1:2 = 0 7s 1 = (J02. Example 20.4. To illustrate the use of Theorem 20.9, we shall prove the following theorem: If H is a normal subgroup of a group G, if K is a normal subgroup of G/H, and if L = ¢p§(K), then L is a normal subgroup of G, and there is an isomorphism ffrom (G/H)/K onto G/L

satisfying f° ‘PK ° p(j). Then the number J, of inversions of p is given by JP: 2 m,,(i,j). (MET

A consideration of the four possible cases shows that

ma,(i,j) E M,('r*(i,j)) + M,(i.j)

(mod 2)

for all (i,j) e T, and hence that Z marani) E 2 ma(7*(isj)) +(i 2Tm1(i9j)'

(1,5)61'

(LIDET

De

(mOd 2)

But since 1" is a permutation of T, 2 m¢(7*(i9j))= 2 ma(i:j)-

(war

(war

Consequently,

JwEJH‘Jn

(mod 2)

whence sgn m- = (—1)Jar = (_1)Ja+.r,

= (—n’rx—n’r = (sgn «0(n 7). In particular,

1 = ssn co“ = (sgn axsgn on, so sgn 6" : sgn a, From Theorem 20.10, we conclude that a product of two even permuta-

tions or of two odd permutations is an even permutation, and that the product of an even permutation and an odd permutation is an odd permutation.

184

GROUPS AND RINGS

Chap. III

Theorem 20.11. A transposition is an odd permutation; more generally, an m-cycle is an odd or even permutation according as m is even or odd.

Proof. An m-cycle is a product of m — l transpositions, since clearly, (a1: - - - 9 am) = (a1, am)(a1: am—l) ' ' - (a1, a2)-

To show that an m-cycle is odd or even according as m is even or odd, it

suflices by Theorem 20.10 and what we have just proved to show that a transposition is an odd permutation. Let -r = (r, s), where 1 S r < s S n. With the notation of Theorem

20.10, mr(r, s) = l, m,(i, r) = m,(i, s) = 0 if i < r, m,(i,s) = l ifrf(x) is a well-defined isomorphism from A/a onto A’ satisfying g o (pa = f. Moreover, f is an isomorphism if and only if a = (0).

The assertion is an immediate consequence of Theorems 21.3 and 20.9.

Theorem 21.5. If f is an epimorphism from a field K onto a ring A, then either A is a zero ring, or A is a field andfis an isomorphism from K onto A.

Sec. 21

GROUPS AND RINGS

191

Proof. The kernel off is an ideal and hence is either K or (0) by the Corollary

of Theorem 19.4. In the former case, f(x) = 0 for all x e K, so A is a zero

ring; in the latter case, f is an isomorphism by Theorem 21.4. We consider next homomorphisms from the ring Z of integers into a ring with identity. We recall that if m > 0, the ring Z/(m) is denoted by 2",.

Theorem 21.6. Let A be a ring with identity 1. The function g from Z into A defined by g(n) = n.1

for all n e Z is a homomorphism from the ring Z into A. The range B of g is the smallest subring of A that contains 1. If A has no proper zero-divisors,

then g is the only nonzero homomorphism from Z into A, and the kernel of g is either (0), in which case g is an isomorphism from Z onto B, or (p) for some prime p, in which case B is isomorphic to the field 2,.

Proof. By (17) of Theorem 14.7, (n + m).1 = n.1+ ml,

and by (1) of §15 and (18) of Theorem 14.7, (n.l)(m.1) = n.(m.l) = (nm).1

for all n, m e Z. Consequently, g is a homomorphism from Z into A, and its range B is, therefore, a subring by the Corollary of Theorem 21.1. An inductive argument establishes that every subring of A containing 1 also contains B, so B is the smallest subring of A that contains 1. Let us assume further that A has no proper zero-divisors. By Theorem 19.8, the kernel of g is (p) for some p e N. By Theorem 21.4, applied to the function obtained

by restricting the codomain of g to B, B is isomorphic to Z, and also has no proper zero-divisors, so either p = 0 or p is a prime by Theorem 19.11.

To show the uniqueness of g, let h be a nonzero homomorphism from Z into A. Since

’10) = ’10“) = 11(1)“, either h(l) = 1 or h(l) = 0 by Theorem 15.2. But

h(n) = h(n.l) = n.h(1)

for all n e Z by the Corollary of Theorem 20.5. Therefore, if h(l) were 0, then h(n) would be n.0 = 0 for all n e Z, and hence, h would be the zero

GROUPS AND RINGS

192

Chap. III

homomorphism, a contradiction. Consequently, h(l) = 1, so

h(n)= n.l = g(n) for all n e Z.

Definition. The characteristic of a ring A with identity 1 is the natural number p such that (p) is the kernel of the homomorphism g: m—z n.l from Z into A.

By Theorem 19.8, ifp E N*, then p is the smallest strictly positive integer belonging to the principal ideal (p) generated by p. Thus a ring A with identity 1 has characteristic p > 0 ifand only ifp is the smallest strictly positive integer such that p.1 = 0; and A has characteristic zero if and only if

n.1 34E 0for all n > 0. For example, for each p e N*, the ring 2, has characteristic p, for clearly p is the smallest of the strictly positive integers m such that m.(l + (p)) = (p), since m.(1 + (p)) = m + (p). By Theorem 21.6, the characteristic of a ring with identity and without

proper zero-divisors is either zero or a prime. In particular, the characteristic of a division ring or integral domain is either zero or a prime. Theorem 21.7. Let A be a ring with identity 1, let p be the characteristic of

A, let a e A, and let g, be the function from Z into A defined by ga(n) = n.a for all n e Z. 1° The function g“ is a homomorphism from the additive group (Z, +) into (A, +), and the kernel of g“ contains (p). Consequently, if

p > 0, then n.a = 0 wheneverp I n. 2° If a is not a zero-divisor of A, then the kernel of g“ is (p), and con-

sequently if p = 0, then n.a = 0 only if n = 0, but if p > 0, then

n.a = 0 ifand only ifp l n. Proof: By Theorem 14.7, g, is a homomorphism. For every n e Z, n.a = (n.a)l = a(n.l) by (l) of §15. Therefore, n.a = 0 whenever n.1 = 0, and if a is not a

zero-divisor, n.a = 0 if and only if n.l = 0, or equivalently, if and only if n e (p). For example, if A is a ring with identity whose characteristic is 2, then

for all a e A, we have 2.a = 0, or equivalently, a = —a.

Sec. 21

GROUPS AND RINGS

193

Definition. A prime field is a field containing no proper subfields. Theorem 21.8. A division ring K contains one and only one prime subfield P, and P is the smallest division subring of K. If the characteristic of K is a prime p, then Z, is isomorphic to P; if the characteristic of K is zero, then

Q is isomorphic to P under an isomorphism g satisfying g(n) = n.1 for all n e 2. Proof. By Theorem 21.6, there is a smallest subring B of K that contains 1. The center C of K is a subfield by the Corollary of Theorem 15.6, so B E C.

Let P be the quotient field of B in C. Any division subring of K contains 1, and hence B, and therefore also P, since each element of P is the product of an element of B and the inverse of a nonzero element of B. Hence, P is

contained in every division subring of K, and in particular, P is the smallest subfield of K. As every subfield of P is a subfield of K, P has no proper subfields and hence is a prime field. Since P is contained in every subfield of K,

P is the only subfield of K that is a prime field. If the characteristic of K is a prime p, then 2,, is isomorphic to B by Theorem 21.6, and hence, P = B as B is a field. If the characteristic of K is zero, then f: n_»—» n.1 is an isomor-

phism from Z onto B by Theorem 21.6, so there is an extension g off that

is an isomorphism from Q onto P by Theorem 17.5. Corollary. To within isomorphism, the only prime field whose characteristic is a prime p is 2,, and the only prime field of characteristic zero is Q.

EXERCISES 21.1. Let (G, +) be a commutative group. We denote by End(G) the set of all endomorphisms of (G,.+). Prove that (End(G), +, o) is a ring, where for all u,veEnd(G), u + v is defined to be the endomorphism XI—_? u(x) + u(x) of G.

*21.2. For each integer p, let L, be the endomorphism of (Z, +) defined by L,(x) = px

for all x e Z, and let L be the function from Z into End(Z) defined by 140’) = Li:

for all p e Z. (a) The function L is an isomorphism from the ring Z of integers onto the ring End(Z) of endomorphisms of the abelian group (Z, +).

GROUPS AND RINGS

194

Chap. 111

(b) What are the automorphisms of (Z, +)? (c) The rings (Z, +, (r)) and (Z, +, (s)) (Exercise 14.7) are isomorphic

if and only if either r = s or r = —s. (d) If (A, +, -) is a ring such that (A, +) is isomorphic to (Z, +), then there is one and only one natural number r such that (A , + , ') is isomorphic to (Z, +, (r)). Thus, to within isomorphism, there are denumerably

many rings whose additive group is isomorphic to the additive group of integers. 21.3. Ifa and b are ideals ofa ring A and ifa S b, then a is an ideal of the ring

b,b/a is an ideal of the ring Ala, and there is an isomorphism g from (A/a)/(b/a) onto A/b satisfying

g° Wit/(1° ‘pa = ‘7’!)[Use Theorem 21.3.]

21.4. Let A be a ring, let a be an ideal of A, and let I) be a subring of A. (a) b + a is a subring of A, and a is an ideal of!) + a.

(b) The function f from 1) onto (13 + a)/a defined by f(x) = x + a for

all x e b is an epimorphism whose kernel is b n a (c) The function g: x + (b n a)n_—__>x + a is an isomorphism from b/(b n a) onto (5 + a)/a. 21.5. Let f be an epimorphism from a ring A onto a ring A’, and let a be the

kernel off. (a) If h is a subring (an ideal) of A, then f* (b) is a subring (an ideal) of

A’, andf*(f*(b)) = b + a. (b) If 5' is a subring (an ideal) of A', then f*(b’) is a subring (an ideal)

of A andf*(f*(b’)) = b'. (c) If I) is an ideal of A and if b' = f(b), then there is one and only one

epimorphism g from All) onto A’lb’ satisfying

g° % = w °f, and moreover, the kernel of g is (b + a)/b.

21.6. Let A be a ring, and for each a e A, let L,I be the endomorphism of (A, +)

defined by La(x) = ax for all x e A. The left regular representation of the ring A is the function L: a I—zL‘I from A into End(A).

(a) The function L is a homomorphism from the ring A into the ring End(A). (b) If A is a ring with identity and if u E End(A), then u = L, for some a e A if and only if u(xy) = u(x)y for all x, y E A.

Sec. 22

GROUPS AND RINGS

195

21.7. (a) If A is a commutative ring containing an element that is not a zerodivisor, then L, is an endomorphism of the ring A if and only if a is a multiplicative idempotent (Exercise 2.17). (b) If A is a ring whose center contains an element that is not a zero-

divisor, then La is an endomorphism of the ring A for all a e A if and only if A is a boolean ring (Exercise 15.21). 21.8. Iffis an epimorphism from a ring onto a division ring, then the kernel of f is a maximal ideal. *21.9. If a is an ideal of a ring A and iff is an epimorphism from the ring a onto a ring with identity B, then there is one and only one epimorphism g from A onto B extending f. [Show first that if e ef*({l}), then f(ex) = f(exe) = f(xe) for all x E A.]

*21.10. Let D be a subdomain of Q. (a) Prove that D 2 Z. (b) Iffis a homomorphism from the group (D, +) into the group (Q, +),

then there exists a efa: (D) such that f(x) = ax for all x e D. (c) The left regular representation (Exercise 21.6) of D is an isomorphism from the ring D onto the ring End(D).

(d) The only nonzero homomorphism from the ring D into the ring Q is the function in: XI—>‘ x. (e) If E is a subdomain of Q isomorphic to D, then E = D. *21.11. If K is a field, then the groups (K, +) and (K*, -) are not isomorphic. [Consider the final example of §6.]

*21.12. If K is a field whose characteristic is not 2 and if f is a function from K

into K satisfying

f(x +y) =f(X) +f(y) for all x,y EKand

f(X)f(x'1) = l for all x EK*, then either f or —f is a monomorphism from K into K. [Assume that f(1) = 1. If x is neither 0 nor —1, show that f(x‘) = f(x)2

by expanding f((1 + x‘1)2) in two ways.]

'

21.13. If A is a totally ordered integral domain whose set of positive elements is well-ordered, then g: m—>n.1 is an isomorphism from the ordered ring of integers onto the ordered ring A. [Use Theorem 14.9.]

22. Principal Ideal Domains

We next wish to extend the definition of divisibility given for the integral domain Z in §l9. Although we shall formulate divisibility concepts in general

196

GROUPS AND RINGS

Chap. 111

for commutative rings with identity, we shall investigate them only for integral domains. Definition. If a and b are elements of a commutative ring with identity A and if b ;E 0, we shall say that b divides a in A, or that b is a divisor or a factor of a in A, or that a is divisible by b in A or a multiple of b in A if there

exists c e A such that be = a.

The symbol I means “divides,” so that bIa means that b divides a.

If A is a subring of a commutative ring with identity K containing the identity element of K and if a and b are nonzero elements of A, it is entirely possible that b divides a in K but not in A. For example, 2 divides 3 in Q but not in Z. In discussing divisibility, therefore, we must always keep in mind what ring is being considered; if it is clear from the context what that ring is,

we shall usually omit explicit reference to it. We recall that if A is a commutative ring with identity and if b e A, the

ideal Ab of A is denoted by (b). Consequently, if b e A*,

(l)

b I a if and only if (a) E (b).

Definition. Let a and b be nonzero elements of a commutative ring with

identity A. We shall say that a is an associate of b in A if b I a and a I b in A. A unit of A is an associate of 1 in A. Theorem 22.1. Let a and b be nonzero elements of an integral domain A. The following statements are equivalent: 1° a is an associate of b.

2° (a) = (b). 3° There is an invertible element u of A such that a = bu. Proof. The equivalence of 1° and 2° follows from (1). Condition 3° implies

1°, for if a = bu where u is invertible, then also b = au‘l, so 17' I a and a I b. Condition 1° implies 3°, for if a = bu and if b = av, then a = avu, so 1 = W and hence u is invertible.

Corollary 22.1.1. Let u be a nonzero element of an integral domain A. The following statements are equivalent:

Sec. 22

GROUPS AND RINGS

197

1° u is a unit of A. 2° (u) = A. 3° u is an invertible element of A.

Corollary 22.1.2. Let A be an integral domain, and let G be the set of is an associate of . . .” on invertible elements of A. Then the relation “ A* is an equivalence relation, and for every a e A"I the equivalence class of a

for this relation is the set Ga. Definition. Let a1, . . . , an be nonzero elements of a commutative ring with identity A. A common divisor of a1, . . . , an is any element of A* dividing each of al, . . . , a”. An element d of A* is a greatest common divisor of a1, . . . , a” if d is a common divisor of a1, . . . , an and if every common divisor of a1, . . . , an is also a divisor of (1.

Since (ab) 9 (d) for all k e [1, n] if and only if (a1) + . . . + (an) E (d) by Theorem 19.3, from (1) we conclude that (2)

d is a common divisor of a1, . . . , a” if and only if

(a1) + . . . + (an) E (d). Definition. Let a1, . . . , a” be nonzero elements of a commutative ring with identity A. A common multiple of al, . . . , an is any element of A* that is a multiple of each of a1, . . . , an. An element m of A* is a least common multiple of a1, . . . , a” if m is a common multiple of a1, . . . , a7. and if every

common multiple of a1, . . . , an is also a multiple of m. Since (m) S (ak) for all k e [1, n] if and only if (m) E (a1) n . . . fl (au , from (1) we also conclude that (3)

m is a common multiple of a1, . . . , an if and only if

(m) 9 (a1) n . . . n (a,,). The following theorem follows easily from (1) and Theorem 22.1.

Theorem 22.2. Let a1, . . . , an be nonzero elements of an integral domain

A. If d is a greatest common divisor of a1, . .- . , an, then the set of all greatest common divisors of a1, . . . , an is the set of all associates of d_. If m is a least common multiple of a1, . . . , a”, then the set of all least common multiples of a1, . . . , an is the set of all associates of m.

198

GROUPS AND RINGS

Chap. III

A paraphrase of Theorem 22.2 is that in an integral domain, a greatest common divisor or a least common multiple of a1, . . . , an is “unique to within an associate.” The use of the superlative adjectives “greatest” and

“least” in the preceding definition thus does not imply that there is at most one greatest common divisor or at most one least common multiple, even though in ordinary English a singular noun modified by a superlative

adjective names at most one member of the class considered. Definition. If a1, . . . , a” are nonzero elements of a commutative ring with

identity, we shall say that a1, . . . , an are relatively prime if 1 is a greatest common divisor of a1, . . . , an.

Theorem 22.3. If m is a least common multiple of nonzero elements a and b of an integral domain A, then there is a greatest common divisor d of a and b satisfying md = ab. Proof. Since ab is a common multiple of a and b, m I ab; it remains for us to show that d = ab/m is a greatest common divisor of a and b. Since

b I m, ab I am, and consequently, d I a; since a I m, ab I mb, and consequently d I b; hence, d is a common divisor of a and b. Let c be a common divisor ofa and b, and let x,y e A* be such that ex = a, cy = b. Then a I cxy and

b I cxy, so m I cxy and

C(fl) = icy = 93' = d, m

m

m

whence c I d. Thus, dis a greatest common divisor of a and b. Corollary. If m is a least common multiple of nonzero elements a and b of an integral domain A, then d’ is a greatest common divisor of a and b if and only if md’ is an associate of ab. Proof. Clearly md’ is an associate of ab if and only if d’ is an associate of ab/m = :1, so the assertion follows from Theorems 22.3 and 22.2.

For a deeper investigation of divisibility, we shall limit our discussion to a special class of integral domains. Definition. A principal ideal domain is an integral domain A satisfying the following two conditions: (PID)

Every ideal of A is a principal ideal.

(N)

Every nonempty set of ideals of A, ordered by E , possesses a maximal element.

Sec. 22

GROUPS AND RINGS

199

Theorem 22.4. The ring Z of integers is a principal ideal domain.

Proof. By Theorem 19.8, Z satisfies (PID). To show that Z satisfies (N), let QI be a nonempty set of ideals of Z. If QI contains only the zero ideal, then that ideal is a maximal element of Qt. In the contrary case, let n be the

smallest of those strictly positive integers m for which (m) 691. If s is a strictly positive integer such that (s) e ‘2! and (s) 2 (n), then s I n, whence s s n and, therefore, .9 = n by the definition of n. Consequently, (n) is a maximal element of ‘21 by Theorem 19.8. Theorem 22.5. Let a1, . . . , an be nonzero elements of a principal ideal domain A. 1° There exists a greatest common divisor of a1, . . . , an. 2° If d is a greatest common divisor of a1, . . . , an, then there exist

x1, . . . , xneA such that d=a1x1+...+a,,x,,.

3° An element d of A is a greatest common divisor of a1, . . . , a" if

and only if (d) = (a1) + . . . + (an). Proof. Since (a1) + . . . + (an) is an ideal of A by Theorem 19.3, there exists d1 6 A such that (d1) = (a1) + . . . + (an). Consequently, d1 is a common divisor of a1, . . . , an by (2). If s is a common divisor of a1, . . . , an, then ((1,) = (a1) + . . . + (an) E (s) by (2), so s is also a divisor of d1 by (1). Thus, all is a greatest common divisor of a1, . . . , an. Let d be any greatest common divisor of a1, . . . , an. By Theorem 22.2, d is an associate of (11.

By Theorem 22.1, therefore, (d) = (d1) = (a1) + . . . + (an), so there exist x1, . . . ,xneA suchthat d=a1x1+...+a.,,xn.

Conversely, if (d) = (a1) + . . . + (an), then (d) = (d1), so d is an associate of d1 and hence is a greatest common divisor of a1, . . . , an by Theorem 22.2.

Since there exist x1, . . . , xn e A such that alxl + . . . + anx” = 1 if and only if (1) = (a1) + . . . + (an), we obtain from 2° and 3° the following corollary. Corollary; (Bezout's Identity) If a1, . . . , a." are nonzero elements of a principal ideal domain A, then a1, . . . , an are relatively prime if and only if there exist x1, . . . , x" e A such that a1x1+...+a,,x,,=l.

200

GROUPS AND RINGS

Chap. III

Theorem 22.6. Let a1, . . . , a” be nonzero elements of a principal ideal domain A.

l° There exists a least common multiple of a1, . . . , an. 2° An element m of A is a least common multiple of a1, . . . , a, if and only if (m) = (a1) 0 . . . n (an). Proof. As (a1) 0 . . . m (an) is an ideal, there exists m1 6 A such that (m) = (a1) n . . . n (an). By (3), m1 is a common multiple of a1, . . . , (1,. Ifs is a common multiple of a1, . . . , an, then (s) S (a1) n . . . n (an) = (ml) by (3), so s is also a multiple of m1 by (1). Therefore, m1 is a least common multiple of a1, . . . , an.

By Theorems 22.1 and 22.2, m is a least common multiple of a1, . . . , a, if and only if (m) = (m) = (a1) n . . . n (an). If A is a commutative ring with identity, a nonzero element of A is divisible by all of its associates and also by every unit of A.

Definition. A nonzero element p of a commutative ring with identity A is an irreducible element of A if p is not a unit of A and if the only divisors of p in A are its associates and the units of A. For example, 2 is an irreducible element of Z, but 2 is not an irreducible element of Q. Indeed, a field contains no irreducible elements, since every

nonzero element of a field is a unit. An associate of an irreducible element is irreducible, for associates have

exactly the same divisors. Theorem 22.7. Let p be a nonzero element of a principal ideal domain A. The following statements are equivalent:

1° p is irreducible. 2° (p) is a maximal ideal. 3° 12 is not a unit, and for all nonzero elements a and b of A, if p I ab,

then eitherp | a orp | b. Proof. Condition l° implies 2°: As p is not a unit, (p) is a proper ideal of A. If (a) 2 (p), then a | 17, so a is either a unit of A or an associate of p, and consequently, (a) is either A or (p). Therefore, as every ideal of A is

principal, (p) is a maximal ideal.

Condition 2° implies 3°: Since (p) is a proper ideal of A by 2°, p is not a unit of A. Suppose that p | ab but that p ,|’ a. Then (a) is not contained in (p), so (p) + (a) is an ideal properly containing (p), and consequently, (p) + (a) = A by 2°. Therefore, there exist s, t E A such that .917 + ta = 1.

GROUPS AND RINGS

Sec. 22

201

Let c e A be such that pc = ab. Then b = spb + tab = spb + tpc = p(sb + tc),

so p | b. Condition 3° implies 1°: Let a be a divisor ofp, and let b be such that

ab = p. By 3°, either p | a or p I b. If p I a, then a and p are associates. If p I b and if c is such thatpc = b, then abc =pc = b,

so ac = 1, and therefore, a is a unit.

Definition. A subset P of an integral domain A is a representative set of irreducible elements of A if each element of P is irreducible and if each irreducible element of A is an associate of one and only one member of P. The word “prime” often occurs in discussions of divisibility. Sometimes “prime” is used as a synonym for “irreducible.” More often, however, there is a “natural” representative set of irreducible elements, and its members are

called primes to distinguish them from irreducible elements not belonging to the representative set. For example, the set P of all positive irreducible integers is a representative set of irreducible elements of Z, for the only units of Z are 1 and —1; elements of P are called prime numbers in accordance with definition of §l9.

The “Fundamental Theorem of Arithmetic,” which we shall prove shortly, is the assertion that every nonzero integer other than 1 and —l is either the product of a sequence of prime numbers in an essentially unique way or the additive inverse of such a product. A natural question is: What other integral domains satisfy a similar statement? This question may be separated into two for an arbitrary integral domain A: 1° Is every element of A* either a unit or a product of a sequence of irreducible elements? 2° Is the factorization of an element of A* into a product of irreducible elements unique ?

In answering the second question, we need a reasonable interpretation of the word “unique”; for example, 15 = 3-5 = 5-3 = (~3)(—5) = (—5)('—3), so (3, 5), (5, 3), (—3, —5), and (—5, —3) are all sequences of irreducible

integers whose product is 15. However, these four sequences are very similar, for they all have two terms, and if (qbqa) is any one of them, there is a

202

GROUPS AND RINGS

Chap. III

permutation (p of {1, 2} such that 9m) is an associate of 3 and gm, is an associate of 5. We are led, therefore, to the following definition:

Definition. An integral domain A is a unique factorization domain (or gaussian domain) if the following two conditions hold: (UFD 1)

Every nonzero element of A is either a unit or a product of

irreducible elements. (UFD 2)

If (p,)15,5" and (q,)1S 15m are any sequences of irreducible elements of A such that

I: Pi = II (1;,

(4)

then n = m and there is a permutation «p of [1, n] such that p,- and 9W) are associates for all i e [1, n].

Theorem 22.8. If A is an integral domain satisfying (UFD 1), then A satisfies (UFD 2) if and only if A satisfies the following condition: (UFD 3)

Ifp is an irreducible element of A, then for all elements a, b of

A*, ifp I ab, then eitherp | a orp | b. Proof. We shall first show that (UFD l) and (UFD 2) imply (UFD 3). Let c EA" be such that pc = ab. If a is a unit, then pca‘1 = b, so 17 | b, and similarly, if b is a unit, then 17 | a. Therefore, by (UFD l) we may assume

that there exist sequences (“7915mm and (bk 15kg“ of irreducible elements whose products are respectively a and b. Consequently, a1, . . . , am, b1, . . . , bn is a sequence of irreducible elements whose product is pc. If c is a unit, then pc is irreducible, so by (UFD 2) pc, and hence also 12, are associates of one of a1, . . . , am, b1, . . . , b”, and consequently, p divides either a or b;

in the contrary case, there is a sequence (0791515, of irreducible elements whose product is c by (UFD 1), so p, c1, . . ., c, is a sequence of irreducible elements whose product is pc, and consequently, by (UFD 2) p is an associate of one of a1, . . . , am, b1, . . . , b” and hence divides either a or b.

Next we shall show that (UFD l) and (UFD 3) imply (UFD 2). First,

we note that by an easy inductive argument, we obtain from (UFD 3) the following assertion: (UFD 3’)

If p is an irreducible element of A, then for every sequence

(09135". of elements of A*, if p I a1 . . . am, then there exists i e [1, m] such that p | a,.

Sec. 22

GROUPS AND RINGS

203

Let (palsy, and (”155", be sequences of irreducible elements such that (4) holds, and let S be the set of all k e [1, n] for which there is an

injection p from [1, k] into [1, m] such that p, and 4pm are associates for

each i e [1, k]. Since p1 divides H q,, by (UFD 3') there exists r e [1, m]

such that p1 divides and hence is an associate of q,. Consequently, I e S. Suppose that s e S and that s < n. Then there exist an injection 6 from [1, s] into [1, m] and, for each is [1, s], a unit u,- such that p, = uflcm. 71

Consequently, u = H u,- is a unit, and i=1

1:11: = “(1: am).

(5)

If a were surjective, then from (4) and (5) we would obtain u (fl 1") (51,111") = u (1: ‘11) = u (E 4am) = E pi, i=1

whence

“(Mn i-H-l

and therefore, p,+l would be a unit, a contradiction. Hence, the complement J of the range of 0' in [1, m] is not empty. From (4) and (5) we obtain

11(1: pi) (ifllps) = u (ill (1;) = u (1:! (law) (l3 (1;) = (E pi) (E 4;), whence

“(113) = g 4,.

(6)

Therefore, by (6), p,” divides H q,, so by (UFD 3’) there exists t eJ such 16.!

thatp,+1 divides and hence is an associate ofq,. The function 1- from [1, s + 1] into [1, m] defined by . _ a(i) for all ie [1,3],

*0)— {tifi=s+l

204

GROUPS AND RINGS

Chap. III

is then an injection, and p, is an associate of gm.) for each is [1,s + 1]. Therefore, .9 + l e S. By induction, 1: e S, so there is an injection (p from [1, n] into [1, m] such that Pi and 94m) are associates for each i e [1, n]. Let 7|

vi be the unit such that 94:“) = mp, for each i e [1, n], and let 0 = H v,. If i=1

the complement L in [1, m] of the range of (p were not empty, we would have

(:11 than) = 00: p.) = 110:1 4;) = 0(1: (low) (1.13;; qr), whence upon cancelling fiqw, we would obtain i=1

1: vale—£49; consequently, q,‘ would be a unit for each k E L, a contradiction. Thus, L = 0, so (p is surjective and hence is a permutation of [1, n].

We may now identify an important class of unique factorization domains:

Theorem 22.9. Every principal ideal domain is a unique factorization domain. Proof. By Theorems 22.7 and 22.8, we need only show that if A is a principal ideal domain, then A satisfies (UFD 1). For this, we shall first show that

a nonzero element a of A that is not a unit is divisible by an irreducible element. Let QI be the set of all proper ideals of A containing (a). As a is not a unit, ((1)691, and therefore, ‘2): is not empty. By (N) there exists 1) e A such that (p) is a maximal element of QI, ordered by E . But then (17) is a maximal ideal of A, for a proper ideal of A containing (p) belongs to ‘II and hence is (p). Consequently, p is irreducible by Theorem 22.7, and

p | a by (1). Finally, we shall show that if a nonzero element b is not a unit, then b

is a product of irreducible elements. Let Y be the set of all y e A* such that sy = b for some products of irreducible elements, and let

23 = {(y)=yE Y}Since b is divisible by an irreducible element, 58 is not empty and hence possesses a maximal element (u). Let p1, . . . , p" be irreducible elements such that p1. . . p” = b. If u were not a unit, there would exist an irreducible

element q such that q I u by what we have just proved. If qv = u, then 121 . . . pnqv = b, so (u) e 23 and (u) C (v) as v divides but is not an associate of

Sec. 22

GROUPS AND RINGS

205

u, a contradiction of the maximality of (u). Hence, u is a unit, so b is the

product of the irreducible elements plu, pa, . . . , p”.

The following lemmas lead to the theorem that any finite sequence of nonzero elements of a unique factorization domain has a greatest common divisor and a least common multiple. Lemma 22.]. Let A be a unique factorization domain, and let a e A*. If

(P1015151. is a sequence of irreducible elements of A no two members of which are associates such that every irreducible divisor of a is an associate of some p7,, then there exist a unit u and a sequence (5)195” of natural numbers such that

a = u U Pick-

(7)

Proof. If a is a unit, we may let u = a and rk = 0 for all k e [1, n]. Therefore, we may assume that a is not a unit. By (UFD 1), there is a sequence

@9155," of irreducible elements such that a = q1. . .qm. By hypothesis, for each j e [1, m], there is a unique k e [1, n] such that q, = ujpk for some

unit u,. Consequently, (7) holds where u = u1 . . .um and where rk is the number of integers j e [1, m] such that q, is an associate ofpk. Lemma 2.2. Let a and b be nonzero elements of a unique factorization domain A. If n

H pr?“ a = u k=l

(7)

where u is a unit, (129195,, is a sequence of irreducible elements no two of which are associates, and (5)195” is a sequence of natural numbers, then b | a if and only if there exist a unit a and a sequence 80195,, of natural numbers such that

b = v fipza Ic=1

(8)

3,, gr, forallke [1,n]. Proof. The condition is clearly sufficient. Necessity: Every irreducible divisor of a is an associate of some pk by (UFD 3’). Therefore, as every

irreducible divisor of b is also an irreducible divisor of a, by Lemma 22.1 there exist a unit v and a sequence (553195,, of natural numbers satisfying (8). If s, > r, for some j e [1, n], then n

p? H pit, k=1

206

GROUPS AND RINGS

Chap. 111

so s—r I pi!

(ii: p?) (.1}.1122*),

whence by (UFD 3’) p, would divide and hence be an associate of one of 121, . . . , p,_1,p,+1, . . ., , p", in contradiction to our hypothesis. Therefore, 3,, Srkforall k6 [1,n].

Lemma 2.3. If A is a unique factorization domain and if fl

‘0

a=u£I1p£h

b=vl'Ip;? k=1

where (129193,, is a sequence of irreducible elements, no two of which are associates, where u and v are units, and where (5)195" and (”9193” are sequences of natural numbers, then the elements m and d defined by n

n

m = H pfilm'k’“) ,

.

d = E pillnifkflk}

are respectively a least common multiple and a greatest common divisor of a and b. The assertion follows easily from Lemma 22.2.

Theorem 22.10. If a1, . . . , an is a sequence of nonzeroelements of a unique factorization domain A, then there exist a sequence (pk)15k$n of irreducible

elements, no two of which are associates, and for each j e [1, h] a unit u, and a sequence (59193,, of natural numbers such that “1: u, H Pig,

(9)

k=1

for each j e [1, h], and the elements m and d defined by n

n

minir It ,...,rMp} _ d_kH1pk

Inexfr 1k ....,rin l ’ _ m_k1'];pk

are respectively a least common multiple and a greatest common divisor of

a1, . . . , ah. Proof: By (UFD l), h

t

H“: = 11% €=

i=1

Sec. 22

GROUPS AND RINGS

207

where ql, . . . , q, are irreducible elements. Let p1, . . . , p” be chosen from

ql, . . . , q, so thatp, andp, are not associates if i 75 j, but eachq1t is an associate of some pk. By Lemma 22.1, there exist for each j E [1, h] a unit u, and a sequence (019195” of natural numbers such that (9) holds. An inductive

argument based on Lemma 22.3 then establishes the final assertion.

Theorem 22.11. Let a and b be relatively prime elements of a unique factorization domain A, and let c e A. 1° Ifa|bc,thena|c.

2° Ifa|candb|c,thenab|c. Proof. By Theorem 22.10, a and b have a least common multiple, and hence ab is a least common multiple by Theorem 2.3. If a I be, then be is a common

multiple of a and b, so ab | bc, whence a | c. If a | c and b I c, then c is a

common multiple of a and b, so ab | c. The following notational convention is frequently convenient: If A is an associative, commutative composition on E for which there is a neutral element e and if (3:1),“ is a family of elements of E indexed by a (possibly infinite) set A, where x, = e for all but finitely many indices ac, then “AA x, is defined as being AB x, where B = {at 6A: x, sé e} ifB 75 0, and AA x, is

defined as being e if B is empty.

If P is a representative system of irreducible elements of a unique factorization domain A, then for each a e A* there exist a unique unit 14 and

a unique family (aeP of natural numbers such that n, = 0 for all but finitely many p e P and a=q"’. 96?

Indeed, the existence is an immediate consequence of Lemma 2.1, and the uniqueness of u and (n9)fiP follows from (UFD 2). From Theorems 19.8 and 22.9, therefore, we obtain the following theorem, since 1 and -—1 are

the only units of Z: Theorem 22.12. (Fundamental Theorem of Arithmetic) Let P be the set of prime numbers. For each a e N* there exists a unique family (momD of natural numbers such that n, = 0 for all but finitely many p e P and

a = ”15112“.

GROUPS AND RINGS

208

Chap. III

EXERCISES 22.1. (a) If a and b are nonzero elements of a commutative ring with identity A, then a and b are associates if and only if (a) = (b). (b) Show the correctness of the statement obtained from Corollary 22.1.1

by replacing “an integral domain” with “a commutative ring with identity.” 22.2. If A is a principal ideal domain and if (m) is a proper ideal of A, then every ideal of A/(m) is principal. [Use Exercise 21.5 (b).]

22.3. Let (m) be a proper nonzero ideal of a principal ideal domain A, and for each x e A, let 1?: denote the coset x + (m). (a) If aeA* and if d is a greatest common divisor of a and m, then

(I?) = (d)(b) If c and d are divisors of m such that (E) = (J), then c and d are associates. (c) Let D be a set of divisors of m such that each divisor of m is an associate

of one and only one element of D. Then d‘i—) (d) is a bijection from D onto the set of all ideals of A/(m). [Use Exercise 22.2.] (d) If a, b 9% (m), then a and 5 are associates in A](m) if and only if there is a unit i of .A/(m) such that (T1: = B. 22.4. Let A be a ring. For each a e A , let (a) be the composition on A defined by

x(a)y = xay for all x, y E A. (a) For each a GA, (A, +, (a)) is a ring.

(b) If A is a commutative ring with identity and if a and b are associates, then (A, +, (a)) and (A, +, (b)) are isomorphic.

22.5. Let m > 1, and for each x e Z, let :2 denote the coset x + (m). (a) The left regular representation of the ring Z,” (Exercise 21.6) is an isomorphism from Z", onto End(Z,,,).

(b) If 0 is a composition on Zm distributive over addition modulo m,

then relative to ordinary multiplication modulo m on 2,”, o is (1‘) (Exercise 22.4) for some r e Z. (c) For all f, 562,", the rings (2",, +, (f)) and (Z,,,, +, (5)) are iso-

morphic if and only if f and s are associates in the ring Z,,,. (d) Let D", be the set of all positive divisors of m. If A is a ring whose

additive group is isomorphic to the additive group of integers modulo m, then there is one and only one r e D", such that A is isomorphic to (Z,,,, +, (1‘)). [Use Exercise 22.3.]

22.6. If p is a nonzero ideal of a principal ideal domain A, then p is a prime ideal of A (Exercise 19.6) if and only if p is amaximal ideal. [Use Theorem 2.7.]

Sec. 22

GROUPS AND RINGS

209

22.7. Let a and b be nonzero elements of a commutative ring with identity A. (a) The element a is irreducible if and only if (a) is a maximal element of

the set of all proper principal ideals of A, ordered by S, (b) There is a greatest common divisor of a and b if and only if the set of all principal ideals of A containing (a) + (b), ordered by S, possesses a smallest member (i.e., a member contained in every other member).

(c) There is a least common multiple of a and b if and only if (a) n (b)

is a principal ideal. 22.8. Let A be an integral domain, and let a1, . . . , a” e A*. (a) Ifb divides each of a1, . . . , a”, then b divides a1 + . . . + an.

(b) If d is a greatest common divisor of a1, . . . , a" and if do, = a, for each ie [1, n], then c1, . . . , c,, are relatively prime.

(0) Let K be a quotient field of A. If every pair of elements of A* admits a greatest common divisor, then for every x e K* there exist relatively prime elements a and b of A* such that x = a/b. 22.9. Let a and b be positive integers. (a) If a = bq + r where 0 s r < b, then a greatest common divisor of

a and b is also a greatest common divisor of b and r. (b) If (rk)o SkS "+1 and @9195" are sequences of positive integers such that r0 = a,

r1 = b,

r" ¢ 0,

rn+1 = 0,

rkl = '73]: + VH1, 0 S rk+1 < rk for all k e [1 , n], then r” is a greatest common divisor of a and b. (c) Use (b) and Theorem 22.3 to calculate the positive greatest common

divisor and least common multiple of the following pairs of integers: 143, 117; 1241, 2336; 4224, 10692. I'22.10. Let A be an integral domain such that every pair of nonzero elements of A

admits a greatest common divisor. We shall denote a greatest common divisor of a and b by (a, b), and we shall write a ~ b if a and b are associates.

For all a, b, c e A*, the following three conditions hold:

1° (a, (b, 6)) ~ ((a, b), 6)2° (ca, cb) ~ c(a, b). 3° If (a, b) ~ 1 and (a, c) ~ 1, then (a, be) ~ 1. [For 3°, use 2° to make a substitution for c in (a, c) ~ 1, and apply 1°.] Infer that A satisfies (UFD 3).

210

GROUPS AND RINGS

Chap. III

22.11. If A is an integral domain satisfying (UFD 3), then for every peA,

(p) is a prime ideal of A (Exercise 19.6) if and only if p is either zero or irreducible. 22.12. If A is a unique factorization domain, for each a e A* the length 1(a) of a is defined as follows: if a is a unit, then 1(a) = 0; if a =p1 . . .p,1 where

(179195,, is a sequence of (not necessarily distinct) irreducible elements, then 1(a) = n. (By (UFD 2), 1(a) is well-defined.)

(a) If A is the length function of a principal ideal domain A, then for all a, b e A* the following three conditions hold:

1° Ifb I a, then Mb) 5 1(a).

2° If bl a and if Mb) = 1(a), then a | b. 31° If b,{’a and a*b, then there exist p,qEA* such that pa + qb #0 and 1(pa + qb) < min{l(a), l.(b)}. (b) If A is an integral domain and if A is a function from A“ into N satis-

fying 1°, 2°, and 3° of (a) for all a, b EA*, then A is a principal ideal domain. [Generalize the proofs of Theorems 22.4 and 19.8.]

22.13. Let A be an integral domain. A euclidean stathm on A is a function A from

A“ into N satisfying the following two conditions for all a, b e A": 1° Ifb I a, then Mb) S 1(a).

2° There existq, r E A such that a = bq + r and eitherr = Oor 1(r) < 1(b). ,The integral domain A is a euclidean domain if there exists a euclidean stathm on A. (a) If A is a euclidean stathm on A, then 1 satisfies conditions 1°, 2° and

3° of Exercise 22.12(a), and hence, A is a principal ideal domain. (b) If A is a euclidean stathm on A, then a nonzero element x of A is a unit if and only if 1(x) = 1(1). (c) The function A: ni:)_ Inl, n E Z*, is a euclidean stathm on Z. (d) If i. is a euclidean stathm on A and iff is a strictly increasing function

from N into N, then f o 1 is also a euclidean stathm on A. 22.14. let A be an integral domain such that every nonempty set of principal

ideals of A, ordered by E , possesses a maximal element, and let a be a nonzero, noninvertible element of A. (3.) There exists an irreducible element of A dividing a. [Modify the first

paragraph of the proof of Theorem 22.9.] (b) The element a is a product of irreducible elements. [Modify the second

paragraph of the proof of Theorem 22.9.] 22.15. If A is an integral domain and if N is a function from A* into N* satisfying 1° N(ab) = N(a)N(b),

2° N(a) = 1 if and only if a is a unit for all a, b EA“, then A satisfies (UFD 1). [Use Exercise 22.14.]

Sec. 22

GROUPS AND RINGS

211

‘22.16. Let m be either the integer 1 or a positive integer that is not a square, and

let lbw/Z] = {a + bit/r7: a, b e 2}. Let Nbe the function defined by N(a + bit/Z) = a2 + mb’ for all nonzero elements a + bi1/r; of Z[iV71]. (a) The function N satisfies 1° and 2° of Exercise 22.15, and hence,

lbw/E] satisfies (UFD 1). (b) If m = 1, then N is a euclidean stathm. (Elements of Z[i] are called gaussian integers.) [If 2 and w are nonzero elements of Z[i], let z/w =

x + iy, and consider b = u + iv where u and v are integers satisfying Iu-x|S%.Iv—ylS%-.] _ _ (c) The numbers 3, 2 + i\/5, 2 — 1V5 are irreducible elements of

zm/S]. [If 7. is a divisor, then N(z) divides 9.] Infer that lbw/3] satisfies (UFD 1) but not (UFD 2). I'22.]7. If p is an ideal of a unique factorization domain A, then 11 = (p) for some irreducible element 1) of A if and only if p is a minimal member of the set of all nonzero prime ideals of A (Exercise 19.6), ordered by 9.. [Let p

be an element of p of smallest possible length, and use Exercise 22.11.] 22.18. Let A be a unique factorization domain.

(a) If a is a nonzero element of A of length n, there are at most 2” principal ideals of A containing a. (b) If a is a nonzero ideal of A, there are only a finite number of principal ideals containing a.

(c) Every nonempty set of principal ideals of A, ordered by s, contains a maximal element. *22.l9. An integral domain A is a principal ideal domain if and only if the following

three conditions hold:

1° A is a unique factorization domain. 2° Every nonzero prime ideal of A (Exercise 19.6) is a maximal ideal.

3° Every proper ideal of A is contained in a maximal ideal. [To prove (PID), first show that every prime ideal of A is principal by use of 2° and Exercise 22.17. If a is an arbitrary nonzero ideal of A, use Exercise 22.1_8(b) to show that there is a minimal member (c) of the set

of principal ideals of A containing a, ordered by S; if b is the set of all xeA such that xcea, then a = be; if b aé A, apply 3° to show that b S (d) where dis not a unit.] 22.20. (a) If P is a finite representative set of irreducible elements of a unique

factorization domain A, then (H p) + l is a unit. neP

(b) There are infinitely many prime numbers in Z. I"22.21. A noetheriantring is a commutative ring with identity satisfying condition

(N) (page 198).

GROUPS AND RINGS

212

Chap. III

(a) Every ideal of a noetherian ring is a sum of principal ideals. (b) If A is an integral domain, then A is a principal ideal domain if and only if A is noetherian and the sum a + b of any two principal ideals in and b of A is a principal ideal. (c) If A is an integral domain, then A is a principal ideal domain if and only if A is noetherian and for all a, b e A* there exist 3, t e A such that

sa + tb is a greatest common divisor of a and b. *22.22. Let A be a principal ideal domain, let P be a representative set of irreducible elements of A, and let K be a quotient field of A. For each subset S of P,

let As' be the set of all m/n e K such that m e A and n is either 1 or a product of elements of S.

(a) As is a subdomain of K containing A. (b) If m and n are relatively prime elements of A such that m/n e As: then

l/n e As(c) If a is an ideal of As and if a n A is the principal ideal Ab of A, then

a is the principal ideal A5b of As. (d) The integral domain A3 is a principal ideal domain, and the complement P — S of S in P is a representative set of irreducible elements of As. (e) If D is a subdomain of K containing A, then D = A8 for some subset S of P. (f) The function S H As is a bijection from $(P) onto the set of all

subdomains of K containing A. In particular, if P has n elements, there are 2" subdomains of K containing A. (g) There is a bijection from the set of all subdomains of Q onto EB(N), and no two subdomains of Q are isomorphic (Exercise 21.10).

CHAPTER IV

FINITE GROUPS

23. Cyclic Groups Let A be a subset of a group (G, ~), and let H be the intersection of all the subgroups of G that contain A. Since the neutral element e of G belongs to each subgroup of G, e e H. If x, y e H, then xy—1 belongs to each subgroup of G containing A by Theorem 8.3, whence xy‘1 e H. Thus, H is a subgroup of G by Theorem 8.3. By its very definition, H is the smallest subgroup of

G that contains A. Hence, we have proved the following theorem: Theorem 23.1. If A is a subset of a group G, then of all the subgroups of G that contain A there is a smallest one, namely, the intersection of all the subgroups of G that contain A. Definition. Let A be a subset of a group G. The smallest subgroup H of G that contains A is called the subgroup generated by A, and A is called a set of generators or a generating set for H. If A is a finite set {an . . . , an}, the

subgroup H generated by A is also said to be generated by a1, . . . , a”, and the elements a1, . . . , an are called generators of H.

The central problem of group theory is to describe all groups in some concrete fashion. The general problem is best attacked, however, by con-

sidering only a part of it at a time, that is, by seeking to describe in some way all groups belonging to some prescribed class. One might seek to describe, for example, all finite abelian groups, or all finite groups having a certain prescribed number of elements. In this section we shall see how the properties

of the ring Z lead to a complete description of one very important class of groups—those generated by a single element. 213

CHAPTER IV

FINITE GROUPS

23. Cyclic Groups Let A be a subset of a group (G, -), and let H be the intersection of all the subgroups of G that contain A. Since the neutral element e of G belongs to each subgroup of G, e e H. If x, y e H, then xy‘1 belongs to each subgroup

of G containing A by Theorem 8.3, whence xy‘1 e H. Thus, H is a subgroup of G by Theorem 8.3. By its very definition, H is the smallest subgroup of

G that contains A. Hence, we have proved the following theorem: Theorem 23.1. If A is a subset of a group G, then of all the subgroups of G that contain A there is a smallest one, namely, the intersection of all the

subgroups of G that contain A.

Definition. Let A be a subset of a group G. The smallest subgroup H of G that contains A is called the subgroup generated by A, and A is called a set of generators or a generating set for H. If A is a finite set {an . . . , an}, the

subgroup H generated by A is also said to be generated by a1, . . . , an, and the elements a1, . . . , an are called generators of H.

The central problem of group theory is to describe all groups in some

concrete fashion. The general problem is best attacked, however, by considering only a part of it at a time, that is, by seeking to describe in some

way all groups belonging to some prescribed class. One might seek to describe,

for example, all finite abelian groups, or all finite groups having a certain prescribed number of elements. In this section we shall see how the properties of the ring Z lead to a complete description of one very important class of groups—those generated by a single element. 213

214

FINITE GROUPS

Chap. IV

Definition. A cyclic group is a group that is generated by one of its elements. If a is an element of a group G, the subgroup of G generated by a will be denoted by [a] in this chapter.

Theorem 23.2. If a is an element of a group (G, -), then [a] = {a”: n e Z}, and the function g from Z into [a] defined by

g(n) = a“ for all n e Z is an epimorphism from (Z, +) onto the subgroup [a]. Proof. An easy inductive argument establishes that for each meN, a’" belongs to any subgroup containing a, and in particular, a'" E [(1]. Since for each m e N*, r“ = (a"‘)—1 by Theorem 14.8, a—m also belongs to any

subgroup of G containing a. Thus, [a] 2 {a"': m 62}. But since a”a”‘ = a”+"‘ and (a")—1 = a-" for all n, m e Z by Theorem 14.8, {a”‘: m e Z} is a subgroup by Theorem 8.3, and it clearly contains a. Hence, [a] S {a”‘: m e Z},

and also g is an epimorphism. If the composition of G is denoted additively, of course,

[a] = {n.a: n e Z}. Theorem 23.3. The group (Z, +) is cyclic. Every subgroup of (Z, +) is an ideal of the ring Z. Proof. For every n e Z, n = n.1 E [1] by Theorems 23.2 and 14.8, so 1 is a generator of (Z, +). Let H be a

subgroup of (Z, +). If n E Z and if h e H, then nh=n.he[h]§ H by Theorems 23.2 and 14.8. Hence, H is an ideal of the ring Z. It is easy to see that —1 is also a generator of (Z, +) and that l and —1

are the only generators of (Z, +).

Sec. 23

FINITE GROUPS

21 5

Theorem 23.4. A cyclic group is abelian. If a is a generator of a cyclic group (G, -) and if H is a subgroup of G, then G/H is a cyclic group, and aH is a generator of G/H. Proof. The first assertion is a consequence of Theorem 23.2 and the corollary of Theorem 20.4. Since G = {a": n e Z} and since (pH: x HxH is an epimorphism from G onto G/H,

G/H = {9911(0"): ’1 E Z} = {(PH(a)”= n E Z} = «4170"! n 62} by the Corollary of Theorem 20.5, whence G/H is cyclic with generator aH by Theorem 23.2. Theorem 23.5. For every m e N, the group (2,", +) is a cyclic group, and every subgroup of (Z,,,, +) is an ideal of the ring 2,". Proof. The first assertion follows from Theorems 23.3 and 23.4. Let H be a subgroup of (2,", +), and let (pm be the canonical epimorphism from Z onto Z,,,. If h + (m) E H and if n e Z, then by Theorem 14.8 and'the Corollary of Theorem 20.5,

(n + (m))(h + (m)) = nh + (m) = x + y from H X K into 2, is an isomorphism. For example,

A((3. 2) + (0, 4)) = A((3, 0)) = 3 + 0 = 3, A((3, 2)) + A((0, 4)) = 5 + 4 =. 3.

Sec. 24

FINITE GROUPS

221

Thus, G1 X G2 is a group, called the cartesian product of the groups 01 and G3. Furthermore, if G1 and G2 are commutative, then so is G1 X G2, since (x1, xa)(yi,ya) = (xlyh xzys) = (y1x1,y2xz) = (y:.ya)(x1, x2).

Of course, if the compositions of G1 and G2 are denoted by symbols similar to +, the composition of 61 X G2 is also customarily denoted by +. This method of forming new groups may be generalized in an obvious way. Let (G, -1), . . . , (Gm -,,) be a sequence of groups. We define a composition- onG1 X X ay (x1, x2» - ' - ’ xn) ' (J’nyz. - ' ' syn) = (351,71, xzya, - ° - s xnyn)'

Equipped with this composition, G1 X . . . X G” is a group, called the cartesian product of the groups G1, . . . , G". Indeed, multiplication is associative as before, the neutral element of G1 X . . . X G” is (e1, . . . , e,),

where e,, is the neutral element of G, for each k e [1, n], and the inverse of (x1, . . . , x”) is (xfl, . . . , x;1). Furthermore, it is easy to verify as before that if G, . . . , G" are commutative groups, then G1 X . . . X G” is also commutative. The following theorem is easy to prove. Theorem 24.1. If GI, . . . , G", G{, . . . , G; are groups and if for each

k e [1, n], 13, is an isomorphism from 0,, onto G,’" then f: (x19 0 - - 9 xn) H (fl.(x1)a - - - 9fn(xn))

is an isomorphism from the group G1 X .. . X G,l onto the group G; X

. .. x G,’,.

Sometimes a group is isomorphic in a “natural” way to the cartesian product of certain of its subgroups. Before we indicate precisely what is meant by “natural,” let us consider an example. The cyclic group Za contains

a subgroup H of order 2, namely, {0, 3}, and a subgroup K of order 3, namely, {0, 2, 4}. A simple calculation shows that the restriction of addition

on Z, to H x K is an isomorphism from the group H X K onto 2,, that is, that the function A: (x, y) »—> x + y from H X K into Z, is an isomorphism.

For example,

A((3, 2)) + A((0, 4)) = 5 + 4 =. 3.

222

FINITE GROUPS

Chap. IV

Thus, Z. is isomorphic to H X K under a “natural” isomorphism, the restriction of its composition, addition, to H X K.

Definition. A group (G, -) is the direct product of subgroups H and K if the function M from H X K into G defined by M(x,y) = xy

for all (x, y) e H X K is an isomorphism from the group H x K onto the group G. If the composition of G is denoted by a symbol similar to +, we say that G is the direct sum rather than the direct product of H and K. Theorem 24.2. If H and K are subgroups of a group (G, ~), then G is the direct product of H and K if and only if the following three conditions hold: 1° Every element of H commutes with every element of K. 2° HK = G. 3° H n K = {e}, where e is the neutral element of G. Proof. Let M be the function from H x K into G defined by M(x, y) = xy.

We shall show that M is a homomorphism if and only if 1° holds. Indeed, if M is a homomorphism and if x e H, y e K, then as

(Jay) = (e,y)(x, e), xy = M(x,y) = M(e,y)M(x, e) = eyxe = yx.

Conversely, if every element of H commutes with every element of K, then for all x, x’ e H, y, y’ e K,

M((X.y)(x', y’)) = M(xx’,yy’) = xx’yy’ = xyx’y’ = M(x, y)M(x’, y'). If M is a homomorphism, then M is a monomorphism if and only if 3° holds. Indeed, if M is a monomorphism and if z e H n K, then 2 E H and r1 e K, so M(z, r1) = e, whence (2, r1) = (e, e) and thus, 2 = e. Conversely, if H n K: {e} and if M(x, y) = e, then xy = e, so y = x‘1 EH

and hence y e H n K, and consequently, y = e, x = e. Clearly, M is

surjective if and only if 2° holds. Theorem 24.3. If H and K are subgroups of a group (G, '), then G is the

direct product of H and K if and only if the following three conditions hold:

Sec. 24

FINITE GROUPS

223

1° H and K are normal subgroups of G. 2° HK = G. 3° H n K = {e}, where e is the neutral element of G.

Proof. By Theorem 24.2, it suffices to show that if 2° and 3° hold, then 1° holds if and only if each element of H commutes with each element of K. Necessity: Let x E H, y e K. As K is normal, xyx-1 e K, so xyx‘1y‘1 e K; since H is normal and since x‘ll e H, yx— y—1 e H, so xyx‘1 y‘1 e H; hence,

xyx— y‘1 e H n K: {e}, so xy= yx. Sufliciency: Let a e G. By 2°, a= hk for some hEH, keK. If e, then axa—1= hkxk—1h—1 = hxkk‘1h‘1— — hxh—1 E H, as x commutes with k, thus, HIS normal. If y e K, then aya 1 = h(kyk—1)h—1=(kyk‘1)hh‘1= kyk‘1 eK as kyk‘1 commutes with h; thus, K is normal. If H and K are subgroups of a group (G, -), it is sometimes desirable

to know when the subgroup generated by the set H U K is the direct product of H and K. To identify that subgroup the following theorem is helpful. Theorem 24.4. If H and K are subgroups of a group (G, -) and if H is a normal subgroup of G, then HK is a subgroup of G and is, moreover, the smallest subgroup of G containing H and K; if H and K are both normal subgroups of G, then HK is a normal subgroup of G. Proof. If x1, x2 6 H, y1, y2 6K, then yly;1r1=zyg;1 for some zeH as H is normal, so (x1y1)(x2yz)“ = x1y1y21x21=z(x1 )(YIJ’21-)EHK Thus, by Theorem 8.3, HK is a subgroup that contains H and K, since

H = He S HK, K = eK _C_ HK. Any subgroup containing H and K clearly contains HK. Thus, HK is the smallest subgroup of G containing H and K. If H and K are both normal subgroups and if a e G, then

a(HK) = (aH)K = (Ha)K = H(aK) = H(Ka) = (1110a, so HK is itself a normal subgroup. An inductive argument establishes the following corollary: Corollary. If H1, H2, . . . , H” are normal subgroups of a group (G, '), then

HIHa . . . H" is a normal subgroup of G and is, moreover, the smallest subgroup of G containing H1, H2, . . . , H”.

It is easy to extend the definition of direct product:

Sec. 24

FINITE GROUPS

223

l° H and K are normal subgroups of G. 2° HK = G.

3° H n K = {e}, where e is the neutral element of G.

Proof. By Theorem 24.2, it suffices to show that if 2° and 3° hold, then 1°

holds if and only if each element of H commutes with each element of K. Necessity: Let x E H, y E K. As K is normal, xyx—1 e K, so xyx—1y‘1 e K; since H is normal and since x—1 e H, yx—1y—1 e H, so xyx‘1y—1 E H; hence,

xyx—1y‘1 e H n K = {e}, so xy = yx. Sufficiency: Let a e G. By 2°, a=hk for some heH, keK. If e, then axa‘1— — hkxk—1h—1— — hxkk‘1h‘1 = hxh—1 e H, as x commutes with k, thus, H is normal. If y E K,

then aya—1— — h(kyk—1)h—1— — (kyk—1)hh—1— — kyk—1 e K as kyk—1 commutes with h, thus, K is normal. If H and K are subgroups of a group (G, °), it is sometimes desirable

to know when the subgroup generated by the set H U K is the direct product of H and K. To identify that subgroup the following theorem is helpful. Theorem 24.4. If H and K are subgroups of a group (G, -) and if H is a

normal subgroup of G, then HK is a subgroup of G and is, moreover, the smallest subgroup of G containing H and K; if H and K are both normal subgroups of G, then HK is a normal subgroup of G. Proof. If x1, x2 6 H, y1, y2 e K, then yLyg1x2-1 = 2:y,y;1 for some z e H as H is normal, so

(x1Y1)(x2.V2)_1 = x1J’1Ya-1xz_1 = (351901351) 5 HK-

Thus, by Theorem 8.3, HK is a subgroup that contains H and K, since H = He S HK, K = eK E HK. Any subgroup containing H and K clearly contains HK. Thus, HK is the smallest subgroup of G containing H and K. If H and K are both normal subgroups and if a e G, then a(HK) = (aH)K = (Ha)K = H(aK) = H(Ka) = (HK)a,

so HK is itself a normal subgroup.

An inductive argument establishes the following corollary: Corollary. If H1, H2, . . . , H” are normal subgroups of a group (G, -), then

H1H2 . . . H" is a normal subgroup of G and is, moreover, the smallest subgroup of G containing H1, H2, . . . , Hn.

It is easy to extend the definition of direct product:

224

FINITE GROUPS

Chap. IV

Definition. A group (G, -) is the direct product (direct sum if the composition of G is denoted by a symbol similar to +) of the sequence H1, H2, . . . , H. of subgroups of G if the function M from H1 X H2 X . .. X H" into G

defined by M(xl, x2, . . . , x") = x1x2 . . . x"

is an isomorphism from the group H1 X H2 X . . . X H” onto the group G. If n = l and HI = G, then M is simply the identity automorphism of G; hence, G is the direct product of the sequence whose only term is G itself. Necessary and sufficient conditions for a group to be the direct product of a sequence of its subgroups analogous to those of Theorem 24.3 may be given (Exercise 24.3), but we shall actually use only the following theorem.

Theorem 24.5. If a group (G, -) is the direct product of subgroups H1, H2, . . . , H”, then H1, H2, . .. . , Hware normal subgroups of G. Proof. By hypothesis, the function M: (x1, x2, . . . , x”) i—rxlx2 . . . xfl from

H1 X H2 X . . . X H" into Gis an isomorphism. To show that H, is normal, let h 6H,, a e G. Then a = M(al, a2, . . . , a,) where a,, 6H,, for each k e [1, n]. Hence,

aha‘1 = M(a1,. . . , an)M(e, . . . , e, h, e, . . . , e)M(a1, . . . , an)-1 = M(al, . . . , an)M(e, . . . , e, h, e, . . . , e)M(af1, . . . , (1:1)

=M((a1,...,a,,)(e,...,e,h,e,...,e)(a1_1,...,a;1)) = M(e, . . . , e, a,ha,‘1, e, . . . , e) = aka:1 6 H, as a, e H,. Thus, H, is normal.

Theorem 24.6. Let H and K be subgroups of a group (G, -). If H is the direct product of subgroups H1, . . . , Hm and if K is the direct product of subgroups K1, . . . , K”, then G is the direct product of H and K if and only if G is the direct product of H1, . . . , Hm, K1, . . . , K”.

Proof. The function J: (x1, ' ' - ,xmayh ' ' - ayn)'-)((x19 ‘ ' - axm)s(y1, ‘ ' ' 2.71.»

is easily seen to be an isomorphism from the group H1 X K1X XKn onto the group (H1 X Theorem 24.1 ,

XHm)X(K1X

X Hm X XK,,). By

P: ((xl, . . . , xm), (yl, . . . ,yn))=!-> (361- - -xm,y1 . - -}’n)

Sec. 24

FINITE GROUPS

225

is an isomorphism from (H1 X . .. X Hm) X (K1 X . . . X K,,) onto H x K. Let M be the function from H X K into G defined by M(x, y) = xy. Then (MoPoJ)(x1,...,xm,y1,...,yn)=x1...xmy1...y,,.

As P and J are isomorphisms, M is an isomorphism if and only if M o P o J is an isomorphism; i.e., G is the direct product of H and K if and only if G is the direct product of H1, . . . , Hm, K1, . . . , Kn. Theorem 24.7. Let r1, . . . , r,,, be natural numbers such that r, and r, are relatively prime whenever i, j E [1, m], i 7E j. If G is a group of order rlrg . . . rm that contains a normal subgroup Hk of order rk for each k e [1, m],

then 1° for each n e [1, m], G" = Hn . . .Hu is a normal subgroup of G and is, moreover, the direct product of H1, H2, . . . , H”,

2° G is the direct product of H1, H2, . . . , Hm.

Proof. By the Corollary of Theorem 24.4, G" is a normal subgroup of G.. To prove 1°, it suffices by induction to show that if G” is the direct product of H1, . . . , H” where n < m, then G,,_,_1 is the direct product of H1, . . . , Hm“. Suppose, therefore, that G” is the direct product of H1, . . .' , H".

Then G" has rlra . . . r,” elements. Since r.,,+1 is relatively prime to r1, . . . , r”, we conclude that r"+1 is relatively prime to rlr2 . . .r” by (UFD 3'). If a e G” n HMI, then the order of a divides rlrz . . .. r,, and r,,+1 and hence is 1, so a = e. Thus, by Theorem 24.3, G,,+1 is the direct product of G” and HM.“ so by Theorem 24.6, Gn+1 is the direct product of H1, . . . , H“, Hn-I-l' Thus, 1° holds. In particular, as Gm is the direct product of H1, . . . , Hm, we

conclude that Gm has rlr2 . . . rm elements and hence is G; thus, 2° holds. Corollary. Let r1, . . . , rm be natural numbers such that r, and r, are

relatively prime whenever i, j e [1, m], i 75 j. A cyclic group of order rlr2 . . . rm is the direct product of its cyclic subgroups of orders r1, . . . , rm.

Theorem 24.8. Let r1, . . . , rm be natural numbers such that r, and r, are

relatively prime whenever i, je [1, m], isé j. If G is a group of order rlrfl . . . rm that contains a normal subgroup H, of order r,, for each k e [1, m], and if ak e Hk for each k e [1, m], then alas. . .am is a generator of G if

and only if a,‘ is a generator of Hk for each k e [1, m]. Proof. For each n e [1, m], G” = H1112 . . . H" is a normal subgroup of G

and is the direct product of H1, . . . , H” by Theorem 24.7. For each n e [1, m] let Sn be the statement: alaz. . .an is a generator of G,, if and only if a,, is a generator of H, for each k e [1, n]. Clearly, S1 is true. Assume

226

FINITE GROUPS

Chap. N

that S” is true where n < m; we shall show that Sn+1 is true. Since S, is true, S,,+1 is equivalent to the following statement: alaa . . . a”+1 is a generator of G,,+1 if and only if alas . . . a” is a generator of Gn and (1,,+1 is a generator of Hn+1' The order of G” is rlr2 . . . r" since G,I is the direct product of H1, . . . , H”; as before, r,,+1 and rlrg. . . r” are relatively prime. Let a = ala, . . . a", b = an“, r = rlra . . . r", s = ”+1. We are, therefore, to prove

that if a e G”, b 6 HM“, then ab is a generator of the group G,,+1 of order rs if and only if a is a generator of the subgroup G" of order r and b is a generator of the subgroup H”+1 of order s.’ Necessity: As ab is a generator of GMI, Gn+1 is cyclic (and hence abelian), and (ab)‘ clearly has order r. But (ab)‘ = a'b‘ = a' by Theorem 23.8, since H,,+1 has order s; therefore, a’ has order r and hence is a generator of G“,

because the cyclic group G."+1 of order rs has only one subgroup of order r by Theorem 23.9; hence, [a] E G7, = [a‘] S [a], so a is a generator of G”.

Similarly, b is a generator of Hn+l' Sufficiency: Since G,,+1 is the direct product of H1, . . . , H,,+1 and since G" is the direct product of H1, . . . , H”, Gn+1 is the direct product of G" and H"+1 by Theorem 24.6. Thus, as a is a generator of G" and as b is a generator of HM“, G."+1 is the direct product of two cyclic (and hence abelian) subgroups and therefore is abelian. By Bezout’s Identity, there exist x, y e Z such that xr + ys = 1. Consequently, because the orders of a and b are r and s respectively, (ab)" = awbw ___ a1—aw = a, (ab)mr ___ acrbxr = bl—u = 1,.

Thus, [ab] 2 [a] = G”, [ab] 2 [b] = HMl, so [ab] 2 G,,H,,+1 = G,,+1. Therefore, ab is a generator of Gn+1‘ By induction, therefore, S,, is true for each n e [1, m]. In particular, Sm

is true. As before, Gm = G, and consequently, the proof is complete. Corollary 24.8.1. Let r1, . . . , r,,, be natural numbers such that r‘ and r, are relatively prime whenever i, j e [1, m], 1'72 j. If G is a group of order

r1r2.. . rm that contains a normal cyclic subgroup of order r,, for each k e [1, m], then G is cyclic. Corollary 24.8.2. Let r1, . . . , r", be natural numbers such that r, and r, are

relatively prime whenever i, je [1, m], igé j. If a group G is the direct product of cyclic subgroups of orders r1, . . . , rm, then G is cyclic. The assertion follows at once from Theorem 24.5 and Corollary 24.8.1. Corollary 24.8.3. Let r1, . . . , rm be natural numbers such that r, and r, are

Sec. 24

FINITB GROUPS

227

relatively prime whenever i, je [1, m], iqé j. The Euler function on then

satisfies 7’01": - . - rm) = ‘7’("-1)‘P("2)- - - ‘P(rm)-

The assertion follows from the Corollary of Theorem 24.7 and Theorem

24.8. Every group (G, -) is the direct product of itself and the subgroup {e} consisting only of the neutral element. Groups which are the direct product of no other pairs of subgroups are called indecomposable: Definition. A group G is decomposable if it is the direct product of two proper subgroups. A group is indecomposable if it is not decomposable. Theorem 24.9. If (G, -) is a cyclic group of order n and if n = p;1 . . . 5;;is the prime decomposition of n, then G is the direct product of H1, . . . , Hm

where H, is the subgroup of G of orderpp for each k e [1, m], and moreover, each H, is indecomposable. Proof. By the Corollary of Theorem 24.7, we need only show that if H is a cyclic group of order p’ where p is a prime, then H is indecomposable. Let H1 and H2 be subgroups of H; their orders are p’1 and 17': respectively for

some natural numbers r1, r2 by Lagrange’s Theorem; we may assume that r2 s r1. Since H1 is cyclic, it has a unique subgroup of order p" by Theorem 23.9, and that subgroup must be the unique subgroup of H of order 1)", namely, H2. Thus, H2 S H1. Consequently, by Theorem 24.3, H is the direct product of H1 and H2 only if H2 = {e}, whence H1 = H1{e} = H.

Thus, H is indecomposable. To illustrate further conditions under which a group is the direct product of certain of its subgroups, we consider finite commutative groups whose elements other than the neutral element all have order p, where p is a prime. We note first that the order of each element of the additive group 2, X . . . >< Z1, other than the zero element is p.

Theorem 24.10. Let p be a prime, and let (G, -) be a finite group containing more than one element. If G is commutative and if all of its elements other than the neutral element e have order p, then the order of G is p’" for some m e N*, and G is the direct product of a sequence of m subgroups, each cyclic of order p. Proof. There exist natural numbers k > 0 such that G contains a subgroup

that is the direct product of k cyclic subgroups of order p; for example, if a 9/: e, then [a] is cyclic of order p by hypothesis, so 1 is such a natural

228

FINITE GROUPS

Chap. IV

number. Let m be the largest natural number such that G contains a subgroup of orderpm that is the direct product of m subgroups, all cyclic of order p, and let G’ be such a subgroup. We wish to prove that G’ = G. Suppose

that G’ is a proper subgroup of G; then there exists b 9? G’. Consequently, G’ n [b] = {e}, for if x were an element of G’ n [b] other than e, then x would be a generator of [b] by the Corollary of Theorem 23.8, so b e [x] S G’, a contradiction. By Theorem 24.4, G’ [b] is a group and is the direct product of G’ and [b] by Theorem 24.2. Since [b] is cyclic of order p, G’ [b] is the direct product of m + 1 subgroups, each cyclic of orderp, by Theorem

24.6. This contradiction shows that G’ = G. The hypothesis of Theorem 24.10 that G be commutative may be omitted

ifp = 2: Theorem 24.11. If (G, -) is a group all of whose elements other than the neutral element e have order 2, then G is commutative.

Proof. By hypothesis, z2 = e for all z e G, whence r1 = 2. In particular, if x, y e G,

xyx‘ly'1 = xyxy = e, so xy = yx. By the Corollary of Theorem 23.8, a group of prime order p is cyclic and hence, by Theorem 23.6, is isomorphic to the (additive) cyclic group Z,. One consequence of Theorems 24.10 and 24.11 is the determination of all groups of order 4: Theorem 24.12. A group G of order 4 is either cyclic or the direct product of two cyclic subgroups of order 2. Thus, to within isomorphism the only groups of order 4 are Z1 and Z2 x 22. Proof. Suppose that G is not cyclic. Then every element of G other than the neutral element has order 2 by Theorem 23.8. Consequently, by Theorems 24.10 and 24.11, G is the direct product of two cyclic subgroups of order 2.

The group Z, x Z, is often called the four group; it is isomorphic to the group of all symmetries of a rectangle that is not a square.

EXERCISES 24.1. Verify that if 61,..., G” are groups, then G1 x .. . x Gn with the

indicated composition is a group, and that G1 x . . . x G" is commutative if each Gk is.

Sec. 24

FINITE GROUPS

229

24.2. Prove Theorem 24.1. 24.3. A group (G, -) is the direct product of a sequence H1,H2, . . . , H,, of

subgroups if and only if the following three conditions hold: 1° H1, H2, . . . , Hn-are normal subgroups of G.

2° H,H,...H,, =G. 3° For each is [2, n], (H1. . . H,_1) n H, = {e}. [Show by induction that HIH,‘ . . . H.i is the direct product of H1, H2, . . . ,

H; for each i6 [2, n].] 24.4. Prove the Corollary of Theorem 24.4. 24.5. Verify that the function J of Theorem 24.5 is an isomorphism. 24.6. If a group (G, -) is the direct product of subgroups H and K, thenf: x 1—) xK,

x e H, is an isomorphism from H onto G/K, and g: yHyH, y 6K, is an isomorphism from K onto G/H. Let (G, -) be the group of symmetries of the square, let G' be the multiplicative group {1, —1}, and let E be the cartesian product group G x G’. Let H = G X {1}, K = {(r0, 1), (h, —l)}. Show that H and K. are subgroups of E, that HK = E and H n K = {(r0, 1)}, that E is isomorphic

to the group H x K, but that E is not the direct product of H and K. 24.8. If K1 and K2 are normal subgroups of groups G1 and G2 respectively, then

K1 >< K2 is a normal subgroup of the group G1 x Ga, and the group (G1 x G3)/(K1 x K5,) is isomorphic to the group (GI/K1) x (G2/K2). 24.9. Iffis a function from a group G1 into a group G2, thenfis a homomorphism

if and only if the graph off is a subgroup of the group G1 x Ga. 24.10. Let (G, -) be a group, and let . be the composition on G x Aut(G)

defined by

(x. 00-0). fl) = (NO). off?)Then (G x Aut(G),.) is a group, xi—> (x, la) is an isomorphism from G onto a normal subgroup of G x Aut(G), and at 1—) (e, a) is an isomorphism from Aut(G) onto a subgroup of G x Aut(G).

24.11. If n > 1 and if n = p51. . .173» is its prime decomposition (where n, 2 l for all k e [1, m]), then m

l

k-l

Pk

92(n) = n H 1 — —) . *24.12. Prove that ¢p(n) = + 00. [Show that ifp is the rth prime and if n 2 2' , n—poo

then fl ,

FINITE GROUPS

238

Chap. IV

is a well-defined isomorphism from the group (Z... x Aut(Zm), .) (Exercise

24.10) onto the group Aut(Dm). 25.8. Let m 2 3, let Jam be the set of all integers in [1, 2m — 1] that are relatively

prime to 2m, and let a and b be generators of the dicyclic group G of order 4m satisfying a” = e, b2 = a“, bab‘1 = a‘l. (a) For each i 6 N2", and each s e J2m, there is one and only one automorphism ft: of G satisfying f,,,(a) = a” and fi,,(b) = a‘b. (b) For each s 612,”, let L, be the automorphism of the additive group Zm defined by L,(x) = s.x (Exercise 23.2). The functionF: (i + (2m), L,) Hf“

is a well-defined isomorphism from the group (Zm x Aut(n), .) onto the group Aut(G).

26. Sylow’s Theorems“

Let G be a finite group of order n. By Lagrange’s Theorem, if H is a sub-

group of G of order m, then m I n. If G is cyclic, the converse holds: if m | n, then a cyclic group of order n contains a subgroup of order m (Theorem 23.9). However, it is not true in general that if m is a divisor of n, then G

has a subgroup of order m (later we shall see, for example, that the alternating group ‘11., of order 12 contains no subgroup of order 6). For those divisors m ofn that are powers of a prime, however, G necessarily contains a subgroup of order m; this is the content of Sylow’s First Theorem. Further information concerning the nature and number of subgroups of G of order m, where m is a power of a prime, is given by Sylow’s Second and Third Theorems. In this section we shall denote by n(X) the number of elements in a

finite set X, and by e the neutral element of a group whose composition is denoted multiplicatively. The proofs we shall give of Sylow’s Theorems depend on certain counting arguments, a general framework for which is contained in the following definition. Definition. Let (G, ') be a group. A G-space is an ordered couple (E, .) where E is a set and . is a function from G x E into E (whose value at (a, z) e G X E is denoted by a.z) satisfying the following two conditions for all a,beGand allzeE: 1° e.z = z. 2° (ab).z = a.(b.z). The function . is called the action of G on E. If (E, .) is a G-space, we also say that the group G acts on E under ., or that E is a set with group of operators G. If (E, .) is a G-space, for each 2 e E the set G.z defined by

G.z = {a.z: a E G} * Except for 547, subsequent sections do not depend upon this section.

Sec. 26

17mm: GROUPS

239

is called the orbit of z, and the subset G, of G defined by

G, = {aEG:a.z=z}

is called the stabilizer of 2; an orbit of the G-space E is simply the orbit of z for some 2 e E.

Theorem 2.6.1. Let (E, .) be a G-space. 1° The orbits of E form a partition of E. 2° For each 2 e E, the stabilizer G, of z is a subgroup of G, and aG‘-.I—> a.z

is a well-defined bijection from the set G/G, onto the orbit of 2. Proof. 1° Since e.z = z, z e G.z. Thus, no orbit is the empty set, and

every element of E belongs to at least one orbit. It remains to show that if x e 6.2 n G.w, then 0.2 = G.w. By our assumption, there exist a, b e G such that x = a.z = b.w. For any 0 e G,

e.z = (ca—1a).z = (ca-1).(a.z) = (ca—1).(b.w) = (ca—1b).w e G.w, and similarly, c.w e G.z. Thus, 6.2 = G.w, so the orbits partition E.

2° Clearly, e e G,, so G, 7’: (0. If a, b E G,, then a.z = z = b.z, so ab—1.z = ab—1.'(b.z) = (ab—1)b.z = a.z = z,

whence ab—1 6 0,. Thus, by Theorem 8.3, G, is a subgroup of G. We have only to show now that aG, = bG, if and only if a.z = b.z. But aG, = bG, if and only if b—la E G,, that is, if and only if b—la.z = 2; but if z = b—la.z, then b.z = b.(b—1a.z) = (bb—la).z = a.z, and if b.z = a.z,

then 2 = (b—lb).z = b—1.(b.z) = b‘1.(a.z) = b‘la.z. Theorem 26.2. If (G, -) is a group, for each a e G the function 19,: xI—>axa'1

is an automorphism of G. The proof is easy.

Sec. 26

FINH'E GROUPS

239

is called the orbit of z, and the subset G, of G defined by

G, = {aeG:a.z=z}

is called the stabilizer of 2; an orbit of the G-space E is simply the orbit of z for some 2 e E. Theorem 26.1. Let (E, .) be a G-space.

1° The orbits of E form a partition of E. 2° For each 2 e E, the stabilizer G, of z is a subgroup of G, and aG,-_I—> a.z

is a well-defined bijection from the set G/G, onto the orbit of 2. Proof. 1° Since e.z = z, z e G.z. Thus, no orbit is the empty set, and

every element of E belongs to at least one orbit. It remains to show that if x e G.z n G.w, then G.z = G.w. By our assumption, there exist a, b e G such that x = a.z = b.w. For any c e G,

c.z = (ca—1a).z = (ca—1).(a.z) = (ca—1).(b.w) = (ca—1b).w e G.w,

and similarly, c.w e G.z. Thus, G.z = G.w, so the orbits partition E. 2° Clearly, e e G,, so G, 7E 0. If a, b e G” then a.z = z = b.z, so ab‘1.z = ab‘1.'(b.z) = (ab-1)b.z = a.z = z,

whence ab“1 6 6,. Thus, by Theorem 8.3, G, is a subgroup of G. We have only to show now that (10, = bG, if and only if a.z = b.z. But aG, = bG, if and only if b—la e G” that is, if and only if b‘la.z = 2; but if Z = b—la.z, then b.z = b.(b—1a.z) = (bb‘la).z = a.z, and if b.z .= a.z,

then 2 = (b—lb).z = b—1.(b.z) = b—1.(a.z) = b‘la.z. Theorem 26.2. If (G, -) is a group, for each a E G the function

K“: x I—> am—1

is an automorphism of G. The proof is easy.

240

FINITB GROUPS

Chap. IV

An automorphism g of (G, -) is called an inner automorphism of G if g = Ka for some a e G. If x E G, an element y of G is a conjugate of x if

y = axa‘1 for some a E G, that is, if y is the image of x under an inner automorphism of G. We shall denote by Conj(x) the set of all conjugates of x; a subset C of G is called a conjugate class of G if C = Conj(x) for some

x e G. If x belongs to the center Z(G) of G, then axa"L = x for all a e G, so x itself is the only conjugate of x; conversely, if Conj(x) = {x}, then axa‘1 = x for all a e G, whence ax = xa for all a e G, so x eZ(G). Therefore, n(Conj(x)) = 1 if and only if x eZ(G). Our first goal is to prove that if G is a group whose order is a power of a prime, then the center of G contains at least one element other than the

neutral element. To prove this, we shall study a suitable action of the group G on the set G.

Let G be a group, and let . be the function from G X G into G defined by a.x = axa‘l. Clearly, e.x = x, and ab.x = abx(ab ‘1 = a(bxb‘1)a‘1 = a.bxb—1 = a.(b.x) for all a, b, x e G. Thus, (G, .) is a G-space; . is called the action of G on

G by conjugation. Clearly, the orbit of x for this action is simply Conj(x).

Thus, by 1° of Theorem 26.1, the sets Conj (x), where x e G, form a partition of G. Suppose further that G is finite. Selecting exactly one element from each conjugate class of G, we obtain a subset F of G such that {Conj(x): x e F} is a partition of G, and Conj(x) 72 Conj(y) if x, y e F and x 9E y. Thus,

n(G) = 2 n(Conj(x)). 26F

Let x1, . . . , x, be the elements of F not in the center Z of G. If x 62, then

Conj(x) = {x}, so x must belong to F; thus, F = Z U {x1, . . . , x,},whence

M) = n c is a well-defined injection from G/S into G/S. Indeed, xS= yS if and only if y—lx e S, and c— — cyS if and only if (cy)‘1(cx) E S; thus, as (cy)—1(cx)=c1—cx — y‘lx, we conclude that xS= yS if and only if c— — cyS. Furthermore, c" is a permutation of G/S, for if z e G, cA(c—1zS) = zS. Clearly,

Sec. 26

FINITE GROUPS

245

Case 1: [a] is a normal subgroup. We shall first show that H contains an element of order 2 that commutes with a. Suppose first that H is cyclic, Let b be a generator of H. Then bab‘1 is either a or a2 as [a] is normal;

in the former case, b commutes with a and hence b” does also; in the latter

case, bi’ab—2 = b(bab‘1)b‘1 = bazb—1 = (bab‘l)2 = a4 = a; thus, in either case b2 is an element of order two that commutes with (1. Suppose next that H is the direct product of two cyclic subgroups of order 2. If b1 and b2 are elements of H of order 2 that do not commute with a, then blab;1 =

a” = bflabg,‘1 as [a] is normal, so (b1b2)a(b1b2)_1 = b1(b2ab;1)bf1 = biazbil = (hub?!)2 = a4 = a,

and moreover blbg is not e and hence is an element of order 2, since each element of H is its own inverse. Thus, in either case H contains an element 6 of order 2 that commutes with a and hence also with a2. By Theorem 24.4, [a] [c] is a subgroup and is, moreover, the direct product of [a] and [c] by

Theorem 24.2; thus,” K = [a] [c] is a cyclic group of order 6 by Corollary 24.8.2. By Theorem 18.6, K is a normal subgroup, and G/K is a (cyclic) group of order 2. Let x be a generator of K, y ¢ K. By Theorem 25.2, there exist r E No, s e Z* suchthaty2 = x’,yxy—1 = x‘, s2 E 1 (mod 6), sr E r (mod 6), and moreover, G is commutative if 3 E1 (mod 6). As s” E 1 (mod 6),

either .9 E 1 (mod 6) or s E 5 E ——1 (mod 6). Case 1(a): s E 1 (mod 6) and r is even. Let r = 2t and let z = yx’“. Then 2 ¢ K and z”: yax‘“ = e, so K[z] is a subgroup by Theorem 24.4 that is the direct product of K and [2] by Theorem 24.2; hence, K[z] has 12 elements and thus is G. Therefore, G is the direct product of cyclic subgroups of orders 6 and 2 and hence is isomorphic to Z, x Z2. Case 1(b): SEI (mod 6) and r is odd. Let r = 2t + l, and let 2 =yx3". Then z“ = y”x“" = x, so the order of z divides 12, but since z“ = x3 qé e and

z4 = x2 E e, the order of z is a divisor of neither 6 nor 4; hence, the order of z is 12, so G is cyclic and thus isomorphic to Z12. Case 1(c): s E —1

(mod 6). Then —r E r (mod 6), so either r = 0 or r = 3. By Theorem

25.2, G is isomorphic to the dihedral group of order 12 in the former case and to the dicyclic group of order 12 in the latter.

Case 2: [a] is not a normal subgroup. Let S = [a], a Sylow 3-subgroup

of G. We shall first show that for each c e G, c": xS.’ n—> c is a well-defined injection from G/S into G/S. Indeed, xS = yS if and only if y‘lx e S, and c = cyS if and only if (cy)—1(cx) e S; thus, as (cy)—1(cx).= y‘lc‘lcx = y‘lx, we conclude that xS = yS if and only if c = cyS. Furthermore, cA is a permutation of G/S, for if z e G, c"(c'l) = zS. Clearly,

246

FINITE GROUPS

Chap. IV

(cd)" = c/‘dA for all e, d e G. Thus, F: c |-> cA is a homomorphism from G

into the group 60,3 of all permutations of G/S. If cA is the identity permutation, then c = xS for all x e G, so in particular c e ceS = eS = S. The kernel of F is, therefore, a normal subgroup of G contained in S.

Since S has prime order, {e} and S are the only subgroups of S; consequently, because S is not normal, the kernel of F is. {e}, so G is isomorphic to a group of permutations of the set G/S. As G/S has four elements, 60,5 is clearly isomorphic to 64 (Exercise 7.5); hence, G is isomorphic to a subgroup of 64 of order 12. By Theorem 26.7, G is isomorphic to ‘11,. Our further study of subgroups of a finite group requires the following general theorem. Theorem 26.9. If K is a normal subgroup of a group (G, .) and if H is a subgroup of G, then f: xh) xK is an epimorphism from H onto HK/K with kernel H n K, and consequently, g: x(H n K)n—->xK is an isomorphism from H/(H n K) onto HK/K. Proof. By Theorem 24.4, HK is indeed a subgroup of G. Clearly, f is a homomorphism. If x e H, y e K, then xyK = xK = f(x); hence, f is surjective. The kernel off is clearly H n K. The final assertion, therefore, follows from Theorem 20.9.

Let S be a subset of a group (G, -). A subset T of G is a coniugate of S if T = aSa—3l for some a e G. Thus, the conjugates of S are simply the images of S under the inner automorphisms of G. If S is a subgroup, then a conjugate of S is a subgroup and hence is called a conjugate subgroup of S. We shall denote by Conj(S) the set of all conjugates of a subset S of G. Let H be a subgroup of G. A subset T of G is an H—conjugate of S if T= aSa‘I for some a EH. Thus, the H-conjugates of S are simply the images of S under the inner automorphisms of G arising from elements of H. We shall denote by ConjH(S) the set of all H-conjugates of S. Definition. If S is a subgroup of a group (G, ~), the normalizer of S is the set N(S) defined by N(S) = {a e G: aSa‘1 = S}. If H and S are subgroups of G, the H-normalizer of S is the set NH(S) defined by NH(S) = {a E H: aSa‘1 = S}.

Thus, NH(S) = N(S) n H.

Sec. 26

FINITE GROUPS

247

Theorem 26.10. If S is a subgroup of a group (G, -), then N(S) is a subgroup

of G that contains S, and S is a normal subgroup of N(S). The proof is easy. Theorem 26.11. Let p be a prime. If H is a subgroup of order p” of a finite group G and if T is a Sylow p-subgroup of G, then

Proof. Clearly H n T S NH(T) E H, so we need only prove that NH(T) S T. Let p‘" be the highest power of p that divides the order g of G, and let g = pmq, so that p ,l’ q. By Theorem 26.10, T is a normal subgroup of N(T), and NH(T) is also a subgroup of N(T). Hence, by Theorem 26.9, the groups

NH(T)/(T n NH(T)) and NH(T)T/Tare isomorphic. The order of the former divides the order of NH(T), a subgroup of H, and hence is p’ for some 3 e [0, r]. The order of NH(T)T is a multiple of the order p’" of T and a divisor of the order q of G; hence, the order of NH(T)T is pmt where

p,}’ t, and consequently, the order of NH(T)T/T is I. Since t = p’ and p ,|’ t, we conclude that t = 1. Thus, T n NH(T) = NH(T), so NH(T) S T. Theorem 26.12. Let p be a prime, let G be a finite group of order g, let pm be the highest power of p that divides g, and let S be a Sylow p-subgroup of G. 1° (Sylow’s Second Theorem) If H is a p-subgroup of G, then H is contained in a conjugate subgroup of S. In particular, every Sylow p-subgroup of G is a conjugate subgroup of S. 2° (Sylow’s Third Theorem) The number n,, of Sylow p-subgroups of G is a divisor of g/p'" and satisfies the congruence n, E 1 (mod p);

moreover, n,=l+k1p+k2p3+...+kmp”‘ where for each r e [1, m], the number of Sylow p-subgroups T for

which S n T has order pm” is k,p'.

Proof. Let g = pmq, whence p*q. For any subgroup H of G, (Conj(S), .) is an H-space where . is defined by a.T = aTa"1L

for all a e H, T e Conj(S). Indeed, if T e Conj(S), clearly, aTa‘1 e Conj(S),

Sec. 26

FINITE GROUPS

247

Theorem 26.10. If S is a subgroup of a group (G, -), then N(.S') is a subgroup of G that contains S, and S is a normal subgroup of N(S). The proof is easy.

Theorem 26.11. Let p be a prime. If H is a subgroup of order p’ of a finite group G and if T is a Sylow p-subgroup of G, then

Proof. Clearly H n T E NH(T) E H, so we need only prove that NH(T) E

T. Let p‘" be the highest power of 1) that divides the order g of G, and let g = pmq, so that p *q. By Theorem 26.10, Tis a normal subgroup of N(T), and NH(T) is also a subgroup of N(T). Hence, by Theorem 26.9, the groups NH(T)/(T n NH(T)) and NH(T)T/T are isomorphic. The order of the former divides the order of NH(T), a subgroup of H, and hence is p‘ for some S E [0, r]. The order of NH(T)T is a multiple of the order p“ of T and a divisor of the order pmq of G; hence, the order of NH(T)T is pmt where

17* t, and consequently, the order of NH(T)T/T is t. Since t = p’ and p *1, we conclude that t = 1. Thus, T n NH(T) = NH(T), so NH(T) E T.

Theorem 26.12. Let p be a prime, let G be a finite group of order g, let pm be the highest power of p that divides g, and let S be a Sylow p—subgroup of G.

1° (Sylow’s Second Theorem) If H is a p-subgroup of G, then H is contained in a conjugate subgroup of S. In particular, every Sylow p-subgroup of G is a conjugate subgroup of S. 2° (Sylow’s Third Theorem) The number n, of Sylow p-subgroups of G is a divisor of g/p'" and satisfies the congruence n, E 1 (mod p);

moreover, np=l+k1p+kgpz+...+kmp"‘ where for each r e [1, m], the number of Sylow p-subgroups T for which S n T has order p'”" is k,p'.

Proof. Let g = pmq, whence p4’q. For any subgroup H of G, (Conj(S), .)

is an H-space where . is defined by a.T = aTa‘1

for all a e H, T e Conj(S). Indeed, if T E Conj(S), clearly, aTa‘1 e Conj(S),

248

FINITE GROUPS

Chap. IV

and it is easy to verify that e.T = T, ab.T = a.(b.T) for all Te Conj(S) and all a, b e H. The orbit of T for this action is clearly ConjH(T), and the stabilizer of T is NH(T). Applying these remarks to the case where H = G, T = S, we conclude

from 2° of Theorem 26.1 that n(Conj(S)) = n(G/N(S)). Since S is a subgroup of N(S), the order of N(S) is a multiple of the order

p’" of S. Therefore, n(G/N(S)) is a divisor of q = g/p’", and in particular,

p Jr n(Conj(S)).

Now let H be a subgroup of order p’, and let Te Conj(S). By 2° of

Theorem 26.1 and by Theorem 26.11,

n(Conm) = n(H/NHa'» = n(H/(H n T», a divisor of 12'. Hence, if H S T, then n(ConjH(T)) = l as H n T = H;

but if H $ T, then n(ConjH(T)) :1)“ for some s 2 1 as H n TC H. By

1° of Theorem 26.1, the sets ConjH(T) where T e Conj(S) form a partition of Conj(S). Hence, if H were not contained in any conjugate subgroup T of S, Conj(S) would be the union of disjoint sets, the number of elements in

each of which being a strictly positive power of17, sop would divide n(Conj(8)), in contradiction to the result of the preceding paragraph. Thus, H E T for some T e Conj(S), and 1° is proved. In particular, Conj(S) is the set of all

Sylow p-subgroups of G. To prove 2°, we apply our remarks of the first paragraph of the proof to the case where H = S. Let T E Conj(S). Then

n(Conj,s(T)) = n(S/Ns(T)) = "(S/(S n T)) by 2° of Theorem 26.1 and by Theorem 26.11. Hence, S n T has 11"” elements if and only if n(Coa(T)) = 17'. For each r e [0, m], let k, be the number of orbits of the S-space Conj(S) that have 12’ members. Then n(Conj(S)) = k0 + klp + . . . + kmp’”. However, S n T has 12’" elements if and only if T = S, so n(Conj8(T)) = 1

if and only if T = S, and consequently, k0 = 1. As Conj(S) is the set of all Sylow p-subgroups of G, therefore, n,=1+k1p+k2p*+...+kmp’", and in particular,

n, E 1 (mod p).

Sec. 26

17mm: GROUPS

249

The set of all Sylow subgroups T such that S n T has 12“" members is thus the union of k, mutually disjoint sets, each containing 17’ members. Hence, there are k,p' Sylow p—subgroups T such that the order of S n T is pm". Our final application of Sylow’s theorems is to classify groups of order pq, where p and q are primes such that p < q. If G is a commutative group

of order pq, G contains (normal) subgroups of orders 1) and q by Sylow’s First Theorem and hence is cyclic by Corollary 24.8.1. Suppose, therefore, that there exists a noncommutative group G of order pq. Let n, be the number of Sylow q—subgroups of G.‘Because each has prime order q, the intersection of any two is a proper subgroup of each and

hence is {e}. Thus, each element of order q belongs to only one Sylow qsubgroup. Since each Sylow q-subgroup contains q — 1 elements of order q, G contains ,nq(q — 1) elements of order q. Suppose that n, > 1. Then n, 2 q + 1 by Sylow’s Third Theorem, so G would have at least (q + l)(q — l) = q“ — 1 elements of order q. The neutral element is not among them, so G would have at least q2 elements, in contradiction to the

fact that the order of G is pq where p < q. Hence, n4 = 1, so G contains only one Sylow q-subgroup S. The only conjugate subgroup of S is, therefore, S itself, so S is a normal subgroup of G, and G/S is a (cyclic) group of order p. Let a-be a generator of S, and let' b ¢ [a]. By Theorem 25.2, there exist s e Z* and r e N, such that b” = a’, bub—1l = a‘, s” E 1 (mod q), sr E r (mod q), and moreover, .9 5L5 1 (mod q), since G is not commutative. Denot-

ing by i the coset x + (q') in 2, corresponding to an integer x, we conclude that s” = l but s ¢ 1, so 5 is an element of order p in the multiplicative group Zg". Hence, p |q —— 1. Thus, if there is a noncommutative group of order pq, then q E 1 (mod p). Conversely, suppose that q E 1 (mod p). Then p | q — 1, so by Sylow’s

First Theorem, applied to the multiplicative group Zf, there exists s e Z* such that s’ E 1 (mod q), s ¢ 1 (mod q). Consequently, G(q, p, 0, s), for example, is a nonabelian group of order pq. Thus, we have proved the

following theorem:

'

Theorem 26.13. Let p and q be primes such that p < q. If q 5.": 1 (mod p),

then a group of order pq is cyclic. If q E 1 (mod p), an abelian group of order pq is cyclic, but there exist nonabelian groups of order pq. In §38 we shall prove that the multiplicative group F* of a finite field F is cyclic. This information enables us to classify completely all groups of order pq. Once again, assume that p I q — 1. By Theorem 23.9, the group Z2‘ contains exactly one subgroup of order p. As we have seen, G(q, p, 0, s)

is a nonabelian group of order pq, where s” E 1 (mod q), s gé 1 (mod q). To show that, to within isomorphism, this is the only nonabelian group of

Sec. 26

FINITE GROUPS

249

The set of all Sylow subgroups T such that S n T has pm" members is thus the union of k, mutually disjoint sets, each containing p' members. Hence, there are k,p' Sylow p-subgroups T such that the order of S n T is pm".

Our final application of Sylow’s theorems is to classify groups of order pq, where p and q are primes such that p < q. If G is a commutative group of order pq, G contains (normal) subgroups of orders p and q by Sylow’s First Theorem and hence is cyclic by Corollary 24.8.1. Suppose, therefore, that there exists a noncommutative group G of order pq. Let n, be the number of Sylow q—subgroups of G.' Because each has prime order q, the intersection of any two is a proper subgroup of each and hence is {e}. Thus, each element of order q belongs to only one Sylow qsubgroup. Since each Sylow q-subgroup contains q — 1 elements of order q,

G contains Inq(q — 1) elements of order q. Suppose that n, > 1. Then 71, 2 q + 1 by Sylow’s Third Theorem, so G would have at least

(q + l)(q — l) = q2 — 1 elements of order q. The neutral element is not among them, so G would have at least qa elements, in contradiction to the fact that the order of G is pq where p < q. Hence, n, = 1, so G contains only one Sylow q—subgroup S. The only conjugate subgroup of S is, therefore, S itself, so S is a normal subgroup of G, and G/S is a (cyclic) group of order 1). Let a-be a generator of S, and let' b 9.5 [a]. By Theorem 25.2, there exist s E Z* and r EN, such that b” = a’, bab—1 = a', s” E 1 (modq), sr E r

(mod q), and moreover, s gé 1 (mod q), since G is not commutative. Denoting by J? the coset x + (q) in Z, corresponding to an integer x, we conclude that 5'” = 1 but 5 75 1, so 5' is an element of order p in the multiplicative

group Zf. Hence, p Iq — 1. Thus, if there is a noncommutative group of order pq, then q E 1 (mod p).

Conversely, suppose that q a 1 (mod p). Then p |q — I, so by Sylow’s First Theorem, applied to the multiplicative group 2;", there exists s e Z* such that s” E 1 (mod q), s gé 1 (mod q). Consequently, G(q, p, 0, s), for example, is a nonabelian group of order pq. Thus, we have proved the following theorem: ‘

Theorem 26.13. Let p and q be primes such that p < q. If q gé 1 (mod p), then a group of order pq is cyclic. If q E 1 (modp), an abelian group of order pq is cyclic, but there exist nonabelian groups of order pq.

In §38 we shall prove that the multiplicative group F* of a finite field F is cyclic. This information enables us to classify completely all groups of order pq. Once again, assume that p | q — 1. By Theorem 23.9, the group 2: contains exactly one subgroup of order p. As we have seen, G(q, p, 0, s)

is a nonabelian group of order pq, where s” E 1 (mod q), s aé 1 (mod q). To show that, to within isomorphism, this is the only nonabelian group of

250

FINITE GROUPS

Chap. IV

order pq, it suffices to consider only the groups G(q, p, r, t) where t” E 1 (mod q), 1‘ 7+. 1 (mod q), tr E r (mod q) by Theorem 25.2, for we have seen that a group of order pq contains a normal subgroup of order q. Since Z, is a field, if r $ 0 (mod q), then from tr 2 r (mod q) we conclude that

t E 1 (mod q), a contradiction; hence, r -=- 0 (mod q), whence r = 0, as r 6 Na. Since Z: has only one subgroup of orderp, both 5 and fare generators of that subgroup, so 5 = t" for some ie [1, q — 1], whence s E t‘ (mod q). Let a, b be generators of G(q, p, 0, t) such that a“ = b" = e, bab—1 = a‘,

and let bl = b‘. Then b1 is also a generator of [b] as [b] has prime order p,

and b',ab;1 = a“ = a’ by Theorem 25.2, so G(q, p, 0, t) is isomorphic to G(q, p, 0, s) by the final assertion of that theorem. In sum, using the fact

that the multiplicative group of nonzero elements of a finite field is cyclic, we have proved the following theorem: Theorem 26.14. If p and q are primes such that p < q and q E 1 (mod p), then a group of order pq is either cyclic or isomorphic to G(q, p, 0, s) where

s e Z* is such that s” 51(modq), s .=,é 1(modq).

EXERCISES 26.1. (a) Prove Theorem 26.2.

(b) Prove that a |—>Ka is a homomorphism from G into Aut(G) whose kernel is the center Z(G) of G.

*26.2. Let p be a prime. For every m e N*, every subgroup of order pm‘l...:iof a group of order p" is a normal subgroup. [Proceed by induction on m by considering G/[a], where a is an element of order p in Z(G); use Exercise 20.12.]

26.3. Verify the assertion made at the beginning of the proof of Theorem 26.7. 26.4. Prove Theorem 26.10.

26.5. Show that the group of symmetries of a cube is isomorphic to 64. [Number the diagonals of the cube.]

26.6. (a) If a group of order 104 contains no normal subgroup of order 8, how many subgroups of order 8 does it contain?

(b) If a group of order 80 contains no normal subgroup of order 5, how many elements. of order 5 does it contain? 26.7. If the Sylow subgroups of a finite group G are all normal, then G is their direct product. 26.8. For each n e [1, 15], list to within isomorphism all the groups of order n.

Sec. 26

FINITE GROUPS

251

*26.9. Let q be a prime > 3, and let G be a group of order 4q. (a) G contains a normal subgroup S of order g.

(b) If G contains a cyclic subgroup of order 2q, then G is isomorphic to one and only one of the following four groups: 2“, 22, x Zg, the dihedral group of order 4q, the dicyclic group of order 4q. [Use Theorem 25.2.] (c) If G contains no cyclic subgroup of order 2g, then q E 1 (mod 4), and G is isomorphic to G(q, 4, 0, s) where 54 E 1 (mod q), 52$ 1 (mod q). [First show that 6/8 is cyclic; argue as in the proof of Theorem 26.14.] ((1) If (1 = 5, which of the five groups mentioned in (b) and (c) is Aut(Ds)

(Exercise 25.7) isomorphic to? *26.10. (a) To within isomorphism there are two groups of order 847, both ofwhich are abelian. (b) State and prove a theorem concerning groups of orderpaq wherep and q are distinct primes that generalizes (a).

*26.11. (a) To within isomorphism there are four groups of order 1225, all of which are abelian. (b) State and prove a theorem concerning groups of order 11292 where p and q are distinct primes that generalizes (a).

CHAPTER V

VECTOR SPACES

An algebraic structure is by definition a set together with one or two compositions on that set. In this chapter we shall consider vector spaces and modules, which are important examples of a new kind of algebraic object, consisting of an algebraic structure E, a ring K, and a function from K x E into E.

27. Vector Spaces and Modules

Addition on the set R of real numbers induces on the plane R2 of analytic geometry a composition, also called addition, defined by (“1: as) + (’31, [32) = (“1 ‘1' I31, 0‘2 + fie)Thus, (R3, +) is simply the cartesian product of (R, +) and (R, +). Under addition, R2 is an abelian group. If («1, org), (8,, 132): and the origin (0, 0)

do not lie on the same straight line, the line segments joining the origin (0, 0) to (0:1, «2) and to ([31, [32) respectively are two adjacent sides of a parallelogram, and (0:1, as) + (’31, fig) and (0,0) are the endpoints of one of the

diagonals of that parallelogram (Figure 13). For every n e N*, by definition n.(at1, 0:2) is the sum of (oil, as) with itself n times; it is easy to see by induction that n.(oc1, a3) = (nag, nag), an equality that holds also if n S 0. Consequently, if we define A'(a1! 0‘2) = (1%, 1012)

for every real number 1, we have generalized for this particular group the notion of adding an element to itself n times. If («1, «2) 9E (0,0), then

252

Sec. 27

VECTOR SPACES

(£1.32) ‘~~

‘~“ ~

~‘~

~~~

253

\(anaz) + (£31.32) \

Figurel3

{1.011, as): 2. e R} is simply the set of all points on the line through («1, a2) and the origin (Figure 14). The directed line segment from the origin to 1.(oc1, a2) is Ill times as long as that from the origin to (al, a2) and has the same direction as the latter if A > 0 but the opposite direction if A < 0. The function (A, x) —> 1.x from R X R2 into R2 is denoted simply by . and is called “scalar multiplication.” It is easy to verify that with addition and scalar multiplication so defined, (R3, +, .) is a vector space over the field

of real numbers: Definition. Let K be a division ring. A vector space over K or a K—vector space is an ordered triple (E, +, .) such that (E, +) is an abelian group and .

is a function from K X E into E satisfying

(vs 1) (vs 2) (vs 3) (vs 4)

1.0: + y) = 2.x + ly

(i. + ,u).x =_/1.x + ,u.x (ly).x = up”) 1.x = x

for all x, y e E and all it, p e K. Elements of E are called vectors, elements

of K are called scalars, and . is called scalar multiplication.

1 fi(ahaz)

-§(a1,az) Figure 14

254

VECTOR SPACES

Chap. V

Similarly, (11", +, .) is a vector space over R for each n EN“ where

addition is defined by (“1:'°':an)+(fi1!"'!fln)=(“1+fi19°'°9an+fin)

and scalar multiplication by 1.011, . . . , or”) = (10:1, . . . , 2.0:”).

For the case n = 3, addition and scalar multiplication may be described geometrically exactly as before. In physics a vector is sometimes described as an entity having magnitude and direction (such as a force acting on a point, a velocity, or an acceleration)

and is represented by a directed line segment or an arrow lying either in the coordinate plane of analytic geometry or in space. In this description, two arrows represent the same vector if they are parallel and similarly directed and if they have the same length. Consequently, each vector is represented uniquely by an arrow emanating from the origin. Geometric definitions of

the sum of two vectors and the product of a real number and a vector are given in such descriptions in terms of arrows and parallelograms, but it is apparent that what essentially is being described in geometric language is the vector space (R‘, +, .) or (R3, +, .), depending on whether the arrows

representing the vectors are considered to lie all in a plane or in space. An important generalization of vector spaces is obtained by relaxing the requirement that K be a division ring. Definition. Let K be a ring with identity. A module over K or a K-module is an ordered triple (E, +, .) such that (E, +) is an abelian group and . is a function from K X E into E satisfying conditions (VS 1), (VS 2), WS 3), and (VS 4) for all x, y e E and all 1, [4 e K. Elements of K are called scalars, and K itself is called the scalar ring. Thus, a vector space is a module whose scalar ring is a division ring. Our primary interest is in vector spaces, but we shall prove theorems for modules when specializing to vector spaces would yield no simplifications. Except in the unusual case where E and K are the same set, scalar multiplication is not a composition on a set, and hence a module is not an algebraic

structure as we have defined the term in §6. To provide a suitable framework for the definition of “isomorphism” we make the following definition: Definition. Let K be a ring. A K-algebraic structure with one composition is an ordered triple (E, +, .) where (E, +) is an algebraic structure with one composition and where . is a function (called scalar multiplication) from

Sec. 27

VECTOR SPACES

255

K x E into E. A K-algebraic structure with two compositions is an ordered quadruple (E, +, -, .) where (E, +, -) is an algebraic structure with two com-

positions and where . is a function from K x E into E. A K-algebraic structure is simply a K-algebraic structure with either one or two compositions. Our definition is quite artificial in allowing only binary operations on E

rather than n-ary operations, in restricting their number to one or two, and in demanding that K be a ring, but it is sufliciently general for our purposes. The following definition formally expresses what is meant by saying that two K-algebraic structures are “just like” each other: Definition. If (E, +, .) and (F, (19, .) are K—algebraic structures with one composition, a bijection f from E onto F is an isomorphism from (E, +, .) onto (F, (B, .) if

f(x + y) =f(x) em) fol-x) = me) for all x,y 6E and all A 6K. If (E, +, -, .) and (F, e, O, .) are Kalgebraic structures with two compositions, a bijection f from E onto F is an isomorphism from (E, +, -, .) onto (F, 6), O, .) if the above two conditions hold and if, in addition,

f(x-y) =f(x) OfO’) for all x, y e E. An automorphism of a K-algebraic structure is an isomor-

phism from itself onto itself. If there exists an isomorphism from one K-algebraic structure onto another, we shall say that they are isomorphic. The analogue of Theorem 6.1 for K-algebraic structures is valid and easily proved: Theorem 27.1. Let K be a ring, and let E, F, and G be K-algebraic structures

with the same number of compositions.

1° The identity function 1,, is an automorphism of the K-algebraic structure E. 2° If f is a bijection from E onto F, then f is an isomorphism from the

K-algebraic structure E onto the K-algebraic structure F if and only iff“ is an isomorphism from F onto E.

3° Iff and g are isomorphisms from E onto F and from F onto G respectively, then g of is an isomorphism from E onto G.

Sec. 27

VECTOR SPACES

255

K x E into E. A K-algebraic structure with two compositions is an ordered quadruple (E, +, -, .) where (E, +, -) is an algebraic structure with two com-

positions and where . is a function from K x E into E. A K-algehraic structure is simply a K-algebraic structure with either one or two compositions. Our definition is quite-artificial in allowing only binary operations on E rather than n-ary operations, in restricting their number to one or two, and in demanding that K be a ring, but it is sufliciently general for our purposes. The following definition formally expresses what is meant by saying that two K-algebraic structures are “just like” each other: Definition. If (E, +, .) and (F, ea, .) are K-algebraic structures with one composition, a bijection f from E onto F is an isomorphism from (E, +, .) onto (F, ea, .) if

f(x + y) =f(x) any) fax) = are) for all x,y 6E and all A 6K. If (E, +, -, .) and (F, (-D, O, .) are Kalgebraic structures with two compositions, a bijection f from E onto F is an isomorphism from (E, +, -, .) onto (F, (-9, O, .) if the above two conditions hold and if, in addition,

f(w) =f(x) Of(y) for all x, y e E. An automorphism of a K-algebraic structure is an isomorphism from itself onto itself. If there exists an isomorphism from one K-algebraic structure onto another, we shall say that they are isomorphic. The analogue of Theorem 6.1 for K-algebraic structures is valid and easily proved: Theorem 27.1. Let K be a ring, and let E, F, and G be K-algebraic structures

with the same number of compositions.

1° The identity function IE is an automorphism of the K-algebraic structure E. 2° If f is a bijection from E onto F, then f is an isomorphism from the K-algebraic structure E onto the K-algebraic structure F if and only iff‘— is an isomorphism from F onto E. 3° Iff and g are isomorphisms from E onto F and from F onto G respectively, then g o f is an isomorphism from E onto G.

256

VECTOR SPACES

Chap. V

It is easy to verify also that a K-algebraic structure isomorphic to a Kmodule is itselfa K-module. In particular, if K is a division ring, a K-algebraic

structure isomorphic to a K-vector space is itself a K-vector space. If (E, +, .) is a vector space over K, it is customary to speak of “the vector space E” when (E, +, .) is meant, “the additive group E” when

(E, +) is meant. A similar convention applies to modules. In any discussion concerning a K-module E, the symbol “0” has two

possible meanings: It denotes the zero element of the scalar ring K and also the zero element of the additive group E. In any context it will be clear which is meant. In both (1) and (8) below, for example, the first and third occurrences of “0” denote the zero element of E, but the second occurrence

denotes the zero element of K. In the notation for scalar multiplication, the dot in “11.x” is usually omitted, and thus, “1.x” often denotes the scalar

product of l and x. Theorem 27.2. Let E be a K-module. If x e E, A e K, and n e Z, then

(1)

10 = 0x = 0

(2)

l(-x) = (—1)x = —.U~X)

(3)

l(n.x) = n.(J.x) = (n.}.)x

(4)

H»: = —x

(5)

n.x = (n.1)x.

If (x9199, is a sequence of elements of E and if (1,, 151:5,» is a sequence of scalars, then

(6)

1(élxk) =21“,

(7)

(72117“) x =21)”;

If K is a division ring, then for every vector x and every scalar 2., (8)

iflx = 0, then either A = 0 or x = 0.

Proof. By (VS 1), yr—niy is an endomorphism of the additive group E. Hence, 10 = 0 and Z(—x) = —(}.x) by Theorem 20.5, (6) holds by induction, and l(n.x) = n.(}.x) by the Corollary of Theorem 20.5. By (VS 2), A v—Mx is a homomorphism from the additive group K into the additive group E. Hence, 0x = 0 and (—A)x = —(}.x) by Theorem 20.5 (in particular,

(— l)x = —x by (VS 4)), (7) holds by induction, and (n.}.)x = n.().x) by

Sec. 27

vrcron SPACES

257

the Corollary of Theorem 20.5 (in particular, (n.l)x = n.x by (VS 4)). Finally, if K is a division ring and if 1x = 0 but A ¢ 0, then by (l),

o = 1-1.0 = 14(1):) = (Hm = 1.x = x. Example 27.1. The definition of the R-vector space R” given earlier may be generalized by replacing R with any ring with identity K: For every n e N*,' addition on K" and scalar multiplication are defined by

(“l""aan)+(‘91:"'-afin)=(°‘1+fll!"'sau+fln)

2.011, . . . , at")=(loc1, . . . ,Zoin). It is easy to verify that (K", +, .) is indeed a K-module, and when we

refer to the K-module K", we shall always have the K-module (K", +, .) just defined in mind.

Example 27.2. If L is a ring with identity and if K is a subring of L containing the identity element 1, then (L, +, .) is a K-module where

+ is the additive composition of L and where scalar multiplication is the restriction to K x L of the multiplicative composition of L. Whenever L is a ring with identity and K is a subring of L containing the identity element 1, by the K—module L we shall always have the K-module (L, +, .) just defined in mind. In particular, if K is a division subring of a ring L and if the identity element of K is that of L, we may regard L as a K-vector space. With this definition of scalar multiplication, for example, both R and C are R-vector spaces, and any division ring is a vector space over its prime subfield. Example 27.3. If L is a ring with identity, if K is a subring of L containing the identity element 1, and if (E, +, .) is an L-module, then (E, +, .K) is a K-module where .K is the restriction of . to K x E. The K-module (E, +, .K) is called the K-module obtained from (E, +, .) by restricting scalar multiplication. In particular, if K

is a division subring of a division ring L and if E is an L-vector space, E may also be regarded in this way as a K-vector space. Example

27.2 is the special case of this example where E is the L-module L. Example 27.4. Let F be a K-module, E a set. Then (FE, +, .) is a K-module where + is defined by

(f + 906) =f(X) + 300

258

VECTOR SPACES

Chap. V

for all x e E, and where . is defined by

(lX) = 1f(x) for all x E E. Indeed, as noted in the discussion of Example 15.5, + on FE is associative, the neutral element for + is the function x |--) 0,

and the additive inverse of f 6 FE is the function —f: x I—> —f(x). The verification of (VS l)—(VS 4) is also easy. The most important

case of this example is that where F is the K-module K. Whenever we refer to the K-module FE of all functions from E into F, we shall

have the K-module (FE, +, .) just defined in mind. Example 27.5. If E1, . . . , E” are K-modules and if (E, +) is the cartesian product of the additive groups E1, . . . , E", then (E, +, .)

is a K-module where scalar multiplication is defined by l.(x1, . . . , x") = (1.x1, . . . , 1.x”).

Example 27.1 is the special case of this example where each E,‘ is the K-module K. Example 27.6. If (E, +) is an abelian group, then (E, +, .) is a Z-module where . is the function from Z X E into E defined in §ll and §l4. The Z-module (E, +, .) is called the Z-module associated with (E, +).

EXERCISES 27.1. Draw a diagram illustrating (VS 1) for the R-vector space R3 similar to that illustrating vector addition on R”. 27.2. Prove Theorem 27.1 and the assertion following.

27.3. Verify that (VS 1)—(VS 4) hold in Examples 27.1—27.5. 27.4. Which of (VS l)—(VS 4) are satisfied if scalar multiplication is the function

from C x C into C defined by z.w = |z| w? by z.w = 3(z)w? by z.w '-—- 0? Which are satisfied if scalar multiplication is the function from C x C2 into C2 defined by (zu, 21)) if v 95 0

"0” v) = [(zu, 0) if v = o? Infer that no one of (VS 1)—(VS 4) is implied by the other three.

Sec. 27

VECTOR SPACES

259

27.5. If E is a K—vector space containing more than one element, then for each

1. EK*, the function XHM is an automorphism of E if and only if 2. belongs to the center of K. 27.6. An isomorphism from an abelian group E onto an abelian group F is also

an isomorphism from the associated Z-module E onto the associated Z-module F. 27.7.

What are 6 and . if f is an isomorphism from the R-vector space R’ onto the R-vector space (R2, GB, .), where f is defined by f(x,y) =(x+y—l,x—y+l)?

wheref is defined by

f(x,y) =

(x2 —y2,x —y)ifx #y

.

_

(x,0)1fx —y?

27.8. If E is a K-module, then {1 SK: Ax = 0 for all xEE} is an ideal of K, called the annihilator of E. 27.9. Let E be a module over an integral domain K. What are the possible

orders of a nonzero element of the abelian group E if the characteristic of K is a prime p? if the characteristic of K is zero ? What are the possible orders if K is a field whose characteristic is zero? 27.10. Let (E, +) be an abelian group. (a) There is exactly one function . from Z x E into E such that (E, +, .)

is a Z-module. (b) If the order of every element of the abelian group E is finite and divides n, there is exactly one function . from 2,, x E into E such that (E, +, .) is a Zn-module.

(c) In particular, if p is a prime and if every nonzero element of E has orderp, there is a unique function . from Z9 x E into E such that (E, + , .)

is a vector space over 2,. *27.11. If (E, +, .) is a Q-algebraic structure satisfying (VS 2), (VS 3), and (VS 4) for all A e Q and all x e Eand if (E,-+) is an abelian group, then (E, +, .)

is a Q-vector space. *27.l2.

An abelian group (E, +) is divisible if n.E = E for every n E Z*; that is,

if for each y e E and each n e Z* there exists x e E such that n.x = y. If (E, +) is an abelian group, then there is a function . from Q x E into E

such that (E, +, .) is a Q-vector space if and only if (E, +) is a divisible

group all of whose nonzero elements have infinite order. *27.13. Let (E, +, .) be a K-module. The module E is torsion-free if for all I. E K and all e, if Ax = 0, then either A = 0 or x = 0. The module E is a

divisible K-module if 1E = E for all AEK”; that is, if for each yeE and each A E K* there exists x EE such that Ax = y. If K is an integral

VECTOR SPACES

260

Chap. V

domain and if L is a quotient field of K, there is a function . from L x E into E such that (E, +, .) is an L-vector space and (E, +, .) is the Kmodule obtained from (E, +, .) by restricting scalar multiplication if and only if (E, +, .) is a divisible torsion-free module. Infer the statement of

Exercise 27.12 as a special case.

28. Subspace: and Bases If E is a K—algebraic structure, a subset M of E is said to be stable for scalar multiplication if 1.x 6 M for all A e K and all x E M, and M is called a stable subset of Eif M is stable for scalar multiplication and the composition(s) of

E. Analogous to the definition of subgroup and subsemigroup is the following definition of vector subspace and submodule: Definition. Let K be a division ring (a ring with identity), let (E, +, .) be a K-algebraic structure with one composition, and let M be a stable subset of E. If (M, +M, .M) is a K-vector space (a K-module) where +M is the

restriction of addition to M x M and where .M is the restriction of scalar multiplication to K x M, we shall call (M, +M, .M) a (vector) subspace

(a submodule) of E. As for subgroups and subsemigroups, however, a stable subset M of E

is called a subspace (a submodule) of E if (M, +M, .M) is a subspace (a submodule) according to our formal definition. Theorem 28.1. If E is a K-module, then a nonempty subset M of E is a submodule if and only if for all x, y e M and all A e K, x + y and 1.x belong to M, that is, if and only if M is a stable subset of E.

Proof. If the condition is fulfilled, then —x = (—l)x e M if x EM, so M is a subgroup of (E, +) by Theorem 8.3 and consequently is a submodule. Example 28.1. If E is a K-module, then E and {0} are submodules of E.

Example 28.2. Let us find all subspaces of the R-vector space R“. Let M be a nonzero subspace of R”. Then M contains a nonzero vector («1, 0(2) and consequently also {Moth org): A e R}, a set that may be described geometrically as the line through (oil, as) and the origin. Suppose that M contains a vector (#1, [32) not on that line. Then

«11?, — eta/31 qé 0, for otherwise ([31, #3) would be £011, 0(2) where §= fillet1 or {= [32/12 according as oil 7S 0 or at, 3:6 0. But then

Sec. 28

vacroa SPACES

261

M = R3, for if (y1, y”) is any vector whatever, (71s 72) = ““1: “2) + #031: [32)

where 1 = 7152 — 7251

= “172 — “271

“1132 —‘ “2}31

“1’32 —‘ “2131

as we see by solving simultaneously the equations

“11 ‘1' I31.“ = 71 “21 + 192” = 72Thus, a subspace of R2 is either the whole plane R2, or a line through the origin, or {0}. Conversely, every line through the origin consists of all scalar multiples of any given nonzero vector on it and hence is easily seen to be a subspace. Example 28.3. If F is a K-module and if E is a set, the set F‘E’ of all functions f from E into F such thatf(x) = 0 for all but finitely many elements x of E is a submodule of the K-module FE.

Example 28.4. Let K be a commutative ring with identity. For each k EN, let pk be the function from K into K defined by pk(x) = x" for all x e K. A function p from K into K is a polynomial function on K if there exists a sequence (099095,. of elements of K such that 11

p = 2 «kphk=0

The set P(K) of all polynomial functions on K is a submodule of the K-module KK. If m is a given natural number, the set Pm(K) of all m—l

the polynomial functions 2 akPk where (“2)oSkSm—1 is any sequence 12-0

of m terms of K is a submodule of P(K).

Example 28.5. Let I = {x e R: a s x s b}. Familiar subspaces of

the R—vector space R" encountered in calculus are the space V0) of all real-valued continuous functions on J, the space 9(J) of all real-valued differentiable functions on J, the space ‘6‘”0) of all

real-valued functions on J having continuous derivatives of order m,

262

VECTOR SPACES

Chap. V

the space @(”)(J) of all real-valued functions on J having derivatives of all orders, and the space .920) of all Riemann-integrable functions on J. Theorem 28.2. Let E be a K-module. If M1, . . . , M” are submodules of E, then Ml + . . . + Mn and M1 n . . . n Mn are also submodules. The intersection of any set of submodules of E is a submodule; consequently, if S is a subset of E, the intersection of the set of all submodules of E

containing S is the smallest submodule of E containing S. The proof is easy. Theorem 28.2 enables us to make the following definition. Definition. If S is a subset of a K-module E, the submodule generated (or spanned) by S is the smallest submodule M of E containing S, and S is called

a set of generators for M. The module E is finitely generated if there is a finite set of generators for E. We shall next characterize the elements of the submodule generated by an arbitrary subset S of a K—module, but before doing so, we need two

preliminary definitions. Definition. Let 01,9199, be a sequence of elements of a K-module E. An element b of E is a linear combination of 91,915,,“ if there exists a sequence

(1,9199, of scalars such that 1|

b = 211.01:k=1

Definition. Let S be a subset of a K-module E. If S is not empty, an element b of E is a linear combination of S if b is a linear combination of some sequence (49195.» of elements of S; the zero element is the only element of E called a linear combination of the empty set. If (@9199, is a sequence of elements of a K-module E, then an element b of E is a linear combination of the sequence (ak)1_ 0. As we saw above, any subset of E containing zero is linearly dependent. Clearly also, any subset of a linearly independent set is linearly independent, and consequently, any set containing a linearly dependent set is linearly dependent. By (8) of Theorem 27.2, if a is a nonzero vector of a vector space, then {a} is a linearly independent set. If (ak)15,, 5,, is a sequence of distinct terms of a K-module E, then (@9199 is a linearly independent sequence if and only if {an . . . , an} is a linearly independent set. Indeed, if {an . . . ,an} is linearly independent,

clearly, (0:013:57. is linearly independent. Conversely, suppose that (@9199 is a linearly independent sequence, let @3195", be a sequence of

distinct terms of {an . . . , an}, and let (119195,“ be a sequence of scalars such that ib, = 0. For each k e [1, n], if a,‘ e{b1, . . . , bm}, let A, = a, where j is in; unique index such that ak = b,, and let 1,, = 0 if a,, ¢ {b1, . . . , bm}. Then clearly, 0 = 2 lhbr = 2 lkak’ 1-1

k=1

so 1,, =‘ 0 for all k e [1, n], whence a, = 0 for allje [1, m], since {(41, . . . , rum} E {119 - - ' a an}

Consequently, if (@9195, is a linearly independent sequence and if (b91519: is a sequence of distinct terms such that {by . . . , b,,,} S {an . . . , an}, then (1791539: is also a linearly independent sequence. Definition. Let E be a K-module. A basis of E is a linearly independent set of generators for E. The module E is free if there exists a basis of E. As we shall shortly see, some modules have infinite bases, others have finite bases, and still others have no bases at all. If a module has a finite basis, it is often convenient to have the elements of that basis arranged in a definite order, and therefore we make the following definition.

Definition. An ordered basis of a K-module E is a linearly independent sequence (@195, of elements of E such that {an . . . , a,,} is a set of generators for E. Thus, by Theorem B.8, each basis of n elements determines n! ordered bases.

VECTOR SPACES

Sec. 28

265

Example 28.6. Let K be a ring with identity, let n be a strictly positive integer, and for eachj e [1, n] let e, be the ordered n-tuple of elements of K whose jth entry is l and all of whose other entries are 0. Then

(@9199 is an ordered basis of the K-module K", since

filtefiwofiw-,0)+(0,z.,o,...,0)+...+(o,o,o,...,z,.)

k-l

= (11112, 3.3, . . . ,2”).

This ordered basis is called the standard ordered basis of K”, and the

corresponding set {e1, . . . , en} is called the standard basis of K". Example 28.7. The set B of all the functions 17,, where n EN is a basis of the R-vector space P(R) of all polynomial functions on R (Example 28.4). By definition, every polynomial function is a linear combination of B. If i akpk = 0 where mm vs 0, we would obtain k=0

by differentiating m times mlocm=0, whence am = 0, a contradiction. Hence, B is linearly independent and therefore is a basis of P(R). Example 28.8. Let K be a ring with identity, let A be a set, and for each a E A let fa be the function from A into K defined by lfix=a fa(x) = 0 ifx 7'5 a.

Then B = {faz a EA} is a basis of KW (Example 28.3); indeed, if (0191n,. is a sequence of distinct terms of A and if (1,, 195,, is a sequence of scalars, then 2 1k is the function whose value at a,,

is 1,, and whose value at any=x not in {an . . . , an} 1s 0; consequently,

B is a linearly independent set of generators of K“). Theorem 28.4. A sequence (11,9195, of elements of a K-module E is an ordered basis of E if and only if for every x e E there is one and only one sequence (1,91 9,5,, of scalars such that x = i Aka” k=1

Chap. V

vacron SPACES

266

Proof. Necessity: Every element of E is a linear combination of the set {an . . . , an} of generators for E by Theorem 28.3. If 1|

1|

2 lira]: = Ema»

k=1

7.1-1

then 7‘



.

‘7}

n

2“]: - #001: = 2(1):“): _ Mkak) = 2 Akak — Zl‘kalc = 0,

k=1

k=1

Ic=l

le=1

so 1,, = pa, for all k e [1, n], since (ak)1 3,, 5,, is alinearly independent sequence. Sufficiency: Clearly {a,, . . . , an} generates E. If

i Aka“,

.= 0,

76:1

then since also 2 O-ak = 0, Its-1

by hypothesis, we have 2,, = 0 for all k e [1, n]. Therefore, (@9195, is a

linearly independent sequence. If (@915kis an ordered basis of a K-module E and if x— — 21,4” the scalars ll, . .

,1 are sometimes called the coordinates of x relative to

the ordered basis (ak)1§,, g”, and (a,,)1 gkén itself is called a coordinate system. The geometric reason for this terminology is that if (a1, a2) is an ordered basis of the plane R2, for example, and if the distance from the origin to a1 (to a,) is taken as the “unit” of length on the line L1 (the line L2) through the origin and a1 ((12), then the coordinates 21 and 12 of x = 11a, + 12412 locate the vector x as the intersection of the line parallel to L2 and 11 units along L1

from it and the line parallel to L1 and 1.2 units along L, from it. Figure 15 illustrates this for the case where a1 = (2, —1), a2 = (—1}, l), and x = (3, '3‘) = 3‘01 + 402-

Theorem 28.5. If (4)15“, is an ordered basis of a K-module E, then 'P: (1791951: HZlkak 10-1

is an isomorphism from the K-module K” onto the K-module E.

267

VECTOR SPACES

Sec. 28

L2

301 ’02\\ ‘2

+4 02

\

L1 Figure 15

Proof. By Theorem 28.4, 1;: is bijective. Since '5

7‘



’3

21%,, + 2.“k = 20%“: ‘l' :ukak) = 2(1): + F1007:

k=1

Ie=1

7c=1

k=1

and since

5 2 lkak = 2 paladin) = 2(131»)am 13—1

75-1

13-1

1p is an isomorphism. Consequently, any two K-modules having bases of n elements are isomorphic, for they are both isomorphic to the K-module K”. The remaining theorems in this section concern only vector spaces, and their proofs depend heavily on the existence of multiplicative inverses of nonzero scalars. Theorem 28.6. A sequence (a915,,5,, of distinct nonzero vectors of a K-Vector space E is linearly dependent if and only if a, is a linear combination of

(@9193?1 for some p e [2, n].

Proof. Necessity: By hypothesis, the set of all integers r e [1, n] such that (@9195, is linearly dependent is not empty; let p be its smallest member. Then p 2 2 as a1 gé 0, and there exist scalars A], . . . , 1,, not all of which are zero, such that D

zlkak = 0. k—l

If A, were zero, then (“0139—1 would be linearly dependent since 2.1, . . . , 1,4 could not then all be zero, a contradiction of the definition of

Sec. 28

vscron SPACES

267

L2 ~

\

\

’02

\ \‘2011'402 2 L 01

\\ \ \

L1 Figure 15

Proof. By Theorem 28.4, 1/) is bijective. Since fl

1‘

'7]



211411: + ZMkak = 2(1):“): + :ukak) = 20% + #19“): k-l k=1 k—l k=1 and since

2(filk)ak, 13 2 has = 213017411) = k—l k=1

k=1

1/) is an isomorphism.

Consequently, any two K—modules having bases of n elements are isomorphic, for they are both isomorphic to the K-module K".

The remaining theorems in this section concern only vector spaces, and their proofs depend heavily on the existence of multiplicative inverses of nonzero scalars. Theorem 28.6. A sequence (“191951. of distinct nonzero vectors of a K-vector space E is linearly dependent if and only if a, is a linear combination of

(ah)15k5,_1 for some p e [2, n]. Proof. Necessity: By hypothesis, the set of all integers r e [1, n] such that (0191 9,5, is linearly dependent is not empty; let p be its smallest member. Then 1) 2 2 as a1 sé 0, and there exist scalars 11, . . . , 1,, not all of which

are zero, such that D 2 lkak = 0.

k=l

If A, were zero, then @9199?1 would be linearly dependent since 11, . . . , 1,4 could not then all be zero, a contradiction of the definition of

268

VECTOR SPACES

Chap. V

p. Hence 1, 75 0, so as 9—1 it’d, = — 2 aka,” k-1

we have p—l

a9 = g1(_1;11k)ak-

Sufliciency: If 9—1

at = Eflkak! k=1

then 211:0]: = 0

k=1

where ,1.,,=,u,b for all ke [1,p— 1], 1,: —l, and 1,,=0 for all ke [p + l, n]. Theorem 28.7. If L is a linearly independent subset of a finitely generated K-vector space E and if G is a finite set of generators for E that contains L, then there is a basis B of E such that L E B E G. Proof. Let 5’ be the set of all the subsets S of E such that S is a set of

generators for E and L s S s G. Since G e 5’, 5’ is not empty, and every member of 5’ is finite as G is. Let n be the smallest of those integers r for which there is a member of 5’ having r elements, and let B be a member of 5’ having n elements. Then 0 ¢ B, for otherwise B — {0} would be a set of generators for E having only n — 1 elements, and B — {0} would, therefore, belong to 5’ as L, being linearly independent, does not contain 0. Let m

be the number of elements in L, and let (“1915157. be a sequence of distinct vectors such that L = {an . . . , am} and B = {a1, . . . , an}. Suppose that B were linearly dependent. By Theorem 28.6, there would then existp e [2, n]

and scalars ,ul, . . . , ,u,_1 such that p—l a, = :flkak-

k=1

As L is linearly independent, 11 > m, and therefore, the set B’ = B — {a,} would contain L. Also, B’ would be a set of generators for E, for if x =

f Aka,” then Ic—l p—1

n

x = 20% + lmuk)ak + 2 1k“):k=l

k=9+1

Consequently, B’ would be a member of 5’ having only 11 — 1 members, a contradiction. Therefore, B is linearly independent and hence is a basis.

Sec. 28

VECTOR SPACES

269

Theorem 28.8. Every finitely generated vector space has a finite basis. Proof. We need only apply Theorem 28.7 to the case where G is a finite

set of generators and L is the empty set. There exist finitely generated modules over integral domains that have no basis at all (Exercise 28.10).

Theorem 28.9. If E is a K-vector space generated by a finite set ofp vectors,

then every linearly independent subset of E is finite and contains at most p vectors.

Proof. Let L be a finite linearly independent subset of E containing m vectors. We shall first prove that every finite set H of generators for E contains at least m vectors, whence in particular, p 2 m. For this proof we

shall proceed by induction on the number of vectors in the relative complementL—HoinL. Let S be the set of all natural numbers n such that for every finite set H

of generators for E, if the relative complement L — H has n vectors, then H contains at least m vectors. We shall prove that S = N. Clearly, 0 e S, for

if L — H = 0, then L S H, so H contains at least m vectors. Assume that n e S, and let H be a finite set of generators for E such that L — H has n + 1 vectors. Let a be a vector in L — H. Then (L n H) U {a} is a subset of L

and hence is linearly independent, so by Theorem 2&7 there is a basis B of E such that (LnH)U{a}§BSHU{a}. Then L — B = (L — H) — {a}, so L — B has n vectors. Consequently, as n E S, B has at least m Vectors. Since a is a linear combination of H but

a ¢ H, H U {a} is linearly dependent, and therefore, B is a proper subset of H U {a}. Thus, if b is the number of vectors in B and if h is the number of vectors in H, we have m0,m >0,andq22.

(a) How many homomorphism are there from the additive group 2," into 2""? (b) If K is a field of q elements and if E and F are respectively an ndimensional and an m-dimensional vector space over K, how many linear transformations are there from E into F? If n S m, how many of them are injective? [Use Exercise 28.23.]

29.13. HE and F are finite-dimensional vector spaces, for every linear transformation-u from E into F there exists a linear transformation 0 from F into E such that u o v o u = 14. [Use Theorem 29.7.] *29.l4. Let u and v be linear operators on a finite-dimensional vector space E.

(a) There exists a linear operator w on E such that u = v o w if and only if u*(E) S 0*(E).

(b) There exists a linear operator w on E such that u = w o v if and only if ker u 2 ker 1). 29.15. Let K be a ring with identity, let E and F be K-modules, and let u be a homomorphism from (E, +) into (F, +).

(a) Let L = {A e K: u(1x) = 2u(x) for all x e E}. Then L is a subring of K, and for every 2. e L, if 1 is invertible in K, then 1‘1 e L. (b) If K is a prime field, then u is a linear transformation from the K-vector

space E into the K-vector space F.

30. Direct Sums and Quotient Spaces

How can a vector space or module be described as put together in some natural way from certain of its subspaces or submodules ? This is the analogue of a question raised concerning groups in §24. There the method of “putting together” was that of forming cartesian products; a group was called the direct product (or direct sum, depending on the notation used for the composition) of certain of its subgroups if it was isomorphic in a natural way to their cartesian product. We have already observed in Example 27.5 that the cartesian product of K-modules may .be made into a K-module in a

natural way, and therefore, we make the following definition. Definition. Let (M9199 be a sequence of submodules of a K-module E.

We shall say that E is the direct sum of (Mk)1 5kg" if the function A: (xk) Hixk k-l

is an isomorphism from the K-module fi Mk onto the K-module E. k=1

Sec. 30

VECTOR SPACES

285

Let (M9196, be a sequence of submodules of a K-module E, and let A be the function of the preceding definition. Then A is a linear transformation from fi Mk into E whose range is the submodule M1 + . . . + M”, for k—l

A((x.) + o.» = A«x,. + y.))=§1(x. + y.) =23, +21y, = Awe.» + Am.» (Theorem A7) and

A(l(x;.)) = A((/1xk)) =§m = zéx = mm.» by Theorem 27.2.

Theorem 30.1. A K-module E is the direct sum of a sequence (M9199, of submodules if and only if the following two conditions hold: 1°

E=M1+...+-Mn

2° If ix» = 0 where xkeMk for each k 6 [1,111, then x, = 0 for all k=1

ke[1,n]. Proof. Clearly, A is surjective if and only if 1° holds. Therefore, by Theorem

20.9, A is an isomorphism if and only if 1° and 2° hold. Theorem 30.2. If M and N are submodules of a K-module E, then the

K-module M + N is the direct sum of M and N if and only if M n N = {0}. Proof. By Theorem 30.1, it suflices to show that M n N = {0} if and only if x + y = 0 implies that x = y'= 0, whenever x e M, y 6N. Necessity: Ifx+y=0wherexeM,yeN, thenx= —yeN, soxEM nN= {0},

whence x = 0 and thus also y = 0. Sufficiency: Given xeM n N, let y= —x;thenxeM,yeN,andx+y=0, sox=0.

Corollary. A K-module E is the direct sum of' submodules M and N if and

onlyifE=M+NandMnN= {0}. Example 30.1. Le’r (ek)1 9,5,, be a basis of a K-module E, and let Mk be the one-dimensional submodule Kek for each k e [1, n]. By the definition of a basis, (M9199 satisfies 1° and 2° of Theorem 30.1, and consequently, E is the direct sum of (Mk)1 5kg".

Sec. 30

VECTOR SPACES

285

Let (M9199 be a sequence of submodules of a K-module E, and let A

be the function of the preceding definition. Then A is a linear transformation from fi Mk into E whose range is the submodule M1 + . . . + M”, for k-l

A«x..) + o.» = A«x.. + y.» =76252:. + y.)

= ix. + in = A((x..» + Aw.» k=1

k=1

(Theorem A7) and 110(k)) = A((}~xk)) =k§11xk

= 22x. = Mac.» 70-1

by Theorem 27.2.

Theorem 30.1. A K-module E is the direct sum of a sequence (Mk 15%,, of submodules if and only if the following two conditions hold:

1° E=M1+...+IM,, 2° If fix, = o where x, 5M, for each k e [1, n], then x, = o for all Ic=1

k e [1, n].

Proof. Clearly, A is surjective if and only if 1° holds. Therefore, by Theorem 20.9, A is an isomorphism if and only if 1° and 2° hold. Theorem 30.2. If M and N are submodules of a K-module E, then the

K-module M + N is the direct sum of M and N if and only if M n N = {0}.

Proof. By Theorem 30.1, it suffices to show that M n N = {0} if and only if x + y = 0 implies that x = y' = 0, whenever x e M, y e N. Necessity: Ifx+y=0wherexeM,yeN,thenx= —yeN, soxeMnN= {0},

whence x = 0 and thus also y = 0. Sufficiency: Given x EM n N, let = —x;thenxeM,yeN,andx+y=0,sox= 0.

Corollary. A K-module E is the direct sum of submodules M and N if and

onlyifE=M+NandMnN= {0}. Example 30.1. Let (@9199, be a basis of a K-module E, and let Mk be the one-dimensional submodule Kek for each k e [1, n]. By the definition of a basis, (M9199, satisfies 1° and 2° of Theorem 30.1,

and consequently, E is the direct sum of (M9199.

286

VECTOR SPACES

Chap. V

Example 30.2. If H is an (n — 1)-dimensional subspace and M a one-

dimensional subspace of an n-dimensional vector space E and if M is not contained in H, then E is the direct sum of M and H. Indeed, as M is not contained in H, Hc M+H and Ma M, so

n — 1 < dim(M + H) and dim(M n H) < 1; therefore, dim(M + H) = n

and

dim(M mm = 0,

soM+H=EandMnH={0}. Definition. A supplement (or complement) of a submodule M of a K-module E is a submodule N of E such that E is the direct sum of M and N. A submodule M of E is a direct snmmand of E if there is a supplement of M in E. By the Corollary of Theorem 30.2, if N is a supplement of M, then M is a supplement of N; in this case, we shall say that M and N are supple-

mentary submodules. An important property of finite-dimensional vector spaces is that every subspace has a supplement: Theorem 30.3. A subspace M of a finite-dimensional vector space E is a

direct summand of E. Proof. Since E and {0} are supplementary subspaces, we may assume that M is a nonzero proper subspace. By Theorems 28.13 and 28.14, there is an ordered basis (@9199 of E such that @9195," is an ordered basis of

M for some m e [1, n — 1]. Let N be the subspace generated by {ak: m + 1 S

k s n}. Clearly, M + N = E and M n N = {0}, so M and N are supplementary subspaces. Theorem 30.4. If E and F are finite-dimensional K-vector spaces, then dim(E X F) = dim E + dim F. Proof: Let {b1, . . . , b,,,} be a basis of E, {c1, . . . , an} a basis of F. We shall

show that {(b1, 0), . . . , (bm, 0), (0, cl), . . . , (0, c,,)} is a basis of E X F, from which the assertion follows. If (x, y) e E x F, there exist 1.1, . . . , 1,", m

n

pl, . . . , [4,, e K such that x = 2 lib), y = 2mg, whence i=1

1-1

(x, y) = (x, 0) + (o, y) = (:1 1.1)., o) + (0,:1 #50,) = 2 jibe 0) + 2 M0, c,). i=1

i=1

Sec. 30

vncron SPACES

287

If

gm, 0) +240, c» = (o, 0), then

(fifi‘biém) = (o, 0), so ifigb‘ =0 and inc, = 0, whence [31 = . . . =13”: 0 and y1 = {-1 :=1 . . . = y” = 0. Thus, {(b1, 0), . . . , (bm, 0), (0, cl), . . . , (0, c”)} is a basis of E x F.

Corollary. If a finite-dimensional vector space E is the direct sum of subspaces M and N, then dim E = dim M + dim N.

Theorem 30.5. If M and N are subspaces of a finite-dimensional vector space E, then

.dim(M+N)+dim(MnN)=dimM+dimN. Proof. Clearly, H = {(z, 2) EM x N: 2 6M n N} is a subspace of the vector space M X N. Moreover, dim H = dim(M n N) since z_:—> (z, z) is clearly an isomorphism from the vector space M n N onto H. The function u: (x, y) I—> x — y from M X N into M + N is easily seen to be an epimor-

phism whose kernel is H. Thus, by the Corollary of Theorem 29.7 and Theorem 30.4,

dim(M + N) + dim(M n N) = dim(M + N) + dimH

= P0!) + 1(a) = dim(M X N) = dim M + dim N.

Let M be a submodule of a K-module E. In §18 we introduced a composition on E/M by defining (x + M) + (y + M) to be x +y + M. We may introduce a scalar multiplication on E/M by defining

(l)

l-(x+M)=/1x+M.

for all A 6K, x + M eE/M. To show that scalar multiplication is well-

defined, we must show that ifx + M = x’ + M, then 1x + M = Ax’ + M for all 26K. Butifx+M=x’+M, then x—x'eM, so lx—lx’=

2(x — x’) e M, whence 2.x + M = lx’ + M. Thus scalar multiplication is

Sec. 30

VECTOR SPACES

287

If

‘21 (31:00 0) +1231“), cg) = (0: 0), then

( '2 Abs 2 m) = (o, 0), i=1

i=1

SO imb, =0 and $i, =0, whence [31: . . . = flm=0 and y1 = i=1

l=1

' ' ' = 7n = 0' Thus,

{(bls 0)) - - - a (but, 0), (0s 61), - - - a (0’ cn)}

is

a

basis

of E X F. Corollary. If a finite-dimensional vector space E is the direct sum of subSPaces M and N, then dim E = dim M + dim N. Theorem 30.5. If M and N are subspaces of a finite-dimensional vector space E, then

.dim(M+N)+dim(MnN)=dimM+dimN. Proof. Clearly, H = {(z, 2) EM X N: 2 SM n N} is a subspace of the vector space M X N. Moreover, dim H = dim(M n N) since 21H (2, z) is

clearly an isomorphism from the vector space M n N onto H. The function u: (x, y);n—>1x — y from M X N into M + N is easily seen to be an epimorphism whose kernel is H. Thus, by the Corollary of Theorem 29.7 and

Theorem 30.4,

dim(M + N) + dim(M n N) = dim(M + N) + dim H

= 9(a) + 1’(u) = dim(M X N) = dim M + dim N. Let M be a submodule of a K-module E. In §18 we introduced a composi-

tion onE/Mbydefining (x+M)+(y+M)to bex+y+M. We may introduce a scalar multiplication on E/M by defining (1)

2.0: + M) = 2x + M.

for all 16K, x + M eE/M. To show that scalar multiplication is well-

defined, we must show that ifx + M = x’ + M, then Ax + M = Ax’ + M for all 16K. But ifx+M=x’+M, then x—x’eM, so lx—lx’= 1(x —— x’) e M, whence Ax + M = Ax’ + M. Thus scalar multiplication is

288

VECTOR SPACES

Chap. V

well-defined by (1). It is easy to verify that (E/M, +, .) is a K-module.

Indeed,

LN"°.

Proof. Let (@9199 be an ordered basis of E such that (59195", is an

ordered basis of M, and let @9133” be the ordered dual basis of E*, If

t’ e M°” and if r' = i hag, then for allj e [1, m], '

k—l

A} = 23140;, (11;) =

k=1

(1,, 21ka£> = (“5, t’) = 0,

k=1

Sec. 31

VECTOR SPACES

295

ordered basis @9195" of E such that (a955,. is the ordered basis of E* dual to @9195”.

Proof. Let @9136” be the ordered basis of E** dual to (“Disksw and let a), = J“(a;) for all k e [1, n]. Then (ah)1 9,5,, is an ordered basis of E by Theorems 31.4 and 29.5, and

(at, 0}) = J(a.-)(a§) = aé'(a§) = 5w for all i, j e [1, n], so (“191515515 is the ordered basis of E* dual to ((10199.

Definition. Let E be a module over a commutative ring with identity K. For each submodule M of E, the annihilator of M is the subset M°* of E* defined

by M0+ = {t’eE*: (x, t’) = 0 for all x e M}, and for each submodule N of E*, the annihilator of N is the subset N"° of E defined by

N“° = {x e E: (x, t’) = 0 for all t’ EN}.

It is easy to verify that M °* is a submodule of E* and that N“° is a submodule of E.

Theorem 31.6. Let E be an n-dimensional vector space over a field K. If M is an m-dimensional subspace of E, then M -°" is an (n — m)-dimensional

subspace of E*, and M °""° = M. If N is a p—dimensional subspace of E*,

then N“° is an (n — p)-dimensional subspace of E, and N"°°" = N. The function M.I—>M0+

is a bijection from the set of all m-dimensional subspaces of E onto the set of all (n — m)-dimensional subspaces of E*, and its inverse is the function NHN“°.

Proof. Let (@9199 be an ordered basis of E such that (ak)1 5kg," is an

ordered basis of M, and let (“191351; be the ordered dual basis of E*. If

t’ e M“ and if t' = i 2%, then for allj e [1, m], I

Ic=1

21" = 2114“” 07") = = (a5, ti) = 0,

i=1

k=1

296

VECTOR SPACES

Chap. V

so t’ is a linear combination of {012: m + 1 g k S n}. But clearly, a; e M°" for each k e [m + 1, n]. Therefore, M °’ has dimension n — m. Similarly, let (4)195” be an ordered basis of E* such that (43195,, is an ordered

basis of N; by Theorem 31.5, (a,',)1 sin 5,, is the ordered basis of E* dual to an ordered basis (@9195, of E, and an argument similar to the preceding shows that N"° has dimension n — p. Clearly, M S M°"‘°. As dim M°""° = n — (n — m) = m, therefore,

M = MN. Similarly, N*°°" = N. The final assertion now follows from Theorem 5.3.

Theorem 31.7. Let K be a field, and let M be a subspace of the n-dimensional vector space K”. The following statements are equivalent: 1° dim M = n — 1. 2° M is the kernel of a nonzero linear form. 3° There exists a sequence (099195,, of scalars not all of which are zero such that M: {0.1“ . . ,ln)eK":a111 + . .. +a,,l,,=0}. Further, if l°— ° hold and if (139195,, is a sequence of scalars such that

M={(ll,...,).,,)eK":fl1}.1 +...+fi,./1,.=0}, then there is a nonzero scalar y such that fl): = V“): for all k e [1, n].

Proof. To show that 1° implies 2°, let N = M°". By Theorem 31.6, N is one-dimensional, and M = N“°-. Let u be a nonzero member of N. Then N is the set of all scalar multiples of u, so

M=N"°={e:(x,lu)=0forallleK}

={e:(x,u)=0}=keru.

By the corollary of Theorem 29.7, 2° implies 1°. Also, 2° and 3° are equivalent, for if («9155” is any sequence of scalars and if u = i “be; where (901515» is the ordered basis of (K”)* dual to the #1

standard ordered basis of K“, a simple calculation shows that

keru={(}.1,...,l,):a111 + ...+ a,1,,=o},

Sec. 31

vncron SPACES

297

and that u 3:5 0 if and only if oh, 75 0 for some k e [1, n]. If also M is the

kernel of v = i fikefi, then 0 e M°* , and so a = ya for some nonzero scalar 15-1

7, since M " is one-dimensional and since 0 7E 0, and therefore fik = yak

for all k €- [1, n]. A hyperplane of a vector space is a coset of a subspace of codimension 1.

Theorem 31.7 allows us-to characterize hyperplanes of the K-vector space X”, where K is a field. Indeed, let H = M + x0, where codim M = l, and let x0 = (,ul, . . . , ,ufl). By Theorem 31.7, there exist «1, . . . , atfl EK, not

all zero, such that M is the set of all (2.1, . . . , h”) e K" satisfying (1)

“111 + e e e + all?! = 0.

Let p = act/h + . . . + any". We shall show that His the set ofall (:1, . . ., z") e K" such that (2)

“1:1 + e e e + “fig” = p.

Indeed, if (2) holds and if 1,, = Q, —- m, for each k e [1, n], then «1111+...+otnhn=ot1§1+...+ocnzn—(0t1/l1+...+anpn) =fi—fl=oa

so ($1,...,§,,)=(1.1,...,h,,)+(#1,...,a,,)eM+xo; conversely, if (Al"-'9ln)€Mandif(€12---n€n)=(Als---’ln)+(”1""5”n)9then

a1§1+...+an§n=0tlhl+...+oc,,h,,+ac1/,t1+...+at,,p,,

=0+p=fl 0n the other hand, let «I, . . . , an be scalars, not all zero, and let ,3 e K. Let M be the set of all (11, . . . , A") GK” satisfying (1). By Theorem 31.7, M is a subspace of K” of codimension 1. If x0 = (,ul, . . . , ,u”) is a vector such that

a1m+..-+w..=fi (if a; are 0, (0, . . . , 0, fi/ot‘, 0, . . . , 0) is such a vector), then by the above argument M + x0 is the set of all (:1, . . . , in) e K" satisfying (2). In sum, a subset H of K" is a hyperplane if and only if there exist a scalar ,6 and a sequence (“9199. of scalars, not all of which are zero, such that H is the set of all (11, . . . , 1,36 K'l satisfying “111+...+dnln=fl,

298

vncron SPACES

Chap. V

in which case H is a coset of the (n — l)-dimensional subspace ofall (1,, . . . , 2,.) e K” satisfying «111+...+0tnh,n=0.

In plane analytic geometry, a line L is either defined or derived to be the set of all (x1, x3) 6 R2 satisfying

(3)

“1351 + 013x, = .8

where oil, as, and ,8 are given real numbers and not both 01.1 and ac, are zero,

and a line L’ is shown to be parallel to L if and only if for some [3’ e R the line L’ is the set of all (x1, x2) 6 R” satisfying “1x1 + “23‘: = fll'

In solid analytic geometry, a plane P is either defined or derived to be the set of all (x1, x2, x3) 6 R3 satisfying

(4)

11x1 + “2x: + “3x3 = 7

where «1, «a, as, and y are given real numbers and not all of 0:1, as, at, are

zero, and a plane P’ is shown to be parallel to P if and only if for some y’ e R the plane P’ is the set of all (x1, x,, x,) e R3 satisfying “1351 + “2952 + “3x3 = 9"-

(Aline or plane is considered parallel to itself, so [3' may be [3 and 31’ may be 9).) A line or plane is called homogeneous if it contains the origin, which is

the zero vector. Clearly, the line of R2 defined by (3) is homogeneous if and only if f] = 0, and the plane of R3 defined by (4) is homogeneous if and only if y = 0. Consequently, we see from the preceding that the lines of plane analytic geometry are precisely the cosets of one-dimensional subspaces of R“, i.e., the hyperplanes of R3, and that two lines are parallel if and only if they

are cosets of the same subspace. Similarly, the planes of solid analytic geometry are precisely the cosets of two-dimensional subspaces of K”, Le, the hyperplanes of R3, and two planes are parallel if and only if they are cosets of the same subspace. In solid analytic geometry, a line is either defined or derived to be the

intersection of two nonparallel planes. Let M and N be two distinct twodimensional subspaces of R3, and let L be the intersection of the cosets P = M + x0 and Q = N + yo of M and N respectively. Then since M 7E N, the subspace M + N properly contains the two-dimensional subspace M,

Sec. 31

vscron SPACES

299

and hence, M + N = R3; therefore, there exist m e M and n e N such that

xo—yo=m+n. Letzo=xo—m. Then P=M+xo=M+m+zo=M+zo, and alsoaszo=yo+n, Q=N+yo=N+n+yo=N+Zo.

Consequently, L=Pn Q=(M+Zo) n(N‘I‘Zo):

which is easily seen to be (M n N) + 20, and furthermore, M n N is one-

dimensional by Theorem 30.5. Thus, a line of solid analytic geometry is a coset of a one-dimensional subspace of R3. Conversely, if L is the coset D + 20 of a one-dimensional subspace D of R3, it follows easily from Theorem

28.7 that there exist two-dimensional subspaces M and N of R3 such that D = M n N, whence L=(MnN)+ZO=(M+Zo)n(N+Zo),

and consequently, L is the intersection of two nonparallel planes. In summary, the lines of solid analytic geometry are precise the cosets of one-dimensional subspaces of R“. If M is a one-dimensional subspace of R“, that is, a homogeneous line of

the plane, by the preceding discussion Rz/M is the set of all lines in the plane parallel to M, and the canonical epimorphism «pM from R2 onto Ila/M is the function associating to each point of R2 the line containing it parallel to M. If N is a homogeneous line distinct from M, then R2 is the direct sum

of M and N by Example 30.2. By Theorem 30.7, therefore, the restriction of em, to N is a bijection from N onto Rz/M; this is just the vector space

translation of the geometric assertion that each line in the plane parallel to M intersects N in one and only one point. Similarly, vector space concepts admit a geometric description if the vector space is R3 (Exercise 31.1).

EXERCISES 31.1. Let L be a one-dimensional subspace and P a two-dimensional subspace of R3. Describe in geometrical language (a) the set R3/P and the canonical epimorphism (pp;

Sec. 31

VECTOR SPACES

299

and hence, M + N = R3; therefore, there exist m e M and n e N such that

xo—yo=m+n. Letzo=xo—m. Then

P=M+xo=M+m+zo=M+zo,

and also as zo=yo+n, Q=N+yo=N+n+yo=N+zo.

Consequently,

L=PnQ= E** is a linear transformation. (b) If M is a submodule of E, then M°" is a submodule of E*. (c) If N is a submodule of E", then N"° is a submodule of E. 31.6. If E and G are isomorphic K-modules, if F and H are isomorphic Kmodules, and if K is commutative, then HomK(E, F) and HomK(G, H)

are isomorphic K-modules. 31.7. If E and F are modules over a commutative ring K and if M is a submodule of E, then {u E HomK(E, F): u(]ll) = {0}} is a submodule of HomK(E, F). What is its dimension if K is a field and if E, F, and M have dimensions n,

m, and p respectively? 31.8. Let n be a strictly positive integer, and let Pn(R) be the n-dimensional subspace of P(R) consisting of all polynomial functions of degree < n. For each real number a, the function M: p Hp(a) is a linear form on

P”(R). If (ak)1 5,55,, is a sequence of distinct real numbers, then (ah/01 5,,“

is an ordered basis of P,,(R)*. To what ordered basis (“915“,, Br Pn(R) is (“I/9199. dual? [Express each pk as a product of polynomial functions of degree 1.] 31.9. Let E and F be modules over a commutative ring with identity K. For each u e HomK(E, F), the transpose of u is the function u‘ from F* into

E* defined by

u‘(y’) = y' o u for all y’ e F*. (a) Show that

(96,14‘0'» = ("(1%)“) for all x 6E, y' e F*, u e HomK(E, F).

vscron SPACES

302 (b) (c) (d) (e)

Chap. V

Show that u‘ e HomK(F*, E*). Show that ker u‘ = u* (E)°". Show that u§k(F*) E (ker u)°". If K is a field and if E and F are n-dimensional vector spaces, then

14!..(F*) = (ker 10"”. 904) = p(u‘). 1'(u) = 1(14‘). 31.10. Let E, F, and G be modules over a commutative ring K. (a) If u e HomK(E, F) and if v e HomK(F, G), then (u o u ‘ = u‘ o 0'.

(b) If u is an isomorphism from E onto F, then u‘ is an isomorphism from F* onto E*, and (u‘)“ = (u“)'. (c) If u E HomK(E, F) and if M is a submodule of E, then u*(M)°" =-

(u‘)*(M°*)31.11. Let E and F be finite-dimensional vector spaces over a field.

(a) The function T: u Hu‘ is an isomorphism from the vector space HomK(E, F) onto the vector space HomK(F*, E*). (b) The function T: it)» u‘ is an anti-isomorphism (Exercise 15.8) from

the ring EndK(E) onto the ring EndK(E*). *31.12. Let E be a finite~dimensional vector space over a field, and let M and N

be subspaces of E. (a) Show that M E Nifand only ifM°" a N°". (b) Show that (M + N)°" = M°" n N°". (c) Show that (M n N)0" = M °" + N 9". [Use (b) and Theorem 30.5.] (d) Generalize (b) and (c) to any finite number of subspaces. n

(e) If u,u1,... ,uneE*, then n keru,c s keru if and only if u is a Ic=1

linear combination of ("1913.51.-

*31.13. Let E be a module over an integral domain K. (a) If the annihilator of E in K (Exercise 27.8) is not the zero ideal, then

E* = {0}. (b) If E is a quotient field of K considered as a K-module, then E* 9% {0} ifand only ifE = K. 31.14. If E is a finite-dimensional vector space over a field and if N is a subspace of E*, then N°" = JAN“). 31.15. Let E be a vector space over a field K. (a) E is a module over the ring EndK(E), where scalar multiplication is defined by u.x = u(x) for all u eEndK(E), e; moreover, the annihilator of the EndK(E)-module E (Exercise 27.8) is {0}. (b) If E is finite-dimensional, then the EndK(E)-module E is generated

by each nonzero element of E. (c) If, in addition, dimKE 2 2, then the EndK(E)-module E has no

basis. Conclude that the statement obtained from Exercise 28.18(b) by omitting the hypothesis that K be commutative is not in general true.

Sec. 32

VECTOR SPACES

303

32. Matrices"

Here we shall show how to describe a linear transformation from one finitedimcnsional module into another in a particularly concrete way by means of matrices—rectangular arrays of scalars—relative to ordered bases of the domain and codomain. If m and n are strictly positive integers and if K is a set, an m by n matrix over K is a function from [1, m] X [1, n] into K. We

shall denote the set of all m by n matrices over K by JK(m, n). Thus, by definition .1K(m, n) is the set Kll’mlxfl'“. Usually, matrices are denoted by means of indices: If “a e K for all (i, j) e [1, m] X [1, n], the m by n matrix whose value at each (i,j) is cg, is denoted by (“ii)(¢.i)e(1.m]x[1.nl [or simply (my) if it is clear what the set of indices is]. If A = (“55) is an m by n matrix over K, for each i e [1, m] the ordered n-tuple (an, «,2, . . . , at“) is called the ith row of A, and for each j e [1, n] the ordered m-tuple (0:1,,

a”, . . . , am) is called the jth column of A. This terminology arises from the fact that an m by n matrix (0%,) is often denoted by the rectangular array 0‘11

0‘12

-

-

.

«1,,

t"in

“as

-

-

-

“2»

“m1

“m2

-

.

.

0cm

in which on“ is the entry appearing in the ith row and thejth column. For a

specific matrix it is often possible to dispense with indices. For example, 2

3

—1

0

7

5

is the2by3 matrix (eta) where ecu = 2, an = 3, 0:13 = —1, an = 0, at” = 7, and or“ = 5. Definition. If A = (my) and B = (5a) are m by n matrices over a ring K,

then the sum A + B of A and Bis the m by n matrix (a:fl + fin), and for each I. e K, the scalar product M of A and A is the m by n matrix (Ma). Thus, by definition,

(mu) + (13“) = (“a + Isa), 3'01“) = (Mia)-

‘ Subsequent chapters do not depend on this section.

304

vncron SPACES

Chap. V

If K is a ring with identity, then with these definitions of addition and scalar multiplication, .1K(m, n) is a K-module, namely, the K-module Knmlxnm]

of Example 27.4, where F of that example is the K-module K and E is the set [1, m] X [1, n]. The neutral element for addition is called the zero

matrix, for all of its entries are zero. The additive inverse of the matrix (01“) is therefore (—a,,).

Definition. If A = (a,,) is an m by n matrix over a ring K and if B = ([3,) ifan n byp matrix over K, the product AB ofA and B is the m byp matrix

(ya) over K where 7“ = ion-H3“ k=1

for all (i,j) e [1, m] X [1,p]. Thus, the product of two matrices over the same ring is defined if and only is the number of columns of the first is the same as the number of rows of the second; in that case, the entry in the ith row andjth column of the product

is the result of multiplying pairwise the entries in the ith row of the first by the entries in the jth column of the second and then adding. The following

are some examples of matric multiplication. [an “12:“:b11 bu] _ [aubu + amber “11171: + ambit] an

an

2 [3

l 0

bai

baa

2

3

5

8

4

8

6

1

—l

7

0

7

0

3

l]

0 6

azibn + aszbn

8

l4

l6

l7

[—l

58

15

73]

=

7]

2 [5

asibn + anbu

2 =[l6]

0 6

[5

3

l]=

10

6

2

0

0

0

3O

18

6

13—23—12 000 —2—6 43 5—4=ooo 412—867—5 000 Definition. Let @9195" and (b,)15,5m be ordered bases respectively of an

Sec. 32

vncron SPACES

305

n-dimensional module E and m-dimensional module F over a commutative

ring with identity K. For each u e HomK(E, F), the matrix of u relative to (“0555. and (balsam is the m by n matrix (0%) where “(‘15) =‘gdubt

for all (i,j) e [1, m] x [1, n]. For example, the matrix of the differentiation linear transformation

D: P I—>Dp from P4(R) into P3(R) relative to the ordered bases (Pfiogsa and (12905.5. (Example 28.7) is 0

1

0

0

0 0 2

0

0 0 0

3 .

To avoid cumbersome notation, we shall sometimes denote an ordered

basis (“19199. of a module simply by (a),, and the matrix of a linear transformation it relative to ordered bases (a),, and (b),,, by [u; (b),,,, (a),,], or simply by [u] if it is understood in the context which ordered bases are in use. [Note the order: In our notation for the matrix of u relative to (a)n and (b),,,, the notation for the given ordered basis (a),, of the domain of u occurs

last and is preceded by the notation for the given ordered basis (b),,, of the codomain.] The entries in thejth column of [u; (b),,,, (a),,] are thus the scalars occurring in the expression of u(aj) as a linear combination of the sequence (b1, . . . , bm). The reason for this choice is given in the following theorem. Theorem 32.1. Let D, E, and F be finite-dimensional modules with ordered

bases (4),, (b),,, and (c),,, respectively over a commutative ring Kwith identity. Then M: u H l“; ((3)715: (b)n]

is an isomorphism from the K-module HomK(E, F) onto the K-module IKOn, n), and for each u e HomK(D, E) and each 12 e HomK(E, F),

[0 ° 14; (0).», (0),] = iv; (0).», (b)n] [14; (b)... (0),]Proof. The proof that M is an isomorphism is straightforward, so we shall consider only the last assertion. If (any) = [u; (b),,, (a),,] and if (#6,) =

Sec. 32

vscron SPACES

305

nrdimensional module E and m-dimensional module F over a commutative ring with identity K. For each u e HomK(E, F), the matrix of u relative to

(“01519: 311d (b6)1sgsm is the m by n matrix (atfi) where

u(a;) = :1 “ubc for all (i,j) e [1, m] x [1, n]. For example, the matrix of the differentiation linear transformation

D: P HDp from P4(R) into P3(R) relative to the ordered bases (119035,

and (Mags2 (Example 28.7) is 0

l

0 0

0 0 2

0

0 0 0

3 .

To avoid cumbersome notation, we shall sometimes denote an ordered

basis (ak)1 5,6,, of a module simply by (a),, and the matrix of a linear transformation u relative to ordered bases (a),. and (b),,, by [11; (b),,,, (a),,], or Simply by [u] if it is understood in the context which ordered bases are in

use. [Note the order: In our notation for the matrix of u relative to (a), and (b),,., the notation for the given ordered basis ((1),, of the domain of u occurs last and is preceded by the notation for the given ordered basis (b)m of the codomain.] The entries in thejth column of [u; (b),,,, (a),,] are thus the scalars occurring in the expression of u(a,) as a linear combination of the sequence

(b1, . . . , bm). The reason for this choice is given in the following theorem. Theorem 32.1. Let D, E, and F be finite-dimensional modules with ordered bases (a),, (b),,, and (c)m respectively over a commutative ring K with identity. Then M: u H [11; (6),", (b)n]

is an isomorphism from the K-module HomK(E, F) onto the K-module lflm, n), and for each u e HomK(D, E) and each 0 e HomK(E, F), [0 ° u; (0)1»! (a)p] = [I]; (0)1», (b)n] l“; (b)m (a)p]-

Proqf. The proof that M is an isomorphism is straightforward, so we shall

con51der only the last assertion. If (cg-j) = [11; (b),,, (a),] and if (13,-,) =

VECTOR SPACES

306

Chap.V

[I]; (0)1»: (b),,], then

(u o u)(a,) = v(u(aa)) = 1’ (élak ibk)

=§1WW =72“, (g aka) =:§1(i:amfimc‘) = in; (éluufiac‘) = i (i fiika'k!) 6‘, SO [U o u; (0),", (a),)] = [0; (0),”, (b)n] [u; (b)m (a)p]'

Multiplication of matrices is purposely defined just to obtain the above equality relating the product of the matrices of two linear transformations

with the matrix of their composite. If E is a K-module, we shall denote by .1!K(n) the set .1!K(n, n) of all n by n matrices over K. An n by n matrix is called a square matrix of order I. Since the product of two square matrices of order n over a ring is a square matrix of order n, matric multiplication is a composition on .1!K(n).

When representing a linear operator on a finite-dimensional module E by a matrix, one usually selects the same ordered basis for E considered both as the domain and the codomain of the linear operator. If we EndK(E) and if ((19199 is an ordered basis of E, we shall abbreviate [u; (a),,, (a),,] to [u; (a),,]. The following theorem is an immediate consequence of Theorem

32.1. Theorem 32.2. If ((49199 is an ordered basis of an nadimensional module E over a commutative ring with identity K, then

M: u H[u; (a)..] is an isomorphism from the ring (EndK(E), +, 0) onto (.1!K(n), +, -). In particular, allK(n) is a ring with identity. By Theorems 31.1 and 32.1, if K is a commutative ring with identity, if U, Veu/lK(m, n), if W, Z EflK(n,p), and if YEIKQ, q), then

(U+ V)W= UW+ VW, U(W+Z)= UW+ UZ, (UW)Y= U(WY).

Sec. 32

vncroa SPACES

307

Actually, a simple calculation shows that these three identities hold even if K lacks an identity element or if K is not commutative (Exercise 32.12). In

particular, therefore,for every ring K and every n e N* , under matric addition and matric multiplication JIK(n) is a ring. If K is a ring with identity 1, it is a simple matter to show that the n by 1: matrix 100

1

-

O

0

0 0 0 . . .

l

is the multiplicative identity of the ring flK(n). The (principal) diagonal of a square matrix (mu) of order n is the sequence an, “22, . . . , a“ of elements occurring on the diagonal joining the upper left and lower right corners of the matrix, written as a square array. Thus, if K is a ring with identity 1, the multiplicative identity of the ring .1!K(n), which is denoted by 1,, and called

the identity matrix is the matrix whose diagonal entries are all 1 and whose nondiagonal entries are all 0. In terms of the Kronecker delta notation, In = “ohms [1,n]X[1,n]'

An invertible matrix of order n is, of course, an invertible element of the

ring JIK(n); if K is a commutative ring with identity, and if E is an ndimensional K-module, then the invertible elements of EndK(E) are precisely the automorphisms of E, so that invertible matrices correspond to automor-

phisms of E under any isomorphism of the rings J/K(n) and EndK(E).

EXERCISES 32.1. Determine the eight products XYZ where each of X, Y, and Z is either 2

0

4

2

3 —1 °r 2 —4 ' 32.2. Make the eight computations of Exercise 32.1 if the numerals denote elements of 25.

Sec. 32

vacron srAcss

307

Actually, a simple calculation shows that these three identities hold even if K lacks an identity element or if K is not commutative (Exercise 32.12). In

particular, therefore,for every ring K and every n e N*, under matric addition and matric multiplication JIK(n) is a ring.

If K is a ring with identity 1, it is a simple matter to show that the n by n matrix

00 0

0_ 0 0 . . .

l

is the multiplicative identity of the ring .1!K(n). The (principal) diagonal of a square matrix (eta) of order n is the sequence can, an, . . . , “m. of elements

occurring on the diagonal joining the upper left and lower right corners of the matrix, written as a square array. Thus, if K is a ring with identity 1, the

multiplicative identity of the ring .1!K(n), which is denoted by 1,, and called the identity matrix is the matrix whose diagonal entries are all 1 and whose nondiagonal entries are all 0. In terms of the Kronecker delta notation, In = (6i!)(i,1)6[1,n]X[1,n]'

An invertible matrix of order n is, of course, an invertible element of the

ring align); if K is a commutative ring with identity, and if E is an ndimensional K—module, then the invertible elements of EndK(E) are precisely

the automorphisms of E, so that invertible matrices correspond to automorphisms of E under any isomorphism of the rings al/K(n) and EndK(E).

EXERCISES

32.1. Determine the eight products X YZ where each of X, Y, and Z is either 2

0

4

2

3 —1 °r 2 —4 ' 32.2. Make the eight computations of Exercise 32.1 if the numerals denote elements of ZS.

VECTOR SPACES

308

Chap. V

32.3. Determine BA — 1;, CB + A, and C’2 + AB if

A =

2

4

l

0 ,

3

2

1

B = [0

2

I],

and

2

l

3

C = 0

2

—2 .

2

2

3

—2

32.4. Make the computations of Exercise 32.3 if the numerals denote elements of Z5. 32.5. Determine the real numbers A for which the following matrices over R are invertible, and for each such 2 determine the inverse.

I.

1

0

1

0

l

1

I. 0

0

A

1

0

1

0

i.

l

I.

1

A 0

2. 0

l

0

I.

1

32.6. Determine the matrices Ra, SM, and PM _N of the linear operators r“, SM,

and pM.N of §29 relative to the standard ordered basis of R“. By matric computation,

show

that

rao r, = rd”, lNopM'N = PM.»

and

SM 0 SM = 1R"

32.7. What are the matrices of ra, sM, and PM.N relative to the ordered basis ((1, l), (0, —l)) of R”? relative to ((1, —l), (2, 4))? 32.8. Let u be the linear operator on P5(R) satisfying

[u(p)l(x) = p(x + l) for all p EP5(R) and all xe R. What is the matrix of u relative to the ordered basis (1790954? 32.9. Determine the matrices of the linear operators defined in Exercises 29.1

and 29.2 relative to the standard ordered bases. *32.10. If u is a linear operator on an n-dimensional vector space E such that u” = 0 but u"‘1 aé 0 [where :4” denotes the linear operator u o u o . . . a u (7: factors)], then for some a E E, (u"’(a))o 5k 513—1 is an ordered basis of E.

What is the matrix of u with respect to this ordered basis? What is such an ordered basis if u is the differential linear operator D on the n-dimensional vector space P”(R) ? 32.11. Let K be a ring with identity, and let (afllsjsfl and 091519» be ordered

bases respectively of an n-dimensional K-module E and an m-dimensional K-module E For each u e HomK(E, F), the matrix of u relative to

(“01951: and (bals‘Sm is the matrix (aw) over the reciprocal ring L of K (Exercise 15.8) where

1401;) = a: “ab; =1 for all (i,j)e [1, m] X [1, n].

(a) Prove the statement obtained by deleting “commutative” from the assertion of Theorem 32.1.

VECTOR SPACES

Sec. 32

309

(b) The function M: uI—> [u; (a),,] of Theorem 32.2 is an isomorphism

from the ring EndK(E) onto 11,01). 32.12. Let K be a ring, let U, VEIK(m, n), let W,Ze./K(n,p), and let

' lie/3(1), q). (a) Verify

that

(UW)Y= U(WY),

(U + V)W = UW+ VW, 'and

U(W+Z) = UW+ UZ.

(b) Infer that under matric addition and multiplication, .IK(n) is a ring. 32.13. Let E and Fbe vector spaces over a field of dimensions n and m respectively,

let (“19151.5» and (bk)1 57:51» be ordered bases of E and F respectively,

and let u e HomK(E. F). If (em) = [u; (b)... (11).] and if (flu) = [u‘; (a’)... (b'),,,] (Exercise 31.9), where (“191951; and (b91391. are the ordered bases dual to (ak)1 51:51: and (4)135». respectively, then flu = a” for all is [1, n],je [1, m]. 32.14. Let (99°54 be a family of distinct terms of a K-module E such that {eaz a: e A} is a basis of E. (a) Extend Theorem 29.4 as follows: If (ball 54 is a family of elements of

a K-module F also indexed by A, there is one and only one linear transformation u from E into F satisfying u(ea) = bu for all a e A. (b) The K-module E* is isomorphic to the K-module K4.

*32.15. Let E be a vector space having a denurnerable basis {e,,: n EN*} (e.g., let E =P(R) and let e" = pn_1 for all neN“), and let K be the ring

EndK(E). Let u1 and u; be the linear operators on E satisfying “1(ean—1) = 0,

"2(e2n—1) = em

“1(e2u) = em

ug(egn) = 0

for all n E N* (Exercise 32.14(a)). (a) Show that {up ua} is a basis of the K-module K. (b) Let vk = '12q for each k 6 1V. Show that

forallq* andthat

"k(32*(2a—1)) = ea v,,(e,) = 0

if r is not an odd multiple of 2". (c) For each m e N*, {00, vl, . . . , vm_1, ulm} is a basis of m + 1 elements of the K-module K.

(d) For each k e N, let 3,, be the linear operator on E satisfying 31193) = 92"(2i—1)

for all jeN*. Prove that 1': wn—)(wsk)k20 is an isomorphism from the K-module K onto the K-module KN.

CHAPTER VI

POLYNOMIALS

Polynomial functions on the field of real numbers, i.e., functionsf satisfying f(x) = anx" + a,,_1x"‘1 + . . . + tax + a0 for all real numbers x, where a0, a1, . . . , and, an are given real numbers,

and the ways of combining them by addition and multiplication are familiar from elementary algebra. The definition of “polynomial function” is easily generalized by replacing the field of real numbers with any commutative ring with identity. In modern algebra it is convenient to study objects called “polynomials" that are closely related to but not identical with polynomial

functions. Such objects are fundamental to the study of fields, and this chapter is primarily devoted to their study. 33. Algebras

Vector spaces and modules, the most important examples of K-algebraic structures with one composition, were investigated in Chapter V. The most

important examples of K-algebraic structures with two compositions are called simply algebras.

Definition. Let K be a commutative ring with identity. A K-algebra, or an algebra over K, is a K-algebraic structure (A, +, -, .) with two compositions such that (A l)

(A, +, .) is a K-module,

(A 2)

(A, +, -) is a ring,

(A 3)

a(xy) = (aux)y =- x(o:y) for all x, y e A and all a: e K. 310

Sec. 33

POLYNOMIALS

31 1

If, in addition, (A, +, -) is a division ring, then (A, +, -, .) is called a division

algebra over K.

Example 33.1. Let E be a module over a commutative ring with

identity K. For every at 6K and every u e EndK(E), om belongs to EndK(E) by Theorem 31.2. One easily verifies that «(140) = (au)v = u(ow) for all at e K and all u, v e EndK(E). Thus, EndK(E) is a K-algebra. Example 33.2. If K is a commutative ring with identity, then .l/K(n) is a K-algebra. If E is an n-dimensional K-module, the K-algebra

.1K(n) is isomorphic to the K-algebra EndK(E) (Theorem 32.2). Example 33.3. If (A, +, -) is a ring with identity and if K is a subring of the center of A containing the identity element of A, by the Kalgebra A we shall mean (A, +, -, .) where . is the restriction to

K x A of the given multiplication ' on A. It is easy to verify that (A, +, -, .) is indeed a K-algebra since K is contained in the center of A. In particular, a commutative ring with identity may be regarded as a one-dimensional algebra over itself. Example 33.4. If (A, +, -) is a ring, then (A, +, -, .) is a Z-algebra where . is the function defined in §11 and §14. Indeed, (A, +, .) is a Z-module by Theorem 14.8, and (A 3) holds by (1) of §15. Thus,

every ring can be made into a Z-algebra in a natural way. Example 33.5. If A is a K-algebra and if E is a set, the compositions

and scalar multiplication induced on the set AE of all functions from E into A by the compositions and scalar multiplication ofthe K-algebra

A (Examples 15.5 and 27.4) convert AE into a K-algebra. Example 33.6. If (A, +, -, .) is a K-algebraic structure such that (A, +, .) is a K-module and (A, +, -) is a trivial ring, then

(aux)y = ot(xy) = x(aty) = 0 for all x, y e A and all at e K, so (A, +, -, .) is a K-algebra. Such algebras

are called trivial algebras. A particularly trivial algebra is a zero algebra, one containing only one element. Definition. If B is a stable subset of a K-algebraic structure (A, +, -, .),

then (B, +3, 3, .3) is a subalgebra of (A, +, -, .) if it is itself a K-algebra.

Sec. 33

POLYNOMIALS

31 1

If, in addition, (A, +, -) is a division ring, then (A, +, -, .) is called a division algebra over K.

Example 33.1. Let E be a module over a commutative ring with identity K. For every o: GK and every u eEndK(E), oiu belongs to EndK(E) by Theorem 31.2. One easily verifies that ac(uv) = (om)v = u(ow) for all at GK and all u, v e EndK(E). Thus, EndK(E) is a K-algebra. Example 33.2. If K is a commutative ring with identity, then u/(K(n) is a K—algebra. If E is an n-dimensional K—module, the K-algebra clan) is isomorphic to the K-algebra EndK(E) (Theorem 32.2). Example 33.3. 'If (A, +, -) is a ring with identity and if K is a subring of the center of A containing the identity element of A, by the Kalgebra A we shall mean (A, +, -, .) where . is the restriction to K X A of the given multiplication - on A. It is easy to verify that (A, +, -, .) is indeed a K-algebra since K is contained in the center of A. In particular, a commutative ring with identity may be regarded as a one-dimensional algebra over itself. Example 33.4. If (A, +, -) is a ring, then (A, +, -, .) is a Z-algebra where . is the function defined in §11 and §14. Indeed, (A, +, .) is a

Z-module by Theorem 14.8, and (A 3) holds by (l) of §15. Thus, every ring can be made into a Z-algebra in a natural way. Example 33.5. If A is a K-algebra and if E is a set, the compositions

and scalar multiplication induced on the set AE of all functions from E into A by the compositions and scalar multiplication of the K-algebra A (Examples 15.5 and 27.4) convert AE into a K-algebra. Example 33.6. If (A, +, -, .) is a K—algebraic structure such that (A, +, .) is a K-module and (A, +, -) is a trivial ring, then

(aux)y = at(xy) = x(oty) = 0 for all x, y e A and all at e K, so (A, +, -, .) is a K-algebra. Such algebras

are called trivial algebras. A particularly trivial algebra is a zero algebra, one containing only One element. Definition. If B is a stable subset of a K-algebraic structure (A, +, -, .), then (B, +3, -3, ~13) is a subalgebra of (A, +, -, .) if it is itself a K-algebra.

3 12

POLYNOMIALS

Chap. VI

A stable subset B of (A, +, -, .) is also called a subalgebra if the Kalgebraic structure (B, +3, '3, .B) is a subalgebra in the sense just defined. One consequence of Theorem 28.1 is the following theorem.

Theorem 33.1. If A is a K-algebra, a nonempty subset B of A is a subalgebra of A if and only if for all x, y e B and all a e K, the elements x + y,

xy, and ocx all belong to B, i.e., if and only if B is a stable subset of A. Theorem 33.2. Let A be a K-algebra. The intersection of any set of subalgebras of A is a subalgebra; consequently, if S is a subset of A, the intersection ofthe set ofall subalgebras ofA containing S is the smallest subalgebra of A containing S.

The proof is easy (see Theorem 28.2). Theorem 33.2 enables us to make the following definition:

Definition. If S is a subset of a K-algebra A, the subalgebra generated by S is the smallest subalgebra B of A containing S, S is called a set of generators of B, and B is said to be generated by S.

Theorem 33.3. Let f and g be homomorphisms from a K-algebra A into a K-algebra B. The set

H = {x e A=f(x) = goo} is a subalgebra of A. If S is a set of generators for the algebra A and if fix) = g(x) for all x e S, then f = g. The proof is easy (see Theorem 29.3). If (A, +, -, .) is a K-algebra, by the ring A we mean (A, +, -), by the

module A (or vector space A if K is a field) we mean (A, +, .), by the (additive)

group A we mean (A, +), and by the multiplicative semigroup A we mean (A, °). An algebra A is commutative if multiplication is a commutative composition, and A is an algebra with identity if the ring A is a ring with identity. Definition. An ideal of an algebra A is a subset of A that is both an ideal of the ring A and a submodule of the module A. Thus, a nonempty subset a of a K-algebra A is an ideal if and only if for all a, b e a, for all x e A, and for all 1 GK, the elements a + b, xa, ax, and in

all belong to a. If there is an identity element e for multiplication, then an

Sec. 33

POLYNOMIALS.

313

ideal a of the ring A is also an ideal of the algebra A, for if a e a and if A e K, then la = 1(ea) = (110a e a.

Otherwise, however, there may exist ideals of the ring A that are not ideals

of the algebra A. For example, the R—vector space R becomes a trivial R-algebra when multiplication is defined by xy = 0; Q is an ideal ofthe trivial ring R but is not an ideal of the trivial R-algebra R. If (11, . . . , a,, are ideals of A, then 'a1+...+a,,,

aln...na,, are ideals of A by Theorems 19.3 and 28.2. If a is an ideal ofa K-algebra A , the compositions and scalar multiplication of A induce compositions and a scalar multiplication on the quotient set

Ala, given by

W+®+0+®=Q+fl+m (x+a)(y+a)=xy+a, at(x+a)=acx+a, as we saw in our discussion of quotient rings and modules. The canonical surjection (pa: x 1—) x + a is thus an epimorphism, so by the following theorem, A/a is a K-algebra, called the quotient algebra defined by a. Theorem 33.4. Iffis an epimorphism from a K-algebra A onto a K-algebraic structure B with two compositions, then B is a K—algebra. Moreover, if A is commutative, so is B.

Proof. By Theorem 21.1 and the remark following Theorem 29.1, it suffices to make the easy verification that 11(zw) = (lz)w = z(}.w) for all A e K, 2, w e B.

Corollary. Iff is a homomorphism from a K-algebra A into a K-algebraic structure (B, +, -, .), then the rangef, (A) offis a subalgebra of (B, +, -, .). Moreover, if A is commutative, so is f, (A). The proof is similar to that of the Corollary of Theorem 21.1.

Theorem 33.5. Letfbe an epimorphism from a K-algebra A onto a K-algebra B. The kernel (1 offis an ideal of A, and there is one and only one isomorphism

314

POLYNOMIALS

Chap. VI

g from the K-algebra A/a onto B such that g ° (pa =f

Furthermore, f is an isomorphism if and only if a = {0}. The assertion is a consequence of Theorems 21.4 and 30.6. Theorem 33.6. If A is a K-algebra with identity e, then h: at 9—) one

is a homomorphism from the K-algebra K into A, and h is a monomorphism if and only if {e} is linearly independent. Proof. That h is a homomorphism follows from the identities (at + fi)e = ace + fie,

(0:13)e = oc(fie), (ueXfle) = «(458» = 04199’) = «(1312) = (time. Furthermore, by the definition of linear independence, {e} is linearly independent if and only if the kernel of h is {0}.

In View of Theorem 33.6, if A is a K-algebra with identity e such that {e} is linearly independent (in particular, if K is a field), K is frequently “identified” with the subalgebra Ke of A; that is, the subalgebra Ke of A is denoted simply by “K,” and for each at e K, the element are of A is denoted simply by “a.” Theorem 33.7. If K is a field and if A is a finite-dimensional K-algebra with identity, then every cancellable element of A is invertible. Proof. Let e be the identity element of A, and let a be a cancellable element. Then La: x I—) ax and R4: x »-—> xa are injective linear operators on the K-vector space A, so L, and R, are permutations of A by Theorem 29.6. In particular, there exist b, c e A such that

ab = La(b) = e = Ra(c) = ca, whence as c = c(ab) = (ca)b = b, a is invertible.

POLYNOMIALS

Sec.33

315

Corollary. If K is a field, if A is a finite-dimensional K-algebra with identity,

and if every nonzero element of A is cancellable, then A is a division algebra.

EXERCISES 33.1. Complete the verifications needed in Examples 33.1, 33.3, and 33.5. 33.2. (a) Prove Theorem 33.2. (b) Prove Theorem 33.3.

(c) Complete the proof of Theorem 33.4. 33.3. Let K be a commutative ring with identity, let (A, + , -, .) be a K-algebraic structure with two compositions, and for each a e A, let L, be the function

from A into A defined by La(x) = ax for all x e A. Then (A, +, -, .) is a K-algebra if and only if (A, +, .) is a K-module, L.I is a linear operator on the K-module A for all a e A, and L: a 1—>La is a homomorphism from (A, + , ‘, .) into the K—algebra EndK(A).

33.4. Let E be a module over a commutative ring K. If (991 m. By our inductive hypothesis, all

terms after the first belong to K. Since the sum is a coeflicient ofgh and hence belongs to K, we have an—k—lflm e K, whence

“n—k—r = (“n—k—Iflm)131_n1 e KConsequently, k + l e S, so by induction S = [0, n]. Therefore g eK[X].

EXERCISES 34.1. Complete the verification of the statement that (KN, + , - , .) is a K-algebra.

34.2. Find the quotient and remainder obtained by dividing 2X2 + 4X — 1 into X4 + 2X3 — 3X” + 4X + l where the scalar field is Q (respectively, 259 Z7)°

34.3. ForwhatprimenumberspisX°+X5+4X4+X3—7X’+4divisibleby —2inZz,[X]? X5+ 2X2—1divisibleb2+ 3inZ”[X]? 34.4. If (fk)0$5" is a sequence of nonzero polynomials over a field K such that degf7, = k for each k e [0, n], then (fk)o_ «5,, is an ordered basis of the

subspace of K[X] of all polynomials of degree < n. *34.5. Let K be a commutative ring with identity, let g be a nonzero polynomial over K of degree n, and let u be the leading coeflicient ofg. (a) Iffis a polynomial over K of degree m and if k is the larger of m — n + 1 and 0, then there exist polynomialsq and r over K such that

“"f=qg+r. eitherr =Oor degr 0and if}? =0,

In: 0 then there is a nonzero polynomial In over K of degree < n such thatfl: = 0.

[Assume [30 9% 0. If «kg = 0 for all k e [0, m — 1], consider h = lioX"; if

akg=0 for all ke[0,p —1] wherep 1(a) and Ma + b) > Mb),

divide a + b into a2 — b2 + b in two ways to obtain remainders b and —a. To show that 1° and 2° imply 3°, use Exercise 22.13(b) and, if the set N of nonzero noninvertible elements of A is not empty, consider c e N where 1(c) is the smallest member of 1*(N).]

36. Irreducibility Criteria

If A is a commutative ring with identity, we shall say that a nonzero polynomial f is irreducible over A iff is an irreducible element of the ring A[X],

and that f is reducible over A if f e A[X] and if f is neither a unit nor an irreducible element of A[X]. An important but often diflicult problem of algebra is to determine whether a given polynomial over A is irreducible. Iffis a nonconstant polynomial over a field K and if n = degf, then fis irreducible over K ifand only if the degree ofevery divisor offin K[X] is either n or 0, for the units of K[X] are the polynomials of degree zero by Theorem 34.4, and therefore, the associates offare those polynomials of degree n that dividef by Theorem 34.3. Every linear polynomial over a field K clearly has

a root in K. Thus, every linearpolynomial over afield K is irreducible and has a

342

POLYNOMIALS

Chap. VI

root in K. By Theorem 35.7, however, a polynomial of degree 2 2 that is irreducible over K has no roots in K. A polynomial may be reducible over K and yet have no roots in K; the polynomial X4 + 2X3 + 1 over R is such a polynomial. However, iff is a quadratic or cubic polynomial over afield K, thenfis irreducible over K ifand only iffhas no roots in K, for iff = gh where g and h are nonconstant polynomials over K, then either g or h is linear and hence has a root in K, which necessarily is also a root off2 Thus, X”l + l is

irreducible over R as it has no roots in R, but X” + 1 is, of course, reducible over C.

The polynomialf = 2X + 6 is irreducible over Q but reducible over Z, for the polynomial 2Xo dividesfin Z[X] but is not a unit of Z[X]. A criterion for irreducibility of nonconstant polynomials over an integral domain that includes the above criterion for irreducibility of nonconstant polynomials over

a field is the following: Iff is a nonconstant polynomial over an integral

domain A and ifn = degf, thenfis irreducible over A ifand only ifthe nonzero coefi‘icients off are relatively prime in A and the degree of every divisor off in A [X] is either n or 0. Indeed, the units of A [X] are those constant polynomials determined by the units of A by Theorem 34.4. The condition is, therefore, necessary, for if at is a common divisor of the nonzero coefficients of a non-

constant polynomialfirreducible over A, then 1X“ dividesfand clearly is not an associate off, so «X0 is a unit ofA [X] and hence at is a unit ofA. Conversely,

if the condition holds and iff = gh, then either g or h is a constant polynomial ocX" by hypothesis, and since ocX° divides f in A[X], clearly at is a common divisor of the nonzero coefficients off, so, by hypothesis, on is a unit of A, and therefore, «X0 is a unit of A[X]. Consequently, if A is a subdomain ofafield K, Iff is a nonconstant polynomial over A that is irreducible over K, and if

the nonzero coefi‘icients off are relatively prime in A, then f is irreducible over A. One of our most important results is the converse for the case where A is a unique factorization domain and K a quotient field of A (Theorem 36.3). Definition. A nonzero polynomial f over an integral domain A is primitive over A if its nonzero coeflicients are relatively prime in A (or,in case f has only one nonzero coeflicient at, if at is a unit).

Lemma 36.1. If f is a nonzero polynomial over a unique factorization domain A, then at is a greatest common divisor of the nonzero coefficients of

f if and only if there is a primitive polynomial f1 over A such that f = ufr Proof. Necessity: Since at is a greatest common divisor of the nonzero coefficients off, clearly there existsf1 6 A [X] such that 0c = fl If [9 is a common divisor of the nonzero coeflicients off1, then up is a common divisor of the nonzero coefficients off, so afl] at, and therefore, [3 is a unit. Hence, f1 is

primitive. Sufliciency: By Theorem 22.10 the nonzero coeflicients of f admit a

Sec. 36

POLYNOMIALS

343

greatest common divisor A, and as observed above, there is a polynomial f,eA[X] such that f = Afi. Since afl = f, at is a common divisor of the nonzero coeflicients off, and therefore, there exists ,3 e A such that on}? = A. Thus, onfifa = f = «fl, so fifz =f1, and therefore, 13 is a cormnon divisor of the nonzero coefficients off1. But sincef1 is primitive, fl is, therefore, a unit, so a and 2. are associates, and consequently, ac is also a greatest common divisor of the nonzero coefficients off. Lemma 36.2. Let A be a unique factorization domain, let 7»

a

h = 2 fix"

1: =25; fikX",

7c=0

be nonzero polynomials over A, let m+a

gh = 2 (Xk, k=0

and let 1r be an irreducible element of A such that 7X0 divides neither g nor h in A [X]. If r is the smallest integer in [0, m] such that 11 4’13, and ifs is the smallest integer in [0, q] such that w zl’ y” then 1r 4’ a,”

Proof. By hypothesis, 11' | £3, for all i< r and 7r l y,+,_, for all i > r, so 11 | fi,y,+,_, for all i sé r. If 1r were a divisor of “H.” then 71' would divide “1+3 _§ fliVHs—i = fir?” I r

whence 11 would divide either [3, or y, by (UFD 3), a contradiction. Theorem 36.1. (Gauss’s Lemma) Iff and g are primitive polynomials over a unique factorization domain A, then fg is primitive over A. Proof. If 71' is an irreducible element of A, then 71X“ divides neither f nor g in A[X] by hypothesis, so by Lemma 36.2, there is a coeflicient of fg not

divisible in A by 17. Consequently, by (UFD 1), the units of A are the only divisors of all the nonzero coeflEicients offg, so fg is a primitive polynomial. Theorem 36.2. Let A be a unique factorization domain, let f be a nonzero polynomial over A, and let K be a quotient field of A. If there exist polynomials g, h e K[X] such that f = gh, then there exist polynomials go, ho e A[X] that are multiples by nonzero scalars of K of g and h respectively such thatf = goho. Proof. As the nonzero coeificients of g and h are quotients of elements of A, there exist nonzero elements at and (3 of A such that ocg and flh are polynomials over A. By Theorem 22.10, the nonzero coefficients of ocg and fih

See. 36

POLYNOMIALS

343

greatest common divisor h, and as observed above, there is a polynomial f,EA[X] such that f = hf... Since ocf1 = f, ac is a common divisor of the nonzero coeflicients off, and therefore, there exists )3 e A such that «,3 = A.

Thus, acflf2 = f = «fl, so flfz = fl, and therefore, ,3 is a common divisor of the nonzero coefficients off1. But sincef1 is primitive, fl is, therefore, a unit, so on and h are associates, and consequently, at is also a greatest common divisor of the nonzero coefficients off.

Lemma 36.2. Let A be a unique factorization domain, let in

(I

g=2flkXE k=0

h=2 M" k=0

be nonzero polynomials over A, let m+a

gh = 2 “kr k=0

and let 71' be an irreducible element of A such that 11X“ divides neither g nor h in A[X]. If r is the smallest integer in [0, m] such that 17 ,f ,3, and if sis the

smallest integer in [0, q] such that 11- * 9),, then a-r ,f «”3. Proof. By hypothesis, 1r | ,3, for all i< r and 11' I 7m...- for all i> r, so 77 l fiwHH for all i 75 r. If 17 were a divisor of apps, then 17 would divide ar+s _ Zfit'yfia—i = [37'7” 6397‘

whence 11' would divide either [3, or y, by (UFD 3), a contradiction. Theorem 36.1. (Gauss’s Lemma) Iff and g are primitive polynomials over a unique factorization domain A, then fg is primitive over A.

Proof. If 11- is an irreducible element of A, then 11X“ divides neither f nor g in A[X] by hypothesis, so by Lemma 36.2, there is a coefficient of fg not

divisible in A by 11'. Consequently, by (UFD 1), the units of A are the only divisors of all the nonzero coefficients offg, so fg is a primitive polynomial.

Theorem 36.2. Let A be a unique factorization domain, let f be a nonzero

polynomial over A, and let K be a quotient field of A. If there exist polynomials g, h e K[X] such that f = gh, then there exist polynomials go,

ho E AIX] that are multiples by nonzero scalars of K of g and h respectively such that f = goho. Proof. As the nonzero coefficients of g and h are quotients of elements of A, there exist nonzero elements at and 5 of A such that org and flh are polynomials over A. By Theorem 22.10, the nonzero coefficients of ocg and [3h

344

POLYNOMIALS

Chap. VI

admit greatest common divisors (11 and [31 respectively in A, and by Lemma 36.1, there exist primitive polynomials g1 and h1 over A such that org = “131 and flh = filhl. By Theorem 36.1, glh1 is primitive, so by Lemma 36.1, cap, is a greatest common divisor of the nonzero coefficients of «1/31,t1 = aflgh =

afif. Consequently, «,3 divides «1/31 in A, so there exists y eA such that «fly = «1,31, whence f = yglhl. The polynomials go = yg1 and h, = hl,

therefore, have the desired properties. Theorem 36.3. Let A be a unique factorization domain, and let K be a

quotient field of A. If f is a nonconstant polynomial over A, then f is ir-

reducible over A if and only iff is irreducible over K and the nonzero coeflicients off are relatively prime in A.

Proofl It follows at once from Theorem 36.2 that the condition is necessary, and we saw in our earlier discussion that it is also sufficient.

Thus, a nonconstant polynomial irreducible over a unique factorization domain is also irreducible over its quotient field. The following theorem is useful in determining the irreducibility of many polynomials. Theorem 36.4. (Eisenstein’s Criterion) Let K be a quotient field of a unique factorization domain A, and let f = 2 ackX" be a nonconstant polynomial k=0

over A of degree n. If there exists an irreducible element 17 of A such that 11' 4’ on”, 71' och for all k e [0, n — l], and 71‘ 4’ 0:0, thenfis irreducible over K. Proof. By Theorem 36.2, it sufiices to prove that iff = g.,ho where go and h, are polynomials over A, then either go or ho is a constant polynomial. Let in

80 = 2 fik:

k=0

a

h0 = 2 Vk-

k=0

As floyo = a0 and as 11" 4’ do, either 7r 4’ ’30 or 11' 4’ yo, but as 7r «0, either 11- | fio or 11 | 70 by (UFD 3). We shall assume that 11' 4’ [30 and that 71' )2... As fimy, = on” and as 11' 4’ at”, we conclude that 1r 4’ ya. By Lemma 36.2, qr 4’ a, wheresis the smallest integer in [0, q] such that 11' 4’ 31,. By hypothesis, therefore, s=n=m+q2q23,

so m = 0 and go is a constant polynomial. Example 36.1. The polynomials 2X5 + 18X3 + 30)!2 — 24 and 4X7 — 20X5 + 100X —— 10 are irreducible over Q, as we see by applying

Eisenstein’s Criterion where 71' = 3 and 71 = 5 respectively. Neither

Sec. 36

POLYNOMIALS

345

polynomial, however, is irreducible over Z. Dividing each of these

polynomials by the constant polynomial 2, we obtain polynomials that are irreducible over Z by Theorem 36.3. Example 36.2. If p is a prime number and if n is an integer 2 2, then VI) is irrational. Indeed, X" — p is irreducible over Q by Eisenstein’s

Criterion and, consequently, has no roots in Q. The root V; of X" — p is, therefore, irrational.

Example 36.3. If u is an automorphism of an integral domain A, then an element a of A is clearly irreducible if and only if u(a) is irreducible, for a = be if and only if u(a) = u(b)u(c), and b is invertible if and only if u(b) is. In particular, if f is a nonzero polynomial over an integral domain K and if at e K, then f (X) is irreducible if and only if f(X + ac) is, for the substitution endomorphism defined by the polynomial X + on is an automorphism of the ring K[X] (corollary of Theorem 35.6). Thus, if f: X3 + 3X + 2, then f(X+ l) =

X3 + 3X2 + 6X + 6, sof (X + 1) is irreducible over Q by Eisenstein’s Criterion, and therefore, f is also irreducible over Q.

Eisenstein’s Criterion is not applicable to many polynomials over Z that are actually irreducible, such as X5 + X3 + 1 (Exercise 36.14). One may

always apply the method of finding factors learned in beginning algebra to polynomials over Z, however, to determine whether they are irreducible.

Example 36.4. We shall show that f = X’f + X3 + 1 is irreducible over Z. Suppose thatf = gh where g and h are polynomials over Z of degree < 5. We may assume that both g and h are monic, for the product of their leading coefficients is l, and hence their leading coefficients are either both 1 or both —1; in the latter case, we may replace g and h by the monic polynomials —-g and —h. Let c and e be the constant terms of g and h respectively. Then ce = 1, so either c = e =1 or c = e = —1. Since neither 1 nor —-1 is a root off, neither g nor h is linear. We may suppose, therefore, that g is cubic and h quadratic. Let g=X3+aX9+bX+c, h=X2+dX+e. Then

(1) (2)

d+a=0, e+ad+b=l,

(3)

ea+bd+c=0,

(4)

be+cd=0,

(5)

ce = 1.

346

POLYNOMIALS

Chap. VI

If c = e = 1, we obtain a = b = —d from (1) and (4), whence

-—d’— d: 0 and —d— (1‘ +1 = 0 from (2) and (3), which is impossible. If c = e = —l, we obtain a = b = —d from (1) and (4), whence —d’—d=2 and d—d9— 1 =0 by (2) and (3); con-

sequently, —2 — d = d2 = d — 1, so 2d = —l, which is impossible as d is an integer. Therefore, X5 + X3 + l is irreducible over 2 and

hence over Q.

EXERCISES 36.1. Let a be the leading coefl‘icient of a linear polynomial f over an integral domain A. (a) The polynomialfhas a root in A if and only iff = mg for some monic linear polynomial g. (b) The polynomial fhas a root in A and is irreducible over A if and only if a: is invertible in A. 36.2. (a) If a and b are inwgers, then X3 + aX’ + bX + l is reducible over 2

ifandonlyifeithera =bora +b = —2. (b) Determine necessary and sufficient conditions on integers a and b for Xa + a2.”I + bX — 1 to be reducible over Z. 36.3. (a) For what integers b is 3X2 + bX + 5 reducible over Z? (b) If a and c are nonzero integers, give an upper bound in terms of the number of positive divisors of a and of c on the number of integers b for

which aX2 + bX + c is reducible over Z. 36.4. Determine all quadratic and cubic irreducible polynomials over 2,. 36.5. Determine all quartic and quintic irreducible polynomials over 2,. [Use Exercise 36.4.]

36.6. (a) Let a be a proper ideal of a commutative ring with identity K, and let in

(p be the canonical epimorphism from K onto K/a. If f = 2 _ 0, then

|a+bl=a+bSla|+lbl by2°,andifa+b n, then 0 < am _ an =

m 2

2—k(lo+1)=

k-n+1

m—n—l _ 2 2-(:+(n+1))(i+(n+2)) i=0

2i, 2 5 go 2—[( +1)( +2)+n _ 2 (n+1)( +2)m_"_1 s "HM 1‘

n

_



< 2—(u+1)(n+2) . 2 < 2—n(n+3)_

n

"

Sec. 39

THE REAL AND COMPLEX NUMBER FIELDS

375

Let (apnzk be a sequence of elements of K that converges in K to an element u, and let e e K} By definition, there exists m 2 k such that e

|a,,—u| n, then m

m—u—l

k-n+1

i=0

0 < am __ a" = 2 2—k(k+1)=

s 2,2 "HM



[( 71+1)( n+2)+n

—2 _

2



2—(1+(n+1))(:i+(n+2))

,202

( +1)( n+2)"'_"_1

— n

< 2—(n+1)(n+2) . 2 < 2—n(n+3)-

-

1

376

THE REAL AND COMPLEX NUMBER FIELDS

Chap. VII

Consequently, (a,,),,20 is a Cauchy sequence. For each n E N, let b

u

.n = 302n(n+1)-k(k+1).

Then b,, is an integer and a” = a—n(n+1)

for all 71 EN. Suppose that (a,,),,20 converged to the rational r. Then for every n e N and every e E Q:, there would exist q > n such that |r — (1,] < e, whence

244nm; I' “ an] S I’ — aql + la, — an| < e + consequently, as Ir — anl < e + 2‘”"‘+3’ for every e 6 Q* and every n EN, we would have |r —— anl g 2—""‘+3’

for all n EN. Let r =p/q wherep eZand q EN“. Since ‘2 _ 2—n(n+1)bn \ = Ir _ an' S 2—n(n+8),

q we have |2n(n+1)P _ qbnl S q.2n(n+1)—n(n+s) = 4-”q_

But as 2"‘”+1’p — qbn is an integer and as 4'"q < l for all n 2 q, we would, therefore, conclude that 2”"‘+1’17 — qbn = 0 and hence that r = a,, for all n 2 q, which is impossible since (1,,+1 > an. Theorem 39.9. A totally ordered field K is Dedekind ordered if and only if K is complete and archimedean ordered.

Proof. Necessity: By Theorem 39.3, K is archimedean ordered. Let (an),,2,, be a Cauchy sequence in K. By Theorem 39.7, there exists 2 e K: such that lafll S 2 and hence —z S a,, g z for all n 2 k. For each n 2 k, let A" =

{am: m 2 n}. Then 2 is an upper bound of A”, so A, admits a supremum b,,. Let B = {b,,: n 2 k}. If m 2 n, then A", S A“, and hence b", s b”. Since —z S a” g b” for all n 2 k, —-z is a lower bound of B, and consequently, B

admits an infimum c. We shall prove that (an),,2,, converges to e. Let e e Ki. Since (ante, is a Cauchy sequence, there exists 1) 2 k such that e

lam — anl |an—a,l=an—a,2e, a contradiction. Hence, as e/2 > 2"", either an > 2"" for all n 2m or an < —2"" for all n 2 m.

Let g be the relation on @(Q) satisfying a S b if and only if an s b,| for all n e N. It is easy to‘ verify that s is an ordering on @(Q) compatible with its ring structure. Lemma 40.3. The set ./V(Q) is a maximal ideal of the ring ?(Q). If he .A/(Q), if k 6 @(Q), and if 0 s k s h, then k eat/(Q). Proof. The second assertion is easy to prove. The multiplicative identity of ?(Q) clearly does not belong to JV(Q), but the zero element ofQ(Q) does, so .IV‘(Q) is a nonempty proper subset of %(Q). Let a, b EJV(Q) and let c e‘flQ). We shall show that a — b and ac belong to JV(Q). By Theorem 39.7, there exists a rational s 2 1 such that |c,,| s s for all n 2 0. Let e 6 Q1, and let m e Nbe such that Ianl < e/Zs and lbnl < e/2s for all n 2 m. Then I“ it

b n I S I“ n I

Ibnl < 25

I

2

S e

for all n 2 m, and e

Iancnl = Ianl Icnl < — - s < e 2s

for all n 2 m. Hence, a — b and ac converge to zero. Thus, JV(Q)_ is a

proper ideal of g(Q). To show that JV(Q) is a maximal ideal of ‘6(Q), let a be an ideal of g(Q) properly containing JV(Q). We shall show that a = €(Q). As a 3 JV(Q),

there is a sequence a E a that is not a null sequence. By Lemma 40.2, there exists m 6N such that 2"" < lanl for all n 2 m. Let h be the sequence defined by {0 if a,I ;E 0,

lifan=0. Then h,,= 0for alln 2m, sohe./V(Q). Letb = a +h. Thenbeasince JV(Q) C (I. Also, bu sé 0 for all n 2 0, so b is invertible in the ring Q", and its inverse is the sequence (b;1),,20. Moreover, [a > 2"", and hence,

Sec. 40

THE REAL AND COMPLEX NUMBER FIELDS

383

|b,,|—1 < 2’" for all n 2 m. Consequently, the inverse b—1 of b belongs to

?(Q), for if e 6 Q:, then there exists q 2 m such that lb” — b,| < 4"”e for all n 2 q, p 2 q, whence

lb? — b;1| = Ib;1(b, — b.)b;‘l = lb..l‘1 lb, — bnl Hul—1 < 2'"(4""e)2"” = e for all n 2 q, p 2 q. Therefore, a contains the invertible element b of the ring ?(Q), so a = @(Q) by Theorem 19.4. Thus, .A/‘(Q) is a maximal ideal of

?(Q)Let .9? be the quotient ring ?(Q)/ .xV'(Q). By Lemma 40.3 and Theorem 19.6, Q is a field. Let S be the relation on 9? satisfying at S [3 if and only if there exist Cauchy sequences a E at and b e [3 such that a S b. Lemma 40.4. Let a and b be elements of %(Q) belonging respectively to the elements at and ,3 of 92. Then at S (3 if and only if there exists h 6 JV(Q) such that a S b + h. If there exists m e N such that an S b,, for all n 2 m, then at S ,6.

Proof. Necessity: As at S ,3, there exist a’ e at and b’ e ,3 such that a’ S b’. Then a — a’ and b’ — b are null sequences, so if h = b’ -— b + a — a’, then h e ./V(Q) and

a=a'+(a—a’)Sb'+(a—a')=b+h. Sufliciency: If a S b + h where h 6 JV(Q), then a S 19 since a e at and b + h e ,3. Suppose, finally, that an S b” for all n 2 m. - If h is defined by h

an—bnforalln 0 again by Lemma 40.6. By Lemmas 40.2 and 40.6, if at aé 0, then either at > 0 or —ot ‘> 0. Conse-

quently, as S is compatible with the ring structure of Q, S is'a total ordering on Q. For each rational r, the sequence (r,,),,20 defined by r” = r for all n 2 0 clearly belongs to %(Q); we shall denote the corresponding element (r,,),,20 + ./V(Q) of Q by [r]. It is easy to verify that the function r~n—-> [r] is the unique isomorphism from the totally ordered field Q onto the totally ordered prime subfield of Q (Theorem 39.4).

Lemma 40.8. If at, [3 6 Q and if 0 S at < 5, then there exist natural numbers k and q such that at < [2—Vk] < ,3. Proof. Since at 2 0, there exist h 6 JV(Q) and a1 6 or such that h S a1. Let a = a1 —— h. Then a e at and a 2 0. Let b 6 fl. By Lemma 40.6, there exists

m EN such that b" — an > 2"" for all n 2 m. As a and b are Cauchy sequences, there exists q 2 m + 2 such that |a,, — up] < 24““) and

lb” — b,l < 2""‘4'3’ for all n 2:], p 2q. By Theorem 39.7, {jeNz on < 2—“(j — l) for all n 2 q} is not empty; let k be its smallest member. Then k — 1 2 l as 2"“(k — l) > 0, and at S [2—‘1(k — 1)] by Lemma 40.4.

Consequently, k — 1 e N, and therefore, 2-"(k — 2) S a, for some r 2g

Sec. 40

THE REAL AND COMPLEX NUMBER FIELDS

385

by the definition of k. If n 2 q, then

b" = br — (b, — b”) 2 b, — lb, — bu| 2 br _ 2—(m+2) > a, + 2"” — 2“"’+”) 2 2—¢(k — 2) + 3 . 2—(m+a)

22406—2) +3 -2-¢=2-¢(k+ 1). By Lemma 40.4, therefore, [2"(k + 1)] S ,3. Consequently, at S [2“(k — 1)] < [2—476] < [2“‘(k + 1)] S [3.

Theorem 40.1. (9?, +, -, g) is a Dedekind ordered field. Proof. By Theorem 39.2, it suflices to show that a subset B of .9? that is

bounded above and contains a strictly positive element admits a supremum in 9?. By Lemma 40.8, there exist natural numbers p and j such that [2— j] is an upper bound of B. For each n e N let K” = {k e N: [2"‘k] is an upper bound of B}. Since [2'”j] > [2"7’] if n < p and since [2"‘(2""fj)] = [2"j] if n 2 p, we

conclude that K" as (D for all n e N; let kn be the smallest member of K”. Since B contains a strictly positive element, k,, > 0 and consequently kn — l e N. If [2—”k,,] e B for some n 2 0, then [2"‘k,,] is clearly the supremum of B. Hence, we may assume that [2"“k,,] > [3 for all ,3 eB and all n 2 0. As in the proof of Theorem 39.9, if n 2 m, then

(1)

o g 24km — 24k, g 2-m,

so the sequence (27%"),20 is a Cauchy sequence. Let a = (2‘"k,,),,20, and let a be the corresponding element a + ./V(Q) of 9?. We shall first show that a is an upper bound of B. Let ,3 e B, and let b e ,3. If on < ,3, by Lemma 40.6, there would exist a natural number m such that b" — an > 2"" for all n 2 m;

as ,3 < [Z‘mkm], by Lemma 40.6, there exists q 6 N such that 2"“km — b” > 2" and, in particular, b” < 2""km for all n 2 q; hence, if n = max{m, q},

we would have 2"" < b” — an = b” — 2‘”k,, < 2""km — 2‘"k,, S 2""

by (l), a contradiction. Consequently, ac is an upper bound of B. To complete the proof, we shall show that if y is an upper bound of B, then on s y. If y < ac, then by Lemma 40.8, there would exist natural numbers k and q such that y < [24k] < or, and so by Lemma 40.6, there

Sec. 40

THE REAL AND COMPLEX NUMBER FIELDS

385

by the definition of k. If n 2 q, then bu = b, — (b, — b”) 2 b, — lb, _ bnl 2 b, _ 2—(m+2)

> a, + 2"" — 2‘4““) 2 24(k _ 2) + 3 . 2-(m+a)

2 2‘“(k — 2) + 3 - 2-4 = 2'“(k +1). By Lemma 40.4, therefore, [2“(k + 1)] s (3. Consequently,

at g [2*(k — 1)] < [2—qk] < [2*(k + 1)] g [3. Theorem 40.1. (9?, +, -, g) is a Dedekind ordered field. Proof. By Theorem 39.2, it suffices to show that a subset B of .9? that is bounded above and contains a strictly positive element admits a supremum

in 9?. By Lemma 40.8, there exist natural numbers p and j such that [2’ j] is an upper bound of B. For each n e N let K, = {k e N: [2—”k] is an upper bound of B}. Since [2—"j] > [2"j] if n < p and since [2‘”(2"—’j)] = [2"j] if n 2 p, we conclude that K” 7'5 (b for all n e N; let kn be the smallest member of K".

Since B contains a strictly positive element, kn > 0 and consequently kfl — l e N. If [2‘"k,,] E B for some n 2 0, then [2'”k,,] is clearly the supremum of B. Hence, we may assume that [2—”k,.] > B for all ,3 e B and all n 2 0. As in the proof of Theorem 39.9, if n 2 m, then

o s 2—mkm — 2-”k,, g 2—1",

(1)

so the sequence (2"‘k,,),,20 is a Cauchy sequence. Let a = (2‘”k,,),,20, and let at be the corresponding element a + ./V(Q) of 9?. We shall first show that at is an upper bound of B. Let ,3 e B, and let b 6 5. If 0!. < ,3, by Lemma 40.6, there would exist a natural number m such that b” — an > 2"" for all n 2 m;

as 13 < [2‘mkm], by Lemma 40.6, there exists q e N such that 2"”km — b” > 2" and, in particular, bn < 2""km for all n 2 q; hence, if n = max{m, q},

we would have

2-“ < b" — an = b" — 2—”k,. < 2""km — 2-”k,, s 2-“ by (l), a contradiction. Consequently, a: is an upper bound of B. To complete the proof, we shall show that if y is an upper bound of B, then at S y. If y < at, then by Lemma 40.8, there would exist natural

numbers k and q such that y < [24k] < ac, and so by Lemma 40.6, there

386

THE REAL AND COMPLEX NUMBER FIELDS

Chap. VII

would exist m 2 q such that 2—"kn — 2‘"k > 2"" and hence 2'”k,, — 2"" > 24k for all n 2 m. We would then have

2—macm — 1) = 2""km — 2-m > 2—«k, so [2""(km — 1)] > [Z‘Vk] > 9), an upper bound of B, whence km — 1 6K, as km — 1 EN, a contradiction of the definition of km. Thus, at = sup B,

and the proof is complete. Our construction of a Dedekind ordered field depends upon the existence of the rational field, which in turn depends upon our fundamental postulate that there exists a naturally ordered semigroup. Actually, the existence of a Dedekind ordered field conversely implies the existence of a naturally ordered semigroup (Exercise 40.5). In sum, there exists a Dedekind ordered field if and only if there exists a naturally ordered semigroup. Since there is an isomorphism f: r I—> [r] from the ordered field Q onto

the ordered prime subfield of 9?, by the corollary of Theorem 17.3, there exist a field R containing the field Q algebraically and an isomorphism g from R onto fl extending f. We define an ordering s on R by x s y if and only if g(x) s g(y). Equipped with this ordering, R is clearly a totally ordered field,

and g is an isomorphism from the ordered field R onto the ordered field 3.

Consequently, R is a Dedekind ordered field. For all x, y 6 Q, x s y if and only iff(x) Sf(y) asfis an isomorphism from the ordered field Q onto the

ordered prime subfield of 971?. Therefore, the total ordering of R induces on Q its given ordering. In §41 we shall see that any two Dedekind ordered fields are isomorphic and, moreover, that there is just one isomorphism from

one onto the other. In view of this, there is nothing arbitrary in our choice if we simply call R the ordered field of real numbers and its elements real numbers. Then Q is the prime subfield of R, and by Theorem 39.5, Q is dense

in R.

EXERCISES 40.1. (a) Prove that the relation S on @(Q) satisfying a s b if and only if a” S b, for all n 2 0 is an ordering compatible with the ring structure of «(Q). (b) Prove the second assertion of Lemma 40.3.

40.2. Verify that r l-—) [r] is the (unique) isomorphism from the ordered field Q onto the ordered prime subfield of .92. 40.3. If Q is the prime subfield of a totally ordered field K, then K is archirnedean

ordered if and only if every convergent sequence in the totally ordered field

Sec. 40

THE RFAL AND COMPLEX NUMBER FIELDS

387

Q is also a convergent sequence in the totally ordered field K. [Consider ("—971 21-]

40.4. Give an example of a null sequence a in Q such that an > 0 for all n 2 0. I"0.5. Let K be a Dedekind ordered field. A subset A of K is called a Peano set if OEA and if for all xeK, xeA implies that x + 1 EA. Let P be the

intersection of all the Peano sets in K. (a) The set P is a Peano set, P contains both 0 and 1, and 0 is the smallest element of P. (b) If xeP and if x _>_ 1, then x —1€P. [Showthat {xEPz if x 21,

then x — 1 EP} is a Peano set.] (c) If0 < x < 1, then x¢P. [Consider {x 61’: either x = 0 or x Z 1}.] (d) For all x,yEK, ifxEPand ifx 0, so

f(59‘) = —f((—S)x) = —f(—S)f(x) =f(S)f(x)-

394

THE REAL AND COMPLEX NUMBER FIELDS

Chap. VII

Therefore, f is a monomorphism from the ordered field K into the ordered field R. Thus, to within isomorphism, the only archimedean ordered fields are the subfields of the ordered field of real numbers. Theorem 41.8. If (G, +, s) is a commutative Dedekind ordered group

possessing no smallest strictly positive element, then for each strictly positive element a, the unique monomorphism j; from (G, +, S) into (R, +, S)

satisfying f;(a) = 1 is surjective and hence is an isomorphism. Proof. Let H be the range offi. Then (H, +, $3) is a Dedeldnd ordered group possessing no smallest strictly positive element as it is isomorphic to (G, +, s). We shall show that H = R. Let s be the infimum in R of all the

strictly positive elements of H. First, we shall prove that s = 0. Suppose that s > 0. Then 3 ¢ H, for otherwise, s would be the smallest strictly positive element of H. Hence, by the definition of s, there would exist x e H such that

s < x < 2s and also there would exist y e H such that s < y < x, whence x — y e H and 0 < x — y < s, a contradiction. Therefore, .9 = 0.

Next, we shall prove that H is a dense subset of R. Let u and v be real numbers such that u < v. Assume first that u 2 0. By what we have just proved, there exists x e H such that 0 < x < v — u. Let n be the smallest of those natural numbers k such that u < k.x. Then n > 0, so n — l e N, and therefore, (n — l).x S u < n.x, whence u 1, then 1 < flc + l) < c. Consequently, by Theorem 41.8, for

each real number a > 1 there is one and only one isomorphism f, from (R1, -, s) onto (R, +, g) satisfying j;(a) = 1. In our remaining discussion, we shall need the following easily proved facts about strictly increasing and strictly decreasing functions: (a) the inverse of a strictly increasing (strictly decreasing) function is strictly increasing

(strictly decreasing); (b) the composite of two functions, one of which is strictly increasing and the other of which is strictly decreasing, is strictly decreasing; (c) the composite of two functions, both of which are either

strictly increasing or strictly decreasing, is strictly increasing. Definition. Let a > 0. If a > 1, the logarithmic function to base a is the

unique isomorphism fl, from (R1, -, 3) onto (R, +, s) such that fl,(a) = 1. If 0 < a < l, the logarithmic function to base a is fl—1 0], where J is the

function from RT. onto R: defined by

J(x) = x—l. The logarithmic function to base a is denoted by log“. Clearly, I is an automorphism of the multiplicative group R1. Consequently, for each strictly positive real number a distinct from 1, log, is an isomorphism from the multiplicative group R1, onto the additive group R. Thus, for all x,

y e R: and all integers n, log, xy = log, x + log, y, log, x" = n log, x, and in particular,

log“ r1 = ——log,, x.

Sec. 41

THE REAL AND COMPLEX NUMBER FIELDS

395

from the ordered field K into the ordered field R is surjective and hence is an isomorphism. Proof. If a is a strictly positive element of K, then a/2 is also strictly positive and a/2 < a, so K has no smallest strictly positive element. The assertion, therefore, follows from Theorems 41.7 and 41.8.

Thus, to within isomorphism, R is the only Dedekind ordered field. By Theorem 41.9, the only monomorphism from the ordered field R into itself is the identity automorphism. Actually, the only monomorphism from the field R into itself is the identity automorphism (Exercise 41.4). Clearly, (Ri, -, s) is a Dedekind ordered group, for the supremum in R:

of a nonempty subset of Rithat is bounded above is simply its supremum in R. The strictly positive elements of the ordered group (R1, -, S) are, of course, the real numbers > 1. There is no smallest real number greater than

1, for if c > 1, then 1 < §(c + 1) < 6. Consequently, by Theorem 41.8, for each real number a > 1 there is one and only one isomorphism fa from (R1, -, s) onto (R, +, s) satisfying fl,(a) = 1. In our remaining discussion, we shall need the following easily proved facts about strictly increasing and strictly decreasing functions: (a) the inverse Of a strictly increasing (strictly decreasing) function is strictly increasing (strictly decreasing); (b) the composite of two functions, one of which is strictly increasing and the other of which is strictly decreasing, is strictly decreasing; (c) the composite of two functions, both of which are either

strictly increasing or strictly decreasing, is strictly increasing. Definition. Let a > 0. If a > 1, the logarithmic function to base a is the

unique isomorphism f, from (R1, -, s) onto (R, +, S) such that fl,(a) = 1. If 0 < a < l, the logarithmic function to base a is fa—1 o J, where J is the

function from R: onto R: defined by J(x) = x—l.

The logarithmic function to base a is denoted by log“. Clearly, J is an automorphism of the multiplicative group R1. Consequently, for each strictly positive real number a distinct from 1, log, is an isomorphism from the multiplicative group R1, onto the additive group R. Thus, for all x, y e R: and all integers n, log, xy = log, x + logay, log, x” = n log, x,

and in particular, log, x—1 = —log,, x.

396

THE REAL AND COMPLEX NUMBER FIELDS

Chap. VII

Furthermore, log, a = l, forifa >1, then log,a = l by definition, and if0 < a 1, then log, is a strictly increasing function. Since J is a strictly decreasing function, log, is strictly decreasing if0 < a < 1. Definition. Let a be a strictly positive real number distinct from 1. The exponential function to base a is the function log;— from R onto R*. We shall denote the exponential function to base a by exp,. As exp, is the

inverse of an isomorphism, exp, is itself an isomorphism from the additive group R onto the multiplicative group R3; Thus, for all x, y ER and all integers n,

eXPaOC + y) = (exPa x)(exn.y), “Paw-x) = (expa 3C)",

and in particular,

exp,(——x) = (exp, x)‘1. Furthermore, as log,a = 1,

exp, = a, so for all integers n, exp, n = (exp, 1)” = a”. For this reason, it is customary to denote exp, x by a” for all real numbers x. We also define l” to be 1 for all x e R. With this notation, am = ama”,

a” = 01”)”, and in particular, “—40 = (aw —1

for all x, y e R, all integers n, and all a e Rx. If a > 1, then exp, is strictly increasing since log, is, and if 0 < a < 1, then exp, is strictly decreasing since log, is.

Theorem 41.10. If a e R: and if a 75 1, then (1)

log, x" = y log, x

THE REAL AND COMPLEX NUMBER FIELDS

Sec. 41

397

for allx > 0and ally ER. Ifa, beRi, then

a“ = 01")"

(2) for all x, y e R, and

(3)

(ab)°’ = a“b°’

for all x e R. Proof. First, let a > 1, let x be a strictly positive real number distinct from 1, and let b = log, x. If x > 1 (respectively, 0 < x < 1), then b > 0 (res-

pectively, b < 0) and hence also b—1 > 0 (respectively, b—1 < 0), so both exp, and Mb—l are strictly increasing (respectively, strictly decreasing) functions. As a > 1, log, is strictly increasing. Hence, the function g defined by

g = Mb—l 0 log, 0 exp, is a strictly increasing function from R onto R. But g is also an automorphism

of the additive group R as it is the composite of isomorphisms. Thus, by Theorem 10.4, g is an automorphism of the ordered group (R, +, g), and

also

g(l) = b‘1 logax = 1. But by Theorem 41.6, there is only one automorphism of the ordered group (R, +, s) taking 1 into 1, and that automorphism is clearly the identity

automorphism. Thus, g is the identity automorphism, so as Mb-l = My, we conclude that log, 0 exp, = M,, whence log, x” = y log, x for all y ER. Also log,,1v = 0 = yloga 1. Therefore, (1) holds if a > 1, x> 0, andyeR. If0 0, there exists d > 0 such that d + d < c. (b) Ifa>0 and b>0, then a+b=b+a. [Ifnot, assume thatc=

(a + b) + [-(b + a)] > 0. If d + d < c, infer from the inequalities m.dSa 0 such that u(x) = x” for all x e R1. [Use Exercise 41.5.]

41.7. If n is a positive odd integer, thenfn: x Hx" is a strictly increasing permutation of R. [Use Theorems 41.2 and 41.11.]

41.8. (a) loglo 5 is irrational. (b) If a and b are integers > 1 and if there is a primep dividing a but not b, then log“ b and logb a are irrational. 41.9. Let a, b, and c be strictly positive real numbers distinct from 1. Show that (log, b)(log,, a) = l and that (log,I b)(log,, c) = log, c.

41.10. Find all strictly» positive real numbers x satisfying the following equalities: (a) log-,2 + 310g” 2 — 610g“ 2 = 0.

(b) (log2 x)(log,1 x)(log8 x) = 36.

(c) xV3= x/x?

(d) x” = (am

41.11. Let K and L be subfields of R. If K and L are isomorphic ordered fields, then K = L.

42. The Field of Complex Numbers. If K is a field over which X2 + l is irreducible, we shall denote by i a root of X‘ + 1 in a stem field K(i) of X2 + 1 over K. Then (1, i) is an ordered basis

of the K-vector space K(i), and addition and multiplication in K(i) are given

by («+fli)+(7+6i)=(a+y)+(fi+6)i, (a + 160(2) + 6i) = («7 — 135) + (a6 + 13WBy the corollary of Theorem 37.7, there is a unique K-automorphism a of K(i) such that o'(i) = —i, since i and —i are the roots of X2 + 1 in K(i). For

each 2 e K(i), we shall denote 0(2) by 2, which is called the conjugate of z. Thus, if z = at + fii where a, ,3 6K, then i = a — fii. Clearly, § = z for all z e K(i). If K is a totally ordered field, then X” + 1 is irreducible over K, for x’ +

1 2 l > 0 for all x EK by 10° of Theorem 17.6. In particular, X‘ + 1 is irreducible over R. By the corollary of Theorem 37.7, any two stem fields of X‘ + 1 over R are R-isomorphic; hence, there is nothing arbitrary in our

choice if we simply select some stem field C of X” + 1 over R and call its members “complex numbers.” We also select one of the‘two roots of X‘ + l

Sec. 42

THE REAL AND COMPLEX NUMBER FIELDS

399

41.5. If u is a function from R into R, then u is an automorphism of the ordered group (R, +, S) if and only if there exists b > 0 such that u(x) = bx for

all x E R. 41.6. If v is a function from R: into R1, then u is automorphism of the ordered group (R1, -, S) if and only if there exists b > 0 such that u(x) = x” for all x E R1. [Use Exercise 41.5.]

41.7. If n is a positive odd integer, thenfn: x |—>x" is a strictly increasing permutation of R. [Use Theorems 41.2 and 41.11.]

41.8. (a) loglo 5 is irrational. (b) If a and b are integers > 1 and if there is a prime p dividing a but not b, then log“ b and log, a are irrational. 41.9. Let a, b, and c be strictly positive real numbers distinct from 1. Show that

(log, b)(logb a) = l and that (log, b)(logb c) = log“ c. 41.10. Find all strictly. positive real numbers x satisfying the following equalities: (a) log, 2 + 3 logm2 — 610g“ 2 = 0.

(b) (log2 x)(log4 x)(logs x) = 36.

(c) xfi: VF.

(d) x” = (3.03.

41.11. Let K and L be subfields of R. If K and L are isomorphic ordered fields, then K = L.

42. The Field of Complex Numbers.

If K is a field over which X3 + 1 is irreducible, we shall denote by i a root of X‘ + 1 in a stem field K(i) of X3 + 1 over K. Then (1 , i) is an ordered basis

of the K-vector space K(i), and addition and multiplication in K(i) are given

by (“+l3i)+(7+5i)=(°¢+7)+(fi+5)i. (at + 130(9) + 5i) = (‘17 — 135) + (a6 + (3701'By the corollary of Theorem 37.7, there is a unique K-automorphism a' of K(i) such that u(i) = —i, since i and ——i are the roots of X“ + 1 in K(i). For

each 2 eK(i), we shall denote 0(2) by 2', which is called the conjugate of 2.

Thus, ifz = at + [3i where a, [3 e K, then 2' = a — (31'. Clearly, E = z for all z e K(i). If K is a totally ordered field, then X2 + 1 is irreducible over K, for x” + l 2 l > 0 for all x GK by 10° of Theorem 17.6. In particular, X2 + l is

irreducible over R. By the corollary of Theorem 37.7, any two stem fields of X‘ + 1 over R are R-isomorphic; hence, there is nothing arbitrary in our

choice if we simply select some stem field C of X2 + 1 over R and call its

members “complex numbers.” We also select one of the ‘two roots of X3 + l

400

THE REAL AND COMPLEX NUMBER FIELDS

Chap. VII

in C and denote it as above by i. The above equalities are then the familiar rules learned in elementary algebra for adding and multiplying complex numbers. If 2 = on + [31' where on, ,B e R, we recall that the real part of z is the real number at and is denoted by .9392, and that the imaginary part of z is the real number ,3 and is denoted by .fz; thus, 2 = .922 + £12. As we saw in Example

20.3, .9? and J are epimorphisms from (C, +) onto (R, +).

A fundamental property of C is that every nonconstant polynomial over C has a root in C, a fact sometimes called the “fundamental theorem of algebra.” 1 Definition. A field K is algebraically closed if every nonconstant polynomia Over K has a root in K. No finite field is algebraically closed. Indeed, if K is a finite field, the

polynomial 1 + H (X — a) is a nonconstant polynomial over K having no K

roots in K.

E

Theorem 42.]. The following conditions are equivalent for any field K:

1° K is algebraically closed. 2° The linear polynomials are the only irreducible polynomials over K.

3° Every nonconstant polynomial over K is a product of linear polynomials. Proof. Condition 1° implies 2° by Theorem 35.7. Since every nonconstant polynomial is a product of irreducible polynomials, 2° implies 3°. Since

every linear polynomial over K has a root in K, 3° implies 1°. In §47, we shall use Galois theory and Sylow’s First Theorem to prove

that C is algebraically closed. Basic to that proof is the fact that R is realclosed in the following sense: Definition. Let K be a totally ordered field. If a, b, and c are elements of K,

we shall say that c is between a and b if either a < c < b or b < c < a. A polynomial f over K changes signs between elements a and b of K if 0 is between f(a) andf(b). Finally, we shall say that K is real-closed if for every polynomial f over K and for all elements a and b of K, if f changes signs between a and b, then f has a root in K between a and b. For example, Q is not a real-closed field, for X' — 2 changes signs between 1 and 2 but has no root in Q between them. Clearly, f changes signs between a and b is and only iff(a)f(b) < 0.

Sec. 42

THE REAL AND COMPLEX NUMBER FIELDS

401

Theorem 42.2. (Bolzano’s Theorem for Polynomials) The totally ordered

field R of real numbers is real-closed.

Proof. Let a and b be real numbers such that a < b, and letfbe a polynomial over R that changes signs between a and b. It suflices to consider the case wheref(a) < 0 0 or both < 0, It changes signs between

a and b and hence has a root 0 between them. But by Theorem 35.10 and its corollary, Hence,

(X — axX — b)n=fh-

(c — axe — b)g(c)(Df)(c) =f(c)h(c) = 0.

Since (c —— a)(c — b)g(c) 75 0, therefore, (Df)(c) = 0. Corollary. Iffis a nonconstant polynomial over a real-closed totally ordered field K and if Df has n roots in K, then f has at most n + l roots in K.

Sec. 42

THE REAL AND COMPLEX NUMBER FIELDS

403

EXERCISES

42.1. If a complex number z is a root of a polynomial f over R, then 2' is also a root off. Iff is a polynomial over R of degree n and iffhas m real roots, multiplicities counted, then n E m (mod 2). 42.2. A complex number 2 is a pure imaginary number if z = ai for some real

number a. A monic cubic polynomial X3 + bX‘ + cX + d over R has one real and two pure imaginary roots if and only if c > 0 and d = bc. *42.3. Iffand g are polynomials over an algebraically closed field K such that the set of roots offis identical with the set of roots ofg and the set of roots of f + l is identical with the set of roots ofg + 1, thenf = g. [Iffhas r roots, iff + 1 has sroots, and ifdegf = n, show that tas at least 2n — s — r roots, multiplicities counted, and infer that f — g = 0.] n

1'42.4. Letp be a prime, let u > 1, and letf” =p + H (X — 3jp). (a) The polynomial fa, has In real roots. i=1 (b) The polynomial fat is irreducible over Q. [Use Eisenstein’s Criterion]

(c) There exist infinitely many prime polynomials over Q of degree n having n real roots. In the remaining exercises, K always denotes a real-closedfield. If x E K, then x is negative if x s 0, and x is strictly negative if x < 0. We shall employ the notation for intervals introduced in the exercises of §39, and in addition we shall

use the following terminology: Let f€- K[X] and let 0 e K. We shall say that f is increasing (decreasing) at c if there exist a, be X such that a < c < b and the

restriction of the polynomial function f~ to [a, b] is a strictly increasing (strictly decreasing) function. We shall say thatfhas a pure local minimum (pure local maximum) at c if there exist a, b e K such that a < c < b, the restriction off~ to [(1, c]

is strictly decreasing (strictly increasing), and the restriction off ” to [c, b] is strictly increasing (strictly decreasing). Elements c and d of K are consecutive roots off if cand dare roots off, ifc < d, andiffhasno roots between candd. Ifx,yeK, we shall say that x and y have the same sign if either x > 0 and y > 0 or x < 0 and y < 0 (or equivalently, if xy > 0) and that x and y have opposite signs if either

x >0andy 0. Apply (a) tog = f — f(c), and use Exercise 42.1 1 (b) in considering the four cases determined by the sign of (Dmf)(c) and the parity of m.] (c) (Budan’s Theorem) If a < b, the number of roots of f in 1a, b],

multiplicities counted, is V101) — Vra’) — for some k e N. [For each root c off or one of its nonzero derivatives,

determine V,(x) — V,.(y) where x and y are close to c and x < c < y.] 1|

‘42.l4. (Descartes’ Rule of Signs) For each polynomial f = 2 :1k of degree k=0

n > 0 over K, we define U+(f) to be the number of variations in sign of the sequence (099095", and we define U_(f) to be the number of variations in the sign of the sequence ((—1)"ock)os,,s,,. Then U_(f) = U+(f( —X)), the number of strictly positive roots off, multiplicities counted, is

(1+0) - 21‘ for somej e N, and the number of strictly negative roots off, multiplicities counted, is

U_(f) — 2k for some k E N. [Use Budan’s Theorem and Exercise 35.14 where a = 0.]

*42.15. Letf

20 kX7‘ be a polynomial of degree n over K.

(a) U+(f)o+ U_(f) 1, then c is a root of

each of f1, . . . ,a of multiplicity m — 1, and consequently, W,(y) = W,(x) — 1. [Use Exercise 42.9 to show that f(x) and Df(x) have opposite

signs] (c) If c is a root offk where k > 0 but not a multiple root off, then c is not a root of eitherfhl orfwd, and the number of variations in sign of the sequencefk_1(x), fk(x), fk+1(x) is identical with that- of the sequencef,,_1(y),

ficUXfi-ruw-

(d) If e is a simple root off, then W,(y) = W,(x) — 1, but if c is not a

root off, then W,(y) = Wf(x).

(e) (Sturm’s Theorem) If a0 < b0 and if neither a0 nor bo is a root of f, then the number of roots of f in [(10, b0] (multiplicities not counted) is

W,(ao) — W,(bo).

42.19. Use Sturm’s Theorem to determine the number of real roots ofthe following polynomials, and locate the roots between consecutive integers. (a) Xa—X+4. (b) X3+3X2—4X—2. (c) X3—7X—7.

(d) X4 + 4X3 — 4X +1. (e) 7X4 + 28X3 + 34X” + 12X + 1. (f) 7X‘ + 28X3 + 34X2 + 12X + 5.

CHAPTER VIII

GALOIS THEORY

In this chapter we shall continue our study of fields and present, in particular, an introduction to Galois theory. As we shall see, the theory offields provides solutions to certain classical problems of geometry and algebra, such as the problem of trisecting an angle by ruler and compass and the problem of solving a polynomial equation “by radicals.” 43. Algebraic Extensions Throughout, K is a field and E is an extension field of K. We recall the

following definitions given in §37. An element c of E is algebraic or transcendental over K according as c is an algebraic or transcendental element of the K-algebra E. If 0 is algebraic over K, the minimal polynomial of c over K is the minimal polynomial of the element c of the K-algebra E. The degree of the minimal polynomial of c over K is called the degree of c over K and is denoted by degK c. The field E is an algebraic extension of the field K if every element of E is algebraic over K, and E is a transcendental extension of K if E is not an algebraic‘extension of K. The simplest example of an element algebraic over K is any element of K itself. If c 6K, then the minimal polynomial of c over K is X — c. The complex number i is algebraic over R, and its minimal polynomial is X2 + l. The real number 13/5 is algebraic over Q, and its minimal polynomial is X3 — 2. Also \/2 + \/2 is algebraic over Q, and its minimal polynomial is X4 — 4X2 + 2. On the other hand, there exist real numbers that are transcen-

dental over Q (Exercise 43.12). Theorems of number theory assert, for example, that the real numbers qr and e are transcendental over Q.

Theorem 43.1. Let c be an element of E algebraic over K, and let g be a 407

CHAPTER VIII

GALOIS THEORY

In this chapter we shall continue our study of fields and present, in particular,

an introduction to Galois theory. As we shall see, the theory offields provides solutions to certain classical problems of geometry and algebra, such as the problem of trisecting an angle by ruler and compass and the problem of solving a polynomial equation “by radicals.” 43. Algebraic Extensions

Throughout, K is a field and E is an extension field of K. We recall the following definitions given in §37. An element c of E is algebraic or transcendental over K according as c is an algebraic or transcendental element of the K-algebra E. If c is algebraic over K, the minimal polynomial of c over K is

the minimal polynomial of the element 6 of the K-algebra E. The degree of the minimal polynomial of c over K is called the degree of c over K and is denoted by degK c. The field E is an algebraic extension of the field K if every element of E is algebraic over K, and E is a transcendental extension of K if E is not an algebraic'extension of K. The simplest example of an element algebraic over K is any element of K itself. If c 6K, then the minimal polynomial of c over K is X — c. The complex number iis algebraic over R, and its minimal polynomial is X2 + l. The real number 13/5 is algebraic over Q, and its minimal polynomial is

X3 — 2. Also \/2 + \/2 is algebraic over Q, and its minimal polynomial is X4 — 4X2 + 2. On the other hand, there exist real numbers that are transcendental over Q (Exercise 43.12). Theorems of number theory assert, for example, that the real numbers 11- and e are transcendental over Q.

Theorem 43.1. Let c be an element of E algebraic over K, and let g be a

407

408

GALOIS THEORY

Chap. VD]

monic polynomial over K satisfying g(c) = 0. The following conditions are equivalent: 1° g is the minimal polynomial of c over K.

2° For every nonzero polynomial f over K, iff(c) = 0, then g I f2 3° For every nonzero polynomial f over K, if f(c) = 0, then degg S

degf4° g is irreducible over K.

Proof. Condition 1° implies 2° and 3° by the definition of minimal polynomial, and 1° implies 4° by Theorem 35.5. Conversely, since g is a multiple

of the minimal polynomialfof c over K, condition 4° implies 1°, and each of 2° and 3° implies that deg g s degf, whence g = f as both are monic polynomials. Theorem 43.2. If e is an element of E algebraic over K whose degree over K is n, then 1° K(c) = K[c], 2° [K(c): K] = n, 3° (1, c, c’, . . . , c"'1) is an ordered basis of the K-vector' space K(c).

The assertion is a consequence of Theorems 35.3 and 35.4.

Theorem 43.3. If e is an element of E algebraic over K and if F is a subfield of E containing K, then c is algebraic over F, and the minimal polynomial of c over F divides the minimal polynomial of c over K in F[X]. Proof: Ifg is the minimal polynomial of c over K, then g e F[X] and g(c) = 0, so the assertion follows from Theorem 43.1.

We have determined the structure of the subfield K(c) of an extension field E of K generated by K and an element c algebraic over K; what is the structure of K(c) if c is transcendental over K? We first recall that K[X] is an

integral domain and hence admits a quotient field, which we shall denote by K(X), whose elements are quotients of polynomials over K. The field K(X) is often called the field of rational functions or the field of rational fractionsin one indeterminate over K. Let c be an element of E transcendental over K. Because the kernel of S, is the zero ideal, S, is an isomorphism from K[X] onto the subring K[c], and the subfield K(c) of E is a quotient field of

K[c], since K(c) is the smallest subfield of E containing K U {c} and afortiori is the smallest subfield of E containing K[c]. By Theorem 17.5, therefore, there is one and only one isomorphism from K(X) onto K(c) extending S,.

Hence, if c is transcendental over K, then K(X) and K(c) are K-isomorphic, and moreover, there is a unique isomorphism from K(X) onto K(c) whose

Sec. 43

GALOIS THEORY

409

restriction to K[X] is So. In the sequel, we shall be almost exclusively con-

cerned with algebraic extensions of fields. Theorem 43.4. If E is an extension of a field K, then E is an algebraic extension

of K if and only if every subring of E containing K is a field.

Proof Necessity: Let A be a subring of E containing K. If e is a nonzero element of A, then A 2 K[c], which is a field by Theorem 43.2, and hence, c—1 e A. Therefore, A is a field. Sufficiency: Let c be a nonzero element of E.

Then K[c] is a subring of E containing K and hence is a field by hypothesis. Therefore, there is a polynomial f e K[X] such that f(c) = c'l, whence g(c) = 0 where g = Xf — 1. Consequently, c is algebraic over K. Theorem 43.5. If E is a finite extension of K, then E is an algebraic extension of K, and the degree over K of every element of E divides [E: K].

Proof. By Theorem 35.2, E is an algebraic extension of K. For each c e E,

[E1 K] = [E1 K(C)][K(C): K] by the corollary of Theorém 37.4, so as degK c = [K(c): K], the degree of c over K divides [E: K].

Corolla'ry. If c is an element of E algebraic over K, then K(c) is an algebraic extension of K.

The assertion is an immediate consequence of Theorems 43.2 and 43.5. Definition. An extension E of a field K is a simple extension of K if there exists c e E such that E = K(c). The following theorem gives a necessary and sufficient condition for a finite extension to be simple. Theorem 43.6. (Steinitz) If E is a finite extension of a field K, then E is a simple extension of K if and only if there are only a finite number of subfields of E containing K. Proof: Necessity: By hypothesis, there exists c e E such that E = K(c). Let fbe the minimal polynomial of c over K. Let L be a subfield of E containing K, and let m-l

g= X'"+ k=0 EakX" be the minimal polynomial of c over L. By Theorem 43.3, g divides f in

Sec. 43

GAmis THEORY

409

restriction to K[X] is S,. In the sequel, we shall be almost exclusively con-

cerned with algebraic extensions of fields. Theorem 43.4. If E is an extension of a field K, then E is an algebraic extension

of K if and only if every subring of E containing K is a field.

Proof. Necessity: Let A be a subring of E containing K. If c is a nonzero element of A, then A 2 K[c], which is a field by Theorem 43.2, and hence,

0‘1 e A. Therefore, A is a field. Sufficiency: Let c be a nonzero element of E. Then K[c] is a subring of E containing K and hence is a field by hypothesis. Therefore, there is a polynomial f e K[X] such that f(c) = c—l, whence g(c) = 0 where g = Xf — 1. Consequently, c is algebraic over K. Theorem 43.5. If E is a finite extension of K, then E is an algebraic extension

of K, and the degree over K of every element of E divides [E: K]. Proof: By Theorem 35.2, E is an algebraic extension of K. For each c e E,

[E: K] = [E: K(C)][K(C)I K] by the corollary of Theorem 37.4, so as degK c = [K(c): K], the degree of c over K divides [E : K].

Corolla'ry. If 6 is an element of E algebraic over K, then K(c) is an algebraic extension of K.

The assertion is an immediate consequence of Theorems 43.2 and 43.5.

Definition. An extension E of a field K is a simple extension of K if there exists c e E such that E = K(c). The following theorem gives a necessary and sufficient condition for a

finite extension to be simple.

Theorem 43.6. (Steinitz) If E is a finite extension of a field K, then E is a simple extension of K if and only if there are only a finite number of subfields of E containing K.

Proof. Necessity: By hypothesis, there exists 0 e E such that E = K(c). Let f be the minimal polynomial of c over K. Let L be a subfield of E containing K, and let m—l

g = X'” + k-O 2 £1k be the minimal polynomial of c over L. By Theorem 43.3, g divides f in

410

GAIDIS THEORY

Chap. V111

L[X] and hence in E[X]. Let 0

L0 = K(ao, . u . , 0W1).

Then Lo 5; L, g e Lo[X], and g is certainly irreducible over L0 as g is irreducible over L. Consequently, by Theorem 43.1, g is the minimal polynomial of c over Lo. Hence,

m = degg = [E: L0] = [E: L][L: L0] = m[L: Lo] by Theorem 43.2 and the corollary of Theorem 37.4, so [Lz L0] = 1 and therefore, L = Lo. Thus, the only subfields of E containing K are the fields m—l

K(ao, . . . , a,,,_1) where X“ + 2 akX" is a divisor off in E[X]. Since E[X] k=0

is a principal ideal domain and hence a unique factorization domain, f has only a finite number of monic divisors in E[X]. Consequently, there are only a finite number of subfields of E containing K. Sufliciency: If K is a finite field, then E is also a finite field since E is finite

dimensional over K. The multiplicative group E* is, therefore, cyclic by the corollary ofTheorem 38.7, and consequently, E = K(c) where c is a generator of E“. We shall assume, therefore, that K is an infinite field. Let m be the

largest of the integers [K(a): K] where a GE, and let c 6E be such that [K(c): K] = m. We shall obtain a contradiction from the assumption that

K(c) ye E. Suppose that b were an element of E not belonging to K(c). By hypothesis, the set of all the fields K(bz + c) where z e K is finite. Since K is infinite, therefore, there exist x, y e K such that x yé y and K(bx + c) =

K(by + c). Let d = bx + c. Then bx + c and by + c belong to K(d), so b e K(d) as

b = (x — y)‘1[(bx + c) — (by + 0)], and hence, also 0 e K(d) as c=(bx+c)—bx.

Therefore, K(d) is a field properly containing K(c), whence

[K001 K] = [K(d)= K(C)][K(C)= K] > m, a contradiction of the definition of m. Therefore, K(c) = E, and the proof is complete.

Sec. 43

GALOIS THEORY

411

Corollary. If E is a finite simple extension of K, then every subfield L of E

containing K is also a finite simple extension of K.

Theorem 43.7. If cl, . . . , c," are elements of an extension E of K that are algebraic over K, then K(c1, . . . , c”) is a finite and hence algebraic extension

of K. Proof. Let S be the set of all integers m e [1, n] such that K(cl, . . . , cm) is a finite extension of K. Then 1 e S by Theorem 43.2. If m e S and m < n, then m + l e S, for cm+1 is algebraic over K and afortiori over K(c1, . . . , cm), so,

since K(c1, . . . , cm“) = [K(c1, . . . , cm)](c,,,+1), K(c1, . . . , cm“) is a finite extension of K(c1, . . . , cm) by Theorem 43.2 and hence is also a finite extension of K by the corollary of Theorem 37.4. By induction, therefore, n e S. Corollary. A splitting field of a polynomial over K is a finite extension of K.

Theorem 43.8. (Transitivity of Algebraic Extensions) If E is an extension of K and if F is a subfield of E containing K, then E is an algebraic extension of K if and only if E is an algebraic extension of F and Fis an algebraic extension of K. Proof. The condition is clearly necessary by Theorem 43.3. Sufiiciency: Let c e E, let m—1

g = X“ + Z akX" TC=0

be the minimal polynomial of c over F, and let L = K(ao, . . . , a,,,_1). Then g e L[X], so c is algebraic over L, and hence, L(c) is a finite extension of L by

Theorem 43.2. But L is a finite extension of K by Theorem 43.7. Hence, L(c) is a finite extension of K by the corollary of Theorem 37.4, so c is

algebraic over K by Theorem 43.5. Theorem 43.9. Let E be an extension of K. The set A of all elements of E algebraic over K is an algebraic extension field of K, and every element of E

algebraic over A belongs to A. Proof. Let x and y be nonzero elements of A. By Theorem 43.7, K(x, y) is an algebraic extension of K containing x —- y, xy, and r1, so those elements are all algebraic over K and hence belong to A. Therefore, A is a field, and consequently, A is an algebraic extension of K. If c is an element of E algebraic over A, then A(c) is an algebraic extension of A by the corollary

of Theorem 43.5 and hence is also an algebraic extension of K by Theorem 43.8, so c is algebraic over K and therefore belongs to A.

Sec. 43

GALOIS THEORY

411

Corollary. If E is a finite simple extension of K, then every subfield L of E containing K is also a finite simple extension of K. Theorem 43.7. If c1, . . . , c,, are elements of an extension E of K that are

algebraic over K, then K(cl, . . . , c") is a finite and hence algebraic extension of K.

Proof. Let S be the set of all integers m e [1, n] such that K(cl, . . . , 0m) is a finite extension of K. Then 1 e S by Theorem 43.2. If m E S and m < n, then m + l e S, for cm+1 is algebraic over K and afortiori over K(cl, . . . , cm), so,

since K(cl, . . . , em“) = [K(cl, . . . , cm)](cm+1), K(cl, . . . , cm“) is a finite extension of K(cl, . . . ,cm) by Theorem 43.2 and hence is also a finite extension of K by the corollary of Theorem 37.4. By induction, therefore, n e S.

Corollary. A splitting field of a polynomial over K is a finite extension of K. Theorem 43.8. (Transitivity of Algebraic Extensions) If E is an extension of K and if F is a subfield of E containing K, then E is an algebraic extension of K if and only if E is an algebraic extension of F and Fis an algebraic extension of K.

Proof. The condition is clearly necessary by Theorem 43.3. Sufl'iciency: Let c e E, let m—l

g = X'” + 2 akX" k—O

be the minimal polynomial of c over F, and let L = K(ao, . . . , am_1). Then g e L[X], so c is algebraic over L, and hence, L(c) is a finite extension of L by Theorem 43.2. But L is a finite extension of K by Theorem 43.7. Hence,

L(c) is a finite extension of K by the corollary of Theorem 37.4, so c is

algebraic over K by Theorem 43.5. Theorem 43.9. Let E be an extension of K. The set A of all elements of E algebraic over K is an algebraic extension field of K, and every element of E algebraic over A belongs to A. Proof. Let x and y be nonzero elements of A. By Theorem 43.7, K(x, y) is

an algebraic extension of K containing x — y, xy, and x—l, so those elements

are all algebraic over K and hence belong to A. Therefore, A is a field, and

consequently, A is an algebraic extension of K. If c is an element of E algebraic over A, then A(c) is an algebraic extension of A by the corollary of Theorem 43.5 and hence is also an algebraic extension of K by Theorem 43.8, so c is algebraic over K and therefore belongs to A.

412

GALOIS THEORY

Chap. VIII

We recall from §42 that a field K is algebraically closed if every nonconstant polynomial over K has a root in K. Theorem 43.10. A field K is algebraically closed if and only if the only algebraic extension of K is K itself.

Proof. Necessity: Let E be an algebraic extension of K, and let c be any element ofE. The minimal polynomial g of c over Kis irreducible byTheorem 43.1 and hence is X — a for some a e K by Theorem 42.1. Therefore, c — a = g(c) = 0, so c = a GK. Consequently, E = K. Sufficiency: If K were

not algebraically closed, there would exist a nonconstant polynomial f over K having no roots in K, and a splitting field off over K would then be an algebraic extension of K properly containing K by the corollary of Theorem

43.7. EXERCISES 43.1. Determine the minimal polynomials over Q of the following numbers:

(an/5+5

(b)i'/§—s

(c) «Aux/E

(d)\/2+1/§

(e)x/§+x‘/§

(f) V5+Vi

Find the other roots in C of the minimal polynomials. [Make a judicious guess, and determine the minimal polynomial of your guess.] 43.2. Determine the minimal polynomials over Q(1/2) of the numbers ofExercise

43.1.

*43.3. For every integer n, the cosine and sine of n degrees are algebraic over Q. [Use Theorem 43.8.] What is the minimal polynomial over Q of 2 cos 15°?

of 2 cos 12°? 43.4. If B is a set of elements of an extension field E of K each of which is algebraic over K, then K(B) is an algebraic extension of K. 43.5. Show that the following extensions of Q are simple: Q(\/2, V5), Q(\/3, i),

Q (Vi, «3/3). 43.6. If E is a splitting field of a polynomial of degree n over K, then [E: K] divides n!. *43.7. If f is an irreducible polynomial of degree 71 over K and if E is a finite extension of K such that [E: K] and n are relatively prime, thenfis irreducible over E. [Compute [E(c): K] in two ways where c is a root off in a

splitting field off over E.]

Sec. 43

GADOIS THEORY

413

I"43.8. (a) If K is a subfield of R and if a and b are positive integers such that neither \/a, 1/3, nor Vab belongs to K, then x/l; does not belong to

K(\/3). (b) If a1, . . . , an are integers > 1 that are not squares of other integers

and if a; and a, are relatively prime whenever i 75 j, then

«Emu/Z, . . . , WE). [Proceed by induction and use (a).] (c) If a1, . . . , a” are integers > 1 that are not squares of other integers

and if a‘ and a, are relatively prime whenever i at: j, then the set of the 2" numbers (air . . . 11;»)1/2 where Si is either 0 or 1 for all j e [1, n] is linearly

independent over Q. (d) A strictly positive integer is square-free if it is not divisible by the square of any integer > 1. The set of all the numbers V; where a is a square-free integer is linearly independent over Q. (e) The field A of complex numbers algebraic over Q is not a finite extension of Q.

“3.9. If fe R[X] and if aeR, there exists a real number M such that for all real numbers x, if 0 < Ix —al n/m is

a permutation of the Set of positive divisors of n, by Theorem 23.9, the only subgroups of I‘ are the cyclic subgroups [am] where m | n, and [am] has order n/m. Let Fa... be the fixed field of [am]. By Theorem 45.6, the only subfields

of E containing K are the fields Fem, and

[.s K] = (P = [6"‘D = nKit/In) = so E1" has q’" elements. By Theorem 38.3, the only. subfield of E having q’” elements is {x 6E: x4" = x}, so F."- is the only subfield of E having q’" elements.

Sec. 45

GALOIS THEORY

431

Replacing E by q where r In and observing that [FQA K] = r, we see that if Fa, contains Fa”, then m | r, and conversely, if m | r, then Fa. contains a subfield of q’" elements, which must be F,“ as that is the only subfield of E

having q’" elements.

Applying Theorem 45.8 to the case where K is a prime field, we obtain the following corollary: Corollary. Let E be a finite field having 12" elements, where p is a prime. For

each positive divisor m of n there is one and only one subfield F”... ofE having p’" elements. The fields F”... where m | n are the only subfields of E, and

17,... S F,, ifand only ifm | r.

EXERCISES 45.1. Prove 2° of Theorem 45.1. 45.2. What is the closure of Q in R? [Use Exercise 41.4.]

45.3. Let E be a finite Galois extension of K, let L1 and L, be subfields of E

containing K, and let A1 and A, be subgroups of the Galois group of E over K. (a) Describe L1(L2)A and (L1 n L2)A in terms of L1A and L}.

(b) Describe (A1 0 AQV and the fixed field of the subgroup generated by A1 U A2 in terms of A17 and A27.

45.4. Let E be a finite field having p” elements where p is a prime, and let K and L be subfields of E having 11’ and 11‘ elements respectively. How many elements does K n L have? How many elements does K(L) have? 45.5. Let a and b be positive integers such that none of V3, 1/b, and \/a—b is

rational, and let E be the splitting field in C of (X2 — a)(X’ — b). Describe

the automorphism grOup I‘ of E over Q. [Use Exercise 43.8(a).] Construct a diagram like that of Figure 18 pairing subgroups of I‘ with their corresponding fixed fields.

*45.6. Let E be the splitting field in C of X4 -— 2 over Q. Describe the automorphism group I‘ of E over Q. Construct a diagram like that of Figure 18 pairing subgroups of I‘ with their corresponding fixed fields. [Observe that E = Q(\4/§, i), and model your discussion after that of Example 45.4.] 45.7. If a is an integer that is not the cube of an integer, then Q(%) is not a Galois extension of Q.

45.8. Let E be a finite Galois extension of K of degree n, and let {01, . . . , an}

432

GALOIS THEORY

Chap. VIII

be the Galois group of E over K. For each e, we define the trace IVE/KG) ofx and the norm NE,K(x) of x over K by

IVE/K0) = 01(x) + . . . + 0,,(x), NE/K(x) = 010:) . . . 6,,(x). (a) For all x e K, TrE,K(x) and NE/K(x) belong to K. (b) For all x,yeE, TrE,K(x + y) = nE,K(x) + Irmxo) and

IVE/KW) = NEIK(x)NE/K(y)'

(c) The function T’E/K is a nonzero linear form on the K-vector space E.

(d) What are the trace and norm of a complex number a + bi over R? of an element a + b\/§ of Q(\/§) over Q ? of an element a + bx/i + c\/§ + (11/3

of Q(1/§, 1/5) over Q?

46. Separable and Normal Extensions

Here we shall characterize finite Galois extensions of fields and determine further relationships between subfields of a finite Galois extension and subgroups of its Galois group which will prove useful in discussing polynomial equations.

Definition. A prime polynomialf over a field K is separable over K if there is an extension field E of K such that for some sequence («0193, of distinct elements of E,

f= fi (X — a.) k=1

A nonconstant polynomial g over K is separable over K if every prime factor of g in K[X] is separable over K.

Thus, a prime polynomial over K of degree n is separable over K if and only if it has n roots in some extension field of K. Ifg is a nonconstant polynomial over K all of whose roots in a splittingfield E ofg are simple, then g is separable over K; indeed, iff is a prime factor of g in K[X], then f divides g in E[X] and so f is the product of distinct linear polynomials in E[X]. However, a polynomial may be separable over K and yet have multiple roots in an extension field; for example, (X2 + 1)2 is separable over Q. Clearly, ifg is a separable polynomial over K and if h is a nonconstant polynomial over K dividing g in K[K], then h is separable over K. Theorem 46.1. Let f be a prime polynomial over a field K. The following

conditions are equivalent:

GALOIS THEORY

Sec. 46

433

1° f is separable over K.

2° ne 0.

3° Every root off in any extension field of K is a simple root.

Proof. Condition 1° implies 2°, for if Df were the zero polynomial, then every root off in any extension field of K would be a multiple root by Theorem 35.11. To show that 2° implies 3°, let c be a root offin an extension

field E of K. If c were a multiple root off, then c would also be a root of Df by Theorem 35.11, and hence, either Df = 0 or degf g deg Df= degf — 1 by Theorem 43.1, a contradiction. To see that 3° implies 1°, we need only

consider a splitting field off over K. Of importance later is the following fact: If L is an extension of afield K and ifg is a nonconstant polynomial separable over K, then g is also separable over L. For let E be a splitting field ofg over L, and let h be a prime factor of g in L[X]. Then in L[X], h divides some prime factor f of g in K[X] by Theorem 22.9 and (UFD 3’). Consequently, as every root off in E is simple by Theorem 46.1, every root of h in E is also simple. Therefore, as E is a

splitting field of g over L, h is a product of distinct linear polynomials in

E[X] and hence is separable over L. Theorem 46.2. Every nonconstant polynomial over a field K of characteristic zero is separable over K. Proof. By Theorem 46.1, it suflices to prove that every monic nonconstant polynomial over K has a nonzero derivative. If 11—].

f= X” + 20:k Ic=0

where n 2 1, then n—l

Df= n.X”'1 + 2(k.a,,)X"‘1, k=1

which is not the zero polynomial since ml at 0. If the characteristic of K is a prime p, however, there may well exist a prime polynomial over K whose derivative is the zero polynomial. For example, let K = Z,(X), the field of rational fractions over 2,. The polynomial h = Y2 — X e K[Y] has no roots in K, for iff and g were nonzero

polynomials over Z2 such that (f/g)* — X = 0, then f2 = Xg”, but the degree off2 is even and that of Xg2 is odd, a contradiction. Consequently, h is irreducible over K, but Dh = 2.Y = 0 as the characteristic of K is 2.

Sec. 46

GALOIS THEORY

433

1° f is separable over K.

rmio 3° Every root off in any extension field of K is a simple root.

Proof. Condition 1° implies 2°, for if Df were the zero polynomial, then every root off in any extension field of K would be a multiple root by

Theorem 35.11. To show that 2° implies 3°, let c be a root offin an extension field E of K. If c were a multiple root off, then c would also be a root of Df by Theorem 35.11, and hence, either Df= 0 or degf s deg Df = degf — 1 by Theorem 43.1, a contradiction. To see that 3° implies 1°, we need only

consider a splitting field off over K.

Of importance later is the following fact: If L is an extension of afield K and ifg is a nonconstant polynomial separable over K, then g is also separable over L. For let E be a splitting field ofg over L, and let h be a prime factor of g in L[X]. Then in L[X], h divides some prime factor f of g in K[X] by Theorem 22.9 and (UFD 3’). Consequently, as every root off in E is simple by Theorem 46.1, every root of h in E is also simple. Therefore, as E is a splitting field of g over L, h is a product of distinct linear polynomials in E[X] and hence is separable over L.

Theorem 46.2. Every nonconstant polynomial over a field K of characteristic zero is separable over K. Proof. By Theorem 46.1, it suffices to prove that every monic nonconstant polynomial over K has a nonzero derivative. If ”—1

f: X" + ZakX" k=0

where n 2 1, then 11—1

w=mm4+zmmfli k=1

which is not the zero polynomial since n.1 7E 0.

If the characteristic of K is a prime p, however, there may well exist a

prime polynomial over K whose derivative is the zero polynomial. For example, let K = Z2(X), the field of rational fractions over 22. The polynomial h = Y2 — X e K[Y] has no roots in K, for iff and g were nonzero

polynomials over Z2 such that (if/g)2 -— X = 0, then f2 = n, but the degree off2 is even and that of Xg” is odd, a contradiction. Consequently, h is irreducible over K, but Dh = 2.Y = 0 as the characteristic of K is 2.

GADOIS THEORY

434

Chap. VIII

In general, when does a polynomial f of degree n over a field K whose characteristic is a prime 12 have a zero derivative? By definition, if

f = i “L'Xk’ 70=0

then

Df = i (k.a,,)Xk-1, k=1

so Df= 0 if and only ifk.oc,c = 0 for all k E [1, n]. If at 9'5 0 and if keZ,

then k.ot = 0 if and only if p I k by Theorem 21.7. Consequently, if Df= 0, then och = 0 whenever p xl’ k, so f= ocwX'” + «(Fm/WP“? + . . . + oc,X’ + an where rp = n. Conversely, if 1'

f =20 fikp:

then

_

Df = Z p.(k.fl,,)X’“"1 = 0. 76-1

In sum, Df = 0 if and only iff belongs to the subdomain K[X1’] of K[X]. Definition. An element a of an extension field E of a field K is separable over

K if a is algebraic over K and if the minimal polynomial of a over K is separable over K. The extension field E is a separable extension of K if every

element of E is separable over K. By a previous observation, ifa is separable over K and ifL is an extension field of K contained in E, then a is separable over L, since by Theorem 43.3

the minimal polynomial of a over L divides the minimal polynomial of a over K in L[X].

By Theorem 46.2, every algebraic extension of a field of characteristic zero is a separable extension. The following two theorems show that every algebraic extension of a finite field is also a separable extension. Theorem 46.3. If K is a field whose characteristic is a prime p, then every

algebraic extension of K is a separable extension of K if and only if the function a: at n—> a." is an automorphism of K. Proof. By Theorem 38.2, 0' is a monomorphism from K into K.' Therefore, we shall show that every algebraic extension of K is separable if and only if a

435

GAmrs THEORY

Sec. 46

is surjective. Necessity: Let {3 e K, let E be a splitting field of X” — ,3 over K, and let 0! be a root of X” — ,3 in E. Then

X’—fi=X”——oc”=(X—-oc)". Consequently, if h is a prime factor of X" — fl in K[X], then h divides (X — on)” in E[X]; hence, h = (X — a)” for some k e [1, p], but, since every

root of h in E is simple by Theorem 46.1, we have k = l and h = X — at. As h e K[X], therefore, cc 6 K. Hence, a is surjective. Sufficiency: We shall show that iff is a nonconstant polynomial over K

whose derivative is the zero polynomial, then f is reducible over K. Since Df = 0, there exist [30, . . . , [3, e K such that

f=ifikxkp k=0

as we saw above. Because or is suijective, there exists akeK such that at,” = [3,, for each k e [0, r]. Therefore,

#2..»k = (w), 1'

T

H

H

9

so f is reducible over K. Definition. A field K is perfect if every algebraic extension of K is separable. Theorem 46.4. All fields of characteristic zero, all finite fields, and all alge-

braically closed fields are perfect. Proof. By Theorem 46.2, all fields of characteristic zero are perfect. If K is a

finite field whose characteristic is a prime p, then 0': act—m” is an auto-

morphism of K by Corollary 38.2.2, so K is perfect by Theorem 46.3. The only algebraic extension of an algebraically closed field K is K itself by Theorem 43.10, so an algebraically closed field is perfect. Definition. An extension field E of a field K is a normal extension of K if E is an algebraic extension of K and if every prime polynomial over K that has a root in E is a product of linear polynomials in E[X]. For example, Q(\3/2) is not a normal extension of Q, for X3 — 2 is a prime polynomial over Q that has a root in Q(\3/2) but is not a product of linear polynomials over Q(\"/2), as we saw in Example 45.3.

Sec. 46

GALOIS THEORY

435

is surjective. Necessity: Let B GK, let E be a splitting field of X1’ —— [3 over K, and let on be a root of X" — [3 in E. Then

XP—fi=X1°—oc’=(X—ac)’. Consequently, if h is a prime factor of X” — [3 in K[X], then h divides (X — at)? in E[X]; hence, h = (X — 1. By 1° of Theorem 45.6, E is a Galois extension of K(i). By Sylow’s First

Theorem, the Galois group P1 of E over K(i) contains a subgroup A1 of order 2"". Let L1 be the fixed field of A1. Then

[L1:K(i)] = (1“,: L9) = (r1: A1) = 2 by Theorem 45.6. Let h, be the minimal polynomial over K(i) of an element belonging to L1 but not to K(i). Then h1 is an irreducible quadratic polynomial over K(i). By hypothesis, h1 has a root in K(i), a contradiction. Therefore, n = 1. Hence, [Ez K(i)] = 1, so E = K(i), and consequently, f has a root in K(i). To complete the proof, let g be a nonconstant monic polynomial over

K(i), and let g be the polynomial such that for each j e N the coefficient of X’ in g is the conjugate of the coefficient a, of X5 in g. Then gg' is a polynomial over K, for if a, = a, + [3,1' for eachj E N, then the coeflicient of X’" in g? is Z (“5 + 1310(“7c — fiki) = ( 2 “1% + fiifik) + ( 2 fifik — aifib)i l+k=m

i+k=m

1+k—m

for all m e N, and M

m

2 fljak —' “#37: = Zfiflm—i —Z°‘m—k k = 0-

r=o

i+k-m

k=0

By what we have just proved, therefore, g? has a root 2 in K(i), and hence, z is a root either of g or of E. In the latter case, i is a root of g, for

g(2) = i akz" = z 5,,2'7‘ = (25,»)— k=0

Ic=0

k=o

=EE=6=Q Every nonconstant monic polynomial over K(i) thus has a root in K(i), so

K(i) is algebraically closed. An element a of a field K is a square of K if there exists x e K such that x2 = a, and any such element x is called a square root ofa. By 10° of Theorem 17.6, if a is a square of a totally ordered field K, then a 2 0. Consequently, there is no ordering on the field K(i) converting K(i) into a totally ordered field, for —l is a square ofK(i). In particular, the field C ofcomplex numbers cannot be made into a totally ordered field. If K is a totally ordered field and if c” = a, then 0 and —c are the only square roots of a, for X3 — a has at most two roots in K. Since exactly one

444

GALOIS THEORY

Chap. VIII

of c and —c is positive, therefore, every square of a totally orderedfield ha: exactly one positive square root. If K is a totally ordered field, then X3 + 1 is irreducible over K, as we observed in §42.

Theorem 47.2. Ifevery positive element of a totally ordered field K is a square of K, then every element of K(i) is a square of K(i) and every quadratic

polynomial over K(i) has a root in K(i). Proof. We shall first show that every element of K is a square of K(i). Indeed, if or. 2 0, then at is a square of K and afortiori of K(i); if at < 0, then —oc > 0, so there exists y e K such that y” = —oc, and consequently, (yi)’l = at. Next, letz= oc+fiiwhere ac, fieK and flyéO. Then u“+fi‘>0,so there exists 7/ > 0 such that y“ = on“ + 13‘. Since 71* > a”, we have 7! > |a|, for otherwise, we would have 91 s locl and hence y” S y la] s lal’ = a‘, a contradiction. Therefore, 7+“27—la‘|>0,

so by hypothesis there exists A > 0 such that 2’ = flat + 71). An easy calculation then establishes that (A + Mi)2 = z where p = [3/21.

Let aX‘ + bX + c be a quadratic polynomial over K(i). By the preceding paragraph, there exists d e K(i) such that d” = b2 — 4ac. An easy calculation shows that (d — b)/2a is then a root of aX’ + ‘bX + c. Theorem 47.3. IfK is a totally ordered field and if K(i) is algebraically closed, then every positive element of K is a square of K, and the irreducible polynomials over K are the linear polynomials and the quadratic polynomials acX‘+fiX+ywherefi“—4ocy 0. As K(i) is algebraically closed, there exist 1., ,4 6K such

that l + [if is a root of X‘ — or. Then 42—13:“: 2}.” = 0,

since (1, i) is an ordered basis of K(i) over K. Because the characteristic of K is zero, either A = 0 or ,u = 0; if 2. = 0, then at = —,u’, which is impossible as a > 0; therefore, {i = 0, so cc = A” and hence is a square of K.

Let f be an irreducible polynomial over K. Since K(i) is algebraically closed, there is a root c off in K(i), so

[K(C)= K] S [K(i)= K] = 2,

Sec. 47

GALors THEORY

445

and hence, [K(c): K] is either 1 or 2. But [K(c): K] is the degree off by

Theorem 37.5, so f is either linear or quadratic. It remains for us to show that a quadratic polynomial «X3 + fiX + y has a root in K if and only if [3’ — 40w 2 0. If ,3“ — 40w 2 0, there exists A e K

satisfying 2’ = fi’ — 40:72 by what we have just proved, and (2. — fi)/2a is easily seen to be a root of ocX’ + 5X + y. Conversely, if x is a root of aX‘+ flX+yinK, then ‘6’ — 40:)! = (200: + [3)” 2 0. Theorem 47.4. (Euler-Lagrange) Let K be a totally ordered field. The following conditions are equivalent: 1° K(i) is algebraically closed.

2° K is real-closed. 3° Every positive element of K is a square of K, and every polynomial over K of odd degree has a root in K.

Proof: To prove that 1° implies 2°, let f be a monic polynomial over K satisfyingf(a)f(b) < 0 where a < b. Sincefis a product of irreducible poly-

nomials, there is a prime polynomial h over K dividingfsuch that h(a)h(b) < 0, for if a product of elements is < 0, at least one of the elements is < 0. If h

were quadratic, then h = X2 + pX + q where p2 — 4q < 0 by Theorem 47.3, and hence 2

2

h(x) = (x + ’32) + (q — P—) 4 >o for all x e K, whence in particular h(a)h(b) > 0, a contradiction. Therefore, by Theorem 47.3, h = X — c for some c e K. Thus, c is a root of h and hence

off. Also, (a—c)(b—c)=h(a)h(b) 0, and hence, f(x) > 0 if and only if aux > 0. Let c = M + 1. Thenf(—c)
4 implies that this case cannot occur. For if m = 2, then a is a product of an even number of transpositions, so 0' = (a1, a2)(a3, (14);), where p is either the identity permutation or a product of mutually disjoint transpositions, each of which is also disjoint from (a1, a2) and from (a3, a4). Then the inverse (a2, a4, as) of (a2, a3, a4) is disjoint from p, so $5 contains [(02, as: “00012, as: day-1°"= (as: as: a4)(a1, a2)(aa, a4)P(a2: as, air-Pkg!” a2)(as, 04) = (a2, “a: “0011: “2)(03: “0(02, (1‘, “3)PP‘_(a1, “2)(asa a4) = (a1, “00% “3)-

Since n 2 5, there exists an integer a, e [1, n] distinct from a1, a2, a3, a4. Consequently, 5 contains {(al! “a: as)[(av “0(42, a3)](a1, a4, day-Kan “4X02, as) = (a1, a4, a5)(a1, a¢)(az, a3)(a1, as, “4)(41: 00(02. as)

= (a1, “5: “4), a contradiction of our assumption that m = 2. Theorem 49.9. If n 75 4, then the only normal subgroups of C5,, are 6", ‘llfl,

and 3”.

Proofl The assertion is easy to prove if n g 3, so we shall assume that n 2 5. Let S) be a normal subgroup of 6”. Then S) n ‘11,, is a normal

Sec. 49

GADOIS THEORY

467

5 would contain [(02, as: 000012, as, air-lo": (a2: “3, “(Xab a2: a3)(a49 a5: (15),)(02, a3, a4)‘-P‘-(a4! a5: day-(ale a2: a3).-

= (as, as, “0011, as: as)(a4. a5, (Texans a4, “3)PP‘-(a4: as, “9011’ as, as) = ((11, a4: a2: “3: as):

a contradiction of our assumption that m = 3. Consequently, each factor of -r is a transposition, so 1" = 1”. Therefore, since 7- commutes with (al, a,, as), 5 contains 0'2 = (a1, “2a (13):“?2 = (a1, as, 02),

and hence, 5 = 9L, by Theorem 49.7.

Case 3: m = 2. We shall see that our assumption that n > 4 implies that

this case cannot occur. For if m = 2, then a is a product of an even number of transpositions, so 0' = (a1, a2)(a3, (14);), where p is either the identity permutation or a product of mutually disjoint transpositions, each of which is also disjoint from (a1, a2) and from (a3, a4). Then the inverse (a2, a4, as) of (a2, 0,, a4) is disjoint from p, so 5 contains [(02, as: 000012: “3a a4)“la" = (as: as, “4)(41: a2)(a3, “4)P(aas “a: day-P9011, az)(aa, a4)

= (a2, a3, a4)(a1, a2)(a3, a4)(a2, a4, a3)pp“(a1, a2)(a3, a4) = (a1, “4)(‘12’ “3)Since n 2 5, there exists an integer a5 6 [1, n] distinct from a1, a2, a3, a4.

Consequently, 5 contains {(01, “a, a5)[(a1, £10012, 03)](41: a4, day—Kala “43012: as) = (a1, “4a a5)(a1, a4)(az, “3)(01: as, a4)(a19 a4)(aa. as)

= (a1, a5» a4),

a contradiction of our assumption that m = 2. Theorem 49.9. If n 73 4, then the only normal subgroups of 6,, are 6”, ‘21",

and 3". Proof. The assertion is easy to prove if n g 3, so we shall assume that n 2 5. Let 5 be a normal subgroup of 6". Then 5 n ‘11,, is a normal

468

GALOIS THEORY

Chap. VIII

subgroup of ‘IIn and hence is either ‘21,, or 3,, by Theorem 49.8. If 55 n ii, = 3”, then $5 = 3,, by Theorem 49.5. If 5 n ‘1!” = 9!”, then 5 2 91,” so the order m of 5 divides n! and is a multiple of n!/2, whence m is either III or

n!/2, and consequently, 5 is either 6,, or ‘IIn.

Definition. A permutation group 6 on E is a transitive permutation group if for all x, y e E there exists a e 6 such that a'(x) = y. It is easy to see that permutation groups that are isomorphic as permuta-

tion groups are either both transitive or both intransitive. The subgroup 9! (Example 49.1) of 64 is intransitive, though 1),, which is isomorphic as

a group to 9%, is transitive. Theorem 49.10. If q is a prime number and if 6 is a transitive group of permutations of [1, q] that contains a transposition (a, b), then 6 = 6,.

Proof. Let M: {je [1,q]: eitherj = a or (a,j) e 6}, and let 0'66. We shall prove that if an, (M) n M =# (I), then (7* (II!) = M. To do so, suppose that there exists 1' e M such that a(i) e M. We shall first show that 0(a) e M. If either a = i or 0(a) = a, then clearly 0(a) e M. Consequently, we shall assume that 0(a) is neither 00') nor a. Then since (a, i) and (a, o'(i)) belong to 6, we also have 7(a, Dr“ E 6 where .r = (a, a'(i))a. But by (2), 7(0, 07‘— : (7(a): 7(0) = (0(a), 0) = (a: 0(a)),

so 6(a) e M. Now letj be any element of M; we shall show that 00) belongs to M. If eitherj = a or a(j) = a, then o'(j) e M by what we havejust proved. Consequently, we shall assume that 0'(j) is neither 6(a) nor a. Then since .(a,j) e 6, we also have p(a,j)p“ e 6 where p = (a, o'(a))d or p '= 6 according as a 7'3 0(a) or a = 0(a). But by (2),

p(a.j)P“ '= (pm. pa» = (a. e0». so 00) e M. Therefore, 6*(M) E M. But since M is finite and since a is a. permutation of [1, q], we conclude that 0*(M) = M.

Thus, for each 0‘ e 6, either 0*(M) = M or 0*(M) n M = (0. From this we may conclude that {0*(M): a e 6} is a partition of [1,q]; indwd, as 6 is transitive, for every k e [1, q] there exists a e 6 such that 6(a) = k,

whence k e 0*(111); moreover, if j 6 13,014) m p*(M) where 7-, p e 6, then

e50) e M n (mum. so M = :(p.04 is a permutation group isomorphism from the Galois group off defined by c onto the Galois group off defined by d. We shall denote by (SKU) the Galois group of a separable polynomial fe K[X] defined by some sequence of the roots off in a splitting field E off over K.

Example 50.1. Let E = Q(J2, J3), letf: (X2 — 2)(X2 — 3), and let g = X4 — 10X“ + 1. We observed in Example 37.2 that E is the splitting field of both f and g over Q. We also saw that X“ —— 3 is irreducible over Q(\/2) and simila_rly that X’— 2 IS irreducible over Q(«/3) hence, there exist a Q(\/2)-automorph1sm a‘ of E such that 0' J3)— — —-«/3 and a Q(«/3)-automorphism 1- of E such that T(\/2) —«/2. Then a‘, 1', 0'7, and the identity automorphism 1E are all the

Q-automorphisms of E as [E: Q] = 4. The Galois group off defined by the sequence (\/3 J2, —\/3, —\/2)1s, therefore, 91— — {1,7, (1,3),

(2, 4), (1, 3)(2, 4)}. The roots of g are \/2 + J3, —\/_2+ J3, x/2— \/3 and —\/2— \/3 Computing the values of o', -r, and mat those roots, we see that the Galois group of g defined by that sequence is 332 = {1m (1, 3)(2, 4), (1, 2)(3, 4), (1, 4)(2, 3)}. As we

saw in Example 49.1, the Galois groups off and g are isomorphic as groups but not as permutation groups. Our next goal is to obtain a condition on (SKU) equivalent to the solvability offby radicals where K is a field whose characteristic is zero. For this we need some preliminary definitions and theorems. Definition. A group (G, -) with neutral element e is solvable if there exists a

sequence (Gauss, of subgroups of G satisfying the following three conditions:

1° G0 = G and G,, = {e}. 2°’ For each ie [1, n], G, is a normal subgroup of GH. 3° For each ie [1,11], the group GH/G, is abelian.

474

GADOIS THEORY

Chap. VIII

h of [1, n] such that ck = d)

for all k e [1, n]; since

dod(h(k)) = “(“7a = “(01) = cacm = dh(a¢(k)) for all k e [1, n], we infer that

adoh=hoac;

therefore, 0', Had is a permutation group isomorphism from the Galois group off defined by c onto the Galois group off defined by d. We shall denote by (SKU) the Galois group of a separable polynomial fe K[X] defined by some sequence of the roots off in a splitting field E off over K.

Example 50.1. Let E = 96/2, J3), letf= (X2 — 2)(X2 — 3), and let g = X4 — 10X2 + 1. We observed in Example 37.2 that E is the

splitting field of bothf and g over Q. We also saw that X2 — 3 is

irreducible over Q(«/2) and similarly that X“ — 2 is irreducible over Q(\/3), hence, there exist a Q(\/2)-automorph1sm a of E such that 0(\/3) — —«/3 and a Q(~/3)-automorph1sm 1- of E such that 1-(«/2) —x/2. Then a‘, 1-, 01-, and the identity automorphism 1E are all the Q-automorphisms of E as [E : Q] = 4. The Galois group off defined by the sequence (J3, x/2, —\/3, —\/2) is, therefore, 91 = {117, (1, 3),

(2, 4), (1, 3)(2, 4)}. The roots of g are «/'2' + J3, —J5 + J3, J2 — J3, and —\/2 — J3. Computing the values of or, 1-, and 61'

at those roots, we see that the Galois group of g defined by that sequence is $2 = {117, (l, 3)(2, 4), (1, 2)(3, 4), (1, 4)(2, 3)}. As we saw in Example 49.1, the Galois groups off and g are isomorphic as groups but not as permutation groups. Our next goal is to obtain a condition on (5K0) equivalent to the solvability offby radicals where K is a field whose characteristic is zero. For this

we need some preliminary definitions and theorems. Definition. A group (G, ') with neutral element e is solvable if there exists a sequence (00051511 of subgroups of G satisfying the following three conditions:

1° G0 = G and G” = {e}. 2°' For each'i e [1, n], Gi is a normal subgroup of GH. 3° For each is [1,11], the group GH/Gi is abelian.

Sec. 50

GALOIS THEORY

475

A sequence of subgroups (G; 055,, satisfying these three conditions is called a solvable sequence for G. As we shall see in Theorem 50.2, for finite groups, condition 3° may be

replaced by the condition that G,_1/Gi be cyclic (and hence abelian) of prime order.

Lemma 50.1. Let f be an epimorphism from a group G onto a group G’, let H’ and K’ be subgroups of G’ such that K’ is a normal subgroup of H’, and let H = f*(H’) and K = f*(K’). Then H and K are subgroups of G, K is a normal subgroup of H, and the group H/K is isomorphic to the group

H’/K’. Proof. If x,y e H, then f(x), f(y) eH’, so f(x.;V—1)=f(x)f(y)—l GH’ and therefore xy‘1 eH; hence, H is a subgroup of G by Theorem 8.3. Also,

f*(H) = H’, for ify .9 H’, then there exists x e G such thatf(x) =y asfis surjective, whence x ef*(H’) = H. Similarly, K is a subgroup of G and f* (K) = K’. Moreover, K is a normal subgroup of H, for if x e H, then

f..(xKr1> =f(x)f*(K)f(x)-1 =f(x)K’f(x)—1 = K’, whence 1:e—1 E f* (K’) = K. The function fH: H —> H’ defined by fH(x). = f(x) for all x 6H is an

epimorphism from H onto H', and (p: x’ Hx’K’ is an epimorphism from H’ onto H’/K’. Consequently, (p ofH is an epimorphism from H onto H'/ ~’ whose kernel is K, for x e K if and only iff(x) e K’. Therefore, H/K

is isomorphic to H’IK’ by Theorem 20.9. Definition. Let H and K be subgroups of a group G such that H 2 K. A cyclic sequence from H to K is a sequence 0,9095, of subgroups of H satisfying the following three conditions:

1° 1., = Hand], = K. 2° For each k e [1, r], 1,, is a normal subgroup of Jk_1. 3° For each k e [1, r], the group 5,4l is a cyclic group whose order is a prime. Theorem 50.2. If G is a finite solvable group, then there is a cyclic sequence from G to {e}. Proof. Let (Gangs, be a solvable sequence for G, and let S be the set of all strictly positive integers m such that for any subgroups H and L of G, if L is a normal subgroup 0f H and if the group H/L is an abelian group of order s m, then there'is a cyclic sequence from H to L. It suflices to

Sec. 50

GALOIS THEORY

475

A sequence of subgroups (G,- 0555,, satisfying these three conditions is called a solvable sequence for G.

As we shall see in Theorem 50.2, for finite groups, condition 3° may be replaced by the condition that GH/G, be cyclic (and hence abelian) of prime order.

Lemma 50.1. Letf be an epimorphism from a group G onto a group G’, let H’ and K’ be subgroups of G’ such that K’ is a normal subgroup of H', and let H = f*(H’) and K = f*(K’). Then H and K are subgroups of G, K is a normal subgroup of H, and the group H/K is isomorphic to the group

H’/K’. Proof. If x,y E H, then f(x), f(y) E H’, so f(xy—1)=f(x)f(y)—1 e H’ and therefore xy—1 eH; hence, H is a subgroup of G by Theorem 8.3. Also,

f*(H) = H', for if y E H’, then there exists x e G such that f(x) = y as f is surjective, whence x ef* (H’ = H. Similarly, K is a subgroup of G and f*(K) = K’. Moreover, K is a normal subgroup of H, for if x EH, then

ANTI) =f(x)};(Kym-1 =fK'f(x)-1 = K', whence xKX‘1 s f*(K’) = K. The function fH: H —> H’ defined by fH(x) = f(x) for all x 6H is an epimorphism from H onto H’, and (p: x’ Hx’K’ is an epimorphism from H’ onto H'/K’. Consequently, 1, so there exist a prime p and a positive integer r such that s = pr. Let b = a", let K’ be the cyclic subgroup of H’ generated by bL, and let K = tpfiK’) where (p1, is the canonical epimorphism from H onto H’. The order of bL is clearly p, so K’ is a subgroup of H’ of order p. By Lemma 50.1, K is a normal subgroup of H as H = 0, fhas at least three real roots and hence, at most, two nonreal complex roots by Theorem 42.2. But

Df= 5X4 — 4 = (#310 + 2)(\/5X’ — 2), so Df has only two real roots, namely, (-91“ and —(§)1/4. Consequently, f has, at most, three real roots by the corollary of Theorem 42.3. Hence, f has exactly two nonreal complex roots. By Theorems 50.6, 50.7, and 50.9, therefore, f is not solvable by radicals over Q.

Sec. 50

GALOIS THEORY

481

Proof. Since (1, 2, 3)(l, 2, 4) .75 (l, 2, 4)(1, 2, 3), ‘11,, is not abelian and hence is not solvable by Theorem 49.8. By Theorem 50.3, therefore, 6,, is also insolvable.

Theorem 50.8. If f is a separable prime polynomial over a field K, then (5K0) is a transitive permutation group. Proof. Let cl, . . . , c” be the roots off in a splitting field E off over K; E is a Galois extension of K by Theorem 46.6. Let i, j e [1, n]. As c, and c, are roots off, there is a K—isomorphism 01 from K(ci) onto K(c,) such that «1(a) = c, by the corollary of Theorem 37.7. By Theorem 46.5, there is a

K-automorphism a of E satisfying 0(x) = 61(x) for all x eK(c,). Consequently, the permutation of [1, n] corresponding to o' is an element of (5K0) taking 1' into j. Therefore, (SEQ) is transitive. Theorem 50.9. Let K be a subfield of R. Iff is a prime polynomial over K

whose degree 1) is a prime and iff has exactly two nonreal roots in C, then (5K0) = 69.

Proof. Sincefis a prime polynomial, (EKG) is a transitive permutation group by Theorem 50.8. Let c1 be a nonreal complex root off: Since the coeflicients off are. real, the complex conjugate c} of c,- is also a root off, and hence C" = c, for somej 6 [1,17] distinct from i. Let E be the splitting field offover K contained in C. Since a: z r—> 2, z e E, is a K-monomorphism from E

into C, by Corollary 46.5.2 2' E E for all z e E, so the function obtained by restricting the codomain of o' to E is a K-automorphism of E. The permuta-

tion in (SEQ) corresponding to this automorphism is (i,j). Hence, (EKG) = 6, by Theorem 49.10. At last we are able to exhibit a quintic that is not solvable by radicals over Q.

Example 50.3. Let f be the polynomial X5 — 4X + 2 over Q. By Eisenstein’s Criterion, f is irreducible over Q. Since f(—2) < 0, f(0) > 0, f(1) < 0, f(2) > 0,fhas at least three real roots and hence, at most, two nonreal complex roots by Theorem 42.2. But

Df= 5X4 — 4 = («/5X‘ + 2)(\/5X‘ — 2), so Df has only two real roots, namely, (91/4 and -(%)1/‘. Consequently, f has, at most, three real roots by the corollary of Theorem 42.3. Hence, f has exactly two nonreal complex roots. By Theorems 50.6, 50.7, and 50.9, therefore, f is not solvable by radicals over Q.

482

GALOIS THEORY

Chap. Vlll

EXERCISES 50.1. A finite group of order p” where p is a prime is solvable. 50.2. Show that the following polynomials are not solvable by radicals over Q.

(a)X5—9X+3 (c) X5—6X"+3 (e) X5—8X+6

(b)X5—5X3—20X+5 (d)X5—10X3+5 (f) X5—15X4+6.

50.3. Is every nonconstant polynomial with complex coefficients solvable by

radicals over C? Is every nonconstant polynomial with real coeflicients solvable by radicals over R? 50.4. If an irreducible polynomialfover a field K has a root in a radical extension

of K, then f is solvable by radicals over K. *50.5. Let f be a separable prime polynomial of prime degree q over a field K such

that the Galois group I‘ of a splitting field E off over K is solvable, and let (I‘k)OS k5,, be a solvable sequence for I‘ such that 1‘,,_1 is a cyclic subgroup containing more than one element. Let c be a root offin E, let p be a generator of I‘,._1, and let a, = p‘(c) for each i e [1, q]. (a) The sequence c = (c1, . . . , ca) is a sequence of distinct terms consisting of all the roots off in E. [Use Exercise 49.7.]

'

(b) Let (5K0) be the Galois group off over K determined by c, and for each k e [0, n] let (5,, be the subgroup of (EEO) corresponding to Ft. Then for each m E [1 , n], the permutation group (5%", is a linear permuta-

tion group (Exercise 49.8), and every cycle of length q belonging to (5,,“ belongs to (5nd. [Use (2) of §49, Exercise 49.8(d), and induction in considering 1' pr“.]

(c) Conclude that (5K0) is a linear permutation group containing the cycle (1, 2, . . . ,q).

If f is a prime polynomial of prime degree q over a field K whose characteristic is zero, thenfis solvable by radicals over K if and only if there exists a sequence c = (Cl, . . . , cc) of distinct terms consisting of all the roots off

in a splitting field off such that the Galois group (5K0) off determined by c is a linear permutation group containing the cycle (1, 2,. . . ,q). [Use Exercises 50.5 and 49.8.]

*50.7. Let f be a prime polynomial of prime degree q over a field K whose charac-

teristic is zero, and let 0 and c’ be two roots off in a splitting field E off

over K. Iff is solvable by radicals over K, then E = K(c, c'). [Use Exercise

50.6 to show that K(c, c')A = {1E}.] 50.8. Let K be a subfield of R, and let f be a prime polynomial over-K of prime

degree q.

482

GALOIS THEORY

Chap. VIII

EXERCISES 50.1. A finite group of order p" where p is a prime is solvable. 50.2. Show that the following polynomials are not solvable by radicals over Q.

(a)X5—9X+3 (c) X5 —6X2 +3 (e) X5 —8X+6

(b)X5—5X3—20X+5 (d) X5 —10X3+5 (r) X5 —15X4+6.

50.3. Is every nonconstant polynomial with complex coefficients solvable by

radicals over C? Is every nonconstant polynomial with real coefiicientl solvable by radicals over R? 50.4. If an irreducible polynomialfover a field K has a root in a radical extension

of K, then f is solvable by radicals over K. I"50.5. Letfbe a separable prime polynomial of prime degree q over a field K such

that the Galois group I‘ of a splitting field E offover K is solvable, and let (1‘90s be a solvable sequence for I‘ such that 1",..1 is a cyclic subgroup containing more than one element. Let c be a root offin E, let p be a generator of I‘n_1, and let c‘ = p‘(c) for each i e [1, q]. (a) The sequence 0 = (c1, . . . , ca) is a sequence of distinct terms consisting

of all the roots off in E. [Use Exercise 49.7.] ' (b) Let (5K0) be the Galois group off over K determined by c, and for each k e [0, n] let (5k be the subgroup of (5K0) corresponding to I“. Then for each m E [1, n], the permutation group _(5,,_,,. is a linear permuta-

tion group (Exercise 49.8), and every cycle of length q belonging to (5%,. belongs to (5M1. [Use (2) of §49, Exercise 49.8(d), and induction in considering 1p1‘_.]

(c) Conclude that (SKU) is a linear permutation group containing the cycle (1, 2, . . . ,q).

50.6. If f is a prime polynomial of prime degree q over a field K whose charac-

teristic is zero, thenfis solvable by radicals over K if and only if there exists a sequence c = (Cl, . . . , cc) of distinct terms consisting of all the roots off

in a splitting field off such that the Galois group (SKU) off determined by c is a linear permutation group containing the cycle (1, 2, . . . ,q). [Use Exercises 50.5 and 49.8.]

*50.7. Let f be a prime polynomial of prime degree q over a field K whose charac-

teristic is zero, and let c and c’ be two roots off in a splitting field E off

over K. Iff is solvable by radicals over K, then E = K(c, 0'). [Use Exercise

50.6 to show that K(c, c’)‘ = {1E}.] 50.8. Let K be a subfield of R, and let f be a prime polynomial over'K of prime

degree q.

Sec. 50

GALOIS THEORY

483

(a) Iffis solvable by radicals over K and iff has two real roots, then every root off in C is a real number. (b) Ifq 2 5 and iffhas n real roots where 2 < n < q, thenfis not solvable by radicals over K. [Use Exercise 50.7.]

50.9. Let q be a prime 2 5. (a) Let f be one of the following polynomials over Q: Xq+aX+b X"+aX2+b

Xa+aq_1X¢—1+a¢_2X¢’2+...+a3X3+ao.

If fis a prime polynomial that is solvable by radicals over Q, then f has exactly one real root. "' (b) Show that Xa — 4X + 2 is a prime polynomial that is not solvable by radicals over Q. Construct two other examples of prime polynomials of degree q that are not solvable by radicals over Q. *50.10. If u is a constructible complex number, then (with the terminology of

Theorem 44.4) there exists an admissible sequence (L905 55¢ of subfields

for u such that La is a normal extension of Q. [Argue as in the proof of Theorem 50.5.]

*50.11. (a) A complex number u is constructible if and only if u belongs to a finite normal extension E of Q such that [EzQ] is a power of 2. [Use Exercise 50.10; for sufficiency, argue as in the proof of Theorem 48.4.] (b) A complex

number u is constructible if and only if u is algebraic over Q and the degree over Q of the splitting field of its minimal polynomial is a power of 2.

Sec. 50

GAIDIS THEORY

483

(a) Iff is solvable by radicals over K and iffhas two real roots, then every root off in C is a real number. (b) If q 2 5 and iffhas n real roots where 2 < n < q, thenfis not solvable by radicals over K. [Use Exercise 50.7.]

50.9. Let q be a prime 2 5.

(a) Let f be one of the following polynomials over Q: X“+aX+b X¢+aX=+b Xv+a¢_1X¢-1+a¢_2X¢-2+

+a3X3+ao.

If f is a prime polynomial that is solvable by radicals over Q, then f has

exactly one real root. L' (b) Show that XV — 4X + 2 is a prime polynomial that is not solvable by radicals over Q. Construct two other examples of prime polynomials of degree q that are not solvable by radicals over Q. *50.10. If u is a constructible complex number, then (with the terminology of

Theorem 44.4) there exists an admissible sequence (L905 55.1 of subfields for u such that L, is a normal extension of Q. [Argue as in the proof of Theorem 50.5.]

*50.11. (a) A complex number u is constructible if and only if u belongs to a finite normal extension E of Q such that [E: Q] is a power of 2. [Use Exercise

50.10; for sufliciency, argue as in the proof ofTheorem 48.4.] (b) A complex number u is constructible if and only if u is algebraic over Q and the degree over Q of the splitting field of its minimal polynomial is a power of 2.

APPENDIX A

INDUCED N-ARY OPERATIONS

Let A be a composition on E. We wish to generalize the definition of N‘a

for every a GE and every n e N* by giving a definition of the expression alAaaA . . . Aa,,_1Aan for every n-tuple (a1, . . . , an) e E", so that alAaaA . . .

Aan_1Aa,, may reasonably be described as the element obtained after n steps by starting first with al, forming then the composite alAag of it with a,, forming next the composite (alAa2)Aa3 of the result with a,, forming next the composite ((alAaQAaQAaa of that result with a4, etc. Thus, our definition of alAagA . . . Aa,_1Aa,, will turn out to be (. . . ((alAa2)Aaa)A . . . Aafl_1)Aa,, where all the left parentheses occur at the beginning. To make our definition

precise we shall use the Principle of Recursive Definition. Definition. Let n e N*. An n-ary operation on a set E is- a function from E" into E. We shall, of course, use the words “unary,” “binary,” “ternary,” etc.,

for “l-ary,” “2-ary,” “3-ary.” Binary operations on E are thus just the compositions on E. Theorem A.1. Let A be a composition on E. There is one and only one sequence (Ak),.21 such that for all n e N*, A” is an n-ary operation on E,

A1(a) = a

(1) fer every a e E, and

(2)

An+1 (a1, - - - a am “n+1) = An(01: - - - : an)Aan+1

485

APPENDIX A

INDUCED N-ARY OPERATIONS

Let A be a composition on E. We wish to generalize the definition of A"a for every a GE and every n e N* by giving a definition of the expression alAagA . . . AaklAan for every n-tuple (a1, . . . , a”) e E", so that alAaaA . . .

Aan_1Aa,, may reasonably be described as the element obtained after n steps by starting first with al, forming then the composite alAa2 of it with a,, forming next the composite (alAaaMaa of the result with as, forming next the composite ((alAaQAaQAa4 of that result with a,,, etc. Thus, our definition of alAanA . . . AaklAan will turn out to be (. . . ((alAa2)Aa3)A . . . Aan_1)Aa,,

where all the left parentheses occur at the beginning. To make our definition precise we shall use the Principle of Recursive Definition. Definition. Let n e N*. An n-ary operation on a set E is a function from E" into E. We shall, of course, use the words “unary,” “binary,’9



ternary,” etc.,

for “l-ary,” “2-ary,” “3-ary.” Binary operations on E are thus just the compositions on E.

Theorem A.1. Let A be a composition on E. There is one and only one sequence (A,,),,21 such that for all n e N*, A” is an n-ary operation on E, A1(a) = a

(l)

fer every a e E, and (2)

A"+1011, ~ -. - s a”, “n+1) = An ((11, - ' ' 9 an)Aan+l

485

486

INDUCED N-ARY OPERATIONS

App. A

for every (n + l)-tuple (a1, . . . , an“) 6 En“. In particular, A2 is the given composition A. Proof. Let 6 = {V: for some n e N*, V is an n-ary operation on E}.

Let s be the function from «9' into 6’ defined as follows: For each n-ary operation V on E, SW) is the (n +'1)-ary operation defined by S(V)(a13 ' ' - s an, an+1) = V011: - - - 9 an)Aa'n+l

for all (a1, . . . , a”, a,,+1)eE"+1. By the Principle of Recursive Definition, there is a unique sequence (A,,),c21 such that A1 is the unary operation defined by (1) and A”+1 = s(A,,) for each n E N*. An easy inductive argument shows that A” is an n-ary operation on E for each n e N*, and by the definition of s, (2) holds for every (a1, . . . , a”, an“) E E"+1. We shall call the nth term A,I of the sequence (A91:21 the n-ary operation defined by A.

Definition. If A is a composition on E, for each (a1, . . . , an) e E” the

composite of (a1,'. . . , an) for A is the value at (al, . . . , an) of the n-ary operation defined by A. However, the composite of (a1, . . . , a,) for a composition denoted by a symbol similar to + is called the sum of (a1, . . . , a”), and

the composite of (a1, . . . , an) for a composition denoted by a symbol similar to - is called the product of (a1, . . . , an). The composite of (a1, . . . , an) for A is ordinarily denoted by Auk

or

a1A . . . Aan.

k=1

The composite of (a1, . . . , a”) for a composition denoted by a symbol similar to + is also denoted by

Zak:

k-l

and the composite of (a1, . . . , an) for a composition denoted by a symbol

similar to - is also denoted by

App. A

INDUCED N-ARY OPERATIONS

487

If an ordered n-tuple is given by a simple formula, that formula may be used to denote its composite. For example, 1!

11

2k2

is

2a,, k=1

k=1

where a,, = k’ for each k e [1, n]. An easy inductive argument establishes the following theorem, which it

shows that our definition of A ah essentially generalizes that of N‘a. 70=1

Theorem A.2. Let A be a composition on a set E, and let a e E. If(a1, . . . , a”) is the ordered n-tuple defined by ak = a for each k e [1, n], then 11

A ak = A"a. k=1

Definition. A sequence is a function whose domain is a subset of N. If the codomain of a sequence is contained in E, the sequence is said to be a

sequence of elements of E, or a sequence in E. A sequence of distinct elements (or terms) of E is an injection from a subset of N into E. A sequence of 1:

terms is a sequence whose domain has n elements, a finite sequence is a sequence whose domain is finite, and an infinite sequence is a sequence whose

domain is infinite. Custom has given the following notation for sequences: Iffis a sequence and if A is the domain off, some letter or symbol is chosen, say “a,” f(k) is denoted by ak for each k e A, and f itself is denoted by 01,)“. Any expres-

sion denoting the domain offmay be used in place of “k e A”; for example, if A is the set of all natural numbers 2 n, the sequence may be denoted by (ak),,2,,, and if A is the integer interval [p, q], the sequence may be denoted

by (@598. If a sequence is defined by a simple formula, that formula is often used to denote the sequence as in the following examples: (#93957 is the sequence (@3395, where ak = k2 for all k e [3, 7], and (log(k3 — 5))“,23 is the sequence (bgkz, where bk = log(k2 — 5) for all k 2 3. With this notation, a sequence (01915.4 is a sequence of distinct elements if and only if a, 7E ah for all j, k e A such that j 7'5 k. If (E, b, 6 f(c), 329 f1), 58 FE, 10

n—>¢)

[L:K], 349 log“, 395

f: E —> F, 7

M°", 295 flK(m, n), 303 .1!K(n), 306 [m, n], 80

f*(X),f*(Y), 173 flat), 6

__ E . . . (mod m), 170 __ E . . . (mod m), 170

g of, 33 GF(q), 365 G/H, 156 G(m, n, r, s), 234 Gr(f), 9 (Gr 1), (Gr 2), 9 G.z, 238 6,, 239

mp(r, s), 183

HomK(E, F), 291

n!, 129

-m, 14 N, 3, 80 N‘— , 295 N“, 80 N*—, 110 (N), 198 n‘, 110

LIST OF SYMBOLS n.a, 87

R, X, 68

9’71, 398 n — m, 81,113

R4, 70 R“, 72 Rn(K), 451

5.4, 130 (”), m

312, 180, 400

n . m, 86

[r1, . . . , n.1,, 108

N", 14, 80 (NO l)—(NO 4), 80 (NO 5), 11.7 (NO 6)—(NO 7), 11.10 .N’(Q), 381 N(S), NH(S), 246 v(u), 281

p(u), 281

(0, 5

k=1

1, 27, 80 1E, 34 (OP), 7 (OR), 147 (OS), 76 (P 1)—(P 4), 148 213(E), 10 (p, 218 (pH, 174 $3, 136 CI)”, 448

sup A, 370

ii a,” 95 k=l

fi E,” 95 k=1

S,, 329 (SE, 53 Gm, 54 ac, 473 sgn a, 182

f a,,, 95 T, 179

[14; (17).,” (am, 305 (UFD l)—(UFD 3), (UFD 3’), 202 u‘, 31.9 V,(x), 42.13 (VS 1)—(VS 4), 253 X, 321 xA, 294 364, —x, 28

x*, 28

ll, 136 n, 68

(x, t'), 293 {x: x has property Q}, 5

(PID), 198 P(K), 261 12,,(K), 261

x o y, 16

(PS), 11.7

X A Y, 53, 155 Z, 3, 110 z', 399

Q, 3, 147 [r], 384 R, 3, 386 31, 383 ‘R, 464

507

x — y, 114 x A y, 12

Z... 169 Z(G), 61 o, 27, so

INDEX

References are to page and exercise numbers. Decimal point numbers refer to exercises.

A Abelian group (see Abelian semigroup) Abelian semigroup, 52 Absolute value, 372 Action, 238

by conjugation, 240 Addition: integers, 110 modulo m, 14

Adjunction of root, 351 Algebra, 310

commutative, 312 division, 311

Algebraically closed field, 400, 412 Alternating group, 184 Annihilator, 27.8, 295

Anticommutative composition, 2.17 Anti-isomorphism, 15.8 Antisymmetric relation, 68, 69 Archimedean ordered field, 371 Archimedean ordered group, 371 Arithmetic, Fundamental Theorem, 207 Associate, 196

Associative composition, 12 Associativity, General Theorem, 489

Automorphism: algebraic structure, 43

formal power series, 320

group of field extension, 421 inner, 240 K-, 353

with identity, 312 K-, 310

polynomial, 321

K-algebraic structure, 255

quaternions, 33.21

ordered ring, 150

quotient, 313 trivial, 311

ordered semigroup, 77

zero, 311

Algebraic closure, 47.2 in extension field, 47.1

Algebraic dual, 293 Algebraic element, 331, 351, 407

Algebraic extension, 407 Algebraic numbers, 47.2

Base, number, 108 Basis, 264 ordered, 264 ordered dual, 294

Algebraic structure, 41

standard, 265

standard ordered, 265

K-, 254, 255

509

510

INDEX

Bears, 69 Bernstein-Cantor-Schroder Theorem, 12.15 Between, 400 Bezout’s Identity, 199

Bijection, 35 Bijective function, 35

Binary notation, 109 Binary operation, 12, 485 Binomial coeflicients, 130 Binomial Theorem, 130

Biquadratic, 323 Bolzano’s Theorem for Polynomials, 401 Boolean ring, 15.21 Bound: greatest lower, 370

least upper, 370 lower, 370

Closed subgroup of automorphism group, 421 Closed subset, 59 Closure of subfield of field extension, 421

Closure of subgroup of automorphism group, 421 Codimension, 289 Codomain, 6

Coeflicient: binomial, 130 constant, 323

leading, 323 Collection, 3 Column of matrix, 303

Common divisor, 197 greatest, 197 Common multiple, 197 least, 197 Commutative algebra, 312

upper, 370 Bounded, 370 Bounded above, 370 Bounded below, 370 Bounded interval, 379

Bounded sequence, 374 Budan’s Theorem, 42.13

C Cancellable element, 54 left, 7.6 right, 7.6

Canonical epimorphism, 175

Commutative composition, 13 Commutative group (see Commutative semigroup) Commutative ring, 123 Commutative semigroup, 52 Commutativity, General Theorem, 49] Compact set, 380

Compatible equivalence relation, 163 Compatible ordering, 76 with ring structure, 147 Complement, 21

Cantor-Bernstein-Schrtider Theorem, 12.15

relative, 21 Complete ordered field, 375 Complex numbers, 4, 399

Cantor’s diagonal mapping, A.11

Composite:

Canonical surjection, 136

Cantor’s Theorem, 12.10 Cartesian product, 8, 95

groups, 221 modules, 258 Cauchy sequence, 374

Cayley’s Theorem, 62 Center: group, 61

ring, 129 Chain Rule, 35.15

Changes signs, 400

elements, 95, 486 functions, 33

n-tuple of elements, 95, 486 sets, 5.8

Composition, 12 (see also Induced composition): anticommutative, 2.17 associative, 12

commutative, 13 distributive, 87

idempotent, 2.17

Characteristic, 192

Congruence relation, 163

Circle group, 179

Conjugate, 240, 246 complex, 240

Class, 3 equivalence, 136 transitivity, 49.5

Closed interval, 379 Closed subfield of field extension, 421

H-, 246 quaternion, 33.22

Conjugate class, 240

Conjugate subgroup, 246

INDEX

511

Conjugation, 240

Dependent (see Linearly dependent)

Connected set, 380 Consecutive roots, 403 Constant coefficient, 323

Constructible point, 414

Derivative, 338 Descartes’ Rule of Signs, 42.14 Development, 107 Diagonal mapping, A.11 Diagonal of matrix, 307 Diagonal subset, 34 Dicyclic group, 236 Difierential operator, 279, 338 Dihedral group, 53, 232 Dimension, 270

Constructible real number, 415 Contains, 5

Dipper, 20.19 Direct product, 222, 224

Constant polynomial, 322 Constructible Constructible Constructible Constructible Constructible

angle, 419 circle, 414 complex number, 416 line, 414 line segment, 418

algebraically, 143 properly, 102

Direct sum:

subgroups, 222, 224

Contraction, 278

Convergent sequence, 374 Converges, 374

Convex subset, 18.20, 379 Coordinate, 266 Coordinate system, 266 Coset, 156 Countable set, 12.7

Counting, Fundamental Principle, 498 Cubic, 323 Cycle, 63, 182 Cyclic extension, 456

Cyclic group, 214 Cyclic sequence, 475 Cyclotomic polynomial, 448, 450

submodules, 284 Direct summand, 286 Directed ordering, 9.17 Disjoint permutations, 462 Disjoint sets, 20

Distributive composition, 87 left, 11.13 right, 11.13 Distributivity, General Theorem, 493 Divides, 93, 170, 196 ’ Divisible element, 196 Divisible group, 27.12

Divisible module, 27.13 Division algebra, 311

Division Algorithm, 106, 168, 22.9, 324 Division ring, 128

D d’Alembert Theorem, 446 Decimal notation, 106

Decomposable group, 227 Decreasing function, 73

at a point, 403

Division subring, 128

Divisor, 170, 196 (see also Zero-divisor): common, 197

greatest common, 197 Domain: euclidean, 22.13

function, 6

Dedekind cut, 40.6 Dedekind ordered field, 370

gaussian, 202 integral, 125

Dedekind ordered group, 370

principal ideal, 198 unique factorization, 202

Dedekind’s Theorem, 424 Deficiency, 289

Degree: element, 332, 351, 407 exponential, 46.6 extension field, 349

Dual, algebraic, 293 Dual ordered basis, 294

E

polynomial, 323

Eisenstein’s Criterion, 344

reduced, 46.6

Element, 3

Delta Kronecker convention, 292 Dense subset, 372 Denumerable set, 12.7

Embedded algebraic structure, 143 Empty set, 5 Endomorphism, 175, 276

512

IND“

Epimorphism, 175, 276 canonical, 175

Field (cont.): Galois, 365

Equipotent sets, 100

obtained by adjoining root, 351

Equivalence class, 136 Equivalence relation, 135 compatible, 163 Euclidean Algorithm (see Division

ordered, 147 perfect, 435

Algorithm) Euclidean domain, 22.13 Euclidean stathm, 22.13 Euler function, 218

Euler-Lagrange Theorem, 445 Evaluation isomorphism, 294 Evaluation linear transformation, 294 Even integer, 170 Even permutation, 182 Exponential degree, 46.6 Exponential function, 396 Extension (see Extension field): algebraic structure, 143

prime, 193

quotient, 139 real-closed, 400 skew, 128

splitting, 355 stem, 351 totally ordered, 147 Finite-dimensional extension, 349 Finite-dimensional module, 270

Finite-dimensional vector space (see Finitedimensional module) Finite extension, 349 Finite sequence, 487 Finite set, 101

Finitely generated module, 262 Finitely generated vector space (see Finitely

function, 58 group, 232

generated module)

relation, 70 Extension field, 349

First term, 7 Form, linear, 293

algebraic, 407

cyclic, 456 finite, 349 finite-dimensional, 349 Galois, 428

Formal power series algebra, 320 Four group, 228 Free module, 264

Frobenius’ Theorem, 47.9 Function, 6, 10

generated by a set, 349

decreasing, 73

normal, 435, 46.5

decreasing at a point, 403

purely inseparable, 46.9 radical, 471

exponential, 396 increasing, 73

separable, 434

increasing at a point, 403

simple, 409

logarithmic, 395

transcendental, 407

polynomial, 261, 337 restriction, 58

F Factor, 196 inseparable, 46.11 separable, 46.11

Factor Theorem:

group, 178

strictly decreasing, 73 strictly increasing, 73 Functional, linear, 293

Fundamental Principle of Counting, 498 Fundamental theorem: algebra (see d’Alernbert’s Theorem)

module, 288

arithmetic, 207

ring, 190

counting (see Fundamental Principle of

Family of elements, 493 Fermat prime, 455

Counting)

Galois theory, 429

Fermat’s Theorem, 23.5, 364 Field, 128

G

algebraically closed, 400 extension, 349 fixed, 420

Galois extension, 428 Galois field, 365

513

INDEX

Galois group, 428 of polynomial, 473 Galois theory, fundamental theorem, 429 Gaussian domain, 202 Gaussian integers, 22.16 Gauss’s Lemma, 342

General Associativity Theorem, 489 General Commutativity Theorem, 491 General Distributivity Theorem, 493 Generated extension field, 349 Generated linear variety, 31.3

Generated principal ideal, 166 Generated subalgebra, 312 Generated subfield, 348

Generated subgroup, 213

H Have the same (opposite) sign, 403 H-conjugate, 246 H-normalizer, 246 Homogeneous line, 298

Homogeneous linear equations, 282 Homogeneous plane, 298 Homomorphism (see Automorphism; Endomorphism; Epimorphism; Isomorphism; Monomorphism): algebraic structure, 174

induced on polynomial ring, 323 K—algebraic structure, 275

substitution, 330

Hyperplane, 297

Generated submodule, 262

Generated subspace (see Generated submodule) Generators (see Generated) Graph, 9 Greatest common divisor, 197 Greatest lower bound, 370 Group, 52

abelian (see Abelian semigroup) acting on a set, 238

alternating, 184 of automorphisms (see Automorphism group of field extension, Galois group) circle, 179 commutative (see Commutative semi-

group) cyclic, 214 decomposable, 227 dicyclic, 236 dihedral, 53, 232 extension, 232 four, 228

Galois, 428, 473 generated by subset, 213

Ideal, 165, 312 left, 19.15 maximal, 167

modular, 19.5 nonzero, 165 prime, 19.6

principal, 166 proper, 165 regular, 19.5 right, 19.15 zero, 165

Idempotent composition, 2.17 Idempotent element, 2.17 Identity, 26, 27

ring with, 124 Identity function, 34

Identity matrix, 307 Image, 6

Imaginary number, 42.2 Imaginary part, 400 Increasing function, 73 at a point, 403

Indecomposable group, 227

indecomposable, 227

Indeterminate, 323

octic, 53 ordered, 76

Index:

11-, 243

subgroup, 157 Indexed family of elements, 493

permutation, 62, 461 quaternio'nic, 236, 33.21

quotient, 160 simple, 466 solvable, 474 symmetric, 54

symmetries of square, 53 tetrahedral, 244

nilpotency, 15.18 Induced composition: quotient group, 160 quotient set, 163 set of subsets, 155

stable subset, 59 Induced homomorphism, 323 Induced relation, 70

514

INDEX

Induction (see Mathematical Induction)

K

Inductive semigroup, 11.6

K-algebra, 310

Infimum, 370 Infinite order, 157, 215

K-algebraic structure, 254, 255 K-automorphism, 353 Kernel, 177, 276 K-isomorphism, 353 Kite, 20.19 K-module, 254 K-monomorphism, 353 Kronecker’s delta convention, 292 Kronecker’s Theorem, 351

Infinite sequence, 487 Infinite set, 101

Infinitely large element, 39.3 Infinitely small element, 39.3 Infinitesimal, 39.3

Injection, 35 Injective function, 35 Inner automorphism, 240 Inseparable factor, 46.4 Integer interval, 80 Integers, 3, 110 gaussian, 22.16 modulo m, 123, 169

prime, 170

K-vector space, 253

L Lagrange’s Theorem, 157 Leading coefficient, 323 Least common multiple, 197

Integral domain, 125

Least upper bound, 370

euclidean, 22.13 gaussian, 202 unique factorization, 202 Intersection, 20, 24 Interval, 80, 379 bounded, 379

Left cancellable element, 7.6

closed, 379 open, 379 Inverse, 27

left, 4.11 multiplicative, 125 right, 4.11

Inverse function, 35 Inverse-completion, 152 Inversions, number of, 182 Invertible element, 27

ring, 125 Invertible matrix, 307 Irreducible element, 200 representative set, 201 Irreducible module, 30.14 Irreducible polynomial, 326, 341 Isomorphic classes, 48

Isomorphism: algebraic structure, 43 evaluation, 294 induced on polynomial ring, 323 K-, 353 K-algebraic structure, 255 ordered ring, 150

ordered semigroup, 77

Left coset, 156

Left distributive composition, 11.13 Left ideal, 19.15 Left inverse, 4.11 Left neutral element, 4.5 Left regular representation, 63, 21.6

Length: cycle, 182

element in unique factorization domain, 22.12

Limit, 375 Linear combination: sequence, 262 set, 262 Linear form, 293

Linear functional, 293 Linear operator, 276

Linear permutation, 49.8 Linear polynomial, 323 Linear subgroup, 49.8 Linear transformation, 276 evaluation, 294 matrix of, 305 Linear variety, 31.2 dimension, 31.2

generated by set, 31.3 parallel, 31.2 proper, 31.2 Linearly dependent sequence, 263 Linearly dependent set, 263

ordered structure, 72

Linearly independent sequence, 263

permutation group, 462

Linearly independent set, 263

515

INDEX Liouville number, 43.11

Multiple root, 335

Logarithmic function, 395

Multiplication :

Lower bound, 370 greatest, 370

M Mathematical Induction, 82 misuse, 84 Matric addition, 303 Matric multiplication, 304 Matrix, 303 column of, 303 identity, 307 invertible, 307 linear transformation, 305, 32.11 row of, 303 square, 306

zero, 304 Maximal element, 167 Maximal ideal, 167 Maximal submodule, 30.14 Mean Value Theorem for Polynomials, 42.7 Member, 3 Mersenne number, 38.4 Mersenne prime, 38.4

formal power series, 319 integers, 116 modulo m, 14 natural numbers, 86

polynomials, 319 ring, 123 scalar, 253 Multiplicative inverse, 125 Multiplicity of root, 335

N N-ary operation, 485 defined by composition, 486 Natural numbers, 3, 80 Natural surjection, 136

Naturally ordered semigroup, 80 N-cycle, 182

N-dimensional, 270 Negative numbers, 4, 403 Neutral element, 26 left, 4.5 right, 4.5

Minimal polynomial 332, 351, 407

Nilpotent element, 15.18

Modular ideal, 19.5 Module, 254

associated to abelian group, 258

Noetherian ring, 22.21 Nontrivial solution, 282 Nonzero idea], 165

cartesian product, 258

Nonzero ring, 123

divisible, 27.13 irreducible, 30.14

Norm: element in finite Galois extension, 45.8

K-, 254

quatemion, 33.22 Normal extension, 435

maximal, 30.14

obtained by restricting scalar multiplication, 257 quotient, 288

simple, 30.14 torsion-free, 27.13 Modulo m compositions, 14 Monic generator, 326

Monic polynomial, 323 Monomial, 321 Monomorphism, 175 algebraic structure, 175

generated by extension field, 46.5

Normal subgroup, 158 Normalizer, 246

Nth root, 398 of unity, 448 N-tuple, 94

composite, 94 Null sequence, 381 Null set, 5 Nullity, 281 Number:

K-, 353

algebraic, 47.2

K-algebraic structure, 276

base, 108

ordered semigroup, 388

complex, 4, 399

Moore’s Theorem, 365 Multiple, 93, 170, 196 common, 197 least common, 197

of elements in a set, 101 integer, 3, 110

of inversions, 182 Liouville, 43.11

516

INDEX

Ordered ring, 147 automorphism, 150

Number (cont): Mersenne, 38.4 natural, 3, 80

isomorphism, 150

pure imaginary, 42.2

Ordered semigroup, 76

rational, 3, 147 real, 3, 386 real algebraic, 47.2 of roots, 337 of roots, multiplicities counted, 337

automorphism, 77 isomorphism, 77 Ordered structure, 71

of variations in sign, 403

O Octic group, 53 Odd integer, 170

Odd permutation, 182 One-to-one function, 35

Ordered subsemigroup, 77 Ordered triad, 8 Ordered triple, 94 Ordering, 69 compatible, 76

compatible with ring structure, 147 defined by subset of ring, 149 directed, 9.17 total, 69 well, 73

One-to-one onto function, 35

Onto, 35 Open interval, 379 Open set, 380

Operation: binary, 12, 485 n-ary, 485 n-ary defined by composition, 486

ternary, 485 unary, 485 Operator:

P Parallel linear varieties, 31.2 Parallel lines, 298 Parallel planes, 298

Partition, 135 relation defined by, 136 Peano model, 11.8

Peano semigroup, 11.7 Peano set, 40.5

differential, 279, 338

Peano’s axioms, 11.8 Perfect field, 435

group of, 238

Permutation, 35

linear, 276

Opposite sign, 403 Orbit, 239 Order:

element of group, 215 group, 157 infinite, 157 square matrix, 306 Ordered basis, 264 Ordered couple, 94 Ordered dual basis, 294 Ordered field, 147 archimedean, 371 complete, 375 Dedekind, 370

Ordered group, 76 archimedean, 371 Dedekind, 370 Ordered n-tuple, 94

defined by sequence, 489 Ordered pair, 7, 1.11 Ordered quadruple, 94

disjoint, 462 even, 182 linear, 49.8 odd, 182

Permutation group, 62, 461

isomorphism, 462 transitive, 468

Permute, 13

P-group, 243 Polynomial, 320

biquadratic, 323 coefficient, 323

constant, 322 constant coefficient, 323 cubic, 323 cyclotomic, 448, 450 degree, 323 function, 261, 337

irreducible, 326, 341 leading coefiicient, 323 linear, 323

minimal, 332, 351, 407

517

INDEX

Polynomial (cont): monic, 323 prime, 326

primitive, 342 pure, 456

quadratic, 323 quartic, 323 reducible, 341

Pullback Theorem, 143 Pure local maximum, 403

Pure local minimum, 403

Purely imaginary number, 42.2 Purely inseparable element, 46.9 Purely inseparable extension, 46.9 Pure polynomial, 456

Q

separable, 432 solvable by radicals, 473 Positive element:

group, 370 ring, 147 Positive nth root, 398 Positive numbers, 3 Power series (see Formal power series) Prime, 170, 201, 326 Fermat, 455

Mersenne, 38.4 polynomial, 326 Prime field, 193 Prime ideal, 19.6

Quadratic, 323 Quartic, 323 Quaternion, 33.21 Quaternionic basis, 33.21

Quaternionic group, 236 Quintic, 323 Quotient, 105, 325 Quotient algebra, 313 Quotient field, 139 Quotient group, 160 Quotient module, 288

Quotient set, 136

R

Prime integer, 170, 201 Prime number, 170, 201

Prime polynomial, 326 Prime subfield, 193 Primitive element, 440 theorem of, 440

Primitive nth root of unity, 448 Primitive polynomial, 342

Radical extension, 471 Range, 173 Rank of linear transformation, 281 Rational fractions, 408 Rational functions, 408 Rational numbers, 3, 147

Principal ideal, 166 Principal ideal domain, 198

Real algebraic numbers, 47.2 Real-closed field, 400 Real numbers, 3, 386

Principle:

Real part, 400

counting, 498 Mathematical Induction (see Mathematical Induction) recursive definition, 83

Product (see also Cartesian product, Direct product): matric, 304 natural numbers, 86 n-tuple of elements, 95

Projection: on line, 278 on submodule, 30.2

Proper idea], 165 Proper linear variety, 31.2 Proper subset, 102 Proper zero-divisor, 125

Properly contains, 102 Pseudo-ring, 15.4 P-subgroup, 243

Reciprocal ring, 15.8 Recursive Definition, 83

Recursive semigroup, 12.13 Reduced degree, 46.6

Reducible polynomial, 341 Reflection, 230, 231, 278 Reflexive relation; 68, 69

Regular ideal, 19.5 Relation, 68, 69 antisymmetric, 68, 69

compatible equivalence, 163 congruence, 163

defined by partition, 136 equivalence, 135

induced on subset, 70 ordering, 69 reflexive, 68, 69 symmetric, 68, 69 transitive, 68, 69

518

INDDK

Relational structure, 71 Relative complement, 21

Scalar multiplication, 253, 254 induced on quotient module, 287

Relatively prime elements, 198

matrices, 303

Remainder, 105, 325 Remainder Theorem, 35.8

restriction, 257 stable for, 260

Repeated root, 335 Representation, left regular, 63, 21.6 Representative set of irreducible elements, 201 Restriction: function, 58

Scalar product, 303 Scalar ring, 254 Schroder-Cantor-Bernstein Theorem, 12.15 Second term, 7

Semigroup, 52 abelian, 52

relation, 70

commutative, 52

scalar multiplications, 257

dipper, 20.19 inductive, 11.6 kite, 20.19 naturally ordered, 80

Right cancellable element, 7.6 Right coset, 156 Right distributive composition, 11.13

Right ideal, 19.15

ordered, 76

Right inverse, 4.11 Right neutral element, 4.5

Peano, 11.7 recursive, 12.13

strictly inductive, 20.15

Rigid motion, 15 Ring, 123

Separable element, 434

boolean, 15.21 center, 129 commutative, 123 division, 128 with identity, 124 integers modulo m, 123, 169 noetherian, 22.21 nonzero, 123

ordered, 147

without proper zero-divisors, 125 pseudo-, 15.4

reciprocal, 15.8 scalar, 254

totally ordered, 147

Sequence, 487 bounded, 374

Cauchy, 374 convergent, 374 distinct terms, 487

finite, 487 infinite, 487 7: terms, 487 null, 381 Set, 3

with group of operators, 238 Signature of permutation, 182

Rolle’s Theorem for Polynomials, 402 Root, 334

adjunction, 351

Similitude, 278 Simple extension, 409

Simple group, 466 Simple module, 30.14

multiple, 335

multiplicity, 335

Simple root, 335 Skew field, 128 Solvable by radicals, 473

repeated, 335

simple, 335

Solvable group, 474

square, 443 of unity, 46, 448 Rotation, 230, 277 Row of matrix, 303

S Scalar, 253, 254

Separable factor, 46.11

Separable polynomial, 432

Sfield, 128

trivial, 12.4 zero, 123

Same sign, 403

Separable extension, 434

Solvable sequence, 475 Space, 3, 238 G-, 238 vector, 253

Spanned submodule, 262 Splits, 355 Splitting field, 355

INDEX Square, 33.13, 443 Square-free integer, 43.8

Square matrix, 306 Square root, 443 Stabilizer, 239

Stable subset, 59, 260 Standard basis, 265 Standard ordered basis, 265 Stathm, 22.13 Steinitz’s Theorem, 409 Stem field, 351 Stretching, 278

Strictly decreasing function, 73 Strictly increasing function, 73 Strictly inductive semigroup, 20.15 Strictly negative numbers, 4, 403 Strictly positive element, 147, 370 Strictly positive number, 3

519

Successor, 79

Sum (see also Direct sum): matric, 303

n-tuple of elements, 95 Supplement, 286 Supplementary submodules, 286 Supremum, 370

Surjection, 35 canonical, 136 natural, 136 Surjective function, 35

Sylow p-subgroup, 243 Sylow’s Theorems, 242, 247 Symmetric difference, 53

Symmetric group, 54 Symmetric relation, 68, 69

Symmetries of square, 16 Symmetry, 15, 230

Structure: algebraic, 41 ordered, 71

T Taylor’s Formula, 35.14

relational, 71 Sturm sequence, 42.17 Sturm’s Theorem, 42.18

Ternary operation, 485 Tetrahedral group, 244

Subalgebra, 311 generated by subset, 312

Theorem (see also Factor Theorem, Fundamental Theorem, Principle):

Term, 7, 94

Subdomain, 126 Subfield, 128

Bezout, 197 Binomial, 130

closed, 421

Bolzano, 401 Budan, 42.13

generated, 348

Subgroup, 60 closed, 421

Cantor, 12.10

Cantor-Bernstein-Schréder, 12.15

conjugate, 246

Cayley, 62

generated, 213

d’Alembert, 446

H-conjugate, 246

Dedekind, 424

normal, 158

Descartes, 42.14 Eisenstein, 344

Submodule, 260

finitely generated, 262

Euler-Lagrange, 445

generated, 262 maximal, 30.14

Frobenius, 47.9

Subring, 126 division, 128

Subsemigroup, 60 ordered, 77 Subset, 4 convex, 18.20, 379 diagonal, 34

proper, 102 stable, 260 Subspace, 260 Substitution, 329

Substitution homomorphism, 330

Fermat, 23.5, 364 Gauss, 343

General Associativity, 489 General Commutativity, 491 General Distributivity, 493 Kronecker, 351

Lagrange, 157 Mean Value, 42.7

Moore, 365 primitive element, 440 Pullback, 143 Remainder, 35.8 Rolle, 402

520

INDEX

Theorem (cont.):' Steinitz, 409

Unique factorization domain, 202 Unit, 196

Sturm, 42.18

Unity, 26 Upper bound, 370

Sylow, 242, 247 Transplanting, 49 Wedderburn, 461 Wilson, 19.14 Torsion-free module, 27.13

Value, 6

least, 370

V

Total ordering, 69

Variations in sign, 403

Totally ordered field, 147

Variety, 31.2

Totally ordered ring, 147

Vector, 253 Vector space, 253

Trace, 45.8 Transcendental element, 331, 351, 407 Transcendental extension, 407 Transformation (see Linear transformation)

Vector subspace, 260 Venn diagram, 20

Transitive permutation group, 468

W

Transitive relation, 68, 69 Transitivity class, 49.5

Transitivity of algebraic extensions, 411 Transplant, 50 Transplanting Theorem, 49 Transpose, 31.9 Transposition, 182 Trichotomy law, 73

Wedderburn’s Theorem, 461

Well-ordering, 73 Wilson’s Theorem, 19.14

Z Zero algebra, 311

Trivial algebra, 311

Zero-divisor, 125 proper, 125

Trivial ring, 124 Truth set, 69

ring without, 125 U

Unary operation, 485 Union, 20, 24

Zero element, 27 Zero idea], 165 Zero matrix, 304

Zero ring, 123