TOPICS IN ALGEBRA

I. N. HERSTEIN
University of Chicago

GINN
WALTHAM,

AND

COMPANY
TORONTO· LONDON

A Xerox Company
MASSACHUSETTS·

To MARIANNE

Copyright

©

1964, by Ginn and Company

AU Rights Reserved
LIBRARY OF CONGRESS CATALOG CARD NUMBER:

63-17982

Preface
The idea to write this book, and more important the desire to do so, is a direct outgrowth of a course I gave in the academic year 1959-1960 at Cornell University. The class taking this course consisted, in large part, of the most gifted sophomores in mathematics at Cornell. It was my desire to experiment by presenting to them material a little beyond that which is usually taught in algebra at the junior-senior leveL I have aimed this book to be, both in content and degree of sophistication, about halfway between two great classics, A Survey oj Modem Algebra by Birkhoff and MacLane and Modem Algebra by Vander Waerden. The last few years have seen marked changes in the instruction given in mathematics at the American universities. This change is most notable at the upper undergraduate and beginning graduate levels. Topics that a few years ago were considered proper subject matter for semiadvanced graduate courses in algebra have filtered down to, and are being taught in, the very first course in abstract algebra. Convinced that this filtration will continue and will become intensified in the next few years, I have put into this book, which is designed to be used as the student's first introduction to algebra, material which hitherto has been considered a little advanced for that stage of the game. There is always a great danger when treating abstract ideas to introduce them too suddenly and without a sufficient base of examples to render them credible or natural. In order to try to mitigate this, I have tried to motivate the concepts beforehand and to illustrate them in concrete situations. One of the most telling proofs of the worth of an abstract concept is what it, and the results about it, tells us in familiar situations. In almost every chapter an attempt is made to bring out the significance of the general results by applying them to particular problems. For instance, in the

"

vi

PREFACE

chapter on rings, the two-square theorem of Fermat is exhibited as a direct consequence of the theory developed for Euclidean rings. The subj ect matter chosen for discussion has been picked not only because it has become standard to present it at this level or because it is important in the whole general development but also with an eye to this "concreteness." For this reason I chose to omit the J ordan-Holder theorem, which certainly could have easily been included in the results derived about groups. However, to appreciate this result for its own sake requires a great deal of hindsight and to see it used effectively would require too great a digression. True, one could develop the whole theory of dimension of a vector space as one of its corollaries, but, for the first time around, this seems like a much too fancy and unnatural approach to something so basic and down-to-earth. Likewise, there is no mention of tensor products or related constructions. There is so much time and opportunity to become abstract; why rush it at the beginning? A word about the problems. There are a great number of them. It would be an extraordinary student indeed who could solve them all. Some are present merely to complete proofs in the text material, others to illustrate and to give practice in the results obtained. Many are introduced not so much to be solved as to be tackled. The value of a problem is not so much in coming up with the answer as in the ideas and attempted ideas it forces on the would-be solver. Others are included in anticipation of material to be developed later, the hope and rationale for this being both to lay the groundwork for the subsequent theory and also to make more natural ideas, definitions, and arguments as they are introduced. Several problems appear more than once. Problems, which for some reason or other seem difficult to me, are often starred (sometimes with two stars). However, even here there will be no agreement among mathematicians; many will feel that some unstarred problems should be starred and vice versa. Naturally, I am indebted to many people for suggestions, comments and criticisms. To mention just a few of these: Charles Curtis, Marshall Hall, Nathan Jacobson, Arthur Mattuck, and Maxwell Rosenlicht. I owe a great deal to Daniel Gorenstein and Irving Kaplansky for the numerous conversations we have had about the book, its material and its approach. Above all, I thank George Seligman for the many incisive suggestions and remarks that he has made about the presentation both as to its style and to its content, I am also grateful to Francis McNary of the staff of Ginn and Company for his help and cooperation. Finally, I should like to express my thanks to the John Simon Guggenheim Memorial Foundation; this book was in part written with their support while the author was in Rome as a Guggenheim Fellow.

Contents
1.
1. Set Theory . 2. Mappings 3. The Integers. 2. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.
GROUP THEORY PRELIMINARY NOTIONS

2

9 17

Definition of a Group . Some Examples of Groups . Some Preliminary Lemmas Subgroups A Counting Principle . Normal Subgroups and Quotient Groups Homomorphisms . Automorphisms . Cayley's Theorem Permutation Groups Another Counting Principle Sylow's Theorem.

26
27

29
32 38

41
46 56 60 64

69 78

3.
2. 3. 4. 5. 6. 7. 8.

RING

THEORY

1. Definition and Examples of Rings

83
88 92
94

Some Special Classes of Rings Homomorphisms. Ideals and Quotient Rings . More Ideals and Quotient Rings The Field of Quotients of an Integral Domain Euclidean Rings . A Particular Euclidean Ring
vii

98
101 104

110

viii

CONTENTS

9. Polynomial Rings 10. Polynomials Over the Rational Field 11. Polynomial Rings Over Commutative Rings . 4. 1. 2. 3. 4. 5.
VECTOR SPACES AND MODULES

113

120 122

Elementary Basic Concepts Linear Independence and Bases Dual Spaces . Inner Product Space Modules

130 136

144

150 160

5.
1. 2. 3. 4:. 5. 6. 7.

FIELDS

Extension Fields . The Transcendence of e Roots of Polynomials . Construction with Straight-Edge and Compass More About Roots The Elements of Galois Theory Solvability by Radicals 6.
LINEAR TRANSFORMATIONS

167 175 179 187 191 195 208

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

The Algebra of Linear Transformations Characteristic Roots . Matrices Canonical Forms: Triangular Form Canonical Forms: Nilpotent Transformations Canonical Forms: A Decomposition of V: Jordan Form Canonical Forms: Rational Canonical Form. Trace and Transpose. Determinants Hermitian, Unitary, and Normal Transformations Real Quadratic Forms

217 225 228 241 248 255 262 271 279 294 308

7.

SELECTED TOPICS

1. Finite Fields 2. Wedderburn's Theorem on Finite Division Rings . 3. A Theorem of Frobenius . 4. Integral Quaternions and the Four-Square Theorem
INDEX

314 318 326 329
337

Topics in Algebra .

from the point of view of someone schooled in the calculus and who is seeing the present material for the first time. topology. number theory. as the occasions arise. and along with this. simple new proofs of difficult classical results. An algebraic system can be described as a set of objects together with some operations for combining them. analysis." Be that as it may. fields. led us to open up whole new areas of mathematics whose very existence had not even been suspected. with operations-it will be necessary to consider sets in general and some notions about them. It is the purpose of this chapter to discuss these and to derive some results about them which we can call upon. This book is intended as an introduction to that part of mathematics that today goes by the name of abstract algebra. rings. The isolation of a problem into its basic essentials has often revealed for us the proper setting. Prior to studying sets restricted in any way whatever-for instance. At the other end of the spectrum. The term "abstract" is a highly subjective one. and vice versa. in fact. what is abstract to one person is very often concrete and down-to-earth to another. the set of integers. This has given rise to a large body of new results and problems and has. The algebra which has evolved as an outgrowth of all this is not only a subject with an independent life and vigor-it is one of the important current research areas in mathematics-but it also serves as the unifying thread which interlaces almost all of mathematics-geometry. we shall need some information about the particular set. of results considered to have been special and apart and has shown us interrelations between areas previously thought to have been unconnected. later in the book. and even applied mathematics. it could be described as "not too abstract". 1 .CHAPTER 1 Preliminary Notions One of the amazing features of twentieth century mathematics has been its recognition of the power of the abstract approach. in the whole scheme of things. vector spaces. we shall concern ourselves with the introduction and development of some of the important algebraic systems-groups. it may very well be described as "quite abstract. In the wake of these developments has come not only a new mathematics but a fresh outlook. In relation to the current research activity in algebra.

as S ::::> A) which may be read as "A is contained in S" (or. what is meant by the equality of two sets? For us this will always mean that they contain the same elements. written A = B. 3}. The null set is the set having no elements. n = 0. . 1. DEFINITION. is the set {xix E:: A or x E:: B}. For those whose tastes run more to the formal and abstract side.." In ordinary English when we say that something is one or the other we imply that it is not both. if a E:: A implies a E:: S. We shall not attempt a formal definition of a set nor shall we try to lay the groundwork for an axiomatic theory of sets. The standard device for proving the equality of two sets. (10). A subset A of S will be called a proper subset of S if A c S but A ~ S (A is not equal to S). purely notational. (4)." For instance. Instead we shall take the operational and intuitive approach that a set is some given collection of objects. then we can describe A as A = {a E:: S I a > O}. The mathematical "or" is quite different. finite or infinite. and in fact we shall. we can carry out the same procedure for any number of sets.2 PRELIMINARY NOTIONS CH. Set Theory. . at least when we are speaking about set theory. is to demonstrate that the two opposite containing relations hold for them. remark: Given a set S we shall constantly use the notation A = {a E:: S I P(a)} to read "A is the set of all elements in S for which the property P holds. (10) could be described by A = {(i) E:: S I i = 3n + 1. if both A c Band B c A. that is. because it illustrates the general construction but is not obscured by the additional notational difficulties. Another example of this' If S is the set consisting of the objects (1). if S is the set of integers and if A is the subset of positive integers. (7). This notation is not meant to preclude the possibility that A = S. the two sets A and B are equal. something we shall be required to do often. we can consider a set as a primitive notion which one does not define. a q_ S will read "a is not an element of S." In the same vein. One final. We shall often describe that a set S is the null set by saying it is empty. then the subset A consisting of (1). We do so for two first. written as A U B. Given two sets we can combine them to form new sets. 1 1. In most of our applications we shall be dealing with rather specific things and the nebulous notion of a set. We shall write this as A c S (or. A few remarks about notation and terminology. 2. For . Given a set S we shall use the notation throughout a E:: S to read "a is an element of S. There is nothing sacred or particular about this number two.. it is a subset of every set. every element which is in one is in the other. In terms of the symbol for the containing relation. and vice versa. sometimes. in these. (2). A word about the use of "or. that is. will emerge as something quite recognizable. By the way. S contains A)." The set A will be said to be a subset of the set S if every element in A is an element of S. The union of the two sets A and B.

Yl>lJ2l. X2. B that on the right. . and may be in both. whenever B is a subset of A. If A is the set of all blondehaired people and if B is the set of all people who smoke. xd. 1 SE'l' THEORY ::l when we say that x is in A or x is in B we mean x is in at least one of A 01' B. A is the circle on the loft. A U A = A. then A n B is the set of all blonde-haired people who smoke. X:3. X2. then A n B = 1xI! (we are supposing no y is an x). If A is the set of all blonde-haired people and if B is the set of all people that smoke. it n B. in fact. A n A = A. Pictorially we can illustrate the intersection of the two sets A and B by: Here A is the circle on the left. written as X E::_ B l. then A U B consists of all the people who either have blonde hair or smoke or both. if B is any subset of A then A n H = B. let us see what the intersections are in those very examples. x3l and B the set 1Ill. If A is the set {Xl> .rt!. . DEFINITION. Y2. In analogy with the examples used to illustrate the union of two sets. B that on the right. For any set A. is the set {x IX E::_ A and The intersection of the two sets A and H. Yz. A U B = A. then A U B = {Xl. Pictorially we can illustrate the union of the two sets A and B by: Here. X2. Xg) and if B is the set (YI.1'2. Let us consider a few examples of the union of two sets.e. the set whose elements are Xl. and A U B is the shaded part. For any set A. while their intersection is the shaded part.. in fact.SEC. xgl (i. If A is the set (Xl. The intersection of A and B is thus the set of all elements which an' both in A and in B.

namely. it is immediate that A nBc A n (B U C). we mean the set {x I x E: A" for at least one a in T}. We continue the discussion of sets to extend the notion of union and of intersection to arbitrary collections of sets. Suppose the former. The index set T can be any set. Because Be B U C. The proof will consist of showing. Thus in either eventuality x E: (A n B) U (A n C). finite or infinite. whence A n (B U C) c (A n B) U (A n C). where a is in T. We first dispose of (A n B) U (A n C) c A. Now for the other direction. some of these can be found in the problems at the end of this section. The two opposite containing relations combine to give us the equality asserted in the proposition. as an element in B U Cit is either in B or in C. n n n first of all it must be an element of A. B. Secondly. the relation (A n B) U (A n C) cAn (B U C) and then the converse relation A (B U C) c (A B) U (A C). T can be any (nonempty) set. C we have A n (B U C) = (A n B) U (A n C). if S is the set of real numbers. 1 Two sets are said to be disjoint if their intersection is empty. leads us to x E: A n C. The second possibility. if A is the set of positive integers and B the set of negative integers. This is the first of a whole host of such results that can be proved. x E: C. we should like to prove a little proposition interrelating union and intersection. we repeat. Note however that if C is the set of nonnegative integers and if D is the set of nonpositive integers. x must be in A n B. For instance. For instance. Very often we use the set of nonnegative integers as an index set.4 PRELIMINARY NOTIONS CH. that is. For any three sets. A n C cAn (B U C). Aa n A (3 is the null set. then A and B are disjoint. A. then as an element both of A and of B. Proof. We shall denote it by U A". Therefore (A n B) U (A n C) c (A n (B U C» U (A n (B U C» = A n (B U C). so is not empty. is the null set. The sets Aa are mutually disjoint if for a r!E /3. to begin with. and if T is the set of rational aCT n . PROPOSITION. Given a set T we say that T serves as an index set for the family 5' = {Aa I of sets if for every a E: T there exists a set Aa in the family 5'. By the intersection aCT of the sets A". we shall denote it by Aa. Before we generalize union and intersection from two sets to an arbitrary number of them. but. where a is in T. then they are not disjoint for their intersection consists of the integer 0. we mean the set {xix E: A" for every a E: T}. By the union of the sets Aa. In a similar manner. n (B U C). Given an element x E: A n (B U C).

1 SE'!' '!'HEORY 5 numbers. their Cartesian product A X E.E) is the null set. Finally.B is the shaded area. Given three sets A.2: a}. we call A .E the complement of E in A.E. b E::_ 15. the set i1 satisfies A = (A n B) U (A .r q_ E}. The set of elements (a.r . we feel that these sets must be closely related. also. E then the difference set. Note that if the set A is a finite set having n elements. To see the exact relation between them we shall have to wait until the next section. where D = 15 X C. where we discuss one-to-one correspondences. As sets these are distinct. the set A X D. Given the two sets Ll and B we could construct the sets 11 X 15 and B X 11from them. when E is a subset of A. then the set A X A is also a finite set. c) where a E::_ A.. A .E). In that case. we do not bother to define it. we can consider the Cartesian product of a set "1 with itself. /1. let. is the set {:r E::_ A!.\. A particular case of interest of the difference of two sets is when OJl(. Given any index set T we could define the Cartesian product of the sets A" as a varies over T. the set E X C. E that on the right. of these is a subset of the other. The sets A" are not mutuallv aC_T "C_T n disjoint. Returning to our little pictures. for a E::_ T. b) where a E::_ A and b E::_ E and where we de dare the pair (aI. Given the two sets A. A" = [:r E::_ 8!. These give us three distinct sets. and c E::_ C. if A is the circle on the left. A few remarks about the Cartesian product. bl) to be equal to (a2' b2) if and only if al = a2 and b] = b2. a) in A X A is called the diagonal of 11 X A. then A . yet we feel that they must be closely related.SEC. C we can construct many Cartesian products from them: for instance. We still want one more construct of two given sets A and 15. This set A X 15 is defined as the set of all ordered pairs (a. . E. but has n2 elements. A X A. It is an easy exercise to see that U A" = 8 whereas A" is the null set. and also the set of all ordered triplet. since we shall not need so general a product. we can continue this process with more and more sets. Of course. Note that for any set E. yet here. where E = 11 X B. (Prove 1) Note further that E n (A . DEFlNITIO. (a.

Instead of speaking about subsets of AX A we can speak about a binary relation (one between two elements of A) on A itself. and the third. (1) Since 0 = a . b. In fact. have corresponding angles equal). b implies b '" a. c) E:: R. a '" b if a .c = (a . Define for a. (3) If a band b c. define a '" b if a . Let S be the set of all integers and let n > 1 be a fixed integer. for a. b. whence b a. . (2). then both a . Given a. r-. r-. (3) of the subset R immediately translate into the properties (1). b E:: S. a = Example 1.J t'.b is a multiple of n. b) E:: Rand (b. transitivity. Let S be the set of all triangles in the plane.. a) E:: R.' Example 3. The properties (1). if a . defining b to be related to a if Ca.(a . The first of these properties is called reflexivity. We illustrate it with a few examples.c) is also even. c) E:: R imply that (1) (a. . proving that a r-« c.b is even. We leave it as an exercise to prove that this defines an equivalence relation on S. (2) If a b.. Example 4. Let S be any set and define a b. The binary relation.b is an even integer. (2). This defines an equivalence relation on S.6 PRELIMINARY NOTIONS CR.band b . whence a ..' r-« t'.b) + (b . The concept of an equivalence relation is an extremely important one and plays a central role in all of mathematics.e.a = . (3) of the definition below. if and only if This clearly defines an equivalence relation on S. that is.a is even. c in A: (1) a (2) a r-« r-v f"'./ (3) a a. a) (a. the second. (3) (a. Two triangles are defined to be equivalent if they are similar (i. 1 A subset R of A X A is said to define an equivalence relation on A if: E:: R for all a E:: A. DEFINITION./ c. b) E:: R. Let S be the set of all integers. b E:: S. an equivalence relation is a generalization of equality.c are even.-.J Example 2..b) is also even. b E:: S.. b) E:: R implies (b. (2) (a. a r-v a.." n A is said to be an equivalence o relation on A if for all a. measuring equality up to some property. band b r-v c imply a 1"'. We verify that this defines an equivalence rela- tion of S. symmetry. then b .

The arguI"V. x b. that is..J I"V I"V . from a band b y. cl(n . Although we have made quite a few definitions. introduced some concepts.. we can define an equivalence relation on A for which these subsets are the distinct equivalence classes.. thus b rv y. Therefore. x whence by the symmetry of the relation. For. a must be in cl( a). However.. the equivalence class of a consists of all integers of the form a len where k = 0.. + + THEOREM Lx.. ±2. every element in cl(b) is in cl(a). we shall encounter a few as we proceed in the book.1). Proof._ . . b.... suppose that cl(a) and cl(b) are not disjoint. in this example there are only two distinct equivalence classes.. . Weare now about to prove the first genuine result in the book. ±l.. There are many more equivalence relations. which proves that cl(b) C cl(a). ±2. then the equivalence class of a E:: A is the set [z E:: A [a . Since x E:: cl(a). namely cl(O) ... The proof of this theorem is not very difficult-actually it is quite easybut nonetheless the result it embodies will be of great use to us. W e write it as z cl(a) . and have even established a simple little proposition. nonempty subsets. cl(l). since z E: cl(b)...) by the transitivity of the relation forces a b. here there are n distinct equivalence classes. In Example 5. namely. A simple check verifies that this defines an equivalence relation on S... If A is a set and if is an equivalence relation on A. ±1. In Example 3. now. Suppose. 1 SET THEORY 7 Example 5. one could say in all fairness that up to this point we have not proved any result of real substance. whence a. we deduce that a "-' y.. . However. given a decomposition of A as a unwn of mutually disjoint. ... Let S he the set of points in the plane. In Example 2 the equivalence class of a consists of all the integers of the form a 2m where m = 0. the equivalence class of a consists of all the points in the plane which lie on the circle which has its center at the origin and passes through a. that y E: cl(b)...SEC. a rOoJ x and x.. ]. r"'V In the examples just discussed.. that y E: cl(a). a . We now assert that given two equivalence classes they are either equal or disjoint. Conversely. I"V r-o. what are the equivalence classes? In Example 1. The distinct equivalence classes of an equivalence relation on A provide us with a decomposition of A as a union of mutually disjoint subsets.. Two points a and b are defined to be equivalent if they are equidistant from the origin. a rOoJ x. DEFINITION. then there is an element x E: cl(a) n cl(b). Let the equivalence relation on A be denoted by We first note that since for any a E:: A.. the equivalence class of a consists merely of a itself. cl(O) and cl(l).. the union of the cl(a)'s is all of A.

(b) Prove that (A n B) n C = A n (B n C). B E: S* (remember.8 PRELIMINARY NOTIONS CH. If A and B are finite sets prove o(A U B) = o(A) + o(B) . (a) If A is a subset of Band B is a subset of C. nonempty sets (a is in some index set T). (c) If B c A prove that for any set C both B U C c A U C and B n C cAn C./s. A survey shows that 63% of the American people likes cheese whereas 76% likes apples. and conversely.oCA n B). Given two sets A and B their symmetnc d'ijJerence is defined to be (A . Prove that A U (B n C) = (A U B) n (A U C). Let S be a set and let S* be the set whose elements are the various subsets of S. This proves the first half of the theorem. b E: A. PROBLEMS 1. 3.A). (b) If B c A prove that A U B = A. In S* we define an addition and multiplication as follows: If A. this means that they are subsets of S) : n n (1) A (2) A·B +B = = (A . We have thus shown that the distinct cl(a)'s are mutually disjoint and that their union is A. 9. For a subset C of S let C' denote the complement of C in S. .A). For any two subsets A. B of S prove the De Morgan rules: (a) (A n B)' = A' U B'.(A n B). (a) Prove that A B =B A and A U B = B U A. whence we conclude that cl(a) c cl(b).. where the A".B) U (B .B) U (B . Now for the other half! Suppose that A = U A". What can you say about the percentage of the American people that likes both cheese and apples? (The given statistics are not meant to be accurate. How shall we use them to define an equivalence relation? The way is clear. 6. Prove that the symmetric difference of A and B equals (A U B) . If A is a finite set having n elements. The two opposite containing relations imply that cl(a) = cl(b). given an element a in A it is in exactly one A".) 8. are mutually disjoint. a r-« b if a and b are in the same A".. prove that A is a subset of C. For a finite set C let o(C) indicate the number of elements in C. A n B.. 4. 7.l ment is clearly symmetric. 5. prove that A has exactly 2n distinct subsets. 2. We leave it as an exercise to prove that this is an equivalence relation on A and that the distinct equivalence classes are the A. (b) (A U B)' = A' n B'. We define for a.

a r-« b if a and b have the same father... (b) S is the set of all people in the world today. 13. (e) If A B =A C then B = C.. a r-« b if a is parallel to b.. 2 MAPPINGS 9 Prove the following laws that govern these operations: (a) (A (b) A·(B (c) A·A + B) + C = A + (B + C). a b if a lives within 100 miles of b. . For the given set and relation below determine which define equivalence relations. namely.) 10. then b .. When we were asked to plot the relation y = x2 we were simply being asked to study the particular mapping which takes every real number onto its square... by property 3 (using a = c). into another./ I""V 2. 11. (f) S is the set of all straight lines in the plane. c then a .a. = A·B + A·C.. (e) S is the set of integers.SEC. cl(n . = + C) A. We shall define a mapping somewhat more formally and precisely but the purpose of the definition is to allow us to think and speak . for we have been considering mappings from the very earliest days of our mathematical training. cl(l). a r--: b if both a > band b > a. Mappings. S. a.../ r--. (a) S is the set of all people in the world today. prove that the relation defined is an equivalence relation and that there are exactly n distinct equivalence classes. What is wrong with the following proof that properties 2 c.. a r-« b if a = ±b. In Example 3 of an equivalence relation given in the text. (The system just described is an example of a Boolean algebra.. property 3 states that if a band b. a b if a and b have an ancestor in common.. cl(O). Without exaggeration this is probably the single most important and universal notion that runs through all of mathematics. whence. We are about to introduce the concept of a mapping of one set into another. (a) Property 2 of an equivalence relation states that if a"" b then b . (c) S is the set of all people in the world today.. and 3 imply property I? Let a r-« b. a"" a.. Loosely speaking. Complete the proof of the second half of Theorem 1.... is a "rule" (whatever that may mean) that associates with each element in S a unique element tin T..a. .. (d) A + A = nuli set.. T... (d) S is the set of real numbers..1). (b) Can you suggest an alternative of property 1 which will insure us that properties 2 and 3 do imply property I? 12. It is hardly a new thing to any of us. a mapping from one set.. + + r--..

DEFINITION. Instead we do prefer to think of a mapping as a rule which associates with any element s in IS some element tinT. If t is the image of sunder we shall sometimes write this as (J":S~ t. the set of all pairs (a. ote that we write the mapping on the right. As a general rule. IS X IS. more often. M. other mathematicians writing them on the left. no more or less. Now for some notation for these things. 1 in the above terms. t) is in M. This definition serves to make the concept of a mapping precise for us but we shall almost never use it in this form. (J" (J" (J" . In this plane we single out all those points whose coordinates are of the form (x." We illustrate this with the familiar example y = x2 defined on the real numbers S and taking its values also in IS. we shall represent this fact by t = S(J". We shall say that t is the image of s under the mapping. If IS and Tare nonempty sets. The point of view we take is to consider the mapping to be defined by its "graph. In fact. For this set IS. the rule being. N There is no over-all consistency in this usage. we shall not be absolutely consistent in this ourselves. the pair Ca. we often denote this by writing (J":S ~ T or IS ~ T. This is. associate (or map) s E:: IS with t E:: T if and only if (s. b) can be viewed as the plane. of IS X T such that for every s E:: IS there is a unique t E:: T such that the ordered pair (s. algebraists write mappings on the right. when we shall want to emphasize the functional nature of a we may very well write t = (J"(s). We should think of them as rules or devices or mechanisms that transport us from one set to another. respectively. the approach we take in the general setting to define a mapping from one set into another.10 PRELIMINARY NOTIONS CH. Let be a mapping from IS to T. x2) and call this set of points the graph of y = x2• We even represent this set pictorially as To find the "value" of the function or mapping at the point x = a we look at the point in the graph whose first coordinate is a and read off the second coordinate as the value of the function at x = a. Let us motivate a little the definition that we will make. many people would write it as t = O"(s). then a mapping from IS to T is a subset. t) E:: M. b) corresponding to the point whose coordinates are a and b.

n) in 8. . We call 8* the set of subsets of 8. Given such a mapping -: 8 X 8 -+ 8 we could use it to define a "product" * in 8 by declaring a * b = c if (c. n)r =- m n for Example 5. Let 8 be the set of integers and let T be the set consisting of the elements E and O. namely. If 8 is any set. define . Define r:8 -+ J by (m. n) -+ E: J X J I n . for any 8 E: 8. Xl) X2. Define r : 8 -+ T = (m. . XI. the set whose elements are the subsets of 8..Define r:8 -+ T by declaring nr = E if n is even and nr = 0 if n is odd. let T be the set of rational numbers. xnl be its subset consisting of the elements ••• . define r:8 every (m. X2r = Xa. Let J be the set of integers and 8 = {(m.:8 ---+ 8 by This mapping' is called the identity mapping of 8. We could similarly define the projection of 8 X T on T. n) Example 4. b)r = a for any Ca.z£ 0 I. if 8 = {Xl. EXAMPLES OF MAPPINGS. X2} then 8* has exactly 4 elements. al = null set. Let 8 be any set. b)r = c Example 6. {x I is the subset of 8 whose only element is x. Xg. 2 MAPPINGS 11 nonempty. . In all the examples the sets are assumed to be Example 1. This -r is called the projection of 8 X Ton 8. Xar = Xl. Given an arbitrary set 8 we call a mapping of 8 X 8 into 8 a binary operation on 8. Xn of 8.SEC. X2. Given a rational number s we can write it as s by Sr =- m n where m and n have no common factor. n)r = m + n. Given 8 we can use it to construct a new set 8*. Let 8 be the set consisting of the elements r:8 -+ 8 by xlr = X2. Let J be the set of integers and 8 = J X J. Let 8 be the set of positive rational numbers and let T = J X J where J is the set of integers. Example 3. Example :2. define r:8 X T -+ 8 by (a. 8 = 8. Let 8 and T be any sets. let {Xl. Example 7. Note that in Example 5 the addition in J itself can be represented in terms of a mapping of J X J into J. Define r:8 ---+ T by r: 8 ---+ to for every 8 E:: 8. Define Example 8. In particular. b) E: 8 X T. Let 8 and T be any sets and let to be an element of T. Thus for instance. T by (m.

~ In terms of inverse images.12 PRELIMINARYOTIONS N CH. S7 = {x E:: T I z = S7 for some S E:: 8} the image of Sunder 7 then 7 is onto if the image of 8 under 7 is all of T.{s}. In the examples discussed. as an element in 8*. the image of any element in 8 under each of these mappings should be the same. Define 7:8 ~ T by S7 = cl(s) . 1 az = the subset. 3. T = 8*. Given a mapping 7:8 ~ T we define for t E: T. the mapping 7 is one-to-one if for any t E:: T the inverse image of t is either empty or is a set consisting of one element. When should we say that two mappings from 8 to T are equal? A natural definition for this is that they should have the same effect on every element of S. If we call the subset. 6. define 7:8 ~ 'I' by of (s} in 8 = 8 . Let S be a set with an equivalence relation. and 10 the mappings used are all onto. in Example 9. DEFINITION. 8. 5. is a very interesting one. The relation of 8 to 8*. is not the image under the 7 used of any element in 8.2) is not the image of any element in 8 under the 7 used. he two mappings T = S7 for every s E:: S. 8. the mappings in Examples 1. of 8. Note that in Examples 1. In Example 3 which was discussed the element (4. In Example 8. and let T be the set of equivalence classes inS (note that T is a subset of 8*). i is not the image under 7 of any element in S. a- and 7 of 8 into T are said to be equal Sa- Consider the following situation: We have a mapping a from S to T and another mapping 7 from T to U. Can we compound these mappings to .The mapping t E: T there exists an element 7 S of S into T is said to be onto T if given E: 8 such that t = S7. Let S be a set. 4. 7. In a little more formal manner: if DEFINITION. a4 = {x2l. S7 = complement Example 10. 7. 8. DEFINITION. a3 = {xd. in general. Another special type of mapping arises often and is important: the oneto-one mapping. that is. and 9 are all one-to-one. We leave the examples to continue the general discussion. It may happen that for some t in T that its inverse image with respect to 7 is empty. that is. the inverse image of E is the subset of S consisting of the even integers. Example 9. the inverse image of t with respect to 7 to be the set (s E: 8 I i = S7 [. some of its properties are examined in the problems. he mapping T if whenever SI ~ S2 then SIT 7 of S into T is said to be a one-to-one mapping S27.

f u:8 -+ T and 7: T -+ U then the composition of a and 7 I (also called their product) is the mapping a 7:8 -+ U defined by means of 7) = (SU)7 for every s E::: 8. n)r)o. = by X37 XlU = X2.0 7) whence a 0 7 . first by applying a to s and then applying 7 to the resulting element ea in T. n)(r u) = ((m.SEC.1). We reiterate.1. 0 -+ 8 by Xl7 = X2. it is not even an onto mapping of T. Note U that in this case roo- cannot be defined. the left-right business is not a uniform one. and suppose u:8 -+ T is defined by me = (m .:C 700-. whereas 70 a-: T -+ ']'. in 8. Define 0-:8 -+ T by sa = largest integer less than or equal to 8. X3(7 0 0-) = (X3T)U = X20- = X3' Xl (0. n7 = 0 if n is odd. even to speak about the equality of 0. in two stages into U. 0 Example 2. a 0 7 reads. X2U = X3. Xsu = = x31 and let T Xl.. and r:8 = 8.1.0 7 and 7 a would make no sense since they do not act on the same space. Thus a 0 7:8 -+ 8 + 0 0 + 0 Example 3. X27 = Xa. whereas = = Xl. because in this case it also makes sense. (m. We now compute 0.1) 1 = m. n)7 = m + n. X3(0. and r: T -+ U defined by n7 = E if n is even. for us a 0 7 will always mean: first apply a and then 7. in reading a given book in mathematics one must make absolutely sure as to what convention is being followed in writing the product of two mappings. too.0 7 as a mapping of 8 into itself and then 7 0 U as one on T into itself. We compute 0- 0 7 for two real . This is the basis of the DEFINITION. Now XI(7 u) = (x17)0-= (X3u)r = Thus Xl (u 0 7) = (XIU)7 = XZ7 = XS. T the set of integers. We illustrate the composition of a and 7 with a few examples. 0 0 seu Note that the order of events reads from left to right.0 7 is the identity mapping of 8 into itself. Mathematicians who write their mappings on the left would read a 7 to mean first perform 7 and then a. O}. = XST = X2.= (m + n)o. Note that 7 0 a is not the identity map of T into itself. X2. Given m E::: 8. and = {E. Accordingly. Let 8 be the set of integers. n) E::: T. n)7 = m n whereby (m. 0 (XIO-) = X2. Let 8 = IXl.1.= (m + n . 1) whence m(o. xz(u 0 7) = (X2U)7 = Xl7 Xl.1.1). Let U = 8 and suppose that 7:T -+ U(= 8) is defined by (m. 2 MAPPINGS 13 produce a mapping from 8 to U? The most natural and obvious way of doing this is to send a given element s. X2(7 0 u) = (X27)00 Note that X2 = = X30Xs XI(7 e). first perform a and then follow it up with 7. Let u:8 -+ 8 be defined Xl. 0 Example 1.T} = (mu)7 = (m . Here. T the set S X 8. Thus 0. What about T u? Given (m. 1)7 = (m .07) At the same time we can compute T a. Let 8 be the set of real numbers. rna = (m .

This proves the lemma. Let 0":8 ~ T T and T: T ~ U. Since T is one-to-one and 810" and 820" are distinct elements of T. by the "one-to-oneness" of a this s is unique. We have shown that a 0"-1 is the identity mapping of S onto itself. if cr:S ~ T is such that there exists a u: T ~ S with the property that a p.: U ~ V. and p. For mappings of sets. by the "onto-ness" of a there exists an element 8 E::: S such that t = se-. proving that a T is indeed one-to-one. Proof. We prove only part (2).) = (S(O"OT»P. We should like to show that if two mappings 0" and T are properly conditioned the very same conditions carryover to 0" T.».14 PRELIMINARY NOTIONS CR. 0 0 0 0 0 0 0 0 .). By the one-to-one nature of 0". a are the identity mappings on Sand T. then (O"OT) u. s«O" T) p. provided the requisite products make sense. This is the content of LEMMA + 1. Suppose that a is a one-to-one mapping of S onto T. To prove the asserted equality we merely must show that for any s E::: 8. we call a a one-to-one correspondence between Sand T. respectively. Let us compute a cr-1 which maps S into itself. then (1) 0" 0 r ie onto if each of 0" and T is onto. (1)0" = 2.and p.) and s(O" (T p. of (0" T) P. (810")T ~ (S20")T whence SI(O" T) = (Slo")T ~ (820")T = 82(cr T).1 o (ASSOCIATIVE LAW). The mapping 0"-1 is called the inverse of cr. Conversely. whence (1) (0" 0 T) = (10")T = (2)T = Ej (7C")0"= 3. thus (0" T) Palso makes sense and takes 8 into V.).. and establishing the lemma.) = 8(0" (r p. 810" ~ S20".» are indeed equal.2.whereass(O" (TOp. or lack of equality. Now since 1= 2 j. Thus. Suppose that S1> S2 E::: 8 and that SI ~ 82. thus 8(0" cr-1) = (8cr)0"-1 = tcr-1 = s. the elements s«O" T) p.) = «So")T)p. If 0":8 ~ T. whence by definition 8 = tcr-1. T: T ~ U. Note first that 0" T makes sense and takes 8 into U. (2) a 0 is one-to-one if each of 0" and T is one-to-one. then we claim that a is a one-to-one correspondence between 8 and T. whence 7C"(0" T) = (7C"o")T= 0 (3)T = O. = «So")T)p. and 0" (T p. We define the mapping cr-1:T ~ 8 by 8 = to"-1 if and only if t = sa. Similarly 0" (T p.)is meaningful and takes 8 into V. Given 8 E::: S let t = se. 1 numbers S = 1 and S = 7C". a general associative law holds. A similar computation reveals that 0"-1 00" is the identity mapping of T onto itself. Given any t E::: T. Thus we can speak about the equality. leaving the proof of part (1) as an exercise. Now by the very definition of the composition of mapastje e ») op.» = (srr)(T p. = 0"0 (TOp. 0 0 0 Proof. 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 LEMMA 1.

Thus 0" 0 r ~ -r o rr. Xs.3. X37' = X2. For this reason we state the next theorem concerning its nature. Next observe that 0" is one-to-one for if S10" = S20". r. (2) (O"or) 0Jl = 0"0 (TOJl).n. This is 0 LEMMA 1. let Xl. and sr = s for any s E: 8 different from X2. and determine the inverse image of any i C T under CT. Xa be three distinct elements in 8. t = t(IL 0 0") = (tJl)O" (since Jl CTis the identity on T) and so t is the image under 0" of the element tJl in 8. The mapping 0":8 --lo T is a one-to-one correspondence between 8 and T if and only if there exists a mapping Jl: T --lo 8 such that 0" Jl and Jl CTare the identity mappings on 8 and T. .SEC. respectively. one-to-one. X20" = Xs. - = L 0 0 0 We close the section with a remark about A(8). T = set of nonnegative real numbers. A simple computation shows that Xl(O" r) = Xs but that Xl(7' 0 0") = X2 ~ Xa. Jl are elements of A(8). (a) 8 = set of real numbers. 0 we can find two elements PROBLEMS 1. define the mapping 0":8 --lo 8 by XICT = X2. (3) there exists an element 0" 0 L L (the identity map) in A (8) such that CT= 0". Suppose that 8 has more than two elements. (4) there exists an element CT1 E:: A (8) such that 0" 0"-1 = CT1 CT= L. Clearly both 0" and 7' are in A(8). determine whether the 0" is onto. then: (1) 0" 0 T is in A (8) . Aside from its own intrinsic interest A (8) plays a central and universal type of role in considering the mathematical system known as a group (Chapter 2). Define the mapping T:8 --lo 8 by X2r = Xa. 2 MAPPINGS 15 First observe that 0" is onto for. SO" = S2. where 0":8 --lo T. X2. All the constituent parts of the theorem have already been proved in the various lemmas.using that 0" Jl is the identity on 8 we have S1 = S1(0" Jl) = (SlO")Jl = (S20")Jl = S2(0" Jl) = S2. Xa. If 8 has more than two elements CT. so we state the theorem without proof. 0 0 DEFINITION. SO" = 8 for any s E: 8 different from Xl. In the following. THEOREM I. We have now proved 0 0 0 0 LEMMA 1. X2.7' in A (8) such that CT 7' ~ TOO".4. mappings If 8 is a nonempty set then A (8) is the set of all one-to-one of 8 onto itself. XsO" = Xl. If 0". given t E:: T.

9. (b) If 0' 0 T is one-to-one.16 PRELIMINARY NOTIONS CH. u) where s E: 8. prove that there exists a one-to-one correspondence between S X T and T X 8. namely: (a) If 0' T is onto. Prove: (a) The set of integers is infinite. (b) The set of real numbers is infinite. prove that A(S) has n! (n factorial) elements. SO' = S2. 11. If 8 is any set. A set S is said to be infinite if there is a one-to-one correspondence between 8 and a proper subset of S. 5. t. t E: T. it need not be that both 0' and T are onto. is defined by ao A = aa for any a E: A. (b) O'A is one-to-one if 0' is. 13. prove that for any 0' E: A(8). (b) If 0' is a one-to-one mapping of 8 into itself. If the set 8 has a finite number of elements. T = set of integers.2 are false. Prove that there is a one-to-one correspondence between the set of integers and the set of rational numbers. 12. 10. (b) If there is a one-to-one correspondence between 8 and T and one between T and U. 0' 0 L *6. If 8. so = 2s. prove that there exists a one-to-one correspondence between: (a) (8 X T) X U and 8 X (T XU). (c) O'A may very well be one-to-one even if a is not. 3. T. u E: U. prove that (0' T)A = O'A TA. by example. (d) 8 = set of integers. that both part (a) and part (b) are false if 8 does not have a finite number of elements. T = set of integers. (c) Prove. 7. 1 (b) 8 = set of nonnegative real numbers. 4. prove that there exists one between T and 8. If 0':8 ~ T and if A is a subset of 8. . Prove that the converse to both parts of Lemma 1. 0 0 0 = L 0 0' = 0'. (a) If there is a one-to-one correspondence between 8 and T. T = set of nonnegative real numbers. (c) 8 = set of integers. (b) Either set in part (a) and the set of ordered triples (s. 8. the reetriaion of 0' to A. 2. U are nonempty sets. O'A. If 0-8 ~ T and A is a subset of 8 such that A a C A. prove that there is a one-to-one correspondence between 8 and U. prove that it is impossible to find a mapping of 8 onto 8*. Prove: (a) O'A defines a mapping of A into T. it need not be that both 0' and T are one-toone. If L is the identity mapping on 8. If Sand Tare nonempty sets. then 0' is onto. prove: (a) If 0' maps S onto 8 then 0' is one-to-one. sa = S2. If the set 8 has n elements.

respectively. etc. we can say that if it exists then it is unique. This will have to be proved. will occur with their usual meaning. (Note: By the result of Problem 8. Another comment: The mere fact that we have defined what is to be meant by the greatest common divisor does not guarantee that it exists. prove that there is one-to-one correspondence between Sand S X S. We denote that b divides a by b I a. The Integers. the familiar symbols: a > b. a ~ b. This fact is known as the Euclidean algorithm and we assume familiarity with it. To avoid repeating that something is an integer. and that b does not divide a by b t a. b) = (-a. b) = (a. then it is a divisor of mg nh for arbitrary integers m and n. then S must be infinite.SEC. We leave ° °~ ° + the verification of these remarks as exercises. -24) = 12. with b r'" 0. If b I a. we can find m and r such that a = mb + r where r < I b I. However. Ia I. b) for the greatest common divisor of a and b. if we had Cl and C2 satisfying both conditions of the definition above. We say that b ~ divides a if a = mb for some m. we can divide a by b to get a nonnegative remainder r which is smaller in size than b. If Sand T are finite sets having m and n elements. Prove that if S < T and T < U then S < U. Given two sets Sand T we declare S < T (S is smaller than T) if there is a mapping of Tonto S but no mapping of S onto T. prove that if m < n then S < T. that when both alb and bla. then a = ±b. written as lower-case Latin letters will be integers. that is. 3. (2) any divisor of a and b is a divisor of c.) *14. 3 THE INTEGERS 17 (c) If a set S has a subset A which is infinite. DEFINITION. We shall use the notation (a. Since we insist that the greatest common divisor be positive. we make the assumption that all symbols. -b). In this number we include the principle of mathematical induction (which will be used freely throughout the book) and the fact that a nonempty set of positive integers always contains a smallest element. The positive integer c is said to be the greatest common divisor of a and b if: (1) c is a divisor of a and of b. *16. Given a and b. (a. Note that if b is a divisor of g and of h. assuming instead that we already have the set of integers and that we know many of the elementary facts about them. .24) = (60. Note that if a 11 then a = ±I. As to notation.. We close this chapter with a brief discussion of the set of integers. We shall make no attempt to construct them axiomatically. (60. 16. If S is infinite and can be brought into one-to-one correspondence with the set of integers. a set finite in the usual sense is not infinite. in this section. and that anyb r'" divides O. For instance. we call b a divisor of a. for. -b) = (-a.

The sequence 2.. we can find integers m and n If BUCh that ma + nb = 1. c. ~ always has in it some positive integers. Since 0 :::. the negative introduces no essential differences. then (a. Writing this out r explicitly. 7. DEFINITION.5. the insistence on positivity would then force CI = C2. We introduce another familiar notion. by the choice of r c. that of prime number.. -5. ma + nb = t(moa + nob) + r. there are nonzero integers in~. moreover. we have proved that c I x for any z E: ~. Since. -3. -x = (-m)a + (-n)b is also in ~. we exclude 1 from the set of prime numbers. equally. b) = moa nob. 1 then cil C2 and c2l CI. As an immediate consequence of Lemma 1. Nate first that if d I a and d I b.tno)b and so must be in ~. LEMMA1.. we can find integers rno and no such that (a. DEFINITION. and r < c.18 PRELIMINARYOTIONS N CR. we actually prove a little more. Given any element x = ma + nb in ~. C has the form c = moa + nob. b). in ~. then d I (moa + nob). are all prime numbers. whence cia and c] b. But a = 1a + Ob E: ~ and b = Oa + 1b E:~. < c. By this we shall mean an integer which has no nontrivial factorization. r = O. b) = 1. namely that (a. x = tc + r where 0 :::. being in ~. b) exists.. the prime numbers are the building blocks of the integers. But then there is a smallest positive integer. 11. 5. a and bare j'elatively prime. Since one of a or b is not 0. . If a and b are integers. he integer p T > 1 is a prime number if its only divisors are Another way of putting this is to say that an integer p (larger than 1) is a prime number if and only if given any other integer n then either (p. Thus x = tc. b) must have a particular form. b) and so we have proved the lemma. But first we need the important observation.The integers a and b are relatively prime if Ca.. Our first business at hand then is to dispose of the existence of (a. b). Let ~ be the set of all integers of the form ma + nb where m and n range freely over the set of integers. . we have the COROLLARY.tmo)a + (n . We have shown that e satisfies the requisite properties to be (a. are prime numbers.. . As we shall soon see. For technical reasons. ±p. We now must show that c I a and c I b. ±1. not both 0. then by the Euclidean algorithm. for us prime numbers will always be positive. therefore. -2. in the next lemma. Because x = ma + nb is in~. in factoring. whence we would have Cl = ±C2. We claim that c = (a. whence r = (m . 3.5. n) = 1 or pin. whence d I c. In doing so. + Proof.

SEC. has a representation as a product of prime powers." THEOREM I.5. By the induction hypothesis. A short word about this. We have asserted that the prime numbers serve as the building blocks for the set of integers.. J Proof. We shall prove the theorem itself by proving each of these subtheorems separately. consequently.r < k implies the truth of P(k).where PI > P2 > ..t. > Pt are prime numbers and . aJnbc. where each a. If k is not a prime number.6. If a is relatively prime to b but a be. which is precisely the assertion of the lemma. Suppose that any integer r. we shall use the following version of mathematical induction: If the proposition P(mo) is true and if the truth of Per) for all r such that mo :::. Now aJmac and. Proof. > O. We first prove that every integer a > 1 can be factored as a product of prime powers. Thus mac + nbc = c. An immediate question presents itself: How shall we go about proving the theorem? A natural method of attack is to use mathematical induction. we conclude that a c. we can find integers m and n such that ma + nb = 1. Thus k = uv is also such a product. The theorem as stated actually consists of two distinct subtheorems. 3 J THE INTEGERS 19 LEMMA 1.lp2Cl2 ptOl. Following immediately from the lemma and the definition of prime number is the important J COROLLARY. If k itself is a prime number.. being a prime number. aJ (mac + nbc). If a prime number divides the product of certain integers it must divide at least one of these integers. then it is a product of prime powers. We have shown that the truth of the proposition for . since both u and v are less than k. each of these can be factored as a product of prime powers. by the corollary to Lemma 1. our approach is via mathematical induction.r < k can be factored as a product of prime powers. Since mac + nbc = c.. The precise statement of this is the "unique factorization theorem. Certainly mo = 2. We leave the proof of the corollary to the reader. the second assures us that this decomposition is unique. then k = uv where 1 < u < k and 1 < v < k. This variant of induction can be shown to be a consequence of the basic property of the integers which asserts that any nonempty set of positive integers has a minimal element (see Problem 10).c. the first asserts the possibility of factoring the given integer as a product of prime powers. 2 :::. Since a and b are relatively prime. by assumption. Any positive integer a > 1 can be factored in a unique way as a = PIOI. then a c. then Pen) is true for all n 2:: mo.

that is.•.• q. > 0... and in the form used above. If b = 8 = 1 then = . by the corollary to Lemma 1. (2) az = !1z. that is.1 = 8 . In consequence. q8f3s and since al > 0. az a P p/. . Our object is to prove: ° (1) r = 8. For a = 2 this is clearly true. ql > qz > .!3s• We claim that this forces al = f31. and we are (1) the number of distinct prime power factors (in b) on both sides are equal... Thus we see that the assumption of the uniqueness of factorization for the integers less than a implied the uniqueness of factorization for a. PT. Now for the uniqueness.'r = qi2 ••• q.. (Prove!) But then b = --a:i = PZOf. . Pr = qr. r = 1.lpza2 . since a = PIOf. Here.. Q8{3s where PI > pz > .r = !1T.. the induction is completed and the assertion of unique factorization is established. = !18 = 0. r . In short. namely.l . Thus ql ?: qi = Pl.. Pr = qT. > and each d.!3'. we shall use mathematical induction.•. this is precisely what we were trying to prove. the proposition is true for all integers n ?: mo = 2.lPZ0f. Consequently. Proceeding by induction we suppose it to be true for all integers u. OI. PI = ql and al = f31. we have shown that PI = ql. implies its truth for k.. .. Together with the information we already have obtained.r = PI{3lql2 . . = ar done...20 PRELIMINARY NOTIONS CH. hence r = 8..{3s. (3) pz = qz..1.. (2) PI = ql. too.6. PTOf. every integer n ?: 2 is a product of prime powers. since ql! a we get ql = Pi for somej. Similarly. (3) al = !1ll az = !1z.. are prime numbers. Therefore a = P1Of. PTOf. .2... that is. since PI is a prime number. r < k..• . .2. > q. If b b to get: > 1 then since b < a we can apply our induction hypothesis to = 0 and !12 = . hence P1! q/h . PI! a. 2 ::. whence PI ?: Pi = qi... However. »-" = q/hql2 . Suppose that a = PIOf. u < a. it follows easily that P1 = qi for some i. 2 ::. Now. Pz = qz.. aT = f3T. 1 all integers r. by the basic induction principle. q.r = q/l . and where each 01..

then n [ (j . (. congruence modulo n. . Given any integer a. the relation that we now introduce is a special case of a much more general one that can be defined in a much broader context. 3 THE INTEGERS 21 We change direction a little to study the important notion of congruence modulo a given integer. DEFINITION.21 == -9 mod 10.b) + (c . these are distinct..1].b). by Lemma 1. n. Further. Suppose that a == b mod nand c == d mod n. n[ (a . This congruence relation enjoys the following basic properties: LEMMA The relation is referred to as "congruence modulo n. [0]. This. o ~ i < j < n. (3) If a == b mod nand c == d mod n then a c == b d mod nand ac == bd mod n.c). [1]. if a == b mod nand b == cmodn. we indeed have that n[ (a . [n . c mod n. a = kn + r where 0 ~ r < n. + . then the fact that n[a(b .6. notice that if ab == ac mod n and if a is relatively prime to n..7 (1) The relation. of a be denoted by [a].b)..i) where j . (2) This equivalence relation has n distinct equivalence classes.b) andn[ (b .c). and so n [{(a + c) . We define a.. for if til = fj] with. Thus there are at most n distinct congruence classes.d) whence n[ {(a .d)b} = ac . If ab == ac mod n and a is relatively prime to n. we call it the congruence class (mod n) of a. that 73 := 4 mod 23. etc.i is a positive integer less than n. which is obviously impossible.. a E: [r] and so [a] = [r].." Note..c)}. if a == b mod n then n[ (a . . say.b) and n[ (c . n[ (a . [1]. However. of course. As we shall see later. namely." n is called the modulus of the relation. and so n[ (b . But then. We have now proved assertions (1) and (2) of the lemma. implies that n[ (b . Finally.c) and so b .SEC.b)c + (c .. Let n> 0 be a fixed integer. there are exactly the n distinct congruence classes [0]. . We now prove part (3). Consequently.bd. b mod n if n[ (a - b).a) whence a == a mod n for every a.d)}.c) whencen[ {(a . But then a + c == b + d mod n.(b + d) }. therefore. Finally. for example. Let the equivalence class. defines an equivalence relation on the set of integers.. that is.. In addition n[ {(a . i) + + Proof. then b == c mod n. thus b. Since n[O. We first verify that the relation "congruence modulo n" is an equivalence relation.1]. and we read a == b mod n as "a is congruent to b modulo 1. . under this relation.a) = -(a . implies that a == c mod n. whence ac == bd mod n. a mod n. [n ..b) (b . then n[ (a . by the Euclidean algorithm.

b) = 1 prove that (ab) [z. (Verify!) These operations in J n have the following interesting properties (whose proofs we leave as exercises) : For any [i]. (2) [a. [1].3 mod 6.:: [0] is in Jp. 4. [i][k] distributive law. b) = P161 ••• Plc6k where Oi = minimum of OI. If alb and b I a.7 opens certain interesting possibilities for us... If a and b are integers. written as [a. . 3. if [i] = [i'] and lJ] = [j']. 2. b] = -. show it is a divisor of mg + nh. One more remark: If n = P is a prime number and if [a] .i 0. . Let J n be the set of the congruence classes mod n. Lemma 1. J n = {[O].. We assert that the lemma assures us that this "addition" and "multiplication" are well-defined.. If a = Pial . ab Prove that [a. [k] in J n (1) [i] (2) [i][J1 + [i] = [i][i] = [i] + [ill [i] (4) ([i][j])[kl = [i]([j][k]) (5) [iJ([j] + [k]) = [t1[j] (6) [0] + [i] = [i]. .. prime numbers and where each OI. b] exists and that [a. (b) [t1[j] = [ij]. Given two elements.] (3) ([i] + [i]) + [k] = = commutative laws. Pic ak and b = Pl{31 PIc{3k where the Pi are distinct . . then [i] + [j] = [i + j] = [i' + j'] = [i'] + [j'] and that [i][J1 = [i'] [j']. It is called the set of integers mod n. the result of part (4) may be false. (7) [1][1.Ill. before we proceed much further we will have become well acquainted with it.22 PRELIMINARY NOTIONS CR. 2. [j]. f3i ~ 0. the least common multiple of a and b. (2) Whenever a I x and b I x then d I x.iandf3doreachi. [i] and [j] in J let us define: n. If al x and b Ix and (a. (a) [i] + [j] = [i +J1.i and {3i for each i. show that a = ±b. 1 If a is not relatively prime to n. [·n. [n . 5. that is. yet 2 ~ 4 mod 6. then there is an element [b] in Jp such that [aJ[b] = [1]. b] is defined as that positive integer d such that: (1) aid and bid. (a) b) . + ([i] + [k])} + associative laws. The set J n plays an important role in algebra and number theory.3 == 4. If b is a divisor of g and of b.. prove: ~ (1) Ca. PROBLEMS 1.. for instance. b] = P1'Yl ••• plc'Ykwhere'Yi = maximum of OI. that is.

Prove the corollary to Lemma 1. 33). Assuming that any nonempty set of positive integers has a minimal element. ro = 7. a < m then pen) is true for all n ~ mo. prove that it is sufficient to show that it is not divisible by any prime number p. here. given a and b. 17. = q2r2 0 :s. 3 THE INTEGERS 23 we 6. Use the method in Problem Ibl·) 6 to calculate: (a) (1128. 12. b). prove that there exists an x such that x == a mod m and x == b mod n. (b) If the proposition P is such that (1) P(mo) is true (2) P(m) is true whenever pea) is true forallasuchthatmo :s.. there is a first integer n such that rn+l = O. 16. Prove the properties 1-7 for the addition and multiplication in J ". Given have: a. 9.SEC. Prove that the addition and multiplication used in J n are welldefined. viTi. r2 < rl < rz rl + r3. 11. 8. Prove that n is a prime number if and only if in J n [a][b] = [0] implies that raj ==: [0] or [b] = [OJ. aP == a mod p. prove: (a) If the proposition P is such that (1) P(mo) is true (2) the truth of P(m . *14. Prove that r« = (c. n) = 1.6. prove that for any integer a. . 0 :s. If p is a prime number. b. prove that one can find [b] ~ In such that [aJ[b] = [1] inJn. If (a. n) = 1. (b) (6540. To check that n is a prime number. n) = 1 or nla. r3 Since the integers rk are decreasing and are all nonnegative. Show that n > 1 is a prime number if and only if for any a either (a. 13.1) implies the truth of P(m) then pen) is true for all n ~ mo. 10. 1206). (We consider. on applying the Euclidean algorithm successively b = qlrl + r2. 15. such that p :s. If (m.

New York. S. and MACLANE.. 1 Supplementary Reading For sets and cardinal numbers: BIRKHOFF. A Brief Survey oj Modern Algebra. 1953. G.24 PRELIMINARY NOTIONS CH.. The Macmillan Company. .

This. Yet despite this simplicity of description the fundamental algebraic concepts such as homomorphism. These are usually sets on whose elements we can operate algebraically-by this we mean that we can combine two elements of the set. groups. because someone finally noted that these special cases were indeed special instances of 25 . let us take a quick look ahead. always hoping that when these results are applied to a particular. being one-operational systems. which play such an important role in all algebraic structures-in fact. One does not just sit down. cogent reasons for this choice. vector spaces. before we become weighted down with details. lend themselves to the simplest formal description. The systems chosen for study are chosen because particular cases of these structures have appeared time and time again. In later chapters we shall have a look at some of the others such as rings. To begin with. there are natural. quotient construction. We should like to stress that these algebraic systems and the axioms which define them must have a certain naturality about them. fields. and linear algebras. and then proceed to study the system so described. and the like. to obtain a third element of the set-and.CHAPTER 2 Group Theory In this chapter we shall embark on the study of the algebraic object known as a group which serves as one of the fundamental building blocks for the subject which is today called abstract algebra. in all of mathematics-already enter here in a pure and revealing form. In this abstract setting we then attempt to prove theorems about these very general structures. special case. have achieved positions of paramount importance. In abstract algebra we have certain basic systems which. we assume that these algebraic operations are subject to certain rules. is done by some. admittedly. At this point. Aside from the fact that it has become traditional to consider groups at the outset. concrete realization of the abstract system there will flow out facts and insights into the example at hand which would have been obscured from us by the mass of inessential information available to us in the particular. they should be rich in meaningful results. but most mathematicians would dismiss these attempts as poor mathematics. in addition. They must come from the experience of looking at many examples. in the history and development of mathematics. which are explicitly spelled out in what we call the axioms or postulates defining the system. list a few axioms. perhaps in several ways.

For any two elements 7 E:: A(S) we introduced a product. namely. E:: A (S). case after case after case of the special object. have been those based on a broad and tall pillar of special cases. such that: (2) a.26 GROUP THEORY CH. (3) There is a very special element L E:: A(S) which satisfies L = a 0 L = a for all IT E:: A (S). that have stood the test of time and have survived to become of importance.I. because one notices analogies between two highly disparate mathematical objects and so is led to a search for the root of these analogies. (4) For every a E: A(S) there is an element. This is usually described by saying that every element in A(S) has an inverse in A(S).2 a general phenomenon. then it follows that a 0 7 is also in A(S). b E:: G implies that a. Amongst mathematicians neither the beauty nor the significance of the first example which we have chosen to discuss-groups-is disputed. To cite an example. For an arbitrary nonempty set S we defined A (S) to be the set of all one-to-one mappings of the set S onto itself. 1. such that a 0 cr-1 = cr-1 0 a = L. was studied toward the end of the eighteenth. yet it was not until relatively late in the nineteenth century that the notion of an abstract group was introduced. a 0 (70 J. that whenever S has three or more elements we can find two elements a. (1) Whenever a.) = (cr 0 7) 0 J. c (1) a. E: G implies that a·(b·c) = (a·b)·c (associative law). also in A (S). which we know today as groups. Such an element is called an identity element for A(S). b. called the product and denoted by .I. {3 E: A (S) such that a (3 . century. denoted by a 0 7. and at the beginning of the nineteenth. 7. At this juncture it is advisable to recall a situation discussed in the first chapter. {3 a. A nonempty set of elements G is said to form a group if in G there is defined a binary operation.. This possibility which runs counter to our usual experience and intuition in mathematics so far. . J. 0 (J' One other fact about A(S) stands out. This is described by saying that A(S) is closed under the product (or. With this example as a model. and on further investigation it turned out that the following facts were true for the elements of A(S) subject to this product: (J'. The only algebraic structures. This relation is called the associative law.b E:: G (closed). introduces a richness into A(S) which would have not been present except for it. and with a great deal of hindsight. written as cr-t.I. so far encountered. Definition of a Group. sometimes. 7 E:: A(S). as closed under multiplication). (2) For any three elements cr. we abstract and make the 0 0 DEFINITION.:c.

SEC. -1 under the multiplication of real numbers. most interesting when it is finite. Another natural characteristic of a group G is the number of elements it contains. a' b = a b. a-b = A group G is said to be abelian (or commutatilJe) if for every b-a. T E::: A(S) such that IT 0 T ~ TO IT. This prompts us to single out a highly special. Thus we already have presented to us an infinite source of interesting. G is then an abelian group of order 2. then the group A(S) has n! elements. a. 2 SOME EXAMPLES OF GROUPS 27 (3) There exists an element e E::: G such that a. In that case we say that G is a finite group. Let G = Sa. ±1. + EXAMPLE 2. If S has three or more elements.. We shall see later (in a theorem due to Cayley) that these A(S)'s constitute. class of groups as in the next definition. We digress a little before returning to Sa. that is. A group which is not abelian is called. G is a group of order 6. To see that finite groups which are not trivial do exist just note that if the set S contains n elements. (Prove!) This highly important example will be denoted whenever it appears in this book by S« and will be called the symmetric group of degree n. but very important. . b E::: G the usual sum of integers. having seen a family of examples of such groups we know that non-abelian groups do indeed exist. where we mean by a· b for a. In the next section we shall more or less dissect Sa which is a non-abelian group of order 6. DEFINITION. ±2. of course. We call this the order of G and denote it by 0 (G). Xa} onto itself. Then the reader can quickly verify that G is an infinite abelian group in which 0 plays the role of e and -a that of a-I. . concrete groups.a = a for all a E::: G (the existence of an identity element in G).. This number is. EXAMPLE 1. EXAMPLE 3. Let G consist of the real numbers 1. Considering the source of this definition it is not surprising that for every nonempty set S the set A(S) is a group. (4) For every a E::: G there exists an element a-I E::: G such that a·a-1 = a-l'a = e (the existence of inverses in G).e = e. the group of all 1-1 mapping of the set {Xl. under the product which we defined in Chapter 1. X2. recall that we can find elements IT. Some Examples of Groups. Let G consist of the integers 0. in some sense. a universal family of groups. bE::: G. non-abelian. naturally enough. 2.

X3 by Xl ~ ~ X2 Xl X3 </>: X2 Xa ~ and the mapping Xa ~ Xl Checking. ~ y.2..2 and we have that .-1 = 1/. a3 = a-a2. X2. The reader may verify that the usual rules of exponents prevail. and a-2 = (a-l)2. X3. but in any group G let us define for any a E: G. and that </>·f: whereas X2 X3 ~ ~ Xl ~ ~ ~ Xl X3 X2 f·</>: X2 X3 It is clear that </>. or zero) m. Since 1/. aU = e.</> on Xl. we readily see that </>2 == e. a1 = a.-1. in this notation. for any two integers (positive. negative. ••• . </> for they do not take Xl into the same image. Since f3 = e.28 GROUP THEORY CH. an means the integer na. Consider the mapping </> defined on the set Xl. not just in 83.) With this notation at our disposal let us examine Sa more closely. Let us now compute the action of 1/.2 For a neater notation. a-3 = (a-l)3. n. y. etc. it follows that 1/. namely.-1 = 1/. a2 = a-a. Xl ~ f3 = X3 X2 Xl e. (1) (2) (It is worthwhile noting that. if G is the group of Example 1. X2. ak = a·ak-1.

(a-1)-1 = a. these are all distinct and are in G (since G is closed). Some Preliminary Lemmas. n . fascinating vistas were opened before us. . One might ask. namely.-.cp) = (f·f-l). We construct a group of order n as fol4.= if. then x = y. a = a-e = e-a = a-] = f·a. EXAMPLE Let n be any integer. and let Pn be a rotation through an angle of 27r/n. Of more interestis the form of (cp·f)·(f·cp) = cp·(f·(f·cp» = cp·(f2. Pn2. cp. since b-] = b for every b E::: G. i = 0. We have now been exposed to the theory of groups for several pages and as yet not a single. they are rather dull) they will be extremely useful.cp) = cp. lows: G will consist of all symbols ai. f· ¢. Learning the alphabet was probably not the most interesting part of our childhood education. if. wearisome chain of equalities here. of radius 1..cp) = cp·(f-1.-1. Then Pn E::: A (8) and Pn in A(S) generates a group of order n. . . Consider the elements e. where all of a. It is the last time we shall be so boringly conscientious. cp.. It is called a cyclic group of order n. e -j :=: f. Some of these will appear in the exercises. (The reader should not be frightened by the long. b E::: G. yare in G. Before we proceed with the proof itself it might be advisable to see what it is that we are going to prove. Since e-a = a for every a E::: G. x. admittedly.--I.1 where we insist that aO = aft = e. for every a E::: G. once this hurdle was cleared.•. But. [e. First let us consider part (a).SEC. . Pn. yet. 3 SOMEPRELIMINARYEMMAS 29 L In other words. = f -1. (cp·f) = cp2·f = e·f = f. which only has six elements. Although the first few results we demonstrate are. Thus this list enumerates all the elements of G. then e = f. It is high time to remedy this situation.. not very exciting (in fact. on the other hand. f2. in the plane. for instance. If. Pn n-11· 3. cp' if. The reader may verify that this is a group. ai·aj = ai+j if i + j ~ nand ai·aj = ai+j-n if i + j > n. cp. We begin with LEMMA 2. 1.2. what is the entry in the list for f· (cp' f)? Using cp' if.. In part (a) we want to show that if two elements e and fin G enjoy the property that for every a E::: G. If G is a group.1. for all a. In part (b) our aim is to show that if x· a = a· x = e and y .a = a' y = e. (a·b)-l = b-1·a-1• Proof.) Using the same techniques as we have used. A geometric realization of the group in Example 4 may be achieved as follows: Let S be the circle. in particular. then (a) (b) (c) Cd) the identity element of G is unique j every a E::: G has a unique inverse in G. solitary fact has been proved about groups.cp = e·cp = cp. . cp we see that f· (cp·f) = f· (if. then. the reader can compute to his heart's content others of the 25 products which do not involve e.

(a) G = set of all integers. however. We have. (a-I)-l = e = a-I. for we cannot conclude that a'x = y·a implies x = y for we have no way of knowing whether a. that is. and so e = f. . Given a. in equations in groups. proved that a·x = a·y in a group G forces x = y.b)-I = b-1·a-1• Certain results obtained in the proof just given are important enough to single out and we do so now in LEMMA 2. Part (d) is the most trivial of these. Thus b·(a·x) = b·(a·y).30 GROUP THEORY CH. a·x = a·y.«b. in fact. x = 1/. Part (c) follows from this by noting that a-I. and so.2. the usual product of integers.2 we must have that e -f = e. Piecing these two bits of information together we obtain f = e -f = e. (b) G = set of all positive integers.(a·y) = (b·a)·y = e·y = y.(. a·x = e and a·y = e. from the same side. a·u and u·a hold in G. PROBLEMS 1. A note of caution. using the associative law this leads to x = e'x = (b·a)·x = b. Rather than proving part (b) we shall prove something stronger which immediately will imply part (b) as a consequence.x = X· a. obviously. assume that a·x = a·y for a. This is the analog in general groups of the familiar result . then the equations a'x = b and y-a = b have unique solutions for x and y in G. This is illustrated in S3 with a = t/>. x. b-I) -a-I) = a' (e·a-I) = a·a-I = e. In the following determine whether the systems described are groups.(a·x) = b. canceling off the a-Ion the left leaves us with (a-I)-1 = a. Similarly we can prove that x·a = u-« implies that x = y. Suppose that for a in G. then. This says that we can cancel. Let us make this our starting point. for (a·b)· (b-1• a-I) = a. by the very definition of the inverse (a. yin G. in the group of real numbers under addition. the two can- cellation laws. There is an element bE:: G such that b-o.. a. a-b = ab. = e (as far as we know yet there may be several such b's). b in the group G. = = a·w implies u = w w·a implies u =w The few details needed for the proof of this lemma are left to the reader. point out which of the group axioms fail to hold. In case not. In particular.-1.5) = 5. say. y = 1/. a-b == a-b.

SEC.

3

SOME PRELIMINARY

LEMMAS

31

(c) G = ao, al, ... , a6 where ai·aj = ai+i if i j < 7 ai'aj = ai+j-7 if i + j 2: 7 (for instance, a5' a4 = a5+4-7 = a2 since 5 + 4 = 9 > 7). (d) G = set of all rational numbers with odd denominators, a-b == a b, the usual addition of rational numbers. 2. Prove that if G is an abelian group, then for all a, b E: G and all integers n, (a·b)n = an·bn. 3. If G is a group such that (a·b)2 = a2·b2 for all a, bE: G, show that G must be abelian. *4. If G is a group in which (a·b)i = ai·bi for three consecutive integers i for all a, b E: G, show that G is abelian. 6. Show that the conclusion of Problem 4 does not follow if we assume the relation (a·b)i = ai·bi for just two consecutive integers. 6. In S3 give an example of two elements z, y such that (z- y)2 ¢ x2. y2. 7. In S3 show that there are four elements satisfying x2 = e and three elements satisfying y3 = e. 8. If G is a finite group, show that there exists a positive integer N such that aN = e for all a E: G. 9. (a) If the group G has three elements, show it must be abelian. (b) Do part (a) if G has four elements. (c) Do part (a) if G has five elements. 10. Show that if every element of the group G is its own inverse, then G is abelian. 11. If G is a group of even order, prove it has an element a ;;::e satisfying a2 = e. 12. Let G be a nonempty set closed under an associative product, which in addition satisfies: (a) There exists an e E: G such that a· e = a for all a E: G. (b) Given a E: G, there exists an element yea) E: G such that a·y(a) = e. Prove that G must be a group under this product. 13. Prove, by an example, that the conclusion of Problem 12 is false if we assume instead: (a') There exists an e E: G such that a·e = a for all a E: G. (b/) Given a E: G, there exists yea) E: G such that y(a)·a = e. 14. Suppose a finite set G is closed under an associative product and that both cancellation laws hold in G. Prove that G must be a group. 16. (a) Using the result of Problem 14, prove that the nonzero integers modulo p, p a prime number, form a group under multiplication modp. (b) Do part (a) for the nonzero integers relatively prime to n under multiplication mod n.

+

+

32

GROUP THEORY

CH.

2

16. In Problem 14 show by an example that if one just assumed one of the cancellation laws, then the conclusion need not follow. 17. Prove that in Problem 14 infinite examples exist, satisfying the conditions, which are not groups. 18. For any n > 2 construct a non-abelian group of order 2n. (Hint: imitate the relations in S3') 19. If S is a set closed under an associative operation, prove that no matter how you bracket ala2 ... an, retaining the order of the elements, you get the same element in S (e.g. (al·a2)·(aa·a4) = al·(a2·(aS·a4))i use induction on n).
4. Subgroups. Before turning to the study of groups we should like to change our notation slightly. It is cumbersome to keep using the· for the group operation; henceforth we shall drop it and instead of writing a-b for a, b E: G we shall simply denote this product as abo In general we shall not be interested in arbitrary subsets of a group G for they do not reflect the fact that G has an algebraic structure imposed on it. Whatever subsets we do consider will be those endowed with algebraic properties derived from those of G. The most natural such subsets are introduced in the
DEFINITION. A subset H of a group G is said to be a subgroup of G if, under the product in G, H itself forms a group.

The following remark is clear: if H is a subgroup of G and K is a subgroup of H, then K is a subgroup of G. It would be useful to have some criterion for deciding whether a given subset of a group is a subgroup. This is the purpose of the next two lemmas.
LEMMA 2.3. A nonempty subset H of the group G is a subgroup of G if and only if (1) a, b E: H implies that ab E: H; (2) a E: H impl~es that a-I E: H.

Proof. If H is a subgroup of G, then it is obvious that (1) and (2) must hold. Suppose conversely that H is a subset of G for which (1) and (2) hold. In order to establish that H is a subgroup all that is needed is to verify that e E: H and that the associative law holds for elements of H. Since the associative law does hold for G, it holds all the more so for H which is a subset of G. If a E: H, by (2) a-I E: H and so by (1) e = aa-1 E: H. This completes the proof. In the special case of a finite group the situation becomes even nicer for there we can dispense with (2).

SEC.

4

SUBGROUPS

33

LEMMA 2.4. If H is a nonempty finite subset of a group G and H is closed under multiplication, then H is a subgroup of G. Proof. In light of Lemma 2.3 we need but show that whenever a E::: H, then a-l E::: H. Suppose that a E::: H; thus a2 = aa E::: H, as = a2a E::: H, ... , am E::: H, since H is closed. Thus the infinite collection of elements a, a2, •.• , am, must all fit into H, which is a finite subset of G. Thus there must be repetitions in this collection of elements; that is, for some integers r, 8 with r > 8 > 0, aT = as. By the cancellation in G, or= = e (whence e is in H); since r - 8 - 1 ;:::0, ar-s-l E::: Hand a-l = ar- 1 l since aa =a = e. Thus a-l E::: H, completing the proof of the lemma. The lemma tells us that to check whether a subset of a finite group is a subgroup we just see whether or not it is closed under multiplication. We should, perhaps, now see some groups and some of their subgroups. G is always a subgroup of itself; likewise the set consisting of e is a subgroup of G. Neither is particularly interesting in the role of a subgroup, so we describe them as trivial subgroups. The subgroups between these two extremes we call nontrivial subgroups and it is in these we shall exhibit the most interest.
8T8 -TS -

EXAMPLE Let G be the group of integers under addition, H the subset 1. consisting of all the multiples of 5. The student should check that H is a subgroup. In this example there is nothing extraordinary about 5; we could similarly define the subgroup H n as the subset of G consisting of all the multiples of n. H n is then a subgroup for every n. What can one say about u; n Hm? It might be wise to try it for He n s; EXAMPLE Let S be any set, A(S) the set of 1-1 mappings of S onto 2. itself, made into a group under the composition of mappings. If Xo E::: S, let H(xo) = {q'> E::: A(S)lxoq'> = xo}. H(xo) is a subgroup of A(S). If for Xl ~ Xo E::: S we similarly define H(xl), what is H(xo) n H(x!)? EXAMPLE Let G be any group, a E::: G. Let (a) = {a' Ii = 0, ±1, 3. ±2, ... }. (a) is a subgroup of G (verify!); it is called the cyclic subgroup generated by a. This provides us with a ready means of producing subgroups of G. If for some choice of a, G = (a), then G is said to be a cyclic group. Such groups are very special but they playa very important role in the theory of groups, especially in that part which deals with abelian groups. Of course, cyclic groups are abelian, but the converse is false. EXAMPLE Let G be a group, W a subset of G. Let (W) be the set of all 4.

34

GROUP THEORY

CH.

2

elements of G representable as a product of elements of W raised to positive, zero, or negative integer exponents. (W) is the subgroup of G generated by Wand is the smallest subgroup of G containing W. In fact, W is the intersection of all the subgroups of G which contain W (this intersection is not vacuous since G is a subgroup of G which contains W). DEFINITION.Let G be a group, H a subgroup of G; for a, b E:: G we say == b mod H if ab-1 E: H. LEMMA 2.5. The relation a

a is congruent to b mod H, written as a

== b mod H is an

equivalence relation.

Proof. If we look back in Chapter 1, we see that to prove Lemma 2.5 we must verify the following three conditions: For all a, b, c E:: G

(1) a (2) a (3) a

== a mod H; == b mod H implies b == a mod Hi == b mod H, b == c mod H implies a == c mod H.

Let's go through each of these in turn. (1) To show that a == a mod H we must prove, using the very definition of congruence mod H, that aa-1 E:: H. Since H is a subgroup of G, e E: H, and since aa-1 = e, aa-1 E:: H, which is what we were required to demonstrate. (2) Suppose that a == b mod H, that is, suppose ab-1 E: Hi we want to get from this b == a mod H, or, equivalently, ba-1 E:: H. Since ab-1 E: H, which is a subgroup of G, (ab-1)-1 E:: H; but, by Lemma 2.1, (ab-1)-1 = (b-1)-la-1 = ba-1, and so ba-1 E:: Hand b == a mod H. (3) Finally we require that a == b mod Hand b == c mod H forces a == c mod H. The first congruence translates into ab-1 E:: H, the second into bc-1 E:: H; using that H is a subgroup of G, (ab-1)(bc-1) E:: H. However, ac-1 = aec-1 = a(b-1b)c-1 = (ab-1)(bc-1), hence ac-1 E: H from which it follows that a == c mod H. This establishes that congruence mod H is a bona fide equivalence relation as defined in Chapter 1, and all results about equivalence relations have become available to us to be used in examining this particular relation. A word about the notation we used. If G were the group of integers under addition, and H = H n were the subgroup consisting of all multiples of n, then in G, the relation a == b mod H, that is, ab-1 E: H, under the additive notation, reads as a - b is a multiple of n. This is the usual number theoretic congruence mod n. In other words, the relation we defined using an arbitrary group and subgroup is the natural generalization of a familiar relation in a familiar group. DEFINITION. f H is a subgroup of G, a I H a is called a right coset of H in G.

E: G, then

Ha = {halh

E: H}.

SEC.

4 2.6. For all a E:: G, Ha

SUBGROUPS

35

LEMMA

= {x E:: Gla == x modH}.

Proof, Let [a] = [z E:: Gla == x mod H}. We first show that Ha C [a]. For if hE:: H, then a(ha)-l = aCa-Ih-l) = h-l, so is in H since H is a subgroup of G. By the definition of congruence mod H this implies that ha E:: [a] for every h E: H, and so Ha C [a]. Suppose, now, that x E: raj. Thus ax-I E: H, so (ax-1)-1 = xa-1 is also in H. That is, xa-I = h for some h E:: H. Multiplying both sides by a from the right we come up with x = ha, and so x E:: Ha. Thus [a] C Ha. Having proved the two inclusions [a] C Ha and Ha C [a], we can conclude that [a] = Ha, which is the assertion of the lemma. In the terminology of Chapter 1, [a], and thus Ha, is the equivalence class of a in G. By Theorem 1.a these equivalence classes yield a decomposition of G into disjoint subsets. Thus any two right coseis of H in G either are identical or have no element in common. We now claim that between any two right cosets Ha and Hb of H in G there exists a 1-1 correspondence, namely, with any element ha E:: Ha, where h E: H, associate the element hb E:: Hb. Clearly this mapping is onto Hb. We aver that it is a one-to-one correspondence, for if hib = h2b, with hI, h2 E: H, then by the cancellation law in G, hI = h2 and so hia = h2a. This proves
LEMMA

2.7. There is a 1-1 correspondence between any two right coseie

oj H in G. Lemma 2.7 is of most interest when H is a finite group, for then it merely states that any two right cosets of H have the same number of elements. How many elements does a right coset of H have? Well, note that H = He is itself a right coset of H, so any right coset of H in G has o(H) elements. Suppose now that G is a finite group, and let k be the number of distinct right cosets of H in G. By Lemmas 2.6 and 2.7 any two distinct right cosets of H in G have no element in common, and each has o(H) elements. Since any a E:: G is in the unique right coset H a, the right cosets fill out G. Thus if k represents the number of distinct right cosets of H in G we must have that koCH) = o(G). We have proved the famous theorem due to Lagrange, namely,
THEOREM

2.A.

If G is a finite group and H is a subgroup oj G, then o(H)

is a divisor oj o(G).
DEFINITION. If H is a subgroup of G, the index oj H in G is the number of distinct right cosets of H in G.

We shall denote it by ia(H). In case G is a finite group, ia(H)

0(G)

=

o(H)'

36

GROUP THEORY

CH.2

as became clear in the proof of Lagrange's theorem. It is quite possible for an infinite group G to have a subgroup H ~ G which is of finite index in G. It might be difficult, at this point, for the student to see the extreme importance of this result. As we penetrate the subject more deeply he will become more and more aware of its basic character. Because the theorem is of such a stature it merits a little closer scrutiny, a little more analysis, and so we give, below, a slightly different way of looking at its proof. In truth, the procedure outlined below is no different from the one already given. The introduction of the congruence mod H smooths out the listing of elements used below, and obviates the need for checking that the new elements introduced at each stage did not appear before. So suppose again that G is a finite group and that H is a subgroup of G. Let hll h2' ... , h; be a complete list of the elements of H, r = o(H). If H = G, there is nothing to prove. Suppose, then, that H ~ G; thus there is an a C G, a t[ H. List all the elements so far in two rows as
hll h2' hI a, h2a, , h; , h-a.

We claim that all the entries in the second line are different from each other and are different from the entries in the first line. If any two in the second line were equal, then hia = hja with i ,..& i, but by the cancellation law this would lead to hi = hj, a contradiction. If an entry in the second line were equal to one in the first line, then h,a = h resulting in a = hi-1hj C H since H is a subgroup of G; this violates a t[ H. Thus we have, so far, listed 20(H) elements; if these elements account for all the elements of G, we are done. If not, there is abe G which did not occur in these two lines. Consider the new list
j,

hI, h2' hl a, h2a,

, li; , h-a

hlb, h2b, ... , hrb.

As before (we are now waving our hands) we could show that no two entries in the third line are equal to each other, and that no entry in the third line occurs in the first or second line. Thus we have listed 30(H) elements. Continuing in this way, every new element introduced, in fact, produces o(H) new elements. Since G is a finite group, we must eventually exhaust all the elements of G. But if we ended up using k lines to list all the elements of the group, we would have written down koCH) distinct elements, and so koCH) = o(G). It is essential to point out that the converse to Lagrange's theorem is false-a group G need not have a subgroup of order m if m is a divisor of o(G). For instance, a group of order 12 exists which has no subgroup of

If n is a positive integer and a is relatively prime . by Lagrange's theorem. If no such integer exists we say that a is of infinite order. A particular case of Corollary 2 is of great interest in number theory. 7 are the numbers less than S which are relatively prime to S. then a</>(n) == 1 mod n. 1. is defined for all integers n by: ¢(1) = 1. o(a) 10(G). whence. COROLLARY 2. By Corollary 1. If G is a finite group and a E:: G. Before we present these we make one definition. If G is a finite group and a E:: G. then ai = aJ for some integers 0 S i < J < oCa)." COROLLARY o(a) for the order of a. If G is a group and a E:: G. 3. In Problem 15 (b) at the end of Section 3 the reader was asked to prove that the numbers less than nand relatively prime to n formed a group under multiplication mod n. v. Therefore. ao(G) = amo(a) = (ao(a))m = em = e. Proof. then o(a) 10(G). the place to look is in 84. The reader might try to find an example of this phenomenon. the order (or period) of a is the least positive integer m such that am = e. a2. o(a) I o(G) i thus o(G) = moCa). ¢(n) .• How many elements are there in (a)? We assert that this number is the order of a. Thus the cyclic subgroup generated by a has oCa) elements. then ao(G) = e. ••. Recall our other notation. 4 SUBGROUPS 37 order 6. With Lagrange's theorem already in hand. this subgroup has at most o(a) elements. the symmetric group of degree 4 which has a subgroup of order 12 which will fulfill our requirement. Clearly. since ao(a) = e. of G generated by ai (a) consists of e. Thus. it seems most natural to prove the corollary by exhibiting a subgroup of G whose order is o(a). This group has order ¢(n).i < oCa) which would contradict the very meaning of o(a). DEFINITION. yet 0 < J . We use the nota- tion u I v reads: "u is a divisor of v. If we apply Corollary 2 to this group we obtain COROLLARY 3 (EULER). ¢(S) = 4 since only 1. for n > 1 ¢(n) = number of positive integers less than n and relatively prime to n. The element a itself furnishes us with this subgroup by considering the cyclic subgroup. a. to n. Lagrange's theorem has some very important corollaries. for two integers u. 5. for instance. Then aJ-' = e. (a). Proof.SEC. If it should actually have fewer than this number of elements. The Euler ¢-function.

This says that G is cyclic and that every element in G is a power of a. Suppose. Note that KH = {e. as we see in the next lemma. that HK = KH. hence 0 == aP == a mod p here also. If P is a prime number and a is any intege1'. cp. and let H = (a). if H is a subgroup of G and a E: G. then by Corollary 3. namely. If G is a finite group whose order is a prime number p. First we claim that G has no nontrivial subgroups H. K = [e. H ~ (e) since a ~ e E: H. now. cpifi}. To prove that HK is a subgroup we must verify that it is closed . Thus H = G. Proof. we could verify this directly but it does not hurt to keep recalling Lagrange's theorem. then hk = klhl for some kl E: K. a is not relatively prime to p. K are two subgroups of G. if hE: Hand k E: K. by Lagrange's theorem HK could not be a subgroup of S3. Let's pause and look at an example.cp2ifi= ifi. hi E: H (it need not be that kl = k or hi = hI). Suppose now that a ~ e E: G. both H and K are subgroups. for o(H) must divide 0(0) = p leaving only two possibilities. whence aP == a mod p. cp. whereas the second implies that H = G. 5. This is precisely why HK fails to be a subgroup. then a P == a mod p. in S3 let H = [e. Proof. o(H) = 1 or o(H) = p. we must have that p I a. o is COROLLARY 5. If H. on the other hand. aP-1 == 1 mod p.) We might try to find out why HK is not a subgroup. then Ha consists of all elements in G of the form ha where h E: H. As we have defined earlier. then a cyclic group. that is. not only for its results but also because the spirit of the proofs occurring here are genuinely group-theoretic. cp}. Section 4 is of great importance in all that comes later. It would be wise for him to assimilate the material and approach thoroughly. first. The first of these implies H = (e). hE: H. If. let HK = {x E: Glx = hk. so that a == 0 mod p. If n should be a prime number p.1. cPVt¢ = ifi-l} ~ HK.cpifi.38 GROUP THEORY CH. cpifi.8. Since cp2 = (cpifi)2= e. the order of S3. Since HK consists of four elements and 4 is not a divisor of 6. H is a subgroup of 0. Let us generalize this notion. rather than a few theorems later when it will be too late. k E: K}. (Of course. 2 In order to apply Corollary 2 one should replace a by its remainder on division by n. LEMMA 2. since p is a prime number. then cp(p) = p . What can we say about HK? Just using the definition of HK we can see that HK consists of the elements e. HK is a subgroup of G if and only if HK = KH. A Counting Principle. The student can expect to encounter other arguments having a similar flavor. Thus COROLLARY 4 (FERMA'l'). If a is an integer relatively prime to p.

Proof. then (1) where hhl E:: H. which is devoid of some of the complexity of the general situation. Similarly. how could this fail to happen? The answer clearly must be that if we list all the elements hk. x-1 = hk E:: HK and so x = (X-1)-1 = (hk)-l = k-1h-1 E:: KH. K are subgroups of the abelian group G. k E:: K. Yet it is a perfect meaningful question to ask: how many distinct elements are there in the subset HK? If we denote this number by o(HK). Although there is no need to pay special attention to the particular case in which H n K = (e). kh' = h2k2 with h2 E:: H. so suppose x = hk E:: HK and y = h'k' E:: HK. that is. Thus HK is a subgroup of G. whence h = hl. k E:: K there should be some collapsing. a contradiction. With this experience behind us we are ready to attack the general case. An interesting special case is the situation when G is an abelian group for in that case trivially HK = KH. o(HK) is indeed o(H)o(K).SEC. is quite revealing. then HK is a subgroup of G. here. h E:: H. k2 E:: K. but since kh' E:: KH = HK.-lk E:: K since hI -1 E:: H n K c K and k E:: K. h1-lh E:: H n K = (e). so h1-lh = e. hk = h1k1. Thus KH c HK. since h E:: H. Thus hk is duplicated in the product at . Thus as a consequence we have the COROLLARY. If Hand o(K). But then h1 -lh = k1k-1. Then xy = hkh'k'. Thus HK = KH. respectively. One should ask oneself. 5 A COUNTING PRINCIPLE 39 and every element in HK has its inverse in HK. so x-1 E:: HK. Hence xy = h(h2k2)7c' = (hh2)(k2k') E:: HK. if HK is a subgroup of G. Let's show the closure first. Now if x is any element of HK. some element in the list must appear at least twice. We have proved that no collapsing can occur.B. then K are finite subgroups of G of orders o(H) and o(HK) = oCH)o(K) o(H n K) . looking at this case. h-1k-1 E:: HK and so kh = (h-1k-1)-1 E:: HK. Also x-1 = (hk)-l = k-1h-1 E:: KH = HK. for some h ~ h1 E:: H. K are subgroups of a group G. Since h1-1h = k1k-1. If H. so HK c KH. k1k-1 E:: K. and so. As above we must ask: how often does a given element hk appear as a product in the list of HK? We assert it must appear o(H n K) times! To see this we first remark that if h1 E:: H n K. If H. Equivalently. Here we should seek to show that o(HK) = o(H)o(K). thus h1-lh E:: H. now since h1 E:: H. we have seen that the subset HK need not be a subgroup of G. and HK is closed. we prove THEOREM 2. then for any h E:: H. h1 E:: H n K cHand i. On the other hand. h1-1 must also be in H.

o(K) > vo(G). This proves the theorem. Can you find an upper bound for the index of H KinG? n . and being a subgroup of H. Later on we shall see that there is at least one subgroup of order p. (e). However. 2. (e). H n K . that is. By the corollary. K are subgroups of the finite group G and o(H) > vo(G) I o(K) > Vo(G). we must conclude that H n K = H. If G is a group and H. which. whence H = K. We apply this corollary to a very special group. will tell us there is exactly one subgroup of order pinG. Since HK c G. and so He H n K c K. and u E:: H n K.z5. h E:: H.z5. o(G) thus o(H 2: o(H)o(K) o(HK) = o(H n K) Therefore. > -vO(G). and so h' = hu. If G has no nontrivial subgroups prove it must have prime order.40 GROUP THEORY CH. Consequently hk appears in the list of HK exactly oCH n K) times.o(G). For a subgroup H of G define a left coset of H in G as the set of all elements of the form ah. proving that there is at most one subgroup of order p. K are subgroups of order p. namely. combined with the above. In Problem 4 what is n. (e). If Hand K are subgroups of G and o(H) then H n K . H > Vo(G) V~ o(H n K) o (G) o(H n K) n K) > 1. We claim that G can have at most one subgroup of order p. K are two subgroups of G of finite index in G. Determine the index of Hn in G and write out all the cosets of Hn in G. then so is H n K a subgroup of G. o(HK) :::. the subgroup consisting of all multiples of a fixed integer n in G. Let G be the group of integers under addition. n K. However. PROBLEMS 1. Suppose H. thus all duplications were accounted for in (1). Suppose G is a finite group of order pq where p and q are prime numbers with p > q. o(H)o(K) divided by the number of times a given element appears. o(H n K). Thus the number of distinct elements in HK is the total number in the listing of HK. 3. H.z5. 5. For suppose H.2 least o(H n K) times. We have proved the COROLLARY. Show that there is a 1-1 correspondence between the set of left cosets of H in G and the set of right cosets of H in G. if hk = h'k'. k' = u-1k. prove that H n K is of finite index in G. If Hand K are subgroups of G. which having prime order has no nontrivial subgroups. From this we shall be able to determine completely the structure of G. n Hm? *6. then h-W = k(k')-l = u. Similarly K C H. 4.

] 16.cP1f. ---')0 6. Let G be a finite abelian group in which the number of solutions in G of the equation xn = e is at most n for every positive integer n. Prove that Z is a subgroup of G.SEC. How many generators does a cyclic group of order n have? [b E:= G is a generator if (b) = G. 13.H = {1f. then ana-1 E:= N. Let the mapping Tab for a. 12. (b) Give an example of a group G and a subgroup H such that not every left coset of H in G is a right coset of H in G. n E:= N. Z is defined by Z = {z E:= Glzx = xz all x E:= G}. cpif. 17... If in the group G. If a E:= G. Let G = {Tab I a ~ O}. We list them: Right Cosets H H1f.2. Find the formula for TabTcd. Since the index of N in G is 2 there are two left cosets and two right cosets of N in G. Since the index of H in G is 3. b real numbers.1f. If an abelian group has subgroups of orders nand m. Prove that G is a group under the composition of mappings. We . (b) If a E:= G. rt>} 1f. a" = e prove that o(a) l?n. 10. map the reals into the reals by the rule.2H = {if. respectively. In Problem 7 let H = {Tab €.2. a5 = e.2} = [e.2 Left Cosets = = = [e. List all the right cosets of H in G. Hi. Show that N(a) is a subgroup of G. Prove that G must be a cyclic group.rt> = rt>1f. there are three right cosets of H in G and three left cosets of H in G. Show that H is a subgroup of G. Hif. In the group of Problem 7 let N = {Tib E:= G}.. aba-1 = b2 for a. Prove: (a) N is a subgroup of G. Thus.2} 1f. rt>}. 6 NORMAL SUBGROUPS AND QUOTIENT GROUPS 41 7. 9. is not a left coset. *18.. 8. In G = S3 let us consider the subgroup N = {e. rt>} {1f. Let G be the group S3 and let H be the subgroup {e.} {if. Prove that any subgroup of a cyclic group is itself a cyclic group. G I a is rational}. then show it has a subgroup whose order is the least common multiple of n and m. (a) In Problem 8 show that every left coset of H in G is a right coset of H in G.} H A quick inspection yields the interesting fact that the right coset H1f. 14. the notions of left and right coset need not coincide. 1f.P.b E:= G find o(b). 11. the center of G.1f.2cp = cf>1f. If G is a group. If a E:= G define N(a) = {x E:= Glxa = ax}.2}. N(a) is usually called the norrnalizer or centralizer of a in G. at least for this subgroup. Normal Subgroups and Quotient Groups. Tab: X ax + b.

42

GROUP THEORY

CR.

2

list these: Right Cosets N

Left Cosets N
=

{e,,p, ,p2} = {¢, ,p¢, 1f2¢}
=

[e, If, ,p2}

¢N = {¢, ¢If, ¢lf2} = {¢, 1f2¢, If¢}

A quick inspection here reveals that every left coset of N in G is a right coset in G and conversely. Thus we see that for some subgroups the notion of left coset coincides with that of right coset, whereas for some subgroups these concepts differ. It is a tribute to the genius of Galois that he recognized that those subgroups for which the left and right cosets coincide are distinguished ones. Very often in mathematics the crucial problem is to recognize and to discover what are the relevant concepts; once this is accomplished the job may be more than half done. We shall define this special class of subgroups in a slightly different way, which we shall then show to be equivalent to the remarks in the above paragraph.
DEFINITION. A subgroup N of G is said to be a normal subgr01lp of G if for every g E::: G and n E::: N, gng-1 E::: N.

Equivalently, if by gNg-1 we mean the set of all gng-I, n E::: N, then N is a normal subgroup of G if and only if gNg-1 eN for every g E::: G.
LEMMA

2.9. N is a normal subgroup of G if and only if gN

every g

E::: G.

«: = N for
so N is
=
C gNg-1

Proof. If gNg-l

=

N for every g

E::: G, certainly gNg-1 eN,
C N, N = g(g-lNg)g-l

normal in G. Suppose thatNis

g-lN(g-l)-l eN. Now, since g-lNg ci N, whence N = gNg-l. does not say that for every n

normal in G. Thus if g E::: G, gNg-1 eN

and g-lNg

In order to avoid a point of confusion here let us stress that Lemma 2.9 E::: N and every g E::: G, gng-1 = n. No! This can be false. Take, for instance, the group G to be S3 and N to be the subgroup Ie, '{I, '{I2}. If we compute ¢N ¢-1 we obtain {e, ¢If¢ -r, ¢lf2¢ -I} = Ie, 1f2, '{I}, yet ¢1f¢-1 r= If. All we require is that the set of elements gNg-1 be the same as the set of elements N. We now can return to the question of the equality of left cosets and right cosets.

LEMMA 2.10. The subgroup N of G is a normal subgroup of G if and only if every left coset of N in G is a right coset of N in G.

SEC.

6

NORMAL SUBGROUPS

AND QUOTIENT

GROUPS

43

Proof. If N is a normal subgroup of G, then for every g E: G, gNg-1 = N, whence (gNg-l)g = Ng; equivalently gN = Ng, and so the left coset gN is the right coset N g. Suppose, conversely, that every left coset of N in G is a right coset of N in G. Thus, for g E:: G, gN, being a left coset, must be a right coset. What right coset can it be? Since g = ge E:: gN, whatever right coset gN turns out to be, it must contain the element g; however, g is in the right coset Ng, and two distinct right cosets have no element in common. (Remember the proof of Lagrange's theorem?) So this right coset is unique. Thus gN = Ng follows. In other words, gNg-1 = Ngg-1 = N, and so N is a normal subgroup of

G.

We have already defined what is meant by HK whenever H, K are subgroups of G. We can easily extend this definition to arbitrary subsets, and we do so by defining, for two subsets, A and B, of G, AB = {x E:: G I x = ab, a E:: A, b E: B). As a special case what can we say when A = B = H, a subgroup of G? HH = {hlh2Ihl, h2 E: H} c H since H is closed under multiplication. But HH ::J He = H since e E:: H. Thus HH = H. Suppose that N is a normal subgroup of G, and that a, b E:: G. Consider (N a) (Nb); since N is normal in G, aN = N a, and so NaNb = N(aN)b

= N(Na)b = NNab = Nab.

What a world of possibilities this little formula opens! But before we get carried away, for emphasis and future reference we record this as
LEMMA 2.11. A subgroup N of G is a normal subgroup of G if and only if the product of two right cosets of N in G is again a right coset of N in G.

Proof. If N is normal in G we have just proved the result. The proof of the other half is one of the problems at the end of this section. Suppose that N is a normal subgroup of G. The formula NaNb = Nab, for a, b € G is highly suggestive; the product of right cosets is a right coset. Can we use this product to make the collection of right cosets into a group? Indeed we can! This type of construction, often occurring in mathematics and usually called forming a quotient structure, is of the utmost importance. Let GIN denote the collection of right cosets of N in G (that is, the elements of G/N are certain subsets of G) and we use the product of subsets of G to yield for us a product in G/N. For this product we claim: (1) X, Y € G/N implies XY € GIN; for X = Na, Y = Nb for some a, b € G, and XY = NaNb = Nab € GIN. (2) X, Y, Z € GIN, then X = Na, Y = Nb, Z = Nc with a, b, c € G,

44

GROUP THEORY

CH.2

and so (XY)Z = (NaNb)Nc = N(ab)Nc = N(ab)c = Na(bc) (since G is associative) = Na(Nbc) = Na(NbNc) = X(YZ). Thus the product in GIN satisfies the associative law. (3) Consider the element N = Ne E:: GIN. If X E:: GIN, X = Na, a E::: G, so XN = NaNe = Nae = Na = X, and similarly NX =; X. Consequently, Ne is an identity element for GIN. (4) Suppose X = Na E:: GIN (where a E:: G); thus Na-1 E:: GIN, and NaNa-1 = Noa/" = Ne. Similarly Na-1Na = Ne. Hence Na-1 is the inverse of Na in GIN. But a system which satisfies 1, 2, 3, 4 is exactly what we called a group. That is,
THEOREM 2.c. If G is a group, N a normal subgroup of G, then GIN is also a group. It is called the quotient group or factor group of G by N.

If, in addition, G is a finite group what is the order of GIN? Since GIN has as its elements the right cosets of N in G, and since there are precisely o(G) iG(N) = -such cosets we can say o(N)
LEMMA

o(N) We close this section with an example. Let G be the group of integers under addition and let N be the set of all multiples of 3. Since the operation in G is addition we shall write the cosets of N in G as N + a rather than as N a. Consider the three cosets N, N + 1, N + 2. We claim that these are all the cosets of N in G. For, given a E:: G, a = 3b + c where b E:: G and c = 0, 1, or 2 (cis the remainder of a on division by 3). Thus N + a = N + 3b + c = + 3b) + c = N + c since 3b E:: N. Thus every coset is, as we stated, one of N, N + I, or N + 2, and GIN = IN, N + 1, N + 2}. How do we add elements in GIN? Our formula NaNb = Nab translates into: (N + 1) + (N + 2) = N + 3 = N since 3 E:: N; (N + 2) + (N + 2) = N + 4 = N + 1 and so on. Without being specific one feels that GIN is closely related to the integers mod 3 under addition. Clearly what we did for 3 we could emulate for any integer n, in which case the factor group should suggest a relation to the integers mod n under addition. This type of relation will be clarified in the next section.

oCGIN)

=_.

2.12. If G is a finite group and N is a normal subgroup of G, then o(G)

eN

PROBLEMS
*1. If H is a subgroup of G such that the product of two right cosets of H in G is again a right coset of H in G, prove that H is normal in G.

SEC.

6

NORMAL SUBGROUPS AND QUOTIENT

GROUPS

45

2. If G is a group and H is a subgroup of index 2 in G, prove that H is a normal subgroup of G. 3. If N is a normal subgroup of G and H is any subgroup of G, prove that NH is a subgroup of G. 4. Show that the intersection of two normal subgroups of G is a normal subgroup of G. 5. If H is a subgroup of G and N is a normal subgroup of G, show that H n N is a normal subgroup of H. 6. Show that every subgroup of an abelian group is normal. *7. Is the converse of Problem 6 true? If yes, prove it, if no, give an example of a non-abelian group all of whose subgroups are normal. 8. Let G be a group, H a subgroup of G. Let, for g C G, gHg-1 = jghg-11 h c. H}. Prove that gHg-1 is a subgroup of G. 9. Suppose H is the only subgroup of order o(H) in the finite group G. Prove that H is a normal subgroup of G. 10. If H is a subgroup of G, let N(H) = jg c. GlgHg-1 == H}. Prove: (1) N(H) is a subgroup of G. (2) H is normal in N(H). (3) If H is a normal subgroup of the subgroup K in G, then K c N(H) (that is, N(H) is the largest subgroup of G in which H is normal). (4) H is normal in G if and only if N(H) = G. 11. If Nand M are normal subgroups of G, prove that NM is also a normal subgroup of G. *12. Suppose that Nand M are two normal subgroups of G and that N n M == (e). Show that for any n eN, 1n C M, nm = mn. 13. If a cyclic subgroup T of G is normal in G, then show that every subgroup of T is normal in G. *14. Prove, by an example, that we can find three groups E c F c G, where E is normal in F, F is normal in G, but E is not normal in G. 15. If N is normal in G and a eGis of order o(a), prove that the order, m, of Na in GIN is a divisor of oCa). 16. If N is a normal subgroup in the finite group such that iG(N) and o(N) are relatively prime, show that any element x C. G satisfying xo(N) = e must be in N. 17. 'Let G be defined as all formal symbols XiyJ, i = 0, 1, j = 0, 1, 2, ... , n - 1 where we assume
xiyi = xi'yJ! if

and only if i = i', j = j'

xy = y-1x. (a) Find the form of the product (xV) (xkye) as x"'yf3. (b) Using this, prove that G is a non-abelian group of order 2n.

46

GROUP THEORY

CH.2

(c) If n is odd, prove that the center of G is (e), while if n is even the center of G is larger than (e). This group is known as a dihedral group. A geometric realization of this is obtained as follows: let y be a rotation of the Euclidean plane about the origin through an angle of 2'lr/n, and x the reflection about the vertical axis. G is the group of motions of the plane generated by y and x.
7. Homomorphisms. The ideas and results in this section are closely interwoven with those of the preceding one. If there is one central idea which is common to all aspects of modern algebra it is the notion of homomorphism. By this one means a mapping from one algebraic system to a like algebraic system which preserves structure. We make this precise, for groups, in the next definition.
DEFINITION.

homomorphism

A mapping cp from a group G into a group () is said to be a if for all a, b E::: G, cp(ab) = cp(a)cp(b).

Notice that on the left side of this relation, namely, in the term cp(ab), the product ab is computed in G using the product of elements of G whereas on the right side of this relation, namely, in the term cp(a) cp(b) , the product is that of elements in (). Example wise cp(x) Example

= x for every x E::: G is a homomorphism.
E:: G is really the real number a

o.

cp(x)

=

e all x E::: G. This is trivially a homomorphism. Like-

ob for a, b

1. Let G be the group of all real numbers under addition (i.e., b) and let () be the group of nonzero real numbers with the product being ordinary multiplication of real numbers. Define cp:G -t () by cp(a) = 2a• In order to verify that this mapping is a homomorphism we must check to see whether cp(ab) = cp(a)cp(b), remembering that by the product on the left side we mean the operation in G (namely, addition), that is, we must check if 2a+b = 2a2b, which indeed is true. Since 2a is always positive the image of cp is not all of (), so cp is a homomorphism of G into (), but not onto G.

+

Example 2. Let G = S3 = [e, cp, y", y,,2, cpy",cpy,,2} and () = {e, cpl. Define the mapping f:G -t G by f(cpiy) = cp'. Thus fee) = e, f(cp) = cp, fey,,) = e, f(y,,2) = e, f(cpy,,) = Cp, f(cpy,,2) = cpo The reader should verify that f so defined is a homomorphism. Example 3. Let G be the group of integers under addition and let G = G. For the integer x E::: G define cp by cp(x) = 2x. That cp is a homomorphism then follows from cp(x y) = 2(x + y) = 2x 2y = cp(x) cp(y).

+

+

+

SEC.

7

HOMOMORPHISMS

47

Example 4. Let G be the group of nonzero real numbers under multiplication, = {I, - II, where 1.1 = 1, (- 1)(- 1) = 1, 1(- 1) = (- 1)1 = -1. Define ep:G -t by cf>(x) = 1 if x is positive, cf>(x) = -1 if x is negative. The fact that epis a homomorphism is equivalent to the statements: positive times positive is positive, positive times negative is negative, negative times negative is positive.

a

a

Example 5. Let G be the group of integers under addition, let be the group of integers under addition modulo n. Define ep by ep(x) = remainder of x on division by n. One can easily verify this is a homomorphism. The result of the following lemma yields, for us, an infinite class of examples of homomorphisms. When we prove Theorem 2.d it will turn out that in some sense this provides us with the most general example of a homomorphism.
LEMMA 2.13. Suppose G is a group, N a normal subgroup of G; define the mapping epfrom G to GIN by ep(x) = Nx for all x E::: G. Then cf> a homois morphism of G onto GIN.

an

Proof. In actuality, there is nothing to prove for we already have proved this fact several times. But for the sake of emphasis we repeat it. That cf>is onto is trivial for every element X E::: GIN is of the form X = Ny, y E::: G, so X = cf>(y). To verify the multiplicative property required in order that cf> a homomorphism one just notes that if x, y E::: G, be cf>(xy) = Nxy = NxNy = ep(x)ep(y). In Lemma 2.13 and in the examples preceding it, a fact which comes through is that a homomorphism need not be 1-1; but there is a certain uniformity in this process of deviating from one-to-oneness. This will become apparent in a few lines. DEFINITION. f ¢ is a homomorphism of G into G, the kernel of cf>,Kq" is I defined by Kq, = {x E::: GJ¢(x) = e, e = identity element of G}. Before investigating any properties of Kq, it is advisable to establish that as a set Kq, is not empty. This is furnished us by the first part of
LEMMA

2.14. If epis a homomorphism

of G into G, then:

(1) epee) = e, the unit element of (2) ep(x-1) = ep(x) -1 for all x E::: G.

a;

Proof. To prove (1) we merely calculate ¢(x)e = ep(x) so by the cancellation property in (j we have that ¢(e)

=

cf>(xe) = cf>(x)cp(e),

=

e.

48

GROUP THEORY

CH.

2

To establish (2) one notes that e = cf>(e) = cp(xx-I) = cp(x)cp(x-1), so by the very definition of cp(X)-1 in Owe obtain the result that cf>(x-1) = cf>(X)-I. The argument used in the proof of Lemma 2.14 should remind any reader who has been exposed to a development of logarithms, of that used in proving the familiar results that log 1 = 0 and log (l/x) = -log x; this is no coincidence for the mapping cp:x .....,.log x is a homomorphism of the group of positive real numbers under multiplication into the group of real numbers under addition. Lemma 2.14 shows that e is in the kernel of any homomorphism, so any such kernel is not empty. But we can say even more.
LEMMA 2.15. If cpis a homomorphism is a normal subgroup of G.

of G into G with kernel K, then K

Proof. First we must check whether K is a subgroup of G. To see this one must show that K is closed under multiplication and has inverses in it for every element belonging to K. If x, y E:: K, then cf>(x) = e, cp(y) = e, where e is the identity element of 0, and so cp(xy) = cf>(x)cp(y) = ee = e, whence xy E:: K. Also, if x E:: K, cf>(x) = e, so, by Lemma 2.14, cf>(x-I) = cp(X)-1 = 13-1 = e; thus X-I E:: K. K is, accordingly, a subgroup of G. To prove the normality of K one must establish that for any g E:: G, k E:: K, gkg-I E:: K; in other words, one must prove that cp(gkg-I) = e whenever cp(k) = e. But cp(gkg-1) = cp(g)cp(k)cp(g-l) = cp(g)ecp(g)-l = cp(g)cp(g)-l = e. This completes the proof of Lemma 2.15. Let cp now be a homomorphism of the group G onto the group V, and suppose that K is the kernel of cp. If g E:: 0, we sayan element x E::: G is an inneree image of g under cp if ¢(x) = g. What are all the inverse images of g? For g = e we have the answer, namely (by its very definition) K. What about elements g ~ e? Well, suppose x E::: G is one inverse image of g; can we write down others? Clearly yes, for if k E:: K, and if y = kx, then ¢(y) = ¢(kx) = ¢(k)cp(x) = eg = g. Thus all the elements Kx are in the inverse image of (j whenever x is. Can there be others? Let us suppose that ¢(z) = g = cf>(x). fgnoring the middle term we are left with ¢(z) = cp(x), and so cp(z)cp(x)- = e. But cp(X)-l = cp(x-1) , whence e = cp(z)cp(X)-l = cp(z)cf>(x-1) = cp(c:x-1), in consequence of which zx-1 E:: K; thus z E:: Kx. In other words, we have shown that Kx accounts for exactly all the inverse images of {j whenever x is a single such inverse image. We record this as
LEMMA 2.16. If cpis a homomorphism set of all inverse images of g E: Gunder any particular inverse image of {j in G.

of G onto G with kernel K, then the cp in G is given by K» where x is

A special case immediately presents itself, namely, the situation when

SEC.

7

HOMOMORPHISMS

49

K = (e). But here, by Lemma 2.16, any {j E: G has exactly one inverse image. That is, cf> is a one-to-one mapping. The converse is trivially true, namely, if cf> is a one-to-one homomorphism of G into (not even onto) G, its kernel must consist exactly of e.
DEFINITION.

phism if

cf>

A homomorphism is one-to-one.

cf>

from G into

G is said to be an isomor-

DEFINITION. Two groups G, G* are said to be isomorphic if there is an isomorphism of G onto G*. In this case we write G "'" G*.

We leave to the reader to verify the following three facts:
(1) G "'" G.

(2) G "'" G* implies G* "'" G. (3) G "'" G*, G* ~ G** implies G "'" G**. When two groups are isomorphic, then, in some sense, they are equal. They differ in that their elements are labeled differently. The isomorphism gives us the key to the labeling, and with it, knowing a given computation in one group, we can carry out the analogous computation in the other. The isomorphism is like a dictionary which enables one to translate a sentence in one language into a sentence, of the same meaning, in another language. (Unfortunately no such perfect dictionary exists, for in languages words do not have single meanings, and nuances do not come through in a literal translation.) But merely to say that a given sentence in one language can be expressed in another is of little consequence; one needs the dictionary to carry out the translation. Similarly it might be of little consequence to know that two groups are isomorphic, the object of interest might very well be the isomorphism itself. So, whenever we prove two groups to be isomorphic, we shall endeavor to exhibit the precise mapping which yields this isomorphism. Returning to Lemma 2.16 for a moment, we see in it a means of characterizing in terms of the kernel when a homomorphism is actually an isomorphism. A homomorphism cf> oj G into G with kernel Kq, is an isomorphism oj G into G ij and only if Kq, = (e).
OOROLLARY.

This corollary provides us with a standard technique for proving two groups to be isomorphic. First we find a homomorphism of one onto the other, and then prove the kernel of this homomorphism consists only of the identity element. This method will be illustrated for us in the proof of the very important

50

GROUP THEORY THEOREM 2.D.

CH.

2

Let cf>be a homomorphism

of

GjK

~

G.
Consider the diagram

G onto G with kernel K. Then

Proof.

where (J(g) = Kg. We should like to complete this to

It seems clear that, in order to construct the mapping if; from G/K to G, we should use G as an intermediary, and also that this construction should be relatively uncomplicated. What is more natural than to complete the diagram using
g

Kg'"

1

cf>(g)
",7f

if;
/",'"

",'"

With this preamble we formally define the mapping if; from G/K to G by: if X E::: G/K, X = Kg, then if;(X) = cf>(g). A problem immediately arises: is this mapping well-defined? If X E:: G/K, it can be written as Kg in several ways (for instance, Kg = Kkg, k E: K); but if X = Kg = Kg', g, g' E:: G, then on one hand if;(X) = cf>(g), and on the other, if;(X) = cf>(g'). For the mapping if; to make sense it had better be true that cf>(g) = cf>(g'). So, suppose Kg = Kg'; then g = kg' where k E: K, hence cf>(g) = cf>(kg') = cf>(k)cf>(g') = ecf>(g') = cf>(g') since k E:: K, the kernel of cf>. We next determine that if; is onto. For, if x E: ii, x = cf>(g), g E::: G (since cf> is onto) so x =. cf>(g) = if;(Kg). If X, Y E: G/K, X = Kg, Y = Kf, g, f E:: G, then XY = KgKf = Kgf, so that if;(Xy) = if;(Kgf) = cf>(gj) = cf>(g)cf>(j) since cf> is a homomorphism of G onto rJ. But if;(X) = if;(Kg) = cf>(g), if;(Y) = if;(Kf) = cf>(f), so we see that if;(XY) = if;(X)if;(Y), and if; is a homomorphism of G/K onto ii. To prove that if; is an isomorphism of G/K onto ii all that remains is to demonstrate that the kernel of if; is the unit element of G/K. Since the unit

SEC.

7

HOMOMORPHISMS

51

element of G/K is K = Ke, we must show that if 1{;(Kg) = e, then Kg = Ke = K. This is now easy, for e = 1J;(Kg) = cp(g), so that cp(g) = e, whence g is in the kernel of cp, namely K. But then Kg = K since K is a subgroup of G. All the pieces have been put together. We have exhibited a one-to-one homomorphism of G/K onto G. Thus G/K ~ G, and Theorem 2.d is established. Theorem 2.d is important for it tells us precisely what groups can be expected to arise as homomorphic images of a given group. These must be expressible in the form G/K where K is normal in G. But, by Lemma 2.13, for any normal subgroup N of G, G/N is a homomorphic image of G. Thus there is a one-to-one correspondence between homomorphic images of G and normal subgroups of G. If one were to seek all homomorphic images of G one could do it by never leaving G as follows: find all normal subgroups N of G and construct all groups G/N. The set of groups so constructed yields all homomorphic images of G (up to isomorphisms). A group is said to be simple if it has no nontrivial homomorphic images, that is, if it has no nontrivial normal subgroups. A famous, long-standing conjecture is that a non-abelian simple group of finite order has an even number of elements. It still awaits proof. t We have stated that the concept of a homomorphism is a very important one. To strengthen this statement we shall now show how the methods and results of this section can be used to prove nontrivial facts about groups. When we construct the group G/N, where N is normal in G, if we should happen to know the structure of G/N we would know that of G "up to N." True, we blot out a certain amount of information about G, but often enough is left so that from facts about G/N we can ascertain certain ones about G. When we photograph a certain scene we transfer a three-dimensional object to a two-dimensional representation of it. Yet, looking at the picture we can derive a great deal of information about the scene photographed. In the two applications of the ideas developed so far, which are given below, the proofs given are not the best possible. In fact, a little later in this chapter these results will be proved in a more general situation in an easier manner. We use the presentation here because it does illustrate effectively many group-theoretic concepts.
ApPLICATION 1 (CAUCHY'S THEOREM FOR ABELIAN GROUPS). Suppose G is ajinite abelian group and p I o(G) where p is a prime number. Then there is an element a ,= e E:: G such that aP = e.

Proof. We proceed by induction over o(G). In other words, we assume that the theorem is true for all abelian groups having fewer elements than G. From this we wish to prove that the result holds for G. To start the

t This important result has just been proved by the two young American mathematicians Walter Feit and John Thompson.

52

GROUP THEORY

CH.2

induction we note that the theorem is vacuously true for groups having a single element. If G has no subgroups H ~ (e), G, by the result of a problem earlier in the chapter G must be cyclic of prime order. This prime must be p and G certainly has p - 1 elements a ~ e satisfying aP = ao(G) = e. So suppose G has a subgroup N ~ (e), G. If pi o(N), by our induction hypothesis, since o(N) < o(G) and N is abelian, there is. an element b E: N, b ~ e, satisfying bP = e; since b E: NeG we would have exhibited an element of the type required. So we may assume that p t o(N). Since Gis abelian, N is a normal subgroup of G, so G/N is a group. Moreover, o(G) . j o(G) .. o(GIN) = --, and SInce p t o(N), p -< o(G). Also, since G IS o(N) o(N) abelian, G/N is abelian. Thus by our induction hypothesis there is an element X E: G/N satisfying XP = e1, the unit element of G/N, X ~ e1. By the very form of the elements of G/N, X = Nb, b E: G, so that XP = (Nb)P = NbP. Since e1 = Ne, XP = ell X ~ e1 translates into NbP = N, Nb ~ N. Thus bP E: N, b t[_ N. Using one of the corollaries to Lagrange's theorem, (bP)o(N) = e. That is, bo(N)p = e. Let c = bo(N). Certainly cP = e. In order to show that c is an element that satisfies the conclusion of the theorem we must finally show that c ~ e. However, if c = e, bo(N) = e, and so (Nb)o(N) = N. Combining this with (Nb)P = N, p t o(N), p a prime number, we find that Nb = N, and so b E: N, a contradiction. Thus c ~ e, cP = e, and we have completed the induction. This proves the result.
APPLICATION 2 (SYLOW'S THEOREM FOR ABELIAN GROUPS). If G is an abelian group of order o(G), and if p is a prime number, such that pc>'o(G), I pIX+! t o(G), then. G has a subgroup of order pIX.

Proof. If IX = 0 the subgroup (e) satisfies the conclusion of the result. So suppose IX ~ O. Then pi o(G). By Application 1, there is an element a ~ e E: G satisfying aP = e. Let S = {x E: G I xP" = e some integer n}. Since a E: S, a ~ e, it follows that S ~ (e). We now assert that S is a subgroup of G. Since G is finite we must only verify that S is closed. If z, n m n-l-m n-l-m n+m y E: S, xP = e, yP = e, so that (xy)P = xP yP = e (we have used that G is abelian), proving that xy E: S. We next claim that o(S) = pf3 with fJ an integer 0 < fJ ::;; IX. For, if some prime qlo(S), q ~ p, by the result of Application 1 there is an element c E: S, c ~ e satisfying cq = e. However, cpr> = e for some n since c E: S. Since p", q are relatively prime, we can find integers "A, Jl. such that "Aq + tsp" = 1, so that c = c1 = cAq+I'P" = (cq)"(cpn)1' = e, contradicting c ~ e. By Lagrange's theorem o(S) 10(G), so that fJ ::;; IX. Suppose that fJ < IX; considertheabeliangroupG/S. SincefJ

< IX and o (G/S)

= o(G) ,plo(G/S), o(S)

Consequently Sx = S contradicting Sx ~ S. {e. our assertion has been established. o(G) = 2. so that K c H follows. with kernel exactly K.b o(ST) = o(S o(S)o(T) p"p" n T) = o(S n T) and since S ~ T. for if x ~ K. S is the required subgroup of order o". that (3 = a. Suppose now that x. Furthermore. p. But S = (Sx)pn = Sxpn. But Sylow's theorem holds for all finite groups. 'Y > a. By Theorem 2. . consequently . 7 HOMOMORPHISMS 53 there is an element Sx. Let H = {x E:: GI <t>(x) ~ Il}.. x satis fies t h e exact requirements x needed to put it in S.+l t o(G). which is non-abelian. since H is normal in G. (Sx)pn = S for some integer n > O. ghg-1 E:: H. We strengthen the application slightly. from which it follows that H is normal in G. If G is abelian of order o(G) and p" 1 o(G). x ~ S. namely. Since ST is a subgroup of G. o(ST) 1 o(G). h E:: H. Suppose. is a unique subgroup of G of order p". <t>(x) = e is in H. We leave the application and return to the general development. Thus no such subgroup T exists.SEC. l ~ L}. ¢1/. xy E:: Hand H is closed under the product in G. Th erefore. leaving us with o(ST) = p"Y. (x ~ G) in GIS satisfying Sx ~ S. Thus (3 < a is impossible and we are left with the only alternative. and suppose that H is a subgroup of G. Can we explicitly describe the subgroup T = (y E:: GI<t>(y) ~ L}? . namely. namely. o(S n T) < p". Suppose ¢ is a homomorphism of G onto G with kernel K. ¢1j. conversely. if x ~ H. that L is a subgroup of G and K c L. All in all. [e. Let L = I x E:: G 1 x = ¢(l). that the homomorphism ¢ from G onto G. We assert that H is a subgroup of G and that H ::) K. by Theorem 2. and S is the unique subgroup of order p". We have proved the COROLLARY. What can we say in addition in case H is normal in G? Let g ~ G. so that ST is a subgroup of G. whence ¢(ghg-l) = ¢(g)¢(h)¢(g)-l E:: H. there If we look at G = S3. and so xpn ~ S. thus p"Y1 o(G) violating the fact that a is the largest power of p which divides o(G). The reader should verify that L is a subgr~up of G. e = ( x pn)o(S) = (pn)p/3 = x pn+/3 . [e. y ~ H. we see that G has 3 distinct subgroups of order 2.3. One other point should be noted. ¢}.}. Otherwise stated. in other words. induces a homomorphism of H onto H. since K c H. Suppose T is another subgroup of G of order pa. Therefore. so that the corollary asserting the uniqueness does not carryover to nonabelian groups. That H :J K is trivial.2}. T ~ S. when just considered on elements of H. hence <t>(x) E:: Il. Since G is abelian ST = TS.d we have that Il "" H IK. then ¢(h) E:: Il. <t>(y) ~ Il from which we deduce that <t>(xy) = <t>(x)<t>(y)~ H. ¢(x) ~ Il and so <t>(x-1) = <t>(X)-1 ~ H from which it follows that X-I E:: H.

(ab) = H¢(ab) by the definition of the mapping 1/. Comparing these two evaluations of 1/I(t). We wish to prove one more general theorem about the relation of two groups which are homomorphic. That is. proving that NeT. Let rjJbe a homomorphdm. so by the very definition of L. bE:: G. On the other hand. there is a homomorphism B of V onto VIN defined by B(g) = Ng. l)IN. N "'" N/K.. so the typical element Ng in VIN can be represented as HrjJ(g) = 1/. and let E:: Nl. Let ¢ be a homomorphism N be a normal subgroup of l).(t) = HrjJ(t). Then GIN"". suppose t E: T. T. The last statement in the theorem is immediate from the observation (following as a consequence of Theorem 2. if n E: N. . for if g E:: l). ij = rjJ(g) for some g E:: G.(g) = N ¢(g) for all 9 E:: G. Thus 1/.(n) = N ¢(n) = N. as been proved h to be equal to N. But then 1/1 is a homomorphism of G onto VIH with kernel N.is a homomorphism of G onto l)IH. By Theorem 2. if H is normal in l). in this correspondence. but 1/. N = [z E:: Equivalently. Then H is a subgroup of G and H ::J K.:G ~ l)IN by 1/. which forces rjJ(t) E:: H. the identity element of VIH. whence a:' E:: K c L. For H a subgroup of V let H be defined by H = Ix E:: GlrjJ(x) E:: HI. thus rjJ(t) E: L.d) that V "'" GIK.E. VIN. Moreover. Proof.d GIN"". VIN "'" (G/K)/(NIK). 1/. but this places t in N by definition of N. 1/. which. of 1/. ¢(ab) = rjJ(a)¢(b). is onto. which is the first part of the theorem.(ab) = H¢(a)¢(b) = N¢(a)HrjJ(b) = 1/. this association sets up a one-to-one mapping from the set of all subgroups of V onto the set of all subgroups of G whwh contain K. a normal subgroup of G corresponds to a normal subgroup of l).?Firstly. If a. Thus rjJ(tl-1) = rjJ(t)rjJ(l)-l = e.(b).54 GROUP THEORY CH. Is there any element t E:: T which is not in L? So. Thus we have set up a one-to-one correspondence between the set of all subgroups of l) and the set of all subgroups of G which contain K. GI¢(x) of G onto G with kernel K. GIN"". then H is normal in G. since ¢ is onto. To begin with. The kernel of 1/.(t) = identity element of VIH = H. So far we have shown that 1/.2 Clearly LeT. ¢(n) E: N. TeN. if t E:: T.(g). combined with LeT yields that L = T.1/. since ¢ is a homomorphism.17. thus t is in Ll = L. Moreover. (GIK)/(NIK). What is the kernel.(a)1/. rjJ(t) = rjJ(l) for some l E:: L. THEOREM 2. As we already know. We define the mapping 1/.However. of G onto G with kernel K. we arrive at N = N ¢(t). Equivalently we have proved that TeL. so that 1/. We summarize these few paragraphs in LEMMA 2.

prove that tJ is a normal subgroup of G. i = 0. y E: G}.provethatNM/M "'" N/N n M. ° ° . . S. Let G be a finite abelian group of order o(G) and suppose the integer n is relatively prime to o(G). Prove that the center of a group is always a normal subgroup. verify if the mappings defined are homomorphisms. prove that N ::::> G'. (c) G is the group of real numbers under addition. then H is normal inG. ¢(x) = x5 all x E: G. ¢(x) = 13x for x E: G. 6. (d) G. Let V be the set of real numbers. n . Prove that a group of order 9 is abelian. 7.1 where x2 = e. (a) G is the group of nonzero real numbers under multiplication.) (b) If gug-I E: U for all 9 E: G. j = 0. (a) Prove that G' is normal in G. g a fixed element in G. In this case 0 is usually written as G' and is called the commutator subgroup of G. and prove this mapping is an isomorphism of G onto G.. 2. ¢(x) = x2 all x E: G. determine the kernel. Prove: (a) The subgroup N = {e. 1. (U is called the subgroup generated by U. y" = e. (b) That GIN"". xy = y-Ix. (e) G is any abelian group. In the following..:o£ let Tab: V -t V defined by Tab(X) = ax + b. 7 PROBLEMS HOMOMORPHISMS 55 1. yn-I} is normal in G. Define ¢: G ---7 G by ¢(x) = gxg-I• Prove that ¢ is an isomorphism of G onto G. V = G. Prove there is such a subgroup U in G. 9. Let G = {Tab [a. 6. a. (Hint: Consider the mapping ¢:G ---7 G defined by ¢(y) = y". . (b) Prove that G/G' is abelian. Let U = {xyx-Iy-I[X. Vas in (c)... y. Prove that J{ is a normal subgroup of G and that GIN "'" group of nonzero real numbers under multiplication. . Prove that every 9 E: G can be written as 9 = x" with x E: G. W. V = G.) 4. V = G. and in those cases in which they are homomorphisms. where W = {I. and for a. 1. IfN. . Vas in (a). Let G be any group. Let G be the dihedral group defined as the set of all formal symbols x'yJ. (d) Prove that if H is a subgroup of G and H :::> G'. y2. let i: be the smallest subgroup of G which contains U. ¢(x) = 2:1>. b real. b real. -I} is the group under the multiplication of the real numbers. 10. (a) Given any group G and a subset U. (c) If GIN is abelian.:o£ I and let N = {TIb E: G}. (b) G. u E: U. ¢(x) = x + 1 allx E: G.MarenormalsubgroupsofG. 3.SEC. a.

2 11. is in (1. There is only one other fact that needs verifying in order that a(G) be a subgroup of A(G). In the preceding section the concept of an isomorphism of one group into another was defined and examined. Automorphisms. We use the word "into" advisedly.(G).(G) we can use the product of A(G).(G) is not empty. thus (xT-1)(yT-1) = (xy)T-r. then xT is the image of x under T.56 GROUP THEORY CH. We proceed to verify this. whenever we talk about mappings of a set into itself we shall write the mappings on the right side. Also I. We want it to be in the smaller set (1. By an automorphism of a group G we shall mean an isomorphism of G onto itself. The special case in which the isomorphism maps a given group into itself should obviously be of some importance. so. ¢ is a homomorphism. the unit element of A(G). DEFINITION. TIT2 E:: (1.(G). T2 are in (1. placing T-1 in a(G). That is. prove that G "'" S3. therefore (xy)T1T2 = «xy)T1)T2 = «xT1)(yT1»T2 = «xT1)T2)«yT1)T2) = (xTS2) (yT1T2). thus if T:S --7 S. namely. being a subset of A (G). y E: G. (1. Also if the image of x and y under ¢ are equal. Trivially I is an automorphism of G.(G) denote the set of all automorphisms of G.( G) we already know that TIT 2 E: A (G). for instance. so (1. x E:: S. xI = x for all x E:: G. Since ¢:x + y --7 2(x + y) = 2x + 2y. ¢ is thus an isomorphism. For all x. If x. Summarizing these remarks we have proved . 8. If G is abelian and if N is any subgroup of G. If G is a non-abelian group of order 6.(G). This product then satisfies the associative law in A(G). and not onto. then T-1 E:: a(G). y E:: G. Yet ¢ is not onto. itself. 1 does not appear an image under ¢ of any element of G. for elements of (1. a fortiori. As we mentioned in Chapter 1. in its own rights. for the image of any integer under ¢ is an even integer.(G) is a subgroup of A(G). namely. 12. then 2x = 2y whence x = y.(G) should be a group. Of greatest interest to us will be the isomorphisms of a group onto itself. that if T E:: a(G). that is. the set of 1-1 mappings of G onto itself. then «xT-1)(yT-1»)T = «xT-1)T)«yT-1)T) = (xI)(yI) = xv. An obvious fact that we should try to establish is that (1. for groups G do exist which have isomorphisms mapping G into. The easiest such example is the following: Let G be the group of integers under addition and define ¢: G --7 G by ¢: x -> 2x for every x E:: G. (xy) Tl = (xT 1) (yT1). and so. Let (1. If T1. and so. (xy) T2 = (XT2) (yT2).(G). prove that GIN is abelian. Let I be the mapping of G which sends every element onto itself. composition of mappings. in (1.

But the center of G. and suppose go E: K. we can write down an explicit automorphism. then g-1xg = g-lyg. T is onto. the set of auiomorphiems Of course. Now consider.(go) = I. there is a pair a.:e: ba. xTgO = x. but then bTa = a-1ba . If G is a group. then 1/. h E: G. If the group G is abelian and there is some element Xo E: G satisfying Xo . equivalently. I). xTgo = go-1xgo. Thus gox = gogo-1xgo = xgoi go must commute with all elements of G. so To is onto. We further assert that Tg is one-to-one. or. in general.That is.:e: I. Let G be a group. If G is a group having only two elements. whence 'I'.:e: Xo-1.) Thus K c Z. was defined to be precisely all elements in G which commute with every element of G. (xy)Tg = g-l(xy)g = g-1(xgg-1y)g = (g-lxg)(g-lyg) = (xTg)(yTg). But this says that for any x E: G. x = y. (See Problem 13. for x. so T. for given y E: G. Also xoT = xo-1 . Let fJ (G) = IT g E:: (t( G) I g E:: G I.? Suppose we call it K. Looking at the start and finish of this chain of equalities we find that Tgh = T gTh. This little remark is both interesting and suggestive.(g) = T g for every 9 E: G. LEMMA is also a group.:e: Xo. and so x = go-1xgo for all x E: G.(gh) = Tglt = TgTh = 1{. Z. y E: G. and we should like to have some automorphisms of non-abelian groups. Tg is called the inner automorphism corresponding to g. (xy)T = (xy)-l = y-1x-1 = x-1y-1 = (xT)(yT). 8 AUTOMORPHISMS 57 of G. Section 5. However. by definition. What we should like is a richer sample of automorphisms than the ones we have (namely.:e: I. Then xTg = g-l(X)g = g-1(gyg-1)g = y. Consequently Tg is a homomorphism of G onto itself.:e: b. Tf( is onto. for s. for any abelian G. might be of some interest.: G ~ (t(G) defined by 1{. = I and so z E: K. so by the cancellation laws in G. First. (Verify!) d(G) is usually called the group of inner automorphisms of G. What is the kernel of 1/. for g E: G define Tg:G ~ G by xTg = g-1xg for all x E: G. Strangely enough the task of finding automorphisms for such groups is easier than for abelian groups. as yet. It is of interest because it immediately yields that d(G) is a subgroup of (t(G). for if xTg = yTg. the mapping T defined by xT = X-I for all x E: G. is a homomorphism of G into (t(G) whose image is d(G). The computation of T gh. If G is non-abelian. For any group G. For groups G with more than two elements (t(G) always has more than one element. so that T a . has elements other than I. b E: G such that ab .(g)1{. .18. 2. for if we consider the mapping 1/. TgO = I. Thus for a non-abelian group G there always exist nontrivial automorphisms. However. So. the reader should convince himself that (t(G) consists only of I. let x = gyg-1. then xT% = z-lxz = Z-1(ZX) (since zx = xz) = x. Then 1/. however. It is suggestive. if z E:: Z. we have no way of knowing that (t(G). then (t(G) . We claim that Tg is an automorphism of G. XTgh = (gh)-IX(gh) = h-1g-1xgh = (g-1xg)Th = (xTg)Th = xTgTh. suppose x E: G.SEC. 1{.(h). the class of abelian groups is a little limited.

which implies.20). -. aT must have the same order as a (Lemma 2. is a homomorphism of G into a(G) with image d(G) and kernel Z. We close the section by determining a(G) for all cyclic groups. If a E:: a is Automorphisms of groups can be used as a means of constructing new groups from the original group. we consider a particular example.19. T}. Let x be a symbol which we formally subject to the following conditions: x3 = e. a a a Example 1. dj r. For instance. a = (a). g E:: a subject to xig = Xi'g' if and only if i == i' mod r. 1. aT = e. Let be a cyclic group of order 7. consists of all a\ where we assume a7 = e. forces t to be relatively prime to r. 1. However. and consider all formal symbols xiai where i = 0.2.. is an automorphism of of order 3.6 and where we declare that xiai = xkal if and only if i == k mod 3. since ep is one-to-one. pick a symbol x and consider all elements xig. The mapping ep:ai ---+ a2i. . This way we obtain a larger group (G. . hence ep(a)n = e. an = e but for no lower positive power). xaxa2 = x(ax)a2 = x(xa3)a2 = x2a5. as can be trivially checked. j = 0. if G is a group. ±1.58 GROUP THEORY CH. Summarizing. that is. a non-abelian group of order 2l. Since aT E:: G. we claim. so gT is determined for all g E:: = (a). that is. l == j mod 7. since T is an automorphism. In order to emphasize this general result we record it as LEMMA 2. i = 0. then ep(am) = ep(a)m = e.d d(G) "'" GjZ. ±2. Then ¢(a)n = ¢(an) = epee) = e.f. x-lax = a3. x-1aix = ep(a)i = a3i. Z c K. T} and (a. We multiply these symbols using the rules: x3 = a7 = e.. Suppose that ep is an automorphism of a group a. and this condition.2. Let G be a finite cyclic group of order r. For if dj t. then o(ep(a» = o(a). Let a be a group and epan automorphism > 0. a is normal in (G.. . epa = I. aiT is determined. of a. thus aT has order a ° . and Z is the center oj G. g = g' and x-lgix = gTi for all i. where d(a) is the group of inner automorphisms oj G. Thus we need consider only possible images of a under T. Thus ° LEMMA oj order o(a) 2. then (aTrld = at (Tid) = aT(tld) = (aT)t/~ = e. aT = at for some integer < t < r. and since every element in G is a power of a. 2 Therefore. Suppose T is an automorphism of G.20. Having proved both K C Z and Z C K we have that Z = K. . that am = e.. Generally. in this way. Before explaining this abstractly. If aT is known. T}/a "'" group generated by T = cyclic group of order r. and suppose that a E:: a has order n (that is. If ¢(ar = e for some < m < n. By Theorem 2. d(a) "'" GjZ. since aiT = (aT)i. The reader can verify that one obtains. T an automorphism of order r of G which is not an inner automorphism. a contradiction.

8 AUTOMORPHISMS 59 a divisor of rid. H a subgroup of G. the . 8. Let G be a group of order 4. b. ±2. Prove a characteristic subgroup of G must be a normal subgroup of G. The question now becomes.SEC. Thus here. for any < 8 < r and relatively prime to r . The mapping i -7 T« exhibits the isomorphism of U. TiTj:a -7 ai -7 (ai)i = aii. Let G be a group. so that a = gT for some g E: G. Since aT = at. T:x -7 x-I. (b) G group of positive reals under multiplication. Are the following mappings automorphisms of their respective groups? (a) G group of integers under addition. t = -1 giving rise to the automorphism T:g -7 g-I for every g in the cyclic group G.1 = 0. 2. what values of t are possible? Since T is an automorphism of G. Suppose that T is an automorphism of G. That is. G = (e. 5. of integers less than r and relatively prime to r under multiplication modulo r. and each of these gives rise to an automorphism. ab}. As in Example 1. (lCG) ~ U'. For any group G prove that £leG) is a normal subgroup of Cl(G) (the group ce(G)/fJ(G) is called the group of outer automorphisms of G). this must force t = ±1. St ° ° Example 2. (b) Prove that the converse of (a) is false. . which. For G = S3 prove that G ~ £l(G). Cl(G) ~ cyclic group of order 2. ±1. Thus ce(G) is in 1-1 correspondence with the group U. Clearly. ti = 1. but there is one which furthermore is an isomorphism. that is. 7. T:» -7 x3• (d) G is the group S3. Prove (H) T is a subgroup of G. Conversely. G consists of all a". i = 1 yielding the identity automorphism I. N a normal subgroup of G. T an automorphism of G. Determine Cl(G). leads us to d = 1. aT = at. we must have that a = at" so that ati-1 = e. Thus a = aiT = (aT)i for some integer i. onto ceCG). T:x -7 x2• (c) G cyclic group of order 12. Here then. < i < rand relatively prime to r. G is an infinite cyclic group. 4. . PROBLEMS 1. T an automorphism of G. a. T:x -7 -x. i = 0. combined with the fact that a/I' has order r . '" where we assume that ai = e if and only if i = 0. Let G be a group. 6. thus TiTi = Tij. a2 = b2 = e. ab = 00. Let us label the elements of ce(G) as Ti where Ti:« -7 ai. Hence ti . (a) A subgroup C of G is said to be a characteristic subgroup of G if CC) T c C for all automorphisms T of G. it maps G onto itself. mapping St a" -7 a is an automorphism of G. Prove that (N) T is a normal subgroup of G. Let (H) T = {hT I h E: H}. We claim not only is there such a 1-1 correspondence. since t and i are integers.

(8n = A (8) when 8 is a finite set with n elements. implies that x = y. Let G be a finite group and suppose the automorphism T sends more than three-quarters of the elements of G onto their inverses. Prove that G must be abelian. consider Tgh. If y E:: G. THEOREM 2. h E:: G. For the set 8 we will use the elements of G. *16. *15. N a normal subgroup of G. Very often it was in the form of a set of transformations of some particular mathematical object. Proof. Let G be a group. 9. then y = (yg-1)g = (yg-1)Tg. We have proved that for every g E: G. prove that I n/ cf>(an - 1). 13. (See Problem 5. 10. Ugh = x(gh) = (xg)h = . Cayley's Theorem. that is.60 GROUP THEORY CH. f a > 1 is an integer. T an automorphism of G with the property that xT = x if and only if x = e. by the cancellation property of groups. Suppose further that T2 = I. In Problem 12. so that re maps 8 onto itself. put 8 = G.) 9. Let G be a finite group. Moreover. Prove that H is of odd order and is an abelian subgroup of G. Let G be a finite group. y E:: 8 and XTg = yTg. in this section. Tg E:: A (8). Suppose that half of the elements of G are of order 2. Let cf>(n) be the Euler cf>-function. as subgroups of 8n. Prove that every g E:: G can be represented as g = x-1(xT) for some z E:: G. can you find an example of a finite group which is non-abelian and which has an automorphism which maps exactly threequarters of the elements of G onto their inverses? *14. Prove that xT = x-1 for all x E:: G and that G is abelian. most finite groups appeared as groups of permutations.F (CAYLEY). and the other half form a subgroup H of order n. If g. T an automorphism of G with the property that xT = x for x E:: G if and only if x = e. For any group G. Prove that every finite group having more than two elements has a nontrivial automorphism. If G is a group. prove that the commutator subgroup G' is a characteristic subgroup of G. Section 7. then xg = yg.2 8. If g E:: G define t e :8( = G) --t 8( = G) by XTg = xg for every x E:: G. Every group is isomorphic to a subgroup of A (8) for some appropriate 8. which. When groups first arose in mathematics they usually came from some specific source and in some very concrete form. Tg is one-to-one. For any z E: 8 = G. will be with a presentation of Cayley's theorem and some related results. for if z. Our concern. *12. M a characteristic subgroup of N. In fact. 11.) The English mathematician Cayley first noted that every group could be realized as a subgroup of A (8) for some 8. that is. Let G be a group of order 2n. prove that M is a normal subgroup of G.

if b E::G is such that Hxb = Hx for every x E: G. Hob = Ha for every a E: G. We wish to show this is the case. Therefore.?If go C K. Can one always say that 0 is an isomorphism? Suppose that K is the kernel of O. and using the definition of tgO' namely. Thus K = lb E::GIHxb = Hx all x E: G}. we must have that HatgO = Ha for every a E: G. We claim that from this characterization of K. is huge in comparison to G. We first explain the use of the word largest. in fact. proving the theorem. We ask: can we find a more economical S. Thus by the corollary to Lemma 2. by this we mean that if N is a normal subgroup of G which is contained in H. as a group of mappings. HaigO = Hago. If go C K.(go) = Tg is the identity map on S. we arrive at the identity Hago = Ha for every a E::G. whence b E::H. whence K = (e). so that Hana-1 = H. if N is a normal subgroup of G which is contained in H. then if. with its o(G)! elements. for if b E: K.16 if. if 1/I:G ~ A(S) is defined by if. Let S be the set whose elements are the right cosets of H in G. as in our proof. That K is a normal subgroup of G follows from the fact that it is the kernel of a homomorphism of G. From XTgh = XTgTh we deduce that Tgh = TgTh. Therefore. is an isomorphism of G into A(S). is a homomorphism. XtgO = X. H a subgroup of G. and. n E: K by our characterization of K. if n E: N.(g) = Tg. What is the kernel K of if. The theorem enables us to exhibit any abstract group as a more concrete object. On the other hand.£ we can easily prove: (1) tg E: A(S) for every g C G. Hb = Heb = He = H. However. we can make our group G act on S in the following natural way: for g C G let tg:S ~ S be defined by (Hx)ig = Hxg. Thus comparing these two expressions for ergo we conclude that go = e. S need not be a group itself. . Finally. Let G be a group. the relation Tgh = TgTh tells us that if. Now we assert that K c H. it has its shortcomings. one for which A(S) is smaller? This we now attempt to accomplish.SEC. in particular. namely. for e C G~ eTgo = e. it would be a group only if H were a normal subgroup of G. using S = G. so that for every X E: S. then ana-1 E: N c H. (2) tgh = tgth' Thus the mapping 8:G ~ A(S) defined by 8(g) = ig is a homomorphism of G into A (S). retracing our argument we could show that bE:: K. then N must be contained in K. 9 CAYLEY'S THEOREM 61 (XTg)Th = XTgTh. That is. a E: G. thus Han = Hafor all a E: G. Our group G of order o(G) is somewhat lost in the group A(S) which. But eTgo = ego = go. A(S) has o(G)! elements. then O(go) = tgO is the identity map on S. for if G is a finite group of order o(G). in particular. then. However. so. so that for x C G. Note that we used the associative law in a very essential way here. S = lHglg E: G}. K must be the largest normal subgroup of G which is contained in H. Emulating the proof of Theorem 2. Since every element of S is a right coset of H in G.

the number of right cosets of H in G) satisfies i(H)! < o(G). of G. as above.21. Suppose that G has a subgroup H whose index i(H) (that is. and also as a means of representing certain finite groups as permutation groups on small sets. that is.G. Then i(H) = 4. and since 99 t 9! there is a nontrivial normal subgroup N r" (e) of Gin H. The mapping. of order a divisor of 9. Applications (1) Let G be a group of order 36. The case H = (e) just yields Cayley's theorem (Theorem 2. However the argument used above has implications even when i(H) ! is not less thano(G).62 GROUP THEORY CH. this kernel being the largest normal subgroup of G which is contained in H. For we shall use this observation both as a means of proving certain finite groups have nontrivial normal subgroups. (2) Let G be a group of order 99 and suppose that H is a subgroup of G of order 11 (we shall also see. implying that N = H. of Theorem 2. for if it were. hence no subgroup isomorphic to G. If H should happen to have no normal subgroup of G.f). . B.2 We have proved THEOREM 2. H itself is a normal subgroup of G. Since H is of order 11. If o(G) does not divide i(H)! then by invoking Lagrange's theorem we know that A(S) can have no subgroup of order o(G). other than (e). 4! = 24 < 36 = 0(G) so that in H there must be a normal subgroup N r" (e). and H r" G is a subgroup of G such that o(G) t i(H)! then H must contain a nontrivial normal subgroup of G. We examine these remarks a little more closely. in it. Suppose that G has a subgroup H of order 9 (we shall see later that this is always the case). which is a prime. later. If G is a finite group. A(S) does contain 8(G). G cannot be simple.g cannot be an isomorphism. That is. But then. H must contain a nontrivial normal subgroup of G.f. ()cannot be an isomorphism. If G is a group. This is interesting mostly for finite groups. 8(G) would have o(G) elements and yet would be a subgroup of A(S) which has i(H)! < o(G) elements. In particular. and S is the set of all right cosets of H in G. In this case we would have cut down the size of the S used in proving Theorem 2. Therefore the kernel of 8 must be larger than (e). then 8 must be an isomorphism of G into A(S). However. that this must be true). Then i(H) = 9. of order 3 or 9. we can conclude that H contains a nontrivial normal subgroup of G. whence B(G) cannot be isomorphic to G. H a subgroup of G. Let S be the set of all right cosets of H in G. its only subgroup other than (e) is itself. then there is a homomorphism B of G into A (S) and the kernel of e is the largest normal subgroup of G which is contained in H. that is. We summarize this as LEMMA 2.

for the moment. defined for g E:: G by XAg = gx for all x E:: G. Let bE:: G.YLEY'S THEOREM 63 (3) Let G be a non-abelian group of order 6. prove that any group of order p2 is abelian. gHg-1 is a subgroup of G. Thus the subgroup H = Ie.6 e. G ~ A(S) = S3. and N(b) :::> Hi N(b) . By Problem 11. making G into an abelian group. Since it might be of some interest we go through a detailed proof of this.SEC. then for every g E:: G. 2 < n ::. (a) If H is a subgroup of G show that for every g E:: G. 8.6 e C G satisfying a2 = e.21 prove that a group of order p2. that ga = ag for every g C G. and consider N(b) = {x E:: Glxb = bx}. We would have proved that any non-abelian group of order 6 is isomorphic to Sa. Let G be a group i consider the mappings of G into itself. 9. Since H has only itself and (e) as subgroups. Ag. there is an a . If (J is a 1-1 mapping of G onto itself such that AgO = (JAg for all g E:: G. prove that any group G of order 2p must have a subgroup of order p. satisfy AgTh = ThAg.6 H since b E:: NCb). b ([_ H. H has no nontrivial normal subgroups of G in it. Section 3. the mappings Ag. a: PROBLEMS 1. 4. contrary to assumption. 9 CA. *6. equivalently. where p is a prime number. Thus H could not have been normal in G. h E:: G. By an earlier problem. Prove that for any g. since gag-1 E:: Hand gag-1 . a} is of order 2. All that remains is to show that H is not normal in G. Since H is a subgroup of N(b). Thus G is isomorphic to a subgroup T of order 6 in A(S). and i(H) = 3. N(b) is a subgroup of G. The only even number n. prove that 0 = Th for some h E:: G. mP == m mod p. o(H) 1o(N(b)) 16. Since o(A(S)) = iCH)! = 3! = 6. If oC G) is pq where p and q are distinct prime numbers and if G has . and that Agh = AhAg. and that this subgroup is normal in G. where S is the set of right cosets of H in G. that we know that H is not normal in G. a} were normal in G. (b) Prove that W = intersection of all gHg-1 is a normal subgroup of G. Using the result of Problem 6. (Hint: if m is an integer. If p is a prime number. or. So o(N (b)) = 6 i whence b commutes with all elements of G. Prove that Ag is one-to-one and onto. Thus every element of G commutes with every other element of G. 2. b H.6 which divides 6 is 6. Let Ag be defined as in Problem 1. (Hint: for x C G consider X(AgTh) and X(ThAg). rt. must have a normal subgroup of order p.) 3. t e as in the proof of Theorem 2. Using Lemma 2. In other words. we would have that gag-1 = a. If H = {e. T = S. This proof is somewhat long-winded but it illustrates some of the ideas already developed. Show that in a group G of order p2 any normal subgroup of order p must lie in the center of G. Suppose.£.) *7. 5.

¢:Xl --+ X2. But this is very cumbersome. 241 3 Given two permutations 0. in particular. (b) If q t p . Returning ¢ might be represented by to our example just above. Xa --+ Xl. then G is cyclic. 0 takes 1 into iI. X2 --+ X4. for some n. Then repeat this . using this symbolic representation of 0 and if. Let o(G) be pq. where Sn is the symmetric group of degree n. xn. say. then ¢ is a 1-1 mapping of S onto itself. We have seen that every group can be represented isomorphically as a subgroup of A(S) for some set S. (c) Given two primes p..g. This clearly shows that the groups Sn themselves merit closer examination. takes 1 into k. for there seems to be no purpose served by the symbol x. X2.1. If ¢ E:: A(S) = Sn. while if. takes il into k. does to Xl (henceforth written as 1). Permutation Groups. Suppose that S is a finite set having n elements Xl. X4 --+ Xa. *10. what would the representation of (Jif..64 GROUP THEORY CH. q. e. While this notation is a little handier there still is waste in it. be? To compute it we could start and see what (Jif. prove: (a) G has a subgroup of order p and a subgroup of order q. ••• . One short cut might be to write ¢ out as where Xik is the image of Xi under ¢.2 a normal subgroup of order p and a normal subgroup of order q.1. if. q I p . 10. p > q are primes. then (Jif. in Sn. and we could write ¢ out by showing what it does to every element. and. prove that G is cyclic. a finite group G can be represented as a subgroup of Sn. (d) Any two non-abelian groups of order pq are isomorphic. there exists a non-abelian group of order pq. We could equally well represent the permutation as Our specific example would read as ( 1 2 3 4).

. so that a = (3) If a == eb.SEC. .by 1/. 3 ~ 3. SOl-I). The orbit of sunder 0 then consists of the elements s. s02. 3.. c se:'. then b = aOi.a induces a decomposition of S into disjoint subsets. If we know all the cycles of 0 we clearly know 0 since we would know the image of any element under e. (aei)Oi which This equivalence relation by Theorem 1. = aO.=G 23 3124 4)C 2 3 4) 1324 23 13 2 :} This is the way we shall multiply the symbols of the form . By a cycle of 0 we mean the ordered set (s. . b == eC. b E: S we define a == eb if and only if b = aOi for some integer i (i can be positive. SOl-I. (2) If a == eb. se. G 12 G 23 32 :) :} 01/. In particular... if e is the permutation represented by 23 and 1/.. thus the orbit of s under 0 consists of all the elements SOi. = b{)i whence b implies that a == eC. 4 ~ 4. the equivalence classes. n) '" kn Let S be a set and e E: A(S). then b = aOi. so k = 2 and ()-..J. For: (1) a == ea since a = aBo = ae.= 23 12 23 32 C 1/.. the representation for eif. ±1. For instance. takes 1 into 2. 10 PERMUTATION GROUPS 65 procedure for 2.J. sO. n. . . Similarly 1. ±2. We call the equivalence class of an element s E: S the orbit of sunder 0.:2 ~ then il = 3 and takes 3 into 2. = == ea. if S is a finite set and s E: S..+j. there is a smallest positive integer l = l(s) depending on s such that SOl = s. We claim this defines an equivalence relation on S. namely. se2. . or 0). is C If we write 3 2 213 :} :) :} C = o=G and then ()-.. Given two elements a. That is.. ••• . negative. i = 0.

18. ~ is the product of its cycles. (9) he will obtain 8.. 8. 104 = 88 = 5. etc. 48 = 5. 3. (3). . 8.. We digress for a moment.6). 4. (1.. 4). that of 4 consists of the elements 4. Let 8 = (1 213 2 3 4 5 6) 5 6 4 where S consists of the elements 1.. (7). 6 2 3 2 3 4 3 1 3 8 1 8) 4 5 5 5 6 6 6 6 7 7 the permutations =G =G 8 8 4 4 1 :)G :)- 45 6 7 8 5 1647 :) 7.3. 106 = 68 = 4.6. and ir into ill and leaves all other elements of S fixed. 182 = 28 = 1. Suppose that by the cycle (iI. The orbits of 7 and 9 can be found to be {7}.. 5. Thus. so the orbit of 1 is the set of elements 1 and 2. 2. then the orbit of 1 consists of 1 = 18°. ••• .. 2. (9). . at least in this case. (4. . 2. 182. 5. . 2.2). if S consists of the elements 1. then the symbol (1. 6. 4}. 1.5. {9}.66 GROUP THEORY CH. This tells us the orbit of 2 is the same set. for instance.).4). ••• . Every permutation is the product of its cycles. But this is no accident for it is now trivial to prove LEMMA 2. given the permutation preceding the last one. leaving our particular 8. That is. and 8 = (1 2 3 4 5 6 7 8 9). iT_1 into i. 2 Before proceeding we illustrate these ideas with an example. Starting with 1. 48a = 68 = 4. if S has 9 elements. 2. 238 164759 what are the cycles of 8? We first find the orbit of 1. 482 = 58 = 6. Thus 2 2 3 3 We multiply cycles by multiplying again. That is. 9. The cycles of 8 are (1. 8 7 5 4 Let us return to the ideas of the paragraph ask. 187 = 48 = 1. iT) we mean the permutation if. 185 = 58 = 6. 6 (remember 1 stands for Xli 2 for X2. 3. The reader should now verify that if he takes the product (as defined in the last paragraph) of (1.8. The cycles of 0 thus are (7). 1(6) = (1. 103 = 38 = 8. which sends i1 into i2.5. 182 = 28 = 3. 3. 18 = 2. 2. namely. The orbit of 3 consists just of 3. 6. . i2. 181 = 2. 6) means the permutation G (1 2 3) (5 2 6 3 4 4 5 6 7 7 8 8 25 1 :)they represent.22. respectively. i2 into ia . the orbit of 1 is the set {I.

SEC. since every permutation is a product of disjoint cycles and every cycle is a product of 2-cycles.22 is usually stated in the form: every permutation can be uniquely expressed as a product of disjoint cycles. xn) It is clear that O:P(XI. 0 . xs) = (Xl - = (134)(25) takes Xa) X5)(Xa . So 0. 3) = (1. We shall refer to 2-cycles as transpositions. sO.. Frankly. Then its cycles are of the form (s.. of all the distinct cycles of e. 3) = (3. is the same as the image of s' under the product.X4) (Xa . .. .<j If 0 i<j II (Xi ±p(XI. aa) (aI' am). Lemma 2. 2.. In doing so the lemma itself will become obvious. A simple computation shows that (1. . and since the cycles of 0 are disjoint.. the image of s' E:: Sunder 0. find its cycles. If the remarks above are still not transparent at this point.. m) = (1. For instance. E:: Sn let e act on the polynomial P(XI' Xi) -+ '<3 II (XO(i) - XO(J)' . the reader should take a given permutation. . m). hence e = if. We first want to show that this cannot happen.• . 1) (3.• .XS) (X2 X4) (X2 - XS) (Xa - X4) (Xa - Xl) (Xa - X2) (X5 X2) X4) (XS (X4 - X (XS - Xl) (X4 - which can easily be verified to be -P(XI. 3) (1. . 2).X4) (Xl . permutation 0 E:: Sn is said to be an even permutation if A it can be represented as a product of an even number of transpositions.. a2. we are not happy with the proof we give of this fact for it introduces a polynomial which seems extraneous to the matter at hand. More generally the m-cycle (aI. xn) bye: P(Xl. take their product. By the multiplication of cycles.. 2. have the same effect on every element of S. . which is what we sought to prove. 2) (1. Let 0 be the permutation.. xn) = -+ . Consider the m-cycle (1. 2. as defined above. m).. if.am) = (ab a2) (aI. Consider the polynomial in n-variables P(Xb . if. Every permutation is a product of 2-cycles. Perhaps it has other representations as a product of an odd number of transpositions. which is s'O. we have proved LEMMA 2. DEFINITION. and verify the lemma.. 10 PERMUTATION GROUPS 67 Proof. For instance.X5) (X4 . xs).23.Xj). The definition given just insists that e have one representation as a product of an even number of transpositions. . Xn). . . Now.X2) p(Xl.. . 2) (1. in Ss. xn) = II(Xi . into (Xa - X2) (Xl X (X2 - Xa) (Xl . .•• . " SOl-I). (1.XS) Xl) X2) (Xl . This decomposition is not unique j by this we mean that an m-cycle can be written as a product of 2-cycles in more than one way.

o(Sn) o(An) we see that o(An) = !n!. 2 If. Perhaps the best way of seeing this is as follows: let W be the group of real numbers 1and -1 under multiplication. in any representation it is a product of an even number of transpositions.. . S. f(s) = -1 if s is an odd permutation. being the kernel of a homomorphism An is a normal subgroup of Sn. is a homomorphism onto W. consistmg of all even permuiatume. in particular. We call a permutation odd if it is not an even permutation. We summarize our remarks in LEMMA 2. The following facts are now clear: (1) The product of two even permutations is an even permutation. An must be a subgroup of Sn. . 234 5 1 679 8 (b) G : : : ~ ~} . 2.. and 3. . xn) fixed. At the end of the next section we shall return to Sn again. The kernel of if.. so that any representation of II as a product of transposition must be such that it leaves P(Xl' . This is not a coincidence since this latter rule is used in establishing 1. The rule for combining even and odd permutations is like that of combining even and odd numbers under addition. xn) fixed. Snl An ~ W. 2.68 GROUP THEORY CH. since 2 = o(W) = 0 ( Sn) - An = --. PROBLEMS 1. An. ():p(Xl.. Since the product of two even permutations is even. We claim it is normal in Sn. has as a normal subgroup of index 2 the alternating group. .24. xn) ~ -P(Xl. (Verify!) Thus if a permutation II can be represented as a product of an even number of transpositions in one representation. so.d. ()is a transposition. (2) The product of an even permutation and an odd one is odd (likewise for the product of an odd and even permutation). By Theorem 2. An is called the alternating group of degree n. Let An be the subset of S« consisting of all even permutations. xn)... By the rules 1. This establishes that the definition given for an even permutation is a significant one. .. (3) The product of two odd permutations is an even permutation. II must leave P(Xl' . that is. Define f:Sn ~ W by f(s) = 1 if sis an even permutation.. Find the orbits and cycles of the following permutations: (a) (1 2 3 4 5 6 7 8 9). 3 above if. is precisely An. .

find a permutation a such that a-Ixa = y. 2.2).4).2. In this great variety one of the most basic tools is counting. Prove that if a normal subgroup of An contains even a single 3-cycle it must be all of An.5)(1.2)(3.2) b = (1.3)(5. . Mathematics is rich in technique and arguments. Determine which of the following are even permutations: (a) (1. 8. (c) Prove that there is no permutation a such that a-I(1. . *13.5).2. n) is Sn.7.4)(1.2. 9) b = (1. " . 2)a = (3.3)(4. 2). Prove that for n :2: 3 the subgroup generated by the 3-cycles is An. 2) and (1.. Thus in various phases of mathematics we find neat counting devices which tell us exactly how many elements. 8). (b) (1. m2. 3) (1. *14. y = (5. . Express as the product of disjoint cycles: (a) (1.5)(1. deft.5). Write the permutations in Problem 1 as the product of disjoint cycles. strangely enough.5)(1. Yet.5).6. Prove that (1. This can sometimes be done by a brute force case-by-case exhaustion but such a routine is invariably dull and violates a mathematician's sense of aesthetics. But the most serious objection to case-by-case division is that it works far too rarely. mk? (c) How do you find the order of a given permutation? 7. delicate touch to the hammer blow. 11. Determine for what m an m-cycle is an even permutation.3.3)(1. (In other words. Assuming the result of Problem 14.9). (b) Prove that there is no a such that a-l(l.3).4. 9.4)(2. by counting we do not mean the creation of tables of logarithms or addition tables. Prove that A5 has no normal subgroups N . 2. As. 2.:e (e). 10. (c) (1. (b) (1. . n)-l = (n.2. 11.8). (a) Given the permutation x = (1.) *12.2. 15. 3. One prefers the light. n . Find the cycle structure of all the powers of (1. 6..2)(1. 7. it is one of the most difficult. n . Another Counting Principle. 11 ANOTHER COUNTING PRINCIPLE 69 2. (a) What is the order of an n-cycle? (b) What is the order of the product of the disjoint cycles of lengths ml. 3)a = (1.1.SEC. 2.2)(1.2.2.3)(4. . we mean the process of precisely accounting for all possibilities in highly complex situations.1). 5. 4. Compute a-Iba where (1) a = (1.8. ..5. prove that any subgroup of A5 has order at most 12. these generate Sn.5).. 3.. Prove that the smallest subgroup of Sn containing (1.9)(1.2.7.6) (1. (2) a = (5. rather. 3). Of course. .3)(1.7.

a . the comparison of the two counts is then used as a means of drawing conclusions. is usually called the conjugate class of a in G. C(a). c = y-1by for some x. C in G. a (3) Suppose that at'. Before doing so. follows. we must prove that (1) a a. b E: G. and shall refer to this relation as conjugacy. a c is a consequence. of course. c C G.' band b "-' c where a. (2) If a b. and then find a neat algebraic description for the size of each equivalence class. Proof. with c = e serving as the c in the definition of conjugacy. From this simple description there will flow a stream of beautiful and powerful results about finite groups. We shall write. one introduces an equivalence relation on a finite set.. b. If a. This remark is. DEFINITION. "-I "-I "-I For a E: G let C(a) = (x E: Gla z}. measures the size of the equivalence classes under this relation.. (1) Since a = e-1ae. A great favorite with mathematicians is the process of counting up a given situation in two different ways. Conjugacy is an equivalence relation on G. satisfy certain conditions.25. Our attention now narrows to the case in which G is a finite group. since xy C G. Then b = x-lax. As usual. then b = x-lax for some xC G.70 GROUP THEORY CH.. for this.. Substituting for b in the expression for c we obtain c = y-1(x-1ax)y = (xy)-la(xy). merely a restatement of the fact that our equivalence relation-con"-I .. We shall introduce a relation. y C G. prove it is an equivalence relation.. Suppose that C(a) has Ca elements.. We seek an alternative description of Ca. Generally speaking. a a. b LEMMA 2._. b. (3) a b. We prove each of these in turn. note that o(G) = ~ca where the sum runs over a set of a E: G using one a from each conjugate class. and then equates the number of elements in the set to the sum of the orders of these equivalence classes. b c implies that a "-I "-I "-I "-I "-I C for all a. b . it consists of the set of all distinct elements of the form y-1ay as y ranges over G. hence. in order to establish this.. then b is said to be a conjugate of a in G if there exists an element c C G such that b = c-1ac. (2) a r-« b implies that b a.. the equivalence class of a in G under our relation. a = (X-1)-lb(x-1) and since y = x-1 C G and a = y-1by.2 in some fairly broad context.. This kind of an approach will be illustrated in this section.

If G is a finite group. y E: N(a). DEFINITION. Our method of proof will be to show that two elements in the same right coset of N(a) in G yield the same conjugate of a whereas two elements in different right cosets of N(a) in G give rise to different conjugates of a. y (: G are in the same right coset of N(a) in G. To begin with. In this way we shall have a one-to-one correspondence between conjugates of a and right cosets of N(a). the conjugate class of a in G. 11 ANOTHER COUNTING PRINCIPLE 71 [ugacy+-induces a decomposition of G into disjoint equivalence classes-the conjugate classes. from x-lax = y-Iay we would deduce that yx-1a = ayx-1i this in turn would imply that .26. so that X-I is also in N(a). x and yare in different right cosets of N(a) in G we claim that x-lax r5. THEOREM 2. We are now in a position to enunciate our counting principle. Therefore. From ax = xa it follows that x-Ia = x-l(ax)x-l = l x-l(xa)x= ax-I. Proof. then N(a). Suppose that x. Were this not the case. N(a) is a subgroup of G. the number of elements conjugate to a in G is the index of the normalizer of a in G. C(a). In this result the order of G. in other words. consists exactly of all the elements x-lax as x ranges over G. and so na = an. whence x and y result in the same conjugate of a.. Ca measures the number of distinct x-Iax's.H. In order to carry this out we recall a concept introduced in Problem 12.SEC. on the other hand. v-lay = x-In-Ianx = x-In-Inax = x-lax.v-lay. Since this concept is important-far too important to leave to the off-chance that the student solved the particular problem-we go over what may very well be familiar ground to many of the readers. is of no relevance. Thus xa = ax and ya = ay.If a (: G. Therefore. N(a) consists of precisely those elements in G which commute with a. Of paramount interest now is an evaluation of Ca. is the set N(a) = Ix (: Glxa = ax}. since v= = (nx)-l = x-In-I. Proof. If. the normalizer of a in G. But then N(a) has been demonstrated to be a subgroup of G. Suppose that z. (xy)a = x(ya) = x(ay) = (xa)y = (ax)y = a(xy). whether it be finite or infinite. Section 5. LEMMA 2. then Ca = o(~~l» . Thus y = nx where n E: N(a). in consequence of which xy (: N(a). and so we put no restrictions on the order of G.

3) = {(I. 2)} = {(I. (1. 3).3).2)}. the group S3. (1.2 C. 2» = [e.3. Proof. 3. N(a) = G.3). (1.3) -1(1. (1.2).2) = {e} = 1(1. 2. We enumerate the conjugate classes: O(e) 0(1. If.72 yx-l GROUP THEORY CH. (2.2.h lends itself to immediate and powerful application.3. Proof.3. G such that ax = xa for all x C. If G is finite.H. There is no point in looking at abelian groups because there two elements are conjugate if and only if they are equal (that is. 2). (1. G. 3). 3)} (Verify!) C(l. Its elements are e. (1. 2) (2. N(a).2. o(G) = L: o(N(a» o (G) where this sum runs over one element a in each con. (2. ApPLICATIONS THEOREM OF 2. 2) (1. so that C(1. So we turn to our familiar friend. this declares x and y to be in the same right coset of N(a) in G.3). Note the SUBLEMMA. c. E: Z if and only if N(a) = G. We need no artificial constructs to illustrate its use for the results below which reveal the strength of the theorem are themselves theorems of stature and importance.2) I and N((l. 2.2. (1. (1. using the theorem the corollary becomes The equation in this corollary is usually referred to as the class equation of G. 3» {e. The student should verify that N((l.3). 2. COROLLARY.2. (1. = ~ca.3. a C. xa = ax for all xc. G.3)-1(1. Theorem 2.3). o(N(a» = o(G) is equivalent to N(a) = G.3) = t = 2. contradicting the fact that they are in different cosets. G. C(1. Z. (1. so that a E: Z.2) = ! = 3.2)-1(1. whence N(a) = G. (2.2)} (after another verification). = 1 for every a). conversely. Before going on to the applications of these results let us examine these concepts in some specific group. (1. The proof is now complete. (1. If G is finite. Since o(G) immediate. 2) (1. xa = ax for all x C. 2). 2). 2. .3) -1(1. 3).2)(1. However. If a C. 3). Let us recall that the center ZeG) of a group G is the set of all a C.iugate class. Z if and a only if o(N(a» = 0(G).

so that o(N(a)) > p yet by Lagrange's theorem o(N(a)) I 0(G) = p2. Therefore. Suppose that o(Z(G)) = p. Proof. We now use Theorem 2. COROLLARY. let a ~ G. a ~ N(a). then G is abelian. Our aim is to show that Z(G) = G.h to prove an important theorem 2 due to Cauchy. then Z(G) ~ (e). Rephrasing. Z(G) c N(a). Write out the class equation for this G. If o(Z(G)) = p2. we find that Now look at this! p is a divisor of the left-hand side. implying that a ~ Z(G). Thus o(Z(G)) = p is not an actual possibility. then Z(G) = G and we are done. At any rate. letting z = o(Z(G)). at[ Z(G). Thus N(a) is a subgroup of G. ApPLICATION. we already know that Z(G) ~ (e) is a subgroup of G so that o(Z(G)) = p or p2.SEC. being a divisor of o(G) = pn must be of the form o(N(a)) = pna. Proof. besides e. the theorem states that a group of prime-power order must always have a nontrivial center. a contradiction. The only way out is for o(N(a)) = p2.(pn jpna). <n for each so that p is a divisor of each term of this sum. Since e ~ Z(G). since na term in the ~ of the right side. Therefore. o(G) If = p2 where p is a prime number. If a ~ G. in Z(G)! This is the contention of the theorem. since there are exactly z elements such that na = n. 11 ANOTHER COUNTING PRINCIPLE 73 ApPLICATION 1 THEOREM If o(G) = pn where p is a prime number. We can now simply prove. o(N(a)).1. as a corollary for this. a result given in an earlier problem. The reader may remember that this theorem was already proved for abelian groups as an application of the results developed in the . 2. We get pn = o(G) = "2:. thus z is a positive integer divisible by the prime p. p I (pn - na<n p a L p:) = Z. since N(a) is a subgroup of G. z > I! But then there must be an element. z ~ 0. hence a divisor of this sum. however. a ~ Z(G) if and only if na = n.

We seek an element a . [The reader should look up the charming. To continue our candor.. despite all these counter-arguments we present it here as a striking illustration of Theorem 2. we shall make use of this special case in the proof below.. page 119. In particular. Thus we may assume that p is not a divisor of the order of any proper subgroup of G. we conclude that pi (O(G) - L: N(a)r'G O(G») o (N(a) ) = o(Z(G».J (CAUCHY). we could prove it. Proof. in the very next section. we shall prove. Z(G) is thus a subgroup of G whose order is divisible by p. which has Cauchy's theorem as an immediate corollary.><G o(G) o(N(a» pi o(G).74 GROUP THEORY CH.><G o(N(a» Since plo(G). since N(a) ¢ G. We are forced to accept the only possibility left us. now we invoke the result already established for abelian groups to complete the induction. were Cauchy's theorem itself our ultimate and only goal. that is. In fact.h. But. then G has an element of order p. to be frank. that Z(G) = G. so that Z(G) cannot be a proper subgroup of G.. VoL 66 (1959). using the barest essentials of group theory in a few lines. and thus there would be such an element in G. namely. due to Sylow. . then by our induction hypothesis there would exist an element of order p in W.i N(a). we assume the theorem to be true for all groups T such that oCT) < o(G). THEOREM 2. after all. If for any subgroup W of G. This proves the theorem. Let us write down the class equation: '" o (G) o(G) = o(Z(G» + £. we have assumed that p is not a divisor of the order of any proper subgroup of G.r£ e E:: G satisfying aP = e. To prove its existence we proceed by induction on o(G).h. . if a ([_Z(G). a much stronger result. p t o(N(a» we have that p and so p since we also have that o(G) I o(N(a» . W .r£ G were it to happen that pi o(W). in a manner which completely avoids Theorem 2. We need not worry about starting the induction for the result is vacuously true for groups of order 1. If P is a prime number and p I o(G). one-paragraph proof of Cauchy's theorem found by McKay and published in the American Mathematical Monthly. 2 section on homomorphisms. But then G is abelian. .J Yet. I L: N(a). But. p t o(N(a».

-+ t... + nr.. To reach our goal we exhibit a very simple rule for computing the conjugate of a given permutation.6)(7) (8. For if = (al. . 2. X2.505. .. and 2 by 3. xnr) and T = (al. 2. note that 1 + 1 + 2 + 2 3 = 9. p(6) = 11. 3} . Once this is proved.5..6. 2 . 6 . then n = nl + n2 + .. 3 . 3 = 1 + 1 + 1 = 5 since 4 = 4. ~ nr constitute a partition.121. a2. 4 = 2 + 2.... 7 .3)(4. nr. For example.-+ t. ••. 4 by 7. to determine 8-11J'0 where () = (1. then e-10'() ends s . .. nr.. ~ nr. ••• . . nl ~ n2 ~ . 4 = 1 + 1 + 2. We now aim to prove that two permutations in Sn are conjugate if and only if they have the same cycle decomposition. How do we find ()-10'() where 0 E:: Sn? Suppose that () sends i .3). + nr. 3 = 1 + 2... n2. (32. With this algorithm for computing conjugates it becomes clear that two permutations having the same cycle decomposition are conjugate.. 4 .. . There is a large mathematical literature on pen). then S« will have exactly pen) conjugate classes. . Given the integer n we say the sequence of positive integers nl. namely.. .. n2.-+ [.SEC. 4 = 1 4 = + 3.. n1 ~ n2 ~ . nr} if it can be written as the product of disjoint cycles of lengths nl.. .. 7 by 4.. . anl) (b1. ()-10'() is obtained. b2.... 3)(4. n2.. X2. 1..7. Thus in S9 IJ' = 1 2 3 4 5 6 7 8 9) (1 3 2 5 6 4 7 9 8 = (1)(2.. . .. 1 + 1 + 1 + 1. of n if n = nl + n2 + .. .. 6 by 6.. from 0' by replacing in IJ' 5 by 5.-+ 4. the symmetric groups Sn...-+ 3.. Suppose that IJ' E:: Sn and that IJ' sends i .. so that o-lO'() = (5. . for if the cycles appearing have lengths nr.. Xnr) . n2. .-+ s and} . ••.6.... ~ nr.. . to compute 0-10'0 replace s every symbol in 0' by its image under o. (Xl. then.. Some others are p(5) = 7.2. bn2) ••• (Xl.. 2). 11 ANOTHER COUNTING PRINCIPLE 75 We conclude this section with a consideration of the conjugacy relation in a specific class of groups.. 7) and 0' = (5. 4. respectively.. p(61) = 1. since ():5 ~ 5.. .... Let pen) denote the number of partitions of n.. then T = ()-1(J() where one (J + ... a2.9) has cycle decomposition 11.. 3 by 1... Let us determine pen) for small values of n: pel) p(2) p(3) p(4) = 1 since 1 = 1 is the only partition of 1 = 2 since 2 = 2 and 2 = 1 + 1 = 3 since 3 = 3. Every time we break a given permutation in S« into a product of disjoint cycles we obtain a partition of n. (3n2) •• . anl) «(31...-+ 1.-+ 7. We shall say a permutation IJ' E:: Sn has the cycle decomposition Inr.4)(1..-+ 6. n] ~ n2 ~ .. In other words.. nr..7)(3.

thus the general element (J commuting with (1. . n(n . n! u= =n. . 8) and (7. 8) can be exhibited as conjugates by using the conjugating permutation ( 1 2 3 4 5 6 7 8) . . The number of conjugate classes in Sn is pen). n) and there are (n . 2. 2. If the order of the normalizer of (1.I)! ..2)!' That is.2 could use as fJ the permutation a2 a2 ant anl c al b1 {3l i: {3n2 Xl xnr).. for instance. We restate the result proved in the previous discussion as LEMMA 2. 7 5 1 3 6 248 That two conjugates have the same cycle decomposition is now trivial for.. Thus if u denotes the order of the normalizer of (1. Thus the conjugate class of (1. (n . 2) has in it n(n . Xnr Xl Thus. ...2)! elements which commute with (1.I)!. .2)! such. 2) is r. 2)iT where i = 0 or 1. 2) commutes with itself. . the order of the normalizer of (1. 2) and the (n . n) in S« = (n .76 GROUP THEORY CH. n) E:: Sn. 3. Given the permutation (1. There are (n . Certainly it does commute with all its powers.2)! elements in the group generated by (1. 4.1)/2 elements. T is a permutation leaving both 1 and 2 fixed. We illustrate this with a very special and simple case. since o(Sn)/u = number of conjugates of (1. 6) (2.. 3. As another application consider the permutation (1. Also (1. 2) is 2(n .I)! distinct n-cycles in Sn. This way we get 2(n . and this gives rise to n elements. We claim this element commutes only with its powers. 2) is (J = (1. . by our counting principle. 2.1) 2 o(Sn) r n! r Thus r = 2(n .1)/2 transpositions and these are precisely all the conjugates of (1. 7. 2) in Sn.2)! permutations leaving 1 ~d 2 fixed. replace every element in a given cycle by its image under the conjugating permutation. 5) (1. by our rule. Are there others? There are n(n . what elements commute with it? Certainly any permutation leaving both 1 and 2 fixed does. Now any n-cycle is conjugate to (1. 4. n) in Sn. 5) (6. 2). the number of partitions of n.2)!' But we exhibited 2(n . 2). 2..27. .. Since we have such an explicit description of the conjugate classes in Sn we can :find all the elements commuting with a given permutation. (1. then. to compute a conjugate. 2) (3.

. 2)(3. . = (e) where Ni is a normal subgroup of Ni-1 and where Ni_dNi is abelian. 9. is normal in G.SEC. p a prime number. Prove that any subgroup of order pn-l in a group G of order pn. (b) Find the form of all elements commuting with (1. 7' is a permutation leaving all of 1. r)i7' where i = 0. prove that there are ~ (n! r n .. Prove that a group of order 28 has a normal subgroup of order 7. prove that G has a normal subgroup N ~ (e). p a prime number. 2. . G. which commutes with (1. 4. If p is a prime number. the alternating group of degree 5. r fixed. and if GjZ is cyclic. . (c) Prove that any element a in S. 6. (b) Prove that o(N) = ~ca for some choices of a in N. show that in Sp there are (p . x ({_H such that x-1Hx = H. If o(G) = v". n) having given us n such elements. .r)! distinct r cycles.. (a) Find the number of conjugates of (1. which are conjugate in S5 but not in A5• (b) Find all the conjugate classes in A5 and the number of elements in each conjugate class. . prove that in A5 there is no normal subgroup N other than (e) and A5• 7. 1. . p a prime number.. p a prime number. p a prime number. PROBLEMS 1. find the number of conjugates that the r-eycle (1.i as a tool. (a) Find two elements in As. . 2.. and H ~ G is a subgroup of G. . 10. 2. show that there exists an x E:: G. 4) in Sn. The powers of (1. *11. r. ::)N.2. . + ° :::. Using Theorem 2.. prove that N n Z ~ (e) where Z is the center of G. .. prove that G must be abelian.. 2. 13.. Z its center. 14. 2) (3. 6. 11 ANOTHER COUNTING PRINCIPLE 77 So the order of the normalizer of (1. prove that there exist subgroups Ni.1)! 1 elements x satisfying xP = e. ... there is no room left for others and we have proved our contention. 4) in S«. n 8. prove that if o(G) = pn. . and if N .. If G is a group.. i = 0. 3. show that every conjugate of a in G is also in N. (a) In S. If in a finite group G an element a has exactly two conjugates.. .. r (for some r) such that G = No::) Nl ::) N2::) . (a) If N is a normal subgroup of G and a E:: N. If o(G) = v". r) has in Sn.2. Prove that any group of order 15 is cyclic. 12. . n) in Sn is n. If o(G) = pn. 1. . (b) Using this. 2.= (e) is a normal subgroup of G. then G has a subgroup of order pC< for all a :::. r) is of the form a = (1. 2. n ~ 4. (c) Using this and the result for Problem 5(b).

i.+l t (Pam) pa . subgroup of a finite group is a divisor of the order of that group. Proof of the Theorem. The converse. Sylow's Theorem. then G has a subgroup of order pol.K (SYLow). one can see that except for the term m in the numerator.pol + 1) (pa _ pol + 1) Looking at this number. Thus mr has (Pam) elements.78 GROUP THEORY CH. is a classic theorem due to the Norwegian mathematician Sylow. is false. If P is a prime number and pal o(G).1) '" (pam . 12. VoL 10 (1959).m: be the set of all subsets of G which have pol elements..m: (M 1 is a subpol . THEOREM 2. consider Pam) ( pol = (pam) ! (pol) !(pam _ pol)! pOlm(pOlm . the power of p dividing (pOlm . the proof in the form we give it is due to Wielandt and appeared in the journal Archiv der M atematik. Before entering the proof of the theorem proper we digress slightly to a brief number-theoretic and combinatorial discussion. The question is: what power of p divides (::m)? written out as we have written it out. Let .2 15. Thus p' I( Pam) pol ' p. (pIX _ i) (pam . Prove that if a group G of order 28 has a normal subgroup of order 4. and widely used.. There are very few theorems which assert the existence of subgroups of prescribed order in arbitrary finite groups. then G is abelian. The number of ways of picking a subset of k elements from a set of n elements can easily be shown to be (~) = k!(n ~ k)! If n = pam where p is a prime number. Given M 11 M 2 E: . We present here a very elegant and elementary proof of Sylow's theorem.i) pOl(pa _ 1) . however. so all powers of p cancel out except the power which divides m. Lagrange's theorem tells us that the order of a.i) is the same as that dividing pa . pages 401--402. and if pr I m but pr+l t m. The most basic.

•. . n. n. Combined with o (H) ~ v" we have that o(H) = p". then for all hE:: H mih E::MI. If v" I o(G).m where pT+l t n. b E::H. it actually has done more-it has constructed the required subgroup before our very eyes! What is usually known as Sylow's theorem is a special case of Theorem 2. Thus p'" ~ o(H).k. But then we have exhibited a subgroup of G having exactly p" elements. and so o(H) ~ p". Let 1M 1. However. 2. namely H.elements. 12 SYLOW'S THEOREM 79 set of G having pC<elements. By our very definition of equivalence in mt. if ml E::MI. We shall not go into a proof of these here. 84.g = MJ for some i. . This proves the theorem.m. M. if g E::G. pm+l t o(G).. We let H = 19 E::G 1M Ig = Md· Clearly H is a subgroup of G. In the supplementary problems. We claim that there is at least one equivalence class of elements in mt such that the number of elements in this class is not a multiple of pT+l. . It is immediate to verify that this defines an equivalence relation on mt. find a 2-Sylow subgroup and a 3-Sylow subgroup. but suggest to him the argument used in the counting principle in Section 11. a combination of the problems 21-24 yields their proofs. namely the I COROLLA. for if a. we leave the proof to the reader. it must follow that p'" I o(H). PROBLEMS 1. then MIa = MI.. . However. where pmlo(G). j s:. We shall be vitally concerned with o(H). then pr+l would be a divisor of the number m of elements in mt. since pT+I t nand pcx+rl p"'m = no(H). We claim that no (H) = o(G).J cannot be the case. In the symmetric group of degree 4. Now no(H) = o(G) = pC!. for if pr+l is a divisor of the size of each equivalence class. M1b = MI whence M1ab = (MIa)b = M1b = MI. and likewise so is M2) define Ml M2 if there exists an element g E::G such that Ml = M2g. Ml was a subset of G containing pC!. Prove that a group of order 108 must have a normal subgroup of order 9 or 27.SEC. then G has a subgroup of order pm. is called a p-8ylow subgroup of G.RY.m has (Pcx ) elements and pT+l t (pam) . Thus Ml has at least o(H) distinct elements. for each i = 1. 1 s:. Since . pm+l t o(G). M n I be such an equivalence class in . A subgroup of G of order pm. . The usual Sylow theorem then has two other parts to it: one of these asserts that any two p-Sylow subgroups of G are conjugate in G and the other states that the number of p-Sylow subgroups of G is of the form 1 + kp.this pcx pC< r-.

2 (c) If q E:: Q. If p is an odd prime and if 1 + t + t + . 10. Prove that a subgroup of a solvable group is solvable.. an is an element whose square is the identity. y. . then nln2 is a residue. 2 SUPPLEMENTARY PROBLEMS There is no relation between the order in which the problems appear and the order of appearance of the sections. (b) If the G in part (a) has no element of order 2 or more than one element of order 2. Prove that the homomorphic image of a solvable group is solvable.. 3. b are integers. prove that G is solvable. then ala2 . a2.. in this chapter. 7. Prove that xax = b is solvable for z in G if and only if ab is the square of some element in G.. No hint is given regarding the difficulty of any problem. (d) (Wilson's theorem). Show that the equation x2ax = a-I is solvable for x in a group G if and only if a is the cube of some element in G. alb DEFINITION. (e) If a is a quadratic residue of p then a(p-l)/2 == 1 mod p. which might be relevant to their solution. If p is an odd prime number.I)! == -lep). = 2. . then (p . Suppose G is a group such that ¢(x) = xn defines an automorphism of G.1) where a.. 2. 12.. prove that ala2 . . Prove: (a) The quadratic residues mod p form a subgroup Q of the group of nonzero integers mod p under multiplication. Find all the automorphisms of S3. prove that pia. Prove that for all a E:: G. (d) If nb n2 are nonresidues. an-l is in the center of G. a ~ 0 is said to be a quadratic residue of p if there exists an integer x such that x2 == a mod p. 6. such that M ~ (e). (b) O(Q) = p 1. If G is a group and N is a normal subgroup of G such that both N and GIN are solvable. an = y. If p is a prime number.80 GROUP THEORY CH. an. (c) If G has one element. prove that ala2 . 9... 3) is not the cube of any element in Sn. 1. If p > 3. then nq is a nonresidue. the symmetric group of degree 3. n ([_ Q (n is called a nonresidue). 8. . 4.. is normal in Ni-l and Ni_dNi is abelian.. 11. ::J Nr = (e) such that N.. prove that p2[ a. of order 2. Prove that (1. an = e. A group G is said to be solvable if there exist subgroups G = No ::J Nl ::J N2 ::J . Prove that a solvable group always has an abelian normal subgroup M. (a) If G is a finite abelian group with elements aI. 5.+ 1/(p .

12 SYLOW'S THEOREM: 81 13. define A to be conjugate to B if B = x-lAx for some x E: G. 18.) 23. prove that the conjugate class of Sp has 1 kp distinct subgroups. (a) If G is a finite group and if Sp is a p-Sylow subgroup of G of order v". is a multiple of n. + . prove that Sp is the only subgroup of G of order pm lying in N(Sp). then a E: (c) Prove that N(N(Sp)) = N(Sp). Same as Problem 17. For A. (b) If Sp is a p-Sylow subgroup of G and if a. (b) If N(A) = {x E: Glx-1Ax = A} then there is a one-to-one correspondence between the right-co sets of N(A) in G and the distinct conjugates of A. 22. 15. A a subgroup of G and if N is a normal subgroup of G. n. 19. is in N(Sp). there are at most n solutions of xn == 1 mod p for every integer n. (b) Breaking up the conjugate class of Sp further by using conjugacy relative to Sp. 16.SEC. Prove that in the integers mod p. of order pk. prove that there is an element c in G whose order is the least common multiple of m and n. (a) If G is a finite group and if Sp is a p-Sylow subgroup of G prove that the number of conjugates of Sp in G is not a multiple of p. 17. Give an example of a non-abelian group in which (xy)3 = X3y3 for all x and y. (a) If Sp is a p-Sylow subgroup and B a subgroup of order pk in G. If G is an abelian group and if a. 20. p a prime number. (b) The number of subgroups of G conjugate to A relative to H equals the index of N(A) n H in H. (This is the second part of Sylow's theorem. Prove: (a) This defines an equivalence relation on the set of subgroups of G. but do not assume the group to be abelian. Prove: (a) The relation of being conjugate is an equivalence relation on the set of subgroups of G. (Hint: Use part (b) of Problem 20 and Problem 21. (c) Prove that any two p-Sylow subgroups of G (for the same prime p) are conjugate. where n I o(G). prove that if both A and N are solvable then so is AN. Let G be a finite group and H a subgroup of G. If G is a group. If A and B are subgroups of G. prove that B must be contained in some conjugate of Sp. 21. prove that if B is not contained in some conjugate of Sp the number of conjugates in G of Sp is a multiple of p. In the finite abelian group G prove that the number of solutions of xn = 8. respectively. (b) Using part (a) and Problem 22.) s. B subgroups of G define A to be conjugate to B relative to H if B = x-lAx for some z E: H. 14. bin G have orders m.

31. prove that a group of order pq is solvable. prove that np. Supplementary Reading BURNSIDE. (b) If q t (p . The Macmillan Company. Theory of Groups.1.. 29. 1911.82 GROUP THEORY CH. page 119. 27. Inc. Making a case-by-case discussion.. prove that the 2-Sylow subgroup of G must be normal in G. JAMES "Another proof of Cauchy's group theorem. (a) Using Problem 24 prove that if o(G) = 36 then G has either 1 or 4 3-Sylow subgroups. Mathematical Monthly. Second Edition.2 24. "The automorphisms of the symmetric group.) 25. SEGAL. . is of the form np = 1 + kp." Bulletin of the American Mathematical Society.. page 565.. Cambridge. CamW bridge University Press. Using Problems 22 and 2:3. Show that a group cannot be written as the set-theoretic union of two proper subgroups. Theory of Groups of Finite Order. 1959. E.1) prove that a group of order pq is cyclic. Vol. If a group G has a proper subgroup of finite-index." American H. 1955. prove it has a normal subgroup of finite-index. HALL.MARSHALL. prove that a group of order less than 60 either is of prime order or has a nontrivial normal subgroup. Topics for Class Discussion McKAy. If o(G) = 42. (This is the third part of Sylow's theorem. 46 (1940). Prove that a group of order 48 must have a normal subgroup of order 8 or 16. If G is a group and if a E::: G has finite order and only a finite number of conjugates in G prove that these conjugates generate a finite normal subgroup of G. (b) If o(G) = 56 prove that G has 1 or 8 7-Sylow subgroups. prove its 7-Sylow subgroup is a normal subgroup. 30. 32. (a) If p > q are distinct primes. 28. 66 (1959). Vol. . using the results developed in the chapter. the number of p-Sylow subgroups in G. England. New York. (c) Prove that any two non-abelian groups of order pq are isomorphic. Dover Publications. New York. 26. In the latter case.

the analysis of rings will follow the pattern already laid out for groups. of a set onto itself. (5) There exists an element -a in R such that a + (-a) = O. (2) a b=b a. At this stage of the development we have learned something about one of these. using the tools developed here. b.CHAPTER 3 Ring Theory 1. these operations are usually called addition and multiplication. etc. normal subgroups. Definition and Examples of Rings. In contrast. namely rings. rings stem from another. To cite merely one instance. 83 (1) a b is in R. We shall require the appropriate analogs of homomorphism. factor groups.(b + c) = a-b + a-c and (b + c)·a = b-o. intertwine them with meaningful theorems. We shall see that they are patterned after. It is our purpose now to introduce and to study a. As we indicated in Chapter 2 there are certain algebraic systems which serve as the building blocks for the structures comprising the subject which is today called modern algebra. and are generalizations of. A nonempty set R is said to be an associative ring if in R there are defined two operations. (6) a-b is in R. later on in the book. the set of integers. denoted by + and . second such. we shall prove that it is impossible to trisect an angle of 60° using only a straight-edge and compass. the algebraic aspects of the ordinary integers. and end up proving results which are both interesting and important about mathematical objects with which we have had long acquaintance. (3) (a + b) + e = a + (b c). (7) a·(b·c) = (a·b)·c. (4) There is an element 0 in R such that a + + + + . and more familiar source. + c-c (the two distributive laws). The abstract concept of a group has its origins in the set of mappings. DEFINITION. Yet. c in R: + 0 = a (for every a in R). despite the differences. or permutations. respectively such that for all a. namely groups. With the experience gained in our study of groups we shall be able to make the requisite definitions. (8) a. In the next paragraph it will become clear that a ring is quite different from a group in that it is a two-operational system.

. The student should verify that R is a commutative ring with unit element. numbers under the usual addition and R is a commutative ring with unit note that the elements of R different multiplication. leaves a remainder of 2). . 4. those in which axiom 7 may fail to hold do occur in mathematics and are studied. R is the set of rational multiplication of rational numbers. N onassociative rings.84 RING THEORY CH. the usual multiplication of integers. EXAMPLE 1. or not happen. Axioms (6) and (7) insist that R be closed under an associative operation. That is. 5·3 = I since 5·3 = 15 has 1 as a remainder on division by 7). Whenever we speak of ring it will be understood we mean associative ring. EXAMPLE 2. we pause to examine some examples.. which we call multiplication. when divided by 7. that there is an element 1 in R such that a·1 = 1·a = a for every a in R. R is the set of integers mod 7 under the addition and multiplication mod 7. element. R is the set of even integers under the usual operations of addition and multiplication. It may very well happen. Before going on to work out some properties of rings. R is a commutative ring with unit element. A ring with this latter EXAMPLE 4. 6 where: (1) 1. If the multiplication of R is such that a. for instance. EXAMPLE 3. R is a commutative ring but has no unit element. from 0 form an abelian group under property is called a field.3 Axioms (1) through (5) merely state that R is an abelian group under the operation + which we call addition. which. R is the set of integers. that is. but we shall have no occasion to consider them. I.b = b. 5. Motivated by these examples we shall define various special types of rings which are of importance. (2) i·J = iii where m is the remainder of ij on division by 7 (thus. + J = k where k is the remainder of i + j on division by 7 (thus. Axiom (8) serves to interrelate the two operations of R. the elements of R are the seven symbols 0. + is the usual addition and . negative. But even more than that. positive. 4 + 5 = 2 since 4 + 5 = 9. b in R then we call R a commutative ring. 3. 2.a for every a. and 0. if there is such we shall describe R as a ring with unit element.

looks rather complicated. L: 2 {3ijeij if and only if for all i.:j=l L: 222 a.. since: 1·1 = I = 6·6 2·4 = I = 4·2 3·5=1=5·3 the nonzero elements of R form an abelian group under multiplication. We now present a noncommutative ring. j (2) = 1. If we denote the elements in R by 0'. first seen. 5. R is thus a field. one sees that 2·3 = 0'. 2. This cannot happen in a field (see Problem 10.SEC. (3) where "Iii = v=l L: ai. I.j=l = i. namely. EXAMPLE R is the set of integers mod 6 under addition and multiplica5. 1 DEFINITION ANDEXAMPLES F RINGS 85 O However. much more can be shown.ieij + L: {3ijeij i.j=l L: 2 aijeij = i. simple rules. = 0 nor b = O.j. e. multiplying using the relations eiJ" ekZ = collecting. tion mod 6. aij = i.j' eiZ = ea in this term by term . end of Section 2). . Every example given so far has been a commutative ring.j=l L: (aij + {3ii)eij.. This multiplication. However. multiply "2aijeij by out term by term. EXAMPLE R will be the set of all symbols 6. . namely.j=l (3. Thus it is possible in a ring R that a-b = 0 with neither 0. when it is founded on relatively "2{3iieii formally.{3vj 2 = ai1{3lj + ai2{32j. thus the ring R in Example 5 is certainly not a field. yet 2 ~ 0' and 3 ~ 0'. and 0 for j ~ k. a22e22 a11811 = L: aije' 2 + a12812 + a21e21 + j where all the aij are rational numbers and where we decree: (1) i. 2. Since it only has a finite number of elements it is called a finite field. and collecting terms.

in order to make of C a ring we still need a multiplication.)" 8) In C we introduce an addition by defining for x (2) x = (a. We define: (1) (a. (3) where a. (3). ao + (3'Y). and its relatives. Y = ('Y. (3 + 0). 0) = (1.86 RING THEORY CH. (3 are real numbers. since (3 are real and not both 0. We assert that C is an abelian group under this operation with (0. 0) is a unit element for C. (3) + (. Note that x + Y is again in C. (X ( a+B 2 2' a+B 2 -(3) 2 = (1. O)·X = X so that (1.a. It is called the ring of 2 X 2 rational matrices. 0) then.(3) as the inverse. Thus the multiplication in R is not commutative. Also it is possible for u·v = 0 with u rf 0 and v rf O. EXAMPLE 7. (3). It. The student should verify that R is indeed a ring. Note that ell' e12 = e12 whereas e12' e11 = O.(30. . (3)'(')'.0). +y = (a.)") 0 = (a + 'Y. Note that X· Y = Y ·X. if X = (a. Y = (. and (. . Also X· (1. (3) = (. Again we notice that X· Y C C. Finally we see that (a.)"0) if and only if a = ')' and (3 = o. Also.3 To illustrate the multiplication. is in C. a2 + (32 rf 0 thus (x. (3). 0) serving as the identity element for addition. (3). We achieve this by defining: for X (3) = (a. Let C be the set of all symbols (a. if then a-b = (ell 621 a = ell - e21 + e22 and b = e22 + 3e12 + 6d·(e22 03e22 + 3612) =0 + 3612 3e22 = 3e12 + e22 = 3e12 + e22 + 0 2e22. 0) in C X·y = (a. 0) = (a'Y . will occupy a good deal of our time later on in the book. (3) rf (0. Now that C is endowed with an addition. under addition. of (a.

called the multiplication table of the quaternion units. This last example is often called the ring of real quaternions. today its primary interest is that of an important example. aI. ±k form a non-abelian group of order 8 under this product. 1. ±i. 1 DEFINITION AND EXAMPLES OF RINGS 87 All in all we have shown that C is a field.O:a(32)i + O:l!33)j + (0:0!33 + 0:3[30 + O:l!32 . In order to make Q into a ring we must define a + and a. ao + ali + azj + a3k and !30 + !3li + !32j + (3ak to be equal if and only if at = !3t for t = 0. Y = (30 + f3li + (32j + (3ak + a2J' + ask) + (!30 + !3li + !32j + !3sk) + (a2 + (32)j + (as + !3a)k 0:2j in = + ali + + 0:3!3l a2!32 .for its elements.0:3(3a) - + + o:ak)· (!30 + !3li + !32j + !33k) = + (O:O!3l + O:l!3o + a2f33 . EXAMPLE 8. Initially it was extensively used in the study of mechanics. ki = j. Admittedly this formula for the product seems rather formidable. while going around counterclockwise you read off the negatives. and as are real numbers. however. although it still plays key roles in geometry and number theory.. 2. ±j. 3. ki = -ik = j. Let Q be the set of all symbols ao ali azj + a3k where aU the numbers aQ. X + Y = (ao + ali (ao + (30) + (al + !3l)i and (2) X· Y = (ao (0:0[30 . If we write (a.g. The latter part of these relations.0:2(3l)k. e. We declare two such symbols. az. It comes from multiplying out two such symbols formally and collecting terms using the relations: i2 = P = k2 = ijk = -1. ij = -ji = k. it looks much more complicated than it actually is. can be remembered by the little diagram.O:l!3l (0:0{32 0:2{30 + a2j + ask. jk = i. This ring was first described by the Irish mathematician Hamilton. jk = -kj = i. Notice that the elements ±1. To this end we define: + + (1) for any X = ao + ali Q. . ij = k.SEC. as you go around clockwise you read off the product. (3) as a + {3i the reader may verify that C is merely a disguised form of the familiar complex numbers.

such that ab = O. All these run counter to our experience heretofore. all a2..band merely write this product as abo DEFINITION. Now if X = ao + ali + ad + ask is not 0.3 o The reader may prove that Q is a noncommutative ring in which + Oi + OJ + Ok and 1 = 1 + Oi + Qi + Ok serve as the zero and unit elements respectively. The examples just discussed in Section 1 point out clearly that although rings are a direct generalization of the integers certain arithmetic facts to which we have become accustomed in the ring of integers need not hold in general rings. DEFINITION. A ring in which the nonzero elements form a group is called a division ring or skewfield. but we would be going too far afield to present one here. a commutative division ring is a field. For instance we have seen the possibility of a. since they are real. Thus the nonzero elements of Q form a non-abelian group under multiplication. is A simple computation now shows that X· Y = 1. Natural examples exist where a. b :. {3 = ao2 a12 + a22 ai rf 0 follows. The unit element under multiplication will be written as 1. and the attempts to classify them form an important part of algebra.z£ b· a. and the inverse of an element a under multiplication will be denoted by a-I. Finally we make the definition of the ultra-important object known as a field. The investigation of the nature of division rings. DEFINITION. Many other examples of noncommutative division rings exist. Q affords us a division ring which is not a field.. naturally enough. .b :.88 RING THEORY CH.b = 0 with neither a nor b being zero.. For simplicity of notation we shall henceforth drop the dot in a. a3 are 0.z£ 0. is an example of an integral domain. Of course. The ring of integers. A commutative ring is an integral domain if it has no zero- divisors.k ~ Q. Thus =0 + + y = . If R is a commutative ring then a :. 2. A ring is said to be a division ring if its nonzero elements form a group under multiplication.z£ 0 ~ R is said to be a zero-divisor if there exists a b ~ R.i.{3 {3 {3 ao al a2 a3 j . then not all of ao. Some Special Classes of Rings.

(5) (-1)(-1)=1.which establishes part (5). But ab + a(-b) = a(b + (-b» = aO = 0 by use of the distributive law and the result of part (1) of this lemma. To this end we prove the next lemma which asserts that certain things we should like to be true in rings are indeed true. 2 SOME SPECIAL CLASSES OF RINGS 89 DEFINITION. Oa = 0 follows. Similarly (3) That (. using the left distributive law. LEMMA 3. = -Cab). we single it out since its analog in the case of real numbers has been so stressed in our early education in grade school. then a (-I)a = la (-l)a = (1 + (-I»a Oa = 0. We wish to be able to compute in rings in much the same manner in which we compute with real numbers. R has a unit element. complex numbers.Oa = (0 + O)a = Oa + Oa. keeping in mind always that there are differences-it may happen that ab F. A field is a commutative division ring.(ab» = ab (by part (2» (by part (2» since -(-x) = xis a consequence of the fact that in any group (U-1)-1 = u. in addition. this equation implies that aO = O. .ba. So on with it: (-a)(-b) = (-a)b = . (1) If a E::: R. Similarly. if a = -1. -Cae-b»~ = .b) = ab is really a special case of part (2). we exhibited the noncommutative division ring of real quaternions and the following fields: the rational numbers. If. then (4) (-I)a = -a. (2) a( -b) = (-a)b (3) (-a)(-b)=ab.(. and since R is a group under addition. and so here too.(ab).a) ( . (2) In order to show that a( -b) = . 1. In particular. + + =. then aO = a(O + 0) = aD + aO (using the right distributive law). If R is a ring.(ab) we must demonstrate that ab + a(-b) = O. In our examples in Section 1. and the integers mod 7. Chapter 5 will concern itself with fields and their properties. (4) Suppose that R has a unit element L. b E::: R (1) aD = Oa = O. (-1)(-1) = -(-1) = 1.SEC.1. then for all a. Proof. or that one cannot divide. whence (-I)a = -a.

but no such argument can be made against this very simple-minded principle given above. stated otherwise. In order to prove that D is a field we must (1) produce an element 1 E: D such that al = a for every a E: D. and one which we shall often use is: If n objects are distributed over n places in such a way that no place receives more than one object then each place receives exactly one object. We claim that they are all distinct! For supposethatxia=xjafori~j. .then (Xi . since a E:: D. . Proof.2.XJ' = 0. This principle says no more or less than the following: if a postman distributes 101 letters to 100 mailboxes then some mailbox must receive at least two letters. The result of Lemma 3. Xna are n distinct elements lying in D which has exactly n elements. every element y E: D can be written as Xia for some Xi. An equivalent formulation. We formalize it and even give it a name..90 RING TREORY CR. Consider the elements Xl a. THE PIGEON-HOLE and if n >m PRINCIPLE. In particular. does it? Yet it will surprise us! Mathematical ideas can often be very difficult and obscure. X2a.. though completely trivial. an integral domain is a commutative ring such that ab = 0 if and only if at least one of a or b is itself O. The lemma just proved. . provides a mighty weapon when wielded properly. X2.. Before we do so we enunciate a principle which. Xn be all the elements of D. is not very exciting. Let xl. A finite integral domain is a field. xna. a = xioa for some Xio E: D. this forces Xi . So let us proceed to results of greater interest.. is a commutative ring with unit element in which every nonzero element has a multiplicative inverse in the ring. a = xioa = . 3 With this lemma out of the way we shall. and so Xi = Xi> contradicting i ~j.1 is our permit to do so. Thus Xla. Since D is an integral domain and a ~ 0. It does not sound very promising as a tool. while it is very useful and important. As we may recall. . By the pigeon-hole principle these must account for all the elements of D. then some place receives at least two objects... X2a. they are all in D. We immediately make use of this idea in proving LEMMA 3. (2) for every element a ~ 0 E: D produce an element b E: D such that ab = 1. If n objects are distributed over m places. and suppose that a ~ 0 E: D. feel free to compute with negatives and 0 as we always have in the past. For convenience. from now on.Xj)a=O. . a + (-b) will be written as a-b. Let D be a finite integral domain. on the other hand. Since D is commutative. A field. .

Find the form of the binomial theorem in a general ring. Prove that if a. PROBLEMS R is a ring in all the problems. since it only has a finite number of elements. b. 4. is a If field. b. If R is a system satisfying all the conditions for a ring with unit element with the possible exception of a + b = b + a.:of 0 the relation ab = ac implies that b = c. Prove that if a. it too is realizable as a multiple of a. P is a prime number then J p. merely considering it as an abelian group under its addition we have. R then (a + b)2 = a2 + ab + ba + b2. hence in Jp one of these is O.) 8.:of 0. COROLLARY. Prove: (a) If D is of characteristic p then px = 0 for all x E: D. as we have seen. rna = O. (Hint: expand (a b)(l 1) in two ways. By the lemma it is enough to prove that Jp is an integral domain. 1.SEC. 3. and so p. b C. c E::D with a . find an expression for (a + b)1< where n is a positive integer. 2. in Chapter 2.:of a E::D and m is an integer can only hold if m = O. For. If a. But then either a == 0 mod p or b == 0 mod p.:of 0 in D and some integer m . y = Xia for some Xi E::D. + + + + . An integral domain D is said to be of characteristic 0 if the relation ma = 0 where 0 . If a.:of 0 in D. prove that the axiom a b=b a must hold in R and that R is thus a ring. Now 1 E::D so by our previous argument. Thus Xio is a unit element for D and we write it as 1. b E: Rand n. (A ring in which X2 = x for all elements is called a Boolean ring. (b) The characteristic of an integral domain is either 0 or a prime number. where by X2 we mean xx. there exists abE:: D such that 1 = ba. in other words. being a prime must divide a or b.) 5. D is said to be of finite characteristic if for some a . dE:: R evaluate (a + b)(c + d). m are integers then (na)(mb) = (nm) (ab). If every x E::R satisfies x2 = x prove that R must be commutative. The lemma is now completely proved. 6. defined what is meant by na where a E: R and n is an integer. If R is a ring. that is. b E::Jp and ab == 0 then p must divide the ordinary integer ab. and so yXio = (Xia)xio = Xi(axio) = Xia = y. c. 7. Show that the commutative ring D is an integral domain if and only if for a. if y E::D. 2 axiO' SOMESPECIAL CLASSES F RINGS 91 O We propose to show that Xio acts as a unit element for every element of D. The characteristic of D is then defined to be the smallest positive integer p such that pa = 0 for some a . Proof. thering of integers mod p.

a m. This suggests that the appropriate analog for rings could also lead to important ideas. To recall. In studying groups we have seen that the concept of a homomorphism turned out to be a fruitful one. Homomorphisms. 3 9.14 for the case of the additive group of a ring yields for us LEMMA 3. b E: R. or if R' is arbitrary but ¢ is onto.l)m on division by n. as far as addition is concerned. all the properties about homomorphisms of groups proved in Chapter 2 carry over.. (Hint: consider the remainders of a. . In particular. Since a ring has two operations. what could be a more natural extension of this type of formula than the DEFINITION. Using the pigeon-hole principle prove that the decimal expansion of a rational number must.) 12. A mapping homomorphism if ¢ from the ring R into the ring R' is said to be a (1) ¢(a (2) ¢(ab) + b) = = ¢(a)¢(b) ¢(a) + ¢(b) for all a. if we totally ignore the multiplications in both these rings. Prove that Lemma 3. + + + 3. Prove that any field is an integral domain. occurring on the left-hand sides of the relations in (1) and (2) are those of R whereas the + and· occurring on the right-hand sides are those of R'. a (n . A useful observation to make is that a homomorphism of one ring. then ¢(l) = l' is indeed true. Therefore. into another. (2) ¢( -a) = -¢(a) for every a E: R. R. become repeating. at least a homomorphism of R into R' when we consider them as abelian groups under their respective additions.3.92 RING THEORY CH.2 is false if we drop the assumption that the integral domain is finite. 11. However. there exists an integer x such that x == a mod m and x == b mod n. As in the case of groups let us again stress here that the + and . if R' is an integral domain. after some point. If ¢ is a homomorphism of R into R' then (1) ¢(O) = O.. for groups a homomorphism was defined as a mapping such that ¢(ab) = ¢(a)¢(b). . . a 2m. is. R'. merely restating Lemma 2. 10. A word of caution: if both Rand R' have the respective unit elements 1 and l' for their multiplications it need not follow that ¢(1) = 1'. Using the pigeon-hole principle prove that if m and n are relatively prime integers and a and b are any integers.

as an additive group. Then cf>(a) = 0 so that cf>(ar) = cf>(a)cf>(r) = Ocf>(r) = 0 by Lemma 3. cf>is called the O-homomorphism. cf>is a homomorphism of JeV2) onto J(V2) and its kernel I(cf».SEC. J( 0) forms a ring under the usual addition and multiplication of real numbers. Trivially cf>is a homomorphism and ICcf» = R. Built into the definition of an arbitrary ring is the condition that the ring forms an abelian group under addition.1. the ring of integers modulo n. the choice is clear. Example 1. Similarly cf>(ra) = O. in particular. However. Example 3. LEMMA 3. If cf>is a homomorphism of R into R' then the kernel of cf>. Before proceeding we examine these concepts for certain examples. 3 HOMOMORPHISMS 93 In the case of groups. given a homomorphism we associated with this homomorphism a certain subset of the group which we called the kernel of the homomorphism. If cf>is a homomorphism of R into R' with kernel ICcf» then: (1) I(cf» is a subgroup of R under addition. (Verify!) Example 4. Let J be the ring of integers. The ring multiplication was left much more unrestricted. I(cf». Proof. Let R be a ring. is the set of all elements a E: R such that cf>(a) = 0. and so. a homomorphism of R.4. The . n are integers. (Verify!) Define cf>:J(y'2) ~ J(y'2) by cf>(m + n0) = m . (2) If a E: I(cf» and r E: R then both ar and ra are in I(cf». addition and multiplication. consists only of O. For this reason the emphasis is given to the operation of addition in the ring. the ring has two operations. as an additive group.ny'2. Define cf>:J ~ In by cf>(a) = remainder of a on division by n. suppose that a E: I(cf». R' = R and define cf>(x) = x for every x Clearly cf> a homomorphism and ICcf» consists only of O. Thus by the defining property of I (cf» both ar and ra are in I (cf». much less under our control than is the addition. in a sense. What should the appropriate definition of the kernel of a homomorphism be for rings? After all. is E: R. and it might be natural to ask which of these should be singled out as the basis for the definition. To see (2). into R'. and we make the DEFINITION. (1) follows directly from our results in group theory. Let J(0) be all real numbers of the form m + n0 where m. r E: R. Example 2. Let Rand R' be two arbitrary rings and define cf>(a) = 0 for all a E: R. J n. Since cf>is. the zero-element of R'.

Sign up to vote on this title
UsefulNot useful

Master Your Semester with Scribd & The New York Times

Special offer for students: Only $4.99/month.

Master Your Semester with a Special Offer from Scribd & The New York Times

Cancel anytime.