Algebra Volume 1 Second Edition P. M. Conn, F.R.S. Bedford College University of London A wo lo, 1807 1982 Ons JOHN WILEY & SONS Chichester - New York - Brisbane - Toronto - SingaporeCopyright © 1974 and 1982 by John Wiley & Sons Ltd. All rights reserved. No part of this book may be reproduced by any means, nor transmitted, nor translated into a machine language without the written permission of the publisher. British Library Cataloguing in Publication Data: Cohn, P. M. Algebra.—2nd ed. Vol. 1 1. Algebra, Abstract I, Title 512'.02 QA162 ISBN 0 471 10169 9 (Paper) Filmset and printed in Northern Ireland at The Universities Press (Belfast) Ltd., and bound at the Pitman Press, Bath, Avon.For Juliet and UrsulaPreface to Second Edition The present edition has offered the first opportunity to make substantial changes to the text. In addition to a complete set of answers to the exercises, these changes are of three kinds. In the first place some additions have been made, dealing with affine spaces (4.8), linear programming (5.3), a more extensive treatment of duality (8.1-2) and a second, more direct derivation of the Jordan normal form, without the use of Ch. 10 (11.5). Secondly, in a number of places concepts have been introduced which should form part of the algebraist’s education, mainly in group theory which is not taken further in Vol. 2 (cf. 9.5, 9.6, 9.8). And thirdly, some obscurities have been removed, proofs expanded, worked examples added and of course some further errors corrected. I am indebted to many colleagues for their helpful criticism of the book, in particular, G. M. Bergman has sent extensive observations after using the book as a text, J. C. Fernau and W. Stephenson have made useful sugges- tions on 5.3 and W. Dicks helped with proof-reading. I am also grateful to many students (including the dedicatees) who by their queries helped me to spot and eradicate obscurities in the text. Finally I would like to thank the publishers for their willingness to undertake this reissue. Bedford College P. M. Conn September 1981 viiFrom the Preface to First Edition Although algebra has a long history, it has undergone some quite striking changes in the past few decades. Not least among these is the way the subject has entered into the development of other branches of mathematics, over and above its new applications elsewhere. Its changing role is reflected in the importance of algebra in the curricula, as well as in the many excellent textbooks that now exist. Most of these are designed for undergraduates at North American universities and are either (a) a very broad introduction to linear algebra, with a little groups and rings, for general students taking mathematics, or (b) a course for graduates, or junior-senior students major- ing in mathematics, who have already taken a course of type (a). The pattern in Britain is a little different: the honours student specializing in mathema- tics takes algebra for two or three years (depending on his ultimate interests) and his need is for a textbook which combines (a) and (b) above and is somewhere between them in level. The object of the present work is to provide such a book: the present first volume includes most of the algebra taught in the first two years to undergraduates at British universities; this will be followed by a second volume covering the third year (and some graduate) topics. The actual prerequisites are quite small: students coming to this book will normally have met calculus and some analytic geometry, complex numbers and a little elementary algebra (binomial theorem, quadratic equations, etc.). In any case, some of the topics will be familiar ideas in a new form. There is no doubt that the chief difficulty for the student is the abstractness of the subject so some pains have been taken to motivate the ideas introduced. Connexions between different parts of the subject have been Stressed and, on occasion, important applications are briefly discussed. .There are numerous exercises, ranging from routine problems to further jdevelopments or alternative proofs of results in the text. Some of the harder jones are marked by an asterisk. The central ideas are group and ring: once the example of the integers and the integers mod m) has been described in Ch. 2, the basic properties f groups and rings are developed in Chapters 3,9 and Chapters 6,10 spectively. The other main topic is linear algebra. Ch. 4 describes vector Paces (without mention of a metric or determinants) and this is followed by ixx From the Preface to First Edition a brief account of methods of solving linear equations (Ch. 5). Determinants are introduced in Ch. 7; although they have not been needed so far, they provide an important invariant and have regained in recent years some of their theoretical importance. Ch. 8 deals with metric questions (quadratic forms, Euclidean spaces) and Ch.11 discusses the various normal forms for matrices. In the first chapter the all-important ideas of set and mapping are briefly described, as well as some notions from formal logic. Although not explicitly used in what follows, the latter has had an important influence and some degree of awareness may preserve the reader from pitfalls. There is no separate chapter on categories in this first volume, but the basic terms are introduced to help systematize the results obtained. It is clearly impossible to be in any sense comprehensive; even if it were possible, this would not be the place. Any selection of material must necessarily be governed by personal taste; my aim has been, if possible, to sustain the reader’s interest, while introducing him to the ideas which are important and useful in present-day mathematics. A more detailed idea of the contents may be gleaned from the table of contents and the index. A book written at this level clearly owes a great deal to other people besides the author (who merely acts as collector, or rather, selector of the material). From the many inspiring lectures I have heard, let me single out the most recent: a delightful series by Paul Halmos in St. Andrews, at the time I was collecting exercises. I am grateful to the Senate of London University for allowing me to draw on past examination papers. The manuscript was read by Walter Ledermann, and by Warren J. Dicks, who also read proofs; my thanks go to both of them for their helpful comments. Finally I should like to thank the publishers for the efficient way they carried out (and often anticipated) my wishes. Bedford College P. M. Coun July 1973Contents Preface to Second Edition vii From the Preface to First Edition ix Table of interdependence of chapters xv Sets and mappings 1.1 The need for logic 1.2 Sets : 1.3. Mappings . 1.4 Equivalence relations . 1.5 Ordered sets Further exercises Integers and rational numbers 2.1 The integers . 2.2 Divisibility and factorization i in Zz 2.3 Congruences 2.4 The rational numbers and some finite fields Further exercises Groups 3.1 Monoids 3.2 Groups; the axioms : . 3.3 Group actions and coset decompositions 3.4. Cyclic groups 3.5 Permutation groups 3.6 Symmetry Further exercises Vector spaces and linear mappings 4.1 Vectors and linear dependence . 4.2 Linear mappings 4.3. Bases and dimension : 4.4 Direct sums and quotient spaces 4.5 The space of linear mappings xi 70 74 76 86xii Contents 4.6 Change of basis 4.7 The rank 4.8 Affine spaces 4.9 Category and functor Further exercises Linear equations 5.1 Systems of linear equations 5.2 _ Elementary operations 5.3 Linear programming . 5.4 PAQ-reduction and the inversion of matrices 5.5 Block multiplication Further exercises - Rings and fields 6.1 Definition and examples 6.2 The field of fractions of an integral domain 6.3. The characteristic = hee 6.4 Polynomials 6.5 Factorization 6.6 The zeros of polynomials 6.7 The factorization of polynomials 6.8 Derivatives 6.9 Symmetric and alternating functions Further exercises Determinants 7.1 Definition and basic properties . 7.2. Expansion of a determinant 7.3. The determinantal rank 7.4 The resultant Further exercises Quadratic forms 8.1 Bilinear forms and pairings 8.2 Dual spaces 8.3 Inner products; quadratic and hermitian forms 8.4 Euclidean and unitary spaces 8.5 Orthogonal and unitary matrices 8.6 Alternating forms 93 101 104 110 112 115 120 129 132 134 136 140 145 147 153 161 164 168 177 182 186 194 199 201 205 208 211 216 224 230 241Contents Further exercises 9 Further group theory The isomorphism theorems The Jordan—Hdlder theorem Groups with operators Automorphisms . The derived group; soluble groups and a simple groups Direct products Abelian groups The Sylow theorems Generators and defining relations; free groups Further exercises 10 Rings and modules 10.1 10.2 10.3 10.4 10.5 10.6 Ideals and quotient rings . Modules over a ring . Direct products and direct sums Free modules Principal ideal domains Modules over a principal ideal domain Further exercises 11 Normal forms for matrices 11.1 11.2 11.3 11.4 11.5 11.6 11.7 Eigenvalues and eigenvectors . . The k{x]-module defined by an endomorphism Cyclic endomorphisms +) i] f= The Jordan normal form The Jordan normal form: another method Normal matrices Linear algebras Further exercises Solutions to the exercises 1 Further reading . 2 Some frequently used notations Index xiii 245 249 255 259 263 266 272 279 287 292 298 301 304 311 314 318 326 332 335 339 344 347 352 356 358 360 362 400 401 403Table of interdependence of chapters (Leitfaden) Sets and mappings Integers and rational numbers 1 | 2 3. Groups d Po eereee Vector spaces Pee | Linear equations ae 5 peso 6 Rings and fields Determinants 7 9 Further group theory Quadratic forms 40 Rings and modules ea 14 Normal forms for matrices1 Sets and mappings This first chapter describes such general background notions as logical terms, sets etc. The notions of set, mapping and equivalence are basic in all that follows; the rest may be skipped (or better skimmed) on a first reading and referred to later when necessary. 1.1 The need for logic Mathematics is the study of relations between certain ideal objects such as numbers, functions and geometrical figures. These objects are not to be regarded as real, but rather as abstract models of physical situations. As examples of the relations that can hold, consider the following assertions that can be made about the natural numbers: (a) every even number is the sum of two odd numbers, (b) every odd number is the sum of two even numbers, (c) every even number greater than 2 is the sum of two primes Of these assertions, (a) is true, (b) is false, while for (c) it is not known whether this is true or false ((c) was conjectured by Goldbach in 1742 and has so far resisted all attempts to prove or disprove it). If our mathematical system is to serve as a model of reality we must know how to recognize true assertions, at least in principle (even though in Practice some may be hard to prove). When the object of discussion is intuitively familiar to us—as in the case of the natural numbers—we take certain assertions recognized to be true as our axioms and try to derive all other assertions from them. Once that is done, we can forget the intuitive interpretation and regard our objects as abstract entities subject to the given axioms. When we come to apply our system to a concrete case, we need to find an interpretation for each notion introduced and verify that each axiom holds in the interpretation; we are then able to conclude that all the assertions derived from the axioms also hold. This underlines the need to keep the axiom system as small as possible. The advantage of this axiomatic method of study is that we can examine the effect on our system of varying the axioms and that the proofs become2 Sets and mappings more transparent the more abstract the system. On the other hand, it takes a little time to familiarize oneself with the abstract notions; here the (more or less concrete) model on which it was based will help, although it is not strictly necessary and certainly no part of the theory. Studying these abstract notions is rather like learning a new language; as in that case we shall find that as our knowledge widens we recognize more landmarks; this makes learning very much easier. But whereas we use a foreign language to talk about the same concepts as in our native tongue, the purpose of the mathematical language is to talk about new ideas which can be expressed only with difficulty (or not at all) in a natural language like English. There is another respect in which the process differs from learning a language: we shall need to reason about the new concepts and this will require careful attention to the logical interrelation of statements. Of course it is true that even in everyday affairs we can spurn logic only at our peril, but there the patent absurdity of our conclusion usually forces us to abandon a faulty line of reasoning. By contrast, when we pursue an abstract line of thought, involving unfamiliar concepts, we may reach conclusions by logical reasoning, but we will no longer be able to check these conclusions by commonsense. It is therefore important to be fully aware of the rules of logic we need and to realize that these rules can be applied without regard for the actual meaning of the statements on which they are used. For this reason we begin by describing very briefly some concepts and notations from logic. Propositional logic describes ways in which true statements (also called assertions or propositions) can be combined to produce other true state- ments. E.g., if it is asserted that ‘Jack was running’ and ‘Jill was singing’, then we may conclude that ‘Jack was running and Jill was singing’. (1) On the other hand, if Jack was not running then statement (1) is false irrespective of what Jill was doing. By enumerating further possibilities we can thus give a precise description of the way the word ‘and’ is used to link assertions. In order to do this concisely, let ‘A’ stand for an assertion, such as ‘Jack was running’, and ‘B’ for a second assertion, not necessarily different from A. Then we can form the expression ‘A and B’, also written ‘A AB’ and called the conjunction of A and B, and make a table which indicates in which cases A A B is true, using “T’ for ‘true’ and ‘F’ for ‘false’: This is called the truth-table for conjunction. It shows that A” B is true11 The need for logic 3 when A is true and B is true, and false in all other cases. For our purposes we may assume that each statement is either true or false; the relevant value T or F is called the truth-value of the statement. Since there are two possible truth-values for A and two for B, we have 2.2=4 possibilities in all, which are listed in the above table. A second way in which assertions can be combined is by using ‘or’: ‘John went to the cinema last night, or to the theatre’. This is a true statement if in fact John last night went to the cinema, and also true if he went to the theatre; the possibility that he went to both is not really envisaged, but if he did, the statement would still be regarded as true. This causes some ambiguity in everyday life: if A and B are both true, is the statement ‘A or B’ to be regarded as true? The situation is usually cleared up by the context (but not always, cf. ‘This summer Jane will go to Italy or Austria’). In mathematics the expression ‘A or B’ is always taken to mean ‘A or B or both’; it is written ‘A vB’ and is called the disjunction of A and B. Its truth-table is A typical use of disjunction in mathematics is the sentence: ‘If a and b are two real numbers whose product is zero: ab = 0, then a = 0 or b = 0’. Clearly we must not exclude the case where a=0 and b=0. With every statement we can associate its opposite or negation by inserting ‘not’ in the appropriate place. Thus ‘Max is the biggest liar’ has the negation ‘Max is not the biggest liar’. Generally, if A is any statement, then its negation is ‘not A’, also written ‘7A’, and it is true precisely when A is false. Its truth-table is A|T F TAF T Here there are only two possibilities because only one statement is involved. The notion of implication is particularly important for us and its use in mathematics differs in some ways from everyday usage, though the underly- ing meaning is of course the same. Thus ‘A implies B’ or ‘if A, then B’, written ‘A = B’ means for us: ‘either A is false or B is true’. It is expressed in the truth-table For example, a mathematical proof might contain the line: ‘If n>5, then n>3’. A parallel use in everyday English would be: ‘If this book was4 Sets and mappings influenced by Shakespeare it must have been written after The Canterbury Tales’. We note from the truth-table for that ‘AB’ holds for any B whenever A is false; in other words, a false statement implies anything. This may seem strange at first, but it has its counterpart in ordinary usage where we underline the absurdity of an assertion by drawing an even more absurd conclusion (‘If Jones wins the election I'll eat my hat’). We also note that implication can be defined in terms of the other connectives by the rule ‘AB’ stands for ‘(7A) v B’. More generally, if we are given a rule for forming a proposition P(A, B) whose truth-value depends only on those of A and B, then P can be defined in terms of —, v alone. E.g., if P is given by the table then P(A, B) =(7A)v(—B). Another connective of frequent use is the bi-implication or equivalence: ‘AB’. This is defined as ‘(A—B)A(B= A)’; thus ‘A = B? is true pre- cisely when A, B are both true or both false. Some composite statements are true for all assignments of truth-values, e.g., AV(7A). Such statements are called tautologies. To check whether a given assertion is a tautology we can again use truth-tables. E.g., consider [An(A=B)]=B. A B AB An(A>B) [Av(A>B)]=B T T T T T T F F F T F T T F T F F T F T This is seen to be a tautology because only Ts occur in the last column. There is a quicker method, based on the fact that in 3 out of 4 cases (PQ) has the value T. Thus assume that for some assignment of truth-values, [Aa(A=B)]=B is F. Then by the truth-table for =, B is F and Aa(A-—>B) is T. By the truth-table for it follows that A is T and AB is T. But we already found B to be F, hence A=B is F, a contradiction. Usually the simple statements discussed above are not enough to deal with mathematical situations. We need, besides the propositions, also prop-11 The need for logic 5 ositional functions or predicates. Consider for example: (a) x is an odd number (x ranges over the natural numbers), (b) x forgot his hat this morning (x ranges over humans), (c) x is married to y (x and y range over humans), (d) x is greater than y (x, y range over natural numbers), Unlike a proposition, a propositional function is no longer true or false but only becomes so when particular values are substituted for the variables (‘2 is an odd number’; ‘Mary’s baby forgot his hat this morning’). In practice one often wants to say that some assertion involving x, say P(x), holds for all x (in the universe of discourse). We write this as (Wx)P(x), which stands for: For all x, P(x) holds. We say that the variable x is bound by the universal quantifier V. To express that P(x) holds for some x, we write (Ax)P(x), which stands for: There exists x such that P(x); here x is bound by the existential quantifier 3. Of course when all the variables occurring in a propositional function have been bound (by univer- sal or existential quantifiers), we have an assertion of the type considered before. For example, in the domain of natural numbers, (WVx)(VWy)(x+y =y +x) expresses the fact that for any x and y the sum x+y is independent of the order of the terms. Similarly, (Vx)(Ay)(x < y) states that for every x, there is a y greater than x, i.e. there is no last number. Note particularly that if we apply the quantifiers in the opposite order we get the proposition (Ay)(Wx) (x (Wx) P(x), UW x)AP(x) <> (Ax) P(x).6 Sets and mappings With the help of these formulae (and noting that — A = A) it is easy to write out the negation of any formula with quantifiers, e.g. LW x)(Ay (V2) F(x, y, 2)] + (Ax)(Wy)(Az)[GF\, y, z)]. As an illustration, consider the assertion that every number has an im- mediate successor. This is expressed by the formula (Wx)(Ay)(Wz)[(x(y =z}. The negation is (Ax)(Wy)(Az)[A@y for —(xz for -(y y)v{@ 2H this says that there is some x which does not have an immediate successor. In any mathematical theory one has axioms from which the assertions of the theory (the theorems) are derived by logical deduction (‘proofs’), using also the logical theorems, i.e. the tautologies. It is not necessary, nor indeed appropriate, to describe in detail the form such a proof would take. The customary presentation of proofs, logically informal though mathematically rigorous, is best assimilated by studying examples. But it may be helpful to end this section with a word or two on the chief methods of proof. A direct proof (or step in a proof) usually takes the form: ‘A’ is true and ‘A= B’ is true, hence ‘B’ is true. This is called modus ponens (which was the term used in Scholastic Logic). It is important to distinguish between ‘A —B’ on the one hand and ‘A, hence B’ on the other. The distinction may seem pedantic in cases where A is true, but to ignore it can easily give rise to confusion. E.g. compare the following two statements about positive real numbers x and y: (a) x>y—(x?>xy and xy>y?)x?>y?, (b) x>y, hence x?>xy and xy>y?, therefore x?>y*. Here (a) is a conditional statement which tells us nothing unless we are told that x>y to begin with. In any case, it is ambiguous; it is of the form A~>B=C and this can mean either (A B)—C, or A(B=>C), or more usually, AB and B=C. With this meaning it is sometimes offered by some slipshod writers who intend (b) (which is unambiguous). However, there is a legitimate use of the expression A~>B—C. When a theorem asserts that a number of statements, say A, B, C, are equivalent, we frequently indicate beforehand the order in which we prove the parts of the theorem. E.g., we may prove in turn AB, BC, C=A; clearly this will establish the claim that A, B, C are equivalent. This is often shortened to ‘A=>B=C=A,11 The need for logic 7 An indirect proof usually takes one of the following forms. In order to prove ‘A = B’ we may prove “1B=—A’ (called the contrapositive), but not ‘B= A’ (the converse), for the latter is not generally equivalent to ‘A = B’, cf. Ex (3). E.g., suppose we wish to prove ‘(Wx)(x? is even > x is even)’. It is not correct to argue: ‘If x is even, then x? is even, hence the result’, but we can argue: ‘If x is odd, then x? is odd, hence the result’. Another form of indirect proof is by contradiction, also called reductio ad absurdum. In order to prove A we show that ‘(7A)=F’, i.e. we show that (not A) leads to a contradiction. Thus to prove V2 irrational, let us assume the contrary, i.e. 2=(m/n)*, where m/n is a rational number. If we take m/n in its lowest terms, m and n cannot both be even and, on multiplying up, we have m?=2n?, i.e. m? is even, hence m is even, say m=2h; now m?= 4h?=2n?, hence n?=2h? and so n? is even, therefore n is also even, a contradiction. This proves V2 to be irrational. This proof goes back to the school of Pythagoras. Finally there is the proof by counter-example. Many statements are of the form (Wx)P(x); if we want to disprove this we must prove its negation, i.e. (4x)-P(x), and this is done by finding a c such that —P(c). E.g., it may be that Goldbach’s conjecture is false; to establish this one would have to find a counter-example, i.e. an even number greater than 2, which cannot be written as the sum of two primes. In mathematics we often have a non-constructive existence proof. This may sound strange at first sight, but on reflection we see that this is no different from everyday life: an assembly of 500 persons must include two with the same birthday, though closer examination is needed to find such a pair. Frequently a theorem is in the form of an implication or an equivalence. ‘We note a number of equivalent ways of saying this: A=B: A holds only if B holds, or A is sufficient for B. A= B (which stands for B=*A): A holds if B holds, or A holds whenever B holds, or A is necessary for B. AB: A holds if and only if B holds, or A is necessary and sufficient for B. In particular, the phrase ‘if and only if’ occurs frequently and it is sometimes abbreviated by ‘iff’. It is also useful to have a sign to indicate the end of a proof (instead of saying each time ‘this is the end of the proof’). We shall follow current usage by employing the sign m at the end of a proof, or at the end of a theorem, either because the proof precedes it or, when the proof is omitted, because it is so easy that it can be supplied by the reader.8 Sets and mappings Exercises (1) Establish the following equivalences: (i) ANA A, (ii) AVA eA, (iii) AA Beo>BAA, (iv) AVBesBVA, (v) (AAB)AC A A(BAC), (vi) (AVB) VC AV (BVC), (vii) AA(BVC)e[(AAB)V(AAC)]}, (viii) Av(BAC)@[(AVB)A (AvO)}. (2) Establish the following tautologies: (i) A=(B—=A), (ii) [A>(B>C)]> [(A=B)>(A>O)], iii) (A=B)-[(B=C)=(A=0)], (iv) “(AnB) oe (AA)v (7B), (v) A(A VB) 2 (AA) A (7B). (3) Show that AB is equivalent to >B——A, but that of the assertions A= B, B-A, neither implies the other. (4) Show that from any tautology expressed in terms of v, A and — another tautology can be derived by interchanging v and A and inserting — in suitable places. Can you formulate a general rule? (5) Define A | B to mean —(AB) (this is the Sheffer stroke function). Show how to express =A and AvB in terms of |; do the same for the other connectives introduced: A, >, <>. (6) P(A, B, C) is defined to be true if precisely one of A, B, C is true. Express P in terms of A, v, 7, and hence in terms of |. (7) Express the following assertions in words, where x, y, z range over the natural numbers (including 0), and replace those that are false by their negations: (i) (Wx\(Ay(x=y+y), (i) (Wx)(Vy)(Az(y2=x—y=x+z), (iii) (x)(Ay)(x# yA (x? = y*)), (iv) Ax)(Wy)(Az Ey > x) (y = xz)]. (8) Express the following assertion in symbols alone: Between any two distinct real numbers there is another real number. (9) What is wrong with the following argument? Any soap is better than no soap; but no soap is better than Wonder-Bubble Soap, hence any soap is better than Wonder-Bubble Soap. 1.2 Sets Many of the objects we shall study are themselves collections of other objects. These collections or sets may be finite or infinite; later we shall meet sets with additional structure, but for the moment we shall look at abstract sets and the ways in which they can be combined to form new sets. By a set we understand, then, any collection of objects. For example, the following are sets: (i) all the stars visible from my house at 9 pm tonight, (ii) all one-legged magicians, (iii) all odd numbers. We see that in some cases it may be difficult to check which objects belong to the set (e.g. (i) above) or whether the set has any members at all (ii) above). All that matters is that the definition is sufficiently clear for us to be able to tell (in principle at least) whether any given object is or is not a member of the set. We usually1.2 Sets 9 denote sets by capitals and their members, also called their elements, by lower case letters. However, it will not always be possible to keep to this convention, especially when we are dealing with sets whose members are themselves sets. If S is a set, we write x € S to indicate that x is a member of S; in the contrary case we write x¢ S. A set, in this sense, is no more and no less than the totality of its members; no considerations of order or multiplicity enter. Thus: Adam and Eve; Eve and Adam; Adam, Eve and the mother of Cain, all describe the same set. In a more precise form this can be stated by saying that S and T denote the same set: S = T, if and only if, for all x,xeSaxeT. Every set encountered in the real world is finite; by this we mean that its members can (at least in principle) be counted, using the natural numbers, and this process stops at a certain number. Otherwise the set is infinite, e.g., the set of all odd numbers is infinite. It is this occurrence of infinite sets in mathematics that requires rather careful analysis. Although no critical situations will arise in these pages, it should be kept in mind that too free a use of set-theoretic notions can easily lead to contradictions. The best known of these is Russell’s paradox: A set may be a member of itself, e.g., the Union of all Registered Charities may be a Registered Charity. Now consider the set M of those sets that are not members of themselves. Is M a member of itself, ie. is Me M? If MeM, then M¢ M (by the definition of M), while if M¢M, then MeM. Thus we reach a contradiction in either case. The paradox is resolved by restricting the ways in which sets can be formed, so that it becomes inadmissible to consider ‘the set of all those sets that are not members of themselves’. There are several ways of doing this, but they need not concern us here; they will not play a role in the rather simple set-theoretical arguments we shall meet. Let S and T be sets. If every member of T also belongs to S, we say that T is a subset of S and write TCS or also S2T. E.g., the odd numbers form a subset of the set of all numbers. Any set S is a subset of itself; this is called an improper subset of S in contrast to the proper subsets of S which are different from S itself. We write TT to indicate that T is a proper subset of S. Frequently we describe a subset of S by means of a propositional function, thus {x € $ | P(x)} denotes the subset of S consisting of those (and only those) elements x for which P(x) holds. E.g., if the set of all natural numbers is denoted by N, then the subset of odd numbers may be denoted by {xeN]|x is odd}. On the other hand, forms like {x | P(x)}, in which the domain over which x ranges is left unspecified, are best avoided. Let S and T be any sets, then the elements which belong to both S and T form a set which is called the intersection of S and T and is denoted by SOT. E.g., if S is the set of all one-legged creatures and T the set of magicians, then ST is the set of all one-legged magicians. It may happen that SMT has no members at all; this means that SMT is the empty set. By10 Sets and mappings definition this is the set with no members; it is generally denoted by @. Two sets whose intersection is the empty set are said to be disjoint. From two sets S, T we can form another set, the union, written SUT, which consists of all the members of S or T. E.g., a public library may admit as borrower anyone who is either (i) a householder in the district or (ii) a resident of at least 3 years’ standing. Denoting the sets of persons named in (i), (ii) by A, B respectively, we see that the set of people eligible as borrowers is A UB. In most cases, the sets under consideration in any given case will all be subsets of some given set U, the ‘universe of discourse’. Thus U might be the set of natural numbers, or of triangles in the plane etc. In this situation we can, for any set S, form its complement, i.e. the set of all members of U that are not in S; it is denoted by S’. Thus if S$ is the set of all odd numbers, then its complement (in the set of all natural numbers) is the set of all even numbers. This example makes it clear why the complement, to be useful, has to be taken within a given set as universe. We also note the following brief way of describing intersection, union and complement, which brings out a certain analogy with the rules for combining propositions. SAT={xeU|xeSaxeT}, MS,={xeU|xeES, for all i}, SUT={xeU|xeSvxeT}, US, ={xeU|xeS, for some i}, S'={xeU|x¢S}. Here U is the universe containing all the objects under discussion. We also define the relative complement $\T={xe S| x¢ T}. Although many of our sets are infinite, we shall also be dealing with finite sets. In particular, with any object x we can associate the set {x} whose only member is x. It is important to distinguish between the set {x} and the object x (which may itself be a set). E.g., the set N of natural numbers is infinite, but the set {N}, whose only member is N, is finite. If S is any finite set, its members can (by the definition of finite set) be labelled or indexed by the integers from 1 to n, for some integer n. Thus if the elements of S are X1,...,X,, then S={x,,...,x,}; if distinct elements have received distinct labels, i.e. if x;# x, for i#j, then S consists of exactly n elements. But it is usually more convenient not to impose this restriction, so that there may be repetitions among x,,..., X,. If we wish to consider the objects x,,..., X, in the order given, we use parentheses: (x,,...,X,), and call the result a Sequence or, more particularly, an n-tuple; e.g., a roster selecting pilots for flying duties is of this form. By contrast, the set {x,,..., x,} where the order is immaterial, is written with curly brackets (braces). A set which has been indexed in some way by the numbers from 1 to n is also called a family; more generally even infinite sets can be indexed if we use an infinite1.2 Sets 11 indexing set. E.g. if A,gc denotes the plane triangle with vertices A, B, C, this provides an indexing of all triangles in the plane by triples of points and we may speak of the family {A,gc} of triangles obtained in this way. From any two objects x and y we can form the sequence (x, y); it is called an ordered pair, and of course is different from (y, x), unless x = y. If S and T are any sets, we denote by S XT the set of all ordered pairs (x, y) with xeS and ye T. When T=S, we also write S? in place of SxS. The set $x T is called the Cartesian product of S and T, after R. Descartes who showed how to describe points of the plane by the Cartesian product of the real line with itself. Examples. (i) At a dance, let S be the set of gentlemen and T the set of ladies, then S x T is the set of possible couples. (ii) If S = {0, 4, 6}, T = {1, 4}, then Sx T={(0, 1), (0,4), (4, 1), (4,4), (6, 1), (6, 4)}. (iii) If S=T=R, the set of real numbers, then R? is the set of pairs of real numbers and these pairs may be used to represent points in the plane. More generally, from n sets S,,...,S, we can form the product S, x S, x +++xS§, whose elements are all the sequences (x,,...,%,), in which x, € S; (i=1,...,n); when S,=---=S, =S, say, one also writes S$" in place of SxSx-+-x§ and S$" is called the nth Cartesian power of S or the Cartesian square when n= 2. Exercises (1) Prove the following formulae for subsets of a set U: (i) ANA =A, (ii) AUA= A, (iii) ANB=BNA, (iv) AUB=BUA, (vy) (ANB)NC=AN(BNO, (vi) (AUB)UC=AU(BUC), (vii), «= AM(BUC)=(ANB)U(ANO), (viii) AU(BNC)=(AUB)M(A UC). (Compare with Ex. (1), 1.1.) (2) Illustrate the formulae of Ex. (1) by taking A, B, C to be the set of all quadrilaterals, all regular polygons and all polygons large enough to cover a penny (not necessarily respectively). (3) If A has @ elements and B has B elements, find the number of elements A x B. If, moreover, A and B are disjoint, find the number of elements in A UB. What is this number when AB has 6 elements? (4) How many subsets are there in a set of n elements? (Do not forget to include @ and the set itself.) (5) Give examples of sets such that (i) all and (ii) none of their members are also subsets. (6) (De Morgan’s laws). Show that (A UB)’ = A’NB’, (ANB)'=A'UB'. (7) Every number can be defined by a sentence in English and since English has a finite vocabulary, the number of numbers definable by a sentence of at most twenty12 Sets and mappings words, say, is finite. So it makes sense to speak of ‘the least number which cannot be defined by a sentence of at most twenty words’. Does it? 1.3 Mappings Let S and T be sets; any subset of S x T is called a correspondence from S to T. E.g., let S be the set of all points in the plane and T the set of lines; the relation of incidence (the point P is incident with the line | if P lies on 1) defines a correspondence from points to lines. To each point P correspond all the lines through P and to each line | correspond all the points on I. This is an example of a ‘many-many correspondence’, where to each element of S correspond many elements of T and vice versa. In general, to each element of S there may correspond many, one or no elements of T. An important special case of a correspondence is that of a bijective or one-one correspondence, also called a bijection. Here there corresponds to each seS just one te T and to each te T just one se S. The following are some examples of bijections: (i) At a gathering of married couples there is a bijection between the set of men and the set of women: to each person there corresponds precisely one spouse of the other sex. (ii) The real numbers may be represented on a line (the x-axis, say, in coordinate geometry) so that to each real number corresponds just one point on the line and to each point on the line corresponds one real number. (iii) The correspondence x <> 2x defines a bijection between real numbers; on the other hand (iv) the correspondence x — x? does not, because x and —x have the same square, for any real x. The last example makes it clear that the notion of bijection is too restrictive to account for such simple functions as x?. But the notion of function, or mapping, in various guises, plays a basic role in mathematics. For this reason the next definition is fundamental in all that follows. A mapping from S to T is a correspondence between S and T such that to each xe€S there corresponds exactly one ye T. If f is the mapping, one writes f: S—> T or S&T and calls § the domain and T the range of f. The unique element y € T that corresponds to x €S is called the image of x and is written f(x) or f, or more often xf. We also write x ++ y to indicate the correspondence between x and its image y. Often the set of all images, namely {y ¢ T | y = xf for some x € S} is also called the image of the mapping f and is written Sf or im f; in practice this double use of the term ‘image’ does not lead to confusion. Examples of mappings. (i) With each newborn baby associate its weight in grams to the nearest gram. This is a mapping from the set of newborn babies to the natural numbers. (ii) Let S be the United Kingdom and T a map of the United Kingdom. There is a ‘mapping’ which associates with each place in the country a point on the map. (iii) and (iv). The examples (iii) and (iv) of1.3 Mappings 13 correspondences considered earlier define mappings from R to itself, namely x+> 2x and x+> x? respectively. (v) If S=(x,) is a family indexed by a set A, then we have a mapping a+» x, from A to S. (vi) Given a Cartesian product P= Sx T, we can define mappings from P to S and T by the rules (x, y) > x and (x, y)+> y; they are called the projections on the factors S and T. Similarly, in an n-fold product P=S,x---xS, we have for each i=1,...,n a projection e,: P— S, given by (x1,...,X,) > xX. Clearly a bijection is a particular type of mapping. On closer examination we see that two properties are required for a mapping to be bijective; it is useful to consider them separately. A mapping f:S— T is said to be injective or an injection or one-one if distinct elements of S have distinct images, i.e. s#s’ implies sf#s'f. The mapping is called surjective or a surjection or onto T if every element of T is an image, i.e. if Sf = T. Thus a mapping is bijective precisely if it is injective and surjective. In the above examples, (i) is neither surjective nor injective (at least if we take enough babies), while (ii) and (iii) are bijective. The mapping x +> x? of R into itself considered in (iv) is neither injective nor surjective, but on the set C of complex numbers it defines a mapping from C to itself which is surjective, though not injective (the surjectivity is just an expression of the fact that every complex number has a square root). Let f: S— T, g: T— U be any mappings, then we can compose them to get a mapping h: S > U, given by xh = (xf)g for allxeS. () A graphic way of expressing this equation is shown in the accompanying diagram. Starting from an element x €S, we reach the same element of U whether we go via T, x > xf +> (xf)g or direct, x + xh; we express this by saying that the triangle shown commutes. r Yad 4 st. u h The mapping h defined by (1) is called the composite or product of f and g and is denoted by fg; in this notation (1) reads x(fg) = Gf). (2) As a rule one omits the parentheses and denotes either side of (2) by xfg. We observe that fg is defined only when the range of f is contained in the domain of g. Further we note that if we had written mappings on the left, (2)14 Sets and mappings would read: (fg)x = g(fx). It is to avoid this reversal of factors that we put mappings on the right of their arguments.+ As an example, let f,g: NN be given by xf=x+1, xg=x?, then xfg =(x+1), xgf=x?+1. We see that fg# gf, so attention must be paid to the order in which the mappings are composed. When S is a finite set, f and g may be given explicitly. Let us indicate each mapping by writing down the elements of S as a sequence and under each element write its image. Thus if S={1,2,3}, and f is given by ¢ : *), while g is ( : >) then fg is ( 7 ), and gf is ( : >). Again fg# gf; on the other hand if h is G 7 : An important rule in composing mappings is the associative law: For any mappings f: S > T, g: T—> U, h: U> V we have (fg)h = f(gh). (3) Observe that both sides of (3) are defined, by what was assumed about f, g, h. More generally, the equation (3) holds for any mappings f, g, h such that both sides of (3) are defined. To prove (3) we apply each side to an element x of S, remembering (2): x[(fg)h] =[x(fg)]h = (xf)g)h and x[f(gh)] = (xf)(gh) = ((xf)g)h; now a com- parison gives (3). With every set S we can associate the identity mapping 1, which maps each element of S to itself: x +> x. Clearly this is always a bijection. Further, for any f:S— T and h: US we have 1,f=f, hls =h. Let S be any set and T a subset, then there is a mapping « from T to S, defined by xi = x for all x € T; this is called the inclusion mapping of T in S. Although « and 1, have the same effect wherever they are defined (namely on T), they must be carefully distinguished; e.g., whereas 1g is bijective, u is injective, but not surjective (except when T= S and so u = 1g). In fact b may be obtained from 1g by restricting the domain to T; this is often expressed by writing «=1,|T. Generally, if f: X—> Y is any mapping and X’ is a subset of X, then the restriction of f to X’, denoted by f | X’, is defined as the mapping f’: X’— Y given by xf'=xf for all xe X'. We observe that this restriction may be written as f’ =f, where « is the inclusion of X’ in X. Using the composition of mappings we can describe bijections. In the first place we have ) then fh = hf. + Another possibility (often used) is to denote the composition of f and g by gf. With this convention, and writing mappings on the left, (2) reads (gf)x = g(fx) and so the order of the factors is the same on both sides. But this would oblige us to read products like gf from right to left and we shall not use this convention.1.3 Mappings 15 LEMMA 1 If f:S—T, g: T—S are any mappings such that fg=Is, (4) then f is injective and g is surjective. 7 Y NN 5—1—+s For let x, yeS and xf=yf, then x =xfg=yfg=y, hence f is injective. Given any xe S, x =xfg =(xf)g and this shows g to be surjective. @ Suppose that f: S—> T, g: T—S satisfy fg=1s gf=1y (5) then f is both injective and surjective, by the lemma, and so is a bijection. Conversely, if f: S— T is a bijection, then we can always find a unique mapping g: T— S to satisfy (5). For, given ue T, we know there exists just one x€S such that xf =u. Put ug = x, then this defines g on T and ugf =u, xfg =x, therefore (5) holds. This proves THEOREM 2 A mapping f: S— T is a bijection if and only if there is a mapping g: T— S to satisfy (5). = There can be at most one mapping g to satisfy (5), for any given f. For assume that (5) holds and that g’: T—S is another mapping such that fg’ =1s, g'f = 17, then by the associative law, g' = g’ls = g'fg = 11g =g. The unique mapping g satisfying (5) is called the inverse of the bijection f and is written f~'. In this notation (5) reads ff-'= 1s, f-'f=17. The distinction between finite and infinite sets is an important one which will be taken up in greater detail in Vol. 2. Here we shall only note one useful property of finite sets (actually it can be used to characterize them): LEMMA 3. An injective mapping from a finite set to itself is also surjective. For let f: S > S be injective and take a € S; we must find b € S such that a=bf. ©) Consider the effect of performing f repeatedly. Let us write f? for ff and generally abbreviate ff---f (with n factors) as f”. In the series of elements a, af, af?,... there must be repetitions, because S is finite, so assume that af” = af*, (7) where r>s say. Since f is injective, xf = yf implies x = y, so we may cancel f16 Sets and mappings in (7). If we do this s times, we get af’*=a, ie. (6) holds with b= af. Later we shall meet many applications of this lemma. It is not really a surprising result and, to newcomers at least, not as surprising as the fact that it no longer holds for infinite sets. To give an illustration, if in a club each member succeeds in borrowing £1 from another member but no two have borrowed from the same person, then everyone has also had to lend £1 (by the lemma) so no one is any better off. But suppose that we have an infinite club, with members A,, Az,... indexed by the positive integers} (where it is assumed that A,,# A, for m#n). If now for each n, A, borrows £1 from A,j+1 then A, is £1 better off, while all the others come out even. The failure of Lemma 3 for infinite sets makes it seem difficult at first sight to extend the notion of counting and cardinality (or ‘number of elements’) to infinite sets. These difficulties were overcome by Cantor who laid the foundations of set theory in the 1870s. This does not concern us directly as we shall (in this volume) use the notion of cardinality only for finite sets. With every finite set S we associate a natural number |S|, the number of its elements (sometimes called the cardinal of S). In a complete account one would have to show that this is uniquely defined, i.e. that different ways of counting S give the same answer. This will be proved when we come to the axiomatic development of numbers in Vol. 2. Exercises (1) S is a set of four elements. Find (i) the number of mappings of S onto itself, (ii) the number of bijections of S to itself. (Hint. Try sets of two and three elements first.) (2) If f: A B and g: BC are both injective (or both surjective), show that fg is 80 too. If fg is injective (or surjective) what can be said about f and g? (3) If f is any bijection and f~" its inverse, show that the domain and range of f-' are the range and domain respectively of f. Show also that f~ is again bijective and GY =f (4) If f: A> B, g: BC, h: CA are three mappings such that fgh =1,, ghf = 1, and hfg = 1c, show that each of f, g,h is a bijection and find their inverses. (5) Let fg=1s; if f is surjective, or if g is injective, show that both f and g are bijective and inverse to each other. (6) Let fg=1s; we say that g is a right inverse of f and that f is a left inverse of g. Show that when the domain of f has more than one element, f is bijective iff it has a 7 It should be pointed out that there are infinite sets that cannot be indexed by the integers (the uncountable sets). This is proved in books on analysis or set theory (see also Vol. 2).1.4 Equivalence relations 17 single right inverse, or also iff it has a single left inverse. (Hint. To get counter- examples, look at mappings from N to N.) (7) (i) For any integer a define a mapping yi, of N into itself by the rule 4, :x+> xa. Show that j4., = Malte. (ii) For any integer a define a mapping a, of N into itself by Gq: X+>x+a. Show that a4, = ap. (8) A mapping f: S — T is said to be constant if xf = yf for all x, y ¢ S. Show that for any two distinct constant mappings of S$ into itself, fg # gf. What happens when S has only one element? (9) In an infinite club indexed by the integers, {A,, Ao, . . .}, how much does A, have to borrow from A,,,, in order that each member shall be £1 better off than before? (10) Let S be the set of finite sequences of Os and 1s and define a mapping f of S into itself by the rule: If a=a,a,---a, (a, =0 or 1), then af=aja}...a;, where 0'=01, 1’=10. Show that af has no block 000 or 111 and that in af?(= aff) any block of length at least five contains 00 or 11. (11)* Show that A x B =Bx A only if A =B or one of A, B is empty. Is it possible for non-empty sets to satisfy (A x B)x C= A X(B x C)? (Hint. Take B to consist of one element.) 1.4 Equivalence relations By a relation on a set S we mean a correspondence of S with itself. E.g., ‘being related’ is a relation on the set of all humans (provided that we are equipped with an exhaustive genealogy). Let w be a relation; we write xwy to express the fact that x stands in the relation w to y, i.e. that the pair (x, y) belongs to w. Frequently relations are denoted by a symbol such as ~, thus in place of xwy we write x ~ y. Many relations have one or more of the following three properties: E.1. For every xe S, x~x (reflexive). E.2 For all x, yéS, if x~y, then y~x (symmetric). E.3 For all x, y,z€S, if x~y and y ~z, then x ~z (transitive). For example, the relation ‘x is father of y’ (on the set of all humans) has none of these properties. On the other hand, ‘x has the same parents as y” has all three, ‘x is ancestor of y’ is transitive and ‘x is brother of y’ is symmetric on the set of all human males, but not on the set of all humans. This last point illustrates that we must always specify the set on which we are operating. A relation on S which is reflexive, symmetric and transitive is called an equivalence on S. This is an important notion, which in some ways generalizes the notion of equality, for the relation of equality (on any set) trivially satisfies E. 1-3. An equivalence on S separates the elements of S18 Sets and mappings into classes, grouping together objects which agree in some particular respect. E.g., ‘x has the same parents as y’ is an equivalence which groups siblings together. Similarly, the relation ‘x has the same remainder after division by 2 as y’ on N groups all the even numbers together and all the odd numbers. Let us see how this can be done generally, for any equivalence on a set S. For any x¢S, we group together all the elements equivalent to x into an equivalence class or block S,, i.e. we put S.={yeS|x~y}. By the reflexivity, xe S,; we claim that any two blocks S,, S, either are disjoint or coincide. Suppose that S, and S, are not disjoint; we must prove that S, = S, and we begin by showing that x ~ y. Since S, MS, # ©, there exists zéS,S,; by definition this means that x ~ z and y ~ z. By symmetry, z ~ y and hence, by transitivity, x ~ y. Now let ue S, then y ~ u, hence x ~u (by transitivity) and so ue S,; this proves that S, ¢ S,. A similar argument shows that S,¢S, and so S,=S,, as claimed. Thus the different S, provide a division of S into non-empty subsets, any two of which are disjoint. This is called a partition of S. Given an equivalence ‘~’ on S, we can form a new set S/~, whose members are the different blocks S, and we then have a mapping A: S—> S/~ which assigns to each xeS the block S,; \ is called the natural mapping from S to S/~. It is surjective, but not injective, unless the equivalence on S was just equality. We note that conversely, every partition on a set S arises in this way from an equivalence. For suppose that S is partitioned into sets A, B,.... Then each x € S belongs to just one set of the partition, say x ¢ A. We put x~y if x and y lie in the same set. This is an equivalence on S with blocks A, B,.... The example considered earlier, ‘x and y leave the same remainder after division by 2’ gives a partition of N into two blocks, the even numbers and the odd numbers. Similarly, in any given year, the relation ‘x and y fall on the same day of the week’ gives a partition of the days of the year into seven blocks, corresponding to the seven days of the week. Any mapping f: S — T gives rise to an equivalence on S by the rule: x ~ y if and only if xf=yf. The reader should verify that this is indeed an equivalence. Exercises (1) Which of the following relations between positive integers are reflexive, which are symmetric and which are transitive? (i) a#b, (ii) a’). Show that greatest, least, maximal,Further exercises on Chapter 1 21 minimal elements for ‘<’ become least, greatest, minimal, maximal elements respec- tively for w. (6) Let < be a relation on S which is transitive and antireflexive (i.e. x T be a mapping. If T is preordered by a relation <, show that the relation w defined on S$ by the rule ‘xwy iff xfx. This relation satisfies the requirements for a total ordering (see 1.5) and is related to the operations of Z by the following rules: Z.7 If x;0, then zx0}. (1) Later we shall see how to reconstruct Z from N; for the moment we note that, for every xeZ, either x=0 or xeN or —xeEN and that these three possibilities are mutually exclusive. In fact, this is true in any totally ordered ring, taking N to be defined by (1). For we know that just one of the following holds (because we have a total order): x =0 or x >0 or x <0. Now x+(—x)=0, hence, if x <0, then 0<—x by Z.7. Thus either x =0 or x>0 or —x>0, as asserted. In order to fix Z completely, we use the following condition on the set N of positive integers: I (Principle of induction): Let S be a subset of N such that 1¢S and n+1e€S whenever ne S. Then S=N. This principle forms the basis of the familiar method of proof by induc- tion. Let P(n) be an assertion about a positive integer n, e.g., P(n) might be ‘the sum of the first n positive integers is n(n+1)/2’. Suppose we wish to prove P(n) for all n, i.e. (¥n)P(n). Then by I it will be enough to prove (i) P(1) and (ii) (Wn)(P(n) = P(n + 1)). For this means that the set S of all n for which P(n) holds contains 1 and contains n+1 whenever it contains 7.21 The integers 25 Hence by I, S=N, i.e. P(n) holds for all neN. There are two alternative forms of I that are often useful. I Let S be a subset of N such that 1¢ S and ne S whenever me S for all m 1—>mn > 1, with ‘>’ unless m=n=1.26 Integers and rational numbers (6) Prove that for any integers a, b the equation a+x=b has a unique solution. (7) If Z is totally ordered in any way, subject only to the conditions Z. 7-8 in the text, show that (i) a >0 implies —a <0 and (ii) 1>0. Deduce that the usual ordering of Z is the only one satisfying Z. 7-8. (8) For any integer a define the absolute value |a| as a if a>0 and —a otherwise. Verify the following rules: (i) ja|=0, with equality iff a = 0, (ii) |a +b|<|a|+\bl, (ii) |ab|=|a| .[b|, Gv) llal—|bl|<|a—b]. Under what conditions on a and b does equality hold in (ii) or (iv)? (9) Show that the sum of the first n positive integers is n(n + 1)/2. (10) Find the sum of the first n odd integers. (11) Find the sum of the first n cubes. (12) Defining (") as the coefficient of a*b"“* in (a+b)", prove that (")= k 7 k -1 “1 (; 7 ‘+ (" \ Hence obtain (")- n(n—1)++-(n—k+1)/k! (by induction), and deduce the binomial theorem: (a +b)" =i nt/k\(n—k)la*b™*, 2.2 Divisibility and factorization in Z Given a,beZ, we write b|a (read: b divides a) to indicate that a is divisible by b, ie. a= be for some c €Z. Since any multiple of 0 is 0, 0| a is true only when a=0. For this reason one usually takes b#0 in bla, although a is allowed to be 0, in fact b | holds for all be Z. The negation of b | a is written b { a; thus b | a means ‘a is not divisible by b’. Divisibility on Z satisfies the following easily verified rules: D. 1 c|b and b|a imply c|a. D.2 ala forall aeZ. D.3 If a|b and b|a, then a=+b. Since 1 is the least element of N, ab =1 (a, b€Z) only holds if a=b=+1. D. 1-3 show that divisibility defines a (partial) ordering on the set of positive integers. D.4 b| a, and b| a, imply b|(a,—a,). D.5 bla implies b | ac for any ceZ. We leave the verification to the reader, but, as an example, let us check D. 3. If a|b, b|a, then a=be, b=ad, hence a = be = adc. It follows that either a =b =0, or a#0 and dc =1, whence c= d =+1, because 1 and —1 are the only integers which have inverses. Two integers a, b such that a|b and b|a are said to be associated. Thus every non-zero integer is associated to exactly one positive integer.2.2 Divisibility and factorization in Z 27 A prime number or prime is an integer p greater than 1 whose only positive factors are p and 1. For example, 2, 3, 19 are prime numbers, but not 1, 9 or 15. The primes may be thought of as the constituents from which every natural number can be constructed by multiplication, just as every natural number can be obtained from 1 by repeated addition. For, as we shall see in Th. 3 below, every positive integer can be written as a product of prime numbers in essentially only one way. Two integers a, b are said to be coprime and a is said to be prime to b if there is no integer other than +1 dividing both a and b; e.g. 12 and 25 are coprime, though neither is a prime number. In fact a prime number p may be characterized by the property that p is greater than 1 and prime to all positive integers less than p. We also note that a, 0 are coprime precisely when a= +1; in particular, two coprime numbers cannot both be 0. The basic tool for studying divisibility in Z is a result going back to Euclid (Prop. 3(ii) below), which depends on the division algorithm: Given a, beZ, if b>0, then there exist q,r¢Z such that a=bq+r, 02 and use inducton: by hypothesis a,|b and by Lemma 2, a, is prime to a,:-- a, while a,---a,|b by induction, hence a,---a,|b by the case r=2. (ii) We shall prove the contrapositive form of this assertion, i.e. p { a; for i=1,...,r implies p { a,--++a,. We remark that for a prime number p and any a&Z, either p|a or p is prime to a. By hypothesis, p | a;, hence p and a; are coprime for i=1,...,1r,s0 p and a,-+-a, are coprime by Lemma 2, ie.pfa,---+a,. = We are now in a position to prove the unique factorization of integers into primes. This is sometimes known as the fundamental theorem of arithmetic: THEOREM 4 Every positive integer can be written as a product of prime numbers: a =PiP2*** Pr (4)2.2 Divisibility and factorization in Z 29 and, if a has a second such factorization, 4=4142"** Gs (S) then s=r and when the q; are suitably renumbered, then p, = qj. Proof. The number 1 can be represented as an empty product, by the convention about such products. If there are positive integers not expressi- ble as a product of primes, let c be the least. Then c is not 1 or a prime, hence c = c,cp, where 11 and by Th. 4, c is divisible by at least one of the ps, say p; (by suitable renumbering): c=p,d. Then Pi(d— pa +++ pn») =pid—pip2**+ p,=1 and this is a contradiction, for no Prime divides 1. =30 Integers and rational numbers The first few primes are 2, 3, 5, 7, 11, 13, 17, 19, 23, 29,.... They become progressively more sparse and are rather irregularly distributed, though the ‘average’ distribution is very regular (Ex. (11), p. 40). Exercises (1) If au+bv = 1, show that u, v are unique up to multiples of b and a respectively. For a given coprime pair a,b, how many pairs u, v exist satisfying au +bv =1 and \u|<|b|, lol 1 is either a prime or has as factor a prime 1, show that the series Yy [a/b"] has at most (log a)/(log b) non-zero terms. (5) Given a prime p, show that for any positive integer c the largest power of p dividing c! is p*, where « =? [c/p"]. (6) Show that for any integer c and any integer n>1, the equation x"=c has no rational solution which is not integral. (7)* If p, is the nth prime (in order of magnitude), show that p, <2". (Use the method of proof of Euclid’s theorem and induction on n.) (8) Show that every product of numbers of the form 4n+1 is again of this form. Deduce that there are infinitely many prime numbers of the form 4n —1. (Observe that every odd prime is of the form 4n+1 or 4n—1.) (9) Show that there are infinitely many prime numbers of the form 6n—1. (Ex. (8) and Ex. (9) are special cases of Dirichlet’s theorem on arithmetic progressions, which states: If a, b are any coprime integers, then there are infinitely many primes of the form an +b. For a proof see, e.g., Serre, Cours d’arithmétique.) 2.3 Congruences In calculations with integers involving division it oftens happens that we are interested in the remainder, but not the quotient. E.g., to find the day of the week on which a given date falls, we can ignore multiples of 7 at any stage, because they represent complete weeks. Similarly, to see if a number is even or odd, we can ignore multiples of 2. The number whose multiples are being ignored is called the modulus. If the modulus is m, we say that a is congruent to b modulo m and write a=b (mod m), (a) if a—b=md for some d€Z. In particular, a=b (mod 0) means a = b. The2.3 Congruences 31 relation (1) between a and b is called a congruence. To express the negation of (1), ie. m | a—b, we write a#b (mod m). This notation (1) was intro- duced by Gauss (1801); although just a different notation for the statement m|a-—b, it has advantages in many situations in that it suggests an analogy to ordinary equality. This is expressed more precisely in Th. 1 below and, in a much more general context, in Ch. 10. THEOREM 1 Let m be any positive integer. Then the relation of congruence mod m is an equivalence on Z. Moreover, for all a, a', b, b'€Z, (i) if a=a’', b=b' (mod m) then a+b=a'+b’' (mod m), (ii) ifa , b=b' (mod m) then ab=a'b’ (mod m), (iii) if ca=cb (mod m) and c is prime to m, then a=b (mod m), (iv) if a=b (mod km) then a=b (mod m). All these assertions follow quite easily from the definition of congruence, except possibly (iii), so we prove the latter and leave the rest to the reader. Assume that ca =cb (mod m). By hypothesis, c and m are coprime, hence cu+mv = 1 for some u, v EZ, i.e. cu=1 (mod m). (2) By (ii), acu =a, bcu=b (mod m) and multiplying the given congruence by u we find acu=bcu (mod m), hence a=acu=bcu=b (mod m). # It is instructive to write this proof out in full: in particular this will give a practical demonstration of the advantage of using the congruence notation. We also note that the proviso in (iii) (that c and m be coprime) cannot be omitted; thus 3=15 (mod 12), but 1#5 (mod 12). Let m again be a fixed positive integer. From the division algorithm we know that for each a¢Z there is an equation a=mqtr, where 01, Y Mx,= M,x,=a, (mod m,), and similarly for i=2,...,r, hence x=YM,x; is a common solution. = As an example, let us find a common solution of x =2 (mod 5), x =1 (mod 6), x =3 (mod 7). Here m=5.6.7=210, M,=42, M,=35, M3= 30; to find the x, we have 42x,=2 (mod 5), so x,=1; 35x,=1 (mod6), so x,=—1, and 30x;=3 (mod 7), so x3=—2. Now 42—35~—2 .30=—53=157 (mod 210). Let us restate Th. 5 in terms of the rings Z/m; for simplicity we take only two factors Z/r, Z/s. If m is any positive integer, we shall write (x),, for the residue class of x mod m. We can define a mapping Z/rs — Z/r by the rule (x)ps > (2), This is a mapping of residue classes, because if (x),,=(x’),, then x=x’ (mod rs) and hence, by Th. 1(iv), x =x’ (mod r). In the same way we can define a mapping Z/rs > Z/s; now we combine these two into a mapping f of Z/rs into the Cartesian product Z/rxZ/s: fi: (X)p > (X),, (2),)- (9) Th. 5 may now be expressed by saying that this mapping is a bijection between Z/rs and Z/rXZ/s; it is surjective because the congruences (8) always have a common solution, for any u, v and it is injective because this solution is unique mod rs. Observe that on the set Z/r x Z/s we again have an addition and multipli- cation if we carry out the operations componentwise: (a, B)+(a’, B’)=2.3 Congruences 35 (a+a’, B +B’), (a, B)(a’, B’) =(aa', BB’). The set Z/r x Z/s with these oper- ations is called the direct product of Z/r and Z/s. Now it may be verified that the mapping (9) preserves the operations, i.e. (x + y)f = xf + yf, (xy)f = xf. yf. A bijection with these properties is called an isomorphism; thus f is an isomorphism between Z/rs and Z/r x Z/s, we also say that Z/rs is isomorphic to Z/rxZ/s and write Z/rs =Z/r XZ/s. It is important to bear in mind that r, s must be coprime for this relation to hold. Of course, Z/r x Z/s can be defined as a direct product without any restriction on r and s and the mapping f can also be defined, but it will not be an isomorphism unless r and s are coprime. The reader is advised to follow this construction through in a particular case, say r=2, s =3, and also to look where it fails for r=2, s=4. More generally it can be shown in the same way that for integers m,,...,m, that are pairwise coprime, if m=m,m,--++m,, we have an isomorphism Z/m=Z/m,X +> > XZ/my. (10) This way of stating Th. 5 is useful because it reduces the structure of Z/m to that of the Z/m;. From Th. 4, 2.2 we know that every positive integer may be written as a product of prime numbers; grouping powers of the same prime together we may thus write m as m= pits *: pis (i) where the p; are distinct primes. If we put m; = p%, we see that the m, are pairwise coprime; thus (10) expresses Z/m as a direct product of the Z/m, where each m; is a prime power. As an illustration we shall calculate the number of invertible elements in Z/m. This number is denoted by ¢(m) and is called Euler’s function. If we recall that the residue class (a),, is invertible in Z/m precisely when a is prime to m, we see that g(m) is just the number of positive integers less than and prime to m; e.g. (2) = 1, ¢(3) =2, p(4) =2, (5) =4. We also put (1) =1. Now an element in a direct product Z/r x Z/s is invertible iff each component is: (x, y)(x’, y’)=(1, 1) iff xx'= yy’ =1. Therefore (rs) = (r)e(s) whenever r and s are coprime. Any function ¢ on Z with this property is said to be multiplicative. It follows that when m is given by (11), then o(m) = e(pt') - ++ (pe), (12) and now it only remains to find g(m) when m is a prime power. Consider p*: the numbers not prime to p* are just the multiples of p, and there are36 Integers and rational numbers p* ' not exceeding p*. The rest are prime to p*, so 9(p*)= p*=p* "= p*(1—p). Combining this result with (12), we obtain the following formula for the Euler function: o(m)=mI] (1-5), where the product is taken over all distinct primes dividing m. Exercises (1) Give a proof of Th. 1. (2) Solve the following congruences: (i) 4x =3 (mod 7), (ii) 3x +2=0 (mod 4), (iii) 2x -1=0 (mod 15), (iv) 3x +6=0 (mod 12). (3) Solve the following systems of congruences: (i) 2x =3 (mod 5), 3x =2 (mod 4), (ii) x=1 (mod 2), x=2 (mod 3), x=3 (mod 5). (4) Prove Th. 5 by induction on r. (5) If a is prime to m, show that a*""=1 (mod m), where g(m) is Euler’s function. (This generalization of Fermat’s theorem is known as Euler’s theorem.) (6) Verify Z. 1-5 for Z/m, and likewise for Z/r0 and m,n are coprime. (2) Given two rational numbers a <8, show that there exist m and r such that a(a+1)(b+1). n (," 1) is p if n is a positive power of a prime p and 1 otherwise. (4) Find a multiple of 7 which has remainder 1, 2, 3 when divided by 2, 3, 4 respectively. (S) If 4 ™m, n are any positive integers, show that q—1|q"~—1; deduce that for m | n, q™—1]q"~-1. (6) Given integers q>1, m,n>0, if q"—1|q"—1, show that m|n. (Hint. Write n=ma+b, where 01, n>0, show that q+1|q"+1 iff n is odd, and 2|q"+1 iff q is odd. Deduce that q"+1 can be prime only when q is even and n=2™. (The numbers F,=2” +1 are the Fermat numbers; the first four Fermat numbers are prime and Fermat conjectured that all are prime, but Euler showed that Fs is composite.) (9) If F, is again the nth Fermat number, verify that for any k=0, Fase — (F, —1)**. Deduce that any two Fs are coprime, and hence obtain another proof of Euclid’s theorem. (Polya) (10) Fill in the details in the following proof of the fundamental theorem of arithmetic (due to Zermelo). Let n be the least number which does not have a unique factorization into primes. Show that the least factor (>1) of n is a prime p; if n= pn’, deduce that n’ has a unique factorization into primes and hence obtain one factorization of n into primes. If there is another one, let q be a prime occurring in it and write n = qm. Now show that n—pm has two distinct factorizations and is less than n, and obtain a contradiction. (11)* For any real x, the number of primes not exceeding x is denoted by 7(x), e.g. 2 a(1)=0, m(m)=2, 7(5-5)=3. Using Ex. (5), 2.2, prove that ( ") divides [I p”, n where », is the largest integer such that p**<2n and p runs over all primes <2n. Deduce the inequality noone *") tar, hn From this inequality and the estimate a (22\_ ao» . ; a ae 2" <(*")<2"", obtained by expanding (1+ 1)” =2°, n deduce the existence of positive constants c,c’ such that et smn ys! . log logn (This inequality was found by CebySev, 1850; Gauss conjectured in 1801 that a(n) .log n/n — 1 as n > (the prime number theorem); this was proved in 1896 by analytical methods.) (12) Show that the product of n terms of an arithmetic progression is divisible by n! provided that the common difference in the progression is prime to n!. (13) If r>0, p is prime and c=1 (mod p’), show that c? =1 (mod p"™*'). (14) Let p be an odd prime and a#0 (mod p); show that the number of roots of a (mod p") is the same for all n>0.Further exercises on Chapter 2 41 (15) Show that the number of proper fractions r/s with denominator (when expres- sed in lowest terms) not exceeding n is Yi g(v). (16)* Let m be an integer and p a prime. Prove that for r=0, (me ) m (mod p). p Deduce that if p{m, the number of subsets with p" elements of a set of mp" elements is not divisible by p. Deduce also that if in some field k, (@+B)" =a"+B" for some n> 1 and all a, Bek, then k has finite characteristic p. Is n necessarily a power of p?3 Groups 3.1 Monoids The number systems described in Ch. 2 had two basic operations: addition and multiplication, and sometimes their inverses: subtraction and division. We now want to study the general properties of an associative but not necessarily commutative operation. The method of doing this is not to look at a particular system such as the integers or the rational numbers, but to assume that we have an arbitrary set with an operation satisfying the associative law and see what consequences can be derived in this way. It will be useful to begin with a formal definition. By a monoid we understand a set S with an element e and a mapping uw: S?> S such that if u(x, y) is the result of applying yw to the pair x, ye S, then M.1 u(x, u(y, z))= w(uG, y), z) for all x, y, zES. M.2 ule, x)= w(x, e)=x for all xeS. Note that by definition, a monoid is never empty. A mapping such as uw which acts on pairs of elements of S is called a binary operation, with values in S, and an element e satisfying M.2 is said to be a neutral element for pw. There cannot be more than one neutral, for if we also have u(e’, x)= u(x, e')=x, then e= ple, e’)=e'. Clearly both addition and multiplication are instances of such a binary operation, and in fact Z is a monoid under addition (with 0 as neutral) as well as multiplication (with 1 as neutral). This example illustrates that in discussing particular instances of monoids one must name not only the set but also the operation, unless this is clear from the context. It will simplify the notation as well as help our intuition if we adopt the multiplicative terminology in general monoids. Thus we shall speak of the multiplication of elements, write xy instead of w(x, y) and call xy the product of x and y. The neutral element e is called the unit-element and is usually denoted by 1. Of course in adopting this notation we must be careful not to assume anything that has not been proved; for example xy will in general be different from yx. If it happens that xy = yx for particular elements x and y, we say that x and y commute; if xy = yx for all x, y, the monoid is said to be commutative.3.1 Monoids 43 Sometimes one may wish to use the additive terminology, e.g., in discus- sing the addition of the integers. Then the element j1(x, y) is called the sum and is written as x+y and the neutral is called zero and is written 0. It is a consequence of the associative law that any product can be written without bracketing. To prove this fact we shall show that a product of n factors, x,x2°* + x, has the same value for any way of bracketing the factors (so long as we take care to keep the factors in the right order). We shall use induction on n; for n= 1 there is nothing to prove, so we assume that n>1 and that for r T, called an isomorphism, such that (xy)f =f). Of) () Isf=11, 6) where 1s, 1;+ are the neutrals of S, T respectively. Thus isomorphic and + This is not to be confused with the notation on p.15. We shall usually denote all multiplicative neutrals by 1, without reference to the monoid, when this is clear from the context.44 Groups monoids are abstractly the same, and differ only in notation; for example, the powers x" of a variable form a monoid isomorphic to the additive monoid of non-negative integers. A mapping f: S > T, not necessarily bijective, but satisfying (5) and (6), is called a homomorphism. Special names are used when T=S; thus a homomorphism of S into itself is called an endomorphism of S and an isomorphism of S with itself is an automorphism. These four names are used quite generally, in discussing algebraic systems, to denote these four types of mapping preserving the structure of the systems. Thus in Ch. 6 we shall define the analogous notion for rings and it will then be clear how the definition would run in other cases. Examples of monoids. (i) We have already seen that Z can be regarded as a monoid (in two ways). The same applies to the other number systems considered in Ch. 2: Q, R, C. The natural numbers N form a monoid under multiplication, but not addition. However, the set Z* of non-negative integers is a monoid under addition. (ii) The set R* of non-negative real numbers is a monoid under the operation max {x, y}, with neutral 0, where max {x, y} stands for the greater of the numbers x, y. (iii) Let X ={x1, X2,...,X,} be a finite set, the alphabet, and denote by X* the set of ‘words’ formed from X, i.e. the set of all finite sequences of elements of X. To multiply two such words we write them one after the other. For example, if X ={x, y}, then X*={1, x, y, x?, xy, yx, y?,x°,...}, where 1 stands for the empty word (which acts as neutral). X™* is called the free monoid on X. An even simpler case is that of X ={x}, then X*= {1, x, x?,...}; we note that {x}*=Z*, an isomorphism being n+> x". An element a of a monoid S is said to be invertible if there exists a'e S such that 7 1 aa'=a'a=1. There cannot be more than one such element a’, for if a" is another with the same properties, then a”=a"1=a"aa'=1a'=a'. We shall call a’ the inverse of a and generally denote it by a~'. The following rules are easily verified: (i) If a is invertible, then so is a~', and (a) "=a. 7) (ii) If a, b are invertible, then so is ab and (ab)*=b"'a"'. (8) It is important to observe the reversal of the factors which occurs in (8). To give an illustration from everyday life, in the morning we first put on socks, then shoes, but at night we have to start by taking off our shoes before we can take off our socks.3.2 Groups; the axioms 45 Exercises (1) Which of the following are monoids under addition? (i) All integers, (ii) all even integers, (iii) all odd integers, (iv) all positive even integers, (v) all integers or halves of integers. (2) Verify that for any set S, the set of all mappings of S into itself is a monoid under composition, with the identity mapping as neutral element. (3) Let M={xo, x1,...,X6} have a multiplication defined by the rule: xx; = X,4)-15 where k is the largest multiple of 3 contained in i+ j—4. Verify that M is a monoid. (4) If M is any monoid and a€M, show that the multiplication defined on M by x * y =xay is associative and that M is a monoid under this multiplication iff a is invertible. What is the neutral in this case? (5) Show that the operation of taking the least common multiple of two numbers defines a monoid structure on N. What about the operation of taking the highest common factor (= greatest common divisor) of two numbers? (6) An element z of a monoid M is called a zero element if xz = zx =z for all xe M. Show that a monoid cannot have more than one zero. (7) Verify the rules (7) and (8) in the text. (8) Show that for any invertible element x in a monoid, and any integer n, Gy =aryt (9) Let f be a surjective mapping of monoids satisfying (5). Show that f is a homomorphism and also give an example of a mapping satisfying (5) but not (6). (10) Let M be a monoid and x € M. Show that there is exactly one homomorphism of Z* (as additive monoid) into M such that 1 x. (11) Let. M be a monoid and aeM such that a°=1. Show that the mapping x axa? is an automorphism of M. (12) Let S be a set with a binary operation xy defined on it satisfying M. 2 (existence of neutral) but not M.1 (associativity). Show that the neutral e is necessary unique, as in the associative case, but that the inverse of x, defined as solution of xx’ =x’x = e, need not be unique. (13) Show that an element a in a monoid M is invertible iff there exists x € M such that axa = 1. 3.2 Groups; the axioms By a group one understands a monoid in which every element is invertible. The commutative law is not assumed; when it holds, i.e. if xy = yx for all x, y, the group is said to be commutative or abelian (after N. H. Abel).46 Groups Thus a group G is defined by the following four conditions: G.1 G has a binary operation xy defined on it. G.2 The operation is associative: (xy)z = x(yz) for all x, y,z€G. G.3 G has a neutral element 1: 1x =x1=x for all xeG. G.4 Every element x €G has an inverse x-': xx"! =x'x=1. It is a remarkable fact that the whole of group theory, which includes many deep results and is far from being fully explored yet, rests ultimately on this simple set of axioms. Of course we shall only be able to touch on the most basic properties in this chapter and in Ch. 9, but it is hoped that even this brief account will convey something of the inherent simplicity and beauty of the theory. We begin by observing that the axioms G.1-4 are to some extent redundant: less is needed to verify that we have a group. This remark is of practical use as it may help to shorten our verifications. Before proving it we note another useful consequence of the group axioms: G.5 Given a,beG, the equations bx =a, yb=a each have a unique solution, namely x =b~‘a, y=ab~'. For if bx =a, then b-'a = b"'bx = 1x =x, so there is at most one solution and, in fact, b(b~'a) = bb-'a = 1a = a. Thus bx =a has exactly one solution; by symmetry the same holds for yb =a. THEOREM 1 Let G be a set with a binary operation which is associative. Then the following conditions on G are equivalent: (a) G is a group, (b) G is not empty and for all a,b€G, the equations bx =a, yb=a each have a solution, (c)_ there exists e € G such that xe = x for all x e G and if we fix such e, then for each x € G there exists x'e G such that xx'=e. Note that we do not assume that the solutions in (b) are unique. We shall prove the theorem according to the scheme (a)—>(b)—(c)—(a). The remarks preceding the theorem show that (a)—(b). Now assume (b), take ae G and find e€G to satisfy ae =a. For any be G there exists xe G such that xa=b (by (b)) and hence b=xa=xae=be; moreover, aa’=e always has a solution a’, therefore (c) follows. Next assume (c); we must show that G is a group. Given x € G, we can find x’e€G with xx'=e and then x"e€G with x'x"=e. Hence x'x =x'xe = x'xx'x" = x'ex"=x'x"=e; thus xx’=e=x'x. Moreover, x = xe =xx'x =ex, ty! = hence e is neutral and x’ is an inverse of x, i.e. G is indeed a group. @3.2 Groups; the axioms 47 As a first application we have PROPOSITION 2. A finite monoid S with either of the properties xz=yz—x=y forall x, y, z€S (right cancellation), (1) zx=zy=x=y forall x, y,z €S (left cancellation). (2) is a group. Proof. Assume that one of (1), (2) holds, say (2); by (2) the mapping Az: x+> zx of S into itself is injective, hence bijective (by Lemma 3, 1.3, because S is finite). Thus bx =1 has a solution for any beS; hence S satisfies (c) of Th. 1 and is therefore a group. @ The notions of homomorphism, isomorphism, endomorphism and au- tomorphism defined for monoids carry over to groups as a special case. We note: to verify that a mapping f: G > H between groups is a homomorph- ism we need only check that (xy)f = xf. yf, for if this holds then (1¢f)* = lef, where 1, is the neutral of G; by division it follows that 1¢f is the neutral of H. Since homomorphisms preserve neutrals it will cause no confusion to denote the neutral in any group by the same symbol 1. We also note that for any group homomorphism f, we have (xf)(x~'f)=(xx~')f=1f=1, and similarly (x~'f)(xf)=1, hence (xf =x7'f. (3) Let G be a group, then by a subgroup of G one understands a subset H of G which is a group relative to the operations in G. Thus H is a subgroup of G iff (i) 1EH, (ii) x, ye H= xy EH, (iii) xe Hx € H. Alternatively, H is a subgroup of G iff (iv) H# @ and x, ye H> xy "'€ H. Clearly (iv) holds for any subgroup of G; conversely, when it holds, take a € H, then 1=aa~'e H, hence for any x € H, x-'=1x~'eH, and whenever x, ye H then y ‘eH by what has just been proved, and so by (iv), xy = x(y')' eH, therefore H is indeed a subgroup of G. In discussing groups the following notation is often useful. Let A,B be any subsets of a group G and write AB ={xy|xeA, ye Bh, Avt={x"!|xe A}. If A reduces to a single element: A = {a}, we write aB instead of {a}B, and similarly for B. To give an example, the condition for a non-empty subset H of G to be a subgroup may be restated as HHCH and H'CH, or equivalently, HH~'< H. Some care must be taken to write HH and not H? (which is sometimes taken to mean {x?| x € H}). PROPOSITION 3 Let G be a group. If H and K are any subgroups of G, then so is their intersection HK.48 Groups To prove this result we need only verify that HMK satisfies the condi- tions (i)-(iii) for a subgroup, a task which may be left to the reader. = In precisely the same way it may be shown that the intersection of any set of subgroups of G (even infinitely many) is again a subgroup. Thus if X is any subset of G we can take the collection of all subgroups H of G such that X ¢H and form their intersection, T, say. Then T will be a subgroup of G and it may be characterized as the least subgroup containing X. A more constructive way of describing T is as follows. Let T’ be the set of all products in the elements of XUX™', ie., the products formed from ele- ments of X and their inverses including 1 as the empty product. It is not hard to verify that T’ is a subgroup and T’> X. Moreover, any subgroup containing X also contains T’. Therefore T > T’, but from the definition of T as the intersection of all subgroups containing X we have T&T’, hence T=T'". This subgroup T is called the subgroup generated by X and is written gp{X}, while X is called a generating set of T. In particular a generating set of G is a subset X such that G = gp{X}. Examples of groups. (i) Z, Q, R, C are all groups under addition, with 0 as neutral and —x as inverse of x; in fact they are abelian groups, each a subgroup of the next. (ii) Let F be any field, such as Q, R, C or F,, and denote the set of non-zero elements in F by F*. Then F* is a group under multiplication, with 1 as neutral and x~'=1/x as inverse. Again these groups are abelian. Note that in the chain C* >R* >Q* each term is a subgroup of the preceding group. In Q* we have a subgroup of two elements, namely {1, —1}, and this (like every group) contains the trivial subgroup {1} consisting of the unit- element alone. We shall usually denote the trivial subgroup by 1 or, for additive groups, by 0. (iii) Let S be any set; by a permutation of S we understand a bijection of S with itself. The set of all permutations of S is denoted by ¥(S). Any permutation has an inverse, again a permutation, and the product of two permutations (obtained by applying them in succession) is also a permuta- tion. In this way ) (S) forms a group, with the identity mapping on S as the unit-element. It is only necessary to check the associative law, and this follows from (3) of 1.3, p. 14. If |S|=1n, we also write Sym, for Y (S). The group ¥ (S) is called the symmetric group on S. We note that as soon as S has more than 2 elements, )(S) is non-abelian. For, given distinct elements x, y, z€S, let a be the permutation interchanging x and y and leaving all other elements fixed, and let 6 be the permutation interchanging y and z and leaving the rest fixed, then a8 maps x, y,z to z, x, y respec- tively, while Ba maps them to y, z, x respectively. (iv) The degree of symmetry of a geometrical figure is expressed by the different ways of moving the figure so as to present the same appearance,3.4 Groups; the axioms 49 e.g., a Square has ‘more’ symmetry than an oblong rectangle. To make this precise we observe that the different movements a figure can undergo without changing its appearance form a group (the symmetry group of the figure), if we ‘multiply’ two such movements by performing them in succes- sion. E.g., the group of symmetries of an equilateral triangle ABC is the symmetric group on A, B, C, while the group of an isosceles triangle with base BC (and not equilateral) is the symmetric group on B, C. The group of the square ABCD consists of 8 elements, the four rotations (by multiples of 7/2), rotations about the diagonals and the bisectors of the sides (cf. p. 65). (v) Any finite group may be described by its multiplication table: write the group elements along the top row and the left-hand column of a square table and put the product ab in the intersection of the row for a and the column for b. From the property G.5 of groups we see that each row and each column must be a permutation of the group elements. This provides a useful check in the construction of multiplication tables. E.g., to construct all groups of 3 elements, we write the elements as 1, a, b. By G.5, ab cannot be a or b, hence ab = 1 and similarly ba = 1. Now there is only one way of completing the table, which looks as follows: It is easily checked that we do indeed have a group; this still has to be checked, all the preceding argument shows is that there is at most one group with 3 elements. Thus we find that (up to isomorphism) there is exactly one group with 3 elements. (vi) The set of all rotations about a fixed point in 3-space forms a group, where the product of two rotations is obtained by carrying them out in succession. This group may also be regarded as the symmetry group of the unit-sphere in 3-space. Unlike the symmetry groups considered earlier, the rotation group is infinite; the rotations about a fixed axis through the given point form a subgroup. (vii) Let G be any group and write C={c €G | ca =ac for all a ¢ G}, then C is a subgroup of G, called the centre of G. By definition, c lies in the centre iff it commutes with all elements of G, therefore C is an abelian group, which coincides with G iff G is itself abelian. Exercises (2) Show that the complex numbers +1, +i(i =~ 1) form a group under multiplica- tion. More generally, show that for any n>1, the complex nth roots of 1 form a group under multiplication.50 Groups (2) Show that the mappings fay: x ax+b(a,beR,a#0) form a group under composition, the 1-dimensional affine group Af,(R). Write down formulae for the product and inverse. (3) Let P be the set of all rotations of a horizontal circular disc about a vertical axis through its centre. Which of the following subsets of P are subgroups? (i) All rotations, (ii) all rotations through less than 7/2, (iii) all rotations through more than 1/2, (iv) all rotations through a multiple of 77/2, (v) all rotations through a multiple of a radian. (4) Let A ={a, b} be a set with a multiplication defined by the table cals sole a b Show that A is a monoid, but not a group. Which rules holding in groups are violated by A? Do the same for the set {1, a, b} with multiplication table (5) Make a group table for the symmetric group on 3 symbols. (6) Ifa group G is generated by elements a, b and ba = ab“, a" = 1, show that every element of G can be written in the form a"b‘(01. (7) Find all groups of four elements and show that they are abelian. (8) Show that x ++ exp x is an isomorphism between the additive group of R and the multiplicative group of the positive elements of R. (9) Show that in any group G, the mapping x +x? is an endomorphism iff G is abelian. What are the conditions on a finite group G for x+»x? to be an automorphism? (10) Let G be any group and aeG. Show that the mapping x+>a™'xa is an automorphism. What is its inverse? (11) Let G be a group and a, b,ceG. If a commutes with b and c, show that a commutes with bc '. Deduce that the set C,(G) ={x € G | xa = ax} is a subgroup of G (C,(G) is called the centralizer of a in G). Deduce that, for any subset X of G, the set Cx(G) = N{C,(G) | x ¢ X} is a subgroup. (12) Find the centre of (i) the group of all rotations about and reflexions in the origin (in 3-space), (ii) Af,(R), (iii) the group of the square.3.3 Group actions and coset decompositions 51 3.3 Group actions and coset decompositions Let S be a set and G any group. By a group action of G or a G-action on S we understand a mapping uw: S x G — S—i.e. a binary operation associating with any pe S, geG an element p(p, g)¢ S—such that A.1 w(p, gh) = u(u(p, g), h) for all pe S, g, heG, A.2 ul(p,1)=p_ forall pes. We express this fact also by saying that G acts on S and call S a G-set. Generally we write pg instead of u(p, g) and rely on the context to indicate whether the group multiplication or the group action is intended; with a little practice and a judicious choice of letters to denote the different types of elements this causes no confusion. When we have a G-action on S, each g¢ G defines a mapping ¢,: p> pg of S into itself. In terms of these mappings the rules A. 1-2 are expressed by the equations: Pech = Oem Qi = 1. () Thus the mapping SO (2) is a homomorphism (of monoids) from G to the monoid of all mappings of S into itself. By (1), Pg@_-1 = _-1@, = 1, which shows that each ¢, is actually a permutation of S. Thus (2) is a homomorphism of G into ¥(S), the symmetric group on S. Conversely, given any set S and any homomorphism g'> @, of G into ¥ (S), we can define a G-action on S by putting pg = pe, (peS); the conditions A. 1-2 follow from the equations (1). The action is said to be faithful if the mapping (2) is injective. Let S be a set, then any group action on S can be used to define an equivalence on S. Let us put x~y iff y =xg for some geG. (3) The group laws just ensure that this is an equivalence relation. It is reflexive because x1 =x for all x eS, symmetric because y = xg implies x = yg™* and transitive because y =xg, z= yh imply z =(xg)h =x(gh). The equivalence classes are called the orbits of the action and the orbit containing x is written xG. If S consists of a single orbit, G is said to act transitively. Examples of group actions. (i) Let G be the group of rotations in 3-space about a fixed point O. This is a G-action on 3-space; the orbits are the spheres with O as centre. (ii) If S is any set, the symmetric group > (S) acts on S in a natural way. We can pass from any element of S to any other by a suitable permutation, thus Y acts transitively.52 Groups (iii) If S is any set and G any group, we can define a G-action by putting xg =x for all xe S, geG. This is called the trivial G-action; the orbits are the 1-element subsets of S. (iv) Let G be any group. We can let G act on itself by right multiplication, i.e. for each ae G we define a mapping of G into itself by the rule Pai X*> xa. This is also called the regular representation of G. The associative law in G shows that this is indeed a group action: pap = Pah». By G.5, G acts transi- tively on itself. If we try to let G act on itself by left multiplication we obtain Nai Xt ax, but this need not define a group action: the effect of A, on x is abx and this is obtained by applying first A, and then A,, thus Nab = Abdar (4) and this need not equal A,A, as required by (1). There are two ways out of this dilemma: either we can distinguish right actions, satisfying (1) and left actions, satisfying (4), or we observe that left multiplication can be used to define a group action satisfying (1), provided that we combine it with the inverse: aixrea'x. Here xpabty = b~'(a~'x) = (ab)'x = xuap, hence fa, = Halt, aS Tequired. For this reason we shall not need to introduce actions satisfying (4). Again G acts transitively under the action so defined. (v) There is another way in which a group G can act on itself. Consider the mapping agi x>aq'xa. We find that xa,a, = b~'(a~!xa)b = (ab)~'!x(ab) = xaqp, thus a, = aa, and we have a group action, called conjugation. This action is never transitive, unless G is trivial, for each a, leaves 1 fixed. It is the trivial action on G precisely when G is abelian. Most of these examples will be used later; for the moment let us consider more closely the group action by right multiplication. Given a group G and a subgroup H, we consider the natural H-action on G obtained by right multiplication. When H is a proper subgroup, this is not transitive; the orbits are of the form xH(xe€G), they are called the right cosetst of H in + Sometimes they are called left cosets and the Hx are called right cosets. The above usage is based on the analogy with ideals (cf. Ch. 10).3.3 Group actions and coset decompositions 53 G; note that H = 1H is itself a right coset. From the definition of cosets as the classes of an equivalence relation we obtain a partition of the group G in the form G=UxH, (5) called the right coset decomposition. A subset of G containing just one element from each right coset xH is called a left transversal of H in G, and right transversals are defined correspondingly. Suppose that G is finite; the number of its elements is called the order of G and will be denoted by |G|. The number of right cosets of H in G is then also finite; it is called the index of H in G and is written (G: H). Further H itself is then finite, of order |H|, and each right coset xH has |H| elements, because the mapping h+> xh (he H) is a bijection between H and xH. The decomposition (5) therefore yields the important formula |G|=(G: H) Al. (6) In particular this proves THEOREM 1 (Lagrange’s theorem) Let G be a finite group, then the order of any subgroup divides the order of G. = This is a useful result which greatly facilitates finding subgroups of finite groups. E.g., it shows that a group of prime order p can have no subgroups apart from itself and the trivial subgroup. Consider the action of H on G by left multiplication; the orbits in this case are the left cosets Hx, and in this way we obtain a decomposition of G into left cosets: G=U Hx. The argument leading to (6) can again be applied, and it shows that in a finite group the number of left cosets of H is the same as the number of right cosets. But this can be proved more generally and more simply by observing that in any group G the mapping x+> x! is a permutation of G, which for any subgroup H of G induces a bijection from the set of right cosets to the set of left cosets, namely xH+> Hx”'. Let us return to an arbitrary G-set S. With each p € S we associate the set of group elements leaving p fixed: G, ={geG| pg =p}. Clearly G, is a subgroup: if g,heG, then p(gh)=(pg)h=ph=p and pg~* =p, moreover p1 =p, so that G, is indeed a subgroup of G. It is called the stabilizer of p under the action of G. It is an important fact that every subgroup of G occurs as a stabilizer of54 Groups some G-set. For, given a subgroup H of G, let us write H\G7 for the set of left cosets of H in G, thus H\G={Hx|xeG}. This is a set of (G:H) distinct elements, and it is a G-set on a natural way: (Hx)y = Hxy. We call H\G the left coset space of G by H and note that it is transitive, for we have Hx. x = Hy. To determine the stabilizer of Hx in this set, suppose that Hx. y= Hx, then xyeHx, hence yex 'Hx, and conversely for any yex 'Hx we have Hxy = Hx. Thus the coset Hx has the stabilizer x~'Hx, in particular, H has stabilizer H. Similarly the set of all right cosets xH forms the right coset space G/H, a left G-set under the left action y . (xH) = yxH. Let us define two G-sets S, T to be isomorphic if there is a bijection g: S—T such that (px)¢ = (pe)x for all pe S, xe G. Then we have THEOREM 2 Let G be a group and H a subgroup, then the left coset space H\G is a transitive G-set with H as stabilizer of a point. Moreover, any transitive G-set with H as stabilizer of a point is isomorphic to H\G. The first part has already been proved. Now let S be a transitive G-set and suppose that pe S has stabilizer H, then every q€ S has the form q = px for some x €G. Let us define a mapping g: H\G— S by Hx+> px. This is well-defined, for Hx = Hy xy '¢ Hes pxy '|=p«px = py; this argument shows in fact that @ is a bijection. Moreover, (Hx)y = Hxy, (px)y = p(xy), hence ¢ is indeed an isomorphism of G-sets. m In particular this theorem completely describes the orbits of an arbitrary G-set. Since the size of H\G is just the index (G:H), we obtain the following useful formula for the size of an orbit: COROLLARY [If a group G acts on a set S and péS is a point lying in a finite orbit pG, then the number of elements of pG equals the index of the stabilizer G, of p: |pG|=(G:G,). = (7) To illustrate this result we consider conjugation in a group. Two elements x, y in a group G are said to be conjugate if there exists ce G such that y=c"'xe. It is easily verified that this defines an equivalence relation on G (cf (v), p. 52); the equivalence classes are called the conjugacy classes of G. Now recall that the mappings a,x co'xc define a group action of G on itself; the orbits under this action are just the + This notation is also used for the relative complement of G in H; it is usually clear from the context which is intended.3.3 Group actions and coset decompositions 55 conjugacy classes. The stabilizer of x € G is the set of elements commuting with x, i.e. the centralizer of x in G: C,(G) ={y €G | xy = yx}. Now the orbit formula (Th. 2, Cor.) shows that the number of elements in the conjugacy class containing x is the index in G of the centralizer of x. By Lagrange’s theorem, the index of any subgroup of G divides the order of G, hence we see that the number of elements in each conjugacy class of a finite group G is a factor of the order of G. If a finite group G has r conjugacy classes C,,..., C,, then we have the class equation IG\=|C,|+ +++ +1G- (8) By what has been said, each |C,| is a factor of |G|. When G is abelian, each C, has just one element; more generally, each element of the centre forms a 1-element conjugacy class. Using these facts we can obtain a great deal of information from (8); e.g. the reader may like to deduce from (8) that a non-trivial group of prime power order has a non-trivial centre. (This is proved in 9.8, Th. 3(i).) Exercises (1) In any G-action, if x is a fixed point of g € G (i.e. xg = x), show that xh is a fixed point of h~'gh. Deduce that conjugate subgroups have the same index (two sub- groups H and K are said to be conjugate if K =g 'Hg, for some geG). (2) Let R.. be the set consisting of the real numbers and a symbol ~. Show that the operations x +> 1—x, xx" (where 07! =, «'=0, 1c =0) generate a group of permutations which is isomorphic to Sym; (thus an action of Sym; on R.. is defined). Show that there are two orbits of 3 points, while all other orbits have 6 points. Describe the action when R is replaced by C, the complex numbers. (3) If a group G acts on sets S and T, we can define a G-action on the Cartesian product S x T by the rule (x, y)g = (xg, yg). Show that the stabilizer of (x, y) is ANB, where A is the stabilizer of x and B the stabilizer of y. (4) Let G be a group and H, K subgroups of index r, s respectively in G. Show that HONK has index at most rs (Poincaré’s theorem). (5) Let G be a group and H, K conjugate subgroups of index r. Show that HNK has index at most r(r—1). (Hint. Find a transitive G-action on an r-element set and consider the action of G on ordered pairs.) (6) Show that in any group, ab is conjugate to ba. (7) Let Cy,..., CG, be the conjugacy classes of a group. Show that each product CG is a union of conjugacy classes.56 Groups (8) Determine all finite groups with only two conjugacy classes. (9) If H, K are subgroups of G, show that |HK|.|H 0 K|=|H].|K|. (Caution. HK need not be a subgroup, cf. 9.1.) Deduce that (H: HN K)<(G: K) with=iff G= HK, and that (G: HNK)<(G: H\(G: K). (10) Let G act on S. If a,be€G are such that ab = ba, and X is the subset of S of points fixed by a, show that b maps X into itself. (11) If G acts on S in two different ways: 9, ¢’, and these two actions commute: a5 = Pb¢a for all a,b € G, show that we obtain another G-action on S by putting Wa = PaP ax Verify that the actions p,, 4. defined in the text commute and identify the action Pabha- 3.4 Cyclic groups A group is said to be cyclic if it can be generated by a single element; thus all its elements can be expressed as powers of the generator. E.g., Z is a cyclic group under addition, with generator 1: every integer can be written as a multiple of 1 (of course Z can equally well be generated by —1). Similarly Z/m for any m > 1 is a cyclic group under addition, again with 1 as generator. We shall now show that every cyclic group is isomorphic to just one of these types. That no two of the groups Z, Z/m (m>1) are isomorphic is clear, because they all have different orders. Let C be a cyclic group with generator c, then two cases are possible: (i) all powers of c are distinct, thus the elements in C are 2¢7,¢'7/,1,607,0,... The mapping n+> c" is an isomorphism from Z to C, so that C=Z. We also say that C is infinite cyclic. (ii) Not all powers of ¢ are distinct. Let c* =c* for h>k, then c’ There is a least positive integer, m say, such that c™ = 1; we claim that the distinct elements in C are 1,c,c?,...,c™~! and that C=Z/m. This will follow if we show that hk], c’=c’ ar=s (mod m). (a) Let r=s (mod m), say r=s+km, then c’ =c**™ =c*(c™)k =c’, because c™=1, Conversely, if c’=c’, let us divide r—s by m: r—s = mq +t, where O0 1). For let H be a3.4 Cyclic groups 57 subgroup of Z and suppose that H is non-trivial, then H contains a non-zero integer k and either k or —k is positive, hence H contains positive integers. Let n be the least positive integer in H, then all multiples of n lie in H, i.e. nZCH; we want to establish equality here. Let he H and write h =nq+r, O1. Exercises (1) Show that C,, has g(m) elements which are generators, where ¢ is the Euler function. (2) Find all cyclic groups which have exactly two elements that are generators. (3) Let G be any group and let a, b be two commuting elements in G, of orders r,s respectively, where r and s are coprime. Show that ab has order rs. If c is an element of order rs in G, where r,s are coprime, show that there exist elements a, b of orders r and s respectively, both powers of c, such that c = ab. If two elements of coprime order in a group commute, show that they are powers of the same element. (4) Show that an abelian group of finite square free order (i.e. of order not di by a square) is cyclic. (5) In any finite group G, show that the equation x* =a has a unique solution for any k prime to the order of G. (6) Show that in any group ab and ba have the same order. (7) Let G be any group and n > 2. Show that the number of elements in G of order n is even. (8) Show that every group of even order has an element of order 2. (9) Show that a group has no non-trivial proper subgroups iff it is cyclic of prime order. (10) Give a direct proof that a finite cyclic group contains at most one subgroup of any given order. (11) Show that every finitely generated subgroup of the additive group of Q is cyclic. 3.5 Permutation groups The action of a group G on a set S may be studied from two points of view. We may be primarily interested in the set S; in that case we have a set with a G-action, already considered in 3.3. Or, we may wish to consider the group G as the primary object. This means that we have a group of permutations where the nature of the set permuted is now immaterial. In3.5 Permutation groups 59 particular, if S is finite, there is no loss of generality in taking the objects of S to be 1, 2,...,n. A permutation of S is said to be of degree n. The group of all such permutations, ie. the symmetric group on {1,2,...,n} is written Sym,, and called the symmetric group of degree n. To find its order we observe that Sym,, acts transitively on {1,2,...,n} and the stabilizer of n is precisely Sym,,_,. Hence by the orbit formula (Sym,: Sym,—4) =n. By induction we find that Sym, has order n(n—1)--++2:1=n!. Of course this order can also be obtained in a more elementary fashion as the number of permutations of n symbols. For a closer study of symmetric groups we need a concise way of writing permutations. There are several notations, useful in different contexts. In the first place, to describe any permutation o we may write down 1,2,...,n in any order, and under i put io. E.g., G 2 y= 1 y-( 2 A) and ( 2 y"=6 2 >) 23 1) \ 2 3) 3 2 23 1/ \3 1 27 This notation makes it easy to multiply permutations: (; 2 SY 2 s)=(3 2 :) G 2 a 2 )-( 2 ;) 13 2/\2 1 3/ \2 3 17 2 1 3/\1 3 2 391 27 However, it is cumbersome to write and it gives no insight into the structure of the permutation. To overcome these defects one uses the cycle notation: a cyclic permutation or cycle (abc -- - f) a permutation in which each symbol is replaced by the next, and the last one replaced by the first; thus a+ b, brec,..., fea. Every permutation o of 1,2,...,n can be written as a product of cycles without a common symbol: the first cycle is 1, lo, 1o?,..., where we stop just before reaching 1 again (we must reach 1 eventually, because o has finite order). If any of 1,2,...,n, say r, is not included in this cycle, we form another cycle starting from r, and so on, until all n symbols are exhausted. E.g., (ri leis 41625 s)=a 4 28 05). A cycle consisting of a single symbol is the identity mapping and may therefore be omitted. Thus the above permutation may also be written as (1 4 2)(3 6). In this way every permutation o of 1,2,...,n can be written as a product of cycles without common symbols: (Aq, 6 A VAp e156 A) Ansty e+ 9 On)» @) This representation is unique, except for the order in which the cycles are60 Groups written, and the symbols in each cycle can be permuted cyclically without affecting o. Various properties, such as the order of a permutation, may be read off from its cycle structure: suppose the permutation o, when written in cycle form (1), has a, cycles of length 1, a, of length 2 and generally a; of length i, We say briefly: o has the cycle structure 12% - + - r%. Any cycle of length k has order k; hence o" =1 whenever m is a common multiple of those numbers i among 1,2,...,r for which a, >0. This condition is also neces- sary, because there can be no cancellation between different cycles. Thus the order of a permutation is the least common multiple of the lengths of its cycles (in the representation (1)). The cycle structure also determines the conjugacy class within Sym, : THEOREM 1 Two permutations are conjugate in Sym,, if and only if they have the same cycle structure. For let G=(a1,..-, A )(Grsiy ++ +5) 0°" Gest + 25%) and T= (Dy, «+5 B)(Bpsay + +5 Bs) +++ (Desay ++ +5 bn), then a=(% a °° “n) by bo ve 2B, is such that a@-'oa =7. For a *oa maps bj+> a,+> a2+> by, and similarly for other symbols. The converse is clear. = A third method of writing permutations is as a product of transpositions. By a transposition one understands a cycle of length 2. Every cycle is a product of transpositions; for instance (1 2+-+n)=(1 2)(1 3)--- (1 nn). (2) It is clear from (2) how to write an arbitrary cycle as a product of transpositions. Since every permutation can be expressed as a product of cycles, it follows that every permutation can be expressed as a product of transpositions. Of course this expression is far from unique, but we can make some economies by restricting the type of transposition used: THEOREM 2 Sym, is generated by (1 2), (1 3),...,(1 n). This means that every permutation is a product of the transpositions shown (and their inverses, which, however, are not needed in this case). Since we have already seen how to express every permutation in terms of transpositions, it only remains to express the general transposition in terms of those of the form (1 i). If i, j#1, this is done by the formula @)=Q NA Ya d; together with the realtion (i 1)=(1 i) this establishes the result. =3.5 Permutation groups 61 Although there are many ways of expressing a given permutation o as a product of transpositions, the parity is always the same, i.e. all such expres- sions for 0 have an odd number of factors, or all have an even number of factors. To prove this fact let us write o in cycle notation (1) and observe that a cycle of length n can be written as a product of n—1 transpositions (by (2)); let us call o even or odd according to whether the number of cycles of even length in (1) is even or odd. Correspondingly we define the sign of o, sgn, to be +1 or —1. Note that a cycle of even length is odd, and one of odd length even. Consider how the sign is affected if we multiply by a transposition, (1 i) say. Two cases are possible: (a) 1 and i occur in the same cycle in (1), say in (1,2° ++r). Then (1 2-s+ie+ A DC 2+++i-1G i41---) (3) and this changes the sign of o: if r is even, both cycles on the right are even or both are odd; if r is odd, just one cycle on the right is even; so in each case the number of even cycles is changed by one. (b) 1 and i occur in different cycles; by suitable renumbering we may assume that they occur in (1 2-+-i—1) and (i i+1---+r) and then (1 2-5 -i-DGEIF1++ +N N= 2---i- ++). (4) This is the inverse of (a) and the same argument shows that the sign is changed. Thus each multiplication by a transposition changes the sign. By induction we see that any odd number of transpositions change the sign and any even number leaves it unchanged. Moreover, the product of two permutations is even if both have the same sign and odd if they have different signs. These results may be summed up as THEOREM 3 With every permutation o we can associate a sign +1 or —1, written sgn o, in such a way that sgn =—1 for any transposition + and sgn ot =sgno.sgn7. In terms of C, the cyclic group of orer 2, we can express this result by saying that there is a homomorphism of Sym, onto C,. The even permuta- tions in Sym,, form a subgroup, called the alternating group of degree n, and denoted by Alt,,. It is of index 2 in Sym,, with the odd permutations forming the other coset. For n =1, 2, Alt, is the trivial group, while Alt, is cyclic of order 3. Generally Alt, has order n!/2, since it is of index 2 in Sym,. From the definition we see that Alt, contains all cycles of length 3. In fact it is generated by these 3-cycles; more precisely, we have THEOREM 4 Alt, is generated by (1 2 3), (1 2 4),...,(1 2 n). Proof. The formulae (i j)(k 1) =(i j K(k iD, (i Pi k)=(i j k) show that62 Groups Alt, is generated by all 3-cycles. Now we use the formulae (i j k)= GQ2)'2j; HA 2), 27 =A 2) 2H 2), Aj w= (1 2 k)(1 2 jf) (1 2k) to express any 3-cycle in terms of the cycles (2%)... The notion of a permutation group—i.e. a group of permutations—is useful because it provides a method of representing group elements, from which the whole structure of the group can in principle be recovered. This makes it important to know what groups occur as permutation groups. In fact all groups do; this is the content of THEOREM 5 (Cayley’s theorem) Every group is isomorphic to a group of permutations on a set. Proof. Let G be the given group; as our set we take G itself. Each acG acts on G by right multiplication p,: x» xa. We have seen that this is a G-action, thus we have a homomorphism G-=L@o@ (5) given by a+>9,, the regular representation of G, and it only remains to show that this is injective. Suppose that p,=p,, then a=1p,=1p,=5, hence (5) is injective and so G is isomorphic to a subgroup of ¥ (G), i.e. toa group of permutations of G. = The result shows in particular that every finite group, of order n say, is isomorphic to a subgroup of Sym,. Of course this applies only to finite groups, because Sym,, is itself finite. Exercises (1) Find a *B~*aB in the following cases: (i) a=(1 2 3), B=(1 4 5), (ii) a= (1 2 3 4), B=(1 3 5), (iii) @=(1 2 3), B=(4 5 6). (2) Show that the elements 1, (1 2)(3 4), (1 3)(2 4), (1 4)(2 3) form a subgroup of Alt,. (This permutation group, and sometimes any abstract group isomorphic to it, is called the four group.) (3) Show that Sym, is generated by (1 2), (2 3),...,("—1n). (4) Show that Sym, is generated by (1 2+++n) and (1 2). (5) Show that Alt, is generated by (1 2 3) and (1 2---n) or by (1 2 3) and (2 3+++n) according to whether n is odd or even. (6) Show that an n-cycle (1 2---n) can be expressed as a product of n—1 transpositions but no fewer. (7) Show that every finite group can be represented as a group of even permutations.3.6 Symmetry 63 (8) Let G be a subgroup of Sym, not contained in Alt,. Show that exactly half the permutations in G are even. (9) Show that any group of order 4n +2 has a subgroup of index 2. (Hint. Use Ex. (8).) (10) A permutation is said to be regular if (in cycle notation) all its cycles have the same length. Show that a permutation is regular iff it is a power of a cycle that moves all points. (11) A permutation group is said to be regular if it consists of regular permutations. Show that a permutation group is regular iff each permutation moves all points or no point. Use Cayley’s theorem to show that every finite group can be represented as a regular permutation group. (12) Show that every abelian transitive permutation group is regular. (13) Show that the n! arrangements of n symbols can be numbered in such a way that each is obtained from the preceding one by a single transposition. Moreover, the first term can be chosen arbitrarily. (Hint. Use induction. This result has an application to change ringing.) 3.6 Symmetry Groups form an important tool in the study of geometrical symmetry. Many mosaics and tapestries show symmetries of varying kinds and the natural way to classify them is by means of the groups they admit. Below we give some examples. For a fuller discussion we refer to the books by Speiser, Weyl and Coxeter (see Appendix 1). These examples will not be needed elsewhere in the book. We shall need a formula for the number of orbits in a G-set: THEOREM 1 Let G be a finite group acting on a finite G-set S. For each ge€G let c, be the number of points fixed by g, then the number of orbits is 1 =a YD ce qd) geG Thus t is the ‘average’ number of points fixed by a permutation. To prove (1) we count the number of pairs (x, g)¢ SX G such that xg = x in two ways: on the one hand, for each geG, the number of pairs occurring is c,; on the other hand, for each orbit, of k points say, each point x is fixed by the elements of its stabilizer, which by the orbit formula has |G|/k elements. Thus each orbit contributes |G| pairs in all and so Y c,=|G|. 1, where t is the number of orbits. Now (1) follows on dividing by |G|. = As a first example consider all finite rotation groups about a fixed point O in 3-space. Any rotation a# 1 about O has an axis through O, whose points64 Groups it leaves fixed. In fact the rotation is completely determined by the way it permutes the points of the unit-sphere S$? about O as centre. The points where the axis of a meets S? (and which are left fixed by a) are called the poles of a. Let G be a finite group of rotations and write |G|=n. A pole P is said to have multiplicity k or be k-tuple if it is a pole of precisely k rotations (including the identity) of G. Thus the stabilizer of P has order k and the G-orbit of P has n/k members. Clearly these are all poles: if P is a pole of a, then PB is a pole of 8B ‘a. Let the poles fall into t orbits, and let the ith orbit have 7, poles; if n=1m, these poles must be m,-tuple. Each non- identity rotation fixes two points, while the identity fixes all. Thus, letting G act on the set of poles of elements #1 of G, we have, by Th. 1, n.t=2(n-1)+ Dn. On dividing by n and remembering that r,=n/m,, we find (after a little rearrangement) 1 1 a(1-4)=5 (1-4). (2) n m, Excluding the trivial group, we have n> 1, hence the left-hand side of (2) is at least 1 but less than 2. Now m; =2, so on the right each term is at least } but less than 1, so there are at least 2 terms but not more than 3 on the right of (2). We take these two cases separately: (i) There are two orbits. Then (2) becomes Bala n m, m Since each r; is a positive integer, r, = r2= 1 and m, = m,=n. There are just two poles; if is a rotation through the least angle 6 say, then any rotation in G is a power of a. For if 6 is a rotation through an angle A, then r93 and we find 1_1,2 m; 6 n° This holds when m3=3, n=12; m3;=4, n=24; m3=5, n=60 and for no other values. (ii, b) n=12, the multiplicities m, are (2,3,3). There is an orbit of 4 points A, B, C, D. If we join these 4 points, we obtain a regular tetrahedron; its 4 faces are equal and equilateral triangles. A second orbit of 4 points is obtained by intersecting the unit-sphere S? with the perpendiculars in the midpoints of the faces, and an orbit of 6 points by joining the midpoints of opposite edges and intersecting these joins with S*. The group is called the tetrahedral group. If we represent it as a permutation group on A, B, C, D we see that it contains all 3-cycles, and hence it contains the alternating group on these 4 letters, Alt,. Since it has the same order as Alt,, viz. 12, it coincides with Alt, as abstract group. (ii, c) n = 24, multiplicities (2, 3, 4). There is an orbit of 6 points; they form the vertices of a regular octahedron. Its faces are 8 equal and equilateral triangles and the perpendiculars in their midpoints meet S? in 8 points of a second orbit. The joins of midpoints of opposite sides meet S? in 12 points of a third orbit. This group is called the octahedral group. (ii,d) n=60, multiplicities (2,3,5). The smallest orbit has 12 points, which form the vertices of a regular icosahedron, whose faces are 20 equal equilateral triangles. Again we get an orbit of 20 points from the faces and an orbit of 30 points from the edges. The group is called the icosahedral group. We shall show that it can be regarded as a permutation group of degree 5. Consider the orbit of 30 poles; they fall into 15 pairs of opposite poles. Each pole belongs to just one rotation apart from the identity and the pole pairs are permuted transitively by G. Hence the stabilizer H of such a pair (PQ) say, has order 4. If a lies in H but does not fix P, it interchanges P and Q, hence a? fixes P, Q as well as the poles of a and so must be 1. Thus H ={1, a, B, y} consists of 3 elements of order 2 besides the unit element. Since a, B, y all have even order, their poles are all double and form a set of66 Groups 3 pairs among the 15. In this way the 30 double poles fall into 5 sets of 6, which are permuted transitively by G. Tt may be shown that G is Alt;, by identifying G as the alternating group on the 5 sets of 6 poles, by a closer analysis of the icosahedron (Ex. (7)). The finite rotation groups are closely connected with the five regular solids, which were already determined by Euclid. Some of them appeared naturally in the above discussion; the rest are obtained by dualizing, i.e. interpreting faces as vertices and vice versa. The complete list is as follows, with the number of faces, edges and vertices listed after each name: (i) tetrahedron (4,6, 4), (ii) hexahedron or cube, (6, 12,8), (iii) octahedron (8, 12, 6), (iv) dodecahedron (12, 30, 20), (v) icosahedron (20, 30, 12). From the numbers given we see that the icosahedron and dodecahedron have the same group, as do the octahedron and cube. A second example is taken from topology. In any topological space X fix a point x and consider loops on x, i.e. continuous images of a finite interval beginning and ending at x. Two loops are multiplied by traversing them in succession. An equivalence between loops is defined by declaring a equiv- alent to B if a can be deformed continuously into 8. It is now possible to consider the multiplication of equivalence classes and so show that we have a group; the neutral element consists of all loops that can be contracted to x. The resulting group 7,(X) is called the fundamental group or the first homotopy group of X; when X is connected, the isomorphism type of 7,(X) does not depend on the choice of the base point x. This group is an important topological invariant of the space, e.g., for the plane or a 2-sphere it is trivial, but for the circumference x?+y?=1 in the plane (or also for the infinite cylinder in 3-space given by the same equation) it is the infinite cyclic group. Exercises (1) Let G be a group acting transitively on a set of n elements. Show that the average number of points moved by an element of G is n—1. Deduce that G is a regular permutation group iff |G|=n. (2) Let D,, be the dihedral group of order 2n. Show that for n > 2 the centre of D,, is 1 or C, according to whether n is odd or even. (3) In the dihedral group D,, of order 2n, considered as acting on a regular n-gon, let a be a rotation through 27/n, and B a rotation through 7 about an axis in the plane of the n-gon. Prove that a"=8?=1, BaB=a' and deduce that every element of D,, can be written uniquely as a'B* where 0 ax +b, where a,b € F, a #0, form a group Af,(F). Find the order of Af,(F) when F is finite. (9) Show that the mapping x+> x"! in a group is an automorphism iff the group is abelian. (10) Show that a group satisfying (xy)?=xy? is abelian. In any group express a”'b~‘ab as a product of 3 squares.68 Groups (11) Let G be an infinite group and H a subgroup of finite index. Show that H intersects every infinite subgroup of G non-trivially. (12) If a finite group acts transitively on a set S with more than one point, show that some g¢G moves all points. (13) Let G be a group and H, K subgroups. If Hx = Ky for some x, y € G, show that H=K. What can we conclude from Hx = yK? (14) Let G be a group and H, K two subgroups (not necessarily distinct). Define a relation on G by putting x~y whenever HxK=HyK; verify that this is an equivalence relation on G. Show that |HxK|=|H].|K\/|H* K|, where H™ = x"'Hx. (The sets HxK are called double cosets of G with respect to H and K, and the expression of G as a disjoint union G = J Hx;,K is called the double coset decompos- ition.) (15) Let c be an element of order n in a group; show that c* has order n/d, where d is the highest common factor of n and k. (16) Show that any group with exactly one maximal proper subgroup is cyclic of prime power order. (17) Let G be a group, H a subgroup and c an element of order n in G. If r is the least positive integer such that c’ ¢ H, show that r|n. (18)* Let G be a group and H a subgroup of index n. If c € G, show that c‘ € H for some k such that 0 3, and Alt, has trivial centre for n>4. (21) For any o € Sym, an inversion is defined as a pair (i, j) in the range {1,...,n} such that i jo. By considering the action of o on the product [Jj <; (x; — xj) show that sgn o =(—1)"®, where v(c) is the number of inversions of o. (22) Describe the group of the square as a subgroup of Sym,. Find its conjugates in Sym, and show that they all have the four group in common. (23) Let G be a group of odd order. Show that for each element a of G, the equation x?=a has a unique solution. Denoting this solution by a}, define a new operation ‘O’ on the set G by the formula xOy =(xly!?, Prove that G forms a group under this operation and that the two group operations on G coincide iff the original group is abelian.4 Vector spaces and linear mappings Many mathematical problems have the property that for any two solutions u and v, u+v is also a solution. Such problems are called linear and are usually very much easier to solve than more general problems. Furthermore, many problems arising in applications are in fact linear, at least in the first approximation, so that this method is of great practical use. For example, the principle of superposition in physics is just an expression of the fact that the differential equations satisfied by heat, light, electricity and other phenomena are linear. Since these problems have a common mathematical content, it is natural to study this in its abstract form, guided at first by geometrical analogy. We recall that in geometry, vectors are used to represent ‘directed lengths’; more generally, any physical quantity such as velocity, which is described by a magnitude and direction, can be represented by a line segment OP of the appropriate length and direction, from the origin O of coordinates. If the endpoint P has coordinates (a,, a2, a3), the vector a= OP is completely specified by these coordinates and we may write a = (qj, do, a3). Any triple of real numbers a,, a2, a3 defines a vector in this way and distinct triples correspond to distinct vectors, e.g. (2, 1, 3) # (2, 3, 1). If a=(aj, a2, a3) and b=(bj, by, b3) are two vectors, then the sum of a and b is formed by adding the components: a +b = (a,+b,, a2+b», a3+bs). This corresponds to the well-known parallelogram rule illustrated in Fig. 1. Figure 170 Vector spaces and linear mappings Similarly, on multiplying a by a real number a (a ‘scalar’) we obtain another vector aa =(ad,,ad2,aa3) in the same straight line as a. Any linear combination of a and b such as aa+ fb, where a and 8£ are scalars, represents a vector in the same plane as a and b and, provided that a, b define a plane (i.e. if they do not lie in the same straight line), every vector in this plane can be written as a linear combination of a and b; more briefly, a and b span the plane. In one’s first encounter with vectors it is of course helpful to keep the geometrical picture in mind. But it is important to realize that this particular interpretation, although it aids one’s intuition, is not an essential part of the theory. This becomes clear if, e.g., we make a statistical study of the measurements of 17 different characteristics of an organism. These are most naturally plotted in 17-dimensional space, in which the vectors are se- quences with 17 components. Here the geometrical analogy, though still highly suggestive, is no longer relevant. All that matters are the relations between vectors and the operations by which they are combined. This is brought out most clearly by an axiomatic development of vectors, a course which has other advantages too. In the first place, it allows us to treat vectors of any finite number of dimensions with the same ease. Secondly, we deal directly with the vectors, as far as possible, without referring to a basis. Thirdly, we can allow the scalars to lie in any field, not necessarily the real numbers. 4.1 Vectors and linear dependence We begin by defining a vector space over a field F of scalars. Here F may be any field, although the reader is advised to keep in mind a concrete case such as F=R at first. Likewise he should keep in mind the case of geometrical vectors as an illustration, although, when it comes to proofs, we can of course only use the axioms or results proved earlier. A vector space over a field F is an abelian group V, written additively, with a mapping of FX V into V (to represent multiplication by scalars), subject to the axioms V. 1-4 below. We shall write the sum of the vectors x and y as x+y and the image of (a,x), for ae F, xe V, as ax; then the axioms read: vl a(x+y)=axtay. v.2 (a+B)x-axtpx, (wBShuyeV) V.3 (aB)x = «(Bx). vV.4 1.x=x. The elements of V are called vectors; by contrast the field elements are called scalars and F itself is called the ground field. We shall mostly use the latin alphabet for vectors and greek characters for scalars.41 Vectors and linear dependence 1 Of the above axioms, V. 1 states that each a € F defines an endomorphism of V, qua abelian group, while V.2-4 just show that the field operations correspond to the way scalars act on V. Note that in V.2 the addition on the left is in F and on the right in V; similarly on the left of V.3 we have the field multiplication, while on the right we have the action of F on V. We note the following consequences of the definition: (i) Ox =a0=0. Here the first zero is a scalar, while the other two are vectors. This rule is analogous to the corresponding rule for integers or indeed field elements. To prove the rule, we write 0x =(0+0)x =0x +0x by V.2, hence 0x =0; the other part follows similarly, using V. 1. (ii) ax =0 implies a =0 or x =0. For if a#0, then x=1.x=(a7'a)x = a"(ax)=0. (iii) If m is any positive integer, then F contains an element m. 1, and (m.1)x=x+x-+++x (m summands). This is easily proved by induction; the details are left to the reader. As a result there is no ambiguity in writing mx for (m.1)x; note however, that for a field of prime characteristic p, where p.1=0, we have px =0 for all x. (iv) If the negative of x (i.e. the additive inverse of x in V) is written —x, then (—1)x =—x. For we have x +(—1)x =(1—-1)x =0x =0. Examples of vector spaces. (i) The geometrical vectors in the plane or in space form a vector space over the field R of real numbers. (ii) For any field F and any integer n=1, denote by F" the set of all n-tuples (é;,...,&) where &€F and (&,...,&) and (m,...,7,) are distinct unless €;=17,...,&, =n. This set F" forms a vector space under the operations (1-22, &) (m+. mm) = (rtm... & +n) a) aE... &)=(ak,...,08) we F (2) This is an important example, which in many ways is typical (cf. Th. 1, 4.3). In particular, for n = 1 this shows that each field may be regarded as a vector space over itself. (iii) For any field F consider the set of infinite sequences (€1, &2,...)- Addition and multiplication by scalars can be defined as in (1) and (2) and it is easily seen that we then have a vector space, F™ say. Similarly the subset F™ consisting of all sequences with only finitely many non-zero coordinates is a vector space. (iv) The set of all real functions is a vector space if we define f +g as the function whose value at x is f(x)+ g(x), and af as the function whose value at x is af(x). (v) The complex numbers C form a vector space over R; both C and R are vector spaces over the rational numbers Q.72 Vector spaces and linear mappings (vi) The set consisting of a single element 0 is a vector space, called the trivial space. The abelian group structure is clear and we define «0=0 for allacF. Let V be a vector space over F. A subspace of V is a subgroup of V admitting multiplication by scalars. Thus U is a subspace of V if U is a subset of V such that (i) Oe U, (ii) x, ye Ux +yeU, and ax€U for all aéF (for on taking a=~—1 we see that —xeU, so the conditions for a subgroup are satisfied). E.g., V is always a subspace of itself; a subspace of V other than V is called a proper subspace. Of course the trivial space is always a subspace of V. Other examples of subspaces: In Example (v) above, R is a subspace of C, in Example (ii) F" has a subspace obtained by restricting the last coordinate to be 0, and in Example (iii) F* is a subspace of FN. The intersection of two (or indeed, any number of) subspaces of a vector space V is again a subspace: if U,, U, are subspaces, then clearly 0é U,NU,, and if x, ye U; (i= 1,2), then x+y and ax belong to U, and hence to U,NU;. A similar argument works for more than two subspaces. On the other hand, the union of two subspaces is not generally a subspace, as a concrete example will quickly show: in Example (i) above, the union of two distinct lines clearly does not form a subspace. However, for any subspaces U, and U, there is a least subspace containing U, and U>, called their sum and written U,+U>. It consists of all vectors u,+uUz (u; € U,); for clearly any subspace containing U, and U, must contain all vectors u,€ U, and u.€ U2, and hence all vectors u,+uz. To show that U, + U, is actually the least subspace containing U, and U, it only remains to show that the set U,+U, is a subspace, a task that may be left to the reader (cf. 3.2 for a similar discussion on subgroups). More generally, the sum of any finite set of subspaces U,,...,U, is U,+-+-+U,= {Ix |x € Uj}. Let V be a vector space and take vectors x,,...,x, in V. Then there is a least subspace containing x,,..., X,, namely the intersection of all subspaces containing x;,...,X,. This space is denoted by (x,,...,X,); here is a more constructive way of forming it: By definition, (x,,...,x,) contains X1,.-+,X,3 being a subspace, it must contain a1X,,..-,Q,%,, for any Q,...,@, €F, and hence also the linear combination a@,x,;+ +++ +a,X,. But the set of all these linear combinations is already a subspace and therefore coincides with (x,,...,X,). Thus Ax, 6625 Xu) = Lexy to FOXy | O1,... 5 Oy EF} E.g., (x) consists of all ax, (a € F) and (), the least subspace containing the empty set, is the zero space 0. The space (x,,...,X,) is called the subspace spanned by x,,...,%,- If41 Vectors and linear dependence 73 {X1,-++5%n) = V, we say that x,,...,x, span V, or that they form a spanning set for V. If x €(x,,..., X,), So that x is a linear combination of x,,...,X,: X=QyXyt + +OXy for some a; € F, (3) we also say that x is linearly dependent on x,,...,X,. E.g., to say that a vector in 3-space is linearly dependent on a pair of vectors a, b means that it lies in a subspace spanned by a and b; this subspace may be a plane, or a line (if a and b are parallel) or a point (if a = b =0). Thus (3, 5, 7) is linearly dependent on the pair (1, 1, 1) and (1, 2, 3), as the equation (3, 5,7) =(1, 1, 1)+2(1, 2, 3) shows; but (3, 5,6) is not, as some trials will convince the reader. Soon we shall meet general methods for testing when a vector is linearly dependent on a given set. Since we shall frequently have to write such expressions as (3) for x, we shall abbreviate it by using the ¥ notation already encountered in Ch. 2. Thus instead of (3) we write x =i a,x; or simply ¥ a,x; or also Y ax, ist when the precise range is not clear from the context. A family of vectors x),..., X, is said to be linearly dependent if there exist scalars a,,...,a@, not all 0, such that a,x,;+---+a,x, =0; otherwise the family is linearly independent. E.g. any family including the zero vector is linearly dependent, for if x, =0, say, then 1.x, +0.x.+---+0.x%,=Oisa non-trivial relation. This holds even if the family consists only of the single vector 0. Similarly any family in which a vector occurs more than once is linearly dependent. On the other hand, the empty set is linearly indepen- dent. We shall often use the following criterion for the linear dependence of vectors: THEOREM 1 A family of vectors x,,...,X, is linearly dependent if and only if (at least) one of the vectors is linearly dependent on the rest. For if one of the xs, say xj, is linearly dependent on the rest: x;= AoXa+++++A,X,, then the relation Xy—AgXyg— ++ —AyxX, =O is non-trivial, because the coefficient of x, is non-zero. Conversely, suppose that we have a non-trivial relation ¥ a,x; =0, say a, #0, then the equation Xp = ay (aQXy t+ ++ +.O%n) shows the linear dependence of x, on X2,...,%,-74 Vector spaces and linear mappings Exercises (1) Which of the following families of vectors are linearly dependent? (i) (1,2), (4,-2), Gi) 1,2), (4,-2), 3,5), (iii) (-2, 1), (4,-2), (iv) G, 1,4), (2,5, -D, (4, -3, 7), (v) (1, 0, 0, 0), (1, 1,0, 0), (1, 1, 1,0), (1, 1, 1, 1), (vi) (3, 1, 2, 3), (2, 1, 3, 2), 3, 1, 2,3), (1, 3,2, D. (2) Test the following sets for linear dependence and, when they are linearly dependent, express each vector when possible as a linear combination of the rest: (i) (1,2, -3, 1), (2,0,4, 1), (5,—4, 14, -3), (ii) (2, 1, 2, D, (6, 3, 6, 3), (5, 1, 4, 3), (iii) (1, 1, 0), (1,0, 1), (0, 1, 1). (3) If x, y, z are linearly independent vectors in a space, show that x+y, y+z, z+x are linearly independent, provided that the field of coefficients has characteristic #2 (so that we can divide by 2). Are x—y, y—z, z—x ever linearly independent? (4) Check the assertions made in Examples (i)—(vi) in the text. (5) Prove the general distributive law for vector spaces: (Y a)(¥ xj) = Liy aux}. (6) Which of the following sets are subspaces of R": (i) all vectors a =(a1,..-, Qn) satisfying a, = 22, (ii) all vectors a satisfying Y a; =0, (iii) all vectors a satisfying Ya; = 1, (iv) all vectors a satisfying Y a? =0. Do any of these answers change if R is replaced by C? (7) Let F be a field and I a non-empty set and define F' as the set of all mappings from I to F. Given two elements f,g¢F", define their sum f+g as the mapping im fi +g (where f,, g are the images of i under f, g) and define af as i+ af,. Show that F' is a vector space over F and verify that Examples (ii), (iii) and (iv) are special cases of this construction. (8) If F is a finite field with q elements, show that F" has q" elements. Show that every non-zero space over an infinite field is infinite. 4.2 Linear mappings For a closer study of vector spaces we need to define their ‘homo- morphisms’, i.e. mappings preserving the vector space structure. For histori- cal reasons they are usually called ‘linear mappings’. Let U, V be vector spaces over the same field F. A mapping 0: U— V is said to be linear if (x+y)0=x0+y0 x,yeU, (1) (Ax)0 = A(x0) AEF. (2) We note that (1) just expresses the fact that 6 is a homomorphism of U into V, regarded as additive groups. Any linear mapping satisfies (yxy t+ + + Oy%q)O = 1(X10) +++ +0n(%9) (EU, EF), (3)4.2 Linear mappings 75 a remark that will often be used. To prove it we note that for n=1 it reduces to (2), and so we may use induction on n. Let n> 1, then (¥ @,x;)0 = (a1x1)6+(X5 a;x;)@ by (1); applying (2) to the first term and induction on n to the sum, we find (Y @,x,)@ = a,(x,0)+¥3 a;(x,0), i.e. (3). In particular, 6 satisfies (Ax + py)0=A(x0)+ u(y) (x, yeU, A, EF). (4) We observe that linear mappings can also be characterized by (4), for, when this holds, we obtain (1) by taking A = w= 1 and (2) by taking y =0. Examples of linear mappings. (i) For any a € F, x + ax is a linear mapping. Note in particular the cases a = 0, 1, which show the zero mapping and the identity mapping to be linear. (i) (&, &) > (&, & &+&), as a mapping of F* > F°. (iii) (&:, &, &) > (E,, &), as a mapping of F*—> F*. Geometrically this represents a projection on the plane é;=0. An important example of a linear mapping is obtained as follows. Let V be a vector space and take any x,,..., x, € V. Then the mapping 6: F" > V defined by the rule (Ei, 0-65 Ea) > Erma bo FE Xn (5) is easily seen to be linear; the verification is left to the reader. Let us denote by e, the element of F" whose ith component is 1 while the others are 0. Then it is clear from (5) that €0 = x; (i=1,...,n). (6) We assert that 6 is the only linear mapping from F" to V with the property (6): for if (6) holds, then by (3), (G+ 8)0= (Léea)o=E seo)=L és and so @ satisfies (5), as claimed. In other words, every mapping of the family (e),...,@,) into V (such as (6)) can be extended to a unique linear mapping of F" into V. This property of the space F” is an instance of the universal mapping property, which we shall meet in many forms later. Exercises (1) Let A be the space of all real-valued functions on R. Which of the following mappings of A into itself are linear? (i) f+ f; where f,(x)=f(x+1), (ii) fof+1, (ili) fr f?, (iv) fr F(x), where f(x) = f(f(x)). (2) Show that a mapping 6: U+> V is linear iff (x+Ay)@=x0+A(y@) for all x, ye U, AEF.716 Vector spaces and linear mappings (3) Show that every homomorphism of vector spaces over Q (regarded as abelian groups) is linear. Do the same for spaces over F,. (4) Given a linear mapping @: U— V, show that the set {x € U | x0 = 0} is a subspace of U. 4.3 Bases and dimension In any vector space V, we define a basis to be a linearly independent spanning set of V. The importance of a basis v,,..., v, is that every vector of V can be written as a linear combination of v,,...,v,, with uniquely determined scalar coefficients. For, given x € V, since the v; span V, we have X= EV ts + Ens qd) for some &€F and, if we also had x=) &v,, then ¥ (& —&))v, =x—-x =0, hence by the linear independence, é{= &. Thus the hohmtalete & in (1) are uniquely determined by x. Conversely, if every x ¢ V has a unique expres- sion (1), then the v; span V and, by uniqueness, no linear combination of the vs is 0, except the trivial one in which all the coefficients are 0; thus the v; are linearly independent. This shows that the existence and uniqueness of the expression (1) for each x € V characterizes a basis of V. For example, the vectors e;,..., ¢, defined in 4.2 form a basis of F", since every vector (,,..., &,) can be uniquely written as Y &¢;. This basis will be called the standard basis of F". Given a vector space V, let us take any vectors x;,...,x,¢€V and consider again the linear mapping @: F" > V defined in (5), 4.2: 0: (Eis. Ge) > D&M. (2) We observe that @ is surjective iff x;,...,x, span V, while it is injective iff X1,...,X, are linearly independent, hence 6 is bijective precisely when X,,...,X, form a basis of V. Thus when V has a basis of n elements, we can use the mapping (2) to establish a linear bijection between F" and V. A linear bijection between vector spaces has the property that its inverse is again linear (as the reader will verify without difficulty). A linear bijection is also called an isomorphism; if an isomorphism exists between spaces U and V, we write U= V and say that the spaces are isomorphic. The importance of this notion is that any abstract property of vector spaces is preserved under isomorphism. The observations made earlier can now be summed up as THEOREM 1 Any vector space over a field F, with a basis of n elements, is isomorphic to F". As in the case of groups, the notion of isomorphism between vector4.3 Bases and dimension 77 spaces is an equivalence relation. Thus if two spaces each have a basis of n elements, for the same n, then both are isomorphic to F" and hence to each other; this proves the COROLLARY Any two vector spaces with bases of the same number of elements are isomorphic. @ Th. 1 shows the importance of the spaces F", but it leaves open two questions: (i) When does a space have a (finite) basis? (ii) Can F™ be isomorphic to F" for m# n? Certainly not every space has a finite basis; as an example we cite the spaces of sequences FN and FY) defined earlier. The reader should have no difficulty in showing that neither has a finite basis (this will in any case follow easily from Th. 5). Of course we have not defined what is meant by an infinite basis, but it is not hard to give appropriate definitions and show that F™’ has a basis (cf. Ex. (8)). As a matter of fact, we shall see in Vol. 2 that every vector space has a basis. The answer to question 2 is ‘no’ (as one would hope), see Th. 5. We begin with an obvious remark which is sometimes useful. We recall (p. 20) that a maximal subset with a certain property P is a subset X having P such that X is not properly contained in any subset with the property P. PROPOSITION 2 If X ={x,,..., X,} is a maximal linearly independent set in a vector space V, then X is a basis of V. Proof. We must show that X spans V, i.e. that every xe V is linearly dependent on x,,...,X,. This is clear if x is one of the x;; otherwise {x, x1,...,X,} contains X as proper subset and so is linearly dependent. Let axt+¥ ax, =0 be a non-trivial relation. If a=0, then by the linear independence of X4,...,X, all the a; must also vanish and the relation is trivial. Hence a#0 and on dividing by a, we can express x as a linear combination of x,,..., Xn as claimed. = We now set about the task of showing that every vector space with a finite spanning set has a basis, and of comparing the different bases of a space. A first (and important) method of getting a basis is given in PROPOSITION 3 Let V be a vector space with a finite spanning set X={x,,...,x,}, and assume that x,,...,X, are linearly independent, for Some k in the range 0O and assume that x1,..., 4-1 Yks--+> Ys Span V. We can then express x, as a linear combination of these4.3 Bases and dimension 79 vectors: Xi = Ky to FO 1 Xe 1 + Bu + + BV @) Since the xs are linearly independent, at least one of the Bs must be non-zero; by renumbering the ys if necessary we may suppose that 6, #0, and then by (3), Vie = Bic (oka ++ Fete aXe He F Beer erst + Bs) (4) Thus y, is linearly dependent on x,,...,Xk, Yeris+-+> Ys hence this set spans V, as we had to show. This shows that for suitable y,€ Y, X1,-...%, Yr+is---» Ys Span V, there- fore r

You might also like

• Geometric Harmonic Analysis I A Sharp Divergence Theorem & NontMathematics, 72) Dorina Mitrea, Irina Mitrea, Marius Mitrea

  

  100% (2)

  Geometric Harmonic Analysis I A Sharp Divergence Theorem & NontMathematics, 72) Dorina Mitrea, Irina Mitrea, Marius Mitrea

  940 pages
• Algebraic Number Theory-J. W. S. Cassels, A. Frohlich

  

  100% (2)

  Algebraic Number Theory-J. W. S. Cassels, A. Frohlich

  392 pages
• B.H. Matzat Inverse Galois Theory

  

  100% (2)

  B.H. Matzat Inverse Galois Theory

  450 pages
• Cassels, Froehlich (Eds) - Algebraic Number Theory (378s)

  

  100% (3)

  Cassels, Froehlich (Eds) - Algebraic Number Theory (378s)

  378 pages
• Szamuely T. Galois Groups and Fundamental Groups (CUP, 2008) (ISBN 0521888506) (276s) - MA

  

  No ratings yet

  Szamuely T. Galois Groups and Fundamental Groups (CUP, 2008) (ISBN 0521888506) (276s) - MA

  276 pages
• Paul M. Cohn Basic Algebra Springer 282004 29

  

  No ratings yet

  Paul M. Cohn Basic Algebra Springer 282004 29

  468 pages
• Commutative Algebra Bourbaki

  

  No ratings yet

  Commutative Algebra Bourbaki

  324 pages
• Weil - Number Theory For Beginners

  

  No ratings yet

  Weil - Number Theory For Beginners

  39 pages
• Method of Real Analysis PDF

  

  No ratings yet

  Method of Real Analysis PDF

  410 pages
• Elliptic Curves, Modular Forms and Fermat's Last Theorem

  

  0% (1)

  Elliptic Curves, Modular Forms and Fermat's Last Theorem

  342 pages
• Cohn Algebra Volume 2

  

  No ratings yet

  Cohn Algebra Volume 2

  439 pages
• Casper Goffman, George Pedrick-First Course in Functional Analysis

  

  86% (7)

  Casper Goffman, George Pedrick-First Course in Functional Analysis

  290 pages
• B L Van Der Waerden Modern Algebra Vol 2 PDF

  

  No ratings yet

  B L Van Der Waerden Modern Algebra Vol 2 PDF

  227 pages
• Cohn Measure Theory

  

  No ratings yet

  Cohn Measure Theory

  384 pages
• Graetzer G. Lattice Theory.. First Concepts and Distributive Lattices (Dover, 2009) (ISBN 048647173X) (600dpi) (T) (237s) MAa

  

  No ratings yet

  Graetzer G. Lattice Theory.. First Concepts and Distributive Lattices (Dover, 2009) (ISBN 048647173X) (600dpi) (T) (237s) MAa

  237 pages
• Topics in Ring Theory (Jacob Barshay)

  

  100% (1)

  Topics in Ring Theory (Jacob Barshay)

  150 pages
• Cyclotomic Polynomials

  

  No ratings yet

  Cyclotomic Polynomials

  13 pages
• Permutation Patterns

  

  100% (2)

  Permutation Patterns

  353 pages
• Linear Algebraic Groups and Their Lie Algebras

  

  No ratings yet

  Linear Algebraic Groups and Their Lie Algebras

  81 pages
• Algebraic Number Theory - Computational Approach PDF

  

  No ratings yet

  Algebraic Number Theory - Computational Approach PDF

  215 pages
• Class Field Theory, Milne

  

  100% (1)

  Class Field Theory, Milne

  230 pages
• Timothy Gowers - Further Mathematical Analysis

  

  No ratings yet

  Timothy Gowers - Further Mathematical Analysis

  53 pages
• Northcott - Ideal Theory

  

  100% (2)

  Northcott - Ideal Theory

  118 pages
• Point-Set Topology With Topics Basic General Topology

  

  No ratings yet

  Point-Set Topology With Topics Basic General Topology

  822 pages
• Chuaqui R.B. Axiomatic Set Theory (NH, 1981) (ISBN 0444861785) (T) (380s) - MAml

  

  100% (1)

  Chuaqui R.B. Axiomatic Set Theory (NH, 1981) (ISBN 0444861785) (T) (380s) - MAml

  380 pages
• (London Mathematical Society Lecture Note Series 44) M. C. Crabb - ZZ - 2, Homotopy Theory (1980, Cambridge University Press)

  

  No ratings yet

  (London Mathematical Society Lecture Note Series 44) M. C. Crabb - ZZ - 2, Homotopy Theory (1980, Cambridge University Press)

  135 pages
• Adler: Random Fields and Geometry

  

  No ratings yet

  Adler: Random Fields and Geometry

  471 pages
• The Theory of Groups

  

  No ratings yet

  The Theory of Groups

  265 pages
• Siegfried Carl, Seppo Heikkilä) Fixed Point Theo (Book4You)

  

  No ratings yet

  Siegfried Carl, Seppo Heikkilä) Fixed Point Theo (Book4You)

  492 pages
• Graduate Texts in Mathematics: S. Axler F.W. Gehring K.A. Ribet

  

  No ratings yet

  Graduate Texts in Mathematics: S. Axler F.W. Gehring K.A. Ribet

  17 pages
• Notes On Class Field Theory

  

  No ratings yet

  Notes On Class Field Theory

  135 pages
• Gaisi Introduction To Axiomatic Set Theory

  

  100% (1)

  Gaisi Introduction To Axiomatic Set Theory

  258 pages
• Dirichlet's Unit Theorem PDF

  

  No ratings yet

  Dirichlet's Unit Theorem PDF

  11 pages
• Metric Spaces of Non-Positive Curvature: Martin R. Bridson Andr e Haefliger

  

  No ratings yet

  Metric Spaces of Non-Positive Curvature: Martin R. Bridson Andr e Haefliger

  662 pages
• Set Theory

  

  50% (2)

  Set Theory

  255 pages
• (Evans L., Thompson R.) Introduction To Algebraic

  

  100% (1)

  (Evans L., Thompson R.) Introduction To Algebraic

  248 pages
• Geometric Harmonic Analysis III - Dorina Mitrea & Irina Mitrea & Marius Mitrea

  

  100% (1)

  Geometric Harmonic Analysis III - Dorina Mitrea & Irina Mitrea & Marius Mitrea

  980 pages
• Open Problems in Topology by Jan Van Mill

  

  No ratings yet

  Open Problems in Topology by Jan Van Mill

  642 pages
• Functional Analysis I PDF

  

  100% (2)

  Functional Analysis I PDF

  286 pages
• Advanced Set Theory - J.D. Monk

  

  100% (3)

  Advanced Set Theory - J.D. Monk

  420 pages
• Emanuel Parzen Modern Probability Theory and Its Applications

  

  100% (2)

  Emanuel Parzen Modern Probability Theory and Its Applications

  480 pages
• Jech Set Theory Solutions Kindle

  

  No ratings yet

  Jech Set Theory Solutions Kindle

  29 pages
• M. H. A. Newman - Elements of The Topology of Plane Sets of Points (1964) PDF

  

  No ratings yet

  M. H. A. Newman - Elements of The Topology of Plane Sets of Points (1964) PDF

  110 pages
• Weil A.-Number Theory For Beginners

  

  No ratings yet

  Weil A.-Number Theory For Beginners

  37 pages
• D.H. Fremlin - Measure Theory - Measure Algebras (Vol. 3) (2002)

  

  No ratings yet

  D.H. Fremlin - Measure Theory - Measure Algebras (Vol. 3) (2002)

  672 pages
• Matsumura Commutative Algebra

  

  No ratings yet

  Matsumura Commutative Algebra

  54 pages
• (Graduate Texts in Mathematics) Serge Lang - SL2 (R) With 33 Figures-Springer Science & Business Media (1985)

  

  100% (2)

  (Graduate Texts in Mathematics) Serge Lang - SL2 (R) With 33 Figures-Springer Science & Business Media (1985)

  432 pages
• (Ergebnisse Der Mathematik Und Ihrer Grenzgebiete) Bosch, Siegfried, Lütkebohmert, Werner, Raynaud, Michel - Néron Models-Springer Verlag (1990) PDF

  

  No ratings yet

  (Ergebnisse Der Mathematik Und Ihrer Grenzgebiete) Bosch, Siegfried, Lütkebohmert, Werner, Raynaud, Michel - Néron Models-Springer Verlag (1990) PDF

  336 pages
• Charalambos D. Aliprantis, Owen Burkinshaw - Problems in Real Analysis - A Workbook With Solutions-Academic Press (1999)

  

  No ratings yet

  Charalambos D. Aliprantis, Owen Burkinshaw - Problems in Real Analysis - A Workbook With Solutions-Academic Press (1999)

  412 pages
• Artin - Galois Title

  

  No ratings yet

  Artin - Galois Title

  2 pages
• Stichtenoth-Algebraic Function Fields and Codes-2008

  

  100% (1)

  Stichtenoth-Algebraic Function Fields and Codes-2008

  363 pages
• Matrix Combinatorics and Algebra

  

  100% (1)

  Matrix Combinatorics and Algebra

  320 pages
• 9780203755419

  

  100% (1)

  9780203755419

  431 pages
• A Classical Introduction To Modern Number Theory

  

  No ratings yet

  A Classical Introduction To Modern Number Theory

  13 pages
• Second Course in Linear Algebra

  

  100% (2)

  Second Course in Linear Algebra

  36 pages
• On The History of Geometrization From Minkowski To Finsler Geometry H Goenner - Ps

  

  No ratings yet

  On The History of Geometrization From Minkowski To Finsler Geometry H Goenner - Ps

  24 pages
• (Mostly) Commutative Algebra

  

  100% (2)

  (Mostly) Commutative Algebra

  480 pages
• A Complete Solution Guide To Complex Analysis 3rd Bak

  

  No ratings yet

  A Complete Solution Guide To Complex Analysis 3rd Bak

  253 pages
• Abstract Algebra - An Introduction With Applications (2ed.) by Derek J. S. Robinson 2015

  

  No ratings yet

  Abstract Algebra - An Introduction With Applications (2ed.) by Derek J. S. Robinson 2015

  347 pages
• Notices of The AMS February 2022 Article

  

  No ratings yet

  Notices of The AMS February 2022 Article

  10 pages
• Kurosh - Theory of Groups PDF

  

  No ratings yet

  Kurosh - Theory of Groups PDF

  272 pages
• Cardinalities of Infinite Antichains in Products of Chains: Mailbox

  

  No ratings yet

  Cardinalities of Infinite Antichains in Products of Chains: Mailbox

  4 pages
• Evolutionary Dynamics of Bee Colony Collapse Disorder: First Steps Toward A Mathematical Model of The Contagion Hypothesis

  

  No ratings yet

  Evolutionary Dynamics of Bee Colony Collapse Disorder: First Steps Toward A Mathematical Model of The Contagion Hypothesis

  7 pages
• Strictly Order-Preserving Maps Into Z, I: A Problem of Daykin From The 1984 Banff Conference On Graphs and Order

  

  No ratings yet

  Strictly Order-Preserving Maps Into Z, I: A Problem of Daykin From The 1984 Banff Conference On Graphs and Order

  23 pages