Rowan Garnier and John TaylorPage iii 100% Mathematical Proof Rowan Garnier Richmond College, London John Taylor University of Brighton, Sussex JOHN WILEY & SONS, LTD Chichester « New York « Brisbane « Toronto # Singapore ‘Stat of Citation[PU Bohn Wiley & Sons, Lid. (UK)|/PU I[DP]1996/DP JEad of CitationPage iv Copyright © 1996 by John Wiley & Sons Ltd, Baffins Lane, Chichester, West Sussex PO19 IUD, England National (01243) 779777 International (+44) 1243 779777 Alll rights reserved. No part of this book may be reproduced by any means, or transmitted, or translated into a machine language without the written permission of the publisher. Other Wiley Editorial Offices John Wiley & Sons, Inc., 605 Third Avenue, New York, NY 10158-0012, USA Jacaranda Wiley Ltd, 33 Park Road, Milton, Queensland 4064, Australia John Wiley & Sons (Canada) Ltd, 22 Worcester Road, Rexdale, Ontario M9W 1L1, Canada John Wiley & Sons (Asia) Pte Ltd, 2 Clementi Loop #02-01, Jin Xing Distripark, Singapore 0512 British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN 0 471 96198 1; 0.471 96199 X (pbk) Typeset in 11/13pt Palatino by Laser Words, Madras, India Printed and bound in Great Britain by Bookcraft (Bath) Ltd. ‘This book is printed on acid-free paper responsibly manufactured from sustainable forestation, for which at least two trees are planted for each one used for paper production. Start of Citaion|PU ohn Wiley & Sons, Lid. (UK)|/PUI[DP}1996)/DP JEad of CitationContents Preface 1 Proofs, Mathematical and Non-Mathem: 1,1 Introduction 1.2 Inductive and Deductive Reasoning 1.3 A Proof or Not a Proof? 2 Propositional Logic 2.1 Propositions and Truth Values 2.2 Logical Connectives 2.3 Tautologies and Contradictions 2.4 Logical Implication and Logical Equivalence 2.5 Arguments and Argument Forms 2.6 Formal Proof of the Validity of Arguments 2.7 The Method of Conditional Proof 3 Predicate Logic 3.1 Introduction Page v3.2. Quantification of Propositional Functions 6 3.3 Two-Place Predicates un 3.4 Validation of Arguments in Predicate Logic 2» 4 89 Axiom Systems and Formal Proof 4.1 Introduction 39 4.2 Case Study of a Proof 0 43 Axiom Systems 93 4.4 Theorems and Formal Proofs 105, ja 4.5 Informal Proofs 45 Direct Proof 5.1 The Method of Direct Proof 5.2 Finding Proofs 5.3 More Advanced Examples 6 Direct Proof: Variations 6.1 Introduction 6.2 Proof Using the Contrapositive 6.3 Proof by Contradiction 6.4 Proof of a Biconditional 7 Existence and Uniqueness Proofs 7.1 Introduction 7.2 Proof by Construction 7.3 Non-Constructive Existence Proofs 7.4 Use of Counter-Examples 7.5 Uniqueness Proofs 8 Further Proof Techniques 8.1 Introduction Page vi0.2 Proofs of Identities 8.3 Use of Counting Arguments 8.4 The Method of Exhaustion 9 Mathematical Induction 9.1 The Principle of Mathematical Induction 9.2 The Second Principle of Mathematical Induction Appendix: Some Definitions and Terminology References and Further Reading Hints and Solutions to Selected Exercises Index 265 23 2s a EPreface Although there have been suggestions that mathematics is becoming more ‘experimental,’ and the notion of proof less salient, this has provoked vigorous denials from the mathematical community which continues to maintain that proof is one of the key concepts which characterise the discipline. It is surprising therefore, that the fundamentals of mathematical proof have rarely been taught in a systematic way. Most students of mathematics are expected to develop their understanding of proof and the associated theorem-proving skills by a process of ‘osmosis’ through encounters with the various techniques and methods. The result is that students frequently have an inadequate appreciation of the underlying structure of proofs and a consequent inability to distinguish a correct proof from a flawed one. It is therefore no wonder that they have considerable difficulty in constructing their own proofs, often not knowing how to start a suitable line of reasoning. Because it has not been the custom to teach the principles of mathe- matical proof systematically, there are few, if any, books on ‘structured theorem proving’. As a consequence some authors of textbooks in abstract algebra and analysis, for example, have found it necessary to include appendices outlining some of the principal methods of proof. Naturally, such summaries tend to concentrate on describing proof techniques without looking too deeply at their underlying structure. On the other hand, the notion of ‘formal proof’ is covered well in many books on mathematical logic, but the proofs that mathematicians write are not formal proofs in this sense. It is our intention in this book to explore the principles which underpin the various methods of mathematical proof and to describe how proofs may be discovered and communicated. We aim to examine thevili Preface structural features common to all mathematical proofs as well as those which are specific to particular techniques. We also consider some of the less tangible skills associated with the discovery and communication of proofs. These aspects are, of course, explored at greater depth in Polya’s much admired book (Polya, 1957). Whilst our primary aim has been to write a book for students of math- ematics, we hope this text will also be useful to others interested in unravelling and understanding the nature of mathematical proof—for instance, teachers of mathematics or students of philosophy. The math- ematical background we assume is little more than would be provided by a GCSE course in the UK or a high-school algebra course in the USA, although some degree of mathematical sophistication beyond this is necessary. We have included, in examples and exercises, some proofs drawn from more advanced mathematics but these can be omitted without jeopardising the comprehensibility of the text. University mathematics departments are becoming increasingly concerned about their students’ inability to understand and write mathematical proofs so that a number of institutions are introducing courses in which proof is taught systematically. We hope that this text will prove useful to those designing, teaching and studying such courses, Of course, not everyone will agree with our approach to proof and how (indeed, whether) it should be taught. However, we do believe there is a need to address the question of ‘proof education’ more directly than has hitherto been the case and we hope this book will make some contribution. Our sincere thanks are due to those colleagues who commented on various parts of the manuscript, to Clifford Mould and Elizabeth Taylor for their continued support, to Alice Tomié for her hospitality following many ‘authors’ meetings’ and to Pam Taylor for, once again, providing excellent cartoons on a less than promising subject at short notice. RG and JT Me pa Opewwnt August 19951 Proofs, Mathematical and Non-mathematical 1.1 Introduction This book is unusual so far as mathematics textbooks are concerned in that its primary purpose is not to teach any specific body of mathe- matics. Instead, the text considers mathematics itself in an attempt to understand its modus operandi. In particular, we shall explore the notion of rigorous proof which is unique to mathematics and logic. (It could be argued that rigorous proofs are also employed in software engineering where so-called ‘safety critical’ programs need to be proved correct. For the purposes of this book, these aspects of software engineering are regarded as being part of mathematics itself.) The mathematician’s concept of proof is rather different from, say, the lawyer's. In criminal law, the prosecution seeks to prove ‘beyond reasonable doubt’ that the accused is guilty of the alleged offence. Since the courts deal with people and events in the real world, the criterion of ‘proof beyond reasonable doubt’ is the most rigorous practical require- ment. Rarely, if ever, can guilt be established beyond all doubt. By contrast, the mathematical notion of proof is (in principle) far more rigorous. We do not speak, for example, of Pythagoras’ theorem being true ‘beyond reasonable doubt.’ This is not to say that mathematical proofs are completely reliable; they are not. In practice, proofs do not always conform to the ideal standard because they are constructed by fallible human beings. However, in principle, there can be no room for doubt in a formal mathematical proof. Our purpose in this book is two-fold. One aim is to explain what mathe- matical proofs are so that they may be better understood. Since proofs are written by human beings (or in a very few limited situations byProofs, Mathematical and Non-mathematical computer programs which are themselves written by human beings) there are many different styles and approaches adopted. Proofs are written for a variety of different audiences in various cultures and in many languages. We shall attempt to delve beneath surface features such as style and language to examine the underlying logical struc- tures of proofs. This aspect of the book could be summarised as ‘proofs dissected, examined and their structure revealed’. Proofs dissected, examined and their structure revealed Our second goal is more difficult to achieve. Put simply, it is to show how to construct proofs. On one level, such a goal is unattainable. There is no magic formula which, if learned, will enable us to construct a proof of any result we care to choose. However, we can provide hints and guidance which we hope will be useful. Traditionally, there has been little systematic attempt to teach ‘theorem-proving’. Instead, it is expected that proof techniques will be absorbed simply through repeated exposure to examples of proofs. We hope this text will go some way to redressing the balance. 1.2 Inductive and Deductive Reasoning Most readers would, no doubt, be able to give an informal and reasonably accurate description of what is, say, physics or psychology, astronomy or anthropology. The educated person’s definition of any of these disciplines is likely to be fairly close to the accepted definition of the experts in the field. But, what of mathematics? What is mathematics? And what is it that professional mathematicians do whenInductive and Deductive Reasoning 3 they are engaged in their discipline? Perhaps a reasonable brief job description of the (pure) mathematician is someone who proves things about abstract objects such as numbers or geometrical configurations, their interrelations and their generalisations. We can safely assume that our readers will know that in mathematics we prove things. It is really the notion of rigorous proof which distinguishes mathematics from other fields of study. In most academic disciplines, theories are proposed and explored, evidence for and against is accumulated, differing opinions are expressed and so on. It is true that certain facts are established: Christopher Columbus discovered the New World in 1492, the metal copper will conduct an electric current, the majority of the Earth's surface is covered in water, etc. But such facts are discovered by observation and collation of data and not because they have been proved in the sense that mathematicians use the word. In science, theories are judged on the basis of how well they explain and predict observable phenomena; in other words, by how well they ‘fit’ experimental data. The scientist draws on a mass of observations to make inferences which become the building blocks for new theories which are tested and then modified in the light of further experimental data. The method of reasoning which makes inferences and draws conclusions from observations is known as inductive reasoning. There are limits to the power of inductive reasoning, however. It is sometimes said that scientific theories are not provable, they are only falsifiable. At best, experimental observations can be consistent with a particular theory and give the scientist greater confidence in it. However, no amount of experimental data can prove that a theory is correct, because there is always the possibility that another experiment will turn out to be inconsistent with the theory and hence show it to be false. (We are assuming here that experiments accord with the rigorous standards of scientific practice. One important criterion, for example, is that experimental results should be essentially repeatable. A single experiment could produce freak results and a successful theory would not be modified on the basis of a single unrepeatable experiment.) In short, data which agree with theoretical predictions increase confidence in the theory whereas data which disagree with theoretical predictions destroy the theory. It is not only scientists who employ inductive reasoning—it is prob- ably the basis of most human belief and knowledge. To illustrate theProofs, Mathematical and Non-mathematical point, we would all presumably accept as a fact that the Sun will rise tomorrow morning. (We ignore the problem that we may not be able to see the Sun rise due to climatic conditions. Equally we ignore possi- bilities such as being situated north of the arctic circle in mid-winter.) Our belief in this ‘fact’ is based on inductive reasoning: we and our forebears have observed the Sun to rise on many thousands of morn- ings. Indeed, we have never known the Sun not to rise! In this way, we have established ‘beyond all reasonable doubt’ that the Sun will rise tomorrow morning. In mathematics, on the other hand, conclusions based solely on obser- vation are not sanctioned. Thus inductive reasoning is not acceptable in a mathematical proof. For instance, we may observe that when- ever we square an odd positive integer (whole number) the result is always another odd positive integer. (For example, 3? = 9, 19? = 361, 321? = 103041, etc.) But, no matter how many times we perform the ‘experiment’ and obtain the expected outcome, this will not consti- tute a mathematical proof that the square of an odd positive integer is also an odd positive integer. The reasoning acceptable in a mathemat- ical proof is of a different kind altogether. It is deductive reasoning, whereby a conclusion is reached by logical inference from a collection of assumptions. Using correct deductive reasoning, we can be confident that a conclu- sion does indeed follow from the assumptions in force at the time. One can immediately appreciate the appeal of deductive reasoning when compared to inductive reasoning, It appears to offer us certainty. However, two words of caution are in order. The first is that any conclu- sion obtained deductively can only be as ‘sound’ as the premises on which it is based. Deductive reasoning allows us to pass with confi- dence from assumptions to conclusions but any such conclusion will be useless if it is based on incorrect assumptions. If the assumptions are false then we cannot guarantee the truth of any conclusion deduced from them. It is said that a chain is only as strong as its weakest link. Similarly, a conclusion obtained by deduction is only as ‘strong’ as the assumptions upon which it is based. There is a more theoretical reason why we must temper any enthusiasm for deductive arguments, though. In the 1930s, the Austrian logician Kurt Gédel showed that there are certain limitations to the power and scope of deductive reasoning. Gédel’s results, known as his incom- pleteness theorems, are amongst the most profound of the twentiethA Proof or Not a Proof? 5 century. In one of the theorems, Gédel showed that there must always be true results about the arithmetic of the positive integers which we will never be able to prove using strict deductive reasoning. In other words, even if deductive arguments offer some form of certainty, that certainty can never extend to include all true facts about a system as familiar to us as the positive integers. Despite these limitations to the power of deduction, it is the basis of the vast, powerful and applicable body of human knowledge known as mathematics. It is our purpose in this book to explore the inner workings of this discipline which is, regrettably, poorly understood by that mythical being, the person in the street. 1.3 A Proof or Not a Proof? In much of the remainder of the book we shall be illustrating various techniques of proof using fairly short, elementary proofs. Naturally, in ‘real’ mathematics, not all proofs are like this. In this section, we shall examine briefly three well-known theorems whose proofs are neither short nor elementary. Each of the theorems, or more accurately their proofs, will tell us something about the nature of mathematical proof. In many areas of human endeavour, actual practice does not always conform to some theoretical ideal. So, too, it is in mathematics. Although none of our three examples could be regarded as a typical mathematical proof, the lessons we can learn from them are relevant. They serve to keep in perspective the main thrust of this book which is to explore the theoretical framework of proof. The choice of the three theorems to consider is a personal one—there are many we could have examined. Each theorem is either well-known, has an interesting history or is remarkable in its own right. Each proof is particularly demanding in some aspect; each is a remarkable achievement worthy of examination. Fermat's last theorem Pierre de Fermat (1601-1665) was a French jurist and amateur mathe- matician who made many significant contributions to number theory. The statement which was to become known as ‘Fermat's last theorem’ originates from a note Fermat made in the margin of his copy of the works of the ancient Greek mathematician Diophantus. Having statedProofs, Mathematical and Non-mathematical the ‘theorem’ Fermat wrote, ‘I have discovered a truly remarkable proof which this margin is too small to contain’. It would appear that Fermat did not write down his ‘remarkable proof’ elsewhere either, as he never communicated a proof to anyone, nor was a proof found amongst his papers. In fact, Fermat rarely wrote down proofs of his discoveries although virtually every one has subsequently been proved by others. Fermat's last theorem can be understood by anyone who has studied a little elementary algebra. We all know Pythagoras’ theorem. Symbol- ically, it states that the sides x, y and hypotenuse z of a right-angled triangle satisfy the equation x? + y? = z”. Furthermore, there are known to be many solutions (infinitely many, in fact) of this equation where x, y and z are integers. (Examples of solutions include (x,y,z) = (3, 4,5), (5, 12, 13), (517, 1044, 1165).) Fermat's last theorem states that, for n > 2, the corresponding equation x" + y" =z" has no solutions where x, y and z are integers. It is not surprising that such an easy-to-state theorem about familiar objects, the integers, has attracted the attention of many mathemati- cians, both professional and amateur. And yet, for some 350 years the theorem defied all attempts at a proof (and all attempts at a disproof, too). It earned the status of the most famous unsolved problem in math- ematics. Lack of success in finding a proof or disproof was not for the want of trying, though. Many thousands of person-hours have been devoted to the problem since Fermat's original marginal note. In the nineteenth century, the Académie de Sciences de Paris twice offered a prize for a solution to the problem. Later, in 1908, under the terms of the will of Dr. Paul Wolfskehl, a prize of 100000 Marks was offered for a proof of the theorem. The prize was to be conferred by the Kénigliche Gellschaft der Wissenschaften in Géttingen on or before September 13, 2007. In the early years of the Wolfskehl Prize several hundred attempts were received each year. In recent years, this has dwindled to a few dozen. Over the years inflation and financial charges have taken their toll on the value of the prize too; it is now worth a little over DM 10000, a fraction of its original value. Of course, the importance of Fermat's last theorem is not to be measured in mone- tary terms. In purely mathematical terms, too, the theorem itself is not of great importance, although a great deal of significant mathematics has been developed in the search for a proof. The importance of the theoremA Proof or Not a Proof? 7 lies in its fame, the fact that it turned out to be inordinately difficult to prove and, most importantly, in the mathematical spin-offs resulting from attempts to prove the theorem. By the 1980s much had been achieved, although a proof was still not in sight. For example, it was known that the theorem was true for all values of the index n less than 125000. Furthermore, if the equation x" +y" =z" did have a solution (x,y,z) for some value of n, then the number x would have to be at least 10'809, 4 truly unimaginably huge number which would take several hundred pages just to write down in the usual decimal notation. In 1983 a German mathematician Gerd Faltings proved that the equation had at most a finite number of solutions where x, y and z have no common factors!. Faltings’ theorem was the first major step towards a proof for several decades although it was still far from a proof—proving that there are only a finite number of solutions is a long way from showing that there are none. Given the long history of unsuccessful attempts it was with much delight and surprise (and not a little scepticism) that the mathematical community received the news in June 1993 that a British mathematician, Andrew Wiles, had finally succeeded in proving Fermat's theorem. (Actually, Wiles had proved a highly technical theorem, known as the Shimura-Taniyama-Weil conjecture; however Fermat's last theorem had previously been shown to follow from this result.) The ideas used in the proof are very deep and complicated—a far cry, indeed, from the simplicity of the last theorem itself. Until Wiles’ arguments have been thoroughly examined and understood, there remains the possi- bility that there is a flaw somewhere in his reasoning. However, the experts agree that any flaw is likely to be minor and relatively easily corrected. It appears, then, that mathematics’ most famous unsolved problem has become one of its most celebrated solved problems. (In fact, a few months after announcement of the proof a flaw was discov- ered. However, in October 1994, Wiles and a colleague, Richard Taylor, issued a manuscript which appears to have repaired the gap in the orig- inal proof. At the time of writing—summer 1995—it is believed that the proof is now complete.) What are the major lessons we can learn from the history of Fermat's last theorem? An obvious lesson is that theorems which are simple to " Faltings actually proved a result known as the Mordell conjecture, which implies the stated result. For his work in this area, Faltings was awarded the Fields Medal in 1986. The Fields Medal is mathematics’ highest prize, equivalent in status to, if not so well known as, the Nobel prizes.Proofs, Mathematical and Non-mathematical state may not be simple to prove. Fermat's last theorem is probably the supreme example—a little algebra is all that is needed to understand the statement of the theorem, but it defeated mathematicians for three and a half centuries. Another important ‘moral’ is that failure to produce a proof does not necessarily mean outright failure. It is undoubtedly the case that the cause of number theory has been very well-served by Fermat's last theorem. Much interesting and useful mathematics has resulted from some of the unsuccessful attempts to prove the theorem. In less mathematical terms, it is clear that the fame of a theorem is not related to its importance. Naturally, it is difficult to define precisely what makes a particular theorem important. Among the criteria are its applicability within and outside mathematics, whether it provides new insights, to what extent it paves the way for further work, and so on. Most mathematicians would agree that there are more important unsolved problems (the Riemann hypothesis, for instance) which are completely unknown outside the mathematical community. Fermat's last theorem captured the imagination of generations primarily for non- mathematical, even romantic, reasons. Important among these was, of course, the possibility that an amateur mathematician would beat the experts in the race to discover a proof. The four-colour theorem Like Fermat's last theorem, the four-colour theorem is simple to state and, as it transpired, extraordinarily difficult to prove. Also like Fermat's theorem (or will it now be the Fermat-Wiles theorem?), it became very well-known and defied proof for a considerable period of time. In contrast to Fermat's last theorem, the proof, when it eventually appeared in 1976, was not greeted with universal acclaim and delight in the mathematical community. Indeed, the proof sparked a vigorous debate amongst mathematicians about the very nature of proof itself. Consider a map of countries drawn on the plane or the surface of a sphere. Is it always possible to colour the map in such a way that two countries which share a common border are coloured differently and to do so using only four colours? The four-colour theorem says that it is always possible. The question of whether four colours are always sufficient to colour a map in this way originated with a young mathematician, FrancisA Proof or Not a Proof? 9 Guthrie, in 1852. Via Guthrie’s brother, the problem was drawn to the attention of Augustus De Morgan, then Professor of Mathematics at the University of London. Unable to prove the theorem, De Morgan passed the problem on to fellow mathematicians, but it did not gain widespread attention until 1878 when Arthur Cayley asked for a proof at a meeting of the London Mathematical Society. Within a year, a barrister called Alfred Kempe had published a ‘proof’ which was to be accepted for eleven years. In 1890, Percy Heawood pointed out a fatal flaw in Kempe’s ‘proof’. Heawood was able to salvage enough of Kempe’s argument to prove that five colours would always be enough to colour a map in the appropriate manner. Indeed, Heawood gener- alised the problem and considered maps drawn on other surfaces with ‘handles’ and ‘twists’—see Figure 1.1. Heawood conjectured a formula for the number of colours which are sufficient to colour a map on any such surface (excluding the sphere and plane). Subsequently it was shown that Heawood’s formula does indeed give the minimum number of colours required for all these more complicated surfaces except one, the so-called Klein bottle. Although Kempe’s argument turned out to be flawed, the ideas contained in his unsuccessful attempt were to be the basis of the subsequent work on the problem, including the final proof itself Despite much work on the problem in the first half of the twentieth century, there was little real progress; until 1976 that is, when Kenneth Appel and Wolfgang Haken, working at the University of Ilinois, announced that they had proved the theorem. Appel and Haken’s proof, though, is very unusual in that it used some 1200 hours of computer time to examine thousands of possible configurations of countries. It must be emphasised that they were not using inductive reasoning. Their argument was definitely not along Sphere with handle Mébius band Klein bottle Figure 1.110 Proofs, Mathematical and Non-mathematical the following lines: we have examined thousands of possible maps and have {found that all are four-colourable, therefore four colours must surely always be sufficient. As we have mentioned, such reasoning is not acceptable in a mathematical proof. (Their proof was deductive, but based on a method confusingly called ‘proof by mathematical induction’—see Chapter 9.) Nevertheless, the quantity of direct computation required by their method was such that the use of a large amount of computer time was essential. Thus the correctness of their proof could not be checked by ‘hand calculation’ alone. This was the first example of a proof which used the computer in an essential way and it generated considerable debate amongst math- ematicians about its acceptability. The mathematical community had been used to proofs which could be verified by direct examination of the arguments. Some simply refused to accept that Appel and Haken’s work did amount to a proof of the four-colour theorem. It is true that software used in the proof, though complicated, could be checked by others—but what of the occasional hardware errors to which every machine is prone? Could not one such error render the proof useless? In defence of the proof, it was argued that the possibility of error exists in any long and complicated proof and that the likelihood of computer error was considerably less than that of human error. Since its first appearance, Appel and Haken’s proof has come under considerable scrutiny and as a result it is now generally accepted. There are a number of lessons to be learnt from the history and even- tual proof of the four-colour theorem. Again we see a situation where an incorrect ‘proof’ was not worthless. Although Kempe’s attempt was unsuccessful, the underlying ideas were useful. They were the starting point for much subsequent work as they were extensively modified and extended over the years. Some of the important strands of the final proof can trace their ancestors back to Kempe’s original incorrect ‘proof’. Of course, Kempe’s attempt was widely accepted for a number of years. Such a situation is not uncommon—the level of complexity of many proofs is such that errors may lie undetected for a consider- able time. It may be that the four-colour theorem is the first of many examples of theorems whose proof will involve computers in an essential way. Certainly Appel and Haken believed this to be the case although to date there have been no further examples of major theorems of this type. However, as hardware becomes increasingly powerful and softwareA Proof or Not a Proof? n increasingly sophisticated, few would say with confidence that the four- colour theorem is a unique example of this kind of theorem The classification of finite simple groups Classification theorems must rank amongst the most satisfying of all mathematical theorems. Given any class of mathematical objects, an obvious question is: what examples are there? A classification theorem answers this question in the most complete possible way. It provides a list (often infinite) of all the examples of the particular mathemat- ical object. In a sense, a classification theorem says two things—these (the objects in the list) are all examples of the particular class of object and there are no others. It is not only in mathematics where classifica- tion ‘theorems’ are important. For instance, when physicists discover that matter is composed of fundamental particles, they want to know precisely what such particles there are. Exactly which particles are regarded as fundamental has changed over time—in the early part of this century, atoms were regarded a fundamental whereas now quarks and leptons are given this status. Unfortunately, in mathematics classification theorems are all too rare. Our aim now is to consider some features of a particularly remarkable classification theorem, that for the class of objects called finite simple groups. Unlike the two previous theorems, we cannot give a precise statement of the classification theorem of finite simple groups. However, we can give an idea of what the theorem is saying. A ‘group’ is a particular Kind of algebraic structure—it comprises a set with a binary operation (like addition, multiplication or composition of functions) satisfying three or four simple properties. (A definition is given in the appendix.) It turns out that some groups can be ‘broken down’ or ‘factored’ into simpler pieces; on the other hand, there are various ways of combining, two or more groups to form new groups. A loose analogy is often used here: some integers (greater than 1) can be factored into the product of smaller ones; similarly, multiplying two or more such inte- gers together produces a larger integer. Molecular physics provides a non-mathematical analogy. Molecules can be split into atoms of various kinds and atoms can combine together to form complex molecules. In the number theory analogy, those numbers which cannot be factored are, of course, the prime numbers. Similarly, in molecular physics it is12 Proofs, Mathematical and Non-mathematical the atoms which cannot be split into smaller entities of a similar nature. There is an analogous concept in group theory. A simple group is one which cannot be ‘factored’ into the ‘product’ of two groups. Note that the word ‘simple’ here has a technical meaning. It most definitely does not mean that such groups have a simple structure, as we shall see. The finite simple groups are the basic building blocks for all finite groups in the same way that prime numbers are the building blocks for all positive integers and atoms are the basic building blocks for molecules. To classify finite simple groups, we need to know exactly what exam- ples there are. What, then, is the complete list of finite simple groups? Firstly, we can identify various ‘families’ of simple groups, each family containing groups of the same kind but with different numbers of elements. In fact there are 18 such families, each containing an infinite number of different groups. Then, curiously, there are 26 finite simple groups which do not fit into any of the families. These ‘outsiders’ are called the sporadic groups. They range considerably in size from the smallest which has 7290 elements to the largest which has about 8 x 10°3 elements, more elements than there are atoms in the Earth! (This latter group, which is known to group theorists as ‘the monster’ or ‘the friendly giant’ can be represented as a certain group of matrices of dimension 196883 x 196 883!) That is the complete list: 18 infinite families and 26 sporadic groups. Already, we can appreciate something of the scale of the theorem, but the proof is even more remarkable. When the proof was finally completed in 1981, it was estimated to run to between 10000 and 15000 pages spread over many articles published in mathematical journals during the previous four decades or so. Many mathematicians had contributed to the proof and (we need hardly add) no one person had read the whole proof. In contrast to the four-colour theorem, though, computers were used very little in the overall proof What does the existence of such a monumental proof tell us about the nature of proof itself? The most startling point to realise is that when the proof was finally completed, it was almost certainly wrong! With such a large and complicated proof, the chance that it was error-free was very small indeed. To quote Michael Aschbacher (1981), one of the major contributors to the proof, writing at the time: The probability of an error in the Classification Theorem is virtu- ally 1. On the other hand the probability that any single error cannotA Proof or Not a Proof? 13 easily be corrected is virtually zero, and as the proof is finite, the probability that the theorem is incorrect is close to zero. As time passes and we have an opportunity to assimilate the proof, that confidence level can only increase. Aschbacher’s comments seem to run counter to the view of mathematics as a precise, deductive science. It appears odd, to say the least, for a highly regarded professional mathematician to be writing about the probability that a theorem is correct and our confidence level in its correctness. Where is the certainty of deductive reasoning? In reality, Aschbacher is only stating the obvious, Mathematics is a human endeavour and human beings make mistakes; therefore a very large mathematical enterprise is almost certain to contain errors. In other words, real mathematics like the classification theorem for finite simple groups, remarkable though it certainly is, does not always live up to the austere standards of the ideal. Human beings make mistakes The monumental nature of the proof is also testimony to the persever- ance, dedication and collaborative spirit of those involved. Without the collaboration of many mathematicians in several countries the venture would not yet have been brought to a successful conclusion. There are no immediate applications of the theorem which will benefit mankind— this was a search for knowledge for its own sake. Whether or not one regards that as a noble aspiration, it is impossible not to admire the achievement itself. There is one further lesson we can learn from our discussion of these three famous theorems. In practice mathematical proofs do not always conform to the ideal of a completely rigorous, logical argument. Perhaps14 Proofs, Mathematical and Non-mathematical a better description of most mathematical proofs is that of a plau- sible argument sufficient to convince the mathematical community of the truth of the particular theorem. Whether we regard a mathemat- ical proof as completely rigorous or ‘merely’ an argument of sufficient power and persuasiveness to convince the experts, there are standard techniques and methods which are employed. It is the purpose of the remaining chapters to explore and understand these methods.2 Propositional Logic 2.1 Propositions and Truth Values We shall consider the detailed structure of a mathematical proof later but, very broadly speaking, constructing a proof consists of showing that, given the truth of certain statements, the truth of the theorem to be proved inevitably follows. Normally such a proof takes the form of a sequence of statements whose truth is guaranteed either by the truth of earlier statements in the sequence or because they follow from other statements whose truth is assumed. For example, if we accept that ‘Today is Tuesday and it is raining’ is a true statement then we could not dispute the truth of the statement ‘It is raining’. ‘It is raining’ follows from (or is implied by) ‘Today is Tuesday and it is raining’ In claiming that the truth of one statement is guaranteed by the truth of others we shall need to supply some justification. The only justification which is acceptable in a mathematical proof is one which is sanctioned by the laws of logic. It is these laws which govern what can be deduced from what and as such they provide us with a means of distinguishing a proof which is valid from one which has some fault in the sequence of steps which purports to establish the inevitable truth of the ‘theorem’ The statements which appear in a mathematical proof are ones which can (under appropriate circumstances) be declared true or false. We refer to such statements as propositions and we denote particular propositions using upper case letters, e.g. P,Q,R,.... We use lower case letters (e.g. p,q,1,--.) to denote propositional variables, i.e. vari- ables for which any proposition may be substituted. The following are examples of propositions: (a) P: Two is an even number. (b) Q: [have three brothers.16 Propositional Logic (©) R: 427 (d) S: V=T does not exist. The truth (T) or falsity (F) of a proposition is termed its truth value. For our purposes a proposition which is not true will be regarded as. false and one which is not false will be viewed as true. It is important to note that the truth value of a proposition may depend on the context in which it is stated. Of the propositions listed above P is true and R is false. However, Q is true only if uttered by someone who has three brothers and is false otherwise. Proposition S is true if we have agreed to restrict our discussion to the real numbers. However, if we make the statement within the context of the complex numbers, then it is a false proposition. Sentences which cannot be viewed as true or false are not propositions. These include questions, demands, exhortations and exclamations. Hence the following are not propositions. (e) Show your working clearly. (f) Has a trapezium got four sides? (g) Vote for Mickey Mouse! It is clear that, for any proposition, adding the prefix ‘It is not the case that ...’ or inserting ‘not’ appropriately results in another proposition with the reverse truth value. For example, if the proposition ‘I have three brothers’ is true then the proposition ‘It is not the case that I have three brothers’ or ‘I do not have three brothers’ is false and vice versa. If we reverse the truth value of any proposition P in this way, the resulting proposition, denoted by P (or ~P or “P) is called the negation of P. There are a variety of different ways of stating the negation of a proposition but what is important about P is that it is true in all circumstances that P is false and false whenever P is true. We can summarise this in a table where we show, for each of the possible truth values of the propositional variable p, the corresponding truth value of ji, the negation of p. pf? rit A table which summarises truth values in this way is called a truth table.Logical Connectives 17 Truth table 2.2 Logical Connectives Each of the propositions P, Q, R and $ defined above makes a single statement about an object or individual. Such propositions are called simple propositions. The proposition ‘Today is Tuesday and it is raining’ is not a simple proposition since it makes two statements, one concerning the day of the week and the other about the state of the weather. However, it can be viewed as being composed of the two simple propositions ‘Today is Tuesday’ and ‘It is raining’ conjoined using the word ‘and’. The truth value of ‘Today is Tuesday and it is raining’ is dependent upon the truth values of these two component simple propositions. It is true if both components are true and false otherwise. We summarise this in the table below where we now use Q and R denote ‘Today is Tuesday’ and ‘It is raining’ respectively. [Q | R | QandR T | T T T|F F F/T F FILE FE The right hand column of the table gives the truth value of the proposition ‘Q and R’ for each possible pair of truth values of the individual propositions Q and R. For instance, the last line indicates that, if Q and R are both false, then the proposition ‘Q and R’ is also false.18 Propositional Logic Propositions which are formed by joining two or more simple proposi- tions are called compound propositions. The items which are used to join the simple components are called logical connectives. The truth value of a compound proposition is determined by two factors: the truth value of each of its component simple propositions and how particular logical connectives are used to link them. There are five connectives which are important: conjunction, inclusive disjunction, exclusive disjunction, conditional and biconditional. We now look at the properties of each of these. Conjunction The truth table for the conjunction of any two propositional variables p and q, denoted by p “q (or by p.q), is given below. aaa mamas momma] > mamas The table shows that the conjunction p A q is true only when true propo- sitions are substituted for each of p and q. Otherwise the conjunction is false. As we have already seen, this sense is conveyed using the word ‘and’ between the two component propositions (often termed the conjuncts). For example, if B denotes the proposition ‘Bob is a foot- baller’ and S denotes ‘Sue is a student’ then the conjunction of B and S is expressed by ‘Bob is a footballer and Sue is a student’ and is denoted by BAS. Although ‘and’ is the most common linguistic expression for logical conjunction, there are alternatives. The following would also be denoted by B AS although they do have nuances which are slightly different from when the two components are joined using ‘and’: Bob is a footballer although Sue is a student; Bob is a footballer whereas Sue is a student; Bob is a footballer but Sue is a student. These are conjunctions because, in each case, we would view the compound proposition as true only when both of its two components are true.Logical Connectives 19 Note that the conjunction p Aq (read as ‘p and q’) is symmetric in the sense that it has exactly the same set of truth values as q A p. Both are true only when each of the components p and q are true. Hence, from the point of view of logic, ‘Bob is a footballer and Sue is a student’ and ‘Sue is a student and Bob is a footballer’ are equivalent propositions. (We shall give a more precise definition of ‘equivalent propositions’ later.) Disjunction There are two forms of logical disjunction—the inclusive and exclu- sive forms. The inclusive disjunction of two propositional variables p and q is denoted by p vq and the exclusive disjunction of the two components by p vq. The truth table for each of these is given below. ea eae eae ea T T T F T F T T F T T T FI F F F The only difference between the truth values of p v q and p vq is when p and q (often termed the disjuncts) are both true propositions. In this case the inclusive disjunction p v q is true but the exclusive disjunction p vq is false. An inclusive disjunction is true only when either or both of its disjuncts are true whereas an exclusive disjunction is true only when one disjunct is true and the other is false. Unfortunately, in English the word ‘or’ is used between disjuncts (sometimes with the first disjunct preceded by ‘either’) as the linguistic expression for both inclusive and exclusive disjunction. Therefore a proposition containing ‘or’ is often ambiguous as to whether the inclusive or exclusive sense is intended. Sometimes the context suggests which form of disjunction is appropriate. For instance, ‘On Monday I shall stay in London or visit a friend in Paris’ has components which seem to be mutually exclusive and so we would interpret the proposition as true only when just one of the disjuncts is true. Exclusive disjunction therefore seems to be the intended interpretation of ‘or’. On the other hand, ‘Applicants for the job must have a degree or three years relevant experience’ does not seem to preclude applicants who