A COURSE IN MATHEMATICAL LOGIC by J. L. BELL London School of Economics and Political Science and M. MACHOVER Chelsea College of Science and Technology at bas ni 1977 NORTH-HOLLAND PUBLISHING COMPANY — AMSTERDAM - NEW YORK - OXFORD© North-Holland Publishing Company — 1977 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. Library of Congress Catalog Card Number: 75-33890 North-Holland ISBN: 0 7204 28440 Published by: North-Holland Publishing Company Amsterdam - New York - Oxford Sole distributors for the U.S.A. and Canada: Elsevier North-Holland, Inc. 52 Vanderbilt Avenue New York, N. Y. 10017 Library of Congress Cataloging in Publication Data Bell. John Lane. A course in mathematical logic: Bibliography: p. 576. Includes index. 1. Logic, Symbolic and mathematical. I. Machover, Moshé, joint author. IT. Title. QA9.B3953 511.3 75-33890 Printed in HungaryOn the contrary, I find nothing in logistic for the discoverer but shackles. It does not help us at all in the direction of conciseness, far from it; and if it requires 27 equations to es- tablish that 1 is a number, how many will it require to demonstrate a real theorem? H. Pomncaré Although it is a distinctly minor issue, we must mention Fibonacci’s famous recurring se- ries.... There is an extensive literature, some of it bordering on the eccentric, concerning these numbers... Some professional and dilettant esthetes have applied Fibonacci’s num bers to the mathematical dissection of masterpieces in painting and sculpture with results not always agreeable, although sometimes ludicrous, to creative artists. Others have dis- covered these protean numbers in religion, phyllotaxis, and the convolutions of sea shells E. T. Bete Infinitesimals as explaining continuity must be regarded as unnecessary, erroneous, and self-contradictory. B. RussetDedicated to the memory of ‘Andrzej Mostowski and Abraham RobinsonACKNOWLEDGEMENTS In the course of producing this book we have become indebted to many people. To begin with, we would like to put on record our intellectual debt to those logicians and mathematicians whose work we have expo- unded: in a work of this kind it would not be feasible to attribute each result to its creator, but we hope that the historical references at the end of each chapter will furnish a general (if sketchy) guide to who proved what. Our students have been very helpful in furnishing comments and cri- ticism; in this connection we would particularly like to thank Michael Bate, Narciso Garcia and Ali al-Nowaihi. We are also grateful to our colleagues Daniel Leivant, Dirk de Jongh and Brian Rotman for reading sections of the manuscript and making useful suggestions. The job of typing the manuscript was, to say the least, an arduous and tricky one and for skilfully carrying out this operation we are indeb- ted to Buffy Fennelly, Diane Roberts, Barbara Silver and Marie-Louise Varichon. We would also like to thank the printers, and the staff of North-Hol- land Publishing Company, in particular Einar Fredriksson and Thomas van den Heuvel, for the competent and friendly way they have handled all stages in the production of this book. Finally, we offer our warmest thanks to our wives Mimi and Ilana for, among many other things, their patience and encouragement throughout the long writing period and for their help in preparing the bibliography. John Bell Moshé MachoverAcknowledgements... 2.2... nterdependence scheme for the chapters Introduction... 6 ee ee et ee Recommended reading... ... 2-2-2222 004 CuapTer 0, PREREQUISITES... . ...~ - “CHAPTER 1. BEGINNING MATHEMATICAL LOGIC $1. §2. $3. $4. §5. §6. §7. §8. §9. $10. Sil. §12. $13. $14. sis. $16. General considerations ..... 2... cont Structures and formal languages. . . . . eae Higher-order languages... .......-..- Basic syntax... ee Notational conventions... ...-. 2.20 e- Propositional semantics... .......---- Propositional tableaux... 2... ee ee The Elimination Theorem for propositional tableaux . Completeness of propositional tableaux see The propositional calculus... . 2... 2. The propositional calculus and tableaux oe Weak completeness of the propositional calculus Strong completeness of the propositional calculus Propositional logic based on TV and A... .. - Propositional logic based on 4, >, A and V .. . Historical and bibliographical remarks Bese eee “CHAPTER 2. FIRST-ORDER LOGIC... 2. 2 1. ee ee $1. §2. §3. $4. §5. First-order semantics .. 2... 1... 200+ Freedom and bondage... 2.2... 00-0 Substitution ©... 2... ee First-order tableaux. . 2 2... 2. ae Some “book-keeping” lemmas. ......--.-- . The Elimination Theorem for first-order tableaux eae ORR EIR Ct atta ttt fat ott . Completeness of first-order tableaux ........- . Prenex and Skolem forms. ........--- . Elimination of function symbols... ...... .gu. §12. §13. §14. CONTENTS Elimination of equality... ......-..--- Peres See a SESS Virtualterms. .. 2... .....-.- Se Historical and bibliographical remarks... ... (CuaPTER 3. FIRST-ORDER LOGIC (CONTINUED) $l. g2. 3. 4. §5. §6. The first-order predicate calculus .......... The first-order predicate calculus and tableaux . . . . Completeness of the first-order predicate calculus . . . First-order logic based ON ee eee Historical and bibliographical remarks... . . . Cuapter 4, BOOLEAN ALGEBRAS... . 2... 1 2s gl. §2. 8. $4. §5. §6. 7. §8. Lattices . 2... tee teEee Booleanalgebras. . 2 2... ee Filters and homomorphisms... ..-.--.---- The Stone Representation Theorem... ... . Atoms .....------.--5 Duality for homomorphisms and continuous mappings. . . . The Rasiowa-Sikorski Theorem... .. 2.2... | Historical and bibliographical remarks . . . ‘CHAPTER 5. MODEL THEORY... . . EE See eet §1. 2. 93. $4. §. 86. §7. §8. Basic ideas of model theory... 2. 2. 7 ee The Léwenheim-Skolem Theorems Ultraproducts . . Completeness and categoricity Lindenbaum algebras... . . EERE eerie Element types and ®o-categoricity .. 2... 2... Indiscernibles and models with automorphisms . . . . Historical and bibliographical remarks . ‘CHAPTER 6. RECURSION THEORY . . - §l. 82. 93. 4. §5. §6. §7. §8. 99. $10. Su. $12. $13. $14. Basic notation and terminology... ......- Algorithmic functions and functionals . . The computer URIM 2... 0... eee Computable functionals and functions . . . Recursive functionals and functions ........ A stockpile of examples... ... 2... 2-05 Church’s Thesis... ... ~~ Recursiveness of computable functionals . . Functionals with several sequence arguments... . . Fundamental theorems... . 2... Recursively enumerable sets . . . . . BEE Diophantine relations... 2 2... 2 ee The Fibonacci sequence. . . . Sesser The power function... . . . . XL 101 102 . 104 107 . 108 108 beer _ U7 122 122 124 125 125 . 129 133 141 150 . 153 157 159 161 161 168 . 174 . 184 191 . 203 214 224 226 . 226 230 232 . 237 239 247 257 - 265XI CONTENTS §15. Bounded universal quantification... 2.2... 2... 2.020024 305 §16. The MRDP Theorem and Hilbert's Tenth Problem... 2.2... .. ul §17. Historical and bibliographical remarks... 2... 2... 2... 314 Cuapter 7. Locic — LIMITATIVE RESULTS... 2 2 ee 316 §1. General notation and terminology... ............... 316 §2. Nonstandard models of@ 2.2... .......0.200. - . 318 §3. Arithmeticity 2... 2... PEPE eters ele . s+ 324 §4. Tarski’s Theorem. 2... eee 327 §5. Axiomatic theories... 2... ee ee : - 332 §6. Babyarithmetic. . 2. 2 2 ee ee ee we 334 §7. Junior arithmetic... 2... ee Ls. 336 §8. A finitely axiomatized theory... . 2... . pa ee teat ba §9. First-order Peano arithmetic... 2.2... ....-.00. » . 342 §10. Undecidability ©... 2. 0. 347 §11. Incompleteness . . . . . . Me ee 353 §12. Historical and bibliographical remarks... .. 2... 0 2... 2 Ye 359 Cwaprer 8, RECURSION THEORY (CONTINUED). ©... ee 361 §1. Thearithmeticalhierarchy. . 2... 2. ee eee 361 §2. A result concerning To... . . SE Eee Sete eb 369 §3. Encodedtheories. 2.2... 2 ee . 370 $4. Inseparable pairs of sets... . EE bac ei He ete raebaee i 372 §5. Productive andcteativesets;reducibility. ...............~, 376 §6. One-one reducibility; recursive isémorphism .............. 384 §7. Turingdegrees. 2 ee eee 388 §8. Post’s problem and its solution... 2... 2... ee - 392 §9. Historical and bibliographical remarks... 2... 2. 2 we 398 CHAPTER 9, INTUITIONISTIC FIRST-ORDER LOGIC. 2... ee ee 400 §1. Preliminary discussion... 2-2. ee ee ee 400 §2. Philosophical remark . . SS a a ea et ate tat 403 §3. Constructive meaning of sentences... . 2.2.2... ..... . 403 §4. Constructive interpretations . . . . . Pee a etree ee ret 404 §5. Intuitionistic tableaux. 2... eee 408 §6. Kripke’s semantics ©. 2 2. 2. 2 ee 416 §7. The Elimination Theorem for intuitionistic tableaux. |... 11... 422 §8. Intuitionistic propositional calculus... 2... ........... 433 §9. Intuitionistic predicate calculus. ©. 2... ee ee 434 §10. Completeness 2... 22. ee 438 §11. Translations from classical to intuitionistic logic... 2.2... 2. 442 §12. The Interpol §13. Some results in classical logic . 2... ee ee ee ee 452 §14. Historical and bibliographical remarks... 2.2... 2.0.0. 457 Cuaprer 10, AXIOMATIC SET THEORY. ©. 2. ee 459 §1. Basic developments. . 2. 2. 1. ee ee ee 459 §2. Ordinals 6 1 ee 46833. §4. §5. §6. §7. §8. §9. §10. $11. CONTENTS The Axiom of Regularity. ...... Cardinality and the Axiom of Choice. . . . Reflection Principles . The formalization of satisfaction pelea Absoluteness . 7 Constructible sets... . . The consistency of AC and GCH Problems Historical and bibliographical remarks Cuapter 11, NONSTANDARD ANALYSIS gl. §2. 8. 4. $5. §6. §7. §8. Enlargements 2... ee Zermelo structures and their enlargements . Filtersandmonads. 2... 2... Topology . . . ; Bee Topological groups The real numbers... an A methodological discussion... 2... Historical and bibliographical remarks BIBLIOGRAPHY GENERAL INDEX... - eee ee INDEX OF SYMBOLS 516 . 529 . 531 532 536 543 . 553 . 561 566 572 . 573 . 576 584 595INTERDEPENDENCE SCHEME FOR THE CHAPTERSINTRODUCTION For the past seven years, the authors have conducted a one-year M.Sc. programme in mathematical logic and foundations of mathematics at London University. The present book developed from our lecture notes. for this programme, and the student should therefore be able to work through the text in (roughly) one academic year. The main problem that we faced in constructing the programme was the following. First, we wanted it to be an integrated and balanced account of the most important aspects of logic and foundations. But secondly, since parts of our programme are taken by mathematics and philosophy of science students who for one reason or another do not want to cover all the topics we discuss, we were led to arrange it in such a way that parts could be taken as separate smaller courses. Accordingly, the book itself falls naturally into several units: 1. Chapters 1-3. These together constitute an elementary introduction to mathematical logic up to the Gédel-Henkin completeness theorem. We teach this part in a fairly leisurely way (four hours per week for ten weeks, including problem classes), and accordingly the pace of the text here is rather gentle. There is one feature which deserves special mention and that is the use of Smullyan’s tableau method. This method serves a dual purpose. First, it is a proof-theoretic instrument that allows us to obtain constructive proofs of various results. In this respect it is equi- valent to Gentzen’s calculus and to various systems of natural deduction. Secondly, our teaching experience shows that Smullyan’s method has the great advantage of being a practical tool — after a little practice, it furnishes a quick, efficient and almost computational method of actually detecting the truth or falsehood of formulas. (This efficiency stems in part from the fact that, unlike Gentzen’s calculus, it does not require the same formula to be copied again and again.) However, the materialXVI INTRODUCTION on tableaux has in fact been isolated in separate (starred) sections so that the reader who does not want to use this material can simply ignore it; what remains is a self-contained standard account of first-order logic. A middle course is also possible: a reader wishing to enjoy the practical advantages of tableaux but who lacks the time or patience for the somewhat complex constructive proofs of elimination theorems (Ch. 1, §8, and Ch. 2, §§5, 6) can skip the latter because the same results are also obtained in an easier but non-constructive way elsewhere (Ch. 1, §9 and Ch. 2, §8). We should like to point out that the somewhat rebarbative complexities of Ch. 2, §5 could have been avoided by using different symbols for free and bound variables (as is often done in texts devoted mainly to proof theory). This, however, would detract slightly from the practical utility of the method and in any case would be contrary to accepted usage in most other branches of logic. 2. Chapters 4 and 5. The contents of Chapter 4 are taught for 1 hour a week over 10 weeks, concurrently with the material in Chapters 1-3 (of which it is totally independent). It in fact constitutes a separate short course on Boolean algebras. The material in Chapter 5 — model theory — is taught over the following 10 weeks for 2 hours a week. It depends heavily on Chapter 4 but only slightly on Chapters 1-3 inasmuch as it can be read by anyone having modest acquaintance with the notation and main results of first-order logic. 3. Chapters 6 and 8. These two chapters constitute a self-contained course on recursion theory. The material in Chapter 6 is taught for 2 hours a week over 10 weeks, concurrently with Chs. 1-4, of which it is totally independent. There are two points here which call for comment. First, we employ register machines instead of Turing machines, because the former are much closer in spirit to actual digital computers, and are also smoother theoretically. Secondly, this chapter includes a full proof of the Matiyasevich-Robinson—Davis-Putnam (MRDP) theorem that every recursively enumerable relation is diophantine. We believe that — despite the length and tedium of the proof — this result is of such importance that no modern account of recursion theory can afford to omit it. In teaching this part of the chapter, we have found that some of the material in §§13, 14, and the first half of §15 can be omitted in class and given to the student to study at home. As for Chapter 8, it is taught for the following eight weeks at a rate of 2 hours per week. The material here of course depends entirely on Chapter 6, but in this book it appears after Chapter 7,INTRODUCTION XVII because it is motivated by and illuminates the results contained there. However, no detailed knowledge of Chapter 7 is required to understand Chapter 8. In Chapters 6 and 8 we have adopted a somewhat formal approach: in proving that such-and-such a function is recursive, we employ the precise apparatus furnished by the recursion theorem, rather than the intuitive “proof by Church’s thesis”. We have chosen this course because we believe that the beginning student has not yet developed sufficient experience in the subject to be totally convinced by intuitive proofs which employ Church’s thesis. 4. Chapter 7. This chapter contains an account of the limitative results about formal mathematical systems. Reliance on the MRDP theorems has enabled us to simplify some of the proofs and obtain somewhat sharper results than usual. The chapter presupposes a good knowledge of first- order logic and some knowledge of recursion theory. However, it can be and is taken by students who have no detailed acquaintance with the latter. We have found it feasible to develop all the requisite results from recursion theory — except the MRDP theorem — using Church’s, thesis, the MRDP theorem itself being stated without proof. This approach enables us to teach the material in this chapter intelligibly to students who do not want take a full-fledged course in recursion theory. (The material here is in fact taught concurrently with Chapter 5 for 10 weeks at 2 hours per week.) 5. Chapter 9. Here we have an elementary introduction to first-order intuitionistic logic. While neither of the authors claim to be an expert on intuitionism, we nevertheless believe that the constructivist approach to mathematics is of such great importance that some discussion of it is essential. (The material in this chapter is taught concurrently with Chapters 5, 7 and 8 for 10 weeks at 1 hour per week.) 6. Chapter ‘10. This is devoted to an axiomatic investigation of Zermelo— Fraenkel set theory, up to the relative consistency of the axiom of choice and the generalized continuum hypothesis. It requires modest familiarity with first-order logic and with the Lowenheim-Skolem theorem in Chapter 5. (This material is taught over roughly 10 weeks at 2 hours per week at the end of the second term and the beginning of the final term.) 7. Chapter 11, This chapter contains an introduction to nonstandard analysis, which is an important method of applying model theory to mathematics. The material here is taught over 10 weeks at 2 hours per week, 2XVII INTRODUCTION during the latter part of the year. Although this chapter presupposes a few results of model theory, these results can be stated concisely without proof for the benefit of those students who wish to study the subject without doing a special course on model theory. In fact it is possible to teach nonstandard. analysis to students who have only a slender acquaintance with logic. As can be seen from the foregoing synopsis, the material in the book can be regarded as forming several relatively independent units. However, the book has been conceived as an organic whole, and provides what is in our view a “balanced diet”. We have striven to reveal the interplay between “structural” (i.e. set-theoretical) ideas and “constructive” methods. The latter play a particularly prominent role in mathematical logic, and we have therefore stressed the constructive approach where appropriate but without, we hope, undue fanaticism. The problems constitute an essential part of the book. They are not mere brainteasers, nor should they be too difficult for the student to solve, given the hints that are provided. Many of them contain results which are later employed in proofs of theorems. Accordingly, no unstarred problem should be skipped! Certain sections and problems are starred. This does not necessarily indicate that they are more difficult, but rather that they may be omitted at a first reading. Some problems have been starred because they require more knowledge or skill than is needed for understanding the text in the same section. Each chapter is divided into sections. When we want to refer to a theorem, problem, definition, etc., within the same chapter, we give the number of the section in which it occurs, followed by its number in that section. Thus, e.g., Def. 2.10.1 is the first numbered statement in §10 of Ch. 2 and within Ch. 2 it is referred to as “Def. 10.1.” (or simply “10.1”). We use the convenient abbreviation “iff” for “if and only if”. The mark Jj is used either to signify the end of a proof or, when it appears immediately after the statement of a result, to indicate that the proof is immediate and is accordingly omitted. References to the bibliography are given thus: KeLtey [1955]. The overwhelming majority of references to the bibliography are given in a separate section at the end of each chapter.RECOMMENDED READING In addition to the special works referred to in the text, we would like to recommend the following books as general reading in logic and the foundations of mathematics. BENACERRAF and Putnam [1964]. (An excellent anthology of seminal works in the philosophy of mathematics.) Davis [1965]. (An anthology of important papers on limitative theorems and recursion theory.) FRAENKEL, BAR-HILLEL and Levy [1973]. (A comprehensive survey of the foundations of mathematics, and set theory in particular.) Van HEVENOORT [1967]. (Another excellent anthology containing many of the classic papers in logic and foundations.) KNEEBONE [1963]. (An elementary but scholarly general introduction to and survey of the foundations and philosophy of mathematics.) KreisEL and Krivine [1967]. (The appendices to this work contain a penetrating discussion of the philosophical foundations of mathematics.) Lakatos [1967]. (A collection of papers on the foundations and philo- sophy of mathematics delivered at the 1965 London colloquium.) Mostowski [1966]. (A wide-ranging survey of the development of mathe- matical logic from 1930 to the 1960's.)CHAPTER 0 PREREQUISITES In this book we assume that the reader is familiar with the basic facts of naive set theory (including the fundamentals of cardinal and ordinal arithmetic) as presented, e.g., in FRAENKEL [1961], Hatmos [1960] or KuraTowski-Mostowsk! [1968]. Facts about cardinals and ordinals are used at the end of Ch. 3, occasionally in Chs. 4 and 9, and throughout Chs. 5 and 10. In some places (especially in Chs. 4 and 11), we assume a slender acquaintance with the basic notions of general topology as presented, e.g., in Ch. 1 of BourBAKi [1961] or the first few chapters of Kettey [1955]. We distinguish between classes and sets. Except in Chs. 10 and 11 (where the terms “‘class” and “set” are assigned a more precise technical meaning), a class is understood to be an arbitrary collection of objects, while a set is a class which can be a member of another class. (Another distinguishing feature of sets is that only they have cardinalities.) Given an object x and a class X, we write as usual x€ X for “x is a member (element, point) of X and say “X contains x” or “‘x is in X”. If X contains every member of a class Y, we say “X includes Y” and write YS X. Two classes are regarded as identical if they have the same members. The set of natural numbers (which contains 0) is denoted by N or @. Except in Ch. 10, the empty set is denoted by 9. If A is a set, the power set PA of A is the set of all subsets of A. Given n>1 objects x,,...,X,, we write (x1,...,x,) for the ordered n-tuple Of X1,...,X,- Thus (x,y) is the ordered pair of x and y. By convention, we put (x)=x (the ordered singleton of x). The Cartesian product of a finite sequence of classes A,,...,4, (with n>1), denoted by A4,X...XA,, is the collection of all n-tuples (a,,...,d,) with a,€ Aj,....4,€4,. If each A; is identical with a fixed class A, we write A” for A,X...XA,. By convention, we set A°={9}; thus A° has exactly one member, namely 6.2 PREREQUISITES For n>1, an n-ary relation on a class A is a collection of n-tuples of members of A, i.e. a subclass of A". A unary relation on A is called a property; it is just a subclass of A. The identity (or diagonal) relation on A is the binary relation {(x,x): x€ A}. The membership relation on A is the binary relation {(x,y): x€A and yeA and x€y}. If R is an n-ary relation on A and BC A, the restriction of R to B is defined to be the n-ary relation ROB" on B. If R is a binary relation, we often write xRy for (x,y)€ R. A function (map, mapping) is a class f of ordered pairs such that, whenever (x, y)€f and (x,z)€f, we have y=z. The domain dom(f) of f is the class {x: for some y, (x, y)€f} and the range ran(f) of f is the class {y: for some x, (x,y)€f}. If f is a function, and x¢€dom(f), then the unique y for which (x,y)¢f is denoted (except in Ch. 10) by f(x), or sometimes fx, etc., and is called the value of f at x. Sometimes we specify a function fin terms of its values; under these conditions we write x-+f(x). (Thus, for example, x-+x+1 describes the successor function on N.) If fis a function such that dom(f)=4 and ran(f)¢ B, we say that f is a function from A to (into) B and write f:A-B. If f: A+B and XA, we define the restriction {|X : XB by (F|NE)=S@) for xe X. If XSA and YCB, we put SUXI=(F): XE X}, FLY] = fe: SOE Y}, and, for y¢ Y, we put FO) =F LOH. A function f : A--B is one-one (an injection) if f(x)=f(y) implies x=y for all x,y€ A; onto (a surjection) if f[A]=B; and a one-one correspondence (a bijection) between A and B if both of these conditions hold. A and B are said to be equipollent if there is a bijection between A and B. If f: A~B and g : BC, the composition gof : A~C is defined by (go f)(x)=g(f(x))PREREQUISITES 3 for x¢€ A. We sometimes omit the o and write simply gf instead of gof. Observe that, for each class A, the identity relation on A is a bijection between A and 4; for this reason it is also called the identity map on A. If ACB, the natural injection of A into B is the map i: A+B defined by i(x)=x for x€ A. If A is a class and / is a set, we write A’ for the collection of all functions from J into A. (Notice that this definition implies 49 = {0}=A°.) If {4,: i€ J} i an indexed family of sets, we write [];<7A; for the collection of all functions f with domain J such that f(i)¢€ A; for all i€ J. The axiom of choice asserts that, if each A,;~0, then []jcr4,49. For any n€ N, an n-ary operation on a class A is a function from A" to A. In particular, a 0-ary operation on A is a function from {@} to A, and therefore has a unique value which we identify with the given 0-ary operation. Thus a 0-ary operation on A is just a member of A. If fis an n-ary operation on A, we write f(a,,...,4,) for f((ay,..-.4,)). A subclass B of A is said to be closed or stable under f if f(b,,....b,)€B whenever 5y,...,b,€ B. If B is closed under f, we define the restriction f|B of f to B by (FB) b15---be)=LOrs---2bn) for byy---sbn€ Be A binary relation R on a class A is called an equivalence relation if it satisfies: (a) xRx forall x€A, (b) xRy implies yRx forall x,y€A, (c) xRy and yRz implies xRz forall x,y,z€ A. If R is an equivalence relation on A, for each x€A the set Xp={y€ A: xRy} is called the R-class of x. Calling a family @ of subsets of A a partition of A if U@=A and Xn Y= for any distinct members X,Y of @, we see immediately that, if R is an equivalence relation on A, the family of all R-classes of members of A constitutes a partition of A. A partially ordered set is an ordered pair (A, <<) in which A is a set and < is a binary relation on A satisfying: (a) x2”) and in making general statements about numbers (such as “for all x and y, x+-y=y+ x”). Similarly, we shall use metalinguistic symbols called syntactic variables, each of which will range over a prescribed collection of symbols or combi- nations of symbols of some formal language. § 2. Structures and formal languages In this section we discuss the various kinds of symbols and expressions that a formal language may be expected to have, if it is to be used as a medium for expressing mathematical statements. A great many — some would say all — mathematical statements are about structures. A structure consists of the following ingredients: (1) A non-empty class, called the universe or domain of the structure. The members of this universe are called the individuals of the structure. (2) Various operations on the universe. These are called the basic operations of the structure. (3) Various relations on the universe. These are called the basic relations of the structure. Here (2) is optional: we admit structures having no basic operations. On the other hand, a structure must have at least one basic relation. The identity relation on the universe is normally taken as one of the basic relations. Among the basic operations of a structure there may be 0-ary operations. According to our convention (see Ch. 0) such an operation is simply an individual. The 0-ary basic operations of a structure are called its designated individuals. Let us give a few examples of structures associated with various branches of mathematics. Elementary arithmetic may be defined as the study of one particular structure — the elementary structure of natural numbers. It has the set of natural numbers as universe, two designated individuals (viz. 0 and 1) and two basic operations (viz. addition and multiplication), Here the only basic relation is the identity relation. Set theory is concerned with a structure whose individuals are all sets * In this book syntactic constants and syntactic variables are printed in bold type, unless otherwise stated.10 BEGINNING MATHEMATICAL LOGIC ICH. 1, §2 (i.e., its universe is the class of all sets). In addition to the identity relation, this structure has the basic relation of membership: the relation that holds for all pairs (x,y) such that x and y are sets and x is a member of y. Elementary Euclidean plane geometry may be regarded as the study of a structure — the elementary Euclidean plane — whose individuals are points and straight lines. Among the basic relations of this structure are the property’ of being a point, the property of being a straight line, and the ternary relation of “‘betweenness” which holds for all triples (x,y,Z) Such that x,y and z are collinear points and y lies between x and z. A topological space may be thought of as a structure whose individuals are all points of the space, as well as all sets of points, all sets of sets of points, etc. There is one designated individual — the set of all open sets of points (this designated individual is the topology of the space); and the basic relations are the property of being a point, the identity relation and the membership relation between individuals. Similarly, a structure can be associated with each group, ring and with other entities studied in algebra. Suppose we are given a structure UW and we want to set up a formal language £ in which statements about U are to be expressed. What symbols should & have? First, we would like to have symbols that may be used as variables ranging over the universe of Uf. The need for variables is obvious to anyone acquainted with mathematics. Variables ranging over the universe of U are used, e.g., in expressing conditions which individuals of WU may or may not satisfy, and in making general statements about U. Next, we expect ¥ to have symbols that may be used to denote the various basic operations of UW. Such symbols are called function symbols. More specifically, a symbol designed to denote an n-ary operation is called an n-ary function symbol. In particular, if U has designated individuals then & should have symbols for denoting them. Such symbols are called individual constants or, more briefly, just constants. Since designated individuals are regarded as 0-ary operations, constants are to be regarded as O-ary function symbols. Using variables and function symbols, we can construct expressions called terms. Roughly speaking, terms are the nounlike expressions of Y. For example, in a formal language suitable for elementary arithmetic we should have variables, say x,y, etc., intended to range over the set » Recall that we have agreed (Ch. 0) that a property is a unary relation.CH. 1, 821. STRUCTURES AND FORMAL LANGUAGES. it N of natural numbers; and function symbols, say 0,1,+,x, intended to denote the numbers zero and one and the operations of addition and multiplication, respectively. Then x, 1, 14x, Oxy, ((1-++x)x(y+)-+0) are some of the terms we can form. Of course, different interpretations can be applied to one and the same language. For example, the language described in the preceding paragraph can be re-interpreted by letting its variables range over some arbitrary non-empty class and letting 0,1,-+,% denote two arbitrarily chosen members. of that class and two arbitrarily chosen binary operations on it. But suppose we have fixed one particular interpretation for a formal language Y, by means of a structure Uf. Then those terms of ¥ that do not contain variables will denote individuals of U. A term containing variables will not denote any particular individual, but will assume various individuals as values, depending on which individuals are assigned as. values to the variables. For example, in the particular language described above (taken with its originally intended interpretation) the term 1441 denotes the natural number two, while the term (1-+-1)xx-++y has as value the number obtained by adding whatever number is assigned as value to y, to twice the number that happens to be assigned as value to x. Variables and function symbols alone do not suffice for formulating in & statements about a structure U. For this, & must have symbols that can be used to denote the basic relations of Uf. A symbol designed to denote an n-ary relation is called an n-ary predicate symbol. If — as is usually the case — the identity relation is one of the basic relations of U, then Y needs to have a predicate symbol to serve as a name for it. It is convenient to earmark one particular symbol, =, for this role.* To illustrate how predicate symbols are used, let us return to our example of the language with function symbols 0,1,4-,x. Assuming this language to have the predicate symbol =, we can write formulas like a) 1=1x1, @ 14+0=1x0, (3) X-++>X=xXx, 4) X-++x=y. 1 This amounts to treating = as a logical symbol: it is always interpreted as denoting the identity relation on the universe of discourse.12 BEGINNING MATHEMATICAL LOGIC (CH. 1, 2 Formulas (1) and (2) are sentences: they express propositions. Under the originally intended interpretation, (1) expresses the true proposition that the number one is the same as its own square; and (2) expresses the false proposition that the sum of one and zero is the same as their product. Thus, under the interpretation which we are assuming, (1) is a true sentence and (2) is a false sentence. We also say that (1) has the truth value ‘truth’ and (2) has the truth value ‘falsehood’. Formulas (3) and (4) are not sentences; they do not express propositions, but conditions regarding the values which may be assigned to the variables involved. Thus, (3) expresses the condition that when the value of x is added to itself the result is the same as when it is multiplied by itself. It makes no sense to say that (3) is true (or false) outright. Rather, (3) will assume a truth value — either truth or falsehood — depending on which natural number is assigned to x as value. If that value is zero or two, (3) will have the truth value truth or, in other words, (3) will be satisfied. For all other values of x, (3) has the truth value falsehood. Similarly, (4) is satisfied (has the truth-value truth) iff the value assigned to y is double that assigned to x. This example illustrates a general fact. Any n-ary predicate symbol of a formal language # can be combined with n terms of ¥ to form a formula. Once an interpretation of £ is fixed, some formulas of # (viz. those that are sentences) become true or false outright; all other formulas acquire truth values when the variables have been assigned values (belonging to the universe of the structure used for interpreting £). Using variables and function symbols we can construct terms; using these and predicate symbols we can construct formulas — but only rather simple ones. In order to combine simpler formulas into more complex ones, we shall require # to have two new kinds of symbol, called connectives and quantifiers. We would like negation to be expressible in &. Thus, for any formula a of £ we want £ to have a formula “a (read: “not «” or “it is not the case that «’’) which will be true whenever « is false, and false whenever @ is true. (Thus « and “a always have opposite truth values.) Next, we want the conjunction and to be expressible in Y. Thus, for any two formulas a, B we need a formula &AB (read: “a and B”) which will be true iff both @ and f are true.CH. 1, $2). STRUCTURES AND FORMAL LANGUAGES, 13 Similarly, we want ¥ to have, for any formulas « and B, a formula avB (read: “a or B”) which is false iff both « and B are false.” Further, we want # to be capable of expressing conditional statements. Therefore, for any formulas a, B of ¥ there should be in ¥ a formula a—>B (read: “a implies B” or “if a, then B”). This formula will be false iff is true but B is false. Finally, @ should have, for any formulas a and B a formula 2B (read: “a iff B”) which is true whenever & and B have the same truth values and false whenever their truth values are different. We could satisfy all these demands by requiring £ to have five logical symbols, called connectives, viz. 1, A, V, >, <. But as a matter of fact we can make do with less. To start with, («>B)A(B—>«) behaves in just the way we want a¢B to behave; so we do not require ¥ to have the symbol «+». Next, (a> 1B) is easily seen to behave exactly as aAB should; so A too can be eliminated. Finally, ("1a)->B behaves just like «VB; so we can eliminate v as well. Thus # need only have two connectives “1 and ->. (For reasons that will become apparent later on, it is convenient to economize in the stock of logical symbols. No sacrifice is involved, since we can define aAB to be 7(%—> —B) etc.) The last demand we shall make on # is that for any formula a and any variable x, Y should have a formula Wxe (read: “for every [value of] x,@”) and a formula 4xa (read: “for some [value of] x, a”). To explain the meaning of such formulas, suppose first that « expresses some condition regarding the value of the variable x. Then Wxa expresses the proposition that every possible value of x (ic., all individuals of the structure) satisfy that condition. And 3xa says that the condition expressed by a is satisfied by some (i.c., at least one) individual. Thus, e.g., in the particular language discussed above, taken with its intended interpretation, the formula x=xxx expresses the condition that the value of x be equal to its own square; Wx(x=x Xx) then expresses the false proposition that every natural number is equal to its own square, and. 4x(x=xxx) expresses the true proposition that some natural number is equal to its own square.* 11In mathematical usage, “or” has the meaning which jurists express by “and/or”. We are following this usage here. 2 Here too we are following normal mathematical usage, which dictates that whenever a is false a->B should be regarded as (vacuously) true, irrespective of the truth value of B. "In fact, there are of course two such numbers. 314 BEGINNING MATHEMATICAL LOGIC ICH. 1, §3 More generally, suppose that @ expresses a condition regarding the values of the variables x,y,,...,y,. Then Wxe and 4xa express conditions regarding the values of y,,...,y,- The values ,,...,5, (assigned to y,,...,Y, respectively) satisfy [the condition expressed by] Wxe iff for every possible value a of x the values a,b,,...,5, satisfy «. And the values 5,,...,b, satisfy xo iff there is at least one value a of x such that the values a,by,...,b, satisfy a. Thus, to return to the particular language we have been using as an example, the formula xxy=z expresses the condition that the product of the values of x and y equals that of z. Then Wx(xxy=z) expresses the condition that if any number be multiplied by the value of y, the result is always equal to the value of z. (This is satisfied iff both y and z are assigned the value zero.) And 3x(xxy=z) expresses the condition that the value of z is a multiple of that of y. We could require Y to have two logical symbols Y and 3, called universal quantifier and existential quantifier respectively. But again we can make do with less. As a matter of fact, it is enough if # only has the symbol YW (universal quantifier); for the formula 7'\Wx—1@ behaves just as we want xa to behave, so that we can define the latter to be the former. Formal languages possessing just the equipment sketched above (with the possible omission of function symbols) are called first-order languages. *§ 3. Higher-order languages Logicians sometimes deal with (apparently) richer formal languages. For example, a second-order language has, in addition to the equipment of a first-order language, another type of variable, ranging not over the individuals of a structure U but over all sets of individuals (i.e., all subsets of the universe). Among the symbols of such a language there is a special one used to denote the relation of membership between an individual and a set of individuals. Quantifiers are allowed to apply not only to the individual variables (those ranging over individuals) but also to the new set variables. Thus, if « is a formula and X is a set variable, WXa is also a formula. Similarly, in a third-order language there are variables ranging over sets of sets of individuals, and so on for languages of still higher orders. However, most logicians agree that such languages are — at least in principle — dispensable. Indeed, let U be any structure and let B be a structure obtained from U in the following way. The universe of B consistsCH. 1, $4]. BASIC SYNTAX 15 of all individuals of U plus all sets of individuals of U. The basic operations of % are defined in such a way that when they are restricted to the universe of U (i.e., when they are applied to individuals of UW) they behave exactly as the corresponding basic operations of U. Finally, the basic relations of & are all the basic relations of U plus two additional relations: the property of being an individual of U, and the relation of membership between an individual of U and a set of individuals of U. Then any statement about U expressed in a second-order language with set variables, can easily be “translated” into a statement about % in a first-order language. In this sense, second-order languages are dispensable. A similar argument applies to other higher-order languages. We therefore do not lose much by confining our attention to first-order languages only. § 4. Basic syntax We shall now begin to put into practice some of the ideas discussed in §1 and §2. We proceed to describe an arbitrary first-order language #. Throughout this book, unless the contrary is stated, # will be kept fixed. We shall introduce various notions relating to Y. These will be labelled by the prefix “Y-” or by phrases “of #”, “in #”, etc. However, once such a notion has been introduced, we shall omit these labels except where they are needed for emphasis or clarity. The symbols of & are: (a) An infinite sequence of (individual) variables, namely Vy sVasVase+ + (b) For each natural number #, a set of n-ary function symbols. (c) For each positive natural number n, a set of n-ary predicate symbols. For at least one this set must be non-empty. (d) The connectives “1 (negation) and -> (implication). (e) The universal quantifier Y. The 0-ary function symbols (if any) are called (individual) constants. If & has the special binary predicate =, we say that Y is a language with equality. We stipulate that if & has at least one function symbol that is not an individual constant, then £ must be a language with equality.’ * Notice the difference between “=” and “=”. We use the former as a syntactic constant denoting the equality symbol of Y, while the latter is used in this book (as in most other mathematical texts) as short for “is the same as”. Pa16 BEGINNING MATHEMATICAL LOGIC [CH.1, 84 The variables, the connectives, the universal quantifier and = are called logical symbols. They are assumed to be the same in all first-order languages (or, in the case of =, in all first-order languages with equality). The function symbols and the predicate symbols other than = are called extralogical symbols and may differ from one language to another. Notice that we have fixed a particular ordering of the variables: v, is the n' variable’ of &. This ordering is called the alphabetic ordering of the variables. A finite (possibly empty) sequence of -symbols is called an ¥-string. (A given symbol may, of course, occur several times in the same string.) The length of a string is the total number of (occurrences of) symbols in it. In particular, the empty string has length 0; and any single symbol is a string of length 1. If s and T are strings, we define st as the string obtained by concatenat- ing s and 1, in this order. Similarly for three or more strings. If R=st, where R,S,T are strings, then s is an initial segment of R. If T is non-empty, then s is a proper initial segment of R. Similarly, T is a terminal segment of R, and it is a proper one if s is non-empty. Obviously, a string of length n has n+1 different initial segments (including the empty string and the entire string itself). We shall only be interested in two kinds of strings, called terms and formulas. &-terms are &-strings formed according to the following two rules: (1) Any &-string consisting of (a single occurrence of) a variable is an £-term. (2) If f is an n-ary function symbol of ¥ and t,...,t, are L-terms, then ft,...t, isan #-term. Notice that, for n=0, (2) says that any constant is a term. In a term ft,...t, formed according to (2), t,,...,t, are called arguments of f. By the degree of complexity of a term t (briefly deg t) we mean the total number of occurrences of function symbols in t. The stipulation that in a term formed according to (2) the arguments always follow the function symbol, makes it unnecessary for Y to have punctuation marks to indicate the grouping of arguments. To show this let us define the weight of a string s as the sum obtained by adding up —1 1 In accordance with a conyention introduced in §1, the bold arrow “>”, eg., is a syntactic constant belonging to our metalanguage and denoting a certain connective of & Similarly, “v,”, eg., is a syntactic constant denoting the first variable of Z.