Yves Nievergelt

Logic, Mathematics,
and Computer
Science
Modern Foundations
with Practical Applications
Second Edition
 www.pdfgrip.com

Logic, Mathematics, and Computer Science

www.pdfgrip.com
 www.pdfgrip.com

Yves Nievergelt

Logic, Mathematics,
and Computer Science
Modern Foundations with Practical
Applications

Second Edition

123
 www.pdfgrip.com

Yves Nievergelt
Department of Mathematics
Eastern Washington University
Cheney, WA, USA

Cover art excerpted from triadic truth tables from Charles Sanders Peirce’s Logic Notebook.

ISBN 978-1-4939-3222-1 ISBN 978-1-4939-3223-8 (eBook)

DOI 10.1007/978-1-4939-3223-8

Library of Congress Control Number: 2015949237

Mathematics Subject Classification (2010): Primary 03-01; Secondary: 68-01, 91-01

Springer New York Heidelberg Dordrecht London

© Springer Science+Business Media New York 2002, 2015
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made.

Printed on acid-free paper

Springer Science+Business Media LLC New York is part of Springer Science+Business Media (www.
springer.com)
 www.pdfgrip.com

Preface

This second edition, entitled Logic, Mathematics, and Computer Science: Modern
Foundations with Practical Applications, has been adapted from Foundations of
Logic and Mathematics: Applications to Computer Science and Cryptography, ©
2002 by Birkhäuser, from which Chapters 1–5 have been retained but extensively
revised. Chapters 6 and 7 have been added.
This text discusses the foundations where logic, mathematics, and computer
science begin. The intended readership consists of undergraduate students majoring
in mathematics or computer science who must learn such foundations either for their
own interest or for further studies. For a motivated reader, there are no technical
prerequisites: you need not know any technical subject to start reading this text.
Although the text does not focus on the history and philosophy of the founda-
tions, the material cites copious references to the literature, where the reader may
find additional historical context. Consulting such references is neither suggested
nor necessary to study the theory or to work on the exercises, but individual citations
document the material by original sources, and all the citations together provide a
guide to the variations and chronological developments of logic, mathematics, and
computer science. For example, Chapter 1 traces the origin of Truth tables to Charles
Sanders Peirce’s unpublished 1909 Logic Notebook on philosophy and points out
their applications over one half of a century later to the design of computers for use
on Earth and on board the Apollo lunar spacecraft.
Along informal arguments, this text also shows the corresponding purely sym-
bolic manipulations of formulae, because they clarify the reasoning [11] and can
reveal hitherto hidden logical properties, such as the mutual independence of
different patterns of reasoning, or the impossibility of some proofs within some
logics:
As for algebra [of logic], the very idea of the art is that it presents formulae which can be
manipulated, and that by observing the effects of such manipulation we find properties not
to be otherwise discerned (Charles Sanders Peirce [104, p. 182]).

If professionals are unable to learn some topics by any means other than the pure
manipulation of symbols, then it would seem unfair to claim that all learning must be
intuitive and hide from students such purely manipulative but successful methods.
v
 www.pdfgrip.com

vi Preface

The selection of topics also reflects major accomplishments from the twentieth
century: the foundation of all of mathematics, and later computer science, as well
as computer-assisted proofs of mathematical theorems, on a formal logic based on
only a few axioms, transformation rules, and postulates for set theory [47, 50, 54,
105, 139]. Also, while not written in formal logic, Nobel-Prize winning applications
to the social sciences rely on the same foundations, as shown in Chapter 7.
To introduce the foundations of logic, the provability theorem in Chapter 1
provides an algorithm to design proofs in propositional logic. Chapter 1 also
explains the concept of undecidability with multi-valued (“fuzzy”) logic and
presents a proof of unprovability. Chapter 2 introduces logical quantifiers.
A working knowledge of logical quantifiers is crucial for the study of basic concepts
in modern mathematical analysis and topology, such as the uniform convergence
of a sequence of equicontinuous functions. Continuing with the foundations of
mathematics, Chapter 3 presents a version of the Zermelo–Fraenkel set theory. At
the juncture of mathematics and computer science, Chapter 4 develops the concepts
of definition and proof by induction. Chapter 4 then uses induction with set theory
to define the integers and rational numbers and derive the associative, commutative,
and distributive laws, as well as algorithms, for their arithmetics. To give readers
some idea of topics at an intermediate level, Chapter 5 shows that in a well-formed
theory some paradoxes do not occur, while Chapter 6 completes the foundations of
set theory with the axiom of choice.
No extragalactic asteroid has yet been found with the universal laws of logic
engraved in it. Consequently, not just one logic but many different logics have
been invented. Different logics lead to different mathematics and different computer
sciences. However, the acid test for adopting a particular logic is its ability
to make predictions that are born by subsequent experiments. Formal logic is
thus a mathematical model of rational thought processes. In this aspect, logic,
mathematics, and computer science are experimental sciences. Only one logic has
passed all such tests, which is the one used throughout this text. Other logics
are outlined in Chapter 1 as a pedagogical device and to show some of their
shortcomings.

Acknowledgments

I thank Dr. Stephen P. Keeler at the Boeing Company for having corrected many
errors in the first edition. New and remaining errors are mine. I also thank Dr. Mary
Keeler for guiding me to Charles Sanders Peirce’s unpublished work on logic and
philosophy. I commend the editors at Birkhäuser and Springer, in particular, Ann
Kostant and Elizabeth Loew, for their hard work, patience, encouragements, and
positive attitude through the development and production of both texts.

Cheney, WA, USA Yves Nievergelt

30 June 2015
 www.pdfgrip.com

Contents

1 Propositional Logic: Proofs from Axioms and Inference Rules. . . . . . . . . 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 An Example Demonstrating the Use of Logic in
Real Life . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 The Pure Propositional Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Formulae, Axioms, Inference Rules, and Proofs. . . . . . . . . . . . . 5
1.3 The Pure Positive Implicational Propositional Calculus . . . . . . . . . . . . . . 9
1.3.1 Examples of Proofs in the Implicational Calculus . . . . . . . . . . . 9
1.3.2 Derived Rules: Implications Subject to Hypotheses . . . . . . . . . 11
1.3.3 A Guide for Proofs: an Implicational Deduction Theorem . . 14
1.3.4 Example: Law of Assertion from the Deduction Theorem . . 18
1.3.5 More Examples to Design Proofs of
Implicational Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.3.6 Another Guide for Proofs: Substitutivity of Equivalences . . 23
1.3.7 More Derived Rules of Inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.3.8 The Laws of Commutation and of Assertion. . . . . . . . . . . . . . . . . 27
1.3.9 Exercises on the Classical Implicational Calculus . . . . . . . . . . . 28
1.3.10 Equivalent Implicational Axiom Systems . . . . . . . . . . . . . . . . . . . . 29
1.3.11 Exercises on Kleene’s Axioms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.3.12 Exercises on Tarski’s Axioms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.4 Proofs by the Converse Law of Contraposition. . . . . . . . . . . . . . . . . . . . . . . . 32
1.4.1 Examples of Proofs in the Full Propositional Calculus . . . . . . 32
1.4.2 Guides for Proofs in the Propositional Calculus . . . . . . . . . . . . . 34
1.4.3 Proofs by Reductio ad Absurdum . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.4.4 Proofs by Cases. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.4.5 Exercises on Frege’s and Church’s Axioms . . . . . . . . . . . . . . . . . . 37
1.5 Other Connectives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.5.1 Definitions of Other Connectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.5.2 Examples of Proofs of Theorems with Conjunctions . . . . . . . . 38
1.5.3 Examples of Proofs of Theorems with Equivalences . . . . . . . . 41
1.5.4 Examples of Proofs of Theorems with Disjunctions. . . . . . . . . 44

vii
 www.pdfgrip.com

viii Contents

1.5.5 Examples of Proofs with Conjunctions and Disjunctions . . . 46

1.5.6 Exercises on Other Connectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
1.6 Patterns of Deduction with Other Connectives . . . . . . . . . . . . . . . . . . . . . . . . 48
1.6.1 Conjunctions of Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.6.2 Proofs by Cases or by Contradiction . . . . . . . . . . . . . . . . . . . . . . . . . 53
1.6.3 Exercises on Patterns of Deduction. . . . . . . . . . . . . . . . . . . . . . . . . . . 54
1.6.4 Equivalent Classical Axiom Systems. . . . . . . . . . . . . . . . . . . . . . . . . 55
1.6.5 Exercises on Kleene’s, Rosser’s, and Tarski’s Axioms . . . . . . 56
1.7 Completeness, Decidability, Independence, Provability,
and Soundness. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
1.7.1 Multi-Valued Fuzzy Logics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
1.7.2 Sound Multi-Valued Fuzzy Logics . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
1.7.3 Independence and Unprovability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
1.7.4 Complete Multi-Valued Fuzzy Logics. . . . . . . . . . . . . . . . . . . . . . . . 61
1.7.5 Peirce’s Law as a Denial of the Antecedent . . . . . . . . . . . . . . . . . . 62
1.7.6 Exercises on Church’s and Łukasiewicz’s
Triadic Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
1.8 Boolean Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
1.8.1 The Truth Table of the Logical Implication . . . . . . . . . . . . . . . . . . 63
1.8.2 Boolean Logic on Earth and in Space . . . . . . . . . . . . . . . . . . . . . . . . 65
1.9 Automated Theorem Proving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
1.9.1 The Provability Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
1.9.2 The Completeness Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
1.9.3 Example: Peirce’s Law from the Completeness Theorem . . . 69
1.9.4 Exercises on the Deduction Theorem . . . . . . . . . . . . . . . . . . . . . . . . 72
2 First-Order Logic: Proofs with Quantifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.2 The Pure Predicate Calculus of First Order . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.2.1 Logical Predicates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.2.2 Variables, Quantifiers, and Formulae . . . . . . . . . . . . . . . . . . . . . . . . . 77
2.2.3 Proper Substitutions of Free or Bound Variables . . . . . . . . . . . . 78
2.2.4 Axioms and Rules for the Pure Predicate Calculus . . . . . . . . . . 80
2.2.5 Exercises on Quantifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
2.2.6 Examples with Implicational and Predicate Calculi . . . . . . . . . 82
2.2.7 Examples with Pure Propositional and Predicate Calculi . . . 86
2.2.8 Other Axiomatic Systems for the Pure Predicate Calculus . . 87
2.2.9 Exercises on Kleene’s, Margaris’s, and Rosser’s Axioms . . . 89
2.3 Methods of Proof for the Pure Predicate Calculus . . . . . . . . . . . . . . . . . . . . 90
2.3.1 Substituting Equivalent Formulae . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2.3.2 Discharging Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
2.3.3 Prenex Normal Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
2.3.4 Proofs with More than One Quantifier . . . . . . . . . . . . . . . . . . . . . . . 96
2.3.5 Exercises on the Substitutivity of Equivalence. . . . . . . . . . . . . . . 97
 www.pdfgrip.com

Contents ix

2.4 Predicate Calculus with Other Connectives. . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

2.4.1 Universal Quantifiers and Conjunctions or Disjunctions . . . . 98
2.4.2 Existential Quantifiers and Conjunctions or Disjunctions . . . 100
2.4.3 Exercises on Quantifiers with Other Connectives. . . . . . . . . . . . 101
2.5 Equality-Predicates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
2.5.1 First-Order Predicate Calculi with an Equality-Predicate . . . 102
2.5.2 Simple Applied Predicate Calculi
with an Equality-Predicate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
2.5.3 Other Axiom Systems for the Equality-Predicate. . . . . . . . . . . . 106
2.5.4 Defined Ranking-Predicates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
2.5.5 Exercises on Equality-Predicates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3 Set Theory: Proofs by Detachment, Contraposition, and
Contradiction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.2 Sets and Subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.2.1 Equality and Extensionality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.2.2 The Empty Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.2.3 Subsets and Supersets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.2.4 Exercises on Sets and Subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.3 Pairing, Power, and Separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.3.1 Pairing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.3.2 Power Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
3.3.3 Separation of Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.3.4 Exercises on Pairing, Power, and Separation of Sets . . . . . . . . 126
3.4 Unions and Intersections of Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
3.4.1 Unions of Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
3.4.2 Intersections of Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
3.4.3 Unions and Intersections of Sets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3.4.4 Exercises on Unions and Intersections of Sets . . . . . . . . . . . . . . . 139
3.5 Cartesian Products and Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
3.5.1 Cartesian Products of Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
3.5.2 Cartesian Products of Unions and Intersections . . . . . . . . . . . . . 147
3.5.3 Mathematical Relations and Directed Graphs . . . . . . . . . . . . . . . 149
3.5.4 Exercises on Cartesian Products of Sets. . . . . . . . . . . . . . . . . . . . . . 153
3.6 Mathematical Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
3.6.1 Mathematical Functions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
3.6.2 Images and Inverse Images of Sets by Functions . . . . . . . . . . . . 159
3.6.3 Exercises on Mathematical Functions . . . . . . . . . . . . . . . . . . . . . . . . 162
3.7 Composite and Inverse Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
3.7.1 Compositions of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
3.7.2 Injective, Surjective, Bijective, and Inverse Functions . . . . . . 166
3.7.3 The Set of all Functions from a Set to a Set . . . . . . . . . . . . . . . . . . 171
3.7.4 Exercises on Injective, Surjective, and Inverse Functions . . . 173
 www.pdfgrip.com

x Contents

3.8 Equivalence Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174

3.8.1 Reflexive, Symmetric, Transitive, or
Anti-Symmetric Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
3.8.2 Partitions and Equivalence Relations . . . . . . . . . . . . . . . . . . . . . . . . . 175
3.8.3 Exercises on Equivalence Relations . . . . . . . . . . . . . . . . . . . . . . . . . . 179
3.9 Ordering Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
3.9.1 Preorders and Partial Orders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
3.9.2 Total Orders and Well-Orderings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
3.9.3 Exercises on Ordering Relations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
4 Mathematical Induction: Definitions and Proofs by Induction . . . . . . . . . 189
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
4.2 Mathematical Induction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
4.2.1 The Axiom of Infinity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
4.2.2 The Principle of Mathematical Induction . . . . . . . . . . . . . . . . . . . . 193
4.2.3 Definitions by Mathematical Induction . . . . . . . . . . . . . . . . . . . . . . 195
4.2.4 Exercises on Mathematical Induction . . . . . . . . . . . . . . . . . . . . . . . . 197
4.3 Arithmetic with Natural Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
4.3.1 Addition with Natural Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
4.3.2 Multiplication with Natural Numbers . . . . . . . . . . . . . . . . . . . . . . . . 200
4.3.3 Exercises on Arithmetic by Induction . . . . . . . . . . . . . . . . . . . . . . . . 203
4.4 Orders and Cancellations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
4.4.1 Orders on the Natural Numbers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
4.4.2 Laws of Arithmetic Cancellations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
4.4.3 Exercises on Orders and Cancellations . . . . . . . . . . . . . . . . . . . . . . . 213
4.5 Integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
4.5.1 Negative Integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
4.5.2 Arithmetic with Integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
4.5.3 Order on the Integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
4.5.4 Nonnegative Integral Powers of Integers . . . . . . . . . . . . . . . . . . . . . 226
4.5.5 Exercises on Integers with Induction . . . . . . . . . . . . . . . . . . . . . . . . . 228
4.6 Rational Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
4.6.1 Definition of Rational Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
4.6.2 Arithmetic with Rational Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
4.6.3 Notation for Sums and Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
4.6.4 Order on the Rational Numbers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
4.6.5 Exercises on Rational Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
4.7 Finite Cardinality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
4.7.1 Equal Cardinalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
4.7.2 Finite Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
4.7.3 Exercises on Finite Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
4.8 Infinite Cardinality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
4.8.1 Infinite Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
4.8.2 Denumerable Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
4.8.3 The Bernstein–Cantor–Schröder Theorem . . . . . . . . . . . . . . . . . . . 258
 www.pdfgrip.com

Contents xi

4.8.4 Denumerability of all Finite Sequences

of Natural Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
4.8.5 Other Infinite Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
4.8.6 Further Issues in Cardinality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
4.8.7 Exercises on Infinite Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
5 Well-Formed Sets: Proofs by Transfinite Induction with
Already Well-Ordered Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
5.2 Transfinite Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
5.2.1 Transfinite Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
5.2.2 Transfinite Construction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
5.2.3 Exercises on Transfinite Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
5.3 Transfinite Sets and Ordinals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
5.3.1 Transitive Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
5.3.2 Ordinals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
5.3.3 Well-Ordered Sets of Ordinals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
5.3.4 Unions and Intersections of Sets of Ordinals . . . . . . . . . . . . . . . . 275
5.3.5 Exercises on Ordinals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
5.4 Regularity of Well-Formed Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
5.4.1 Well-Formed Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
5.4.2 Regularity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
5.4.3 Exercises on Well-Formed Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
6 The Axiom of Choice: Proofs by Transfinite Induction . . . . . . . . . . . . . . . . . 283
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
6.2 The Choice Principle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
6.2.1 The Choice-Function Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
6.2.2 The Choice-Set Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
6.2.3 Exercises on Choice Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
6.3 Maximality and Well-Ordering Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
6.3.1 Zermelo’s Well-Ordering Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
6.3.2 Zorn’s Maximal-Element Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
6.3.3 Exercises on Maximality and Well-Orderings . . . . . . . . . . . . . . . 292
6.4 Unions, Intersections, and Products of Families of Sets . . . . . . . . . . . . . . 292
6.4.1 The Multiplicative Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
6.4.2 The Distributive Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
6.4.3 Exercises on the Distributive and Multiplicative
Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
6.5 Equivalence of the Choice, Zorn’s, and Zermelo’s Principles . . . . . . . . 295
6.5.1 Towers of Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
6.5.2 Zorn’s Maximality from the Choice Principle . . . . . . . . . . . . . . . 297
6.5.3 Exercises on Towers of Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
6.6 Yet Other Principles Related to the Axiom of Choice . . . . . . . . . . . . . . . . . 299
6.6.1 Yet Other Principles Equivalent to the Axiom of Choice. . . . 299
 www.pdfgrip.com

xii Contents

6.6.2 Consequences of the Axiom of Choice. . . . . . . . . . . . . . . . . . . . . . . 300

6.6.3 Exercises on Related Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
7 Applications: Nobel-Prize Winning Applications of Sets,
Functions, and Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
7.2 Game Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
7.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
7.2.2 Mathematical Models for The Prisoner’s Dilemma. . . . . . . . . . 305
7.2.3 Dominant Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
7.2.4 Mixed Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
7.2.5 Existence of Nash Equilibria for Two Players
with Two Mixed Strategies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
7.2.6 Exercises on Mathematical Games . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
7.3 Match Making . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
7.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
7.3.2 A Mathematical Model for Optimal Match Making . . . . . . . . . 316
7.3.3 An Algorithm for Optimal Match Making
with a Match Maker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
7.3.4 An Algorithm for Optimal Match Making
Without a Match Maker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
7.3.5 Exercises on Gale & Shapley’s Algorithms . . . . . . . . . . . . . . . . . . 320
7.3.6 Projects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
7.4 Arrow’s Impossibility Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
7.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
7.4.2 A Mathematical Model for Arrow’s
Impossibility Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
7.4.3 Statement and Proof of Arrow’s Impossibility Theorem . . . . 326
7.4.4 Exercises on Arrow’s Impossibility Theorem . . . . . . . . . . . . . . . . 330

Solutions to Some Odd-Numbered Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
 www.pdfgrip.com

Chapter 1
Propositional Logic: Proofs from Axioms
and Inference Rules

1.1 Introduction

This chapter introduces propositional logics, which consist of starting formulae

called axioms and rules of inference to derive from the axioms other formulae called
theorems. Axioms and rules of inference form a mathematical model of rational
thinking processes; theorems are their consequences. Different such logics, which
are also called calculi, rely on different axioms or different rules of inference.
For example, the Pure Positive Implicational Propositional Calculus focuses only
on the logical implication. The first few sections derive some of its theorems, for
instance, the transitivity of the logical implication and the law of commutation,
using the rule of Detachment with the laws of affirmation of the consequent and
self-distributivity of the logical implication taken as axiom schemata from Frege and
Łukasiewicz. A preliminary version of the Deduction Theorem for the propositional
calculus provides a method for designing proofs. Another section shows the mutual
equivalence of these axioms with Kleene’s axioms and Tarski’s axioms. Adding
the converse law of contraposition, subsequent sections focus on the Classical
Propositional Calculus, deriving the laws of double negation, reductio ad absurdum,
proofs by contradiction, and proofs by cases. Yet another section outlines the
equivalence of Frege and Łukasiewicz’s axioms with Church’s, Kleene’s, Tarski’s,
and Rosser’s axioms, respectively. A final section demonstrates the contrast between
logics that admit of a recipe for constructing proofs of all “valid” formulae, and
logics where some formulae are “valid” but unprovable. The prerequisite for this
chapter is the ability to read, compare, and substitute sequences and tables of
symbols. The goal of this chapter is merely to develop a working knowledge of
propositional logic:
Young man, in mathematics you don’t understand things, you just get used to them. — John
von Neumann, cited by Felix Smith, cited by Gary Zukav [148, p. 226].

© Springer Science+Business Media New York 2015 1

Y. Nievergelt, Logic, Mathematics, and Computer Science,
DOI 10.1007/978-1-4939-3223-8_1
 www.pdfgrip.com

2 1 Propositional Logic: Proofs from Axioms and Inference Rules

Logic, mathematics, and computations can be traced through several millennia

to ancient civilizations in Babylonia, China, and India. Documents attributed to
them show methods to calculate such items as taxes, the dimensions of altars,
and the dates of future solstices or eclipses. More complicated problems arose, for
example, the determination of the shapes and sizes of the Earth and the Moon, or the
distances from Earth to the Moon and the Sun. (For a survey of these ancient records,
consult, for instance, the texts by Dreyer [27], Evans [32], Neugebauer [98–100],
Van Brummelen [133], and van der Waerden [134].) The solutions to such problems
require methods more sophisticated than mere calculation, and hence arises the need
for a study of logic itself, which can be traced to the Greece of a couple of millennia
ago. This study of logic continues: the ambiguity of the classical verbal exposition
of logic and the need for unambiguity in complex situations led to algebraic and
symbolic treatments of logic, for instance, the Truth tables presented here. The point
of logic is not only “Truth” but also “relevance” to its users [24, p. 6]. Yet relevance
is subjective. Therefore, the following subsection presents an example that is not
only claimed but documented to be relevant in the real lives of real people.

1.1.1 An Example Demonstrating the Use of Logic in Real Life

For the purpose of an introduction, the following example demonstrates how logic
can help in resolving practical issues in real life, how questions arise about the
validity of logical methods to reach conclusions, and eventually what thought
processes are acceptable or successful in explanations and predictions. Yet an
understanding of this example will not be necessary for any of the subsequent
material.
1.1 Example. The planetary status of Pluto has been debated in newspapers:
Is Pluto really a planet? Like all civil wars, this has even split families apart [67].

The public’s interest in Pluto’s planet-hood is sufficient to devote an entire book

explaining the question to children [73]. Textbooks have classified Pluto as a planet
since its discovery in 1930 [69, p. 213], but the question remains whether Pluto
might rather be an asteroid [75]. Various answers rely on various definitions and on
logic [101]. For instance, one definition states that planets are bigger than moons.
This definition can be stated in terms of a hypothesis (abbreviated by H),
hypothesis H: a celestial object P is a planet,

a conclusion (abbreviated by C),

conclusion C: the celestial object P must be bigger than every moon,

and a logical implication of the form “if H, then C”:

If a celestial object P is a planet (if H),
then the celestial object P must be bigger than every moon (then C).
 www.pdfgrip.com

1.1 Introduction 3

The logical implication “if H, then C” (also worded as “H implies C”) is true by this
definition of planets. If the hypothesis H is also true, then the conclusion C is true,
too: C can be “detached from the implication” [80, p. 124].
With a logical implication (“if H, then C”) there are two other useful statements:
its converse (“if C, then H”) and its contraposition (“if not C, then not H”).
The hypothesis H can be tested by the contraposition “if not C, then not H”:
If a celestial object P is not bigger than every moon (if not C),
then P is not a planet (then not H).

In 1978 measurements revealed that Pluto was smaller than the Moon [69, p. 213];
consequently, Pluto would no longer be a planet, by the foregoing definition.
The definition “if H, then C” can also be tested in practice. For instance,
textbooks classify Mercury as a planet, but they also classify Ganymede as a moon
(of Jupiter), even though Mercury is smaller than Ganymede [69, p. 182 & 203].
Thus the statement “if Mercury is a planet, then Mercury is bigger than every moon”
is false. Therefore the foregoing definition “if H, then C” is false, and Pluto can
remain a planet. Logic has thus resolved the issue by revealing that the question
pertains not to the status of Pluto but to the definition of planets.
Other definitions and logical arguments have also been debated [73, 101]. Very
shortly thereafter, all existing definitions of planets were again put into question:
Scientifically, we are unable to define a planet in a sufficiently meaningful way such as to
include Pluto without including many other objects Œ: : : we are also unable to develop
a definition based on principles of astronomy and physics that excludes Pluto in any
nonarbitrary way [137, p. 219, summarizing chapter 14, p. 185–221].

The definition of planet will not be settled here, but some patterns of logical
reasoning will. For instance, the preceding arguments show that detachment and
contraposition are valid popular and scientific modes of reasoning; converse is not.
The preceding discussion contains one logical principle that has been so suc-
cessful that it remains widely accepted in theory and in practice: the law of
contraposition, which states that if an implication “if H, then C” holds, then its
contraposition “if not C, then not H” also holds.
The converse law of contraposition — which states that if the contraposition
“if not C, then not H” holds, then the implication “if H, then C” also holds — was
not used in the preceding discussion, and it is not a part of some logical systems.
In this example, the converse statement “if C, then H” (also worded as “C implies
H”) is false, because the Sun is bigger than every moon (C is true), but the Sun is
not a planet (H is false). Nevertheless, implications of the form “if H, then C” and
their converse “if C, then H” have been confused by professional scientists, so that
the difference between an implication and its converse bears being emphasized:
Leontovich explained to me why the paper could not be published in ZETP [“the main
Russian physics journal”]:
the paper claims that “A implies B” while every physicist knows examples showing
that B does not imply A; [...]
 www.pdfgrip.com

4 1 Propositional Logic: Proofs from Axioms and Inference Rules

An author, claiming that A implies B, must say whether the converse holds, otherwise the
reader who is not spoiled by the mathematical slang would interpret the claim as “A is
equivalent to B.” If mathematicians do not follow this rule, they are wrong [3, p. 619–620].

Such confusion among world-class scientists shows the necessity of specifying

patterns of rational thought processes exactly, for instance, as done in this chapter.
1.2 Remark (difficulties with real examples). Some examples of uses of logic may
enhance the effectiveness of the exposition. Such examples might consist of English
sentences, for example, “Pluto is a planet.” However, difficulties arise in determining
whether and why such a practical sentence is true or false. Indeed, a statement
as simple as “Pluto is a planet” can immediately be challenged to no end, as
demonstrated in example 1.1. Moreover, the word “planet” comes from the Greek
word for “wanderer” and therefore in antiquity the Sun was also considered a planet
[133, p. 3]. Thus the argument about such an elementary practical logical issue as
Pluto’s planetary status really has no ends in any direction, without any mention of
other more complex practical questions. Therefore, to focus on logic, mathematics,
and computing, instead of endlessly debatable issues, the following discussion will
also use “toy” examples with truth or falsity decided in advance.

1.2 The Pure Propositional Calculus

Propositional logic focuses on the derivation of conclusions from hypotheses, by

means of rules of inference and initial hypotheses called axioms that state patterns of
rational thinking precisely. Different axioms or different rules of inference may lead
to different logics or to mutually equivalent logics, for instance, the Pure Positive
Implicational Propositional Logic, which is a part of the full Classical Propositional
Logic.
Different readers may prefer prose or formulae. Jan Łukasiewicz stated concisely
the disadvantage of prose in formulating and solving logical problems:
Alles zerfließt in vagen philosophischen Spekulationen [80, p. 125]
(“everything melts into vague philosophical speculations”).

In contrast, formulae provide greater precision in complex situations [11]. To this

end, the Pure Propositional Calculus (also called Sentential Calculus) presented
here can be traced to Gottlob Frege [38, 39]. The adjective “pure” means that the
simplest (“atomic”) formulae are letters or symbols that do not denote anything:
mathematical logic is a meaningless game with symbols [108, p. xi].

Yet such symbols may later denote various types of atomic formulae that apply
to such various fields as algebra, arithmetic, geometry, or set theory, and therein
lies the power of the pure propositional calculus. Furthermore, purely symbolic
manipulations of formulae can reveal hitherto hidden logical properties:
 www.pdfgrip.com

1.2 The Pure Propositional Calculus 5

As for algebra [of logic], the very idea of the art is that it presents formulae which can be
manipulated, and that by observing the effects of such manipulation we find properties not
to be otherwise discerned (Charles Sanders Peirce, [104, p. 182]).

If famous logicians find manipulations of formulae indispensable in clarifying logic,

then such manipulations might also help nonspecialists in studying logic.

1.2.1 Formulae, Axioms, Inference Rules, and Proofs

Common instances of logical reasoning consist of sentences. For example, the first
axiom of Euclidean geometry is a sentence (paraphrased from [61, p. 3]):
For each pair of distinct points there exists exactly one line passing through both points.

Similarly, propositional logic starts with certain logical formulae called axioms.
The word “axiom” can mean “self-evident truth” but in the present context, which
focuses on patterns of reasoning, the word “axiom” means “initial” or “starting”
logical pattern [110, p. 55]. Different selections of axioms can lead to different
kinds of logic, but the present chapter focuses mainly on classical logic, which
has been successful for several millennia. Several choices for the initial axioms and
formulae lead to the same classical logic. Because the principal concepts of logic
consist of “negation” and “implication” several common choices of initial axioms
and formulae involve only the connectives for negation .:/ and implication .)/.
Also, to allow for applications in various areas, the pure propositional calculus
replaces the “atomic formulae” from algebra, arithmetic, geometry, or set theory
by general symbols called propositional variables or sentence symbols.
1.3 Definition (Well-formed formulae). Select two disjoint lists of symbols.
Every symbol from the first list of symbols, which may consist of one or more
letter(s) from a specified alphabet, P, Q, : : :, optionally with subscript(s) P[ , P[[ ; : : :,
superscript(s) P] , P]] , : : :, or “middlescript(s)” Pj, Pjj, : : :, is called a formulaic
letter. Such formulaic letters are not parts of the propositional calculus, but they
help in describing the following rules to define well-formed formulae.
Also, every symbol from the second list of symbols, which may consist of one
or more letter(s) from a specified alphabet, A, B, : : :, optionally with subscript(s)
A[ , A[[ , : : :, superscript(s) A] , A]] , : : :, or “middlescript(s)” Aj, Ajj, : : :, is called a
propositional variable or a sentence symbol [31, p. 17]. (Propositional variables
may later be replaced by atomic formulae specific to applications.)
Every propositional variable is a well-formed formula. For all well-formed
formulae P and Q, the following strings of symbols are also well-formed formulae:

.W1/ :.P/ (read “not P”);

.W2/ .P/ ) .Q/ (read “P implies Q” or “if P, then Q”):
 www.pdfgrip.com

6 1 Propositional Logic: Proofs from Axioms and Inference Rules

Furthermore, only strings of symbols built from letters or variables through

applications of the rules W1 and W2 can be well-formed formulae. Equivalent
definitions apply to other connectives and to prefix and postfix notations.
The use of parentheses in definition 1.3 conforms to [22, p. 7]. and [114, p. 185].
Parentheses without rules of precedence reflect the motto
more parentheses but less memorizing [114, p. 216].

By definition 1.3, a string of symbols such as .P/ ) .Q/ is not yet a well-
formed formula of the propositional calculus. It only becomes so after P and Q
have been replaced by propositional variables or well-formed formulae, for instance,
.A/ ) .B/. In this section, any capital letter may denote a propositional letter,
variable, or formula. In subsequent sections, however, the distinction may matter.
From definition 1.3, propositional letters or variables are “atoms” or “atomic”
propositional formulae in the sense that they are the simplest well-formed proposi-
tional formulae [72, p. 5], in contrast to more elaborate “composite” or “compound”
formulae, also called propositional forms, built from rules W1 and W2.
Several choices of well-formed propositional formulae can serve as axioms.
A system that remains concise and differentiates the roles of separate connectives
consists of the following three axioms. Subsection 1.3.10 explains their popularity
[18, 20, p. 119], [84, p. 49], [81, p. 31], [85, p. 165], [122, p. 165].
1.4 Definition (Jan Łukasiewicz’s axioms). A logical formula is an axiom of the
version of the classical propositional calculus considered here if and only if it is one
of the following three formulae, attributed to Łukasiewicz [62, p. 29], where P, Q,
and R may be any well-formed propositional formulae. The first two axioms are also
all the axioms of the Pure Positive Implicational Propositional Calculus:
Axiom P1 .P/ ) Œ.Q/ ) .P/.
Axiom P2 f.P/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g.
Axiom P3 fŒ:.Q/ ) Œ:.P/g ) Œ.P/ ) .Q/.
The first two axioms are also in Frege’s work [38], [39, p. 137, eq. (1) & (2)],
where they reflect a common mathematical model of rational thinking:
• Axiom P1, .P/ ) Œ.Q/ ) .P/, is called the law of affirmation of the
consequent. In Frege’s (translated) words, axiom P1 states that
If a proposition ŒP holds, it also holds in case an arbitrary proposition ŒQ holds [39,
p. 137].

• Axiom P2, f.P/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g, is called
the law of self-distributivity of implication. In Frege’s words, axiom P2 states
that
If a proposition ŒR is the necessary consequence of two propositions (ŒQ and ŒP),
that is, if Œ.P/ ) f.Q/ ) .R/g, and if the first term ŒQ is again the necessary
consequence of the other ŒP, then the proposition ŒR is the necessary consequence of
the last proposition ŒP alone [39, p. 139].
 www.pdfgrip.com

1.2 The Pure Propositional Calculus 7

• Axiom P3, fŒ:.Q/ ) Œ:.P/g ) Œ.P/ ) .Q/, called the converse law of
contraposition [18, 20, p. 119], states that a contraposition, Œ:.Q/ ) Œ:.P/
suffices to establish a classical logical implication .P/ ) .Q/.
The converse law of contraposition distinguishes classical logic from several
other systems of logic, for instance, Hilbert’s Positive Propositional Calculus,
Brouwer & Heyting’s Intuitionistic Logic, and Kolmogorov & Johansson’s Minimal
Logic [18, 26, p. 140–146]. Still, all these logical systems include Łukasiewicz’s
first two axioms, P1 and P2.
The propositional calculus includes the following concepts of theorem and proof.
1.5 Definition. A well-formed propositional form is a theorem of a propositional
calculus if and only if it is obtained by the following rules of inference:
1.6 Rule (Axioms).
Every axiom (of a logic) is a theorem (of the same logic).

1.7 Rule (“Modus Ponens” (abbreviated by M. P.), or “Detachment”).

For all propositional forms H and C,
if H is a theorem and
if .H/ ) .C/ is a theorem,
then C is a theorem.

The name of this rule will be printed here as “Detachment” to avoid unintended
awkward sentences. With Detachment, H is the minor premiss while .H/ ) .C/
is the major premiss (so spelled to distinguish its plural from “premises” [18, p. 1,
footnote 3]). Rule 1.7 states that if a hypothesis H and an implication .H/ ) .C/
hold, then the conclusion C may be “detached from the implication” (“von der
Implikation abgetrennt” in Łukasiewicz’s language [80, p. 124]).
1.8 Remark. Each use of the rule of Detachment requires two previously proved
well-formed formula: a proved hypothesis H and a proved implication .H/ ) .C/.
Definitions 1.3 and 1.4 allow P, Q, and R to denote any propositional letters
or well-formed propositional formulae, so that axioms P1–P3 are templates, or
schemas, to generate axioms. Alternatively, allowing only propositional variables
in axioms P1–P3 but introducing the substitution rule 1.9 gives equivalent axioms:
1.9 Rule (Substitution).
For each propositional variable K in a theorem R (which is a propositional form),
and for each well-formed propositional form L,
the propositional form obtained by replacing in R every occurrence of K by L is again a
theorem.

A proof, or deduction, of a theorem R is a finite sequence of logical formulae

P; Q; : : : ; R, in which each formula is either a substitution in an axiom or in a
previously proven formula, or results from the rule of Detachment (Modus Ponens).
1.10 Example (Substitution). The formula .A/ ) Œ.B/ ) .A/ is an instance of
axiom P1; hence it is a theorem. Substituting :.C/ for A in .A/ ) Œ.B/ ) .A/
 www.pdfgrip.com

8 1 Propositional Logic: Proofs from Axioms and Inference Rules

yields Œ:.C/ ) f.B/ ) Œ:.C/g, which is another instance of axiom P1, and
hence also a theorem. Because such substitutions in an axiom yield other axioms,
each axiom is also called an axiom schema. Thus both axioms .A/ ) Œ.B/ ) .A/
and Œ:.C/ ) f.B/ ) Œ:.C/g result from the axiom schema P1.
1.11 Definition. For every logical formula R, the symbol ` (called a “turnstyle”
and read “yield(s)” [110, p. 57]) and the notation

`R

means that there exists a proof of R. An alternative notation, P1, P2, P3 ` R, also
specifies the list of axioms, here P1, P2, P3, from which R is a theorem.
More generally, for all logical formulae P and R, the notation P ` R means that
with P added to the list of axioms, there exists a proof of R. The corresponding
alternative notation, P1, P2, P3, P ` R, again specifies the list of axioms. In other
words, R is a theorem for the logic with axioms P1, P2, P3, P. With a different
terminology, P ` R means that R is derivable from P and the axioms.
Yet more generally, for all logical formulae P; Q; : : : ; R, either notation
P; Q; : : : ` R or P1, P2, P3, P; Q; : : : ` R, means that with P; Q; : : : added to
the list of axioms, there exists a proof of R. In other words, R is a theorem for
the logic with axioms P1, P2, P3, P; Q; : : : The formula R is then derivable from
P; Q; : : :, if and only if P; Q; : : : ` R. In the notation of Smullyan [117, p. 17] and
Stolyar [123, p. 63], P; Q; : : : ` R is also denoted by

P; Q; : : :
:
R

Verifying a proof reduces to checking that each step conforms to the foregoing
definition of proof. In contrast, constructing a proof may require some creativity,
which may involve trying some rules and some axioms in various combinations,
some of which may fail whereas others may succeed [72, p. 55, lines 1–3], [114,
p. 31]. For the propositional calculus presented here, there is an algorithm (a recipe)
to design proofs, but its justification first requires most of the proofs presented here
[123, p. 193–197]. Moreover, the algorithm is cumbersome and would generate
proofs longer than the ones explained here. Nevertheless, the collection of all the
proofs shown here will demonstrate the steps that the algorithm would involve.
With such an understanding of the algorithm, a user might then automate the
algorithm with a computer [47, 50, 54, 139]. Furthermore, the following proofs also
provide some practice in creating proofs without using an algorithm, a practice that
corresponds more closely to the situation in mathematics, and for which there can
be no algorithm [46].
 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 9

1.3 The Pure Positive Implicational Propositional Calculus

With the two rules of inference (Detachment and substitution), the first two axioms
of classical propositional calculus (P1 and P2) pertain only to logical implications
.)/; they form the Pure Positive Implicational Propositional Calculus, which is
common to other logics [18,  29, p. 161]. In contrast, the concept of negation .:/
does not belong to the Positive Implicational Propositional Calculus. Axioms about
the negation, for instance, axiom P3, differentiate classical logic from other logics.

1.3.1 Examples of Proofs in the Implicational Calculus

The following three theorems provide examples of proofs about or in the Pure
Positive Implicational Propositional Calculus, which is the propositional calculus
with implications but no negations. The first theorem is called a “derived rule”
(of inference) because it involves a hypothesis, T, which can be any axiom or
previous theorem. Such a derived rule of inference is a theorem about rather than
in the implicational calculus, but it provides a recipe to shorten subsequent proofs.
Specifically, theorem 1.12 shows a derivation of .S/ ) .T/ from T and axiom P1.
The proof of theorem 1.12 is also a building block of the Deduction Theorem 1.22.
1.12 Theorem (derived rule). For each well-formed formula S and for each
theorem T, the implication .S/ ) .T/ is a theorem: P1, T ` .S/ ) .T/.
Proof. Apply axiom P1 and Detachment as follows:
`T hypothesis (minor premiss),
` .T/ ) Œ.S/ ) .T/ substitution in axiom P1 (major premiss),
` .S/ ) .T/ Detachment and preceding two formulae.
Thus .S/ ) .T/ is a theorem derivable from axiom P1, the hypothesis T, and
Detachment. t
u
The second theorem uses axiom P2 and also involves a hypothesis, .H/ )
Œ.K/ ) .L/, which can be any formula of this form that has already been proved.
1.13 Theorem (derived rule). For all well-formed formulae H, K, L, if
.H/ ) Œ.K/ ) .L/
is a theorem, then
Œ.H/ ) .K/ ) Œ.H/ ) .L/
is also a theorem: P2, .H/ ) Œ.K/ ) .L/ ` Œ.H/ ) .K/ ) Œ.H/ ) .L/.
Proof. Apply axiom P2 and Detachment:
` .H/ ) Œ.K/ ) .L/ hypothesis,
` f.H/ ) Œ.K/ ) .L/g ) fŒ.H/ ) .K/ ) Œ.H/ ) .L/g axiom P2,
` Œ.H/ ) .K/ ) Œ.H/ ) .L/ Detachment.
t
u
 www.pdfgrip.com

10 1 Propositional Logic: Proofs from Axioms and Inference Rules

The third theorem (1.14) involves no hypotheses other than the axioms. Thus
theorem 1.14 is a theorem of the Pure Implicational Propositional Calculus. More
accurately, theorem 1.14 is a theorem schema representing a different theorem for
each different formula P. The proof of theorem 1.14 is also a building block of the
Deduction Theorem 1.22.
1.14 Theorem (reflexive law of implication). For each well-formed propositional
formula P, the formula

.P/ ) .P/

is a theorem derivable from axioms P1, P2, and Detachment.

Proof. Apply axioms P1, theorem 1.13, and Detachment, with lines numbered for
clarity (the line numbers are not parts of the proof):
Q
‚ …„ ƒ
1 ` .„ƒ‚…
P / ) fŒ.P/ ) .P/ ) .„ƒ‚…
P /g substitution in axiom P1,
„ ƒ‚ …
H K L

H K H L
‚…„ƒ ‚ …„ ƒ ‚…„ƒ ‚…„ƒ
2 ` f. P / ) Œ.P/ ) .P/g ) Œ. P / ) . P / theorem 1.13,

3 ` .P/ ) Œ.P/ ) .P/ substitution in axiom P1,

4 ` .P/ ) .P/ lines 2, 3, Detachment.
A formal proof would replace line 2 by the details of the proof of theorem 1.13,
labeled here as lines 2a and 2b, which gives the following proof:
P Q R
‚…„ƒ ‚ …„ ƒ ‚…„ƒ
1 ` . P /)fŒ.P/).P/). P /g P1,
„ ƒ‚ …
S

P Q R P Q P R
 ‚…„ƒ ‚ …„ ƒ ‚…„ƒ   ‚…„ƒ ‚ …„ ƒ ‚…„ƒ ‚…„ƒ 
2a ` . P /)fŒ.P/).P/). P /g ) f. P /)Œ.P/).P/g)Œ. P /). P / P2,
„ ƒ‚ … „ ƒ‚ …
S U

U
‚ …„ ƒ
2b ` f.P/)Œ.P/).P/g)Œ.P/).P/ 1, 2a, M.P.,
„ ƒ‚ … „ ƒ‚ …
V W

V
‚ …„ ƒ
4 ` .P/)Œ.P/).P/ P1,

5 ` .P/).P/ 2b, 4, M.P.

„ ƒ‚ …
W
Thus, P1, P2 ` .P/ ) .P/. t
u
 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 11

1.3.2 Derived Rules: Implications Subject to Hypotheses

The following theorems about the Pure Positive Implicational Propositional

Calculus are variants of the transitivity of implications: they shorten sequences of
implications into one implication, from the initial hypothesis to the last conclusion.
The same theorems also allow for the substitution of any well-formed propositional
formula by any logically equivalent formula. Moreover, their proofs constitute steps
of an algorithm known as the Deduction Theorem 1.22. For instance, the following
derived rules of inference extend the rule of Detachment to situations where K and
.K/ ) .L/ might hold only under some hypothesis H: then the conclusion L also
holds under the same hypothesis H.
1.15 Theorem (derived rule). For all well-formed formulae H, K, L, if
.H/ ) .K/ and
.H/ ) Œ.K/ ) .L/
are theorems, then
.H/ ) .L/
is also a theorem. Thus,
.H/ ) .K/; .H/ ) Œ.K/ ) .L/ ` .H/ ) .L/.
Proof. Apply theorem 1.13 and Detachment:
1 ` .H/ ) Œ.K/ ) .L/ hypothesis;
2 ` Œ.H/ ) .K/ ) Œ.H/ ) .L/ theorem 1.13, major premiss;
3 ` .H/ ) .K/ hypothesis, minor premiss;
4 ` .H/ ) .L/ lines 3, 2, and Detachment.
t
u
The following theorem simplifies the use of theorem 1.15 if one of the logical
implications is a theorem. In particular, in an implication .K/ ) .L/, theorem 1.16
allows for the substitution of a stronger hypothesis H implying (or equivalent to) K.
Alternatively, in an implication .H/ ) .K/, theorem 1.16 allows for the substitution
of a weaker conclusion L implied by (or equivalent to) K.
1.16 Theorem (derived rule). For all well-formed formulae H, K, L, if
.H/ ) .K/ and
.K/ ) .L/
are theorems, then
.H/ ) .L/
is also a theorem. Thus, .H/ ) .K/; .K/ ) .L/ ` .H/ ) .L/.
Proof. Apply theorems 1.12 and 1.15:
` .K/ ) .L/ hypothesis,
` .H/ ) Œ.K/ ) .L/ theorem 1.12,
` .H/ ) .K/ hypothesis,
` .H/ ) .L/ theorem 1.15.
t
u
 www.pdfgrip.com

12 1 Propositional Logic: Proofs from Axioms and Inference Rules

Similarly, the following theorem simplifies the use of theorem 1.15 if one of the
components is already a theorem.
1.17 Theorem (derived rule). For all well-formed formulae H, K, L, if
.K/ and
.H/ ) Œ.K/ ) .L/
are theorems, then
.H/ ) .L/
is also a theorem. Thus, K; .H/ ) Œ.K/ ) .L/ ` .H/ ) .L/.
Proof. Apply theorems 1.12 and 1.15:
`K hypothesis,
` .H/ ) .K/ theorem 1.12,
` .H/ ) Œ.K/ ) .L/ hypothesis,
` .H/ ) .L/ theorem 1.15.
t
u
The following theorem demonstrates how the rule of Detachment extends to a
sequence of several logical implications.
1.18 Theorem (derived rule). For all well-formed formulae P, Q, R, S, if
.P/ ) .Q/,
.P/ ) Œ.Q/ ) .R/, and
.P/ ) Œ.R/ ) .S/
are theorems, then
.P/ ) .S/
is also a theorem. Thus,

.P/ ) .Q/; .P/ ) Œ.Q/ ) .R/; .P/ ) Œ.R/ ) .S/ ` .P/ ) .S/:

Proof. Apply theorem 1.15 twice:

` .P/ ) .Q/ hypothesis,
` .P/ ) Œ.Q/ ) .R/ hypothesis,
` .P/ ) .R/ theorem 1.15,
` .P/ ) Œ.R/ ) .S/ hypothesis,
` .P/ ) .S/ theorem 1.15.
t
u
Similar theorems hold for a sequence of more than three consecutive implica-
tions, but their need will not arise here. The following theorem simplifies the use of
theorem 1.18 if two of the logical implications are already theorems.
1.19 Theorem (derived rule). For all well-formed formulae P, Q, R, S, if
.P/ ) .Q/,
.Q/ ) .R/, and
.R/ ) .S/
 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 13

are theorems, then

.P/ ) .S/
is also a theorem. Thus,

.P/ ) .Q/; .Q/ ) .R/; .R/ ) .S/ ` .P/ ) .S/:

Proof. Apply theorem 1.16 twice:

` .P/ ) .Q/ hypothesis,
` .Q/ ) .R/ hypothesis,
` .P/ ) .R/ theorem 1.16, t
u
` .R/ ) .S/ hypothesis,
` .P/ ) .S/ theorem 1.16.
The following theorems demonstrate the transitivity of implications subject to a
common hypothesis.
1.20 Theorem (derived rule). For all well-formed formulae H, K, L, M, if
.H/ ) Œ.K/ ) .L/ and
.H/ ) f.K/ ) Œ.L/ ) .M/g
are theorems, then
.H/ ) Œ.K/ ) .M/
is also a theorem. Thus,

.H/ ) Œ.K/ ) .L/; .H/ ) f.K/ ) Œ.L/ ) .M/g ` .H/ ) Œ.K/ ) .M/:

Proof. Use axiom P2 with theorems 1.16 and 1.15:

` .H/ ) f.K/ ) Œ.L/ ) .M/g hypothesis,
` f.K/ ) Œ.L/ ) .M/g ) fŒ.K/ ) .L/ ) Œ.K/ ) .M/g axiom P2,
` .H/ ) fŒ.K/ ) .L/ ) Œ.K/ ) .M/g theorem 1.16;
` .H/ ) Œ.K/ ) .L/ hypothesis,
` .H/ ) Œ.K/ ) .M/ theorem 1.15.
t
u
1.21 Theorem (derived rule). For all well-formed formulae H, K, L, M, if
.H/ ) .K/,
.H/ ) .L/, and
.K/ ) Œ.L/ ) .M/
are theorems, then
.H/ ) .M/
is also a theorem. Thus,

.H/ ) .K/; .H/ ) .L/ .K/ ) Œ.L/ ) .M/ ` .H/ ) .M/:

 www.pdfgrip.com

14 1 Propositional Logic: Proofs from Axioms and Inference Rules

Proof. Apply theorems 1.12, 1.13, and 1.15:

1 ` .K/ ) Œ.L/ ) .M/ hypothesis,
2 ` .H/ ) f.K/ ) Œ.L/ ) .M/g theorems 1.12,
3 ` Œ.H/ ) .K/ ) f.H/ ) Œ.L/ ) .M/g theorem 1.13, major premiss,
4 ` .H/ ) .K/ hypothesis, minor premiss,
5 ` .H/ ) Œ.L/ ) .M/ lines 4, 3, and Detachment,
6 ` .H/ ) .L/ hypothesis,
7 ` .H/ ) .M/ theorem 1.15.
t
u

1.3.3 A Guide for Proofs: an Implicational Deduction Theorem

In general the question arises, how to find a proof of a theorem. One guide to
design a proof of an implication .H/ ) .C/, where H and C denote well-formed
propositional formulae, begins by deriving a proof for a derivation H ` C of the
conclusion C from the hypothesis H and the axioms. For instance, all the proofs
of derived rules in subsection 1.3.2 are examples of such derivations. The method
for designing a proof then proceeds to “discharge” the hypothesis H to produce a
proof of .H/ ) .C/, as described in the proof of the Deduction Theorem 1.22,
which is not in but about the implicational calculus. More generally, from any proof
that a logical proposition S is derivable from proved hypotheses H; K; : : : ; M; N the
Deduction Theorem provides a recipe to turn that proof into a proof of

.H/ ) f.K/ ) : : : .M/ ) Œ.N/ ) .S/ : : :g:

The Deduction Theorem presented here is also a part of an algorithm to design

proofs within the full Classical Propositional Calculus.
1.22 Theorem (Deduction Theorem for the Pure Classical Propositional Calcu-
lus, preliminary version). With any axiom system for which axioms P1, P2, and
.P/ ) .P/ are axioms or theorems (or schema thereof), there is an algorithm to
transform any proof of

H; K; : : : ; M; N ` S

within the Classical Propositional Calculus into a proof of

.H/ ) f.K/ ) : : : .M/ ) Œ.N/ ) .S/ : : :g:

Proof (Outline). The Deduction Theorem removes the hypotheses one at a time, for
instance, beginning with the last one listed, here N, from all the steps in the proof.
 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 15

(D1) If the step ` P in the initial proof is the hypothesis N being removed, then
in the new proof the Deduction Theorem replaces the old step
` N (current hypothesis)
by a complete proof of .N/ ) .N/, for instance, that of theorem 1.14:
` .N/ ) fŒ.N/ ) .N/ ) .N/g axiom P1,
 
` .N/ ) fŒ.N/ ) .N/ ) .N/g  axiom P2,
) f.N/ ) Œ.N/ ) .N/g ) Œ.N/ ) .N/

` f.N/ ) Œ.N/ ) .N/g ) Œ.N/ ) .N/ Detachment,

` .N/ ) Œ.N/ ) .N/ axiom P1,

` .N/ ) .N/ Detachment.

(D2) If a step ` P of the initial proof is a substitution in one of the axioms, or
in a previously proved theorem, or one of the hypotheses other than the one N
being removed here, then in the new proof the Deduction Theorem replaces the
old step
` P (axiom or hypothesis)
by a complete proof of .N/ ) .P/, for instance, as in theorem 1.12:
`P axiom or hypothesis,
` .P/ ) Œ.N/ ) .P/ axiom P1,
` .N/ ) .P/ Detachment.
(D3) If the step ` P is derived in the initial proof by Detachment from previously
proved propositions M and .M/ ) .P/, then (D2) and (D1) allow for their
replacement by complete proofs of .N/ ) .M/ and .N/ ) Œ.M/ )
.P/ respectively. Specifically, in the new proof, the Deduction Theorem then
replaces the old steps
` .M/ ) .P/ (previously proven True),
`M (previously proven True),
`P (Detachment),
by a complete proof that .N/ ) .P/, for instance, as in theorem 1.15:
` .N/ ) Œ.M/ ) .P/ theorem 1.12,

` f.N/ ) Œ.M/ ) .P/g axiom P2 : : :

) fŒ.N/ ) .M/ ) Œ.N/ ) .P/g : : : continued,

` fŒ.N/ ) .M/ ) Œ.N/ ) .P/g Detachment,

` .N/ ) .M/ theorem 1.12,
` .N/ ) .P/ Detachment,
with the proof of each instance of theorem 1.12 completely written out.
(D4) If a step ` P results from a previously proved derived rule, then it may be
necessary first to replace the step ` P by a complete proof of the derived rule,
and only then to replace each step of the proof of the derived rule as instructed
by directives (D1), (D2), (D3).
 www.pdfgrip.com

16 1 Propositional Logic: Proofs from Axioms and Inference Rules

Still with the hypothesis N, after the completion of any operation (D1)–(D3) on step
P, the Deduction Theorem then performs the same operations (D1)–(D3) on each of
the following steps, Q; : : : ; R. After the completion of operations (D1)–(D3) on all
the steps P; Q; : : : ; R, for the hypothesis N, the Deduction Theorem gives a proof of

H; K; : : : ; M ` Œ.N/ ) .S/:

Then the Deduction Theorem repeats the whole process with the preceding hypothe-
ses, H; : : : ; M. The Deduction Theorem terminates with a proof of

.H/ ) f.K/ ) : : : .M/ ) Œ.N/ ) .S/ : : :g:

The general case follows by several applications of the previous cases, in a way
that may be specified more explicitly after the availability of the Principle of
Mathematical Induction in chapter 4. t
u
Example 1.23 shows how to use theorem 1.22.
1.23 Example (Tarski’s axiom II). To prove Tarski’s axiom II, f.P/ ) Œ.P/ )
.Q/g ) Œ.P/ ) .Q/, define H and C by

f.P/ ) Œ.P/ ) .Q/g ) Œ.P/ ) .Q/:

„ ƒ‚ … „ ƒ‚ …
H C

Phase 1: a proof of H ` C.

For H ` C, in other words, to derive C from H, substitute H in axiom P2:

1 ` .P/)Œ.P/).Q/ hypothesis,
„ ƒ‚ …
H
2 ` f.P/)Œ.P/).Q/g)fŒ.P/).P/)Œ.P/).Q/g axiom P2,
„ ƒ‚ … „ ƒ‚ …
H L
L
‚ …„ ƒ
3 ` Œ.P/).P/)Œ.P/).Q/ 1, 2, Detachment,
„ ƒ‚ … „ ƒ‚ …
K C
4 ` .P/).P/ theorem 1.14,
„ ƒ‚ …
K
5 ` .P/).Q/ 3, 4, Detachment,
„ ƒ‚ …
C
which completes the derivation H ` C of C from H.
 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 17

Phase 2: a proof of .H/ ) .C/ from H ` C.

To transform this derivation .H ` C/ into a proof of .H/ ) .C/, apply to each step
of phase 1 the procedure described in the Deduction Theorem (1.22):
` f.P/)Œ.P/).Q/g)f.P/)Œ.P/).Q/g (D1),
„ ƒ‚ … „ ƒ‚ …
H H
H
‚ …„ ƒ
` f.P/)Œ.P/).Q/g
 
) f.P/)Œ.P/).Q/g)fŒ.P/).P/)Œ.P/).Q/g (D2), P2,
„ ƒ‚ … „ ƒ‚ …
H L
H L
‚ …„ ƒ ‚ …„ ƒ
` f.P/)Œ.P/).Q/g)fŒ.P/).P/)Œ.P/).Q/g (D3),
„ ƒ‚ … „ ƒ‚ …
K C
H
‚ …„ ƒ
` f.P/)Œ.P/).Q/g)Œ.P/).P/ (D2), 1.12,
„ ƒ‚ …
K
H
‚ …„ ƒ
` f.P/)Œ.P/).Q/g)Œ.P/).Q/ (D3),
„ ƒ‚ …
C
which completes the proof of f.P/ ) Œ.P/ ) .Q/g ) Œ.P/ ) .Q/. A completely
formal proof would also expand each of the steps just listed into its own proof,
as specified in the proof of the Deduction Theorem 1.22, for instance, expanding
each use of the directive (D3), which uses theorem 1.15, into a complete proof of
theorem 1.15.
Although theorem 1.22 can produce lengthy proofs, the resulting long proofs can
also suggest shorter proofs, for instance, as in theorem 1.24.
1.24 Theorem. The formula f.P/ ) Œ.P/ ) .Q/g ) Œ.P/ ) .Q/ is a theorem.
Proof. Apply axiom P2 with theorems 1.14 and 1.17:
` .P/ ) .P/ theorem 1.14,
„ ƒ‚ …
K

` f.P/ ) Œ.P/ ) .Q/g ) fŒ.P/ ) .P/ ) Œ.P/ ) .Q/g axiom P2,

„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
H K L

` f.P/ ) Œ.P/ ) .Q/g ) Œ.P/ ) .Q/ theorem 1.17.

„ ƒ‚ … „ ƒ‚ …
H L
t
u
Theorem 1.24 is also used in the form of a derived rule.
1.25 Theorem (derived rule). If .P/ ) Œ.P/ ) .Q/ is a theorem, then .P/ )
.Q/ is also a theorem.
 www.pdfgrip.com

18 1 Propositional Logic: Proofs from Axioms and Inference Rules

Proof. Apply theorem 1.24 and Detachment:

` f.P/ ) Œ.P/ ) .Q/g ) Œ.P/ ) .Q/ theorem 1.24,
„ ƒ‚ … „ ƒ‚ …
H C
` .P/ ) Œ.P/ ) .Q/ hypothesis,
„ ƒ‚ …
H
` .P/ ) .Q/ Detachment.
„ ƒ‚ …
C
t
u
The design of proofs of formulae of the form .A/ ) Œ.B/ ) .C/ can start with a
derivation A; B ` C of C from two hypotheses A and B. With A treated as an axiom,
and B as the hypothesis H, a first application of the Deduction Theorem (1.22)
leads to a derivation A ` .B/ ) .C/. Then, with A as the hypothesis H, a second
application of the Deduction Theorem (1.22) yields a proof of .A/ ) Œ.B/ ) .C/.
The general case with several hypotheses follows by several applications of the
Deduction Theorem (1.22), in a way that may be specified more explicitly after the
availability of the Principle of Mathematical Induction in chapter 4.

1.3.4 Example: Law of Assertion from the Deduction Theorem

The following proof shows the use of the Deduction Theorem in designing proofs.
1.26 Theorem. The law of assertion .A/ ) fŒ.A/ ) .B/ ) .B/g is a theorem.
Proof. A finished proof can proceed as follows:
` Œ.A/ ) .B/ ) Œ.A/ ) .B/ theorem 1.14,
` .A/ ) fŒ.A/ ) .B/ ) Œ.A/ ) .B/g theorem 1.12,
` .A/ ) fŒ.A/ ) .B/ ) .A/g axiom P1,
` .A/ ) fŒ.A/ ) .B/ ) .B/g theorem 1.20.
The following considerations explain how to design such a proof.
The formula .A/ ) fŒ.A/ ) .B/ ) .B/g has the pattern .H/ ) Œ.K/ ) .S/
of the Deduction Theorem, with A for H, .A/ ) .B/ for K, and B for S.

Step 1.

As in the Deduction Theorem, assume first that the hypotheses H and K are proved,
and from them derive the conclusion S by proving H; K ` S. Here, assume that the
hypotheses A and .A/ ) .B/ are both proved, and prove A; Œ.A/ ) .B/ ` B:
`A first temporary hypothesis,
` .A/ ) .B/ second temporary hypothesis,
`B Detachment.
 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 19

The foregoing derivation shows that if A and .A/ ) .B/ are proved, then B is
proved. Still under the first hypothesis A, the Deduction Theorem allows for the
removal of the second hypothesis, .A/ ) .B/, as follows.

Step 2.

Step 2.1
The first line in step 1 consists of the other hypothesis, A, which is assumed proved,
whence instructions (D2) in the Deduction Theorem replace A with a complete proof
of Œ.A/ ) .B/ ) .A/ as in theorem 1.12. In other words, replace the first line, ` A,
by the following three lines:
Proof of theorem 1.12:
`A temporary hypothesis,
` .A/ ) fŒ.A/ ) .B/ ) .A/g axiom P1,
„ ƒ‚ …
Q

` Œ.A/ ) .B/ ) .A/ Detachment.

End of proof of theorem 1.12.

Step 2.2
Similarly, the second line in step 1 consists of the hypothesis K being currently
removed, here .A/ ) .B/, which instructions (D1) in the Deduction Theorem
replace with a complete proof of .K/ ) .K/, here Œ.A/ ) .B/ ) Œ.A/ ) .B/,
as in theorem 1.14. Thus, replace the second line, ` .A/ ) .B/, by the following
lines:
Proof of theorem 1.14
  with ŒK for Œ.A/).B/:
` ŒK) fŒK)ŒKg)ŒK axiom P1,
˚  
` ŒK)
 fŒK)ŒKg)ŒK
  axiom P2 : : :
) ŒK)fŒK)ŒKg )fŒK)ŒKg : : : continued,
 
` ŒK)fŒK)ŒKg )fŒK)ŒKg Detachment,
` ŒK)fŒK)ŒKg axiom P1,
` ŒK)ŒK Detachment,
` Œ.A/).B/)Œ.A/).B/ substitution.
End of proof of theorem 1.14.
 www.pdfgrip.com

20 1 Propositional Logic: Proofs from Axioms and Inference Rules

Step 2.3
Finally, the third line in step 1 invokes Detachment, which instructions (D3) replace
by an instance of (the proof of) theorem 1.15:
` Œ.A/ ) .B/ ) .A/ step 2.1,
` Œ.A/ ) .B/ ) Œ.A/ ) .B/ step 2.2,
` Œ.A/ ) .B/ ) .B/ theorem 1.15.
Hence the proof no longer assumes .A/ ) .B/ as a hypothesis, but it still
assumes A as a hypothesis, thus proving that

A ` fŒ.A/ ) .B/ ) .B/g:

Step 3.

Finally, the Deduction Theorem allows for the removal of the first hypothesis, A,
from step 2. Here step 2 consists of steps 2.1, 2.2, and 2.3.

Step 3.1
In step 2.1 the first line consists of this hypothesis, A, whence instructions (D1)
replace A with .A/ ) .A/ by a complete proof of theorem 1.14. In other words,
replace the first line in step 2, ` A, by the following lines:
  Proof of theorem 1.14:
` ŒA
˚ ) fŒA
 ) ŒAg ) ŒA  axiom P1,
` ŒA) fŒA ) ŒAg )ŒA  axiom P1 : : :
) ŒA ) fŒA ) ŒAg ) fŒA ) ŒAg : : : continued,
 
` ŒA ) fŒA ) ŒAg ) fŒA ) ŒAg Detachment,
` ŒA ) fŒA ) ŒAg axiom P1,
` ŒA ) ŒA Detachment.
End of proof of theorem 1.14.

Step 3.2
The second line in step 2.1 is an instance of axiom
 P1, which instructions (D2)
replace by .A/ ) .A/ ) fŒ.A/ ) .B/ ) .A/g .

Step 3.3
The third line in step 2.1 yields Œ.A/ ) .B/ ) .A/, from Detachment, which
instructions (D3) replace by a complete proof of .A/ ) fŒ.A/ ) .B/ ) .A/g,
as in theorem 1.15. In this case, however, such a proof would be correct but not
 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 21

necessary, because .A/ ) fŒ.A/ ) .B/ ) .A/g is merely an instance of axiom

P1. Because it is an axiom, all the preceding lines also become superfluous.

Step 3.4
The result of step 2.2, Œ.A/ ) .B/ ) Œ.A/ ) .B/, is proved by theorem 1.14.
Hence, instructions (D2) replace it by .A/ ) Œ.A/ ) .B/ ) Œ.A/ ) .B/.

Step 3.5
Fully written out, the remaining lines in step 2.3 would follow the proof of
theorem 1.15. Removing the hypothesis A then amounts to theorem 1.20, which
forms the last line of the final proof:
` .A/)fŒ.A/).B/).A/g axiom P1 (from 3.3, replacing 2.1),
` Œ.A/).B/)Œ.A/).B/ theorem 1.14 (from step 2.2),
` .A/)fŒ.A/).B/)Œ.A/).B/g theorem 1.12 (from 3.4, replacing 2.2),
` .A/)fŒ.A/).B/).B/g Theorem 1.20 (from 3.5, replacing 2.3).
Thus the Deduction Theorem has provided some guidance for the construction
of a proof of the theorem .A/ ) fŒ.A/ ) .B/ ) .B/g. t
u

1.3.5 More Examples to Design Proofs of Implicational

Theorems

In patterns of deductive reasoning involving two premisses, the following theorems

confirm that the order of the premisses does not matter.
1.27 Theorem (transitive law of implication). The following formula is a theo-
rem:

Œ.Q/ ) .R/ ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g:

Proof. The formula to be proved has the form .A/ ) Œ.B/ ) .C/, with A denoting
.Q/ ) .R/, B denoting .P/ ) .Q/, and C denoting .P/ ) .R/:

A B C
‚ …„ ƒ ‚ …„ ƒ ‚ …„ ƒ
Œ.Q/ ) .R/ ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g:

Phase 1: deriving A; B ` C and discharging B.

Designing a proof of .A/ ) Œ.B/ ) .C/ can start with a derivation A; B ` C, in

this case .Q/ ) .R/; .P/ ) .Q/ ` .P/ ) .R/, which is exactly theorem 1.16.
 www.pdfgrip.com

22 1 Propositional Logic: Proofs from Axioms and Inference Rules

Expanding the proof of theorem 1.16 which invokes theorems 1.12 and 1.15,
helps discharging the hypothesis .P/ ) .Q/:
` .Q/ ) .R/ hypothesis A,
` .P/ ) Œ.Q/ ) .R/ theorem 1.12,
` .P/ ) .R/ conclusion C, by theorem 1.15.
The preceding three steps form a derivation A; B ` C of C from A and B,
but without invoking B, which plays a hidden rôle in the proof of theorem 1.15.
Replacing the citation of theorem 1.15 by its proof leads to a derivation of .B/ )
.C/ from A:
` .Q/ ) .R/ hypothesis A,
` .P/ ) Œ.Q/ ) .R/ theorem 1.12,
` Œ.P/ ) .Q/ ) Œ.Q/ ) .R/ .B/ ) .C/ by theorem 1.13.
Replacing the citation of theorem 1.13 by its proof, which uses only axioms and
Detachment, gives a derivation of .B/ ) .C/ from A directly from the axioms:
1 ` .Q/).R/ hypothesis A,
2 ` .P/)Œ.Q/).R/ theorem 1.12,
3 ` f.P/)Œ.Q/).R/g)fŒ.P/).Q/)Œ.P/).R/g axiom P2,
4 `Œ.P/).Q/ ) Œ.P/).R/ 2, 3, Detachment.
Lines 1–4 complete the derivation of Œ.P/ ) .Q/ ) Œ.Q/ ) .R/ from the first
hypothesis .Q/ ) .R/ with axiom P2 and Detachment. Since the second hypothesis,
.P/ ) .Q/, has not been used, it need not be discharged.

Phase 2: discharging A.

An application of the Deduction Theorem (1.22) discharges the first hypothesis and
yields a proof of .A/ ) Œ.B/ ) .C/. The resulting proof can be shortened, or
alternative proofs may result from trial and error. To this end, H, K, L refer to
theorem 1.15:
` Œ.Q/ ) .R/ ) f.P/ ) Œ.Q/ ) .R/g axiom P1,
„ ƒ‚ … „ ƒ‚ …
H K

` f.P/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g axiom P2,

„ ƒ‚ … „ ƒ‚ …
K L

` Œ.Q/ ) .R/ ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g theorem 1.15.

„ ƒ‚ … „ ƒ‚ …
H L
t
u
Swapping the premisses .P/ ) .Q/ and .Q/ ) .R/ also yields .P/ ) .R/:
1.28 Theorem (transitive law of implication, law of the hypothetical syllogism).
The following formula is a theorem:

Œ.P/ ) .Q/ ) fŒ.Q/ ) .R/ ) Œ.P/ ) .R/g:

 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 23

Proof. With notation as in the proof of theorem 1.27, the formula to be proved has
the form .B/ ) Œ.A/ ) .C/, with A denoting .Q/ ) .R/, B denoting .P/ ) .Q/,
and C denoting .P/ ) .R/:

B A C
‚ …„ ƒ ‚ …„ ƒ ‚ …„ ƒ
Œ.P/ ) .Q/ ) fŒ.Q/ ) .R/ ) Œ.P/ ) .R/g:

Thus designing a proof of .B/ ) Œ.A/ ) .C/ can start with a derivation B; A ` C,
in this case .P/ ) .Q/; .Q/ ) .R/ ` .P/ ) .R/, which is exactly theorem 1.16.
Hence steps as in the proof of theorem 1.27 discharge the hypotheses, but in the
reverse order. The resulting proof might then be shortened or give clues for a shorter
alternative proof. For example, apply axiom P1 with theorems 1.27, 1.13, and 1.16:
U V
‚ …„ ƒ ‚ …„ ƒ
` Œ.P/ ) .Q/ ) fŒ.Q/ ) .R/ ) Œ.P/ ) .Q/g axiom P1,
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
K H K

` f.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g 1.27,

„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
H K L

H K H L
‚ …„ ƒ ‚ …„ ƒ ‚ …„ ƒ ‚ …„ ƒ
` fŒ.Q/ ) .R/ ) Œ.P/ ) .Q/g ) fŒ.Q/ ) .R/ ) Œ.P/ ) .R/g 1.13,
„ ƒ‚ … „ ƒ‚ …
V W

U W
‚ …„ ƒ ‚ …„ ƒ
` Œ.P/ ) .Q/ ) fŒ.Q/ ) .R/ ) Œ.P/ ) .R/g 1.16.
t
u

1.3.6 Another Guide for Proofs: Substitutivity of Equivalences

Besides the Deduction Theorem (theorem 1.22), another guide to design proofs
consists of replacing any occurrence of a formula by an equivalent formula, thanks
to theorem 1.29 [18, p. 101, 124, 189], [108, p. 48].
1.29 Theorem (Substitutivity of Equivalence in the Pure Positive Implicational
Propositional Calculus, preliminary version). For all well-formed implicational
logical formulae U and V, if ` .U/ ) .V/ and ` .V/ ) .U/, and if a formula Q
results from substituting any (zero, one, several, or all) occurrence(s) of U by V in
a well-formed formula P, then ` .P/ ) .Q/ and ` .Q/ ) .P/.
Proof (Outline). This proof proceeds by cases and subcases.
In all cases, if Q is P, which results by substituting none of the occurrences of U
by V, then each of .P/ ) .Q/ and .Q/ ) .P/ is .P/ ) .P/, which is theorem 1.14.
 www.pdfgrip.com

24 1 Propositional Logic: Proofs from Axioms and Inference Rules

1. If P is U, then Q is either U or V. If Q is V, then .P/ ) .Q/ and .Q/ ) .P/

become .U/ ) .V/ and .V/ ) .U/, which are the hypotheses.
2. If P is .U/ ) .W/, then Q is either .U/ ) .W/, which is P, or Q is .V/ ) .W/.
If Q is .V/ ) .W/, then the hypothesis ` .V/ ) .U/ yields ` .P/ ) .Q/:
` .V/ ) .U/ hypothesis,
` Œ.V/ ) .U/ ) fŒ.U/ ) .W/ ) Œ.V/ ) .W/g theorem 1.28,
` Œ.U/ ) .W/ ) Œ.V/ ) .W/ Detachment.
„ ƒ‚ … „ ƒ‚ …
P Q
Similarly, the hypothesis ` .U/ ) .V/ yields ` .Q/ ) .P/:
` .U/ ) .V/ hypothesis,
` Œ.U/ ) .V/ ) fŒ.V/ ) .W/ ) Œ.U/ ) .W/g theorem 1.28,
` Œ.V/ ) .W/ ) Œ.U/ ) .W/ Detachment.
„ ƒ‚ … „ ƒ‚ …
Q P
3. If P is .W/ ) .U/, then Q is either .W/ ) .U/ or .W/ ) .V/. If Q is .W/ )
.V/, then P with .U/ ) .V/ yield Q, and Q with .V/ ) .U/ yield P, by
transitivity (theorem 1.16).
If P is .W/ ) .U/, then Q is either .W/ ) .U/, which is P, or Q is .W/ )
.V/. If Q is .W/ ) .V/, then the hypothesis ` .V/ ) .U/ yields ` .Q/ ) .P/:
` .V/ ) .U/ hypothesis,
` Œ.V/ ) .U/ ) fŒ.W/ ) .V/ ) Œ.W/ ) .U/g theorem 1.27,
` Œ.W/ ) .V/ ) Œ.W/ ) .U/ Detachment.
„ ƒ‚ … „ ƒ‚ …
Q P
Similarly, the hypothesis ` .U/ ) .V/ yields ` .P/ ) .Q/:
` .U/ ) .V/ hypothesis,
` Œ.U/ ) .V/ ) fŒ.W/ ) .U/ ) Œ.W/ ) .V/g theorem 1.27,
` Œ.W/ ) .U/ ) Œ.W/ ) .V/ Detachment.
„ ƒ‚ … „ ƒ‚ …
P Q

The general case follows by several applications of the previous cases, in a

way that may be specified more explicitly after the availability of the Principle of
Mathematical Induction in chapter 4. t
u
1.30 Example. If U denotes .H/ ) .K/, and V denotes .H/ ) Œ.H/ ) .K/, then
` .U/ ) .V/ and ` .V/ ) .U/:
U V
‚ …„ ƒ ‚ …„ ƒ
` Œ.H/ ) .K/ ) f.H/ ) Œ.H/ ) .K/g axiom P1,
` f.H/ ) Œ.H/ ) .K/g ) Œ.H/ ) .K/ theorem 1.24.
„ ƒ‚ … „ ƒ‚ …
V U
Also, if P denotes .L/ ) Œ.H/ ) .K/, which is .L/ ) .U/, then ` .P/ ) .Q/
and ` .Q/ ) .P/ become
P Q
‚ …„ ƒ ‚ …„ ƒ
` f.L/ ) Œ.H/ ) .K/g ) .L/ ) f.H/ ) Œ.H/ ) .K/g theorem 1.29,
` .L/ ) f.H/ ) Œ.H/ ) .K/g ) f.L/ ) Œ.H/ ) .K/g theorem 1.29,
„ ƒ‚ … „ ƒ‚ …
Q P
by axiom P1 with theorems 1.24 and 1.29.
 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 25

1.3.7 More Derived Rules of Inference

The following derived rules allow for substitutions within implications subject to
hypotheses, for instance, a substitution within an intermediate hypothesis.
1.31 Theorem (derived rule). For all well-formed formulae H, K, L, M, if
.H/ ) Œ.L/ ) .M/ and
.K/ ) .L/
are theorems, then
.H/ ) Œ.K/ ) .M/
is also a theorem.
Proof. Apply theorems 1.28 and 1.16:
` .K/ ) .L/ hypothesis,
` Œ.K/ ) .L/ ) fŒ.L/ ) .M/ ) Œ.K/ ) .M/g theorem 1.28,
` Œ.L/ ) .M/ ) Œ.K/ ) .M/ Detachment. t
u
` .H/ ) Œ.L/ ) .M/ hypothesis,
` .H/ ) Œ.K/ ) .M/ theorem 1.16.
The second theorem allows for a substitution in the conclusion.
1.32 Theorem (derived rule). For all well-formed formulae H, L, M, N, if
.H/ ) Œ.L/ ) .M/ and
.M/ ) .N/
are theorems, then
.H/ ) Œ.L/ ) .N/
is also a theorem.
Proof. Apply theorems 1.28, 1.16, and 1.17:
` .H/ ) Œ.L/ ) .M/ hypothesis,
„ ƒ‚ …
P
` Œ.L/ ) .M/ ) fŒ.M/ ) .N/ ) Œ.L/ ) .N/g theorem 1.28,
„ ƒ‚ … „ ƒ‚ …
P Q
` .H/ ) fŒ.M/ ) .N/ ) Œ.L/ ) .N/g theorem 1.16,
„ ƒ‚ …
Q
` .M/ ) .N/ hypothesis,
` .H/ ) Œ.L/ ) .N/ theorem 1.17.
t
u
The following three derived rules of inference will simplify subsequent proofs.
1.33 Theorem (derived rule). If .K/ ) .L/ is a theorem, then the following
formula is also a theorem: Œ.H/ ) .K/ ) Œ.H/ ) .L/.
 www.pdfgrip.com

26 1 Propositional Logic: Proofs from Axioms and Inference Rules

Proof. Apply the transitivity of implication in the form of theorem 1.28:

` .K/ ) .L/ hypothesis,
` Œ.K/ ) .L/ ) fŒ.H/ ) .K/ ) Œ.H/ ) .L/g theorem 1.28,
` Œ.H/ ) .K/ ) Œ.H/ ) .L/ Detachment.
t
u
1.34 Theorem (derived rule). If .I/ ) .H/ is a theorem, then the following
formula is also a theorem: Œ.H/ ) .K/ ) Œ.I/ ) .K/.
Proof. Apply the transitivity of implication in the form of theorem 1.27:
` .I/ ) .H/ hypothesis,
` Œ.I/ ) .H/ ) fŒ.H/ ) .K/ ) Œ.I/ ) .K/g theorem 1.27,
` Œ.H/ ) .K/ ) Œ.I/ ) .K/ Detachment.
t
u
1.35 Theorem (derived rule). If .A/ ) .B/ and .C/ ) .D/ are theorems, then
the following formula is also a theorem: Œ.B/ ) .C/ ) Œ.A/ ) .D/ .
Proof. Apply theorems 1.34 and 1.33, with H, K, L as in theorem 1.16:
` .A/ ) .B/ hypothesis,

` Œ.B/ ) .C/ ) Œ.A/ ) .C/ theorem 1.34,

„ ƒ‚ … „ ƒ‚ …
H K

` .C/ ) .D/ hypothesis,

` Œ.A/ ) .C/ ) Œ.A/ ) .D/ theorem 1.33,

„ ƒ‚ … „ ƒ‚ …
K L
` Œ.B/ ) .C/ ) Œ.A/ ) .D/ theorem 1.16.
„ ƒ‚ … „ ƒ‚ …
H L
t
u
In contrast to the preceding derived rules, theorem 1.36 reveals a different pattern.
1.36 Theorem (derived rule). If Œ.H/ ) .L/ ) .M/ and .H/ ) .K/ are
theorems, then the following formula is also a theorem: Œ.K/ ) .L/ ) .M/ .
Proof. Apply theorems 1.27, 1.13, 1.12, and 1.16:
` Œ.K/).L/)fŒ.H/).K/)Œ.H/).L/g 1.27,
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
P Q R
` fŒ.K/).L/)Œ.H/).K/g)fŒ.K/).L/)Œ.H/).L/g 1.13,
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
P Q P R
` Œ.H/).K/ hypothesis,
„ ƒ‚ …
Q
` Œ.K/).L/)Œ.H/).K/ 1.12,
„ ƒ‚ … „ ƒ‚ …
P Q
 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 27

` Œ.K/).L/)Œ.H/).L/ Detachment,
„ ƒ‚ … „ ƒ‚ …
P R
` Œ.H/).L/).M/ hypothesis,
„ ƒ‚ …
R
` Œ.K/).L/).M/ 1.16.
„ ƒ‚ …
P
t
u

1.3.8 The Laws of Commutation and of Assertion

The following “Law of Commutation” allows for yet another change in the order of
hypotheses:

f.P/ ) Œ.Q/ ) .R/g ) f.Q/ ) Œ.P/ ) .R/g:

The Law of Commutation is one of Frege’s axioms [39, p. 146, eq. (8)]. In Frege’s
1879 words, the Law of Commutation states that
If a proposition is the consequence of two propositions, their order is immaterial [39,
p. 147].

In 1935, however, Łukasiewicz showed that the Law of Commutation is a theorem

derivable from Detachment with axioms P1 and P2 [80, p. 127], as proved by
theorem 1.37. (The time lapse between 1879 and 1935 indicates that recognizing
whether a formula is a theorem can also be difficult for specialists.)
1.37 Theorem (law of commutation). The following formula is a theorem:

f.P/ ) Œ.Q/ ) .R/g ) f.Q/ ) Œ.P/ ) .R/g:

Proof. Apply theorem 1.31:

` f.P/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g axiom P2,
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
H L M

` . Q / ) Œ.P/ ) .Q/ axiom P1,

„ƒ‚… „ ƒ‚ …
K L

` f.P/ ) Œ.Q/ ) .R/g ) f. Q / ) Œ.P/ ) .R/g theorem 1.31.

„ ƒ‚ … „ƒ‚… „ ƒ‚ …
H K M
t
u
With a shorter proof relying on the law of commutation, the following theorem
combines the rule of Detachment into a single formula.
 www.pdfgrip.com

28 1 Propositional Logic: Proofs from Axioms and Inference Rules

1.38 Theorem (law of assertion). The formula .H/ ) fŒ.H/ ) .C/ ) .C/g is
a theorem.
Proof. Apply theorems 1.14 and 1.37 with Detachment:
P P
‚ …„ ƒ ‚ …„ ƒ
` Œ.H/ ) .C/ ) Œ.„ƒ‚…
H / ) .„ƒ‚…
C / theorem 1.14,
„ ƒ‚ …
P Q R
` .„ƒ‚…
H / ) fŒ.H/ ) .C/ ) .„ƒ‚…
C /g commutation (1.37) and Detachment.
„ ƒ‚ …
Q P R
t
u
Also relying on the law of commutation, the following derived rule will shorten
the proof of subsequent results in particular, theorem 1.50.
1.39 Theorem (derived rule). If ` .U/ ) Œ.V/ ) .W/
and ` .H/ ) Œ.W/ ) .R/,
then ` .H/ ) f.U/ ) Œ.V/ ) .R/g.
Proof. Apply the transitivity of implication (theorem 1.27):
` Œ.V/).W/)fŒ.W/).R/)Œ.V/).R/g
  theorem 1.27,
` .U/) Œ.V/).W/)fŒ.W/).R/)Œ.V/).R/g theorem 1.12,
` .U/)Œ.V/).W/ hypothesis,
` .U/)fŒ.W/).R/)Œ.V/).R/g theorem 1.15,
` f.U/)Œ.W/).R/g)f.U/)Œ.V/).R/g theorem 1.13,
` .H/)Œ.W/).R/ hypothesis,
` .U/)f.H/)Œ.W/).R/g theorem 1.12,
` .H/)f.U/)Œ.W/).R/g commutation (1.37),
` .H/)f.U/)Œ.V/).R/g transitivity.
t
u

1.3.9 Exercises on the Classical Implicational Calculus

The foregoing theorems involve only implications but no negation, and their
proofs do not involve any negation either. Nevertheless, there exist other theorems
involving only ) but not : for which there does not exist any proof involving only
implications. Examples of such theorems are hidden in the exercises, to be revealed
later. Investigate whether the formulae in the following exercises are theorems, using
any of the axioms P1 and P2, any rules of inference, and any of the theorems just
proved.
1.1 . Œ.H/ ) .L/ ) f.H/ ) Œ.K/ ) .L/g
1.2 . Œ.K/ ) .L/ ) f.H/ ) Œ.K/ ) .L/g
 
1.3 . Œ.A/ ) .B/ ) fŒ.A/ ) .B/ ) .B/g ) .B/
 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 29

 
1.4 . Œ.A/ ) .B/ ) fŒ.A/ ) .B/ ) .A/g ) .A/
1.5 . fŒ.P/ ) .P/ ) .P/g ) .P/
1.6 . .P/ ) fŒ.P/ ) .P/ ) .P/g
1.7 . fŒ.P/ ) .Q/ ) .P/g ) .P/ (Peirce’s law.)
 
1.8 . Œ.P/ ) .R/ ) fŒ.P/ ) .Q/ ) .R/g ) .R/
1.9 . fŒ.R/ ) .Q/ ) .P/g ) fŒ.R/ ) .Q/ ) Œ.S/ ) .P/g
 
1.10 . Œ.R/ ) .Q/ ) fŒ.R/ ) .Q/ ) .P/g ) Œ.S/ ) .P/

1.3.10 Equivalent Implicational Axiom Systems

The Classical Implicational Calculus just presented rests on the rules of Detachment
and Substitution with Frege’s axioms P1 and P2:
Frege’s axiom P1 .P/ ) Œ.Q/ ) .P/.
Frege’s axiom P2 f.P/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g.
Other selections of axioms exist, for instance, Stephen Cole Kleene’s [72, p. 15],
Kleene’s axiom 1a .A/ ) Œ.B/ ) .A/,
Kleene’s axiom 1b Œ.A/ ) .B/ ) .f.A/ ) Œ.B/ ) .C/g ) Œ.A/ ) .C//,
and Tarski’s [129, p. 147],
Tarski’s axiom I .P/ ) Œ.Q/ ) .P/,
Tarski’s axiom II f.P/ ) Œ.P/ ) .Q/g ) Œ.P/ ) .Q/,
Tarski’s axiom III Œ.P/ ) .Q/ ) fŒ.Q/ ) .R/ ) Œ.P/ ) .R/g.
Frege’s, Kleene’s, and Tarski’s implicational axiom systems are mutually equiva-
lent, in the sense that each system leads to the same Pure Positive Implicational
Propositional Calculus.
Indeed, their first axioms, P1, 1a, and I are mutually identical.
Second, Kleene’s axiom 1b results from applying the law of commutation
(theorem 1.37) to Frege’s axiom P2. Consequently, both of Kleene’s axioms 1a
and 1b are theorems derivable from Frege’s axioms P1 and P2. Thus, prepending
derivations of Kleene’s axioms from Frege’s axioms to any proof of any theorem
from Kleene’s axioms yields a proof of the same theorem from Frege’s axioms.
Therefore, every theorem derivable from Kleene’s axioms is also derivable from
Frege’s axioms.
Similarly, Tarski’s axiom II is theorem 1.24, while Tarski’s axiom III is the-
orem 1.28. Consequently, all three of Tarski’s axioms I, II, and III are theorems
derivable from Frege’s axioms P1 and P2. Therefore, every theorem derivable from
Tarski’s axioms is also derivable from Frege’s axioms.
 www.pdfgrip.com

30 1 Propositional Logic: Proofs from Axioms and Inference Rules

The exercises establish the converse derivations. From Kleene’s axioms,

exercises 1.11 and 1.13 derive Tarski’s axiom II, whereas exercises 1.14–1.22 derive
Tarski’s axiom III. Thus exercises 1.11–1.22 show that every theorem derivable
from Tarski’s axioms is also derivable from Kleene’s axioms. Hence from Tarski’s
axioms, exercises 1.23–1.32 derive the law of commutation following Tarski’s
outline [129, p. 148–149], and thence Frege’s axioms P2 from Tarski’s axiom III.
Thus exercises 1.23–1.36 show that every theorem derivable from Frege’s axioms
is derivable from Tarski’s axioms, and thus also from Kleene’s axioms.
However, after the proof of .P/ ) .P/ in theorem 1.14, Frege’s axioms lead
immediately to the Deduction Theorem (1.22), whereas the exercises reveal that
several intermediate theorems stand between Kleene’s or Tarski’s axioms and the
Deduction Theorem, which may explain the popularity of Frege’s axioms, already
announced before definition 1.4 on page 6.

1.3.11 Exercises on Kleene’s Axioms

For the following exercises, use only Kleene’s axioms 1a and 1b with Substitution
and Detachment.
1.11 . Prove .P/ ) .P/.
1.12 . Establish the derived rule of inference that if T is a theorem, then .S/ ) .T/
is also a theorem.
1.13 . Prove fŒ.P/ ) Œ.P/ ) .Q/g ) Œ.P/ ) .Q/.
1.14 . Establish the derived rule of inference that if .H/ ) .K/ and .K/ ) .L/ are
theorems, then .H/ ) .L/ is also a theorem.
1.15 . Establish the derived rule of inference that if T is a theorem, then f.A/ )
Œ.T/ ) .C/g ) Œ.A/ ) .C/ is also a theorem.
1.16 . Prove Œ.B/ ) .C/ ) f.A/ ) Œ.B/ ) .C/g.
1.17 . Prove
n  o
Œ.B/ ) .C/ ) f.A/ ) Œ.B/ ) .C/g ) Œ.A/ ) .C/

) fŒ.B/ ) .C/ ) Œ.A/ ) .C/g:

1.18 . Prove

fŒ.B/).C/)Œ.A/).B/g
h n  o
) Œ.B/).C/) Œ.A/).B/) f.A/)Œ.B/).C/g)Œ.A/).C/
n  oi
) Œ.B/).C/) f.A/)Œ.B/).C/g)Œ.A/).C/ :
 www.pdfgrip.com

1.3 The Pure Positive Implicational Propositional Calculus 31

1.19 . Prove

fŒ.B/ ) .C/ ) Œ.A/ ) .B/g

n  o
) Œ.B/ ) .C/ ) f.A/ ) Œ.B/ ) .C/g ) Œ.A/ ) .C/

1.20 . Prove

fŒ.B/ ) .C/ ) Œ.A/ ) .B/g ) fŒ.B/ ) .C/ ) Œ.A/ ) .C/g :

1.21 . Prove Œ.A/ ) .B/ ) fŒ.B/ ) .C/ ) Œ.A/ ) .B/g.

1.22 . Prove Œ.A/ ) .B/ ) fŒ.B/ ) .C/ ) Œ.A/ ) .C/g.

1.3.12 Exercises on Tarski’s Axioms

For the following exercises, use only Tarski’s axioms I, II, and III, with Substitution
and Detachment.
1.23 . Establish the derived rule of inference that if .H/ ) .K/ and .K/ ) .L/ are
theorems, then .H/ ) .L/ is also a theorem.
1.24 . Establish the derived rule of inference that if T is a theorem, then .S/ ) .T/
is also a theorem.
1.25 . Prove .P/ ) fŒ.P/ ) .Q/ ) .P/g.
1.26 . Prove
 
fŒ.P/ ) .Q/ ) .P/g ) Œ.P/ ) .Q/ ) fŒ.P/ ) .Q/ ) .Q/g :

1.27 . Prove
 
.P/ ) Œ.P/ ) .Q/ ) fŒ.P/ ) .Q/ ) .Q/g :

1.28 . Prove
 
Œ.P/ ) .Q/ ) fŒ.P/ ) .Q/ ) .Q/g ) fŒ.P/ ) .Q/ ) .Q/g:

1.29 . Prove the law of assertion:

.P/ ) fŒ.P/ ) .Q/ ) .Q/g:

 www.pdfgrip.com

32 1 Propositional Logic: Proofs from Axioms and Inference Rules

1.30 . Prove
 
f.P/ ) Œ.Q/ ) .R/g ) fŒ.Q/ ) .R/ ) .R/g ) Œ.P/ ) .R/ :

1.31 . Prove
 
fŒ.Q/ ) .R/ ) .R/g ) Œ.P/ ) .R/ ) f.Q/ ) Œ.P/ ) .R/g:

1.32 . Prove the law of commutation:

f.P/ ) Œ.Q/ ) .R/g ) f.Q/ ) Œ.P/ ) .R/g:

1.33 . Prove Frege’s axiom 2:

Œ.Q/ ) .R/ ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g:

1.34 . Prove
 
f.Q/ ) Œ.P/ ) .R/g ) Œ.P/ ) .Q/ ) f.P/ ) Œ.P/ ) .R/g :

1.35 . Prove
 
Œ.P/ ) .Q/ ) f.P/ ) Œ.Q/ ) .R/g ) Œ.P/ ) .R/g :

1.36 . Prove

f.P/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g:

1.4 Proofs by the Converse Law of Contraposition

The Pure Positive Implicational Propositional Calculus belongs to several logical

systems, which differ from one another by their different axioms about negation.
For instance, classical logic defines its concept of negation by the converse law of
contraposition:

Axiom P3: fŒ:.Q/ ) Œ:.P/g ) Œ.P/ ) .Q/:

1.4.1 Examples of Proofs in the Full Propositional Calculus

The following proofs demonstrate the use of the converse law of contraposition.
 www.pdfgrip.com

1.4 Proofs by the Converse Law of Contraposition 33

1.40 Theorem (law of denial of the antecedent). For all well-formed formulae P
and Q, the following formula is a theorem: Œ:.P/ ) Œ.P/ ) .Q/.
Proof. Apply axioms P1 and P3 with the transitivity of implication (theorem 1.16):
` Œ:.P/ ) fŒ:.Q/ ) Œ:.P/g substitution in axiom P1,
` fŒ:.Q/ ) Œ:.P/g ) Œ.P/ ) .Q/ axiom P3,
` Œ:.P/ ) Œ.P/ ) .Q/ theorem 1.16.
t
u
1.41 Theorem. The formula .P/ ) fŒ:.P/ ) .Q/g is a theorem (schema).
Proof. Apply theorem 1.40 and the law of commutation (theorem 1.37):
` Œ:.P/ ) Œ.P/ ) .Q/ theorem 1.40,
 
` fŒ:.P/ ) Œ.P/ ) .Q/g ) .P/ ) fŒ:.P/ ) .Q/g commutation (1.37),
` .P/ ) fŒ:.P/ ) .Q/g Detachment.
t
u
The following two theorems establish the complete law of double negation.
1.42 Theorem (law of double negation). The formula Œ::.P/ ) .P/ is a
theorem (schema).
Proof. Apply the transitivity of implication (theorem 1.19) and theorem 1.25:
` f:Œ:.P/g
 ) fŒ::::.P/ )  Œ::.P/g axiom P1,
` f:Œ:::.P/g ) f:Œ:.P/g  ) fŒ:.P/ ) Œ:::.P/g axiom P3,
` Œ:.P/ ) f:Œ::.P/g ) fŒ::.P/ ) .P/g axiom P3,
` Œ::.P/ ) fŒ::.P/ ) .P/g theorem 1.19,
` Œ::.P/ ) .P/ theorem 1.25.
t
u
1.43 Theorem (converse law of double negation). The formula .P/ ) Œ::.P/
is a theorem (schema).
Proof. Apply the law
 of double negation (theorem 1.42) and contraposition (P3):
` ˚:f:Œ:.P/g ) Œ:.P/    theorem 1.42,
` :f:Œ:.P/g ) Œ:.P/ ) .P/ ) f:Œ:.P/g axiom P3,
` .P/ ) f:Œ:.P/g Detachment.
t
u
With axiom P3, the following theorem gives the complete law of contraposition.
1.44 Theorem (law of contraposition, principle of transposition). The follow-
ing formula is a theorem (schema): Œ.P/ ) .Q/ ) fŒ:.Q/ ) Œ:.P/g .
Proof. Apply transitivity (theorem 1.16) with A, B, C, D as in theorem 1.35:
` f:Œ:.P/g ) .„ƒ‚… P / theorem 1.42,
„ ƒ‚ …
A B
` . Q / ) f:Œ:.Q/g theorem 1.43,
„ƒ‚… „ ƒ‚ …
C D
 www.pdfgrip.com

34 1 Propositional Logic: Proofs from Axioms and Inference Rules

` Œ.„ƒ‚…
P / ) . Q / ) fŒ::.P/ ) Œ::.Q/g theorem 1.35,
„ƒ‚… „ ƒ‚ … „ ƒ‚ …
B C A D
` fŒ::.P/ ) Œ::.Q/g ) fŒ:.Q/ ) Œ:.P/g axiom P3,
` Œ.P/ ) .Q/ ) fŒ:.Q/ ) Œ:.P/g theorem 1.16.
t
u
Theorem 1.44 is the theoretical basis for reasoning by contraposition. Theo-
rem 1.45 syncopates several steps for later use.
1.45 Theorem. For all well-formed formulae P and Q, the following formula is a
theorem: fŒ:.P/ ) .Q/g ) fŒ:.Q/ ) .P/g.
Proof. Apply the principle of transposition and the law of double negation:
` fŒ:.P/ ) .Q/g ) Œ:.Q/ ) f:Œ:.P/g transposition (1.44) ,
` f:Œ:.P/g ) .P/ double negation (1.42) ,
` fŒ:.P/ ) .Q/g ) fŒ:.Q/ ) .P/g derived rule (1.32).
t
u

1.4.2 Guides for Proofs in the Propositional Calculus

Besides the Deduction Theorem (theorem 1.22 ), an extension of the substitutivity

of equivalence in the implicational calculus (theorem 1.29) also allows for the
replacement of any occurrence of a formula by an equivalent formula containing
negations, thanks to theorem 1.46 [18, p. 101, 124, 189], [108, p. 48].
1.46 Theorem (Substitutivity of Equivalence in the Pure Propositional Calcu-
lus, preliminary version). For all well-formed propositional formulae U and V, if
` .U/ ) .V/ and ` .V/ ) .U/, and if a formula Q results from substituting any
(zero, one, several, or all) occurrence(s) of U by V in a well-formed formula P, then
` .P/ ) .Q/ and ` .Q/ ) .P/.
Proof (Outline). Theorem 1.29 has already established the conclusions for
implications.
For negations, if P is :.U/, then Q is either :.U/ or :.V/. If Q is :.V/, then
.P/ ) .Q/ and .Q/ ) .P/ become Œ:.U/ ) Œ:.V/ and Œ:.V/ ) Œ:.U/,
which hold by the law of contraposition (theorem 1.44 ):
` .U/ ) .V/ hypothesis ,
` Œ.U/ ) .V/ ) fŒ:.V/ ) Œ:.U/g theorem 1.44 ,
` Π:.V/  ) Π:.U/  Detachment;
„ƒ‚… „ƒ‚…
Q P

` .V/ ) .U/ hypothesis ,

` Œ.V/ ) .U/ ) fŒ:.U/ ) Œ:.V/g theorem 1.44 ,
` Π:.U/  ) Π:.V/  Detachment;
„ƒ‚… „ƒ‚…
P Q
 www.pdfgrip.com

1.4 Proofs by the Converse Law of Contraposition 35

The general case follows by several applications of the previous case and the
cases in theorem 1.29, in a way that may be specified more explicitly after the
availability of the Principle of Mathematical Induction in chapter 4. t
u

1.4.3 Proofs by Reductio ad Absurdum

Within classical logic, a proposition and its negation together form an “absurdity”
that cannot hold. In particular, if a hypothesis implies a conclusion and its
negation — an absurdity — then the hypothesis may be rejected. The following
theorems establish the validity of such a pattern of reasoning, called reduction to
the absurd.
1.47 Theorem (special law of reductio ad absurdum). For each well-formed
formula P, the following formula is a theorem: f.P/ ) Œ:.P/g ) Œ:.P/ .
Proof. Start with theorem
 1.44 and the denialof the antecedent (theorem 1.40):
` f.P/ ) Œ:.P/g  ) f:Œ:.P/g ) Œ:.P/ (1.44),
` f:Œ:.P/g
 ) Œ:.P/
 ) f:Œ.P/ ) .P/g  (1.40),
` f:Œ:.P/g
 ) Œ:.P/ ) f:Œ.P/ ) .P/g 
 ) f:Œ:.P/g ) Œ:.P/
  ) f:Œ:.P/g ) f:Œ.P/ )  .P/g (P2),
` f:Œ:.P/g ) Œ:.P/ ) f:Œ:.P/g  ) f:Œ.P/ ) .P/g (M.P.),
` f:Œ:.P/g ) f:Œ.P/ ) .P/g ) fŒ.P/ ) .P/ ) Œ:.P/g (P3),
` f.P/ ) Œ:.P/g ) fŒ.P/ ) .P/ ) Œ:.P/g (1.19),
` .P/ ) .P/ (1.14),
` f.P/ ) Œ:.P/g ) Œ:.P/ (1.17).
t
u

1.48 Theorem (law of reductio ad absurdum). For all well-formed formulae P

and Q, the following formula is a theorem:
 
Œ.P/ ) .Q/ ) f.P/ ) Œ:.Q/g ) Œ:.P/ :

Proof. Apply theorems 1.44, 1.12, 1.37, 1.16, 1.47, 1.32:

` Œ.P/).Q/)fŒ:.Q/)Œ:.P/g
  (1.44),
` .P/) Œ.P/).Q/)fŒ:.Q/)Œ:.P/g
  (1.12),
` Œ.P/).Q/) .P/)fŒ:.Q/)Œ:.P/g
   (1.37),
` .P/)fŒ:.Q/)Œ:.P/g
 ) f.P/)Œ:.Q/g)f.P/)Œ:.P/g
 (P2),
` Œ.P/).Q/) f.P/)Œ:.Q/g)f.P/)Œ:.P/g (1.16),
` f.P/)Œ:.P/g)Œ:.P/
  (1.47),
` Œ.P/).Q/) f.P/)Œ:.Q/g)Œ:.P/ (1.32). t
u
Theorem 1.48 is the theoretical basis for the pattern of reasoning by reduction to
the absurd: if .P/ ) .Q/ and .P/ ) Œ:.Q/ are theorems, then theorem 1.48 and
Detachment twice prove that :.P/ is a theorem. As a special case of theorem 1.48,
theorem 1.47 shows that if a statement P implies its negation :.P/, then :.P/ is a
theorem.
 www.pdfgrip.com

36 1 Propositional Logic: Proofs from Axioms and Inference Rules

1.4.4 Proofs by Cases

Theorem 1.49 provides a variation on the theoretical foundation of proofs by cases:

if, in the first case, a conclusion R follows from a hypothesis P, and if, in the
second case, the same conclusion R also follows from the negation :.P/ of the
same hypothesis P, then R is a theorem, derivable from the axioms and inference
rule, so that the hypothesis is superfluous.
1.49 Theorem (proof by cases). For all well-formed formulae P and R,
P1, P2, P3, ` fŒ:.P/ ) .R/g ) fŒ.P/ ) .R/ ) .R/g.
Proof. Apply the laws of double negation (theorem 1.42), contraposition (theo-
rem 1.44), and reduction
 to the absurd (theorem 1.48):

` fŒ:.R/).P/g) fŒ:.R/)Œ:.P/g)f:Œ:.R/g theorem 1.48,
` f:Œ:.R/g).R/ double negation (1.42),
` fŒ:.R/).P/g) ŒfŒ:.R/)Œ:.P/g).R/ derived rule (1.32),
` Œ.P/).R/)fŒ:.R/)Œ:.P/g transposition (1.44),
` fŒ:.R/).P/g) ŒfŒ.P/).R/g).R/ derived rule (1.31),
` fŒ:.P/).R/g)fŒ:.R/).P/g theorem 1.45,
` fŒ:.P/).R/g)fŒ.P/).R/).R/g derived rule (1.16).
t
u
Theorem 1.50 generalizes theorem 1.49 to the situation with an intermediate
hypothesis [18, p. 205, footnote 355].
1.50 Theorem (proof by cases subject to hypotheses). For all well-formed for-
mulae P, Q, and R,   
P1, P2, P3, ` fŒ:.P/ ) .R/g ) Œ.Q/ ) .R/ ) fŒ.P/ ) .Q/ ) .R/g .
Proof. Apply the law of proof by cases (theorem 1.49), the transitivity of implica-
tion (theorem 1.27), and a derived rule (theorem 1.39):
` fŒ:.P/ ) .R/g ) fŒ.P/ ) .R/ ) .R/g theorem 1.49,
„ ƒ‚ … „ ƒ‚ …
H W

` Œ.Q/ ) .R/ ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g theorem 1.27,

„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
U V W

  
` fŒ:.P/ ) .R/g ) Œ.Q/ ) .R/ ) fŒ.P/ ) .Q/ ) .R/g theorem 1.39.
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
H U V
t
u
 www.pdfgrip.com

1.4 Proofs by the Converse Law of Contraposition 37

1.4.5 Exercises on Frege’s and Church’s Axioms

For the following four exercises, use the rules of inference and only the following
six axioms, due to Frege [39]:

Axiom F1 [39, p. 137, eq. (1)]: .P/ ) Œ.Q/ ) .P/.

Axiom F2 [39, p. 137, eq. (2)]: f.P/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ )
Œ.P/ ) .R/g.

Axiom F3 [39, p. 146, eq. (8)]: f.P/ ) Œ.Q/ ) .R/g ) f.Q/ ) Œ.P/ ) .R/g.

Axiom F4 [39, p. 154, eq. (28)]: Œ.P/ ) .Q/ ) fŒ:.Q/ ) Œ:.P/g.

Axiom F5 [39, p. 156, eq. (31)]: f:Œ:.P/g ) .P/.

Axiom F6 [39, p. 158, eq. (41)]: .P/ ) f:Œ:.P/g.

1.37 . Prove that F1–F6 ` .P/ ) fŒ:.P/ ) Œ:.Q/g.

1.38 . Prove that F1–F6 ` Œ:.P/ ) f.P/ ) Œ:.P/g.
1.39 . Prove that F1–F6 ` fŒ:.Q/ ) Œ:.P/g ) Œ.P/ ) .Q/.
1.40 . Prove that Frege’s six axioms F1–F6 are logically equivalent to
Łukasiewicz’s three axioms P1, P2, P3.
The following exercises outline a proof that the axioms P1, P2, P3 of classical
logic are logically equivalent to the following three axioms C1, C2, C3, used by
Church [18, 10, p. 72] and Robbin [108, p. 14]. Because the two logical systems
have the same first two axioms, they also have the same implicational calculus. The
two logical systems differ from each other only by their third axiom, where F stands
for False, so that :.F/ is a theorem.

Axiom C1 .P/ ) Œ.Q/ ) .P/.

Axiom C2 f.P/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g.

Axiom C3 fŒ.P/ ) .F/ ) .F/g ) .P/.

1.41 . This exercise establishes the converse law of contraposition in Church’s

system. Prove the tautology fŒ.B/ ) .F/ ) Œ.A/ ) .F/g ) Œ.A/ ) .B/ within
Church’s system, using only results from the implicational calculus (axioms C1 and
C2) and axiom C3 (not axiom P3). Hint: start from axiom C2, with .B/ ) .F/
for P, with A for Q, and with F for R. Then use the transitivity of implications.
To show the equivalence of Church’s logical system and classical logic, the
following exercise establishes the equivalence of :.P/ and Church’s .P/ ) .F/.
1.42 . Prove that Œ.P/ ) .F/ ) Œ:.P/ and Œ:.P/ ) Œ.P/ ) .F/ are theorems,
using the theorem :.F/, and the classical axioms P1, P2, P3.
 www.pdfgrip.com

38 1 Propositional Logic: Proofs from Axioms and Inference Rules

1.43 . Define a “negation”  .P/ to be an abbreviation for .P/ ) .F/. Prove that
every theorem of classical logic is a theorem of Church’s logic.
1.44 . Using the theorem :.F/ and axioms P1, P2, P3, prove the theorems fŒ.P/ )
.F/ ) .F/g ) f:Œ:.P/g and f:Œ:.P/g ) fŒ.P/ ) .F/ ) .F/g .
1.45 . Using axioms
 P1, P2, P3 and anyof the classical theorems already
 proved,
prove Œ:.P/ ) .P/ ) f:Œ.S/ ) .S/g and .P/ ) f:Œ.S/ ) .S/g ) Œ:.P/.
1.46 . In classical logic define a constant f to be an abbreviation for :Œ.S/ ) .S/.
Prove that every theorem of Church’s logic is a theorem of classical logic.

1.5 Other Connectives

There exist logical connectives other than the negation .:/ and implication .)/,
for example, the conjunction .^/, disjunction ._/, and equivalence .,/. Such other
connectives can be specified by other axioms, for instance, by Tarski’s axioms IV–
VI to specify the equivalence, as in example 1.87 on page 55. Alternatively, other
connectives can be defined as abbreviations of longer formulae in terms of negations
and implications, as presented in this section.

1.5.1 Definitions of Other Connectives

The logical connectives : and ) suffice to define all the other logical connectives,
for instance, the conjunction ^, the disjunction _, and the equivalence ,, as
outlined here.
1.51 Definition (conjunction, disjunction, and equivalence). In the full Classi-
cal Propositional Calculus, the connectives ^, _, and , may be defined as follows:
.P/ ^ .Q/ stands for :f.P/ ) Œ:.Q/g;
.P/ _ .Q/ stands for Œ.P/ ) .Q/ ) .Q/;
.P/ , .Q/ stands for Œ.P/ ) .Q/ ^ Œ.Q/ ) .P/.

1.5.2 Examples of Proofs of Theorems with Conjunctions

The logical conjunction .^/ can also be introduced into a logic by additional
axioms, for instance, as in Hilbert’s Positive Propositional Calculus, Brouwer &
Heyting’s Intuitionistic Logic and Kolmogorov & Johansson’s Minimal Logic [18,
26, p. 140–146]:
 www.pdfgrip.com

1.5 Other Connectives 39

(AND.1) ` Œ.P/ ^ .Q/ ) .Q/

(AND.2) ` Œ.P/ ^ .Q/ ) .P/
(AND.3) ` .P/ ) f.Q/ ) Œ.P/ ^ .Q/g

The following theorems reveal that within the Classical Propositional Calculus,
such axioms also follow from definition 1.51.
The first theorems show that if .P/ ^ .Q/ holds, then P holds and Q holds.
1.52 Theorem (AND.1). The formula Œ.P/ ^ .Q/ ) .Q/ is a theorem (schema).
Proof. Apply transposition, double negation, transitivity, and the definition:
` Œ:.Q/
 ) f.P/ ) Œ:.Q/g axiom P1,
` :f.P/ ) Œ:.Q/g ) f:Œ:.Q/g contraposition (1.44) of P1, and M.P.,
` Œ.P/ ^ .Q/ ) f:Œ:.Q/g definition 1.51 of ^,
` f:Œ:.Q/g ) .Q/ theorem 1.42,
` Œ.P/ ^ .Q/ ) .Q/ theorem 1.16.
t
u
1.53 Theorem (AND.2). The formula Œ.P/ ^ .Q/ ) .P/ is a theorem.
Proof. Apply theorems 1.40, 1.44, 1.42, and 1.16:
` Œ:.P/
 ) Œ.P/ ) .Q/
 denial of the antecedent (1.40),
` :f.P/ ) Œ:.Q/g ) f:Œ:.P/g contraposition (1.44) of 1.40,
` Œ::.P/ ) .P/ double negation (1.42),
` Œ.P/ ^ .Q/ ) .P/ transitivity (theorem 1.16).
t
u
The next theorem demonstrates a “converse” to the preceding two theorems. so
that if P and Q hold, then their conjunction .P/ ^ .Q/ also holds.
1.54 Theorem (AND.3). If P and Q are both theorems, then .P/^.Q/ is a theorem;
equivalently, for all well-formed formulae P and Q, ` .P/ ) f.Q/ ) Œ.P/ ^ .Q/g.
Proof. Apply the
 law of assertion, contraposition,
 and transitivity:
` .P/ ) f.P/ ) Œ:.Q/g ) Œ:.Q/ assertion (theorem 1.38),
 
` f.P/
˚ ) Œ:.Q/g
 ) Œ:.Q/  contraposition : : :
) .Q/ ) :f.P/ ) Œ:.Q/g : : : continued,
˚  
` .P/ ) .Q/ ) :f.P/ ) Œ:.Q/g theorem 1.16,
` .P/ ) f.Q/ ) Œ.P/ ^ .Q/g definition of ^.
Hence, if P and Q are both theorems, then .P/ ^ .Q/ is a theorem:
` .P/ ) f.Q/ ) Œ.P/ ^ .Q/g just proved,
` .P/ hypothesis,
` .Q/ ) Œ.P/ ^ .Q/ Detachment,
 www.pdfgrip.com

40 1 Propositional Logic: Proofs from Axioms and Inference Rules

` .Q/ hypothesis,
` .P/ ^ .Q/ Detachment.
t
u
Theorem 1.54 allows for the following derived rule of inference.
1.55 Theorem (derived rule). For all well-formed formulae H, K, and L, if .H/ )
.K/ and .H/ ) .L/ are theorems, then .H/ ) Œ.K/ ^ .L/ is a theorem:
 
` Œ.H/ ) .K/ ) Œ.H/ ) .L/ ) f.H/ ) Œ.K/ ^ .L/g I

conversely, if .H/ ) Œ.K/ ^ .L/ is a theorem, then .H/ ) .K/ and .H/ ) .L/ are
theorems.
Proof. Apply theorem 1.54 and transitivity
 (theorem 1.32) with M for .K/ ^ .L/:
` .H/ ) f.K/ ) Œ.L/ ) .M/g 
) Œ.H/ ) .K/ ) f.H/ ) Œ.L/ ) .M/g axiom P2,

` f.H/ ) Œ.L/ ) .M/g

) fŒ.H/ ) .L/ ) Œ.H/ ) .M/g axiom P2,
 
` .H/ ) f.K/ ) Œ.L/ ) .M/g 
) Œ.H/ ) .K/ ) fŒ.H/ ) .L/ ) Œ.H/ ) .M/g theorem 1.32,

` .K/ ) Œ.L/ ) .M/ theorem 1.54,

` .H/ ) f.K/ ) Œ.L/ ) .M/g theorem 1.12,
` Œ.H/ ) .K/ ) fŒ.H/ ) .L/ ) Œ.H/ ) .M/g Detachment,
with M for .K/ ^ .L/. The converse results from theorems 1.52 and 1.12:
` Œ.K/ ^ .L/ ) .L/ theorem 1.52,
` .H/ ) fŒ.K/ ^ .L/ ) .L/g theorem 1.12,
` f.H/ ) Œ.Q/ ) .R/g ) fŒ.H/ ) .Q/ ) Œ.H/ ) .R/g axiom P2,
` f.H/ ) Œ.K/ ^ .L/g ) Œ.H/ ) .L/ Detachment.
Replacing Œ.K/^.L/ ) .L/ (theorem 1.52) by Œ.K/^.L/ ) .K/ (theorem 1.53)
in the foregoing proof yields a proof of f.H/ ) Œ.K/ ^ .L/g ) Œ.H/ ) .K/. u t
For instance, theorem 1.55 yields the following theorem.
1.56 Theorem (idempotency of ^). For each well-formed formula P, the formulae
.P/ ) Œ.P/ ^ .P/ and Œ.P/ ^ .P/ ) .P/ are theorems.
Proof. Substitute P for each
 of H, K, L in theorem 1.55: 
` Œ.H/ ) .K/ ) Œ.H/ ) .L/ ) f.H/ ) Œ.K/ ^ .L/g  theorem 1.55,
` Œ.P/ ) .P/ ) Œ.P/ ) .P/ ) f.P/ ) Œ.P/ ^ .P/g substitutions,
 www.pdfgrip.com

1.5 Other Connectives 41

` .P/ ) .P/ theorem 1.14,

` Œ.P/ ) .P/ ) f.P/ ) Œ.P/ ^ .P/g Detachment,
` .P/ ) Œ.P/ ^ .P/ Detachment.
The converse implication is a substitution in theorem 1.53. t
u
The next theorem shows that the conjunction ^ commutes.
1.57 Theorem (commutativity of ^). For all well-formed formulae P and Q, the
formula Œ.P/ ^ .Q/ ) Œ.Q/ ^ .P/ is a theorem.
Proof. Apply theorems 1.52, 1.53, and 1.55:
` Œ.P/ ^ .Q/ ) .Q/ theorem 1.52,
` Œ.P/ ^ .Q/ ) .P/ theorem 1.53,
` Œ.P/ ^ .Q/ ) Œ.Q/ ^ .P/ theorem 1.55.
Swapping the roles of P and Q yields the converse: ` Œ.Q/ ^ .P/ ) Œ.P/ ^ .Q/:
t
u
1.58 Theorem (derived rule). If ` Œ.H/ ^ .L/ ) .N/, then ` .H/ )
Œ.L/ ) .N/.
Proof. Apply theorems 1.54 and 1.32.
` .H/ ) f.L/ ) Œ.H/ ^ .L/g theorem 1.54,
` Œ.H/ ^ .L/ ) .N/ hypothesis,
` .H/ ) Œ.L/ ) .N/ theorem 1.32.
t
u
1.59 Theorem (derived rule). If ` .H/ ) Œ.L/ ) .N/, then ` Œ.H/ ^ .L/
) .N/.
Proof. Apply theorems 1.54 and 1.32.
1 ` Œ.H/ ^ .L/ ) .H/ theorem 1.53,
2 ` .H/ ) Œ.L/ ) .N/ hypothesis,
3 ` Œ.H/ ^ .L/ ) Œ.L/ ) .N/ lines 1, 2, theorem 1.16;
4 ` Œ.H/ ^ .L/ ) .L/ theorem 1.52,
5 ` Œ.H/ ^ .L/ ) fŒ.H/ ^ .L/ ) .N/g lines 3, 4, theorem 1.31,
6 ` Œ.H/ ^ .L/ ) .N/ theorem 1.25.
t
u

1.5.3 Examples of Proofs of Theorems with Equivalences

The logical equivalence .,/ can also be introduced into a logic by additional
axioms, for instance, Hilbert’s Positive Propositional Calculus, Brouwer & Heyt-
ing’s Intuitionistic Logic and Kolmogorov & Johansson’s Minimal Logic [18, 26,
p. 140–146], by Tarski’s axioms IV, V, and VI [129, p. 147]:

(EQ.1) Tarski’s Axiom IV: ` Œ.P/ , .Q/ ) Œ.P/ ) .Q/

 www.pdfgrip.com

42 1 Propositional Logic: Proofs from Axioms and Inference Rules

(EQ.2) Tarski’s Axiom V: ` Œ.P/ , .Q/ ) Œ.Q/ ) .P/

(EQ.3) Tarski’s Axiom VI: ` Œ.P/ ) .Q/ ) fŒ.Q/ ) .P/ ) Œ.P/ , .Q/g

The following theorems reveal that within the Classical Propositional Calculus,
such axioms also follow from definition 1.51.
Combining theorem 1.54 with the definition of , gives Tarksi’s Axiom VI.
1.60 Theorem (Tarski’s axiom VI). For all well-formed formulae U and V,
Tarski’s axioms VI is a theorem: ` Œ.U/ ) .V/ ) fŒ.V/ ) .U/ ) Œ.U/ ,
.V/g.
Proof. Combine theorem 1.54 with the definition of ,:

.U/,.V/
 ‚ …„ ƒ
Œ.U/ ) .V/ ) Œ.V/ ) .U/ ) fŒ.U/ ) .V/ ^ Œ.V/ ) .U/g :
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
P Q P Q

t
u
A particular instance of the foregoing theorem allows for any proof of any
equivalence .I/ , .J/ to be split into two separate proofs of .I/ ) .J/ and
.J/ ) .I/.
1.61 Theorem (derived rule). If .I/ ) .J/ and .J/ ) .I/ are theorems, then so
is .I/ , .J/. Conversely, if .I/ , .J/ is a theorem, then so are .I/ ) .J/ and
.J/ ) .I/.

Proof. Apply theorem 1.54 with .I/ ) .J/ for P, and with .J/ ) .I/ for Q, so that
if .I/ ) .J/ and .J/ ) .I/ are theorems, then Œ.I/ ) .J/ ^ Œ.J/ ) .I/ is also a
theorem, which is .I/ , .J/ by definition, and conversely. t
u
Theorem 1.62 discharges the hypothesis .I/ , .J/ in theorem 1.61, which yields
Tarski’s axioms IV and V.
1.62 Theorem (Tarski’s axioms IV and V). For all well-formed formulae I and
J, Tarski’s axioms IV Œ.I/ , .J/ ) Œ.I/ ) .J/ and Tarski’s axioms V Œ.I/ ,
.J/ ) Œ.J/ ) .I/ are theorems.
Proof. By definition 1.51, the formula .I/ , .J/ is an abbreviation for Œ.I/ )
.J/ ^ Œ.J/ ) .I/. The present theorem results from substitutions in theorems 1.52
and 1.53. t
u
Hence the following theorem establishes the reflexivity of equivalence.
1.63 Theorem (reflexivity of ,). For each well-formed formula P, ` .P/ , .P/.
Proof. Apply the definition of , and the reflexivity of ) (theorem 1.14) with
theorem 1.61:
` .P/ ) .P/ theorem 1.14,
` .P/ ) .P/ theorem 1.14,
` .P/ , .P/ theorem 1.61.
t
u
 www.pdfgrip.com

1.5 Other Connectives 43

The following theorem establishes the symmetry of equivalence.

1.64 Theorem (symmetry of ,). For all well-formed formulae H and K, if
.H/ , .K/ is a theorem, then so is .K/ , .H/.
Proof. Apply the definition of , and the commutativity of ^ (theorem 1.57):
` .H/ , .K/ hypothesis,
` Œ.H/ ) .K/ ^ Œ.K/ ) .H/ definition of ,,
` Œ.K/ ) .H/ ^ Œ.H/ ) .K/ commutativity of ^ and Detachment,
` .K/ , .H/ definition of ,.
t
u
Similarly, the following theorem establishes the transitivity of equivalence.
1.65 Theorem (transitivity of ,). For all well-formed formulae H, K, and L, if
.H/ , .K/ and .K/ , .L/ are theorems, then .H/ , .L/ is a theorem.
Proof. Apply the definition of , and theorems 1.61 and 1.16:
` .H/ , .K/ hypothesis,
` .H/ ) .K/ theorem 1.61,
` .K/ , .L/ hypothesis,
` .K/ ) .L/ theorem 1.61,
` .H/ ) .L/ theorem 1.16,
By symmetry (theorem 1.64) .K/ , .H/ and .L/ , .K/ are also theorems,
whence .L/ ) .H/ is a theorem. From .H/ ) .L/ and .L/ ) .H/ it then follows
that .H/ , .L/ is a theorem. t
u
The following theorem demonstrates the use of the transitivity of equivalence in
the proof that the conjunction ^ is associative.
1.66 Theorem (associativity of ^). For all well-formed formulae P, Q, and R,

` f.P/ ^ Œ.Q/ ^ .R/g ) fŒ.P/ ^ .Q/ ^ .R/g:

Proof. Apply the law of commutation (theorem

 1.37) and theorem 1.55:
` .P/  ) f.R/ ) Œ:.Q/g  twice theorem 1.37 : : :
, .R/ ) f.P/ ) Œ:.Q/g : : : and theorem 1.55,
˚   m contrapositions,
` : ˚.P/  ) f.Q/ ) Œ:.R/g 
, : .R/ ) f.P/ ) Œ:.Q/g
˚   m definition of ^,
` : ˚.P/  ) f:Œ.Q/ ^ .R/g 
, : .R/ ) f:Œ.P/ ^ .Q/g
m definition of ^,
 www.pdfgrip.com

44 1 Propositional Logic: Proofs from Axioms and Inference Rules

` f.P/ ^ Œ.Q/ ^ .R/g

, f.R/ ^ Œ.P/ ^ .Q/g
m commutativity of ^.
` f.P/ ^ Œ.Q/ ^ .R/g
, fŒ.P/ ^ .Q/ ^ .R/g
t
u

1.5.4 Examples of Proofs of Theorems with Disjunctions

The logical disjunction ._/ can also be introduced into a logic by additional
axioms, for instance, as in Hilbert’s Positive Propositional Calculus, Brouwer &
Heyting’s Intuitionistic Logic and Kolmogorov & Johansson’s Minimal Logic [18,
26, p. 140–146]:

(OR.1) ` .P/ ) Œ.P/ _ .Q/

(OR.2) ` .Q/ ) Œ.P/ _ .Q/
 
(OR.3) ` Œ.P/ ) .R/ ) Œ.Q/ ) .R/ ) fŒ.P/ _ .Q/ ) .R/g

The following theorems reveal that within the Classical Propositional Calculus,
the first two axioms also follow from definition 1.51. The third axiom is derived in
theorem 1.78.
1.67 Theorem (OR.1, OR.2). For all well-formed formulae P and Q, the formulae
.Q/ ) Œ.P/ _ .Q/ and .P/ ) Œ.P/ _ .Q/ are theorems.
Proof. For the first formula, apply axiom P1 and definition of _:
` .Q/ ) fŒ.P/ ) .Q/ ) .Q/g axiom P1,
` .Q/ ) Œ.P/ _ .Q/ definition of _.
For the second formula, apply theorem 1.14, the Law of Commutation, and the
definitions:
` Œ.P/ ) .Q/ ) Œ.P/ ) .Q/ theorem 1.14,
` .P/ ) fŒ.P/ ) .Q/ ) .Q/g Law of Commutation (theorem 1.37),
` .P/ ) Œ.P/ _ .Q/ definition 1.51 of _.
t
u
The next theorem shows that the disjunction _ is idempotent:
1.68 Theorem (idempotency of _). For each well-formed formula Q,

` .Q/ , Œ.Q/ _ .Q/:

Proof. Substituting Q for P in theorem 1.67 yields ` .Q/ ) Œ.Q/ _ .Q/.

For the converse, apply theorem 1.14, the Law of Assertion (theorem 1.38), and
Detachment:
 www.pdfgrip.com

1.5 Other Connectives 45

` Œ.Q/ ) .Q/ theorem 1.14,

„ ƒ‚ …
H  
` Œ.Q/ ) .Q/ ) fŒ.Q/ ) .Q/ ) .Q/ g ) .Q/ theorem 1.38,
„ ƒ‚ … „ ƒ‚ … „ƒ‚… „ƒ‚…
H H C C
` fŒ.Q/ ) .Q/ ) .Q/ g ) .Q/ Detachment,
„ ƒ‚ … „ƒ‚… „ƒ‚…
H C C
` Œ.Q/ _ .Q/ ) .Q/ definition of _.
Hence the equivalence ` .Q/ , Œ.Q/ _ .Q/ results from theorem 1.61. t
u
Theorem 1.69 provides an alternative definition of the disjunction in the Classical
Propositional Calculus.
1.69 Theorem. For all well-formed formulae P and Q, the following formulae are
theorems:

` fŒ:.P/ ) .Q/g ) fŒ.P/ ) .Q/ ) .Q/g;

` fŒ.P/ ) .Q/ ) .Q/g ) fŒ:.P/ ) .Q/g:

Proof. The first formula is a substitution in the proof by cases (theorem 1.49).
The second formula results from the law of denial of the antecedent (theo-
rem 1.40) and a derived rule (theorem 1.33):
` Œ:.P/ ) Œ.P/ ) .Q/ theorem 1.40,
„ƒ‚… „ ƒ‚ …
I H
` fŒ.P/ ) .Q/ ) . Q /g ) fŒ:.P/ ) . Q /g theorem 1.33.
„ ƒ‚ … „ƒ‚… „ƒ‚… „ƒ‚…
H K I K
t
u
The next theorem establishes the law of excluded middle in classical logic.
1.70 Theorem (law of excluded middle). For each well-formed formula B, the
formula .B/ _ Œ:.B/ is a theorem.
Proof. Apply theorems 1.14 and 1.69, and the definition of the disjunction ._/:
` Œ:.B/ ) Œ:.B/ theorem 1.14,
` .B/ _ Œ:.B/ theorem 1.69 and definition of _ with : and ).
t
u
The following theorem shows that the disjunction _ is associative:
1.71 Theorem (associativity of _). For all well-formed formulae P, Q, and R,

` f.P/ _ Œ.Q/ _ .R/g , fŒ.P/ _ .Q/ _ .R/g:

 www.pdfgrip.com

46 1 Propositional Logic: Proofs from Axioms and Inference Rules

Proof. Apply the law of commutation (theorem

 1.37) and theorem 1.55:
` Œ:.P/
 ) fŒ:.R/ ) .Q/g  twice theorem 1.37 : : :
, Œ:.R/ ) fŒ:.P/ ) .Q/g : : : and theorem 1.55,
m definition of _,
` f.P/ _ Œ.R/ _ .Q/g , f.R/ _ Œ.P/ _ .Q/g
m commutativity of _, etc.
` f.P/ _ Œ.Q/ _ .R/g , fŒ.P/ _ .Q/ _ .R/g
t
u

1.5.5 Examples of Proofs with Conjunctions and Disjunctions

The following theorems establish De Morgan’s laws in classical logic.

1.72 Theorem (De Morgan’s first law). For all well-formed formulae P and Q,

` f:Œ.P/ ^ .Q/g , fŒ:.P/ _ Œ:.Q/g:

Proof. Apply the definitions of ^ and _ and double negations:

:Œ.P/ ^ .Q/
  m definition of ^,

: :f.P/ ) Œ:.Q/g
m double negation (theorems 1.42 and 1.43),
.P/ ) Œ:.Q/
m double negation (theorems 1.42 and 1.43),
f:Œ:.P/g ) Œ:.Q/
m definition of _ by theorem 1.69.
Œ:.P/ _ Œ:.Q/.
t
u
1.73 Theorem (De Morgan’s second law). For all well-formed formulae P and Q,

` f:Œ.P/ _ .Q/g , fŒ:.P/ ^ Œ:.Q/g:

Proof. Apply the definitions of ^ and _ and double negations:

:Œ.P/ _ .Q/
m definition of _ by theorem 1.69,
:fŒ:.P/ ) .Q/g
 m double negation (theorems 1.42 and 1.43),
: Œ:.P/ ) f:Œ:.Q/g
m definition of ^.
Œ:.P/ ^ Œ:.Q/.
t
u
The following theorem shows that disjunctions distribute over conjunctions.
 www.pdfgrip.com

1.5 Other Connectives 47

1.74 Theorem (distributivity of _ over ^). For all well-formed formulae P, Q,

and R,

` f.P/ _ Œ.Q/ ^ .R/g , fŒ.P/ _ .Q/ ^ Œ.P/ _ .R/g:

Proof. Apply the definition of _ with De Morgan’s laws and theorem 1.55:
Œ.P/ _ .Q/ ^ Œ.P/ _ .R/
m definition of _ by theorem 1.69,
fŒ:.P/ ) .Q/g ^ fŒ:.P/ ) .R/g
m theorem 1.55,
Œ:.P/ ) Œ.Q/ ^ .R/
m definition of _ by theorem 1.69.
.P/ _ Œ.Q/ ^ .R/
t
u
The following theorem shows that conjunctions distribute over disjunctions.
1.75 Theorem (Distributivity of ^ over _). The following formula is a theorem:

f.P/ ^ Œ.Q/ _ .R/g , fŒ.P/ ^ .Q/ _ Œ.P/ ^ .R/g:

Proof. Apply the definition of ^ and theorem 1.55:

Œ.P/ ^ .Q/ _ Œ.P/ ^ .R/
   m definition of ^,
:f.P/ ) Œ:.Q/g _ :f.P/ ) Œ:.R/g
 m De Morgan’s first law,
: f.P/ ) Œ:.Q/g ^ f.P/ ) Œ:.R/g
 m theorem 1.55,
: .P/ ) fŒ:.Q/ ^ Œ:.R/g
 m De Morgan’s second law,
: .P/ ) f:Œ.Q/ _ .R/g
m definition of ^.
.P/ ^ Œ.Q/ _ .R/
t
u

1.5.6 Exercises on Other Connectives

For the following exercises, prove that the stated formulae are theorems (schemas),
using the classical propositional calculus and any of the results just proved.
1.47 . fŒ.P/ ) .Q/ ) .P/g ) .P/
 
1.48 . Œ.P/ ) .R/ ) fŒ.P/ ) .Q/ ) .R/g ) .R/
 www.pdfgrip.com

48 1 Propositional Logic: Proofs from Axioms and Inference Rules

1.49 . fŒ.P/ ) .Q/ ) .R/g ) fŒ.R/ ) .P/ ) .P/g

1.50 . fŒ.P/ ) .Q/ ) .R/g ) fŒ.P/ ) .R/ ) .R/g
 
1.51 . .P/ ) Œ:.Q/ ) f:Œ.P/ ) .Q/g
1.52 . Œ:.P/ ) f:Œ.P/ ^ .Q/g
1.53 . .P/ , fŒ.P/ ) .Q/ ) .P/g
 
1.54 . Œ:.P/ ) Œ:.Q/ ) f:Œ.P/ _ .Q/g
 
1.55 . Œ.P/ ) .Q/ , :f.P/ ^ Œ:.Q/g
1.56 . Œ.P/ ) .Q/ , fŒ:.P/ _ .Q/g

1.6 Patterns of Deduction with Other Connectives

The preceding sections have demonstrated the theoretical foundations for patterns of
deduction in the implicational calculus, for example, the transitivity of implication
and the law of commutation, and with contraposition, for example, the law of
reductio ad absurdum. Similarly, this section presents patterns of deduction with
conjunctions and disjunctions, for instance, proofs by cases or by contradiction.

1.6.1 Conjunctions of Implications

The first theorem shows yet another form of the transitivity of the logical implica-
tion.
1.76 Theorem. For all well-formed formulae P, Q, and R,

` fŒ.P/ ) .Q/ ^ Œ.Q/ ) .R/g ) Œ.P/ ) .R/:

Proof. Apply theorems 1.52, 1.53, 1.27, and 1.21:

` fŒ.P/ ) .Q/ ^ Œ.Q/ ) .R/g ) Œ.Q/ ) .R/ theorem 1.52,
„ ƒ‚ … „ ƒ‚ …
H K
` fŒ.P/ ) .Q/ ^ Œ.Q/ ) .R/g ) Œ.P/ ) .Q/ theorem 1.53,
„ ƒ‚ … „ ƒ‚ …
H L
` Œ.Q/ ) .R/ ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g theorem 1.27,
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
K L M
` fŒ.P/ ) .Q/ ^ Œ.Q/ ) .R/g ) Œ.P/ ) .R/ theorem 1.21.
„ ƒ‚ … „ ƒ‚ …
H M
t
u
 www.pdfgrip.com

1.6 Patterns of Deduction with Other Connectives 49

The following theorems give derived rules of inference with conjunctions of

implications. The first theorem shows that if either of two hypotheses leads to
a conclusion, then the disjunction of both hypotheses also leads to the same
conclusion.
1.77 Theorem. If ` .U/ ) .W/ and ` .V/ ) .W/, then ` Œ.U/ _ .V/ ) .W/:

` fŒ.U/ ) .W/ ^ Œ.V/ ) .W/g ) fŒ.U/ _ .V/ ) .W/g:

Proof. The first part of the proof assumes the two hypotheses.
` .U/ ) .W/ hypothesis,
` Œ:.W/ ) Œ:.U/ contraposition and Detachment,
` .V/ ) .W/ hypothesis,
` Œ:.W/ ) Œ:.V/ contraposition and Detachment,
` Œ:.W/ ) fŒ:.U/ ^ Œ:.V/g theorem 1.55,
` Œ:.W/ ) f:Œ.U/ _ .V/g De Morgan’s second law,
` Œ.U/ _ .V/ ) .W/ axiom P3 and Detachment.
The second part of the proof dispenses with the two hypotheses.
` fŒ.U/).W/ ^ Œ.V/).W/g)Œ.U/).W/ theorem 1.53,
` Œ.U/).W/)fŒ:.W/)Œ:.U/g transposition,
` fŒ.U/).W/ ^ Œ.V/).W/g)fŒ:.W/)Œ:.U/g theorem 1.16;

` fŒ.U/).W/ ^ Œ.V/).W/g)Œ.V/).W/ theorem 1.53,

` Œ.V/).W/)fŒ:.W/)Œ:.V/g transposition,
` fŒ.U/).W/ ^ Œ.V/).W/g)fŒ:.W/)Œ:.V/g theorem 1.16;

` fŒ.U/).W/
 ^ Œ.V/).W/g 
) fŒ:.W/)Œ:.U/g ^ fŒ:.W/)Œ:.V/g theorem 1.55;
 
` fŒ:.W/)Œ:.U/g
 ^ fŒ:.W/)Œ:.V/g

) Œ:.W/)fŒ:.U/ ^ Œ:.V/g theorem 1.55;
   
` Œ:.W/)fŒ:.U/ ^ Œ:.V/g ) fŒ.U/ _ .V/g).W/ contraposition;
 
` fŒ.U/).W/ ^ Œ.V/).W/g) fŒ.U/ _ .V/g).W/ theorem 1.55.
t
u
Theorem 1.78 shows that an intuitionistic and minimalist axiom for the disjunc-
tion is a theorem in the Classical Propositional Calculus.
1.78 Theorem (OR.3). For all propositional forms P, Q, and R,
 
` Œ.P/ ) .R/ ) Œ.Q/ ) .R/ ) fŒ.P/ _ .Q/ ) .R/g :
 www.pdfgrip.com

50 1 Propositional Logic: Proofs from Axioms and Inference Rules

Proof. Apply theorems 1.58 and 1.77:  

` fŒ.P/ ) .R/ ^ Œ.Q/ ) .R/g ) fŒ.P/ _ .Q/g ) .R/ theorem 1.77,
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
H  L N 
` Œ.P/ ) .R/ ) Œ.Q/ ) .R/ ) ffŒ.P/ _ .Q/g ) .R/g theorem 1.58.
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
H L N
t
u
The next theorem shows that the disjunction _ commutes:
1.79 Theorem (commutativity of _). For all well-formed formulae P and Q, the
formula Œ.P/ _ .Q/ ) Œ.Q/ _ .P/ is a theorem.
Proof. Substitute .Q/ _ .P/ for R in theorem 1.78 and apply theorem 1.67 with
Detachment: 
` f.P/)Œ.Q/ _ .P/g) f.Q/)Œ.Q/ _ .P/g
)fŒ.P/ _ .Q/)Œ.Q/ _ .P/g theorem 1.78 ,
` .P/)Œ.Q/ _ .P/ theorem 1.67,
` f.Q/)Œ.Q/ _ .P/g)fŒ.P/ _ .Q/)Œ.Q/ _ .P/g Detachment ,
` .Q/)Œ.Q/ _ .P/ theorem 1.67, ,
` Œ.P/ _ .Q/)Œ.Q/ _ .P/ Detachment.
The reverse implication results from swapping the rôles of P and Q. t
u
Similarly, the second theorem shows that if either of two hypotheses leads to
a conclusion, then the conjunction of both hypotheses also leads to the same
conclusion.
1.80 Theorem. If ` .U/ ) .W/ and ` .V/ ) .W/, then ` Œ.U/ ^ .V/ ) .W/:

` fŒ.U/ ) .W/ ^ Œ.V/ ) .W/g ) fŒ.U/ ^ .V/ ) .W/g:

Proof. This proof relies on theorems 1.77, 1.53, 1.52, 1.67, 1.31:
` fŒ.U/).W/ ^ Œ.V/).W/g
„ ƒ‚ …
H
)fŒ.U/ _ .V/).„ƒ‚…
W /g theorem 1.77,
„ ƒ‚ …
L M
` Œ.U/ ^ .V/)Œ.U/ _ .V/ 1.53, 1.52, 1.67,
„ ƒ‚ … „ ƒ‚ …
K L
` fŒ.U/).W/ ^ Œ.V/).W/g)fŒ.U/ ^ .V/).„ƒ‚…
W /g theorem 1.31.
„ ƒ‚ … „ ƒ‚ …
H K M
t
u
The third theorem shows that if one hypothesis leads to either of two conclusions,
then the same hypothesis also leads to the disjunction of both conclusions.
1.81 Theorem. For all well-formed formulae P, Q, and S,

` fŒ.P/ ) .Q/ _ Œ.P/ ) .S/g , f.P/ ) Œ.Q/ _ .S/g:

 www.pdfgrip.com

1.6 Patterns of Deduction with Other Connectives 51

Proof. This proof uses contraposition and previously established equivalences.

:f.P/ ) Œ.Q/ _ .S/g
m definition of ^ and equivalences,
.P/ ^ f:Œ.Q/ _ .S/g
m De Morgan’s second law,
.P/ ^ fŒ:.Q/ ^ Œ:.S/g
m idempotence of ^,
Œ.P/ ^ .P/ ^ fŒ:.Q/ ^ Œ:.S/g
m associativity and commutativity of ^,
f.P/ ^ Œ:.Q/g ^ f.P/ ^ Œ:.S/g
m definition of ^,
f:Œ.P/ ) .Q/g ^ f:Œ.P/ ) .S/g
m De Morgan’s first law.
:fŒ.P/ ) .Q/ _ Œ.P/ ) .S/g
t
u
The fourth theorem shows that if two implications hold, then the conjunction of
their hypotheses leads to the conjunction of their conclusions.
1.82 Theorem. If ` .P/ ) .Q/ and ` .R/ ) .S/, then ` Œ.P/^.R/ ) Œ.Q/^.S/:

` fŒ.P/ ) .Q/ ^ Œ.R/ ) .S/g ) fŒ.P/ ^ .R/ ) Œ.Q/ ^ .S/g:

Proof. The proof relies repeatedly on theorem 1.53:

` Œ.P/ ^ .R/ ) .P/ theorem 1.53,

` fŒ.P/ ^
 .R/ ) .P/g 
) Œ.P/ ) .Q/ ) fŒ.P/ ^ .R/ ) .Q/g theorem 1.28,
` Œ.P/ ) .Q/ ) fŒ.P/ ^ .R/ ) .Q/g Detachment;

` fŒ.P/ ) .Q/ ^ Œ.R/ ) .S/g ) Œ.P/ ) .Q/ theorem 1.53,

` fŒ.P/ ) .Q/ ^ Œ.R/ ) .S/g ) fŒ.P/ ^ .R/ ) .Q/g theorem 1.15;
„ ƒ‚ … „ ƒ‚ …
H K

` Œ.P/ ^ .R/ ) .R/ theorem 1.53,

` fŒ.P/ ^
 .R/ ) .R/g 
) Œ.R/ ) .S/ ) fŒ.P/ ^ .R/ ) .S/g theorem 1.28,
` Œ.R/ ) .S/ ) fŒ.P/ ^ .R/ ) .S/g Modus Ponens;

` fŒ.P/ ) .Q/ ^ Œ.R/ ) .S/g ) Œ.R/ ) .S/ theorem 1.53,

` fŒ.P/ ) .Q/ ^ Œ.R/ ) .S/g ) fŒ.P/ ^ .R/ ) .S/g theorem 1.15;
„ ƒ‚ … „ ƒ‚ …
H L
 www.pdfgrip.com

52 1 Propositional Logic: Proofs from Axioms and Inference Rules

` fŒ.P/ ) .Q/ ^ Œ.R/ ) .S/g

„ ƒ‚ …
 H 
) fŒ.P/ ^ .R/ ) .Q/g ^ fŒ.P/ ^ .R/ ) .S/g theorem 1.55;
„ ƒ‚ … „ ƒ‚ …
K L

 
` fŒ.P/ ^ .R/ ) .Q/g ^ fŒ.P/ ^ .R/ ) .S/g
„ ƒ‚ … „ ƒ‚ …
H H
) fŒ.P/ ^ .R/ ) Œ.Q/ ^ .S/g theorem 1.55;
„ ƒ‚ …
H

` fŒ.P/ ) .Q/ ^ Œ.R/ ) .S/g

) fŒ.P/ ^ .R/ ) Œ.Q/ ^ .S/g theorem 1.15.
t
u
Similarly, the fifth theorem shows that if two implications hold, then the
conjunction of their hypotheses leads to the disjunction of their conclusions.
1.83 Theorem. If ` .P/ ) .Q/ and ` .R/ ) .S/, then ` Œ.P/^.R/ ) Œ.Q/_.S/:

` fŒ.P/ ) .Q/ ^ Œ.R/ ) .S/g ) fŒ.P/ ^ .R/ ) Œ.Q/ _ .S/g:

Proof. This proof relies on theorem 1.82:

` fŒ.P/).Q/ ^ Œ.R/).S/g)fŒ.P/ ^ .R/)Œ.Q/ ^ .S/g 1.82,
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
H L M
` Œ.Q/ ^ .S/)Œ.Q/ _ .S/ 1.53, 1.52, 1.67,
„ ƒ‚ … „ ƒ‚ …
M N
` fŒ.P/).Q/ ^ Œ.R/).S/g)fŒ.P/ ^ .R/)Œ.Q/ _ .S/g 1.32.
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
H L N
t
u
1.84 Theorem. For all well-formed formulae P, Q, and R,

` fŒ.P/ ) .Q/ ^ .R/g ) f.P/ ) Œ.Q/ ^ .R/g:

Proof. Apply theorems 1.14, 1.82 and 1.56:

` Œ.P/).Q/)Œ.P/).Q/ theorem 1.14,
` .R/)Œ.P/).R/ axiom P1,
` fŒ.P/).Q/ ^ .R/g)fŒ.P/).Q/ ^ Œ.P/).R/g theorem 1.82,
` fŒ.P/).Q/ ^ Œ.P/).R/g)f.P/)Œ.Q/ ^ .R/ theorems 1.82 and 1.56.
t
u
 www.pdfgrip.com

1.6 Patterns of Deduction with Other Connectives 53

1.6.2 Proofs by Cases or by Contradiction

The following theorems establish further derived rules of inference. The first
theorem forms a part of the basis for an algorithm — called the Completeness
Theorem — to design proofs within the propositional calculus, as explained in
section 1.9.
1.85 Theorem (proof by cases). If .H/ ) .R/ and Œ:.H/ ) .R/ are theorems,
then R is a theorem. Hence the following formula is also a theorem (schema):
 
Œ.H/ ) .R/ ^ fŒ:.H/ ) .R/g ) .R/:

Proof. Apply theorem 1.77 and the law of excluded middle (theorem 1.70):
` .H/ ) .R/ hypothesis,
` Œ:.H/ ) .R/ hypothesis,
` f.H/ _ Œ:.H/g ) .R/ theorem 1.77,
` .H/ _ Œ:.H/ theorem 1.70,
`R Detachment;
   
` Œ.H/ ) .R/ ^ fŒ:.H/ ) .R/g ) f.H/ _ Œ:.H/g ) .R/ theorem 1.77,
` .H/
 _ Œ:.H/  theorem 1.70,
` Œ.H/ ) .R/ ^ fŒ:.H/ ) .R/g ) .R/ theorem 1.17.
t
u
Theorem 1.85 is the theoretical basis for proofs by cases, in the sense that if a
conclusion R is a necessary consequence of a case H, and if R is also a necessary
consequence of all the other cases, lumped together into :.H/, then R is a theorem.
The second theorem establishes the classical principle of proof by contradiction.
1.86 Theorem (proof by contradiction). If Œ:.R/ ) .S/ and Œ:.R/ ) Œ:.S/
are theorems, then R is a theorem: for all well-formed formulae R and S,
 
` fŒ:.R/ ) .S/g ^ fŒ:.R/ ) Œ:.S/g ) .R/:

Proof. Apply theorem 1.82, De Morgan’s first law, and contraposition:

` Œ:.R/).S/ hypothesis,
` Œ:.R/)Œ:.S/ hypothesis,
 
` fŒ:.R/).S/g
 ^ fŒ:.R/)Œ:.S/g 
) fŒ:.R/ ^ Œ:.R/g)f.S/ ^ Œ:.S/g theorem 1.82,

` Œ:.R/)fŒ:.R/ ^ Œ:.R/g theorem 1.56,

 
` fŒ:.R/).S/g
 ^ fŒ:.R/)Œ:.S/g

) Œ:.R/)f.S/ ^ Œ:.S/g theorem 1.31,
 www.pdfgrip.com

54 1 Propositional Logic: Proofs from Axioms and Inference Rules

    
` Œ:.R/)f.S/ ^ Œ:.S/g) :f.S/ ^ Œ:.S/g )f:Œ:.R/g
 contraposition,
` Œ:.R/)f.S/ ^ Œ:.S/g ) fŒ:.S/ _ .S/g).R/ De Morgan,
` Œ:.S/
 _ .S/  theorem 1.70 ,
` Œ:.R/)f.S/ ^ Œ:.S/g ).R/  theorem 1.17 ,
` fŒ:.R/).S/g ^ fŒ:.R/)Œ:.S/g ).R/ theorem 1.16.
t
u
Theorem 1.86 is the theoretical basis for proofs by contradiction: if a conclusion
S and its negation :.S/ are both necessary consequences of the negation :.R/ of a
statement R, then R is a theorem.

1.6.3 Exercises on Patterns of Deduction

1.57 . Determine whether the following derived rule holds. “If .V/ ) .W/ and
.R/ ) .S/ are theorems, then Œ.V/ _ .R/ ) Œ.W/ _ .S/ is also a theorem”:

fŒ.V/ ) .W/ ^ Œ.R/ ) .S/g ) fŒ.V/ _ .R/ ) Œ.W/ _ .S/g:

1.58 . Determine whether the following derived rule holds. “If .I/ ) .R/ and
.I/ ) Œ:.S/ are both theorems, then :Œ.R/ ) .S/ is also a theorem”:
 
Œ.I/ ) .R/ ^ f.I/ ) Œ:.S/g ) f:Œ.R/ ) .S/g:

1.59 . Determine whether the following derived rule holds. “If .U/ ) .W/ or
.V/ ) .W/ is a theorem, then Œ.U/ _ .V/ ) .W/ is also a theorem”:

fŒ.U/ ) .W/ _ Œ.V/ ) .W/g ) fŒ.U/ _ .V/ ) .W/g:

1.60 . Determine whether the following derived rule holds. “If .P/ ) .Q/ or
.R/ ) .S/ is a theorem, then Œ.P/ _ .R/ ) Œ.Q/ _ .S/ is also a theorem”:

fŒ.P/ ) .Q/ _ Œ.R/ ) .S/g ) fŒ.P/ _ .R/ ) Œ.Q/ _ .S/g:

1.61 . Determine whether the following derived rule holds. “If .P/ ) .S/ and
Œ:.Q/ ) Œ:.S/ is a theorem, then .P/ ) .Q/ is also a theorem”:
 
Œ.P/ ) .S/ ^ fŒ:.Q/ ) Œ:.S/g ) Œ.P/ ) .Q/:

1.62 . Prove that if T is a theorem, then .R/ ) Œ.T/ ) .R/ is a theorem.

1.63 . Prove that if T is a theorem, then Œ.T/ ) .R/ ) .R/ is a theorem.
1.64 . Prove that if :.F/ is a theorem, then so is fŒ:.R/ ) .F/g ) .R/.
 www.pdfgrip.com

1.6 Patterns of Deduction with Other Connectives 55

1.65 . Prove that if :.F/ is a theorem, then so is .R/ ) fŒ:.R/ ) .F/g.

1.66 . Determine whether the following derived rule of inference holds. “If :.F/
and f.P/ ^ Œ:.Q/g ) .F/ are theorems, then .P/ ) .Q/ is also a theorem.”

1.6.4 Equivalent Classical Axiom Systems

The Pure Positive Implicational Propositional Calculus just presented rests on

the rules of Detachment and Substitution with axioms P1, P2, and P3. Other
equivalent selections of axioms exist, for instance, Tarksi’s axioms I–VII listed in
example 1.87.
1.87 Example (Tarski’s Axioms). Alfred Tarski lists seven axioms [129, p. 147]:
Tarski’s Axiom I. .P/ ) Œ.Q/ ) .P/.
Tarski’s Axiom II. f.P/ ) Œ.P/ ) .Q/g ) Œ.P/ ) .Q/.
Tarski’s Axiom III. Œ.P/ ) .Q/ ) fŒ.Q/ ) .R/ ) Œ.P/ ) .R/g.
Tarski’s Axiom IV. Œ.P/ , .Q/ ) Œ.P/ ) .Q/.
Tarski’s Axiom V. Œ.P/ , .Q/ ) Œ.Q/ ) .P/.
Tarski’s Axiom VI. Œ.P/ ) .Q/ ) fŒ.Q/ ) .P/ ) Œ.P/ , .Q/g.
Tarski’s Axiom VII. fŒ:.Q/ ) Œ:.P/g ) Œ.P/ ) .Q/.
The exercises show that Tarksi’s axioms I–VII are equivalent to axioms P1,
P2, and P3. Other equivalent selections of axioms include Kleene’s, listed in
example 1.88 [72, p. 15].
1.88 Example (Kleene’s Axioms). Kleene lists four axioms [72, p. 15]:
Kleene’s axiom 1a. .A/ ) Œ.B/ ) .A/.
Kleene’s axiom 1b. Œ.A/ ) .B/ ) .f.A/
 ) Œ.B/ ) .C/g ) Œ.A/  ) .C//.
Kleene’s axiom 7. Œ.A/ ) .B/ ) f.A/ ) Œ:.B/g ) Œ:.A/ .
Kleene’s axiom 8. f:Œ:.A/g ) .A/.
The exercises also show that Kleene’s axioms are derivable from axioms P1, P2,
and P3. Yet other equivalent selections of axioms include John Barkley Rosser’s,
listed in example 1.89 [110, p. 55]:
1.89 Example (Rosser’s Axioms). Rosser lists three axioms [110, p. 55]:
Axiom R1 .P/ ) Œ.P/ ^ .P/.
Axiom R2 Œ.P/ ^ .Q/ ) .P/.
 
Axiom R3 Œ.P/ ) .Q/ ) f:Œ.Q/ ^ .R/g ) f:Œ.R/ ^ .P/g .
The exercises show that Rosser’s axioms are derivable from axioms P1, P2, and
P3. Rosser’s reference [110] establishes the converse.
 www.pdfgrip.com

56 1 Propositional Logic: Proofs from Axioms and Inference Rules

1.6.5 Exercises on Kleene’s, Rosser’s, and Tarski’s Axioms

For the following exercises, define _, ^, and , as in definition 1.51.

1.67 . Prove that Tarski’s axiom IV, Œ.P/ , .Q/ ) Œ.P/ ) .Q/, is derivable
from axioms P1, P2, and P3.
1.68 . Prove that Tarski’s axiom V, Œ.P/ , .Q/ ) Œ.Q/ ) .P/, is derivable from
axioms P1, P2, and P3.
1.69 . Prove that Tarski’s axiom VI, Œ.P/ ) .Q/ ) fŒ.Q/ ) .P/ ) Œ.P/ ,
.Q/g, is derivable from axioms P1, P2, and P3.
1.70 . Prove that every theorem derivable from Tarski’s axioms I– VII is also
derivable from axioms P1, P2, and P3.
1.71 . Prove that every theorem derivable from axioms P1, P2, and P3 is also
derivable from Tarski’s axioms I– VII.
1.72 . Prove that Rosser’s axiom R1, .P/ ) Œ.P/ ^ .P/, is derivable from
axioms P1, P2, and P3.

1.73 . Prove that Rosser’s axiom R3, Œ.P/ ) .Q/ ) f:Œ.Q/ ^ .R/g ) f:Œ.R/ ^
.P/g , is derivable from axioms P1, P2, and P3.
1.74 . Prove that Rosser’s axiom R2, Œ.P/ ^ .Q/ ) .P/, is derivable from
axioms P1, P2, and P3.
 
1.75 . Prove that Kleene’s axiom 7, Œ.A/ ) .B/ ) f.A/ ) Œ:.B/g ) Œ:.A/ ,
is derivable from axioms P1, P2, and P3.
1.76 . Prove that Kleene’s axiom 8 f:Œ:.A/g ) .A/, is derivable from axioms P1,
P2, and P3.

1.7 Completeness, Decidability, Independence, Provability,

and Soundness

This section addresses the question how to determine whether a formula admits of
a proof from selected axioms, and, if so, how to find such a proof. The main tool to
this end consists of multi-valued logics, also called “fuzzy” logics [102].

1.7.1 Multi-Valued Fuzzy Logics

Multi-valued fuzzy logics are “models” of propositional logics that assign a “value”
to each propositional variable, and hence also to each formulaic letter and each well-
formed propositional form, by means of a table or a formula.
 www.pdfgrip.com

1.7 Completeness, Decidability, Independence, Provability, and Soundness 57

Table 1.94 Church’s logic :.P/ P Q .P/ ) .Q/ .Q/ ) Œ.P/ ) .Q/
with values u, w, and
distinguished value v [18, u v v v v v v
p. 113]. u v w w w v w
u v u u u v u
w w v v v v v
w w w v w v v
w w u u u v u
v u v v v v v
v u w v w v v
v u u v u v v

1.90 Definition. A value in propositional logic may be any symbol that is not
already allowed in well-formed propositional formulae. Thus a value may not be
a connective, formulaic or propositional letter, parenthesis, bracket, or brace.
For the present purposes three values suffice, denoted here by u, v, and w. A logic
with exactly three values is also called a triadic logic. A multi-valued logic also
designates any one of its values as the distinguished value, for example, v (Fig. 1.1).
1.91 Definition. A propositional form holds, or is valid or a tautology, in a multi-
valued logic if and only if it has the distinguished value regardless of the values of
its components. The notation ˆ P indicates that P is valid [72, p. 12 & p. 14].
1.92 Remark. Such software packages as John Harrison’s program [54, 55] and
Stephen Wolfram’s Mathematica [142] provide facilities to calculate and print
multi-valued Truth tables of propositional forms, as explained in [102].
1.93 Example (Church’s triadic logic). Table 1.94 defines (by fiat) the values of
the negation :.P/ and of the implication .P/ ) .Q/ from the values of P and Q in
Church’s triadic logic [18, p. 113, un-numbered table, penultimate column]. The last
column shows how to derive the values of the compound formula .Q/ ) Œ.P/ )
.Q/ from the values of its components. The formula .Q/ ) Œ.P/ ) .Q/ is valid
because it has the distinguished value v regardless of the values of P and Q.
1.95 Example (Łukasiewicz’s triadic logic). Table 1.96 defines (by fiat) the values
of the negation :.P/ and of the implication .P/ ) .Q/ from the values of P and
Q in Łukasiewicz’s triadic logic [18, 79, p. 113, un-numbered table, last column].
The last column of table 1.96 shows how to derive the values of the compound
formula .Q/ ) Œ.P/ ) .Q/ from the values of its components. The formula
.Q/ ) Œ.P/ ) .Q/ is valid because it has the distinguished value v regardless of
the values of P and Q.

1.7.2 Sound Multi-Valued Fuzzy Logics

In some logics, validity provides a necessary criterion for provability: if there is a

proof of a theorem L, then L is valid. Such logics are called sound.
 www.pdfgrip.com

58 1 Propositional Logic: Proofs from Axioms and Inference Rules

Table 1.96 Łukasiewicz’s :.P/ P Q .P/ ) .Q/ .Q/ ) Œ.P/ ) .Q/

logic with values u, w,
and distinguished u v v v v v v
value v [79]. u v w w w v w
u v u u u v u
w w v v v v v
w w w v w v v
w w u w u v w
v u v v v v v
v u w v w v v
v u u v u v v

Fig. 1.1 Triadic Truth-tables

on page 640 of Charles
Sanders Peirce’s 1909 Logic
Notebook: MS Am 1632
(339) Houghton Library,
Harvard University, used by
permission. (http://pds.lib.
harvard.edu/pds/view/
15255301?n=640&
printThumbnails=no).

1.97 Definition. A multi-valued fuzzy logic is sound if and only if all its theorems
are valid: for every formula L, if ` L (if L is a theorem), then ˆ L (then L is valid).
Every axiom is also a theorem and thus must be valid in a sound logic.
1.98 Example. Table 1.94 in example 1.93 shows that Frege’s axiom P1 (the law
of affirmation of the consequent) is valid in Church’s triadic logic. Similarly,
exercise 1.83 confirms that Frege’s axiom P2 (the law of self-distributivity of
implication) is also valid in Church’s triadic logic. Thus all the axioms of the Pure
Positive Implicational Propositional Calculus are valid in Church’s triadic logic.
If a sound logic allows for inferences by Detachment, then Detachment must be
“sound” or “preserve” valid formulae, in the sense that from valid premisses P and
.P/ ) .Q/ Detachment must produce a valid conclusion Q.
 www.pdfgrip.com

1.7 Completeness, Decidability, Independence, Provability, and Soundness 59

1.99 Example. Table 1.94 in example 1.93 contains only one line where P and
.P/ ) .Q/ both have the distinguished value .v/: only in the first line. In that line
Q also has the distinguished value .v/. Thus Detachment preserves valid formulae
in Church’s triadic logic.
1.100 Example. Table 1.96 in example 1.95 contains only one line where P and
.P/ ) .Q/ both have the distinguished value .v/: only in the first line. In that line
Q also has the distinguished value .v/. Thus Detachment preserves valid formulae
in Łukasiewicz’s triadic logic.
Conversely, theorem 1.101 confirms that sound axioms and a sound Detachment
suffice for a propositional logic to be sound.
1.101 Theorem. Suppose that a multi-valued propositional logic is such that its
axioms are valid and Detachment from valid premisses produces a valid conclusion:
for all propositional forms H and C, if H is valid, and if .H/ ) .C/ is valid, then C
must be valid. Then such a logic is sound.
Proof (Outline). This proof shows that each step of a proof is valid. Each step C of
a proof is either an axiom or the result of Detachment. In particular, every formal
proof starts with an axiom.
If C is an axiom, then C is valid by hypothesis.
Suppose that all the steps up to but not including C have already been proved
valid. If C results from Detachment from previous steps H and .H/ ) .C/, then H
and .H/ ) .C/ are valid, and hence C is valid by the hypotheses of the theorem.
Thus every step of a proof is a valid formula. In particular, the last step of a proof
is also a valid formula.
A rigorous proof uses the Principle of Mathematical Induction in chapter 4. u t
1.102 Example. Examples 1.98 and 1.99 show that in Church’s triadic logic
Detachment is sound and all the axioms of the Pure Positive Implicational Proposi-
tional Calculus are valid. Consequently, theorem 1.101 shows that the Pure Positive
Implicational Propositional Calculus with Church’s triadic logic is sound: every
theorem of the Pure Positive Implicational Propositional Calculus is a valid formula
in Church’s triadic logic.

1.7.3 Independence and Unprovability

In a sound logic, every theorem is valid. By contraposition, if a formula is not valid,

then it is not a theorem. Consequently, to prove that there are no proofs of a formula
L from specified axioms and rules of inference, it suffices to find a system of logical
values where the logic is sound but the formula L is not valid.
1.103 Example. Example 1.102 shows that the Pure Positive Implicational Proposi-
tional Calculus is sound with Church’s triadic logic defined by table 1.94. However,
table 1.104 shows that Peirce’s Law is not valid in this logic. Consequently, Peirce’s
 www.pdfgrip.com

60 1 Propositional Logic: Proofs from Axioms and Inference Rules

Table 1.104 Peirce’s Law in P Q fŒ.P/ ) .Q/ ) .P/g ) .P/

Church’s triadic logic [18,
p. 113]. v v v v v v v
v w w v v v v
v u u v v v v
w v v w w v w
w w v w w v w
w u u v w w w
u v v u u v u
u w v u u v u
u u v u u v u

Law is unprovable in — is not a theorem of — the Pure Positive Implicational

Propositional Calculus: there are no proofs of Peirce’s Law using only Detachment
and Frege’s axiom P1 (the law of affirmation of the consequent) and axiom P2 (the
law of self-distributivity of implication).
Different analyses of different formulae may require different multi-valued fuzzy
logics. For instance, the question whether Frege’s axiom P2 (the law of self-
distributivity of implication) can be proved from axiom P1 (the law of affirmation
of the consequent) cannot be answered from Church’s triadic logic defined by
table 1.94, because both axioms are valid there. However, another multi-valued
fuzzy logic answers the question.
1.105 Example. Table 1.96 in example 1.95 shows that Frege’s axiom P1 (the law
of affirmation of the consequent) and Detachment are valid in Łukasiewicz’s triadic
logic. However, exercise 1.83 reveals that axiom P2 (the law of self-distributivity
of implication) is not valid in the same logic. Therefore, axiom P2 is not provable
in — is not a theorem of — the implicational propositional logic with Detachment
and only one axiom: the single axiom P1.
1.106 Definition. A logical formula L is logically independent from a system of
axioms and rules of inference if and only if L is not provable in — is not a theorem
of — the logic defined by the system of axioms and rules of inference.
1.107 Example. Example 1.103 shows that Peirce’s Law is independent from the
Pure Positive Implicational Propositional Calculus.
1.108 Example. Example 1.105 shows that Frege’s axiom P2 (the law of self-
distributivity of implication) is independent from the implicational propositional
logic defined by Frege’s axiom P1 (the law of affirmation of the consequent) and
Detachment.
In general, the concept of mathematical as opposed to practical impossibility
seems difficult to explain to the general public [28, 29]. Nevertheless, as just
demonstrated, multi-valued logic provides elementary examples of mathematical
impossibilities, in the form of rigorous proofs of the nonexistence of proofs of
 www.pdfgrip.com

1.7 Completeness, Decidability, Independence, Provability, and Soundness 61

specific formulae, and thereby an explanation of the concept — though not the
proof — of such impossibilities as Arrow’s Impossibility Theorem in voting
presented in chapter 7, ruler-and-compass angle trisection, duplication of the cube,
or a proof of Euclid’s fifth postulate in geometry, and a solution to the decision
problem in first-order logic, or Gödel’s Incompleteness Theorem in logic.

1.7.4 Complete Multi-Valued Fuzzy Logics

In some logics, validity provides a sufficient criterion for provability: if a formula L

is valid, then there is a proof of L. Such logics are called complete.
1.109 Definition. A multi-valued logic is complete if and only if all its valid
formulae are theorems: for every formula L, if ˆ L (if L is valid), then ` L (then L
is a theorem). Otherwise, if there are valid formulae that are not theorems, then the
logic is called incomplete.
1.110 Example. In table 1.94 from example 1.93, deleting all the lines where P or
Q may take the value w, thus keeping only the lines where P and Q may take only
the value u or v, gives a two-valued logic where Frege’s axiom P2 (the law of self-
distributivity of implication) and axiom P1 (the law of affirmation of the consequent)
and Detachment are still valid, because they are valid in the larger table 1.94.
Similarly, in table 1.104 from example 1.103, deleting all the lines where P or Q
may take the value w, thus keeping only the lines where P and Q may take only the
value u or v, gives table 1.111, where Peirce’s Law is valid.
Yet example 1.103 shows that Peirce’s Law is not a theorem of the Pure Positive
Implicational Propositional Calculus. Consequently, the Pure Positive Implicational
Propositional Calculus with the dyadic logic defined by table 1.111 is incomplete.
In a two-valued (dyadic) Boolean logic, the distinguished value v is also called
True or denoted by 1, while the other value u is also called False or denoted by 0.
1.111 Example. The three axioms formed by Peirce’s Law with Axioms P1 and P2
in addition to Detachment lead to a complete logic, where every Boolean dyadic
tautology has a proof [108, p. 25].

Table 1.111 Peirce’s Law is P Q fŒ.P/ ) .Q/ ) .P/g ) .P/

valid in Boolean dyadic logic.
v v v v v v v
v u u v v v v
u v v u u v u
u u v u u v u
 www.pdfgrip.com

62 1 Propositional Logic: Proofs from Axioms and Inference Rules

1.7.5 Peirce’s Law as a Denial of the Antecedent

Theorems 1.67, 1.68, and 1.78 show that the formula Œ.P/ ) .Q/ ) .Q/ uses
only the connective ) but is equivalent to .P/ _ .Q/ in the Classical Propositional
Calculus. Similarly, without negations, using only implications, Charles Sanders
Peirce’s Law

(Peirce’s Law) fŒ.P/ ) .Q/ ) .P/g ) .P/ (1.1)

[104, p. 189–190] corresponds to the pattern of reasoning by the Law of the

Excluded Middle .B/ _ Œ:.B/ as follows. With Boolean logic, in Peirce’s Law (1.1)
the formula .P/ ) .Q/ is True for every proposition Q if and only if P is False. For
this reason, .P/ ) .Q/ is a form of a denial of the antecedent P, as in the Law of
Denial of the Antecedent Œ:.P/ ) Œ.P/ ) .Q/ (theorem 1.40):
Peirce took ‘p  ˛’ as the denial of whatever statement ‘p’ abbreviates on the ground that
to say of a statement ‘It implies everything,’ was tantamount to saying of it ‘It is false.’ —[9,
p. 157]

With P and Q replaced by .P/ ) .Q/ and P respectively, if the denial .P/ ) .Q/
is False, then Œ.P/ ) .Q/ ) .P/ is True, whence, by Peirce’s Law (1.1) and
Detachment, the consequent P is True. In other words, from the Falsity of the denial
of P follows the Truth of P. Equivalently, if :.P/ is False, then P must be True. In
this sense, using only the implicational connective ) Peirce’s Law (1.1) expresses
the law of the excluded middle .P/ _ Œ:.P/ [104, p. 189–190].

1.7.6 Exercises on Church’s and Łukasiewicz’s Triadic Systems

For each of the following formulae, determine whether it is a triadic tautology with
Church’s table 1.94, or Łukasiewicz’s table 1.96, or both, or neither.
1.77 . Œ.P/ ) Œ.Q/ ) fŒ.Q/ ) .R/ ) Œ.P/ ) .R/g.
1.78 . .P/ ) .P/.
1.79 . fŒ.P/ ) .Q/ ) .P/g ) .P/.
 
1.80 . Œ.P/ ) .Q/ ) f.P/ ) Œ:.Q/g ) Œ:.P/ .
1.81 . fŒ.P/ ) .Q/ ) .Q/g ) Œ.Q/ ) .P/ ) .P/g.
1.82 . Œ:.P/ ) Œ.P/ ) .Q/.
1.83 . f.P/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g.
 
1.84 . Œ.P/ ) .Q/ ) f.P/ ) Œ:.Q/g ) Œ:.P/ .
1.85 . f.P/ ) Œ.Q/ ) .R/g ) f.Q/ ) Œ.P/ ) .R/g.
1.86 . fŒ:.Q/ ) Œ:.P/g ) Œ.P/ ) .Q/.
 www.pdfgrip.com

1.8 Boolean Logic 63

1.8 Boolean Logic

Boolean logic is a dyadic logic: a multi-valued logic with exactly two values. The
distinguished value may be called “True” and denoted by v, whereas the other value
may be called “False” and denoted by u. Table 1.113 defines the Boolean Truth
values of compound propositional formulae with selected connectives.
Boolean logic will lead to an algorithm to design proofs in the Full Propositional
Calculus in section 1.9.

1.8.1 The Truth Table of the Logical Implication

Based on [101], this subsection clarifies the Truth table of the logical implication.
True logical implications with a False hypothesis are convenient in classical
mathematics, but other versions of mathematics do not include them [33]. Moreover,
True logical implications with a False hypothesis rarely occur in practical reasoning:
Actually, the rule that any conditional is true if its antecedent is known to be false has almost
no parallel in natural logic. Examples of the type “if snow is black, then 2  2 D 5,” which
keep cropping up in textbooks, are only capable of confusing the student, since no natural
subsystem in our language has expressions with this semantics [81, p. 36].

Correspondingly, some computers and logical circuits do not include any facility to
test the Truth value of logical implications. Therefore the following four examples
serve solely to demonstrate the difference between Boolean algebraic logic and
practical reasoning. There exist other logics, but they are less used and more
complicated than Boolean logic [18, p. 142, p. 146, 26:11].
The present subsection shows that the Truth table for the Boolean logical
implication is the only Truth table that satisfies certain requirements. Specifically,
the present considerations confirm that the complete law of contraposition

Œ.P/ ) .Q/ , fŒ:.Q/ ) Œ:.P/g

and the nonequivalence of an implication .P/ ) .Q/ with its converse .Q/ ) .P/
hold only with implications defined as in table 1.113. To this end, denote by  any
candidate connective for a logical implication. To reflect common experience, as in

Table 1.113 Boolean dyadic logic with values u and distinguished value v.
:.P/ P Q .P/ ) .Q/ .P/ ^ .Q/ .P/ _ .Q/ Œ.P/ NOR .Q/ .P/ , .Q/
u v v v v v u v
u v u u u v u u
v u v v u v u u
v u u v u u v v
 www.pdfgrip.com

64 1 Propositional Logic: Proofs from Axioms and Inference Rules

Table 1.114 A partial Truth table for a

connective .P/  .Q/ conforming to
(Implication. 1) & (Implication. 2).
REQUIREMENT P Q .P/  .Q/
(Implication. 1) T T T
(Implication. 2) T F F

Table 1.115 The connectives #, , Õ, and ).

P Q .P/ # .Q/ P Q .P/ .Q/ P Q .P/ Õ .Q/ P Q .P/ ) .Q/
T T T T T T T T T T T T
T F F T F F T F F T F F
F T T F T F F T F F T T
T F F F F F F F T F F T

example 1.1, any concept of logical implication may have to satisfy the following
two requirements.

(Implication.1) To allow for reasoning by Detachment, if a hypothesis P holds, and

if the implication .P/  .Q/ holds, then the conclusion Q also holds.
(Implication.2) To avoid faulty reasoning, if a hypothesis P holds, but if the
conclusion Q fails, then the implication .P/  .Q/ also fails.
(Implication.3) To allow reasoning by contraposition, an implication and its con-
traposition must have the same Truth table.
(Implication.4) Again to prevent faulty reasoning, an implication and its converse
must have different Truth tables.

The first two requirements, (Implication. 1) & (Implication. 2), dictate the first two
lines of Truth table 1.114, where the hypothesis P is True. There remain only
four possibilities for  in the last two rows, where the hypothesis is False. For
convenience, denote these four connectives by ), #, , and Õ respectively
(these last three symbols are used in this manner only in the present discussion).
Table 1.115 shows their Truth values. For comparison, the last two requirements,
(Implication. 3) & (Implication. 4), dictate the last two lines of the desired Truth
table.
Verifications based on table 1.113. confirm that the logical implication ) has
the same Truth values as its contraposition has, but not as its converse has.
In contrast, .P/ # .Q/ and its contraposition Œ:.Q/ # Œ:.P/ have different
Truth values, as in table 1.116. Also, for the connective #, neither the law of
contraposition nor its converse holds.
Similarly, .P/ .Q/ and its contraposition Œ:.Q/ Œ:.P/ do not have the
same Truth values, as in table 1.117. Moreover, for the connective , neither the
law of contraposition nor its converse hold.
 www.pdfgrip.com

1.8 Boolean Logic 65

Table 1.116 The connective P Q .P/ # .Q/ Œ:.Q/ # Œ:.P/

# and its contraposition do
not have the same Truth T T T F
values. T F F F
F T T T
F F F T

Table 1.117 The connective P Q .P/ .Q/ Œ:.Q/ Œ:.P/

and its contraposition do
not have the same Truth T T T F
values. T F F F
F T F F
F F F T

Table 1.118 The connective P Q .P/ Õ .Q/ .Q/ Õ .P/

Õ and its converse have the
same Truth values. T T T T
T F F F
F T F F
F F T T

Finally, .P/ Õ .Q/ and its converse .Q/ Õ .P/ have the same Truth values, as
in table 1.118. However, for the connective Õ, both the law of contraposition and
its converse hold.
Thus the Truth table 1.113 specified for .P/ ) .Q/ is the only one that reflects
experience. There exist other concepts of logical implication, but they do not lend
themselves to Truth tables [18, p. 146, #26.12].

1.8.2 Boolean Logic on Earth and in Space

Some logical connectives — for instance, NOR — can combine so as to play the rôle
of every logical connective.
1.119 Definition. A logical connective is called primitive, or also universal, if
and only if every propositional form is logically equivalent to a propositional form
containing only that connective.
1.120 Example. The logical connective NOR, defined so that .A/ NOR .B/ stands
for :.A _ B/, is universal; in particular, the following equivalences are tautologies:

Œ:.P/ , Œ.P/ NOR .P/;

 www.pdfgrip.com

66 1 Propositional Logic: Proofs from Axioms and Inference Rules

Œ.P/ ^ .Q/ , f Œ.P/ NOR .P/ NOR Œ.Q/ NOR .Q/ g;

Œ.P/ _ .Q/ , fŒ.P/ NOR .Q/ NOR Œ.P/ NOR .Q/g;
Œ.P/ ) .Q/ , fŒ.P/NOR.P/NOR.Q/gNORfŒ.P/NOR.P/NOR.Q/g:

1.121 Example. The logical connective NAND, defined so that .A/ NAND .B/ stands
for :.A ^ B/, is universal; in particular, the following equivalences are tautologies:

Œ:.P/ , Œ.P/ NAND .P/;

Œ.P/ ^ .Q/ , f Œ.P/ NAND .Q/ NAND Œ.P/ NAND .Q/ g;
Œ.P/ _ .Q/ , fŒ.P/ NAND .P/ NAND Œ.Q/ NAND .Q/g;
Œ.P/ ) .Q/ , .P/ NAND Œ.Q/ NAND .Q/:

1.122 Example. For logic and arithmetic, Westinghouse DPS-2402 Computers used
only NAND gates:
Its function can be considered fundamental in that all sequential and combinational logic
functions can be performed entirely by NAND gates [138, Section 4,  4–1(7)(a), p. 4–4].

1.123 Example. The Apollo spacecraft contained two electrically identical Apollo
Guidance Computers (AGC): a Command Module Computer (CMC) in the Com-
mand Module (CM), and a Lunar Module Computer (LMC) in the Lunar Module
(LM) [51,  2.1, p. 23], pictured in figure 1.2.

Fig. 1.2 The logic in the

present chapter was used in
the Apollo Guidance
Computer onboard the
Command and Lunar
Modules. (Neil A.
Armstrong’s photograph of
Edwin E. Aldrin Jr., 20 July
1969, NASA ID
AS11-40-5927.)
 www.pdfgrip.com

1.9 Automated Theorem Proving 67

Each AGC used only one type of logical circuit: a NOR gate with three variables,
such that NOR .A; B; C/ is equivalent to :.A _ B _ C/ [51,  3.2.1, p. 60, fig. 3–1].
Setting C to False .0V/ shows that NOR .A; B; F/ is equivalent to :.A _ B/ and
hence to .A/ NOR .B/. By universality of NOR, every logical, arithmetic, reading,
writing, and copying operation necessary during space flight was implemented by
circuits consisting entirely and exclusively of NOR gates [51,  3.2.1, p. 62]. The
universality of NOR provides a reliability greater than several different connectives
would:
The single logic type simplified packaging, manufacturing, and testing, and gave higher
confidence to the reliability predictions [51,  1.1, p. 10].

1.9 Automated Theorem Proving

Relying on the Deduction Theorem 1.22, the Provability Theorem and the Com-
pleteness Theorem will provide not only guidance but an algorithm to design proofs
within the propositional calculus.

1.9.1 The Provability Theorem

The full Classical Propositional Calculus based on axioms P1, P2, and P3 is
absolutely complete: every tautology is a theorem. Moreover, there are algorithms
to determine for each propositional form whether it is a theorem, and, if it is, to
design a proof of it. The demonstration relies on the following notation.
1.124 Definition. For each proposition P, define a proposition P0 by

P if P is True;
P0 WD
:.P/ if P is False:

The following theorem constitutes a first step in a proof of completeness.

1.125 Theorem (Provability Theorem). For each propositional form S with
propositional variables from a finite list P; : : : ; R, there exists a proof of S0 from
P0 ; : : : ; R0 :

P0 ; : : : ; R0 ` S0 :

Proof (Outline). This proof proceeds by cases, removing from S one connective at
each step.
 www.pdfgrip.com

68 1 Propositional Logic: Proofs from Axioms and Inference Rules

Negation

If S is :.V/, then V has one fewer connective than S has. Suppose that P0 ; : : : ; R0 `
V 0 has already been proved, and consider two cases.
S True If S is True, then S0 is S. However, if S is True, then V is False, and V 0 is
:.V/, which is S and hence also S0 . Thus P0 ; : : : ; R0 ` V 0 by the hypothesis on
V, and substituting S0 for V 0 yields P0 ; : : : ; R0 ` S0 .
S False In contrast, if S is False, then S0 is :.S/. However, if S is False, then V is
True, and V 0 is V. Thus P0 ; : : : ; R0 ` V 0 by the hypothesis on V, and substituting
V for V 0 yields P0 ; : : : ; R0 ` V. Hence, appending a proof of the converse law
of double negation (theorem 1.43) produces a proof of P0 ; : : : ; R0 ` f:Œ:.V/g,
and substituting S for :.V/ gives P0 ; : : : ; R0 ` Œ:.S/, whence substituting S0 for
:.S/ yields P0 ; : : : ; R0 ` S0 .

Implication

If S is .V/ ) .W/, then V and W have fewer connectives than S has. Suppose that
P0 ; : : : ; R0 ` V 0 and P0 ; : : : ; R0 ` W 0 have already been proved, and consider two
cases. The first two cases occur if S is True, which occurs if W is True or V is False.
S True, W True If W is True, then W is W 0 and by the hypothesis on W there exists
a proof of P0 ; : : : ; R0 ` W. Again because W is True, it follows from theorem 1.12
that .V/ ) .W/ is also True, and appending a proof of theorem 1.12 after
the proof of P0 ; : : : ; R0 ` W produces a proof of P0 ; : : : ; R0 ` Œ.V/ ) .W/.
However, because .V/ ) .W/ is True and .V/ ) .W/ is S, it also follows that S
is S0 , whence P0 ; : : : ; R0 ` Œ.V/ ) .W/ is P0 ; : : : ; R0 ` S0 .
S True, W False If V is False, then V 0 is :.V/ and True. Thus there exists a proof
of P0 ; : : : ; R0 ` V 0 by the hypothesis on V, which is thus a proof of P0 ; : : : ; R0 `
Œ:.V/. Hence the law of denial of the antecedent (theorem 1.40) gives a proof
of Œ:.V/ ) Œ.V/ ) .W/, which is Œ:.V/ ) .S/, and thence the transitivity
of implications (theorem 1.16) yields a proof of P0 ; : : : ; R0 ` S, which is also a
proof of P0 ; : : : ; R0 ` S0 .
S False The third case occurs if S is False, which occurs if and only if V is True
and W is False. Then S0 is :.S/ and W 0 is :.W/ but V 0 is V. By the hypotheses
on V and W, there exist proofs of P0 ; : : : ; R0 ` V 0 and P0 ; : : : ; R0 ` W 0 , which
are thus proofs of P0 ; : : : ; R0 ` V and P0 ; : : : ; R0 ` Œ:.W/. Appending a proof
of theorem 1.54 then gives a proof of P0 ; : : : ; R0 ` f.V/ ˚ ^ Œ:.W/g, whence the
definition (1.51) of ^ produces a proof of P0 ; : : : ; R0 ` : .V/ ) f:Œ:.W/g .
Thence the converse law of double negation, transitivity applied to
˚   
Œ.V/ ) .W/ ) .W/ ) f:Œ:.W/g ) .V/ ) f:Œ:.W/g

and contraposition yield a proof of P0 ; : : : ; R0 ` f:Œ.V/ ) .W/g, which is a

proof of P0 ; : : : ; R0 ` Œ:.S/ and hence also a proof of P0 ; : : : ; R0 ` S0 .
 www.pdfgrip.com

1.9 Automated Theorem Proving 69

The general case follows by several applications of the previous cases, in a way
that may be specified more explicitly after the availability of the Principle of
Mathematical Induction in chapter 4. t
u

1.9.2 The Completeness Theorem

The Completeness Theorem shows that within the full classical propositional
calculus every tautology is a theorem, provable from the axioms and the rules of
inference.
1.126 Theorem (Completeness Theorem). Within the full classical propositional
calculus, every tautology is a theorem.
Proof. This proof uses the Deduction Theorem 1.22 and the Provability The-
orem 1.125, removing at each step one propositional variable that occurs in a
tautology.
For every tautology S with propositional variables P; : : : ; Q; R, theorem 1.125
produces a proof of P0 ; : : : ; Q0 ; R0 ` S, because S0 is S. Two cases arise with the last
variable R.
R True If R is True, then R0 is R, whence from the proof of P0 ; : : : ; Q0 ; R ` S, the
Deduction Theorem gives a proof of P0 ; : : : ; Q0 ` Œ.R/ ) .S/.
R False If R is False, then R0 is :.R/, whence from the proof of P0 ; : : : ; Q0 ; R0 ` S,
the Deduction Theorem gives a proof of P0 ; : : : ; Q0 ` fŒ:.R/ ) .S/g.
A proof of P0 ; : : : ; Q0 ` S follows by the principle of proofs by cases (theorem 1.85):
 
` Œ.R/ ) .S/ ^ fŒ:.R/ ) .S/g ) .S/:

Thus the Deduction Theorem reduces the number of propositional variables by

1. Therefore, applying the Deduction Theorem as many times as the number of
propositional variables in S yields a proof of ` S. t
u

1.9.3 Example: Peirce’s Law from the Completeness Theorem

The following considerations demonstrate how to plan the design of a proof by the
Completeness Theorem (theorem 1.126), here with the example of Peirce’s Law:

fŒ.P/ ) .Q/ ) .P/g ) .P/:

To apply the Completeness Theorem, let S designate Peirce’s Law. Because S

involves only two propositional variables, P and Q, for each of the four combi-
nations of Truth values of P and Q, the Completeness Theorem first invokes the
Provability Theorem (theorem 1.125) for a separate proof of P0 ; Q0 ` S0 , here
 www.pdfgrip.com

70 1 Propositional Logic: Proofs from Axioms and Inference Rules

 0
P0 ; Q0 ` fŒ.P/ ) .Q/ ) .P/g ) .P/ :

In all cases, S has the propositional form .V/ ) .W/, where W is P, and where V
is .H/ ) .K/, with .P/ ) .Q/ for H, and P for K:

S
‚ …„ ƒ
H K
‚ …„ ƒ ‚…„ƒ
fŒ.P/ ) .Q/ ) .P/ g ) .P/ :
„ ƒ‚ … „ƒ‚…
V W

P True, Q True If P is True, then W is also True, because W is P. Hence the

Provability Theorem calls for a proof of P ` W, which is here P ` P. Thence the
Deduction Theorem provides a proof of .P/ ) .W/, in effect here the proof of
theorem 1.14. Because S has the form .V/ ) .W/, the proof just obtained gives
the following main steps (the final complete proof replaces every theorem cited
by a complete proof of that theorem).

` .P/ ) .W/ theorem 1.14,

` .W/ ) Œ.V/ ) .W/ axiom P1,

` .P/ ) Œ.V/ ) .W/  theorem 1.16,

` .P/ ) fŒ.P/ ) .Q/ ) .P/g ) .P/ substitutions.
„ ƒ‚ … „ƒ‚…
V W

Alternatively axiom P1 yields the conclusion directly, but the foregoing deriva-
tion serves to illustrate the use of the Completeness Theorem.
P True, Q False Because P is again True, the preceding reasoning remains valid
because it does not use the Truth value of Q.
P False, Q True If P is False, then so is W. Hence the Provability Theorem calls
for a proof of V 0 . Here V is Œ.P/ ) .Q/ ) .P/, which has the form .H/ ) .K/.
With P False, H is True and K is False, whence V is False. Consequently, V 0 is
:.V/, which has the form :Œ.H/ ) .K/. Therefore, the Provability Theorem
calls for proofs of P0 ; Q0 ` H and P0 ; Q0 ` Œ:.K/.
Here P0 ; Q0 ` Œ:.K/ is Œ:.P/; Q0 ` f:Œ:.P/g, which follows from the
substitution Œ:.P/ ) Œ:.P/ in the proof of theorem 1.14.
Also, P0 ; Q0 ` H is Œ:.P/; Q0 ` Œ.P/ ) .Q/, where P is False. Thus the
Provability Theorem calls for a proof of Œ:.P/; Q0 ` Œ:.P/, which again
follows from the substitution Œ:.P/ ) Œ:.P/ in the proof of theorem 1.14.
Hence Œ:.P/ ) Œ.P/ ) .Q/ by the law of denial of the antecedent
(theorem 1.40).
These proofs of Œ:.P/; Q0 ` H and Œ:.P/; Q0 ` Œ:.K/ complete the proof of
Œ:.P/; Q0 ` f:Œ.H/ ) .K/g, which is Œ:.P/; Q0 ` Œ:.V/. Again the law of
denial of the antecedent gives a proof of Œ:.P/; Q0 ` Œ.V/ ) .W/, which is
 www.pdfgrip.com

1.9 Automated Theorem Proving 71

Œ:.P/; Q0 ` S. The proof just obtained gives the following main steps (the final
proof replaces every theorem cited by a complete proof of that theorem).

` Œ:.P/ ) Œ:.K/ theorem 1.14,

` Œ:.P/ ) Œ:.P/ theorem 1.14,
` Œ:.P/ ) fŒ:.Q/ ) Œ:.P/g axiom P1,
` fŒ:.Q/ ) Œ:.P/g ) Œ.P/ ) .Q/ axiom P3,
` Œ:.P/ ) Œ.P/ ) .Q/ theorem 1.16,
„ ƒ‚ …
H
` Œ:.P/ ) .H/ substitution;

` Œ:.P/ ) f.H/ ^ Œ:.K/g theorem 1.55,

` Œ:.P/ ) f: Œ.H/ ) .K/g definition of ^,
„ ƒ‚ …
V
` Œ:.P/ ) Œ:.V/ substitution;
` Œ:.V/ ) fŒ:.W/ ) Œ:.V/g axiom P1,
` Œ:.P/ ) fŒ:.W/ ) Œ:.V/g theorem 1.16,
` fŒ:.W/ ) Œ:.V/g ) Œ.V/ ) .W/ axiom P3,
` Œ:.P/ ) Œ.V/ ) .W/ theorem 1.16,
„ ƒ‚ …
S

 
` Œ:.P/ ) fŒ.P/ ) .Q/ ) .P/g ) .P/ substitutions.
„ ƒ‚ … „ƒ‚…
V W

P False, Q False Because P is again False, the preceding reasoning remains valid
because it does not use the Truth value of Q.

From the preceding proofs of .P/ ) .S/ and Œ:.P/ ) .S/, the principle of
proofs by cases (theorem 1.85) yields a proof of Peirce’s Law (S). Subsequent
examinations of the proof produced by the Completeness Theorem can yield
simplifications.

 
` .P/ ) fŒ.P/ ) .Q/ ) .P/g ) .P/ axiom P1,
` .P/ ) .S/ substitution;

` Œ:.P/ ) f.H/ ^ Œ:.P/g axiom P1,

` Œ:.P/ ) f: Œ.H/ ) .K/g definition of ^,
„ ƒ‚ …
V
` Œ:.P/ ) Œ:.V/ substitution;
 www.pdfgrip.com

72 1 Propositional Logic: Proofs from Axioms and Inference Rules

` Œ:.V/ ) fŒ:.W/ ) Œ:.V/g axiom P1,

` Œ:.P/ ) fŒ:.W/ ) Œ:.V/g theorem 1.16,
` fŒ:.W/ ) Œ:.V/g ) Œ.V/ ) .W/ axiom P3,
` Œ:.P/ ) Œ.V/ ) .W/ theorem 1.16,
„ ƒ‚ …
S
` Œ:.P/ ) .S/ substitution;

` .P/ ) .S/ previous result;

`S rule of inference;
` fŒ.P/ ) .Q/ ) .P/g ) .P/ substitution.

1.9.4 Exercises on the Deduction Theorem

1.87 . Assume that Œ:.P/ ) .P/ holds and prove that P holds. In other words,
prove that fŒ:.P/ ) .P/g ` .P/.
1.88 . Assume that .P/ ) Œ:.P/ holds and prove that :.P/ holds. In other words,
prove that f.P/ ) Œ:.P/g ` Œ:.P/.
1.89 . Apply the Deduction Theorem to prove fŒ:.P/ ) .P/g ) .P/ .
1.90 . Apply the Deduction Theorem to prove f.P/ ) Œ:.P/g ) Œ:.P/ .
1.91 . Assume .P/ ) .Q/ and prove f.P/ ) Œ:.Q/g
 ) Œ:.P/. In other words,
prove that Œ.P/ ) .Q/ ` f.P/ ) Œ:.Q/g ` Œ:.P/ .
1.92 . Apply the Deduction Theorem to prove the law of reductio ad absurdum
 
Œ.P/ ) .Q/ ) f.P/ ) Œ:.Q/g ) Œ:.P/ :

1.93 . Assume Œ:.P/ ) .Q/ and prove

 fŒ:.P/ ) Œ:.Q/g ) .P/. In other
words, prove that fŒ:.P/ ) .Q/g ` fŒ:.P/ ) Œ:.Q/g ) .P/ .
1.94 . Apply the Deduction Theorem to prove the law of indirect proof
 
fŒ:.P/ ) Œ:.Q/g ) fŒ:.P/ ) Œ:.Q/g ) .P/ :

1.95 . Apply the Deduction Theorem to prove the tautology

fŒ.P/ ) .Q/ ) Œ.Q/ ) .R/g ) Œ.Q/ ) .R/:

 www.pdfgrip.com

1.9 Automated Theorem Proving 73

1.96 . Apply the Deduction Theorem to prove the tautology

fŒ.P/ ) .Q/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g:

1.97 . Apply the Completeness Theorem to prove fŒ:.P/ ) .P/g ) .P/.

1.98 . Apply the Completeness Theorem to prove the special law of reductio ad
absurdum: f.P/ ) Œ:.P/g ) Œ:.P/ .
1.99 . Apply theCompleteness Theorem to prove
 the law of reductio ad absurdum:
Œ.P/ ) .Q/ ) f.P/ ) Œ:.Q/g ) Œ:.P/ .
1.100 . Apply the Completeness Theorem to prove the law of assertion:

.P/ ) fŒ.P/ ) .Q/ ) .Q/g:

1.101 . Apply the Completeness Theorem to prove

fŒ.P/ ) .Q/ ) .R/g ) fŒ.R/ ) .P/ ) .P/g:

1.102 . Apply the Completeness Theorem to prove

fŒ.P/ ) .Q/ ) .R/g ) fŒ.R/ ) .P/ ) Œ.S/ ) .P/g:

1.103 . Apply the Completeness Theorem to prove

 
fŒ.P/ ) .R/ ) .Q/g ) .Q/ ) fŒ.Q/ ) .R/ ) Œ.P/ ) .R/g:

1.104 . Apply the Completeness Theorem to prove

 
Œ.R/ ) .Q/ ) fŒ.R/ ) .Q/ ) .P/g ) Œ.S/ ) .P/ :

1.105 . Apply the Completeness Theorem to prove

fŒ.R/ ) .Q/ ) Œ.S/ ) .P/g ) fŒ.R/ ) .P/ ) Œ.S/ ) .P/g:

1.106 . Apply the Completeness Theorem to prove

   
fŒ.R/).P/).P/g)Œ.S/).P/ ) fŒ.P/).Q/).R/g)Œ.S/).P/ :
 www.pdfgrip.com

Chapter 2
First-Order Logic: Proofs with Quantifiers

2.1 Introduction

This chapter introduces quantifiers and first-order logic. The first few sections
demonstrate methods for designing proofs through preliminary versions of the
Deduction Theorem for first-order logic, Substitutivity of Equivalences, and trans-
formations into prenex forms. A final section derives features of predicates for
equality and inequality, either as primitive predicate constants, or predicates defined
from other primitive binary predicate constants. The prerequisite for this chapter
is a working knowledge of the Classical Propositional Logic for instance, as in
chapter 1.
Pure first-order logic includes quantifiers corresponding to phrases such as “for
each object” or “there exists an (at least one) object” with templates for functions
of objects and relations between objects. Applied first-order logic replaces such
templates with functions and relations specific to areas such as algebra, arithmetic,
geometry, or set theory.

2.2 The Pure Predicate Calculus of First Order

In grammar, the noun “predicate” designates the verb or verbal phrase that makes a
statement about the subject of a clause. In logic, similarly, a predicate is a part of an
atomic formula that makes a statement about individual objects in applications.

2.2.1 Logical Predicates

The logical concept of predicate depends upon the theory under consideration.

© Springer Science+Business Media New York 2015 75

Y. Nievergelt, Logic, Mathematics, and Computer Science,
DOI 10.1007/978-1-4939-3223-8_2
 www.pdfgrip.com

76 2 First-Order Logic: Proofs with Quantifiers

2.1 Example (predicates in arithmetic). Some versions of arithmetic have only two
predicates, which state that a number is the sum or product of two numbers:
M D K C L (read “M equals the sum of K and L”),
N DKL (read “N equals the product of K and L”),
or equivalent formulae with a different notation [18, p. 318], [72, p. 202–203].
These predicates are called “ternary” because each involves three variables.
2.2 Example (predicates in geometry). In geometry, a predicate may state that a
point is on a line, or that a point lies between two other points on the same line, or
that two segments, or two angles, are congruent, or that a line lies in a plane:
P2L (read “the point P is on the line L”),
X<Y<Z (read “the point Y is between the points X and Z”),
PQ  RS (read “the segment PQ is congruent to the segment RS”),
†ABC  †PQR (read “the angle ABC is congruent to the angle PQR”),
LE (read “the line L lies in the plane E”),
or equivalent formulations with a different notation [61, Ch. I].
2.3 Example (predicates in set theory). Some versions of set theory have only one
predicate, which states that a set is an element of a set:
X2Y (read “X is an element of Y”),
¿ 2 Y (read “the empty set is an element of Y”),
X 2 ¿ (read “X is an element of the empty set”),
¿ 2 ¿ (read ‘the empty set is an element of the empty set”).
This predicate is called “binary” because it involves two variables, X and Y.
The formulae in the foregoing examples are called terms or atomic formulae
because they are the simplest formulae in arithmetic, geometry, and set theory. Thus,
X 2 Y is a term, or, in other words, an atomic formula.
In arithmetic, the symbols 0 and 1 are called individual constants, because they
always denote the numbers zero and one, respectively. Similarly, in set theory, the
symbol ¿ is an individual constant, because it always denotes the empty set.
In contrast, the symbols D and 2 are called predicate constants, because they
always denote the relations of equality and set membership, respectively. Logics that
include such constants are called applied predicate calculi; they may also include
other functional constants or relational constants corresponding to other relations
between objects. In contrast, logics that do not include any constants but allow for
variables representing arbitrary individuals, predicates, functions, and relations are
called pure predicate calculi. Thus a pure predicate calculus is a general logic that
may later apply to algebra, arithmetic, geometry, and set theory as well.
In applied logics, if an atomic formula contains a variable, then it may, but need
not, have a Truth value. For example, the formula X 2 Y has no Truth value, because
different substitutions for X and Y can yield different Truth values. However some
formulae may contain variables and yet have a Truth value.
 www.pdfgrip.com

2.2 The Pure Predicate Calculus of First Order 77

2.4 Example. In logics with an “equality” relation, the formula X D X is True for
every X.
2.5 Example. In binary arithmetic the formula 0  X D 0 is True for every X.
2.6 Example. In binary arithmetic 0  X D 1 is False for every X.
2.7 Example. In set theory the formula X 2 ¿ is False for every X.
2.8 Example. In the theory of well-formed sets the formula X 2 X is False for
every X.

2.2.2 Variables, Quantifiers, and Formulae

The formulae studied here are those specified in definition 2.9.

2.9 Definition (well-formed formulae). Select three disjoint lists of symbols.
Every symbol from the first list of symbols, which may consist of one or more
letter(s) from a specified alphabet, P, Q, : : :, optionally with subscript(s) P[ , P[[ ; : : :,
superscript(s) P] , P]] , : : :, or “middlescript(s)” Pj, Pjj, : : :, is called a formulaic
letter. Such formulaic letters are not parts of the predicate calculus, but they help in
describing the following rules to define well-formed formulae.
Also, every symbol from the second list of symbols, which may consist of one
or more letter(s) from a specified alphabet, A, B, : : :, optionally with subscript(s)
A[ , A[[ , : : :, superscript(s) A] , A]] , : : :, or “middlescript(s)” Aj, Ajj, : : :, is called a
propositional variable or a sentence symbol [31, p. 17]. (Propositional variables
may later be replaced in pure calculi by functional or relational variables, or in
applied calculi by atomic formulae, which may include individual variables or
constants specific to applications.)
Moreover, every symbol from the third list of symbols, which may consist of one
or more letter(s) from a specified alphabet, X, Y, : : :, optionally with subscript(s)
X[ , X[[ , : : :, superscript(s) X ] , X ]] , : : :, or “middlescript(s)” Xj, Xjj, : : :, is called an
individual variable. (Individual variables may later be replaced by items specific
to applications, for instance, numbers in arithmetic, or points in geometry.)
Every propositional variable or atomic formula is a well-formed formula. For
all well-formed formulae P and Q, and for every individual variable X, the following
four strings of symbols are also well-formed formulae:

.W1/ :.P/ (read “not P”);

.W2/ .P/ ) .Q/ (read “P implies Q” or “if P, then Q”);
.W3/ 8X.P/ (read “for each X, P”);
.W4/ 9X.P/ (read “there exists X such that P”):
 www.pdfgrip.com

78 2 First-Order Logic: Proofs with Quantifiers

Furthermore, only strings of symbols built from letters or variables through appli-
cations of the rules W1–W4 can be well-formed formulae. Equivalent definitions
apply to other connectives and to prefix and postfix notations.

2.2.3 Proper Substitutions of Free or Bound Variables

In the logic presented here, only individual variables may appear immediately
after either quantifier, 8 (read “for each”) or 9 (read “there exists”). Because
of this restriction, this logic is a first order logic. Logical systems allowing for
propositional variables to appear immediately after a quantifier are of second or
higher order.
In Boolean logic, if a formula P is True regardless of X, but if P also contains
another variable Z, then substituting Z for X can change the Truth value of P.
2.10 Counterexample. Consider any context with at least two different objects,
for instance, two binary numbers in arithmetic, two points in geometry, or two
sets in set theory. Thus for each object X there exists a different object Z,
whence 8Xf9ZŒ:.X D Z/g is True. Replacing Z by X in 9ZŒ:.X D Z/ gives
9XŒ:.X D X/, which is False, because each object equals itself, by example 2.4.
Thus, replacing Z by X in the True formula 8Xf9ZŒ:.X D Z/g yields the False
formula 8Xf9XŒ:.X D X/g. Similarly, replacing X by Z in the True formula
8Xf9ZŒ:.X D Z/g gives the False formula 8Zf9ZŒ:.Z D Z/g.
One way (not pursued here) to avoid the phenomenon exhibited in counterexam-
ple 2.10 consists of substituting parameters other than variables [117]. Alternatively,
counterexample 2.10 shows that substitutions of a variable by another must obey
certain rules, for instance, with the concepts introduced in definition 2.11.
2.11 Definition (free or bound variables). For each individual variable X and for
each logical formula P, an occurrence of the variable X is bound in the formula P
if and only if in P that occurrence of the variable X immediately follows 8 or 9,
or if it appears in the scope of the quantifier, which is defined to be between either
8X. or 9X. and the corresponding right parenthesis //. An occurrence of the variable
X is free in P if and only if that occurrence of X is not bound in P. A logical formula
is closed if and only if it does not contain any free occurrence of any variable.
A logical sentence is a closed logical formula.
2.12 Example. This example focuses on the formula from counterexample 2.10.
In the formula 9ZŒ:.X D Z/, both occurrences of the variable Z are bound.
In the formula 9ZŒ:.X D Z/, the only occurrence of the variable X is free.
The formula 9ZŒ:.X D Z/ is not closed, because it contains a free occurrence of X.

In contrast, the formula 8Xf9ZŒ:.X D Z/g is closed, because all occurrences of X

and Z are bound.
The following definitions avoid the phenomenon in counterexample 2.10.
 www.pdfgrip.com

2.2 The Pure Predicate Calculus of First Order 79

2.13 Definition (change of bound variables). The substitution of the variable Z

for each bound occurrence of the variable X in P is denoted by SubbXZ .P/. Such a
substitution is called a change of bound variables if and only if Z does not occur
in P. Such a change of bound variables is proper if and only if X does not occur
freely in P and Z does not occur in P.
2.14 Example. In the formula 9ZŒ:.X D Z/ from counterexample 2.10, the
substitution of X for the bound occurrences of Z is not a change of bound variables,
because X already occurs in the formula.
Definition 2.13 explicitly applies only to variables. In particular, it does not allow
for substitutions of constants for bound variables.
2.15 Remark. If SubbXZ .P/ is a change of bound variables in a formula P, then
SubbZX ŒSubbXZ .P/ reproduces P.
Indeed, if SubbXZ .P/ is a change of bound variables, then Z does not occur
in P. Consequently, the only occurrences of Z in SubbXZ .P/ are those replacing
bound occurrences of X. Therefore, Z does not occur freely in SubbXZ .P/. Thus,
SubbZX ŒSubbXZ .P/ replaces all the occurrences of Z in SubbXZ .P/, all of which are
bound, by the initially bound occurrences of X, and reverts to the initial formula P.
Depending on the axiomatic system, an axiom, theorem, or inference rule may
declare that two formulae that differ from each other only by changes of bound
variables are mutually equivalent, so that .P/ , ŒSubbXZ .P/ [89, p. 181].
2.16 Definition (change of free variables). A formula P admits (the substitution
of) Z for an individual variable X, or, in other words, Z is free (to be substituted) for
X in P, if and only if in the substitution of Z for every free occurrence of X
• every free occurrence of X becomes a free occurrence of Z,
or, equivalently, if X, Z, and P satisfy the following condition:
• either Z is an individual constant, or
• Z is an individual variable, and substituting Z for every free occurrence of X in P
does not convert any free occurrence of X into a bound occurrence of Z.
In the present exposition, the notation SubfXZ .P/ states that P admits Z for X, and
substitutes Z for each free occurrence of X in P.
2.17 Example. In counterexample 2.10, the formula 9ZŒ:.X D Z/ does not admit
Z for X, or, in other words, Z is not free for X, because substituting Z for every free
occurrence of X converts the free occurrence of X into a bound occurrence of Z in
9ZŒ:.Z D Z/.
Thus another way to avoid the phenomenon exhibited in counterexample 2.10
consists in allowing only substitutions admitted in the sense of definition 2.16 [72,
p. 94], [84, p. 48] [108, p. 37], [110, p. 101].
 www.pdfgrip.com

80 2 First-Order Logic: Proofs with Quantifiers

2.18 Remark. If an individual variable Z does not occur in a formula P, then P

admits Z for each individual variable X. Indeed, Z does not appear in P after any
quantifier (8 or 9); thus Z is not bound in P, so that every free occurrence of X is
replaced by a free occurrence of Z in SubfXZ .P/.
2.19 Remark. If an individual variable Z does not occur in a formula P, and
if P does not contain any bound occurrences of an individual variable X, then
SubfZX ŒSubfXZ .P/ is P. Indeed, by hypothesis all occurrences of X are free in P, and
by remark 2.18 they all become free occurrences of Z in SubfXZ .P/. Also, X does not
occur in SubfXZ .P/. Again by remark 2.18 but with X and Z swapped, all occurrences
of Z in SubfXZ .P/ become free occurrences of X in SubfZX ŒSubfXZ .P/, which no longer
contain any occurrences of Z. Thus SubfZX ŒSubfXZ .P/ is P.
2.20 Remark. All well-formed formulae P and Q result from the construction
specified in definition 2.9, so that individual variables occur only immediately after
quantifiers, or in terms (atomic formulae) that are then combined with connectives.
Consequently, the following pairs of formulae are not only mutually equivalent but
also mutually identical [117, p. 44]:
• SubfXZ Œ:.P/ and :ŒSubfXZ .P/.
• SubfXZ Œ.P/ ) .Q/ and ŒSubfXZ .P/ ) ŒSubfXZ .Q/.
• SubbXZ Œ:.P/ and :ŒSubbXZ .P/.
• SubbXZ Œ.P/ ) .Q/ and ŒSubbXZ .P/ ) ŒSubbXZ .Q/.
The following abbreviation is convenient.
2.21 Definition (abbreviation). The notation 9ŠX.P/ (read “there exists a unique
X such that P” or “there exists exactly one X such that P”) abbreviates the formula
  
9X .P/ ^ 8ZfŒSubfXZ .P/ ) .Z D X/g :

2.2.4 Axioms and Rules for the Pure Predicate Calculus

As the axioms of the propositional calculus reflect patterns of deductive reasoning

with implications and negations, the axioms of the predicate calculus reflect
patterns of deductive reasoning with quantifiers. There also exist several choices
of initial axioms for use with quantifiers, for instance, the following axioms
[18, 30, p. 171–172], [122, p. 170].
2.22 Definition. The following axioms of the Pure Predicate Calculus govern the
use of the universal quantifier 8 and the existential quantifier 9.
Axiom Q0 Axioms of the Propositional Calculus, but here with well-formed
formulae as in definition 2.9, are axioms of the predicate calculus.
Axiom Q1 (specialization) Œ8X.P/ ) ŒSubfXZ .P/, if P admits Z for X.
 www.pdfgrip.com

2.2 The Pure Predicate Calculus of First Order 81

Axiom Q2 f8XŒ.P/ ) .Q/g ) f.P/ ) Œ8X.Q/g, if P contains no free X.

Axiom Q3 f8XŒ:.P/g , f:Œ9X.P/g.
Axiom Q4 f9XŒ:.P/g , f:Œ8X.P/g.
The first axiom (schema) of the predicate calculus (Q0) and the rules of inference
carry all the theorems from the propositional calculus over to theorems of the
predicate calculus. In particular, different propositional calculi, which may result
from different axiom systems, may lead to different predicate calculi.
2.23 Example. From definition 1.4, the following formulae (P1) and (P2) form a
system of two axioms for the Pure Positive Implicational Propositional Calculus:

(P1) .P/ ) Œ.Q/ ) .P/;

(P2) f.P/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g:

Formulae (P1), (P2), and

(P3) fŒ:.Q/ ) Œ:.P/g ) Œ.P/ ) .Q/;

form a system of three axioms for the Pure Classical Propositional Calculus.
The second axiom (schema) (Q1) corresponds to the notion that if an individual
variable X may occur in a formula P, and if P is True regardless of X, in other words,
if 8X.P/ is True, then P remains True with X replaced by any individual variable or
constant Z. If X and Z are the same variable, then axiom Q1 gives Œ8X.P/ ) .P/.
The third axiom (schema) (Q2) describes the relation between the universal
quantifier (“for each”) and the logical connective of the Pure Positive Implicational
Propositional Calculus (“if : : : then”).
The fourth axiom, for the existential quantifier (Q3), states that a formula P is
False for every X if and only if there does not exist any X for which P is True.
Similarly, the fifth axiom (schema), for the existential quantifier (Q4), states that
there exists some X for which P is False if and only if it is False that P is True
for every X. Axiom Q4 asserts the existence of an object. Consequently, axiom Q4
applies neither to “empty” theories where nothing exists, nor to logics that require
not only existence but also the determination of which objects satisfy a formula.
Besides the axioms, the predicate calculus allows for proofs of theorems through
the following rules of inference.
2.24 Definition (rules of inference). The following rules of inference hold.
2.25 Rule (“Modus Ponens” (abbreviated by M. P.), or “Detachment”).
If P is a theorem, and
if .P/ ) .Q/ is a theorem,
then Q is a theorem.
 www.pdfgrip.com

82 2 First-Order Logic: Proofs with Quantifiers

2.26 Rule (Generalization).

If P is a theorem,
then 8X.P/ is also a theorem.

2.27 Definition (theorems and proofs). A proof is a sequence of well-formed

formulae H, K, L, : : : P, Q, R, where each formula is either (a substitution in) an
axiom (schema), or results from a previous formula in the sequence by any rule of
inference (Detachment, Generalization, or Substitution).
A formula is a theorem if and only if it is a (usually the last) formula in a proof.
The notation ` R means that R is a theorem.
2.28 Example. Every axiom of the predicate calculus is a theorem.

2.2.5 Exercises on Quantifiers

Each of the following ten exercises lists one formula P. Identify a formula that is
logically equivalent to :.P/ among the same ten exercises.
2.1 . 8XŒ9Y.X 2 Y/
2.2 . 8XŒ9Y.Y 2 X/
2.3 . 8XŒ.X 2 A/ ) .X 2 B/
2.4 . 8Xf.X 2 C/ , Œ.X 2 A/ ^ .X 2 B/g
2.5 . 8Xf.X 2 C/ , Œ.X 2 A/ _ .X 2 B/g
 
2.6 . 9X f.X 2 C/^Œ:.X 2 A/^Œ:.X 2 B/g_fŒ:.X 2 C/^Œ.X 2 A/_.X 2 B/g
2.7 . 9Xf.X 2 A/ ^ Œ:.X 2 B/g
2.8 . 9Xf8YŒ:.Y 2 X/g
 
2.9 . 9X .X 2 C/ ^ fŒ:.X 2 A/ _ Œ:.X 2 B/g _ fŒ:.X 2 C/ ^ Œ.X 2 A/ ^ .X 2
B/g
2.10 . 9Xf8YŒ:.X 2 Y/g

2.2.6 Examples with Implicational and Predicate Calculi

The examples of theorems and proofs selected for this and the subsequent subsec-
tions gradually build up a tool to design proofs by substituting mutually equivalent
formulae for one another. As a first step, the following derived rules of inference
will simplify proofs by avoiding potentially lengthy instances of axiom Q2.
 www.pdfgrip.com

2.2 The Pure Predicate Calculus of First Order 83

2.29 Theorem (derived rule). If X is not free in R, and if 8XŒ.R/ ) .S/ is a

theorem, then .R/ ) Œ8X.S/ is also a theorem.
Proof. Apply axiom Q2 and Detachment:
` 8XŒ.R/ ) .S/ hypothesis,
` f8XŒ.R/ ) .S/g ) f.R/ ) Œ8X.S/g axiom Q2, no free X in R,
` .R/ ) Œ8X.S/ Detachment.
t
u
2.30 Theorem (derived rule). If X does not occur freely in R, and if .R/ ) .S/ is
a theorem, then .R/ ) Œ8X.S/ is also a theorem.
Proof. Apply Generalization and theorem 2.29:
` .R/ ) .S/ hypothesis,
` 8XŒ.R/ ) .S/ Generalization,
` .R/ ) Œ8X.S/ theorem 2.29, no free X in R.
t
u
Theorem 2.31 reveals a situation where 8YŒSubfXY .U/ may be replaced by
8XŒSubfXX .U/, which is 8X.U/.
2.31 Theorem (change of bound variables). If Y does not occur in U, then `
Œ8X.U/ ) f8YŒSubfXY .U/g.
If Y does not occur in U, and if U does not contain any bound occurrence of X,
then conversely ` Œ8X.U/ ( f8YŒSubfXY .U/g.
Proof. This proof follows Monk’s [89, p. 180, thm. 10.55].
` Œ8X.U/ ) ŒSubfXY .U/ specialization (axiom Q1), no Y in U,
„ ƒ‚ … „ ƒ‚ …
R S
R 8YŒS
‚ …„ ƒ ‚ …„ ƒ
` Œ8X.U/ ) f8YŒSubfXY .U/g theorem 2.29, no Y in U, so no Y in R.
For the converse, if Y does not occur in U, and if U does not contain any bound
occurrence of X, then each occurrence of X in U is replaced by a free occurrence of
Y in SubfXY .U/. Consequently, if V denotes the formula SubfXY .U/, then V contains
no occurrences of X and no bound occurrences of Y. Moreover, SubfYX ŒSubfXY .U/
reproduces U, by remark 2.19. Thus, applying the previous result to the formula V
and swapping the roles of X and Y give ` Œ8Y.V/ ) f8XŒSubfYX .V/g. t
u
2.32 Remark (change of bound variables). Theorem 2.31 shows that two formulae
P and Q that differ from each other only by the names of their bound variables are
mutually equivalent. Indeed, if the variable Y does not occur in a formula P, and if U
is any atomic formula (for instance, X 2 Z) that occurs in P, then U does not contain
any bound variables; in particular, U does not contain any bound occurrence of X.
Consequently ` Œ8X.U/ , f8YŒSubfXY .U/g by theorem 2.31. This substitution
may use different new variables Y, Y[ , Y[[ , : : :, for different atomic formulae U, U[ ,
U[[ , : : : and then proceed to more complicated components of P. Using the same
new variables in P and Q results in two identical formulae.
 www.pdfgrip.com

84 2 First-Order Logic: Proofs with Quantifiers

2.33 Example. Let P denote the formula 8Xf9ZŒ:.X D Z/g, and let Q denote the
formula 8Wf9YŒ:.W D Y/g. The variables Y[ and Y[[ occur in neither P nor Q.
In P, let U[ denote the atomic formula X D Z. Then ` Œ9Z.U[ / ,
f9Y[ ŒSubfZY[ .U[ /g by theorem 2.31. Let U[[ denote the resulting formula 9Y[ Œ:.X D
Y[ /. Then theorem 2.31 shows that ` Œ8X.U[[ / , f8Y[[ ŒSubfXY[[ .U[[ /g, which is
8Y[[ f9Y[ Œ:.Y[[ D Y[ /g. The same formula results from the same procedure applied
to Q.
The following selection of theorems also relates the present axioms to other
axiom systems in subsection 2.2.8. Their proofs follow Church’s [18, p. 186–188].
2.34 Theorem. For all P, Q, and X, ` f8XŒ.P/ ) .Q/g ) fŒ8X.P/ ) .Q/g.
Proof. Apply the Implicational Calculus with axiom Q1:
` f8XŒ.P/ ) .Q/g ) Œ.P/ ) .Q/ axiom Q1,
` Œ8X.P/ ) .P/ axiom Q1,
` f8XŒ.P/ ) .Q/g ) fŒ8X.P/ ) .Q/g derived rule (theorem 1.31).
t
u
2.35 Theorem. For all P, Q, and X, ` f8XŒ.P/ ) .Q/g ) fŒ8X.P/ )
Œ8X.Q/g.
Proof. Apply the Implicational Calculus with Generalization, axiom Q2, and
theorems 2.30 and 2.34:
` f8XŒ.P/ ) .Q/g ) fŒ8X.P/ ) .Q/g theorem 2.34,
„ ƒ‚ … „ ƒ‚ …
R S
R 8XfSg
‚ …„ ƒ ‚ …„ ƒ
` f8XŒ.P/ ) .Q/g ) 8XfŒ8X.P/ ) .Q/g theorem 2.30,
` 8XfŒ8X.P/ ) .Q/g ) fŒ8X.P/ ) Œ8X.Q/g axiom Q2,
` f8XŒ.P/ ) .Q/g ) fŒ8X.P/ ) Œ8X.Q/g transitivity (1.16).
t
u
2.36 Counterexample. The converse of theorem 2.35, which would be

Œ8X.P/ ) 8X.Q/ ) f8XŒ.P/ ) .Q/g;

is False in contexts with two different objects Y and Z, so that :.Y D Z/ is True:
• 8XŒ.X D Y/ ) .X D Z/ is False, because substituting Y for X gives Œ.Y D
Y/ ) .Y D Z/, which is False, because of the True hypothesis Y D Y and the
False conclusion Y D Z.
• 8X.X D Y/ is False, because substituting Z for X gives .Z D Y/, which is False
by the assumption that :.Y D Z/.
• Œ8X.X D Y/ ) Œ8X.X D Z/ is True, because of its False hypothesis.
• fŒ8X.X D Y/ ) Œ8X.X D Z/g ) f8XŒ.X D Y/ ) .X D Z/g is False,
because of the True hypothesis and the False conclusion.
 www.pdfgrip.com

2.2 The Pure Predicate Calculus of First Order 85

2.37 Theorem (derived rule). For all P, Q, and X, if ` .P/ ) .Q/, then
` Œ8X.P/ ) .Q/ and ` Œ8X.P/ ) Œ8X.Q/ .
Proof. Apply theorems 2.34 and 2.35:
` .P/ ) .Q/ hypothesis,
` 8XŒ.P/ ) .Q/ Generalization,
` fŒ8X.P/ ) .Q/g theorem 2.34 and Detachment,
` Œ8X.P/ ) Œ8X.Q/ theorem 2.35 and Detachment.
t
u
2.38 Theorem. For all P, Q, and X, if P does not contain any free occurrence of X,
then ` f.P/ ) Œ8X.Q/g , f8XŒ.P/ ) .Q/g .
Proof. Axiom Q2 gives ` f8XŒ.P/ ) .Q/g ) f.P/ ) Œ8X.Q/g. For the
converse, use the Pure Positive Implicational Propositional Calculus with axioms Q1
and Generalization:
` Œ8X.Q/).Q/ axiom Q1,
` f.P/)Œ8X.Q/g)f.P/)Œ8X.Q/g theorem 1.14,
` f.P/)Œ8X.Q/g)Œ.P/).Q/ theorem 1.32,
 
` 8X f.P/)Œ8X.Q/g)Œ.P/).Q/ Generalization,
„ ƒ‚ … „ ƒ‚ …
R S
R 8XfSg
‚ …„ ƒ ‚ …„ ƒ
` f.P/)Œ8X.Q/g) 8XfŒ.P/).Q/g theorem 2.29, no free X in .P/)Œ8X.Q/.
t
u
2.39 Theorem. For all P and X, if P does not contain any free occurrence of X,
then ` Œ8X.P/ ) .P/ and ` Œ8X.P/ ( .P/ .
Proof. Axiom Q1 gives ` Œ8X.P/ ) .P/. For the converse, apply theorems 1.14
and 2.30:
` .P/ ) .P/ theorem 1.14,
` .P/ ) Œ8X.P/ theorem 2.30.
t
u
2.40 Remark. The statement of theorem 2.39 suggests that if P contains a free
occurrence of X, then the implication .P/ ) Œ8X.P/ may differ from the
Generalization rule, from ` P to infer ` 8X.P/, which applied only if P is a
theorem.
2.41 Example. If P denotes the formula X D ¿, then .P/ ) Œ8X.P/ becomes
.X D ¿/ ) Œ8X.X D ¿/, which is not a theorem. Indeed, if .X D ¿/ )
Œ8X.X D ¿/ were a theorem, then substituting ¿ for the free occurrences of X by
specialization and Detachment would yield .¿ D ¿/ ) Œ8X.X D ¿/, which is
not a theorem, because ¿ D ¿ is True while 8X.X D ¿/ is False in set theory.
Theorem 2.42 provides a converse for theorem 2.35 if X is not free in P.
2.42 Theorem. For all P, Q, and X, if X does not occur freely in P, then `
fŒ8X.P/ ) Œ8X.Q/g ) f8XŒ.P/ ) .Q/g.
 www.pdfgrip.com

86 2 First-Order Logic: Proofs with Quantifiers

Proof. Use theorems 2.38 and 2.39, with H, K, L, M as in theorem 1.36:

P / ) Œ8X.Q/g ) f8XŒ.P/ ) .Q/g theorem 2.38, no free X in P,
` f.„ƒ‚…
„ ƒ‚ … „ ƒ‚ …
H L M
` .„ƒ‚…
P / ) Œ8X.P/ theorem 2.39, no free X in P,
„ƒ‚…
H K
` fŒ8X.P/ ) Œ8X.Q/g ) f8XŒ.P/ ) .Q/g theorem 1.36.
„ƒ‚… „ ƒ‚ … „ ƒ‚ …
K L M
t
u

2.2.7 Examples with Pure Propositional and Predicate Calculi

The following theorems invoke the full Classical Propositional Calculus, including
contraposition and its converse for negations, or Tarski’s axioms for equivalences.
2.43 Theorem. For all P, Q, and X,
` f8XŒ.P/ , .Q/g ) fŒ8X.P/ , Œ8X.Q/g.
Proof. Apply theorems 2.37 and 2.35 with the transitivity of implication:
` Œ.P/ , .Q/ ) Œ.P/ ) .Q/ definition of ,,
` f8XŒ.P/ , .Q/g ) f8XŒ.P/ ) .Q/g theorem 2.37,
` f8XŒ.P/ ) .Q/g ) fŒ8X.P/ ) Œ8X.Q/g theorem 2.35,
` f8XŒ.P/ , .Q/g ) fŒ8X.P/ ) Œ8X.Q/g transitivity.
The converse conclusion results from the symmetry of , and swapping P and Q.
The final result then follows from theorem 1.55. t
u
For the records, theorem 2.44 combines theorems 2.35, 2.42 and 2.43.
2.44 Theorem. For all P, Q, and X, if X does not occur freely in P, then `
fŒ8X.P/ ) Œ8X.Q/g , f8XŒ.P/ ) .Q/g.
Proof. Apply theorems 2.35, 2.42 and 2.43. t
u
2.45 Theorem (derived rule). For all P, Q, and X, if ` .P/ , .Q/, then
` Œ8X.P/ , Œ8X.Q/.
Proof. Apply Generalization, theorem 2.43, and Detachment:
` .P/ , .Q/ hypothesis,
` 8XŒ.P/ , .Q/ Generalization,
` f8XŒ.P/ , .Q/g ) fŒ8X.P/ , Œ8X.Q/g theorem 2.43,
` Œ8X.P/ , Œ8X.Q/ Detachment.
t
u
Theorems 2.46 and 2.47 show that 9 could be defined in terms of 8 and
double negation, or vice versa, provided that axiom Q0 includes the full Classical
Propositional Calculus.
 www.pdfgrip.com

2.2 The Pure Predicate Calculus of First Order 87

 
2.46 Theorem. For all P and X, ` Œ9X.P/ , :fŒ8XŒ:.P/g :
Proof. Apply the full propositional calculus and axiom Q3:
` f:Œ9X.P/g
 , f8XŒ:.P/g
  axiom Q3,
` :f:Œ9X.P/g , :f8XŒ:.P/g contraposition and its converse, t
u
 
` Œ9X.P/ , :f8XŒ:.P/g double negation and transitivity.
 
2.47 Theorem. For all P and X, ` Œ8X.P/ , :fŒ9XŒ:.P/g :
Proof. Apply the full propositional calculus and axiom Q4:
` f:Œ8X.P/g
 , f9XŒ:.P/g
  axiom Q4,
` :f:Œ8X.P/g , :f9XŒ:.P/g contraposition and its converse, t
u
 
` Œ8X.P/ , :f9XŒ:.P/g double negation and transitivity.
2.48 Theorem (existential generalization). For all X, Y, and P, ` ŒSubfXY .P/ )
Œ9X.P/. In particular, ` .P/ ) Œ9X.P/.
Proof. Apply the propositional calculus with axioms Q1 and Q3:
` f8XŒ:.P/g ) fSubfXY Œ:.P/g axiom Q1,
` f:Œ9X.P/g ) f8XŒ:.P/g axiom Q3,
` fSubfXY Œ:.P/g ) f:ŒSubfXY .P/g remark 2.20,
` f:Œ9X.P/g ) f:ŒSubfXY .P/g transitivity,
` ŒSubfY .P/ ) Œ9X.P/
X
converse contraposition & Detachment.
t
u
Theorem 2.49 provides a converse to theorem 2.48 if X is not free in P.
2.49 Theorem. For all P and X, if X is not free in P, then ` Œ9X.P/ , .P/.
Proof. Apply the propositional calculus with theorems 2.46 and 2.39:
` Œ:.P/
 ) f8XŒ:.P/g
 theorem 2.39, no free X in P,
` :f8XŒ:.P/g  ) f:Œ:.P/g
 contraposition,
` Œ9X.P/ ) :f8XŒ:.P/g theorem 2.46,
` Œ9X.P/ ) f:Œ:.P/g transitivity,
` f:Œ:.P/g ) .P/ double negation,
` Œ9X.P/ ) .P/ transitivity.
The converse is theorem 2.48. t
u

2.2.8 Other Axiomatic Systems for the Pure Predicate Calculus

With the rules of Detachment and Generalization, and equivalent propositional

calculi, Margaris [84, p. 49] and Rosser [110, p. 101] use the following axiom
schemata for the predicate calculus:

Axiom A4 f8XŒ.P/ ) .Q/g ) fŒ8X.P/ ) Œ8X.Q/g.

 www.pdfgrip.com

88 2 First-Order Logic: Proofs with Quantifiers

Axiom A5 Œ8X.P/ ) ŒSubfXY .P/.

Axiom A6 .P/ ) Œ8X.P/ if X does not occur freely in P.
Margaris allows Generalizations only of axioms but proves a deduction theorem
that then leads to the same rule of Generalization [84, p. 49].
In contrast, Kleene [72, p. 107] uses two axiom schemata
8-schema Œ8X.P/ ) ŒSubfXY .P/,
9-schema ŒSubfXY .P/ ) Œ9X.P/,
paired with two inference rules, where X does not occur freely in P:
8-rule from .P/ ) .Q/ infer .P/ ) Œ8X.Q/, where X is not free in P,
9-rule from .P/ ) .Q/ infer Œ9X.P/ ) .Q/, where X is not free in Q.
In both systems, 9X.P/ is merely an abbreviation for :f8XŒ:.P/g, as also in
the systems of Church [18, p. 171] and Stoll [122, p. 115]. Reversely, in other
systems 8X.P/ is merely an abbreviation for :f9XŒ:.P/g, for instance, in Kunen’s
[74, p. 3]. Axioms such as Q3 and Q4 partially dissociate the quantifiers from the
axiom(s) for the negation in the selected propositional calculus.
Also, yet another way to define substitutions of free variables consists in
performing substitutions with a different procedure, as outlined in definition 2.50.
2.50 Definition (proper substitution of free variables). A proper substitution
of a variable Z for each free occurrence of a different variable X in a logical formula
P consists of the following three steps:
(1) Identify a variable that does not occur in P, for example, Y.
(2) In P, replace each bound occurrence of Z by Y.
(3) Then replace each free occurrence of X by Z.
Steps (1) and (2) produce a change of bound variables according to definition 2.13.
After step (2), P no longer contains any bound occurrence of Z, and hence no strings
of the form 8Z or 9Z. Consequently, P now admits Z for X in step (3).
Definitions 2.16 and 2.50 thus provide two ways to avoid the phenomenon in
counterexample 2.10. Both ways lend themselves to the same notation.
2.51 Definition. The notation SubfXZ .P/ states that P admits Z for X, and substitutes
Z for each free occurrence of X in P. Alternatively, the same notation SubfXZ .P/
denotes the proper substitution of Z for each free occurrence of X in P. The two
alternatives are compatible, by remark 2.15 and definition 2.50. For convenience,
SubfXX .P/ is defined to be P, and if X does not occur freely in P, then SubfXZ .P/ is
also defined to be P. The concept and notation for proper substitutions also apply to
the substitution of a constant for a free variable. Because constants cannot appear
immediately after a quantifier, they are not bound. Consequently, only the last step
applies to the proper substitution of constants for free variables. Thus, SubfX¿ .P/
merely substitutes ¿ for every free occurrence of X in P.
 
2.52 Example. For P consider the formula 8Xf9ZŒ:.X D Z/g _ .X D ¿/.
 www.pdfgrip.com

2.2 The Pure Predicate Calculus of First Order 89

(1) Verify that the variable Y does not occur in P.

(2) Replace the bound occurrences of Z by Y, which gives the formula
.8Xf9YŒ:.X D Y/g/ _ .X D ¿/.
(3) Replace the free occurrence of X by Z, which gives the formula
.8Xf9YŒ:.X D Y/g/ _ .Z D ¿/ for SubfXZ .P/.
In  the constant ¿ for every free occurrence of X in P yields
 contrast, substituting
8Xf9ZŒ:.X D Z/g _ .¿ D ¿/ for SubfX¿ .P/.
2.53 Example. A situation like that in definition 2.50 occurs with computer algo-
rithms to swap two variables X and Z, which typically use a third variable Y distinct
from X and Z as a temporary storage. First, the algorithm assigns X to Y, an
operation denoted by Y WD X. Second, the algorithm assigns Z to X, an operation
denoted by X WD Z. Finally, the algorithm assigns Y to Z, an operation denoted by
Z WD Y.

2.2.9 Exercises on Kleene’s, Margaris’s, and Rosser’s Axioms

The following exercises show that Margaris’s and Rosser’s axioms A4–A6 are
derivable from the rules of inference with axioms Q1–Q4 and the Classical
Propositional Calculus.
2.11 . Prove that the abbreviation 9X.P/ for :f8XŒ:.P/g is derivable from the
rules of inference with axioms Q1–Q4 and the Classical Propositional Calculus.
2.12 . Prove that Margaris’s and Rosser’s axioms A4, A5, and A6 are theorems
derivable from the rules of inference with axioms Q1–Q4 and the Classical
Propositional Calculus.
The following exercises show that axioms Q1–Q4 are derivable from Margaris’s
and Rosser’s axioms A4–A6 and the Classical Propositional Calculus.
2.13 . Prove that axiom Q2 is a theorem derivable from the rules of inference with
Margaris’s and Rosser’s axioms A4–A6 and the Classical Propositional Calculus.
2.14 . Prove that axiom Q1 is a theorem derivable from the rules of inference with
Margaris’s and Rosser’s axioms A4–A6 and the Classical Propositional Calculus.
2.15 . Prove that axiom Q4 is a theorem derivable from the rules of inference with
Margaris’s and Rosser’s axioms A4–A6 and the Classical Propositional Calculus.
2.16 . Prove that axiom Q3 is a theorem derivable from the rules of inference with
Margaris’s and Rosser’s axioms A4–A6 and the Classical Propositional Calculus.
The following exercises show that Kleene’s schema and rules are derivable from
the rules of inference with axioms Q1–Q4 and the Classical Propositional Calculus.
 www.pdfgrip.com

90 2 First-Order Logic: Proofs with Quantifiers

2.17 . Prove that Kleene’s 9-rule is derivable from the rules of inference with
axioms Q1–Q4 and the Classical Propositional Calculus.
2.18 . Prove that Kleene’s 8-rule is derivable from the rules of inference with
axioms Q1–Q4 and the Classical Propositional Calculus.
2.19 . Prove that Kleene’s 9-schema is derivable from the rules of inference with
axioms Q1–Q4 and the Classical Propositional Calculus.
2.20 . Prove that Kleene’s 8-schema is derivable from the rules of inference with
axioms Q1–Q4 and the Classical Propositional Calculus.

2.3 Methods of Proof for the Pure Predicate Calculus

If other considerations guarantee that a well-formed formula P has a proof but do not
produce any proof of it, then writing down all the proofs of the predicate calculus,
for instance, in increasing order of complexity, eventually yields among all such
proofs a proof of P [18, p. 99–100, footnote 183]. However, if the shortest proof of
P is very long, then this method may take longer than the time available to the user
to arrive at any proof of P. Thus for all practical purposes this method may also fail
to determine whether a formula is a theorem.
The problem of deciding whether a well-formed formula is a theorem, derivable
from specified axioms and inference rules, is called the decision problem. For the
pure predicate calculus, no algorithms can provide a step-by-step recipe applicable
to all well-formed formulae to determine whether any such formula is a theorem, as
proved by Church [16, 17]. Nevertheless, methods exist to help in deciding whether
a well-formed formula is a theorem.
Trial and error is an option [114, p. 31], sometimes working backward from the
particular well-formed formula as a final goal, or forward from the axioms, inference
rules, and previous theorems as starting points or intermediate steps [72, p. 54–55].
The methods presented in this section guide this method of designing proofs.

2.3.1 Substituting Equivalent Formulae

One method of proof consists of replacing any occurrence of a formula by an

equivalent formula, thanks to theorem 2.54 [18, p. 101, 124, 189], [108, p. 48].
2.54 Theorem (Substitutivity of Equivalence in the Pure Predicate Calculus,
preliminary version). For all well-formed logical formulae U and V, if ` .U/ ,
.V/, and if a formula Q results from substituting any (zero, one, several, or all)
occurrence(s) of U by V in a well-formed formula P, then ` .P/ , .Q/.
 www.pdfgrip.com

2.3 Methods of Proof for the Pure Predicate Calculus 91

Proof (Outline). Theorems 1.29 and 1.46 have already established the conclusions
for logical implications and negations.
In all cases, if Q is P, which results by substituting none of the occurrences of U
by V, then .P/ , .Q/ is .P/ , .P/, which is theorem 1.63.
For the universal quantifier, if P is 8X.U/, then Q is either 8X.U/ or 8X.V/.
If Q is 8X.V/, then P with .U/ ) .V/ yield Q, by theorem 2.37, and conversely, Q
with .V/ ) .U/ yield P, by theorems 2.37, or also by theorem 2.45.
For the existential quantifier, theorem 2.46 reduces to the previous cases a
formula P of the form 9X.U/.
The general case follows by several applications of the previous case and the
cases in theorems 1.29 and 1.46, in a way that may be specified more explicitly
after the availability of the Principle of Mathematical Induction in chapter 4. t
u
 
2.55 Example. Let P denote the formula 8X Œ9Y.Y 2 X/ _ f8ZŒ:.Z 2 X/g , U
the formula 9Y.Y 2 X/, and W the formula :f8YŒ:.Y 2 X/g. Then .U/ , .W/
by theorem 2.46. Moreover, let V denote the formula :f8ZŒ:.Z 2 X/g. Because
Z does not occur in :.Y 2 X/, theorem 2.31 shows that .W/ , .V/. Hence
.U/, .V/ by transitivity. Consequently, .P/ , .Q/ whereQ denotes  the formula
8X :f8ZŒ:.Z 2 X/g _ f8ZŒ:.Z 2 X/g , which is 8X :fVg _ fVg . Since
.V/ _ Œ:.V/ is a theorem, by Generalization Q and hence P is also a theorem.

2.3.2 Discharging Hypotheses

A method to design a proof of an implication .H/ ) .C/ consists of first designing

a derivation H ` C of C from H, and then transforming the derivation H ` C into a
proof of .H/ ) .C/ by theorem 2.56 [108, p. 46–47].
2.56 Theorem (Deduction Theorem, preliminary version). With any axiom sys-
tem for which axioms P1, P2, and .P/ ) .P/ are axioms or theorems (or schema
thereof), and for every derivation H ` C of a formula C from a formula H with
the propositional calculus, the rules of inference, and axioms Q1–Q4, but without
Generalization on any free variable in H, there exists a proof of .H/ ) .C/.
Proof (Outline). Every step of the derivation H ` C is a formula, denoted here
by S. If S is H, or an axiom, or results from two previous steps and Detachment,
then the proof of the deduction theorem 1.22 for the Pure Positive Implicational
Propositional Calculus shows how to replace S by .H/ ) .S/.
If S is a Generalization 8X.R/ of a previous step R with a variable X that does
not occur freely in H, then R has already been replaced by .H/ ) .R/. Hence
Generalization gives 8XŒ.H/ ) .R/, whence, because X does not occur freely in
H, theorem 2.29 yields a proof of .H/ ) Œ8X.R/, which is .H/ ) .S/.
The general case follows by several applications of the previous cases in a
way that may be specified more explicitly after the availability of the Principle of
Mathematical Induction in chapter 4. t
u
 www.pdfgrip.com

92 2 First-Order Logic: Proofs with Quantifiers

Thus, the selection of axioms P1–P3 and Q1–Q4 leads to the Deduction
Theorem (2.56) more directly than would other selections of otherwise equivalent
axioms [108, p. 47]. In practice, however, a derivation of H ` C of C from H
may already suggest other logical steps that shortcut or bypass the entire procedure
outlined in the proof of the Deduction Theorem 2.56.
To demonstrate such shortcuts, the following theorems provide means for
bringing quantifiers to the “front” of a formula. For example, axioms Q3 and Q4
with theorem 2.54 already allow the replacement of :Œ9X.P/ by 8XŒ:.P/, and of
:Œ8X.P/ by 9XŒ:.P/. In an implication .R/ ) .S/, each of R and S can be of the
form .P/, or 8X.P/, or 9X.P/, starting with 8, or 9, or no quantifiers, which leads
to nine different cases. In the case where neither R nor S begins with a quantifier,
then no quantifiers need to be brought to the front of .R/ ) .S/. The other eight
cases form the object of the following theorems.
Theorem 2.57 handles a case where R is 9X.P/ while S is Q.
2.57 Theorem. If X does not occur freely in Q, then
` f8XŒ.P/ ) .Q/g , fŒ9X.P/ ) .Q/g.
Proof. Let H denote 8XŒ.P/ ) .Q/, and let C denote Œ9X.P/ ) .Q/.
` 8XŒ.P/ ) .Q/ hypothesis,
` f8XŒ.P/ ) .Q/g ) Œ.P/ ) .Q/ specialization (Q1),
` .P/ ) .Q/ Detachment,
` Œ:.Q/ ) Œ:.P/ contraposition,
` 8XfŒ:.Q/ ) Œ:.P/g Generalization,
` Œ:.Q/ ) f8XŒ:.P/g
 theorem 2.29, no free X in Q,
` :f8XŒ:.P/g ) f:Œ:.Q/g contraposition,
` Œ9X.P/ ) .Q/ double negation (1.42) and 2.46.
Hence the Deduction Theorem (2.56) leads to a proof of f8XŒ.P/ ) .Q/g )
fŒ9X.P/ ) .Q/g. Yet the foregoing derivation suggests shortcuts:
` f8XŒ.P/).Q/g)Œ.P/).Q/ axiom Q1,
` Œ.P/).Q/)fŒ:.Q/)Œ:.P/g contraposition,
` f8XŒ.P/).Q/g)fŒ:.Q/)Œ:.P/g
  transitivity,
` f8XŒ.P/).Q/g) 8XfŒ:.Q/)Œ:.P/g
   theorem 2.30,
` 8XfŒ:.Q/)Œ:.P/g)Œ:.Q/)f8XŒ:.P/g  theorem 2.29,
` Œ:.Q/)f8XŒ:.P/g
 ) :f8XŒ:.P/g )f:Œ:.Q/g contraposition,
` :f8XŒ:.P/g )f:Œ:.Q/g)fŒ9X.P/).Q/g 1.42, 2.46,
` f8XŒ.P/).Q/g)fŒ9X.P/).Q/g transitivity.
For the converse, let H denote Œ9X.P/ ) .Q/, and let C denote 8XŒ.P/ ) .Q/.
` Œ9X.P/ ) .Q/ hypothesis,
` .P/ ) Œ9X.P/ theorem 2.48,
` .P/ ) .Q/ transitivity,
` 8XŒ.P/ ) .Q/ Generalization.
 www.pdfgrip.com

2.3 Methods of Proof for the Pure Predicate Calculus 93

Again the Deduction Theorem (2.56) leads to a proof of fŒ9X.P/ ) .Q/g )

f8XŒ.P/ ) .Q/g but the foregoing derivation suggests shortcuts:
` .P/
 ) Œ9X.P/  theorem 2.48,
` fŒ9X.P/ ) .Q/g ) Œ.P/ ) .Q/ theorem 1.34,
` fŒ9X.P/ ) .Q/g ) f8XŒ.P/ ) .Q/g theorem 2.30.
t
u
The proofs of the following theorems emerge from similar outlines, starting with
a derivation H ` C of C from H, and transforming it into a proof of ` .H/ ) .C/
by shortcuts suggested by the derivation or by the Deduction Theorem (2.56). To
this end, the following derived rule proves useful.
2.58 Theorem (derived rule). If X is not free in Q, and if ` .P/ ) .Q/, then
` Œ9X.P/ ) .Q/.
Proof. Apply theorem 2.57:
` .P/ ) .Q/ hypothesis,
` 8XŒ.P/ ) .Q/ Generalization,
` f8XŒ.P/ ) .Q/g ) fŒ9X.P/ ) .Q/g theorem 2.57, no free X in Q,
` Œ9X.P/ ) .Q/ Detachment.
t
u
2.59 Theorem. If X does not occur freely in P, then ` .P/ , Œ9X.P/.
Proof. Apply theorems 2.48, 1.12, and 2.58.
` .P/ ) Œ9X.P/ theorem 2.48,
` .P/ ) .P/ theorem 1.12,
` Œ9X.P/ ) .P/ theorem 2.58, no free X in P.
t
u

2.60 Theorem. For all P, Q, and X, ` f8XŒ.P/ ) .Q/g ) fŒ9X.P/ )

Œ9X.Q/g.
Proof. Let H denote 8XŒ.P/ ) .Q/, and let C denote Œ9X.P/ ) Œ9X.Q/.
` f8XŒ.P/ ) .Q/g ) Œ.P/ ) .Q/ axiom Q1,
` Œ.P/ ) .Q/ ) fŒ:.Q/ ) Œ:.P/g contraposition,
` f8XŒ.P/ ) .Q/g ) fŒ:.Q/ ) Œ:.P/g  transitivity,
` f8XŒ.P/ ) .Q/g ) 8XfŒ:.Q/
 ) Œ:.P/g theorem 2.30,
` 8XfŒ:.Q/
 ) Œ:.P/g 
) f8XŒ:.Q/g ) f8XŒ:.P/g theorem 2.35
` f8XŒ:.Q/g
 ) f8XŒ:.P/g
  
) :f8XŒ:.P/g
  ) :f8XŒ:.Q/g
 contraposition,
` :f8XŒ:.P/g ) :f8XŒ:.Q/g
) fŒ9X.P/ ) Œ9X.Q/g 1.42, 2.46,
` f8XŒ.P/ ) .Q/g ) fŒ9X.P/ ) Œ9X.Q/g transitivity.
t
u
2.61 Theorem (derived rule). If .P/ , .Q/ is a theorem, then Œ9X.P/ ,
Œ9X.Q/ is a theorem.
 www.pdfgrip.com

94 2 First-Order Logic: Proofs with Quantifiers

Proof. Apply Generalization, theorem 2.60, and Detachment:

` .P/ , .Q/ hypothesis,
` 8XŒ.P/ , .Q/ Generalization,
` f8XŒ.P/ , .Q/g ) fŒ9X.P/ ) Œ9X.Q/g theorem 2.60,
` Œ9X.P/ ) Œ9X.Q/ Detachment.
t
u
2.62 Theorem. ` f9XŒ.P/ ) .Q/g , fŒ8X.P/ ) Œ9X.Q/g.
Proof. This proof follows Church’s [18, p. 205]. Apply the law of denial of the
antecedent (theorem 1.40) and the law of proof by cases subject to hypotheses
(theorem 1.50):
` Œ:.P/ ) Œ.P/ ) .Q/ theorem 1.40,
` 8XfŒ:.P/ ) Œ.P/ ) .Q/g Generalization,
` f9XŒ:.P/g ) f9XŒ.P/ ) .Q/g theorem 2.60 and Detachment,
` f:Œ8X.P/g ) f9XŒ.P/ ) .Q/g axiom Q4 and Detachment,
` .Q/ ) Œ.P/ ) .Q/ axiom P1,
` 8Xf.Q/ ) Œ.P/ ) .Q/g Generalization,
` Œ9X.Q/ ) f9XŒ.P/ ) .Q/g theorem 2.60 and Detachment,
` fŒ8X.P/ ) Œ9X.Q/g ) f9XŒ.P/ ) .Q/g theorem 1.50 and Detachment.
For the converse, use theorems 2.48 and 2.57:
` .P/ ) fŒ.P/ ) .Q/ ) .Q/g law of assertion (1.38),
` Œ8X.P/ ) .P/ axiom Q1,
` Œ8X.P/ ) fŒ.P/ ) .Q/ ) .Q/g transitivity,
` Œ.P/ ) .Q/ ) fŒ8X.P/ ) .Q/g commutation (1.37),
` .Q/ ) Œ9X.Q/ theorem 2.48,
` Œ.P/ ) .Q/ ) fŒ8X.P/ ) Œ9X.Q/g  derived rule,
` 8X Œ.P/ ) .Q/ ) fŒ8X.P/ ) Œ9X.Q/g Generalization,
` f9XŒ.P/ ) .Q/g ) fŒ8X.P/ ) Œ9X.Q/g theorem 2.57 and Detachment.
t
u
2.63 Theorem. If X does not occur freely in P, then
` f9XŒ.P/ ) .Q/g , f.P/ ) Œ9X.Q/g.
Proof. This proof follows Church’s [18, p. 205]. Apply theorems 2.39 and 2.62:
` f9XŒ.P/ ) .Q/g , fŒ8X.P/ ) Œ9X.Q/g theorem 2.62,
` .P/ , Œ8X.P/ theorem 2.39, no free X in P,
` f9XŒ.P/ ) .Q/g , f.P/ ) Œ9X.Q/g derived rules.
t
u
Similar to theorem 2.57, theorem 2.64 handles a case where R is 8X.P/ while
S is Q.
2.64 Theorem. If X does not occur freely in Q, then
` f8XŒ.P/ ) .Q/g , fŒ8X.P/ ) .Q/g.
 www.pdfgrip.com

2.3 Methods of Proof for the Pure Predicate Calculus 95

Proof. Apply theorems 2.59, 2.62, and 2.54:

` .Q/ , Œ9X.Q/ theorem 2.59, no free X in Q,
` f9XŒ.P/ ) .Q/g , fŒ8X.P/ ) Œ9X.Q/g theorem 2.62,
` f8XŒ.P/ ) .Q/g , fŒ8X.P/ ) .Q/g theorem 2.54.
t
u

2.3.3 Prenex Normal Form

Yet another method of proof consists in transforming a formula into an equivalent

formula in which all the quantifiers, if any, are at the beginning.
2.65 Definition. A formula Q is in prenex normal form if and only if Q is of the
form

Q[ X[ Q[[ X[[ : : : Q[:::[ X[:::[ .R/

optionally with brackets and parentheses, where R is a well-formed formula without

quantifiers and each string Q[:::[ is either 8 or 9. The formula R is called the matrix
of P while the string Q[ X[ Q[[ X[[ : : : Q[:::[ X[:::[ is called the prefix of P.
2.66 Example. The formula 8X9ZŒ:.Z D X/ is in prenex normal form. Its prefix
is 8X9Z while its matrix is :.Z D X/.
Theorem 2.67 reveals that every well-formed formula is equivalent to a formula
in prenex normal form [18, 39], [108, p. 49].
2.67 Theorem (prenex normal form, preliminary version). For every well-
formed formula P there exists a well-formed formula Q in prenex normal form such
that .P/ , .Q/.
Proof (Outline). In P, replace bound variables so that different quantifiers bind
different variables, which gives a formula equivalent to P, by theorems 2.31
and 2.54.
With different quantified variables, theorem 2.38 then provides a means to bring
quantifiers in front of an implication.
Axioms Q3 and Q4 also provide a means to bring any quantifier in front of any
negation.
The general case follows by several applications of the previous cases in a
way that may be specified more explicitly after the availability of the Principle of
Mathematical Induction in chapter 4. t
u
2.68 Example. The formula :f9XŒ8Z.Z D X/g is not in prenex normal form.
Nevertheless, axiom Q4 gives the equivalent formula 8Xf:Œ8Z.Z D X/g, whence
axiom Q3 and theorem 2.54 yield the equivalent formula 8Xf9ZŒ:.Z D X/g, which
is in prenex normal form.
 www.pdfgrip.com

96 2 First-Order Logic: Proofs with Quantifiers

Transforming a logical formula P into an equivalent formula Q in prenex normal

form, or partly so, may reveal a proof of Q and hence also of P, as demonstrated in
example 2.55. Example 2.69 completes the transformation into prenex normal form,
which reveals a propositional theorem in the matrix.
 
2.69 Example. The formula 8X Œ9Y.Y 2 X/ _ f8ZŒ:.Z 2 X/g is not in prenex
normal form. Nevertheless, by the definition of _ in
 terms of : and ) the formula
becomes 8X f:Œ9Y.Y 2  X/g ) f8ZŒ:.Z 2 X/g . 
Axiom Q3 gives 8X f8YŒ:.Y
  2 X/g ) f8ZŒ:.Z 2 X/g . 
Theorem 2.38 gives 8X 8Z  f8YŒ:.Y
 2 X/g ) Œ:.Z 2 X/ . 
Theorem 2.64 yields 8X 8Z 9YfŒ:.Y 2 X/ ) Œ:.Z 2 X/g , which is a
theorem: selecting Z for Y gives Œ:.Z 2 X/ ) Œ:.Z 2 X/, which has the pattern
.P/ ) .P/ of theorem 1.12.
Besides providing transformations that may facilitate proofs, as in example 2.55,
bringing formulae into prenex normal form, in particular, Skolem’s normal form
with all the existential quantifiers preceding all the universal quantifiers,

9X[ : : : 9X[:::[ 8Y ] : : : 8Y ]:::] .R/;

leads to Gödel’s Completeness Theorem, that a formula is a theorem if and only if

it is valid in all applications, even though no mechanical ways to check either may
exist [18, 42–44].

2.3.4 Proofs with More than One Quantifier

The following theorems are examples of theorems involving more than one quanti-
fier. The first theorem allows for the deletion of a redundant universal quantifier.
2.70 Theorem. ` Œ8X.Q/ , f8XŒ8X.Q/g .
Proof. Apply theorem 2.39 to 8X.Q/, which has no free X. t
u
The second theorem allows for the swap of two consecutive universal quantifiers.
2.71 Theorem. ` f8XŒ8Y.P/g , f8YŒ8X.P/g:
Proof. Apply axiom Q1, theorem 2.29, and Generalization:
` f8XŒ8Y.P/g ) Œ8Y.P/ axiom Q1,
` Œ8Y.P/ ) .P/ axiom Q1,
` f8XŒ8Y.P/g ) .P/ transitivity (theorem 1.16),
` f8XŒ8Y.P/g ) Œ8X.P/ theorem 2.37,
` f8XŒ8Y.P/g ) f8YŒ8X.P/g theorem 2.29.
t
u
The third theorem allows for the swap of two consecutive existential quantifiers.
 www.pdfgrip.com

2.3 Methods of Proof for the Pure Predicate Calculus 97

2.72 Theorem. ` f9XŒ9Y.P/g , f9YŒ9X.P/g:

Proof. Apply the complete law of double negation, axiom Q3, and theorem 2.71:
9XŒ9Y.P/
 m double negation,
: :f9XŒ9Y.P/g
 m axiom Q3 and theorem 2.54,
: 8Xf:Œ9Y.P/g
 m axiom Q3 and theorem 2.54,
: 8Xf8YŒ:.P/g
 m theorem 2.71,
: 8Yf8XŒ:.P/g
 m axiom Q3 and theorem 2.54,
: 8Yf:Œ9X.P/g
 m axiom Q3 and theorem 2.54,
: :f9YŒ9X.P/g
m double negation (theorems 1.41 and 1.42).
9YŒ9X.P/
t
u
The fourth theorem allows for the swap of different quantifiers in an implication.
2.73 Theorem. ` f9XŒ8Y.P/g ) f8YŒ9X.P/g.
Proof. Apply theorems 2.48, 2.35, and 2.57:
` .P/ ) Œ9X.P/ theorem 2.48,
` Œ8Y.P/ ) f8YŒ9X.P/g theorem 2.37,
` f9XŒ8Y.P/g ) f8YŒ9X.P/g theorem 2.57, no free X in f8YŒ9X.P/g.
t
u
2.74 Counterexample. The converse of theorem 2.73, which would be

f8YŒ9X.P/g ) f9XŒ8Y.P/g;

can be False. For instance, in every context with at least two different objects V and
W, consider the logical formula X D Y for P.
` 8YŒ9X.X D Y/ for each Y, choose X WD Y;
9XŒ8Y.X D Y/ is False: no X equals V and W;
f8YŒ9X.X D Y/g ) f9XŒ8Y.X D Y/g is False because .T/ 6) .F/.

2.3.5 Exercises on the Substitutivity of Equivalence

The following exercises focus on details of the proof of theorem 2.54, with the
logical equivalence , defined either by Tarski’s axioms IV, V, VI in example 1.87
on page 55, or with ) and ^ in definition 1.51 on page 38.
 www.pdfgrip.com

98 2 First-Order Logic: Proofs with Quantifiers

2.21 . Prove that if P denotes .U/ ) .W/, if Q denotes .V/ ) .W/, and if `
.V/ , .U/, then ` .P/ ) .Q/.
2.22 . Prove that if P denotes .W/ ) .U/, if Q denotes .W/ ) .V/, and if `
.V/ , .U/, then ` .P/ ) .Q/.
2.23 . Prove that if P denotes .U/ ) .W/, if Q denotes .V/ ) .W/, and if `
.V/ , .U/, then ` .Q/ ) .P/.
2.24 . Prove that if P denotes .W/ ) .U/, if Q denotes .W/ ) .V/, and if `
.V/ , .U/, then ` .Q/ ) .P/.
2.25 . Prove that if P denotes 8X.U/, if Q denotes 8X.V/, without free occurrences
of X in U and V, and if ` .V/ , .U/, then ` .P/ ) .Q/.
2.26 . Prove that if P denotes 8X.U/, if Q denotes 8X.V/, without free occurrences
of X in U and V, and if ` .V/ , .U/, then ` .Q/ ) .P/.
2.27 . Prove that if P denotes :.U/, if Q denotes :.V/, and if ` .V/ , .U/, then
` .P/ ) .Q/.
2.28 . Prove that if P denotes :.U/, if Q denotes :.V/, and if ` .V/ , .U/, then
` .Q/ ) .P/.
2.29 . Prove that if P denotes either .U/ ) .W/ or .W/ ) .U/, if Q denotes either
.V/ ) .W/ or .W/ ) .V/, respectively, and if ` .V/ , .U/, then ` .P/ , .Q/.
2.30 . Prove that if P denotes :.U/, if Q denotes :.V/, and if ` .V/ , .U/, then
` .P/ , .Q/.

2.4 Predicate Calculus with Other Connectives

This section introduces theorems with quantifiers and conjunctions or disjunctions.

2.4.1 Universal Quantifiers and Conjunctions or Disjunctions

This subsection presents theorems involving the universal quantifier .8/ and a con-
junction .^/ or disjunction ._/, beginning with an equivalence with a conjunction.
2.75 Theorem. ` f8XŒ.P/ ^ .Q/g ) fŒ8X.P/ ^ Œ8X.Q/g.
Proof. Apply theorems 1.53, 2.37, 1.52, 1.55:
` Œ.P/ ^ .Q/ ) .P/ theorem 1.53,
` f8XŒ.P/ ^ .Q/g ) f8X.P/g theorem 2.37,
` Œ.P/ ^ .Q/ ) .Q/ theorem 1.52,
 www.pdfgrip.com

2.4 Predicate Calculus with Other Connectives 99

` f8XŒ.P/ ^ .Q/g ) f8X.Q/g theorem 2.37,

` f8XŒ.P/ ^ .Q/g ) fŒ8X.P/ ^ Œ8X.Q/ theorem 1.55.
t
u
The converse implication forms the object of the following theorem.
2.76 Theorem. ` fŒ8X.P/ ^ Œ8X.Q/g ) f8XŒ.P/ ^ .Q/g.
Proof. Apply axiom Q1 with theorems 1.82 and 2.30:
` Œ8X.P/ ) .P/ axiom Q1,
` Œ8X.Q/ ) .Q/ axiom Q1,
` fŒ8X.P/ ^ Œ8X.Q/g ) Œ.P/ ^ .Q/ theorem 1.82,
` fŒ8X.P/ ^ Œ8X.Q/g ) f8XŒ.P/ ^ .Q/g theorem 2.30.
t
u
The following theorem gives an implication with a disjunction.
2.77 Theorem. ` fŒ8X.P/ _ Œ8X.Q/g ) f8XŒ.P/ _ .Q/g.
Proof. Apply axiom Q1, exercise 1.57, theorem 2.29, and Generalization:
` Œ8X.P/ ) .P/ axiom Q1,
` Œ8X.Q/ ) .Q/ axiom Q1,
` fŒ8X.P/
 _ Œ8X.Q/g ) Œ.P/ _ .Q/  exercise 1.57,
` 8X fŒ8X.P/ _ Œ8X.Q/g ) Œ.P/ _ .Q/ Generalization,
` fŒ8X.P/ _ Œ8X.Q/g ) f8XŒ.P/ _ .Q/g theorem 2.29.
t
u
2.78 Counterexample. The converse of theorem 2.77, which would be

f8XŒ.P/ _ .Q/g ) fŒ8X.P/ _ Œ8X.Q/g;

may be False. For instance, in every context with exactly two different objects V and
W, consider the formulae X D V for P and X D W for Q:
` 8XŒ.X D V/ _ .X D W/ because either .X D V/ or .X D W/;
8X.X D V/ is False if X WD W;
8X.X D W/ is False if X WD V;
Œ8X.X D V/ _ Œ8X.X D W/ is False by the preceding two lines;
f8XŒ.X D V/ _ .X D W/g
) fŒ8X.X D V/ _ Œ8X.X D W/g is False because .T/ 6) .F/.
However, theorem 2.79 shows a converse of theorem 2.77 in a particular case.
2.79 Theorem. If P has no free X, then ` f8XŒ.P/ _ .Q/g ) f.P/ _ Œ8X.Q/g and
` f8XŒ.P/ _ .Q/g ) fŒ8X.P/ _ Œ8X.Q/g
 www.pdfgrip.com

100 2 First-Order Logic: Proofs with Quantifiers

Proof. Apply the definition of _:

8XŒ.P/ _ .Q/
m definition of .P/ _ .Q/,
8XfŒ.P/ ) .Q/ ) .Q/g
+ theorem 2.35,
f8XŒ.P/ ) .Q/g ) Œ8X.Q/
m theorems 2.38, 2.54, no free X in P,
f.P/ ) Œ8X.Q/g ) Œ8X.Q/
m definition of _,
.P/ _ Œ8X.Q/
m theorem 2.39, 2.54, no free X in P.
Œ8X.P/ _ Œ8X.Q/
t
u

2.4.2 Existential Quantifiers and Conjunctions or Disjunctions

This subsection presents theorems involving the existential quantifier .9/ and a con-
junction .^/ or disjunction ._/, beginning with an equivalence with a disjunction.
2.80 Theorem. ` fŒ9X.P/ _ Œ9X.Q/g , fŒ9XŒ.P/ _ .Q/g:
Proof. Apply
 contraposition with
 theorems
 2.75, 2.76, and 2.45:

` 8XfŒ:.P/ ^ Œ:.Q/g , f8XŒ:.P/g ^ f8XŒ:.Q/g 2.75, 2.76,
m contraposition,
     
: f8XŒ:.P/g ^ f8XŒ:.Q/g , : 8XfŒ:.P/ ^ Œ:.Q/g
m 2.45,
      
:f8XŒ:.P/g _ :f8XŒ:.Q/g , : 8Xf:Œ.P/ _ .Q/g
m axiom Q4,
      
:f:Œ9X.P/g _ :f:Œ9X.Q/g , : :f9XŒ.P/ _ .Q/g
m double negations.
fŒ9X.P/ _ Œ9X.Q/g , f9XŒ.P/ _ .Q/g
t
u
A similar equivalence with a conjunction requires that X be not free in P.
2.81 Theorem. If P has no free X, then ` f9XŒ.P/ ^ .Q/g , f.P/ ^ Œ9X.Q/g.
Proof. Apply the full propositional calculus with theorems 2.38, 2.61, 2.46, and
axiom Q4, and theorem 1.69:
9XŒ.P/ ^ .Q/
 m double negation,
9X :f:Œ.P/ ^ .Q/g
m De Morgan’s first law and theorem 2.61,
 www.pdfgrip.com

2.5 Equality-Predicates 101

 
9X :fŒ:.P/ _ Œ:.Q/g
 m definition of _,
9X :f.P/ ) Œ:.Q/g
 m axiom Q4,
: 8Xf.P/ ) Œ:.Q/g
 m theorem 2.38,
: .P/ ) f8XŒ:.Q/g
 m definition of _ by theorem 1.69,
: Œ:.P/ _ f8XŒ:.Q/g
 m De Morgan’s second law and double negation,
.P/ ^ :f8XŒ:.Q/g
m theorem 2.46.
.P/ ^ Œ9X.Q/
t
u

2.4.3 Exercises on Quantifiers with Other Connectives

For the following exercises, prove that the stated formulae are theorem schema.
2.31 . f9XŒ.P/ _ .Q/g , f9XŒ.Q/ _ .P/g.
2.32 . f8XŒ.P/ ^ .P/g , f8X.P/g.
2.33 . f9XŒ.P/ _ .P/g , f9X.P/g.
2.34 . .9XfŒ.P/ _ .Q/ _ .R/g/ , .9Xf.P/ _ Œ.Q/ _ .R/g/.
2.35 . .8XfŒ.P/ ^ .Q/ _ .R/g/ , .f8XŒ.P/ _ .R/g ^ f8XŒ.Q/ _ .R/g/.
2.36 . .9XfŒ.P/ _ .Q/ ^ .R/g/ , .f9XŒ.P/ ^ .R/g _ f9Œ.Q/ ^ .R/g/.
2.37 . Œ9X.Q/ , Œ9X .9X.Q//.
2.38 . If P has no free X, then f.P/ ^ Œ8X.Q/g , f8XŒ.P/ ^ .Q/g.
2.39 . If P has no free X, then f.P/ _ Œ8X.Q/g , f8XŒ.P/ _ .Q/g.
2.40 . If P has no free X, then f.P/ _ Œ9X.Q/g , f9XŒ.P/ _ .Q/g.

2.5 Equality-Predicates

Applications of logic, for instance, algebra, arithmetic, and geometry, may include
concepts of “equality” that allow for substitutions of mutually equal objects in
statements and formulae, which results in mutually equivalent statements and
formulae.
 www.pdfgrip.com

102 2 First-Order Logic: Proofs with Quantifiers

2.5.1 First-Order Predicate Calculi with an Equality-Predicate

Different applications may define equality differently [8, p. 6–7]. For instance, in
some versions of integer arithmetic, the equality a D b means that a and b are
two symbols for one integer [25, p. 44], [76, p. 1]. In contrast, in some versions
of set theory, the equality A D B means that A and B denote sets with identical
set-theoretical features: they have the same elements, and they are elements of the
same sets [8, 35, p. 6–7]; the question whether A and B denote the same set does
not arise in the theory. Nevertheless, such different concepts of equality happen to
conform to a logical predicate, denoted by I to suggest identity, subject to the
following axioms (which might also be called postulates to distinguish them from
logical axioms) [18,  48].

Axiom I 1 (reflexivity of equality) ` I .X; X/.

Axiom I 2 (substitutivity of equality) ` ŒI .X; Y/ ) Œ.P/ ) .Q/ for all well-
formed formula P and Q such that Q results from the substitution of any one free
occurrence of X in P by Y, provided that the resulting occurrence of Y is also free,
or, in other words, provided that in P this occurrence of X is not within the scope of
a quantifier .8X; 8Y; 9X; 9Y/ bounding X or Y.

The condition stipulated in axiom I 2 is similar to the requirement that P admit

Y for X, or that Y be free for X, but only for one particular occurrence of X in P.
Using only the Pure Positive Implicational Propositional Calculus, theorems 2.82
and 2.83 show that every predicate I satisfying axioms I 1 and I 2 is symmetric
and transitive [84, p. 104].
2.82 Theorem (symmetry of equality). ` ŒI .X; Y/ ) ŒI .Y; X/.
Proof. In axiom I 2, substitute the terms I .X; X/ for P and I .Y; X/ for Q:
` ŒI .X; Y/ ) fŒI .X; X/ ) ŒI .Y; X/g axiom I 2, ,
` I .X; X/ axiom I 1,
` ŒI .X; Y/ ) ŒI .Y; X/ derived rule (theorem 1.15).
t
u
2.83 Theorem (transitivity of equality). ` ŒI .X; Y/ ) fŒI .Y; Z/ )
ŒI .X; Z/g. Hence, if ` I .X; Y/ and ` I .Y; Z/, then ` I .X; Z/.
Proof. Use the symmetry of equality (theorem 2.82) and axiom I 2:
` ŒI .X; Y/ ) ŒI .Y; X/ theorem 2.82,
` ŒI .Y; X/ ) fŒI .Y; Z/ ) ŒI .X; Z/g axiom I 2,
` ŒI .X; Y/ ) fŒI .Y; Z/ ) ŒI .X; Z/g transitivity (theorem 1.16),
` I .X; Y/ hypothesis,
` ŒI .Y; Z/ ) ŒI .X; Z/ Detachment,
` I .Y; Z/ hypothesis,
` I .X; Z/ Detachment.
t
u
 www.pdfgrip.com

2.5 Equality-Predicates 103

Using only the Pure Positive Implicational Propositional Calculus, theorem 2.84
extends axiom I 2 to a converse implication, so that substituting mutually equal
objects results in mutually equivalent formulae.
2.84 Theorem (substitutivity of equality). ` ŒI .X; Y/ ) Œ.P/ , .Q/ for all
well-formed formula P and Q such that Q results from the substitution of any one
free occurrence of X in P provided that in P this occurrence of X is not within the
scope of a quantifier .8X; 8Y; 9X; 9Y/ bounding X or Y.
Proof. The implication ` ŒI .X; Y/ ) Œ.P/ ) .Q/ is axiom I 2.
For the converse, the hypothesis also states that in Q the resulting occurrence
of Y is not within the scope of a quantifier .8X; 8Y; 9X; 9Y/ bounding Y or X,
which allows swapping X and Y, and swapping P and Q, in axiom I 2, so that
` ŒI .Y; X/ ) Œ.Q/ ) .P/. Hence the conclusion follows from the symmetry
` ŒI .X; Y/ ) ŒI .Y; X/ by theorem 2.82 and the transitivity of implication. ut
Repeated applications of theorem 2.84 and the proof of substitutivity of equiv-
alence then show that substituting mutually equal objects in a formula leads to an
equivalent formula.

2.5.2 Simple Applied Predicate Calculi

with an Equality-Predicate

Some applications of logic might omit all propositional variables and instead have
only atomic formulae with a few predicates, or perhaps only one predicate, which
might be denoted by some constant E . Such applications are called simple applied
predicate calculi. For instance, a version of set theory has no propositional variables
and only one predicate, for set membership, so that E .X; Y/ stands for X 2 Y.
In such applications, an additional equality predicate I allows for substitutions
of mutually equivalent objects in statements and formulae if and only if I is
reflexive (a condition that replaces axiom I 1), symmetric, transitive, and satisfies
the following two conditions, which replace axiom I 2 [18, p. 283, exercise 48.3]:

ŒI .A; B/ ) fŒE .X; A/ ) ŒE .X; B/g;

ŒI .A; B/ ) fŒE .A; Y/ ) ŒE .B; Y/g:

In applied logics with other predicates, for instance, predicates for the sum and
products of integers in arithmetic, two similar conditions must be appended for each
predicate to ensure the substitutivity of mutually equal objects. By the postulated
symmetry of the equality predicate I these two conditions are equivalent to

ŒI .A; B/ ) fŒE .X; A/ , ŒE .X; B/g;

ŒI .A; B/ ) fŒE .A; Y/ , ŒE .B; Y/g:
 www.pdfgrip.com

104 2 First-Order Logic: Proofs with Quantifiers

These conditions suffice to ensure that if I .A; B/ holds, then substituting any free
occurrence of A for any free occurrence of B according to the conditions stipulated
by axiom I 2 in any formula P produces an equivalent formula Q, because well-
formed formulae include only atomic formulae of the form E .Z; W/. The proof
follows the pattern of the proof of the substitutivity of equivalence. The resulting
theorem is called the substitutivity of equality.
In particular, if a simple applied predicate calculus has exactly one predicate,
E , which is binary (involving exactly two individual variables), then the same
conditions may serve to define an equality predicate I so that I .A; B/ is merely
an abbreviation for
   
8XfŒE .X; A/ , ŒE .X; B/g ^ 8YfŒE .A; Y/ , ŒE .B; Y/g : (2.1)

Formula (2.1), abbreviated by I .A; B/, satisfies axiom I 2 and is reflexive,

symmetric, and transitive, because so is the equivalence , in the full propositional
calculus. In particular, the theorem on substitutivity of equality holds. An equality
predicate I defined in this manner from another binary predicate E thus does not
add anything to the theory except convenience.
2.85 Example (equality in set theory). In a version of set theory, with the single
binary predicate E .Z; W/, the postulate (or axiom) of extensionality states that
   
8XfŒE .X; A/ , ŒE .X; B/g , 8YfŒE .A; Y/ , ŒE .B; Y/g : (2.2)

The equality I .A; B/ of sets A and B is then an abbreviation of each of the formulae
8XfŒE .X; A/ , ŒE .X; B/g and 8YfŒE .A; Y/ , ŒE .B; Y/g.
The following theorems confirm that every equality-predicate defined by for-
mula (2.1) is reflexive, symmetric, and transitive.
2.86 Theorem (reflexivity of defined equality-predicates). Every equality-
predicate I .A; B/ defined by formula (2.1) is reflexive:

` 8CŒI .C; C/:

Proof. One method to design a formal proof transforms the objective, here the
yet unproved formula 8CŒI .C; C/, first into its defining formula (2.1), and then
into logically equivalent formulae, for instance, in prenex form, until one such
equivalent formula appears that is a theorem, thanks to an axiom or to a previously
proven theorem. For instance, substituting C for A and also C for B in the defining
formula (2.1) gives
I .C; C/ yet unproved,
m definition of I
   
8XfŒE .X; C/ , ŒE .X; C/g ^ 8YfŒE .C; Y/ , ŒE .C; Y/g
m theorem 2.75,
 
8X8Y fŒE .X; C/ , ŒE .X; C/g ^ fŒE .C; Y/ , ŒE .C; Y/g
 www.pdfgrip.com

2.5 Equality-Predicates 105

where each logical formula ŒE .W; Z/ , ŒE .W; Z/ has the pattern of the reflexivity
of the logical implication .P/ , .P/ (theorem 1.63). Thus, a complete proof may
proceed as follows:
` .P/,.P/ theorem 1.63,
` ŒE .X; C/,ŒE .X; C/ substitution in .P/,.P/,
` ŒE .C; Y/,ŒE .C; Y/ substitution in .P/,.P/,
` fŒE .X; C/,ŒE .X; C/g ^ fŒE .C; Y/,ŒE .C; Y/g theorem 1.54,
` I .C; C/ formula (2.1).
Hence ` 8CŒI .C; C/ results by Generalization and theorem 2.75. t
u
2.87 Theorem (symmetry of defined equality-predicates). Every equality-
predicate defined by formula (2.1) is symmetric: if ` I .A; B/, then ` I .B; A/;
moreover.

` 8A8BfŒI .A; B/ ) ŒI .B; A/g:

Proof. One method to design a formal proof transforms the objective, here the
yet unproved formula ` 8A8BfŒI .A; B/ ) ŒI .B; A/g, first into its defining
formula (2.1), and then into logically equivalent formulae, for instance, in prenex
form, until one such equivalent formula appears that is a theorem, thanks to an
axiom or to a previously proven theorem. Here an equivalence will emerge:
ŒI .A; B/,ŒI .B; A/ yet unproved,
   m definition of I
 8XfŒ E .X; A/,ŒE .X; B/g ^
  8YfΠE .A; Y/,ΠE .B; Y/g

, 8XfŒE .X; B/,ŒE .X; A/g ^ 8YfŒE .B; Y/,ŒE .A; Y/g ;
which suggests invoking the symmetry of the logical equivalence Œ.P/ , .Q/ ,
Œ.Q/ , .P/ (theorem 1.64). Thus, a complete proof may proceed as follows:
` Œ.P/,.Q/,Œ.Q/,.P/ theorem 1.64,
` fŒE .X; A/,ŒE .X; B/g,fŒE .X; B/,ŒE .X; A/g substitution,
` Œ.R/,.S/,Œ.S/,.R/ theorem 1.64,
` fŒE .A; Y/,ŒE .B; Y/g,fŒE .B; Y/,ŒE .A; Y/g substitution,
` fŒ.P/,.Q/ ^ Œ.R/,.S/g,fŒ.Q/,.P/ ^ Œ.S/,.R/g theorem 1.82.
Hence the conclusion results by Generalization and theorem 2.75. t
u
2.88 Theorem (transitivity of defined equality-predicates). Every equality-
predicate defined by formula (2.1) is transitive: if ` I .A; B/ and ` I .B; C/,
then ` I .A; C/; moreover,
 
` 8A8B8C fŒI .A; B/ ^ ŒI .B; C/g ) ŒI .A; C/ :

Proof. One method to design aformal proof transforms the objective, here  the yet
unproved formula ` 8A8B8C fŒI .A; B/ ^ ŒI .B; C/g ) ŒI .A; C/ , first into
its defining formula (2.1), and then into logically equivalent formulae, for instance,
 www.pdfgrip.com

106 2 First-Order Logic: Proofs with Quantifiers

in prenex form, until one such equivalent formula appears that is a theorem, thanks
to an axiom or to a previously proven theorem.
fŒI .A; B/ ^ ŒI .B; C/g ) ŒI .A; C/ yet unproved,
   m definition of I
 8XfΠE .X; A/,ΠE .X; B/g ^
  8YfΠE .A; Y/,ΠE .B; Y/g

^ 8XfŒE .X; B/,ŒE .X; C/g  ^ 8YfŒE .B; Y/,ŒE .C; Y/g 
) 8XfŒE .X; A/,ŒE .X; C/g ^ 8YfŒE .A; Y/,ŒE .C; Y/g ;
which suggests invoking the transitivity of the logical equivalence (theorem 1.65):

fŒ.H/ , .K/ ^ Œ.K/ , .L/g ) Œ.H/ , .L/;

fŒ.P/ , .Q/ ^ Œ.Q/ , .R/g ) Œ.P/ , .R/;

with the commutativity and associativity of the logical conjunction (theorems 1.57
and 1.66) combined with theorem 1.82:

 
fŒ.H/ , .K/ ^ Œ.K/ , .L/g ^ fŒ.P/ , .Q/ ^ Œ.Q/ , .R/g
) fŒ.H/ , .L/ ^ Œ.P/ , .R/g:

The conclusion results by Generalization and theorem 2.75. t

u

2.5.3 Other Axiom Systems for the Equality-Predicate

Other axioms systems exist to specify the identity predicate.

Axiom J 1 (reflexivity of equality) ` 8X I .X; X/.

Axiom J 2 (substitutivity of equality) For every unary predicate variable or

predicate constant F , involving only one individual variable,
 
` 8X8Y ŒI .X; Y/ ) fŒF .X/ ) ŒF .Y/g :

For every binary predicate variable or predicate constant F , involving only two
individual variables,
˚  
` 8X8Y8W8Z ŒI .X; Y/ ) ŒI .W; Z/ ) fŒF .X; W/ ) ŒF .Y; Z/g :
 www.pdfgrip.com

2.5 Equality-Predicates 107

For every ternary predicate variable or predicate constant F , involving only three
individual variables,

` 8U8V8X8Y8W8Z
 ˚  
ŒI .U; V/) ŒI .X; Y/) ŒI .W; Z/)fŒF .U; X; W/)ŒF .V; Y; Z/g :

Similar stipulations also hold for predicate variables or constants involving more
than three individual variables.

2.5.4 Defined Ranking-Predicates

Each application with a binary predicate constant E allows for a corresponding

predicate R of ranking, also called ordering or inequality, defined in terms of
E so that R .A; B/ is an abbreviation of the formula

8XfŒE .X; A/ ) ŒE .X; B/g: (2.3)

Formula (2.3) is also denoted by A  B (read “A precedes B”) instead of R .A; B/.
The resulting predicate R is reflexive and transitive, but not necessarily symmetric,
as verified in the exercises.

2.5.5 Exercises on Equality-Predicates

2.41 . Verify that the ranking-predicate R .A; B/ defined by formula (2.3) is reflex-
ive: prove ` 8AŒR .A; A/.
2.42 . Investigate whether the ranking-predicate R .A; B/ defined by formula (2.3)
is symmetric: is 8A8BfŒR .A; B/ ) ŒR .B; A/g is a theorem?
2.43 . Verify that the ranking-predicate
 R .A; B/ defined by formula
 (2.3) is transi-
tive: prove ` 8A8B8C ŒR .A; B/ ^ ŒR .B; C/g ) ŒR .A; C/ .
Exercises 2.45, 2.44, and 2.46 focus on the alternative ranking predicate A .A; B/
defined in terms of the same binary predicate constant E as an abbreviation of
formula (2.4):

8XfŒE .B; Y/ ) ŒE .A; Y/g: (2.4)

2.44 . Verify that the alternative ranking-predicate

 A .A; B/ defined by for-
mula (2.4) is transitive: prove ` 8A8B8C ŒA .A; B/ ^ ŒA .B; C/g ) ŒA .A; C/ .
 www.pdfgrip.com

108 2 First-Order Logic: Proofs with Quantifiers

2.45 . Verify that the alternative ranking-predicate A .A; B/ defined by

formula (2.4) is reflexive: prove ` 8AŒA .A; A/.
2.46 . Investigate whether the alternative ranking-predicate A .A; B/ defined by
formula (2.4) is symmetric: determine whether 8A8BfŒA .A; B/ ) ŒA .B; A/g
is a theorem.
2.47 . Verify that the equality predicate defined as in example 2.85 for set theory
satisfies the alternative axioms J 1 and J 2 from subsection 2.5.3.
2.48 . Verify that the equality predicate defined as in example 2.85 for set theory
satisfies axioms I 1 and I 2. from subsection 2.5.1.
2.49 . Verify that the equality predicate defined by axioms I 1 and I 2 from
subsection 2.5.1 also satisfies the alternative axioms J 1 and J 2 from subsec-
tion 2.5.3.
2.50 . Verify that the equality predicate defined by axioms J 1 and J 2 from
subsection 2.5.3 also satisfies axioms I 1 and I 2 from subsection 2.5.1.
 www.pdfgrip.com

Chapter 3
Set Theory: Proofs by Detachment,
Contraposition, and Contradiction

3.1 Introduction

This chapter introduces set theory from two parallel perspectives: as an intuitive
mathematical theory, and as a simple applied predicate calculus of first order.
Starting from first-order logic and some of the Zermelo-Fraenkel axioms (exten-
sionality, empty set, pairing, power set, separation, and union), where all objects
under consideration are sets, the chapter first derives relations between sets, subsets,
supersets, unions, intersections, and Cartesian products of sets of sets. Subsequent
sections introduce relations, functions, injections, surjections, bijections, composite
functions, and inverse functions. Another section focuses on the duality between
partitions and equivalence relations. The last section deals with pre-orders, partial
orders, linear or total orders, and well-orders. Many proofs begin with an informal
intuitive proof, then demonstrate how to design a more formal proof, and finally
present a detailed outline of such a formal proof in first-order logic. The other
Zermelo-Fraenkel axioms (choice and infinity or substitution) are only mentioned
here, because they form the topic of subsequent chapters. The prerequisites for
this chapter consist of a working knowledge of first-order logic, for instance, as
described in chapters 1 and 2, which contain all the logical theorems cited in this
chapter.
For some practical problems, features that are essential to their solutions can be
specified in terms of sets or collections of objects.
3.1 Example (Binary arithmetic). The binary arithmetic of computers relies on a
set of two symbols, 0 and 1, which will be defined with yet other sets in this chapter.
3.2 Example (Geometries). Geometries can be designed entirely with sets. Points
are sets (of sets of coordinates), while lines, planes, and space are sets of points.
Points, lines, planes, and space are “primitive” objects that may remain undefined,
but relations between them are specified through axioms. For instance, the first
axiom of incidence geometry specifies that through any two distinct points passes

© Springer Science+Business Media New York 2015 109

Y. Nievergelt, Logic, Mathematics, and Computer Science,
DOI 10.1007/978-1-4939-3223-8_3
 www.pdfgrip.com

110 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

exactly one line [61, p. 3]. Likewise in this chapter, mathematical “sets” are
“primitive” objects that remain undefined, while features of sets and relations
between sets are specified by axioms.
The foregoing examples already demonstrate a major difficulty in using problems
about “real” objects to illustrate logical and mathematical concepts: no exact
answers might be available. For an example as elementary as binary arithmetic,
electronic digital computers internally do not use anything like the symbols 0 and
1: indeed, they use two electrical potentials confined to two mutually exclusive
ranges, neither of which need contain any zero [48, p. 642], with different ranges on
different machines [51, p. 60, fig. 3–1; p. 83], [138, p. 4–5, 1], in the reverse order
on other machines [103, p. 1–4], sometimes reversing the order within the same
machine [23,  5]. A precise answer would involve more advanced engineering,
logic, mathematics, and physics. Therefore, the sets in the present exposition will
not contain “real” objects; instead, all the following sets will contain only abstract
objects defined by precise rules. The judicious use of such abstract mathematical
sets in applied disciplines from astronomy to zoology is precisely the task of such
disciplines.

3.2 Sets and Subsets

This section introduces mathematical sets by means of the concept of set member-
ship. The predicate of set membership then allows for the definition of the concepts
of subset, superset, and a derived predicate of equality.

3.2.1 Equality and Extensionality

One of the major mathematical achievements around the beginning of the twentieth
century was the realization that most of mathematics and computer science consists
of logical relations between abstract objects called sets [8, p. 3], [83]. There is no
definition of mathematical sets. Indeed, such a definition would have to define sets in
terms of yet more foundational objects, but sets are the most foundational objects.
Henceforth, here and as in other texts [22, p. 50], [141, p. 60], all mathematical
objects are sets, and all quantified variables designate sets:
andere Objekte als Mengen existieren für uns überhaupt nicht
(“for us objects other than sets simply do not exist”) [36, p. 271].

Instead of a definition of sets, a few “axioms” specify certain characteristics of

sets. Such axioms are also called postulates because they are appended to the list of
axioms for the underlying logic [18,  55]. Such axioms of logic describe universal
patterns of reasoning applicable in all contexts. In contrast, postulates specify the
 www.pdfgrip.com

3.2 Sets and Subsets 111

kind of objects considered in a particular context. The distinction between logical

axioms and applied postulates is convenient to distinguish between reasoning and
objects. Yet this distinction is somewhat artificial, because changing the logical
axioms also changes the provable properties of the applied objects, which amounts
to changing the kind of objects under consideration [18, p. 317, footnote 520].
Consequently, following tradition, the postulates of set theory are also labeled here
as axioms.
The set theory presented here involves one undefined primitive binary relation,
denoted by 2 and called “membership.” The symbol 2 is a typographical variation
on the lowercase Greek letter  (read “epsilon”), selected here as the first letter of the
copula ©K ¢£Jš (pronounced “es-tee”), meaning “is” [36, p. 272]. The notation X 2 Y is
read in various ways as “X is an element of Y,” “X is a member of Y,” or “X belongs
to Y.” For any set X and any set Y, the atomic formula X 2 Y has a Truth value, so
that X 2 Y is either True (if X is an element of Y) or False (if X is not an element of
Y). Thus, the following generalization of the tautology .B/ _ Œ:.B/ is universally
valid:
 
` 8X 8Yf.X 2 Y/ _ Œ:.X 2 Y/g :

The relation 2 of membership is the only foundational relation between sets.

Consequently, the only characteristics of a set are its elements. Because set theory
involves only one relation, every set A can relate to other sets X and Y in only four
ways:

X 2 A;
:.X 2 A/I
A 2 Y;
:.A 2 Y/:

Whether A 2 Y or :.A 2 Y/ might also be considered as a characteristic of

A. Consequently, rather than involving every element of A (every set X such that
X 2 A), another way to define the characteristics of A might involve every set
of which A is an element (every set Y such that A 2 Y). However, if the only
characteristics of a set are its elements, then the two ways ought to be logically
equivalent; this equivalence forms the essence of the axiom of extensionality.

Axiom S1 (Axiom of extensionality)

˚  
` 8A 8B f8XŒ.X 2 A/ , .X 2 B/g , f8YŒ.A 2 Y/ , .B 2 Y/g :

In the axiom of extensionality (S1), the formula 8XŒ.X 2 A/ , .X 2 B/ states

that each set X is an element of A if and only if X is an element of B. The formula
8YŒ.A 2 Y/ , .B 2 Y/ states that A is an element of Y if and only if B is
an element of Y. The axiom of extensionality states that these two formulae are
logically equivalent.
 www.pdfgrip.com

112 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

Yet another way to state that two sets have exactly the same characteristics
involves a derived binary relation (derived from 2) denoted by D and called
“equality.” For each set A and each set B, the formula A D B (read “A equals
B”) means that A and B have exactly the same characteristics. By the axiom of
extensionality, the equality of two sets can be stated in two logically equivalent
ways:
˚  
` 8A 8B .A D B/ , f8XŒ.X 2 A/ , .X 2 B/g ;
˚  
` 8A 8B .A D B/ , f8YŒ.A 2 Y/ , .B 2 Y/g :

The notation A D B is a shorthand to state that the following two formulae hold:

8XŒ.X 2 A/ , .X 2 B/;

8YŒ.A 2 Y/ , .B 2 Y/:

The axiom of extensionality states that these two formulae are logically equivalent.
A variation consists in defining A D B as an abbreviation of the first formula,
8XŒ.X 2 A/ , .X 2 B/, [8, p. 4–5], [36, p. 272–273, Def. 2], and then in adopting
the axiom that if A D B then the second formula holds: 8YŒ.A 2 Y/ , .B 2 Y/
[36, p. 274, Axiom I].
There is another presentation of set theory with two undefined relations, equality
.D/ and membership .2/. Then the axiom of extensionality specifies that two sets
are the “same” set if and only if they contain the “same” elements. In this exposition
the distinction just made does not matter, because equality .D/ serves only as a
shorthand: all operations with sets pertain to elements of those sets. (See also the
discussions by Bernays [8, p. 53] and Fraenkel [8, p. 6–8].) For the negations of
membership and equality, the following abbreviations prove convenient.
3.3 Definition. The symbols … and 6D denote the negations of 2 and D so that

` 8X8Yf.X … Y/ , Œ:.X 2 Y/gI

` 8A8Bf.A 6D B/ , Œ:.A D B/g:

At the elementary stage of set theory, most formal logical proofs of relations
between sets are straightforward, in the sense that they use only axioms and
definitions to establish a sequence of equivalences between the objective of the
proof and a theorem or universally valid formula. Such formal logical proofs are
usually longer than “informal” proofs. To show a first example of a proof within set
theory — a formal version and an informal version — the following theorem states
that each set equals itself.
 www.pdfgrip.com

3.2 Sets and Subsets 113

3.4 Remark. In designing a proof we may at any stage start from the conclusion
— but we may not assume it as True — and then search for logically equivalent
formulae that connect the conclusion to other formulae that we know how to prove.
Smullyan’s method of tableaux uses such an approach [117, Ch. II, p. 15–30].
3.5 Theorem. Each set equals itself: the formula 8S .S D S/ is universally valid.
Proof. An informal proof can consist of the following statements.
• Every set X is an element of S if and only if X is an element of S;
• hence S D S by definition of the equality of sets and extensionality (S1).
One method to design a formal proof transforms the objective, here the yet unproved
formula 8S .S D S/, into logically equivalent formulae, until one such equivalent
formula appears that is a theorem thanks to an axiom or to a previously proven
theorem. For instance, substituting S for A and also S for B in the axiom of
extensionality gives
SDS yet unproved,
m definition of D
8XŒ.X 2 S/ , .X 2 S/,
which is in prenex form, and where the logical formula .X 2 S/ , .X 2 S/ has the
pattern of the theorem .P/ , .P/. Thus, a complete proof may proceed as follows:
` .P/ , .P/ reflexivity of equivalence (theorem 1.63),
` .X 2 S/ , .X 2 S/ substitution in the theorem .P/ , .P/,
` 8Sf8XŒ.X 2 S/ , .X 2 S/g generalizations, first on X, then on S,
` 8S.S D S/ definition of D and extensionality (S1).
The proof just presented relies on one of the two formulae for the axiom of
extensionality: 8XŒ.X 2 A/ , .X 2 B/. Another proof could rely on the other
formula: 8YŒ.A 2 Y/ , .B 2 Y/. t
u
More generally, by the properties of a logical equality-predicate derived from a
binary predicate, here 2, as explained in section 2.5, the equality of sets is
• reflexive: ` 8S.S D S/ (every set equals itself), also proved in theorem 3.5,
• symmetric: ` 8A8BŒ.A D B/ ) .B D A/ (if A D B, then B D A),
• transitive: ` 8A8B8CfŒ.A D B/ ^ .B D C/ ) .A D C/g (if A D B and B D C,
then A D C),
and equality also allows substitutions of mutually equal sets in theorems.
The axiom of extensionality merely provides two logically equivalent criteria to
test whether sets have exactly the same characteristics. However, so far in this theory
there is no “set” yet. The “existence” of at least one set — or, more accurately, a
convention about an abstract concept of a specific set — requires a second axiom.
 www.pdfgrip.com

114 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.2.2 The Empty Set

The second axiom, called the axiom of the empty set, guarantees the existence of
at least one set, denoted by ¿ or also by f g; this set contains no element.

Axiom S2 (Axiom of the empty set) ` 8XŒ:.X 2 ¿/:

Set theory could dispense with the constant ¿ and state the axiom of the
empty set in the alternative form ` 9EŒ8X.X … E/. In the present theory, this
alternative form is a consequence of axiom S2. Indeed, with P denoting the formula
8X.X … E/, theorem 2.48 becomes

ŒSubfE¿ .P/ ) Œ9E.P/;

Œ8X.X … E/ ) f9EŒ8X.X … E/g:

The alternative form ` 9EŒ8X.X … E/ is more cumbersome, because it does not
provide a name for any such set.
The determination of the Truth value of an equality A D B requires prior
definitions of both sets A and B. In contrast, there exists a different use of the
same concept of equality, denoted by C WD D (read “let C equal D”) to specify
a hitherto undefined set C in terms of an already defined set D [59, p. 8], [121, p. 5].
Alternatively, the notation D DW C (also read “let C equal D”) can also serve to
specify C in terms of D, especially where a derivation leads to a lengthy formula D,
which can thus be abbreviated by a shorter variable or string C [121, p. 271, p. 347].
3.6 Example. The notation E WD ¿ specifies that E stands for ¿.

3.2.3 Subsets and Supersets

In some circumstances, only some of the elements of a set prove useful; the
following definition then allows for the grouping of all such elements into a “subset.”
3.7 Definition (Subsets and supersets). For each set A, for each set B, the set A is
a subset of the set B if and only if each element of A is also an element of B. Either
notation A  B or A j B indicates that “A is a subset of B”; thus,
˚  
` 8A 8B .A  B/ , f8XŒ.X 2 A/ ) .X 2 B/g :

Similarly, a set B is a superset of a set A if and only if A  B, a relation also denoted

by B  A or B k A.
In the definition of subsets, the equivalence .,/ states that the relation of subset
.A  B/ is logically equivalent to the formula 8XŒ.X 2 A/ ) .X 2 B/. In this
 www.pdfgrip.com

3.2 Sets and Subsets 115

formula, for each set X the logical implication .)/ states that if X is an element of
A, then X is also an element of B.
The concept of subset is so different from the concept of element as to warrant
different terminologies, for instance, reading A  B as “A is a subset of B” but
reading A 2 B as “A is an element of B.” In contrast, such vague phrases as “A is
in B” or “B contains A” do not have any significance, unless they are supplemented
with “as an element” or “as a subset” [36, p. 272].
There also exist symbols more specific than A  B. For instance, A  B or A ¨ B
or A ¤ B indicate that A is a subset of B different from B; thus,
 ˚ 
` 8A 8B .A ¤ B/ , .A ¨ B/ , .A  B/ , Œ.A  B/ ^ .A 6D B/ :

Similarly, B A or B © A or B ¥ A stand for .B  A/ ^ .B 6D A/.

(However, some authors use  to mean  [140,  3.12, p. 59–60].)
The following theorems provide further examples and practice with proofs in set
theory, and demonstrate how to design a proof. The first three theorems establish
features of the concept of subset: reflexivity, anti-symmetry, and transitivity.
3.8 Theorem (reflexivity of ). Each set is a subset of itself: ` 8S.S  S/.
Proof. An informal proof may consist of the single statement that each element of
S is also an element of S, whence S is a subset of S, by definition 3.7.
The design of a more formal proof can transform the set-theoretic formula S  S
into a logical formula, and verify that it is a logical theorem:
8S.S  S/ yet unproved,
m definition of 
8S f8XŒ.X 2 S/ ) .X 2 S/g
which is in the prenex form 8S f8XŒ.P/ ) .P/g, where the matrix is the theorem
.P/ ) .P/ (theorem 1.14), here with X 2 S instead of P. Thus a complete proof can
proceed as follows.
` .P/ ) .P/ theorem from implicational logic (1.14),
` .X 2 S/ ) .X 2 S/ substitution in theorem 1.14,
` 8Sf8XŒ.X 2 S/ ) .X 2 S/g generalizations, first on X, then on S,
` 8S.S  S/ definition (3.7) of subsets.
t
u
3.9 Theorem (transitivity of ). For all sets A, B, and C, if A  B and B  C,
then A  C:

` 8A8B8CfŒ.A  B/ ^ .B  C/ ) .A  C/g:

Proof. An informal proof can proceed as follows.

• If each element of A is an element of B,
• and if each element of B is an element of C,
 www.pdfgrip.com

116 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

• then each element of A is an element of B and hence also an element of C,

• whence A is a subset of B, by definition of subset.
The design of a formal proof can unravel the set-theoretic formula .A  B/ ^
.B  C/ ) .A  C/ into a logical formula by means of the definition of subset,
and verify that the resulting formula is a theorem from logic:
.A  B/ ^ .B  C/ ) .A  C/ yet unproved,

m definition of 
 
f8XŒ.X 2 A/ ) .X 2 B/g ^ f8XŒ.X 2 B/ ) .X 2 C/g
) f8XŒ.X 2 A/ ) .X 2 C/g

m theorems 2.54, 2.75, and 2.76,

  
8X fŒ.X 2 A/ ) .X 2 B/g ^ fŒ.X 2 B/ ) .X 2 C/g
) f8XŒ.X 2 A/ ) .X 2 C/g

* theorem 2.35,
˚ 
8X fŒ.X 2 A/ ) .X 2 B/g
 ^ fŒ.X 2 B/ ) .X 2 C/g
) Œ.X 2 A/ ) .X 2 C/
which is in prenex normal form, with a matrix of the type

fŒ.P/ ) .Q/ ^ Œ.Q/ ) .R/g ) Œ.P/ ) .R/;

which is another form of the transitivity of the implication (theorem 1.76). Thus the
last line is a theorem. Reversing the order of the steps and inserting 8A and 8B
before each step (generalizing) then completes the proof. t
u
3.10 Theorem (anti-symmetry of ). Two sets are subsets of each other if and
only if they equal each other:
 
` 8A 8Bf.A D B/ , Œ.A  B/ ^ .B  A/g :

Proof. An informal proof can proceed as follows.

• If each element of A is an element of B,
• and if each element of B is an element of A,
• then A and B have exactly the same elements;
• hence A D B by extensionality, and conversely.
 www.pdfgrip.com

3.2 Sets and Subsets 117

A formal proof can establish the following sequence of logical equivalences.

.A  B/ ^ .B  A/

m definition of subset,

f8XŒ.X 2 A/ ) .X 2 B/g ^ f8XŒ.X 2 B/ ) .X 2 A/g

m thanks to ` fŒ8X.P/ ^ Œ8X.Q/g , f8XŒ.P/ ^ .Q/g, by theorem 2.75,

8XfŒ.X 2 A/ ) .X 2 B/ ^ Œ.X 2 B/ ) .X 2 A/g

m definition of ,,

8XŒ.X 2 A/ , .X 2 B/

m axiom of extensionality (S1).

ADB
Inserting 8A and 8B before each step (generalizing) then completes the proof. t
u
The reflexivity, anti-symmetry, and transitivity of the concept of subset also result
more generally from the properties of a logical ranking-predicate, here , derived
from a binary predicate, here 2, as explained in section 2.5.
The next theorems focus on the subsets and supersets of the empty set.
3.11 Theorem. The empty set is a subset of every set: ` 8S.¿  S/.
Proof. An informal proof can use the converse law of contraposition (axiom P3):
• Every set not in S is also not in ¿, because no set belongs to ¿;
• the contraposition then means that every element in ¿ is also in S;
• hence ¿ is a subset of S, by the definition of subsets.
The design of a more formal proof can transform the set-theoretic formula ¿  S
into a logical formula, and verify that it is a logical theorem.
8S.¿  S/ yet unproved,
m definition of 
8S f8XŒ.X 2 ¿/ ) .X 2 S/g
m contraposition and theorem 2.45
8S f8XŒ.X … S/ ) .X … ¿/g
which is in the prenex form 8S8XŒ.P/ ) .Q/, where the matrix .P/ ) .Q/ is a
theorem, because so is Q. Thus a complete proof may consist of the following steps.
 www.pdfgrip.com

118 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

` 8X.X … ¿/ axiom of the empty set (S2),

`X…¿ logical axiom of specialization (Q1),
` .X … S/ ) .X … ¿/ theorem 1.12 from implicational logic,
` 8S f8XŒ.X … S/ ) .X … ¿/g generalizations,
` 8S f8XŒ.X 2 ¿/ ) .X 2 S/g contraposition and theorem 2.45
` 8S.¿  S/ definition 3.7 of .
t
u
3.12 Theorem. Every subset of the empty set is the empty set: ` 8SŒ.S  ¿/ )
.S D ¿/.
Proof. Theorem 3.11 already guarantees that the empty set is a subset of every set:
` 8S.¿  S/, in particular, for every subset S of the empty set, for which the
reverse inclusion also holds .S  ¿/. Hence S D ¿ by theorem 3.10.
` 8S.¿  S/ theorem 3.11,
`¿S specialization (axiom Q1),
` .¿  S/ ) f.S  ¿/ ) Œ.¿  S/ ^ .S  ¿/g theorem 1.54,
` .S  ¿/ ) Œ.¿  S/ ^ .S  ¿/ Detachment,
` .S  ¿/ ) .S D ¿/ theorem 3.10 and transitivity,
` 8SŒ.S  ¿/ ) .S D ¿/ Generalization.
t
u
The axioms of extensionality (S1) and of the empty set (S2) apply to all sets. Yet
so far the theory guarantees the existence of only one set, namely the empty set ¿,
for which ¿ 2 ¿ is False, but ¿ D ¿ and ¿  ¿ are True. Sets other than the
empty set require further axioms, as explained in the next section.

3.2.4 Exercises on Sets and Subsets

3.1 . Write a logical formula stating that a set S is not the empty set.
3.2 . Write a logical formula stating that a set A is not a subset of B.
3.3 . Write a logical formula stating that a set A is not a superset of B.
3.4 . Write a logical formula stating that a set A is not equal to a set B.
3.5 . Prove that ¿ 2 ¿ is not a theorem.
3.6 . Prove that ¿  ¿ is a theorem.
3.7 . Prove that ¿ D ¿ is a theorem.
3.8 . Use the second formula, 8YŒ.A 2 Y/ , .B 2 Y/, in the axiom of
extensionality to write a proof that S D S for each set S.
3.9 . Prove that ¿ is the only subset of ¿: ` 8SŒ.S  ¿/ ) .S D ¿/.
 www.pdfgrip.com

3.3 Pairing, Power, and Separation 119

3.10 . For each set S, prove that S ¥ S is not a theorem.

3.11 . Prove that if a set S is the only subset of itself, then S is empty.
3.12 . Prove that there does not exist a set S such that S 2 Y for every set Y.
3.13 . Prove that if a set S is a subset of every set, then S is empty.
3.14 . For all sets A, B, and C, prove that if A D B and B D C, then A D C.
3.15 . For all sets A, B, and C, prove that if A ¨ B and B ¨ C, then A ¨ C.
3.16 . For all sets A, B, and C, prove that if A  B and B  C, then A  C.
3.17 . For all sets A and B, prove that A D B if and only if for every set Z

.A  Z/ , .B  Z/:

3.18 . For all sets A and B, prove that A  B if and only if for each set Z

.B  Z/ ) .A  Z/:

3.19 . For all sets C and D, prove that C  D if and only if for each set W

.D  W/ ) .C  W/:

3.20 . For all sets R and S, prove that R  S if and only if for each set W

.W 2 S/ ) .W 2 R/:

3.3 Pairing, Power, and Separation

This section introduces axioms to form sets with one or two elements (pairing), all
subsets of a set (power), or selections of specific elements into a subset (separation).

3.3.1 Pairing

A theory allowing for sets other than the empty set requires additional axioms. For
instance, the axiom of pairing states that for every set H and every set K, there
exists a set L, also denoted by fH; Kg, which contains only the elements H and K.

Axiom S3 (Axiom of pairing)

˚   
` 8H 8K 9L 8X f.X 2 L/ , Œ.X D H/ _ .X D K/g :
 www.pdfgrip.com

120 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

In the axiom of pairing (S3), the equivalence .,/ states that a set X is an element
of L if and only if X equals H or X equals K. Because the logical “or” is inclusive,
the axiom of pairing thus allows fH; Kg to contain both H and K. Moreover, because
the logical “or” commutes, the order in which H and K appear does not matter.
3.13 Theorem. For each set H and for each set K, ` fH; Kg D fK; Hg.
Proof. An informal proof can compare the elements of fH; Kg and fK; Hg:
• The set fH; Kg contains the elements H and K, but no other element;
• the set fK; Hg contains the elements K and H, but no other element;
• thus fH; Kg and fK; Hg contain exactly the same elements;
• therefore fH; Kg D fK; Hg by the axiom of extensionality (S1).
A formal proof can use ` Œ.P/ _ .Q/ , Œ.Q/ _ .P/ (theorem 1.79) to show that

Œ.X D H/ _ .X D K/ , Œ.X D K/ _ .X D H/

is a theorem, and hence also

.X 2 fH; Kg/ , Œ.X D H/ _ .X D K/ , Œ.X D K/ _ .X D H/ , .X 2 fK; Hg/;

whence ` fH; Kg D fK; Hg by the axiom of extensionality (S1). t

u
With H WD S and K WD S, the following theorem shows that for each set S, there
exists a set, denoted by fSg, which contains only one element, S.
3.14 Theorem. For each set S there exists a set L, also denoted by fSg, which
contains only the element S; formally, ` 8S 9L f8X Œ.X 2 L/ , .X D S/g .
Proof. An informal proof can use the axiom of pairing with H WD S and K WD S:
• For each set S, the axiom of pairing yields a set fS; Sg;
• by the axiom of pairing, X 2 fS; Sg if and only if X D S or X D S;
• yet “X D S or X D S” merely repeats the same statement “X D S”;
• therefore fS; Sg D fSg by the axiom of extensionality.
A formal proof can rely on ` .P/ , Œ.P/ _ .P/ (theorem 1.68):
˚   
` 8H 8K 9L 8X f.X 2 L/ , Œ.X D H/ _ .X D K/g axiom S3,
 
` 9L 8X f.X 2 L/ , Œ.X D S/ _ .X D S/g SubfH K
S , SubfS ,
  
` 8S 9L 8X f.X 2 L/ , Œ.X D S/ _ .X D S/g Generalization,
 
` 8S 9Lf8XŒ.X 2 L/ , .X D S/g `.P/,Œ.P/_.P/.
t
u
3.15 Definition (singleton). A singleton is a set containing exactly one element.
3.16 Example. In theorem 3.14, substituting ¿ for S gives the set L D f¿g, which
contains the single element ¿. In particular, ¿ 2 f¿g, so that f¿g is not empty.
 www.pdfgrip.com

3.3 Pairing, Power, and Separation 121

3.17 Remark. The distinction between ¿ and f¿g is crucial. With˚ different
 ˚ nota-

tions, to appreciate better the difference between ¿ D f g and ¿ D f g ,
consider ¿ as an empty bag; then f¿g is a bag f: : :g with another empty bag ¿
inside it, also known as a “double bag” in the market place. There, a single empty
bag f g D ¿ might not be sufficiently strong to hold a six-pack of heavy glass
bottles filled with your favorite beverage. (Bottled
˚ water,
 ˚of course,
 what were you
thinking?) That’s why you ask for a double bag ¿ D f g in which to put and
then carry the heavy six-pack. If the six-pack also n˚ comes with
n˚ a wrapping, then
o o
the combined packaging becomes a “triple bag”: ¿ D f g , which is yet
another set.
Theorem 3.18 confirms that the sets ¿ and f¿g have different characteristics.
3.18 Theorem. The sets ¿ and f¿g are two distinct sets: ¿ 6D f¿g.
Proof. An informal proof can utilize substitutions in previous axioms and
theorems:
• By definition of the empty set, ¿ … ¿ (by a substitution in axiom S2);
• moreover, ¿ 2 f¿g (by a substitution in theorem 3.14);
• hence the two sets ¿ and f¿g have different elements: ¿ 2 f¿g but ¿ … ¿;
• consequently, ¿ 6D f¿g, by the axiom of extensionality.
The following formal proof uses contraposition in the theorems
` .P/ ) fŒ.P/ ) .Q/ ) .Q/g  law of assertion (theorem 1.38),
` .P/ ) Œ:.Q/ ) f:Œ.P/ ) .Q/g theorem 1.54.
whence if P and :.Q/ are theorems, then so is :Œ.P/ ) .Q/ by Detachment:

` :.¿ 2 ¿/ :.Q/: substitution in axiom S2,

` ¿ 2 f¿g P: substitution in theorem 3.14,
` :Œ.¿ 2 f¿g/ ) .¿ 2 ¿/ :Œ.P/ ) .Q/ ,
˚ 
` 9X :Œ.X 2 f¿g/ ) .X 2 ¿/ SubfX¿ with theorem 2.48,
˚ 
` : 8XŒ.X 2 f¿g/ ) .X 2 ¿/ f9XŒ:.P/g , f:Œ8X.P/g (axiom Q4),
` :.f¿g  ¿/ definition 3.7 of subsets,
` ¿ 6D f¿g contraposition of theorem 3.10 .
t
u
The distinction between ¿ and f¿g allows for the formation of other sets.
3.19 Example. Substituting H WD ¿ and K WD f¿g in the axiom of pairing gives
the set L D fH; Kg D f¿; f¿gg. This set L has two elements, because ¿ 6D f¿g.
3.20 Example. Substituting S WD f¿g in theorem 3.14 gives the set L D fSg D
f f¿g g. This set L has one element, in effect f¿g; thus, f¿g 2 f f¿g g.
 www.pdfgrip.com

122 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.3.2 Power Sets

For sets with more than two elements, a new axiom becomes necessary. The axiom
of the power set states that for each set A, the collection of all subsets of A forms a
new set, denoted by P or by P.A/ and called the power set of A:

Axiom S4 (Axiom of the power set)

 
` 8A 9P f8SŒ.S 2 P/ , .S  A/g :

In the axiom of the power set (S4), the equivalence .,/ states that a set S is an
element of the power set P.A/ if and only if S is a subset of the set A.
3.21 Example. The empty set ¿ is the only subset of itself, by theorem 3.12. Hence
its power set has only one element, ¿, so that P.¿/ D f¿g:
A set A and its power set P.A/ might have no elements in common. Theo-
rem 3.22 shows that for every element X of A, the singleton fXg is an element
of P.A/.
3.22 Theorem. For all sets A and X, if X 2 A, then fXg 2 P.A/, and conversely:
˚ 
8A8X .X 2 A/ , ŒfXg 2 P.A/ :

Proof. An informal proof can rely on the definitions of subsets and power sets.
• X 2 A if and only if fXg  A, by definitions of fXg and subsets;
• fXg  A if and only if fXg 2 P.A/ by definition of power sets.
A more formal proof carries out similar verifications from the formal definitions.
 
8A 8Xf.X 2 A/ , ŒfXg 2 P.A/g yet unproved,
 m definition of power sets,

8A 8Xf.X 2 A/ , ŒfXg  Ag
˚ m  definition of subsets, 
8A 8X .X 2 A/ , f8ZŒ.Z 2 fXg/ ) .Z 2 A/g
˚ m  definition of fXg (pairing), 
8A 8X .X 2 A/ , f8ZŒ.Z D X/ ) .Z 2 A/g
˚ m theorems 2.34 and 2.38, 
8A 8X8Zf.X 2 A/ , Œ.Z D X/ ) .Z 2 A/g
which holds by extensionality and substitutivity of equality. t
u
Theorem 3.23 shows that the power set of a singleton has exactly two elements.
˚ 
3.23 Theorem. For each set H, P.fHg/ D ¿; fHg .
Proof. An informal proof can list the subsets of fHg by cases:
 www.pdfgrip.com

3.3 Pairing, Power, and Separation 123

• fHg  fHg by theorem 3.8,

• ˚¿  fHg by theorem 3.11,
• ¿; fHg  P.fHg/ by the previous two lines;
• if B  fHg, then B has no element or B has the single element H,
• hence fHg has no˚ subsetsother than ¿ and fHg,
• thus P.fHg/  ¿;˚fHg by the previous two lines,
• whence P.fHg/ D ¿; fHg .
A formal proof can proceed through two cases, using theorem 1.85:
 
` Œ.P/ ) .Q/ ^ fŒ:.P/ ) .Q/g ) .K/;
˚ 
 B D ¿ and Q for
˚with P for ˚ B 2 ¿; f¿g . In the first case, .B D ¿/ ) .B 2
¿; f¿g / by substituting ¿; f¿g for Y in the axiom of extensionality (S1). In
the second case, the definitions of B 6D ¿ and B  fHg give
.B 6D ¿/ ^ .B  fHg/
m definitions of ¿ and ,
Œ9X.X 2 B/ ^ f8ZŒ.Z 2 B/ ) .Z 2 fHg/g
m definition of fHg by theorem 3.14,
Œ9X.X 2 B/ ^ f8ZŒ.Z 2 B/ ) .Z D H/g
+ .ZDH/)f8YŒ.Z2Y/,.H2Y/g
 
9X .X 2 B/ ^ f8ZŒ.Z 2 B/ ) .H 2 B/g
+ specialization SubfZX .
9Xf.X 2 B/ ^ Œ.X 2 B/ ) .H 2 B/g
+ f.P/^Œ.P/).Q/g).Q/,
H2B
Thus B  fHg, but H 2 B, whence fHg  B, and hence B D fHg. t
u
3.24 Example. The singleton f¿g has two subsets: f¿g by theorem 3.8, and ¿ by
theorem 3.11. Moreover, f¿g has no other subsets. Hence, P.f¿g/ D f¿; f¿gg :
˚ 
3.25 Counterexample. nThe set A WD o f¿g has no element in common with its
˚  ˚ 
power set P f¿g D ¿; f¿g : In particular, A is not a subset of P.A/.

The axioms of the empty set (S2), of pairing (S3), and of the power set (S4) allow
for increasingly large sets, for instance,

¿;
P.¿/ D f¿g;
˚ 
P.f¿g/ D n ¿; f¿g ˚;
˚   ˚ o
P ¿; f¿g D ¿; f¿g; f¿g ; ¿; f¿g :
::
:

However, the “selection” or “separation” of subsets requires yet another axiom.

 www.pdfgrip.com

124 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.3.3 Separation of Sets

For each set A and for each logical formula P, the axiom schema of separation
“separates” or “sets aside” from A a subset S consisting of all the elements X of A
for which P is True. The logical formula P may not involve any free occurrence
of the variable S, but it may involve free occurrences of X and of other variables,
as indicated in the following axiom by the ellipsis : : :, which may represent other
free variables. Such a “separation rule” to form subsets differs from other axioms,
because the logic used here allows for the quantification of the elements, with the
symbols 8X, but this logic has no provision for the quantification of formulae: it
does not allow for expressions like 8P with P standing for formulae. Thus, the
“separation rule” provides a schema for infinitely many axioms, in effect one axiom
for each logical formula, whence its name.

Axiom S5 (Axiom schema of separation) For each set A, and for each logical
formula P that does not contain any free occurrence of the variable S, there exists a
subset S  A that consists of only those elements X 2 A for which P is True:
  
` 8A 9S 8X f.X 2 S/ , Œ.X 2 A/ ^ .P/g :

Common notations for the subset S have the formats

S D fX W .X 2 A/ ^ .P/g;
S D fX 2 A W Pg:

An alternative notation replaces the colon (W) by a vertical bar (j), but in the context
of further mathematics such vertical bars become difficult to recognize against
absolute values, norms, and other similar symbols.
3.26 Example. Consider the set

A WD P.f¿;
˚ f¿gg/ 
D ¿; f¿g; ff¿gg; f¿; f¿gg ;

and let P be the formula :.X D ¿/. Then

˚ 
A D ¿; f¿g; ff¿gg; f¿; f¿gg ;
.P/ , Œ:.X D ¿/;
S D fX W .X 2 A/ ^ Œ:.X D ¿/g;
S D fX
˚ 2 A W :.X D ¿/g; 
S D f¿g; ff¿gg; f¿; f¿gg :

The existence of this set S would not have followed from the previous axioms.
 www.pdfgrip.com

3.3 Pairing, Power, and Separation 125

The following examples demonstrate the axiom schema of separation with

various instances of the formula P. Additional examples appear in the next section.
3.27 Example. For each set A and for each set B, let P be .X 2 B/. Then

S D fX W .X 2 A/ ^ .X 2 B/g;

which is usually denoted by S D A \ B and called the intersection of A and B.

3.28 Example. For each set A and for each set B, let P be :.X 2 B/. Then

S D fX W .X 2 A/ ^ Œ:.X 2 B/g;

which is usually denoted by S D A n B and called the difference of A and B.

The symbol n adopted here for the difference of two sets aims at avoiding
confusions with the arithmetic difference of sets A B in such further branches
of mathematics as convexity, linear algebra, and functional analysis.
3.29 Example. For each set A, and for all subsets B  A and C  A, let P be the
formula .X 2 B/ _ .X 2 C/. Then

S D fX W .X 2 A/ ^ Œ.X 2 B/ _ .X 2 C/g;

which is usually denoted by S D B [ C and called the union of B and C.

Unions and intersections of more than two sets form the topic of the next section.
3.30 Remark. Zermelo introduced the axiom of separation (S5) as the “Axiom der
Aussonderung” [145, p. 263, Axiom III], which also translates as the “axiom of
triage” or “sifting” [8, p. 11, footnote 2]. With the axiom of separation, the theory
presented here can be called Zermelo’s set theory. Fraenkel and Skolem substituted
for the axiom of separation a more general axiom called the “axiom of replacement”
(“Axiom der Erzetzung”) [36, p. 309, Axiom VIII], [116], which can be stated with
the notation from definition 2.21 [22, p. 52]:
   
8Z : : : 8W f8XŒ9ŠY.Q/g ) 8A9B8C .C 2 B/ , 9Df.D 2 A/ ^ ŒSubfXD SubfYC .Q/g

or, alternatively [128, p. 202],

˚ 
Y .Q/ ^ ŒSubfZ .Q/g ) .Y D Z/
8A 8W8X8Y8Zf.X 2 A/ ^ ŒSubfW W

  
) 9B8Y .Y 2 B/ , f9X.X 2 A/ ^ ŒSubfW Y .Q/g :

This formula can be better described with the concept of “mathematical function”
defined in section 3.6. The axiom of separation follows from the axiom of
replacement by substituting .P/ ^ .W D X/ for Q.
 www.pdfgrip.com

126 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.3.4 Exercises on Pairing, Power, and Separation of Sets

3.21 . Give examples of sets X, Y, Z with X 2 Y and Y 2 Z but X … Z.

3.22 . Provide examples of sets X, Y such that X 2 Y but X ¤ Y.
3.23 . Provide examples of sets X, Y such that X  Y but X … Y.
3.24 . Provide examples of sets X, Y such that X 2 Y and X  Y.
3.25 . Provide examples of sets A, X such that X 2 A but X … P.A/.
3.26 . Provide examples of sets A, X such that X  A but X ¤ P.A/.
3.27 . Prove that f¿g 6D f f¿g g.
3.28 . Prove that f¿g 6D f ¿; f¿g g.
3.29 . Prove that f f¿g g 6D f ¿; f¿g g.
˚ 
3.30 . Determine whether f¿g  f¿g .
˚  ˚ 
3.31 . Determine whether f¿g  f¿g .
˚ 
3.32 . Determine whether f¿g  ¿; f¿g .
˚  ˚ 
3.33 . Determine whether f¿g  ¿; f¿g .
3.34 . Prove that replacing _ by ^ in axiom S3 yields the False formula
˚   
8H 8K 9L 8X f.X 2 L/ , Œ.X D H/ ^ .X D K/g :

3.35 . For theorem 3.13, explain how the word “and” in the informal proof
corresponds to the logical connective _ in the formal proof.
3.36 . For each set S, prove that f¿; Sg  P.S/.
3.37 . Prove
 that two sets equal each other if and
 only if they have the same power
sets: ` 8A 8B f.A D B/ , ŒP.A/ D P.B/g :
˚ 
3.38 . Identify the set f¿g \ ¿; f¿g .
˚  ˚ 
3.39 . Identify the set f¿g \ ¿; f¿g .
˚ 
3.40 . Identify the set f¿g \ f¿g .
˚ 
3.41 . Identify the set f¿g [ ¿; f¿g and one of its supersets.
˚  ˚ 
3.42 . Identify the set f¿g [ ¿; f¿g and one of its supersets.
3.43 . For each set S, identify the set S \ ¿.
3.44 . For each set S, identify the set S [ ¿.
3.45 . Prove that ` 8AŒ.A n ¿/ D A.
 www.pdfgrip.com

3.4 Unions and Intersections of Sets 127

3.46 . Prove that ` 8AŒ.A n A/ D ¿.

 
3.47 . Prove that ` 8A 8BfŒ.A n B/ D ¿ , .A  B/g .
 
3.48 . Prove that ` 8A 8BfŒ.A n B/ D A , Œ.A \ B/ D ¿/g .
3.49 . Prove or disprove 8Af8BŒ.A n B/ D .B n A/g.
  
3.50 . Prove or disprove 8A 8B 8CfŒ.A n B/ n C D ŒA n .B n C/g .

3.4 Unions and Intersections of Sets

3.4.1 Unions of Sets

Many mathematical situations involve unions and intersections of any “collection”

or “family” of sets, or, in other words, of any set of sets. For instance, the Venn
diagram in figure 3.1 (adapted from Hamburger and Pippert’s [52, 53]) illustrates
all the possible unions and intersections for a set of five sets. Scientists, for instance,
biologists, also use such Venn diagrams [7, p. 402, Fig. 2(B)], [90, p. 1472, Fig. 3].
For the existence of such general unions and intersections, however, new axioms
become necessary. Thus, S the axiom of union asserts that for each set F there exists
a new set U, denoted by F and called the union of F , which consists of all the
elements X of all the elements S of F :

Axiom
S S6 (Axiom of union) For each set F , there exists a set U, also denoted by
F , which consists of all the elements that belong to any element of F :
 ˚  
` 8F 9U 8X .X 2 U/ , f9SŒ.S 2 F / ^ .X 2 S/g :

Fig. 3.1 Venn diagram for

the unions and intersections
of five sets [52, 53].
Scientists, for instance,
biologists, also use such Venn
diagrams [7, p. 402,
Fig. 2(B)], [90, p. 1472,
Fig. 3].
 www.pdfgrip.com

128 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

In the axiom ofSunion (S6), the equivalence .,/ states that a set X is an element
of the union U D F if and only if there exists an element S of F such that X 2 S.
S
3.31 Example. If F D ¿, then . ¿/ D ¿, because in the axiom of union (S6)
the condition
S .S 2 F / is False for everySset S, and then the equivalent condition
X 2 . ¿/ is False for every set X. Thus ¿ contains no element.
For the union of two sets, a special notation proves convenient.
S
3.32 Definition. The notation A [ B stands for fA; Bg.
S
With only two sets in F , the definition of F simplifies considerably.
3.33 Theorem. For all sets A and B, if F D fA; Bg, then
nh [ i o
8X X2 fA; Bg , Œ.X 2 A/ _ .X 2 B/ :

Proof. For F D fA; Bg, the axiom of pairing (S3) shows that

.S 2 F / , Œ.S D A/ _ .S D B/;

whence the axiom of union (S6) gives the following condition for X 2 .A [ B/:

ŒX 2 .A [ B/ , f9SŒ.S
 2 fA; Bg/ ^ .X 2 S/g 
, 9SfŒ.S D A/ _ .S D B/ ^ .X 2 S/g 
, 9SfŒ.S D A/ ^ .X 2 S/ _ Œ.S D B/ ^ .X 2 S/g
, f9SŒ.X 2 A/ _ .X 2 B/g
, Œ.X 2 A/ _ .X 2 B/:

The last equivalences result from extensionality, which gives

Œ.S D A/ ^ .X 2 S/ , .X 2 A/;

Œ.S D B/ ^ .X 2 S/ , .X 2 B/:

Moreover, S does not occur in .X 2 A/ _ .X 2 B/, which makes 9S superfluous, so

that f9SŒ.X 2 A/ _ .X 2 B/g , Œ.X 2 A/ _ .X 2 B/ by theorem 2.49. t
u
˚  n˚ o
3.34 Example. For A WD ¿; f¿g and B WD ¿; f¿g , the union A [ B consists
of the elements that belong to A (these are ¿ and f¿g) ornthat belong to B (where
˚  ˚ o
there is only one element, ¿; f¿g ). Therefore, A [ B D ¿; f¿g; ¿; f¿g .

3.35 Example. Abbreviations are common [81, p. 98], [83, p. 453], [128, p. 129]:
 www.pdfgrip.com

3.4 Unions and Intersections of Sets 129

0 WD ¿;

1 WD 0 [ f0g D ¿ [ f¿g D f¿g;

˚  ˚ 
2 WD 1 [ f1g D f¿g [ f¿g D ¿; f¿g ;

˚  n˚ o n ˚ o
3 WD 2 [ f2g D ¿; f¿g [ ¿; f¿g D ¿; f¿g; ¿; f¿g ;

˚  n ˚ o
4 WD 3 [ f3g D ¿; f¿g; ¿; f¿g ; ¿; f¿g; ¿; f¿g ;

5 WD 4 [ f4g;
6 WD 5 [ f5g;
7 WD 6 [ f6g;
8 WD 7 [ f7g;
9 WD 8 [ f8g:

The sets 0; 1 are the binary digits; 0; 1; 2; 3; 4; 5; 6; 7; 8; 9 are the decimal digits.
3.36 Example. For all sets A, B, and C, there exists a set V whose only elements are
A, B, and C. Indeed, the axiom of pairing yields two sets fA; Bg and fB; Cg. With
˚ 
F D fA; Bg; fB; Cg ;
S
the axiom of union produces a set F D fA; Bg [ fB; Cg such that for each set X,

ŒX 2 .fA; Bg [ fB; Cg/ , Œ.X 2 fA; Bg/ _ .X 2 fB; Cg/

, fŒ.X D A/ _ .X D B/ _ Œ.X D B/ _ .X D C/g
, Œ.X D A/ _ .X D B/ _ .X D C/:

The common notation fA; B; Cg can replace fA; Bg [ fB; Cg, so that

fA; B; Cg D fA; Bg [ fB; Cg:

3.37 Example. SIf the set F contains only three sets A, B, and C, so that F D
fA; B; Cg, then fA; B; Cg is also denoted by A [ B [ C, so that
[ [
A[B[C D fA; B; Cg D F

consists of all the elements that are elements of at least one of A, B, or C.

The following theorems show that unions and intersections of sets have features
similar to — and derived from — those of logical disjunctions and conjunctions.
 www.pdfgrip.com

130 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.38 Theorem. The union of sets commutes: ` 8A f8BŒ.A [ B/ D .B [ A/g.

Proof. An informal proof may consist of the statements that a set belongs to A [ B
if and only if it belongs to A or B, which means that it belongs B or A, and hence
to B [ A, whence A [ B D B [ A by the axiom of extensionality. A formal proof
reveals that the commutativity of the union [ corresponds to the commutativity of
the logical connective _ in the proof of theorem 3.13, which states that fA; Bg D
fB; Ag:
[ [
A[BD fA; Bg D fB; Ag D B [ A:

t
u
3.39 Theorem. The union of sets is associative:
  
` 8A8B 8C fŒ.A [ B/ [ C D ŒA [ .B [ C/g :

Proof. An informal proof can merely point out that a set “A or B, or C” is equivalent
to “A, or B or C.” A formal proof reveals that the associativity of the union [
corresponds to the associativity of the logical connective _ in a translation of
Œ.A [ B/ [ C D ŒA [ .B [ C/ into a theorem with atomic formulae and connectives:
Œ.A [ B/ [ C D ŒA [ .B [ C/ yet unproved,
m axiom of extensionality (S1),
8X .fX 2 Œ.A [ B/ [ Cg
, fX 2 ŒA [ .B [ C/g/
m theorem 3.33 twice,
8X .fŒ.X 2 A/ _ .X 2 B/ _ .X 2 C/g
, f.X 2 A/ _ Œ.X 2 B/ _ .X 2 C/g/
which is in prenex form with a matrix that is a theorem:

fŒ.P/ _ .Q/ _ .R/g , f.P/ _ Œ.Q/ _ .R/g;

in other words, by associativity of the logical disjunction _, by theorem 1.71. t

u
The following theorem
S shows that if a set B is an element of a set F , then B is a
subset of the union F .
S
3.40 Theorem. For each set F and for each set B, if B 2 F , then B  . F /:
 n h [ io
` 8F 8B .B 2 F / ) B  F :

Proof. An informal proof can substitute B for S in the axiom of union (S6):
• If B 2 F , then
S for each X 2 B there exists S 2 F Swith X 2 S, namely S D B;
• hence X 2 F for S every X 2 B, by definition of F (axiom S6);
• consequently B  F by definition of subsets (definition 3.7).
 www.pdfgrip.com

3.4 Unions and Intersections of Sets 131

A formal proof consists in transforming the proposed formula until a theorem

appears. Because two steps involve not an equivalence but an implication, however,
the final proof reorders the investigative five steps in their reverse order:
 S 
(5) 8F 8B fŒB 2 F  ) ŒB  . F /g yet unproved,

m definitions of ,
h  ˚ S i
(4) 8F 8B ŒB 2 F  ) 8X f.X 2 B/ ) ŒX 2 . F /g

S
m definitions of ,
h  ˚  i
(3) 8F 8B ŒB 2 F  ) 8X .X 2 B/ ) f9AŒ.A 2 F / ^ .X 2 A/g

* no X in .B 2 F /, hence f8XŒ.P/ ) .Q/g ) f.P/ ) Œ8X.Q/g

(theorem 2.38),
n  ˚  o
(2) 8F 8B 8X ŒB 2 F  ) .X 2 B/ ) f9AŒ.A 2 F / ^ .X 2 A/g

* ` ŒSubfAB .P/ ) Œ9A.P/,

 ˚  
(1) 8F 8B 8X .B 2 F / ) f.X 2 B/ ) Œ.B 2 F / ^ .X 2 B/g
which holds thanks to ` .P/ ) f.Q/ ) Œ.P/ ^ .Q/g (theorem 1.54). t
u
S
Theorem 3.41 shows that if F is a set of subsets of a set A, then its union F
is a subset of A.
S
3.41 Theorem. For all sets A and F , if F  P.A/, then F  A.
Proof. By the axiom of union (S6)
n [  o
8X X2 F , f9BŒ.X 2 B/ ^ .B 2 F /g :

Yet .B 2 F / ) .B  A/ by the hypothesis that F  P.A/, whence

Œ.X 2 B/ ^ .B 2 F / ) Œ.X 2 B/ ^ .B  A/

by theorem 1.82, and hence X 2 A by definition 3.7 of subset. Consequently,

because X 2 A has no free occurrences of B, theorem 2.57 gives
h [  i
8X X2 F ) .X 2 A/ ;
S
so that F  A again by definition 3.7 of subset. t
u
 www.pdfgrip.com

132 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.4.2 Intersections of Sets

Based on the axiom of union and the axiom of T separation, definition 3.42 specifies
for each nonempty set F a new set, denoted by F and called the intersection of
F , which consists of every element X that is an element of every element of F .
3.42 Definition (intersection
S of sets). For each set F , apply the axiom of union
to define A WD F , and apply the axiom of separation to the set A and to the
formula

8YŒ.Y 2 F / ) .X 2 Y/:
T
Then define the intersection of F , a set denoted by F , through the formula
\ n [ o
F WD X 2 F W 8YŒ.Y 2 F / ) .X 2 Y/ ;

so that
˚ \  [ 
8X .X 2 F / , .X 2 F / ^ f8YŒ.Y 2 F / ) .X 2 Y/g :
T
The definition
T of the intersection F of a set of sets F states that a set X is an
element of F if and only if 8YŒ.Y 2 F / ) .X 2 Y/ holds, which occurs if and
only if X is an element of every element T Y of F . This definition also holdsS if F
is empty because ofTthe requirement S that F first be a subset of the union F.
(This definition of in terms of conforms to Bernays’s
T [8, p. 14].) If the set
F contains only two elements, then the definition of F simplifies considerably.
3.43 Theorem. For all sets A and B, if F D fA; Bg then
\
fA; Bg D fX W ŒX 2 .A [ B/ ^ Œ.X 2 A/ ^ .X 2 B/g :

Proof. Apply theorem 1.55, Œ.H/ ) .K/ ) Œ.H/ ) .L/ ) f.H/ ) Œ.K/ ^
.L/g :
 
` 8Yf.Y 2 fA; Bg/ ) .X 2 Y/g
) Œ.A 2 fA; Bg/ ) .X 2 A/ specialization SubfYA ,
` A 2 fA; Bg axiom of pairing (S3),
` X 2 A  Detachment;
` 8Yf.Y 2 fA; Bg/ ) .X 2 Y/g
) Œ.B 2 fA; Bg/ ) .X 2 B/ specialization SubfYB ,
` B 2 fA; Bg axiom of pairing (S3),
`X2B Detachment;
 www.pdfgrip.com

3.4 Unions and Intersections of Sets 133

 
` 8Yf.Y 2 fA; Bg/ ) .X 2 Y/g
) Œ.X 2 A/ ^ .X 2 B/ theorem 1.55;

` .X 2 A/ ) Œ.A 2 fA; Bg/ ) .X 2 A/ from axiom P1,

` .X 2 B/ ) Œ.B 2 fA; Bg/ ) .X 2 B/ from axiom P1,
` Œ.X 2 A/ ^ .X 2 B/ 
) 8Yf.Y 2 fA; Bg/ ) .X 2 Y/g
which holds thanks to theorem 1.82:

` fŒ.P/ ) .Q/ ^ Œ.R/ ) .S/g ) fŒ.P/ ^ .R/ ) Œ.Q/ ^ .S/g:

t
u
For the intersection of two sets, a special notation proves convenient.
T
3.44 Definition. The notation A \ B stands for fA; Bg.
3.45 Example. For the sets
˚ 
A WD ¿; f¿g; f f¿g g ;
˚ 
B WD f¿g; f f¿g g; ff f¿g gg ;

A \ B contains only the elements belonging “simultaneously” to both A and B:

˚  T˚ 
A\B D ¿; f¿g; f f¿g g f¿g; f f¿g g; f f f¿g g g
˚ 
D f¿g; f f¿g g :

Because the definition of the intersection of sets relies upon the conjunction ^,
the intersection has formal features similar to the logical features of the conjunction,
for instance, commutativity and associativity, as demonstrated in the following
theorems.
3.46 Theorem. The intersection of sets commutes:

` 8A f8BŒ.A \ B/ D .B \ A/g :

Proof. An informal proof can state that a set X is an element of A and B if and only
if X is an element of B and A. A formal proof shows that the commutativity of the
intersection \ corresponds to the commutativity of the logical connective ^:
.A \ B/ D .B \ A/ yet unproved,
m axiom S1,
8X fŒX 2 .A \ B/ , ŒX 2 .B \ A/g
m theorem 3.43,
8X fŒ.X 2 A/ ^ .X 2 B/ , Œ.X 2 B/ ^ .X 2 A/g
 www.pdfgrip.com

134 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

which holds by the commutativity ` Œ.P/ ^ .Q/ , Œ.Q/ ^ .P/ (theorem 1.57). t
u
3.47 Theorem. The intersection of sets is associative:
  
` 8A 8B 8CfŒ.A \ B/ \ C D ŒA \ .B \ C/g :

Proof. An informal proof can rely on the equivalence of “A and B, and C” with
“A, and B and C.” A formal proof shows that the associativity of the intersection \
corresponds to the associativity of the logical connective ^.
Œ.A \ B/ \ C D ŒA \ .B \ C/ yet unproved,
 m axiom of extensionality (S1),
8X fX 2 Œ.A \ B/ \ Cg
, fX 2 ŒŒA \ .B \ C/g
 m theorem 3.43 twice,
8X fŒ.X 2 A/ ^ .X 2 B/ ^ .X 2 C/g
, f.X 2 A/ ^ Œ.X 2 B/ ^ .X 2 C/g
which holds thanks to the associativity of ^ (theorem 1.66):

fŒ.P/^.Q/^.R/g , f.P/^Œ.Q/^.R/g: t
u
T
Theorem 3.48 shows that the intersection F of a set of sets F Tis a subset of
every element of F . In other terms, for each set B, if B 2 F , then . F /  B.
T
3.48 Theorem. For each (nonempty) set F and for each B 2 F , . F /  B:
 n h\  io
` 8F 8B .B 2 F / ) F B :

Proof. An informal proof can consist of the following statements:

T
• If X 2 F , thenT X is an element of every Y 2 F , and hence of B 2 F ;
• consequently, . F /  B, by definition of subsets.
A formal proof can consist in transforming the proposed formula into its prenex
form until a theorem appears in its matrix.
T
8F .8B f.B 2 F / ) Œ. F /  Bg/ yet unproved,
n h T ii m definition of ,
8F 8B .B 2 F / ) f8X Œ.X 2 F / ) .X 2 B/g
T
n h m definition of ,
8F 8B .B 2 F / )
˚  io
8X f8YŒ.Y 2 F / ) .X 2 Y/g ) .X 2 B/
* no X in B 2 F (theorem 2.38),
 www.pdfgrip.com

3.4 Unions and Intersections of Sets 135

n h ˚
8F 8B 8X .B 2 F / )
 io
f8YŒ.Y 2 F / ) .X 2 Y/g ) .X 2 B/
m theorem 1.37,
P
n h  ‚ …„ ƒ
8F 8B 8X f8YŒ.Y 2 F / ) .X 2 Y/g
 io
) .B 2 F…/ ) .X
„ ƒ‚ 2 B/
„ƒ‚…
Q R

which holds by specialization (SubfYB ) and the law of commutation (theorem 1.37):
   
` .P/ ) f.Q/ ) .R/g , .Q/ ) f.P/ ) .R/g :
t
u
3.49 Definition (disjoint sets). Two sets A and B are disjoint if and only if
A \ B D ¿. Similarly, a set of sets F is pairwise disjoint if and only if either
A D B or A \ B D ¿ for all elements A and B of F .
For the union of disjoint sets, a special notation proves convenient.
3.50 Definition (disjoint unions). A union A [ B is disjoint if and only if A \S BD
P stands for A [ B. Similarly, a union F
¿; only for disjoint sets, the notation A[B
is pairwise disjoint if and only if either A D B or A \
SB D ¿ for all elements
S A and
B of F ; only for pairwise disjoint sets, the notation P F stands for F .

3.4.3 Unions and Intersections of Sets

Another notation for unions and intersections of a set G proves convenient,

especially where the set G relates in a specific way to a second set F . The new
notation may then “index” the elements of G with the corresponding elements of F .
3.51 Example. For each set F  P.U/ of subsets of a set U, consider the set G
of all the complements U n S of all the elements S 2 F ; thus,

G WD fB 2 P.U/ W 9S Œ.S 2 F / ^ .B D U n S/g :

The indexed notation then denotes the union and the intersection of G as follows:
[ [
.U n S/ WD G;
S2F

\ \
.U n S/ WD G:
S2F

The notation on the left-hand sides avoids the need to write a formula for the set G .
 www.pdfgrip.com

136 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

Because the definitions of the intersection and union of sets rely on conjunction
and disjunction, the union and intersection have formal features similar to the logical
features of the conjunction and disjunction, for instance, distributivity.
3.52 Theorem (De Morgan’s Laws). For each set U, and for each set F 
P.U/ of subsets of U, the complement of the union equals the intersection of the
complements,
[  \
Un F D .U n A/;
A2F

whereas for each set U, and for each nonempty family F  P.U/ of subsets of U,
the complement of the intersection equals the union of the complements,
\  [
Un F D .U n A/:
A2F

Proof. For the complement of the intersection, an informal proof can proceed as
follows.
T T
• For each set X, X 2TU n . F / if and only if X 2 U but X … . F /;
… . F / if and only if there exists A 2 F with X … A;
• by definition, X T
• hence X 2 U n . F / T if and only if X 2 U and there exists A 2 F with X … A;
• equivalently, X 2TU n . F / if and only ifSthere exists A 2 F with X 2 .U n A/;
• hence X 2 U n . F / if and only if X 2 A2F .U n A/.
The foregoing informal proof does not justify the permutation of the two
statements “X 2 U” and “there exists A 2 F ” but the following formal proof
justifies such a permutation by the absence of any free occurrence of A in X 2 U.
T S
U n . F / D A2F .U n A/
m (axiom S1),
˚ T S 
8X ŒX 2 fU n . F /g , ŒX 2 A2F .U n A/
m .n/,
˚ T S 
8X Œ.X 2 U/ ^ f: .X 2 F /g , ŒX 2 A2F .U n A/
S T
m , ,
8X fŒ.X 2 U/ ^ .: f8A Œ.A 2 F / ) .X 2 A/g/
, f9A Œ.A 2 F / ^ f.X 2 U/ ^ Œ:.X 2 A/ggg
:Œ8A.P/;
m
9AŒ:.P/;
8X fŒ.9A f.X 2 U/ ^ f: fŒf:.A 2 F /g _ .X 2 A/ggg/
, Œ9A f.A 2 F / ^ Œ.X 2 U/ ^ Œ:.X 2 A/gg
m De Morgan,
8X fŒ.9A f.X 2 U/ ^ f.A 2 F / ^ Œ:.X 2 A/gg/
, f9A Œ.A 2 F / ^ f.X 2 U/ ^ Œ:.X 2 A/ggg
 www.pdfgrip.com

3.4 Unions and Intersections of Sets 137

which holds by associativity ` f.P/ ^ Œ.Q/ ^ .R/ , fŒ.P/ ^ .Q/ ^ .R/g

(theorem 1.66).
For the complement of the union, an informal proof can proceed as follows.
S S
• For each set X, X 2SU n . F / if and only if X 2 U but X … . F /;
… . F / if and only if X … A for every A 2 F ;
• by definition, X S
• hence X 2 U n . F / S if and only if X 2 U and X … A for every A 2 F ;
• equivalently, X 2SU n . F / if and only T if X 2 .U n A/ for every A 2 F ;
• hence X 2 U n . F / if and only if X 2 A2F .U n A/.
The foregoing informal proof does not justify the permutation of the two statements
“X 2 U” and “for every A 2 F ” because it hides the permutation by placing the
quantifier at the end of the statements. The following formal proof justifies such a
permutation by the absence of any free occurrence of A in X 2 U.
S T
U n . F / D A2F .U n A/
m (S1),
˚ S T 
8X ŒX 2 U n . F / , ŒX 2 A2F .U n A/
m .n/,
˚ S T 
8X Œ.X 2 U/ ^ f: .X 2 F /g , ŒX 2 A2F .U n A/
S T
m , ,
8X fŒ.X 2 U/ ^ f: .9A f.A 2 F / ^ .X 2 A/g/g
 S  
, Œ X 2 A2F .U n A/ ^ 8A f.A 2 F / ) f.X 2 U/ ^ Œ:.X 2 A/gg
m
8X fŒ.8A f.X 2 U/ ^ Œ: ..A 2 F / ^ .X 2 A//g/
T
, Œ.X 2 U n F / ^ 8A fŒ:.A 2 F / _ f.X 2 U/ ^ Œ:.X 2 A/ggg
m
8X fŒ.8A f.X 2 U/ ^ Œ.Œ:.A 2 F / _ Œ:.X 2 A//g/
, .ŒX 2 U ^ 9B Œ.B 2 F / ^ f:.X 2 B/g/
 
^ 8A fŒ:.A 2 F / _ f.X 2 U/ ^ Œ:.X 2 A/gg
*
8X fŒ.8A f.X 2 U/ ^ Œ.Œ:.A 2 F / _ Œ:.X 2 A//g/
, Œ.X 2 U/ ^ 8A fŒ:.A 2 F / _ f.X 2 U/ ^ Œ:.X 2 A/ggg
where the first implication, *, used the assumption that F 6D ¿ for the formula
beginning with 9B : : :; the last line follows from the distributivity of ^ over _. u
t
3.53 Theorem. For all sets F and G , intersection distributes over their unions:
h[  [ i h[ i
` F \ G D fA \ B W .A 2 F / ^ .B 2 G /g :
 www.pdfgrip.com

138 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

Proof. An informal proof can proceed as follows.

S S S S
• For each set X, X 2 Œ. F / \ . G / if and only if X 2 . F / and X 2 . G /;
• equivalently, X belongs to some A 2 F and X belongs to some B 2 G ,
• equivalently, X 2 .A
S\ B/ for some elements A 2 F and B 2 G ;
• equivalently, X 2 Œ fA \ B W .A 2 F / ^ .B 2 G /g.
A formal proof can proceed as follows.
S S
Œ. F / \ . G /
S
D ΠfA \ B W .A 2 F / ^ .B 2 G /g yet unproved,
m extensionality,
S S
8X .fX 2 Œ. F / \ . G /g
S
, fX 2 ΠfA \ B W .A 2 F / ^ .B 2 G /gg/
S
m definitions: \, ,
S S
8X .fŒX 2 . F / ^ ŒX 2 . G /g
  
, 9A 9Bf.A 2 F / ^ .B 2 G / ^ ŒX 2 .A \ B/g
S
m definition of ,
Œ8X .f9AŒ.A 2 F / ^ .X 2 A/g ^ f9BŒ.B 2 G / ^ .X 2 B/g/
  
, 9A 9Bf.A 2 F / ^ .B 2 G / ^ ŒX 2 .A \ B/g
m no free A or B,
Œ8X .9Af9BŒ.A 2 F / ^ .B 2 G / ^ .X 2 A/ ^ .X 2 B/g/
  
, 9A 9Bf.A 2 F / ^ .B 2 G / ^ ŒX 2 .A \ B/g

which holds by definition of A \ B (theorem 3.43). t

u
3.54 Theorem. For all sets F and G , union distributes over their intersections:
h\  \ i h\ i
` F [ G D fA [ B W .A 2 F / ^ .B 2 G /g :

Proof. An informal proof can proceed as follows.

T T T T
• For each set X, X 2 Œ. F / [ . G / if and only if X 2 . F / or X 2 . G /;
• equivalently, X belongs to every A 2 F or X belongs to every B 2 G ,
• equivalently, X 2 .A
T[ B/ for all elements A 2 F and B 2 G ;
• equivalently, X 2 Œ fA [ B W .A 2 F / ^ .B 2 G /g.
A formal proof can proceed as follows.
 www.pdfgrip.com

3.4 Unions and Intersections of Sets 139

T T
Œ. F / [ . G /
T
D ΠfA [ B W .A 2 F / ^ .B 2 G /g unproved,
m (S1),
T T
8X .fX 2 Œ. F / [ . G /g
T
, fX 2 ΠfA [ B W .A 2 F / ^ .B 2 G /gg/
T
m def. [, ,
T T
8X .fŒX 2 . F / _ ŒX 2 . G /g
  
, 8A 8BfŒ.A 2 F / ^ .B 2 G / ) ŒX 2 .A [ B/g
T
m definition: ,
Œ8X .f8AŒ.A 2 F / ) .X 2 A/g _ f8BŒ.B 2 G / ) .X 2 B/g/
  
, 8A 8BfŒ.A 2 F / ^ .B 2 G / ) ŒX 2 .A [ B/g
m no free A, B,
f8X Œ8A .8BfŒ.A 2 F / ) .X 2 A/ _ Œ.B 2 G / ) .X 2 B/g/g
  
, 8A 8Bf.A 2 F / ^ .B 2 G / ^ ŒX 2 .A [ B/g
m Œ.P/^.Q/).P/,

f8X Œ8A .8BfŒ.A 2 F / ) .X 2 A/ _ Œ.B 2 G / ) .X 2 B/g/g

  
, 8A 8BfŒ.A 2 F / ^ .B 2 G / ) Œ.X 2 A/ _ .X 2 B/g
m definition: [,
f8X Œ8A .8BfŒ.A 2 F / ) .X 2 A/ _ Œ.B 2 G / ) .X 2 B/g/g
  
, 8A 8BfŒ.A 2 F / ^ .B 2 G / ) ŒX 2 .A [ B/g
which holds by definition of A [ B (theorem 3.33). t
u
Besides union, intersection, and difference, another operation with sets is useful.
3.55 Definition (symmetric difference). The symmetric difference of any sets A
and B is denoted by A 4 B and defined by

A 4 B WD .A [ B/ n .A \ B/:

3.4.4 Exercises on Unions and Intersections of Sets

3.51 . List all the elements of f2; 3; 7g [ f3; 5; 7g.

3.52 . List all the elements of f4; 6; 8g [ f4; 8; 9g.
3.53 . List all the elements of f2; 3; 7g \ f3; 5; 7g.
3.54 . List all the elements of f4; 6; 8g \ f4; 8; 9g.
 www.pdfgrip.com

140 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.55 . List all the elements of f2; 3; 7gf3; 5; 7g.

3.56 . List all the elements of f4; 6; 8gf4; 8; 9g.
S
3.57 . Provide an example of a set F such that F D F .
T
3.58 . Provide an example of a set G such that G D G .
S
3.59 . Provide an example of a set F such that F 6D F .
T
3.60 . Provide an example of a set G such that G 6D G .
S
3.61 . Provide examples of sets B and F such that B 2 F but B … F .
T
3.62 . Provide examples of sets B and G such that B 2 G but G … B.
3.63 . Provide examples of sets A, B, X, Y such that X 2 A and Y 2 B but .X [ Y/ …
.A [ B/.
3.64 . Provide examples of sets A, B, X, Y such that X 2 A and Y 2 B but .X \ Y/ …
.A \ B/.
S
3.65 . For each set A, prove that fAg D A.
S
3.66 . For each set A, prove that P.A/ D A.
3.67 . Prove that 8A8B8C fŒ.A [ B/ \ C D Œ.A \ C/ [ .B \ C/g.
3.68 . Prove that 8A8B8C fŒ.A \ B/ [ C D Œ.A [ C/ \ .B [ C/g.
3.69 . Prove that 8AŒ.A [ ¿/ D A.
3.70 . Prove that 8AŒ.A \ ¿/ D ¿.
3.71 . Prove that 8AŒ.A [ A/ D A.
3.72 . Prove that 8AŒ.A \ A/ D A.
Prove the following formulae for all subsets A and B of each set U.
3.73 . ŒU n .A \ B/ D .U n A/ [ .U n B/
3.74 . ŒU n .A [ B/ D .U n A/ \ .U n B/
3.75 . Œ.A n B/ n U D .A n U/ n .B n U/
3.76 . ŒU n .A n B/ D .U n A/ [ .U \ B/
3.77 . Œ.A [ B/ n U D .A n U/ [ .B n U/
3.78 . Œ.A \ B/ n U D .A n U/ \ .B n U/
3.79 . ŒU \ .A n B/ D .U n B/ n .U n A/
3.80 . Prove or disprove that .ŒP.A/ [ ŒP.B//  P.A [ B/.
3.81 . Prove or disprove that .ŒP.A/ \ ŒP.B//  P.A \ B/.
3.82 . Prove or disprove that .ŒP.A/ \ ŒP.B//  P.A \ B/.
 www.pdfgrip.com

3.4 Unions and Intersections of Sets 141

3.83 . Prove or disprove that .ŒP.A/ [ ŒP.B//  P.A [ B/.

3.84 . Prove or disprove that .ŒP.A/ n ŒP.B//  P.A n B/.
3.85 . Prove or disprove that .ŒP.A/ n ŒP.B//  P.A n B/.
3.86 . For each nonempty set F  P.U/, and each B  U, prove that
\  \
F [BD .A [ B/:
A2F

3.87 . For each set F  P.U/, and for each subset B  U, prove that
[  [
F \BD .A \ B/:
A2F

3.88 . Prove that A 4 ¿ D A.

3.89 . Prove that A 4 A D ¿.
3.90 . Prove that the symmetric difference is associative: .A4B/4C D A4.B4C/.
3.91 . Prove that the symmetric difference commutes: A 4 B D B 4 A.
3.92 . Prove or disprove that Œ.AC/ [ .BC/  Œ.A [ B/C.
3.93 . Prove or disprove that Œ.AC/ [ .BC/  Œ.A [ B/C.
3.94 . Prove or disprove that Œ.AC/ \ .BC/  Œ.A \ B/C.
3.95 . Prove or disprove that Œ.AC/ \ .BC/  Œ.A \ B/C.
3.96 . Prove or disprove that Œ.AC/ n .BC/  Œ.A n B/C.
3.97 . Prove or disprove that Œ.AC/ n .BC/  Œ.A n B/C.
3.98 . Prove or disprove that Œ.A [ C/.B [ C/  Œ.AB/ [ C.
3.99 . Prove or disprove that Œ.A [ C/.B [ C/  Œ.AB/ [ C.
3.100 . Prove or disprove that Œ.A \ C/.B \ C/  Œ.AB/ \ C.
3.101 . Prove or disprove that Œ.A \ C/.B \ C/  Œ.AB/ \ C.
3.102 . Prove or disprove that Œ.A n C/.B n C/  Œ.AB/ n C.
3.103 . Prove or disprove that Œ.A n C/.B n C/  Œ.AB/ n C.
3.104 . Prove or disprove that Œ.C n A/.C n B/  ŒC n .AB/.
3.105 . Prove or disprove that Œ.C n A/.C n B/  ŒC n .AB/.
3.106 . Prove or disprove that .ŒP.A/ŒP.B//  P.AB/.
3.107 . Prove or disprove that .ŒP.A/ŒP.B//  P.AB/.
 www.pdfgrip.com

142 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

P n A/.
3.108 . Prove that if A \ B D ¿, then A [ B D .A n B/[.B
P \ B/.
3.109 . Prove that A [ B D .A 4 B/[.A
P n A/.
3.110 . Prove that A 4 B D .A n B/[.B

3.5 Cartesian Products and Relations

Beyond logic and sets, much of mathematics consists of connections between types
of sets called Cartesian products, mathematical functions, and mathematical rela-
tions. These types of sets allow for mathematical specifications, analysis, synthesis,
and processing of such concepts as graphs, maps, algorithms, and rankings.

3.5.1 Cartesian Products of Sets

Cartesian products contain certain sets with two elements. Whereas fX; Yg D fY; Xg
for all sets X and Y, however, some situations require a method for listing the
elements of a set in a specific order by means of “ordered pairs” or otherwise, for
example, in geography and in navigation as in figure 3.2. The following definition
(attributed to Wiener and Kuratowski [83, p. 455]) and theorem 3.58 derive such

X,Y

Y,X

Fig. 3.2 If X 6D Y, then .X; Y/ 6D .Y; X/.

 www.pdfgrip.com

3.5 Cartesian Products and Relations 143

ordered pairs from sets, which shows that the concept of ordered pairs does not
require any additional axiom.
3.56 Definition (ordered pairs). For all sets X and Y, the ordered pair .X; Y/ is
the set defined by three applications of the axiom of pairing as
˚ 
.X; Y/ D fXg; fX; Yg :

X is the first coordinate of .X; Y/, and Y is the second coordinate of .X; Y/.
3.57 Example.
˚ 
• If X D 0 and Y D 1, then .X; Y/ D ˚f0g; f0; 1g. ˚ 
• If X D 1 and Y D 0, then .X; Y/ D ˚f1g; f1; 0g D ˚f1g; f0; 1g . ˚ 
• If X D 0 and Y D 0, then .X; Y/ D f0g; f0; 0g D f0g; f0g D f0g .
The following theorem confirms that, in contrast to sets with two elements,
ordered pairs record the order of their coordinates.
3.58 Theorem. For all sets X and Y, if X 6D Y, then .X; Y/ 6D .Y; X/.
Proof. An informal proof can consist in showing that the two sets .X; Y/ and .Y; X/
contain different elements, whence .X; Y/ 6D .Y; X/.
• For all sets X and Y, X 2 fXg and Y 2 fYg by pairing (S3);
• if X 6D Y, then X … fYg and Y … fXg, by pairing (S3);
• hence if X 6D Y, then fXg 6D fYg by extensionality (S1);
• from X … fYg and X 2 fY; Xg it follows that fXg ˚ 6D fY; Xg by
 (S1);
• from fXg 6D˚ fYg , fXg D
6  fY; Xg follows fXg … fYg; fY; Xg D .Y; X/;
• yet fXg 2 fXg; fX; Yg D .X; Y/;
• from fXg 2 .X; Y/ and fXg … .Y; X/ follows .X; Y/ 6D .Y; X/, by (S1).
A formal proof can parallel the same reasoning.
(1) ` 8ZŒ.Z 2 fXg/ , .Z D X/ axiom of pairing (S3),
(2) ` 8ZŒ.Z 6D X/ , .Z … fXg/ contraposition and transposition,
(3) ` .Y 6D X/ , .Y … fXg/ specialization SubfZY ,
(4) ` Y 6D X hypothesis,
(5) ` Y … fXg Detachment;
(6) ` Y 2 fX; Yg axiom of pairing (S3),
(7) ` fXg 6D fX; Yg (5), (6), and extensionality (S1);
(8) ` Y 2 fYg axiom of pairing (S3),
(9) ` fXg 6D fYg (5), (8), and extensionality (S1);
(10) ` fXg … ffYg; fY; Xgg D .Y; X/ (7), (9), and extensionality (S1);
(11) ` fXg 2 ffXg; fX; Ygg D .X; Y/ pairing (S3),
(12) ` .X; Y/ 6D .Y; X/ (10), (11), and extensionality (S1).
t
u
 www.pdfgrip.com

144 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

The definition of the ordered pair .X; Y/ holds for all sets X and Y, in particular,
for all elements X and Y of two sets A and B. The following theorem shows that all
these ordered pairs .X; Y/ are themselves elements of a set.
3.59 Theorem. For all sets A and B, for each X 2 A and for each Y 2 B, the pair
.X; Y/ belongs to PŒP.A [ B/.
Proof. An informal proof can trace back the definition of the ordered pair .X; Y/:
• From X 2 A follows X 2 A [ B, because A  A [ B, whence fXg 2 P.A [ B/;
• from Y 2 B it follows that Y 2 A [ B, because B  A [ B;
• from X 2 A [ B and Y 2 A [ B it follows that fX; Yg 2 P.A [ B/;
˚ 
• from fXg 2 P.A [ B/ and fX; Yg 2 P.A [ B/, it follows that fXg; fX; Yg 2
PŒP.A [ B/.
A formal proof can parallel the foregoing argument.
`X2A hypothesis,
` A  .A [ B/ theorem 3.40 ,
` X 2 .A [ B/ definition of subsets and specialization,
` fXg  .A [ B/ definitions of subsets and singletons,
` fXg 2 P.A [ B/ definition of power sets.
`Y2B hypothesis,
` Y 2 .A [ B/ as for X 2 .A [ B/,
` fX; Yg 2 P.A [ B/ X 2 .A [ B/ and Y 2 .A [ B/,
` f fXg; fX; Yg g  P.A [ B/ fXg 2 P.A [ B/, fX; Yg 2 P.A [ B/,
` f fXg; fX; Yg g 2 PŒP.A [ B/ definition of power set.
t
u
Because every ordered pair .X; Y/ belongs to the set PŒP.A [ B/, the axiom
of separation guarantees that the collection of all such ordered pairs is a set.
3.60 Definition (Cartesian product). For all sets A and B, the Cartesian product
of A and B is the set A B (read “A cross B”), consisting of all ordered pairs .X; Y/
with X 2 A and Y 2 B. Thus,

A B
n  o
D C 2 P ŒP.A [ B/ W 9X 9Yf.X 2 A/ ^ .Y 2 B/ ^ ŒC D .X; Y/g

D f.X; Y/ W .X 2 A/ ^ .Y 2 B/g
 
with 9X 9Yf.X 2 A/ ^ .Y 2 B/ ^ ŒC D .X; Y/g for P in the axiom of separation
(S5). The sets A and B are the factors of the Cartesian product A B.
 www.pdfgrip.com

3.5 Cartesian Products and Relations 145

A common graphical representation of a Cartesian product A B lists all the

elements of A along a horizontal axis, and all the elements of B along a vertical axis,
so that the element .X; Y/ of A B appears directly above X and across from Y.
3.61 Example. For the sets

A WD f0; 1; 2g;
B WD f0; 1g;

the Cartesian product A B takes the form

A B D f.0; 0/; .1; 0/; .2; 0/; .0; 1/; .1; 1/; .2; 1/g;

with the following graphical representation:

B A B

1 .0; 1/ .1; 1/ .2; 1/

0 .0; 0/ .1; 0/ .2; 0/

0 1 2 A

For practice and for future use, the following theorems establish relations
between Cartesian products and other operations with sets.
3.62 Theorem. For all sets A, B, C, and D,

Œ.A \ C/ .B \ D/ D Œ.A B/ \ .C D/:

Proof. An informal proof can establish that Œ.A \ C/ .B \ D/ and Œ.A B/ \
.C D/ have exactly the same elements. These two sets are Cartesian products,
and consequently their elements are ordered pairs.
• An ordered pair .X; Y/ is an element of .A\C/ .B\D/ if and only if X 2 .A\C/
and Y 2 .B \ D/;
• hence .X; Y/ 2 Œ.A \ C/ .B \ D/ if and only if X 2 A and X 2 C, and Y 2 B
and Y 2 D,
• which is equivalent to X 2 A and Y 2 B, and X 2 C and Y 2 D,
• which is equivalent to .X; Y/ 2 .A B/ and .X; Y/ 2 .C D/,
• which is thus equivalent to .X; Y/ 2 Œ.A B/ \ .C D/.
As the preceding informal proof swapped X 2 C and Y 2 B, a formal proof can rely
on the commutativity Œ.P/ ^ .Q/ , Œ.Q/ ^ .P/ of ^ by theorem 1.57:
 www.pdfgrip.com

146 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

.X; Y/ 2 Œ.A \ C/ .B \ D/

m definition of Cartesian products,
ŒX 2 .A \ C/ ^ ŒY 2 .B \ D/
m definition of intersection,
Œ.X 2 A/ ^ .X 2 C/ ^ Œ.Y 2 B/ ^ .Y 2 D/
m commutativity and associativity of ^,
Œ.X 2 A/ ^ .Y 2 B/ ^ Œ.X 2 C/ ^ .Y 2 D/
m definition of Cartesian products,
Œ.X; Y/ 2 .A B/ ^ Œ.X; Y/ 2 .C D/
m definition of intersection.
.X; Y/ 2 Œ.A B/ \ .C D/
t
u
3.63 Theorem. For all sets A, B, C, and D,

Œ.A B/ [ .C D/ [ .A D/ [ .C B/ D Œ.A [ C/ .B [ D/:

Proof. An informal proof can establish that Œ.A B/ [ .C D/ [ .A D/ [ .C B/

and Œ.A [ C/ .B [ D/ have exactly the same elements.
• An ordered pair .X; Y/ is an element of .A B/ [ .C D/ [ .A D/ [ .C B/
if and only if .X; Y/ 2 .A B/, or .X; Y/ 2 .C D/, or .X; Y/ 2 .A D/, or
.X; Y/ 2 .C B/,
• which is equivalent to X 2 A and Y 2 B, or X 2 C and Y 2 D, or X 2 A and
Y 2 D, or X 2 C and Y 2 B,
• which is equivalent to X 2 A or X 2 C, and Y 2 B or Y 2 D,
• which is equivalent to .X; Y/ 2 Œ.A [ C/ .B [ D/.
A formal proof translates the alleged equivalences into the logical theorem

fŒ.P/ ^ .Q/ _ Œ.P/ ^ .S/ _ Œ.R/ ^ .Q/ _ Œ.R/ ^ .S/g , fŒ.P/ _ .R/ ^ Œ.Q/ _ .S/g;

which follows from the distributivity of ^ over _ (theorem 1.75), with

.P/ , .X 2 A/; .Q/ , .Y 2 B/; .R/ , .X 2 C/; .S/ , .Y 2 D/:

Thus,
.X; Y/ 2 Œ.A B/ [ .C D/ [ .A D/ [ .C
B/
S
m definition of ,
Œ.X; Y/ 2 .A B/ _ Œ.X; Y/ 2 .C D/
_Œ.X; Y/ 2 .A D/ _ Œ.X; Y/ 2 .C B/
m definition of ,
Œ.X 2 A/ ^ .Y 2 B/ _ Œ.X 2 C/ ^ .Y 2 D/
_Œ.X 2 A/ ^ .Y 2 D/ _ Œ.X 2 C/ ^ .Y 2 B/_
m theorem 1.75,
 www.pdfgrip.com

3.5 Cartesian Products and Relations 147

Œ.X 2 A/ _ .X 2 C/ ^ Œ.Y 2 B/ _ .Y 2 D/

S
m definition of ,
ŒX 2 .A [ C/ ^ ŒY 2 .B [ D/
m definition of .
.X; Y/ 2 Œ.A [ C/ .B [ D/
t
u

3.5.2 Cartesian Products of Unions and Intersections

The foregoing theorems generalize to Cartesian products of unions or intersections

of any sets of sets.
3.64 Theorem. For all sets of sets F and G ,
[  [  [
F G D .A B/:
.A;B/2F G

S S S
Proof. An informal proof can show that . F / . G / and .A;B/2F G .A B/
have exactly the same elements.
S S S
• A pairS.X; Y/ is an element of . F / . G / if and only if X 2 . F / and
Y 2 . G /,
• which is equivalent to X 2 A for some A 2 F and Y 2 B for some B 2 G ,
• which is equivalent to .X; Y/ 2 .A B/ for some .A; B/ 2 .F G /.
A formal proof can parallel the foregoing reasoning.
S S
.X; Y/ 2 . F / . G /
m definition of ,
S S
ŒX 2 . F / ^ ŒY 2 . G /
S
m definition of ,
f9AŒ.A 2 F / ^ .X 2 A/g ^ f9BŒ.B 2 G / ^ .Y 2 B/g
 m no B, no A,
9A 9BfŒ.A 2 F / ^ .X 2 A/ ^ Œ.B 2 G / ^ .Y 2 B/g
 m properties of ^,
9A 9BfŒ.A 2 F / ^ .B 2 G / ^ Œ.X 2 A/ ^ .Y 2 B/g
 m definition of ,
9A 9BfŒ.A B/ 2 .F G / ^ Œ.X; Y/ 2 .A B/g
S
m definition of .
S
.X; Y/ 2 .AB/2.F G / .A B/
t
u
 www.pdfgrip.com

148 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.65 Theorem. For all sets of sets F and G ,

\  \  \
F G D .A B/:
.A;B/2F G

T T T
Proof. An informal proof can show that . F / . G / and .A;B/2F G .A B/
have exactly the same elements.
T T T
• A pairT.X; Y/ is an element of . F / . G / if and only if X 2 . F / and
Y 2 . G /,
• which is equivalent to X 2 A for every A 2 F and Y 2 B for every B 2 G ,
• which is equivalent to .X; Y/ 2 .A B/ for every .A; B/ 2 .F G /.
The foregoing informal proof glosses over the case in which at least one of F or G
is empty, which would require invoking the corresponding unions as supersets of the
intersections, because of the definition of intersections. One remedy could consist in
proving such cases separately. Indeed, F D ¿ or G D ¿ if and only if .F G / D
¿. A formal proof can parallel the foregoing reasoning with theorem 1.82,

` fŒ.P/ ) .Q/ ^ Œ.R/ ) .S/g ) fŒ.P/ ^ .R/ ) Œ.Q/ ^ .S/g;

to prove that
\  \  \
F G  .A B/;
.A;B/2F G

with the hypotheses C 2 F for P and D 2 G for R for the converse, so that
 
` fŒ.P/ ) .Q/ ^ Œ.R/ ) .S/g ( Œ.P/ ^ .R/ ^ fŒ.P/ ^ .R/ ) Œ.Q/ ^ .S/g :

Thus,
T T
.X; Y/ 2 . F/ . G/
m definition of ,
T T
ŒX 2 . F / ^ ŒY 2 . G /
T S
m H  H,
 
f8AŒ.A 2 F / ) .X 2 A/g ^ f8BŒ.B 2 G / ) .Y 2 B/g
 
^ f9CŒ.C 2 F / ^ .X 2 C/g ^ f9DŒ.D 2 G / ^ .Y 2 D/g
m no B, no A,
  
8A 8BfŒ.A 2 F / ) .X 2 A/ ^ Œ.B 2 G / ) .Y 2 B/g
  
^ 9C 9DfŒ.C 2 F / ^ .X 2 C/ ^ .D 2 G / ^ .Y 2 D/g
m theorems,
 www.pdfgrip.com

3.5 Cartesian Products and Relations 149

  
8A 8BfŒ.A 2 F / ^ .B 2 G / ) Œ.X 2 A/ ^ .Y 2 B/g
  
^ 9C 9DfŒ.C 2 F / ^ .X 2 C/ ^ .D 2 G / ^ .Y 2 D/g
m definition of ,
  
8A 8BfŒ.A; B/ 2 .F G / ) Œ.X; Y/ 2 .A B/g
  
^ 9C 9DfŒ.C; D/ 2 .F G / ^ Œ.X; Y/ 2 .C D/g
T
m definition of .
T
.X; Y/ 2 .AB/2.F G / .A B/g
t
u

3.5.3 Mathematical Relations and Directed Graphs

The Cartesian product A B provides a means to draw connections, or, in other

words, to specify relations, between elements of the two sets A and B.
3.66 Definition (relation). A relation between elements of sets A and B is a subset
R  A B of their Cartesian product.
• Two elements X 2 A and Y 2 B are related with respect to the relation R if and
only if .X; Y/ 2 R.
• For each relation R  A B, the domain D.R/  A of the relation R consists of
those elements of A related by R to at least one element of B:
˚ 
D.R/ D X 2 A W 9Yf.Y 2 B/ ^ Œ.X; Y/ 2 Rg :

• Similarly, the range R.R/  B of the relation R consists of those elements of B

related by R to at least one element of A:
˚ 
R.R/ WD Y 2 B W 9Xf.X 2 A/ ^ Œ.X; Y/ 2 Rg :

Another common notation for .X; Y/ 2 R is XRY, especially if such a special symbol
as , ¤, or D denotes the relation. Relations may also be represented by graphs.
3.67 Definition (directed graph). A directed graph is an ordered pair G WD
.V; E/ with a relation E  V V between elements of the same set A D V D B.
Elements of V are called vertices. A pair .X; Y/ 2 E is called an edge from X to Y:
(X,Y)
X Y

3.68 Example. For each set A, the diagonal A  A A is the subset

A D f.X; X/ 2 A A W X 2 Ag:
 www.pdfgrip.com

150 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

Thus, the diagonal corresponds to the relation called “equality,” or, equivalently,
“identity”: a pair of elements .X; Y/ lies on the diagonal A if and only if X D Y.
Because X D X, and hence .X; X/ 2 A , for every X 2 A, it follows that the
diagonal A relates every element of A to itself, whence D.R/ D A D R.R/. For
example, if
˚ 
A WD ¿; f¿g ;

then

.¿; f¿g/ .f¿g; f¿g/

A AD ;
.¿; ¿/ .f¿g; ¿/

.f¿g; f¿g/
A D ;
.¿; ¿/

because the only pairs .X; Y/ with X D Y are the pairs .¿; ¿/ and .f¿g; f¿g/. As a
graph, the diagonal consists of a single loop at each vertex:

(X, X) X

3.69 Example. For all sets H and K, consider their respective power sets A WD
P.H/ and B WD P.K/. The relation R  P.H/ P.K/ of inclusion consists
of all, but only those, pairs .V; W/ of subsets V  H and W  K such that V  W:

R D f.V; W/ 2 P.H/ P.K/ W .V  H/ ^ .W  K/ ^ .V  W/g:

The domain of the relation  consists of all subsets of A included as a subset in at

least one subset of B. Similarly, the range of the relation  consists of all subsets of
B that contain as a subset at least one subset of A.
3.70 Example. For all sets H and K, let A WD P.H/ and B WD P.K/. The relation
S  P.H/ P.K/ of strict inclusion consists of all, but only those, pairs .V; W/
of subsets V  H and W  K such that V  W but V 6D W:

S D f.V; W/ 2 P.H/ P.K/ W .V  H/ ^ .W  K/ ^ .V  W/ ^ .V 6D W/g:

The strict inclusion .V  W/ ^ .V 6D W/, is also denoted by V  W or by V ¤ W:

` Œ.V  W/ ^ .V 6D W/ , .V  W/ , .V ¤ W/:

The domain of the relation of strict inclusion consists of all subsets of A included
in, but not equal to, at least one subset of B; its range consists of all subsets of B that
contain, but do not coincide with, at least one subset of A.
 www.pdfgrip.com

3.5 Cartesian Products and Relations 151

3.71 Example. For all sets A and B, the relational constant 2 relates every element
X 2 A that is an element of some Y 2 B. Denote this relation on A B by E.
(The symbol 2 cannot denote a subset E of A B because 2 is neither an individual
variable nor an individual constant.) Thus, E  .A B/ is defined by

Œ.X; Y/ 2 E , Œ.X 2 A/ ^ .X 2 Y/ ^ .Y 2 B/:

For example, if
˚ 
A WD ˚ ¿; f¿g ;
B WD ¿; f¿g ;

then

.¿; f¿g/ .f¿g; f¿g/

A BD ;
.¿; ¿/ .f¿g; ¿/

.¿; f¿g/
ED ;

because in A B the only pair .X; Y/ such that X 2 Y is the pair .¿; f¿g/.
3.72 Example. For the sets
˚ 
A WD n ¿; f¿g ;
˚ o
B WD ¿; f¿g; ¿; f¿g ;

the Cartesian product A B and the relation E take the forms

8  ˚   ˚  9
ˆ
< ¿; ¿; f¿g f¿g; ¿; f¿g >
=
   
A BD ¿; f¿g f¿g; f¿g ;
:̂ >
;
.¿; ¿/ .f¿g; ¿/
8 ˚  ˚  9
< .¿; ¿; f¿g / .f¿g; ¿; f¿g / =
ED .¿; f¿g/ ;
: ;

because in A B the only pairs .X; Y/ such that X 2 Y are those just displayed.
In some contexts a relation R  A B may also prove useful with B and A listed
in the reverse order. Then the “inverse” relation Rı1  B A contains similar pairs
as R does but with coordinates listed in the reverse order. Some texts denote the
inverse relation by R1 , which can cause confusion because the same notation also
represents reciprocals in arithmetic and algebra. The notation adopted here for the
inverse relation, Rı1 , conforms to [70].
 www.pdfgrip.com

152 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.73 Definition (inverse relation). The inverse of a relation R  A B between

two sets A and B is the relation Rı1  B A defined by

Rı1 D f.Y; X/ 2 B A W .X; Y/ 2 Rg:

3.74 Example. The inverse of the diagonal is the same diagonal.

3.75 Example. The inverse of the subset relation  is the superset relation .
3.76 Definition (composite relation). The composition of two relations R  A B
and S  B C is the relation S ı R on A C defined as follows. The composite
relation S ı R relates an element U 2 A to an element W 2 C if and only if R relates
U to some element V 2 B and S relates the same element V to W 2 C:

.S ı R/ D f.U; W/ 2 .A C/ W 9V.ŒV 2 B ^ Œ.U; V/ 2 R ^ Œ.V; W/ 2 S/g:

3.77 Example. For each relation R  A B between any sets A and B, the
composition R ı Rı1 contains the diagonal 4R.R/ for the range of R. Indeed, by
definition of its range, R relates every Y 2 R.R/ to an element X 2 A so that
.X; Y/ 2 R; then the inverse Rı1 relates Y to X so that .Y; X/ 2 Rı1 . From
.Y; X/ 2 Rı1 and .X; Y/ 2 R it follows that .Y; Y/ 2 .R ı Rı1 / for every
.Y; Y/ 2 4R.R/ . Therefore 4R.R/  .R ı Rı1 /. Similarly, Rı1 ı R contains the
diagonal 4D.R/ for the domain of R. Indeed, by definition of its domain, R relates
every X 2 D.R/ to an element Y 2 B so that .X; Y/ 2 R; then the inverse Rı1
relates Y to X so that .Y; X/ 2 Rı1 . From .X; Y/ 2 R and .Y; X/ 2 Rı1 it follows
that .X; X/ 2 .Rı1 ı R/ for every .X; X/ 2 4D.R/ . Therefore 4D.R/  .Rı1 ı R/.
Because every relation R between sets A and B is a subset R  .A B/, operations
with sets apply to all relations.
3.78 Definition (unions and intersections of relations). For all sets A, B, C, E,
and for all relations R  A B and T  C E, the union of the relations R and T
is the relation R [ T between A [ C and B [ E, so that

.R [ T/ WD f.X; Y/ 2 Œ.A [ C/ .B [ E/ W Œ.X; Y/ 2 R _ Œ.X; Y/ 2 Tg:

Similarly, the intersection of the relations R and T is the relation R \ T between

A \ C and B \ E, so that

.R \ T/ D f.X; Y/ 2 Œ.A \ C/ .B \ E/ W Œ.X; Y/ 2 R ^ Œ.X; Y/ 2 Tg:

A particular instance of intersections of relations R  .A B/ and T  .A B/

consists of the intersection of R with a subset S  A and its Cartesian product with
B. The “restriction” of R to S is then the relation R\T with T D .S B/. The concept
of the “restriction” of a relation to a subset is useful if the subset has characteristics
that are useful for the purpose at hand while the complement of the subset does not.
 www.pdfgrip.com

3.5 Cartesian Products and Relations 153

3.79 Definition (restricted relation). For each relation R  .A B/, and for each
subset S  A of A, the restriction of R to S is the relation RjS  .S B/ defined by

RjS D R \ .S B/
D f.X; Y/ 2 R W X 2 Sg:

Thus, RjS restricts its first coordinates only to those elements of S.

There also exists a similar instance of intersections of relations R  .A B/ and
T  .A B/ as the intersection of R with a subset V  B and its Cartesian product
with A. The “restriction” of R to V is then the relation R \ T with T D .A V/.
Similarly, the restriction of R to S  A and V  B is the relation R \ .S V/.

3.5.4 Exercises on Cartesian Products of Sets

3.111 . Determine whether the Cartesian product is associative.

3.112 . For each set A, prove that A ¿ D ¿, and that ¿ A D ¿.
3.113 . Prove that A B D ¿ if and only if A D ¿ or B D ¿.
3.114 . Prove that distributes over \: Œ.A B/ \ .C B/ D Œ.A \ C/ B:
3.115 . Prove that distributes over [: Œ.A B/ [ .C B/ D Œ.A [ C/ B:
3.116 . Prove that Œ.A n C/ B D Œ.A B/ n .C B/.
3.117 . Prove that ŒA .B n D/ D Œ.A B/ n .A D/.
3.118 . Prove or disprove that Œ.A n C/ .B n D/  Œ.A B/ n .C D/.
3.119 . Prove or disprove that Œ.A n C/ .B n D/  Œ.A B/ n .C D/.
3.120 . Prove or disprove that Œ.AC/ .BD/  Œ.A B/.C D/.
3.121 . Prove or disprove that Œ.AC/ .BD/  Œ.A B/.C D/.
3.122 . Prove or disprove that .ŒP.A/ ŒP.B//  P.A B/.
3.123 . Prove or disprove that .ŒP.A/ ŒP.B//  P.A B/.
3.124 . Provide a formula for the inverse of the relation of inclusion.
3.125 . Give a formula for the inverse of the relation of strict inclusion.
3.126 . For each relation R  A B, and for each subset S  A of A, prove that
RjS D R \ .S B/.
3.127 . For all sets A and B, prove that ¿ is an element of the domain D of the
relation  on P.A/ P.B/.
 www.pdfgrip.com

154 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.128 . For all sets A and B, prove that P.B/ is the range R of the relation  on
P.A/ P.B/.
3.129 . Provide examples of sets A and B, such that P.A/ is the domain D of the
relation  on P.A/ P.B/.
3.130 . Provide examples of sets A and B, such that P.A/ is not the domain D of
the relation  on P.A/ P.B/, so that D ¤ P.A/.

3.6 Mathematical Functions

Functions are relations relating exactly one element of their domain to each element
of their range.

3.6.1 Mathematical Functions

In some applications of relations, the domain and the range can contain measure-
ments.
3.80 Example. Results from astronomical observations can consist of a relation
between two coordinates of position, with ordered pairs .X; Y/ where X is the
observed ascension (elevation) and Y is the observed declination (azimuth) of an
asteroid. For example, the following pairs .X; Y/ record the ascension X and the
declination Y of the asteroid Pallas, measured by Baron von Zach about 1800 A.D.
[14, p. 5]:

.0; 408/ .30; 89/ .60; 66/ .90; 10/ .120; 338/ .150; 807/
.180; 1238/ .210; 1511/ .240; 1583/ .270; 1462/ .300; 1183/ .330; 804/

with the corresponding graphical representation in figure 3.3.

Results from astronomical observations can also consist of a relation between
time and position, with ordered pairs .T; Y/ where Y is the observed position
(declination or azimuth) of a planet at time T. Such a relation has the following
properties.
• If no observation was made at some time T, then the results do not contain any
ordered pair with T as their first coordinate.
• Each observation yields only one position Y at any time T.
• Observations can yield the same position Y at several times, for instance, if the
motion is periodic.
Mathematical “functions” are relations corresponding to such applications.
 www.pdfgrip.com

3.6 Mathematical Functions 155

Positions (+) of Pallas

1583
1511
1462

1238
1183
Declination [minutes]

806

408
338

89
10
-66

0 30 60 90 120 150 180 210 240 270 300 330

Ascension [degrees]

Fig. 3.3 Positions include at most one declination for each ascension.

3.81 Definition. A function from a set A to a set B is a relation F  .A B/ such

that for each X 2 A there exists at most one Y 2 B for which .X; Y/ 2 F.
The domain of F is the subset D.F/  A consisting of every X 2 A for which
there exists some Y 2 B such that .X; Y/ 2 F.
˚ 
D.F/ D X 2 A W 9YŒ.Y 2 B/ ^ f.X; Y/ 2 Fg :

The range of F, denoted by R.F/, consists of every Y 2 B such that there exists
some X 2 A such that .X; Y/ 2 F:
˚ 
R.F/ D Y 2 B W 9XŒ.X 2 A/ ^ f.X; Y/ 2 Fg :

Moreover, B is called the co-domain of F.

For each X 2 D.F/, the unique Y such that .X; Y/ 2 F is called the value of F
at X, or also the image of X by F. The same Y is denoted by F.X/ (read “F of X”);
thus, Y D F.X/ if and only if .X; Y/ 2 F. The element X is called the argument of
F in the expression F.X/. The notation

FW A!B

(read “F maps A to B”) means that F is a function from A to B; the notation X 7!

F.X/ (read “F maps X to F.X/”) may specify F.X/ by a formula or otherwise.
 www.pdfgrip.com

156 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.82 Remark. A variant of the concept of a function from a set A to a set B is a

subset F  .A B/ such that for each X 2 A there exists exactly one Y 2 B for
which .X; Y/ 2 F, and then the notation F W A ! B implies that D.F/ D A.
The requirement that D.F/ D A remains harmless with simple examples of
functions, but it presents unnecessary obstacles with more realistic examples of
functions, whose complexity can make the domain difficult or impossible to identify,
especially if the identification of the domain is irrelevant to the task at hand.
The specification with “at most” in definition 3.81 is common in set theory
[128, p. 58 & 86], algebraic geometry (especially with “rational” functions)
[136, p. 34–35], complex analysis (especially with “meromorphic” functions) [1,
p. 128], and functional analysis [130, p. 4 & 18], in instances where several
functions called “operators” have different domains but arise from a relation
common to all of them.
3.83 Example. The pairs in example 3.80 define a function F W A ! B from the
set of ascensions A WD f0; 30; 60; 90; 120; 150; 180; 210; 240; 270; 300; 330g to the
set of declinations B WD f 66; 10; 89; 338; 408; 804; 807; 1183; 1238; 1462; 1511;
1583g. This function F W A ! B has domain A, codomain B, and range B.
3.84 Example. With A WD f6; 7; 8; 9g and B WD f1; 2; 3; 4g, let

F WD f.6; 2/; .7; 1/; .8; 2/; .9; 3/g:

Then F W A ! B is a function with domain A, codomain B, and range f1; 2; 3g.

3.85 Example. For all nonempty sets A and B, and for any element Z 2 B, there
exists a constant function
CZ W A ! B;
X 7! Z:
The constant function CZ maps every element X 2 A to the same value Z, so that

CZ WD f.X; Z/ W X 2 Ag:

In particular, D.CZ / D A for the domain of CZ , and R.CZ / D fZg for its range.
3.86 Example. For each set A, there exists an identity function

IA W A ! A;
X 7! X:

Thus the identity function IA maps every element to itself, so that

IA WD f.X; X/ W X 2 Ag:

In particular, D.IA / D A and R.IA / D A, because .X; X/ 2 IA for every X 2 A. The

function IA is also denoted by A and is called the the “diagonal” of A A.
 www.pdfgrip.com

3.6 Mathematical Functions 157

3.87 Example. For all sets A and B, the canonical projection functions from the
Cartesian product A B into its factors A and B are the functions PA and PB with

PA W .A B/ ! A;
.X; Y/ 7! XI

PB W .A B/ ! B;
.X; Y/ 7! Y:

Thus, PA maps .X; Y/ to its first coordinate X in A, whereas PB maps .X; Y/ to its
second coordinate Y in B. The domain of PA and PB is A B, but the range of PA is
A, whereas the range of PB is B.
3.88 Example. For all sets A and B, for any subset V  A and any element Z 2 B,
the slice function SV;Z maps each element X 2 V to .X; Z/ 2 .A B/:

SV;Z W V ! .A B/;
X 7! .X; Z/:

Thus, SV;Z maps its domain V to the slice V fZg in A B.

3.89 Example. For each set A and each subset S  A, the characteristic function
S , with the Greek letter  (read “chi”), maps every element of the subset S to 1, and
every element outside the subset S to 0, (with 0 D ¿, 1 D f¿g, and 2 D f0; 1g):

S W A ! 2;
1 if X 2 S,
X 7!
0 if X … S.

The following theorem provides a means to compare two functions to each other.
3.90 Theorem. Two functions F W A ! B and G W A ! B are equal, F D G, if and
only if they have the same domain D  A and F.X/ D G.X/ for every X 2 D.
Proof. This proof rewrites F.X/ D G.X/ in terms of the definition of functions:
FDG
 m extensionality,
8X 8Y fŒ.X; Y/ 2 F , Œ.X; Y/ 2 Gg
 m functional notation.
8X 8Y fŒY D F.X/ , ŒY D G.X/g
t
u
Some situations involve only parts of a function, or combinations of several
functions, for instance, as defined by the concepts introduced here.
 www.pdfgrip.com

158 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.91 Definition (restriction of functions). For each function F W A ! B, and for

each subset S  A of A, the restriction of F to S is the function FjS  S B
defined by

FjS W S ! B;
X 7! F.X/:

Thus, FjS .X/ D F.X/, but FjS maps only the elements of S \ D.F/.
3.92 Example. With A WD f6; 7; 8; 9g and B WD f1; 2; 3; 4g, let

F WD f.6; 2/; .7; 1/; .8; 2/; .9; 3/g:

Then the restriction of F to the subset S WD f6; 8g is FjS D f.6; 2/; .8; 2/g.
3.93 Example. For each set B, and for each subset W  B, the inclusion function,
denoted by the Greek letter  (“iota”), is the restriction of the identity function to W:

W W W ! B;
X 7! X:

Thus W W W ! B is the restriction of IB W B ! B to the subset W  B.

3.94 Definition (union of functions). For all disjoint sets A and C, so that A \ C D
¿, for all sets B and E, and for all functions F W A ! B and G W C ! E, the union
P of the two sets F  A B and G  C E:
of the functions F and G is the union F [G

P W .A[C/
F [G P ! .B [ E/;

F.X/ if X 2 D.F/  A,
X 7!
G.X/ if X 2 D.G/  C.

3.95 Example (overloading operators). In a computer language such as C++

“overloading” the addition .C/ from numbers to pairs of numbers corresponds to
forming the union of the addition function F defined on a set A of numbers and an
addition function G defined on a set C of pairs of numbers.
3.96 Definition (intersection of functions). For all sets A, B, C, E, and all
functions F W A ! B and G W C ! E, the intersection of the functions F and
G is the intersection F \ G of the two sets F  A B and G  C E, so that

F \ G W D ! .B \ E/;
X 7! F.X/ D G.X/;

with domain D WD D.F \ G/ D fX 2 A \ C W F.X/ D G.X/g.

 www.pdfgrip.com

3.6 Mathematical Functions 159

3.6.2 Images and Inverse Images of Sets by Functions

Some situations involve the images by functions of not only single elements in the
domain, but also subsets of the domain, as defined here.
3.97 Definition (image of a set). For each function F W A ! B and each subset
V  A, the image of V by F is the subset F”.V/ consisting of all images of each
X 2 V \ D.F/ by F:

F”.V/ D fY 2 B W 9X Œ.X 2 V/ ^ .F.X/ D Y/g

D fF.X/ W X 2 V \ D.F/g :

For each function F W A ! B, F” is a function of subsets: F” W P.A/ ! P.B/.

3.98 Remark. The notation F”.V/ adopted in definition 3.97 is common in set
theory [74, p. 14], [128, p. 65]. Informal usage employs the notation F.V/ for the
image of a subset V  A by a function
˚ F WA ! B, but this usage is ambiguous. For
example, consider the set A WD ¿; f¿g and the constant function
˚  ˚ 
C¿ W ¿; f¿g ! ¿; f¿g ;
X 7! ¿:

The set f¿g is an element of A, whence C¿ .f¿g/ D ¿ because C¿ .X/ D ¿ for

every X 2 A. Yet f¿g is also a subset of A, containing the single element ¿ 2 f¿g;
because C¿ .¿/ D ¿ it follows that C”¿ .f¿g/ D f¿g as the image of a subset:

C¿ .f¿g/ D ¿;
C”¿ .f¿g/ D f¿g:

The common informal notation leads to the contradiction

C¿ .f¿g/ D ¿;
C¿ .f¿g/ D f¿g:

In a formal theory containing this contradiction, every proposition would be True.

3.99 Example. If CZ W A ! B is a constant function that maps each X 2 A to the
same Z 2 B, then CZ ”.V/ D fZg for each nonempty subset V  A.
3.100 Example. If IA W A ! A, X 7! X is an identity function, so that IA .X/ D X.
for each X 2 A then IA ”.V/ D V for each subset V  A. Thus, IA ” is the identity
function IA ” D IP.A/ W P.A/ ! P.A/, with V 7! V for every V 2 P.A/.
Besides images of subsets, such problems as the solution of equations involve
the identification of a subset, called a pre-image, mapped to a specified image.
 www.pdfgrip.com

160 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.101 Definition (pre-image). For each function F W A ! B, and for each element
Y 2 B, the pre-image of Y by F is the subset of A denoted by F ı1 ”.fYg/ and
consisting of all elements X 2 A such that F.X/ D Y:

F ı1 ”.fYg/ D fX 2 A W F.X/ D Yg :

For each subset W  B, the inverse image, or pre-image, of W by F is the subset

of F ı1 ”.W/ in A and consisting of all pre-images of all elements of V by F:

F ı1 ”.W/ D fX 2 A W F.X/ 2 Wg :

3.102 Example. If CZ W A ! B is a constant function that maps each X 2 A to the

same Z 2 B, then CZı1 ”.fZg/ D A.
Also, CZı1 ”.W/ D A for each subset W  B for which Z 2 W.
In contrast, CZı1 ”.S/ D ¿ for each subset S  B for which Z … S. Thus,

A if Z 2 W,
CZı1 ”.W/ D
¿ if Z … W.

3.103 Example. If IA W A ! A is an identity function, with IA .X/ D X for each

X 2 A, then IAı1 ”.W/ D W for each subset W  A.
Theorem 3.104 relates images and pre-images to unions and intersections.
3.104 Theorem. For each function F W A ! B, for each set F of subsets of A, and
for each set G of subsets of B, the following relations hold.

[  [
F ı1 ” G D F ı1 ”.W/
W2G
\  \
F ı1 ” G D F ı1 ”.W/
W2G
[  [
F” F D F”.V/
V2F
\  \
F” F  F”.V/:
V2F

S
Proof. Apply the definitions of union and intersection. For F ı1 ” . G /,
S S
F ı1 ” . G / D W2G F ı1 ”.W/ yet unproved,
m extensionality (S1),
˚ ı1
S 
8X X 2 F ” . G /
 S 
, X 2 W2G F ı1 ”.W/
 www.pdfgrip.com

3.6 Mathematical Functions 161

S
m definitions of F ı1 ” and ,
S
8X fŒF.X/ 2 .
G /
 
, 9W f.W 2 G / ^ ŒF.X/ 2 Wg
S
m definition of ,
˚ 
8X 9W f.W 2 G / ^ ŒF.X/ 2 Wg
 
, 9W f.W 2 G / ^ ŒF.X/ 2 Wg
T
which holds thanks to theorem 1.63: .P/ , .P/. For F ı1 ” . G /,
T T
F ı1 ” . G / D W2G F ı1 ”.W/ yet unproved,
m extensionality (S1),
˚ ı1

T
8X X 2 F ” . G /
 T 
, X 2 W2G F ı1 ”.W/
T
m definitions of F ı1 and ,
T
8X fŒF.X/ 2 . G / ,
 
8W f.W 2 G / ) ŒF.X/ 2 Wg
T
m definitions of ,
˚ 
8X 8W f.W 2 G / ) ŒF.X/ 2 Wg
 
, 8W f.W 2 G / ) ŒF.X/ 2 Wg
S
which holds thanks to theorem 1.63: .P/ , .P/. For F” . F /,
S S
F” . F / D V2F F”.V/ yet unproved,
m extensionality (S1),
˚ S  S 
8Y ŒY 2 F” . F / , Y 2 V2F F”.V/
S
m definitions: F”, ,
 ˚ S 
8Y 9X ŒX 2 . F / ^ ŒY D F.X/
 
, 9V fŒV 2 F  ^ ŒY 2 F”.V/g
S
m definitions: , F”,
 
8Y 9X f9VŒ.V 2 F / ^ .X 2 V/g ^ ŒY D F.X/
˚   
, 9V .V 2 F / ^ 9Xf.X 2 V/ ^ ŒY D F.X/g
m no X in V 2 F ,
 
8Y 9X f9VŒ.V 2 F / ^ .X 2 V/g ^ ŒY D F.X/
 ˚  
, 9V 9X .V 2 F / ^ f.X 2 V/ ^ ŒY D F.X/g
T
which holds thanks to the associativity of ^ (theorem 1.66). For F” . F /,
 www.pdfgrip.com

162 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

T T
F” . F/  V2F F”.V/ yet unproved,
m definition 3.7,
˚ T  T 
8Y ŒY 2 F” . F / ) Y 2 V2F F”.V/
T
m definitions: F”, ,
 T
8Y f9X ŒX 2 . F / ^ ŒY D F.X/g
 
) 8V f.V 2 F / ) ŒY 2 F”.V/g
T
 m definitions: , F”,
8Y f9X Œ8V .ŒV 2 F  ) ŒX 2 V/ ^ ŒY D F.X/g
˚   
) 8V .V 2 F / ) 9Xf.X 2 V/ ^ ŒY D F.X/g

 theorem 2.63: m no X in .V 2 F /,
8Y f9X Œ8V .ŒV 2 F  ) ŒX 2 V/ ^ ŒY D F.X/g
˚   
) 8V9X .V 2 F
„ ƒ‚ …/ ) f.X 2
„ƒ‚…V / ^ ŒY D F.X/g
„ ƒ‚ …
P Q R

which holds thanks to ` fŒ.P/ ) .Q/ ^ .R/g ) f.P/ ) Œ.Q/ ^ .R/g by

theorem 1.84, and ` f9XŒ8V.W/g ) f8VŒ9X.W/g by theorem 2.73. u
t

3.6.3 Exercises on Mathematical Functions

3.131 . Determine whether F WD f.0; 1/; .2; 3/; .1; 2/; .0; 4/g is a function.
3.132 . Determine whether G WD f.9; 2/; .7; 3/; .8; 2/; .6; 1/g is a function.
3.133 . Determine whether the following relation is a function:

R WD f.0; 1/; .1; 2/; .2; 4/; .3; 8/; .4; 8/; .5; 4/; .6; 2/; .7; 1/; .8; 0/g:

3.134 . Determine whether the following relation is a function:

S WD f.0; 2/; .1; 5/; .2; 7/; .3; 5/; .4; 2/; .5; 5/; .6; 7/; .7; 5/; .8; 2/g:

3.135 . Determine whether the following relation is a function:

Z WD f.0; 0/; .1; 0/; .2; 0/; .3; 0/; .4; 0/; .5; 0/; .6; 0/; .7; 0/; .8; 0/; .9; 0/g:

3.136 . Prove that exactly one function exists from A WD ¿ to B WD ¿.

3.137 . Investigate whether a function F W ¿ ! B exists from ¿ to any set B.
3.138 . Investigate whether a function F W A ! ¿ exists from any set A to ¿.
 www.pdfgrip.com

3.6 Mathematical Functions 163

3.139 . For each set A let 1A denote the constant function 1:

1A W A ! f1g; 1A .a/ D 1:

For each subset B  A, prove that 1A jB coincides with B .

3.140 . For all subsets V, W of each set A, investigate whether the characteristic
function of the intersection, V\W W A ! 2, is the intersection of the two
characteristic functions V W A ! 2 and W W A ! 2, so that V\W D V \ W .
3.141 . For all subsets V, W of each set A, investigate whether the characteristic
function of the union, V[W W A ! 2, is the union of the two characteristic functions
V W A ! 2 and W W A ! 2, so that V[W D V [ W .
3.142 . Prove that for each function F W A ! B, for all subsets R; S  A, and for all
subsets V; W  B, the following relations hold.

F ı1 ”.V [ W/ D F ı1 ”.V/ [ F ı1 ”.W/

F ı1 ”.V \ W/ D F ı1 ”.V/ \ F ı1 ”.W/
F”.R [ S/ D F”.R/ [ F”.S/
F”.R \ S/  F”.R/ \ F”.S/:

3.143 . Provide an example for which F”.R \ S/ ¤ F”.R/ \ F”.S/.

3.144 . For each function F W A ! B and for all subsets V; W  B, prove that
F ı1 ”.W n V/ D ŒF ı1 ”.W/ n ŒF ı1 ”.V/.
3.145 . For each function F W A ! B and for all subsets H; K  A, investigate
whether inclusion or equality holds for F”.K n H/ and ŒF”.K/ n ŒF”.H/.
3.146 . For each function F W A ! B, prove that F”.V/ D R .FjV /.
3.147 . For each function F W A ! B, prove that F” W P.A/ ! P.B/ contains all
the information about F, in the sense that fF.X/g D F”.fXg/ for every X 2 A.
3.148 . For each function F W P.A/ ! P.B/, investigate whether there exists a
function F W A ! B such that F” D F .
3.149 . Consider the function F W A ! B defined by
n ˚ o
A WD 3 D f0; 1; 2g D ¿; f¿g; ¿; f¿g ;
B WD A;
F WD f.0; 2/; .1; 2/; .2; 0/g:

Recall that the superscript ” indicates images of subsets (rather than of elements).
˚  ˚ 
(1) ˚¿; f¿g is a subset of A. Find its image: F” ¿; f¿g ˚: 
(2) ¿; f¿g is a subset of B. Find its inverse image F ı1 ” ¿; f¿g :
 www.pdfgrip.com

164 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

˚  ˚ 
(3) ¿; f¿g is an element of A. Find its image
n F ¿; f¿g
o :
˚  ı1
˚ 
(4) ¿; f¿g is an element of B. Find F ” ¿; f¿g :

3.150 . Consider the function G W C ! D defined by

C WD 4 D f0; 1; 2; 3g
˚  n ˚ o
D ¿; f¿g; ¿; f¿g ; ¿; f¿g; ¿; f¿g ;
D WD C;
G WD f.0; 3/; .1; 0/; .2; 3/; .3; 1/g:

Recall that the superscript ” indicates images of subsets (rather than of elements).
(1) 3 is a subset of C. Find its image: G”.3/:
(2) 3 is a subset of D. Find its inverse image Gı1 ”.3/:
(3) 3 is an element of C. Find its image G.3/:
(4) 3 is an element of D. Find Gı1 ”.f3g/:

3.7 Composite and Inverse Functions

3.7.1 Compositions of Functions

Some situations involve sequences of operations corresponding to sequences of

functions. For instance, if a first function consists of pairs .T; X/ with the ascension
(elevation) X of a planet at time T, and if a second function consists of pairs .X; Y/
with the declination (azimuth) Y of the planet at ascension X, then the composition
of the two functions consists of pairs .T; Y/ with the declination Y of the planet at
time T.
3.105 Definition (Composition of functions). For all functions F W A ! B and
G W B ! C, the composite function G ı F (read “G preceded by F” or “F followed
by G” or “the composition of G and F”) is the function G ı F W A ! C defined by

.G ı F/ .X/ WD G ŒF.X/

for each X 2 F ı1 ”ŒD.G/. Thus,

Œ.X; Z/ 2 .G ı F/ , fŒ9Y.Y 2 B/ ^ Œ.X; Y/ 2 F ^ Œ.Y; Z/ 2 Gg :

 www.pdfgrip.com

3.7 Composite and Inverse Functions 165

3.106 Example. Consider the following functions F and G:

A D f0; 1g; B D f0; 1; 2g;

B D f0; 1; 2g; C D f0; 1g;
FW A ! B; GW B ! C;
F D f.0; 0/; .1; 2/gI G D f.0; 1/; .1; 0/; .2; 1/g:

Their composition

.G ı F/W A ! C;

has values

.G ı F/.0/ D GŒF.0/ D GŒ0 D 1;

.G ı F/.1/ D GŒF.1/ D GŒ2 D 1;

so that

.G ı F/ D f.0; 1/; .1; 1/g:

3.107 Theorem. The composition of functions is associative: For all functions

F W A ! B, G W B ! C, and H W C ! D,

ŒH ı .G ı F/ D Œ.H ı G/ ı F:

Proof. For each X 2 F ı1 ”fGı1 ”ŒD.H/g, apply the definition of ı repeatedly:

ŒH ı .G ı F/.X/ D Hf.G ı F/.X/g

D HfGŒF.X/g
D ŒH ı GŒF.X/
D .ŒH ı G ı F/ .X/:

t
u
In contrast to its associativity, the composition of functions is not commutative.
3.108 Counterexample. Consider the following functions F and G:

A D f0; 1g; B D f0; 1; 2g;

B D f0; 1; 2g; C D f0; 1g;
FW A ! B; GW B ! C;
F D f.0; 0/; .1; 2/gI G D f.0; 1/; .1; 0/; .2; 1/g:

Their composition

.F ı G/W C ! A;
 www.pdfgrip.com

166 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

has values

.F ı G/.0/ D FŒG.0/ D FŒ1 D 2;

.F ı G/.1/ D FŒG.1/ D FŒ0 D 0;

so that

.F ı G/ D f.0; 2/; .1; 0/g:

In contrast,

.G ı F/ D f.0; 1/; .1; 1/g

from example 3.106, which confirms that .F ı G/ 6D .G ı F/.

3.7.2 Injective, Surjective, Bijective, and Inverse Functions

Such problems as the solution of equations involve the determination of whether

an equation has no solution, exactly one solution, or more than one solution, which
correspond to the features of functions introduced here.
3.109 Definition (injectivity). A function F W A ! B is injective if and only if for
all W 2 D.F/ and X 2 D.F/, if W 6D X, then F.W/ 6D F.X/:
  
8W 8X fŒW 2 D.F/ ^ ŒX 2 D.F/g ) fŒ.W 6D X/ ) ŒF.X/ 6D F.W/g :

By contraposition, the condition just stated is equivalent to the following alternative

condition: for all W 2 D.F/ and X 2 D.F/, if F.W/ D F.X/, then W D X:
  
8W 8X fŒW 2 D.F/ ^ ŒX 2 D.F/g ) fŒF.X/ D F.W/ ) .W D X/g :

The notation F W A ,! B indicates that F is an injection.

Another common usage consists of saying that F maps only one X to each one
Y; yet this alternative terminology fails to indicate which “one” it emphasizes (the
first “one”), which leads to confusion, and, therefore, will not be used here.
3.110 Example. With A WD f0; 1; 2g and B WD f0; 1; 2; 3; 4; 5; 6; 7; 8; 9g, the
function G WD f.0; 1/; .1; 3/; .2; 9/g is injective.
3.111 Example. With A WD f4; 6; 8; 9g and B WD f2; 3g, the function

H WD f.4; 2/; .6; 3/; .8; 2/; .9; 3/g

is not injective, because 4 6D 8 but H.4/ D 2 D H.8/.

 www.pdfgrip.com

3.7 Composite and Inverse Functions 167

3.112 Definition (surjectivity). A function F W A ! B is surjective if and only if

for each Y 2 B, there exists some X 2 A for which Y D F.X/. In other words, the
condition just stated means that the range of F consists of all of the co-domain B:
˚  
8Y .Y 2 B/ ) 9Xf.X 2 A/ ^ ŒF.X/ D Yg :

The notation F W A  B indicates that F is a surjection.

Another common mathematical usage consists of saying that F maps A onto B.
3.113 Example. With A WD f0; 1; 2g and B WD f0; 1; 2; 3; 4; 5; 6; 7; 8; 9g, the
function G WD f.0; 1/; .1; 3/; .2; 9/g is not surjective: there is no X 2 A with
G.X/ D 6.
3.114 Example. With A WD f4; 6; 8; 9g and B WD f2; 3g, the function

H WD f.4; 2/; .6; 3/; .8; 2/; .9; 3/g

is surjective. Indeed, H.4/ D 2 for Y WD 2, and H.6/ D 3 for Y WD 3,

3.115 Example. With A WD f6; 7; 8; 9g and B WD f1; 2; 3; 4g, the function

F WD f.6; 2/; .7; 1/; .8; 2/; .9; 3/g

is neither injective, because F.6/ D 2 D F.8/ with 6 6D 8, nor surjective, because

there does not exist any X 2 A with F.X/ D 4.

3.116 Definition. For all sets A, B, C, D, and for all functions F W A ! B and
G W C ! D, define the function

FGW A C ! B D;
 
.W; Z/ 7! F.W/; G.Z/ :
    
Thus .W; Z/; F.W/; G.Z/ 2 F  G if and only if W; F.W/ 2 F and
     
Z; G.Z// 2 G, so that W; F.W/ ; Z; G.Z// 2 F G.

3.117 Theorem. If F W A ! B and G W C ! D are both injective, or both

surjective, then F  G is injective, or surjective, respectively.
Proof. If F W A ! B and G W C ! D are both surjective, then the function F  G
is surjective: indeed, for each .U; V/ 2 B D, or, equivalently, for all U 2 B and
V 2 D, there exist X 2 A and Y 2 C such that F.X/ D U and G.Y/ D V, by
surjectivity of F and G. Hence
 
.F  G/.X; Y/ D F.X/; G.Y/ D .U; V/:

If F W A ! B and G W C ! D are both injective, then the function F G is injective:

indeed, if .F  G/.W; Z/ D .F  G/.R; S/, then
 www.pdfgrip.com

168 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

   
F.W/; G.Z/ D .F  G/.W; Z/ D .F  G/.R; S/ D F.R/; G.S/ :

Hence F.W/ D F.R/ and G.Z/ D G.S/, by equality of ordered pairs. Consequently,
W D R and Z D S, by injectivity of F and G. t
u
3.118 Definition (Bijectivity). A function F W A ! B is bijective if and only if F
is both injective and surjective, which can be denoted by F W A  B or F W A • B.
3.119 Example. With A WD f0; 1; 2; 3g and B WD f1; 2; 4; 8g, the function P W A !
B defined by P WD f.0; 1/; .1; 2/; .2; 4/; .3; 8/g is bijective.
3.120 Example. For each set A the identity function IA W A ! A, X 7! X is bijective.
Indeed, IA is injective, because if IA .W/ D IA .X/ then W D IA .W/ D IA .X/ D X.
Similarly, IA is surjective, because for each Y 2 A there exists X 2 A, in effect
X WD Y, with Y D IA .Y/.
3.121 Example. For each proper subset S ¤ A, the inclusion function  W S ! A,
X 7! X, is injective but not surjective. Indeed,  is injective, for if W; X 2 S and
W 6D X, then .W/ D W 6D X D .X/, whence .W/ 6D .X/. However,  is not
surjective: because S is a proper subset of A, there exists some element Z in A n S; in
particular, Z 6D X for each X 2 S, and, consequently, .X/ D X 6D Z, which means
that  is not surjective. Thus,  is not bijective either.
3.122 Example. For each nonempty set A and for each set B containing more than
one element, the canonical projection PA W .A B/ ! A is surjective but not
injective. Indeed, B contains at least one element Y 2 B, because B 6D ¿; hence,
X D PA .X; Y/ for each X 2 A. However, B also contains some Z 2 B such that
Y 6D Z. Consequently, .X; Y/ 6D .X; Z/, and yet PA .X; Y/ D X D PA .X; Z/, which
means that PA is not injective. Thus, PA is not bijective either.
3.123 Theorem. For all functions F W A ! B and G W B ! C,
• if F and G are both injective, then G ı F is also injective;
• if F and G are both surjective, then G ı F is also surjective;
• if F and G are both bijective, then G ı F is also bijective;
• if G ı F is injective, then F is injective;
• if G ı F is surjective, then G is surjective;
• if G ı F is bijective, then F is injective and G is surjective.
Proof. Assume that F and G are both injective. For all distinct elements W 6D X in
F ı1 ”ŒD.G/, the injectivity of F ensures that F.W/ 6D F.X/. Hence G.F.W// 6D
G.F.X// by the injectivity of G. Hence, .G ı F/.W/ D G.F.W// 6D G.F.X// D
.G ı F/.X/, whence .G ı F/.W/ 6D .G ı F/.X/, so that G ı F is injective.
Assume that F and G are both surjective. For each Z 2 C, the surjectivity of
G ensures the existence of an element Y 2 B such that G.Y/ D Z. Hence, by the
surjectivity of F, there exists an element X 2 A for which F.X/ D Y. Therefore,
.G ı F/.X/ D G.F.X// D G.Y/ D Z, which means that .G ı F/ is surjective.
In particular, if F and G are both bijective, then G ı F is also bijective.
 www.pdfgrip.com

3.7 Composite and Inverse Functions 169

Assume that G ı F is injective. For all distinct elements W 6D X in F ı1 ”ŒD.G/,

the injectivity of G ı F ensures that .G ı F/.W/ 6D .G ı F/.X/, whence G.F.W// 6D
G.F.X//. Because G is a function, G cannot take different values at the same
argument, whence it follows that F.W/ 6D F.X/, so that F is injective.
Assume that G ı F is surjective. For each element Z 2 C, the surjectivity of G ı F
ensures the existence of an element X 2 A such that .G ı F/.X/ D Z. Hence, letting
Y WD F.X/ demonstrates the existence of an element Y 2 B for which G.Y/ D
G.F.X// D .G ı F/.X/ D Z, which means that G is surjective.
In particular, if G ı F is bijective, then F is injective and G is surjective. t
u
3.124 Definition (invertibility). A function F W A ! B is invertible if and only if
there exists a function G W B ! A for which G ı F D ID.F/ and F ı G D ID.G/ . Such
a function G is denoted by F ı1 and called the inverse function of F. Thus,

F ı1 ı F D ID.F/ ;
F ı F ı1 D ID.G/ :

3.125 Example. With A WD f0; 1; 2; 3g and B WD f1; 2; 4; 8g, the function

F WD f.0; 1/; .1; 2/; .2; 4/; .3; 8/g

is invertible, with inverse F ı1 WD f.1; 0/; .2; 1/; .4; 2/; .8; 3/g.
3.126 Theorem. Each function F W A ! B admits at most one inverse function.
Moreover, if G W B ! A is an inverse function for F, then G consists of all pairs
obtained by swapping the coordinates in each pair of F.
Proof. Assume that G is an inverse function for F, which means that G ı F D ID.F/
and F ı G D ID.G/ , and consider the set

H WD f.Y; X/ W .X; Y/ 2 Fg :

If .X; Y/ 2 F, then X 2 D.F/. Because G ı F D ID.F/ , it follows that there exists

some Z 2 B such that .X; Z/ 2 F and .Z; X/ 2 G. With .X; Y/ 2 F and .X; Z/ 2 F,
it follows that Y D Z, because F is a function. This shows that if .Y; X/ 2 H, so that
.X; Y/ 2 F, then .Y; X/ D .Z; X/ 2 G, whence G  H.
Conversely, If .Z; X/ 2 G, then Z 2 D.G/. Because F ı G D ID.G/ it follows that
there exists some W 2 D.F/ such that .Z; W/ 2 G and .W; Z/ 2 F. With .Z; X/ 2 G
and .Z; W/ 2 G, it follows that W D X, because G is a function. This shows that if
.Z; X/ 2 G, then .X; Z/ 2 F, whence .Z; X/ 2 H, so that G  H.
Finally, G D H, which shows that if F has an inverse function G, then the only
possibility is G D H. Thus, G D H D F ı1 D f.Y; X/ W .X; Y/ 2 Fg. t
u
3.127 Theorem. For each function F W A ! B with D.F/ D A, the function F is
invertible if and only if F is bijective.
 www.pdfgrip.com

170 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

Proof. Assume that F is invertible, with inverse G WD F ı1 . Because F ı F ı1 D

ID.G/ is surjective, it follows from theorem 3.123 that F is surjective. Similarly,
because F ı1 ı F D ID.F/ is injective, it follows from theorem 3.123 that F is
injective.
Conversely, assume that F is bijective. Construct an inverse function by means
of the set G defined by swapping both coordinates in each pair of the function F:

G WD f.Y; X/ W .X; Y/ 2 Fg :

Then verify that G is the inverse function of F.

First, the injectivity of F ensures that G is a function: if .Y; X/ 2 G and .Y; W/ 2
G, then .X; Y/ 2 F and .W; Y/ 2 F, whence X D W by injectivity.
Second, G ı F D ID.F/ . Indeed, if .X; Z/ 2 .G ı F/, then there exists some Y 2 B
for which .X; Y/ 2 F and .Y; Z/ 2 G. Consequently, .Z; Y/ 2 F, and again the
injectivity of F shows that X D Z, whence .X; Z/ D .X; X/. Thus, .G ı F/  ID.F/ .
Because the foregoing reasoning holds for each X 2 D.F/, however, it also follows
that ID.F/  .G ı F/, and thus .G ı F/ D ID.F/ .
Finally, F ı G D ID.G/ . Indeed, if .Y; W/ 2 .F ı G/, then there exists some X 2
D.G/ with .Y; X/ 2 G and .X; W/ 2 F. From .Y; X/ 2 G, it follows that .X; Y/ 2 F.
Because F is a function, it also follows that Y D W, whence .Y; W/ D .Y; Y/. Thus,
.F ı G/  ID.G/ . Then the surjectivity of F guarantees that for each Y 2 D.G/ there
exists some X 2 A with .X; Y/ 2 F. Hence, .Y; X/ 2 G and then .Y; Y/ 2 .F ı G/,
so that IB  .F ı G/. Therefore, .F ı G/ D IB . t
u
Some situations involve a concept more general than invertibility.
3.128 Definition (left or right invertibility). A function F W A ! B is invertible
on its left, or left invertible, if and only if there exists a function G W B ! A for
which G ı F D ID.F/ . Such a function G is called a left inverse function for F.
Similarly, a function F W A ! B is invertible on its right, or right invertible,
if and only if there exists a function G W B ! A for which F ı G D ID.G/ . Such a
function G is called a right inverse function for F.
3.129 Example. With A WD f0; 1; 2g and B WD f0; 1; 2; 3; 4; 5; 6; 7; 8; 9g, the
function F W A ! B defined by F WD f.0; 1/; .1; 3/; .2; 9/g has a left inverse function

G WD f.1; 0/; .2; 0/; .3; 1/; .4; 0/; .5; 0/; .6; 0/; .7; 0/; .8; 0/; .9; 2/g:

3.130 Example. For A WD f4; 6; 8; 9g and B WD f2; 3g, the function F W A ! B with
F WD f.4; 2/; .6; 3/; .8; 2/; .9; 3/g has a right inverse G WD f.2; 4/; .3; 6/g
3.131 Example. Perspectives to draw a picture of space A on a flat screen B  A
can be represented by a function F W A ! B, mapping each point X in space A to its
image F.X/ on the screen B. For such perspectives, each point Y 2 B on the screen
B is its own image, so that F.Y/ D Y. Thus the inclusion function B W B ! A is a
right inverse for F, because .F ı B /.Y/ D FŒB .Y/ D F.Y/ D Y for every Y 2 B,
 www.pdfgrip.com

3.7 Composite and Inverse Functions 171

so that F ı B D IB . In contrast, such a perspective F has no left inverse, because F

maps many points in space to the same image on the screen.
The existence of a left-inverse G W B ! A for a function F W A ! B indicates
that for each Y 2 B the equation F.X/ D Y has at most one solution. In contrast, the
existence of a right-inverse G W B ! A for a function F W A ! B indicates that for
each Y 2 B the equation F.X/ D Y has at least one solution.

3.7.3 The Set of all Functions from a Set to a Set

This subsection shows that all the functions from any fixed set into any fixed co-
domain form a set.
Theorem 3.132 shows that all the functions between two fixed sets form a set.
3.132 Theorem. For all sets A and B, all the functions from any subset of A to B are
the elements of a set, of which all such functions defined on all of A form a subset.
Proof. By definition 3.66, the power set P.A B/ is the set of all the relations
between A and B. By definition 3.81, every function F W A ! B defined on any
subset of A is also a relation, so that F 2 P.A B/. By the axiom of separation
(page 124), with a formula stating that a relation F 2 P.A B/ is a function, all
such functions form a subset of P.A B/, denoted here by FA!B :

FA!B
˚   
WD F 2 P.A B/ W 8W8X8Y Œ.X; Y/ 2 F ^ Œ.W; Y/ 2 F ) .X D W/ :

For each subset D  A, again by the axiom of separation (page 124), with a formula
stating that the domain of F is D, the functions F W D ! B defined on all of D form
a subset of FD!B , denoted here by BD :
˚   
BD WD F 2 FD!B W 8X .X 2 D/ ) 9YŒ.X; Y/ 2 F :

The case D WD A corresponds to the set BA of all functions from all of A to B. t

u
Definition 3.133 sets the notation from the proof of theorem 3.132.
3.133 Definition. For all sets A and B, the set of all functions F W A ! B defined
on any subset of A is denoted by FA!B . The set of all functions F W A ! B defined
on all of A is denoted by BA .
3.134 Example. If A D ¿, then A B D ¿ B D ¿ for every set B regardless of
B. The only subset F WD ¿  ¿ D A B is a function from A to B with domain A.
Hence if K D ¿ D 0, then BK D B0 D f¿g D f0g. Thus there exists a function of
zero variable with zero value in every set B, and this function is the natural number
 www.pdfgrip.com

172 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

zero. This function of zero variable allows for a minimal set of starting functions in
some contexts, for instance, with primitive recursive functions [109, p. 926].
[
3.135 Theorem. For all sets A and B, BD D FA!B .
DA

Proof. For each subset D  A, the set BD consists of all the functions F W D ! B
defined on all of D  A, whence BD  FA!B , or, equivalently, BD 2 P.FA!B /
and hence fBD 2 P.FA!B / W D 2 P.A/g S is also a set. Consequently so is its union
by the axiom of union (page 127): DA BD  FA!B . Conversely, each function
F 2 FA!B is defined Son all of its domain, which is a subset D  A, so that F 2 BD .
Hence also FA!B  DA BD . t
u
3.136 Theorem. For all sets A and B, for each set D  P.A/ of subsets of A, all
the D
S sets B Sfor every D 2 D also form a set FD;B  P.FA!B /. Hence the union
FD;B D D2D BD is also a set.
Proof. For each set D  P.A/ of subsets of A, all the sets BD for every D 2 D
also form a subset of P.FA!B /, denoted here by FD;B :
˚ 
FD;B WD E 2 PŒP.A B/ W 9DŒ.D 2 D/ ^ .E D BD / ;

where E D BD is an abbreviation of the formula resulting from the axiom of

extensionality (page 111) and the preceding
S twoSapplications of the axiom of
separation (page 124). Hence the union FD;B D D2D BD is also a set,
[ ˚ 
FD;B D F 2 FA!B W 9DŒ.D 2 D/ ^ .F 2 BD / ;

by the axiom of union (page 127). t

u
Theorem 3.137 reveals a bijection between the set CAB of all functions defined
on a Cartesian product A B into a set C and the set .CB /A of all functions from A
into the set CB of all functions defined on B into C.
3.137 Theorem. For all sets A, B, C, there is a bijection from CAB onto .CB /A .
Proof. Define a function H W CAB ! .CB /A as follows. For each element F 2
CAB , which is a function F W A B ! C defined on all of A B, define an element
H.F/ 2 .CB /A , which is a function H.F/ W A ! CB defined on all of A, so that for
each X 2 A the image ŒH.F/.X/ 2 CB is a function ŒH.F/.X/ W B ! C defined on
all of B, defined for each Y 2 B by

H W CAB ! .CB /A ;
F 7! H.F/;
fŒH.F/.X/g.Y/ WD F.X; Y/:
 www.pdfgrip.com

3.7 Composite and Inverse Functions 173

Similarly, define a function L W .CB /A ! CAB as follows. For each element G 2

.CB /A , define an element L.G/ 2 CAB , which is a function L.G/ W A B ! C,
defined by ŒL.G/.X; Y/ WD ŒG.X/.Y/.

L W .CB /A ! CAB ;
G 7! L.G/;
ŒL.G/.X; Y/ WD ŒG.X/.Y/:

Then H ı L D I.CB /A and L ı H D ICAB , whence H and L are inverses of each other:

Œ.L ı H/.F/.X; Y/ D fLŒH.F/g.X; Y/

D fŒH.F/.X/g.Y/
D F.X; Y/I
fŒ.H ı L/.G/.X/g.Y/ D ŒfHŒL.G/g.X/.Y/
D ŒL.G/.X; Y/
D ŒG.X/.Y/:

Thus .L ı H/.F/ D F and .H ı L/.G/ D G. Hence H and L are bijections. t

u

3.7.4 Exercises on Injective, Surjective, and Inverse Functions

3.151 . For each F W A ! B and IB W B ! B, prove that IB ı F D F.

3.152 . For each F W A ! B and IA W A ! A, prove that F ı IA D F.
3.153 . For each F W A ! B and ¿ W ¿ ! A, prove that F ı ¿ D ¿.
3.154 . Prove that if a function has a left inverse, then it is injective.
3.155 . Prove that if a function is injective, then it has a left inverse.
3.156 . Prove that if a function has a right inverse, then it is surjective.
3.157 . Provide an example of a function with a right inverse but no left inverse.
 www.pdfgrip.com

174 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.158 . Provide a function with a left inverse but no right inverse.

3.159 . Provide a function that has more than one left inverse.
3.160 . Provide a function that has more than one right inverse.

3.8 Equivalence Relations

3.8.1 Reflexive, Symmetric, Transitive, or Anti-Symmetric

Relations

Besides functions, mathematics contains several other types of relations. For

instance, ordering relations define orders or rankings, whereas equivalence relations
define equivalences relative to certain criteria. Such various types of relations can
be defined by combinations of several features called reflexivity, symmetry, and
transitivity.
3.138 Definition (Reflexivity). For each set A, a relation R  A A is reflexive if
and only if .X; X/ 2 R for each X 2 A:

8Xf.X 2 A/ ) Œ.X; X/ 2 Rg:

As a graph, a reflexive relation contains at least a single loop at each vertex:

(X, X) X Y (Y,Y)

3.139 Theorem. A relation R  A A is reflexive if and only if A  R.

Proof. A relation R  A A is reflexive if and only if .X; X/ 2 R for every X 2 A,
in other words, if and only if R contains the set f.X; X/ W X 2 Ag D A . u
t
3.140 Example. For each set A, the diagonal A is a reflexive relation.
3.141 Example. For each set A and its power set P.A/, the relation  on P.A/ is
reflexive. Indeed, B  B for each B 2 P.A/.
3.142 Counterexample. The relation of membership 2 is not reflexive. For
instance, if A D f¿g, then ¿ 2 A, but ¿ … ¿. Thus 2 is not reflexive on
A A.
3.143 Definition (symmetry). For each set A, a relation R  A A is symmetric
if and only if .X; Y/ 2 R is equivalent to .Y; X/ 2 R for all X 2 A and Y 2 A:

8X8Y.Œ.X; Y/ 2 R , Œ.Y; X/ 2 R/:

 www.pdfgrip.com

3.8 Equivalence Relations 175

As a graph, a symmetric relation contains edges in either both directions or neither:

(X,Y)

X Y Z

(Y, X)

3.144 Example. For each set A, the diagonal A is a symmetric relation.

 The relation of membership 2 is not symmetric in gen-

3.145 Counterexample.
˚
eral: if A WD ¿; f¿g , then ¿ 2 f¿g but f¿g … ¿; thus 2 is not symmetric.
3.146 Definition (transitivity). For each set A, a relation R  A A is transitive
if and only if .X; Y/ 2 R and .Y; Z/ 2 R imply .X; Z/ 2 R for all X; Y; Z 2 A:
 
8X8Y8Z fŒ.X; Y/ 2 R ^ Œ.Y; Z/ 2 Rg ) Œ.X; Z/ 2 R

As a graph, a transitive relation completes two consecutive edges into a triangle:

(X,Y)
X Y

(X, Z) (Y, Z)

3.147 Example. For each set A, the diagonal A is a transitive relation.

3.148 Example. For each set A, the relation  on P.A/ is transitive: for all U 2
P.A/, V 2 P.A/, W 2 P.A/, if U  V and V  W, then U  W.
3.149 Counterexample.
n The relation of membership 2 is not transitive in general:
˚ o ˚  ˚ 
if A D ¿; f¿g; f¿g , then ¿ 2 f¿g and f¿g 2 f¿g , but ¿ … f¿g .

3.8.2 Partitions and Equivalence Relations

The concept of an equivalence relation on a set corresponds to a “partition” of that

set into a union of disjoint subsets called “equivalence” classes.
3.150 Definition (equivalence relations). For each set A, a relation R  A A is
an equivalence relation if and only if R is reflexive, symmetric, and transitive.

3.151 Example. For each set A, the diagonal A is an equivalence relation.

3.152 Example. The set A WD f0; 1; 2; 3g admits an equivalence relation

R WD f.0; 0/; .1; 1/; .2; 2/; .3; 3/; .0; 2/; .2; 0/; .1; 3/; .3; 1/g
D A [ f.0; 2/; .2; 0/; .1; 3/; .3; 1/g:
 www.pdfgrip.com

176 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

• The relation R is reflexive because it contains the diagonal

A D f.0; 0/; .1; 1/; .2; 2/; .3; 3/g:

• The relation R is symmetric: it contains .0; 2/, .2; 0/, as well as .1; 3/, .3; 1/.
• The relation R is transitive because if it contains .X; Y/ and .Y; Z/, then it
contains .X; Z/, for instance, .0; 2/ and .2; 0/ hence .0; 0/; .2; 0/ and .0; 2/ hence
.2; 2/, as well as .1; 3/ and .3; 1/ hence .1; 1/; .3; 1/ and .1; 3/ hence .3; 3/.
3.153 Counterexample. The relation of membership 2 is not an equivalence
relation in general, because it is not symmetric and not transitive.
3.154 Definition (partition). A partition of a set A is a set F  P.A/ of subsets
of A with all of the following properties.
• No member of F is empty: 8Vf.V 2 F / ) .V 6D ¿/g: S
• The union of all the members of F “covers” A, which means that A D . F / :
• All pairs of distinct members of F are disjoint:
 
8V 8WfŒ.V 2 F / ^ .W 2 F / ^ .V 6D W/ ) Œ.V \ W/ D ¿g :

3.155 Example. The empty set ¿ admits only one partition: the empty set F D ¿.
3.156 Example. Each nonempty set S has a partition: the singleton F D fSg.
3.157 Example.
˚ The set AWD f0; 1; 2; 3g admits a partition F into two disjoint
sets, F WD f0; 2g; f1; 3g :
• No member of F is empty: f0; 2g 6D ¿ and f1; 3g 6D ¿.
• The unions of the members of F covers A, so that f0; 1; 2; 3g D f0; 2g [ f1; 3g.
• Distinct members of F are disjoint: f0; 2g \ f1; 3g D ¿.
3.158 Definition (relations from partitions). For each partition F  P.A/ of
each set A, define a relation RF on A so that R relates elements X 2 A and Y 2 A if
and only if X and Y belong to the same element of the partition F :

Œ.X; Y/ 2 RF  , f9BŒ.B 2 F / ^ .X 2 B/ ^ .Y 2 B/g:

3.159 Theorem (equivalence relations from partitions). For each set A and for
each partition F of A, the relation RF is an equivalence relation.
Proof. The relation RF is reflexive: for each X 2 A, the partition F has an element
B 2 F that contains X, because F covers A. Thus X and X both belong to the same
B 2 F , whence .X; X/ 2 RF . Symbolically,
 www.pdfgrip.com

3.8 Equivalence Relations 177

8Xf.X 2 A/ ) Œ.X; X/ 2 RF g yet unproved,

 m definition 3.158 for RF ,
8X fX 2 Ag ) 
f9BŒ.B 2 F / ^ .X 2 B/ ^ .X 2 B/g
 m tautology Œ.P/ ^ .P/ , .P/,
8X fX 2 Ag ) 
f9BŒ.B 2 F / ^ .X 2 B/g
S
m definition of .
S
A  . F/
which is universally valid by definition of a partition.
The relation RF is symmetric: .X; Y/ 2 RF if and only if the partition F has an
element B 2 F that contains X and Y, whence B contains Y and X, which means
that .Y; X/ 2 RF . Symbolically, from the tautology Œ.P/ ^ .Q/ , Œ.Q/ ^ .P/,
n
` 8X8Y Œ.X 2 A/ ^ .Y 2 A/ )
˚
9BŒ.B 2 F / ^ .X 2 B/ ^ .Y 2 B/g ,
o
f9BŒ.B 2 F / ^ .Y 2 B/ ^ .X 2 B/g

  m definition 3.158 for RF .

8X 8Y Œ.X 2 A/ ^ .Y 2 A/ ) 
fŒ.X; Y/ 2 RF  , Œ.Y; X/ 2 RF g
The relation RF is transitive: if .X; Y/ 2 RF , then the partition F has an element
B 2 F that contains X and Y. If also .Y; Z/ 2 RF , then the partition F has an
element C 2 F that contains Y and Z. However, from X 2 B and Y 2 C it follows
that X 2 .B \ C/, whence B and C are not disjoint and hence B D C, by definition
of a partition. Consequently, X 2 B and Z 2 B, and, therefore, .X; Z/ 2 RF . t
u
3.160 Definition (equivalence classes). For each equivalence relation R  .A A/,
and for each element X 2 A, define the subset ŒXR of all the elements Y 2 A
equivalent to X with respect to R, called the equivalence class of X:

ŒXR WD fY 2 A W .X; Y/ 2 Rg:

Then let FR consist of all such equivalence classes:

FR WD fŒXR 2 P.A/ W X 2 Ag:

Another common notation for FR is A=R, so that FR D A=R D fŒXR W X 2 Ag.

The set FR D A=R of all equivalence classes is also called the quotient of the set A
by the relation R.
3.161 Example. For each set A, the equivalence classes of the “diagonal” equiva-
lence relation
˚ A consist
 of every singleton ŒXA D fXg for every X 2 A. Thus
A=A D fXg W X 2 A .
 www.pdfgrip.com

178 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.162 Example. For the set A WD f0; 1; 2; 3g, the equivalence relation

R WD f.0; 0/; .1; 1/; .2; 2/; .3; 3/; .0; 2/; .2; 0/; .1; 3/; .3; 1/g
D A [ f.0; 2/; .2; 0/; .1; 3/; .3; 1/g

corresponds to the equivalence classes

Œ0R D f0; 2g;

Œ1R D f1; 3g:

Thus A=R D fŒ0R ; Œ1R g.

3.163 Theorem (partitions from equivalence relations). For each equivalence
relation R  .A A/, the set of subsets FR  P.A/ is a partition of A.
Proof. The partition FR covers A: the reflexivity of RSguarantees thatS.X; X/ 2 R
for each
S X 2 A, whence X 2 ŒXR , and hence X S2 . B
Z2ASZ / D . FR /. Thus
A  . F /. The reverse inclusion follows from . F /  ΠP.A/ D A:
` 8Xf.X 2 A/ ) Œ.X; X/ 2 Rg reflexivity of R,
m definition of ŒXR ,
8Xf.X 2 A/ ) .X 2 ŒXR /g
m definition of FR ,
8Xf.X 2 A/ ) Œ.ŒXR 2 FR / ^ .X 2 ŒXR /g
m definition of 9,
8XŒ.X 2 A/ ) f9BŒ.B 2 FR / ^ .X 2 B/g
S
m definition of ,
S
8Xf.X 2 A/ ) ŒX 2 . F /g
m definition of .
S
A  . F /
Any two distinct elements of FR are disjoint: if two members B and C of FR are
not disjoint, then their intersection B \ C contains an element X 2 A; by definition
of FR , however, every element of B is equivalent to X, and so is every element of C,
whence B D ŒXR D C, which is a negation of the distinctness of B and C. Finally,
FR does not contain any empty element. Indeed, if A D ¿, then R  A A, whence
R D ¿ and FR D ¿, which does not contain any element, and hence no empty
element. If A 6D ¿, then FR D fŒXR W X 2 Ag where X 2 ŒXR , whence ŒXR 6D ¿.
t
u
3.164 Example. For each equivalence relation R  A A on a set A, the canonical
map, also called the quotient map, is the function P W A ! A=R that maps each
element X 2 A to its equivalence class ŒXR D fY 2 A W .X; Y/ 2 Rg:

PW A ! A=R;
X 7! ŒXR :
 www.pdfgrip.com

3.8 Equivalence Relations 179

3.8.3 Exercises on Equivalence Relations

3.161 . Prove that the empty relation ¿  A A is reflexive, symmetric, and

transitive.
3.162 . Prove that for each set A the relation of strict inclusion ¤ on P.A/ is not
reflexive.
3.163 . Prove that for each set A the relation of strict inclusion ¤ on P.A/ is anti-
symmetric.
3.164 . Prove that for each set A the relation of strict inclusion ¤ on P.A/ is
transitive.
3.165 . For the set A WD f0; 1; 2; 3; 4; 5g, verify that the following relation R is an
equivalence relation, and list all its equivalence classes:
8 9
ˆ
ˆ .1; 5/ .3; 5/ .5; 5/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 4/ .2; 4/ .4; 4/ >
>
ˆ
< >
=
.1; 3/ .3; 3/ .5; 3/
R WD :
ˆ
ˆ .0; 2/ .2; 2/ .4; 2/ >
>
ˆ
ˆ >
ˆ
ˆ .1; 1/ .3; 1/ .5; 1/ >
>
>
>
:̂ ;
.0; 0/ .2; 0/ .4; 0/

3.166 . For the set A WD f0; 1; 2; 3; 4; 5g, verify that the following relation S is an
equivalence relation, and list all its equivalence classes:
8 9
ˆ
ˆ .5; 2/ .5; 5/ >
>
ˆ
ˆ >
>
ˆ
ˆ .1; 4/ .4; 4/ >
>
ˆ
< >
=
.0; 3/ .3; 3/
S WD :
ˆ
ˆ .2; 2/ .2; 5/ >
>
ˆ
ˆ >
>
ˆ
ˆ .1; 1/ .4; 1/ >
>
:̂ >
;
.0; 0/ .3; 0/

3.167 . For the set B WD f0; 1; 2; 3; 4; 5; 6; 7g, verify that the following set F of
subsets is a partition, and list the corresponding equivalence relation:
˚ 
F WD f0; 2; 4; 6g; f1; 3; 5; 7g :

3.168 . For the set B WD f0; 1; 2; 3; 4; 5; 6; 7g, verify that the following set G of
subsets is a partition, and list the corresponding equivalence relation:
˚ 
G WD f0; 4g; f1; 5g; f2; 6g; f3; 7g :

3.169 . Prove that R.FR / D R for each equivalence relation R.

3.170 . Prove that F.RF / D F for each partition F .
 www.pdfgrip.com

180 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.9 Ordering Relations

3.9.1 Preorders and Partial Orders

Besides the reflexivity, symmetry, and transitivity used to define equivalence

relations, such other types of relations as rankings also require different variations
of these concepts, as introduced here (with the terminology of Suppes [128, 3:2,
p. 69]).
3.165 Definition (strict, irreflexivity). For each set A, a relation R  A A is
irreflexive, or, equivalently, strict, if and only if R does not relate any element of A
to itself:

8Xf.X 2 A/ ) Œ.X; X/ … Rg:

3.166 Example. The empty relation ¿ is strict, because the conclusion .X; X/ … ¿
is universally valid.
3.167 Example. For each set A, the relation of strict inclusion  is strict on P.A/.
Indeed, for all subsets V  A and W  A, the definition of V  W includes the
requirement that V 6D W, so that .V  W/ ) .V 6D W/. Contraposition then
confirms that .V D W/ ) .V 6 W/, so that V 6 V.
One method to specify a ranking or direction on a set removes the requirement
of symmetry from the concept of equivalence, which gives the following concept of
preorder.
3.168 Definition (preorder or quasi-order). For each set A, a relation R  A A
is a preorder or a quasi-order, if and only if R is reflexive and transitive. It is a
strict preorder if and only if R is irreflexive and transitive.
3.169 Example. Consider the set A WD f0; 1; 2g.
The relation Q WD f.0; 0/; .0; 1/; .1; 0/; .1; 1/; .2; 2/g is a preorder.
The relation R WD f.0; 0/; .0; 1/; .1; 1/; .2; 2/g is a preorder.
The relation S WD f.0; 1/g is a strict preorder.
3.170 Example. For each set A, the diagonal A is a preorder. If A 6D ¿, then A is
not strict, because there exists X 2 A and then .X; X/ 2 A .
3.171 Example. For each set A the relation  is a preorder on P.A/. The relation
 is not strict because ¿ 2 P.A/ and ¿  ¿.
3.172 Example. For each set A, the relation  is a strict preorder on P.A/.
The concept of a preorder R allows for “circular” rankings, with .X; Y/ 2 R and
.Y; X/ 2 R even though X 6D Y. To specify different types of rankings, a different
concept — anti-symmetry — becomes necessary.
 www.pdfgrip.com

3.9 Ordering Relations 181

3.173 Definition (anti-Symmetry). For each set A, a relation R  A A is anti-

symmetric if and only if .X; Y/ 2 R and .Y; X/ 2 R imply X D Y:
  
8X 8Y fŒ.X; Y/ 2 R ^ Œ.Y; X/ 2 Rg ) .X D Y/ :

3.174 Example. For each set A, the diagonal A is an anti-symmetric relation.

3.175 Example. For each set A, the relation  on P.A/ is anti-symmetric. Indeed,
for each B 2 P.A/ and for each C 2 P.A/, if B  C and C  B, then B D C.
Similar features can also be defined through the following concept of asymmetry.
3.176 Definition (asymmetry). For each set A, a relation R  A A is asymmetric
if and only if .X; Y/ 2 R implies .Y; X/ … R for each X 2 A and each Y 2 A:
 
8X 8YfŒ.X; Y/ 2 R ) Œ.Y; X/ … Rg :

3.177 Example. Consider the set A WD f0; 1; 2g.

The relation Q WD f.0; 0/; .0; 1/; .1; 0/; .1; 1/; .2; 2/g is not asymmetric: Q
contains .0; 1/ and .1; 0/.
The relation R WD f.0; 0/; .0; 1/; .1; 1/; .2; 2/g is not asymmetric: R contains
.0; 0/, .1; 1/, and .2; 2/.
The relation S WD f.0; 1/g is asymmetric.
3.178 Example. The following relation is not asymmetric, not anti-symmetric, not
irreflexive, not reflexive, not symmetric, and not transitive:
(X,Y)
(Y, Z)
(X, X) X Y Z

(Y, X)

In contrast to preorders, “partial orders” do not allow for circular rankings.

3.179 Definition (partial order). For each set A, a relation R  A A is a partial
order if and only if R is reflexive, anti-symmetric, and transitive. It is a strict partial
order if and only if R is irreflexive and transitive.
3.180 Example. Consider the set A WD f0; 1; 2g.
The relation Q WD f.0; 0/; .0; 1/; .1; 0/; .1; 1/; .2; 2/g is not a partial order,
because it is not anti-symmetric: Q contains .0; 1/ and .1; 0/ but 0 6D 1.
The relation R WD f.0; 0/; .0; 1/; .1; 1/; .2; 2/g is a partial order.
The relation S WD f.0; 1/g is a strict partial order.
3.181 Example. For each set A, the diagonal A is a partial order.
3.182 Example. For each set A the relation  is a partial order on P.A/.
3.183 Example. For each set A the relation ¤ is a strict partial order on P.A/.
 www.pdfgrip.com

182 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

As their names might suggest, neither preorders nor partial orders need relate all
elements to one another. Relations that do so are called strongly connected.
3.184 Definition (Strong connectivity). For each set A, a relation R  A A
is strongly connected if and only if .X; Y/ 2 R or .Y; X/ 2 R (including both
possibilities) for each X 2 A and for each Y 2 A:
  
8X 8Y Œ.X 2 A/ ^ .Y 2 A/ ) fŒ.X; Y/ 2 R _ Œ.Y; X/ 2 Rg :

Strict rankings replace strong connectivity by connectivity.

3.185 Definition (connectivity). For each set A, a relation R  A A is connected
if and only if .X; Y/ 2 R or .Y; X/ 2 R (including both possibilities) for all distinct
X 2 A and Y 2 A (such that X 6D Y):
  
8X 8Y Œ.X 2 A/ ^ .Y 2 A/ ^ .X 6D Y/ ) fŒ.X; Y/ 2 R _ Œ.Y; X/ 2 Rg :

3.186 Example. Consider the set A WD f0; 1; 2g.

The relation Q WD f.0; 0/; .0; 1/; .1; 0/; .1; 1/; .2; 2/g is neither connected nor
strongly connected, because it contains neither .0; 2/ nor .2; 0/.
The relation T WD f.0; 0/; .0; 1/; .0; 2/; .1; 1/; .1; 2/; .2; 2/g is strongly
connected.

3.9.2 Total Orders and Well-Orderings

The geometric “direction along a line” corresponds to a total order, also called
linear order, complete order [71, p. 14], or simple order [71, p. 14], [128, p. 69].
3.187 Definition (total order). For each set A, a relation R  A A is a total order,
or total ordering, if and only if R is a strongly connected partial order (strongly
connected, reflexive, anti-symmetric, and transitive). It is a strict total order if and
only if R is connected, irreflexive, and transitive.
3.188 Example. The empty relation on the empty set is a strict total order.
3.189 Example. For each set S and the singleton A D fSg, the relation  on
P.A/ D f¿; fSgg, is a total order. Indeed, ¿  ¿, ¿  fSg, and fSg  fSg.
3.190 Counterexample. The relation  is not a total order in general. Thus, if
˚ 
A WD ¿; f¿g ;

then  is not a total order on the power set

n ˚  ˚ o
P.A/ D ¿; f¿g; f¿g ; ¿; f¿g ;
 www.pdfgrip.com

3.9 Ordering Relations 183

because
˚ 
˚ f¿g
 
6 f¿g ;
f¿g 6 f¿g:
 ˚  
Thus
˚ the   contains neither the pair f¿g;
 relation f¿g nor the pair
f¿g ; f¿g . Instead, for this set A the relation  takes the following form:
˚  ˚ 
f¿g  ¿; f¿g

[ [

¿  f¿g

Some relations that are not total orders can restrict to total orders on subsets.
3.191 Definition (chain). For each set A, and for each relation R  A A, a subset
B  A is a chain if and only if the restriction RjB is a total order on B.
In particular, for each total order R on each set A, the set A is a chain relative to R.
In partially ordered sets, a subset might have a first element, which precedes
every element of that subset, or a last element, which follows every element of that
subset.
3.192 Definition (first or last element). For each set A, for each subset B  A,
and for each partial order R  A A, an element X 2 B is a minimum, or first, or
smallest, element in B if and only if .X; Y/ 2 R for each Y 2 B. A first element of
B is also denoted by min.B/. Similarly, an element Z 2 B is a maximum, or last,
or largest, element in B if and only if .Y; Z/ 2 R for each Y 2 B. A last element of
B is also denoted by max.B/.
3.193 Example. For each set A and for the relation  on P.A/, the element ¿ 2
P.A/ is a first element of P.A/. Indeed, ¿  C for each C 2 P.A/. Also, the
element A 2 P.A/ is a last element of P.A/. Indeed, C  A for each C 2 P.A/.
3.194 Definition (well-ordering). For each set A and for each partial order R 
A A, the set A is well-ordered by R if and only if each nonempty subset B  A has
a first element. A relation R is called a well-ordering (for the lack of a grammatical
and logically equivalent terminology) if and only if A is well-ordered by R.
3.195 Remark. Instead of requiring that a well-ordering be a partial order [30,
p. 31], some texts impose the stronger requirement that a well-ordering be a total
order [71, p. 29]. Yet the insistence on a total order is redundant; indeed, if each
nonempty subset has a first element, then a partial order is automatically a total
order [128, p. 74–76].
 www.pdfgrip.com

184 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.196 Example. The set f¿g D P.¿/ is well-ordered by the relation of inclusion
. Indeed, the only nonempty subset of f¿g D P.¿/ is f¿g, and it has a first
element, namely ¿, because ¿  C for each C 2 f¿g.
3.197 Theorem. Every subset of a well-ordered set is well-ordered.
Proof. Each nonempty subset E  B of a subset B  A of a set A with a well-order
 is also a subset E  A and hence has a first element. Thus  induces a well-order
on B. t
u
3.198 Definition (upper or lower bound). For each set A, for each subset B  A,
and for each partial order R  A A, an element Z 2 A is an upper bound for B
if and only if .Y; Z/ 2 R for each Y 2 B. Similarly an element X 2 A is a lower
bound for B if and only if .X; Y/ 2 R for each Y 2 B.
Thus a last element can differ from an upper bound because an upper bound
need not belong to the same subset. Similarly a first element can differ from a lower
bound because a lower bound need not belong to the same subset.
3.199 Definition (maximal element). For each set A, for each subset B  A, and
for each partial order R  A A, an element Z 2 B is a maximal element of B if
and only if Œ.Z; Y/ 2 R ) Œ.Y; Z/ 2 R for each Y 2 B. In other words, Z 2 B
is a maximal element of B if and only if every element Y 2 B that follows Z also
precedes Z (which allows for the possibility that no Y 2 B follows Z).
Similarly, an element Z 2 B is a minimal element of B if and only if Œ.Y; Z/ 2
R ) Œ.Z; Y/ 2 R for each Y 2 B.
Thus, either a maximal element precedes no element, or it follows every element
that it precedes, but in either case it belongs to the same subset. For instance, a
last element is a maximal element. However, a maximal element need not be a last
element.
3.200 Example. Consider the set A WD f0; 1; 2g and the relation

Q WD f.0; 0/; .0; 1/; .1; 0/; .1; 1/; .2; 2/g:

The element Z WD 2 is maximal in A, because the hypothesis .Z; Y/ 2 Q in Œ.Z; Y/ 2

Q ) Œ.Y; Z/ 2 Q is False for each Y 6D 2 in A. Yet 2 is not a last element in A,
because 2 does not follow every element: .0; 2/ … Q and .1; 2/ … Q.
Such notions as the geometric direction on an “infinite line, plane, or space”
require another axiom, the axiom of infinity, which forms the subject of chapter 4.
Axiom S7 (Axiom of infinity) There exists a set I, such that ¿ 2 I, and for each
element C 2 I its successor C [ fCg is also an element of I:
  
` 9I .¿ 2 I/ ^ 8C f.C 2 I/ ) Œ.C [ fCg/ 2 Ig :

The concept of well-ordered sets allows for many logically equivalent statements
of the last axiom of Zermelo-Frænkel set theory [30, Ch. 1, 9, p. 23, #9.2(3)].
 www.pdfgrip.com

3.9 Ordering Relations 185

Axiom S8 (Axiom of choice) For each set F of nonempty

S sets (A 6D ¿ for each
A 2 F ), there is a “choice” function F W F ! F with F.A/ 2 A for each
A 2 F.
The axiom of choice is logically equivalent to each of the following two
theorems, as proved in chapter 6.
3.201 Theorem (Zermelo’s Theorem). Every set is well-ordered by some
relation.
3.202 Theorem (Zorn’s Lemma). In a pre-ordered set A, if every chain in A has
an upper bound, then A has a maximal element.

3.9.3 Exercises on Ordering Relations

For the following exercises, determine all of the characteristics — among connec-
tivity, strong connectivity, reflexivity, irreflexivity, symmetry, anti-symmetry, asym-
metry, transitivity — of the given relation on the set A WD f0; 1; 2; 3; 4; 5; 6; 7; 8; 9g.
3.171 .
8 9
ˆ
ˆ .0; 9/ .1; 9/ .3; 9/ .6; 9/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 8/ .1; 8/ .2; 8/ .4; 8/ .6; 8/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 7/ .1; 7/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 6/ .1; 6/ .2; 6/ .3; 6/ .4; 6/ >
>
ˆ
< >
=
.0; 5/ .1; 5/
Q WD :
ˆ
ˆ .0; 4/ .1; 4/ .2; 4/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 3/ .1; 3/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 2/ .1; 2/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 1/ >
>
:̂ >
;

3.172 .
8 9
ˆ
ˆ .0; 9/ .1; 9/ .3; 9/ .6; 9/ .9; 9/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 8/ .1; 8/ .2; 8/ .4; 8/ .6; 8/ .8; 8/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 7/ .1; 7/ .7; 7/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 6/ .1; 6/ .2; 6/ .3; 6/ .4; 6/ .6; 6/ >
>
ˆ
< >
=
.0; 5/ .1; 5/ .5; 5/
R WD :
ˆ .0; 4/
ˆ .1; 4/ .2; 4/ .4; 4/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 3/ .1; 3/ .3; 3/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 2/ .1; 2/ .2; 2/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 1/ .1; 1/ >
>
:̂ >
;
.0; 0/
 www.pdfgrip.com

186 3 Set Theory: Proofs by Detachment, Contraposition, and Contradiction

3.173 .
8 9
ˆ
ˆ .0; 9/ .1; 9/ .3; 9/ .6; 9/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 8/ .1; 8/ .2; 8/ .4; 8/ .6; 8/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 7/ .1; 7/ >
>
ˆ
ˆ >
ˆ
ˆ
ˆ .0; 6/ .1; 6/ .2; 6/ .3; 6/ .4; 6/ .8; 6/ .9; 6/ >
>
>
>
< =
.0; 5/ .1; 5/
S WD :
ˆ .0; 4/
ˆ .1; 4/ .2; 4/ .6; 4/ .8; 4/ >
>
ˆ
ˆ >
ˆ
ˆ
ˆ .0; 3/ .1; 3/ .6; 3/ .9; 3/ >
>
>
>
ˆ
ˆ >
>
ˆ
ˆ .0; 2/ .1; 2/ .4; 2/ .6; 2/ .8; 2/ >
>
ˆ
ˆ >
ˆ .0; 1/ .2; 1/ .3; 1/ .4; 1/ .5; 1/ .6; 1/ .7; 1/ .8; 1/ .9; 1/ >
>
>
:̂ ;
.1; 0/ .2; 0/ .3; 0/ .4; 0/ .5; 0/ .6; 0/ .7; 0/ .8; 0/ .9; 0/

3.174 .
8 9
ˆ
ˆ .0; 9/ .1; 9/ .3; 9/ .6; 9/ .9; 9/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 8/ .1; 8/ .2; 8/ .4; 8/ .6; 8/ .8; 8/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 7/ .1; 7/ .7; 7/ >
>
ˆ
ˆ >
ˆ
ˆ
ˆ .0; 6/ .1; 6/ .2; 6/ .3; 6/ .4; 6/ .6; 6/ .8; 6/ .9; 6/ >
>
>
>
< =
.0; 5/ .1; 5/ .5; 5/
U WD :
ˆ
ˆ .0; 4/ .1; 4/ .2; 4/ .4; 4/ .6; 4/ .8; 4/ >
>
ˆ
ˆ >
ˆ
ˆ
ˆ .0; 3/ .1; 3/ .3; 3/ .6; 3/ .9; 3/ >
>
>
>
ˆ
ˆ >
>
ˆ
ˆ .0; 2/ .1; 2/ .2; 2/ .4; 2/ .6; 2/ .8; 2/ >
>
ˆ
ˆ >
ˆ .0; 1/ .1; 1/ .2; 1/ .3; 1/ .4; 1/ .5; 1/ .6; 1/ .7; 1/ .8; 1/ .9; 1/ >
>
>
:̂ ;
.0; 0/ .1; 0/ .2; 0/ .3; 0/ .4; 0/ .5; 0/ .6; 0/ .7; 0/ .8; 0/ .9; 0/

3.175 .
8 9
ˆ
ˆ .1; 9/ .9; 9/ >
>
ˆ
ˆ >
>
ˆ
ˆ .1; 8/ .8; 8/ >
>
ˆ
ˆ >
>
ˆ
ˆ .1; 7/ .7; 7/ >
>
ˆ
ˆ >
>
ˆ
ˆ .1; 6/ .6; 6/ >
>
ˆ
< >
=
.1; 5/ .5; 5/
V WD :
ˆ
ˆ .1; 4/ .4; 4/ >
>
ˆ
ˆ >
ˆ
ˆ
ˆ .1; 3/ .3; 3/ .9; 3/ >
>
>
>
ˆ
ˆ >
>
ˆ
ˆ .1; 2/ .2; 2/ .4; 2/ .8; 2/ >
>
ˆ
ˆ >
>
ˆ .0; 1/ .1; 1/ >
>
:̂ ;
.1; 0/
 www.pdfgrip.com

3.9 Ordering Relations 187

3.176 .
8 9
ˆ
ˆ .1; 9/ .3; 9/ .9; 9/ >
>
ˆ
ˆ >
>
ˆ
ˆ .1; 8/ .2; 8/ .8; 8/ >
>
ˆ
ˆ >
>
ˆ
ˆ .1; 7/ .7; 7/ >
>
ˆ
ˆ >
>
ˆ
ˆ .1; 6/ .6; 6/ >
>
ˆ
< >
=
.1; 5/ .5; 5/
W WD :
ˆ
ˆ .1; 4/ .2; 4/ .4; 4/ >
>
ˆ
ˆ >
ˆ
ˆ
ˆ .1; 3/ .3; 3/ .9; 3/ >
>
>
>
ˆ
ˆ >
>
ˆ
ˆ .1; 2/ .2; 2/ .4; 2/ .8; 2/ >
>
ˆ
ˆ >
ˆ .0; 1/ .1; 1/ .2; 1/ .3; 1/ .4; 1/ .5; 1/ .6; 1/ .7; 1/ .8; 1/ .9; 1/ >
>
>
:̂ ;
.1; 0/

3.177 . Prove that for each set A the empty relation ¿  A A is a partial order.
3.178 . Prove that a relation R is symmetric if and only if R D R ı1 .
3.179 . Prove that a relation R on A is irreflexive if and only if R \ A D ¿.
3.180 . Prove that a relation R on a set A is anti-symmetric if and only if R \
R ı1  A .
3.181 . Prove that every strict partial order is asymmetric
3.182 . Prove that if a relation is irreflexive and anti-symmetric, then it is also
asymmetric
3.183 . Prove that every asymmetric relation is also anti-symmetric.
3.184 . Prove that every asymmetric relation is also irreflexive.
3.185 . Prove that a relation is a strict partial order if and only if it is asymmetric
and transitive
3.186 . Provide an example of an anti-symmetric but not asymmetric relation.
3.187 . Provide an example of an asymmetric and strongly connected relation.
3.188 . Exhibit an example of a set A for which the relation  on P.A/ is not a
total order: supply subsets B and C such that neither B  C nor C  B.
3.189 . Exhibit an example of a set A and a subset S  P.A/ that is a chain with
respect to the relation  on P.A/.
3.190 . Prove that if a function is surjective, then it has a right inverse.
 www.pdfgrip.com

Chapter 4
Mathematical Induction: Definitions and Proofs
by Induction

4.1 Introduction

This chapter introduces the concepts of mathematical induction and recursion.

Starting from first-order logic and the Zermelo-Fraenkel axioms of set theory,
including the axiom of infinity, the chapter establishes the theoretical basis for
proofs by the Principle of Mathematical Induction, followed by definitions through
mathematical induction, which forms the basis for the concept of primitive-
recursive functions. As an extended example, one section defines integer addition
and multiplication as primitive-recursive functions, and hence derives the basic
properties of integer arithmetic (associativity, commutativity, and distributivity) by
induction. Subsequent sections define ordering relations and derive cancellation
laws and exponential laws for integer and rational arithmetic. Another section
then applies arithmetic to the cardinalities of unions, intersections, differences, and
Cartesian products of finite sets. Yet another section focuses on denumerable and
other not-necessarily finite sets, leading to J. M. Whitaker’s proof of the Bernstein-
Cantor-Schröder Theorem within Zermelo-Fraenkel set theory. The prerequisites
for this chapter consist of a working knowledge of first-order logic and Zermelo-
Fraenkel set theory, for instance, as described in chapters 1, 2, and 3, which contain
all the logical and set-theoretical theorems cited in this chapter.
The concepts of “numbers” and “counting” allow for the determination and
comparison of the “sizes” of various objects. The same concepts also lead to precise
definitions of “infinite” sets, which then reveals an “infinite” variety of “infinite”
sets. For example, the number of particles in the universe is related to its mass,
volume, and density, which quantities might affect the fate of the universe, in
particular, whether the universe will remain finite or will expand to “infinity” [69,
p. 445].
Mathematically, this chapter shows how the concepts of “numbers,” “arithmetic,”
“counting,” “computing,” and “infinity” fit within set theory, without any new
axiomatic system.

© Springer Science+Business Media New York 2015 189

Y. Nievergelt, Logic, Mathematics, and Computer Science,
DOI 10.1007/978-1-4939-3223-8_4
 www.pdfgrip.com

190 4 Mathematical Induction: Definitions and Proofs by Induction

4.2 Mathematical Induction

The principle of “mathematical induction” provides a method for proving theorems,

and also a method for specifying and analyzing many practical numerical calcula-
tions.

4.2.1 The Axiom of Infinity

The Principle of Mathematical Induction forms the theoretical basis for such
algorithms as counting and arithmetic, by means of sequences of definitions,
computations, and verifications, where the length of the sequence depends on the
situation. One framework to allow for sequences of yet unspecified lengths consists
in embedding all such sequences into one set N that already allows for sequences
of all lengths. To this end, with a construction attributed to John von Neumann [8,
p. 22], [128, p. 129], [135], for each element X 2 N, the set N also contains a “next”
element X [ fXg.
4.1 Definition. For each set X the successor of X is X [ fXg.
4.2 Example. The successor of the empty set ¿ is the set ¿ [ f¿g D f¿g.
A subsequent theorem will verify that for each set X 2 N the successor X [ fXg
is strictly larger than X, so that X ¤ .X [ fXg/. Within the theory presented so
far, however, nothing guarantees the existence of a set containing the successor of
each of its elements. To this end, a new axiom becomes necessary [8, p. 21]. The
following version is identical to that of axiom S7 already mentioned in chapter 3.

Axiom S7 (Axiom of infinity) There exists a set I, such that ¿ 2 I, and for each
element C 2 I its successor C [ fCg is also an element of I:
  
` 9I .¿ 2 I/ ^ 8C f.C 2 I/ ) Œ.C [ fCg/ 2 Ig :

Digits or other symbols can abbreviate the set notation for the elements of I.
4.3 Example. The set I described in axiom S7 contains the following elements.

0 WD ¿;

1 WD 0 [ f0g D ¿ [ f¿g D f¿g;

˚  ˚ 
2 WD 1 [ f1g D f¿g [ f¿g D ¿; f¿g ;
 www.pdfgrip.com

4.2 Mathematical Induction 191

˚  n˚ o n ˚ o
3 WD 2 [ f2g D ¿; f¿g [ ¿; f¿g D ¿; f¿g; ¿; f¿g ;

˚  n ˚ o
4 WD 3 [ f3g D ¿; f¿g; ¿; f¿g ; ¿; f¿g; ¿; f¿g ;
::
:

Such elements are called “natural numbers” (see remark 4.7 about 0).
The axiom of infinity does not restrict I to contain only the elements in
example 4.3, so that the set I might also contain other elements. Therefore, a further
construction with the preceding axioms becomes necessary to form a set containing
only the natural numbers. The construction selected here forms the intersection of
subsets of I, using a method common to several parts of mathematics [30, p. 66],
[34, p. 21], [60, p. 132]. Specifically, denote by P the logical formula defined by
  
.P/ , .¿ 2 A/ ^ 8X f.X 2 A/ ) Œ.X [ fXg/ 2 Ag :

Thus the formula P asserts that a set A contains among its elements the empty set
and the successor of every one of its elements. The Axiom of Infinity asserts that
there exists a set I for which SubfAI .P/ is True:
  
` 9I .¿ 2 I/ ^ 8X f.X 2 I/ ) Œ.X [ fXg/ 2 Ig :

The following construction forms a subset N  I containing only elements such as

those in example 4.3. To this end, consider the set F  P.I/ of all subsets B  I
for which SubfAB .P/ is also True:

F WD fB 2 P.I/ W SubfAB .P/g

n
D B 2 P.I/ W
  o
.¿ 2 B/ ^ 8X f.X 2 B/ ) Œ.X [ fXg/ 2 Bg :
T
The following definition and theorems will confirm that N can be defined as F.
4.4 Theorem. The set F D fB 2 P.I/ W SubfAB .P/g is not empty: I 2 F .
Proof. By the Axiom of Infinity (S7), the formula P is True for I, so that SubfAI .P/
is True. Moreover, I  I. Therefore, I is a subset of I for which P is True, so that
.I  I/ ^ ŒSubfAI .P/ holds, which means that I 2 F . t
u
From I the following definition extracts N as the smallest subset for which P
holds.
 www.pdfgrip.com

192 4 Mathematical Induction: Definitions and Proofs by Induction

4.5 Definition (natural numbers). The set of natural numbers, denoted by N, is

the intersection of all the elements of F :
\
N WD F

A natural number is an element of N. Also, define N WD N n f¿g.

4.6 Remark. In the present theory, every natural number is a set.
4.7 Remark. Some texts exclude the element ¿ from the set N. The two definitions
differ only in their terminology and lead to the same theory. Yet the definition of N
with ¿ 2 N has proved more convenient than without it in the context of set theory
[128, p. 121], ordinals [74], and in such situations as Kurt Gödel’s work [46] on
logic and mathematics. Therefore, the definition adopted here includes ¿ 2 N.
The following theorems verify that P also holds for N.
T
4.8 Theorem. The set N D F is not empty; indeed, ¿ 2 N.
S
Proof. First, ¿ 2 I and I 2 F by theorem 4.4, whence ¿ 2 F . Second,
T for each
B 2 F the formula SubfAB .P/ holds, hence ¿ 2 B; consequently, ¿ 2 F D N.
t
u
4.9 Theorem. The formula P is True for N: 8X f.X 2 N/ ) Œ.X [ fXg/ 2 Ng.
T
Proof. For each X, if X 2 N D F , then X 2 BTfor each B 2 F , but then
.X [ fXg/ 2 B for each B 2 F , whence .X [ fXg/ 2 F D N. t
u
4.10 Theorem. The set N is an element of F .
Proof.TTheorems 4.8 and 4.9 show that ¿ 2 N and SubfAN .P/ are True. Moreover,
N D F  I. Thus, .N  I/ ^ ŒSubfAN .P/ holds, whence N 2 F . t
u
The concept of “successor” can serve to specify functions defined on N.
4.11 Example. The “successor” function is defined by
˚ 
G WD .X; Y/ 2 .N N/ W Y D .X [ fXg/ ;

so that
G W N ! N;
X 7! X [ fXg:

4.12 Definition (sequence). For each set E, a sequence in E is a function F W N !

E. For each N 2 N, the value F.N/ can also be denoted by FN , and then the function
F can also be denoted by .FN / or .FN /N2N . Also, a finite sequence of length L in E
is a function S W L ! E defined on some L 2 N. The value S.K/ is also denoted by
SK ; then the function S is also denoted by .SK / or .SK /K2L .
4.13 Example. The “successor” function in example 4.11 is a sequence in N.
For convenience, mathematical usage abbreviates the successor N [fNg as N C1.
 www.pdfgrip.com

4.2 Mathematical Induction 193

4.14 Definition. For each natural number N, let N C 1 WD N [ fNg .

In general, the specification of functions defined on N requires a method known
as the Principle of Mathematical Induction, as described in the next subsections.

4.2.2 The Principle of Mathematical Induction

The following theorem forms the theoretical basis for the methods of proof by
induction and recursive computation. Specifically, the “Principle of Mathematical
Induction” shows that if a subset S  N contains ¿, and if S also contains the
successor X [ fXg of each of its elements X, then S contains all the natural numbers:
S D N.
4.15 Theorem (Principle of Mathematical Induction). For each subset S  N, if
S contains the
˚ empty set and the successor
  one of its elements, then S D N.
of every
Thus, ` 8S .S  N/ ^ SubfAS .P/ ) .S D N/ ; or, with SubfAS .P/ spelled out:
   
` 8S .S  N/ ^ .¿ 2 S/ ^ 8X f.X 2 S/ ) Œ.X [ fXg/ 2 Sg ) .S D N/ :
T
Proof. If S  N then S  N D F  I. If moreover SubfAST .P/ is True, then
ST2 F by definition
T of F . From S 2 F , it then follows that TF  S. Thus,
F  S  F , whence (by theorem 3.10) equality holds: S D F D N. u t
4.16 Remark. The proof of theorem 4.15 derives, or, equivalently, deduces the
Principle of Mathematical Induction from the preceding axioms and inference rules
of logic and axioms or postulates of set theory. Therefore, in mathematics, all
inductive proofs are deductive:
In mathematical English, the words “inductive” and “deductive” may be used interchange-
ably for any kind of reasoning, except that the principle listed in [theorem 4.15] is usually
called the Induction Principle [114, p. 195, Remark].

The following example demonstrates a pattern amenable to induction.

4.17 Example. The elements of N in example 4.3 reveal the following pattern:
• Every element of 0 is also a subset of 0, so that ` 8XŒ.X 2 0/ ) .X  0/. The
hypothesis X 2 0 is False, whence the implication is universally valid.
• Every element of 1 D f¿g  a subset of 1, because ¿ 2 1 and ¿  1.
˚ is also
• Every element of 2 D ¿; f¿g is also a subset of 2. Indeed ¿ 2 2 and ¿  2;
also, f¿g 2 2 and f¿g  2. n ˚ o
• Similarly, every element of 3 D ¿; f¿g; ¿; f¿g is also a subset of 3. Indeed
˚ 
¿ 2˚3 and ¿  3; moreover, f¿g 2 3 and f¿g  3; furthermore, ¿; f¿g 2 3
and ¿; f¿g  3.
 www.pdfgrip.com

194 4 Mathematical Induction: Definitions and Proofs by Induction

The proof of the following theorem shows the use of the Principle of Mathemat-
ical Induction to verify the pattern of example 4.17 for all natural numbers.
4.18 Theorem. For all L 2 N and N 2 N, if L 2 N then L  N.
Proof. This proof proceeds by induction with N. Define a set S  N by

S WD fN 2 N W 8LŒ.L 2 N/ ) .L  N/g:

Initial step

If N WD ¿, then L 2 ¿ is False, whence .L 2 ¿/ ) .L  ¿/ is universally valid;

thus ¿ 2 S, by definition of S.

Inductive hypothesis

Assume K 2 S; thus 8LŒ.L 2 K/ ) .L  K/ is True.

Inductive step

To verify the theorem for N WD K C 1, assume that L 2 .K C 1/ D .K [ fKg/; then

L 2 K or L 2 fKg, by definition of the union K [ fKg.
In the case L 2 K, it follows from the inductive hypothesis that L  K, whence
L  K  .K [ fKg/ D .K C 1/ and then L  .K C 1/.
In the case L 2 fKg, it follows that L D K, whence L D K  .K[fKg/ D .KC1/
and then L  .K C 1/.
Consequently L  .K C 1/ in either case. Because L  .K C 1/ for every L 2 K,
it follows that .K C 1/ 2 S.

Completion of the proof by induction

From the initial step ¿ 2 S, and from the inductive step, if K 2 S then .K C 1/ 2 S;
it follows from the Principle of Mathematical Induction (theorem 4.15) that S D N.
Thus N 2 S for every N 2 N: for every N 2 N, if L 2 N then L  N. t
u
Theorem 4.18 holds for all natural numbers, but it can fail for other sets, as
demonstrated in counterexample 4.19.
˚ 
4.19 Counterexample. If X D f¿g and Y D f¿g , then X 2 Y but X 6 Y,
because ¿ 2 X but ¿ … Y . Theorem 4.18 does not apply to Y because Y … N.
Similarly, theorem 4.20 shows that every natural number is also a subset of N.
4.20 Theorem. For every set N, if N 2 N, then N  N.
 www.pdfgrip.com

4.2 Mathematical Induction 195

Proof. This proof proceeds by induction with N.

For N D 0 D ¿, ¿  N by theorem 3.11.
As an induction hypothesis, assume that there exists K 2 N such that K  N.
For the induction step, from K 2 N follows fKg  N by theorem 3.22. Also,
K  N by induction hypothesis. Hence K C 1 D K [ fKg  N. t
u
Theorem 4.20 holds for N but fails for other sets, as already demonstrated in
counterexample 4.19.
The following subsection demonstrates how to use the Principle of Mathematical
Induction to prove the “existence” of certain functions.

4.2.3 Definitions by Mathematical Induction

The preceding discussion has introduced the Principle of Mathematical Induction

as a method to prove theorems and verify formulae involving natural numbers. In
addition, the following theorem shows how the Principle of Mathematical Induction,
and the concept of unions of functions, also form the foundation of a method to
define by induction the values of some functions. The same method is also known
as recursion [18, p. 322, n. 526], [63, p. 10]; the resulting functions are also called
recursive functions, with a terminology attributed [49, p. 167] to Kurt Gödel [43],
[46, p. 46].
4.21 Theorem (Definition by Induction or Recursion). For each nonempty set C,
for each element A 2 C, and for each function G W C ! C with domain D.G/ D
C, there exists exactly one function F W N ! C that satisfies the following two
conditions:
(DMI.0) F.0/ D A.
(DMI.1) F.N C 1/ D GŒF.N/ for each N 2 N.
Proof. This proof verifies by induction that there exists a sequence of functions
.FN /, with each function FN  .N C/ defined on the set SN WD .N [ fNg/ so that
the following two conditions hold:
(N:0) FN .0/ D A.
(N:1) FN .I C 1/ D GŒFN .I/ for each I 2 SN n fNg D N.
S
Then the proof ends by verifying that the function F defined by F WD N2N FN ,
so that F.N/ D FN .N/, satisfies all the requirements.

Initial step

For N D 0, consider the set S0 D f0g, and consider the function

 www.pdfgrip.com

196 4 Mathematical Induction: Definitions and Proofs by Induction

F0 W f0g ! C;
0 7! A:

Then the function F0 just defined satisfies the following two conditions:
(0:0) F0 .0/ D A.
(0:1) F0 .I C 1/ D GŒF0 .I/ for each I 2 S0 n f0g D ¿.
Moreover, there exists only one function of singletons F0 W f0g ! fAg.

Induction hypothesis

Assume that there exists a natural number J 2 N such that the theorem holds for
N WD J, so that for each L 2 SJ D J [ fJg there exists exactly one function FL
satisfying the two conditions .N:0/ and .N:1/ for N WD L.

Induction step

Let S  N consist of every N 2 N for which there exists exactly one function FN
satisfying .N:0/ and .N:1/.
There exists exactly one function of singletons HJ W fJ C 1g ! fGŒFJ .J/g,
because there exists exactly one function FJ and hence exactly one value FJ .J/.
Define FJC1 WD FJ [H P J , which is a function because of the disjoint domains
SJ \ fJ C 1g D ¿ (by definition 3.94 in chapter 3).
For I WD J C 1, the definition of FJC1 gives FJC1 .J C 1/ D HJ .J C 1/ D
GŒFJ .J/ D GŒFJC1 .J/. For each L 2 SJ n fJg D J, the definition of FJC1 gives
FJC1 .L C 1/ D FJ .L C 1/ D GŒFJ .L/ D GŒFJC1 .L/ so that the following two
conditions hold:
(J C 1:0) FJC1 .0/ D FJ .0/ D A,
(J C 1:1) FJC1 .L C 1/ D GŒFJC1 .L/ for every L 2 SJC1 n fJ C 1g.
Moreover, there exists exactly one such function FJC1 because there exists exactly
one restriction FJ jSJ , namely FJ , and exactly one value for FJC1 .J C 1/ D GŒFJ .J/.

Completion of the proof of the theorem

Having constructed the sequence of functions .FN /, define a function F W N ! C

by F.N/ WD FN .N/. Specifically, let
[
F WD FN :
N2N

Then verify that

 www.pdfgrip.com

4.2 Mathematical Induction 197

(DMI.1) F.0/ D F0 .0/ D 0.

(DMI.2) F.N C 1/ D FNC1 .N C 1/ D GŒFNC1 .N/ D GŒFN .N/ D GŒF.N/.
The uniqueness of F follows from the uniqueness of each restriction FjNC1 D FN .
t
u
Because the proof of the validity of the method of recursion (theorem 4.21) relies
on mathematical induction (theorem 4.15), recursion is a logical consequence of
induction. The following sections show how recursion suffices to define arithmetic
with natural numbers.

4.2.4 Exercises on Mathematical Induction

4.1 . With 5 WD 4 [ f4g, write 5 in terms of sets, as in example 4.3.

4.2 . With 6 WD 5 [ f5g, write 6 in terms of sets, as in example 4.3.
4.3 . Prove that every set is an element and a subset of its successor.
4.4 . Provide an example of a nonempty proper subset S ¤ N that contains the
successor of every one of its elements.
Prove or disprove each of the following statements for all sets A and B.
4.5 . Prove or disprove that if A  B, then .A [ fAg/  .B [ fBg/ .
4.6 . Prove or disprove that .A [ fAg/ [ .B [ fBg/ D .A [ B/ [ fA [ Bg.
4.7 . Prove or disprove that .A [ fAg/ \ .B [ fBg/ D .A \ B/ [ fA \ Bg.
4.8 . Prove or disprove that .A [ fAg/ n .B [ fBg/ D .A n B/ [ fA n Bg.
Prove or disprove each of the following statements for all K; L; M; 2 N.
4.9 . Prove that if K 2 M and L 2 M, then K [ L 2 M.
4.10 . Prove that if K 2 M and L 2 M, then K \ L 2 M.
4.11 . Outline a formal proof of theorem 4.4.
4.12 . Outline a formal proof of theorem 4.8.
4.13 . Outline a formal proof of theorem 4.9.
4.14 . Outline a formal proof of theorem 4.10.
4.15 . Outline a formal proof of theorem 4.15.
S
4.16 . Prove that N D N.
T
4.17 . Prove that N D ¿.
4.18 . For the successor function, prove that N D GıN .¿/ for each N 2 N .
 www.pdfgrip.com

198 4 Mathematical Induction: Definitions and Proofs by Induction

4.19 . Prove that if a subset V  N contains 2, and if V also contains the

“successor” X [fXg of each of its element X, then V contains all the natural numbers
except 0 and 1: V D N n f ¿; f¿g g. In other words, replace ¿ by f ¿; f¿g g in P
to obtain the logical formula R defined by
 
.R/ , .f ¿; f¿g g 2 A/ ^ 8X f.X 2 A/ ) Œ.X [ fXg/ 2 Ag :
˚  
Prove that ` 8V .V  N/ ^ ŒSubfAV .R/ ) .V D N n f ¿; f¿g g/ .
4.20 . Prove that if a subset U  N contains 1, and if U also contains the
“successor” X [ fXg of each of its elements X, then U contains all the positive
natural numbers: U D N n f¿g. In other words, replace ¿ by f¿g in P to obtain the
logical formula Q defined by
 
.Q/ , .f¿g 2 A/ ^ 8X f.X 2 A/ ) Œ.X [ fXg/ 2 Ag :
˚  
Prove that ` 8U .U  N/ ^ ŒSubfAU .Q/ ) .U D N n f¿g/ .

4.3 Arithmetic with Natural Numbers

4.3.1 Addition with Natural Numbers

This subsection defines the addition of natural numbers and establishes some of
its properties, all by induction. Besides explaining the foundations of integer arith-
metic, the following considerations also provide examples of proofs by induction.
4.22 Definition (Addition). For every M 2 N, define

M C 0 WD M; .A0/
M C 1 WD M [ fMg: .A1/

Then for every M 2 N, define an addition function by induction, so that for every
N 2 N,

M C .N C 1/ WD .M C N/ C 1: .A2/

4.23 Remark. According to definition 4.22, adding N amounts to adding 1 repeat-

edly N times. With an alternative but logically equivalent notation, for each M 2 N
definition 4.22 uses theorem 4.21 and the successor function

G W N ! N;
N 7! N C 1;
 www.pdfgrip.com

4.3 Arithmetic with Natural Numbers 199

to specify by induction (recursively) an addition function F .M/ W N ! N such that

F .M/ .0/ WD M; .A0/

G.L/ WD L C 1;  .A1/
F .M/ .N C 1/ WD G F .M/ .N/ : .A2/

The following theorem shows that addition is associative.

4.24 Theorem. For all P; Q; R 2 N, .P C Q/ C R D P C .Q C R/.
Proof. For each P and each Q, proceed by induction with R. For R WD 0,

.P C Q/ C 0 D P C Q by (A0) with M WD .P C Q/;

D P C .Q C 0/ by (A0) with M WD Q:

Second, assume that there exists some K 2 N such that the theorem holds for R WD
K, so that .P C Q/ C K D P C .Q C K/ for each P and each Q; then

.P C Q/ C .K C 1/ D Œ.P C Q/ C K C 1 (A2), M WD .P C Q/, N WD K;

D ŒP C .Q C K/ C 1 induction hypothesis;
D P C Œ.Q C K/ C 1 (A2), M WD P, N WD .Q C K/;
D P C ŒQ C .K C 1/ (A2), M WD Q, N WD K:

t
u
The following three theorems show that addition commutes. The first theorem
shows that adding 0 commutes.
4.25 Theorem. For each N 2 N, 0 C N D N.
Proof. Proceed by induction with N. First, establish the conclusion for N WD 0:

0 C 0 D 0 by (A0) with M WD 0:

Second, assume that 0 C K D K for some K 2 N, so that the theorem holds for
N WD K; then

0 C .K C 1/ D .0 C K/ C 1 by (A2);
D KC1 induction hypothesis:

t
u
The second theorem shows that adding 1 commutes.
4.26 Theorem. For each P 2 N, 1 C P D P C 1.
 www.pdfgrip.com

200 4 Mathematical Induction: Definitions and Proofs by Induction

Proof. Proceed by induction with P. First, for P WD 0,

1C0 D 1 by (A0) with M WD 1;

D 0 C 1 by theorem 4.25:

Second, assume that 1 C K D K C 1 for some K 2 N, so that the theorem holds for
P WD K; then

1 C .K C 1/ D .1 C K/ C 1 (A2) with M WD 1 and N WD K;

D .K C 1/ C 1 induction hypothesis:
t
u
Finally, the third theorem shows that addition commutes.
4.27 Theorem. For all natural numbers P and Q, P C Q D Q C P.
Proof. For each P 2 N, proceed by induction with Q. First,

PC0 D P by (A0);
D 0 C P by theorem 4.25:

Second, assume that there exists K 2 N such that the theorem holds for Q WD K, so
that R C K D K C R for each R 2 N; then for each P 2 N,

P C .K C 1/ D P C .1 C K/ induction hypothesis with R WD 1;

D .P C 1/ C K theorem 4.24;
D K C .P C 1/ induction hypothesis with R WD .P C 1/;
D K C .1 C P/ theorem 4.26;
D .K C 1/ C P theorem 4.24:

t
u

4.3.2 Multiplication with Natural Numbers

This subsection defines by induction the multiplication of natural numbers and

establishes some of its properties, as well as some relations between addition and
multiplication. Besides a continuation of the foundations of integer arithmetic, the
following considerations also provide further examples of proofs by induction.
4.28 Definition (Multiplication). For every nonnegative integer M 2 N, define a
multiplication function by induction based on the following specifications:

M  0 WD 0; .M0/
M  .N C 1/ WD .M  N/ C M: .M1/
 www.pdfgrip.com

4.3 Arithmetic with Natural Numbers 201

4.29 Remark. By definition 4.28, multiplying M and N amounts to starting from

0 and adding M repeatedly N times. With an alternative but logically equivalent
notation, for each M 2 N, definition 4.28 uses a recursive definition with G (in
theorem 4.21) replaced by the addition function F .M/ from remark 4.23,

G WD F .M/ W N ! N;
N 7! N C M;

to specify by induction a multiplication function H .M/ W N ! N such that

G.L/ WD F .M/ .L/ D L C M;

H .M/ .0/ WD 0;   .M0/
H .N C 1/ WD G H .M/ .N/ : .M1/
.M/

4.30 Example. Multiplication by repeated addition, as in definition 4.28, was used

in the IBM
R
604
R
[64, p. 23 & 71] and IBM 650 R
[65, p. 35 & 36, 2].
The following theorem shows that multiplication distributes over addition on the
right-hand side.
4.31 Theorem. For all P; Q; R 2 N, .P C Q/  R D .P  R/ C .Q  R/.
Proof. For each P and each Q, proceed by induction with R. For R WD 0,

.P C Q/  0 D 0 (M0) with M WD .P C Q/;

D 0C0 (A0) with M WD 0;
D .P  0/ C .Q  0/ (M0) with M WD P and (M0) with M WD Q:

Second, assume that there exists K 2 N such that the theorem holds for R WD K, so
that .P C Q/  K D .P  K/ C .Q  K/ for each P and each Q; then

.P C Q/  .K C 1/ D Œ.P C Q/  K C .P C Q/ (M1), M WD P C Q;
D Œ.P  K/ C .Q  K/ C .P C Q/ induction hypothesis;
D .P  K/ C Œ.Q  K/ C .P C Q/ theorem 4.24;
D .P  K/ C Œf.Q  K/ C Pg C Q theorem 4.24;
D .P  K/ C ŒfP C .Q  K/g C Q theorem 4.27;
D .P  K/ C ŒP C f.Q  K/ C Qg theorem 4.24;
D Œ.P  K/ C P C Œ.Q  K/ C Q theorem 4.24;
D ŒP  .K C 1/ C ŒQ  .K C 1/ (M1) twice:

t
u
The following three theorems show that multiplication commutes. The first
theorem shows that multiplication by 0 commutes.
 www.pdfgrip.com

202 4 Mathematical Induction: Definitions and Proofs by Induction

4.32 Theorem. For each natural number N 2 N, 0  N D 0.

Proof. Proceed by induction with N. First, establish the conclusion for N WD 0:

0  0 D 0 .M0/:

Second, assume that there exists K 2 N such that the theorem holds for N WD K, so
that 0  K D 0 ; then

0  .K C 1/ D .0  K/ C 0 .M1/;
D 0C0 induction hypothesis;
D0 .A0/:
t
u
The second theorem shows that multiplication by 1 commutes.
4.33 Theorem. For each nonnegative integer N 2 N, 1  N D N.
Proof. Proceed by induction with N. First, establish the conclusion for N WD 0:

1  0 D 0 .M0/:

Second, assume that there exists K 2 N such that the theorem holds for N WD K, so
that 1  K D K ; then

1  .K C 1/ D .1  K/ C 1 .M1/;
D KC1 induction hypothesis:
t
u
Finally, the third theorem shows that multiplication commutes.
4.34 Theorem. For all nonnegative integers P; Q 2 N, P  Q D Q  P.
Proof. For each P 2 N, proceed by induction with Q. First, for Q WD 0,

P0 D 0 .M0/;
D 0  P theorem 4.32:

Second, assume that there exists K 2 N such that the theorem holds for Q WD K, so
that P  K D K  P for each P 2 N; then

P  .K C 1/ D .P  K/ C P .M1/;
D .K  P/ C P induction hypothesis;
D .K  P/ C .1  P/ theorem 4.33;
D .K C 1/  P theorem 4.31:

t
u
The next theorem shows that multiplication distributes over addition also on the
left-hand side.
 www.pdfgrip.com

4.3 Arithmetic with Natural Numbers 203

4.35 Theorem. For all P; Q; R 2 N, P  .Q C R/ D .P  Q/ C .P  R/.

Proof. Use commutativity and distributivity on the right-hand side (theo-
rems 4.34, 4.31):

P  .Q C R/ D .Q C R/  P theorem 4.34;
D .Q  P/ C .R  P/ theorem 4.31;
D .P  Q/ C .P  R/ theorem 4.34:

t
u
The following theorem shows that multiplication is associative.
4.36 Theorem. For all P; Q; R 2 N, P  .Q  R/ D .P  Q/  R.
Proof. For all P; Q 2 N, proceed by induction with R. For R WD 0,

P  .Q  0/ D P  0 .M0/;
D0 .M0/;
D .P  Q/  0 (M0) with M WD .P  Q/:

Second, assume that there exists K 2 N such that the theorem holds for R WD K, so
that P  .Q  K/ D .P  Q/  K for all P; Q 2 N ; then

.P  Q/  .K C 1/ D Œ.P  Q/  K C .P  Q/ (M1) with M WD .P  Q/;

D ŒP  .Q  K/ C .P  Q/ induction hypothesis;
D P  Œ.Q  K/ C Q theorem 4.35;
D P  ŒQ  .K C 1/ .M1/:

t
u
There are other arithmetic operations with natural numbers, such as the factorial.
4.37 Definition. For each N 2 N, define NŠ (read “N factorial”) recursively by

0Š WD 1;
.N C 1/Š WD .NŠ/  .N C 1/:

4.3.3 Exercises on Arithmetic by Induction

The following exercises involve the following sets (also defined in example 3.35):

0 WD ¿;
1 WD f0g;
2 WD f0; 1g;
 www.pdfgrip.com

204 4 Mathematical Induction: Definitions and Proofs by Induction

3 WD f0; 1; 2g;
4 WD f0; 1; 2; 3g;
5 WD f0; 1; 2; 3; 4g;
6 WD f0; 1; 2; 3; 4; 5g;
7 WD f0; 1; 2; 3; 4; 5; 6g;
8 WD f0; 1; 2; 3; 4; 5; 6; 7g;
9 WD f0; 1; 2; 3; 4; 5; 6; 7; 8g:

4.21 . Prove that 1 C 1 D 2.

4.22 . Prove that 2 C 1 D 3.
4.23 . Prove that 3 C 1 D 4.
4.24 . Prove that 4 C 1 D 5.
4.25 . Prove that 5 C 1 D 6.
4.26 . Prove that 6 C 1 D 7.
4.27 . Prove that 7 C 1 D 8.
4.28 . Prove that 8 C 1 D 9.
4.29 . Prove that 2 C 2 D 4.
4.30 . Prove that 3 C 2 D 5.
4.31 . Prove that 4 C 2 D 6.
4.32 . Prove that 5 C 2 D 7.
4.33 . Prove that 6 C 2 D 8.
4.34 . Prove that 7 C 2 D 9.
4.35 . Prove that 3 C 3 D 6.
4.36 . Prove that 4 C 3 D 7.
4.37 . Prove that 5 C 3 D 8.
4.38 . Prove that 6 C 3 D 9.
4.39 . Prove that 4 C 4 D 8.
4.40 . Prove that 5 C 4 D 9.
4.41 . Prove that 2  2 D 4.
4.42 . Prove that 3  2 D 6.
4.43 . Prove that 4  2 D 8.
4.44 . Prove that 3  3 D 9.
 www.pdfgrip.com

4.4 Orders and Cancellations 205

4.45 . Prove or disprove that addition distributes over multiplication on the left.
4.46 . Prove or disprove that addition distributes over multiplication on the right.
The following exercises refer to the factorial specified by definition 4.37.
4.47 . Identify a set C and functions F and G that fit theorem 4.21 to justify
definition 4.37.
4.48 . Calculate 0Š, 1Š, 2Š, 3Š.
4.49 . Prove or disprove that .P C Q/Š D .PŠ/ C .QŠ/ for all P; Q 2 N.
4.50 . Prove or disprove that .P  Q/Š D .PŠ/  .QŠ/ for all P; Q 2 N.

4.4 Orders and Cancellations

4.4.1 Orders on the Natural Numbers

The set of natural numbers can model geometric concepts, for instance, a direction
from left to right, and algebraic concepts, for instance, increasing magnitudes:

0 < 1 < 2 < 3 < 4 < 5 < 6 < 7 < 8 < 9 < :::

For both types of concepts — geometric and algebraic — it suffices to introduce an

ordering relation on the natural numbers. To this end, this subsection shows that the
natural numbers are well-ordered by a relation defined by

.M N/ , Œ.M D N/ _ .M 2 N/:

From this well-ordering relation will result the laws of arithmetic cancellations,
which will also allow for the solutions of certain equations. The first results define
a strict order < in terms of the foundational relation 2 of set membership.
4.38 Definition. For all M; N 2 N, define M < N (read “M is less than N”) by

.M < N/ , .M 2 N/:

4.39 Example. 0 < 1 because 0 2 1: 0 ˚D ¿ and1 D f¿g, so ¿ 2˚f¿g. 

1 < 2 because 1 2 2; 1 D f¿g, 2 D ¿; f¿g , whence f¿g 2 ¿; f¿g .
The following theorem establishes the transitivity of the relation 2 on N.
 www.pdfgrip.com

206 4 Mathematical Induction: Definitions and Proofs by Induction

4.40 Theorem. For all L; M; N 2 N, if L 2 M and M 2 N, then L 2 N:

8L8M8NfŒ.L 2 N/ ^ .M 2 N/ ^ .N 2 N/

) Œf.L 2 M/ ^ .M 2 N/g ) .L 2 N/g:

Proof. Proceed by induction with N.

Initial step

If N WD 0 D ¿, then for all nonnegative integers L; M 2 N the proposition M 2 ¿

is False, whence so is the conjunction .L 2 M/^.M 2 0/, and hence the implication
Œ.L 2 M/ ^ .M 2 0/ ) .L 2 0/ is True.

Induction hypothesis

Assume that there exists some K 2 N such that the theorem holds for N WD K, so
that Œ.L 2 M/ ^ .M 2 K/ ) .L 2 K/ for all L; M 2 N.

Induction step

If L 2 M and M 2 K C 1 D K [ fKg, then two cases arise: M 2 fKg or M 2 K.

In the first case, M 2 fKg, whence M D K, and the hypothesis L 2 M then yields
L 2 M D K  K [ fKg, so that L 2 K C 1.
In the second case, M 2 K, whence the induction hypothesis yields L 2 K 
K [ fKg D K C 1, and then again L 2 K C 1. t
u
The transitivity of the relation 2 holds on the set N, but it can fail on other sets.
˚ 
4.41 Counterexample. If X WD ¿, Y WD f¿g, ˚ and Z WD f¿g ˚ , then
 X 2 Y and
Y 2 Z, but X … Z. Theorem 4.40 fails for Z D f¿g because f¿g … N.
The following theorem shows that on the natural numbers the relation 2 is neither
reflexive nor symmetric, but, instead, 2 is both irreflexive and asymmetric.
4.42 Theorem. For all I; L; M 2 N, M … M, and .I … L/ _ .L … I/.
Proof. This proof of the first result (M … M) proceeds by induction with M.

Initial step

If M D 0, then M D ¿. From ¿ … ¿ it follows that M … M.

 www.pdfgrip.com

4.4 Orders and Cancellations 207

Induction hypothesis

Assume that there exists some K 2 N such that the theorem holds for M WD K, so
that K … K.

Induction step

This step of the proof proceeds by contraposition. If .K [ fKg/ 2 .K [ fKg/, then

two cases can arise: .K [ fKg/ 2 K or .K [ fKg/ 2 fKg.
First, if .K [ fKg/ 2 K, then the transitivity of the relation 2 (theorem 4.40) and
K 2 .K [ fKg/ give K 2 K, which contradicts the induction hypothesis.
Second, if .K [ fKg/ 2 fKg, then .K [ fKg/ D K, whence K 2 fKg  .K [
fKg/ D K gives K 2 K, which contradicts the induction hypothesis.
The proof of the second result proceeds by contraposition. For all I 2 N and
L 2 N, the conjunction .I 2 L/ ^ .L 2 I/ and the transitivity of 2 (theorem 4.40)
give I 2 I, which contradicts the result just proved. Therefore, :Œ.I 2 L/ ^ .L 2 I/.
t
u
The following theorem shows that each natural number differs from its successor.
4.43 Theorem. For each natural number N 2 N, N 6D N C 1.
Proof. From N 2 fNg  .N [ fNg/ D N C 1 it follows that N 2 .N C 1/. Yet N … N
by theorem 4.42. Therefore N 6D N C 1 by the axiom of extensionality (S1). u
t
The following theorem shows that adding 1 to both sides of a valid inequality
gives a valid inequality, so that if M < N then .M C 1/ < .N C 1/.
4.44 Theorem. For all M; N 2 N, if M 2 N, then .M C 1/ 2 .N C 1/:

` 8M8N .Œ.M 2 N/ ^ .N 2 N/ ) f.M 2 N/ ) Œ.M C 1/ 2 .N C 1/g/ :

Proof. Apply induction with N. For N WD 0 D ¿, and for each M 2 N, the

proposition M 2 ¿ is False; hence the implication .M 2 ¿/ ) Œ.M C 1/ 2 .N C 1/
is True.
Next, assume that there exists K 2 N such that the theorem holds for N WD K, so
that .M 2 K/ ) Œ.M C 1/ 2 .N C 1/ holds for every M 2 N. If M 2 K C 1 D
K [ fKg, then two cases occur.
In the first case, M 2 fKg, and then M D K, whence M C 1 D K C 1 2
.K C 1/ [ fK C 1g D .K C 1/ C 1.
In the second case, M 2 K, whence M C 1 2 K C 1 by induction hypothesis; with
K C 1  .K C 1/ [ fK C 1g D .K C 1/ C 1, it follows that M C 1 2 .K C 1/ C 1. u t
The next theorem shows that 0 D ¿ is the smallest natural number.
4.45 Theorem. For each natural number M 2 N, if M 6D 0, then 0 2 M.
 www.pdfgrip.com

208 4 Mathematical Induction: Definitions and Proofs by Induction

Proof. This proof proceeds by induction with M.

For M WD 0, the proposition M 6D 0 is False, whence the implication .M 6D 0/ )
.0 2 M/ is True.
Next, assume that there exists K 2 N such that the theorem holds for M WD K, so
that .K 6D 0/ ) .0 2 K/. Two cases arise: K D 0 or K 6D 0.
In the first case, if K D 0, then 0 D ¿ 2 .¿ [ f¿g/ D K [ fKg D K C 1; thus,
0 2 K C 1 is True.
In the second case, if K 6D 0, then 0 2 K, by the induction hypothesis, whence
0 2 K  K [ fKg D K C 1, and hence 0 2 K C 1. t
u
The following theorem shows that the relation 2 is connected on the natural
numbers: if M 6D N, then either M 2 N or N 2 M.
4.46 Theorem. For all natural numbers M; N 2 N, exactly one of the following
three formulae is True, while the other two are False:

M 2 N;
M D N;
N 2 M:

Proof. First, observe that at most one of the three formulae may hold. Indeed, M …
M by theorem 4.42, whence .M 2 N/ ^ .M D N/ is False, and .M D N/ ^ .N 2 M/
is False. Similarly, also by theorem 4.42, .M 2 N/ ^ .N 2 M/ is also False. Second,
at least one of the three formulae must hold. This proof uses induction with N.
For N WD 0, and for each M 2 N either M D 0 D N is True, or M 6D 0, and then
0 2 M is True by theorem 4.45. To complete the induction, assume that there exists
K 2 N such that the theorem holds for N WD K, so that .M 2 K/ _ .M D K/ _ .K 2
M/ for each M 2 N, and examine all three formulae.
If M 2 K, then M 2 K  K C 1, whence M 2 K C 1.
If M D K, then M D K 2 fKg  K [ fKg, whence M 2 K C 1.
If K 2 M, then K C 1 2 M C 1 D M [ fMg, (theorem 4.44); two cases occur.
In the first case, K C1 2 fMg, whence K C1 D M. In the second case, K C1 2 M
already. Thus, .M 2 ŒK C 1/ _ .M D ŒK C 1/ _ .ŒK C 1 2 M/ is True. t
u
The foregoing result completes the proof that < is a strict total order (irreflexive,
asymmetric, and transitive) on the natural numbers. The following theorem shows
that the natural numbers are well-ordered by the relation .
4.47 Theorem (Well-Ordering Principle). Each nonempty subset of the natural
numbers has a smallest element.
Proof. By contraposition, this proof establishes that every subset of the natural
numbers without a smallest element (definition 3.192) is empty. To this end, assume
that a subset E  N has no smallest element. Thus every N 2 N is not the smallest
element of E, which means that N … E or that E contains an element M such that
M < N.
 www.pdfgrip.com

4.4 Orders and Cancellations 209

To prove that E D ¿, the proof proceeds by induction with the set S  N

consisting of every N 2 N such that the complement N n E contains every M N:

S WD fN 2 N W .N [ fNg/  .N n E/g:

Initial step

The set S D N n E contains 0; indeed, contraposition shows that if E contained 0,

then 0 would be the smallest element of E, because 0 N for every N 2 N by
theorem 4.45 and hence also for every N 2 E.

Induction step

Assume that K 2 S for some K 2 N; thus, every M K belongs to the complement

of E. Contraposition again confirms that if E contained K C 1, then K C 1 would be
the smallest element of E; consequently, K C 1 … E, whence K C 1 2 S. Therefore,
S D N, whence E D N n S D ¿. t
u
4.48 Definition (minimum). For each nonempty subset E  N, the minimum of
E is the smallest element of E and it is denoted by min.E/.
In contrast to the Well-Ordering Principle, some nonempty subsets of the natural
numbers have no maximum. Yet every nonempty subset of the natural numbers that
has an upper bound (definition 3.198) also has a largest element (definition 3.192).
4.49 Theorem. Each nonempty subset of the natural numbers with an upper bound
in the natural numbers has a largest element.
Proof. If there exists an upper bound M 2 N for a nonempty subset E  N, then
I M for every I 2 E. Let S  N be the set of all the upper bounds for E:

S WD fN 2 N W 8IŒ.I 2 E/ ) .I N/g:

Then M 2 S by hypothesis on E; in particular, S 6D ¿. Consequently, by the Well-

Ordering Principle (theorem 4.47) S has a smallest element K WD min.S/. Because
K 2 S, it follows that I K for every I 2 E.
Moreover, K 2 E, as proved by the following induction. If K D 0, then E D f0g
because K D 0 < I for every I 2 N n f0g; hence K D 0 2 E.
Suppose that there exists L 2 N such that the theorem holds for K WD L, so that
for each E  N if L D min.S/, then L 2 E. If min.S/ D L C 1, then L C 1 2 E,
for otherwise I L for every I 2 E and then L 2 S; thus the theorem holds for
L C 1. t
u
 www.pdfgrip.com

210 4 Mathematical Induction: Definitions and Proofs by Induction

4.50 Definition (maximum). For each nonempty subset E  N with an upper

bound in N, the maximum of E is the largest element of E and it is denoted by
max.E/.
The following abbreviations occasionally prove convenient.
4.51 Definition. For all I 2 N and N 2 N define a set fI; : : : ; Ng by

fI; : : : ; Ng WD fK 2 N W .I K/ ^ .K N/g;

which thus consists of all the natural numbers from I through N . Similarly,

fI; : : :g WD fI; I C 1; I C 2; : : :g WD fK 2 N W I Kg;

which thus consists of all the natural numbers larger than or equal to I .
4.52 Example. If E WD f2; : : : ; 7g, then E D f2; 3; 4; 5; 6; 7g. Also, min.E/ D 2
and max.E/ D 7.
4.53 Example. If E WD f2; 3; : : :g, then min.E/ D 2 but E has no maximum.

4.4.2 Laws of Arithmetic Cancellations

This subsection establishes rules to cancel terms in equations with additions or mul-
tiplications. The material also provides more practice with proofs by mathematical
induction. The first rule shows how to solve equations of the form M C 1 D L C 1.
4.54 Theorem. For all L; M 2 N, if M C 1 D L C 1, then M D L.
Proof. If M C 1 D L C 1, then M [ fMg D L [ fLg, and the sets on both sides have
the same elements. In particular, L 2 .L [ fLg/ and thus L 2 .M [ fMg/, whence
two cases arise: L 2 fMg, or L 2 M.
In the first case, L 2 fMg, and then L D M indeed.
In the second case, L 2 M; then M … fLg, for otherwise M D L and L 2 L would
contradict theorem 4.42. However, M 2 L [ fLg, whence M 2 L, but that would also
contradict theorem 4.42. Therefore, this second case does not occur. t
u
The following theorem allows for the cancellation of an additive term N common
to both sides of an equation of the type M C N D L C N.
4.55 Theorem. For all M; N; L 2 N, if M C N D L C N, then M D L.
Proof. For all natural numbers L; M 2 N, proceed by induction with N. For N WD 0,
if M C 0 D L C 0, then M D M C 0 D L C 0 D L, by hypothesis and by (A0).
Second, assume that there exists K 2 N such that the theorem holds for N WD K,
so that for all natural numbers L; M 2 N, if M C K D L C K then M D L.
 www.pdfgrip.com

4.4 Orders and Cancellations 211

If M C .K C 1/ D L C .K C 1/, then the associativity and the commutativity of

addition lead to .M C 1/ C K D .L C 1/ C K, whence M C 1 D L C 1 by induction
hypothesis, whence finally M D L by theorem 4.54. t
u
The following theorem shows that addition preserves the ordering.
4.56 Theorem. For all L; M; N 2 N, if L < M then L C N < M C N.
Proof. For all L and M, proceed by induction with N.
For N WD 0, and for all L and M, if LC0 D M C0, then L D LC0 D M C0 D M.
To complete the induction, assume that there exists K 2 N such that the theorem
holds for N WD K, so that for all L and M, if L < M then L C K < M C K . Hence,
if L < M then L C 1 < M C 1 by theorem 4.44, whence

L C .K C 1/ D L C .1 C K/
D .L C 1/ C K
< .M C 1/ C K
D M C .1 C K/
D M C .K C 1/:

t
u
4.57 Theorem. For all M; N 2 N, if M 6D 0, then N < N C M.
Proof. Set L WD 0 in theorem 4.56. If M 6D 0 then 0 < M by theorem 4.45, whence
L < M. Hence N D 0 C N < M C N by theorem 4.56. t
u
The following theorem forms the basis for the concept of subtraction of a natural
number from a larger natural number.
4.58 Theorem. For all M; N 2 N, if M < N, then there exists exactly one natural
number L 2 N such that M C L D N.
Proof. This proof establishes the existence and the uniqueness separately.

Existence

First, establish the existence of such a number L. For each M 2 N, proceed by

induction with N.
If N WD 0, then .M < 0/ is False, because .M < 0/ , .M 2 ¿/. Therefore
 
` 8M Œ.M 2 N/ ^ .M < 0/ ) f9LŒ.L 2 N/ ^ .M C L D N/g

is universally valid, and hence the theorem holds for N D 0.

For each M 2 N, assume that there exists K 2 N such that the theorem holds
for N WD K, so that for each M 2 N, if M < K then there exists L 2 N such that
M C L D K. If M < .K C 1/ D K [ fKg, then two cases arise.
 www.pdfgrip.com

212 4 Mathematical Induction: Definitions and Proofs by Induction

In the first case, M 2 fKg, whence M D K, and then M C 1 D K C 1, so that

L D 1. In the second case, M 2 K, that is, M < K, and the induction hypothesis
yields some L 2 N for which MCL D K; hence MC.LC1/ D .MCL/C1 D KC1.

Uniqueness

Second, verify the uniqueness of L, which results from the theorem that if M C L D
N D M C K, then L D K (theorem 4.55). t
u
The following definition specifies the concept of subtraction of a natural number
from a larger natural number.
4.59 Definition (Subtraction). For all L; M; N; 2 N, if M < N, then N M WD L
is the natural number L defined in theorem 4.58 such that M C L D N
4.60 Example. 8 5 D 3 because 5 C 3 D 8 by exercise 4.37.
The following theorem shows that multiplication by a nonzero natural number
preserves the ordering.
4.61 Theorem. For all L; M; N 2 N, if M < N and 0 < L, then L  M < L  N.
Proof. For all nonnegative integers M; N 2 N, use induction with L, the smallest
nonzero value of L being 1. For L WD 1, if M < N, then 1  M D M < N D 1  N.
Assume that there exists K 2 N with 0 < K such that the theorem holds for
L WD K, so that for all natural numbers M; N 2 N, if M < N, then K  M < K  N.
Thus, if M < N, then apply theorem 4.56 twice:

.K C 1/  M D .K  M/ C M < .K  M/ C N < .K  N/ C N D .K C 1/  N:

t
u
The following theorem allows for the cancellation of a nonzero multiplicative
term on both sides of an equation, which forms the basis for the division of a natural
number by a nonzero natural number, and the solution of equations of the type
L  M D L  N.
4.62 Theorem. For all L; M; N 2 N, if 0 < L and L  M D L  N, then M D N.
Proof. Proceed by contraposition.
If M 6D N, then either M < N or N < M.
If M < N, then L  M < L  N and L  M 6D L  N.
Similarly, if N < M, then L  N < L  M and L  M 6D L  N. t
u
4.63 Definition (division). For all K; M; N 2 N, if 0 < L and K  L D N, then
define N=L by N=L WD K, as defined uniquely by theorem 4.62.
 www.pdfgrip.com

4.4 Orders and Cancellations 213

4.64 Example. 8=4 D 2 because 2  4 D 8 by exercise 4.43.

4.65 Example. Division by repeated subtraction was used in the Whirlwind I
computer [82, p. 3-7–3-9].

4.4.3 Exercises on Orders and Cancellations

The following exercises involve the natural numbers (sets) 0; 1; 2; 3; 4; 5; 6; 7; 8; 9

defined in example 3.35, page 128, and reviewed in subsection 4.3.3.
4.51 . Prove that 2 < 5.
4.52 . Prove that 5 < 7.
4.53 . Prove that 2 < 7.
4.54 . Prove that 3 < 5.
4.55 . Prove that 3 < 7.
4.56 . Prove that 4 < 9.
4.57 . Prove that 2 < 8.
4.58 . Prove: there are no natural numbers K > 1, L > 1 with K  L D 2.
4.59 . Prove: there are no natural numbers K > 1, L > 1 with K  L D 3.
4.60 . Prove: there are no natural numbers K > 1, L > 1 with K  L D 5.
4.61 . Prove: there are no natural numbers K > 1, L > 1 with K  L D 7.
4.62 . Determine all the natural numbers K > 1 and L > 1 such that K  L D 6 and
prove that there are no other such natural numbers.
4.63 . Determine all the natural numbers K > 1 and L > 1 such that K  L D 9 and
prove that there are no other such natural numbers.
4.64 . Prove that the relation is a total ordering on N, with

.M N/ , Œ.M D N/ _ .M 2 N/:

4.65 . Prove: for each nonempty S  N there exists I 2 S with I \ S D ¿.

4.66 . Prove that for each L 2 N there exists I 2 L such that I \ L D ¿.
4.67 . Prove that .I … K/ _ .K … L/ _ .L … I/ for all I; K; L 2 N.
4.68 . Prove that 1 < N for each N 2 .N n f0; 1g/.
4.69 . Prove that N … N.
4.70 . Prove that N 6D .N [ fNg/.
 www.pdfgrip.com

214 4 Mathematical Induction: Definitions and Proofs by Induction

4.5 Integers

4.5.1 Negative Integers

Several operations remain undefined with natural numbers. For instance, the
ordering relation does not include elements smaller than the empty set, which
precludes the use of natural numbers to model relations extending in two opposite
directions. Also, the arithmetic of natural numbers does not contain provisions
for the “difference” from a larger natural number to a smaller one. Finally, the
arithmetic of natural numbers does not include any concept of division other than
special cases.
There exist several methods to extend the ordering and the arithmetic of natural
numbers, to allow for elements “smaller” than zero, and for differences of any two
elements. Such methods begin with the specification of a larger set of “integers” Z
[from the German “Zahl(en)” for “number(s)”].
One method of defining a larger set, outlined by Kunen [74, p. 35], is sufficiently
general to produce not only the integers, but also the rational numbers and the real
numbers. Essentially, the method consists in defining the new numbers in terms of
equivalence classes. For the integers, the strategy consists in introducing the concept
of the “difference” between two natural numbers M and N by means of the pair
.M; N/. Then relate .M; N/ to every other pair .P; Q/ with the same “difference”
between P and Q. Because the concept of “difference” has not yet been defined for
all pairs of natural numbers, however, a precise definition uses sums instead.
Two cases arise: either N M, or N > M. In the first case, if N M, then
definition 4.59 specifies their difference J WD M N. If also .M; N/ and .P; Q/
represent the same difference, then P Q D J D M N. An equivalent statement
without subtractions results from adding J to both extremes:

M N D J; M D N C JI
P Q D J; P D Q C J:

In the second case, if N > M with J WD N M and also Q PDJDN M, then

N M D J; N D M C JI
Q P D J; Q D P C J:

In either case, the statement that .M; N/ and .P; Q/ represent the same “difference”
can be reworded without subtractions but with sums instead: there exists J 2 N
with
• either M D N C J and P D Q C J (if M > N and P > Q),
 www.pdfgrip.com

4.5 Integers 215

M N
      
P Q

• or N D M C J and Q D P C J (if M < N and P < Q),

N M
      
Q P

Swapping coordinates, from .M; N/ to .N; M/, will amount geometrically to

reversing an order or a direction, which will amount algebraically to passing from a
positive to a negative integer. Such considerations lead to the following definition.
4.66 Definition. Define a relation l on the set A WD N N for all M; N; P; Q 2 N by

.M; N/ l .P; Q/
 m 
9J .J 2 N/ ^ fŒ.M D N C J/ ^ .P D Q C J/ _ Œ.N D M C J/ ^ .Q D P C J/g W

there is a J 2 N with M D N C J and P D Q C J, or N D M C J and Q D P C J.

4.67 Example. .3; 6/ l .5; 8/ because J WD 3 confirms that 6 D 3 C 3 and 8 D
5 C 3, by exercises 4.35 and 4.37.
The following theorem provides an equivalent formulation of this relation.
4.68 Theorem. For all pairs of natural numbers,

.M; N/ l .P; Q/
m
M C Q D N C P:

Proof. This proof establishes the two implications () and () separately.

Assume first that .M; N/ l .P; Q/. Then two cases can arise.
If there exists I 2 N with M D N C I and P D Q C I, then

M C Q D .N C I/ C Q D N C .I C Q/ D N C .Q C I/ D N C P:

If there exists I 2 N with N D M C I and Q D P C I, then

M C Q D M C .P C I/ D .M C .I C P/ D .M C I/ C P D N C P:

For the converse, assume that M C Q D N C P. Then two cases can arise.
If M N, then there exists I 2 N such that N D M C I. Hence
 www.pdfgrip.com

216 4 Mathematical Induction: Definitions and Proofs by Induction

M C Q D N C P D .M C I/ C P D M C .P C I/

whence (by theorem 4.55) cancelling M yields Q D P C I.

If M > N, then there exists I 2 N such that M D N C I. Hence

N C P D M C Q D .N C I/ C Q D N C .Q C I/

whence (by theorem 4.55) cancelling N yields P D Q C I. t

u
4.69 Example. .5; 4/ l .2; 1/ because 5 C 1 D 4 C 2, by exercises 4.25 and 4.31.
Instead of definition 4.66 the derivations can hence utilize theorem 4.68.
Thus the following theorem shows that the relation l is an equivalence relation
(definition 3.150).
4.70 Theorem. The relation l is an equivalence relation on N N.
Proof. This proof verifies that the relation l is reflexive, symmetric, and transitive.

Reflexivity

For each pair .M; N/ 2 .N N/, the equality M C N D N C M holds by the

commutativity of addition, whence .M; N/ l .M; N/.

Symmetry

If .M; N/ l .P; Q/, then M C Q D N C P, whence Q C M D P C N, which means

that .P; Q/ l .M; N/.

Transitivity

If .K; L/ l .M; N/, then K C N D L C M. If also .M; N/ l .P; Q/, then M C Q D

N C P. Consequently,

K C .M C Q/ D K C .N C P/ D .K C N/ C P D .L C M/ C P D L C .M C P/

whence (theorem 4.55) cancelling M yields K C Q D L C P, and .K; L/ l .P; Q/.

t
u
4.71 Definition (Kunen’s definition of the integers). The set of integers is the set
Z WD .N N/= l of all equivalence classes Œ.M; N/l for the relation l [74, p. 35].
4.72 Example. Setting I WD 3 shows that .0; 3/ l .1; 4/ l .2; 5/ l .3; 6/ l
.4; 7/. Thus the pairs .0; 3/, .1; 4/, .2; 5/, .3; 6/, .4; 7/ are elements of the
equivalence class Œ.0; 3/l . Similarly, .3; 0/ l .4; 1/ l .5; 2/ l .6; 3/ l .7; 4/.
 www.pdfgrip.com

4.5 Integers 217

Thus the pairs .3; 0/, .4; 1/, .5; 2/, .6; 3/, .7; 4/ are elements of the equivalence class
Œ.3; 0/l .
Each pair .I; J/ 2 N N is equivalent to a pair of the type .K; 0/ or .0; K/.
4.73 Theorem. If I > J, then .I; J/ l .I J; 0/.
If I J, then .I; J/ l .0; J I/.
Proof. If I > J, then I D .I J/ C J and J D 0 C J whence .I; J/ l .I J; 0/.
If I J, then I D 0 C I and J D .J I/ C I whence .I; J/ l .0; J I/. t
u
The following diagram shows a few elements from three equivalence classes:
Œ(0, 2)l , Œ.0; 0/l , and Œ(3, 0)l relative to the relation l on A WD N N:

(1, 3) .3; 3/ (6, 3)

(0, 2) .2; 2/ (5, 2)
.1; 1/ (4, 1)
.0; 0/ (3, 0)

4.5.2 Arithmetic with Integers

From the preceding definition of integers in terms of equivalent pairs of natural

numbers — which represent equivalent differences — follows a definition of
arithmetic in terms of such pairs.
4.74 Definition (Kunen’s definition of integer arithmetic). Define an arithmetic
with pairs of natural numbers as follows:

.M; N/ C .P; Q/ WD .M C P; N C Q/;

.M; N/  .P; Q/ WD .ŒM  P C ŒN  Q; ŒM  Q C ŒN  P/:

The following theorem verifies that addition and multiplication of pairs com-
mute.
4.75 Theorem. For all M; N; P; Q 2 N,

.M; N/ C .P; Q/ D .P; Q/ C .M; N/;

.M; N/  .P; Q/ D .P; Q/  .M; N/:

Proof. Apply the commutativity of addition and multiplication with natural

numbers:
 www.pdfgrip.com

218 4 Mathematical Induction: Definitions and Proofs by Induction

.M; N/ C .P; Q/ D .M C P; N C Q/
D .P C M; Q C N/
D .P; Q/ C .M; N/I

.M; N/  .P; Q/ D .ŒM  P C ŒN  Q; ŒM  Q C ŒN  P/

D .ŒP  M C ŒQ  N; ŒQ  M C ŒP  N/
D .ŒP  M C ŒQ  N; ŒP  N C ŒQ  M/
D .P; Q/  .M; N/:

t
u
The next theorem checks that arithmetic with equivalent pairs yields equivalent
results: different pairs representing the same difference yield the same sum or
product.
4.76 Theorem. For all I; J; K; L; M; N; P; Q 2 N, if

.I; J/ l .K; L/;

.M; N/ l .P; Q/;

then

.I; J/ C .M; N/ l .K; L/ C .P; Q/;

.I; J/  .M; N/ l .K; L/  .P; Q/:

Proof. If .I; J/ l .K; L/, then I C L D J C K. If also .M; N/ l .P; Q/, then
M C Q D N C P. Moreover,

.I; J/ C .M; N/ D .I C M; J C N/;

.K; L/ C .P; Q/ D .K C P; L C Q/;

whence

.I C M/ C .L C Q/ D .I C L/ C .M C Q/
D .J C K/ C .N C P/
D .J C N/ C .K C P/;

which means that .I; J/ C .M; N/ l .K; L/ C .P; Q/. For the multiplication,
 www.pdfgrip.com

4.5 Integers 219

.I; J/  .M; N/ D .ŒI  M C ŒJ  N; ŒI  N C ŒJ  M/;

.K; L/  .P; Q/ D .ŒK  P C ŒL  Q; ŒK  Q C ŒL  P/;

so that .I; J/  .M; N/ l .K; L/  .P; Q/ if and only if

.ŒI  M C ŒJ  N/ C .ŒK  Q C ŒL  P/ D .ŒI  N C ŒJ  M/ C .ŒK  P C ŒL  Q/:

Additions and cancellations (theorem 4.55) give

IMCJNCKQCLP D INCJMCKPCLQ
m
I  MCI  QCJ  N CK  QCL  P D I  N CI  QCJ  MCK  PCL  Q
m
I  .M C Q/ C J  N C K  Q C L  P D I  N C J  M C K  P C .I C L/  Q
m
I  .N C P/ C J  N C K  Q C L  P D I  N C J  M C K  P C .J C K/  Q
m
IPCJNCLP D JMCKPCJQ
m
IPCJNCJPCLP D JMCKPCJPCJQ
m
I  P C J  .N C P/ C L  P D J  M C .K C J/  P C J  Q
m
I  P C J  .M C Q/ C L  P D J  M C .I C L/  P C J  Q
m
0 D 0;

which is universally valid, whence .I; J/  .M; N/ l .K; L/  .P; Q/. t

u
Hence the following definition specifies an arithmetic with equivalence classes
based on the arithmetic of representative pairs.
4.77 Definition (Integer addition and multiplication).
     
Œ.M; N/l C Œ.P; Q/l WD Œ.M C P; N C Q/l ;
     
Œ.M; N/l  Œ.P; Q/l WD Œ.M  P C N  Q; M  Q C N  P/l :

4.78 Theorem. For Z, and for all K; L; M; N; P; Q 2 N,

addition commutes,
addition is associative,
Œ.0; 0/l is the additive unit: Œ.M; N/l C Œ.0; 0/l D Œ.M; N/l ,
Œ.N; M/l is the additive inverse: Œ.M; N/l C Œ.N; M/l D Œ.0; 0/l ,
 www.pdfgrip.com

220 4 Mathematical Induction: Definitions and Proofs by Induction

multiplication commutes,
multiplication is associative,
multiplication distributes over addition, and
Œ.1; 0/l is a multiplicative unit: Œ.M; N/l  Œ.1; 0/l D Œ.M; N/l .
Proof. The proof forms the object of exercises. t
u
The following definition specifies a subtraction.
4.79 Definition (Integer subtraction). Define a subtraction of integers by
       
Œ.M; N/l Œ.P; Q/l WD Œ.M; N/l C Œ.Q; P/l ;

which is a binary operation from Z Z to Z. Define an “opposite” by

   
Œ.P; Q/l WD Œ.Q; P/l ;

which is a unary operation (unfortunately also denoted by ) from Z to Z.

4.80 Example. Reducing every pair .I; J/ to either form .I J; 0/ or .0; J I/ can
facilitate calculations:

Œ.2; 3/l C Œ.5; 1/l D Œ.0; 1/l C Œ.4; 0/l

D Œ.0 C 4; 1 C 0/l
D Œ.4; 1/l
D Œ.3; 0/l I

Œ.2; 1/l Œ.7; 3/l D Œ.2; 1/l C Œ.3; 7/l

D Œ.1; 0/l C Œ.0; 4/l
D Œ.1; 4/l
D Œ.0; 3/l
D Œ.3; 0/l I

Œ.3; 1/l  Œ.2; 5/l D Œ.2; 0/l  Œ.0; 3/l

D Œ.2  0 C 0  3; 2  3 C 0  0/l
D Œ.0; 6/l
D Œ.6; 0/l :

4.5.3 Order on the Integers

The order < on N extends to Z by a definition of “positive” and “negative” on Z.

By theorem 4.73, every integer Œ.M; N/l is the equivalence class of a pair .I; 0/
or .0; I/ for some I 2 N. In the first case M N D I 0  0, so that M N
 www.pdfgrip.com

4.5 Integers 221

is a nonnegative difference. In the second case N M D I 0  0, so that the

opposite, N M, is a nonnegative difference, and then M N represents a non-
positive difference.
4.81 Definition. An integer Œ.M; N/l is positive if and only if M > N; an integer
Œ.M; N/l is negative if and only if M < N.
4.82 Definition. The sets of nonzero integers .Z /, negative integers .Z /, positive
integers .ZC /, non-positive integers .Z /, and nonnegative integers .ZC / are

Z WD Z n fŒ.0; 0/l g;

Z WD fŒ.M;
˚ N/l 2 Z W M Ng; 
WD Œ.M; N/l 2 Z W 9If.I 2 N/ ^ Œ.M; N/ l .0; I/g ;

ZC WD fŒ.M;
˚ N/l 2 Z W M  Ng; 
WD Œ.M; N/l 2 Z W 9If.I 2 N/ ^ Œ.M; N/ l .I; 0/g ;

Z WD fŒ.M;
˚ N/l 2 Z W M < Ng; 
WD Œ.M; N/l 2 Z W 9If.I 2 N / ^ Œ.M; N/ l .0; I/g ;

ZC WD fŒ.M;
˚ N/l 2 Z W M > Ng 
WD Œ.M; N/l 2 Z W 9If.I 2 N / ^ Œ.M; N/ l .I; 0/g :

The following relation on pairs will lead to an order on equivalence classes.

4.83 Definition. Define a relation < on N N as follows:

.M; N/ < .0; 0/

m
M<N

and

.M; N/ < .P; Q/

m
.M; N/ C .Q; P/ < .0; 0/
m
M C Q < N C P:

The following theorem verifies that the ordering does not depend on the choice
of the pairs representing the equivalence classes.
 www.pdfgrip.com

222 4 Mathematical Induction: Definitions and Proofs by Induction

4.84 Theorem. For all I; J; K; L; M; N; P; Q 2 N, if

.I; J/ l .K; L/;

.M; N/ l .P; Q/;

then

.I; J/ < .M; N/

m
.K; L/ < .P; Q/:

Proof. If .I; J/ l .K; L/, then I C L D J C K. If also .M; N/ l .P; Q/, then also
M C Q D N C P. Consequently, adding L C P to each side of the inequalities yields

.I; J/ < .M; N/

m
ICN < JCM
m
.I C L/ C .N C P/ < .L C P/ C .J C M/
m
.J C K/ C .M C Q/ < .L C P/ C .J C M/
m
.J C M/ C .K C Q/ < .L C P/ C .J C M/
m
KCQ < LCP
m
.K; L/ < .P; Q/:

t
u
A definition of an order on Z can thus use any pair from an equivalence class.
4.85 Definition. Define an order < on Z by

Œ.M; N/l < Œ.P; Q/l

m
.M; N/ < .P; Q/
m
M C Q < N C P:

Theorem 4.86 shows that < is a strict total order (definition 3.187).
 www.pdfgrip.com

4.5 Integers 223

4.86 Theorem. The relation < is a strict total order on Z Z. In particular, if

Œ.K; L/l < Œ.M; N/l

and

Œ.M; N/l < Œ.P; Q/l ;

then

Œ.K; L/l < Œ.P; Q/l :

Proof. This proof verifies that < is irreflexive, connected, and transitive.

Irreflexivity

(The relation does not relate any element to itself.) For each pair .M; N/ 2 N N,
M C N – N C M, whence .Œ.M; N/l / – .Œ.M; N/l /.

Connectedness

(Each element is related to every different element.) If .Œ.M; N/l / 6D

.Œ.P; Q/l /, then M C Q 6D N C P whence either M C Q < N C P, and then
.Œ.M; N/l / < .Œ.P; Q/l /, or M CQ > N CP, and then .Œ.M; N/l / > .Œ.P; Q/l /.

Transitivity

If .Œ.K; L/l / < .Œ.M; N/l /, then

K C N < L C MI

if also .Œ.M; N/l / < .Œ.P; Q/l /, then also

M C Q < N C P:

Consequently, adding the preceding two inequalities gives

.K C N/ C .M C Q/ < .L C M/ C .N C P/

whence the commutativity and associativity of addition yields

K C Œ.M C N/ C Q < L C Œ.M C N/ C P

 www.pdfgrip.com

224 4 Mathematical Induction: Definitions and Proofs by Induction

whence (by theorem 4.56) cancelling M C N yields

KCQ<LCP

so that Œ.K; L/l < Œ.P; Q/l : t

u
The next theorem shows that multiplication by a positive integer keeps the order.
4.87 Theorem. If

Œ.M; N/l < Œ.P; Q/l ;

Œ.0; 0/l < Œ.I; J/l < Œ.K; L/l ;

then

.Œ.I; J/l  Œ.M; N/l / < .Œ.I; J/l  Œ.P; Q/l / ;

whence, if also .P; Q/ > .0; 0/, then

.Œ.I; J/l  Œ.M; N/l / < .Œ.K; L/l  Œ.P; Q/l / :

Proof. The proof uses the equivalence .I; J/ l .I J; 0/. First, from Œ.M; N/l <
Œ.P; Q/l it follows that M C Q < N C P by definition of < on the pairs. Hence a
multiplication throughout by I J > 0 yields

.I J/  .M C Q/ < .I J/  .N C P/

whence

.I; J/  .M; N/ l .I J; 0/  .M; N/

D .ŒI J  M; ŒI J  N/
< .ŒI J  P; ŒI J  Q/
D .I J; 0/  .P; Q/
l .I; J/  .P; Q/:

Swapping the roles of .I; J/ with .K; L/, and .M; N/ with .P; Q/, then yields

.I; J/  .P; Q/ D .P; Q/  .I; J/ < .P; Q/  .K; L/ D .K; L/  .P; Q/:

Consequently,

.I; J/  .M; N/ < .I; J/  .P; Q/ < .K; L/  .P; Q/:

t
u
 www.pdfgrip.com

4.5 Integers 225

The following theorem shows that the square of every integer is nonnegative.
4.88 Theorem. For every X 2 Z, X  X  0.
Proof. For every X 2 Z there exists K 2 N such that X D Œ.K; 0/l or X D
Œ.0; K/l , by theorem 4.73. However,

.K; 0/  .K; 0/ D .K  K C 0  0; K  0 C 0  K/
D .K  K; 0/
D .0  0 C K  K; 0  K C K  0/
D .0; K/  .0; K/

and in either case X  X D Œ.K  K; 0/l  Œ.0; 0/l because K  K C 0  0 C 0. t

u
4.89 Example. . Œ.1; 0/l /  . Œ.1; 0/l / D .Œ.0; 1/l /  .Œ.0; 1/l / D Œ.1; 0/l :
An alternative method to define all the integers specifies from N a similar but
disjoint set Z for all the “negative” integers, and then defines Z WD Z [ N. For
instance, apply the axioms of the power set, pairing, and separation to define

Z WD fK
˚ 2 P.N/ W 9N Œ.N 2 N/ ^ .N 6D 0/ ^ .K D fNg/g
D fNg W N 2 N
D f: : : ; ff¿; f¿gg; ff¿ggg
D f: : : ; f3g; f2g; f1gg :

Then change the notation to define N WD fNg for each N 2 N . In particular,

Z \ N D ¿. Indeed, ¿ 2 N by theorem 4.45 and ¿ … fNg for each positive
natural number N 2 N n f¿g. Consequently, fNg … N for every fNg 2 Z .
4.90 Definition (Landau’s definition of integer arithmetic). The set Z WD Z [N
is the set of integers. Arithmetic extends from N to Z by setting

. M/ C . N/ WD .M C N/;
. N/ C M WD M C . N/ WD .N M/ if M < N;
. N/ C M WD M C . N/ WD M N if N < M:

Similarly,

. N/  . M/ WD M  N;
. N/  M WD M  . N/ WD .M  N/:

Moreover, the ordering 2 on N extends to an ordering Z by the definition

. N/ < . M/ if and only M < N;

. N/ < K

for all K 2 N and M; N 2 N .

 www.pdfgrip.com

226 4 Mathematical Induction: Definitions and Proofs by Induction

The addition and multiplication thus defined for integers remain associative and
commutative, multiplication distributes over addition, N C0 D N and 1N D N for
each integer N 2 Z. Proofs proceed by cases, depending on the sign of the operands,
and are straightforward but lengthy. See [76, Ch. IV] or the exercises.
4.91 Remark. Common usage abbreviates each nonnegative integer Œ.I; 0/l by I.
Also, any variable can denote an integer, for instance, M D Œ.P; Q/l .

4.5.4 Nonnegative Integral Powers of Integers

The J-th power M J of an integer M 2 Z is the product M   M of J factors M.

Specifically, from the convention M 0 WD 1, induction produces higher powers.
4.92 Definition (Integral powers). For each integer M, define

M 0 WD 1:

Then for each integer M and for each nonnegative integer J define

M JC1 WD .M J /  M:

In M J , the number M is the base while J is the exponent.

4.93 Remark. According to definition 4.92, the Jth power of M amounts to J
multiplications of M, beginning with 1. With an alternative but logically equivalent
notation, for each M 2 N definition 4.92 uses a recursive definition with G in
theorem 4.21 replaced by the multiplication function H .M/ from remark 4.29:

G WD H .M/ W N ! N;
L 7! M  L;

to specify by induction an exponentiation function E.M/ W N ! Z, J 7! M J , with

G.L/ WD H .M/ .L/ D M  L;

.M/
E .0/ WD 1;  
E.M/ .J C 1/ WD G E.M/ .J/ :

4.94 Example. Here are the first four nonnegative powers of the integer 2:

20 D 1I
21 D .20 /  2 D 1  2 D 2I
22 D .21 /  2 D 2  2 D 4I
23 D .22 /  2 D 4  2 D 8:
 www.pdfgrip.com

4.5 Integers 227

The following theorem establishes relations between product of bases, sums of

exponents, and integral powers.
4.95 Theorem. For all M 2 Z and N 2 Z, and for all I 2 N and J 2 N,

.M  N/J D .M J /  .N J /;
N ICJ D .N I /  .N J /:

Proof. This proof proceeds by induction with the exponent J. The first equation,
.M  N/J D .M J /  .N J /, holds for all integers M and N, and for J 2 f0; 1g:

.M  N/0 D 1 D 1  1 D .M 0 /  .N 0 /I

.M  N/1 D M  N D .M 1 /  .N 1 /:

Hence, to prove that .M  N/J D .M J /  .N J / for every J 2 N, let

˚ WD
S 
J 2 N W 8M8NfŒ.M 2 N/ ^ .N 2 N/ ) Œ.M  N/J D .M J /  .N J /g :

Thus, assume that there exists K 2 S, or, equivalently, K 2 N such that the theorem
holds for J WD K; then

.M  N/KC1 D Œ.M  N/K   .M  N/ definition 4.92;

D Œ.M K /  .N K /  .M  N/ induction hypothesis;
D Œ.M K /  M  Œ.N K /  N associativity and commutativity;
D .M KC1 /  .N KC1 / definition 4.92 twice;

whence K C 1 2 S, and hence, S D N, which means that the first equation holds.
The second equation, N ICJ D .N I /  .N J /, holds for each integer N, for each
nonnegative integer I, and for J 2 f0; 1g:

N IC0 D N I D .N I /  1 D .N I /  .N 0 /;
N IC1 D .N I /  N D .N I /  .N 1 /:

Next, let

S WD fJ 2 N W 8I8NŒN ICJ D .N I /  .N J /g

and assume that there exists K 2 S, or, equivalently, K 2 N such that the theorem
holds for J WD K, so that N ICK D .N I /  .N K / for all I 2 N and N 2 N; then
 www.pdfgrip.com

228 4 Mathematical Induction: Definitions and Proofs by Induction

N IC.KC1/ D N .IC1/CK associativity and commutativity of C;

D .N IC1 /  .N K / induction hypothesis;
D Œ.N I /  N  .N K / definition 4.92;
D .N I /  Œ.N K /  N associativity and commutativity of ;
D .N I /  .N KC1 / definition 4.92;

whence K C 1 2 S, and, consequently, S D N, which proves the second equation.

t
u

4.5.5 Exercises on Integers with Induction

4.71 . Calculate Œ.2; 4/l C Œ.6; 3/l .

4.72 . Calculate Œ.5; 3/l C Œ.1; 6/l .
4.73 . Calculate Œ.3; 1/l Œ.5; 2/l .
4.74 . Calculate Œ.7; 3/l  Œ.2; 5/l .
4.75 . Prove that Kunen’s addition commutes.
4.76 . Prove that Landau’s addition commutes.
4.77 . Prove that Kunen’s addition is associative.
4.78 . Prove that Landau’s addition is associative.
4.79 . Prove that Kunen’s multiplication commutes.
4.80 . Prove that Landau’s multiplication commutes.
4.81 . Prove that Kunen’s multiplication is associative.
4.82 . Prove that Landau’s multiplication is associative.
4.83 . Prove that Kunen’s multiplication distributes over addition.
4.84 . Prove that Landau’s multiplication distributes over addition.
4.85 . Prove that subtraction does not commute.
4.86 . Prove that subtraction is not associative.
4.87 . Prove that multiplication distributes over subtraction.
4.88 . Prove that subtraction does not distribute over multiplication.
4.89 . Prove that subtraction does not distribute over addition.
4.90 . Prove that addition does not distribute over subtraction.
 www.pdfgrip.com

4.6 Rational Numbers 229

4.6 Rational Numbers

4.6.1 Definition of Rational Numbers

Some practical situations involve comparisons of proportions. Integer arithmetic

does not allow for proportions, but a method similar to that for passing from natural
numbers to differences also leads from integers to proportions. As a pair of natural
numbers .I; J/ can represent a difference, a pair of integers .P; Q/ can represent a
proportion.
For example, the density of the universe at one location involves the mass or
number of particles P in a volume Q, which can be summarized by the ordered pair
.P; Q/. The density at another location involves the mass or number of particles M
in a volume N, as summarized by the pair .M; N/. A comparison of the densities
.P; Q/ and .M; N/ can proceed through a multiplication by a common factor, for
instance, Q or N.
Specifically, in a volume N times larger than Q, the first density of P particles in
a volume Q becomes P  N particles in a volume Q  N, or .P  N; Q  N/.
Likewise in a volume Q times larger than N, the second density of M particles in
a volume N becomes Q  M particles in a volume Q  N, or .Q  M; Q  N/.
Both densities .P  N; Q  N/ and .Q  M; Q  N/ refer to the same volume Q  N;
thus, they are identical if and only if their masses equal each other: P  N D Q  M.
The foregoing reasoning leads to a relation between pairs of integers.
4.96 Definition. On the set Z Z of all pairs of integers .P; Q/ with Q 6D 0, define
a relation  by

.P; Q/  .M; N/
m
PN DQM

The relation  on Z Z in definition 4.96 differs from the relation l on N N

in definition 4.66. Nevertheless, it is also an equivalence relation (definition 3.150).
4.97 Theorem. The relation  in definition 4.96 is an equivalence relation on
Z Z .
Proof. This proof checks algebraically that  is reflexive, symmetric, and transitive.

Reflexivity

For each pair, .I; J/  .I; J/ because I  J D J  I.

 www.pdfgrip.com

230 4 Mathematical Induction: Definitions and Proofs by Induction

Symmetry

If .I; J/  .K; L/, then I  L D J  K by definition of  whence K  J D L  I by

commutativity, so that .K; L/  .I; J/.

Transitivity

If

.I; J/  .K; L/;

.K; L/  .M; N/;

then

I  L D J  K;
K  N D L  M;

whence multiplying the left-hand sides and the right-hand sides together gives

.I  L/  .K  N/ D .J  K/  .L  M/;
.I  N/  .L  K/ D .J  M/  .K  L/;

and hence cancelling the nonzero factor L  K D K  L yields

IN DJM

which means that .I; J/  .M; N/. t

u
4.98 Definition. The set of rational numbers, denoted by Q (for “quotients”), is the
set of all equivalence classes for the relation . A rational number is an element
of Q.
The equivalence class of a pair .I; J/ is called the ratio of I to J, and it is denoted
by Œ.I; J/, or also by I=J, or also by JI . A fraction is a pair .I; J/ 2 .Z Z /, where
I is the numerator and J is the denominator.
The following theorem provides a means to select a specific fraction from a
rational number, or, equivalently, a specific pair .M; N/ from an equivalence class
P=Q.
4.99 Theorem. For each P=Q 2 Q, there exists M=N  P=Q such that

M D minfI 2 N W 9JŒ.J 2 N/ ^ .I=J  P=Q/g:

Proof. This proof applies the Well-Ordering Principle to the set of all nonnegative
numerators of a rational number. To this end, for each P=Q 2 Q define the set
 www.pdfgrip.com

4.6 Rational Numbers 231

E WD fI 2 N W 9JŒ.J 2 N/ ^ .I=J  P=Q/g:

Then E 6D ¿, because if P  0 then P=Q  P=Q, whence P 2 E, whereas if

P < 0 then . P/=. Q/  P=Q, whence . P/ 2 E. The Well-Ordering Principle
(theorem 4.47) then guarantees the existence of a first element in E. t
u

4.6.2 Arithmetic with Rational Numbers

The comparison of two rational numbers P=Q and M=N can proceed through
equivalent fractions with a common denominator, for instance, .P  N/=.Q  N/ and
.Q  M/=.Q  N/. Common denominators also lead to an arithmetic with fractions
(ordered pairs) and then with rational numbers (equivalence classes).
4.100 Definition. For all pairs .I; J/ and .K; L/ in Z Z , define functions C and
 on .Z Z / .Z Z / by their counterparts C and  already defined on Z Z :

.I; J/ C .K; L/ WD .ŒI  L C ŒJ  K; J  L/ ;

.I; J/  .K; L/ WD .I  K; J  L/ :

The symbols C and  on the left-hand sides are the functions being defined, whereas
the symbols C and  on the right-hand sides are the addition and multiplication of
integers. Yet common usage employs C and  for both.
The following theorem shows that equivalent fractions lead to equivalent results.
4.101 Theorem. If

.I; J/  .K; L/;

.M; N/  .P; Q/;

then

.I; J/ C .M; N/  .K; L/ C .P; Q/;

.I; J/  .M; N/  .K; L/  .P; Q/:

Proof. This proof proceeds through algebraic verifications. By hypotheses, I  L D

J  K and M  Q D N  P, whence

.I; J/ C .M; N/ D .I  N C J  M; J  N/;

.K; L/ C .P; Q/ D .K  Q C L  P; L  Q/;

 www.pdfgrip.com

232 4 Mathematical Induction: Definitions and Proofs by Induction

where

.I  N C J  M/  .L  Q/ .J  N/  .K  Q C L  P/

D ŒI  N  L  Q C J  M  L  Q ŒJ  N  K  Q C J  N  L  P

D Œ.I  L/  .N  Q/ C .J  L/  .M  Q/
Œ.J  K/  .N  Q/ C .J  L/  .N  P/

D Œ.I  L/  .N  Q/ C .J  L/  .M  Q/
Œ.I  L/  .N  Q/ C .J  L/  .M  Q/

D 0:

Thus, .I; J/ C .M; N/  .K; L/ C .P; Q/. Similarly,

.I; J/  .M; N/ D .I  M; J  N/;

.K; L/  .P; Q/ D .K  P; L  Q/;

where, by hypotheses,

.I  M/  .L  Q/ .J  N/  .K  P/
D .I  L/  .M  Q/ .J  K/  .N  P/
D .I  L/  .M  Q/ .I  L/  .M  Q/
D 0:

Thus, .I; J/  .M; N/  .K; L/  .P; Q/. t

u
4.102 Definition. Define the addition and the multiplication of rational numbers by

I K ILCJK
C WD ;
J L JL
I K IK
 WD :
J L JL
The following theorems establish algebraic characteristics of rational arithmetic.
4.103 Theorem. The addition of rational numbers commutes.
Proof. For all I=J and K=L in Q,

I K ILCJK LICKJ K I
C D D D C :
J L JL LJ L J
t
u
 www.pdfgrip.com

4.6 Rational Numbers 233

4.104 Theorem. The multiplication of rational numbers commutes.

Proof. For all I=J and K=L in Q,

I K IK KI K I
 D D D  :
J L JL LJ L J
t
u
4.105 Theorem. The addition of rational numbers is associative.
Proof. For all I=J, K=L, and M=N in Q,

I K M ILCJK M
C C D C
J L N JL N

Œ.I  L C J  K/  N C Œ.J  L/  M
D
.J  L/  N

ŒI  .L  N/ C J  .K  N/ C ŒJ  .L  M/
D
J  .L  N/

ŒI  .L  N/ C ŒJ  .K  N C L  M/
D
J  .L  N/

I KNCLM
D C
J LN

I K M
D C C :
J L N

t
u
4.106 Theorem. The multiplication of rational numbers is associative.
Proof. For all I=J, K=L, and M=N in Q,

I K M IK M
  D 
J L N JL N

.I  K/  M
D
.J  L/  N

I  .K  M/
D
J  .L  N/
 www.pdfgrip.com

234 4 Mathematical Induction: Definitions and Proofs by Induction

I KM
D 
J LN

I K M
D   :
J L N

t
u
4.107 Theorem. For each N 2 N and each I=J 2 Q, multiplications of the
numerator and denominator by a nonzero common factor yields the same rational
number: I=J D .I  N/=.J  N/.
Proof. Verify the criterion for equivalent fractions: I  .J  N/ D J  .I  N/. t
u
4.108 Theorem. The multiplication of rational numbers distributes over addition.
Proof. For all I=J, K=L, and M=N in Q,

I K M ILCJK M
C  D 
J L N JL N

.I  L C J  K/  M
D
.J  L/  N

.I  L/  M C .J  K/  M/
D
J  .L  N/

.I  M/  L C J  .K  M/
D
J  .L  N/

.I  M/  .L  N/ C .J  N/  .K  M/
D
.J  N/  .L  N/

IM KM
D C
JN LN

I M K M
D  C  :
J N L N

t
u
The following theorem shows that adding 0=1 does not produce any change.
4.109 Theorem. For each K=L 2 Q, .K=L/ C .0=1/ D .K=L/.
Proof. .K=L/ C .0=1/ D .ŒK  1 C L  0=L  1/ D .K=L/. t
u
Thus, the rational number 0=1 plays the same role as the integer 0 in additions.
The following theorem shows that each rational number has an additive inverse.
 www.pdfgrip.com

4.6 Rational Numbers 235

4.110 Theorem. Each I=J 2 Q has an additive inverse: .I=J/ C .ΠI=J/ D .0=1/.
Proof.

I I I  J C J  . I/
C D
J J JJ

J  I C J  . I/
D
JJ

J  ŒI C . I/
D
JJ
J0
D
JJ
0
D
JJ

0  .J  J/
D
1  .J  J/

0
D :
1
t
u
The following theorem shows that multiplying by 1=1 changes nothing.
4.111 Theorem. For each K=L 2 Q, .K=L/  .1=1/ D .K=L/.
Proof. .K=L/  .1=1/ D .ŒK  1=ŒL  1/ D .K=L/. t
u
Thus, the rational number 1=1 plays the same role as the integer 1 in multiplica-
tions. Similarly, each nonzero rational number has a multiplicative inverse.
4.112 Theorem. Each I=J 2 Q such that I 6D 0 has a multiplicative inverse: .I=J/
.J=I/ D .1=1/.
Proof.

I J IJ IJ 1  .I  J/ 1
 D D D D :
J I JI IJ 1  .I  J/ 1

t
u
Rational arithmetic thus satisfies the algebraic properties in table 4.115.
4.113 Definition (Field). A field (of numbers) is a set F with at least two different
elements 0 and 1, so that 0 6D 1, and binary operations C and , satisfying the
algebraic properties in table 4.115.
 www.pdfgrip.com

236 4 Mathematical Induction: Definitions and Proofs by Induction

Table 4.115 These properties hold for all I=J; K=L; P=Q 2 Q.
(1) Associativity of C Œ.I=J/ C .K=L/ C .P=Q/ D .I=J/ C Œ.K=L/ C .P=Q/
(2) Commutativity of C .I=J/ C .K=L/ D .K=L/ C .I=J/
(3) Additive identity .K=L/ C .0=1/ D .K=L/ D .0=1/ C .K=L/
(4) Additive inverse .K=L/ C .ŒK=L/ D .0=1/
(5) Associativity of  Œ.I=J/.K=L/.P=Q/ D .I=J/Œ.K=L/.P=Q/
(6) Commutativity of  .I=J/.K=L/ D .K=L/.I=J/
(7) Multiplicative identity .K=L/.1=1/ D .K=L/ D .1=1/.K=L/
(8) Multiplicative inverse If .K=L/ 6D 0,
then .K=L/.L=K/ D .1=1/
(9) Distributivity .I=J/Œ.K=L/ C .P=Q/ D Œ.I=J/.K=L/ C Œ.I=J/.P=Q/

Technically, the pair .C; / suffices to identify the set F and the elements 0 and
1, but for emphasis a field can be defined as the quintuple .F; C; 0; ; 1/.
4.114 Example. The quintuple .Q; C; 0; 1/ is a field (of numbers).
The next theorem forms the basis for a concept of division of rational numbers.
4.116 Theorem. For each K=L 2 Q, and for each I=J 2 Q such that I 6D 0,
Œ.K=L/  .J=I/  .I=J/ D .K=L/.
Proof.

K J I KJ I .K  J/  I K  .J  I/ K  .I  J/ K
  D  D D D D :
L I J LI J .L  I/  J L  .I  J/ L  .I  J/ L

t
u
4.117 Definition. For each K=L 2 Q, and for each I=J 2 Q such that I 6D 0, define

K I K J
 WD 
L J L I

4.118 Definition. For each K=L 2 Q, and for each J 2 N, define

0
K 1
WD ;
L 1

JC1 J
K K K
WD  :
L L L

Moreover, if K=L 6D 0=1, then for each J 2 Z, define

J1 J
K K L
WD  :
L L K
 www.pdfgrip.com

4.6 Rational Numbers 237

4.6.3 Notation for Sums and Products

The notation introduced here proves convenient to define and investigate sums and
products of finite sequences of numbers.
4.119 Definition. A finite sequence of rational numbers is a function S W N ! Q
defined on some N 2 N. The value S.K/ is also denoted by SK ; then the function S
is also denoted by .SK /.
4.120 Example. The function S W 9 ! Q defined by SK WD . 2=3/K is a finite
sequence of numbers:

S0 D . 2=3/0 D 1;
S1 D . 2=3/1 D 2=3;

S2 D . 2=3/2 D 4=9;

S3 D . 2=3/3 D 8=27;

S4 D . 2=3/4 D 16=81;

S5 D . 2=3/5 D 32=243;

S3 D . 2=3/6 D 64=729;

S4 D . 2=3/7 D 128=2187;

S5 D . 2=3/8 D 256=6561:

The next definition gives a notation for the product of a finite sequence of
numbers.
4.121 Definition (Product notation). For each finite sequence of numbers
S W N ! Q, define the “empty product” to be 1:
Y
SK WD 1:
K<0

Then define the product of the first value to be the first value:

0
Y
SK WD S0 :
KD0

Hence for each L 2 N, such that 0 < L < N, define the product of the first L values
of the sequence S “inductively” [77, p. 5] or “recursively” [49, p. 133] by
!
Y
L Y
L1
SK WD SK  SL :
KD0 KD0
 www.pdfgrip.com

238 4 Mathematical Induction: Definitions and Proofs by Induction

4.122 Example. Consider the finite sequence S W 9 ! Q defined by SK WD . 2=3/K :

Q
K<0 SK D 1;
Q0
KD0 SK D S0
D 1;

Q1 Q 
0
KD0 SK D KD0 SK  S1
D .1/  2=3;

Q2 Q 
1
KD0 SK D KD0 SK  S2
D .1  2=3/  4=9;

Q3 Q 
2
KD0 SK D KD0 SK  S3
D .1  2=3  4=9/  8=27;

Q4 Q 
3
KD0 SK D KD0 SK  S4
D .1  2=3  4=9  8=27/  16=81;
::
:  
Q8 Q7
S
KD0 K D KD0 SK  S8
256
81  243  729  2187 / 
D .1  23  49  278  16 32 64 128
6561
:

The next definition gives a notation for the sum of a finite sequence of numbers.
4.123 Definition (Sum notation). For each finite sequence of numbers S W N ! Q,
define the “empty sum” to be 0:
X
SK WD 0:
K<0

Then define the sum of the first value to be the first value:
0
X
SK WD S0 :
KD0

Hence for each L 2 N, such that 0 < L < N, define the sum of the first L values of
the sequence S inductively by
!
X
L X
L1
SK WD SK C SL :
KD0 KD0
 www.pdfgrip.com

4.6 Rational Numbers 239

4.124 Example. Consider the finite sequence S W 9 ! Q defined by SK WD . 2=3/K :

P
K<0 SK D 0;
P0
KD0 SK D S0
D 1;

P1 P 
0
KD0 SK D KD0 SK C S1
D .1/ C 2=3;

P2 P 
1
KD0 SK D KD0 SK C S2
D .1 C 2=3/ C 4=9;

P3 P 
2
KD0 SK D KD0 SK C S3
D .1 C 2=3 C 4=9/ C 8=27;

P4 P 
3
KD0 SK D KD0 SK C S4
D .1 C 2=3 C 4=9 C 8=27/ C 16=81;
::
:  
P8 P7
S
KD0 K D KD0 SK C S8
256
81 C 243 C 729 C 2187 / C
D .1 C 23 C 49 C 278 C 16 32 64 128
6561
:

The pattern in the foregoing example, called a geometric series, is amenable to

an alternative formula, which expresses the entire sum as one ratio.
4.125 Theorem (geometric series). For every N 2 N and every X 2 Q n f1g,

X
N1
1 XN
XK D :
KD0
1 X

Proof. This proof uses induction with N. For N WD 1, and for every X 2 Q n f1g,

11
X 0
X 1 X1
XK D XK D X0 D 1 D :
KD0 KD0
1 X

Assume that there exists I 2 N such that the theorem holds for N WD I and for
every X 2 Q n f1g, so that
 www.pdfgrip.com

240 4 Mathematical Induction: Definitions and Proofs by Induction

X
I1
1 XI
XK D :
KD0
1 X

Then
.IC1/1
!
X X
I1
X DK
X K
C XI
KD0 KD0

1 XI
D C XI
1 X

1 XI .1 X/  X I
D C
1 X 1 X

.1 X I / C .X I X IC1 /
D
1 X

1 X IC1
D :
1 X
t
u
An alternative proof of the same formula proceeds along the following outline:

PN1
KD0 XK D1C X C C X N2 C X N1
PN1
X KD0 XK D X C X2 C C X N1 C X N
PN1
.1 X/  KD0 XK D 1 C 0 C C 0 XN

whence dividing both sides by .1 X/ yields

X
N1
1 XN
XK D :
KD0
1 X

Yet such a proof also requires induction to rearrange the terms in the subtraction.
 www.pdfgrip.com

4.6 Rational Numbers 241

4.126 Example. Consider the finite sequence S W 9 ! Q defined by SK WD . 2=3/K :

P8 256
KD0 81 C 243 C 729 C 2187 / C
SK D .1 C 23 C 49 C 278 C 16 32 64 128
6561

1. 2=3/9
D
1 2=3

1 512=19,683
D 1=3

19,171=19683
D 1=3

D 3=1  19171=19683

D 19171=6561:

4.6.4 Order on the Rational Numbers

The determination of whether two rational numbers P=Q and M=N coincide can
utilize equivalent fractions with a common denominator, for instance, .P  N/=.Q 
N/ and .Q  M/=.Q  N/, and then with the comparison of the numerators P  N and
Q  M. The same comparison leads to a concept of order on the rational numbers.
4.127 Definition. Define a relation < on Q as follows. First, 0 < .P=Q/ if and only
if 0 < P  Q, so that either P and Q are both positive, or P and Q are both negative:

Œ0 < .P=Q/ , Œ0 < .P  Q/:

Second, .I=J/ < .P=Q/ if and only if 0 < Œ.P=Q/ .I=J/:

Œ.I=J/ < .P=Q/ , f0 < Œ.P=Q/ .I=J/g:

The following theorem shows that the square of any nonzero rational number,
and the sum and product of positive rational numbers, are positive rational numbers.
4.128 Theorem. If .M=N/ > 0 and .I=J/ > 0, then .M=N/ C .I=J/ > 0 and
.M=N/  .I=J/ > 0. Moreover, .K=L/  .K=L/ > 0 for every .K=L/ 6D .0=1/.
Proof. For the square, let P=Q WD .K=L/2 D .K 2 /=.L2 /. Then P  Q D .K 2 / 
.L2 / D .K  L/2 > 0 (theorem 4.88). Thus .K=L/2 D P=Q > 0 (definition 4.127).
For the product, let P=Q WD .M=N/.I=J/ D .M I/=.N J/. By the hypotheses,
M  N > 0 and I  J > 0. Hence P  Q D .M  I/  .N  J/ D .M  N/  .I  J/ > 0,
whence .M=N/  .I=J/ D P=Q > 0 (definition 4.127). For the sum, let
 www.pdfgrip.com

242 4 Mathematical Induction: Definitions and Proofs by Induction

P M I .M  J/ C .N  I/
WD C D :
Q N J NJ

Then

P  Q D Œ.M  J/ C .N  I/  .N  J/ D .M  N/  .J  J/ C .N  N/  .I  J/ > 0:

Indeed, J 2 > 0 and N 2 > 0 (theorem 4.88), .M N/.J 2 / > 0 and .N 2 /.I J/ > 0
by hypothesis, whence P  Q D .M  N/  .J 2 / C .N 2 /  .I  J/ > 0. t
u
The next theorem shows that < is a strict total order (definition 3.187) on Q.
4.129 Theorem. The relation < is a strict total order on the rational numbers.
Proof. This proof verifies that < is connected, irreflexive, and transitive.

Irreflexivity

.P=Q/ – .P=Q/ because 0 – 0 D .P=Q/ .P=Q/.

Connectedness

If .M=N/ 6D .P=Q/, then P  N 6D Q  M whence .P  N Q  M/ 6D 0, whence

.QN/.PN QM/ 6D 0, and then either .QN/.PN QM/ < 0, in which
case .P=Q/ .M=N/ < .0=1/, so that .P=Q/ < .M=N/, or .QN/.PN QM/ >
0, in which case .P=Q/ .M=N/ > .0=1/, so that .P=Q/ > .M=N/.

Transitivity

If .I=J/ < .K=L/ and .K=L/ < .M=N/, then .K=L/ .I=J/ > 0 and .M=N/
.K=L/ > 0, whence .M=N/ .I=J/ D Œ.M=N/ .K=L/ C Œ.K=L/ .I=J/ > 0
(theorem 4.128) and then .M=N/ > .I=J/ (definition 4.127). t
u
4.130 Definition. The sets of nonzero rationals .Q /, negative rationals .Q /,
positive rationals .QC /, non-positive rationals .Q /, and nonnegative rationals
.QC / are

Q WD Q n f0=1g;
Q WD fP=Q 2 Q W P=Q 0=1g;
QC WD fP=Q 2 Q W P=Q  0=1g;
Q WD fP=Q 2 Q W P=Q < 0=1g;
QC WD fP=Q 2 Q W P=Q > 0=1g:
 www.pdfgrip.com

4.6 Rational Numbers 243

4.131 Definition. Define the absolute value jP=Qj of a rational number P=Q by

P=Q if P=Q  0;
jP=Qj WD
.P=Q/ if P=Q < 0:

4.132 Theorem (Triangle Inequality). For all K=L 2 Q and M=N 2 Q,

ˇ ˇ ˇ ˇ ˇ ˇ
ˇK ˇ ˇK ˇ ˇM ˇ
ˇ C Mˇ ˇ ˇCˇ ˇ
ˇL Nˇ ˇLˇ ˇN ˇ

with equality if and only if .K=L/.M=N/  0. Also,

ˇˇ ˇ ˇ ˇˇ ˇ ˇ
ˇ ˇK ˇ ˇM ˇ ˇ ˇK M ˇˇ
ˇˇ ˇ ˇ ˇˇ ˇ
ˇˇLˇ ˇN ˇˇ ˇL Nˇ

with equality if and only if .K=L/.M=N/  0.

Proof. Apply the definition of the absolute value to four cases. t
u
4.133 Theorem (Archimedean Property of the Rationals). For each rational
P=Q 2 Q there exists a natural number N 2 N such that P=Q < N=1.
Proof. If P=Q 0, let N WD 1. If P=Q > 0, let N WD jPj C 1; then

N P N jPj fŒ.jPj C 1/  jQj .1  jPj/g

D D > 0;
1 Q 1 jQj 1  jQj

because Œ.jPj C 1/  jQj .1  jPj/ D jPj  .jQj 1/ C jQj  jQj > 0. t

u

4.6.5 Exercises on Rational Numbers

4.91 . Calculate .2=3/ C .7=5/.

4.92 . Calculate .5=2/ C .1=7/.
4.93 . Calculate .7=3/ .2=5/.
4.94 . Calculate .1=2/ .1=3/.
4.95 . Calculate .2=3/  .7=5/.
4.96 . Calculate .5=2/  .1=7/.
4.97 . Calculate .2=3/  .7=5/.
4.98 . Calculate .5=2/  .1=7/.
4.99 . Prove that on Q the division does not commute.
 www.pdfgrip.com

244 4 Mathematical Induction: Definitions and Proofs by Induction

4.100 . Prove that on Q division is not associative.

4.101 . Prove that on Q division does not distribute over addition.
4.102 . Prove that on Q division does not distribute over multiplication.
4.103 . Prove that on Q addition does not distribute over division.
4.104 . Prove that on Q multiplication does not distribute over division.
4.105 . Prove that if 0 < .I=J/ and 0 < .P=Q/, then 0 < Œ.I=J/ C .P=Q/.
4.106 . Prove that if 0 < .I=J/ and 0 < .P=Q/, then 0 < Œ.I=J/  .P=Q/.
4.107 . Prove that if Q > 0 and R > 0, then P=.Q=R/ D .P  R/=Q.
4.108 . Prove that for each P=Q 2 Q there exists a smallest N 2 N such that there
exists M 2 Z with P=Q D M=N.
4.109 . Prove that for each K=L 2 QC there exists a smallest N 2 N with K=L
N=1.
4.110 . Find rational numbers K=L and M=N with .K=L/2 C .M=N/2 D 1.

4.7 Finite Cardinality

4.7.1 Equal Cardinalities

The adjective “cardinal” means “principal” or “of greatest importance” . In the

context of sets, the cardinal feature of sets is their size. One way to define the “size”
of a set consists in establishing a correspondence with another set of known size,
for instance, a natural number, as in figure 4.1. Such a natural number is then the
“number” of elements in the set, and the correspondence amounts to an operation
of counting. Thus the natural numbers constitute the “standard” sizes with which to
“count” sets. More generally, two sets have the same cardinality if and only if there
exists a bijection between those two sets.
4.134 Definition. For all sets A and B, the sets A and B have the same cardinality
if and only if there exists a bijection F W A ! B, a situation denoted by

A Ð B;
ANN D B;
NN
jAj D jBj;
#.A/ D #.B/;
card .A/ D card .B/:

Definition 4.134 does not yet define the concept of cardinality; it only defines
the concept of same cardinality. Because such a definition leaves the notation #.A/
 www.pdfgrip.com

4.7 Finite Cardinality 245

˚ 
Fig. 4.1 Same cardinality: Earth and Moon, or Pluto and Charon, or ¿; f¿g .

yet undefined, this exposition adopts the notation A Ð B, which merely means that
there exists a bijection from A to B.
4.135 Example. All empty sets have the same cardinality. Indeed, by the axiom of
extensionality there exists only one empty set, namely ¿, and the empty function
¿ W ¿ ! ¿ is a bijection, whence ¿ Ð ¿.
4.136 Example. All singletons have the same cardinality. Indeed, for all sets X and
Y and all singletons fXg and fYg, the function F W fXg ! fYg defined by F WD
f.X; Y/g is a bijection. Thus fXg Ð fYg.
4.137 Example. The sets A WD f4; 9g and B WD f2; 3g have the same cardinality,
thanks to the bijection F W A ! B with F WD f.4; 2/; .9; 3/g. The other bijection,
G WD f.4; 3/; .9; 2/g, could also serve to prove that A and B have the same
cardinality.
The following theorem forms the basis for the relation between the addition of
natural numbers and the union of disjoint sets.
4.138 Theorem. For all sets A, B, C, and D, if

A Ð C;
B Ð D;
A\B D ¿;
C\D D ¿;

then

P Ð .C[D/:
.A[B/ P
 www.pdfgrip.com

246 4 Mathematical Induction: Definitions and Proofs by Induction

Proof. The hypotheses A Ð C and B Ð D mean that there exist bijections

F W A ! C;
G W B ! D:

Such bijections lead to a bijection

P ! .C[D/
H W .A[B/ P

defined by

P
H WD .F [G/ P
 .A[B/ P
.C[D/

so that
8
< Y D F.X/ if X 2 A;
Œ.X; Y/ 2 H ,
:
Y D G.X/ if X 2 B:

The relation H just defined is a function, because for each X 2 .A[B/ P the relation
H contains only one pair .X; Y/. Indeed, thanks to A \ B D ¿, either X 2 A and
then Y D F.X/, or X 2 B and then Y D G.X/, but not both (definition 3.94).
To verify the injectivity of H, assume that W 2 .A[B/ P and X 2 .A[B/ P have the
same image H.W/ D Y D H.X/ in C [ D. Because of C \ D D ¿, either both
images lie in C or both lie in D. In the first case, if both images lie in C, then W 2 A
and X 2 A, and then

F.W/ D H.W/ D H.X/ D F.X/;

whence W D X by injectivity of F. In the second case, if both images lie in D, then

W 2 B and X 2 B, and then

G.W/ D H.W/ D H.X/ D G.X/;

whence W D X by injectivity of G. To verify the surjectivity of H, assume that

P
Z 2 .C[D/. Then either Z 2 C or Z 2 D. In the first case, if Z 2 C, then the
surjectivity of F guarantees the existence of an element W 2 A such that

Z D F.W/ D H.W/:

In the second case, if Z 2 D, then the surjectivity of G guarantees the existence of

an element X 2 B such that

Z D G.X/ D H.X/:

P ! .C[D/
Therefore, H W .A[B/ P is a bijection. t
u
 www.pdfgrip.com

4.7 Finite Cardinality 247

The following theorem forms the basis for the relation between multiplication of
natural numbers and Cartesian products of sets.
4.139 Theorem. For all sets A, B, C, and D, if

A Ð C;
B Ð D;

then

.A B/ Ð .C D/:

Proof. The hypotheses A Ð C and B Ð D mean that there exist bijections

F W A ! C;
G W B ! D:

Such bijections lead to a bijection H defined by

H W .A B/ ! .C D/; 
.W; X/ 7! F.W/; G.X/ :

The relation H is a function, because F and G are functions, so that for each W 2 A
and each X 2 B there exists at most one Y 2 C and at most one Z 2 D with
.W;
 Y/ 2 F and .X; Z/ 2 G, so that there exists at most one .Y; Z/ 2 C D with
.W; X/; .Y; Z/ 2 H. To verify the injectivity of H, assume that .W; X/ 2 .A B/
and .U; V/ 2 .A B/ have the same image H.W; X/ D H.U; V/ in C D:

H.W; X/ D H.U; V/ hypothesis;

m definition of H;
.F.W/; G.X// D .F.U/; G.V//
m equality of pairs;
ŒF.W/ D F.U/ ^ ŒG.X/ D G.V/
m injectivity of F and G;
ŒW D U ^ ŒX D V
m equality of pairs:
.W; X/ D .U; V/

To verify the surjectivity of H, assume that .R; S/ 2 .C D/. Then the surjectivity
of F guarantees the existence of an element W 2 A such that

R D F.W/;
 www.pdfgrip.com

248 4 Mathematical Induction: Definitions and Proofs by Induction

and the surjectivity of G guarantees the existence of an element X 2 B such that

S D G.X/:

Consequently,

.R; S/ D .F.W/; G.X// D H.W; X/:

Therefore, H W .A B/ ! .C D/ is a bijection. t
u

4.7.2 Finite Sets

The following definition establishes the concept of cardinality for finite sets.
4.140 Definition. For each set S, the set S is finite if and only if there exists N 2 N
and a bijection F W N ! S. Such a natural number N is then called the number of
elements in S, or the cardinality of S, which is denoted by #.S/, jSj, or S. NN

4.141 Example. For each natural number N 2 N the set N is finite and has N
elements, because the identity function IN W N ! N; K 7! K is a bijection.
4.142 Remark. Because every bijection has an inverse function, a set S is finite if
and only if there exist a natural number N 2 N and a bijection G W S ! N, for
instance, the inverse function G WD F ı1 for any bijection F W N ! S.
The following theorem shows that the insertion of a new element into a set
corresponds to the arithmetic addition of 1 to its cardinality.
P
4.143 Theorem. The equality #.A[fZg/ D Œ#.A/ C 1 holds for each finite set A,
and for each set Z … A.
Proof. For each finite set A there is a natural number N and a bijection F W N ! A.
Moreover, for each set Z there exists a bijection of singletons G W fNg ! fZg.
Consequently, because A \ fZg D ¿ by hypothesis, and because N \ fNg D ¿
P
by theorem 4.42, it follows that theorem 4.138 gives a bijection N C 1 ! A[fZg:

P
H D .F [G/ P
W N C 1 D .N [fNg/ P
! .A[fZg/:

t
u
The following theorem shows that the cardinality of the union of two disjoint
finite sets equals the arithmetic sum of their two cardinalities.
P D Œ#.A/CŒ#.B/ holds for all disjoint finite
4.144 Theorem. The equality #.A[B/
sets A and B.
 www.pdfgrip.com

4.7 Finite Cardinality 249

Proof. This proof proceeds by induction with the cardinality of the second set.
If #.B/ D 0, then B D ¿ by definition, whence for each finite set A,

P D #.A[¿/
#.A[B/ P D #.A/ D Œ#.A/ C 0 D Œ#.A/ C Œ#.B/:

Hence, assume that there exists a natural number N 2 N for which the theorem
P D Œ#.A/ C Œ#.B/ holds for all disjoint finite sets
holds, so that the equality #.A[B/
A and B with #.B/ D N. For each set C with N C 1 elements, there exists a bijection
F W N C 1 ! C. Consequently, the subset B WD F".N/ D F".f0; : : : ; N 1g/ has
N elements, because the restriction FjN W N ! B is a bijection. Hence, with the
P
element Z WD F.N/, it follows that C D B[fZg with Z … B, whence

P
#.A[C/
D P
because C D B[fZg;
P [fZg/
#.A[ŒB P
D associativity of [;
P [fZg
#Œ.A[B/ P
D theorem 4.143;
P C #.fZg/
#.A[B/
D induction hypothesis;
Œ#.A/ C #.B/ C #.fZg/
D associativity of C;
#.A/ C Œ#.B/ C #.fZg/
D theorem 4.143;
#.A/ C #.C/:

t
u
The following two theorems confirm that every subset of a finite set is also finite.
4.145 Theorem. For each N 2 N, every subset S  N is also a finite set, with at
most N elements. Moreover, each proper subset S  N has fewer than N elements.
Proof. This proof proceeds by induction with N.
If N WD 0, then N D ¿, and the only subset S  N is S D ¿, which is finite.
As an induction hypothesis, assume that there exists a natural number K 2 N
such that the theorem holds for N WD K, so that each subset S  K is finite with
at most K elements, and that each proper subset S  K has fewer than K elements.
Hence, consider a subset R  K C 1. Two cases arise: either K … R or K 2 R.
If K … R, then R  Œ.K C 1/ n fKg D K whence R is finite with at most K
elements by induction hypothesis.
If K 2 R, then the set C WD R n fKg is a subset of Œ.K C 1/ n fKg D K, whence
C is finite with at most K elements by induction hypothesis. Thus, there exists a
 www.pdfgrip.com

250 4 Mathematical Induction: Definitions and Proofs by Induction

natural number L K such that C D R n fKg has cardinality L K, and L D K

P
if and only if C D K. Hence, theorem 4.143 shows that R D C[fKg is finite with
cardinality L C 1 K C 1, and L C 1 D K C 1 if and only if R D K C 1. t
u
4.146 Theorem. Every subset of a finite set is also finite, with at most as many
elements.
Proof. For each finite set A there exists a natural number N 2 N and a bijection
F W A ! N. Hence, for each subset B  A, the restriction FjB W B ! N is a
bijection from B onto a subset S WD F".B/  N. Because every such subset S  N
is also finite with at most N elements, it follows that there exists a natural number
L N and a bijection G W S ! L. Consequently, the composition GıFjB establishes
a bijection from B onto L, which means that B is finite with L N elements. t
u
4.147 Theorem. The equality #.A n B/ D Œ#.A/ Œ#.B/ holds for every subset
B  A of every finite set A.
Proof. The result follows from the disjointness of B and A n B, and from theo-
rem 4.146, which ensures that both B and A n B are finite:

B \ .A n B/ D ¿;
B [ .A n B/ D A;
Œ#.B/ C Œ#.A n B/ D #.A/;

whence #.A n B/ D Œ#.A/ Œ#.B/ by definition (4.59) of subtraction. t

u
Theorem 4.145 shows that for each subset S  N there is a bijection F W S ! L
with L N, but it does not yet prevent the existence of other bijections G W S ! L
with L > N. The following theorem confirms that there is no such bijection.
4.148 Theorem. For all K; N 2 N with K < N, there are no injections F W N ! K.
Proof. This proof uses induction with N. For N WD 1, the only smaller natural
number is K WD 0, and there exists no function F W f1g ! ¿, whence no injection
either.
As an induction hypothesis, assume that there exists a positive integer L 2 N
such that the theorem holds for N WD L, so that for every natural number K < L
there are no injections F W L ! K. The proof that for every K < L C 1 there
are no injections F W .L C 1/ ! K proceeds by contraposition. Thus, assume that
there is such an injection F W .L C 1/ ! K. Let Z WD F.L/ and S WD K n fZg. By
theorem 4.147, S is a finite set with K 1 elements, and there is a bijection G W S !
.K 1/. Then the restriction FjL W L ! S is an injection, and the composition
G ı FjL W L ! .K 1/ is an injection. Hence K 1  L by induction hypotheses,
whence K  L C 1. t
u
The following theorem shows the equivalence of the concepts of injection,
surjection, and bijection between sets with the same finite cardinality.
 www.pdfgrip.com

4.7 Finite Cardinality 251

4.149 Theorem. For all sets A and B with the same finite cardinality, and for each
function F W A ! B with domain D.F/ D A, the following conditions are mutually
equivalent:
.P/ F is injective,
.Q/ F is surjective,
.R/ F is bijective.
Proof. If F is bijective, then F is also injective and surjective, because .R/ , Œ.P/^
.Q/ by definition of bijectivity (definition 3.118); therefore both .R/ ) .P/ and
.R/ ) .Q/ hold, by theorems 1.53 and 1.54.
For the reverse implications, because A and B have the same finite cardinality,
there exist a natural number N 2 N and a bijection G W N ! A.
If F is injective, then F".A/  B is a finite set with cardinality L N, so that
there exists a bijection H W F".A/ ! L. Then the composition

G F H
H ı F ı G W N ! A ! F".A/ ! L

is an injection, whence L  N by theorem 4.148. From L N and N L it follows

that L D N. Hence H, G, and .H ı F ı G/ W N ! N are bijections, whence F is also
surjective, whence bijective, which proves the implication .P/ ) .Q/; therefore
.P/ ) Œ.P/ ^ .Q/ holds (by theorem 1.82), whence also .P/ ) .R/.
The proof of the converse, .Q/ ) .P/, uses the contraposition Œ:.P/ ) Œ:.Q/.
If F is not injective, then A contains two distinct elements X 6D Z such that F.X/ D
F.Z/. Let S WD A n fZg, so that F".A/ D F".S/ and there exists a bijection J W .N
1/ ! S, by theorem 4.147. Also, there is a bijection I W B ! N, because A and B
have the same cardinality, N. Then the composition

J F I
I ı F ı J W .N 1/ ! S ! F".S/ ! N

cannot be a bijection, because by theorem 4.148 its inverse could not be an injection
from N to N 1. Hence F cannot be surjective, which proves .Q/ ) .P/. t
u
The following theorem shows that the number of elements in a Cartesian product
equals the product of the numbers of elements in its factors.
4.150 Theorem. For all K; L 2 N, the Cartesian product K L has K  L elements.
Proof. By induction with L, if L D 0, then L D ¿, whence for every K 2 N

K 0 D K ¿ D ¿;
#.K 0/ D #.¿/ D 0 D K  0:

As induction hypothesis, assume that there exists a natural number M 2 N such

that the theorem holds for L WD M, so that the equality

#.K M/ D K  M
 www.pdfgrip.com

252 4 Mathematical Induction: Definitions and Proofs by Induction

holds for every K 2 N. Hence, from the disjoint union M C 1 D M [fMg,P and from
the distributivity of Cartesian products over unions (theorem 3.63), it follows that

K P
.M C 1/ D K .M [fMg/
P
D .K M/[.K fMg/;

#ŒK P
.M C 1/ D #Œ.K M/[.K fMg/
D Œ#.K M/ C Œ#.K fMg/
D .K  M/ C K
D K  .M C 1/;

thanks to the disjoint union .K P

M/[.K fMg/. t
u

4.7.3 Exercises on Finite Sets

4.111 . Determine #.¿/.

4.112 . Determine #ŒP.¿/.
 
4.113 . Determine # PŒP.¿/ .
  
4.114 . Determine # P PŒP.¿/ .
   
4.115 . Determine # P P PŒP.¿/ .
h n ˚ oi
4.116 . Determine # P ¿; f¿g; ¿; f¿g .

4.117 . Prove or disprove that all ordered pairs have the same cardinality.
4.118 . For all finite sets A and B, prove that .A B/ Ð .B A/.
4.119 . For all finite sets A, B, C, prove that Œ.A B/ C Ð ŒA .B C/.
4.120 . For all finite sets A and B, prove Œ#.AB/ D Œ#.A [ B/ Œ#.A \ B/.

4.8 Infinite Cardinality

4.8.1 Infinite Sets

Some sets are not finite, for they do not admit any bijection onto any natural number.
4.151 Definition (infinite sets). A set Z is infinite if and only if Z is not finite,
which means that there are no bijections from Z onto any natural number.
For instance, the set N is infinite.
 www.pdfgrip.com

4.8 Infinite Cardinality 253

4.152 Theorem. The set N of all natural numbers is not finite.

Proof. For each natural number N and for each function F W N ! N, the restriction
FjNC1 .N C 1/ ! N cannot be injective, by theorem 4.148, whence F cannot be
injective. Therefore, there are no such bijections, which means that N is not finite.
t
u
As there exist finite sets with different cardinalities, there also exist infinite sets
with different cardinalities. For instance, the following considerations lead to infinite
sets with cardinalities different from the cardinality of the natural numbers.
4.153 Definition. For all sets X, Y, let Y X denote the set of all functions from X to
Y, with domain X.
4.154 Example. If X WD ¿, then 2X D f¿g, because ¿ W ¿ ! f0; 1g is the only
function from X D ¿ to 2 D f0; 1g.
4.155 Example. If X WD fSg is a singleton, then 2X consists of two elements,
because there are exactly two functions from X D fSg to 2 D f0; 1g:

F W fSg ! f0; 1g;

S 7! 0I

G W fSg ! f0; 1g;

S 7! 1:

Thus, 2fSg D fF; Gg has exactly two elements, F and G.

The next theorem shows that each set X is “strictly smaller” than 2X .
4.156 Theorem. For each set X, there exists an injection from X to 2X . Yet there
does not exist any surjection from X to 2X .
Proof. To establish the existence of an injection X ,! 2X , consider the function
J W X ,! 2X defined as follows. The function J maps each element N 2 X to a
function J.N/ W X ! 2 specified by

1 if K D N;
ŒJ.N/.K/ WD
0 if K 6D N:

In other words, the function J.N/ is the characteristic function fNg of the singleton
fNg (from example 3.89). Consequently, J is injective; indeed if M 6D N, then
ŒJ.M/.M/ D 1 but ŒJ.N/.M/ D 0, whence J.M/ 6D J.N/.
To prove the absence of any surjection X  2X , for each function J W X ! 2X ,
this proof demonstrates a method known as Cantor’s diagonalization to show that
J is not surjective. Each such function J W X ! 2X maps each element N 2 X to a
function J.N/ 2 2X , so that J.N/ W X ! 2. In particular, J.N/ maps N to an element
ŒJ.N/.N/ 2 2 D f0; 1g. Thus, define a function F W X ! 2 by
 www.pdfgrip.com

254 4 Mathematical Induction: Definitions and Proofs by Induction

0 if ŒJ.N/.N/ D 1;
F.N/ WD
1 if ŒJ.N/.N/ D 0:

Thus, F.N/ 6D ŒJ.N/.N/ for every N 2 X, whence F 6D ŒJ.N/ for every N 2 X.

Consequently F … J".X/, whence J is not surjective. u
t
4.157 Example. One among several methods to define the set R of all real numbers
consists in defining R as a set of infinite sequences of digits [118, p. 565–566]. Thus
the set of all real numbers between 0 and 1 can be defined as the set of all sequences
R 2 3N (subject to the constraint that there does not exist any K 2 N such that
R.N/ D 2 for every N  K). Then every function J W N ! R fails to be surjective.
Indeed, as in the proof of theorem 4.156, for each function J W N ! R define a
function G W N ! R by
8
< 1 if ŒJ.N/.N/ D 2;
G.N/ WD 0 if ŒJ.N/.N/ D 1:
:
1 if ŒJ.N/.N/ D 0:

Thus, G 2 R but G.N/ 6D ŒJ.N/.N/ for every N 2 N, whence G 6D ŒJ.N/ for every
N 2 N. Consequently G … J".N/, and therefore J is not surjective. Hence there does
not exist any bijection J W N ! R.
Example 4.157 reveals that the set R of all real numbers is “more infinite” than
the set N of all natural numbers. Moreover, applying theorem 4.156 to X WD R
R
shows that 2R is also “more infinite” than R. Then 2.2 / is also “more infinite” than
2R . And so forth, thus there exists an “infinite” variety of “infinite” sets.

4.8.2 Denumerable Sets

There exist several infinite sets that have the same cardinality as N has. For instance,
using only addition and multiplication from integer arithmetic, this subsection
presents a proof that the set of all nonnegative integers N and the set of all integers
Z have the same cardinality; similarly, N and the Cartesian product N N have the
same cardinality. The following terminology conforms to [8, p. 152], [30, p. 47], and
[128, p. 151].
4.158 Definition. A set is denumerable — or has cardinality @0 (read “aleph
zero”) — if and only if it has the same cardinality as the set N of all natural numbers.
A set is countable if and only if it is either finite or denumerable.
The following theorem shows that the set Z of all the integers has the same
cardinality as the set N of all the nonnegative integers.
 www.pdfgrip.com

4.8 Infinite Cardinality 255

4.159 Theorem. The sets N and Z have the same cardinality.

Proof. Define F W N ! Z by

N=2 if there exists K 2 N with N D 2  K;

F.N/ WD
.N C 1/=2 if there exists K 2 N with N C 1 D 2  K:

The function F is surjective. Indeed, if L 2 N, then L D F.2  L/. Similarly, if

L 2 Z n N, then L D F.Œ2  L C 1/.
The function F is also injective. Indeed, if F.M/ D F.N/, then either both or
neither of F.M/ and F.N/ are elements of N. If F.M/ 2 N and F.N/ 2 N, then
M=2 D F.M/ D F.N/ D N=2, whence M D N. If F.M/ … N and F.N/ … N, then
.M C 1/=2 D F.M/ D F.N/ D .N C 1/=2, whence M D N. t
u
P is denumerable.
4.160 Theorem. For all disjoint denumerable sets A and B, A[B
Proof. By the hypotheses there exist bijections I W A ! N and J W B ! N. The
function K W N ! Z with K.N/ WD .N C1/ is injective, and hence the composition
P is a bijection from
K ı I is also injective. Therefore, the function G WD .K ı I/[J
P
A[B to Z. Hence F ı1 P
ı G W A[B ! N is a bijection, with F as in theorem 4.159.
t
u
To facilitate the proof that N and the Cartesian product N N have the same
cardinality, the following definition specifies an inductive method to define and
compute the sum 1 C 2 C C .N 1/ C N, known as an arithmetic series.
4.161 Definition. Define a function T W N ! N by

T.0/ WD 0;
T.N C 1/ WD T.N/ C .N C 1/:

Also, define the notation 0 C 1 C C N WD T.N/.

Thus,

T.0/ WD 0;
T.1/ WD T.0/ C .0 C 1/ D 0 C .0 C 1/ D 1;
T.2/ WD T.1/ C .1 C 1/ D 1 C .1 C 1/ D 3;
T.3/ WD T.2/ C .2 C 1/ D 3 C .2 C 1/ D 6;
::
:

The values of T are called triangular numbers because they correspond to the
number of elements in the following patterns:
 www.pdfgrip.com

256 4 Mathematical Induction: Definitions and Proofs by Induction

T.0/ T.1/ T.2/ T.3/

  
   
  

The following theorem provides a different formula to compute the same function
T.
4.162 Theorem (arithmetic series). For each natural number N 2 N,

N  .N C 1/
0C1C2C C .N 1/ C N D T.N/ D :
2

Proof. This proof uses induction with N. If N WD 0, then T.0/ D 0 D 0  .0 C 1/=2.

Assume that there exists M 2 N such that the theorem holds for N WD M, so that
T.M/ 2 N and 2  T.M/ D M  .M C 1/. Then T.M C 1/ D T.M/ C .M C 1/ 2 N,
and

2  T.M C 1/ D 2  ŒT.M/ C .M C 1/

D Œ2  T.M/ C Œ2  .M C 1/
D M  .M C 1/ C 2  .M C 1/
D .M C 2/  .M C 1/
D .M C 1/  Œ.M C 1/ C 1:

Consequently, 2  T.M C 1/ D .M C 1/  Œ.M C 1/ C 1, but T.M C 1/ 2 N, whence

.M C 1/  Œ.M C 1/ C 1=2 2 N and T.M C 1/ D .M C 1/  Œ.M C 1/ C 1=2. t u
An alternative proof of the same formula proceeds along the following outline:

T.N/ D 0 C 1 C C .N 1/ C N;
T.N/ D N C .N 1/ C C 1 C 0;
T.N/ C T.N/ D .N C 1/ C .N C 1/ C C .N C 1/ C .N C 1/;

whence 2  T.N/ D N  .N C 1/. However, a proof along this outline also requires
induction to rearrange the terms of the sum with associativity and commutativity.
The following definition provides a formula for Cantor’s diagonal enumeration
of N N, and the subsequent theorems will verify that indeed it enumerates N N.
4.163 Definition. Define a function T W .N N/ ! N by

T .M; N/ WD M C T.M C N/

.M C N/  .M C N C 1/
D MC :
2
 www.pdfgrip.com

4.8 Infinite Cardinality 257

The value T .M; N/ corresponds to the sum of the number of elements in the
triangular pattern counted by the “triangular number” T.M C N/ and a last partial
row with M elements (instead of a complete last row of M C N elements for the next
triangular number). For example, with M WD 1 and N WD 2,

T .1; 2/


T.M C N/  
  

M 

The following theorem shows that the function T W .N N/ ! N is surjective.

4.164 Theorem. For each I 2 N there exist M 2 N and N 2 N such that

.M C N/  .M C N C 1/
IDMC :
2
Proof. This proof proceeds by induction with I.
First, if I WD 0, then I D 0 C T.0 C 0/ with M WD 0 and N WD 0.
Second, assume that there exists K 2 N such that the theorem holds for I WD K,
so that there exist M; N 2 N with K D M C T.M C N/.
If N > 0, then K C 1 D 1 C ŒM C T.M C N/ D .M C 1/ C T.ŒM C 1 C ŒN 1/.
If N D 0, then K C 1 D 1 C ŒM C T.M C 0/ D .M C 1/ C T.M C 0/ D
0 C T.0 C ŒM C 1/. t
u
The following theorem shows that the function T W .N N/ ! N is injective.
4.165 Theorem. For all K; L; M; N 2 N, if K C T.K C L/ D M C T.M C N/, then
both K D M and L D N.
Proof. If

K C T.K C L/ D M C T.M C N/;

then subtracting M and T.K C L/ from both sides gives

K M D T.M C N/ T.K C L/:

If K > M, then K M > 0, so that the left-hand side is positive, but then the right-
hand side must also be positive: T.M C N/ T.K C L/ > 0. Hence M C N > K C L,
but then

T.M C N/ T.K C L/  T.M C N/ T.M C N 1/ D M C N

 www.pdfgrip.com

258 4 Mathematical Induction: Definitions and Proofs by Induction

by definition of T. Thus

K M D T.M C N/ T.K C L/  M C N

whence K  .2  M/ C N. With M C N > K C L, this gives

M C N > K C L  Œ.2  M/ C N C L

whence subtracting .2  M/ C N from all sides gives

M  L;

which contradicts the hypothesis that L  0. t

u
The following theorem confirms that N N is denumerable.
4.166 Theorem. The function T W .N N/ ! N is bijective.
Proof. The bijectivity results from theorems 4.165 and 4.164. t
u
The computation of the inverse function T 1 W N ! N N can proceed
according to the straightforward algorithm provided by the proof of theorem 4.164:
for each I 2 N, observe that T.0/ D 0 I, and compute T.0/; T.1/; : : : ; T.L/ until
T.L 1/ I < T.L/. Then let M WD I T.L 1/ and N WD .L 1/ M.

4.8.3 The Bernstein–Cantor–Schröder Theorem

The following theorem guarantees the existence of a bijection between two sets,
provided that there exist injections from one set to the other and vice versa. Accord-
ing to Suppes [128, p. 95], Cantor conjectured the theorem and then Bernstein and
Schröder proved it independently of each other in the 1890s. Fraenkel [37, p. 102]
credits the following proof to J. M. Whitaker.
4.167 Theorem (Bernstein–Cantor–Schröder). For all sets A and B, if there are
injections F W A ! B and G W B ! A, then there is a bijection H W A ! B.
Proof. The strategy of this proof consists in producing a subset E  A such that

G"ŒB n F".E/ D .A n E/

and then in setting

 
P Gı1 jAnE :
H WD .FjE / [

To this end, define

 www.pdfgrip.com

4.8 Infinite Cardinality 259

D WD fC  A W G"ŒB n F".C/  .A n C/g

D fC  A W C  A n G"ŒB n F".C/gI
S
the proof will verify that D satisfies the requirements.
First, for all subsets V; W  A, if V  W, then F".V/  F".W/, whence ŒB n
F".W/  ŒB n F".V/, and hence G"ŒB n F".W/  G"ŒB n F".V/, so that

fA n G"ŒB n F".V/g  fA n G"ŒB n F".W/g:

S if V 2 D, then VS A n G"ŒB n F".V/ by definition of D, and

In particular,
V  W WD D by definition of D; consequently
n h [ io
V  fA n G"ŒB n F".V/g  A n G" B n F" D :

Because these inclusions hold for every element V 2 D, it follows that they also
hold for their union:
[ n h [ io
D  A n G" B n F" D :

Let
h [ i
E WD A n G" B n F" D :
S
From D  E it follows that
h [ i
A n G" B n F" D  fA n G"ŒB n F".E/g;

so that

E  fA n G"ŒB n F".E/g
S S
whence E 2 D. Consequently, E  D, whence E D D, but then the definition
of E gives

E D A n G" ŒB n F" .E/ :

t
u
The next theorem shows a use of the Bernstein–Cantor–Schröder Theorem.
4.168 Theorem. There exists a bijection H W Z ! Q.
Proof. First, there exists an injection F W Z ! Q with F.N/ WD N=1.
Second, there exists an injection I W Q ! .Z Z/ such that I.P=Q/ WD .P; Q/
with P 2 N minimum (theorem 4.99).
 www.pdfgrip.com

260 4 Mathematical Induction: Definitions and Proofs by Induction

Also, there exists an injection J W .Z Z/ ! Z, by theorem 4.166.

Consequently, the composition G WD J ı I is an injection Q ! Z.
Therefore, the Bernstein–Cantor–Schröder Theorem guarantees the existence of
a bijection H W Z ! Q. t
u
Theorem 4.169 shows that every infinite subset of a denumerable set is denumer-
able. Hence every subset of a countable set is also countable.
4.169 Theorem. For each infinite subset S  N, there exists an injection H W N !
S. Hence S is denumerable.
Proof. Define G W P.S/ ! P.S/ by G.S/ WD S and G.B/ WD B [ fmin.S n B/g for
every B 6D S.
Define F W N ! P.S/ by F.0/ WD fmin.S/g, and F.N C 1/ WD GŒF.N/.
Thus F.N C 1/ inserts the smallest element of S that has not yet been included by
F.0/; : : : ; F.N/.
Let H.M/ WD maxŒF.M/.
Also, the inclusion function  W S ! N defined in example 3.93 by .X/ D X is
injective by example 3.121.
Hence there exists a bijection between S and N, by the Cantor–Bernstein–
Schröder Theorem (4.167). t
u
Theorem 4.170 shows that every subset of the range of a function defined on a
countable domain is also countable.
4.170 Theorem. For each set E, if there exists a surjection Q W N ! E, then E is
countable.
Proof. If E is finite, then E is countable, by definition.
If E is not finite, then define an injection G W E ! N by G.X/ WD
minŒQı1" .fXg/, so that G.X/ is the smallest natural number mapped to X by
Q. The function G is well-defined, because Q is surjective, so that the pre-image
Qı1" .fXg/ 6D ¿ contains a unique smallest element. The function G is also
injective; indeed, if X 6D Y, then Qı1" .fXg/ \ Qı1" .fYg/ D ¿, because Q is a
function. Consequently, the image S WD G".E/  N is an infinite subset of N.
By theorem 4.169 and the Bernstein–Cantor–Schröder Theorem (4.167), it
follows that S and hence E are denumerable. t
u

4.8.4 Denumerability of all Finite Sequences

of Natural Numbers

This subsection shows that the set of all finite sequences of natural numbers is
denumerable.
Theorem 4.171 shows that allowing the number of variables to change from one
function to another still produces a set of functions.
 www.pdfgrip.com

4.8 Infinite Cardinality 261

4.171 Theorem. For all sets E and C, all the functions F W EK ! C of any finite
number of variables from the domain E and with values in the co-domain C form a
set
n K o
C.E / W K 2 N :

Proof. Let A WD N, B WD E, and D WD N. Thus D D N  P.N/ D P.A/ because

K  N for each K 2 N, by theorem 4.20. Then, with the notation of theorem 3.136,
˚ 
FN;E WD EK W K 2 N

is the set of all finite sequences of elements of E. S

Next, again with the notation of theorem 3.136, let A WD FN;E . Then EK 2
FN;E whence E  A for each K 2 N, by theorem 3.40, and hence FN;E  P.A/.
K

Also let D WD FN;E and B WD C, the set C in the hypothesis of the theorem. Thus
C.E / is the set of all functions F W EK ! C defined on all of EK , with EK 2 D for
K

each K 2 N, and
n K o
FD;C WD C.E / W K 2 N

is the set of all functions defined on any EK into C.

4.172 Theorem. For each K 2 N , the set NK of all finite natural sequences of
length K is denumerable.
Proof. Following [86, p. 799], this proof proceeds by induction with K.
For K D 1, the identity function IN W N ! N, N 7! N defined in example 3.86 is
a bijection, by example 3.120.
For K D 2, Cantor’s diagonal enumeration T W N N ! N from definition 4.163
is a bijection, by theorem 4.166.
Denote the identity function IN by T1 , and Cantor’s diagonal enumeration T
by T2 .
As an induction hypothesis, assume that for some K D M there exists a bijection
TM W NM ! N.
For the induction step from K D M to K D M C 1, define TMC1 W NMC1 ! N
by

TMC1 .N0 ; : : : ; NM1 ; NM / WD T2 ŒTM .N0 ; : : : ; NM1 /; NM :

Equivalently, TMC1 D T2 ı .TM  IN /, with  from definition 3.116, and where

TM is bijective by induction hypothesis, whence TM  IN is also bijective, by
theorem 3.117, and hence the composition of the bijection T2 preceded by the
bijection TM  IN is again bijective, by theorem 3.123. t
u
 www.pdfgrip.com

262 4 Mathematical Induction: Definitions and Proofs by Induction

[
4.173 Theorem. The set NK of all finite sequences of natural numbers is
K2N
denumerable.
Proof. For each K 2 N , denote by TK W NK ! N any enumeration of all natural
sequences of length K, as in theorem 4.172. Again following [86, p. 799], this proof
[ all the bijections TK of N by means of K and T2 . Specifically,
K
pieces together
define S W N ! N by
K

K2N

S .N0 ; : : : ; NK1 / WD T2 ŒTK .N0 ; : : : ; NK1 /; K:

The map S is injective: indeed, if S .N0 ; : : : ; NK1 / D S .J0 ; : : : ; JL1 /,

then T2 ŒTK .N0 ; : : : ; NK1 /; K D T2 ŒTL .J0 ; : : : ; JL1 /; L, whence K D L and
TK .N0 ; : : : ; NK1 / D TL .J0 ; : : : ; JL1 /, because T2 is injective by theorem 4.165.
Consequently, TL .N0 ; : : : ; NL1 / D TL .J0 ; : : : ; JL1 /, whence .N0 ; : : : ; NL1 / D
.J0 ; : : : ; JL1 /, because TL is injective, by theorem 4.172.
The map S is also surjective: indeed, for each M 2 N there exists .I; K/ 2 N2
such that T2 .I; K/ D M, because T2 is surjective by theorem 4.164. Consequently,
there exists .N0 ; : : : ; NK1 / 2 NK such that TK .N0 ; : : : ; NK1 / D I, because TK is
surjective, also by theorem 4.172. Thus S .N0 ; : : : ; NK1 / D M. t
u

4.8.5 Other Infinite Sets

The present text has defined a set to be finite if and only if there exists a bijection
onto a natural number, and infinite if and only if there does not exist any such
bijection. There exists a different definition of infinite sets, called “Dedekind-
infinite” [128, p. 107], corresponding to Dedekind’s definition [25, V, #64, p. 63].
4.174 Definition. A set Z is Dedekind-infinite if and only if there exists a proper
subset Y ¤ Z and a bijection F W Z ! Y, or, equivalently, F ı1 W Y ! Z.
Thus, a set is Dedekind-infinite if and only if it has the same cardinality as that
of one of its proper subsets.
4.175 Example. The set N is Dedekind-infinite because there is a proper subset
N  N and a bijection defined by the successor function:

G W N ! N;
X 7! X [ fXg:

4.176 Theorem. Every Dedekind-infinite set is also infinite.

Proof. This proof proceeds by contradiction. If a set Z is not infinite, in other words,
if Z is finite, then there exist a natural number N 2 N and a bijection G W N ! Z.
 www.pdfgrip.com

4.8 Infinite Cardinality 263

If Z is also Dedekind-infinite , then there exists a proper subset Y ¤ Z and a bijection

F W Z ! Y. Then the composition

G F Gı1
H WD Gı1 ı F ı G W N ! Z ! Y ! N

would be a bijection from N onto a proper subset Gı1 ".Y/ ¤ N, which would
contradict theorem 4.148. t
u
The proof of the converse requires the Axiom of Choice (exercises 4.139, 4.140).
4.177 Theorem. Every infinite set contains a denumerable subset.
Proof. Apply recursion (theorem 4.21). If Z is infinite, then Z 6D ¿, whence there
exists some X 2 Z. Define F0 W f0g ! Z by 0 7! X. Assume that there exists
an injection FN W N ! Z. Then Z 6D FN ".N/ because Z is infinite, whence Z n
FN ".N/ 6D ¿ and there exists X 2S Z n FN ".N/. Let HN W fNg ! fXg and let
FNC1 WD FN [HP N . Finally, let F WD N2N FN . Then F W N ! Z is an injection. u t
Theorem 4.177 explains the subscript 0 in the notation @0 for the cardinality of N.
Because in the Zermelo-Frænkel set theory with the Axiom of Choice every infinite
set Z contains a denumerable subset, there exists an injection F W N ! Z, so that
the cardinality of N cannot exceed that of Z. If there is also an injection G W Z ! N,
then the Bernstein–Cantor–Schröder Theorem (theorem 4.167) guarantees that there
is also a bijection H W N ! Z; thus, the cardinality of Z cannot be strictly smaller
than that of N. Thus @0 represents the “smallest” infinite cardinality.
4.178 Theorem. Every infinite set is also Dedekind-infinite.
Proof. If W is infinite, then W contains a denumerable subset Z  W, by
 F W Z ! Y onto a proper
theorem 4.177. By example 4.175, there exists a bijection
subset Y ¤ Z. Extend F to all of W by setting H WD F [P IW jWnZ . t
u
4.179 Theorem. Every denumerable union of disjoint denumerable sets is denu-
merable.
Proof. If F is denumerable, then there is a bijection A W N ! F with I 7! A.I/. If
each A.I/ is denumerable, then by the Axiom of Choice there is a bijection
S FI W N !
A.I/ with J 7! FI .J/ 2 A.I/. Hence the function G W N N ! F defined by
G.I; J/ WD FI .J/ is a bijection. t
u

4.8.6 Further Issues in Cardinality

4.8.6.1 Other Axioms of Infinity

Bolzano [10] and Dedekind [25, V, Theorem 66, p. 64] argued for the “existence”
of an infinite set from practical considerations. Yet the “existence” of an infinite set
does not follow from the other axioms, but requires an axiom of infinity [8, p. 21].
 www.pdfgrip.com

264 4 Mathematical Induction: Definitions and Proofs by Induction

Zermelo [145] introduced such an axiom with a set such that if X is an element, then
fXg is also an element. The variant adopted here, with fXg replaced by X [ fXg, is
attributed to John von Neumann [135] and has proved more convenient [8, p. 22].
There also exist “infinitely” many other nonequivalent axioms of infinity [18, 57,
p. 342–346].

4.8.6.2 Peano’s Axioms

There is an alternative method to introduce natural numbers, published by Giuseppe

Peano, which does not involve set theory. Instead, Peano’s system consists of the
following axioms, stated here beginning with 0, as done in [46, 2, #II], [128,
p. 121].

Axiom A0 0 is a natural number.

Axiom A1 K D L means that K and L are the same natural numbers.

Axiom A2 For each natural number N there exists exactly one natural number
denoted by N 0 and called the successor of N.

Axiom A3 N 0 6D 0 for each natural number N.

Axiom A4 If K 0 D L0 , then K D L.

Axiom A5 For each set S of natural numbers, if

• 0 is an element of S, and
• if N belongs to S, then K 0 belongs to S,
then every natural number is an element of S.

From Peano’s axioms, and two additional axioms for recursive definitions of
addition and multiplication [128, p. 136], the same proofs as in this chapter verify
all the algebraic properties of arithmetic and ordering with natural numbers [76,
Ch. 1, p. 1–18]. However, in situations that involve other topics, for examples,
rational numbers or cardinality of sets, the use of Peano’s axioms would require
some theoretical link between Peano’s natural numbers and other sets, in other
words, some means of including Peano’s arithmetic and applications within the
same framework. This theoretical link can be the development of arithmetic from
within set theory, as done here.

4.8.6.3 Alternative Sequence of Developments

The progression from N to Z and then to Q allows for subtractions with integers
without requiring rational numbers, and then allows for divisions and subtractions
with rational numbers without requiring the development of “real” numbers.
 www.pdfgrip.com

4.8 Infinite Cardinality 265

An alternative development proceeds from N to QC , then to the nonnegative “real”

numbers RC and finally to the “real” numbers R and the “complex” numbers C, as
outlined by Edmund Landau [76]. Yet other constructions of the set of real numbers
from the axioms are detailed by Michael Henle [58] and John Stillwell [120].

4.8.6.4 The Generalized Continuum Hypothesis

Because there is no surjection from N to 2N (theorem 4.156), the question arises

whether there is any set S with a cardinality between the cardinalities of N and 2N .
In other words, the question pertains to the existence of a set S for which there exist
injections

I J
N ,! S ,! 2N

but no injections from 2N back to S and no injections from S back to N. The

hypothesis that no such set as S exists is called the continuum hypothesis.
More generally, because for every infinite set X there does not exist any surjection
from X to 2X , the question arises whether there exists any set S with a cardinality
strictly between the cardinalities of X and 2X . In other words, the question pertains
to the existence of a set S for which there exist injections

I J
X ,! S ,! 2X

but no injections from 2X back to S and no injections from S back to X. The

hypothesis that no such set as S exists is called the generalized continuum
hypothesis.
The axioms of set theory (S1–S8) are consistent with both the generalized
continuum hypothesis (as proved by K. Gödel [44, 45]) and with the negation of the
generalized continuum hypothesis (as proved by P. J. Cohen [19, 20, 22]). Therefore,
if the axioms of set theory are consistent, then the generalized continuum hypothesis
can be neither proved nor disproved within set theory. The generalized continuum
hypothesis is thus logically independent from set theory. Hence there exist two
mutually exclusive extensions of set theory, one extension with the generalized
continuum hypothesis, the other extension with the negation of the generalized
continuum hypothesis.

4.8.7 Exercises on Infinite Sets

4.121 . Prove that if A is infinite and if A  B, then B is also infinite.

P
4.122 . Prove that if A is denumerable and Z … A, then A[fZg is denumerable.
 www.pdfgrip.com

266 4 Mathematical Induction: Definitions and Proofs by Induction

P is denumerable.
4.123 . With A denumerable, B finite, disjoint, prove that A[B
4.124 . Prove that if A is denumerable and B finite, then A [ B is denumerable.
4.125 . Prove that if A and B are denumerable, then A [ B is denumerable.
4.126 . Prove that there exists a bijection from 2N to P.N/.
4.127 . Prove that P.N/ is not countable.
4.128 . Prove: if A is denumerable and B is finite, then A B is countable.
4.129 . For each nonempty set X prove that there is no injection 2X ,! X.
4.130 . Prove that if Œ#.X/ D Œ#.Y/, then Œ#.2X / D Œ#.2Y /.
4.131 . Prove: if Œ#.X/ < Œ#.Y/ are both finite, then Œ#.2X / < Œ#.2Y /.
4.132 . Prove that if X is a finite set, then 2X is also finite.
4.133 . Prove that if X is a finite set, then Œ#.X/ < Œ#.2X /.
4.134 . Prove that every infinite subset S  N is denumerable.
4.135 . For each denumerable set A prove that every subset S  A is countable.
4.136 . Prove: if A is uncountable and A  B, then B is uncountable.
4.137 . Prove that if A, B, and C are denumerable, then so is .A B/ C.
4.138 . Prove that if A B is denumerable, then A and B are countable.
4.139 . Show where the proof of theorem 4.179 invokes the Axiom of Choice.
4.140 . Show where the proof of theorem 4.177 invokes the Axiom of Choice.
 www.pdfgrip.com

Chapter 5
Well-Formed Sets: Proofs by Transfinite
Induction with Already Well-Ordered Sets

5.1 Introduction

This chapter focuses on “well-formed” sets, which are defined by means restricted
to the axioms of Zermelo-Fraenkel set theory from chapters 1, 2, 3, and 4. The
main result states that no two well-formed sets are members of each other, and
consequently that every well-formed set is not an element of itself.
This chapter shows the dependence of one axiom on the others for sets — called
well-formed sets — defined in specific ways solely through the axioms of set theory.
Such well-formed sets suffice for most of logic, mathematics, computer science, and
their applications to the sciences and engineering. Among other features, the result
shows that no well-formed set is an element of itself, and that this result is provable
from the other axioms of set theory [74]. This chapter also provides a way to revisit
induction (chapter 4) on a different level.

5.2 Transfinite Methods

5.2.1 Transfinite Induction

Transfinite methods lead to an example of decidability in set theory. On the set

N, the Principle of Mathematical Induction (theorem 4.15) is logically equivalent
to the Well-Ordering Principle (theorem 4.47), which states that every nonempty
subset S  N has a smallest element. All well-ordered sets also lend themselves to
a method of proof known as transfinite induction, which relies on “initial intervals”
in well-ordered sets. All sets in this chapter are already well-ordered.
5.1 Definition. For each set W well-ordered (definition 3.194) by a relation  and
for each C 2 W, the initial interval determined by C is the subset

© Springer Science+Business Media New York 2015 267

Y. Nievergelt, Logic, Mathematics, and Computer Science,
DOI 10.1007/978-1-4939-3223-8_5
 www.pdfgrip.com

268 5 Well-Formed Sets: Proofs by Transfinite Induction with Already Well-Ordered Sets

WC WD fB 2 W W .B  C/ ^ .B 6D C/g;

which consists of all elements preceding C but different from C.

5.2 Example. If W D N, with < for , then NN D N for each N 2 N.
The Principle of Mathematical Induction extends to all well-ordered sets.
5.3 Theorem (Transfinite Induction). For each set W well-ordered by  and for
each set V, if V  W, then V D W if and only if

8CfŒ.C 2 W/ ^ .WC  V/ ) .C 2 V/g:

Proof. If V D W, then the formula is a tautology: Œ.P/ ^ .Q/ ) .P/. For the
converse, let U WD W n V. If U 6D ¿, then U has a first element A 2 U. Thus, if
B  A but B 6D A, then B … U, whence B 2 W n U D V. Hence the initial interval
WA  V, but then A 2 V by hypothesis on V, which contradicts A 2 U D W nV. u t
Well-ordered sets also lend themselves to a method of definition known as
transfinite construction, which relies on the concept of “ideal” in a well-ordered
set.
5.4 Definition (ideal). For each set W well-ordered by a relation  a subset V 
W is an ideal of W if and only if V contains every element preceding any of its
elements. Thus V is an ideal if and only if WC  V for every C 2 V:

8B8CfŒ.B 2 W/ ^ .B  C/ ^ .C 2 V/ ) .B 2 V/g:

The set of all ideals of W relative to  is denoted by I .W/.

5.5 Example. If W D N, with < for , then N is an ideal for each N 2 N.
The following two theorems yield relations between ideals and initial intervals.
5.6 Theorem. For each set W well-ordered by a relation  and for each ideal
V  W, if B 2 W n V, then V  WB .
Proof. By definition of an ideal, if C 2 V and B 2 W with B  C, then B 2 V.
By contraposition, C … V follows from B 2 W n V, and C 2 W with B  C.
Because  totally orders W (remark 3.195), it follows that if B 2 W n V, then
.W n V/  ŒW n .WB [ fBg/, whence V  WB [ fBg. Yet B … V, whence V  WB .
t
u
5.7 Theorem. For each set W well-ordered by a relation  and for each ideal
V ¤ W, there exists a smallest element A 2 W n V, and for this element WA D V.
Proof. If V ¤ W, then W n V 6D ¿. Because W is well-ordered by  it follows that
W n V contains a smallest element A 2 W n V. Thus for every B 2 W such that
B  A and B 6D A, it follows that B 2 V by minimality of A in W n V. Therefore
WA  V. Moreover, V  WA by theorem 5.6. t
u
 www.pdfgrip.com

5.2 Transfinite Methods 269

The following theorems show that the set of all ideals is well-ordered by
inclusion.
5.8 Theorem. For each set W well-ordered
T by a relation  and for each nonempty
set F of ideals of W, the intersection F is also an ideal of W.
T
Proof. If C 2 F , then C 2 V for each ideal V 2 F . Hence, if B 2 W and B  C,
B 2 V because
then T T V is an ideal. This conclusion holds for each V 2 F , whence
B 2 F . Thus F is an ideal of W. t
u
5.9 Theorem. For each set W well-ordered by a relation
T  and for each nonempty
set F of ideals of W, the smallest element of F is F . Therefore the set I .W/
is well-ordered by inclusion ./.
Proof. For each nonempty setT F of ideals of W, there exists at least one ideal
U 2 F , and the intersection F is also an ideal, by theorem 5.8. Let
n [  \ o
Z WD C 2 W W C 2 F n F :
S T
If Z D ¿, then F D F , whence F contains only one ideal, in effect U D
T
F. T
If Z 6D ¿ then it has a smallestTelement A 2 Z  W. BecauseT F is an ideal
by theorem 5.8, and because
T A … F , it follows that WA D F by theorem 5.7.
Also because A … F , there exists an ideal V 2 F with A … V. Hence V  WA
by theorem 5.6. T T T
From V  TWA and WA D F it follows that V  F , but F  V.
Consequently F D V 2 F .
T Therefore every nonempty set F  I .W/ has a smallest element, in effect
F , so that I .W/ is well-ordered by set inclusions. t
u

5.2.2 Transfinite Construction

As the Principle of Mathematical Induction leads to a method of definition by

induction (theorem 4.21) — also called a recursive definition or recursion —
similarly transfinite induction also yields a method of definition by transfinite
induction.
5.10 Theorem (Transfinite Construction). For each nonempty set W well-
ordered by a relation  with first element A 2 W, and for each nonempty set
E, let
[
Y WD EWB
B2W
 www.pdfgrip.com

270 5 Well-Formed Sets: Proofs by Transfinite Induction with Already Well-Ordered Sets

denote the set of all functions with domain equal to an initial interval WB and with
range in E (definition 4.153). For each Z 2 E, and for each function P W Y ! E,
there exists exactly one function F W W ! E such that F.A/ D Z and F.B/ D
P .FjWB /.
Proof. This proof establishes the uniqueness and existence separately.

Uniqueness

There is at most one such function. Indeed if F W W ! E and G W W ! E are two

such functions, with F.A/ D Z D G.A/ and F.B/ D P .FjWB /, G.B/ D P .GjWB /,
then let

S WD fC 2 W W F.C/ 6D G.C/g:

If S 6D ¿, then S has a smallest element, D 2 S. Hence F.B/ D G.B/ for every

B  D in W, which means that FjWD D GjWD , but then

F.D/ D P .FjWD / D P .GjWD / D G.D/

would contradict D 2 S. Consequently, S D ¿, so that F.B/ D G.B/ for every

B 2 W, which means that F D G.

Existence

Let F denote the set of all ideals V  W for which there exists a function
FV W V ! E such that FV .A/ D Z and FV .B/ D P .FV jWB /.
For each ideal U 2 F , applying the uniqueness just proved to the well-ordered
set U \ V instead of W shows that the functions FU and FV coincide on the well-
ordered subset U \ V in W. S
Hence, define a function FF W F ! E by settingS FF .B/ WD FU .B/ for
any ideal U 2 F with B 2 U. In other terms, FF D U2F FU . The preceding
argument confirms that this definition does not depend on which ideal U contains
B, because if B 2 U and B 2 V, then FU .B/ D FV .B/.
Next, if an ideal U is an initial interval, U D WB for some B 2 W, and if WB 2 F ,
then WB [ fBg 2 F . Indeed, a function FU W U ! E extends to WB [ fBg by the
definition FWB [fBg .B/ WD P.FU /.
Suppose that W … F . Then let G denote the set of Tall the ideals of W that are not
elements of F . In particular, W 2 G . Define V WD G , which is then the smallest
ideal of W in G . If V had a last element D, then V D W TD [ fDg by definition of
WD and of an ideal; however, WD 2 F , otherwise V 6D G , but from WD 2 F it
follows that WD [ fDg 2 F . Thus V cannot have a last element.
 www.pdfgrip.com

5.3 Transfinite Sets and Ordinals 271

S
If V does not have a last element, then V D B2V WB by definition
S of an ideal.
S
Again, it follows that WB 2 F for each B 2 V, whence V D B2V WB  F
and then V 2 F because of the existence of FF , contradicting the definition of V.
Therefore, W 2 F , which means that F extends to all of W. t
u

5.2.3 Exercises on Transfinite Methods

5.1 . Prove that in each well-ordered set each initial interval is an ideal.
5.2 . Provide an example of an ideal that is not an initial interval in a well-
ordered set.
5.3 . Prove that if Z is well-ordered by 4, then 4 differs from .
5.4 . Prove that if Q is well-ordered by 4, then 4 differs from .
5.5 . Provide an example of a well-order  on a set of modular integers ZM D
fŒ0M ; : : : ; ŒM 1M g and modular integers ŒIM , ŒKM , ŒLM , such that ŒKM  ŒLM
but ŒIM C ŒKM 6 ŒIM C ŒLM .
5.6 . Provide an example of a well-order  on a set of modular integers ZM D
fŒ0M ; : : : ; ŒM 1M g and modular integers ŒIM , ŒKM , ŒLM , such that Œ0M  ŒIM
and ŒKM  ŒLM but ŒIM  ŒKM 6 ŒIM  ŒLM .
5.7 . Prove that every subset of a well-ordered set is also well-ordered.
S
5.8 . Determine whether for each set F of ideals in W the union F is also an
ideal in W.
S
5.9 . Determine whether for each set G of initial intervals in W the union G is
also an initial interval in W.
5.10 . Determine whether for each set G of initial intervals in W the intersection
T
G is also an initial interval in W.

5.3 Transfinite Sets and Ordinals

5.3.1 Transitive Sets

Sets defined exclusively through the axioms of set theory adopted here are called
well-formed sets. They have the advantage of avoiding certain contradictions that
would arise from defining sets by means not so strict. The definition of well-formed
sets involves the concept of sets that are “transitive” relative to the relation 2.
 www.pdfgrip.com

272 5 Well-Formed Sets: Proofs by Transfinite Induction with Already Well-Ordered Sets

5.11 Definition (Transitive Sets). A set A is transitive if and only if every element
of A is also a subset of A, so that 8XŒ.X 2 A/ ) .X  A/, or, equivalently,

8Y8XfŒ.Y 2 X/ ^ .X 2 A/ ) .Y 2 A/g:

5.12 Example. The following sets are transitive:

¿;
f¿g;
˚ 
n¿; f¿g ;˚ o
¿; f¿g; ¿; f¿g ;
n ˚  ˚ o
¿; f¿g; f¿g ; ¿; f¿g :
˚ 
5.13 Counterexample. The set A WD f¿g is not transitive, because it contains
an element X WD f¿g that is not a subset of A: X 6 A, because ¿ 2 X but ¿ … A.
Power sets, unions, and intersections of transitive sets are also transitive.
5.14 Theorem. If a set A is transitive, then P.A/ is also transitive.
Proof. If S 2 P.A/, then S  A. Thus if X 2 S, then X 2 A, and X  A by
transitivity of A. Hence X 2 P.A/ for each X 2 S, whence S  P.A/. t
u
S
5.15 Theorem. If a set F is transitive, then F is also transitive.
S
Proof. If S 2 F , then there exists A 2 F with S 2 A. YetS A  F by transitivity
of F . From S 2 A and A  F follows S 2 F , whence S  F . t
u
S T
5.16 Theorem. If F is a nonempty set of transitive sets, then F and F are
also transitive.
S
Proof. If S 2 F , then S S 2 A for some
S A 2 F , whence S  A by transitivity of
A, and hence
T S  A  F , so that F is transitive.
If S 2 F T , then S 2 A for
T every A 2 F , whence S  A by transitivity of A,
and hence S  F , so that F is transitive. t
u

5.3.2 Ordinals

Well-formed sets will rely on the concept of ordinals (also called “ordinal numbers”
[30, p. 42]). The following definition conforms to Kunen’s [74, p. 16].
5.17 Definition (ordinals). A set A is an ordinal if and only if it is a transitive set,
and the relation 2 is an irreflexive well-ordering on the set A.
 www.pdfgrip.com

5.3 Transfinite Sets and Ordinals 273

5.18 Example. The following sets are ordinals:

¿;
f¿g;
˚ 
n¿; f¿g ;˚ o
¿; f¿g; ¿; f¿g :

5.19 Counterexample. The set

n ˚  ˚ o ˚ 
A WD ¿; f¿g; f¿g ; ¿; f¿g D P ¿; f¿g

is transitive but not an ordinal. Indeed, if

X WD ¿;
˚ 
Y WD f¿g ;

then X 2 A and Y 2 A, so that fX; Yg  A, but fX; Yg does not have any smallest
element relative to the relation 2, because X … Y and Y … X.
In particular the subset fX; Yg is not an ordinal.
5.20 Theorem. The empty set is an element of every nonempty ordinal.
Proof. By definition, every ordinal A is transitive, so that if X 2 A then X  A.
Consequently, if Y 2 X and X 2 A, then Y 2 A. Therefore, if X 2 A and X 6D ¿,
then X is not the smallest element of A. Yet every nonempty ordinal A has a smallest
element. Hence contraposition shows that the smallest element must be ¿. t
u
5.21 Theorem. If A is an ordinal, then A … A. Moreover, A … X for each X 2 A. In
particular, if A and X are ordinals, then A … X or X … A.
Proof. If A is an ordinal, then 2 is connected (definitions 3.185, 3.194) and
irreflexive (definition 5.17): exactly one of X 2 Y, X D Y, or Y 2 X holds for
all X; Y 2 A. Because A D A, it follows that A … A. Moreover, if A is an ordinal
and X 2 A, then X  A, whence if Y 2 X then Y 2 A. With Y WD A, it follows by
contraposition that A … X. t
u
5.22 Theorem. Every element of an ordinal is an ordinal.
Proof. If A is an ordinal and X 2 A, then X  A because A is a transitive set. Hence
2 well-orders X, because 2 well-orders A. If moreover Z 2 Y and Y 2 X, then
Z 2 X because A is a transitive set, whence Y  X. Thus X is also a transitive set.
Furthermore, the relation 2 remains irreflexive on the subset X  A. u
t
5.23 Theorem. If A isSan ordinal, then either A contains a last element D and
A D D [ fDg, or A D A is the union of all its elements.
 www.pdfgrip.com

274 5 Well-Formed Sets: Proofs by Transfinite Induction with Already Well-Ordered Sets

S
an ordinal and B 2 A, then B  A; consequently, A  A.
Proof. If A is S
Let D WD A. If D 6D A, then there exists X 2 A n D. Then X  D because
X 2 A;
S hence X is an ordinal by theorem 5.22, whence X 6D A by theorem 5.21, and
X  A D D.
Conversely, still with X 2 A nS D, for each B 2 A it follows
S thatSX … B, whence
B 2 X, and hence B  X, so that B2A B  X. Thus, D D A D B2A B  X.
Therefore, if D 6D A, then D  X  D, whence X D D is the only element of
A n D, and hence A D D [ fDg. t
u
5.24 Theorem. If B is an ordinal, then B [ fBg is also an ordinal.
Proof. First, B [ fBgis transitive. Indeed, if X 2 B [ fBg, then either X 2 B, whence
X  B  B [ fBg, or X 2 fBg, whence X D B  B [ fBg.
Second, 2 well-orders B [ fBg. Indeed, for each nonempty subset S  B [ fBg,
two situations can occur: S  B or S \ fBg 6D ¿. If S  B then S has a smallest
element, because B is an ordinal. If S \ fBg 6D ¿, then either S D fBg has the
smallest element B, or S \ B 6D ¿ and then S \ B is a subset of B and hence has a
smallest element, which is then also a smallest element of S D .S \ B/ [ fBg.
Moreover, the relation 2 remains irreflexive and transitive on B [ fBg. Indeed,
B … B by theorem 5.21. Furthermore, if X 2 Y and Y 2 Z in B [ fBg, then Y 2 B
from either Z 2 B or Z 2 fBg. Also, Y 6D B by theorem 5.21, which forbids B 2 Z
and Z 2 B [ fBg, and hence X 6D B also by theorem 5.21, which forbids B 2 Y
Y 2 Z, and Z 2 B [ fBg. Consequently only two cases can occur: Z 2 B or Z D B.
If Z D B, then X 6D B and Y 6D B, whence X 2 Z.
If Z 2 B, then X, Y, and Z all three lie in B, whence X 2 Z because 2 is transitive
on the well-ordered set B.
Finally, 2 is strict on B [ fBg. Indeed, because 2 is strict on B, it follows that if
X 2 B [ fBg and Y 2 B [ fBg, then two different cases can arise.
If X 2 B and Y 2 B, then X 2 Y and Y 2 X cannot both hold, for transitivity
would yield X 2 X which cannot hold by strictness of 2 on B.
If X 2 B and Y 2 fBg, then X 2 Y and Y 2 X cannot both hold. Otherwise Y D B
and then X 2 B and B 2 X. However, X is also an ordinal by theorem 5.22, whence
2 is also transitive on X, so that B 2 X and X 2 B yield B 2 B, which cannot hold
by strictness of 2 on X. t
u

5.3.3 Well-Ordered Sets of Ordinals

The following theorems show that every set of ordinals is well-ordered. First  is
strongly connected (definition 3.184) on every set of ordinals.
5.25 Theorem. For all ordinals A and B, either A  B, or A D B, or B  A.
Proof. If B 6 A, then there exists X 2 B n A, and hence there exists a smallest such
element: X 2 B n A, so that X 2 Y for every Y 2 B n A with Y 6D X. Also, X 6D ¿,
 www.pdfgrip.com

5.3 Transfinite Sets and Ordinals 275

because X … A but ¿ 2 A. From X 2 B it follows that X  B, whence if Y 2 X, then

Y 2 B, but then Y 2 A because of the minimality of X. Thus X  A \ B.
Conversely, if Y 2 A \ B, then Y 2 B, whence either Y 2 X or X 2 Y, because
X 2 B also. However, X 2 Y cannot occur, because X 2 Y and Y 2 A \ B would
yield X 2 A. Thus, if Y 2 A \ B, then Y 2 X, which means that A \ B  X.
Consequently, X D A \ B.
The foregoing argument with A and B switched shows that if A 6 B, then Z WD
B \ A is the smallest element in A n B. In particular, Z D B \ A D A \ B D X.
Consequently, if A 6 B and B 6 A both held, then X WD A \ B DW Z would be
the smallest element in both A n B and B n A.
However, because .A n B/ \ .B n A/ D ¿, it follows that X … .A n B/ \ .B n A/.
Thus at least one of B 6 A or A 6 B must fail to hold, which means that B  A or
A  B or both, so that either A  B, or A D B, or B  A. t
u
The second theorem shows that 2 is strongly connected on every set of ordinals.
5.26 Theorem. For all ordinals A and B, either A D B, or A 2 B, or B 2 A.
Proof. Consider the sets C WD A [ fAg and D WD B [ fBg, which are ordinals by
theorem 5.24. Applying theorem 5.25 to C and D instead of A and B shows that
D  C or C  D. If C  D, then A 2 C  D D B [ fBg, so that either A 2 B or
A D B. If D  C, then B 2 D  C D A [ fAg, so that either B 2 A or B D A. t
u
5.27 Theorem. Every set F of ordinals is well-ordered by 2.
Proof. The relation 2 is irreflexive on F by theorem 5.21, and it is connected by
theorem 5.26. The relation 2 is also transitive on F . Indeed, for all A; B; C 2 F , if
A 2 B and B 2 C, then B  C whence A 2 C.
Moreover, for each nonempty subset G  F , there exists some C 2 G . For each
B 2 G , either B 2 C, or B D C, or C 2 B, by theorem 5.26. Define

E WD fB 2 G W B 2 Cg D C \ G :

If E D ¿, then C is the smallest element of G , because then C 2 B for each B 2 G

with B 6D C. If E 6D ¿, then E has a smallest element A 2 E, because E is a
subset of the ordinal C. If B 2 G , then B … A, because B 2 A would yield B 2 C,
contradicting the minimality of A. Hence A 2 B for every B 2 G with B 6D A. t
u

5.3.4 Unions and Intersections of Sets of Ordinals

The union and the intersection of every nonempty set of ordinals is an ordinal.
S
5.28 Theorem. For each set F of ordinals, F is also an ordinal.
S
Proof. The union F is transitive, by theorem 5.15.
 www.pdfgrip.com

276 5 Well-Formed Sets: Proofs by Transfinite Induction with Already Well-Ordered Sets

S S
The union F is well-ordered. Indeed, for each nonempty subset S  F,
let

E WD fA 2 F W A \ S 6D ¿g:

From S 6D ¿ follows E 6D ¿. Hence E  F has a smallest element A, because

F is an ordinal, by theorem 5.27. Thus S \ A 6D ¿. Therefore, S \ A  A also
has a smallest element B 2 S \ A. Also, for each C 2 S there exists D 2 F with
C 2 D. Then either A D D, or A 2 D, or D 2 A. Yet D 2 A cannot occur, by
minimality of A. From B 2 A, with A 2 D or A D D, follows B 2 D; hence B 2 D
and C 2 D. Consequently, either B 2 C or C 2 B or B D C, but C 2 B cannot occur
by minimality of B and because B 2 A. Thus, B 2 C or B DSC, which shows that B
is the smallest element of S. By theorem 5.21 2 is strict on F , for every element
is an ordinal by theorem 5.22. t
u
T
5.29 Theorem. For each nonempty set F of ordinals, F is also an ordinal.
T
Proof. The intersection F is a transitive set by theorem 5.16.
T T The intersection
F is also well-ordered. Indeed, each nonempty subset S  F is also a subset
S  A of some A 2 F and hence S has a smallest
T element, because A is an ordinal.
The relation 2 is a strict totalTorder on FT, as it is on every subset of A 2 F ,
where it is strict. Indeed, if X 2 F and Y 2 F , then X 2 A and Y 2 A whence
X 2 Y and Y 2 X cannot both hold. t
u

5.3.5 Exercises on Ordinals

5.11 . Prove that N is a transitive set.

5.12 . Prove that every natural number N 2 N is a transitive set.
5.13 . Prove that N is an ordinal.
5.14 . Prove that every natural number N 2 N is an ordinal.
5.15 . Investigate whether 2 is a transitive relation on every transitive set.
5.16 . Prove that every ordinal is an element of some ordinal.
5.17 . Prove that every ordinal is a subset of some ordinal.
5.18 . Determine whether every ordinal is a subset of some transitive set.
5.19 . Determine whether every transitive set is a subset of some ordinal.
5.20 . Verify that it is not necessary to assume that V 6D ¿ for Transfinite Induction.
In other words, prove that for each set W well-ordered by a relation  and for each
subset V  W, if the following formula is True,

8CfŒ.C 2 W/ ^ .WC  V/ ) .C 2 V/g;

then either W D ¿ or V 6D ¿.
 www.pdfgrip.com

5.4 Regularity of Well-Formed Sets 277

5.21 . Determine whether every singleton fAg with an ordinal A is an ordinal.

5.22 . Determine whether fA; Bg is an ordinal for all ordinals A and B.
5.23 . Determine whether every set of ordinals is an ordinal.
5.24 . Determine whether every subset of every ordinal is an ordinal.
5.25 . Prove that there is an ordinal whose power set is not an ordinal.
5.26 . Prove that every countable set admits a well-ordering.

5.4 Regularity of Well-Formed Sets

5.4.1 Well-Formed Sets

The following definition establishes sets that contain all the well-formed sets.
5.30 Definition. For each ordinal A, define a set R.A/ by transfinite construction:
• R.¿/ WD ¿;
• R.A [ fAg/
S WD PŒR.A/;
• R.A/ WD B2A R.B/ if there does not exist any ordinal B such that A D B [ fBg,
but if R.B/ has been defined for every B 2 A.
5.31 Definition. A set X is well-formed if and only if there exists an ordinal A such
that X 2 R.A/.
5.32 Remark. The transfinite construction proceeds as follows. For each ordinal W,
let

E WD PŒP.W/;
S
Y WD A2W EWA ;
PW Y ! E;
S
P.RjWA / WD B2A R.B/:

The values of P remain in E, because if RjWA W WA ! E, then R.B/ S 2 E D

PŒP.W/ for eachS B 2 W A D A, whence R.B/  P.W/, and hence B2A R.B/ 
P.W/, so that B2A R.B/ 2 PŒP.W/. Yet for all ordinals U and V, U  V
or V  U by theorem 5.25, so that PŒP.U/  PŒP.V/ or PŒP.V/ 
PŒP.U/. Therefore, the transfinite construction defined with W WD U or W WD V
gives the same definition on U \ V. In other words, for each ordinal A, the definition
of R.A/ can proceed from any ordinal W containing A, for example, W WD A [ fAg.
 www.pdfgrip.com

278 5 Well-Formed Sets: Proofs by Transfinite Induction with Already Well-Ordered Sets

5.33 Example. The first few sets of the form R.A/ are also ordinals:

R.¿/ D ¿;
R.f¿g/ D R.¿ [ f¿g/ D PŒR.¿/ D PŒ¿ D f¿g;
˚   ˚  ˚ 
R ¿; f¿g D R f¿g [ f¿g D PŒR.f¿g/ D PŒf¿g D ¿; f¿g :

The next ordinal,

n ˚ o
A WD ¿; f¿g; ¿; f¿g ;
˚ 
is not of the form R.B/ for any set B, but A D B[fBg for the ordinal B WD ¿; f¿g .
Hence, the list continues with
n ˚  o ˚  n˚ o
R ¿; f¿g; ¿; f¿g D R ¿; f¿g [ ¿; f¿g
 ˚ 
D P R ¿; f¿g
˚ 
DP n ¿; f¿g
˚  ˚ o
D ¿; f¿g; f¿g ; ¿; f¿g ;

which is of the form R.A/, but it is not well-ordered by 2 and hence not an ordinal,
by counterexample 5.19. The set R.A/ is also different from the ordinal

˚  n ˚ o
¿; f¿g; ¿; f¿g ; ¿; f¿g; ¿; f¿g :

5.34 Theorem. For each ordinal Q, the set R.Q/ is transitive.

Proof. This proof proceeds by transfinite induction (theorem 5.3).
Choose an ordinal W with Q 2 W, for instance, W WD Q [ fQg. Then consider
the subset V  W of all the elements C 2 W for which R.C/ is transitive.
Consider any element C 2 W such that WC  V. By transfinite induction, it
suffices to verify that C 2 V to establish that V D W. Because

WC D fB 2 W W .B 2 C/ ^ .B 6D C/g

by definition 5.1 for the relation 2 instead of , it follows that WC D C. From

C D WC  V it then Sfollows that R.B/ is transitive for each B 2 C by definition
of V. Consequently, B2C R.B/ is also transitive, by theorem 5.15.
Two cases can arise, either C does not have a lastS
element, or C has a last element.
If C does not have a last element, then R.C/ D B2C R.B/ is transitive.
If C has a last element Z 2 C, then C D Z [ fZg. From Z 2 C D WC  V, it
follows that R.Z/ is transitive by definition of V. Hence PŒR.Z/ is also transitive,
by theorem 5.14. Yet R.C/ D R.Z [ fZg/ D PŒR.Z/, whence R.C/ is transitive.
Thus in either case R.C/ is transitive, whence C 2 V, and thence V D W.
Finally, from Q 2 W D V it follows that R.Q/ is transitive by definition of V. u t
 www.pdfgrip.com

5.4 Regularity of Well-Formed Sets 279

5.4.2 Regularity

The following theorems confirm that every element, subset, pairing, power set,
union, intersection, and Cartesian product of well-formed sets is again a well-
formed set.
5.35 Theorem. For each well-formed set X there is a smallest ordinal A with
X 2 R.A/.
Proof. If X is a well-formed set, then there exists an ordinal C such that X 2 R.C/ 2
PŒR.C/ D R.C [ fCg/. Hence X 2 R.C [ fCg/ by transitivity (theorem 5.34). By
theorem 5.27, there exists a smallest ordinal A 2 C [ fCg such that X 2 R.A/.
For every ordinal D, either D D A, or D 2 A 2 C [ fCg and then X … R.D/, or
A 2 D and then D is not the smallest such ordinal. u
t
5.36 Theorem. If X is well-formed, then every Y 2 X is well-formed.
Proof. If X is well-formed, then there exists an ordinal A such that X 2 R.A/.
If A D ¿, then X 2 R.A/ D R.¿/ D ¿, whence Y 2 X is vacuously well-
formed.
If there exists an ordinal B such that A D B [ fBg, then X 2 R.A/ D PŒR.B/,
whence X  PŒR.B/ by transitivity of PŒR.B/, Consequently,
S Y 2 R.A/ D
PŒR.B/ is well-formed for each Y 2 X. If R.A/ D B2A R.B/ and the theorem
holds for each Z 2 R.B/ for each B 2 A, then for each X 2 R.A/ there exists B 2 A
such that X 2 R.B/ whence every Y 2 X is also well-formed. t
u
S
5.37 Theorem. If X and Y are T well-formed sets, then so are fX; Yg, P.X/, X,
X Y, every subset of X, and X provided X 6D ¿.
Proof. If X and Y are well-formed sets, then there exist ordinals A and B such that
X 2 R.A/ and Y 2 R.B/. Either A D B (whence R.A/ D R.B/), or A 2 B (whence
R.A/  R.B/), or B 2 A (whence R.B/  R.A/). For instance, assume that R.A/ 
R.B/. Thus X 2 R.B/ and Y 2 R.B/, so that fX; Yg 2 PŒR.B/ D R.B [ fBg/,
whence fX; Yg is well-formed, because B [ fBg is an ordinal.
Similarly, if X 2 R.B/ is a well-formed set, then X  R.B/ by transitivity, whence
P.X/  PŒR.B/ D R.B [ fBg/ and hence P.X/ 2 PŒR.B [ fBg/ D R.A [ fAg/
with A WD B [ fBg. Thus, P.X/ is well-formed.
Consequently, from P.X/ 2 R.A [ fAg/ it follows that P.X/  PŒR.A [ fAg/,
whence if S  X,Tthen S 2 PŒR.A [ fAg/ is T well-formed.
S S
In particular, X is well-formed because X  X and X is well-formed.
Also, if X 2 R.B/ is a well-formed set, then X  R.B/. Hence, S if Z 2 Y and
Y 2 X, then Y S  X whence Z 2 X and Z 2 R.B/. This shows that X  R.B/.
Consequently, X 2 PŒR.B/ D R.B [ fBg/ is well-formed. S
In particular, because fX; Yg is well-formed, it follows that X [ Y D fX; Yg is
well-formed, whence P.X [ Y/, PŒP.X [ Y/, and PfPŒP.X [ Y/g are also
well-formed. Therefore, X Y 2 PfPŒP.X [ Y/g is also well-formed. t
u
 www.pdfgrip.com

280 5 Well-Formed Sets: Proofs by Transfinite Induction with Already Well-Ordered Sets

5.38 Theorem. For all well-formed sets X and Y, :Œ.X 2 Y/ ^ .Y 2 X/. In

particular, X … X for each well-formed set X.
Proof. If X D ¿, then Y … X and X … X.
If X 6D ¿ and Y 2 X, then it suffices to verify that X … Y. To this end, let A be
the first ordinal such that X 2 R.A/; in particular, X  R.A/ by transitivity of R.A/.
If A D Z [ fZg, then X 2 R.A/ D R.Z [ fZg/ D PŒR.Z/, so X  R.Z/,
whence Y 2 X  R.Z/, and hence Y  R.Z/. However, X … R.Z/ because A is
S first ordinal with X 2 R.A/. If A does not contain a last element, then R.A/ D
the
B2A R.B/, whence X 2 R.A/ means that there exists B 2 A with X 2 R.B/, and
then X  R.B/ by transitivity of R.B/. Yet X … R.B/ because B 2 A and A is the
first ordinal with X 2 R.A/;
Thus, in either case there exists C 2 A such that X  R.C/ and X … R.C/ and
Y 2 R.C/. (In the first case C WD Z, while C WD B in the second case.)
Therefore X … Y, because X 2 Y 2 R.C/ and the transitivity of R.C/ would yield
X 2 R.C/. In particular, X … X, because the foregoing argument applied to Y WD X
shows that if X 2 X, then X … X, contrary to the axioms governing 2, which state
that for all sets X and Y either X 2 Y or X … Y but not both. t
u

5.4.2.1 Independence of the axiom of regularity

As an extension of theorem 5.38, every nonempty well-formed set X contains

an element Y that does not contain any element of X, so that Y \ X D ¿, or,
equivalently, so that there does not exist any Z 2 Y such that Z 2 X (exercise 5.31).
Outside of well-formed sets, however, there exist systems of set theory in which the
relation A 2 A may hold for some set. One way of preventing the relation A 2 A
from holding for any set consists in the axiom of regularity,
˚   
8X .X 6D ¿/ ) 9Y .Y 2 X/ ^ f8ZŒ.Z 2 Y/ ) .Z … X/g ;

attributed independently to Zermelo and von Neumann [128, p. 53]. The axiom of
regularity has the disadvantage of asserting a condition about sets already defined by
previous axioms. Yet within the theory of well-formed sets, exercise 5.31 confirms
that the axiom of regularity is not independent but is a theorem derivable from
the other axioms. Because well-formed sets suffice for most of logic, mathematics,
computer science, and their applications, the foundations of these fields can restrict
themselves to well-formed sets [74]. In contrast to the derivability of the axiom
of regularity from the other axioms of the theory of well-formed sets, neither the
generalized continuum hypothesis nor its negation are derivable from the other
axioms of the theory of well-formed sets [19, 20, 44, 45]. Thus, the “axiom of
regularity” is an example of an axiom that is “dependent” on the other axioms,
whereas the generalized continuum hypothesis is an example of an axiom that is
“independent” from the other axioms.
 www.pdfgrip.com

5.4 Regularity of Well-Formed Sets 281

5.4.3 Exercises on Well-Formed Sets

5.27 . Prove that the set N of all natural numbers is a well-formed set.
5.28 . Prove that every natural number N 2 N is a well-formed set.
5.29 . Prove that the set Z of all integers is a well-formed set.
5.30 . Prove that the set Q of all rational numbers is a well-formed set.
5.31 . Prove that for each well-formed set X there exists Y 2 X such that Y \X D ¿.
5.32 . Prove that if X is a well-formed set, then so is fXg.
5.33 . For each well-formed set X, prove that if A is the smallest ordinal such that
X 2 R.A/, then there exists an ordinal B such that A D B [ fBg.
5.34 . Prove that every finite set of well-formed sets is a well-formed set.
5.35 . Determine whether every countable set of well-formed sets is a well-formed
set.
5.36 . Determine whether every set of well-formed sets is a well-formed set.
 www.pdfgrip.com

Chapter 6
The Axiom of Choice: Proofs
by Transfinite Induction

6.1 Introduction

This chapter presents several statements, which are called “principles” because
they are well-formed formulae but not propositions, in the sense that neither of
them nor their negations are theorems, in the Zermelo-Frænkel set theory. The
first sections show how Zorn’s Maximal-Element Principle implies Zermelo’s Well-
Ordering Principle, which in turn implies the Choice Principle. Thus any extension
of the Zermelo-Frænkel set theory that includes Zorn’s Maximal-Element Principle
as an axiom also includes the other two principles as theorems. From the Choice
Principle, subsequent sections demonstrate the converse implications, known as
Zorn’s Lemma and Zermelo’s Theorem, so that all three principles are logically
equivalent within the Zermelo-Frænkel set theory. Hence all three principles are
theorems in the Zermelo-Frænkel-Choice set theory, which includes the Choice
Principle as the Axiom of Choice. The material also introduces yet other principles
that are logically equivalent to the Axiom of Choice, for example, the principle of
the distributivity of intersections over unions of families of sets. Any theory that
requires any such equivalent principle thus also requires the Axiom of Choice.
Other consequences of the Axiom of Choice include the existence of extrema for
continuous functions on closed and bounded sets in Euclidean spaces.

6.2 The Choice Principle

This section presents several mutually equivalent forms of the Choice Principle.

© Springer Science+Business Media New York 2015 283

Y. Nievergelt, Logic, Mathematics, and Computer Science,
DOI 10.1007/978-1-4939-3223-8_6
 www.pdfgrip.com

284 6 The Axiom of Choice: Proofs by Transfinite Induction

6.2.1 The Choice-Function Principle

One version of the Choice Principle relies on the concept of a “choice function”:
6.1 Definition (choice function).S For each set F of nonempty sets, a choice
function is a function C W F ! F such that C.A/ 2 A for every set A 2 F .
S S
S If F D ¿, then F D ¿ D ¿: there exists exactly one function
6.2 Example.
C W F ! F , namely ¿ W ¿ ! ¿, which is “vacuously” a choice function.
S S
6.3 Example. With 1 D f¿g, if F DSf1g, then F D f1g D 1 D f¿g: there
exists exactly one function C W F ! F , namely C W f1g ! f¿g with C.1/ D ¿,
which is a choice function because C.1/ 2 1.
6.4 Example. For eachSnonempty set A, if F D fAg, then there exists X 2 A,
whence .A; X/ 2 F F D fAg A, and C WD f.A; X/g is a choice function.
S
6.5 Example. With 1 D f¿g and 2 D f¿; 1g, if F D f1; 2g, then
S F D
f1; 2g D 1[2 D f¿; 1g D 2. There are two choice functions C W f1; 2g ! f¿; 1g:
The requirement that C.1/ 2 1 D f¿g imposes that C.1/ D ¿.
The requirement that C.2/ 2 2 D f¿; 1g allows for the two possibilities:
either C.2/ D ¿ 2 f¿; 1g, or C.2/ D 1 2 f¿;S1g.
Thus the two
S choice functions are F W F ! F with F.1/ D 0 and F.2/ D 0,
or G W F ! F with G.1/ D 0 and G.2/ D 1.
More generally, each finite set of nonempty sets has a choice function.
6.6 Theorem. Within the Zermelo-Frænkel set theory, each finite set of nonempty
sets has a choice function.
Proof. This proof proceeds by induction on the number N 2 N of sets.
For N D 0, example 6.2 shows that the empty set has a choice function.
Example 6.4 proves the theorem for N D 1.
As an induction hypothesis, assume that the theorem holds for some N D M 2 N
and every finite set F with exactly M elements, all nonempty. For each set G with
exactly M C 1 elements, all nonempty, there exists B 2 G , so that F WD G n fBg has
exactly M elements, all
S nonempty. By the induction hypothesis, there exists a choice
function F SW F ! F . Example 6.4 shows that there exists a choice function
G W fBgS! fBg D B.SBecause F \ fBg D ¿, the union C WD F [ G W F [ fBg D
G ! . F / [ B D G is a function. Also, C.A/ D F.A/ 2 A for each A 2 F
and C.B/ D G.B/ 2 B for each B 2 fBg, whence C is a choice function for G . u t
Still, theorem 6.6 applies only to finite sets.
6.7 Example. No choice functions are known for G D P ŒP.N/ n f¿g, which
is the set of all nonempty sets of subsets of the set N of natural numbers, which
corresponds to the set of all nonempty sets of real numbers between 0 and 1.
The statement of the existence of choice functions is one version of the Choice
Principle, called the Choice-Function Principle to distinguish it from other versions.
 www.pdfgrip.com

6.2 The Choice Principle 285

6.8 Definition (Choice-Function Principle). Each set of nonempty sets has a

choice function. Specifically, the Choice-Function Principle is formula (6.1):

8F f8A Œ.A 2 F / ) .A 6D ¿/g (6.1)

!
h n [   oi
) 9C C W F ! F ^ 8A f.A 2 F / ) ŒC.A/ 2 Ag :

Neither the Choice-Function Principle 6.8 nor its negation are propositions in the
Zermelo-Frænkel set theory: there it is merely formula (6.1).
The concept of a “family” of sets provides a convenient way to specify more than
one operation on a set, for instance, to choose more than one element from a set.
6.9 Definition (family of sets). A family of sets F D fAi W i 2 I g is a set (of sets)
with a function I W I ! F , i 7! Ai , from a set I , called the indexing set or set of
indices, to F .
6.10 Example (Self-Indexed Family of Sets). Each set (of sets) F is a family of
sets: the same set I WD F serves as an indexing set, and the identity function
 W F ! F , E 7! AE WD E shows that F D fE W E 2 F g D fAE W E 2 I g.
Still other versions of the Choice Principle rely on families of sets, as in
definitions 6.11 and 6.12.

S family fAi W i 2 I g of sets, a

6.11 Definition (family choice-function). For each
family choice-function is a function C W I ! i2I Ai such that C.i/ 2 Ai for
each index i 2 I .

6.12 Definition (Family Choice-Function Principle). Each family of nonempty

sets has a family choice-function.
The Family Choice-Function Principle 6.12 and its negation are formulae but not
propositions in the Zermelo-Frænkel set theory. However, theorem 6.13 shows that
choice functions are equivalent to family choice-functions.
6.13 Theorem. Within the Zermelo-Frænkel set theory, the Choice-Function Prin-
ciple 6.8 is logically equivalent to the Family Choice-Function Principle 6.12
Proof. If the Family Choice-Function Principle 6.12 holds and F is a set of
S F D fE S
nonempty sets, then the self-indexed family W E 2 F g from example 6.10
has a family choice-function C W F ! E2F E D F such that C.E/ 2 E for
each E 2 F . Thus C is a choice function for F .
Conversely, if the Choice-Function Principle 6.8 holds and F WD fAi W i 2 I g is
a nonempty family of nonempty sets,Sindexed by a function I W I ! F , then there
exists a choice function C W F ! S F such that C.E/ 2 E for each E 2 F . The
composite function C ı I W I ! i2I Ai , i 7! CŒI.i/ D CŒAi  2 Ai , is a family
choice-function. t
u
 www.pdfgrip.com

286 6 The Axiom of Choice: Proofs by Transfinite Induction

Another version of the Choice Principle relies on choice functions only for
pairwise disjoint sets, as specified by definition 6.14.
6.14 Definition (Choice-Function Principle for Pairwise Disjoint Sets). Each
set of pairwise disjoint nonempty sets has a choice function.
Yet another version of the Choice Principle relies on functions and relations, as
specified by definition 6.15.
6.15 Definition (Choice-Relation Principle). For each relation R there exists a
function F  R such that F and R have the same domain [128, p. 243, AC3 ].
In the Zermelo-Frænkel set theory, a relation is a subset of a Cartesian product:
R  A B. There is a different principle, called the Relational Axiom of Choice,
where R may be a relation defined for all sets [88, p. 22].

6.2.2 The Choice-Set Principle

This subsection provides different statements of the Choice Principle that are
logically mutually equivalent within the Zermelo-Frænkel set theory. One version
of the Choice Principle relies on “choice sets” rather than choice functions.
Definition (choice set). For each set F of nonempty sets, a choice set is a set
6.16 S
S  F such that for each A 2 F , S \ A contains exactly one element.
As with the Choice-Function Principle 6.8, neither the Choice-Set Principle nor
its negation are propositions in the Zermelo-Frænkel set theory: it is merely a
formula, stated in words in definition 6.17.
6.17 Definition (Choice-Set Principle). Each set of nonempty sets has a
choice set.
S S
Example. If F D ¿, then F D ¿ D ¿: there exists exactly one subset
6.18 S
S  F D ¿, namely S D ¿, which is “vacuously” a choice set, because S \ A
“vacuously” contains exactly one element for each A 2 F D ¿.
S S
6.19 Example. With 1 D f¿g, if F D f1g, then S F D f1g D 1 D f¿g,
whence there exists exactly one choice set S  F D 1, namely S D 1. Indeed,
the only element in F D f1g is 1, and S \ 1 DS1 \ 1 D f¿g, which contains exactly
one element, namely ¿ 2 1. In contrast ¿  F D 1 is not a choice set.
Theorem 6.20 shows that choice sets are equivalent to choice functions.
6.20 Theorem. Within the Zermelo-Frænkel set theory, the Choice-Function Prin-
ciple 6.8 is logically equivalent to the Choice-Set Principle 6.17.
 www.pdfgrip.com

6.2 The Choice Principle 287

S
Proof. Each choice set S  F corresponds to the choice function FS defined by
n [ o
FS WD .A; X/ 2 F F W X 2A\S :
S
Conversely, each choice function is a subset of a Cartesian product: F  F F.
Thus F consists of pairs .A; X/ with exactly one X 2 A for each A 2 F . Hence
n [ o
SF WD X 2 F W 9Af.A 2 F / ^ Œ.A; X/ 2 Fg

is a choice set, obtained from the second projection (example 3.87) of F. t

u
Still another version of the Choice Principle relies on choice sets only for
pairwise disjoint sets, as specified by definition 6.21.
6.21 Definition (“Pairwise Disjoint” Choice-Set Principle). Each set of pairwise
disjoint nonempty sets has a choice set.
Theorem 6.22 shows that choice sets for any sets, and choice sets for pairwise
disjoints sets, are mutually equivalent in the Zermelo-Frænkel set theory.
6.22 Theorem. Within the Zermelo-Frænkel set theory, the Choice-Set Prin-
ciple 6.17 is logically equivalent to the “Pairwise Disjoint” Choice-Set
Principle 6.21.
Proof. The “Pairwise Disjoint” Choice-Set Principle 6.21 is a particular case of the
Choice-Set Principle 6.17. Thus the latter implies the former.
To establish the converse implication, to each set E of nonempty sets corresponds
the set F of pairwise disjoint nonempty sets defined by
[ 
A0 WD f.A; X/ W .X 2 A/g  E E ;
h [ i
F WD fA0 W A 2 E g  P E E :

The elements of F are pairwise disjoint, because if A; B 2 E and A 6D B, then

.A; X/ 6D .B; Y/ regardless of X 2 A and Y 2 B, whence A0 \ B0 D ¿. Also,
A0 6D ¿ for each A0 2 F , because A contains at least one element X by hypothesis,
so that A0 contains at least one element .A; X/. If the “Pairwise
S Disjoint” Choice-Set
S
Principle 6.21. holds, then there exists a choice set S  F  E . E / such
that S \ A0 is a singleton for each A0 2 F : there exists a unique X 2
SA for which
.A; X/ 2 A0 \ S. The second projection of S yields a choice set T  E such that
T \ A D fXg is a singleton for each A 2 F . t
u
The statements of the Choice-Set Principle 6.17 and the “Pairwise Disjoint”
Choice-Set Principle 6.21 with a proof of their mutual equivalence in the Zermelo-
Frænkel set theory similar to theorem 6.22 are in Zermelo’s [145].
 www.pdfgrip.com

288 6 The Axiom of Choice: Proofs by Transfinite Induction

6.23 Definition (Axiom of Choice). The Axiom of Choice is any of the principles
6.8, 6.14, 6.15, 6.17, 6.21, included as an axiom in a set theory.

6.2.3 Exercises on Choice Principles

6.1 . Find the number of choice functions for each finite set F D fA0 ; : : : ; AN1 g
of nonempty finite sets where each element Aj has exactly Nj elements.
6.2 . Find the number of choice sets for each finite set F D fA0 ; : : : ; AN1 g of
nonempty finite sets where each element Aj has exactly Nj elements.
6.3 . Prove that each finite set of nonempty finite sets has a choice set.
6.4 . Prove that the Choice-Function Principle 6.8 is logically equivalent to the
“Pairwise Disjoint” Choice-Function Principle 6.14. within the Zermelo-Frænkel
set theory.
6.5 . Translate the Choice-Set Principle 6.17 into a logical formula, similar to
formula (6.1) for the Choice-Function Principle 6.8.
6.6 . Translate the “Pairwise Disjoint” Choice-Set Principle 6.21. into a logical
formula, similar to formula (6.1) for the Choice-Function Principle 6.8.
6.7 . Prove that the Choice-Relation Principle 6.15 is logically equivalent to the
Choice-Function Principle 6.8 in the Zermelo-Frænkel set theory.
6.8 . Translate the “Pairwise Disjoint” Choice-Function Principle 6.14 into a
logical formula, similar to formula (6.1) for the Choice-Function Principle 6.8.

6.3 Maximality and Well-Ordering Principles

This section introduces two principles and shows that each of them implies the
Choice Principle.

6.3.1 Zermelo’s Well-Ordering Principle

As the Choice-Function Principle 6.8, neither Zermelo’s Well-Ordering Principle

nor its negation are propositions in the Zermelo-Frænkel set theory: it is merely a
formula, stated in words in definition 6.24.
6.24 Definition (Zermelo’s Well-Ordering Principle). Each set has a well-order.
[88, p. 117].
 www.pdfgrip.com

6.3 Maximality and Well-Ordering Principles 289

6.25 Example. The order is a well-order on the set N of all natural numbers.
6.26 Example. Every finite set has a well-order. Indeed, by definition of “finite” for
each finite set E there exists a natural number N 2 N and a bijection F W E ! N.
The subset N  N is well-ordered by example 6.25 and theorem 3.197. Hence the
relation  defined on E by X  Y if and only if F.X/ F.Y/ well-orders E.
Theorem 6.27 shows that the existence of a choice function on a set follows from
the existence of a well-ordering on its union.
S
6.27 Theorem. For each set F of S nonempty sets with a well-ordered union F ,
there exists a function C W F ! F such that C.A/ 2 A for every set A 2 F .
S
Proof. Let C.A/ be the first element of A relative to the well-order on F . t
u
6.28 Example. If F D P.N/ nSf¿g, which is the set of all nonempty subsets of the
set N of natural numbers, then F D N, where is a well-order. By the proof of
theorem 6.27, the function C W F ! N with C.A/ D min.A/ chooses the smallest
element of A and thus is a choice function for F .
6.29 Example. No well-orders are known for R D P.N/, which is the set of all
subsets of the set N of natural numbers, which
Sis also isomorphic to the set of all real
numbers from 0 through 1. Because R D G with G D P ŒP.N/ n f¿g from
example 6.7, any specific well-order on R would yield a specific choice function on
G , by theorem 6.27.
Theorem 6.30 shows a logical relation between the foregoing principles.
6.30 Theorem. Zermelo’s Well-Ordering Principle 6.24 implies the Choice-
Function Principle 6.8.
Proof. Theorem
S 6.27 shows that if every set admits a well-ordering relation, in
particular, F , then every set F of nonempty sets has a choice function. t
u
As sets (of pairs), relations are partially ordered by inclusion. Theorem 6.31
shows that every total order is maximal relative to inclusion among partial orders.
6.31 Theorem. Each reflexive total order on a set is maximal among all partial
orders on that set.
Proof. If R is a reflexive total order on a set A, and if T is a relation on A such
that R ¨ T, then there are X; Y 2 A such that .X; Y/ 2 T but .X; Y/ … R, whence
.Y; X/ 2 R by totality of R. Hence X 6D Y by reflexivity of R. Consequently, .Y; X/ 2
R  T and .X; Y/ 2 T with X 6D Y. Thus T is not anti-symmetric and therefore not
a partial order. Therefore, R is not properly contained in any partial order on A. ut
The concept of maximality in a partially ordered set leads to Zorn’s Maximal-
Element Principle.
 www.pdfgrip.com

290 6 The Axiom of Choice: Proofs by Transfinite Induction

6.3.2 Zorn’s Maximal-Element Principle

As the Choice-Function Principle 6.8 and Zermelo’s Well-Ordering Principle 6.24,

neither Zorn’s Maximal-Element Principle nor its negation are propositions in
the Zermelo-Frænkel set theory: it is merely a formula, stated in words in
definition 6.32.
6.32 Definition (Zorn’s Maximal-Element Principle). In each partially ordered
set where each chain has an upper bound, there is a maximal element [88, p. 117],
[128, p. 248, Z2 ].
Theorem 6.33 shows another logical relation between the foregoing principles.
6.33 Theorem. Zorn’s Maximal-Element Principle 6.32 implies Zermelo’s Well-
Ordering Principle 6.24.
Proof. This proof follows Dugundji’s [30, p. 34, .2/ ) .3/]. For each set X, let
X be the set of all pairs .A; R/ where R weakly well-orders a subset A  X. Thus
X  ŒP.X/ ŒP.X X/ is well-formed. Also, X 6D ¿, because .¿; ¿/ 2 X .
Define a relation  on X by .A; R/  .B; S/ if and only if A  B and R  S
with .X; Y/ 2 S for all X 2 A and Y 2 B n A. Then  partially orders X :
Reflexivity. .A; R/  .A; R/ because A  A and R  R with A n A D ¿.
Antisymmetry. If .A; R/  .B; S/ with .B; S/  .A; R/, then A  B and R  S
with B  A and S  R, whence A D B and R D S, so that .A; R/ D .B; S/.
Transitivity. If .A; R/  .B; S/ and .B; S/  .C; T/, then A  B  C and R 
S  T. If also X 2 A and Z 2 C n A, then two cases can arise: either Z 2 B n A,
or Z 2 C n B.
In the first case, if Z 2 B n A, then .X; Z/ 2 S  T because X 2 A.
In the second case, if Z 2 C n B, then .X; Z/ 2 T because X 2 B.
Thus .A; R/  .C; T/.
For each chain Y  X linearly ordered by  in X define
[
C WD A;
.A;R/2Y
[
T WD R:
.A;R/2Y

Then T weakly well-orders C:

Reflexivity. For each X 2 C there exists .A; R/ 2 Y such that X 2 A, whence
.X; X/ 2 R  T by reflexivity of R.
Antisymmetry. For all X; Y 2 C there exist .A; R/; .B; S/ 2 Y such that X 2 A
and Y 2 B. Because Y is a chain, two cases can arise: either .A; R/  .B; S/,
or .B; S/  .A; R/. In the first case, if .A; R/  .B; S/, then X; Y 2 B; if also
 www.pdfgrip.com

6.3 Maximality and Well-Ordering Principles 291

.X; Y/ 2 T and .Y; X/ 2 T, then .X; Y/ 2 S and .Y; X/ 2 S by maximality of S

on B from theorem 6.31, whence X D Y by anti-symmetry of S. The second case
is similar.
Transitivity. For all X; Y; Z 2 C there exist .A; R/; .B; S/; .D; U/ 2 Y such that
X 2 A, Y 2 B, Z 2 D. Because Y is a chain, assume A  B  D and R  S  U.
(The other five orders are similar.) If also .X; Y/; .Y; Z/ 2 T, then there exist
.E; V/; .F; W/ 2 Y such that .X; Y/ 2 V and .Y; Z/ 2 W. Again because Y is
a chain, assume V  W, whence .X; Y/; .Y; Z/ 2 W, and hence .X; Y/; .Y; Z/ 2
U by maximality of U on D from theorem 6.31, whence .X; Z/ 2 U  T by
transitivity of U.
Well-ordering. S Each nonempty subset E  C contains at least one element Z 2
E  C D .A;R/2Y A. Hence there exists .A; R/ 2 Y such that Z 2 A.
Thus A \ E 6D ¿, whence A \ E has a least element X 2 A \ E, because R
well-orders A.
Hence X is also the least element of E. Indeed, if Y 2 E, then there exists .B; S/ 2
Y such that Y 2 B. Either Y 2 A, or Y 2 B n A.
In the first case, if Y 2 A, then .X; Y/ 2 R  T, because X is the least element of
A \ E.
In the second case if Y 2 B n A, then .X; Y/ 2 S by definition of  and hence
.X; Y/ 2 S  T.
In either case .X; Y/ 2 T. Therefore X is the least element of E.
Thus .C; T/ 2 X , and .C; T/ is an upper bound relative to  for Y .
Hence, if Zorn’s Maximal-Element Principle holds, then X contains a maximal
element .D; U/. In particular, D D X: by contraposition, if D 6D X, then there would
exist K 2 X n D. Setting E WD D [ fKg and V WD U [ f.H; K/ W H 2 Dg would
define a strict upper bound such that .D; U/  .E; V/, contradicting the maximality
of .D; U/. Therefore, U well-orders D D X. t
u
Definition 6.34 states another version of Zorn’s Maximal Principle.
6.34 Definition (Zorn’s Maximal-Set Principle). In each nonempty set (of sets)
that contains the union of each chain of elements (which are sets) relative to
inclusion, there is a maximal element relative to inclusion: an element that is not
a proper subset of any other element. [128, p. 245, Z1 ].
Max Zorn stated the Maximal-Set Principle 6.34 for sets of sets partially ordered
by inclusion in reference [147, p. 667, (MP)], where the axiom of the power-set is
also in question [147, p. 669, (MP)], but did not prove it there from any other axioms.
Indeed, Max Zorn also stated that he would show its relation and equivalence with
the Axiom of Choice in “another paper” [147, p. 669, (MP)]. In other words, “Zorn’s
Lemma” is not in [147].
The Axiom of Choice is logically independent from the Zermelo-Frænkel axioms
of set theory [21, 66, Ch. 5]. Consequently, appending Zorn’s Maximal-Element
Principle to the Zermelo-Frænkel axioms of set theory yields a larger “Zermelo-
Frænkel-Choice” set theory that also includes Zermelo’s Well-Ordering Principle
and the Axiom of Choice.
 www.pdfgrip.com

292 6 The Axiom of Choice: Proofs by Transfinite Induction

6.3.3 Exercises on Maximality and Well-Orderings

6.9 . Prove that the hypothesis of Zorn’s Maximal-Element Principle 6.32 implies
that the partially ordered set is not empty.
6.10 . Prove that the conclusion of Zorn’s Maximal-Set Principle 6.34 requires the
hypothesis that the set be not empty.
6.11 . Translate Zorn’s Maximal-Set Principle 6.34 into a logical formula, similar
to formula (6.1) for the Choice-Function Principle 6.8.
6.12 . Translate Zermelo’s Well-Ordering Principle 6.24 into a logical formula,
similar to formula (6.1) for the Choice-Function Principle 6.8.
6.13 . For each partial order  on a set E, let C  P.E/ be the set of all chains
relative to  in E. Thus each element of C is a subset of E on which  is a linear
order. Partially order C by inclusion. Prove that the union of each chain in C relative
to inclusion is an element of C .
6.14 . Prove that Zorn’s Maximal-Set Principle 6.34 is logically equivalent to
Zorn’s Maximal-Element Principle 6.32 within the Zermelo-Frænkel set theory.

6.4 Unions, Intersections, and Products of Families of Sets

This section shows that the Choice Principle is equivalent to the distributivity of
intersections over unions of sets.

6.4.1 The Multiplicative Principle

Yet other versions of the Choice Principle rely on Cartesian products, as in

definitions 6.35 and 6.36.
Q
6.35 Definition (Cartesian Product). The Cartesian productS i2I Ai of a fam-
ily of sets fAi W i 2 I g is the set of all functions C W I ! i2I Ai , such that
C.i/ 2 Ai for each i 2 I

6.36 Definition (Multiplicative Principle). For each

Q nonempty family of
nonempty sets fAi W i 2 I g, the Cartesian product i2I Ai is not empty [88,
p. 117].
The Multiplicative Principle 6.36 and its negation are formulae but not propo-
sitions in the Zermelo-Frænkel set theory. However, theorem 6.40 shows that the
Multiplicative Principle 6.36 is equivalent to the foregoing choice principles.
 www.pdfgrip.com

6.4 Unions, Intersections, and Products of Families of Sets 293

6.37 Theorem. Within the Zermelo-Frænkel set theory, the Family Choice-
Function Principle 6.12 is logically equivalent to the Multiplicative Principle 6.36.
Proof. Definitions 6.11 and 6.35 show that the Cartesian product consists of all the
family choice-functions as its elements. Thus the Cartesian product is nonempty if
and only if there exists a family choice-function. t
u

6.4.2 The Distributive Principle

This subsection shows that the Choice Principle is equivalent to the distributivity
of intersections over unions of families of sets. The following development follows
Dugundji’s [30, Ch. I, § 9.7–9.8, p. 25].
6.38 Definition (Distributive Principle). For each family of sets fAi W i 2 I g and
each partition fI` W ` Q2 L g of the indexing set I , which sets up equivalence
classes in I , let K WD `2L I` ; then
! 0 1
[ \ \ [
Ak.`/ D @ Ai A : (6.2)
k2K `2L `2L i2I`

6.39 Example. For the self-indexed family F WD fA; B; C; Dg and the self-indexed
partition G WD ffA; Bg; fC; Dgg, the right-hand side of equation (6.2) becomes
[
E D A [ B;
E2fA;Bg
[
E D C [ D;
E2fC;Dg
0 1
\ [
@ Ai A D .A [ B/ \ .C [ D/:
`2L i2I`

The Cartesian product of the partitioning equivalence classes is K WD fA; Bg

fC; Dg D f.A; C/; .A; D/; .B; C/; .B; D/g, and the left-hand side of equation (6.2)
becomes
!
[ \
Ak.`/ D .A \ C/ [ .A \ D/ [ .B \ C/ [ .B \ D/:
k2K `2L

The Distributive Principle 6.38 and its negation are not propositions in the
Zermelo-Frænkel set theory; there they are mere formulae. Nevertheless, theo-
rem 6.37 shows that one of the inclusions, but only one, implied by equation (6.2)
is a theorem in the Zermelo-Frænkel set theory.
 www.pdfgrip.com

294 6 The Axiom of Choice: Proofs by Transfinite Induction

6.40 Theorem. Within the Zermelo-Frænkel set theory, for each family of sets
fAi W i 2 I g and each partition fI` W ` Q 2 L g of the indexing set I , which
sets up equivalence classes in I , let K WD `2L I` ; then
! 0 1
[ \ \ [
Ak.`/  @ Ai A : (6.3)
k2K `2L `2L i2I`

Proof. This proof unravels the definitions of unions, intersections, and Cartesian
products: T 
S
X 2 k2K `2L Ak.`/ S
m definition of
  T 
9k .k 2 K / ^ X 2 `2L Ak.`/
T
m definition of
 ˚   
9k .k 2 K / ^ 8` .` 2 L / ) X 2 Ak.`/
.k 2 K / )
+
.8`f.` 2 L / ) Œk.`/ 2 I` g/
8` Œ.` 2 L / ) f9i Œ.i 2 I` / ^ .X 2 Ai /g
S
m h  i definition of
S
8` .` 2 L / ) X 2 i2I` Ai
T
m S  definition of
T
X2 `2L i2I` Ai
t
u
Theorem 6.41 shows that in the Zermelo-Frænkel set theory, the Distributive
Principle 6.38 is equivalent to the foregoing choice principles.
6.41 Theorem. Within the Zermelo-Frænkel set theory, the Multiplicative
Principle 6.36 is logically equivalent to the Distributive Principle 6.38.
Proof. Reversing the last two equivalences in the proof of theorem 6.40 gives
2 0 13
\ [
4X 2 @ Ai A5 , f8` Œ.` 2 L / ) f9i Œ.i 2 I` / ^ .X 2 Ai /gg ;
`2L i2I`

which implies that for every ` 2 L ,

fi 2 I` W X 2 Ai g 6D ¿:

If the Multiplicative Principle

Q 6.36 holds, then the Cartesian product contains a fam-
ily choice function k 2 `2L fi 2 I` W X 2 Ai g, so that k.`/ 2 fi 2 I` W X 2 Ai g,
 www.pdfgrip.com

6.5 Equivalence of the Choice, Zorn’s, and Zermelo’s Principles 295

whence X 2 Ak.`/ , for each ` 2 L , which is precisely the converse of the

implication in the proof of theorem 6.40. Thus equation (6.2) holds, and so does
the Distributivity of Intersection over Union Principle 6.38.
Q Conversely, for each nonempty
S family of nonempty sets fI` W ` 2 L g, let K WD
`2L I ` , and for each i 2 `2L I` define Ai WD f¿g or any other singleton. Thus
for every ` 2 L ,
[
Ai D f¿g;
i2I`
0 1
\ [
@ Ai A D f¿g:
`2L i2I`

If the Distributivity of Intersection over Union Principle 6.38 holds, then equa-
tion (6.2) gives
!
[ \
Ak.`/ D f¿g;
k2K `2L

Q
which is not empty, whence neither is K D `2L I` , so that the Multiplicative
Principle 6.36 holds. t
u

6.4.3 Exercises on the Distributive and Multiplicative

Principles

6.15 . Translate the Multiplicative Principle 6.36 into a logical formula, similar to
formula (6.1) for the Choice-Function Principle 6.8.
6.16 . Translate the Distributive Principle 6.38 into a logical formula, similar to
formula (6.1) for the Choice-Function Principle 6.8.

6.5 Equivalence of the Choice, Zorn’s, and Zermelo’s

Principles

Based on Dugundji’s proof [30, Ch. II, § 2], this section proves that the Choice-
Function Principle 6.8 implies Zorn’s Maximal-Element Principle 6.32 by means of
the concept of a “tower” of sets relative to a function.
 www.pdfgrip.com

296 6 The Axiom of Choice: Proofs by Transfinite Induction

6.5.1 Towers of Sets

The subsection introduces the concept of “towers” of sets relative to a function.

6.42 Definition (Tower). [30, Ch. II, p. 32, def. 2.2] A nonempty set T  P.X/
of subsets of a set X is a tower relative to a function F W T ! X if and only if
(T.A) ¿ 2 T , S
(T.B) if A  T is linearly ordered by inclusion, then A 2 T , and
(T.C) if A 2 T , then A [ fF.A/g 2 T .
A sub-tower is a nonempty subset S  T that is also a tower relative to the
restriction FjS W S ! X.
An element M 2 S is medial in a sub-tower S if and only if A  M or M  A
for every element A 2 S .
A sub-tower is minimal if and only if it does not properly contain any sub-tower.
6.43 Example. The empty set ¿ is medial in every tower T : indeed, ¿ 2 T by the
definition 6.42 of towers, and ¿  A for every A 2 T . u
t
6.44STheorem. For each nonempty set T F of towers relative to a common function
FW F ! X, the intersection M WD F is also a tower relative to FjM .
Tset F of all sub-towers of a tower T contains a unique smallest
In particular, the
sub-tower M WD F .
T
Proof. This proof verifies that M D F satisfies the definition 6.42 of towers.
T
(T.A) From ¿ 2 T for every T 2 F follows ¿ 2 F D M .
(T.B) If A  M , then
S A  T for every T 2 F ; if A isSalso linearlyT ordered
by inclusion, then A 2 T for every T 2 F , whence A 2 F D M .
(T.C) If A 2 M , then A 2 T for every TT 2 F , whence A [ fF.A/g 2 T for
every T 2 F , and hence A [ fF.A/g 2 F D M .
In particular,
T S set F of all sub-towers of a tower T contains the sub-tower
the
M D FT  F D T , because T 2 F and. S  T for every S 2 F .
Also, M D F  S for every S 2 F , so that M is the smallest sub-tower. u t
6.45 Theorem. If M 2 M is medial in the minimal sub-tower M of a tower T ,
then either A  M or M [ fF.M/g  A for every element A 2 M .
Proof. This proof verifies that the set S WD fA 2 M W .A  M/ _ .M [ fF.M/g 
A/g is also a sub-tower of the smallest sub-tower M .
(T.A) ¿ 2 S because ¿  M regardless of M. Thus S 6D ¿.
(T.B) If R is a chain in S , then for each A 2 R  S , either M [ fF.M/g  A
or A  M. Hence two cases can arise:
S exists any A 2 R such that M [ fF.M/g  A, then
In the first case, if there
M [ fF.M/g  A  R.
In the second case, if there does not exist any A 2 R such that M S
[fF.M/g  A,
then A  M for every A 2 R  S by definition of S , whence R  M.
 www.pdfgrip.com

6.5 Equivalence of the Choice, Zorn’s, and Zermelo’s Principles 297

S
In either case R 2 S .
(T.C) For each A 2 S , either M [ fF.M/g  A or A  M, by definition of S .
In the first case, if M [ fF.M/g  A, then M [ fF.M/g  A  A [ fF.A/g.
In the second case, if A  M and A 2 S  M , then A [ fF.A/g 2 M because
M is a tower, whence either M [ fF.M/g  A [ fF.A/g or A [ fF.A/g  M
because M is medial in M . Thus A [ fF.A/g 2 S .
Thus S is a sub-tower of the smallest sub-tower M . Hence S D M , so that
A  M or M [ fF.M/g  A for each medial element M 2 M and each A 2 M . u
t
6.46 Theorem. Every element of the smallest sub-tower of any tower is medial in
the smallest sub-tower. The smallest sub-tower is linearly ordered by inclusion.
Proof. This proof consists of verifying that the set V of all medial element of M is
a sub-tower of M .
(T.A) ¿ 2 V because ¿ is medial in M by example 6.43.
(T.B) If U  V is a chain in V and B 2 U  V , then B is medial in M , so
that B [ fF.B/g  A or A  B for each A 2 M , by theorem 6.45.
there exists any A 2 U such that B [ fF.B/g  A, then
In the first case, if S
B [ fF.B/g  A  U .
In the second case, if there does not exist any A 2 U such that B
S[ fF.B/g  A,
then A  B forSevery A 2 U  V , by definition of V whence U  B.
In either case U 2 V .
(T.C) Every A 2 V is medial in M by definition of V . By theorem 6.45 for
every B 2 M either B  A, in which case B  A  A[fF.A/g, or A[fF.A/g 
B. In either case A [ fF.A/g is medial in M , so that A [ fF.A/g 2 V .
Thus V is a sub-tower of the minimal sub-tower M . Hence V D M , so that every
element of M is medial in M . Hence M is linearly ordered by inclusion. t
u
6.47 Theorem. For every tower T  P.X/ relative to a function F W T ! X
there exists A 2 T such that F.A/ 2 A.
S
Proof. Let M be the smallest sub-tower, and let A WD M . Then A 2 M by
definition 6.42 of a tower and because M  M and M is linearly ordered, by
theorem 6.46. Hence
S also A [ fF.A/g 2 M by definition 6.42 of a tower, whence
A [ fF.A/g  M D A, and hence F.A/ 2 A. t
u

6.5.2 Zorn’s Maximality from the Choice Principle

This subsection shows that Zorn’s Maximality Principle follows from the Choice
Principle.
6.48 Theorem. The Choice-Function Principle 6.8 implies Zorn’s Maximal-
Element Principle 6.32.
 www.pdfgrip.com

298 6 The Axiom of Choice: Proofs by Transfinite Induction

Proof. For each set X with a reflexive partial order  let C be the set of all chains
relative to  in X.
For each nonempty chain C 2 C , let UC be the set of all upper bounds for C
relative to  in X. The hypotheses of Zorn’s Maximal-Element Principle 6.32 imply
that UC 6D ¿.
If the Choice-Function Principle 6.8 holds, then there exists a function
F W fUC W C 2 C g ! X such that F.UC / 2 UC , which chooses one upper bound
F.UC / for each nonempty chain C 2 C . Hence define DC WD fK 2 X W ŒF.UC / 
K ^ f:ŒK  F.UC /gg, which is the set of all the elements of X that strictly follow
the upper bound F.UC /. If there exists C 2 C such that DC D ¿, then F.UC / is a
maximal element of X.
Alternatively, if DC 6D ¿ for every C 2 C , then, again by the Axiom of
Choice 6.8, there is a function G W fDC W C 2 C g ! X such that G.DC / 2 DC
for each C 2 C . The hypotheses of Zorn’s Maximal-Element Principle 6.32 also
imply X 6D ¿, so that there exists Z 2 X. Define H.¿/ WD Z and H.C/ WD G.DC /.
The remainder of the proof verifies that T WD C [ f¿g is a tower relative to H.
(T.A) ¿ 2 T by the definition T WD C [ f¿g. S
(T.B) If S is a chain in T relative
S to inclusion, then S is a chain relative to
 in X. Indeed, for all K; L 2 S there exist A; B 2 S such that K 2 A and
L 2 B. Also, S is linearly ordered by inclusion, so that A  B or B  A. Hence
K; L 2 B or K; L 2 A. Moreover, A; B 2 S  T are chains relative to  in X,
S A and B are linearly ordered, and hence K  L or L  K, in B or A.
whence
Thus S 2 T .
(T.C) If C 2 T , then H.C/ is an upper bound for C relative to  in X, whence
C [ fH.C/g is also a chain relative to  in X. Hence C [ fH.C/g 2 T .
By theorem 6.47 there is a chain C 2 T such that H.C/ 2 C. Hence H.C/ 
F.UC / because F.UC / is an upper bound for C. Yet H.C/ 2 DC is a strict upper
bound with :ŒH.C/  F.UC /. This contradiction completes the proof: there is a
chain C 2 C for which DC D ¿, so that F.DC / is maximal relative to  in X. t
u
Zermelo’s Theorem [144] consists of the logical implication that the Well-
Ordering Principle follows from the Axiom of Choice within the Zermelo-Frænkel-
Choice set theory.
Zorn’s Lemma consists of the logical implication that Zorn’s Maximal-Element
Principle follows from the Axiom of Choice within the Zermelo-Frænkel-Choice
set theory.
6.49 Definition (Interval). In a set E pre-ordered by a relation  a subset S  E
is an interval if and only if for all U; W 2 S and every V 2 E, if U  V  W, then
V 2 S.
6.50 Example. The empty subset ¿ and the whole set E are intervals.
 www.pdfgrip.com

6.6 Yet Other Principles Related to the Axiom of Choice 299

6.5.3 Exercises on Towers of Sets

6.17 . Define F W P.N/ ! N by F.¿/ WD 0 and F.A/ WD min.A/ for A 6D ¿.

Prove that the set T of all intervals of N is a tower relative to F.
6.18 . For each set E well-ordered by a reflexive relation  define F W P.E/ ! E
by F.¿/ WD min.E/ and F.A/ WD min.A/ for A 6D ¿. Prove that the set T of all
intervals of E is a tower relative to F.
6.19 . Prove that P.E/ is a tower relative to any function P.E/ ! E.
6.20 . Prove or disprove that every union of towers is a tower.

6.6 Yet Other Principles Related to the Axiom of Choice

This section states principles that are related to the Axiom of Choice, in particular,
in the sense that they are not propositions within the Zermelo-Frænkel set theory.

6.6.1 Yet Other Principles Equivalent to the Axiom of Choice

This subsection states principles that are equivalent to the Choice Principle in the
Zermelo-Frænkel set theory.
6.51 Definition (Hausdorff’s Particular Maximal-Chain Principle). In a set of
sets partially ordered by inclusion, each chain is a subset of a maximal chain [128,
p. 248, H1 ].

6.52 Definition (Hausdorff’s Maximal-Chain Principle). Each set of sets has a

maximal chain relative to inclusion [88, p. 118], [128, p. 248, H2 ].

6.53 Definition (Hausdorff’s Maximal-Subset Principle). In each partially

ordered set there is a maximal linearly ordered subset [56, 57, p. 339], [91, p. 69].
6.54 Definition (Zermelo’s Principle). For each partition F of a set (of sets) A,
there exists a subset C  A such that C \ B is a singleton for each B 2 F [88,
p. 117].

6.55 Definition (Counting Principle). For each set E there is an ordinal O and a
bijective function F W O ! E with domain O [88, p. 117], [128, p. 241].
 www.pdfgrip.com

300 6 The Axiom of Choice: Proofs by Transfinite Induction

6.56 Definition (Kuratowski’s Maximal-Order Principle). For each partial

order R and for each linear order L  R there exists a maximal (relative to inclusion)
linear order M such that L  M  R [88, p. 118].

6.57 Definition (Trichonomy Principle). For all sets A and B, there exists an
injection F with F W A ,! B with domain A, or F W B ,! A with domain B [88,
p. 118], [128, p. 241].

6.58 Definition (Mapping Principle). For all nonempty sets A and B, there exists
a surjection F with F W A  B, or F W B  A [88, p. 118].

6.59 Definition. A set (of sets) A is of finite character if and only if

(FC.A) A 6D ¿,
(FC.B) if X 2 A and B  X is finite, then B 2 A, and
(FC.C) for each set E, if every finite subset of E is a member of A, then E 2 A.

6.60 Definition (Teichmüller-Tukey Maximal-Element Principle). Every set of

finite character has a maximal element [128, p. 249].

6.6.2 Consequences of the Axiom of Choice

This subsection states principles that follow from the Choice Principle in the
Zermelo-Frænkel set theory.
6.61 Definition (Finite Sets and Infinite Sets). A set E is finite if and only if there
is N 2 N and a bijection F W N ! E; it is infinite if and only if it is not finite.
6.62 Theorem. In the Zermelo-Frænkel-Choice set theory, every infinite set con-
tains a denumerable subset.
Proof. For each set E, let E be the set of all injections from any subset of N into E.
Thus each element F 2 E is a subset of N E , and E  P.N E/ is partially
ordered by inclusion. If E is infinite, then for each N 2 N there exists an injection
F W N ,! E, which is also a subset of N E.S In particular E 6D ¿.
S Each chain F  E defines an injection F 2 E . Indeed, for each .X; Y/ 2
F , there exists F 2 F such that .X; Y/ 2 F. If also .X; Z/ 2 G 2 F , then F  G
S G  F, because F is a chain, whence Y D Z. Moreover, if .U; Y/; .X; Y/ 2
or
F , then there exist F; G 2 F such that .X; Y/ 2 F and S .U; Y/ 2 G, with F  G
or G  SF, whence U D X by injectivity of F or G. Thus F is an upper bound for
F and F 2 E . By Zorn’s Maximal-Element Principle 6.32 the set E contains a
maximal element F.
If the domain of F was finite, then a reparametrization of its domain would give
an injection G W N ,! N, which would not be injective, by definition of infinite.
Hence the domain of F is an infinite subset of N.
 www.pdfgrip.com

6.6 Yet Other Principles Related to the Axiom of Choice 301

Therefore, the image of F is a denumerable subset of E. t

u
Without the Axiom of Choice, theorem 6.62 fails [66, p. 141].
6.63 Definition (Dedekind Infinite Sets). A set D is Dedekind infinite if and only
if there exist a proper subset B ¨ D and an injection F W D ,! B. A set is Dedekind
finite if there are no injections from it into any of its proper subsets.
6.64 Theorem. In the Zermelo-Frænkel set theory, every Dedekind infinite set is
infinite.
Proof. By contraposition, if a set is finite, then there are no injections from it into
any of its proper subsets, whence it is Dedekind finite. t
u
6.65 Theorem. In the Zermelo-Frænkel-Choice set theory, every infinite set is
Dedekind infinite.
Proof. If a set E is infinite, then by theorem 6.62 it contains a denumerable subset
D  E. In particular, there is a bijection C W N ! D. Define B WD E n fC.0/g, and
F W E ! B with F.X/ WD X for every X 2 E n D and FŒC.N/ WD FŒC.N C 1/ if
X D C.N/ 2 D. Then F is an injection from E to B. t
u

6.66 Definition (Countable Axiom of Choice). Each countable set of nonempty

sets has a choice function.
Proofs of the equivalence of compactness and sequential compactness, or
continuity and sequential continuity — and hence of the existence of extrema of
continuous functions on closed and bounded intervals — invoke the Countable
Axiom of Choice [66, p. 21, § 2.4.3] but do not require (the results do not logically
imply) the Axiom of Choice [111, p. 178].

6.6.3 Exercises on Related Principles

6.21 . Prove that in the Zermelo-Frænkel-Choice set theory every denumerable

union of pairwise disjoint denumerable sets is denumerable.
6.22 . Following the R. L. Moore method [15, 146], prove all logical implications
between all “principles” stated in the present chapter.
 www.pdfgrip.com

Chapter 7
Applications: Nobel-Prize Winning Applications
of Sets, Functions, and Relations

7.1 Introduction

This chapter shows concrete applications of sets, functions, and relations:

1. Arrow’s Impossibility Theorem. Kenneth J. Arrow received the Nobel Prize in
Economic Science in 1972, mainly for his Impossibility Theorem, from work at
the RAND Corporation in 1948 [4, p. 328, footnote 1].
2. Gale and Shapley’s Matching Algorithm. Gale and Shapley’s Ph.D. advisor was
Princeton’s Albert William Tucker; Lloyd S. Shapley received the Nobel Prize in
Economic Science in 2012, for work that can be traced back to a lecture by John
von Neumann in 1948 at the RAND Corporation [68, p. 384].
3. Nash’s Equilibrium. Nash’s Ph.D. advisor was Princeton’s Albert William
Tucker; John Forbes Nash, Jr., received the Nobel Prize in Economic Sciences
in 1994, for work that can be traced to Melvin Dresher and Merrill Meeks Flood
in 1950 at the RAND Corporation [12, 124, 125]. He received the Abel Prize in
2015.
David Gale and Lloyd S. Shapley pointed out that readers without a technical
background may miss the logic in their Nobel-Prize winning work even though it is
written in plain English:
“Yet any mathematician will recognize the argument as mathematical, while people without
mathematical training will probably find difficulty in following the argument” — [41,
p. 391].

This chapter shows what mathematicians recognize in the argument: links between
applications and mathematical sets, functions, and relations. Expositions considered
concise and elegant by mathematicians are listed in the references.

© Springer Science+Business Media New York 2015 303

Y. Nievergelt, Logic, Mathematics, and Computer Science,
DOI 10.1007/978-1-4939-3223-8_7
 www.pdfgrip.com

304 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

7.2 Game Theory

The mathematical analysis of games, called game theory, explains why and
predicts how people will rationally make decisions against their interest. Examples
include decisions whether to arm one’s self. For their contributions to game theory,
John Harsanyi, Reinhard Selten, and John Forbes Nash, Jr., received the Nobel
Prize in Economics in 1994 [93–95]. This introduction to game theory relies on
mathematical functions and ordering relations.

7.2.1 Introduction

Publicized by Albert W. Tucker as A Two-Person Dilemma [131], but now known

as The Prisoner’s Dilemma [132], a prototype of the subject of game theory
originated through the work of Melvin Dresher and Merrill Meeks Flood in 1950
at the RAND Corporation [12, 124, 125]. Similar games explain the nuclear arm
race (see Figure 7.1) [13, 126].

Fig. 7.1 Early days of the nuclear arm race. Left: a Soviet modified SS-6 (Sapwood) with
Sputnik 2 on 3 November 1957, photograph courtesy NASA. Right: an Atlas rocket lifts off,
photograph courtesy U.S. Air Force. (http://www.history.nasa.gov/SP-4408pt1.pdf) (http://www.
nationalmuseum.af.mil/shared/media/photodb/photos/050406-F-1234P-014.jpg)
 www.pdfgrip.com

7.2 Game Theory 305

7.1 Example (The Prisoner’s Dilemma [131, 132]). The police charge Al and Bo
with a crime and hold them in separate cells, so that Al and Bo cannot communicate
with each other. Al and Bo believe that they have at least a first strategy:
(PD.1) If Al and Bo do not confess to the crime, then they both go free.
However, the police gives each of them another strategy, to confess:
(PD.2) If both Al and Bo confess to the crime, then each gets sentenced to prison.
(PD.3) If one confesses but the other does not confess, then whoever confesses
gets a reward and goes frees, while whoever does not confess gets a death
sentence.
Al and Bo’s dilemma is whether to confess or not to confess.
The Prisoner’s Dilemma also applies to two countries, or their rulers, or any two
individuals, Al and Bo, deciding whether to cooperate not to arm themselves, or to
defect the agreement of cooperation and arm themselves. The Prisoner’s Dilemma
predicts that they will both arm themselves, as shown in figure 7.1, and explains
why [13, 126], as demonstrated in subsection 7.2.2. Game theory also explains
animal behavior, for example musth in male African elephants [87]. To get a sense
of Nash’s work, the reader may at this stage try to design a method to analyze
mathematical games such as The Prisoner’s Dilemma but with any numbers of
players and strategies.

7.2.2 Mathematical Models for The Prisoner’s Dilemma

This subsection describes ways to design a mathematical model and analyze games
such as The Prisoner’s Dilemma.
7.2 Example (The Prisoner’s Dilemma, continued [131, 132]). What Al and Bo
eventually get (reward, freedom, prison, or death) is called their payoff. The first
step in designing a mathematical model of the game consists in modeling the
players’ preferences with ordering relations on each player’s set of payoffs. In this
example, it seems reasonable to assume that Al and Bo rank all four payoffs in the
same order of preference, with > meaning “better than”:

reward & freedom > freedom > prison > death.

The second step in designing a mathematical model of the game consists in

modeling every way the players may play by organizing their actions, called
strategies, and corresponding payoffs in a table. Table 7.3 summarizes Al’s and
Bo’s strategies and their resulting payoffs in The Prisoner’s Dilemma. Al has only
two strategies available: either to confess or not to confess. Similarly, Bo may either
confess or not confess. Moreover, The Prisoner’s Dilemma is a noncooperative
game in the sense that neither player knows the other player’s action in advance.
 www.pdfgrip.com

306 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

The analysis of the game then examines every combination of strategies in the
table. For instance, Al may examine each of Al’s own strategies and seek to avoid
the worst payoff.
Al confesses. If Al confesses, then meanwhile two cases can occur: Bo can either
confess or not confess.
Bo confesses. If Bo also confesses, then Al faces a prison term,
Bo does not confess If Bo does not confess, then Al goes free with a reward.
Thus, if Al confesses, then the worst that can happen to Al is a prison term.
Al does not confess. Similarly, if Al does not confess, then two cases can occur:
Bo can either confess or not confess.
Bo confesses. If Bo confesses, then Al faces death.
Bo does not confess If Bo does not confess either, then Al goes free.
Thus, if Al does not confess, then the worst that can happen to Al is death.
Consequently, to avoid death with certainty but without knowing Bo’s action, Al
must confess.
Table 7.3 is symmetric in the sense that swapping Al’s and Bo’s rôles does not
change their payoffs. Thus Al’s reasoning also applies to Bo, who must also confess
to avoid death with certainty.
Therefore, to avoid death with certainty, Al and Bo both confess, even though
they would both be better off by not confessing.
7.4 Example (Arm race [13, 126]). If two individuals or countries cooperate not to
arm themselves, then they remain free and safe. If both defect and arm themselves,
then they remain safe but incur a penalty with the cost of armament. If either
one defects and arms itself while the other one does not arm itself, then whoever
arms itself stays free and gets a reward by conquering and looting the other, who
suffers from death, destruction, and loosing the property to the conqueror. The
ordering of the outcomes just described are the same as those in table 7.3 for
The Prisoner’s Dilemma. Therefore, the analysis of The Prisoner’s Dilemma in
example 7.2 predicts and explains why they will both arm themselves.
Example 7.5 shows another way to analyze The Prisoner’s Dilemma.

Table 7.3 Payoffs for the BO’S STRATEGIES

Prisoner’s Dilemma.
CONFESS NOT CONFESS
PAYOFFS TO PAYOFFS TO
AL BO AL BO
CONFESS prison prison reward death
AL’S STRATEGIES
NOT CONFESS death reward freedom freedom
 www.pdfgrip.com

7.2 Game Theory 307

7.5 Example (The Prisoner’s Dilemma, continued [131, 132]). In contrast to

example 7.2, Al may focus first on Bo’s strategies and seek to get the best payoff.
Again, Bo may either confess or not confess.
Bo confesses. If Bo confesses, then Al has two choices.
Al confesses. If Al confesses, too, then Al gets a prison term.
Al does not confess. If Al does not confess, then Al gets death.
Thus, if Bo confesses, then Al gets the better of the two payoffs available (prison
or death) by confessing.
Bo does not confess. If Bo does not confess, then Al has two choices.
Al confesses. If Al confesses, then Al goes free with a reward.
Al does not confess. If Al does not confess, then Al goes free.
Thus, if Bo does not confess, then Al also gets the better payoff by confessing.
Thus, regardless of Bo’s action, Al gets the better payoff available by confessing.
Therefore, Al confesses. The same analysis applies to Bo, who also confesses.

7.2.3 Dominant Strategies

Some of the modeling and analysis of The Prisoner’s Dilemma carry over to other
games. To this end, this subsection shows that a special type of strategy contains an
equilibrium, from which no players has any incentive to switch to another strategy.
For a game with exactly two players, Al and Bo, assume that Al has a nonempty set
of strategies SAl and a nonempty set of payoffs PAl , while Bo has a nonempty set of
strategies SBo and a nonempty set of payoffs PBo . Assume also that Al and Bo have a
linear order “” on their sets of payoffs. If Al plays a strategy A 2 SAl and Bo plays
a strategy B 2 SBo , then Al gets a payoff pAlA ;B
while Bo gets a payoff pBo
A ;B
. Thus for
each Ci 2 fAl; Bog, p is a function p W SAl SBo ! PCi . Table 7.6 shows the payoff
Ci Ci

table of a game for two players with two strategies.

7.7 Definition (dominant strategy). A strategy is weakly dominant for a
player (Al, for instance) if and only if, for each fixed combination of the other
players’ strategies, that player (Al) cannot get a higher payoff by switching to

Table 7.6 Payoffs for BO’S STRATEGIES

noncooperative two-person
L (“LEFT”) R (“RIGHT”)
games with two strategies.
PAYOFFS TO PAYOFFS TO
AL BO AL BO
T (“TOP”) pAl
T;L pBo
T;L pAl
T;R pBo
T;R
AL’S STRATEGIES
B (“BOTTOM”) pAl
B;L pBo
B;L pAl
B;R pBo
B;R
 www.pdfgrip.com

308 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

another strategy. A strategy is strongly dominant for a player (Al) if and only if,
for each fixed combination of the other players’ strategies, that player (Al) always
gets a lower payoff by switching to another strategy.
7.8 Example. In The Prisoner’s Dilemma (examples 7.1 and 7.5), Al’s strongly
dominant strategy is to confess: If Bo confesses, then Al gets a higher payoff by
confessing (prison) than by not confessing (death). Similarly, if Bo does not confess,
then Al also gets a higher payoff by confessing (freedom and a reward) than by
not confessing (freedom). By symmetry of The Prisoner’s Dilemma, Bo’s strongly
dominant strategy is also to confess.
7.9 Definition (dominant strategy equilibrium). A position is an array of strate-
gies, with one strategy for each player. A weakly (respectively strongly) dominant
strategy equilibrium is a position where each player plays a weakly (respectively
strongly) dominant strategy.
7.10 Example. In The Prisoner’s Dilemma, the position where Al and Bo both
confess is a dominant strategy equilibrium, where their dominant strategies meet.
Neither Al nor Bo has any incentive to switch to another strategy (not to confess),
because whoever decides not to confess while the other still confesses faces death.
7.11 Definition (Nash equilibrium). A Nash equilibrium is a position from
which no players can get a higher payoff by changing only their own strategy, while
all the other players’ strategies remain fixed [93, p. 287].
7.12 Example. In The Prisoner’s Dilemma, the position where Al and Bo both
confess is a Nash equilibrium: If Al decides not to confess but Bo still confesses,
then Al’s payoff drops. Similarly, if Bo decides not to confess but Al still confesses,
then Bo’s payoff drops. In contrast, the position where neither Al nor Bo confess is
not a Nash equilibrium, because whoever decides to confess gets a higher payoff if
the other one still does not confess.
7.13 Theorem. Every weakly dominant equilibrium is a Nash equilibrium.
Proof. Table 7.14 shows the conditions for either of Bo’s strategic to be weakly
dominant in a game with only one other player (Al).

Table 7.14 Bo’s potentially weakly dominant strategies.

BO’S STRATEGIES
L IS WEAKLY R IS WEAKLY
DOMINANT: OR DOMINANT:
PAYOFFS TO PAYOFFS TO
BO BO BO BO
T pBo
T;L pBo
T;R pBo
T;L pBo
T;R
AL’S STRATEGIES AND OR AND
B pBo
B;L pBo
B;R pBo
B;L pBo
B;R
 www.pdfgrip.com

7.2 Game Theory 309

If R is Bo’s single weakly dominant strategy (Bo’s R column in table 7.14), then
Bo plays R. Indeed, if Al plays T , then R gets Bo a larger payoff, because pBo T ;L
pBo
T;R
;
Bo Bo
if Al plays B , then R also gets Bo a larger payoff, because pB;L pB;R .
Since Al knows the game, Al knows that R is Bo’s single weakly dominant
strategy. Thus, Al knows that Bo will play R. Consequently, Al has only two payoffs
available: pAl
T;R
or pAl
B;R
. Therefore, Al chooses a strategy that yields the larger available
payoff. For instance, if pAlT ;R
pAl
B;R
, then Al plays B.
Hence the game ends at (B,R) with payoffs pAl B ;R
to Al and pBo B ;R
to Bo. From (B,R)
Bo cannot get a higher payoff by switching to L because pBo B ;L
pBo B ;R
. From (B,R) Al
cannot get a higher payoff by switching to T because pAl T;R
pAlB ;R
. Thus (B,R) is a
Nash equilibrium.
If R and L are both weakly dominant strategies for Bo, then Bo’s payoffs are
identical in R and L. If pAl B ;R
is the largest of Al’s four payoffs in Bo’s weakly
dominant strategies, then (B,R) is still a Nash equilibrium. Yet Al no longer knows
which of R or L Bo chooses: the game might not end at a Nash equilibrium. t
u

7.2.4 Mixed Strategies

Some games or situations might not occur more than once in the players’ life time,
for instance, The Prisoner’s Dilemma. Yet other games or situations may occur more
than once in the players’ life time. Each player may then adopt a mixed strategy,
defined by choosing one strategy some of the time and another strategy at other
times. Each player’s payoff may then be a weighted average of the payoffs from
the different strategies in the mixed strategy. For emphasis, strategies that are not
mixed, for instance, as in The Prisoner’s Dilemma, are called pure strategies.
7.15 Example (The Battle of the Sexes [92]). Al and Bo would like to go to a
show together. Al prefers a Jazz concert while Bo prefers a play at the theater.
Nevertheless, they would prefer to go to the same show together, rather than to
different shows separately. After some negotiation they agree on a show. On the
day of the show, however, they each forget which show they had agreed on, but
cannot communicate with each other. So they each must decide to which show to
go, without knowing in advance where the other will go. Table 7.16 shows their
payoffs. In hope of finding Bo, Al goes to the Jazz concert one half of the time, and

Table 7.16 Payoffs for the BO GOES TO

Battle of the Sexes.
JAZZ (J) PLAY (P)
PAYOFFS TO PAYOFFS TO
AL BO AL BO
JAZZ (J) 3 2 1 1
AL GOES TO
PLAY (P) 0 0 2 3
 www.pdfgrip.com

310 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

Table 7.17 Payoffs for the Battle of the Sexes over four days.
BO GOES TO
JAZZ (J) PLAY (P)
PAYOFFS TO PAYOFFS TO
AL BO AL BO
JAZZ (J) 3 (DAY 1) 2 1 (DAY 3) 1
AL GOES TO
PLAY (P) 0 (DAY 2) 0 2 (DAY 4) 3

to the play at the theater the other half of the time. In hope of finding Al, Bo does
the same. Table 7.17 shows the four combinations of strategies; for this example,
assume that they occur equally frequently. Thus Al’s and Bo’s average payoffs over
four days are

3C0C1C2 3
pAl .1=2; 1=2/ D D I
4 2
2C0C1C3 3
pBo .1=2; 1=2/ D D :
4 2
Al now decides to go to the Jazz concert every day, while Bo still goes to Jazz every
other day and to the play every other day. There are then only two kinds of days,
and their average payoffs become

3C1
pAl .1; 0/ DD 2I
2
2C1 3
pBo .1=2; 1=2/ D D :
2 2
Hence arises the question whether other mixed strategies can increase both Al’s
and Bo’s payoffs.

7.2.5 Existence of Nash Equilibria for Two Players

with Two Mixed Strategies

This section shows that if neither player has any weakly dominant pure strategy, then
the game still has a Nash equilibrium for some combination of mixed strategies.
7.18 Definition (mixed strategy). The same players might play multiple rounds
of the same game and independently choose any available strategy for each round.
Over M rounds, Al may play T exactly T times and B exactly B times, with T C
B D M. Thus Al plays T with a frequency t D T=M and B with a frequency
b D B=M D 1 t. The ordered pair of frequencies .t; 1 t/ is called a mixed
 www.pdfgrip.com

7.2 Game Theory 311

strategy for Al. Meanwhile, Bo may play L exactly L times and R exactly R times,
with LCR D M. Thus Bo plays L with a frequency ` D L=M and R with a frequency
r D R=M D 1 `. Thus .`; 1 `/ is a mixed strategy for Bo. Mixed strategies are
also subject to the condition that Al and Bo play (T,L) with frequency t ` and hence
also receive payoffs pAl
T ;L
and pBo
T ;L
with frequency t `. Similar products of frequencies
apply to the other plays (T,R), (B,L), and (B,R). The sum of such payoffs to Al and
Bo are

pAl .t; `/ D t ` pAl

T;L
C t .1 `/ pAl
T ;R
C .1 t/ ` pAl
B ;L
C .1 t/ .1 `/ pAl
B;R
;
pBo .t; `/ D t ` pBo
T;L
C t .1 `/ pBo
T ;R
C .1 t/ ` pBo
B ;L
C .1 t/ .1 `/ pBo
B;R
:

7.19 Theorem. Every game for two players with two pure strategies has a Nash
equilibrium.
Proof. For Bo’s payoff, collecting similar powers of t and ` leads after some algebra
to the equivalent formula

pBo .t; `/ D ` ft Œ.pBo

T;L T;R / C .pB;R
pBo Bo
B;L /
pBo .pBo
B;R B;L /g C t .pT;R
pBo Bo
B;R / C pB;R :
pBo Bo

Al has a problem: if Al can choose a frequency t such that

t Œ.pBo
T ;L
pBo
T ;R
/ C .pBo
B;R
pBo
B ;L
/ .pBo
B;R
pBo
B ;L
/ D 0;

then Bo’s payoff becomes

pBo .t ; `/ D t .pBo

T ;R
pBo
B ;R
/ C pBo
B ;R
;

which does not depend on ` and thus strips away from Bo all controls over Bo’s
own payoff. In particular, Bo cannot get a higher payoff by switching to a different
frequency `. Yet if Bo can choose a frequency ` so that Al cannot control Al’s own
payoff, then they are at a Nash equilibrium .t ; ` /. The following considerations
show that such a Nash equilibrium exists provided that neither Al nor Bo has any
weakly dominant strategy. Table 7.20 shows the conditions for Bo not to have any
weakly dominant strategy, obtained from the logical negation of Table 7.14. Two
cases emerge: either pBo B ;L
< pBo
B ;R
and pBo
T ;L
> pBo
T ;R
, or pBo
T ;L
< pBo
T ;R
and pBo
B ;L
> pBo
B ;R
.
In the first case, pB;L < pB;R and pT;L > pT;R . Hence pB;R
Bo Bo Bo Bo Bo
pBo
B;L
> 0 and
pBo
T ;L
pBo
T;R
> 0. Consequently,

.pBo
T;L
pBo
T ;R
/ C .pBo
B ;R
pBo
B ;L
/ > .pBo
T;L
pBo
T ;R
/ > 0:

Thus, Al can choose a frequency t between 0 and 1 where Bo has no controls over
Bo’s payoff:
 www.pdfgrip.com

312 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

Table 7.20 Conditions for Bo to have no weakly dominant pure strategies.

BO’S STRATEGIES
L IS NOT WEAKLY R IS NOT WEAKLY
DOMINANT: AND DOMINANT:
PAYOFFS TO PAYOFFS TO
BO BO BO BO
T pBo
T;L < pBo
T;R pBo
T;L > pBo
T;R
AL’S STRATEGIES OR AND OR
B pBo
B;L < pBo
B;R pBo
B;L > pBo
B;R

.pBo
B;R
pBo
B ;L
/
1 > t D > 0:
.pBo
T;L
pT;R / C .pB;R
Bo Bo
pBo
B ;L
/

Similarly, if Al has no weakly dominant strategies, then Bo can choose a frequency

` between 0 and 1 where Al has no control over Al’s payoff, because

.pAl
B;R
pAl
T ;R
/
1 > ` D > 0:
.pAl
B;L
pT;L / C .pB;R
Al Al
pAl
T ;R
/

With such frequencies,

pAl .t ; ` / D ` pAl

B;L
C .1 ` / pAl
B ;R
;
pBo .t ; ` / D t pBo
T ;R
C .1 t / pBo
B ;R
:

Neither Al nor Bo can change their own payoff by switching to another frequency
while the other player’s frequency is fixed. Thus .t ; ` / is a Nash equilibrium. The
second case, where pBo
T;L
< pBo
T ;R
and pBo
B ;L
> pBo
B ;R
, is entirely similar.
If a player has a dominant strategy, then such a strategy contains a Nash
equilibrium, by theorem 7.19. t
u
For Nash’s equilibria with two players and two strategies, see also
[106, p. 138–139], [107], [119, p. 155–168].
7.21 Example (The Professors’ Problem). Al and Bo are on the tenured faculty
at King Game’s College. Each may either teach students or sit on committees.
Teaching does not hamper any one’s work, but committees hamper other faculty
members’ work, to the extent summarized by Al’s and Bo’s end-of-the-year bonus
payoffs in table 7.22. The sum of Al’s and Bo’s payoffs reflects the value of their
work to the College. Al and Bo know table 7.22 but work in different offices and
thus do not cooperate with each other; hence they must choose a strategy without
knowing in advance what the other is doing.
Bo has exactly one weakly dominant strategy, which is to sit on committees;
consequently, Bo declines to teach and decides to sit. Al knows table 7.22 and thus
 www.pdfgrip.com

7.2 Game Theory 313

Table 7.22 Payoffs for The BO’S STRATEGIES

Professors’ Problem.
TEACH SIT
PAYOFFS TO PAYOFFS TO
AL BO AL BO
TEACH 6 9 2 9
AL’S STRATEGIES
SIT 5 1 3 2

Table 7.23 Payoffs for The BO’S STRATEGIES

Administration’s Response.
TEACH SIT
PAYOFFS TO PAYOFFS TO
AL BO AL BO
TEACH 4 9 4 5
AL’S STRATEGIES
SIT 5 1 3 2

knows that Bo will sit on committees; therefore, to get the higher payoff available
to Al under Bo’s decision to sit, Al also declines to teach and decides to sit on
committees.
Not only do Al and Bo choose the Nash equilibrium where they are both worse
off than in the other Nash equilibrium, but the sum of the values of their work to the
College is the worst of all possibilities.
The administration’s challenge is to entice the faculty to teach by modifying the
payoff table.
Table 7.23 shows the new game on campus after the administration capped
payoffs from committees to 5 units. If Bo sits on committees, then Al may also
sit for a payoff of 4 rather than teach for 6: Al now values teaching only up to 4.
To get a deeper sense of one of Nash’s many contributions, the reader may
attempt proving the existence of a Nash equilibrium in games for any number of
players with any number of strategies.
7.24 Example (Blue against Red [26]). In Melvin Dresher’s initial context, the
Blue and Red commanders led opposing armed forces [26, p. 4], but they might also
lead sports teams [12]. Table 7.25 shows only the payoffs to the Blue commander.

7.2.6 Exercises on Mathematical Games

7.1 . Identify all the weakly dominant pure strategies in The Battle of the Sexes
(table 7.16 in example 7.15).
 www.pdfgrip.com

314 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

Table 7.25 Payoffs to Blue for a noncooperative two-commander game with

three strategies (adapted from [26, p. 4]).
RED’S STRATEGIES
L (“LEFT”) C (“CENTER”) R (“RIGHT”)
T (“TOP”) Failure Success Success
BLUE’S
M (“MIDDLE”) Draw Success Draw
STRATEGIES
B (“BOTTOM”) Success Draw Failure

7.2 . Identify all weakly dominant pure strategies in the Administration’s Solution
to the Professors’ Problem (table 7.23 in example 7.21).
7.3 . Identify all Nash equilibria with pure strategies in The Battle of the Sexes
(table 7.16 in example 7.15).
7.4 . Identify all Nash equilibria with mixed strategies in the Administration’s
Solution to the Professors’ Problem (table 7.23 in example 7.21).
7.5 . Prove or disprove that every Nash equilibrium is a dominant strategy
equilibrium.
7.6 . Prove or disprove that every two-player game restricted to pure strategies has
a Nash equilibrium.
7.7 . For each function f W A B ! C such that C is linearly ordered and for each
nonempty subset E  A B the image f 00 .E/ has a first element minŒf 00 .E/ and a
last element maxŒf 00 .E/, prove that

maxfminff .x; y/ W y 2 Bg W x 2 Ag minfmaxff .x; y/ W x 2 Ag W y 2 Bg:

7.8 . Design a function f W A B ! C such that C is linearly ordered and for each
nonempty subset E  A B the image f 00 .E/ has a first element minŒf 00 .E/ and a
last element maxŒf 00 .E/, with a strict inequality

maxfminff .x; y/ W y 2 Bg W x 2 Ag < minfmaxff .x; y/ W x 2 Ag W y 2 Bg:

Denote by “Al’s strategies” and “Bo’s strategies” the set of all strategies available
to Al and Bo respectively.
7.9 . Prove that if each player plays so as to avoid the worst payoff available, then
they get the payoffs

payoff to Al D maxfminfpAl
SAl ;SBo W SBo 2 Bo’s strategiesg W SAl 2 Al’s strategiesg;

payoff to Bo D maxfminfpBo
SAl ;SBo W SAl 2 Al’s strategiesg W SBo 2 Bo’s strategiesg:
 www.pdfgrip.com

7.3 Match Making 315

7.10 . Assume that in a game for two players with two strategies pAl SAl ;SBo  pSAl
Al
0 0
;SBo
if and only if pSAl ;SBo pS0 ;S0 for all positions .SAl ; SBo / and .SAl0 ; SBo
Bo Bo 0
/. Prove that if
Al Bo
each player plays so as to avoid the worst payoff available, then they get the payoffs

payoff to Al D minfmaxfpAl
SAl ;SBo W SBo 2 Bo’s strategiesg W SAl 2 Al’s strategiesg;

payoff to Bo D minfmaxfpBo
SAl ;SBo W SAl 2 Al’s strategiesg W SBo 2 Bo’s strategiesg:

7.11 . For noncooperative games with any number of players and any number of
pure strategies, prove or disprove that if at least one player has at least one weakly
dominant strategy, then such a strategy contains a Nash equilibrium.
7.12 . For noncooperative games with two players but any number of pure strate-
gies, prove or disprove that if at least one player has at least one weakly dominant
strategy, then such a strategy contains a Nash equilibrium.
7.13 . Analyze the game of Blue against Red defined by table 7.25 in example 7.24.
7.14 . Prove or disprove that the formulae from exercise 7.10 still hold for more
than two strategies.

7.3 Match Making

Match making pairs up items or individuals from two groups. For example, the
National Resident Matching Program (NRMP) uses an algorithm developed by
David Gale, Alvin E. Roth, and Lloyd S. Shapley to match medical doctors with
internships in hospitals [68, 96]. For their work on such problems, Alvin E. Roth
and Lloyd S. Shapley received the Nobel Prize in Economics in 2012. The precise
statements and proofs of their algorithms involve the mathematical concepts of sets,
functions, relations, and induction.

7.3.1 Introduction

The problems considered here have been documented for millennia since Plato’s
time (Figure 7.2): how to arrange for proper marriages [97, p. 27], and how to admit
students to schools [127, p. 44, footnote 11]. The problems consist in matching in
some “optimal” way individuals from two groups, for example, boys and girls for
marriage, students and schools for education, doctors and hospitals for internships,
or, more generally, beggars and choosers (table 7.26). “Optimality” may mean, for
instance, that there are no beggar from one couple and chooser from another couple
that prefer each other to their current mate. A similar optimality applies to schools
 www.pdfgrip.com

316 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

Fig. 7.2 Heracles and

Athena, the goddess of
wisdom, 480–470 B.C. (about
the time of the Battles of
Thermopylae and Salamis
between Greek and Persian
forces), by an olive tree
presumably at the then future
site of Plato’s Academy;
photograph courtesy
“User:Bibi Saint-Pol” via
Wikipedia. (http://en.
wikipedia.org/wiki/File:
Athena_Herakles_Staatliche_
Antikensammlungen_2648.
jpg)

Table 7.26 Applications of BEGGARS B CHOOSERS C RELATION M

Gale and Shapley’s
algorithms [68]. Boy Girl Marriage
Student School Admission
Doctor Hospital Residency
Recipient Donor Transplant

admitting several students. To get a sense of Gale and Shapley’s work, the reader
may at this stage try to design such optimal match-making procedures.

7.3.2 A Mathematical Model for Optimal Match Making

The first step in producing a successful match-making method consists in designing

a mathematical model of the situation. With marriages, for instance, the group of
boys may be modeled by a set B and the girls by a set C. To allow for applications
more general than marriages and to shorten the prose, call the elements of B beggars
and those of C choosers. Moreover, in the contexts considered here, the two sets B
and C must be disjoint: B \ C D ¿. The goal consists in marrying each boy exactly
one girl and each girl to exactly one boy. In general, the goal consists in establishing
a bijection between the two sets B and C, or, yet more generally, a relation M  B
C. Yet not every relation leads to successful marriages, because each chooser prefers
some beggars over others while each beggar prefers some choosers over others. Thus
a mathematical model of the situation must also include such preferences.
To specify preferences for certain choosers over others, each beggar b 2 B ranks
all choosers by a strict well-order b on C. Thus relative to b each nonempty
subset W  C has a unique first-ranked element, denoted by firstb .W/.
 www.pdfgrip.com

7.3 Match Making 317

To specify preferences for certain beggars over others, each chooser c 2 C ranks
all beggars by a strict well-order c on B. Thus relative to c each nonempty subset
V  B has a unique first-ranked element, denoted by firstc .V/.
The two disjoint sets B and C, with B well-ordered by each chooser and C well-
ordered by each beggar, complete the mathematical model of the situation. The
second step consists in designing a model of a successful relation.
For each relation M  B C, denote its domain by Domain.M/ and its range by
Range.M/. Also, call “single” each beggar b0 2 B0M WD B n Domain.M/ and each
0
chooser c0 2 CM WD C n Range.M/: neither b0 nor c0 are related to anyone by M.
Definition 7.27 specifies the notion of a successful relation by a concept of stability.
7.27 Definition. A relation M  B C is unstable if and only if at least one of the
following conditions holds:
(US.1) A beggar and a chooser prefer each other to their current mates: there
are different couples .b1 ; c1 /; .b2 ; c2 / 2 M for whom b2 c1 b1 and c1 b2 c2 , a
condition denoted by .b1 ; c1 / ‰ .b2 ; c2 /.
(US.2) There is a couple .b; c/ 2 M and a “single” beggar b0 2 B0M D
B n Domain.M/ for whom b0 c b and c b0 c0 for every “single” chooser
0
c0 2 CM D C n Range.M/, a condition denoted by .b; c/ ‰ b0 .
0
(US.3) There is a couple .b; c/ 2 M and a “single” chooser c0 2 CM D Cn
Range.M/ for whom c b c and b c0 b for every “single” beggar b 2 B0M D
0 0 0

B n Domain.M/, a condition denoted by c0 ‰ .b; c/.

A relation M  B C is stable if and only if it is not unstable, so that none of the
preceding three conditions holds.
7.28 Definition. A stable relation M  B C is total if and only if Domain.M/ D B
and Range.M/ D C.
7.29 Example. The empty relation ¿  B C is vacuously stable. It is total if and
only if B D ¿ D C.
Specifications of the situation and goal do not yet suffice to reach the goal. To this
end, the next step in mathematical modeling consists in developing an algorithm,
method, or procedure to reach the goal from the current situation. One algorithm
uses a match maker, another algorithm is carried out by the participants without a
match maker.

7.3.3 An Algorithm for Optimal Match Making

with a Match Maker

This subsection defines and demonstrates an algorithm for a match maker (person or
machine) to find a stable relation. First, a match maker with access to all well-orders
from all beggars and choosers can extend any stable relation.
 www.pdfgrip.com

318 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

7.30 Theorem. For each stable relation M  B C with Domain.M/  B and

O with M  M
Range.M/  C there exists a stable proper extension M O  B C.

Proof. By the hypothesis that Domain.M/  B, there exists a “single” beggar b0 2

B0M D B n Domain.M/ 6D ¿. Since the well-order b0 restricts to a well-order on
0
CM D C n Range.M/ 6D ¿, the “single” beggar b0 has a first-ranked single chooser
0 0 0
c D firstb0 .CM /. In particular, the set C1 of those single choosers in CM that are
0 0
ranked first among CM by some single beggar in BM is not empty:
0
C1 WD fc1 2 CM W 9b0 .b0 2 B0M / ^ Œc1 D firstb0 .CM
0
/g 6D ¿:

Also, for each first-ranked single chooser c1 2 C1 , the set Bc1  B0M of single
0
beggars who rank c1 first in CM ,

Bc1 WD fb0 2 B0M W c1 D firstb0 .CM

0
/g;

is not empty by definition and hence has a unique first element bc1 WD firstc1 .Bc1 /.
Moreover, if c1 6D c2 2 C1 , then bc1 6D bc2 , because bc1 ranks c1 first, ahead of c2 .
Thus the function C ,! B0M , c1 ! bc1 is injective. Define

M1 WD f.bc1 ; c1 / 2 B0M 0
CM W .c1 2 C1 / ^ Œbc1 D firstc1 .Bc1 /gI
O WD M [ M1
M M:

First, the relation M1 is stable in B0M CM 0

:
0 0 0
If .bc1 ; c1 /; .bc1 ; c1 / 2 M1 , then c1 D firstbc1 .CM /, whence c1 bc c0 for every
1
0 0 0 0 0
c 2 CM , so that .bc1 ; c1 / ‰ .bc1 ; c1 / and c ‰ .bc1 ; c1 / do not occur. Moreover,
bc1 D firstc1 .Bc1 /, so that if b0 2 B0M and b0 c1 bc1 , then b0 … Bc1 , which means that
0
there exists c0 2 CM with c0 b0 c1 , so that .bc1 ; c1 / ‰ b0 does not occur either.
Second, M O is stable in B C: if .b; c/ 2 M and .bc1 ; c1 / 2 M1 , then either
b c bc1 , in which case .b; c/ ‰ .bc1 ; c1 / fails, or bc1 c b but then by the assumed
0
stability of M in B C there exists a “single” chooser c0 2 CM for whom c0 bc c, yet
1
0
c1 D firstbc1 .CM /, so that either c1 D c0 or c1 bc c0 bc c. In either case .b; c/ ‰
1 1
.bc1 ; c1 / fails.
Similarly, if c b c1 , then .bc1 ; c1 / ‰ .b; c/ fails; also, if c1 b c, then by the
0
assumed stability of M there exists some “single” beggar b0 2 CM such that b0 c b,
0 0
whence either b D bc1 or bc1 c b c b, and then .bc1 ; c1 / ‰ b fails. t
u
Second, if the sets B of beggars and C of choosers have the same finite number of
elements, then the algorithm that starts from the stable empty relation and iterates
theorem 7.30 yields a total stable relation after finitely many iterations.
7.31 Algorithm (Match Maker’s Algorithm).
Initially, for all disjoint sets B and C, set M WD ¿  B C.
O by
While Domain.M/ 6D B and Range.M/ 6D C, find a proper extension M
Theorem 7.30, and re-set M WD M. O
 www.pdfgrip.com

7.3 Match Making 319

Table 7.33 Gale and BEGGARS CHOOSERS C

Shapley’s unlabeled example,
B c1 c2 c3 c4
adapted from [41, p. 389].
b1 1, 3 2, 2 3, 1 4, 3
b2 1, 4 2, 3 3, 3 4, 4
b3 3, 1 1, 4 2, 3 4, 2
b4 2, 2 3, 1 1, 4 4, 1

7.32 Example. Consider Gale and Shapley’s unlabeled example [40, 41, p. 389],
adapted in table 7.33. At the intersection of the row for bi and the column for cj ,
the ordered pair .mi ; nj / states that bi ranks cj in position mi , whereas cj ranks bi in
position nj . For instance, at the intersection of the row for b3 and the column for c2 ,
the ordered pair .1; 4/ states that b3 ranks c2 in position 1, whereas c2 ranks b3 in
position 4.
Start from the stable empty relation ¿  B C. Thus B0¿ D B and C¿ 0
D C. The
set of first-ranked choosers is C1 D fc1 ; c2 ; c3 g. They are ranked first by

Bc1 D fb1 ; b2 g; Bc2 D fb3 g; Bc3 D fb4 g:

Among those beggars who ranked them first, their first-ranked beggars are

bc1 D firstc1 .fb1 ; b2 g/ D b1 ; bc2 D b3 ; bc3 D b4 :

Theorem 7.30 extends M WD ¿ to

O D f.b1 ; c1 /; .b3 ; c2 /; .b4 ; c3 /g;

M

O to
where all beggars have their first choice. Theorem 7.30 then extends M

OO D f.b ; c /; .b ; c /; .b ; c /g [ f.b ; c /g;

M 1 1 3 2 4 3 2 4

where b2 and c4 have their worst choice, but then c4 was ranked last by every beggar.

7.3.4 An Algorithm for Optimal Match Making

Without a Match Maker

This subsection describes Gale & Shapley’s algorithm to find a stable relation
without any match maker [40, 41]. In the context of marriages, boys and girls carry
out the algorithm themselves through rounds of proposals and rejections.
 www.pdfgrip.com

320 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

7.34 Algorithm (Gale & Shapley’s Deferred Acceptance Procedure).

Initially, for each girl c 2 C, the set Bc of boys who proposed to her is empty.
Similarly, for each boy b 2 B, the set Cb of girls who rejected him is empty.
Then each round of proposals and rejections proceeds as follows:
(BG.1) Each girl c 2 C has a set Bc , which is either empty, or is a singleton
containing only (the name of) her top-ranked boy among those boys who have
already proposed to her.
(BG.2) Each boy b 2 B has a set Cb , which is either empty, or contains all the
girls who have already rejected him.
(BG.3) Each boy b 2 B proposes to his top-ranked girl cb WD first.CnCb / among
those girls who have not yet rejected him.
(BG.4) From the boys’ proposals, each girl c 2 C receives a set B0c of proposals,
which may be empty, she forms the union B00c WD B0c [ Bc , and rejects all but her
top-ranked boy in that set, in effect re-setting it to Bc WD ffirstc .B00c /g.
(BG.5) Each boy b 2 B who receives a rejection from a girl c 2 C appends her
to his set of rejections, in effect replacing Cb by Cb [ fcg.
If the sets B of beggars and C of choosers have the same finite number of
elements, then Gale & Shapley’s algorithm yields a stable relation after finitely
many iterations, as proved in exercises 7.15 and 7.16.

7.3.5 Exercises on Gale & Shapley’s Algorithms

7.15 . For disjoints sets B and C with the same finite number N of elements, prove
that if in some round any boy b 2 B receives his .N 1/-th rejection, then Gale &
Shapley’s algorithm terminates at the next round.
7.16 . For disjoints sets B and C with the same finite number N of elements, prove
that Gale & Shapley’s algorithm terminates after at most N 2 C 2 2N rounds.
7.17 . Prove or disprove that for all disjoint sets B and C with the same infinite
cardinality there must exist a total stable relation.
7.18 . Carry out Gale & Shapley’s algorithm with Gale & Shapley’s example 7.32.
7.19 . Extend algorithm 7.31 (with a match maker) to stable relations that may
relate each chooser to more than one beggar, with a quota qc of beggars for chooser
c (polygamy, polyandry, schools admitting more than one student, etc.).
7.20 . Extend algorithm 7.34 (without a match maker) to stable relations that may
relate each chooser to more than one beggar, with a quota qc of beggars for chooser
c (polygamy, polyandry, schools admitting more than one student, etc.).
7.21 . Denote by S the subset of the power set P .B C/ consisting of all stable
relations,Spartially ordered by inclusion. Prove that for each chain T  S , its union
UT WD T 2 S is an upper bound for T in S .
 www.pdfgrip.com

7.4 Arrow’s Impossibility Theorem 321

7.22 . Determine the maximum number of iterations of theorem 7.30 (match

making with a match maker) necessary to complete a total stable relation between
disjoints sets B and C with the same finite number N of elements.
7.23 . Design and test a program to iterate theorem 7.30 (match making with a
match maker) for disjoints sets B and C with the same finite number N of elements.
7.24 . Design and test a program to iterate theorem 7.30 (match making with a
match maker) for disjoints finite sets B and C with quota.

7.3.6 Projects

7.35 Project. Develop concepts, theorems, and algorithms for match making where
beggars might order any subset, not necessarily the whole set, of choosers, and
choosers might order any subset, not necessarily the whole set, of beggars. For
example, the National Resident Matching Program (NRMP) evidently uses an
algorithm to this effect [96].
7.36 Project. Develop concepts, theorems, and algorithms for tri-partite match
making. (Tri-partite reproduction occurs in Isaac Asimov’s novel The Gods Them-
selves [5]. See also the “three-parent” therapy [2].)

7.4 Arrow’s Impossibility Theorem

Arrow’s Impossibility Theorem exposes some of the limitations inherent to voting:

several desired features of voting procedures are mutually incompatible. Jointly with
John R. Hicks, Kenneth J. Arrow received the Nobel Prize in Economics in 1972,
in part for his Impossibility Theorem. Its precise statement and proof involve the
mathematical concepts of sets, functions, and ranking relations.

7.4.1 Introduction

The problem considered here consists in designing a procedure to rank several

mutually exclusive alternatives, or merely to choose exactly one such alternative. By
law or otherwise, the voting procedure may have to take into account many decision
criteria from the voters, and may have to conform to rules imposed in advance by
the voters.
In the political arena, the decision criteria may be ballots submitted by voters,
while the alternatives may be persons who are candidates for public office. Different
voting procedures may lead to the election of different candidates.
 www.pdfgrip.com

322 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

Fig. 7.3 The Mars Climate

Orbiter is conjectured to have
followed too low a trajectory
and burned up in Mars
atmosphere; art work
courtesy Jet Propulsion
Laboratory (JPL) and NASA.
(http://www.jpl.nasa.gov/
jplhistory/the90/images/
climate-orbiter-browse.jpg)

7.37 Example. In the United States presidential election of 1824, the popular and
electoral votes had ranked Andrew Jackson (Democrat-Republican Party) ahead of
John Quincy Adams (Coalition Party), William H. Crawford, and Henry Clay (Whig
Party); nevertheless the House of Representatives elected John Quincy Adams ahead
of Andrew Jackson (http://www.archives.gov/federal-register/electoral-college/
scores.html).
In the scientific arena, the decision criteria may be measurements from sensors,
which may act as voters, while the alternatives, which play the rôles of candidates,
may be decisions on how to proceed with a mission. In some missions, a simple
majority of votes may fail to select the best alternative (Figure 7.3).
7.38 Example. The Mars Climate Orbiter spacecraft was lost on 23 September
1999, due to small thrusters controlled by faulty unit conversions that compounded
during the year-long flight, which had been tracked by three telemetric systems:
Doppler only, range only, and Doppler and range combined. Two out of the three
systems submitted the same vote and won. Range only, and Doppler and range
combined, both showed a flight path clearing the planet, allowing the mission to
proceed without corrections. Both systems were wrong and the probe crashed.
The minority vote was right: “The Doppler-only solutions consistently indicated
a flight path insertion closer to the planet. These discrepancies were not resolved”
[78, p. 13]. In some situations, discrepancies must be investigated, not voted away.
Among other procedures, the voting procedures considered here are based on
voters ranking all the candidates.
7.39 Example. Table 7.40 shows Judges’ (voters’) ranking of three skaters (candi-
dates) at the 1994 Winter Olympic Games in Lillehammer, Norway [112, p. 22].
The judges’ plurality voting procedure selects the skater rated first by the largest
number of judges, here Baiul ahead of Kerrigan.
 www.pdfgrip.com

7.4 Arrow’s Impossibility Theorem 323

Table 7.40 Judges’ rankings of skaters [112, p. 22].

VOTERS
SKATERS D PL CZ UA PRC USA J CDN UK
Baiul 1 1 1 1 1 2 2 3 3
Kerrigan 2 2 2 2 2 1 1 1 1
Chen Lu 3 3 3 3 3 3 3 2 2

Table 7.42 Voters’ rankings of three candidates [113, p. 448].

VOTERS’ RANKINGS OF CANDIDATES
RANKS 0 1 2 3 4 5 6 7 8 9 X
Top Al Al Al Al Al Bo Bo Ci Ci Ci Ci
Second Bo Bo Bo Ci Ci Ci Ci Bo Bo Bo Bo
Last Ci Ci Ci Bo Bo Al Al Al Al Al Al

Table 7.43 Voters’ rankings of the two remaining candidates.

VOTERS’ RANKINGS OF CANDIDATES
RANKS 0 1 2 3 4 5 6 7 8 9 X
Top Al Al Al Al Al Bo Bo Bo Bo Bo Bo
Second Bo Bo Bo Bo Bo Al Al Al Al Al Al

Example 7.41 exposes some peculiarities of plurality voting.

7.41 Example. Table 7.42 shows hypothetical rankings of three candidates by
eleven voters, adapted from [113, p. 448].
The method of election called Borda’s count attempts to take into account
voters’ rankings by allocating a candidate two points for each top choice and one
point for each second choice from the voters. Table 7.42 shows that Ci leads with
with .4 2/ C .4 1/ D 12 points, followed by Bo .2 2/ C .7 1/ D 11 points,
and Al with 5 2 D 10 points. The press might list the outcome as Ci > Bo > Al.
Thus Ci is elected.
The method of election called plurality voting ignores voters’ second and
subsequent choices, takes into account only each voter’s top choice, and elects the
candidate who is the top choice of most voters. Table 7.42 shows Al leading with 5
top choices, followed by Ci with 4, and Bo with 2. Thus Al > Ci > Bo, and Al is
elected.
Suppose now that Ci drops out (as did Ross Perrot in 1992). Table 7.43 shows
the voters’ rankings of Al and Bo from table 7.42. With plurality voting, Bo wins
with 6 votes and Al loses with 5 votes: Bo > Al, and Bo is elected.
Such a reversal of the election result from Al to Bo, caused by a change in
the ranking of a seemingly irrelevant third candidate, Ci, is one of the features of
plurality voting that is deemed undesirable by some voters.
 www.pdfgrip.com

324 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

7.4.2 A Mathematical Model for Arrow’s Impossibility

Theorem

Many features have been deemed desirable from a voting procedure, for instance,
the following features.
(AIT.1) Unrestricted Domain. Each voter may submit any ranking of the candi-
dates: no rankings are forbidden.
(AIT.2) Unanimity. If every voter prefers candidate Al to Bo, then the voting
procedure must rank Al ahead of Bo.
(AIT.3) Independence of Irrelevant Alternatives. If every voter prefers candidate
Al to Bo in two different elections, then the voting procedure must rank Al
ahead of Bo in both elections, independently (regardless) of changes in the
ranking of any candidate Ci other than Al and Bo between the two elections.
(AIT.4) Absence of Dictators. The ranking from the voting procedure does not
coincide with the ranking of any fixed voter: no voters can dictate the outcome
of all elections.
(AIT.5) Anonymity. The ranking from the voting procedure does not depend
on who cast what vote. (Anonymity may be desirable for public votes, but
undesirable for the votes of elected representatives or scientific sensors.)
To get a sense of Arrow’s work, the reader may at this stage try to design a voting
procedure with all the features just listed to elect one among three candidates: either
find such a voting procedure, or prove that there are none.
Arrow’s Impossibility Theorem states that the first four features, (AIT.1),
(AIT.2), (AIT.3), and (AIT.4) are already mutually incompatible: no such voting
procedures are possible. As just stated, the four features are somewhat vague. For
instance, a voting procedure might need only to accommodate finitely many voters.
A precision sufficient for rigorous reasoning may have to be inserted and culminate
in a mathematical model of the descriptions of the features. To get a deeper sense
of Arrow’s work, the reader may at this stage try to formulate those four features
mathematically.
In general, a set V of voters must rank a set C of mutually exclusive alternatives,
for instance, candidates, decisions, laws, policies, etc. To this end, each voter
submits one ranking of the set C of candidates. From the rankings submitted by
the voters a “social welfare function” F produces a final “aggregate” ranking of the
set C of candidates (Table 7.44).
7.45 Definition. A weak ranking allowing for ties on a set C is a transitive and
strongly connected relation R  C C .
To allow for ties, weak rankings need not be anti-symmetric; thus weak rankings
are not partial orders.
The notation .X; Y/ 2 R, also abbreviated by XRY, means that R ranks X before
or tied with Y.
The transitivity of R is defined by 8X8Y8Z fŒ.XRY/ ^ .YRZ/ ) .XRZ/g.
 www.pdfgrip.com

7.4 Arrow’s Impossibility Theorem 325

Table 7.44 Symbols from logic.

SYMBOL DEFINITION
:.P/ “not P”: True if and only if P is False.
.P/ ^ .Q/ “P and Q”: True if and only if P and Q are both True.
.P/ _ .Q/ “P or Q”: False if and only if P and Q are both False.
.P/ ) .Q/ “P implies Q”: False if and only if P is True and Q is False.
8 “for each”;
9 “there exists”.

The strong connectedness of R is defined by

8X8Y fŒ.X 2 C / ^ .Y 2 C / ) Œ.XRY/ _ .YRX/g :

In particular, for all elements X and Y in C , if X D Y, then XRX, so that R is

reflexive.
The inverse ranking is the inverse relation Rı1 WD f.Y; X/ 2 C C W
.X; Y/ 2 Rg.
The preference PR associated with the ranking R is the set-theoretic difference
PR WD R n Rı1 .
A weak ranking R may also be denoted by such a symbol as  so that X  Y
means XRY.
The associated preference PR may then be denoted by  so that X  Y means
.X  Y/ ^ Œ:.Y  X/.
The set of all weak rankings of a set C is denoted by R .C /.
A voters’ profile is a function r W V ! R .C / with domain V .
To each voter V 2 V , a voters’ profile r associates that voter’s ranking, r.V/ 2
R .C /, also denoted by V .
The voter’s preference associated with that voter’s ranking r.V/ is then also
denoted by Pr.V/ or by V .
The set of all voters’ profiles, which are all functions from V to R .C / with
domain V , is denoted by R .C /V .
A social welfare function is a function F W R .C /V ! R .C / with domain
R .C /V .
To each voters’ profile r, a social welfare function F associates an “aggregate”
ranking F .r/ 2 R .C /.
7.46 Definition. A ranking R  C C induces a ranking RX;Y WD R \ .fX; Yg
fX; Yg/ of two candidates X; Y 2 C relative to each other. Two rankings R; S 
C C rank two candidates X; Y 2 C in the same order if and only if RX;Y D SX;Y .
Two voters rank two candidates in the same order if and only if so do their rankings.
The problem considered here consists in investigating the feasibility and design
of a social welfare function.
 www.pdfgrip.com

326 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

Table 7.47 Symbols for Arrow’s Impossibility Theorem.

SYMBOL DEFINITION
C Set of all candidates.
R .C / Set of all rankings on C .
V Set of all voters.
r Voters’ profile: a function r W V ! R .C / with domain V .
R .C /V Set of all voters’ profiles: all functions V ! R .C / with domain V .
F Social welfare function: a function F W R .C /V ! R .C /.
V
R .C /ŒR.C /  Set of all social welfare functions: all functions R .C /V ! R .C /.

X V
Y Alternative notation for .X; Y/ 2 Pr.V/ : voter V prefers X to Y.

7.4.3 Statement and Proof of Arrow’s Impossibility Theorem

Not all social welfare functions are considered here: only those that satisfy the
following four conditions.
(AIT.1) Unrestricted Domain. The domain of each social welfare function F is
all of R .C /V . This means that each voter may submit any ranking of the
candidates for the social welfare function to produce an aggregate ranking
(Table 7.47).
(AIT.2) Unanimity. For all candidates X and Y in C , if in a voters’ profile r each
voter V in V prefers X to Y, then so does the aggregate ranking from the social
welfare function F :
˚  
8X8Y8r Œr 2 R .C /V  ^ 8VŒ.X; Y/ 2 Pr.V/  ) Œ.X; Y/ 2 PF .r/  :

(AIT.3) Independence of Irrelevant Alternatives. If every voter V 2 V ranks

candidates X and Y in the same order in two voting profiles r and s, then the
aggregate rankings F .r/ and F .s/ also rank X and Y in the same order:
˚ ˚ 
8X8Y8r8s 8V Œr; s 2 R .C /V  ^ Œ.X; Y/ 2 r.V/ \ s.V/
) Œ.X; Y/ 2 F .r/ \ F .s/g :

(AIT.4) Absence of Dictators. The social welfare function does not coincide with
the ranking of any voter:
˚  ˚ 
: 9V 8r Œr 2 R .C /V  ) ŒF .r/ D r.V/ :

7.48 Definition. A permutation of a set V is a bijection  W V ! V with domain

V and range V .
The set of all permutations of a set V is denoted by SV and called the symmetric
group of V .
 www.pdfgrip.com

7.4 Arrow’s Impossibility Theorem 327

For all elements Bo; Ci 2 C , the transposition Bo;Ci is the permutation Bo;Ci 2
SV that swaps Bo and Ci but fixes every other element Al 2 C n fBo; Cig:

Bo;Ci .Bo/ WD Ci;

Bo;Ci .Ci/ WD BoI
8Alf.Al 2 C n fBo; Cig/ ) ΠBo;Ci .Al/ WD Alg:

A voter V 2 V swaps two candidates Bo and Ci by changing from a ranking V

to
the ranking defined by


f.X; Y/ 2 C C W Bo;Ci .X/ Bo;Ci .Y/g:
V

Arrow’s Impossibility Theorem applies to every set V of voters that is, or can be,
strictly well-ordered by a relation  so that each nonempty subset E  V has a first
element first.E / and V has a last element last.V /. The strict well-order  remains
fixed for the entire proof. For instance, some sets of voters may be in alphabetical
order.
7.49 Theorem (Arrow’s Impossibility Theorem). No social welfare functions
satisfy all four conditions (AIT.1)–(AIT.4).
Proof. This proof expands on [4, 6, 42, 143], showing that every function satisfying
(AIT.1), (AIT.2), (AIT.3) violates (AIT.4). For all distinct candidates Al; Bo 2 C ,
at one extreme focus on any profile r where every voter V 2 V prefers Al to Bo, so
that .Al; Bo/ 2 Pr.V/ for every V 2 V :

8Vf.V 2 V / ) Œ.Al; Bo/ 2 Pr.V/ g:

By the rule of unanimity so does the aggregate ranking: .Al; Bo/ 2 PF .r/ .
At the other extreme, focus on any profile s where every voter V 2 V prefers Bo
to Al, so that .Al; Bo/ 2 Ps.V/ for every V 2 V :

8Vf.V 2 V / ) Œ.Bo; Al/ 2 Ps.V/ g:

By the rule of unanimity so does the aggregate ranking: .Bo; Al/ 2 PF .s/ .
Between extremes, for each voter V 2 V define a profile rV 2 R .C / such that
each voter U  V prefers Bo to Al, so that Bo U Al for each U  V, while each
voter W  V prefers Al to Bo, so that Al W Bo for each W  V:

U  V ) .Bo; Al/ 2 PrV .U/ I

V  W ) .Al; Bo/ 2 PrV .W/ :

If V D last.V /, then rV D s and F .rV / D F .s/ ranks Bo ahead of Al.

 www.pdfgrip.com

328 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

Let EBo=Al be the subset of voters for whom F .rV / ranks Bo ahead of Al:

EBo=Al WD fV 2 V W .Bo; Al/ 2 PF .rV / g:

Thus last.V / 2 EBo=Al 6D ¿ and EBo=Al has a first elements VBo=Al WD first.EBo=Al /. This
first element VBo=Al is called a pivotal voter for Bo over Al.
By the rule of Independence of Irrelevant Alternatives VBo=Al is a pivotal voter
for Bo over Al starting from every profile r0 where every voter ranks Al before Bo.
Indeed, for each voter V 2 V , in r0 and in r, also in s0 and in s, as well as in r0V and
in rV , each voter U  V, V, and W  V ranks Al and Bo relative to each other in
the same way.
The following considerations show that VBo=Al can also dictate the ranking of Bo
relative to any third candidate Ci 2 C n fAl; Bog. To this end, consider the voters’
profile qV defined for V WD VBo=Al by
8
<U V W Bo U
Ci U
Al;
qV VW Al Bo Ci;
: V V
V  W W Al W
Bo W
Ci:

Since V D VBo=Al is pivotal for Bo over Al, the aggregate ranking F .qV / still ranks
Al  Bo. Also, by unanimity F .qV / ranks Bo  Ci. By transitivity the aggregate
ranking F .qV / ranks Al  Bo  Ci. If V D VBo=Al swaps Al and Bo, then the voting
profile changes from qV to rV , defined by
8
<U V W Bo U
Ci U
Al;
rV VW Bo Al Ci;
: V V
V  W W Al W
Bo W
Ci:

Since V D VBo=Al is pivotal for Bo over Al, the aggregate ranking F .rV / now ranks
Bo  Al. By Independence of Irrelevant Alternatives F .rV / still ranks Al  Ci,
because each voter ranks Al and Ci in the same order in both profiles qV and rV . By
transitivity, the aggregate ranking F .rV / ranks Bo  Al  Ci. If any, some, or all
voters other than V D VBo=Al swap Bo and Ci, then the voting profile changes from
rV to sV , defined by
8
<U V W Ci U
Bo U
Al or Bo U
Ci U
Al;
sV VW Bo Al Ci;
: V V
V  W W Al W
Ci W
Bo or Al W
Bo W
Ci:

Then Al and Bo are ranked in the same order by every voter in rV and sV , whence the
aggregate ranking F .sV / still ranks Bo  Al. Similarly, Al and Ci are ranked in the
same order by every voter in rV and sV , whence the aggregate ranking F .sV / still
ranks Al  Ci. By transitivity the aggregate ranking F .sV / ranks Bo  Al  Ci,
even though every voter other than V ranks Ci  Bo.
 www.pdfgrip.com

7.4 Arrow’s Impossibility Theorem 329

Candidate Al plays only an auxiliary rôle in the proof that VBo=Ci is a dictator for
Bo over Ci. If any, some, or all voters swap Al and any candidate Ig other than Bo
and Ci, then the voting profile changes from sV to pV , defined by
8
<U V W Ci U Bo U Al or U
Ci U Bo U ;
pV VW Bo V Al V Ci or Bo Ci ;
: V V V
V  W W Al W Ci W Bo or W
Ci W
Bo W
:
Then Bo and Ci are ranked in the same order by every voter in sV and pV , whence
the aggregate ranking F .pV / still ranks Bo  Ci, even though every voter other
than V ranks Ci  Bo.
For the uniqueness of the dictator, in the strict well-ordering  of the voters V ,
a pivotal voter VBo=Ci for Bo over Ci cannot appear later than such a dictator for Bo
over Ci as VBo=Al , because such an earlier dictator VBo=Al would determine the ranking
of Bo over Ci before a pivotal voter VBo=Ci does; consequently, VBo=Ci  VBo=Al .
Similarly, a pivotal voter VCi=Bo for Ci over Bo cannot come earlier than such a
dictator for Bo over Ci as VBo=Al , for otherwise such a later dictator could reverse a
pivotal vote from the pivotal voter; therefore VBo=Al  VCi=Bo .
Combining VBo=Ci  VBo=Al with VBo=Al  VCi=Bo gives VBo=Ci  VBo=Al  VCi=Bo by
transitivity.
Reversing the rôles of Bo and Ci gives the reverse ranks VCi=Bo  VCi=Al  VBo=Ci .
Consequently, VBo=Ci D VBo=Al D VCi=Bo .
Therefore there is exactly one pivotal voter, who is the pivotal voter and the
dictator for every pair of candidates.
Thus every social welfare function satisfying (AIT.1), (AIT.2), (AIT.3) violates
(AIT.4). t
u
Another rule that might be imposed on a social welfare function pertains to the
anonymity of voters. One way to model the concept of anonymity of voters consists
in requiring that a social welfare function remains invariant under all permutations
of the voters.
7.50 Definition. A social welfare function F W R .C /V ! R .C / is invariant
under permutations if and only if F .r ı  / D F .r/ for every profile r W V !
R .C / and every permutation  2 SV .
(AIT.5) Anonymity. A social welfare function respecting voters’ anonymity is
invariant under permutations.
Yet another condition attempts to model the condition that no voters have a right
of veto.
(AIT.6) No Veto. Except perhaps for voter Val, if none of the other voters prefer
any other candidate to Al, in other words, if all the other voters prefer Al to all
the other candidates, or are indifferent between Al and all the other candidates,
then the social welfare function does not rank any other candidate ahead of Al,
regardless of Val’s rankings [115, p. 386].

There is an alternative variation of the concept of unanimity.

 www.pdfgrip.com

330 7 Applications: Nobel-Prize Winning Applications of Sets, Functions, and Relations

(AIT.7) Weak Unanimity. Except perhaps for voter Val, if all the other voters do
not prefer Bo to Al, so that they might prefer Al to Bo or be indifferent between
Al and Bo, but Val prefers Al to Bo, then the social welfare function prefers Al
to Bo:
n o
8X8Y8r Œr 2 R .C /V  ^ Œ.X; Y/ 2 Pr.Val/  ^ 8VŒ.Y; X/ … Pr.V/ 

) Œ.X; Y/ 2 PF .r/  :

7.4.4 Exercises on Arrow’s Impossibility Theorem

7.25 . Prove that for each relation R  C C the relation PR WD R n Rı1 is

irreflexive.
7.26 . Prove that for each relation R  C C the relation PR WD R n Rı1 is
asymmetric.
7.27 . Prove that if a relation R  C C is transitive, then PR WD R n Rı1 is also
transitive.
7.28 . Prove that for each relation R  C C the relation PR WD R n Rı1 is
anti-symmetric.
7.29 . Prove that if a relation R  C C is transitive, then PR WD R n Rı1 is a
partial order.
7.30 . Prove that if a relation R  C C is transitive, then the inverse relation Rı1
is also transitive.
7.31 . With at least two voters and at least two alternatives, prove that every social
welfare function conforming to the rules of Unrestricted Domain and Anonymity
has no dictators.
7.32 . With at least two voters and at least three alternatives, prove that every social
welfare function conforming to the rules of Unrestricted Domain, Unanimity, and
Independence of Irrelevant Alternatives is not invariant under permutations of the
voters.
7.33 . Determine whether either of Unanimity and Weak Unanimity implies the
other.
7.34 . Determine whether Weak Unanimity is compatible with No Veto.
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises

Exercises from Chapter 1

1.1 Use axioms P1 and P2, theorem 1.12, and Modus Ponens:
` .L/ ) Œ.K/ ) .L/ axiom P1,
` .H/ ) f.L/ ) Œ.K/ ) .L/g theorem 1.12,
   
` .H/ ) f.L/ ) Œ.K/ ) .L/g ) Œ.H/ ) .L/ ) f.H/ ) Œ.K/ ) .L/g axiom P2,
` Œ.H/ ) .L/ ) f.H/ ) Œ.K/ ) .L/g Modus Ponens.

1.3 Use the reflexivity of implications and the law of commutation:

` fŒ.A/ ) .B/ ) .B/g ) fŒ.A/ ) .B/ ) .B/g P WD fŒ.A/ ) .B/
  ) .B/g in 1.14,
` Œ.A/ ) .B/ ) fŒ.A/ ) .B/ ) .B/g ) .B/ substitution in the law
„ ƒ‚ … „ ƒ‚ … „ƒ‚… of commutation.
Q P R

1.5 Substitute P for A and also P for B in the preceding tautology:

 
` Œ.A/ ) .B/ ) fŒ.A/ ) .B/ ) .B/g ) .B/ preceding tautology,
` Œ.P/ ) .P/ ) fŒ.P/ ) .P/ ) .P/g ) .P/ substitutions,
` .P/ ) .P/ theorem 1.14,
` fŒ.P/ ) .P/ ) .P/g ) .P/ Modus Ponens.

1.7 The formula fŒ.P/ ) .Q/ ) .P/g ) .P/ cannot be proved using only
implications.
1.9 Use the tautology Œ.H/ ) .L/ ) f.H/ ) Œ.K/ ) .L/g:

` fŒ.R/ ) .Q/ ) .P/ g ) fŒ.R/ ) .Q/ ) Œ .S/ ) .P/ g

„ ƒ‚ … „ƒ‚… „ ƒ‚ … „ƒ‚… „ƒ‚…
H L H K L

© Springer Science+Business Media New York 2015 331

Y. Nievergelt, Logic, Mathematics, and Computer Science,
DOI 10.1007/978-1-4939-3223-8
 www.pdfgrip.com

332 Solutions to Some Odd-Numbered Exercises

1.11 See also [72, p. 34]:

` .P/ ) fŒ.P/ ) .P/ ) .P/g substitution in axiom 1a,

` f.P/) Œ.P/ ) .P/g 

) .P/ ) fŒ.P/  ) .P/ ) .P/g
) Œ.P/ ) .P/ substitution in axiom 1b,

` .P/
 ) Œ.P/ ) .P/  substitution in axiom 1a,
` .P/ ) fŒ.P/ ) .P/ ) .P/g ) Œ.P/ ) .P/ Detachment.
` Œ.P/ ) .P/ Detachment.
1.13 Apply exercise 1.11
 and Detachment:
` Œ.P/ ) .P/ ) fŒ.P/ ) Œ.P/ ) .Q/g substitution in 1b,

) Œ.P/ ) .Q/
` fŒ.P/ ) Œ.P/ ) .Q/g ) Œ.P/ ) .Q/ Detachment.
1.15 Apply exercise 1.12 and Detachment:
`T hypothesis,
` .A/ ) .T/  exercise 1.12,
` Œ.A/ ) .T/ )  f.A/ ) Œ.T/ ) .C/g substitution in 1b,
) Œ.A/ ) .C/
` f.A/ ) Œ.T/ ) .C/g ) Œ.A/ ) .C/ Detachment.
1.17
` Œ.B/
 ) .C/ ) f.A/ ) Œ.B/ ) .C/g  1a,
` Œ.B/ ) .C/ ) f.A/ ) Œ.B/ ) .C/g )
˚  
Œ.B/ ) .C/ ) f.A/ ) Œ.B/ ) .C/g ) Œ.A/ ) .C/

) fŒ.B/ ) .C/ ) Œ.A/ ) .C/g 1b,
˚  
` Œ.B/ ) .C/ ) f.A/ ) Œ.B/ ) .C/g ) Œ.A/ ) .C/
) fŒ.B/ ) .C/ ) Œ.A/ ) .C/g Detachment.
1.19 Apply exercises 1.15,
 1.18, and Detachment: 
` Œ.A/ ) .B/ ) f.A/ ) Œ.B/ ) .C/g ) Œ.A/ ) .C/ 1b,
` Œ.B/˚) .C/  
) Œ.A/ ) .B/ ) f.A/ ) Œ.B/ ) .C/g ) Œ.A/ ) .C/ 1.12,
` fŒ.B/
n ) .C/ ) Œ.A/) .B/g o
) Œ.B/ ) .C/ ) f.A/ ) Œ.B/ ) .C/g ) Œ.A/ ) .C/

1.21 ` Œ.A/ ) .B/ ) fŒ.B/ ) .C/ ) Œ.A/ ) .B/g by substitution in 1a.

 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 333

1.23 Apply Tarski’s axiom III and Detachment:

` .H/ ) .K/ hypothesis,
` Œ.H/ ) .K/ ) fŒ.K/ ) .L/ ) Œ.H/ ) .L/g substitution in axiom III,
` Œ.K/ ) .L/ ) Œ.H/ ) .L/ Detachment,
` .K/ ) .L/ hypothesis,
` .H/ ) .L/ Detachment.
1.25 ` .P/ ) fŒ.P/ ) .Q/ ) .P/g by substitution in axiom I.
1.27 Apply exercises 1.23, 1.25, and 1.26:
 
P / ) Œ.P/ ) .Q/ ) fŒ.P/ ) .Q/ ) .Q/g :
.„ƒ‚…
„ ƒ‚ …
H L
1.29 Apply exercises 1.27, 1.28, and 1.23:
.P/ ) fŒ.P/ ) .Q/ ) .Q/g:
1.31 Apply exercise 1.29 and axiom III:
` f.Q/
h ) Œ.Q/ ) .R/ ) .R/g ) 
fŒ.Q/ ) .R/ ) .R/g ) Œ.P/ ) .R/
i
) f.Q/ ) Œ.P/ ) .R/g III,
` .Q/ ) Œ.Q/ ) .R/ ) .R/  1.29,
` fŒ.Q/ ) .R/ ) .R/g ) Œ.P/ ) .R/
) f.Q/ ) Œ.P/ ) .R/g Detachment.
1.33 Apply axiom III, exercise 1.32, and Detachment:
` Œ.P/ ) .Q/ ) fŒ.Q/ ) .R/ ) Œ.P/ ) .R/g axiom III,
` Œ.Q/ ) .R/ ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g 1.32, Detachment.
1.35

Apply axiom II with exercises 1.24,

1.33, and 1.23:
` Œ.P/ ) .Q/ ) f.P/ ) Œ.P/ ) .R/g
 
) f.P/ ) Œ.P/ ) .R/g ) Œ.P/ ) .R/ ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g III,
 
` f.P/ ) Œ.P/ ) .R/g ) Œ.P/ ) .R/
 
) Œ.P/ ) .Q/ ) f.P/ ) Œ.P/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g 1.32,

` f.P/ ) Œ.P/ ) .R/g ) Œ.P/ ) .R/ II,

 
` Œ.P/ ) .Q/ ) f.P/ ) Œ.P/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g Detachment,

` f.Q/ ) Œ.P/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g 1.34.

 
Œ.P/ ) .Q/ ) f.P/ ) Œ.Q/ ) .R/g ) Œ.P/ ) .R/g :
 www.pdfgrip.com

334 Solutions to Some Odd-Numbered Exercises

1.37 Proceed as in theorem 1.40:

` .P/ ) Œ.Q/ ) .P/ axiom F1;

` Œ.Q/ ) .P/ ) fŒ:.P/ ) Œ:.Q/g axiom F4;
` .P/ ) fŒ:.P/ ) Œ:.Q/g transitivity of implications (theorem 1.16):

1.39 A proof of fŒ:.Q/ ) Œ:.P/g ) Œ.P/ ) .Q/ can proceed as follows.

 
` fŒ:.Q/ ) Œ:.P/g ) f:Œ:.P/g ) f:Œ:.Q/g axiom F4;
` .P/ ) f:Œ:.P/g   axiom F6;
` fŒ:.Q/ ) Œ:.P/g ) .P/ ) f:Œ:.Q/g transitivity (theorem 1.31);
` f:Œ:.Q/g ) .Q/ axiom F5;
` fŒ:.Q/ ) Œ:.P/g ) Œ.P/ ) .Q/ transitivity (theorem 1.32):

1.41
 
fŒ.B/ ) .F/ ) Œ.A/ ) .F/g ) fŒ.B/ ) .F/ ) .A/g ) fŒ.B/ ) .F/ ) .F/g axiom C2;
fŒ.B/ ) .F/ ) .F/g ) .B/ axiom C3;
 
fŒ.B/ ) .F/ ) Œ.A/ ) .F/g ) fŒ.B/ ) .F/ ) .A/g ) .B/ transitivity;
.A/ ) fŒ.B/ ) .F/ ) .A/g C1;
fŒ.B/ ) .F/ ) Œ.A/ ) .F/g ) Œ.A/ ) .B/ transitivity:

1.43 Substitutions of :.B/ for .B/ ) .F/ and :.A/ for .A/ ) .F/ into the solution
of exercise 1.41 transform Church’s third axiom into fŒ:.Q/ ) Œ:.P/g )
Œ.P/ ) .Q/. Consequently, all three axioms of classical logic remain valid in
Church’s logic, and, therefore, Church’s logic allows for proofs of all the theorems
of classical logic.
1.45 For the first theorem,

` Œ:.P/ ) fŒ.S/ ) .S/ ) Œ:.P/g axiom P1;

 
` fŒ.S/ ) .S/ ) Œ:.P/g ) f:Œ:.P/g ) f:Œ.S/ ) .S/g law of contraposition;
 
` Œ:.P/ ) f:Œ:.P/g ) f:Œ.S/ ) .S/g transitivity;
` .P/ ) f:Œ:.P/g converse double negation;
 
` Œ:.P/ ) .P/ ) f:Œ.S/ ) .S/g transitivity:

For the second theorem,

` .S/ ) .S/ theorem 1.14;

` .P/ ) Œ.S/ ) .S/ theorem 1.12;
 
` Œ.P/ ) .Q/ ) f.P/ ) Œ:.Q/g ) Œ:.P/ reduction ad absurdum;
˚  
` f.P/ ) Œ.S/ ) .S/g ) .P/ ) f:Œ.S/ ) .S/g ) Œ:.P/ substitution;
 
` .P/ ) f:Œ.S/ ) .S/g ) Œ:.P/ Modus Ponens:
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 335

1.47 Pierce’s law:

fŒ.P/ ) .Q/ ) .P/g ) .P/
   m definition of _,
: :fŒ:.P/ _ .Q/g _ .P/ _ .P/
˚    m de Morgan’s second law,
: :fŒ:.P/ _ .Q/g ^ Œ:.P/ _ .P/
  m double negations,
fŒ:.P/ _ .Q/g ^ Œ:.P/ _ .P/
  m distributivity,
fŒ:.P/ _ .Q/g _ .P/ ^ fŒ:.P/ _ .P/g
m commutativity, associativity,
excluded middle,
Œ.Q/ _ .T/ ^ .T/
m identity,
.T/ ^ .T/
m idempotence,
.T/

1.49
fŒ.P/ ) .Q/ ) .R/g ) fŒ.R/ ) .P/ ) .P/g
m definition of _,
    
: :fŒ:.P/ _ .Q/g _ .R/g _ :fŒ:.R/ _ .P/g _ .P/
m de Morgan’s second law,
˚      
: :fŒ:.P/ _ .Q/g ^ Œ:.R/ _ f:Œ:.R/g ^ Œ:.P/ _ .P/
m double negations,
   
fŒ:.P/ _ .Q/g ^ Œ:.R/ _ f.R/ ^ Œ:.P/g _ .P/
m distributivity,
   
fŒ:.P/ ^ Œ:.R/g _ f.Q/ ^ Œ:.R/g _ Œ.R/ _ .P/ ^ fŒ:.P/ _ .P/g
m excluded middle,
  ˚ 
fŒ:.P/ ^ Œ:.R/g _ f.Q/ ^ Œ:.R/g _ Œ.R/ _ .P/ ^ .T/
m identity,
 
fŒ:.P/ ^ Œ:.R/g _ f.Q/ ^ Œ:.R/g _ Œ.R/ _ .P/
m de Morgan’s second law,
 
f:Œ.R/ _ .P/g _ f.Q/ ^ Œ:.R/g _ Œ.R/ _ .P/
m commutativity,
associativity,
 
f:Œ.R/ _ .P/g _ Œ.R/ _ .P/ _ f.Q/ ^ Œ:.R/g
m excluded middle,

.T/ _ f.Q/ ^ Œ:.R/g
m identity,
.T/
 www.pdfgrip.com

336 Solutions to Some Odd-Numbered Exercises

1.51

` .P/ ) fŒ.P/ ) .Q/ ) .Q/g

  theorem 1.26;
` fŒ.P/ ) .Q/ ) .Q/g ) Œ:.Q/ )
 f:Œ.P/ ) .Q/g theorem 1.44;
` .P/ ) Œ:.Q/ ) f:Œ.P/ ) .Q/g theorem 1.16:

1.53 One implication is axiom P1 and the other implication is Pierce’s law.
 
1.55 Œ.P/ ) .Q/ , :f.P/ ^ Œ:.Q/g :

.P/ ) .Q/
m double negations;
.P/ ) f:Œ:.Q/g
m   double negations;
: : .P/ ) f:Œ:.Q/g
m
  definition of ^;
:f.P/ ^ Œ:.Q/g

1.57

` .V/ ) .W/ hypothesis;

` .:.W// ) .:.V// contraposition;
` .R/ ) .S/ hypothesis;
` Œ:.V/ ) Œ.R/ ) .S/ theorem 1.12;

` fŒ:.V/ ) Œ.R/ ) .S/g

) fŒ.:.V// ) .R/ ) Œ.:.V// ) .S/g axiom P2;

` Œ.:.V// ) .R/ ) Œ.:.V// ) .S/ Modus Ponens;

` Œ.:.V// ) .R/ ) Œ.:.W// ) .S/ second line and theorem 1.16;
` Œ.V/ _ .R/ ) Œ.W/ _ .S/ definition of _:

1.59 No, the suggested rule fails if U and W are False but V is True.
1.61
 
` Œ.P/ )
 .S/ ^ fŒ:.Q/ ) Œ:.S/g 
) f.P/ ^ Œ:.Q/g ) f.S/ ^ Œ:.S/g theorem,
 
` f.P/
 ^ Œ:.Q/g ) f.S/
 ^ Œ:.S/g ) 
:f.S/ ^ Œ:.S/g ) :f.P/ ^ Œ:.Q/g contraposition,
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 337

   
` :f.S/
 ^ Œ:.S/g ) :f.P/ ^ Œ:.Q/g

) fŒ:.S/ _ .S/g ) Œ.P/ ) .Q/ de Morgan, definition of ^,

` Œ:.S/ _ .S/ excluded middle,

` .P/ ) .Q/ Modus Ponens.
1.63
` .R/ ) Œ.T/ ) .R/ axiom P1;

` .A/ ) fŒ.A/ ) .B/ ) .B/g theorem,

` .T/ ) fŒ.T/ ) .R/ ) .R/g substitutions in theorem,
`T hypothesis,
Œ.T/ ) .R/ ) .R/ Modus Ponens.

1.65 Use fŒ:.R/ ) .F/g ) fŒ:.F/ ) .R/g and Œ.T/ ) .R/ ) .R/.
1.67 Definition 1.51 of .A/ , .B/ as Œ.A/ ) .B/^Œ.B/ ) .A/ with theorems 1.57
Œ.P/ ^ .Q/ ) Œ.Q/ ^ .P/ and 1.52 Œ.P/ ^ .Q/ ) .Q/ yield Tarski’s axiom IV.
1.69 Theorem 1.61 gives a derivation of .I/ , .J/. from .I/ ) .J/ and .J/ ) .I/.
The Deduction Theorem (1.22) then yields a proof of Œ.I/ ) .J/ ) fŒ.J/ )
.I/ ) Œ.I/ , .J/g.
1.71 Subsections 1.3.10, 1.3.11, and 1.3.12 show that axioms P1 and P2 are
derivable from Tarski’s axioms I–III. Moreover, axiom P3 coincides with Tarski’s
axioms VII. Consequently, axioms P1, P2, and P3 are derivable from Tarski’s
axioms I–VII. Therefore, every theorem of the Classical Propositional Calculus is
also derivable from Tarski’s axioms I–VII.
1.73 Substitute R for S in theorem 1.82, which gives fŒ.P/ ) .Q/ ^ Œ.R/ )
.R/g ) fŒ.P/ ^ .R/ ) Œ.Q/ ^ .R/g. Then apply the reflexivity of the
logical implication (theorem 1.14), the law of contraposition (theorem 1.44), and
transitivity, to derive Rosser’s axiom R3 from axioms P1, P2, and P3.
1.75 Kleene’s axiom 7 is the law of reductio ad absurdum (theorem 1.48).
1.77 fŒ.P/ ) .Q/ ) .P/g ) .P/ is a triadic tautology in Łukasiewicz’s triadic
logic.
1.79 fŒ.P/ ) .Q/ ) .P/g ) .P/ is not a triadic tautology in Church’s system or
in Łukasiewicz’s.
1.81 fŒ.P/ ) .Q/ ) .Q/g ) Œ.Q/ ) .P/ ) .P/g is a triadic tautology in
Łukasiewicz’s system, but not in Church’s.
1.83 f.P/ ) Œ.Q/ ) .R/g ) fŒ.P/ ) .Q/ ) Œ.P/ ) .R/g is a triadic
tautology in Church’s system, but not in Łukasiewicz’s.
 www.pdfgrip.com

338 Solutions to Some Odd-Numbered Exercises

1.85 f.P/ ) Œ.Q/ ) .R/g ) f.Q/ ) Œ.P/ ) .R/g is a triadic tautology in both
Church’s system, and in Łukasiewicz’s.
1.87

` Œ:.P/ ) .P/ hypothesis;

` .P/ ) .P/ theorem 1.14;
` P deduction rule:

1.89 In the proof of fŒ:.P/ ) .P/g ` .P/, the first step lists the hypothesis H, here
` fŒ:.P/ ) .P/g. The Deduction Theorem replaces this step, ` H, by a complete
proof of .H/ ) .H/, from theorem 1.14, with Œ:.P/ ) .P/ substituted for H
everywhere. The second step invokes theorem 1.14, which the Deduction Theorem
replaces by a complete proof of theorem 1.14. The third step uses the deduction rule

` Œ:.R/ ) .S/ hypothesis;

` .R/ ) .S/ theorem 1.15;
` S deduction rule;

which the Deduction Theorem replaces by a complete proof of this deduction rule.

1.95 The commutation law establishes the logical equivalence

fŒ.P/ ) .Q/ ) Œ.Q/ ) .R/g ) Œ .Q/ ) .R/ 

„ ƒ‚ … „ƒ‚… „ƒ‚…
H K L

m
 
.Q/ ) fŒ.P/ ) .Q/ ) Œ.Q/ ) .R/g ) .R/ :
„ƒ‚… „ ƒ‚ … „ƒ‚…
K H L

To prove either formula with the Deduction Theorem, this proof starts by assuming
that the hypotheses H and K, here Q and Œ.P/ ) .Q/ ) Œ.Q/ ) .R/, are True.
` Œ.P/ ) .Q/ ) Œ.Q/ ) .R/
   
` .Q/ ) fŒ.P/ ) .Q/ ) Œ.Q/ ) .R/g ) f.Q/ ) Œ.P/ ) .Q/g ) f.Q/ ) Œ.Q/ ) .R/g
` f.Q/ ) Œ.P/ ) .Q/g ) f.Q/ ) Œ.Q/ ) .R/g
` .Q/ ) Œ.P/ ) .Q/
` .Q/ ) Œ.Q/ ) .R/
`Q
` .Q/ ) .R/
`Q
`R
The result then follows from the Deduction Theorem.
1.97 The formula S defined by fŒ:.P/ ) .P/g ) .P/ has only one propositional
variable, P. Moreover, S has the form .V/ ) .W/, with Œ:.P/ ) .P/ for V,
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 339

and P for W. Thus the first step consists in applying the Provability Theorem
(theorem 1.125) to prove P0 ` S0 .

P True If P is True, then so is W, and the single line ` P constitutes a proof

of ` W. Hence a copy of the proof of theorem 1.12 forms a complete proof of
P ` .V/ ) .W/, which is P0 ` S0 .
P False If P is False, then W is also False, but so is V. Thus, V 0 is :.V/, which
is :fŒ:.P/ ) .P/g. In this case the Provability Theorem calls for a proof of
P0 ` V 0 . However, V is False and has the form .H/ ) .K/, with :.P/ for H, and
P for K. Hence the Provability Theorem calls for proofs of P0 ` H 0 and P0 ` K 0 ,
which are here Œ:.P/ ` Œ:.P/ and Œ:.P/ ` Œ:.P/. In both cases the proof
of Œ:.P/ ` Œ:.P/ is a substitution in the proof of theorem 1.14. Hence follows
a proof of P0 ` .H 0 / ^ .K 0 /, which is here Œ:.P/ ` Œ:.P/ ^ Œ:.P/, and, by
definition of ^ the same proof shows that Œ:.P/ ` :fŒ:.P/ ) .P/g, which is
Œ:.P/ ` Œ:.V/, or, equivalently, P0 ` V 0 .
Thence follows a proof of P0 ` fŒ:.W/ ) Œ:.V/g and hence by contraposition
a proof of P0 ` Œ.V/ ) .W/, which is again P0 ` S0 .

1.99 The law of reductio ad absurdum, S,

 
Œ.P/ ) .Q/ ) f.P/ ) Œ:.Q/g ) Œ:.P/ ;

has the form .V/ ) .W/, with .P/ ) .Q/ for V, and f.P/ ) Œ:.Q/g ) Œ:.P/
for W.

P True, Q True If P is True and Q is True, then W is True. However, W has the
form .H/ ) .K/, with .P/ ) Œ:.Q/ for H, which is False, and :.P/ for K,
which is also False. Hence the Provability Theorem calls for a proof of P0 ; Q0 `
H 0 , here P; Q ` :.H/. Because H has the form .P/ ) Œ:.Q/, and hence :.H/
has the form .P/ ^ f:Œ:.Q/g, the Provability Theorem calls for proofs of P; Q `
P and P; Q ` Q, which follow from substitutions in the proof of theorem 1.14:

` .P/ ) .P/ theorem 1.14,

` .Q/ ) .Q/ theorem 1.14,
` .P/ ) f.Q/ ) Œ.P/ ^ .Q/g  theorem 1.82,
` .P/ ) .Q/ ) f:Œ.P/ ) Œ:.Q/g definition of ^,
` .P/ ) f.Q/
 ) Œ:.H/g  substitution,
` .P/ ) .Q/ ) fŒ:.K/ ) Œ:.H/g axiom P1 and theorem 1.16,
` .P/ ) f.Q/ ) Œ.H/ ) .K/g axiom P3 and theorem 1.16;

` .P/ ) Œ.Q/ ) .W/ substitution,

` .W/ ) Œ.V/ ) .W/ axiom P1,
` .P/ ) f.Q/ ) Œ.V/ ) .W/g theorem 1.32,
` .P/ ) Œ.Q/ ) .S/ substitution.
 www.pdfgrip.com

340 Solutions to Some Odd-Numbered Exercises

P False If P is False, then W is True, regardless of whether Q is True or False, and

in either case the same proof of theorem 1.14 gives a proof of Œ:.P/ ) Œ:.P/,
whence a proof of Œ:.P/ ) .W/, and hence a proof of Œ:.P/ ) Œ.V/ ) .W/,
which is Œ:.P/ ) .S/.
P True, Q False If P is True but Q is False, then W is False. Because S is a
tautology, V is also False. Hence the Provability Theorem calls for a proof of
P0 ; Q0 ` V 0 , here P; Œ:.Q/ ` :.V/. Because V has the form .H/ ) .K/, the
Provability Theorem calls for proofs of P0 ; Q0 ` H and P0 ; Q0 ` Œ:.K/, here
P; Œ:.Q/ ` P and P; Œ:.Q/ ` Œ:.Q/, both of which follow from substitutions
in the proof of theorem 1.14. Thence the proof of theorem 1.82 gives a proof of
.H/ ^ Œ:.K/, which is :.V/:

` .P/ ) .P/ theorem 1.14,

` Œ:.Q/ ) Œ:.Q/  theorem 1.14,
` .P/ ) Œ:.Q/ ) f.P/ ^ Œ:.Q/g  theorem 1.82,
` .P/ ) Œ:.Q/ ) f:Œ.P/ ) .Q/g definition of ^,
` .P/ ) fŒ:.Q/
 ) Œ:.V/g  substitution,
` .P/ ) Œ:.Q/ ) fŒ:.W/ ) Œ:.V/g axiom P1 and theorem 1.16,
` .P/ ) fŒ:.Q/ ) Œ.V/ ) .W/g axiom P3 and theorem 1.16.

From the proofs of .P/ ) Œ.Q/ ) .S/ and .P/ ) fŒ:.Q/ ) .S/g follows a proof
of .P/ ) .S/, and thence from the proof of Œ:.P/ ) .S/ follows a proof of S.
1.101 The propositional form fŒ.P/ ) .Q/ ) .R/g ) fŒ.R/ ) .P/ ) .P/g has
the form .V/ ) .W/.

P True If P is True, then axiom P1 gives a proof of .P/ ) .W/, and hence a proof
of .P/ ) .S/.
P False, R False If P is False, then .P/ ) .Q/ holds, by the law of denial of the
antecedent (theorem 1.40):

` Œ:.P/ ) Œ.P/ ) .Q/:

If R is also False, then Œ:.R/ ) Œ:.R/, whence Œ.P/ ) .Q/ ^ Œ:.R/ holds,
whence also :fŒ.P/ ) .Q/ ) .R/g, which is :.V/.
P False, R True If P is False and R is True, then Œ.R/ ) .P/ is False, but so is P,
whence Œ.R/ ) .P/ ) .P/, which is W, is True.
 
1.103 The propositional form fŒ.P/ ) .R/ ) .Q/g ) .Q/ ) fŒ.Q/ ) .R/ )
Œ.P/ ) .R/g has the form .V/ ) .W/.

R True If R is True, then axiom P1 gives a proof of W, whence a proof of S.

R False, P False If P is False, then the proof of Œ:.P/ ) Œ.P/ ) .R/ gives a
proof of W, whence a proof of S.
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 341

R False, P True, Q True

  then the definition of ^ gives
If Q is True and R is False,
a proof of .Q/ ) Œ:.R/ ) f:Œ.Q/ ) .R/g and hence a proof of W, whence
a proof of S.
R False, P True, Q False With R False, P True, Q False,
 the definition of ^ gives
a proof of .P/ ) Œ:.R/ ) f:Œ.P/ ) .R/g , and hence also a proof of
fŒ.P/
 ) .R/g ) .Q/, whence a proof of :.V/, because Q is False and V is
fŒ.P/ ) .R/g ) .Q/ ) .Q/.

1.105 The propositional form U defined by

fŒ.R/ ) .Q/ ) Œ.S/ ) .P/g ) fŒ.R/ ) .P/ ) Œ.S/ ) .P/g

has the form .V/ ) .W/.

P True If P is True, then axiom P1 gives a proof of .P/ ) Œ.S/ ) .P/, whence
a proof of .P/ ) .W/, and hence a proof of .P/ ) .U/.
P False, S False If S is False, then the law of denial of the antecedent, Œ:.S/ )
Œ.S/ ) .P/ gives a proof of Œ:.S/ ) .W/ and hence of Œ:.S/ ) .U/.
P False, S True, R False With P False, S True, R False, the law of denial of the
antecedent gives Œ:.R/ ) Œ.R/ ) .Q/, while .S/ ) .P/ is False, whence V is
False, and then the law of denial of the antecedent gives .V/ ) .W/, which is U.
P False, S True, R True With P False, S True, R True, .R/ ) .P/ is False, whence
the law of denial of the antecedent gives Œ.R/ ) .P/ ) Œ.S/ ) .P/, which is
W; hence the law of denial of the antecedent gives .V/ ) .W/, which is U.
P False, S True The foregoing two cases give a proof of Œ:.P/ ) Œ.S/ ) .U/.
P False The proofs of Œ:.P/ ) Œ.S/ ) .U/ and Œ:.P/ ) fŒ:.S/ ) .U/g
then combine into a proof of Œ:.P/ ) .U/.

Finally, the proofs of .P/ ) .U/ and Œ:.P/ ) .U/ combine into a proof of U.

Exercises from Chapter 2

2.1 9Xf8YŒ:.X 2 Y/g:

2.3 9Xf.X 2 A/ ^ Œ:.X 2 B/g:
 
2.5 9X f.X 2 C/^Œ:.X 2 A/^Œ:.X 2 B/g_fŒ:.X 2 C/^Œ.X 2 A/_.X 2 B/g :
2.7 8XŒ.X 2 A/ ) .X 2 B/:
2.9 8Xf.X 2 C/ , Œ.X 2 A/ ^ .X 2 B/g:
 
2.11 Theorem 2.46 establishes the equivalence Œ9X.P/ , :f8XŒ:.P/g .
 www.pdfgrip.com

342 Solutions to Some Odd-Numbered Exercises

2.13 Axiom Q2 is a theorem derivable from Margaris’s and Rosser’s axioms:

` f8XŒ.P/ ) .Q/g ) fŒ8X.P/ ) Œ8X.Q/g axiom A4,
` .P/ ) Œ8X.P/ axiom A6, no free X in P,
` f8XŒ.P/ ) .Q/g ) f.P/ ) Œ8X.Q/g derived rule.

2.15 Axiom Q4 follows from the abbreviation :f8XŒ:.P/g, double negation and
theorem 2.45:
` Œ8X.P/ , f8XŒ::.P/g
  theorem 2.45,
` f:Œ8X.P/g ,  :f8XŒ::.P/g
  contraposition and its converse,
` f9XŒ:.P/g , : 8Xf:Œ:.P/g abbreviation, .R/ , .R/,
` f9XŒ:.P/g , f:Œ8X.P/g transitivity.

2.17 Kleene’s 9-rule is derivable from the rules of inference with axioms Q1– Q4
and the propositional calculus.
` .P/ ) .Q/ hypothesis,
` Œ.P/ ) .Q/ ) fŒ:.Q/ ) Œ:.P/g contraposition,
` Œ:.Q/ ) Œ:.P/ Detachment,
` 8XfŒ:.Q/ ) Œ:.P/g Generalization,
` Œ:.Q/ ) f8XŒ:.P/g theorem 2.29,
` f8XŒ:.P/g ) f:Œ9X.P/g axiom Q3,
` Œ:.Q/ ) f:Œ9X.P/g transitivity,
` Œ9X.P/ ) .Q/ converse contraposition
and Detachment.

2.19 Kleene’s 9-schema is derivable from the rules of inference with axioms
Q1– Q4 and the propositional calculus.
` f8XŒ:.P/g ) fSubfXY Œ:.P/g axiom Q1,
` fSubfXY Œ:.P/g ) f:ŒSubfXY .P/g remark 2.20,
` f8XŒ:.P/g ) f:ŒSubfXY .P/g Detachment,
` ŒSubf
 Y .P/ ) :f8XŒ:.P/g
X
contraposition and double negation,
` :fŒ8XŒ:.P/g ) Œ9X.P/ theorem 2.46,
` ŒSubfXY .P/ ) Œ9X.P/ transitivity.

2.21
` .V/ , .U/ hypothesis,
` Œ.V/ , .U/ ) Œ.V/ ) .U/ theorem 1.62,
` .V/ ) .U/ Detachment,
` Œ.V/ ) .U/ ) fŒ.U/ ) .W/ ) Œ.V/ ) .W/g transitivity
(theorem 1.27),
` Œ.U/ ) .W/ ) Œ.V/ ) .W/ Detachment.
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 343

2.23
` .V/ , .U/ hypothesis,
` Œ.V/ , .U/ ) Œ.U/ ) .V/ theorem 1.62,
` .U/ ) .V/ Detachment,
` Œ.U/ ) .V/ ) fŒ.V/ ) .W/ ) Œ.U/ ) .W/g transitivity
(theorem 1.28),
` Œ.V/ ) .W/ ) Œ.U/ ) .W/ Detachment.

2.25 Theorem 2.45 shows that if ` .V/ , .U/, then ` .P/ ) .Q/.
2.27 If ` .V/ , .U/, then ` Œ:.V/ , Œ:.U/, by contraposition and
transposition.
2.31

f9X Œ.P/ _ .Q/g , fŒ9X.P/ _ Œ9X .Q/g , fŒ9X.Q/ _ Œ9X .P/g

 
, f9X Œ.Q/ _ .P/g

2.33

f9XŒ.P/ _ .P/g , fŒ9X.P/ _ Œ9X.P/g , Œ9X.P/

2.35
 
8X fŒ.P/ ^ .Q/ _ .R/g  ,
 8X fŒ.P/ _ .R/ ^ Œ.Q/ _ .R/g ,
f8XŒ.P/ _ .R/g ^ f8XŒ.Q/ _ .R/g

2.37
 
Œ9X.Q/ , 9Xf:Œ:.Q/g 
, :f8XŒ:.Q/g 
, :8X f8XŒ:.Q/g
, :8X f: Œ9X.Q/g
, : : f9X Œ9X.Q/g
, f9XŒ9X.Q/g

2.39 The implication

f8XŒ.P/ _ .Q/g ) f.P/ _ Œ8X.Q/g

 www.pdfgrip.com

344 Solutions to Some Odd-Numbered Exercises

is theorem 2.79. For the converse,

` .P/ ) Œ8X.P/ axiom Q1;

` Œ8X.Q/ ) Œ8X.Q/ theorem 1.14;
` f.P/ _ Œ8X.Q/g ) fŒ8X.P/ _ Œ8X.Q/g theorem 2.39;
` fŒ8X.P/ _ Œ8X.Q/g ) f8XŒ.P/ _ .Q/g theorem 2.77;
` f.P/ _ Œ8X.Q/g ) f8XŒ.P/ _ .Q/g theorem 1.16:

2.41 Invoke the reflexivity of the logical implication (theorem 1.63): by definition
of R,
 
ŒR.A; B/ , 8XfŒE .X; A/ ) ŒE .X; A/g ;

which is in prenex form, and its matrix is an instance of theorem 1.63: .P/ ) .P/.
2.43 Invoke the transitivity of the logical implication (theorem 1.65):

` ŒE .X; A/ ) ŒE .X; B/ specialization of the hypothesisR.A; B/;

` ŒE .X; B/ ) ŒE .X; C/ specialization of the hypothesisR.B; C/;
` ŒE .X; A/ ) ŒE .X; C/ transitivity of the implication (theorem 1.65/;

whence the conclusion R.A; C/ follows by Generalization.

2.45 Invoke the reflexivity of the logical implication (theorem 1.63): by definition
of A ,
 
ŒA .A; A/ , 8XfŒE .A; Y/ ) ŒE .A; Y/g ;

which is in prenex form, and its matrix is an instance of theorem 1.63: .P/ ) .P/.
2.47 The equality predicate I defined as in example 2.85 for set theory is
reflexive, as proved by the solutions to exercises 2.41 and 2.45, so that I satisfies
axioms J 1. Also, the set theory described in example 2.85 has only one predicate,
E , and formula (2.2) shows that I satisfies axiom J 2.
2.49 Axiom J 1 from subsection 2.5.3 coincides with axiom I 1 from subsec-
tion 2.5.1. Axiom J 2 from subsection 2.5.3 is a special case of axiom I 2 from
subsection 2.5.1, which allows for atomic formulae as particular cases of P and Q.

Exercises from Chapter 3

3.1 Negate the definition of the empty set: 9X.X 2 S/.

3.3 Negate the definition of supersets: 9XŒ.X 2 B/ ^ .X … A/.
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 345

3.5 By definition of the empty set (axiom S2), the formula :.X 2 ¿/ is universally
valid. By specialization, with ¿ substituted for X, it follows that :.¿ 2 ¿/ is True,
whence ¿ 2 ¿ is False, by definition of False.
3.7 Use substitutions in the axiom of extensionality and the definition of equality.
3.9 For each set S, ¿  S, by theorem 3.11. Consequently, ¿ is a subset of ¿.
Conversely, if S is a subset of ¿, so that S  ¿, then S D ¿ by theorem 3.10:
`S¿ hypothesis on S,
`¿S theorem 3.11,
`SD¿ theorem 3.10.
3.11 This proof proceeds by contraposition, showing that if S 6D ¿, then S has a
subset different from S. Because ¿  S by theorem 3.11, it follows that if S 6D ¿,
then S has a subset, ¿, different from S.
3.13 If S is a subset of every set, then S is a subset of the empty set: S  ¿.
Moreover, ¿  S by theorem 3.11. Consequently, S D ¿, by theorem 3.10.
3.15 If A ¨ B and B ¨ C, then A  B and B  C, whence A  C, by theorem 3.9.
However, because A ¨ B, there exists some Z 2 B such that Z … A. Consequently,
Z 2 C but Z … A, so that A 6D C, whence A ¨ C.
3.17 This proof establishes each implication () and () independently.
For one implication, assume ` 8YŒ.A  Y/ , .B  Y/.
` 8YŒ.A  Y/ , .B  Y/ hypothesis,
+ SubfYA ,
` .A  A/ , .B  A/
+ ),
` .A  A/ ) .B  A/
`AA theorem 3.8,
`BA Modus Ponens;

` 8YŒ.A  Y/ , .B  Y/ hypothesis,

+ SubfYB ,
` .A  B/ , .B  B/
+ ),
` .A  B/ ( .B  B/
`BB theorem 3.8,
`AB Modus Ponens;
`ADB theorem 3.10.
For the converse, assume ` A D B, and begin with any superset Y.
 www.pdfgrip.com

346 Solutions to Some Odd-Numbered Exercises

`AY hypothesis,
` 8XŒ.X 2 A/ ) .X 2 Y/ definition of subsets,
`BA hypothesis,
` 8XŒ.X 2 B/ ) .X 2 Y/ transitivity of ),
`BY definition of subsets.
Swapping A and B yields the converse. Hence ` 8YŒ.A  Y/ , .B  Y/.
3.19 If C  D and D  W, then D  C and W  D, whence W  C. For the
converse, let W WD D, whence D  D, the hypothesis .D  D/ ) .C  D/, and
Modus Ponens yield C  D.
˚ 
3.21 Let X WD ¿, Y WD f¿g, Z WD f¿g .
˚ 
3.23 Let X WD ¿ and Y WD f¿g .
˚ 
3.25 Let X WD f¿g and A WD f¿g .
3.27 .X 2 f¿g/ , .X D ¿/ whereas .X 2 f f¿g g/ , .X D f¿g/. Yet ¿ 6D f¿g.
Consequently, f¿g and f f¿g g have different elements. Therefore f¿g 6D f f¿g g.
3.29 f¿g 2 f ¿; f¿g g. Yet .X 2 f¿g/ , .X D ¿/ and ¿ 6D f¿g, whence
f¿g … f¿g. Hence f¿g 6D f ¿; f¿g g.
˚  ˚  ˚ 
3.31 From 8S.S  S/ specialization with S WD f¿g gives f¿g  f¿g .
˚ 
3.33
˚ The set f¿g ˚ has only
 ˚one element f¿g, which is also an element of the set
¿; f¿g . Hence f¿g  ¿; f¿g .
3.35 For theorem 3.13, the word “and” in the informal proof corresponds to
the logical connective _ in the formal proof through the universal quantifier,
specializations with H first and then K, and the axioms of extensionality and pairing,
along the following outline:
` .X D H/ ) Œ.X D H/ _ .X D K/ .P/ ) Œ.P/ _ .Q/,
` Œ.X D H/ _ .X D K/ ) .X 2 L/ pairing,
` .X D H/ ) .X 2 L/ transitivity,
` .X D H/ ) Œ.X 2 L/ ) .H 2 L/ extensionality,
` .X D H/ ) .H 2 L/ transitivity,
` .H D H/ ) .H 2 L/ specialization SubfXH ,
`HDH extensionality,
`H2L Modus Ponens;
`K2L as for H;
` .H 2 L/ ^ .K 2 L/ .P/ ) f.Q/ ) Œ.P/ ^ .Q/g.
3.37 If A D B, and if S  A, then each element of S is also an element of B, whence
S  B. Thus P.A/  P.B/, and conversely with the rôles of A and B switched.
If P.A/ D P.B/, and if X 2 A, then fXg 2 P.A/, whence fXg 2 P.B/, so that
fXg  B, and hence X 2 B. Thus A  B, and conversely B  A with the rôles of A
and B switched. Therefore A D B.
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 347

˚  ˚  ˚ 
3.39 f¿g \ ¿; f¿g D f¿g .
˚  ˚  ˚  ˚ 
3.41 f¿g [ ¿; f¿g D ¿; f¿g in the superset ¿; f¿g
3.43 8SŒ.S \ ¿/ D ¿.

 A n ¿ D fX 2 A W X … ¿g whereX … ¿ holds for every

3.45 By definition,
set X. Hence 8A 8Xf.X 2 A/ , Œ.X 2 A/ ^ .X … ¿/g whence A D A n ¿ by
extensionality.
3.47 By definition, AnB D fX 2 A W X … Bg. Thus .AnB/ D ¿ if and only if X … B
fails, and hence X 2 B holds, for every X 2 A, which is the definition of A  B.
3.49 The formula 8A8BŒ.A n B/ D .B n A/ is False. Indeed, with A WD ¿ and
B WD f¿g, it follows that

A WD ¿;
B WD f¿g;
A n B D ¿ n f¿g
D ¿
6D f¿g
D f¿g n ¿
D B n A:

3.51 f2; 3; 7g [ f3; 5; 7g D f2; 3; 5; 7g.

3.53 f2; 3; 7g \ f3; 5; 7g D f3; 7g.
3.55 f2; 3; 7gf3; 5; 7g D f2; 5g.
S
3.57 ¿ D ¿.
S˚  ˚ 
3.59 ¿; f¿g D f¿g 6D ¿; f¿g .
˚  S˚ 
3.61 f¿g 2 ¿; f¿g but f¿g … f¿g D ¿; f¿g .
3.63

X WD f¿g;
˚ 
A WD f¿g I
˚ 
Y WD f¿g ;
n˚ o
B WD f¿g I
˚ 
X [ Y D ¿; f¿g ;
n ˚ o
A [ B D f¿g; f¿g :
 www.pdfgrip.com

348 Solutions to Some Odd-Numbered Exercises

3.65
 [ 
X2 fAg , f9YŒ.Y 2 fAg/ ^ .X 2 Y/g

, f9YŒ.Y D A/ ^ .X 2 Y/g
, Œ9Y.X 2 A/
, X 2 A:

3.67 Use the tautology fŒ.P/ _ .Q/ ^ .R/g , fŒ.P/ ^ .R/ _ Œ.Q/ ^ .R/g.
3.69 Use the tautology Œ.P/ _ .F/ , .P/:

ŒX 2 .A [ ¿/ , Œ.X 2 A/ _ .X 2 ¿/

, X 2 A:

3.71 Use the tautology Œ.P/ _ .P/ , .P/.

3.73 Use de Morgan’s second law:
 
fX 2 ŒU n .A \ B/g , .X 2 U/ ^ f:ŒX 2 .A \ B/g
 
, .X 2 U/ ^ f:Œ.X 2 A/ ^ .X 2 B/g
, f.X 2 U/ ^ Œ.X … A/ _ .X … B/g
, fŒ.X 2 U/ ^ .X … A/ _ Œ.X 2 U/ ^ .X … B/g
, fŒ.X 2 .U n A/ _ ŒX 2 .U n B/g
, fX 2 Œ.U n A/cup.U n B/g

3.75 Use the tautology .B/ _ Œ:.B/ and the contradiction .B/ ^ Œ:.B/, with X 2 U
for B:

fX 2 Œ.A n B/ n Ug
, fŒX 2 .A n B/ ^ Œ:.X 2 U/g
˚ 
, .X 2 A/ ^ f:Œ.X 2 B/g ^ Œ:.X 2 U/
 
, f.X 2 A/ ^ Œ:.X 2 U/g ^ Œ:.X 2 B/
˚  
, ŒX 2 .A n U/ ^ Œ:.X 2 B/ _ f.X 2 U/ ^ Œ:.X 2 U/g
  
, f.X 2 A/ ^ Œ:.X 2 U/g ^ fŒ:.X 2 B/ _ .X 2 U/g ^ fŒ:.X 2 B/ _ Œ:.X 2 U/g
  
, f.X 2 A/ ^ Œ:.X 2 U/g ^ fŒ:.X 2 B/g ^ fŒ:.X 2 B/ _ Œ:.X 2 U/g
  
, f.X 2 A/ ^ Œ:.X 2 U/g ^ fŒ:.X 2 B/ _ Œ:.X 2 U/g
  
, f.X 2 A/ ^ Œ:.X 2 U/g ^ f:f.X 2 B/ ^ Œ:.X 2 U/g
 
, ŒX 2 .A n U/ ^ f:ŒX 2 .B n U/g
, fX 2 Œ.A n U/ n .B n U/g:
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 349

3.77

fX 2 Œ.A [ B/ n Ug , fX 2 .A [ B/ ^ Œ:.X 2 U/g

, fŒ.X 2 A/ _ .X 2 B/ ^ Œ:.X 2 U/g
 
, fŒ.X 2 A/ ^ Œ:.X 2 U/g _ fŒ.X 2 B/ ^ Œ:.X 2 U/g
, fX 2 Œ.A n U/ [ .B n U/g:

3.79

fX 2 ŒU \ .A n B/g
, f.X 2 U/ ^ ŒX 2 .A n B/g
 
, Œ.X 2 U/ ^ f.X 2 A/ ^ Œ:.X 2 B/g
 
, fŒ.X 2 U/ ^ Œ:.X 2 B/g ^ .X 2 A/
˚   
, fŒ.X 2 U/ ^ Œ:.X 2 B/g ^ .X 2 A/ _ fŒ.X 2 U/ ^ Œ:.X 2 B/g ^ Œ:.X 2 U/
  
, fŒ.X 2 U/ ^ Œ:.X 2 B/g ^ f.X 2 A/ ^ Œ:.X 2 U/g
   
, fŒ.X 2 U/ ^ Œ:.X 2 B/g ^ :fŒ:.X 2 A/ _ .X 2 U/g
, fX 2 Œ.U n B/ n .U n A/g:

3.81 If C  .A \ B/, then C  A and C  B.

3.83

A WD f¿g;
˚ 
B WD f¿g ;
˚ 
A [ B WD ¿; f¿g ;
C WD A [ B;
C 2 P.A [ B/;
C … P.A/;
C … P.B/;
C … P.A/ [ P.B/:

3.85 ¿ 2 P.A n B/ but ¿ 2 P.B/ whence ¿ … ŒP.A/ n P.B/.

 www.pdfgrip.com

350 Solutions to Some Odd-Numbered Exercises

3.87
S S
. F / \ B D A2F .A \ B/
m ˚  
S S
8X ŒX 2 . F / \ B , X 2 A2F .A \ B/
m
S
8X .fŒX 2 . F / ^ .X 2 B/g , Œ9A f.A 2 F / ^ ŒX 2 .A \ B/g/
m
8X .fŒ9A.A 2 F / ^ .X 2 A/ ^ .X 2 B/g
, Œ9A f.A 2 F / ^ Œ.X 2 A/ ^ .X 2 B/g/

which is a tautology by associativity of ^.

3.89 AA D .A [ A/ n .A \ A/ D A n A D ¿.
3.91 AB D .A [ B/ n .A \ B/ D .B [ A/ n .B \ A/ D BA.
3.93 Yes, Œ.AC/ [ .BC/  Œ.A [ B/C.

.A [ B/C D Œ.A [ B/ [ C n Œ.A [ B/ \ C

D Œ.A [ B/ [ C n Œ.A \ C/ [ .B \ C/
D Œ.A [ B [ C/ n .A \ C/ \ Œ.A [ B [ C/ n .B \ C/
D fŒ.A [ B [ C/ n A [ Œ.A [ B [ C/ n Cg \ fŒ.A [ B [ C/ n B
[ Œ.A [ B [ C/ n Cg
 fŒ.A [ C/ n A [ Œ.A [ C/ n Cg \ fŒ.B [ C/ n B [ Œ.B [ C/ n Cg
 fŒ.A [ C/ n A [ Œ.A [ C/ n Cg [ fŒ.B [ C/ n B [ Œ.B [ C/ n Cg
D Œ.A [ C/ n .A \ C/ [ Œ.B [ C/ n .B \ C/
D Œ.AC/ [ .BC/ :

3.95 Yes, Œ.AC/ \ .BC/  Œ.A \ B/C.

.A \ B/C D Œ.A \ B/ [ C n Œ.A \ B/ \ C

D fŒ.A \ B/ [ C n .A \ B/g [ fŒ.A \ B/ [ C n Cg
D fŒ.A \ B/ [ C n Ag [ fŒ.A \ B/ [ C n Bg [ fŒ.A \ B/ [ C n Cg
 fŒ.A [ C/ n A [ Œ.A [ C/ n Cg \ fŒ.B [ C/ n B [ Œ.B [ C/ n Cg
D Œ.A [ C/ n .A \ C/ \ Œ.B [ C/ n .B \ C/
D Œ.AC/ \ .BC/ :

3.97 Yes, Œ.AC/ n .BC/  Œ.A n B/C.

 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 351

3.99 No, Œ.A [ C/.B [ C/ 6 Œ.AB/ [ C, because the left-hand side does not
contain .A [ C/ \ .B [ C/ while the right-hand side contains all of C.
3.101 Yes, Œ.A \ C/.B \ C/  Œ.AB/ \ C.
3.103 Yes, Œ.A n C/.B n C/  Œ.AB/ n C.
3.105 No, Œ.C n A/.C n B/ 6 ŒC n .AB/, because the left-hand side does not
contain C n .A [ B/ while the right-hand side contains all of C n .A [ B/.
3.107 No, ŒP.A/P.B/ 6 ŒP.AB/, because the left-hand side does not
contain ¿ 2 P.A/ \ P.B/ while the right-hand side contains ¿ 2 P.AB/.
3.109

.AB/ \ .A \ B/ D Œ.A [ B/ n .A \ B/ \ .A \ B/

D ¿I
.AB/ [ .A \ B/ D Œ.A [ B/ n .A \ B/ [ .A \ B/
D A [ B:

3.111 As defined here the Cartesian product of sets is not associative (but a slightly
different version of the Cartesian product is associative). For example, .A B/ C 6D
A .B C/ for the sets A D f0g, B D f1g, and C D f2g, because

.A B/ C D f..0; 1/; 2/g D ff.0; 1/g; f.0; 1/; 2gg

A .B C/ D f.0; .1; 2//g D f f f0g; f0; .1; 2/g g g

which reveals that f0g 2 .0; .1; 2// 2 A .B C/ whereas f0g … ..0; 1/; 2/ 2
.A B/ C. Consequently, .0; .1; 2// 6D ..0; 1/; 2/ by extensionality, and then
.A B/ C 6D A .B C/ again by extensionality. (A slightly different definition of
the Cartesian product through functions of integers will later provide an associative
Cartesian product.)
3.113 (The last step still requires further symbolic substeps.)
.A D ¿/ _ .B D ¿/
m extensionality,
f:Œ9X.X 2 A/g _ f:Œ9Y.Y 2 B/g
m de Morgan’s Law,
:fŒ9X.X 2 A/ ^ Œ9Y.Y 2 B/g
m definition of A B,
:f9X9YŒ.X; Y/ 2 .A B/g
m uniqueness of ¿.
.A B/ D ¿
3.115 The Cartesian product distributes over unions: for all sets A, B, and C,

Œ.A B/ [ .C B/ D Œ.A [ C/ B:

 www.pdfgrip.com

352 Solutions to Some Odd-Numbered Exercises

An informal proof can establish that Œ.A B/ [ .C B/ and Œ.A [ C/ .B/ have
exactly the same elements. These two sets are Cartesian products, and, consequently
their elements are ordered pairs.
• An ordered pair .X; Y/ is an element of .A B/ [ .C B/ if and only if .X; Y/ 2
.A B/ or .X; Y/ 2 .C B/;
• hence .X; Y/ 2 Œ.A B/ [ .C B/ if and only if X 2 A and Y 2 B, or X 2 C and
Y 2 B,
• which is equivalent to X 2 A or X 2 C, and Y 2 B;
• thus .X; Y/ 2 Œ.A B/ [ .C B/ if and only if X 2 .A [ C/ and Y 2 B, which
is equivalent to .X; Y/ 2 Œ.A [ C/ B.
Just as the preceding informal proof concatenated the two occurrences of Y 2 B into
one such occurrence, a formal proof can rely on the distributivity of ^ over _ by the
tautology

fŒ.P/ ^ .R/ _ Œ.Q/ ^ .R/g , fŒ.P/ _ .Q/ ^ .R/g:

.X; Y/ 2 Œ.A B/ [ .C
B/
m definition of union,
Œ.X; Y/ 2 .A B/ _ Œ.X; Y/ 2 .C B/
m definition of Cartesian products,
Œ.X 2 A/ ^ .Y 2 B/ _ Œ.X 2 C/ ^ .Y 2 B/
m distributivity of ^ over _,
Œ.X 2 A/ _ .X 2 C/ ^ .Y 2 B/
m definition of union,
ŒX 2 .A [ C/ ^ Œ.Y 2 B/
m definition of Cartesian products.
.X; Y/ 2 Œ.A [ C/ B

3.117

ŒA  .B n D/ D Œ.A  B/ n .A  D/

8X8Yf.X; Y/ 2 ŒA  .B n D/ , .X; Y/ 2 Œ.A  B/ n .A  D/g

8X8YfŒ.X 2 A/ ^ .Y 2 B/ ^ f:.Y 2 D/gg , Œ.X 2 A/ ^ .Y 2 B/ ^ f:Œ.X 2 A/ ^ .Y 2 D/gg

8X8YfŒ.X 2 A/ ^ .Y 2 B/ ^ f:.Y 2 D/gg , Œ.X 2 A/ ^ .Y 2 B/ ^ fŒ:.X 2 A/ _ Œ:.Y 2 D/gg

8X8YfŒ.X 2 A/ ^ .Y 2 B/ ^ f:.Y 2 D/gg , Œ.X 2 A/ ^ .Y 2 B/ ^ f:.Y 2 D/g _ Œ.X 2 A/ ^ .Y 2 B/ ^ f:.X 2 A/gg

.P/ , Œ.P/ _ .False/ m

8X8YfŒ.X 2 A/ ^ .Y 2 B/ ^ f:.Y 2 D/gg , Œ.X 2 A/ ^ .Y 2 B/ ^ f:.Y 2 D/gg

 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 353

3.119 No, Œ.A n C/ .B n D/ 6 Œ.A B/ n .C D/, because the right-hand side
contains all of .A n C/ B.
3.121 No, Œ.AC/ .BD/ 6 Œ.A B/.C D/. For instance, if B D D, then
BD D ¿, whence .AC/ .BD/ D ¿ on the left-hand side. Yet on the right-
hand side, still with B D D, if C D ¿, then C D D ¿, whence .A B/.C D/ D
.A B/.
3.123 No, .ŒP.A/ ŒP.B// 6 P.A B/, because the left-hand side consists of
pairs of subsets of A and B, whereas the right-hand side consists of subsets of A B.
For instance, if A D ¿ D B, then A B D ¿, whence P.A B/ D P.¿/ D f¿g,
whereas ŒP.A/ ŒP.B/ D ŒP.¿/ ŒP.¿/ D f¿g f¿g D f.¿; ¿/g.
3.125 If S denotes the relation of strict inclusion on A WD P.H/, then

Sı1 D f.V; W/ 2 P.H/ P.H/ W W ¨ Vg:

3.127 From ¿  A and ¿  B it follows that ¿ ¿ 2 P.A/ P.B/, and also

¿ D ¿ ¿  ¿ ¿. Thus ¿ 2 D.
3.129 If A D ¿ D B, then P.A/ D P.¿/ D f¿g D D, because P.A/ P.B/ D
P.¿/ P.¿/ D f¿g f¿g D f.¿; ¿/g. From the solution to exercise 3.127, it
follows that P.A/ D f¿g is all of D.
3.131 No, F is not a function, because it contains two pairs, .0; 1/ and .0; 4/, with
the same first coordinate but different second coordinates.
3.133 Yes, R is a function.
3.135 Yes, Z is a function (in effect the zero function).
3.137 Yes, the empty set ¿  ¿ B is a function from ¿ to B.
3.139 For each X 2 B, 1A jB .X/ D 1 D B .X/, so that 1A jB coincides with B .X/
on B (but not on A).
3.141 If V 6D W, then V [ W is not a function, because of multiple values on
VW, where one of V or W has the value 0 while the other has the value 1.
Therefore, V [ W does not coincide with the function V[W .
3.143 Let

R WD f¿g;
˚ 
S WD f¿g ;
P
A WD R[S;
B WD f¿g;
F W A ! B;
˚ 
F WD .¿; ¿/; .f¿g; ¿/ :
 www.pdfgrip.com

354 Solutions to Some Odd-Numbered Exercises

Then

R \ S D ¿;
F".R \ S/ D F".¿/ D ¿;
F".R/ D F".f¿g/ D fF.¿/g D f¿g D B;
˚  ˚ 
F".S/ D F" f¿g D F.f¿g/ D f¿g D B;
F".R/ \ F".S/ D B \ B D B 6D F".R \ S/:

3.145 For all sets and for every function, ŒF".K/ n ŒF".H/  F".K n H/.
3.147 For each X 2 A, from fXg  A it follows that F".fXg/  B, with F".fXg/ D
fF.X/g by definition of F". Consequently,  X 2 A, the pair fXg; F".fXg/ 2
 for each
P.A/ P.B/ corresponds to the pair X; F.X/ 2 F  A B.
˚  ˚ 
3.149 F" ˚¿; f¿g D F.¿/; F.f¿g/
˚ D fF.0/; F.1/g D f2g.
Fı˚ 1" ¿; f¿g D Fı 1" 0; 1 D f2g.
F ¿; f¿g
n˚ D F.2/ D 0 D ¿.
o
Fı 1" ¿; f¿g D Fı 1".f2g/ D f0; 1g.

3.151 For each X 2 A, .IB ı F/.X/ D IB ŒF.X/ D F.X/.

3.153 ¿  F ı ¿  ¿ B  ¿.
3.155 Define a function G W F".A/ ! A as follows. For each Y 2 F".A/ there exists
exactly one X 2 D.F/  A with Y D F.X/; define G.Y/ WD X. Then GŒF.X/ D X,
so that G ı F D ID.F/ , with D.G/ D F".A/  B.
3.157 Let
˚ 
A WD ¿; f¿g ;
B WD f¿g;
˚ 
F WD .¿; ¿/; .f¿g; ¿/ ;
G WD f.¿; ¿/g:

Then .F ı G/.¿/ D ¿, so that F ı G D IB , and hence G is a right inverse for F. Yet

F has no left inverse, because F is not injective.
3.159 Let

A WD f0; 1g;
B WD f0; 1; 2g;
F WD f.0; 0/; .1; 1/g;
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 355

G WD f.0; 0/; .1; 1/; .2; 0/g;

H WD f.0; 0/; .1; 1/; .2; 1/g:

Then G ı F D IA , and H ı F D IA .

3.161 The empty relation is vacuously reflexive, symmetric, and transitive because
.X; Y/ 2 ¿ is False>
Reflexivity:

8Xf.X 2 ¿/ ) Œ.X; X/ 2 ¿g

is True (a theorem) by the tautology (theorem) .False/ ) .P/.

Symmetry:

8X8YfŒ.X; Y/ 2 ¿ ) Œ.Y; X/ 2 ¿g

is True (a theorem) by the tautology (theorem) .False/ ) .P/.

Transitivity:
 
8X8Y8Z fŒ.X; Y/ 2 ¿ ^ Œ.Y; Z/ 2 ¿ ) Œ.X; Z/ 2 ¿g

is True (a theorem) by the tautology (theorem) .False/ ) .P/.

3.163 The relation of strict inclusion is vacuously anti-symmetric, because the
hypothesis .V ¤ W/ ^ .W ¤ V/ is False for all sets V and W:
.V ¤ W/ ^ .W ¤ V/

m definition of ¤,

Œ.V  W/ ^ .V 6D W/ ^ Œ.W  V/ ^ .W 6D V/

m definitions of  and D,

f8XŒ.X 2 V/ ) .X 2 W/g ^ f9YŒ.Y 2 W/ ^ .Y … V/g

^f8XŒ.X 2 W/ ) .X 2 V/g ^ f9ZŒ.Z 2 V/ ^ .Z … W/g

m: Œ9Y.P/ , f:8YŒ:.P/g

f8XŒ.X 2 V/ ) .X 2 W/g ^ f:8Y:Œ.Y 2 W/ ^ .Y … V/g

^f8XŒ.X 2 W/ ) .X 2 V/g ^ f:8Z:Œ.Z 2 V/ ^ .Z … W/g
 www.pdfgrip.com

356 Solutions to Some Odd-Numbered Exercises

m de Morgan’s Law,

f8XŒ.X 2 V/ ) .X 2 W/g ^ f:8Y.Œ:.Y 2 W/ _ Œ:.Y … V//g

^f8XŒ.X 2 W/ ) .X 2 V/g ^ f:8Z.Œ:.Z 2 V/ _ Œ:.Z … W//g

m: Œ.P/ ) .Q/ , fŒ:.P/ _ .Q/g,

f8XŒ.X 2 V/ ) .X 2 W/g ^ f:8YŒ.Y 2 W/ ) .Y 2 V/g

^f8XŒ.X 2 W/ ) .X 2 V/g ^ f:8ZŒ.Z 2 V/ ) .Z 2 W/g

m substitutions,

f8XŒ.X 2 V/ ) .X 2 W/g ^ f:8YŒ.Y 2 W/ ) .Y 2 V/g

^f8YŒ.Y 2 W/ ) .Y 2 V/g ^ f:8XŒ.X 2 V/ ) .X 2 W/g

m: .P/ ^ Œ:.P/ is False,

False
Consequently, with .V ¤ W/ ^ .W ¤ V/ False, the implication

Œ.V ¤ W/ ^ .W ¤ V/ ) .V D W/

is True.
3.165 Here are the equivalence classes:

A=R D fŒ0; Œ1g;

Œ0 D f0; 2; 4g;
Œ1 D f1; 3; 5g:
S
3.167 Verify that F D B and that the elements of F are pairwise disjoint:
[
F D f0; 2; 4; 6g [ f1; 3; 5; 7g D f0; 1; 2; 3; 4; 5; 6; 7g D B;
f0; 2; 4; 6g \ f1; 3; 5; 7g D ¿:

Here is the corresponding equivalence relation:

 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 357

8 9
ˆ
ˆ .1; 7/ .3; 7/ .5; 7/ .7; 7/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 6/ .2; 6/ .4; 6/ .6; 6/ >
>
ˆ
ˆ >
ˆ
ˆ
ˆ .1; 5/ .3; 5/ .5; 5/ .7; 5/ >
>
>
>
< =
.0; 4/ .2; 4/ .4; 4/ .6; 4/
RF D :
ˆ
ˆ .1; 3/ .3; 3/ .5; 3/ .7; 3/ >
>
ˆ
ˆ >
>
ˆ
ˆ .0; 2/ .2; 2/ .4; 2/ .6; 2/ >
>
ˆ
ˆ >
>
ˆ
ˆ .1; 1/ .3; 1/ .5; 1/ .7; 1/ >
>
:̂ >
;
.0; 0/ .2; 0/ .4; 0/ .6; 0/

3.169 Outline: For each equivalence relation R, the partition FR consists of all the
equivalence classes corresponding to R. The relation RFR then consists of all the
pairs from all the equivalence classes of R, which is then R again.
3.171 The relation Q is
antisymmetric, vacuously, because it does not contain any two pairs .X; Y/ and
.Y; X/,
asymmetric, because it does not contain any two pairs .X; Y/ and .Y; X/,
not connected, because .2; 9/ … Q and .9; 2/ … Q,
irreflexive, because it does not contain any element on the diagonal,
not reflexive, because .0; 0/ … Q,
not strongly connected, because .2; 9/ … Q and .9; 2/ … Q,
not symmetric, because .0; 1/ 2 Q but .1; 0/ … Q,
not transitive, because .2; 6/ 2 Q and .6; 9/ 2 Q but .2; 9/ … Q.
3.173 The relation S is
not antisymmetric, because .0; 1/ 2 S and .1; 0/ 2 S but 0 6D 1,
not asymmetric, because it does not contain any two pairs .0; 1/ 2 S and
.1; 0/ 2 S ,
not connected, because .8; 9/ … S and .9; 8/ … S ,
irreflexive, because it does not contain any element on the diagonal,
not reflexive, because .0; 0/ … S ,
not strongly connected, because .8; 9/ … S and .9; 8/ … S ,
symmetric, because if .X; Y/ 2 S then .Y; X/ … S ,
not transitive, because .9; 6/ 2 S and .6; 8/ 2 S but .9; 8/ … S .
3.175 The relation V is
not antisymmetric, .0; 1/ 2 V and .1; 0/ 2 V but 0 6D 1,
not asymmetric, because it does not contain any two pairs .0; 1/ … V and
.1; 0/ … V ,
not connected, because .0; 2/ … V and .2; 0/ … V ,
not irreflexive, because it contains at least one element on the diagonal,
not reflexive, because .0; 0/ … V ,
not strongly connected, because .0; 2/ … V and .2; 0/ … V ,
not symmetric, because .1; 2/ 2 V but .2; 1/ … V ,
not transitive, because .0; 1/ 2 V and .1; 2/ 2 V but .0; 2/ … V .
 www.pdfgrip.com

358 Solutions to Some Odd-Numbered Exercises

3.177 The empty relation is vacuously a partial order.

3.179 See the definitions of the “diagonal” (example 3.68) and “irreflexive”
(definition 3.165).
3.181 By definition, every strict partial order is irreflexive and transitive. If it
contained .X; Y/ and .Y; X/, then it would also contain .X; X/ by transitivity, but
it cannot contain .X; X/ by irreflexivity; consequently, it cannot contain both .X; Y/
and .Y; X/, which makes it asymmetric.
3.183 By definition 3.176, an asymmetric relation cannot contain both .X; Y/
and .Y; X/, so that the hypothesis in the definition (3.173) of “anti-symmetric” is
vacuously False.
3.185 By the preceding exercise (3.184), every asymmetric relation is also
irreflexive; consequently, an asymmetric and transitive relation is also irreflexive and
transitive, which is the definition of a strict partial order. Conversely, the solution
to exercise 3.181 shows that every strict partial order is asymmetric, whence also
asymmetric and transitive.
3.187 The empty relation is vacuously asymmetric and strongly connected.
3.189 The empty subset of the empty set is vacuously a chain with respect to the
relation of inclusion.

Exercises from Chapter 4

4.1
( )
˚  n ˚ o ˚  n ˚ o
5D ¿; f¿g; ¿;f¿g ; ¿;f¿g; ¿;f¿g ; ¿; f¿g; ¿;f¿g ; ¿;f¿g; ¿;f¿g

4.3 For each set, X 2 fXg by pairing, whence X 2 .X [ fXg/ by union.

Moreover, for each set, X  X, whence X  .X [ fXg/, also by union.
˚ 
4.5 The equation may fail. Let A WD ¿ and B WD f¿g . Then A  B but

A [ fAg D ¿ [ f¿g D f¿g;

˚  ˚  ˚ 
B [ fBg D f¿g [ f¿g D f¿g; f¿g ;
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 359

˚ 
but ¿ … f¿g; f¿g whence

˚ 
f¿g 6 f¿g; f¿g ;

so that A [ fAg 6 B [ fBg.

˚ 
4.7 The equation may fail. Let A WD ¿ and B WD f¿g . Then

˚  ˚ 
.A [ fAg/ \ .B [ fBg/ D .¿ [ f¿g/ \ f¿g [ f¿g D ¿;
 ˚  ˚ 
.A \ B/ [ fA \ Bg D ¿ \ f¿g [ ¿ \ f¿g D f¿g;

so that .A [ fAg/ \ .B [ fBg/ 6D .A \ B/ [ fA \ Bg.

4.9 This proof proceeds by induction with M.

Initial Step

If M D 0, then M D ¿, whence the hypothesis “K 2 M and L 2 M” is False for all

K; L 2 N, and hence the implication is True.

Induction hypothesis

Assume that there exists N 2 N such that the theorem holds for M WD N, so that for
all K; L; 2 N, if K 2 N and L 2 N, then K [ L 2 N.

Induction step

For all K; L; 2 N, if K 2 .N [ fNg/ and L 2 .N [ fNg/, then only four cases can
arise.
If K 2 N and L 2 N, then .K[L/ 2 N by induction hypothesis; from N  N[fNg
it then follows that .K [ L/ 2 .N [ fNg/.
 www.pdfgrip.com

360 Solutions to Some Odd-Numbered Exercises

If K 2 fNg and L 2 N, then K D N and L  N, whence K [ L D N [ L D N,

and hence .K [ L/ D N 2 fNg  .N [ fNg/; swapping K and L yields a proof for
K 2 N and L 2 fNg.
If K 2 fNg and L 2 fNg, then K D N and L D N, whence K [ L D N, and hence
.K [ L/ D N 2 fNg  .N [ fNg/.
4.11 Let H WD I be as defined by the Axiom of Infinity.
H2F yet unproved,

m definition of F ,

  ŒH 2 P.H/
^ .¿ 2 H/ ^ 8Xf.X 2 H/ ) Œ.X [ fXg/ 2 Hg
which is True by the definition of H and the Axiom of Infinity.
4.13
8Xf.X 2 N/ ) Œ.X [ fXg/ 2 Ng yet unproved,

m definition of N,
T T
8Xf.X 2 F / ) Œ.X [ fXg/ 2 F g
T
m definition of ,

 8X f8BŒ.B 2 F / ) .X 2 B/g
) 8Bf.B 2 F / ) Œ.X [ fXg/ 2 Bg

*
 
8X 8B fŒ.B 2 F / ) .X 2 B/g
) f.B 2 F / ) Œ.X [ fXg/ 2 Bg

˚  * axiom P2,
8X 8B .B 2 F/
) f.X 2 B/ ) Œ.X [ fXg/ 2 Bg
which is True by the definition of F . Moreover, ¿ 2 N by theorem 4.8.
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 361

4.15
`I2F theorem 4.4,
T
` .I 2 F / ) . F  I/ theorem 3.48,
T
F I Modus Ponens;
T
` .S  N/ , .S  F/ definition of F ,
SI transitivity of ;
  
` .S  N/ ^ .¿ 2 S/ ^ 8Xf.X 2 S/ ) Œ.X [ fXg/ 2 Sg hypothesis on S,
 
` .¿ 2 S/ ^ 8Xf.X 2 S/ ) Œ.X [ fXg/ 2 Sg Œ.P/ ^ .Q/ ) .Q/ and M.P.,
  
` .¿ 2 S/ ^ 8Xf.X 2 S/ ) Œ.X [ fXg/ 2 Sg ) .S 2 F / definition of F ,
`S2F M.P.
T
` .S 2 F / ) . F  S/ theorem 3.48,
T
F S Modus Ponens;
T T T
F DS F  S and S  F .
`SDN definition of N.

T T
4.17 Because ¿ 2 N it follows that N  ¿ whence N D ¿.
˚ 
P
4.19
˚  S WD ˚V [ ¿; f¿g
Let  . This proof shows that S D N, whence V D S n
¿; f¿g D N n ¿; f¿g . ˚  ˚ 
For each˚ X 2 S, either X 2 ¿; f¿g or X … ¿; f¿g .
If X 2 ¿; f¿g , then either X D ¿ or X D f¿g.
In the first case (if X D ¿), then ¿ 2 S, by pairing and union. Moreover, ¿ [
f¿g D f¿g 2 V by hypothesis, whence ¿ [ f¿g ˚ 2 S. ˚ 
In the second case (if ˚X D f¿g),
 then f¿g[ f¿g 2 V whence f¿g[ f¿g 2 S.
Otherwise (if X 2 Sn ¿; f¿g ), then X 2 V, whence X [fXg 2 V by hypothesis,
and hence X [ fXg 2 S by union.
Thus, X [ fXg 2 S for each X 2 S, and ¿ 2 S. Consequently, S D N by the
Principle of Mathematical Induction.
Therefore, V D S n f¿g D N n f¿g.
4.21 1 C 1 D 1 [ f1g by definition of M C 1 D M [ fMg, and 1 [ f1g D 2 by
definition of 2.
4.23 3 C 1 D 3 [ f3g by definition of M C 1 D M [ fMg, and 3 [ f3g D 4 by
definition of 4.
4.25 5 C 1 D 5 [ f5g by definition of M C 1 D M [ fMg, and 5 [ f5g D 6 by
definition of 6.
4.27 7 C 1 D 7 [ f7g by definition of M C 1 D M [ fMg, and 7 [ f7g D 8 by
definition of 8.
4.29 2C2 D 2C.1C1/ D .2C1/C1 by definition of M C.N C1/ D .M CN/C1,
and .2 C 1/ C 1 D 3 C 1 D 4 by the preceding exercises.
 www.pdfgrip.com

362 Solutions to Some Odd-Numbered Exercises

4.31 4C2 D 4C.1C1/ D .4C1/C1 by definition of M C.N C1/ D .M CN/C1,

and .4 C 1/ C 1 D 5 C 1 D 6 by the preceding exercises.
4.33 6C2 D 6C.1C1/ D .6C1/C1 by definition of M C.N C1/ D .M CN/C1,
and .6 C 1/ C 1 D 7 C 1 D 8 by the preceding exercises.
4.35 3C3 D 3C.2C1/ D .3C2/C1 by definition of M C.N C1/ D .M CN/C1,
and .3 C 2/ C 1 D 5 C 1 D 6 by the preceding exercises.
4.37 5C3 D 5C.2C1/ D .5C2/C1 by definition of M C.N C1/ D .M CN/C1,
and .5 C 2/ C 1 D 7 C 1 D 8 by the preceding exercises.
4.39 4C4 D 4C.3C1/ D .4C3/C1 by definition of M C.N C1/ D .M CN/C1,
and .4 C 3/ C 1 D 7 C 1 D 8 by the preceding exercises.
4.41 2  2 D 2  .1 C 1/ D .2  1/ C 2 by definition of M  .N C 1/ D .M  N/ C M,
and 2  1 D 2 by definition of M  1 D M, whence .2  1/ C 2 D 2 C 2 D 4 by the
preceding exercises.
4.43 4  2 D 4  .1 C 1/ D .4  1/ C 4 by definition of M  .N C 1/ D .M  N/ C M,
and 4  1 D 4 by definition of M  1 D M, whence .4  1/ C 4 D 4 C 4 D 8 by the
preceding exercises.
4.45 No, because 1 C .1  1/ D 1 C 1 D 2 but .1 C 1/  .1 C 1/ D 2  2 D 4.
4.47 Let C WD N N and specify G W C ! C by G.K; L/ WD .K  L; L C 1/. Then
define F W N ! C by F.0/ WD .1; 1/ and F.I C 1/ WD GŒF.I/. Thus,

F.0/ D .0; 1/;

F.1/ D GŒF.0/ D G.1; 1/ D .1  1; 1 C 1/ D .1; 2/;
F.2/ D GŒF.1/ D G.1; 2/ D .1  2; 2 C 1/ D .2; 3/;
F.3/ D GŒF.2/ D G.2; 3/ D .2  3; 3 C 1/ D .6; 4/;
::
:
F.I C 1/ D GŒF.I/ D G.IŠ; I C 1/ D . .IŠ/  .I C 1/; .I C 1/ C 1 /
D . .I C 1/Š; I C 2 /;
::
:

Thus NŠ equals the first projection of F.N/.

4.49 No, because .1 C 2/Š D 3Š D 6 but .1Š/ C .2Š/ D 1 C 2 D 3.
4.51 2 < 5 because 2 2 5:
˚ 
¿;f¿g 2

( )
˚  n ˚ o ˚  n ˚ o
¿; f¿g; ¿;f¿g ; ¿;f¿g; ¿;f¿g ; ¿; f¿g; ¿;f¿g ; ¿;f¿g; ¿;f¿g
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 363

4.53 2 < 5 and 5 < 7 by previous exercises whence 2 < 7 by transitivity.

4.55 3 < 4, and 4 < 5, by definition of <, 5 < 7 by previous exercises, whence
3 < 7 by transitivity.
4.57 2 < 7 by previous exercises, and 7 < 8 by definition of < whence 2 < 8 by
transitivity.
4.59 If 1 < K and 1 < L, then by theorem 4.58 there exist I 2 N and J 2 N with
K D I C 1 and L D J C 1. Hence K  L D .I C 1/  .J C 1/ D I  J C I C J C 1 >
0 C 1 C 1 C 1.
4.61 If 3 < K and 3 < L, then 3  3 < K  L, whence 7 < 9 < K  L; thus K < 4
and L < 4. The only possibilities with 1 < K < 4 and 1 < L < 4 are K 2 f2; 3g
and L 2 f2; 3g. Yet 2  2 D 4 6D 7, 2  3 D 3  2 D 6 6D 7, and 3  3 D 9 6D 7.
4.63 Similarly, K  L D 9 if and only if .K; L/ 2 f.1; 9/; .3; 3/; .9; 1/g.
4.65 If S 6D ¿ and S  N then S has a smallest element I 2 S, by the Well-Ordering
Principle. Consequently, for every J 2 S it follows that I J, which means that
.I D J/ _ .I < J/, whence .I < J/ for I 6D J. Thus :.J < I/ because only one of
the three relations (<, D, >) holds. From :.J 2 I/ for every J 2 S it follows that
I \ S D ¿.
4.67 This proof proceeds by contradiction. Negating the conclusion gives

:Œ.I … K/ _ .K … L/ _ .L … I/ D Œ:.I … K/ ^ Œ:.K … L/ ^ Œ:.L … I/

D .I 2 K/ ^ .K 2 L/ ^ .L 2 I/

If .I 2 K/ ^ .K 2 L/ ^ .L 2 I/ for some I; K; L 2 N, then I 2 I by transitivity of 2

on N, but I … I for every I 2 N. Therefore the negation of the conclusion is False.
4.69 This proof proceeds by contradiction.
First, I … I for every I 2 N: ` 8If.I 2 N/ ) Œ:.I 2 I/g.
Second, SubfIN gives .N 2 N/ ) Œ:.N 2 N/, which has the pattern .P/ )
Œ:.P/.
From the tautology f.P/ ) Œ:.P/g ) Œ:.P/ it then follows that :.N 2 N/.
4.71 Calculate Œ.2; 4/l C Œ.6; 3/l D Œ.0; 2/l C Œ.3; 0/l D Œ.0 C 3; 2 C 0/l D
Œ.3; 2/l D Œ.1; 0/l .
4.73 Calculate Œ.3; 1/l Œ.5; 2/l D Œ.3; 1/l CŒ.2; 5/l D Œ.2; 0/l CŒ.0; 3/l D
Œ.2 C 0; 0 C 3/l D Œ.2; 3/l D Œ.0; 1/l .
 www.pdfgrip.com

364 Solutions to Some Odd-Numbered Exercises

4.75 Kunen’s addition of pairs of natural numbers commutes.

.Œ.M; N/l / C .Œ.P; Q/l / D Œ.M C N; P C Q/l

D Œ.N C M; Q C P/l
D .Œ.P; Q/l / C .Œ.M; N/l / :

4.77 Kunen’s addition of pairs of natural numbers is associative.

.Œ.K; L/l C Œ.M; N/l / C Œ.P; Q/l D Œ.K C M; L C N/l C Œ.P; Q/l
D Œ..K C M/ C P; .L C N/ C Q/l
D Œ.K C .M C P/; L C .N C Q//l
WD Œ.K; L/l C Œ.N C M; Q C P/l
WD Œ.K; L/l C .Œ.P; Q/l C Œ.M; N/l / :

4.79 Kunen’s multiplication of pairs of natural numbers commutes.

.Œ.M; N/l /  .Œ.P; Q/l / D Œ.M  P C N  Q; M  Q C N  P/l

D Œ.P  M C Q  N; Q  M C P  N/l
D Œ.P  M C Q  N; P  N C Q  M/l
D .Œ.P; Q/l /  .Œ.M; N/l /
:

4.81 Kunen’s multiplication of pairs of natural numbers is associative.

Œ.K; L/l  .Œ.M; N/l /  .Œ.P; Q/l /

D Œ.K; L/l  Œ.M  P C N  Q; M  Q C N  P/l

D Œ.K  .M  P C N  Q/ C L  .M  Q C N  P/;
K  .M  Q C N  P/ C L  .M  P C N  Q//l

D Œ..K  P C L  Q/  M C .K  Q C L  P/  N;
.K  P C L  Q/  N C .K  Q C L  P/  M/l

D Œ.K  P C L  Q; K  Q C L  P/l  Œ.M; N/l

D .Œ.K; L/l  Œ.P; Q/l /  Œ.M; N/l :

 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 365

4.83 Kunen’s multiplication of pairs of natural numbers distributes over addition.

Œ.K; L/l  .Œ.M; N/l C Œ.P; Q/l /

D Œ.K; L/l  Œ.M C P; N C Q/l

D Œ.K  .M C P/ C L  .N C Q/; K  .N C Q/ C L  .M C P/l

D Œ.K  P C L  Q C K  M C L  N; K  Q C L  P C K  N C L  Ml

D Œ.K  P C L  Q; K  Q C L  P/l C Œ.K  M C L  N; K  N C L  M/l

D .Œ.K; L/l  Œ.P; Q/l / C .Œ.K; L/l  Œ.M; N/l / :

4.85 Subtraction does not commute:

Œ.0; 0/l Œ.1; 0/l D Œ.0; 0/l C Œ.0; 1/l

D Œ.0 C 0; 0 C 1/l
D Œ.0; 1/l I

Œ.1; 0/l Œ.0; 0/l D Œ.1; 0/l C Œ.0; 1/l

D Œ.1 C 0; 0 C 0/l
D Œ.1; 0/l

where Œ.0; 1/l 6D Œ.1; 0/l because M C Q D 0 C 0 6D 1 C 1 D N C P.

4.87 Kunen’s multiplication of pairs of natural numbers distributes over subtraction.

X  .Y Z/ D X  ŒY C . Z/ D .X  Y/ C ŒX  . Z/ D .X  Y/ C Œ .X  Z/

D .X  Y/ .X  Z/:

4.89 Subtraction does not distribute over addition.

1 .1 C 1/ D 1 2 D 1;
.1 1/ C .1 1/ D 0 C 0 D 0:

4.91 By definition 4.102, .2=3/ C .7=5/ D Œ.2  5/ C .3  7/=.3  5/ D 31=15.

 www.pdfgrip.com

366 Solutions to Some Odd-Numbered Exercises

4.93 By theorem 4.110, .7=3/ .2=5/ D .7=3/ C Œ. 2/=5 D f.7  5/ C Œ3 

. 2/g=.3  5/ D 29=15.
4.95 By definition 4.102, .2=3/  .7=5/ D .2  7/=.3  5/ D 14=15.
4.97 By definition 4.117, .2=3/.7=5/ D .2=3/.5=7/ D .25/=.37/ D 10=21.
4.99 By exercise 4.97, .2=3/  .7=5/ D 10=21. Similarly, .7=5/  .2=3/ D .7=5/ 
.3=2/ D .7  3/=.5  2/ D 21=10. However, 10=21 6D 21=10, because, according to
the relation in definition 557, .10; 21/ 6 .21; 10/ since 10  10 6D 21  21 in N.
4.101 .2=3/  Œ.7=5/ C .1=1/ D .2=3/  fŒ.7  1/ C .1  5/=.5  1/g D .2=3/ 
.12=5/ D .2=3/  .5=12/ D .2  5/=.3  12/ D 5=18. However, Œ.2=3/  .7=5/ C
Œ.2=3/  .1=1/ D .21=10/ C .2=3/ D Œ.21  3/ C .10  2/=.10  3/ D 83=30, and
83=30 6D 5=18 because 83  18 6D 30  5.
4.103 .1=1/ C Œ.2=3/  .7=5/ D .1=1/ C .10=21/ D Œ.1  21/ C .1  10/=.1 
21/ D 31=21. However, Œ.1=1/ C .2=3/  Œ.1=1/ C .7=5/ D .5=3/  .12=5/ D
.5=3/  .5=12/ D 25=36, and 25=36 6D 31=21 because 25  21 6D 36  31.
4.105 If 0 < .I=J/ and 0 < .P=Q/, then 0 < I  J and 0 < P  Q. Also, .I=J/ C
.P=Q/ D .I  Q C J  P/=.J  Q/, with

.I  Q C J  P/  .J  Q/ D .I  J/  .Q  Q/ C .J  J/  .P  Q/ > 0

because Q  Q > 0 and J  J > 0. Hence 0 < Œ.I=J/ C .P=Q/.

4.107 .P=1/=.Q=R/ D .P=1/  .R=Q/ D .P  R/=.1  Q/ D .P  R/=Q.
4.109 If there exists M 2 N such that K=L D M=1, then N D M. In the alternative,
by theorem 4.133, for each K=L 2 Q, there exists N 2 N with K=L < N=1.
Consequently, the set S WD fN 2 N W K=L < N=1g is not empty. By the Well-
Ordering Principle, S contains a smallest element, denoted here by N.
4.111 By example 4.135 with N WD ¿ it follows that #.¿/ D ¿ D 0.
Alternatively, #.¿/ D 0 because of the bijection ¿ W ¿ ! ¿ and ¿ D 0.
˚ 
4.113 By example 3.24, P.f¿g/ D ¿; f¿g D 2. By definition 4.140 with N WD
2, it follows that #.2/ D 2.
 
4.115 From # PŒP.¿/ D 2 follows #ŒP.PŒP.¿// D 4 whence
#fPŒP.PŒP
.¿//g D 16.
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 367

  
Alternatively, # P PfPŒP.¿/g D 16 because
 
P PfPŒP.¿/g

D P hn
PfPŒf¿gg
˚  ˚ oi
D P ¿; f¿g; f¿g ; ¿; f¿g

D ¿;
˚  n˚ o n˚ o
f¿g; f¿g ; f¿g ; ¿; f¿g ;
˚  n ˚ o n ˚ o
¿; f¿g ; ¿; f¿g ; ¿; ¿; f¿g ;
n ˚ o n ˚  o n˚  ˚ o
f¿g; f¿g ; f¿g; ¿; f¿g ; f¿g ; ¿; f¿g ;
n ˚ o n ˚ o
¿; f¿g; f¿g ; ¿; f¿g; ¿; f¿g ;
n ˚  ˚ o n ˚  ˚ o
¿; f¿g ; ¿; f¿g ; f¿g; f¿g ; ¿; f¿g ;
 f0; 1; 2; 3; 4; 5; 6; 7; 8; 9; 10; 11; 12; 13; 14; 15g:

4.117 No, not ˚all ordered pairs

 need to have the same cardinality. If X 6D Y,
then .X; Y/˚ D fXg, fX; Yg˚ with fXg
 6D˚ fX;Yg. In contrast, if X D Y, then
.X; X/ D fXg; fX; Xg D fXg; fXg D fXg . There exist only two functions
from .X; X/ to .X; Y/, neither of which is a bijection: F W fXg 7! fXg, and
G W fXg 7! fX; Yg. Consequently, there does not exist any bijection from .X; X/
to .X; Y/, whence .X; X/ and .X; Y/ do not have the same cardinality.
4.119 By theorem 4.150 and by associativity of , #Œ.A B/ C D Œ#.A B/ 
Œ#.C/ D fŒ#.A/Œ#.B/gŒ#.C/ D Œ#.A/fŒ#.B/Œ#.C/g D Œ#.A/Œ#.B C/ D
#ŒA .B C/.
4.121 Apply contraposition to theorem 4.146.
4.123 If A is denumerable, then there is a bijection F W N ! A. Consequently, the
function G N ! A defined by G.N/ WD F.N C 1/ is also a bijection.
If moreover X … A, then H W f0g ! fXg is also a bijection.
Thus, G [ H is a bijection .G [ H/ W .f0g [ N/ ! .fXg [ A/. Hence A [ fXg is
denumerable.
Now proceed by induction with the number of elements in B.
4.125 Use theorem 4.160 and .A [ B/ D .A n B/ [ .B n A/.
4.127 Apply theorem 4.156 and exercise 4.126.
4.129 Apply theorems 4.156 and 4.167.
4.131 Any non-surjective injection F W X ,! Y induces a non-surjective injection
G W 2X ,! 2Y by restriction from Y to X.
4.133 #.2X / D 2#.X/ > #.X/ by induction with #.X/ 2 N.
 www.pdfgrip.com

368 Solutions to Some Odd-Numbered Exercises

4.135 If A is denumerable, then there exists a bijection F W A ,! N, which restricts

to an injection on each subset S  A.
4.137 Apply theorem 4.166 twice.
4.139 The existence of every function FI relies on the Axiom of Choice.

Exercises from Chapter 5

5.1 For each X 2 F let F.X/ be the smallest element of X.

5.3 The set Z is nonempty but has no smallest element relative to < or .
5.5 With M WD 2, define Œ02  Œ12 . Let ŒIM WD Œ12 , ŒKM WD Œ02 , ŒLM WD Œ12 .
Then Œ02  Œ12 but Œ12 C Œ02 6 Œ12 C Œ12 .
5.7 For each countable set C there exists an injection F W C ! N. Define a relation
R on C by XRY if and only if F.X/ F.Y/.
5.9 If W does not contain a last element,

:f9Z8YŒ.Y  Z/ _ .Y D Z/g

whence

8Z9YfŒ:.Y  Z/ ^ .Y 6D Z/:g

In other words, for each Z 2 W there exists Y 2 W such that :.Y  Z/ and Y 6D Z.
In particular, Y … WZ . In yet other words, there does not exist any Z 2 W such that
W D WZ . Because  totally orders W, however, it follows
S that Z  Y must hold.
Let G consists of all the initial intervals in W. Then SG D W, but there does not
exist any Z 2 W such that W D WZ . Consequently, G is not an initial interval
of W.
5.15 There is a transitive set on which 2 is not a transitive relation. For example,
the set
n ˚ o
A WD ¿; f¿g; f¿g

is transitive, because every element of A is also

˚ a subset of A. Nevertheless, the
relation 2 is not transitive on A, because ¿ … f¿g , even though
˚ 
¿ 2 f¿g; f¿g 2 f¿g :
 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 369

Exercises from Chapter 6

F D fA0 ; : : : ; AN1 g where each

6.1 Each finite set of pairwise disjoint sets Q
element Aj has exactly Nj elements has exactly N1
kD0 .Nk / choice functions.

6.3 Theorem 6.6 shows that each finite set of nonempty sets has a choice function.
The proof of theorem 6.20 shows that each choice function corresponds to a choice
set.
6.5 The Choice-Set Principle 6.17 translates into formula (1):

8F f8A Œ.A 2 F / ) .A 6D ¿/g (1)

!
n h [   io
) 9S S  F ^ 8A f.A 2 F / ) Œ9X.S \ A D fXg/g :

S
6.7 For each set F of nonempty sets, define a relation R  F F by R WD
f.A; X/ W .A 2 F / ^ .X 2 A/g. If the Choice-Relation Principle 6.15 holds, then
there exists a function F  R with the same domain as that of R. Hence F is a choice
function. Thus the Choice-Function Principle 6.8 holds.
Conversely, for each relation R  A B, and for each set X in the domain of R,
define the vertical section of R at X by RX WD fY 2 B W .X; Y/ 2 Rg. In particular,
RX 6D ¿ because X is in the domain of R. If the
SChoice-Function Principle 6.8 holds,
then there exists a choice function C W F ! F for F WD fRX W 9YŒ.X; Y/ 2 Rg.
Thus C.RX / 2 RX for every X. Hence parametrize the vertical sections of R by
S W A ! P.B/ with S.X/ WD RX , and set F WD C ı S. Thus F is a function,
with F.X/ D CŒS.X/ D C.RX / 2 RX , whence F  R. Thus the Choice-Relation
Principle 6.15 holds.
6.9 In every partially ordered set A the empty subset ¿  A is a chain. If the
hypothesis of Zorn’s Maximal-Element Principle 6.32 holds, then A contains an
upper bound for ¿. Hence A 6D ¿ [88, p. 118].
6.11 Zorn’s Maximal-Set Principle 6.34 translates into formula (3):

˚ ˚ 
8F .F 6D ¿/ ^ 8G ŒG 2 P.F / ^ 8A8B fŒ.A 2 G / ^ .B 2 G /

 [ 
) Œ.A  B/ _ .B  A/ ) G 2F (2)
!
˚  
) 9S .S 2 F / ^ f8A Œ.A 2 F / ) .S 6 A/g :
 www.pdfgrip.com

370 Solutions to Some Odd-Numbered Exercises

6.15 The Multiplicative Principle 6.36 is formula (3):

n  
8F 8I 8I .I 6D ¿/ ^ .I W I ! F / ^ 8i f.i 2 I / ) ŒI.i/ 6D ¿g
(3)
Q  o
) i2I I.i/ D
6 ¿ :

6.17
(T.A) ¿ 2 T by example 6.50; S
(T.B) if A  T is linearly ordered by inclusion, and if U; W 2 A , then there
exist intervals H; K 2 A such that U 2 H and W 2 K, with H  K or K  H.
case, if H SK, then U; W 2 K. If also V 2 N and U V W, then
In the firstS
V 2 K  A ; Thus A 2 T ; the second case is similar;
(T.C) if A 2 T , then either A D ¿ and A [ fF.A/g D f0g 2 T , or A 6D ¿ and
A [ fF .A/g D A 2 T , because F.A/ D min.A/ 2 A.
6.19
(T.A) ¿ 2 P.E/; S S
(T.B) if A  P.E/ then A  E whence A 2 P.E/;
(T.C) if A 2 P.E/, then A [ fF.A/g 2 P.E/, because F.A/ 2 E.
6.21 For each denumerable element D` of a denumerable family F D fDk k 2 Ng
there exists a bijection F W N ! D` . Thus the set B` of all such bijections is not
empty. Let B WD fB` ` 2 Ng. In the Zermelo-Frænkel-Choice
S set theory, there
exists a family choice-function F W N ! B such that F` WD F.`/ 2 B` for S each
` 2 N; thus each F` W N ! D` is a bijection. Hence the function N N ! B
with .k; `/ 7! F` .k/ is a bijection.

Exercises from Chapter 7

7.1 There are no dominant pure strategies in The Battle of the Sexes.
7.3 The two positions where they both go to the same show are Nash equilibria.
7.5 There are two-person games with a Nash equilibrium but without any dominant
strategy and hence without dominant strategy equilibrium, for instance, The Battle
of the Sexes (exercises 7.1 and 7.3).
7.7 For all x 2 A and y 2 B,

minff .x; y/ W y 2 Bg f .x; y/ maxff .x; y/ W x 2 Ag:

Consequently,

minff .x; y/ W y 2 Bg maxff .x; y/ W x 2 Ag;

 www.pdfgrip.com

Solutions to Some Odd-Numbered Exercises 371

where the left-hand side depends only on x while the right-hand side depends only
on y. Therefore,

maxfminff .x; y/ W y 2 Bg W x 2 Ag minfmaxff .x; y/ W x 2 Ag W y 2 Bg:

7.9 For each strategy SAl available to Al, Al identifies the worst payoff pAl
SAl ;SBo under
every strategy SBo available to Bo:

SAl ;SBo W SBo 2 Bo’s strategiesg:

minfpAl

To avoid the worst of the worst, Al plays the strategy SAl that returns the best among
the worst payoffs:

SAl ;SBo W SBo 2 Bo’s strategiesg W SAl 2 Al’s strategiesg:

payoff to Al D maxfminfpAl

7.11 If a third player Ci has a single weakly dominant strategy, then Ci will play
that strategy, leaving Al and Bo with a game for two players and any number of pure
strategies, which need not have any Nash equilibrium.
7.13 To avoid failure, the Blue commander must play Middle. Thus the Red
commander knows that Blue plays Middle; to avoid failure, Red must then play
either Left or Right. In either case, (M,L) or (M,R), both commanders get a Draw.

7.15 Each of the N 1 girls who rejected him did so because she received and holds
a better proposal. In particular, the last girl who rejected him received at least two
proposals. Consequently, there is still at least one girl c who has not yet received
any proposal. In the next round, b must then propose to c, which terminates the
algorithm, because all the other girls have received at least one proposal (from b)
and therefore hold one.
7.17 Depending on the beggars’ and choosers’ rankings, infinite sets need not admit
of any total stable relations. Let the beggars and choosers consist of all negative and
positive integers, indexed by their values:

B WD Z D f: : : ; 3; 2; 1gI C WD ZC D f1; 2; 3; : : :gI

bm WD mI cn WD n:

Because beggar bm begs for m, every chooser c ranks all beggars with the same
well-order bk c b` if and only if k < `. In contrast, beggar bm ranks choosers
by swapping 1; : : : ; 2m with 2mC1; : : : ; 4m, and otherwise leaving the natural order
on ZC unchanged:

2m C 1; : : : ; 4m; 1; : : : ; 2m; 4m C 1; 4m C 2; : : :
 www.pdfgrip.com

372 Solutions to Some Odd-Numbered Exercises

With such rankings, there are no total stable relations. The proof proceeds by
contraposition, showing that every total relation is unstable. Indeed, if a relation
M  B C is total, then there exists k 2 ZC with .bk ; c1 / 2 M, preceded by
k 1  0 couples

.b1 ; c`1 /; : : : ; .b1k ; c`1k /; .bk ; c1 /; : : : .bn ; c`n /; : : :

Yet bk prefers 2k > k 1 choosers c2kC1 ; : : : ; c4k to c1 . Hence there exists n > k
such that .bn ; c`n / 2 M with 2k C 1 `n 4k, whence .bn ; c`n / ‰ .bk ; c1 /, so
that M is unstable.
7.19 Extend algorithm 7.31 (with a match maker) to stable relations that may relate
each chooser to more than one beggar, with a quota qc of beggars for chooser c
(polygamy, polyandry, schools admitting more than one student, etc.). See Gale &
Shapley’s references [40, 41].
7.21 For all .b1 ; c1 /; .b2 ; c2 / 2 UT there exist stable relations M1 ; M2 2 T with
.b1 ; c1 / 2 M1 and .b2 ; c2 / 2 M2 . Yet M1  M2 or M2  M1 because T is a chain.
Thus .b1 ; c1 /; .b2 ; c2 / 2 M WD maxfM1 ; M2 g 2 T , but M is stable, so that none of
the three conditions in definition 7.27 can hold. Therefore UT is stable.

7.25 From .X; X/ 2 R ı1 if and only if .X; X/ 2 R follows .X; X/ … R n R ı1 .
7.27 If .X; Y/; .Y; Z/ 2 PR D R n R ı1  R, then .X; Z/ 2 R, by transitivity
of R.
If also .X; Z/ 2 R ı1 , then .Z; X/ 2 R, but .X; Y/ 2 R by hypothesis, whence
.Z; Y/ 2 R by transitivity of R, so that .Y; Z/ 2 R ı1 , contradicting the hypothesis
that .Y; Z/ … R ı1 .
7.29 Combine the preceding solutions.
7.31 With at least two voters Val and Vic and at least two alternatives Al and Bo,
there is a voters’ profile r such that Val prefers Al to Bo while Vic prefers Bo to Al:

r.Val/ D f.Al; Bo/gI r.Vic/ D f.Bo; Al/g:

If  is a permutation that swaps Val and Vic, then

.r ı  /.Val/ D rŒ .Val/ D r.Vic/ D f.Bo; Al/gI

.r ı  /.Vic/ D rŒ .Vic/ D r.Val/ D f.Al; Bo/g:

If Val is a dictator, so that F .s/ D s.Val/ for every voters’ profile s, then F .r/ D
r.Val/ D f.Al; Bo/g while F .r ı  / D .r ı  /.Val/ D f.Bo; Al/g. Thus F .r/ 6D
F .r ı  /, so that F is not invariant under permutations.
7.33 Weak unanimity implies unanimity.
 www.pdfgrip.com

References

1. L.V. Ahlfors, Complex Analysis, 3rd edn. (McGraw-Hill, New York, 1979). ISBN 0-07-
000657-1. QA331.A45 1979. LCCC No. 78-17078. 515’.93; MR0510197 (80c:30001)
2. Anonymous (ed.), ‘Three-parent’ therapy approved. Science 347(6222), 592 (2015). http://
scim.ag/_mtvote
3. V.I. Arnold, From Hilbert’s superposition problem to dynamical systems. Am. Math. Mon.
111(7), 608–624 (2004). ISSN 0002-9890
4. K.J. Arrow, A difficulty in the concept of social welfare. J. Polit. Econ. 58(4), 328–346 (1950)
5. I. Asimov, The Gods Themselves (Spectra, New York, 1990). ISBN-10: 0553288105. ISBN-
13: 978-0553288100
6. S. Barberá, Pivotal voters: a new proof of Arrow’s theorem. Econ. Lett. 6(1), 13–16 (1980)
7. S. Becattini1, D. Latorre1, F. Mele1, M. Foglierini1, C. De Gregorio1, A. Cassotta1, B.
Fernandez1, S. Kelderman, T.N. Schumacher, D. Corti1, A. Lanzavecchia1, F. Sallusto1,
Functional heterogeneity of human memory cd4+ t cell clones primed by pathogens or
vaccines. Science 347(6220), 400–406 (2015). doi:10.1126/science.1260668. http://www.
sciencemag.org/content/347/6220/400.long
8. P. Bernays, Axiomatic Set Theory with a Historical Introduction by Abraham A. Fraenkel
(Dover, Mineola, 1991). ISBN 0-486-66637-9. QA248.B47 1991. LCCC No. 90-25812.
511.3’22–dc20
9. G.D.W. Berry, Peirce’s contributions to the logic of statements and quantifiers, in Studies
in the Philosophy of Charles Sanders Peirce: First Series, ed. by P.P. Wiener, F.H. Young
(Harvard University Press, Cambridge, 1952), pp. 153–165
10. B. Bolzano, Paradoxien des Unendlichen, in Essays on the Theory of Numbers, ed. by R.
Dedekind (Dover, Leipzig, 1851)
11. R.T. Boute, How to calculate proofs: bridging the cultural divide. Not. Am. Math. Soc. 60(2),
173–191 (2013)
12. S.J. Brams, P.D. Straffin Jr., Prisoner’s dilemma and professional sports drafts. Am. Math.
Mon. 86(2), 80–88 (1979)
13. S.J. Brams, M.D. Davis, P.D. Straffin Jr., Module 311: Geometry of the Arm Race (COMAP,
Lexington, 1978)
14. W.L. Briggs, V.E. Henson, The DFT: An Owner’s Manual for the Discrete Fourier Transform
(Society for Industrial and Applied Mathematics, Philadelphia, 1995). ISBN 0-89871-342-0;
QA403.5.B75 1995; 95-3232; 515’.723–dc20
15. D.R. Chalice, How to teach a class by the modified Moore method. Am. Math. Mon. 102(4),
317–321 (1995)

© Springer Science+Business Media New York 2015 373

Y. Nievergelt, Logic, Mathematics, and Computer Science,
DOI 10.1007/978-1-4939-3223-8
 www.pdfgrip.com

374 References

16. A. Church, Correction to a note on the Entscheidungsproblem. J. Symb. Log. 1(3), 101–102
(1936)
17. A. Church, A note on the Entscheidungsproblem. J. Symb. Log. 1(1), 40–41 (1936)
18. A. Church, Introduction to Mathematical Logic. Princeton Landmarks in Mathematics and
Physics (Princeton University Press, Princeton, 1996). Tenth printing. ISBN 0-691-02906-7
19. P.J. Cohen, The independence of the continuum hypothesis. Proc. Natl. Acad. Sci. U. S. A.
50(6), 1143–1148 (1963)
20. P.J. Cohen, The independence of the continuum hypothesis, II. Proc. Natl. Acad. Sci. U. S. A.
51(1), 105–110 (1964)
21. P.J. Cohen, Set Theory and the Continuum Hypothesis. (W. A. Benjamin, New York, 1966)
QA9.C69
22. P.J. Cohen, Set Theory and the Continuum Hypothesis (Dover, Mineola, 2008). ISBN-10 0-
486-46921-2; QA248.C614 2008; LCCC No. 2008042847 511.3’22—dc22
23. L. Dadda, Some schemes for parallel multipliers. Alta Frequenza 34(2), 349–356 (1965)
24. C. Davis, The role of the untrue in mathematics. Math. Intell. 31(3), 4–8 (2009). Text of a talk
presented at the Joint Mathematics Meeting in Toronto on Monday 5 January 2009
25. R. Dedekind, Essays on the Theory of Numbers (Dover, New York, 1963). ISBN 0-486-21010-
3; 63-3681. (Also Open Court Publishing, Chicago, 1901; QA248.D3)
26. M. Dresher, Games of Strategy: Theory and Applications. Prentice-Hall Applied Mathematics
Series (Prentice-Hall Inc., Englewood Cliffs, 1961)
27. J.L.E. Dreyer, A History of Astronomy from Thales to Kepler, 2nd edn. (Dover, New York,
1953). SBN 486-60079-3; QB15.D77 1953; 53-12387
28. U. Dudley, Mathematical Cranks. MAA Spectrum (Mathematical Association of America,
Washington, DC, 1992)
29. U. Dudley, The Trisectors. MAA Spectrum, revised edition (Mathematical Association of
America, Washington, DC, 1994)
30. J. Dugundji, Topology (Allyn and Bacon, Boston, 1966). ISBN 0-205-00271-4. LCCC No.
66-10940
31. H.B. Enderton, A Mathematical Introduction to Logic (Academic Press, New York, 1972).
LCCC No. 78-182659
32. J. Evans, The History and Practice of Ancient Astronomy (Oxford University Press,
New York, 1998). ISBN 0-19-509539-1; QB16.E93 1998; LCCC No. 97-16539; 520’.
938–dc21
33. S. Feferman, Does mathematics need new axioms? Am. Math. Mon. 106(2), 99–111 (1999)
34. G.B. Folland, Real Analysis: Modern Techniques and Their Applications (Wiley, New York,
1984)
35. A. Fraenkel, Axiomatische theorie der geordneten mengen. (untersuchungen über die grund-
lagen der mengenlehre. ii.). J. Reine Angew. Math. (Crelle’s Journal) 1926(155), 129–158
(1926)
36. A. Fraenkel, Einleitung in die Mengenlehre, volume IX of Die Grundlehren der math-
ematischen Wissenschaften in Einzerldarstellungen mit besonderer Berücksichtigung der
Anwendungsgebiete (Dover, New York, 1946). First edition published by Julius (Springer,
Berlin, 1919)
37. A. Fraenkel, Abstract Set Theory (North-Holland, Amsterdam, 1953)
38. G. Frege, Begriffsschrift, eine der arithmetischen nachgebildete Formelsprache des reinen
Denkens (Nebert, Halle, 1879)
39. G. Frege, Conceptual Notation, and Related Articles; Translated [from the German] and
Edited with a Biography and Introduction by Terrell Ward Bynum (Clarendon Press, Oxford,
1972). ISBN 0198243596; B3245.F22 B94
40. D. Gale, L.S. Shapley, College admissions and the stability of marriage. Am. Math. Mon.
69(1), 9–15 (1962)
41. D. Gale, L.S. Shapley, College admissions and the stability of marriage. Am. Math. Mon.
120(5), 386–391 (2013). Reprint of MR1531503
 www.pdfgrip.com

References 375

42. J. Geanakoplos, Three brief proofs of Arrow’s impossibility theorem. Econ. Theory 26(1),
211–215 (2005)
43. K. Gödel, Über formal unentscheidbare Sätze der Principia Mathematica und verwandter
Systeme I. Mon. Math. Phys. 38, 173–198 (1931) Leipzig
44. K. Gödel, The consistency of the axiom of choice and of the generalized continuum
hypothesis. Proc. Natl. Acad. Sci. U. S. A. 24, 556–557 (1938)
45. K. Gödel, The Consistency of the Axiom of Choice and of the Generalized Continuum
Hypothesis with the Axioms of Set Theory. Annals of Mathematics Studies, vol. 3 (Princeton
University Press, Princeton, 1940). QA9.G54
46. K. Gödel, On formally Undecidable Propositions of Principia Mathematica and Related
Systems (Dover, Mineola, 1992). ISBN 0-486-66980-7. QA248.G573 1992. LCCC No.
91-45947. 511.3–dc20
47. G. Gonthier, Formal proof—the four-color theorem. Not. Am. Math. Soc. 55(11), 1382–1393
(2008)
48. C. Gram, O. Hestvik, H. Isaksson, P.T. Jacobsen, J. Jensen, P. Naur, B.S. Petersen, B. Svej-
gaard, GIER—a Danish computer of medium size. IEEE Trans. Electron. Comput. EC-12(5),
629–650 (1963)
49. W. Grassmann, J.-P. Tremblay, Logic and Discrete Mathematics: A Computer Science Per-
spective. (Prentice Hall, Upper Saddle River, 1996). ISBN 0-13-501296-6; QA76.9.M35G725
1996; LCCC No. 95-38351; 005.1’01’5113–dc20
50. T.C. Hales, Formal proof. Not. Am. Math. Soc. 55(11), 1370–1380 (2008)
51. E.C. Hall, MIT’s role in project Apollo: final report on contracts NAS 9-163 and NAS 94065.
Technical report R-700 (Charles Stark Draper Laboratory, MIT, Cambridge) (1972)
52. P. Hamburger, R.E. Pippert, Venn said it couldn’t be done. Math. Mag. 73(2), 105–110 (2000)
53. P. Hamburger, R.E. Pippert, A symmetrical beauty: a non-simple 7-Venn diagram with a
minimum vertex set. Ars Combinatoria 66, 129–137 (2003)
54. J. Harrison, Formal proof—theory and practice. Not. Am. Math. Soc. 55(11), 1395–1406
(2008)
55. J.R. Harrison, Handbook of Practical Logic and Automated Reasoning (Cambridge University
Press, London, 2009). ISBN-13: 9780521899574
56. F. Hausdorff, Bemerkung §ber den Inhalt von Punktmengen. Math. Ann. 75(3), 428–433
(1914)
57. F. Hausdorff, Set Theory, 2nd edn. (Chelsea, New York, 1962). Translated from the German
by John R. Aumann, et al. QA248.H353 1962; LCCC No. 62-19176
58. M. Henle, Which Numbers Are Real? Classroom Resource Materials Series (Mathematical
Association of America, Washington, DC, 2012)
59. P. Henrici, Essentials of Numerical Analysis with Pocket Calculator Demonstrations (Wiley,
New York, 1982). ISBN 0-471-05904-8; QA297.H42; 81-10468; 519.4
60. E. Hewitt, K. Stromberg, Real and Abstract Analysis (Springer, New York, 1965). LCCC No.
65-26609
61. D. Hilbert, Foundations of Geometry, 10th edn. (Open Court, La Salle, 1971). Eighth printing,
1996. Translated by Leo Unger and revised by Paul Bernays. ISBN 0-87548-164-7; LCCC
No. 73-110344
62. D. Hilbert, W. Ackermann, Principles of Mathematical Logic (Chelsea Publishing Company,
New York, 1950). BC135.H514. (Translation of David Hilbert & Wilhelm Ackermann,
Grundzüge der theoretischen Logik, Julius Springer, Berlin, Germany, 1928 & 1938)
63. T.W. Hungerford, Algebra (Holt, Rinehart and Winston, New York, 1974). ISBN 0-03-
086078-4; QA155.H83; LCCC No. 73-15693; 512
64. International Business Machines Corporation, 590 Madison Avenue, New York 22, NY. IBM
604 Electronic Calculating Punch Manual of Operation, 1954. http://www.bitsavers.org/pdf/
ibm/604
65. International Business Machines Corporation, 590 Madison Avenue, New York 22, NY. IBM
650 Magnetic-Drum Data Processing Machine Manual of Operation, revised edition, June
1955. http://www.bitsavers.org/pdf/ibm/650
 www.pdfgrip.com

376 References

66. T.J. Jech, The Axiom of Choice, II. Studies in Logic and the Foundations of Mathematics,
vol. 75 (North Holland, Amsterdam, 1973). ISBN 0-444-10484-4; QA248.J4 1973; LCCC
No. 73-75535
67. M. Kaku, Venus envy. Wall Street J. CCXLVIII(40), A8 (2006). Western Edition
68. E. Kalai, Foreword: the high priest of game theory. Am. Math. Mon. 120(5), 384–385 (2013)
69. H. Karttunen, P. Kröger, H. Oja, M. Poutanen, K.J. Donner (eds.), Fundamental Astronomy
(Springer, New York, 1994). ISBN 0-387-57203; QB43.2.T2613 1993; 93-31098; 520–dc20
(second enlarged edition)
70. L. Keen, Julia sets, in Chaos and Fractals: The Mathematics Behind the Computer Graphics
(American Mathematical Society, Providence, 1989), pp. 57–74
71. C.T. Kelley, in Iterative Methods for Linear and Nonlinear Equations. Frontiers in Applied
Mathematics, vol. 16 (Society for Industrial and Applied Mathematics, Philadelphia, 1995).
0-89871-352-8; QA297.8.K45 1995; 95-32249; 519.4–dc20
72. S.C. Kleene, Mathematical Logic (Dover, Mineola, 2002). ISBN 0-486-42533-9; QA9.A1
K54 2002; LCCC No. 2002034823 (originally published by Wiley in 1967)
73. S. Kortenkamp, Why Isn’t Pluto a Planet?: A Book about Planets (Capstone Press, Mankato,
2007). ISBN 0-7368-6753-8; QB701.K57 2007; LCCC No. 2006025648; 523.4–dc22
74. K. Kunen, Set Theory: An Introduction to Independence Proofs (North Holland, Amsterdam,
1980). ISBN 0-444-85401-0. QA248.K75 1980. LCCC No. 80-20375. 510.3’22
75. E. Lakdawalla, Pluto and the Kuiper belt. Sky Telesc. 127(2), 18–25 (2014)
76. E. Landau, Foundations of Analysis (The Tan Chiang Book Co., 1951)
77. S. Lang, Algebra (Addison-Wesley, Reading, 1965). LCCC No. 65-23677
78. L.S. LaPiana, F.H. Bauer, Mars Climate Orbiter mishap investigation board phase I report.
Technical report, National Aeronautics and Space Administration, 10 Nov 1999. http://sse.
jpl.nasa.gov.news
79. J. Łukasiewicz, Zur vollen dreiwertigen Aussagenlogik. Erkenntniss 5, 176 (1935)
80. J. Łukasiewicz, Zur Geschichte der Aussagenlogik. Erkenntniss 5, 111–131 (1935)
81. Y.I. Manin, A Course in Mathematical Logic (Springer, New York, 1977). ISBN 0-387-90243-
0; QA9.M296; LCCC No. 77-1838; 511’.3
82. M.F. Mann, R.R. Rathbone, J.B. Bennett, Whirlwind I Operation Logic, vol. 29 (Massachus-
setts Institute of Technology, Digital Computer Laboratory, Cambridge, 1954). http://www.
bitsavers.org/pdf/mit/whirlwind
83. V.W. Marek, J. Mycielski, Foundations of mathematics in the twentieth century. Am. Math.
Mon. 108(5), 449–468 (2001). ISSN 0002-9890
84. A. Margaris, First Order Mathematical Logic (Dover Publications Inc., New York, 1990).
Corrected reprint of the 1967 edition
85. B. Mates, Elementary Logic, 2nd edn. (Oxford University Press, New York, 1972).
BC135.M37 1972. LCCC No. 74-166004
86. R. Miller, Computable fields and Galois theory. Not. Math. Am. Soc. 55(7), 798–807 (2008).
http://www.ams.org/notices
87. K. Mitchell, J. Ryan, Game theory and models of animal behavior, in UMAP/ILAP Modules
2002–03: Tools for Teaching, ed. by P.J. Campbell (COMAP, Bedford, 2003), pp. 1–48.
Reprinted as Module 783, COMAP, Bedford, MA, 2003
88. J.D. Monk, Introduction to Set Theory (McGraw-Hill, New York, 1969). LCCC No. 68-20056
89. J.D. Monk, Mathematical Logic (Springer, New York, 1976). ISBN 0-387-90170-1.
QA9.M68
90. M.R. Mugnier, G.A.M. Cross, F.N. Papavasiliou, The in vivo dynamics of
antigenic variation in Trypanosoma brucei. Science 347(6229), 1470–1473 (2015).
doi:10.1126/science.aaa4502. http://www.sciencemag.org/content/347/6229/1470.full
91. J.R. Munkres, Topology, 2nd edn. (Prentice-Hall, Englewood Cliffs, 2000). ISBN 0-13-
181629-2; QA611.M82 2000; 514–dc21; LCCC No. 99-052942
92. R.B. Myerson, An Introduction to Game Theory. Studies in Mathematical Economics. MAA
Studies in Mathematics, vol. 25 (Mathematical Association of America, Washington, DC,
1986), pp. 1–61
 www.pdfgrip.com

References 377

93. J. Nash, Non-cooperative games. Ann. Math. Second Ser. 54(2), 286–295 (1951)
94. J.F. Nash Jr., The bargaining problem. Econometrica 18(2), 155–162 (1950)
95. J.F. Nash Jr., Equilibrium points in n-person games. Proc. Natl. Acad. Sci. U. S. A. 36(1),
48–49 (1950)
96. National Resident Matching Program, http://www.nrmp.org/match-process/match-algorithm/
97. O. Neugebauer, The Exact Sciences in Antiquity, 2nd edn. (Dover, New York, 1969).
SBN 486-22332-0; LCCC No. 69-20421. Reprint of the 1957 second edition from Brown
University Press
98. O. Neugebauer, A History of Ancient Mathematical Astronomy. Part I. Studies in the History
of Mathematics and Physical Sciences, vol. 1 (Springer, New York, 1975)
99. O. Neugebauer, A History of Ancient Mathematical Astronomy. Part II. Studies in the History
of Mathematics and Physical Sciences, vol. 1 (Springer, New York, 1975)
100. O. Neugebauer, A History of Ancient Mathematical Astronomy. Part III. Studies in the History
of Mathematics and Physical Sciences, vol. 1 (Springer, New York, 1975)
101. Y. Nievergelt, The truth table of the logical implication. Math. Gaz. 94(531), 509–513 (2010).
ISSN 0025-5572
102. Y. Nievergelt, H. Sullivan, Undecidability in fuzzy logic. UMAP J. 31(4), 323–359 (2010).
Also reprinted as UMAP Unit 804
103. Packard Bell Electronics, Packard Bell Computer, 1905 Armacost Avenue, Los
Angeles 25, California. pb 250 Technical Manual, Volume 1, 15 July 1961.
http://www.bitsavers.org/pdf/packardBell
104. C.S. Peirce, On the algebra of logic: a contribution to the philosophy of notation. Am. J. Math.
7(2), 180–196 (1885)
105. F. Quinn, A revolution in mathematics? What really happened a century ago and why it
matters today. Not. Am. Math. Soc. 59(1), 31–37 (2012)
106. A. Rapoport, Two-Person Game Theory: The Essential Ideas (The University of Michigan
Press, Ann Arbor, 1966)
107. A. Rapoport, Two-Person Game Theory (Dover Publications Inc., Mineola, 1999). Reprint of
the 1966 original [Univ. Michigan Press, Ann Arbor, MI, 1966; MR0210463 (35 #1356)]
108. J.W. Robbin, Mathematical Logic: A First Course (Dover, Mineola, 2006). SBN 486-61272-
4; LCCC No. 65-12253 (originally published by W. A. Benjamin, New York, 1969)
109. R.M. Robinson, Primitive recursive functions. Bull. Am. Math. Soc. 53, 925–942 (1947)
110. J.B. Rosser, Logic For Mathematicians (McGraw-Hill, New York, 1953). BC135.R58; LCCC
No. 51-12640
111. H. Rubin, J.E. Rubin, in Equivalents of the Axiom of Choice, II. Studies in Logic and the
Foundations of Mathematics, vol. 116 (North Holland, Amsterdam, 1980). ISBN 0-444-
87708-8; QA248.R8 1985; LCCC No. 84-28692; 511.3’22
112. D.G. Saari, Chaotic Elections: A Mathematician Looks at Voting (American Mathematical
Society, Providence, 2001). ISBN 0-8218-2847-9
113. D.G. Saari, Mathematics and voting. Not. Am. Math. Soc. 55(4), 448–455 (2008). http://
www.ams.org/notices
114. E. Schechter, Classical and Nonclassical Logics (Princeton University Press, Princeton 2005).
ISBN 0-691-12279-2; QA9.3.S39 2005; 2004066030; 160–dc22
115. R. Serrano, The theory of implementation of social choice rules. SIAM Rev. 46(3), 377–414
(2004) (electronic)
116. T.A. Skolem, Einige Bemerkungen zur axiomatischen Begrundung der Mengenlehre, in
Matematikerkongressen i Helsingfors den 4–7 Juli 1922, Den femte skandinaviska matem-
atikercongressen, Redogšrelse (Helsinki, SF, 1923), pp. 217–237. Akademiska Bokhandeln.
Cited in [36] and [128]
117. R.M. Smullyan, First-Order Logic (Dover, Mineola, 1995). ISBN 0-486-68370-2. QA9.S57
1994. LCCC No. 94-39736. 511.3–dc20
118. M. Spivak, Calculus, 2nd edn. (Publish or Perish, Wilmington, DE, 1980). ISBN 0-914098-
77-2. LCCC No. 80-82517
 www.pdfgrip.com

378 References

119. S. Stahl, A Gentle Introduction to Game Theory. Mathematical World, vol. 13 (American
Mathematical Society, Providence, 1999)
120. J. Stillwell, The Real Numbers. Undergraduate Texts in Mathematics. An Introduction to Set
Theory and Analysis (Springer, Cham, 2013)
121. J. Stoer, R. Bulirsch, Introduction to Numerical Analysis, 2nd edn. (Springer, New York,
1993). ISBN 0-387-97878-X; QA297.S8213 1992; 92-20536; 519.4–dc20
122. R.R. Stoll, Sets, Logic, and Axiomatic Theories (W. H. Freeman, San Francisco, 1961). LCCC
No. 61-6784
123. A.A. Stolyar, Introduction to Elementary Mathematical Logic (Dover, Mineola, 1983). ISBN
0-486-64561-4. BC135.S7613 1983. LCCC No. 83-5223. 511.3
124. P.D. Straffin Jr., The prisoner’s dilemma. UMAP J. 1(1), 102–103 (1980)
125. P.D. Straffin Jr., Changing the way we think about the social world. Two Year Coll. Math. J.
14(3), 229–232 (1983)
126. P.D. Straffin Jr., Game theory and nuclear deterrence. UMAP J. 1(1), 87–92 (1989)
127. D.J. Struik, A Concise History of Mathematics, 4th edn. (Dover, New York, 1987). ISBN
0-486-60255-9; QA21.S87 1987; LCCC No. 86-8855; 510’.09 (revised edition)
128. P. Suppes, Axiomatic Set Theory (Dover, Mineola, 1972). ISBN 0-486-61630-4. LCCC No.
72-86226
129. A. Tarski, Introduction to Logic and to the Methodology of Deductive Sciences (Oxford
University Press, Oxford, 1941)
130. A.E. Taylor, Introduction to Functional Analysis (Wiley, New York, 1958) 58-12704
131. A.W. Tucker, The prisoner’s dilemma: a two-person dilemma. UMAP J. 1, 101 (1980). Dated
“Stanford, May 1950”
132. A.W. Tucker, A two-person dilemma: the prisoner’s dilemma. Two Year Coll. Math. J. 14(3),
228 (1983). Dated “Stanford, May 1950”
133. G. van Brummelen, The Mathematics of the Heavens and the Earth: The Early History
of Trigonometry (Princeton University Press, Princeton, 2009). ISBN 978-0-691-12973-0;
QA24.V36 2009; 516.2409–dc22; LCCC No. 2008032521
134. B.L. van der Waerden, Geometry and Algebra in Ancient Civilizations (Springer, Berlin,
1983). ISBN 3-540-12159-5. QA151.W34 1983. LCCC No. 83-501. 512’.009
135. J. von Neumann, Zur einführung der transfiniten Zahlen. Acta Szeged 1, 199–208 (1923)
Cited in [128, p. 129]
136. R.J. Walker, Algebraic Curves (Springer, New York, 1978). ISBN 0-387-90361-5.
QA564.W35 1978. LCCC No. 78-11956. 516’.352. (First published by Princeton University
Press, Princeton, 1950)
137. D.A. Weintraub, Is Pluto a Planet?: A Historical Journey through the Solar System (Princeton
University Press, Princeton, 2007). ISBN 0-691-12348-9; QB602.9.W45 2007; LCCC No.
2006929630
138. Westinghouse Electric Corporation, Surface Division, P.O. Box 1897, Baltimore, MD
21203. Technical Manual for DPS-2402 Computer, Volume I, 1 February 1966.
http://www.bitsavers.org/pdf/westinghouse
139. F. Wiedijk, Formal proof—getting started. Not. Am. Math. Soc. 55(11), 1408–1417 (2008)
140. R.L. Wilder, Introduction to the Foundations of Mathematics, 2nd edn. (Dover, Mineola,
2012)
141. R.S. Wolf, A Tour Through Mathematical Logic. The Carus Mathematical Monographs, vol.
30. (Mathematical Association of America, Washington, DC, 2005). ISBN 0-88385-036-2;
LCCC No. 2004113540
142. S. Wolfram, The Mathematica Book, 3rd edn. (Wolfram Media, Champaign, 1996). ISBN
0-9650532-1-0; QA76.95.W65 1996; LCCC No. 96-7218; 510’.285’53–dc20
143. N.N. Yu, A one-shot proof of Arrow’s impossibility theorem. Econ. Theory 50(2), 523–525
(2012)
144. E. Zermelo, Beweis, daß jede Menge wohlgeordnet werden kann. Math. Ann. 59(4), 514–516,
(1904)
 www.pdfgrip.com

References 379

145. E. Zermelo, Untersuchungen über die Grundlagen der Mengenlehre, I. Math. Ann. 65(2),
261–281 (1908). Cited in [8, p. 21]
146. D.E. Zitarelli, The origin and early impact of the Moore method. Am. Math. Mon. 111(6),
465–486 (2004)
147. M. Zorn, A remark on method in transfinite algebra. Bull. Am. Math. Soc. 41(10), 667–670
(1935)
148. G. Zukav, The Dancing Wu Li Masters: An Overview of the New Physics (Morrow, New York,
1979). ISBN: 0-688-03402-0; QC173.98.Z84 1979; 530.1’2; LCCC No. 78-25827
 www.pdfgrip.com

Index

Symbols ,! (injective function), 166

.X; Y/ (ordered pair), 143 2 (set membership), 111
2X , 253 #, 64, 65
WD, 114 #, 64
< 7!, 155
on N  N, 221 AND (^), 38
on N, 205 NAND
on Q, 241 connective
on Z, 222 universal, 66
D, 112 Westinghouse DPS-2402 computer, 66
DW, 114 NOR
A  B (Cartesian product), 144 Apollo Guidance Computer (AGC), 66
Fjs (restriction of a function), 158 connective
F" (images of sets), 159 universal, 65
F ı1 (inverse function), 169 OR (_), 38
F ı1 " (inverse images of sets), 160 max, 183, 210
M J (integral power of an integer), 226 SubfXZ .P/, 79, 88
NŠ (factorial), 203 min, 183, 209
N C 1 (successor), 193 6D, 112
Rı1 (inverse relation), 152 …, 112
Y X (set of all functions from X to Y), 253  (proper subset), 115
 (diagonal), 149  (subset or equal), 114
 (symmetric difference of sets), 139 ¨ (proper subset), 115
@0 (cardinality of N), 254
T ¤ (proper subset), 115
F (intersection of sets), 132  (superset or equal), 114
\ (intersection of sets), 133 Ð (same cardinality), 244
ı ¿ (empty set), 114
composition of functions, 164 ` (derivable), 8
composition of relations, 152 _ (OR), 38
[ (union of sets), 129 j (restriction of relations), 153
Õ, 65 ^ (AND), 38
Õ, 64 N, 192
, 229 N (set of all natural numbers), 190, 192
9 (there exists), 80 N , 192
8 (for each), 80 Q (set of all rationals), 230

© Springer Science+Business Media New York 2015 381

Y. Nievergelt, Logic, Mathematics, and Computer Science,
DOI 10.1007/978-1-4939-3223-8
 www.pdfgrip.com

382 Index

R (set of all reals), 254 of multiplication

Z (set of all integers), 216, 225 of natural numbers, 203
P.A/, 122 of union, 130
Generalization, rule, 82 asymmetric, relation, 181
Łukasiewicz asymmetry, of a relation, 181
triadic logic, 57 atomic
formula, 76
axiom
A (P3) converse contraposition, 7
Absence of Dictators, 324 (P1) affirmation of the consequent, 6
absolute value (P2) self-distributivity of implication, 6
of rational numbers, 243 (Q3), 81
absurd (Q4), 81
reduction to the, 35 (Q1) specialization, 80
absurdum (Q2), 80
law of reductio ad, 35 (Q0), 80
reductio ad, 53 (S6) union
addition of a set of sets, 127
of natural numbers (S2) empty set, 114
associativity, 199 (S1) extensionality, 111
commutativity, 200 (S7) of infinity, 184
definition, 198 (S7) of infinity, 190
admit, 79 (S8) of choice, 184
AGC (Apollo Guidance Computer) (S3) pairing, 119
NOR, 66 (S4) power set, 122
algorithm (S5) separation, 124
Deferred Acceptance, 320 choice, 287
match maker’s, 318 countable, 301
Anonymity, 324 relation, 286
anti-symmetric, relation, 181 continuum, 280
anti-symmetry, of a relation, 181 Kleene’s 1a, 29, 55
Apollo Guidance Computer (AGC), 66 Kleene’s 1b, 29, 55
application Kleene’s 7, 55
logical reasoning, 2 Kleene’s 8, 55
applied regularity, 280
predicate calculus, 76 rule, 7
Archimedean, property of rationals, schema, 8
243 specialization, 80
argument, of a function, 155 Tarski’s I, 29, 55
arithmetic Tarski’s II, 29, 55
predicate, 76 Tarski’s III, 29, 55
with integers Tarski’s IV, 55, 56
Kunen’s, 217 Tarski’s V, 55, 56
Landau’s, 225 Tarski’s VI, 55, 56
with rational numbers, 232 Tarski’s VII, 55
arithmetic series, 255 axioms
Arrow’s Impossibility Theorem, (A) Margaris’s and Rosser’s, 87
327 (A) Peano’s, 264
Arrow, Kenneth J., 321 (C) Church’s, 37
assertion, law, 18, 28 (F) Frege’s, 37
associativity (P) Łukasiewicz’s, 6
of addition (Q) of the Pure Predicate Calculus, 80
of natural numbers, 199 (R) Rosser’s, 55
of intersection, 134 equality, 102
 www.pdfgrip.com

Index 383

Kleene’s, 29, 55 principle, 284

Rosser’s, 55 principle
Tarski’s, 29, 55 function, 284
set, 286
set, 286
B principle, 286
base Church
of a power, 226 triadic logic, 57
Bernays, Paul, 112 Church’s axioms (C), 37
Bernstein–Cantor–Schröder Theorem, classical, logic, 5
258 closed
bijective, function, 168 formula, 78
bijectivity, 168 co-domain, of a function, 155
Borda’s count, 323 Cohen, Paul J., 265
bound commutation, law, 27
lower, 184 commutativity
upper, 184 of addition
variable, 78 of natural numbers, 200
change, 79 of intersection, 133
substitution, 79 of multiplication
Brouwer, L. E. J., 7, 38, 41, 44 of natural numbers, 202
of union, 130
complete
C propositional logic, 61
calculus Completeness Theorem, 69
implicational composite
positive, 9 function, 164
canonical relation, 152
map, 178 composition
projection function, 157 of functions, 164
Cantor’s of relations, 152
diagonal enumeration, 256 computers
diagonalization, 253 Apollo Guidance (AGC), 66
Cantor–Bernstein–Schröder Theorem, 258 IBM
cardinality 604, 201
finite, 248 650, 201
same, 244 Westinghouse DPS-2402, 66
Cartesian Whirlwind I, 213
product, 292 conclusion
Cartesian product, 144 about Pluto, 2
cases, proof by, 53 connected, 182
chain, 183 connective
change primitive, 65
of bound variables, 79 universal, 65
character connectivity, 182
finite, 300 connectivity, strong, 182
characteristic, function, 157 constant
choice function, 156
axiom, 287 functional, 76
countable, 301 individual, 76
relation, 286 predicate, 76
axiom (S8) of, 184 relational, 76
function, 185, 284 construction
family, 285 transfinite, 269
 www.pdfgrip.com

384 Index

continuum directed graph, 149

axiom, 280 disjoint
hypothesis, 265, 280 pairwise, 135
contradiction sets, 135
proof by, 53 union, 135
contraposition distributivity
about Pluto, 3 intersection over union
law, 33 principle, 293
applied, 3 of multiplication
converse, 3 of natural numbers, 201, 203
about Pluto, 3 division
law of double negation, 33 of natural numbers
Truth table, 64 definition, 212
countable, set, 254 existence, 212
counting Whirlwind I, 213
principle, 299 domain
of a function, 155
of a relation, 149
D dominant
De Morgan strategy, 307
laws for sets, 136 double negation
decision converse law, 33
problem, 90 law, 33
Dedekind
finite
set, 301 E
infinite edge, 149
set, 301 element
Dedekind infinite sets, 262 first, 183
Dedekind, Richard, 262 largest, 183
deduction last, 183
definition, 7 maximal, 184
theorem minimal, 184
for predicates, 91 of sets, 111
Deduction Theorem, 14 smallest, 183
Deferred Acceptance Algorithm, empty set, 114
320 axiom (S2), 114
definition by equality
induction, 195 axioms, 102
recursionn, 195 reflexivity, 102, 106
transfinite substitutivity, 102, 104, 106
construction, 269 equality, of sets, 112
induction, 269 equilibrium
denial of the antecedent, law, 33 dominant strategy, 308
denominator, 230 Nash, 308
denumerable, set, 254 equivalence
derivable, theorem, 8 relation, 175
derived rule, 11 equivalence class, 177
Detachment, 7, 81 equivalent
diagonal, 149 pairs of integers, 229
diagonalization, Cantor’s, 253 pairs of natural numbers, 215
difference existential quantifier, 80
of two sets, 125 exponent
symmetric (), 139 of a power, 226
 www.pdfgrip.com

Index 385

exponentiation, 226 social welfare, 325

extensionality, axiom (S1), 111 surjective, 167
union of, 158
functional
F constant, 76
factor, of a Cartesian product, 144 fuzzy logic, 56
factorial, 203
family
choice-function, 285 G
fof sets, 285 Gale, David, 303, 315
field, of numbers, 235 Game
finite Prisoner’s Dilemma, 304
cardinality, 248 Theory, 304
character, 300 Two-Person Dilemma, 304
sequence, 192 Ganymede, 3
length, 192 generalized continuum hypothesis, 265
sequence of numbers, 237 geometric series, 239
set, 248, 300 geometry
Dedekind, 301 predicate, 76
first element, 183 Gödel, Kurt, 192, 195, 265
first order, logic, 78 graph
for each, 80 directed, 149
formula
atomic, 76
closed, 78 H
well-formed, 5, 77 Hausdorff
formulaic Maximal-Chain Principle, 299
letter, 5, 77 Maximal-Subset Principle, 299
fraction, 230 Heyting, Arend, 7, 38, 41, 44
Fraenkel, A. A., 112, 258 Hilbert
free Positive Propositional Calculus, 7, 38, 41,
for, 79 44
variable, 78 Hilbert, David, 7, 38, 41, 44
substitution, 88 holds
Frege’s axioms (F), 37 propositional form, 57
Frege, Gottlob hypothesis
axioms, 37 about Pluto, 2
function continuum, 265, 280
bijective, 168 generalized continuum, 265
canonical projection, 157
characteristic, 157
choice, 185, 284 I
composite, 164 IBM
composition, 164 604, 201
constant, 156 multiplication, 201
definition, 155 650, 201
identity, 156 multiplication, 201
inclusion, 158 ideal
injective, 166 for a preorder, 268
inverse, 169 identity
left inverse, 170 function, 156
recursive, 195 image
right inverse, 170 by a function, 155
slice, 157 of sets, 159
 www.pdfgrip.com

386 Index

implication relation, 152

about Pluto, 2 right, 170
reflexive law, 10 invertibility of functions, 169
transitive law, 21, 22 irreflexive relation, 180
implicational calculus
positive, 9
inclusion J
function, 158 Johansson, Ingebrigt, 7, 38, 41, 44
incomplete
propositional logic, 61
Independence of Irrelevant Alternatives, K
324 Kleene’s axioms, 55
independent Kleene, Stephen, 55
logically, 60 Kleene, Stephen Cole
individual axioms, 29, 55
constant, 76 Kolmogorov, A., 7, 38, 41, 44
variable, 77 Kunen, Kenneth, 192, 214, 216, 217, 267, 272,
induction 280
definition by, 195 Kuratowski
principle of, 193 Maximal-Order Principle, 299
proof by, 193
transfinite, 268
inference L
rules, 7, 81 Landau, Edmund, 225, 226, 264, 265
infinite largest element, 183
set, 252, 300 last element, 183
Dedekind, 301 law
infinite sets affirmation of the consequent, 6
Dedekind, 262 assertion, 18, 28
infinity commutation, 27
axiom (S7), 184 contraposition, 33
axiom (S7), 190 applied, 3
initial interval, 267 converse contraposition, 7
injective, function, 166 converse of double negation, 33
injectivity, 166 denial of the antecedent, 33
integer double negation, 33
arithmetic hypothetical syllogism, 22
Kunen’s, 217 reductio ad absurdum, 35
Landau’s, 225 special law, 35
definition reflexive implication, 10
Kunen’s, 216 self-distributivity of implication, 6
negative, 221 transitivity of implication, 21, 22
positive, 221 left invertibility
intersection of functions, 170
associativity, 134 lemma
commutativity, 133 Zorn’s, 185
of a set of sets, 132 length
of two sets, 125 of a finite sequence, 192
interval, 298 letter
interval, initial, 267 formulaic, 5, 77
Intuitionistic Logic, 7, 38, 41, 44 logic
inverse classical, 5
function, 169 complete
left, 170 propositional, 61
 www.pdfgrip.com

Index 387

first order, 78 minor

fuzzy, 56 premiss, 7
incomplete mixed strategy, 309, 310
propositional, 61 Modus Ponens, 7, 81
intuitionistic, 7, 38, 41, 44 Moon, 3
minimal , 7, 38, 41, 44 multiplication
propositional IBM
complete, 61 604, 201
incomplete, 61 650, 201
sound, 58 of natural numbers
sound associativity, 203
propositional, 58 commutativity, 202
logically independent, 60 definition, 200
lower bound, 184 distributivity, 201, 203
Łukasiewicz’s axioms (P), 6 Multiplicative
Łukasiewicz, Jan, 6 Principle, 292

M N
major Nash equilibrium, 308
premiss, 7 Nash, John Forbes, Jr., 304
map National Resident Matching Program, 315
canonical, 178 natural numbers, 192
quotient, 178 negative integer, 221
mapping notation
principle, 300 product, 237
match maker’s algorithm, 318 sum, 238
match making, 315 number
matrix field, 235
of a prenex normal form, 95 of elements, 248
maximal rational, 230
element, 184 real, 254
Maximal-Chain Principle triangular, 255
Hausdorff, 299 numerator, 230
Maximal-Element Principle
Zorn, 290
Maximal-Order Principle O
Kuratowski, 299 operator
Maximal-Set Principle overloading, 158
Zorn, 291 order
Maximal-Subset Principle of natural numbers
Hausdorff, 299 definition, 205
maximum, 183, 210 on Z, 221, 222
member of a set, 111 ordered pairs, 143
membership ordinal, 272
of sets, 111 overloading operators, 158
Mercury, 3
Minimal
Logic, 7, 38, 41, 44 P
minimal pairing, axiom (S3), 119
element, 184 pairwise disjoint
propositional calculus, 7, 38, set, 135
41, 44 union, 135
minimum, 183, 209 partial order, 181
 www.pdfgrip.com

388 Index

partition, 176 Kuratowski, 299

payoff, 305 mapping, 300
Peano’s axioms (A), 264 maximal
Peano, Giuseppe, 264 Hausdorff, 299
Peirce’s law, 29 Zorn, 290, 291
Peirce, Charles Sanders, 5 of induction, 193
permutation, 326 of transposition, 33
plurality voting, 322 Teichmüller-Tukey, 300
Pluto, 2 trichonomy, 300
positive integer, 221 Zermelo, 288
Positive Propositional Calculus, 7, 38, 41, 44 problem
postulates decision, 90
of set theory, 110 product
power Cartesian, 144, 292
base, 226 notation, 237
exponent, 226 profile
of an integer, 226 voters’, 325
power set, 122 proof, 82
axiom (S4), 122 definition, 7
pre-image proof by
of a subset, 160 cases, 36, 53
of an element, 160 contradiction, 53
predicate, 75 Converse Law of Contraposition, 32
arithmetic, 76 induction, 193
constant, 76 reductio ad absurdum, 35, 53
geometry, 76 transfinite induction, 268
ranking, 107 propositional
set theory, 76 calculus
predicate calculus minimal , 7, 38, 41, 44
applied, 76 variable, 5, 77
pure, 76 Provability Theorem, 67
preference, 325 pure
prefix predicate calculus, 76
of a prenex normal form, 95 pure strategy, 309
premiss
major, 7
minor, 7 Q
prenex quantifier
normal form existential, 80
Skolem’s, 96 scope, 78
prenex normal form, 95 universal, 80
matrix, 95 quasi-order, 180
prefix, 95 quotient
preorder, 180 map, 178
primitive connective, 65
Principle
Multiplicative, 292 R
principle range
choice of a function, 155
function, 284 of a relation, 149
set, 286 ranking, 324
counting, 299 predicate, 107
distributivity ratio of integers, 230
intersection over union, 293 rational number, 230
 www.pdfgrip.com

Index 389

real numbers, 254 derived, 11

recursion, 195 of Generalization, 82
definition by, 195 of Detachment, 7, 81
recursive function, 195 of inference, 7, 81
reductio ad absurdum, 53 of Modus Ponens, 7, 81
law, 35 of Substitution, 7
special, 35
reduction
to the absurd, 35 S
reflexive scope of a quantifier, 78
implication, law, 10 sentence, 78
relation, 174 symbol, 5, 77
reflexivity separation axiom (S5), 124
equality, 102, 106 sequence, 192
reflexivity of a relation, 174 finite, 192
regularity axiom, 280 length, 192
relation, 149 of numbers, 237
anti-symmetric, 181 series
asymmetric, 181 arithmetic, 255
composite, 152 geometric, 239
composition of, 152 set
domain of a, 149 choice, 286
equivalence, 175 countable, 254
inverse, 152 Dedekind
irreflexive, 180 finite, 301
partial order, 181 infinite, 301
preorder, 180 denumerable, 254
quasi-order, 180 empty, 114
range of a, 149 equality, 112
reflexive, 174 finite, 248, 300
restricted, 153 Dedekind, 301
stable, 317 infinite, 252, 300
strict, 180 Dedekind, 301
symmetric, 174 membership, 111
total, 317 pairwise disjoint, 135
total order, 182 power, 122
transitive, 175 successor, 190
union of, 152 theory
unstable, 317 postulates, 110
well-order, 183 set theory
relational predicate, 76
constant, 76 Shapley, Lloyd S., 303, 315
restricted singleton, 120
function, 158 Skolem’s
relation, 153 prenex normal form, 96
restriction slice function, 157
of a function, 158 smallest element, 183
of a relation, 153 social welfare function, 325
right invertibility of functions, 170 sound
Rosser’s axioms, 55 propositional logic, 58
Rosser, John Barkley, 55 specialization axiom, 80
axioms, 55 square of an integer, 225
Roth, Alvin E., 315 stable
rule relation, 317
 www.pdfgrip.com

390 Index

strategy, 305 definition, 8

dominant, 307 derivable, 8
mixed, 309, 310 provability, 67
pure, 309 substitutivity of equivalence
strict relation, 180 for implications, 23
strong connectivity, 182 for predicates, 90
strongly connected, 182 for propositional formulae, 34
subset, 114 Zermelo’s, 185
substitution there exists, 80
of bound variables, 79 total
of free variables, 88 order, 182
Substitution rule, 7 relation, 317
substitutivity tower, 296
equality, 102, 104, 106 transfinite
substitutivity of equivalence construction, 269, 277
theorem definition by, 269
for implications, 23 definition, 269
for predicates, 90 induction, 268
for propositional formulae, 34 definition by, 269
successor of a set, 190 proof by, 268
sum transitive
notation, 238 relation, 175
Sun, 3 transitive law of implication, 21, 22
superset, 114 transitivity
Suppes, Patrick, 258 of a relation, 175
surjective, function, 167 transposition, 327
surjectivity, 167 principle of, 33
syllogism triangle inequality, 243
hypothetical triangular number, 255
law, 22 trichonomy
symbol principle, 300
sentence, 5, 77
symmetric
relation, 174
symmetric difference of sets (), 139 U
symmetric group, 326 Unanimity, 324
symmetry of a relation, 174 Weak, 330
union
associativity, 130
T axiom (S6), 127
Tarski’s axioms, 55 commutativity, 130
Tarski, Alfred, 55 disjoint, 135
axioms, 29 of a set of sets, 127
tautology, 57 of functions, 158
Teichmüller-Tukey of relations, 152
principle, 300 of two subsets, 125
term, 76 pairwise disjoint, 135
Theorem universal
Arrow’s Impossibility, 327 connective, 65
theorem, 82 quantifier, 80
Bernstein–Cantor–Schröder, 258 Unrestricted Domain, 324
completeness, 69 unstable
deduction, 14 relation, 317
for predicates, 91 upper bound, 184
 www.pdfgrip.com

Index 391

V well-order, 183
valid induced on subsets, 184
propositional form, 57 Well-Ordering
value Principle
distinguished, 57 Zermelo, 288
propositional, 57 Westinghouse DPS-2402 computer
value of a function, 155 NAND, 66
variable Whirlwind I computer
bound, 78 division, 213
free, 78 Whitaker, J. M., 258
individual, 77
propositional, 5, 77
Venn diagram, 127 Z
vertices, 149 Zermelo
veto, 329 Theorem, 185
von Neumann, John, 1, 303 Well-Ordering Principle, 288
voters’ profile, 325 Zorn
Maximal-Element Principle, 290
Maximal-Set Principle, 291
W Zorn’s Lemma, 185
well-formed formula, 5, 77 Zorn, M., 185