0% found this document useful (0 votes)
471 views160 pagesSet Theory and The Continuum Hypothesis
A good book on set theory
Uploaded by
Dibyendu PramanikCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
471 views160 pagesSet Theory and The Continuum Hypothesis
A good book on set theory
Uploaded by
Dibyendu PramanikCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 160
SET THEORY
AND THE
CONTINUUM HYPOTHESIS
Paul J. Cohen
Stanford University
W. A. BENJAMIN, INC.
New York 1966 Amsterdam SET THEORY AND THE CONTINUUM HYPOTHESIS
Copyright © 1966 by W. A. Benjamin, Inc.
Alll rights reserved
Library of Congress Catalog Card Number 66-2866 4
Manufactured in the United States of America
Final copy ready for camera was received from the author
on July 21, 1966; this volume was published on
September 16, 1966
W. A. Benjamin, Inc.
New York, New York 10016 ‘PREFACE
The notes that follow are based on a course given at Harvard Uni-
versity, Spring 1965. ‘the main objective was to give the proof of the
independence of the continuum hypothesis. To keep the course as self-
contained as possible we included background material in logic and ax-
jomatic set theory as well as an account of Gédel's proof of the con-
sistency of the continuum hypothesis. Our review of logic is of neces-
sity rather sketchy although we have tried to cover some of the funda-
mental concepts such as formal systems, undecidable statements and re-
cursive functions. Actually, with the exception of the Léwenheim-Skolem
theorem, none of the results of the first chapter are used in the later
work and the reader who has had an introductory course in logic may
omit this chapter. Its primary purpose is to accustom mathematicians
who are not specialists in logic to the strictly precise point of view
which is necessary when dealing with questions in the foundations of
mathematics. Also, it is intended to clarify certain common confusions
such as that of the concept of an undecidable statement in a particular
axiom system with the concept of an unsolvable problem, which concerns
nethods of computation,
Since our very sincere hope is to make these notes intelligible to
the large body of non-specialists who are interested in the problem, we
have not adopted the very formalistic style which is to be found in some
textbooks on logic. Rather we have tried to emphasize the intuitive mo-
tivations while at the same time giving as complete proofs as possible.
No specific background is assumed, although we occasionally refer to ex-
amples from other parts of mathematics, Of course, it would be helpful
if the reader were familiar with the development of "naive" set theory
as it is customarily taught in courses on real variables or point set
topology. We would like to thank most heartily L. Corwin, D. Pincus, T.
Scanlon, R. Walton, and J. Xenakis for taking notes for various sec-
tions and Jon Barwise for helping in the preparation of the final
version. To Azriel Lévy we are deeply indebted for corrceting many
errors, for many improvements in presentation, and for helping to
bring the entire stylistic level of these notes above their rather
primitive original state. ‘thanks are also duc to the members of
the course at large who consistently challenged all mistakes and who
by their stimulating discussion made teaching the course a distinct
pleasure. ‘the notes are certainly not in the polished form the sub-
ject warrants, but since there is no reasonably complete account of
these questions in the literature, we thought it best to publish
them in the present informal manner rather than to procrastinate in-
definitely. Lastly, our sincere thanks go to the typists who worked
on the notes, principally Sue Golan and Mari Wilson, for their fine
efforts. CHAPTER I.
WON AU FW
10.
ne
CONTENTS
GENERAL BACKGROUND IN LOGIC .
Introduction ...
Formal Languages
Universally Valid Statements .
Gddel Completeness Theorem ..
The Léwenheim-Skolem Theorem .
Examples of Formal Systems ..
Primitive Recursive Functions
General Recursive Functions ..
Gédel Incompleteness Theorem .
Generalized Incompleteness Theorem
Further Results in Recursive Functions ....
CHAPTER II. ZERMELO-FRAENKEL SET THEORY ...-..-+eseeeeeeeee
OIA FWP
AXLOME ceceseseeseceees .
Discussion of the Axioms
Ordinal Numbers ....
Cardinal Numbers .
The Axion of Regularity ....
The System of Gédel-Bernays ..
Higher Axioms and Models for Set Theory .
Léwenheim-Skolem Theorem Revisited ...
CHAPTER III. THE CONSISTENCY OF THE CONTINUUM HYPOTHESIS
Ou FUR
AND THE AXIOM OF CHOICE
Introduction .
Proof of Theorem 1 .
Absoluteness ....
Proof of AC and GCH in L
Relations with GB
‘The Minimal Model
CHAPTER IV. ‘THE INDEPENDENCE OF THE CONTINUUM HYPOTHESIS
AND THE AXIOM OF CHOICE ......
Introduction ..
Intuitive Motivation .
The Forcing Concept
‘The Main Lemmas ..
85
85
92
95
99
107
1o7
109
113
8 120
5. Definability of Forcing
6. The Model No .sssseeeveeseee wal
7. The General Forcing Concept . 127
8. The Continuum Hypothesis . 1e9
9. The Axiom of Choice ... 136
10. Changing Cardinalities . 143
11. Avoiding SM... 1k7
12. GCH Implies AC . 148
13. Conclusion .....4. 150
REVERENCES 0s .eseeseeeeeeeees Ron UUS nS ESEnocnaE an oe 153 CHAPTER I
GENERAL BACKGROUND IN LOGIC
1. IwwrROpUCTION
It is now known that the truth or falsity of the continuum hypoth-
esis and other related conjectures cannot be determined by set theory
as we know it today. This state of affairs regarding a classical and
presumably well-posed problem must certainly appear rather wsatisfac-
tory to the average mathematician. One is tempted to look more closely
and perhaps more critically at the foundations of mathematics. Although
our present "Cantorian" mathematics is highly successful in its treat-
ment of abstractions, one must not overlook the fact that from the vory
beginning the use of infinite processes was regarded with suspicion by
many people. In the 19th century, the objections regarding the use of
convergent series and real numbers were met by Cauchy, Dedekind, Cantor
and others, only to be followed by more profound criticisms from later
mathematicians such as Brouwer, Poincaré and Weyl. The controversy
which followed resulted in the formation of various schools of thought
concerning the foundations. It is safe to say that no attitude has been
completely successful in answering the fundamental questions, but rather
that the difficulties seem to be inherent in the very nature of mathe-
matics. Despite the fact that the continuum hypothesis is a very dra-
matic example of what might be called an absolutely undecidable state-
ment (in our present scheme of things), Gédel's incompleteness theorem
still represents the greatest obstacle to a satisfactory philosophy of
mathematics. These fundamental difficulties, often dismissed by mathe-
maticians, make the independence of the continuum hypothesis less sur-
prising. 2 GENERAL BACKGROUND IN LOGIC ch. I
Gauss seems to have been the first mathematician to have expressed
doubts about too free a use of infinities. In 1831, he wrote, "I protest
..-egainst the use of an infinite magnitude as something completed, which
is never permissible." Later, Kronecker expressed views which were crit-
ical of definitions that required an infinite process to verify that an
object satisfied them. Cantor's work on set theory was the subject of
much criticism to the effect that it dealt with fictions. Nevertheless,
infinite sets are accepted today with few reservations. The traditional
attitude accepts the construction of the real number system from the ra-
tionals as the last and final step in the long series of criticisms and
re-examinations which have marked the history of mathematics. What pos-
sible objections can be raised to the construction of the real numbers?
Simply this: although the reals are based on the integers, the vague
notion of an arbitrary set of integers (or, equivalently, an arbitrary
sequence of integers) must be introduced. Mathematicians inclined to a
finitist point of view might hold that only sets which have an explicit
rule to determine which integers are in the set should be admitted. For
example, the school of Brouwer (Intuitionism) would only admit finite
sets as legitimate objects of study, and even a single integer would not
be considered defined unless a very definite rule for computing it was
given. (For instance, the set consisting of 5 if Fermat's Last Theorem
is true and 7 if it is false is not well-defined, according to Brouwer.)
The criticism of Weyl and Poincaré was directed against "impredicative"
definitions. Although their objections were not as extreme as Brouwer's,
the acceptance of these’ criticisms would mean the destruction of large
portions of mathematics.
Another source of objections was the paradoxes or antinomies of set
theory. In Cantor's set theory a set was thought of as being defined by
a property. Cantor himself pointed out that the set of all sets leads
to an absurdity. Although this type of paradox (along with those of
Russell, Burali-Forti and others), seems entirely remote from ordinary
mathematical reasoning, the paradoxes did point out the necessity of cx-
treme care when attempting to describe which properties describe sets.
In 1908, Zermclo presented a formal set of axioms for set theory which
encompassed all the present day reasonings in mathematics and yet which
is presumably free from paradoxes, This axiouatization of set theory $2 FORMAL LANGUAGES 2
was in keoping with the spirit of the school of Formalism, led by David
Hilbert. According to the Formalist point of view, mathematics should
be regarded as a purcly formal game played with marks on paper, and the
only requirement this game need fulfill is that it does not lead to an
inconsistency. To completely describe the game required setting down
the rules of mathematical logic with much greater precision than had
‘een previously done. This was done, and the Formalists turned their
attention to showing that various systems were consistent. As is well
known, this hope was destroyed by Gédel's discovery of the incomplete-
ness theorem, which implies that the consistency of a mathematical sys-
tem cannot be proved except by methods more powerful than those of the
system itself.
Despite this failure, the Formalist program contributed greatly to
the development of logic by establishing a systematic study of mathe-
matical systems. In these notes, our first object will be to describe
how a mathematical system can be completely reduced to a purely formal
game involving the manipulation of symbols on paper. By a formal system
we shall mean a finite collection of symbols and perfectly precise rules
for manipulating these symbols to form certain combinations called "the-
orems". Of course, these rules must be given in informal mathematical
language. However, we shall demand that they be completely explicit
rules requiring no infinite processes to check and that in principle
they can be coded into computing machine. In this way questions con-
cerning infinite sets are replaced by questions concerning the combina-
torial possibilities of a certain formal game. Then we will be able to
say that certain statements are not decidable within given formal sys-
tems.
2. FORMAL LANGUAGES
If we examine Peano's axioms for the integers, we find that they are
not capable of being transcribed in a form acceptable to a computing ma-
chine. This is because the crucial axiom of induction speaks about "sets"
of integers but the axioms do not give rules for forming sets nor other
basic properties of sets, Here is an example of the difficulty in satis-
fying the stringent requirements for a formal system outlined above. When 4 GENERAL BACKGROUND IN LOGIC ch. L
we do construct a formal system corresponding to Peano's axions we shall
find that the result can not quite live up to all our expectations. This
difficulty is associated with any attempt at formalization.
‘two types of symbols will appear in our formal language. First,
there are those symbols common to all mathematical systems. Then there
are those used to denote particular concepts in special branches of math-
ematics, such as the symbols for addition, group multiplication, adjoint
of a matrix, etc. The general symbols we shall use are the following:
& v 4 o
not and, or implies if and only if
v = G) ee
for all there exists equals parentheses variable symbols
Thc words beneath the symbols have, in principle, nothing to do with our
“Qo! as formally
formal language. Rather, we are thinking of the symbol
representing the words "if and only if". The ordinary meanings of these
words will suggest certain rules concerning the symbols, but the formal
game must be played by the rules without recourse to any meanings which
may have suggested then.
The first five symbols are known as propositional connectives. The
symbols which formally represent "for all" and "there exist” are know
as universal and existential quantifiers respectively. Parentheses are
used in the formation of expressions to insure their unique readability.
In any given discussion one may need arbitrarily many variables so that
we shall use x, x', x",..-, as symbols for variables. In practice we
will use the letters x,, xp) x)+-- or x, y, 2 although in theory
these should be replaced by x, x', x",... . In this way we need only
a finite set of symbols. Our list of symbols is in no sense the most
economical for our purposes since (as will follow from the rules for
our formal system) some of the symbols can be avoided by combinations
of others. For example, it will follow from rules A and G below that
V could be replaced by ~~ . We shall sometimes omit parentheses
or employ other abbreviations in our formulas if there is no danger
of confusion. $2 FORMAL LANGUAGES 5
To express interesting statements we need special symbols which
will be used to formally represent particular relations under consider-
ation. For this purpose we admit a finite set of relation symbols R,,
Royer + To each R, is assigned an integer ny 21, ny intuitively
indicating that the relation represented by R, is a relation between
Ry objects. For example, if me 2, then R, is a binary relation
(such as < between real numbers), and in our rules for forming formal
expressions we will require that R, be folloved by two variables or
constants as in Rj (x,y) and Ry (c)2)- We will say that R, is an n,-
ary predicate symbol, and sometimes write Ry (x 9+0%,)) to indicate
n, explicitly. We shall also use particular symbols to represent con-
stants which play special role in the system. For example, in group
theory it is convenient to have a symbol for the identity element. We
use the letter "c" together with "' " to generate the symbols c,
c', c", ete. to represent constants. In practice, we write more simply
ey oy ete.
Before we give the precise rules for forming the formal expressions
of our formal language we give some examples. Instead of thinking of
"addition" an an operation on pairs of numbers, we can think of it as
a relation between triples of numbers, namely the relation
rt+s=t.
Thus if we are considering » formal language in which R, is a ternary
relation symbol, we could let R represent this relation. Then the
uniqueness of addition would be insured by
Vx Vy VzVu ((R, (x,y,z) & R, (x,y5u)) ~z=u
If we let multiplication (r+ s = +) be represented by R,(x,y,z), then
the associativity of multiplication becomes
Va Wy We St, St, 3t,(,(oy,t,) @R(ty »2/t)) &R,(¥,2,t3) OR, (20t, 5tp))
‘The existence of an additive identity can be stated as Ax Wy (R,(x,y,y))
tut this would be inconvenient in many cases. For, whenever an argument
used this additive identity we would have to restate its existence. It
is simpler to use a constant symbol "0" and introduce as an axiom 6 GENERAL BACKGROUND IN LOGIC Ch. I
Vx (x + 0+ x). In general the use of abbreviations is indispensable
for making mathematics comprehensible and we shall not hesitate to use
them when their meaning is clear and when it is also clear how to re-
place them by the strict formal language.
Let us examine further our formal system using only the special
relation symbols R, end R, representing addition and miltiplication.
It is not hard to see that we can express all the axioms of Field Theory
within the system. (Observe that we are not yet discussing proofs or
“true” statements, merely what can be expressed.) More complicated as-
sertions such as the fact that a quadratic equation has at most two roots
can be written as follows?
Neyo du dv Wx (x? + ax + b= 0) 4 (x=uvx=v))-
We leave as an exercise to show how the forma x°
+2x+b=0 can
be transcribed in the formal system. One encounters difficulty if one
tries to say that an equation of degree n has at most n roots. This
is because we have no notation (within the formal system) for an inte-
ger to denote the degree of an arbitrary polynomial nor an induction
procedure to define the notion of a polynomial of arbitrary degree. Thus
we cannot handle properties which involve the notion of an arbitrary in-
teger. Nevertheless, one could formulate the theorems of Galois theory
for quadratic, cubic, etc. extensions by speaking only of root perm
tations. Gelois theory in its customary formation speaks about sets,
such as sub-fields, sub-groups, etc. and so cannot be expressed directly
in our system.
We now give the precise rules for forming grammatically correct
statements. These are called well-formed formilas (wff). Recall that
our formal language consists of the general symbols given above includ-
ing variable symbols and constant symbols, as well as a finite number
of relation symbols each associated with an integer.
Rules for well-formed formulas
1. x=yx=c,c=c! are wff's where x and y are variable sym-
pols and c and c' are any constant symbols.
2. If R is an n-ary relation symbol and each of t,,...,t, is either
s& variable or constant symbol, then R(t) y++.5t,) is a wef. §2 FORMAL LANGUAGE 7
3. If A and Bare wff's so are ~(A), (A) & (B), (A) v (B), (A) (8),
and (A) <> (B).
4, If A isa wf, then soare AxA and Vxa.
Observe that our wff's may have loose constant and variable symbols
dangling about. Also Rule 4 allows an expression such as Vxy=z as
avwff. This is to be interpreted intuitively as meaning that if Wx
or Hx occurs before a wff A which does not involve x then the ef-
fect of the quantifiers is nil, and they can be omitted. This brings us
to the precise distinction between bound and free variables.
Definition. Fach occurrence of a variable symbol in a wff is defined as
free or bounded as follows:
1. Every variable occurring in a formula of the form mentioned in Rules
land 2 is free,
2. The free and bound occurrences of variables in the wff's mentioned
in Rule 3 are precisely the samc as those for A and B separately.
3. the free and bound occurrences of a variable in a formula JxA or
VxA are the same as those of A except that every free occurrence
of x is now considered bound.
‘The reader may object that we have been too carcless in Rule 4 by
allowing bound variables to be bound again. Actually, no confusion is
possible if we agree that such quantification has no effect, but to be
more specific we can make the following definition:
Definition. A wff is called "good" if in the application of Rule 3, A
and B have only free variables in common, and in the application of
Rule 4, x occurs as a free variable in A.
The conventions concerning the use of parentheses are such that no
ambiguity can arise, This would actually require a simple coubinatorial
argument, which we omit,
Definition. A statement is a formula with no free variables. 8 GENERAL BACKGROUND IN LOGIC ch. I
3. UNIVERSALLY VALID STATEMENTS
The object of mathematics is to discover “true” theorems. We shall
use the term "valid" to describe statements formed according to certain
rules and then shall discuss how this notion compares with the intuitive
idea of "true". When we write down a statement involving certain con-
stant and relation symbols, we have not indicated any specific interpre-
tation of these symbols. If such a statement is to be intuitively con-
sidered as "true" it must be true independently of how these symbols are
interpreted. For example, (f A 1s the conjunction of the usual axioms
for field theory and B is the statement that a quadratic equation has
at most two roots, then A+B isa true statement, because in any sys-
tem where the axioms hold with regard to two ternary relations (which
we may call addition and multaplication), a quadratic equation may have
at most two roots. We shall introduce the concept of a model for a for-
mal system and a fundamental result will be the identification of the
valid statements with those that are true in overy model.
The rulcs for forming valid statements arc known as the Predicate
Calculus. A certain simple sub-system of these rules, whose importance
was realized before the other rules, is known as the Propositional Cal-
culus, ‘These are the rules concerning the manipulation of the symbols
~s&,v, 4,4. For the moment, let Ay, 4 be variable let-
pee
ters, not to be confused with the variables uscd in the formal ianguage.
These variables will eventually be replaced by wff's.
Definition. A propositional function is a formal string of symbols de-
fined as follows:
l. If A isa variable letter then A is a propositional function.
2. If P and Q are propositional functions so arc ~ (P), (P) & (Q),
(P) v (Q), (P) +(@), and (P) (a).
If P is a propositional function of the variables Apr A, (that
is, P only involves variable letters among the A,,...,A,, and not nec-
essarily all of these), we wish lo indicate how the truth or falsity of
P depends on that of A,. To this end we associate wth P a function
defined on the set of all n-tuples of O and 1 (i.e., (epee), where
is either 0 or 1) taking the values 0 and 1. This is done as follows:
4 §3 FORMAL LANGUAGE 9
1, The propositional function A, corresponds to the projection func-
tion (eys--4¢,) +e.
2. If f is the function corresponding to P and g that for Q, the
functions corresponding to ~ (P), (P) & (Q), (P) v (a), (P) +(@),
(P) 4 (Q) are respectively, 9,(f), 9,(f,8)) 93(f 6), 9,(f8),
9,(f8) where the 9, are given by the “truth tables":
fe oe eee ee
oO 1 oO ae oO = oO
x x x x x
1 Ol of o/o] +s ofof}1z} 4 of2fal ¢ oO
lio 1[ 0 1 1jl 1 ae - 1) 0 1
930409) 9, (59)
~ & v > o
The reader will casily check that this corresponds to the usual usage
of the propositional connectives, with the understanding that 1 represents
truth and 0 falsity. We define a propositional function as identically
true if the corresponding function takes only the value 1. This means
intuitively that the propositional function is "true" no matter what
statements are substituted for the variables. We can now state the
basic rule of the Propositional Calculus
Rule A (Rule of the Propositional Calculus)
If P is a propositional function of the variable letters A,,...,4,
wnich is identically true, then the result of replacing each A, by any
statement is a valid statement.
Recall that our intended meaning of the term "valid statement" will
be a statement which is intuitively true in any interpretation of the re-
lation and constant symbols used to form it. There is another rule of
the predicate calculus related to A which allows us to form new valid
statements from old ones. It has often been illustrated by statements
concerning the mortality of Socrates.
Rule B (Rule of Inference)
If A and (A) +(B) are valid statements so is B.
We next give the rules for manipulating the sign of equality. 30 GENERAL BACKGROUND IN LOGIC ch. I
Rule C (Rules of Equality)
lL. cee, (ce =e') 3(e' =e), and ((c =c') & (ec = c")) a(e = 0")
are valid statements where c, c', c" are any three constant symbols.
2, If A isa statement, c and c' constant symbols, and if A’ re-
presents A with every occurrence of c replaced by c', then
(ec = ¢')—+((A) +(A')) is a valid statement.
Rule D (Change of Variables)
If A is any statement and A' results from A by replacing cach
occurrence of the symbol x with the symbol x', where x and x' are
any two variable symbols, then the statement (A) «+(A') is a valid
statement.
Both of these rules are obvious and require no discussion. It is a
simple exercise to show that for every statement A, there is a "good"
statement A' obtained by replacing the variable symbols in A by other
variable symbols, and such that (A) ¢+(A') is a valid statement. For
the next rule, let A(x) represent a formula with one free variable x
and in which every occurrence of x is free and let A(c) represent the
result of replacing every occurrence of x by the constant symbol c.
Rule E (Rule of Specialization)
(Vx A(x)) +(A(c)) is a valid statement where c is any constant
symbol.
The next rule is a bit misleading and requires some explanation.
Often in arguments we say "let c be an arbitrary but fixed integer”.
We then proceed to reason about ¢ and come to @ certain conclusion
A(c). We can then deduce that Vx A(x) since we used no special prop-
erties of c. In reality, we have treated ¢ as a variable, even though
we called it a constant, This is because our valid statements will be
true in every interpretation of the constant and relation symbols. We
could express this by: if A(c) is a valid statement sois Vx A(x).
However, rule F is in a form more convenient for our purposes and yields
the above, as we shall see.
Rule F
Let B be a statement not involving c or x. Then if A(c) 9B
is valid, so is HxA(x) +B. gh GODEL COMPLETENESS THEOREM ul
The next rule will allow us to bring every formula involving quanti-
fiers into a form in which it begins with a quantifier.
Rule G
Let A(x) have x as the only free variable and let every occur-
rence of x be free, Let B be a statement which does not contain x.
Then the following are valid statements.
(~ (Wea(x))) > (Ax ~ (A(x)))
((MeA(x)) & (B)) > (eC (A(x)) & (B)))
(Ca xA(x)) & (B)) > (Bx(A(x)) & (B))
Definition. Let S be a collection of statements. We say that A is
derivable from 8, if for some B,,...,B, in S, the statement ((B) ieee
& (B,)) (A) is valid.
Fact: If S and §' are collections of statements such that every
statement in S' is derivable from 8, then every statement derivable
from $' is derivable from S.
Proof. Exercise. In the course of a proof we often reason by con-
tradiction or assume certain hypotheses temporarily. Such steps are all
justified by the propositional caleuJus, although we shall not give all
the formal details in applications. Assume for example we know that A(c)
is a valid statement. We shall show that V(x) A(x) is valid. (This is not
the same as saying A(c) + Vx) A(x) is valid, which it is not in gen-
eral.) It is sufficient to show that ~ W(x) A(x) leads to a contra-
diction (i.e., B&~B for any B). Equivalently by Rule G we show that
Hx~ A(x) leads ta a contradiction. But ~A(c) does lead to a con-
tradiction since A(c) is valid and hence by Rule F so does Sx ~ A(x).
4, GODEL, COMPLETENESS THEOREM
Having now given rules for forming valid statements we come to the
problem of identifying these statements with the intuitively "true"
statements. This discussion will be carried out in the spirit of tradition 12 GENERAL BACKGROUND IN LOGIC ch. I
mathematics, that is to say, outside of any formal language. We shall
use some elementary notions of set theory. After we have formalized set
theory itself, then of course this discussion can be expressed in that
formal system. In our original discussion, we had a finite number of
symbols. This was important for foundational purposes, in order to re-
duce mathematics to a formal game playable by a computing machine. How-
ever, for some purposes it is of interest to allow arbitrarily many con-
stant and relation symbols. We write the proofs of Sections 4 and 5 for
this more general case.
Assume now that we are dealing with a collection S of statements
involving constants ¢,, @¢ I, and relation symbols R,, B ¢ J, vhere
each R, has a fixed number of variables. Let M be @ non-empty set and le
Cy fq be a map from the constant symbols to elements of M, not nec-
essarily distinct, and R, +R, @ map which associates to @ k-ary relatio
symbol, a subset of the k-fold direct product, MXMx +++ XM, We shall
then say that we have an interpretation of the constants fy and the re-
lation symbols R, in the set M. To every statement using only these
constant symbols and relation symbols, we shall associate its “truth
value" under this interpretation. Intuitively, of course, we merely mean
whether or not the statement is true in M under the given interpreta-
tion of the constant and relation symbols. However, a precise definition
is easy to give if we proceed by induction on the iength of formulas.
Definition. Let A be a formula with free variables among Rp Xs
n> 0, and let Ryo er ok pe elements of M. We define the truth value
of A (in M) at xy.
1, If A is of the form x,
in
mm Ky se ___——s~_—=™=TE
at Ky.ok if Xk, =x,, % = c, or =6
a t
7 pF 1 = Sy respectively.
eeceeeeeeraeeee eae R(t» ++ yt,) where R is an m-ary relation symbol and
each t, is a constant symbol or one of the x,,...,x,, then A is
true at X),..+,%, if the m-tuple By ot) isin R (the sub-
set of M™ associated with R under the given interpretation).
3. If A is a propositional function of formulas, we evaluate the
truth of A at x
a ok, by means of the propostional calculus. gh GODEL COMPLETENESS THEOREM 13
4, If A is of the form V(y)B(y,x,,.+.5x,)[resp. AyB(y,x,,.-+5%,)]
then A is true at york, if, forall y in M [resp. for
some y in M] BY, x1 9+++9%,) is true at J, X.-+o%,+
To avoid any ambiguity over substitution of variables we may assume that
all the formulas are good wff's as defined in Section 3. We note that if
A is a statement we can take n = 0, and our definition is just truth
in M under the given interpretation.
Definition. oe S isa set statements containing Ly and Ry Ma
set and ey +6, and R, +R, are maps as above, ve say that M isa
model for § (under the interpretation) if all the statements of $ are
true in M.
Strictly speaking a model is a set M together with an interpreta-
tion of some (possibly none) constant and relation symbols. Unless it
is necessary for clarity, we shall suppress any mention of the explicit
interpretation.
Definition. A set of statements is said to be consistent if the state-
ment A&~A cannot be derived from S$ for any A.
The point of these definitions is the following obvious fact:
THEORFM 1. If A is a valid statement, it is true in every model.
If a set of statements § has a model then it is consistent.
We omit the tedious proof, which is merely a verification that the
rules of the predicate calculus correspond to correct methods of deduc-
tion. A much more interesting question is whether the rules we have
given exhaust all possible deductions. With out notion of models we can
phrase this quite precisely.
THEOREM 2, Godel Completeness Theorem. Let S be any consistent
set of statements, ‘Then there exists a model for §$ whose cardinality
does not exceed the cardinality of the number of statements in S if S
is infinite, and is countable if S is finite.
The proof uses the axiom of choice (unless the set S is already
well-ordered). Given S we shall explicitly show how to construct a wy GENERAL BACKGROUND IN LOGIC ch. I
model for S. The proof however is non-constructive in the sense that
the construction of the model for S may depend upon examining an in-
finite number of possibilities. Neverthcless, if S is finite, the
model can be taken as the set of integers and the resulting rclations
will be arithmetically definable (i.e, definable by formulas using only
addition and multiplication). This will follow from the form of the
proof although we shall not go into the matter.
The first step is an examination of the special case in which S
does not contain any quantifiers.
THEOREM 3. (Completeness of the Propositional Calculus). If S
contains no quantifiers and is consistent, then there is a model M for
§ in which every element of M is of the form , for some ¢, appear-
ing in 8.
We need one lemma.
LEMMA. If T is a consistent set of statements, A an arbitrary
statement, either Tu (A) or TU (~A) is consistent.
Proof. If TU (A) is inconsistent, then for some B, in 1,
A&B, & +++ &B »C&~C for some C, is valid. If TU (~A) is in
consistent, then for some By in T, (~A)& BL Bee & BL ~C&-Cis
valid, The propositional calculus now implies that B, & ++: &B. & BY
Bere & BL 3C&~C is valid, so that T must be inconsistent.
Now to prove Theorem 3. Let § be well-ordered. This induces a
well-ordering on all the constant and relation symbols which appear in
8, This in turn induces a well-ordering of all possible statements of
the forn c, = ‘, and Ba(eys+--2e,) where c, and RB are constant
and relation symbols occurring in 8. Call these statements F,. We now
define statements G, by induction on a. If F, is consistent with
Su (c,|B a, € oO This motivates the following definition;
Definition. x is an integer if
1) Wealycx&zexsy=zvyezviey)
2) Wiz (yex&zey +z x)
One easily sees that in our intuitive model M, the sets o, are pre-
cisely the sets satisfying this definition. Namely, first observe that
in M every non-empty set x contains a member y such that if ze x
then ~zey. That is, y is "minimal" with respect to the e-relation.
Now if x is an integer in M, a "minimal" element of x must be $ by
2). If y is the minimal element of x - $ , then clearly by 1) and 2)
y must be (g]. By repeating this operation, one can show that x must
be equal to some o:. Thus our definition of integer is a reasonable one.
Definition. If x is an integer, let x +1 denote x U (x).
We show that x +1 is an integer. If y and z belong to x+l,
either both belong to (x}, both belong to x, or one belongs to x and
the other to (x}. Ifa set belongs to {x} it must equal x, so that
in the first case y = 2, in the second since both belong to x and x
is an integer 1) holds, and in the third we trivially have either ye z
or zey. Thus 1) holds. If ye x+1 and zey, then if y =x,
wehave zex so zex+l. If yex we again have 2 x, since
x isan integer so zex+l.
We can now state the induction axiom, Again AL (x, tyy rth) range
over all formulas with at least one free variable. We write © in place
of § and Int x as an abbreviation for the definition that x is an
integer.
: Ntpese ke [A(t s+ sty) & Wy (Int y &A (yt ot) >
Ayly Ly ty ++sst))) 9 VeUInt x 9 AL(x,ty,---5t,))]
Om $6 EXAMPLES OF FORMAL SYSTEMS Ea)
This axiom scheme is essentially identical to the one given for zy.
Again, we have not intuitively captured the complete strength of the in-
formal induction principle since not every possible set of integers,
is described by a property A, for suitable t,. Since as has been al-
ready remarked, no set of axioms for the integers can be categorical,
this is not an overpovering objection to our system. We now show how to
develop conventional number theory in our system Z,. First, we define
a function as a set of ordered pairs such that a set occurs as a left
member at most once. The domain of the function is the set of left mem-
bers. Now, whenever a function is defined in elementary number theory
it usually is defined as follows: there is a certain condition B(x)
which says that x is a set of ordered pairs defining a function on
the integers y as well as constants
Axtons for Real-Closed Fields
1. Field Axioms These are the well-known axioms for a field which
we do not bother to repeat here.
2. Order Axioms:
i) Vay (xs yvxO&y>O>xty>O&x+y>O)
3,° Completeness Axioms. Let f(x) denote age + total. The
axion states
Va +28 eXpy( f(x) >O& ty) x(d,i) >M & x(d,j) >M & (x(a,4)
and x(d,Jj) are relatively prime)].
Proof. In words, ii) says that forall n and M, we can find a
a, such that x(d,i) for iy is an abbreviation for 3z(z/0 & x=y+z).
Also "x and y are relatively prime" can be written as
Vavelkeu-schyev-zazell.
We now define B(dji,x) as x=1+(i+1)d. If n and M are
given, we can take d = (max(n,M))!, so that clearly x(d,i) >M. Also, 30 GENERAL BACKGROUND OF LOGIC ch. I
if 1
iii) for each N, there is a number (in the proof taken as N!)
which is divisible by all n
