-
NOTES ON SET THEORYJ. Donald MonkMarch 6, 2018
TABLE OF CONTENTS
LOGIC
1. Sentential logic . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . .12. First-order logic . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .163. Proofs . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . 294. The completeness theorem . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . 49
ELEMENTARY SET THEORY
5. The axioms of set theory . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . 776. Elementary set theory . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . .807. Ordinals, I . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . 898. Recursion .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . 949. Ordinals, II . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . 10310. The axiom of choice. . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .12711. Cardinals . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
135
MODELS OF SET THEORY
12. The set-theoretical hierarchy . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . 16113. Absoluteness . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . .18114. Checking the axioms . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . 19115. Reflection theorems . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . 19716.
Consistency of no inaccessibles . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20417. Constructible sets . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . .205
INFINITE COMBINATORICS
18. Real numbers in set theory . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . .22119. The Cichon diagram . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . 27220. Continuum cardinals . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . 29121. Linear orders . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . 31622. Trees . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . 35423. Clubs and stationary sets . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . 39024. Infinite combinatorics . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . 42325. Martins axiom . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . 43426. Large cardinals
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
476
FORCING
27. Boolean algebras and forcing orders . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
50528. Generic extensions and forcing . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. 53829. Forcing and cardinal arithmetic . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . 56230. General theory of forcing . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . .577
i
-
31. Iterated forcing . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . 60632. p = t . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . 690
PCF
33. Cofinality of posets . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . 72534. Basic properties of PCF . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . .75935. Main cofinality theorems . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . 781
ADDITIONAL SECTIONS
36. Various forcing orders . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . .80537. More examples of iterated forcing . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . 82838. Consistency results concerning P()/fin . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
838
REAL NUMBERS
39. The integers . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . 84740. The rationals . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . 85441. The reals . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
860Index of symbols . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . .870Index of words . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .875
ii
-
1. Sentential logic
We go into the mathematical theory of the simplest logical
notions: the meaning of and,or, implies, if and only if and related
notions. The basic idea here is to describe aformal language for
these notions, and say precisely what it means for statements in
thislanguage to be true. The first step is to describe the
language, without saying anythingmathematical about meanings. We
need very little background to carry out this develop-ment. is the
set of all natural numbers 0, 1, 2, . . .. Let + be the set of all
positiveintegers. For each positive integer m let m = {0, . . . , m
1}. A finite sequence is afunction whose domain is m for some
positive integer m; the values of the function canbe arbitary.
To keep the treatment strictly mathematical, we will define the
basic symbols ofthe language to just be certain positive integers,
as follows:
Negation symbol: the integer 1.Implication symbol: the integer
2.Sentential variables: all integers 3.
Let Expr be the collection of all finite sequences of positive
integers; we think of thesesequences as expressions. Thus an
expression is a function mapping m into +, for somepositive integer
m. Such sequences are frequently indicated by 0, . . . , m1. The
casem = 1 is important; here the notation is .
The one-place function mapping Expr into Expr is defined by = 1
for anyexpression . Here in general is the sequence followed by the
sequence .
The two-place functionmapping ExprExpr into Expr is defined by =
2for any expressions , . (For any sets A,B, A B is the set of all
ordered pairs (a, b)with a A and b B. So Expr Expr is the set of
all ordered pairs (, ) with , expressions.)
For any natural number n, let Sn = n+ 3.Now we define the notion
of a sentential formulaan expression which, suitably inter-
preted, makes sense. We do this definition by defining a
sentential formula construction,which by definition is a sequence
0, . . . , m1 with the following property: for eachi < m, one of
the following holds:
i = Sj for some natural number j.
There is a k < i such that i = k.
There exist k, l < i such that i = (k l).
Then a sentential formula is an expression which appears in some
sentential formula con-struction.
The following proposition formulates the principle of induction
on sentential formulas.
Proposition 1.1. Suppose that M is a collection of sentential
formulas, satisfying thefollowing conditions.
1
-
(i) Si is in M , for every natural number i.(ii) If is in M ,
then so is .(iii) If and are in M , then so is .
Then M consists of all sentential formulas.
Proof. Suppose that is a sentential formula; we want to show
that M . Let0, . . . , m be a sentential formula construction with
t = , where 0 t m. We proveby complete induction on i that for
every i m, i M . Hence by applying this to i = twe get M .
So assume that for every j < i, the sentential formula j is
in M .Case 1. i is Ss for some s. By (i), i M .Case 2. i is j for
some j < i. By the inductive hypothesis, j M , so i M by
(ii).Case 3. i is j k for some j, k < i. By the inductive
hypothesis, j M and
k M , so i M by (iii).
Proposition 1.2. (i) Any sentential formula is a nonempty
sequence.(ii) For any sentential formula , exactly one of the
following conditions holds:
(a) is Si for some i .(b) begins with 1, and there is a
sentential formula such that = .(c) begins with 2, and there are
sentential formulas , such that = .
(iii) No proper initial segment of a sentential formula is a
sentential formula.(iv) If and are sentential formulas and = , then
= .(v) If , , , are sentential formulas and = , then = and
= .
Proof. (i): Clearly every entry in a sentential formula
construction is nonempty, so(i) holds.
(ii): First we prove by induction that one of (a)(c) holds. This
is true of sententialvariablesin this case, (a) holds. If it is
true of a sentential formula , it is obviously trueof ; so (b)
holds. Similarly for , where (c) holds.
Second, the first entry of a formula differs in cases
(a),(b),(c), so exactly one of themholds.
(iii): We prove this by complete induction on the length of the
formula. So, supposethat is a sentential formula and we know for
any formula shorter than that no properinitial segment of is a
formula. We consider cases according to (ii).
Case 1. is Si for some i. Only the empty sequence is a proper
initial segment of in this case, and the empty sequence is not a
sentential formula, by (i).
Case 2. is for some formula . If is a proper initial segment of
and it is aformula, then begins with 1 and so by (ii), has the form
for some formula . Butthen is a proper initial segment of and is
shorter than , so the inductive hypothesisis contradicted.
Case 3. is for some formulas and . That is, is 2. If is a
properinitial segment of which is a formula, then by (ii), has the
form 2 for someformulas , . Now = , so is an initial segment of or
is an initial segment
2
-
of . Since and are both shorter than , it follows from the
inductive hypothesis that = . Hence = , and = , contradiction.
(iv) is rather obvious; if = , then and are both the sequence
obtained bydeleting the first entry.
(v): Assume the hypothesis. Then is the sequence 2, and isthe
sequence 2
. Since these are equal, and start at the same place in
thesequence. By (iii) it follows that = . Deleting the initial
segment 2 from thesequence, we then get = .
Parts (iv) and (v) of this proposition enable us to define
values of sentential formulas,which supplies a mathematical meaning
for the truth of formulas. A sentential assignmentis a function
mapping the set {0, 1, . . .} of natural numbers into the set {0,
1}. Intuitivelywe think of 0 as false and 1 as true. The definition
of values of sentential formulas isa special case of definition by
recursion:
Proposition 1.3. For any sentential assignment f there is a
function F mapping the setof sentential formulas into {0, 1} such
that the following conditions hold:
(i) F (Sn) = f(n) for every natural number n.(ii) F () = 1 F ()
for any sentential formula .(iii) F ( ) = 0 iff F () = 1 and F () =
0.
Proof. An f -sequence is a finite sequence (0, 0), . . . , (m1,
m1) such that eachi is 0 or 1, and such that for each i < m one
of the following holds:
(1) i is Sn for some n , and i = f(n).
(2) There is a k < i such that i = k and i = 1 k.
(3) There are k, l < i such that i = k l, and i = 0 iff k = 1
and l = 0.
Now we claim:
(4) For any sentential formula and any f -sequences (0.0), . . .
, (m1, m1) and(0.
0), . . . , (
n1,
n1) such that m1 =
n1 = we have m1 =
n1.
We prove (4) by induction on , thus using Proposition 1.1. If =
Sn, then m1 = f(n) =n1. Assume that the condition holds for , and
consider . There is a k < m 1such that = m1 = k. By Proposition
1.2(iv) we have k = . Similarly, thereis an l < n 1 such that =
n1 =
l and so
l = . Applying the inductive
hypothesis to and the sequences 0, . . . , k and 0, . . . , l we
get k =
l. Hence
m1 = 1 k = 1 l = n1.
Now suppose that the condition holds for and , and consider .
There arek, l < m 1 such that ( ) = (k l). By Proposition 1.2(v)
we have k = and l = . Similarly there are s, t < n 1 such that (
) = (s
t). By
Proposition 1.2(v) we have s = and t = . Applying the inductive
hypotheis to
and the sequences 0, . . . , k and 0, . . . ,
s we get k =
s. Similarly, we get l =
t.
Hence
m1 = 0 iff k = 1 and l = 0
3
-
iff s = 1 and t = 0
iff n1 = 0.
This finishes the proof of (4).
(5) If 0, . . . , m1 is a sentential formula construction, then
there is an f -sequence ofthe form (0, 0), . . . , (m1, m1).
We prove this by induction on m. First suppose that m = 1. Then
0 must equal Sn forsome n, and (0, f(n)) is as desired. Now suppose
that m > 1 and the statement is true
for m 1. So let def= (0, 0), . . . .(m2, m2) be an f
-sequence.
Case 1. m1 = Sp. Then (m1, f(p)) is as desired.
Case 2. There is a k < m such that m1 = k. Then (m1, 1 k) is
asdesired.
Case 3. There are k, l < m such that m1 = k l. Then (m1, ) is
asdesired, where = 0 iff k = 1 and l = 0.
Thus (5) holds. Now we can define the function F required in the
Proposition. Let be any sentential formula. Let 0, . . . , m1 be a
sentential formula construction suchthat m1 = . By (5), let (0, 0),
. . . , (m1, m1) be an f -sequence. We defineF () = m1. This is
unambiguous by (4).
Case 1. = Sn for some n. Then by the definition of f -sequence
we have F () =f(n).
Case 2. There is a k < m such that = m1 = k. Then (0, 0), . .
. , (k, k)is an f -sequence, so F (k) = k. So
F () = F (m1) = m1 = 1 k = 1 F (k).
Case 3. There are k, l < m such that = m1 = k l. Then (0, 0),
. . . ,(k, k) is an f -sequence, so F (k) = k; and (0, 0), . . . ,
(l, k) is an f -sequence, soF (l) = l. So
F () = 0 iff F (m1) = 0 iff em1 = 0 iff
k = 1 and l = 0 iff F (k) = 1 and F (l) = 0.
With f a sentential assignment, and with F as in this
proposition, for any sententialformula we let [f ] = F ().
Thus:
Si[f ] = f(i);
()[f ] = 1 [f ];
( )[f ] =
{0 if [f ] = 1 and [f ] = 0,
1 otherwise.
The definition can be recalled by using truth tables:
4
-
1 0
0 1
1 1 1
1 0 0
0 1 1
0 0 1
Other logical notions can be defined in terms of and . We
define
= ( ). = . = ( ) ( ).
Working out the truth tables for these new notions shows that
they mean approximatelywhat you would expect:
1 1 0 0 1 0 1 1 1 1
1 0 1 1 0 0 1 0 1 0
0 1 0 1 0 1 1 1 0 0
0 0 1 1 0 1 0 1 1 1
(Note that corresponds to non-exclusive or: or , or both.)
The following simple proposition is frequently useful.
Proposition 1.4. If f and g map {0, 1, . . .} into {0, 1} and
f(m) = g(m) for every msuch that Sm occurs in , then [f ] =
[g].
Proof. Induction on . If is Si for some i, then the hypothesis
says that f(i) = g(i);hence Si[f ] = f(i) = g(i) = Si[g]. Assume
that it is true for . Now Sm occurs in iff it occurs in . Hence if
we assume that f(m) = g(m) for every m such thatSm occurs in , then
also f(m) = g(m) for every m such that Sm occurs in , so()[f ] = 1
[f ] = 1 [g] = ()[g]. Assume that it is true for both and , andf(m)
= g(m) for every m such that Sm occurs in . Now if Sm occurs in ,
then italso occurs in , and hence f(m) = g(m). Similarly for . It
follows that
( )[f ] = 0 iff ([f ] = 1 and [f ] = 0) iff ([g] = 1 and [g] =
0) iff ( )[g] = 0.
This proposition justifies writing [f ] for a finite sequence f
, provided that f is long enoughso that m is in its domain for
every m for which Sm occurs in .
A sentential formula is a tautology iff it is true under every
assignment, i.e., [f ] = 1for every assignment f .
5
-
Here is a list of common tautologies:
(T1) .(T2) .(T3) ( ) .(T4) ( ) ( ).(T5) ( ).(T6) ( ) [( ) (
)].(T7) [ ( )] [( ) ( )].(T8) ( ) ( ).(T9) ( ) .(T10) ( ) .(T11) [
( )].(T12) ( ).(T13) ( ).(T14) ( ) [( ) (( ) )].(T15) ( ) ( ).(T16)
( ) ( ).(T17) [ ( )] [( ) ].(T18) [ ( )] [( ) ].(T19) [ ( )] [( ) (
)].(T20) [ ( )] [( ) ( )].(T21) ( ) ( ).(T22) ( ).(T23) ( ).
Now we describe a proof system for sentential logic. Formulas of
the following form aresentential axioms; , , are arbitrary
sentential formulas.
(1) ( ).
(2) [ ( )] [( ) ( )].
(3) ( ) ( ).
Proposition 1.5. Every sentential axiom is a tautology.
Proof. For (1):
( )
1 1 1 1
1 0 1 1
0 1 0 1
0 0 1 1
6
-
For (2): Let denote this formula:
( ) ( ) ( )
1 1 1 1 1 1 1 1 1
1 1 0 0 0 1 0 0 1
1 0 1 1 1 0 1 1 1
1 0 0 1 1 0 0 1 1
0 1 1 1 1 1 1 1 1
0 1 0 0 1 1 1 1 1
0 0 1 1 1 1 1 1 1
0 0 0 1 1 1 1 1 1
For (3):
(3)
1 1 0 0 1 1 1
1 0 0 1 1 1 1
0 1 1 0 0 0 1
0 0 1 1 1 1 1
If is a collection of sentential formulas, then a -proof is a
finite sequence 0, . . . , msuch that for each i m one of the
following conditions holds:
(a) i is a sentential axiom.
(b) i .
(c) There exist j, k < i such that k is j i. (Rule of modus
ponens, abbreviated MP).
We write if there is a -proof with last entry . We also write in
place of .
Proposition 1.6. (i) If , f is a sentential assignment, and [f ]
= 1 for all ,then [f ] = 1.
(ii) If , then is a tautology.
Proof. For (i), let 0, . . . , m be a -proof. Suppose that f is
a sentential assign-ment and [f ] = 1 for all . We show by complete
induction that i[f ] = 1 for alli m. Suppose that this is true for
all j < i.
7
-
Case 1. i is a sentential axiom. Then i[f ] = 1 by Proposition
1.5.Case 2. i . Then i[f ] = 1 by assumption.Case 3. There exist j,
k < i such that k is j i. By the inductive assumption,
k[f ] = j[f ] = 1. Hence i[f ] = 1.(ii) clearly follows from
(i),
Now we are going to show that, conversely, if is a tautology
then . This is a kindof completeness theorem, and the proof is a
highly simplified version of the proof of thecompleteness theorem
for first-order logic which will be given later.
Lemma 1.7. .
Proof.(a) [ [( ) ]] [[ ( )] ( )] (2)(b) [( ) ] (1)(c) [ ( )] ( )
(a), (b), MP(d) ( ) (1)(e) (c), (d), MP
Theorem 1.8. (The deduction theorem) If {} , then .
Proof. Let 0, . . . , m be a ( {})-proof with last entry . We
replace each iby several formulas so that the result is a -proof
with last entry .
If i is a logical axiom or a member of , we replace it by the
two formulas i ( i), i.
If i is , we replace it by the five formulas in the proof of
Lemma 1.7; the last entryis .
If i is obtained from j and k by modus ponens, so that k is j i,
we replacei by the formulas
[ (j i)] [( j) ( i)]
( j) ( i)
i
Clearly this is as desired.
Lemma 1.9. ( ).
Proof. By axiom (1) we have {,} . Hence axiom (3) gives {,} ,
and hence {,} . Now two applications of Theorem 1.8 give the
desiredresult.
Lemma 1.10. ( ) [( ) ( )].
Proof. Clearly { , , } , so three applications of Theorem 1.8
give thedesired result.
8
-
Lemma 1.11. ( ) .
Proof. Clearly { ,} and also { ,} , so by Lemma 1.9,{( ,} ( ).
Then Theorem 1.8 gives { } ( ), andso using axiom (3), { } ( ) .
Hence by Lemma 1.7, { } , andso Theorem 1.8 gives the desired
result.
Lemma 1.12. ( ) [( ) ].
Proof.
{ , ,} using axiom (1)
{ , ,} using axiom (3)
{ , ,} using Lemma 1.10
{ , ,} by Lemma 1.11
{ , ,}
{ , } by Theorem 1.8
{ , } by Lemma 1.11
Now two applications of Theorem 1.8 give the desired result.
Theorem 1.13. There is a sequence 0, 1, . . . containing all
sentential formulas.
Proof. One can obtain such a sequence by the following
procedure.
(1) Start with S0.
(2) List all sentential formulas of length at most two which
involve only S0 or S1; they areS0, S1, S0, and S1.
(3) List all sentential formulas of length at most three which
involve only S0, S1, or S2;they are S0, S1, S2, S0, S1, S2, S0, S1,
S2, S0 S0, S0 S1, S0 S2,S1 S0, S1 S1, S1 S2, S2 S0, S2 S1, S2
S2.
(4) Etc.
Theorem 1.14. If not( ), then there is a sentential assignment f
such that [f ] = 1for all , while [f ] = 0.
Proof. Let 0, 1, . . . list all the sentential formulas. We now
define 0,1, . . . byrecursion. Let 0 = . Suppose that i has been
defined. If not(i {i} ) then weset i+1 = i {i}. Otherwise we set
i+1 = i.
Here is a detailed proof that exists. Let M = { : is a function
with domain m
for some positive integer m, 1 = , and for every positive
integer i with i + 1 m wehave
i+1 =
{i {i} if not(i {i} ),i otherwise.}
9
-
(1) If , M with domains m, n respectively, with m n, then i m[i
= i].
This is easily proved by induction on i.
(2) For every positive integer m there is a M with domain m.
Again this is easily proved by induction on m.Now we define i =
i, where M and i < dmn(). This is justified by (1) and
(2).Now it is easily verified that the defining conditions for
hold.Let =
i i. By induction we have not(i ) for each i . In fact, wehave 0
= , so not(0 ) by assumption. If not(i ), then not(i+1 )
byconstruction.
Hence also not( ), since means that there is a -proof with last
entry ,and any -proof involves only finitely many formulas i, and
they all appear in some j ,giving j , contradiction.
() For any formula i, either i or i .
In fact, suppose that i / and i / . Say i = j . Then by
construction,i {i} and j {i} . So {i} and {i} . Henceby Theorem
1.8, i and i . So by Lemma 1.12 we get ,contradiction.
() If , then .
In fact, clearly not( {} ) by Theorem 1.8, so () follows.Now let
f be the sentential assignment such that f(i) = 1 iff Si . Now we
claim
( ) For every sentential formula , [f ] = 1 iff .
We prove this by induction on . It is true for = Si by
definition. Now suppose that itholds for . Suppose that ()[f ] = 1.
Thus [f ] = 0, so by the inductive assumption, / , and hence by (),
. Conversely, suppose that . If ()[f ] = 0,then [f ] = 1, hence by
the inductive hypothesis. Hence by Lemma 1.9, ,contradiction. So
()[f ] = 1.
Next suppose that () holds for and ; we show that it holds for .
Supposethat ( )[f ] = 1. If [f ] = 1, then by the inductive
hypothesis. By axiom(1), . Hence by (), ( ) . Suppose that [f ] =
0. Then [f ] = 0also, since ( )[f ] = 1. By the inductive
hypothesis and () we have . Hence by axiom (1), so by axiom (3). So
( ) by ().
Conversely, suppose that ( ) . Working for a contradiction,
suppose that( )[f ] = 0. Thus [f ] = 1 and [f ] = 0. So and by the
inductivehypothesis and (). Since ( ) and , we get . Since also ,
weget by Lemma 1.9, contradiction.
This finishes the proof of ( ).Since , ( ) implies that [f ] = 1
for all . Also [f ] = 0 since
/ .
Corollary 1.15. If [f ] = 1 whenever [f ] = 1 for all , then
.
10
-
Proof. This is the contrapositive of Theorem 1.14.
Theorem 1.16. iff is a tautology.
Proof. is given by Proposition 1.6(ii). follows from Corollary
1.15 by taking = .
Proposition 1.17.S0 S1 = 2, 3, 1, 4
and(S0 S1) (S1 S0) = 2, 2, 3, 4, 2, 1, 4, 1, 3.
Proof.
S0 S1 = 2S0 S1
= 231S1
= 2, 3, 1, 4;
(S0 S1) (S1 S0) = 2(S0 S1)
(S1 S0)
= 22S0 S1 2
S1 S0
= 2, 2, 3, 4, 21S1 1S0
= 2, 2, 3, 4, 2, 1, 4, 1, 3.
Proposition 1.18. There is a sentential formula of each positive
integer length.
Proof. If m is a positive integer, then
m1 times
1, 1, . . . , 1, S0
is a formula of length m, it ism1 times S0.
Proposition 1.19. m is the length of a sentential formula not
involving iff m is odd.
Proof. : We prove by induction on that if is a sentential
formula not involving, then the length of is odd. This is true of
sentential variables, which have length 1.Suppose that it is true
of and , which have length 2m+1 and 2n+1 respectively. Then , which
is 1, has length 1 + 2m+ 1 + 2n+ 1 = 2(m+ n+ 1) + 1, which isagain
odd. This finishes the inductive proof.. We construct formulas
without with length any odd integer by induction. S0
is a formula of length 1. If has been constructed of length 2m+
1, then S0 , whichis 1, S0, has length 2m+ 3. This finishes the
inductive construction.
11
-
Proposition 1.20. The truth table for a sentential formula
involving n basic formulashas 2n rows.
Proof. We prove this by induction on n. For n = 1, there are two
rows. Assume thatfor n basic formulas there are 2n rows. Given n+ 1
basic formulas, let be one of them.For the others, by the inductive
hypothesis there are 2n rows. For each such row thereare two
possibilities, 0 or 1, for . So for the n + 1 basic formulas there
are 2n 2 = 2n+1
rows.
Proposition 1.21. The formula
( ) ( )
is a tautology.
Proof.
( ) ( )
1 1 1 0 1 1
1 0 0 0 0 1
0 1 1 1 1 1
0 0 1 1 1 1
Proposition 1.22. The formula
[ ( )] [( ) ( )]
is a tautology.
Proof. Let be the indicated formula.
( ) ( ) ( )
1 1 1 1 1 1 1 1 1
1 1 0 1 1 1 0 1 1
1 0 1 1 1 1 0 1 1
1 0 0 1 1 1 0 1 1
0 1 1 1 1 1 1 1 1
0 1 0 1 0 0 0 0 1
0 0 1 0 1 0 0 0 1
0 0 0 0 0 0 0 0 1
12
-
Proposition 1.23. The formula
( ) ( )
is not a tautology.
Proof.
( ) ( )
1 1 1 0 0 0
Proposition 1.24. The following is a tautology:
S0 (S1 (S2 (S3 S1))).
Proof. Suppose that f is an assignment making the indicated
formula false; we worktowards a contradiction. Thus
(1) S0[f ] = 1 and
(2) (S1 (S2 (S3 S1)))[f ] = 0.
From (2) we get
(3) S1[f ] = 1 and
(4) (S2 (S3 S1))[f ] = 0.
From (4) we get
(5) S2[f ] = 1 and
(6) (S3 S1)[f ] = 0.
From (6) we get S1[f ] = 0, contradicting (3).
Proposition 1.25. The following is a tautology.
({[( ) ( )] } ) [( ) ( )].
Proof. Suppose that f is an assignment which makes the given
formula false; wewant to get a contradiction. Thus we have
(1) ({[( ) ( )] } )[f ] = 1 and
(2) [( ) ( )][f ] = 0.
By (2) we have
(3) ( )[f ] = 1 and
13
-
(4) ( )[f ] = 0.
By (4) we have
(5) [f ] = 1 and
(6) [f ] = 0.
By (3) and (6) we get
(7) [f ] = 0.
By (1) and (7) we get
(8) {[( ) ( )] }[f ] = 0.
It follows that
(9) [( ) ( )][f ] = 1 and
(10) [f ] = 0.
Now by (6) we have
(11) ( )[f ] = 1,
and hence by (9),
(12) ( )[f ] = 1.
By (5) we have
(13) ()[f ] = 0,
and hence by (12),
(14) ()[f ] = 0.
This contradicts (10).
Proposition 1.26. The following statements are logically
consistent: If the contract isvalid, then Horatio is liable. If
Horation is liable, he will go bankrupt. Either Horatio willgo
bankrupt or the bank will lend him money. However, the bank will
definitely not lendhim money.
Proof. Let S0 correspond to the contract is valid, S1 to Horatio
is liable, S2 toHoratio will go bankrupt, and S3 to the bank will
lend him money. Then we want tosee if there is an assignment of
values which makes the following sentence true:
(S0 S1) (S1 S2) (S2 S3) S3.
We can let f(0) = f(1) = f(2) = 1 and f(3) = 0, and this gives
the sentence the value1.
Proposition 1.27. {} .
14
-
Proof. Following the proof of Lemma 1.9, the following is a
{,}-proof:
(a) (b) ( ) (1)(c) (a), (b), MP(d) ( ) ( ) (3)(e) (c), (d),
MP(f) (g) (e), (f), MP
Now applying the proof of the deduction theorem, the following
is a {}-proof:
(a) [ [( ) ]] [[ ( )] ( )] (2)
(b) [( ) ] (1)(c) [ ( )] ( ) (a), (b), MP(d) ( ) (1)(e) (c),
(d), MP(f) [ ( )] [ [ ( )]] (1)(g) ( ) (1)(h) [ ( )]] (f), (g),
MP(i) [( ) ( )] [ [( ) ( )]] (1)(j) ( ) ( ) (3)(k) [( ) ( )] (i),
(j), MP(l) [ [( ) ( )]] [[ ( )]
[ ( )]] (2)(m) [ ( )] [ ( )] (k), (l), MP(n) ( ) (g), (m), MP(o)
( ) (1)(p) (q) (o), (p), MP(r) [ ( )] [( ) ( )] (2)(s) ( ) ( ) (n),
(r), MP(t) (q), (s), MP
15
-
2. First-order logic
Although set theory can be considered within a single
first-order language, with only non-logical constant , it is
convenient to have more complicated languages, corresponding tothe
many definitions introduced in mathematics.
All first-order languages have the following symbols in common.
Again, as for senten-tial logic, we take these to be certain
natural numbers.
1 (negation)2 (implication)3 (the equality symbol)4 (the
universal quantifier)5m for each positive integer m (variables
ranging over elements, but not subsets, of a givenstructure) We
denote 5m by vm1. Thus v0 is 5, v1 is 10, and in general vi is 5(i+
1).
Special first-order languages have additional symbols for the
functions and relations andspecial elements involved. These will
always be taken to be some positive integers notamong the above;
thus they are positive integers greater than 4 but not divisible by
5. Sowe have in addition to the above logical symbols some
non-logical symbols:
Relation symbols, each of a certain positive rank.Function
symbols, also each having a specified positive rank.Individual
constants.
Formally, a first-order language is a quadruple (Rel, Fcn, Cn,
rnk) such that Rel, Fcn, Cnare pairwise disjoint subsets of M (the
sets of relation symbols, function symbols, andindividual
constants), and rnk is a function mapping RelFcn into the positive
integers;rnk(S) gives the rank of the symbol S.
Now we will define the notions of terms and formulas, which give
a precise formu-lation of meaningful expressions. Terms are certain
finite sequences of symbols. A termconstruction sequence is a
sequence 0, . . . , m1, m > 0, with the following properties:for
each i < m one of the following holds:
i is vj for some natural number j.
i is c for some individual constant c.
i is F0 1
n1 for some n-place function symbol F, with each j equal to kfor
some k < j, depending upon j.
A term is a sequence appearing in some term construction
sequence. Note the similarityof this definition with that of
sentential formula given in Chapter 1.
Frequently we will slightly simplify the notation for terms.
Thus we might writesimply vj , or c, or F0 . . . n1 for the
above.
The following two propositons are very similar, in statement and
proof, to Propositions1.1 and 1.2. The first one is the principle
of induction on terms.
Proposition 2.1. Let T be a collection of terms satisfying the
following conditions:
16
-
(i) Each variable is in T .(ii) Each individual constant is in T
.(iii) If F is a function symbol of rank m and 0, . . . , m1 T ,
then also F0 . . . m1
T .
Then T consists of all terms.
Proof. Let be a term. Say that 0, . . . , m1 is a term
construction sequenceand i = . We prove by complete induction on j
that j T for all j < m; hence T . Suppose that j < m and k T
for all k < j. If j = vs for some s, thenj T . If j = c for some
individual constant c, then sj T . Finally, suppose thatj is Fk0 .
. . kn1 with each kt < j. Then kt T for each t < n by the
inductivehypothesis, and it follows that j T . This completes the
inductive proof.
Proposition 2.2. (i) Every term is a nonempty sequence.(ii) If
is a term, then exactly one of the following conditions holds:
(a) is an individual constant.(b) is a variable.(c) There exist
a function symbol F, say of rank m, and terms 0, . . . , m1
such
that is F0 . . . m1.(iii) No proper initial segment of a term is
a term.(iv) If F and G are function symbols, say of ranks m and n
respectively, and if
0, . . . , m1, 0, . . . , n1 are terms, and if F0 . . . m1 is
equal to G0, . . . n1, thenF = G, m = n, and i = i for all i <
m.
Proof. (i): This is clear since any entry in a term construction
sequence is nonempty.(ii): Also clear.(iii): We prove this by
complete induction on the length of a term. So suppose that
is a term, and for any term shorter than , no proper initial
segment of is a term.We consider cases according to (ii).
Case 1. is an individual constant. Then has length 1, and any
proper initialsegment of is empty; by (i) the empty sequence is not
a term.
Case 2. is a variable. Similarly.Case 3. There exist an m-ary
function symbol F and terms 0, . . . , m1 such that
is F0 . . . m1. Suppose that is a term which is a proper initial
segment of . By(i), is nonempty, and the first entry of is F. By
(ii), has the form F0 . . . m1 forcertain terms 0, . . . , m1.
Since both 0 and 0 are shorter terms than , and one ofthem is an
initial segment of the other, the induction hypothesis gives 0 = 0.
Let i < mbe maximum such that i = i. Since is a proper initial
segment of , we must havei < m 1. But i+1 and i+1 are shorter
terms than and one is a segment of the other,so by the inductive
hypthesis i+1 = i+1, contradicting the choice of i.
(iv): F is the first entry of F0 . . . m1 and G is the first
entry of G0, . . . n1, soF = G. Then by (ii) we get m = n. By
induction using (iii), each i = i.
We now give the general notion of a structure. This will be
modified and extended for settheory later. For a given first-order
language L = (Rel, Fcn, Cn, rnk), an L -structure isa quadruple A =
(A,Rel, F cn, Cn) such that A is a nonempty set (the universe of
the
17
-
structure), Rel is a function assigning to each relation symbol
R a rnk(R)-ary relationon A, i.e., a collection of rnk(R)-tuples of
elements of A, Fcn is a function assigning toeach function symbol F
a rnk(F)-ary opeation on A, i.e., a function assigning a value in
Ato each rnk(F)-tuple of elements of A, and Cn is a function
assigning to each individual
constant c an element of A. Usually instead of Rel(R), Fcn(F)
and Cn(c) we write RA,
FA, and cA.Now we define the meaning of terms. This is a
recursive definition, similar to the
definition of the values of sentential formulas under
assignments:
Proposition 2.3. Let A be a structure, and a a function mapping
into A. (A is theuniverse of A.) Then there is a function F mapping
the set of terms into A with thefollowing properties:
(i) F (vi) = ai for each i .
(ii) F (c) = cA for each individual constant c.
(iii) F (F0 . . . m1) = FA(F (0), . . . , F (m1)) for every
m-ary function symbol F
and all terms 0, . . . , m1
Proof. An (A, a)-term sequence is a sequence (0, b0), . . . ,
(m1, bm1) such thateach bi A and for each i < m one of the
following conditions holds:
(1) i is vj and bi = aj .
(2) i is c for some individual constant c, and bi = cA.
(3) i = Fk(0) k(n1) and bi = F
A(bk(0), . . . , bk(n1)) for some n-ary function
symbol F and some k(0), . . . , k(n 1) < i.
Now we claim
(4) For any term and any (A, a)-term sequences
(0, b0), . . . , (m1, bm1) and (0, b
0), . . . , (
n1, b
n1)
such that m1 = n1 = we have bm1 = b
n1.
We prove (4) by induction on , thus using Proposition 2.1. If =
vi, then bm1 =
ai = bn1. If is an individual constant c, then bm1 = cA = bn1.
Finally, if =
F0 p1, then we have:
m1 = Fk(0)
k(p1) and bm1 = FA(bk(0), . . . , bk(p1));
n1 = F l(0)
l(p1) and bm1 = F
A(bl(0), . . . , bl(p1))
with each k(s) and l(t) less than i. By Proposition 2.2(iv) we
have k(s) = l(s) for every
s < p. Now for every s < p we can apply the inductive
hypothesis to k(s) and thesequences
(0, b0), . . . , (k(s), bk(s)) and (0, b
0), . . . , (
l(s), b
l(s))
18
-
to obtain bk(s) = bl(s). Hence
bm1 = FA(bk(0), . . . , bk(p1)) = F
A(bl(0), . . . , bl(p1)) = b
n1,
completing the inductive proof of (4).
(5) If 0, . . . , m1 is a term construction sequence, then there
is an (A, a)-term sequenceof the form (0, b0), . . . , (m1,
bm1.
We prove this by induction on m. For m = 1 we have two
possibilities.Case 1. 0 is vj for some j . Then (0, bj) is as
desired.
Case 2. 0 is c, an individual constant. Then (0, cA) is as
desired.
Now assume the statement for m 1 1. By the induction hypothesis
there is an
(A, a)-term sequence of the form def= (0, b0), . . . , (m2,
bm2). Then we have three
possibilities:Case 1. m1 is vj for some j . Then (m1, bj) is as
desired.
Case 2. m1 is c, an individual constant. Then (m1, cA) is as
desired.
Case 3. m1 is Fk(0) k(p1) for some p-ary function symbol F with
each
k(s) < i. Then (m1,FA(bk(0), . . . , bk(p1)) is as
desired.
So (5) holds.Now we can define F as needed in the Proposition.
Let be a term. Let 0, . . . , m1
be a term construction sequence with m1 = . By (5), let (0, b0),
. . . , (m1bm1) bean (A, a)-term sequence. Then we define F () =
bm1. This definition is unambigu-ous by (4). Now we check the
conditions of the Proposition. Let be a term, and let(0, b0), . . .
, (m1, bm1) be an (A, a)-term sequence with m1 = .
Case 1. = vj for some j . Then F () = bm1 = aj .
Case 2. = c for some individual constant c. Then F () = bm1 =
cA.
Case 3. = F0 p1 with F a p-ary function symbol and each s a
term.
Then there exist c(0), . . . , c(p 1) < m 1 such that s =
c(s) for every s < p. Then
F (c(s)) = bc(s) = Ac(s) for each s < p, and hence
F () = bm1 = FA(bs(0), . . . , bs(p1)) = F
A(As(0), . . . , As(p1)) = F
A(A0 , . . . , Ap1).
With F as in Proposition 2.3, we denote F () by A(a). Thus
vAi (a) = ai;
cA(a) = cA;
(F0 . . . m1)A(a) = FA(A0 (a), . . . ,
Am1(a)).
Here vi is any variable, c any individual constant, and F any
function symbol (of somerank, say m).
19
-
What A(a) means intuitively is: replace the individual constants
and function sym-bols by the actual members of A and functions on A
given by the structure A, and replacethe variables vi by
coresponding elements ai of A; calculate the result, giving an
elementof A.
Proposition 2.4. Suppose that is a term, A is a structure, a, b
assignments, and
a(i) = b(i) for all i such that vi occurs in . Then A(a) =
A(b).
Proof. By induction on :
cA(a) = cA = cA(b);
vAi (a) = a(i) = b(i) = vAi (b);
(F0 . . . m1)A(a) = FA(A0 (a), . . . ,
Am1(a))
= FA(A0 (b), . . . , Am1(b))
= (F0 . . . m1)A(b).
The last step here is the induction step (many of them, one for
each function symbol andassociated terms). The inductive assumption
is that a(i) = b(i) for all i for which vi occursin F0 . . . m1;
hence also for each j < m, a(i) = b(i) for all i for which vi
occurs in j ,so that the inductive hypothesis can be applied.
This proposition enables us to simplify our notation a little
bit. If n is such that each
variable occurring in has index less than n, then in the
notation A(a) we can just usethe first n entries of a rather than
the entire infinite sequence.
We turn to the definition of formulas. For any terms , we define
= to be the sequence3 . Such a sequence is called an atomic
equality formula. An atomic non-equalityformula is a sequence of
the form R0
m1 where R is an m-ary relation symboland 0, . . . m1 are terms.
An atomic formula is either an atomic equality formula or anatomic
non-equality formula.
We define , a function assigning to each sequence of symbols of
a first-order
language the sequence def= 1. is the function assigning to each
pair (, ) of
sequences of symbols the sequence def= 2. is the function
assigning to
each pair (i, ) with i and a sequence of symbols the sequence
videf= 4, 5i+5.
A formula construction sequence is a sequence 0, . . . , m1 such
that for each i < mone of the following holds:
(1) i is an atomic formula.
(2) There is a j < i such that i is j
(3) There are j, k < i such that i is j k.
(4) There exist j < i and k such that i is vkj .
A formula is an expression which appears as an entry in some
formula construction se-quence.
20
-
The following is the principle of induction on formulas.
Proposition 2.5. Suppose that is a set of formulas satisfying
the following conditions:(i) Every atomic formula is in .(ii) If ,
then .(iii) If , , then ( ) .(iv) If and i , then vi .
Then is the set of all formulas.
Proof. It suffices to take any formula construction sequence 0,
. . . , m1 and showby complete induction on i that i for all i .
So, suppose that i < m and j for all j < i. By the definition
of formula construction sequence, we have the followingcases.
Case 1. i is an atomic formula. Then i by (i).Case 2. There is a
j < i such that i is j . By the inductive hypothesis, j .
Hence by (ii), i .Case 2. There are j, k < i such that i is j
k. By the inductive hypothesis,
j and k . Hence by (iii), i .Case 4. There exist j < i and k
such that i is vkj . By the inductive
hypothesis, j . Hence by (iv), i .This completes the inductive
proof.
Proposition 2.6. (i) Every formula is a nonempty sequence.(ii)
If is a formula, then exactly one of the following conditions
holds:
(a) is an atomic equality formula, and there are terms , such
that is = .(b) is an atomic non-equality formula, and there exist a
positive integer m, a
relation symbol R of rank m, and terms 0, . . . , m1, such that
is R0 . . . m1.(c) There is a formula such that is .(d) There are
formulas , such that is .(e) There exist a formula and a natural
number i such that is vi.
(iii) No proper initial segment of a formula is a formula.(iv)
(a) If is an atomic equality formula, then there are unique terms ,
such that
is = .(b) If is an atomic non-equality formula, then there exist
a unique positive integer
m, a unique relation symbol R of rank m, and unique terms 0, . .
. , m1, such that isR0 . . . m1.
(c) If is a formula and the first symbol of is 1, then there is
a unique formula such that is .
(d) If is a formula and the first symbol of is 2, then there are
unique formulas, such that is .
(e) If is a formula and the first symbol of is 4, then there
exist a unique naturalnumber i and a unique formula such that is
vi.
Proof. (i): First note that this is true of atomic formulas,
since an atomic formulamust have at least a first symbol 3 or some
relation symbol. Knowing this about atomic
21
-
formulas, any entry in a formula construction sequence is
nonempty, since the entry iseither an atomic formula or else begins
with 1,2, or 4.
(ii): This is true on looking at any entry in a formula
construction sequence: eitherthe entry begins with 3 or a relation
symbol and hence (a) or (b) holds, or it begins with1, 2, or 4,
giving (c), (d) or (e). Only one of (a)(e) holds because of the
first symbol inthe entry.
(iii): We prove this by complete induction on the length of the
formula. Thus supposethat is a formula of length m, and for any
formula of length less than m, no properinitial segment of is a
formula. Suppose that is a proper initial segment of and isa
formula; we want to get a contradiction. By (ii) we have several
cases.
Case 1. is an atomic equality formula = for certain terms , .
Thus is3 . Since is a formula which begins with 3 (since is an
initial segment of andis nonempty by (i)), (ii) yields that is 3
for some terms , . Hence = .Thus is an initial segment of or is an
initial segment of . By Proposition 2.2(iii) itfollows that = .
Then also = , so = , contradiction.
Case 2. is an atomic non-equality formula R0 . . . m1 for some
m-ary relationsymbol R and some terms 0, . . . , m1. Then is a
formula which begins with R, and sothere exist terms 0, . . . , m1
such that is R0 . . . m1. By induction using Proposition2.2(iii), i
= i for all i < m, so = , contradiction.
Case 3. is for some formula . Then 1 is the first entry of , so
by (ii) hasthe form for some formula . Thus is a proper initial
segment of , contradicting theinductive hypothesis, since is
shorter than .
Case 4. is for some formulas , , i.e., it is 2. Then starts with
2,so by (ii) has the form 2 for some formulas , . Now both and are
shorterthan , and one is an initial segment of the other. So = by
the inductive assumption.Then is a proper initial segment of ,
contradicting the inductive assumption.
Case 5. is 4, 5(i+ 1) for some i and some formula . Then by
(ii), is4, 5(i+ 1) for some formula . So is a proper initial
segment of , contradiction.
(iv): These conditions follow from Proposition 2.2(iii) and
(iii).
Now we come to a fundamental definition connecting language with
structures. Again thisis a definition by recursion; it is given in
the following proposition. First a bit of notation.If a : A, i ,
and s A, then by ais we mean the sequence which is just like
aexcept that ais(i) = s.
Proposition 2.7. Suppose that A is an L -structure. Then there
is a function G assigningto each formula and each sequence a : A a
value G(, a) {0, 1}, such that
(i) For any terms , , G( = , a) = 1 iff A(a) = A(a).
(ii) For each m-ary relation symbol R and terms 0, . . . , m1,
G(R0 . . . m1, a) =
1 iff A0 (a), . . . , Am1(a) R
A.
(iii) For every formula , G(, a) = 1G(, a).
(iv) For all formulas , , G( , a) = 0 iff G(, a) = 1 and G(, a)
= 0.
(v) For all formulas and any i , G(vi, a) = 1 iff for every s A,
G(, ais) = 1.
22
-
Proof. An (A, a)-formula sequence is a sequence (0, b0), . . . ,
(m1, bm1) such
that each bs is a function mapping Mdef= {a : a : A} into {0, 1}
and for each i < m
one of the following holds:
(1) i is an atomic equality formula = , and a M [bi(a) = 1 iff
A(a) = A(a)].
(2) i is an atomic nonequality formula R0 . . . m1, and
a M [bi(a) = 1 iff A0 (a), . . . ,
Am1(a) R
A].
(3) There is a j < i such that i = j , and a M [bi(a) = 1
bj(a)].
(4) There are j, k < i such that i = j k, and a M [bi(a) = 0
iff bj(a) = 1 andbk(a) = 0].
(5) There are j < i and k such that i = vkj , and a M [bi(a)
= 1 iff u A[bj(a
ku) = 1]].
Now we claim
(6) For any formula and any (A, a)-formula sequences
(0, b0), . . . , (m1, bm1) and (0, b
0), . . . , (
n1, b
n1)
such that m1 = n1 = we have bm1 = b
n1.
We prove (6) by induction on , thus using Proposition 2.5. First
suppose that is anatomic equality formula = . Then the desired
conclusion is clear. Similarly for atomicnonequality formulas. Now
suppose that is . Then by Proposition 2.6(c) there arej < m and
k < n such that = j =
k. By the inductive hypothesis we have bj = b
k,
and hence a M [bm1(a) = 1 bj(a) = 1 bk(a) = bn1(a)], so that bm1
= bn1. Next
suppose that is . Then by Proposition 2.6(d) there are j, k <
m 1 such that = j and = k, and there are s, t < n 1 such that =
s and =
t. Then bj = b
s
and bk = bt by the inductive hypothesis. Hence for any a M ,
bm1(a) = 0 iff bj(a) = 1 and bk = 0 iff bs = 1 and b
t = 0 iff b
n1(a) = 0.
Thus bm1 = bn1. Finally, suppose that is vk. Then by Proposition
2.6(e) there are
j, s < i such that j = and s = . So by the inductive
hypothesis bj = b
s. Hence for
any a M we have
bm1(a) = 1 iff for every u A[bj(aku) = 1]
iff for every u A[bs(aku) = 1]
iff bn1(a) = 1.
Thus bm1 = bn1, finishing the proof of (6).
(7) If 0, . . . , m1 is a formula construction sequence, then
there is an (A, a)-formulasequence of the form (0, b0), . . . ,
(m1, bm1).
23
-
We prove (7) by induction on m. For m = 1 we have two
possibilities.
Case 1. 0 is an atomic equality formula = . Let b0(a) = 1 iff
A(a) = A(a).
Case 2. 0 is an atomic nonequality formula R0 . . . , xm1. Let
b0(a) = 1 iff
A0 (a), . . . , Am1 R
A.
Now assume the statement in (7) for m 1 1. By the inductive
hypothesis there is an
(A, a)-formula sequence of the form def= (0, b0), . . . , (m2,
bm2). Then we have these
possibilities for m1.
Case 1. m1 is = for some terms , . Define bm1(a) = 1 iff A(a) =
A(a).
Then (m1, bm1) is as desired.Case 2. m1 is R0 . . . p1 for some
terms 0, . . . , p1. Define bm1(a) = 1 iff
A0 (a), . . . , Ap1(a) R
A. Then (m1, bm1) is as desired.Case 3. m1 is i with i < m 1.
Define bm1(a) = 1 bi(a) for any a. Then
(m1, bm1) is as desired.Case 4. m1 is i j with i, j < m 1.
Define bm1(a) = 0 iff bi(a) = 1 and
bj(a) = 0. Then (m1, bm1) is as desired.
Case 5. m1 is vki with i < m 1. Define bm1(a) iff for all u
A, bi(aku) = 1.Then (m1, bm1) is as desired.
This completes the proof of (7).Now we can define the function G
needed in the Proposition. Let be a formula
and a : A. Let 0, . . . , m1 be a formula construction sequence
with m1 = .By (7) let (0, b0), . . . , (m1, bm1) be an (A,
a)-formula sequence. Then we defineG(, a) = bm1(a). The conditions
in the Proposition are clear.
With G as in Proposition 2.7, we write A |= [a] iff G(, a) = 1.
A |= [a] is read: A isa model of under a or A models under a or is
satisfied by a in A or holdsin A under the assignment a. In
summary:
A |= ( = )[a] iff A(a) = A(b). Here and are terms.
A |= (R0 . . . m1)[a] iff the m-tuple A0 , . . . , Am1 is in the
relation R
A. Here R is anm-ary relation symbol, and 0, . . . , m1 are
terms.
A |= ()[a] iff it is not the case that A |= [a].
A |= ( )[a] iff either it is not true that A |= [a], or it is
true that A |= [a].(Equivalently, iff (A |= [a] implies that A |=
[a]).
A |= (vi)[a] iff A |= [ais] for every s A.
We define some additional logical notions:
is the formula ; is called the disjunction of and .
is the formula ( ); is called the conjunction of and .
is the formula ( ) ( ); is called the equivalence between
and.
24
-
vi is the formula vi; is the existential quantifier.
These notions mean the following.
Proposition 2.8. Let A be a structure and a : A.(i) A |= ( )[a]
iff A |= [a] or A |= [a] (or both).(ii) A |= ( )[a] iff A |= [a]
and A |= [a].(iii) A |= ( )[a] iff (A |= [a] iff A |= [a]).(iv) A
|= vi[a] iff there is a b A such that A |= [a
ib].
Proof. The proof consists in reducing the statements to ordinary
mathematical usage.(i):
A |= ( )[a] iff A |= ( )[a]
iff either it is not true that A |= ()[a] or it is true that A
|= [a]
iff not(not(A |= [a])) or A |= [a]
iff A |= [a] or A |= [a].
(ii):
A |= ( )[a] iff not(A |= ( )[a])
iff not(not(A |= [a]) or A |= [a])
iff not(not(A |= [a]) or not(A |= [a]))
iff A |= [a] and A |= [a].
(iii):
A |= ( )[a] iff A |= (( ) ( ))[a]
iff A |= (( )[a] and A |= ( ))[a]
iff (A |= [a] implies that A |= [a]) and
(A |= [a] implies that A |= [a])
iff (A |= [a] iff A |= [a]).
(iv):
A |= vi[a] iff A |= vi[a]
iff not(for all b A(A |= [aib]))
iff not(for all b A(not(A |= [aib]))
iff there is a b A such that A |= [aib].
We say that A is a model of iff A |= [a] for every a : A. If is
a set of formulas,we write |= iff every structure which models each
member of also models . |= means that every structure models . is
then called universally valid.
25
-
Now we want to apply the material of Chapter 1 concerning
sentential logic. By definition,a tautology in a first-order
language is a formula such that there exist formulas 0, 1, . . .and
a sentential tautology such that is obtained from by replacing each
symbol Sioccurring in by i, for each i < .
Theorem 2.9. If is a tautology in a first-order language, then
holds in every structurefor that language.
Proof. Let A be any structure, and b : A any assignment. We want
to showthat A |= [b]. Let formulas 0, 1, . . . , be given as in the
above definition. For eachsentential formula , let be the
first-order formula obtained from by replacing eachsentential
variable Si by i. Thus
is . We define a sentential assignment f by setting,for each i
,
f(i) =
{
1 if A |= i[b],0 otherwise.
Then we claim:
(*) For any sentential formula , A |= [b] iff [f ] = 1.
We prove this by induction on :
If is Si, then is i, and our condition holds by definition. If
inductively is , then
is , and
A |= [b] iff not(A |= [b])
iff not( [f ] = 1)
iff [f ] = 0
iff [f ] = 1.
Finally if inductively is , then is , and
A |= [b] iff (A |= [b] implies that A |= [b]
iff [f ] = 1 implies that [f ] = 1
iff [f ] = 1.
This finishes the proof of (*).Applying (*) to , we get A |=
[b], i.e., A |= [b].
A language for the structure (,
-
Proposition 2.11. In any first-order language, the sequence v0,
v0 is not a term.
Proof. Suppose that v0, v0 is a term. This contradicts
Proposition 2.2(ii).
Proposition 2.12. In the language for (, S, 0,+, ), the sequence
+, v0, v1, v2 is not aterm. Here S(i) = i+ 1 for any i .
Proof. Suppose it is a term. By Proposition 2.2(ii)(c), there
are terms , such that+, v0, v1, v2 is + . Thus v0, v1, v2 = . So
the term v0 is an initial segmentof the term . By Proposition
2.2(iii) it follows that v0 = . Hence v1, v2 = . Thiscontradicts
Proposition 2.2(ii).
The structure (, S, 0,+, ) can be put in the general framework
of structures as follows.It can be considered to be the structure
(,Rel, F cn, Cn) where Rel = , Cn is thefunction with domain {8}
such that Cn(8) = 0, and Fcn is the function with domain{6, 7, 9}
such that Fcn(6) = S, Fcn(7) = +, and Fcn(9) = .
Proposition 2.13. In the language for the structure (,+), a term
has length m iff m isodd.
Proof. First we show by induction on terms that every term has
odd length. Thisis true for variables. Suppose that it is true for
terms and . Then also + has oddlength. Hence every term has odd
length.
Second we prove by induction on m that for all m, there is a
term of length 2m+ 1.A variable has length 1, so our assertion
holds for m = 0. Assume that there is a term of length 2m+ 1. Then
+ v0 has length 2m+ 3. This finishes the inductive proof.
Proposition 2.14. In the language for (Q,+, ) the formula def=
v1[v0 v1 = v1] is such
that for any a : Q, (Q,+, ) |= [a] iff a0 = 1.
Proposition 2.15. The following formula holds in a structure,
under any assignment,iff the structure has at least 3 elements.
v0v1v2((v0 = v1) (v0 = v2) (v1 = v2)).
Proposition 2.16. The following formula holds in a structure,
under any assignment,iff the structure has exactly 4 elements.
v0v1v2v3((v0 = v1) (v0 = v2) (v0 = v3) (v1 = v2) (v1 = v3) (v2 =
v3)
v4(v0 = v4 v1 = v4 v2 = v4 v3 = v4)).
Proposition 2.17. The following formula in the language for
(,
-
Proposition 2.18. The formula
v0 = v1 (Rv0v2 Rv1v2)
is universally valid, where R is a binary relation symbol.
Proof. Let A be a structure and a : A an assignment. Suppose
that A |= (v0 =
v1)[a]. Then a0 = a1. Also suppose that A |= Rv0v2[a]. Then (a0,
a2) RA. Hence
(a1, a2) RA. Hence A |= Rv1v2[a], as desired.
Proposition 2.19. The formula
v0 = v1 v0(v0 = v1)
is not universally valid.
Proof. Consider the structure Adef= (,
-
3. Proofs
The purpose of this chapter is give the definition of a
mathematical proof, and give thesimplest proofs which will be
needed in proving the completeness theorem in the nextchapter.
Given a set of formulas in a first-order language, and a formula in
thatlanguage, we explain what it means to have a proof of from
.
The following formulas are the logical axioms. Here , , are
arbitrary formulasunless otherwise indicated.
(L1a) ( ).(L1b) [ ( )] [( ) ( )].(L1c) ( ) ( ).(L2) vi( ) (vi
vi), for any i .(L3) vi for any i such that vi does not occur in
.(L4) vi(vi = ) if is a term and vi does not occur in .(L5) = ( = =
), where , , are terms.(L6) = ( = = ), where , , are terms.(L7) =
F0 . . . i1i+1 . . . m1 = F0 . . . i1i+1 . . . m1, where F is an
m-aryfunction symbol, i < m, and , , 0, . . . , i1, i+1, . . .
m1 are terms.(L8) = (R0 . . . i1i+1 . . . m1 R0 . . . i1i+1 . . .
m1), where R is anm-ary relation symbol, i < m, and , , 0, . . .
, i1, i+1, . . . m1 are terms.
Theorem 3.1. Every logical axiom is universally valid.
Proof. (L1ac): Universally valid by Theorem 2.9.(L2): Assume
that
(1) A |= vi( )[a] and(2) A |= vi[a];
We want to show that A |= vi[a]. To this end, take any b A; we
want to show thatA |= [aib]. Now by (1) we have A |= ( )[a
ib], hence A |= [a
ib] implies that A |= [a
ib].
Now by (2) we have A |= [aib], so A |= [aib].
(L3): We prove by induction on that if vi does not occur in ,
and if a, b : Aare such that a(j) = b(j) for all j 6= i, then A |=
[a] iff A |= [b]. This will imply that(L3) is universally
valid.
is = . Thus vi does not occur in or in . Then
A |= ( = )[a] iff A(a) = A(a)
iff A(b) = A(b) by Proposition 2.4
iff A |= ( = )[b].
is R0 . . . m1 for some m-ary relation symbol and some terms 0,
. . . , m1. We areassuming that vi does not occur in R0 . . . m1;
hence it does not occur in any term i.
A |= (R0 . . . m1)[a] iff A0 (a), . . . ,
Rm1(a) R
A
29
-
iff A0 (b), . . . , Rm1(b) R
A
(by Proposition 2.4)
iff A |= (R0 . . . m1)[b].
is (inductively).
A |= [a] iff not(A |= [a])
iff not(A |= [b]) (inductive hypothesis)
iff A |= [b].
is (inductively).
A |= [a] iff (A |= [a] implies that A |= [a])
iff (A |= [b] implies that A |= [b])
(inductive hypothesis)
iff A |= [b].
is vk (inductively). By symmetry it suffices to prove just one
direction. Supposethat A |= [a]; we want to show that A |= [b]. To
this end, suppose that u A; we wantto show that A |= [bku]. Since A
|= [a], we have A |= [a
ku]. Now k 6= i, since vi does
not occur in . Hence (aku)(j) = (bku)(j) for all j 6= i. Hence A
|= [b
ku] by the inductive
hypothesis, as desired.This finishes our proof by induction of
the statement made above. Now assume that
A |= [a] and u A; we want to show that A |= [aiu]. This holds by
the statement above.This finishes the proof of (L3).(L4): Suppose
that is a term and vi does not occur in . To prove that A |=
(vi(vi = ))[a], we want to find u A such that A |= (vi = )[aiu].
Let u = A(a). Then
(vi)A[aiu] = u =
A(a) = A(aiu)
by Proposition 2.4 (since vi does not occur in , hence a(j) =
aiu(j) for all j such that vj
occurs in ). Hence A |= (vi = )[aiu].
(L5): Assume that A |= ( = )[a] and A |= ( = )[a]. Then A(a) =
A(a) and
A(a) = A(a), so A(a) = A(a), hence A |= ( = )[a].
(L6): Assume that A |= ( = )[a] and A |= ( = )[a]. Then A(a) =
A(a) and
A(a) = A(a), so A(a) = A(a), hence A |= ( = )[a].
(L7): Assume that A |= ( = )[a]. Then A(a) = A(a), and so
(F0 . . . i1i+1 . . . m1)A(a) = FA(A0 (a), . . . ,
Ai1(a),
A(a), Ai+1(a), . . . , Am1(a))
= FA(A0 (a), . . . , Ai1(a),
A(a), Ai+1(a), . . . , Am1(a))
= (F0 . . . i1i+1 . . . m1)A(a);
30
-
it follows that A |= (F0 . . . i1i+1 . . . m1 = F0 . . . i1i+1 .
. . m1)[a], hence (L7)is universally valid.
(L8): Assume that A |= ( = )[a]. Then A(a) = A(a). Assume
that
A |= (R0 . . . i1i+1 . . . m1)[a]; hence
A0 (a), . . . , Ai1(a),
A(a), Ai+1(a), . . . , Am1(a) R
A; hence
A0 (a), . . . , Ai1(a),
A(a), Ai+1(a), . . . , Am1(a) R
A; hence
A |= (R0 . . . i1i+1 . . . m1)[a];
hence (L8) is universally valid.
Now let be a set of formulas. A -proof is a finite sequence 0, .
. . , m1 of formulassuch that for each i < m one of the
following conditions holds:
(I1) i is a logical axiom(I2) i .(I3) (modus ponens) There are
j, k < i such that j is the formula k i.(I4) (generalization)
There exist j < i and k such that i is the formula vkj .
Then we say that proves , in symbols , provided that is an entry
in some-proof. We write in place of .
Theorem 3.2. If , then |= .
Proof. Recall the notion |= from Chapter 2: it says that for
every structure Afor the implicit language we are dealing with, if
A |= [a] for all and all a : A,then A |= [a] for every a : A. Now
it suffices to take a -proof 0, . . . , m1 andprove by complete
induction on i that |= i for each i < m.
Case 1. i is a logical axiom. Then the result follows by Theorem
3.1.Case 2. i . Obviously then |= i.Case 3. There are j, k < i
such that j is k i. Suppose that A is a model of
and a : A. Then A |= k[a] by the inductive hypothesis, and also
A |= (k i)[a]by the inductive hypothesis. Thus A |= k[a] implies
that A |= i[a], so A |= i[a].
Case 3. There exist j < i and k such that i is vkj . Given u
A, we want toshow that A |= j [aku]; but this follows from the
inductive hypothesis.
One form of the completeness theorem, proved in the next
chapter, is that, conversely, |= implies that .
In this chapter we will show that many definite formulas are
such that . We beginwith tautologies.
Lemma 3.3. for any first-order tautology .
Proof. Let be a sentential tautology, and let 0, 1, . . . be a
sequence of first-orderformulas such that is obtained from by
replacing each sentential variable Si by i. Foreach sentential
formula , let be obtained from by replacing each sentential
variable
31
-
Si by i. By Theorem 1.16, (in the sentential sense). Hence there
is a sententialproof 0, . . . , m with m = . We claim that 0, . . .
,
m is a first-order proof. Since
m = = , this will prove the lemma. If i m and i is a
(sentential) axiom, then i is
the corresponding first-order axiom:
[ ( )] = [ ( )];
[[ ( ] [( ) ( )]] =
[[ ( )] [( ) ( )]];
[( ) ( )] = [( ) ( )].
If j, k < i and k is j i, then k is j
i.
We proceed with simple theorems concerning equality.
Proposition 3.4. = for any term .
Proof. The following is a -proof; on the left is the entry
number, and on the righta justification. Let vi be a variable not
occurring in .
(1) vi = (vi = = ) (L5)(2) [vi = (vi = = ) [( = ) (vi = )]
(taut.)(3) ( = ) (vi = ) ((1), (2), MP)(4) vi[( = ) (vi = )] ((3),
gen.)(5) vi[( = ) (vi = )] [vi( = ) vi(vi = )] (L2)(6) vi( = )
vi(vi = ) (4), (5), MP(7) ( = ) vi( = ) (L3)(8) (7) [(6) [( = )
vi(vi = )] (taut.)(9) (6) [( = ) vi(vi = )] (7), (8), MP(10) ( = )
vi(vi = ) (6), (9), MP(11) (10) [vi(vi = ) = ] (taut.)(12) vi(vi =
) = (10), (11), MP(13) vi(vi = ) (L4)(14) (13) [(12) = ] (L1)(15)
(12) = ((13), (14), MP)(16) = ((12), (15), MP)
Proposition 3.5. = = for any terms , .
Proof. By (L5) we have
= ( = = );
and by Proposition 3.4 we have = . Now
= ([ = ( = = )] ( = = ))
32
-
is a tautology, so = = .
Proposition 3.6. = ( = = ) for any terms , , .
Proof. By (L5), = ( = = ). By Proposition 3.5, = = .Now
( = = ) ([ = ( = = )] [ = ( = = )])
is a tautology, so = ( = = ).
We now give several results expressing the principle of
substitution of equals for equals.The main fact is expressed in
Theorem 3.16, which says that under certain conditions theformula =
( ) is provable, where is obtained from by replacing
someoccurrences of by .
Lemma 3.7. If and are terms, and are formulas, vi is a variable
not occurringin or , and = ( ), then = (vi vi).
Proof.
(1) vi[ = ( )] (hypothesis, gen.)(2) vi( = ) vi( )] (from (1),
using (L2))(3) vi( ) (vi vi) ((L2))(4) = vi( = ). ((L3))
Now putting (2)(4) together with a tautology gives the
lemma.
To proceed further we need to discuss the notion of free and
bound occurrences of variablesand terms. This depends on the notion
of a subformula. Recall that a formula is just afinite sequence of
positive integers, subject to certain conditions. Atomic equality
formulashave the form = for some terms , , and = is defined to be 3
. Atomic non-equality formulas have the form R0 . . . m1 for some
m, some m-ary relation symbol R,and some terms 0, . . . , m1. R is
actually some positive integer k greater than 5 and notdivisible by
5, and R0 . . . m1 is the sequence k0
m1. Non-atomic formulashave the form
= 1, = 2, orvs = 4, 5(s+ 1).
Thus every formula begins with one of the integers 1,2,3,4 or
some positive integer greaterthan 5 not divisible by 5 which is a
relation symbol. This helps motivate the followingpropositions.
Proposition 3.8. If = 0, . . . , k1 is a term, then each i is
either of the form 5mwith m a positive integer, or it is an odd
integer greater than 5 which is a function symbolor individual
constant.
33
-
Proof. We prove this by induction on , thus using Proposition
2.1. The propositionis obvious if is a variable or individual
constant. Suppose that F is a function symbol ofrank m, 0, . . . ,
m1 are terms, and is F0 . . . m1, where we assume the truth of
theproposition for 0, . . . , m1. Suppose that i < k. If i = 0,
then i is F, a function symbol.If i > 0, then i is an entry in
some j, and the desired conclusion follows by the
inductivehypothesis.
Proposition 3.9. Let = 0, . . . k1 be a formula, suppose that i
< k, and i is oneof the integers 1,2,3,4 or a positive integer
greater than 5 which is a relation symbol. Thenthere is a unique
segment i, i+1, . . . , j of which is a formula.
Proof. We prove this by induction on , thus using Proposition
2.5. We assume thehypothesis of the proposition. First suppose that
is an atomic equality formula = with and terms. Thus = is the
sequence 1 . Now by Proposition 2.2(ii), noentry of a term is among
the integers 1, 2, 3, 4 or is a positive integer greater than 5
whichis a relation symbol. It follows from the assumption about i
that i = 0, and hence thedesired segment of is itself. It is unique
by Proposition 2.6(iii). Second suppose that is an atomic
non-equality formula R0 . . . m1 with R an m-ary relation symbol
and0, . . . , m1 terms. This is very similar to the first case. R0
. . . m1 is the sequenceR0
m1. By Proposition 2.2(ii) i must be 0, and hence the desired
segment of is itself. It is unique by Proposition 2.6(iii).
Now assume inductively that is ; so is 1. If i = 0, then itself
isthe desired segment, unique by Proposition 2.6(iii). If i > 0,
then i = i1, where = 0, . . . , k1. By the inductive hypothesis
there is a segment i1, i, . . . , j of which is a formula. This
gives a segment i, i+1, . . . , j+1 of which is a formula; itis
unique by Proposition 2.6(iii).
Assume inductively that is for some formulas , . So is 2. Ifi =
0, then itself is the required segment, unique by Proposition
2.6(iii). Now supposethat i > 0. Now we have = 1, . . . , m and
= m+1, . . . , k1 for some m. If1 i m, then by the inductive
assumption there is a segment i, i+1, . . . , n of which is a
formula. This is also a segment of , and it is unique by
Proposition 2.6(iii). Ifm+ 1 i k 1, a similar argument with gives
the desired result.
Finally, assume inductively that is vs with some formula and s .
If i = 0then itself is the desired segment, unique by Proposition
2.6(iii). If i > 0 then actuallyi > 1 so that i is within ,
and the inductive hypothesis applies.
The segment of asserted to exist in Proposition 3.9 is called
the subformula of beginning
at i. For example, consider the formula def= v0[v0 = v2 v0 =
v2]. The formula v0 = v2
occurs in two places in . In detail, is the sequence 4, 5, 2, 3,
5, 15, 3, 5, 15. Thus
0 = 4;1 = 5;2 = 2;3 = 3;4 = 5;5 = 15;
34
-
6 = 3;7 = 5;8 = 15;
On the other hand, v0 = v2 is the formula 3, 5, 15. It occurs in
beginning at 3, andalso beginning at 6.
Now a variable vs is said to occur bound in at the j-th position
iff with =0, . . . , m1, we have j = vs and there is a subformula
of of the form vs =i, i+1, . . . , m with i + 1 j m. If a variable
vs occurs at the j-th position of but does not occur bound there,
then that occurrence is said to be free. We give someexamples. Let
be the formula v0 = v1 v1 = v2. All the occurrences of v0, v1, v2
arefree occurrences in . Note that as a sequence is 2, 3, 5, 10, 3,
10, 15; so 0 = 2, 1 = 3,2 = 5, 3 = 10, 4 = 3, 5 = 10, and 6 = 15.
The variable v0, which is the integer 5,occurs free at the 2-nd
position. The variable v1, which is the integer 10, occurs free
atthe 3rd and 5th positions. The variable v2, which is the integer
15, occurs free at the 6thposition.
Now let be the formula v0 = v1 v1(v1 = v2). Then the first
oc-curence of v1 is free, but the other two occurrences are bound.
As a sequence, is2, 3, 5, 10, 4, 10, 3, 10, 15. The variable v1
occurs free at the 3rd position, and bound atthe 5th and 7th
positions.
We also need the notion of a term occurring in another term, or
in a formula. The followingtwo propositions are proved much like
3.9.
Proposition 3.10. If = 0, . . . , m1 is a term and i < m,
then there is a uniqueterm which is a segment of beginning at
i.
Proof. We prove this by induction on . For a variable or
individual constant,we have m = 1 and so i = 0, and itself is the
only possibility for . Now supposethat the proposition is true for
terms 0, . . . n1, F is an n-ary function symbol, and is F0 . . .
n1. If i = 0, then itself begins at i, and it is the only term
beginning at iby Proposition 2.2(iii). If i > 0, then i is
inside some term k, and so by the inductiveassumption there is a
term which is a segment of k beginning there; this term is a
segmentof too, and it is unique by Proposition 2.2(iii).
Under the assumptions of Proposition 3.10, we say that occurs in
beginning at i.
Proposition 3.11. If = 0, . . . , m1 is a formula, i < m, and
i is a variable, anindividual constant, or a function symbol, then
there is a unique segment of beginningat i which is a term.
Proof. We prove this by induction on . First suppose that is an
atomic equalityformula = for some terms , . Thus is 3 . So i >
0, and hence i is inside or . If i is inside , then by Proposition
3.10, there is a term which is a segment of beginning at i; it is
also a segment of , and it is unique by Proposition 2.2(iii).
Similarlyfor .
Suppose inductively that is . Thus is 1. It follows that i >
0, so that iappears in ; then the inductive hypothesis applies.
35
-
Suppose inductively that is . Thus is 2. It follows that i >
0, sothat i appears in or ; then the inductive hypothesis
applies.
Finally, suppose that is vs with a formula and s . Thus is 4,
5(s+1).Hence i > 0. If i = 1, then 5(s+1) is the desired
segment, unique by Proposition 2.6(iii).Suppose that i > 1. So i
is an entry in and hence by the inductive assumption, thereis a
segment i, i+1, . . . m which is a term; this is also a segment of
, and it is uniqueby Proposition 2.6(iii).
Under the assumptions of Proposition 3.11, we say that the
indicated segment occurs in beginning at i.
We now extend the notions of free and bound occurrences to
terms. Let be a term whichoccurs as a segment in a formula . Say
that = 0, . . . , m1 and = i, . . . k. Wesay that this occurrence
of in is bound iff there is a variable vs which occurs bound in at
some place t with i t k; the occurrence of is free iff there is no
such variable.
We give some examples. The term v0 + v1 is bound in its only
occurrence in theformula v0(v0 + v1 = v2). The same term is bound
in its first occurrence and free in itssecond occurrence in the
formula v0(v0 + v1 = v2) v0 + v1 = v0.
Suppose that , , are terms, and occurs in beginning at i. By the
result ofreplacing that occurrence of by we mean the following
sequence . Say , , havedomains (lengths) m,n, p respectively. Then
is the sequence
0, . . . , i1, 0, . . . , p1, i+n, . . . , m1.
Put another way, if is with of length i, then is .
Proposition 3.12. Suppose that , , are terms, and the sequence
is obtained from by replacing one occurrence of by . Then is a
term.
Proof. We prove this by induction on , thus by using Proposition
2.1. If is avariable or an individual constant, then must be
itself, and is , which is a term.Now suppose that is F0 . . . m1
for some m-ary function symbol F and some terms0, . . . , m1, and
the proposition holds for 0, . . . , m1. Say the occurrence of in
begins at i. If i = 0, then equals , and hence equals , which is a
term. If i > 0,then i is inside some j , and hence the
occurrence of is actually an occurrence in j byProposition
2.2(iii). Replacing this occurrence of in j by we obtain a term by
theinductive hypothesis; call this term j . It follows that is F0 .
. . j1
j , j+1 . . . m1,
which is a term.
As an example, consider the term v0 (v1 + v2) in the language
for (Q,+, ). Replacingthe occurrence of v1 by v0 v1 we obtain the
term v0 ((v0 v1) + v2). Writing this outin detail, assuming that
corresponds to 9 and + corresponds to 7, we start with thesequence
9, 5, 7, 10, 15 and end with the sequence 9, 5, 7, 9, 5, 10,
15.
Our first form of subsitution of equals for equals only involves
terms:
Theorem 3.13. If , , are terms, and is a sequence obtained from
by replacing anoccurrence of in by , then is a term and = = .
36
-
Proof. is a term by Proposition 3.12. Now we proceed by
induction on . If isa variable or an individual constant, then must
be the same as , since has length 1and occurs in . Then is , and =
= is = = , a tautology. So theproposition is true in this case.
Now assume inductively that is F0 . . . m1 with F an m-ary
function symbol and0, . . . , m1 terms. There are two possibilities
for the occurrence of . First, possibly isthe same as . Then is ,
and again we have the tautology = = , Second, theoccurrence of is
within some i. Then by the inductive hypothesis, = i = i,where i is
obtained from i by replacing the indicated occurrence of by . Now
aninstance of (L7) is
i = i F0 . . . i1 . . . ii+1 . . . m1 = F0 . . . i1 . . .
ii+1 . . . m1.
Putting this together with = i = i and a tautology gives = =
.
Proposition 3.14. Suppose that is a formula and , are terms.
Suppose that occursat the i-th place in , and if i > 0 and i1 =
, then is a variable. Let the sequence be obtained from by
replacing that occurrence of by . Then is a formula.
Proof. Induction on . Suppose that is = . Then by Proposition
3.13, occurs in or . Suppose that it occurs in . Let be obtained
from by replacing thatoccurrence of by . Then is a term by
Proposition 3.14. Since is = , is aformula. The case in which
occurs in is similar. Now suppose that is R0 . . . m1with R an
m-ary relation symbol and 0, . . . , m1 are terms. Then the
occurrence of is within some i. Let
i be obtained from i by replacing that occurrence by . Now
is R0 . . . i1i . . . m1, so is a formula.
Now suppose that the result holds for , and is . Then occurs in
, so if is obtained from by replacing the occurrence of by , then
is a formula by theinductive assumption. Since is also is a
formula.
Next, suppose that the result holds for and , and is . Then
theoccurrence of is within or is within . If it is within , let be
obtained from
by replacing that occurrence of by . Then is a formula by the
inductive hypothesis.Since is , also is a formula. If the
occurrence is within , let be obtainedfrom by replacing that
occurrence of by . Then is a formula by the inductivehypothesis.
Since is , also is a formula.
Finally, suppose that the result holds for , and is vk. If i =
1, then is vk,and by hypothesis is some variable vl. Then is vl,
which is a formula. If i > 1, then occurs in , so if is obtained
from by replacing the occurrence of by , then
is a formula by the inductive assumption. Since is vk also is a
formula.
For the exact definition of see the description before
Proposition 3.12.
Lemma 3.15. Suppose that and are terms, is a formula, and is
obtained from by replacing one free occurrence of in by , such that
the occurrence of that resultsis free in . Then = ( ).
37
-
Proof. We proceed by induction on . First suppose that is an
atomic equalityformula = . If the occurrence of that is replaced is
in , let be the resulting term.Then by Proposition 3.13, = = . Now
(L5) gives = ( = = ).Putting these two together with a tautology
gives = ( = = ). Bysymmetry, = ( = = ). Hence = ( = = ).
If the occurrence of that is replaced is in , a similar argument
using (L6) works.Second, suppose that is an atomic non-equality
formula R0 . . . m1, with R an m-
ary relation symbol and 0, . . . , m1 terms. Say that the
occurrence of that is replacedby is in i, the resulting term
being
i. Then by Proposition 3.13, = i =
i.
By (L8) we have
i = i (R0 . . . m1 R0 . . . i1
ii+1 . . . m1),
so by a tautology we get from these two facts
= (R0 . . . m1 R0 . . . i1ii+1 . . . m1),
and by symmetry
= (R0 . . . i1ii+1 . . . m1 R0 . . . m1),
and then another tautology gives
= (R0 . . . m1 R0 . . . i1ii+1 . . . m1),
This finishes the atomic cases. Now suppose inductively that is
. The occurrence of in that is replaced actually occurs in ; let be
the result of replacing that occurrenceof by . Now the occurrence
of in is free in . In fact, suppose that vi is asubformula of which
has as a segment the indicated occurrence of , and vi occurs in.
Then vi is also a subformula of , contradicting the assumption that
the occurrenceof is free in . Similarly the occurrence of in which
replaced the occurrence of is free. So by the inductive hypothesis,
= ( ), and hence a tautology gives = ( ), i.e., = ( ).
Suppose inductively that is .Case 1. The occurrence of in is
within . Let be obtained from by replacing
that occurrence by , such that that occurrence is free in ,
hence free in . By theinductive hypothesis, = ( ). Since is , a
tautology gives thedesired result.
Case 2. The occurrence of in is within . Let be obtained from by
replacingthat occurrence by , such that that occurrence is free in
, hence free in . By theinductive hypothesis, = ( ). Since is , a
tautology gives the desiredresult.
Finally, suppose that is vi. Then the occurrence of in that is
replaced isin . Let be obtained from by replacing that occurrence
of by . The occurrenceof in must be free since it is free in , as
in the treatment of above; similarly
38
-
for and . Hence by the inductive hypothesis, = ( ). Now since
theoccurrence of in is free, the variable vi does not occur in .
Similarly, it does notoccur in . Hence by Proposition 3.7 and
tautologies we get = (vi vi),i.e., = ( ).
The hypothesis that the term is still free in the result of the
replacement in this propo-sition is necessary for the truth of the
proposition. This hypothesis is equivalent to sayingthat the
occurrence of which is replaced is not inside a subformula of of
the form viwith vi a variable occurring in .
Theorem 3.16. (Substitution of equals for equals) Suppose that
is a formula, is aterm, and occurs freely in starting at indices
i(0) < < i(m 1). Also suppose that is a term. Let be obtained
from by replacing each of these occurrences of by ,and each such
occurrence of is free in . Then = ( ).
Proof. We prove this by induction on m. If m = 0, then is the
same as , andthe conclusion is clear. Now assume the result for m,
for any . Now assume that occurs freely in starting at indices i(0)
< < i(m), and no such occurrence is inside asubformula of of
the form vj with vj a variable occurring in . Let be obtained from
by replacing the last occurrence of , the one beginning at i(m), by
. By Proposition3.15, = ( ). Now we apply the inductive hypothesis
to and the occurrencesof starting at i(0), . . . , i(m 1); this
gives = ( ). Hence a tautology gives = ( ), finishing the inductive
proof.
Proposition 3.17. Suppose that , , are formulas, and the
sequence is obtained from by replacing an occurrence of in by .
Then is a formula.
Proof. Induction on . If is atomic, then is equal to , and is
equal to andhence is a formula. Suppose the result is true for and
is . If = , again thedesired conclusion is clear. Otherwise the
occurrence of is within the subformula .If is obtained from by
replacing that occurrence by , then is a formula by theinductive
hypothesis. Since is , also is a formula.
Now suppose the result is true for and , and is . If = , again
thedesired conclusion is clear. Otherwise the occurrence of is
within the subformula oris within the subformula . If it is within
and is obtained from by replacing thatoccurrence by , then is a
formula by the inductive hypothesis. Since is , also is a formula.
If it is within and is obtained from by replacing that occurrenceby
, then is a formula by the inductive hypothesis. Since is , also is
aformula.
Finally, suppose the result is true for and is vi. If = , again
the desiredconclusion is clear. Otherwise the occurrence of is
within the subformula . If isobtained from by replacing that
occurrence by , then is a formula by the inductivehypothesis. Since
is vi, also is a formula.
For the exact meaning of see the description before Proposition
3.12.
Another form of the substitution of equals by equals principle
is as follows:
39
-
Theorem 3.18. Let , , be formulas, and let be obtained from by
replacing anoccurrence of in by . Suppose that . Then .
Proof. Induction on . If is atomic, then is the same as , and
the conclusion isclear. Suppose inductively that is . If is equal
to , then is equal to and theconclusion is clear. Suppose that
occurs within , and let be obtained from byreplacing that
occurrence by . Assume that . Then by the inductive hypothesis , so
, as desired.
The case in which is is similar. Finally, suppose that is vi,
and occurs within . Let be obtained from by replacing that
occurrence by . Assumethat . Then by the inductive assumption. So
by a tautology, ,and then by generalization vi( ). Using (L2) we
then get vi vi.Similarly, vi vi. Hence using a tautology, vi
vi.
Now we work to prove two important logical principles: changing
bound variables, anddropping a universal quantifier in favor of a
term.
For any formula , i , and term by Subfvi we mean the result of
replacingeach free occurrence of vi in by . We now work towards
showing that under suitableconditions, the formula vi Subf
vi is provable. The supposition expressed in the first
sentence of the following lemma will be eliminated later on.
Lemma 3.19. Suppose that vi does not occur bound in , and does
not occur in the term.
Assume that no free occurrence of vi in is within a subformula
of of the formvj with vj a variable occurring in . Then vi Subf
vi .
Proof.
(1) vi = ( Subfvi ) (by Proposition 3.16 and a tautology)
(2) (Subfvi (vi = )) (using a tautology)(3) vi[ (Subf
vi (vi = ))] (generalization)
(4) vi vi(Subfvi (vi = )) (using (L2))
(5) vi(Subfvi (vi = )) (viSubf
vi vi(vi = )) ((L2))
(6) vi (viSubfvi vi(vi = )) ((4), (5), a tautology)
(7) vi(vi = ) (vi viSubfvi ) ((6), a tautology)
(8) vi(vi = ) ((L4))(9) vi viSubf
vi ) ((7), (8), modus ponens)
(10) Subfvi viSubfvi ((L3))
(11) vi Subfvi ((9), (10), a tautology)
Lemma 3.20. If i 6= j, is a formula, vi does not occur bound in
, and vj does notoccur in at all, then vi vjSubf
vivj.
Proof.
vi Subfvivj (by Lemma 3.19)
vjvi vjSubfvivj (using (L2) and a tautology)
40
-
vi vjvi (by (L3)) vi vjSubf
vivj
Lemma 3.21. If i 6= j, is a formula, vi does not occur bound in
, and vj does notoccur in at all, then vi vjSubf
vivj.
Proof. By Proposition 3.20 we have vi vjSubfvivj. Now vj does
not occur
bound in Subfvivj and vi does not occur in Subfvivj at all.
Hence by Proposition 3.20
again, vjSubfvivj viSubf
vjvi
Subfvivj. Now Subfvjvi
Subfvivj is actually just itself; so vjSubf
vivj vi. Hence the proposition follows.
For i, j and a formula, by Subbvivj we mean the result of
replacing all boundoccurrences of vi in by vj . By Proposition 3.14
this gives another formula.
Proposition 3.22. If vi occurs bound in a formula , then there
is a subformula vi of such that vi does not occur bound in .
Proof. Induction on . Note that the statement to be proved is an
implication. If is atomic, then vi cannot occur bound in ; thus the
hypothesis of the implication isfalse, and so the implication
itself is true. Now suppose inductively that is , and vioccurs
bound in . Then it occurs bound in , and so by the inductive
hypothesis, hasa subformula vi such that vi does not occur bound in
. T