OXFORD UNIVERSITY

Part C Mathematics, Mathematics and Philosophy, Mathematics and Computation
M.Sc. in Mathematics and Foundations of Computer Science
B.Phil. and M.St. in Philosophy
16 lectures on
G¨ odel’s Incompleteness Theorems
Michaelmas Term 2010
Daniel Isaacson
Faculty of Philosophy
Oxford University
1st December 2010
Copyright c ( 2010 by Daniel Isaacson
All rights reserved. No part of this publication may be reproduced
without prior permission by anyone other than for their own use in
studying this subject. Enquiries and corrections to
daniel.isaacson@philosophy.ox.ac.uk
Contents
0 Background: first-order logic and formal systems 2
0.1 First-order formal languages with identity . . . . . . . . . . . . . . . 2
0.2 Interpretations of first-order formal languages with identity; truth of
a first-order formula in an interpretation; logical validity and logical
consequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
0.3 A system of natural deduction for first-order logic with identity . . . 4
0.4 Prenex normal form . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
0.4.1 Model theory and proof theory . . . . . . . . . . . . . . . . . 11
0.5 Completeness of a system of natural deduction with respect to first-
order logical consequence . . . . . . . . . . . . . . . . . . . . . . . . . 11
0.5.1 Lindenbaum’s Lemma . . . . . . . . . . . . . . . . . . . . . . 11
0.5.2 The Completeness Theorem . . . . . . . . . . . . . . . . . . . 11
0.5.3 The Compactness Theorem for first-order logic . . . . . . . . . 11
0.5.4 The existence of non-standard models of the truths of arithmetic 11
0.5.5 Completeness of other systems of derivation with respect to
first-order logical consequence . . . . . . . . . . . . . . . . . . 11
1 Introduction: a weak form of G¨odel’s First Incompleteness Theo-
rem; the symbols and expressions of a language for arithmetic L
E
;
G¨odel numbering of the expressions of L
E
12
1.1 Introduction: a weak form of G¨odel’s First Incompleteness Theorem 12
1.2 The symbols and expressions of a language for arithmetic L
E
. . . . . 16
1.3 G¨odel numbering of the expressions of L
E
. . . . . . . . . . . . . . . 17
2 Terms and formulas of the language L
E
; expressibility of diagonal
substitution in the language L
E
20
2.1 Terms and formulas of the language L
E
. . . . . . . . . . . . . . . . . 20
2.1.1 Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.1.2 Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.1.3 Free and bound variables; open formulas and sentences . . . . 22
1
CONTENTS 2
2.2 Designation by terms in L
E
, truth of sentences of L
E
, and express-
ibility of sets and relations of natural numbers by formulas of L
E
. . 23
2.2.1 Designation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2.2 Truth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.2.3 Expressibility . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 Concatenation of numbers in a given base notation is Arithmetical. . 24
2.4 Substitution and diagonal functions, and their arithmetization . . . . 25
3 The Diagonal Lemma; expressibility of properties of sequence num-
bers 28
3.1 The Diagonal Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2 Expressibility of properties of sequence numbers in the language L
E
29
3.2.1 Properties of sequences of digits . . . . . . . . . . . . . . . . . 29
3.2.2 Sequence numbers . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 Coding of finite sequences of G¨odel numbers . . . . . . . . . . . . . . 31
4 A formal system PA
E
for arithmetic; an Arithmetical proof predi-
cate for PA
E
; a weak version of G¨odel’s first incompleteness theo-
rem for PA
E
34
4.1 A formal system PA
E
for arithmetic . . . . . . . . . . . . . . . . . . . 34
4.2 An Arithmetical proof predicate for PA
E
. . . . . . . . . . . . . . . . 37
4.3 An inefficient and a weak version of G¨odel’s First Incompleteness
Theorem for PA
E
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5 The system PA with zero, successor, addition, multiplication, and
≤ as primitive; Σ
0
- and Σ
1
-formulas; a Σ
0
-coding of finite sets of
ordered pairs; the relation x
y
= z is ∆
1
-expressible in the language
of PA 42
5.1 The system PA with zero, successor, addition, multiplication, and ≤
as primitive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.2 Σ
0
-formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.3 Σ
1
- and Π
1
-formulas; Σ
1
- Π
1
- and ∆
1
-relations . . . . . . . . . . . . . 44
5.4 Arithmetization of syntax in the language of PA . . . . . . . . . . . . 46
5.5 A Σ
0
-coding of finite sets of ordered pairs of numbers . . . . . . . . . 47
5.6 The relation x
y
= z is ∆
1
-expressible in the language of PA . . . . . . 49
6 Every Σ-formula is provably equivalent to a Σ
1
-formula; the arith-
metized proof predicate for PA is Σ
1
; the arithmetical hierarchy 50
6.1 Σ-formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.2 The arithmetized proof predicate for PA is Σ
1
. . . . . . . . . . . . . 52
6.3 The arithmetical hierarchy . . . . . . . . . . . . . . . . . . . . . . . . 53
CONTENTS 3
7 Σ
0
-completeness and Σ
1
-completeness; weak systems of arithmetic
Q and R (without induction); Σ
0
-completeness of systems R, Q, and
PA; Σ
0
-soundness and Σ
1
-soundness 55
7.1 Σ
0
-completeness and Σ
1
-completeness . . . . . . . . . . . . . . . . . . 55
7.2 Weak systems of arithmetic Q and R (without induction) . . . . . . . 57
7.3 Σ
0
-completeness of systems R, Q, and PA . . . . . . . . . . . . . . . 59
7.4 Σ
0
-soundness and Σ
1
-soundness . . . . . . . . . . . . . . . . . . . . . 62
8 The notions of consistency, ω-consistency and 1-consistency; incom-
pleteness from the assumption of 1-consistency; truth of the G¨odel
sentence; ω-incompleteness. 63
8.1 The notions of consistency, ω-consistency and 1-consistency. . . . . . 63
8.2 Incompleteness of PA from the assumption of 1-consistency . . . . . . 67
8.3 Truth of the G¨odel sentence . . . . . . . . . . . . . . . . . . . . . . . 68
8.4 PA is ω-incomplete . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
9 Enumerability and the Separation Lemma; incompleteness of PA
from the assumption of consistency (Rosser’s Theorem); weak and
strong definability of a function in a system; formal provability of
the Diagonal Lemma 71
9.1 Enumerability and the Separation Lemma . . . . . . . . . . . . . . . 72
9.2 Incompleteness of PA from the assumption of consistency (Rosser’s
Theorem) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
9.3 Weak and strong definability of a function in a system . . . . . . . . 74
9.4 Formal provability of the Diagonal Lemma . . . . . . . . . . . . . . . 76
10 Arithemization of consistency; provability predicates; G¨odel’s Sec-
ond Incompleteness Theorem; L¨ob’s Theorem; analyzing and strength-
ening the First Incompleteness Theorem 79
10.1 Arithmetization of the statement that a system S is consistent . . . . 79
10.2 Provability predicates. . . . . . . . . . . . . . . . . . . . . . . . . . . 80
10.3 G¨odel’s Second Incompleteness Theorem . . . . . . . . . . . . . . . . 82
10.4 L¨ob’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
10.5 Analyzing and strengthening the First Incompleteness Theorem . . . 84
10.5.1 S ∼ G
S
cannot be proved from the consistency of S . . . . . 85
10.5.2 Strengthened second half of the First Incompleteness Theorem 85
10.5.3 Consistency of S∪¦ConS¦ is strictly weaker than 1-consistency
of S . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
11 Provable Σ
1
-completeness 88
CONTENTS 4
12 The ω-rule and uniform reflection; PA proves that PA proves every
instance of the G¨odel sentence; Π
1
-uniform reflection and consis-
tency; PA is Π
1
-conservative over PA
Π
2
∪ ¦Con
PA
¦ 98
12.1 The ω-rule and uniform reflection . . . . . . . . . . . . . . . . . . . . 98
12.2 PA proves that PA proves every instance of the G¨odel sentence . . . . 100
12.3 Equivalence of Π
1
-Uniform Reflection and consistency . . . . . . . . . 101
12.4 PA is Π
1
-conservative over PA
Π
2
∪ ¦Con
PA
¦ . . . . . . . . . . . . . . 102
13 Provability logic: the system GL 104
13.1 The language of GL . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
13.2 The axioms and inference rules of GL . . . . . . . . . . . . . . . . . . 105
13.3 Some derivations in GL . . . . . . . . . . . . . . . . . . . . . . . . . . 107
13.4 Closure of GL under substitution by provably equivalent formulas . . 108
13.5 Closure of GL under substitution of provably equivalent formulas is
provable in GL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
13.6 Strengthened proof that the closure of GL under substitution of prov-
ably equivalent formulas is provable in GL . . . . . . . . . . . . . . . 111
14 The fixed-point theorem for GL 114
14.1 The notion of a sentence letter modalized in a sentence, and arithme-
tized substitution for modalized sentences . . . . . . . . . . . . . . . 114
14.2 The fixed point theorem for GL . . . . . . . . . . . . . . . . . . . . . 117
Lecture 0
Background: first-order logic and
formal systems
This unspoken lecture reviews essential results on first-order logic and formal systems
as background to the development of G¨odel’s incompleteness theorems.
The notion of logical consequence is the starting point. A rough characterization
of this notion is the following. A sentence φ is a logical consequence of a set of
sentences Γ, symbolized as Γ φ, if and only if φ is true in every interpretation of
the language of Γ∪¦φ¦ in which all the sentences of Γ are true. If Γ is empty, Γ φ
reduces to the condition that φ is true in every interpretation of the language of φ,
which is to say, φ is logically valid, symbolized by φ.
Accordingly, we need to make precise the notion of a formal language and the
notion of interpretation of a formal langauge.
0.1 First-order formal languages with identity
The notion of a formal language begins from the specification of a finite alphabet
of symbols which are strung together (concatenated) to produce the expressions of
the language. The symbols of the alphabet of a formal language have no intrinsic
meaning (they are purely formal symbols) but are chosen with the intention that
they should be interpretable in certain ways (which in general does not rule out
their being interpreted in other ways). Some but not all of the expressions of the
language constitute well-formed expressions. Which expressions are well-formed is
a matter of stipulation, according to our intended use of the formal language. The
symbols of a first-order language are of three sorts, logical, non-logical, and syntactic.
The logical symbols include two sorts: propositional, and quantificational, and may
include a third, identity. The non-logical are of two sorts, function symbols and
predicate or relation symbols. Functions and relations are of a particular ’arity’
5
LECTURE 0 6
(unary, binary, ternary, and so on). A zero-ary function symbol is a constant term,
a zero-ary relation symbol represents a sentence. The shape of a symbol is completely
arbitrary, though there are some conventional choices, e.g. the universal quantifier
is usually written “∀”, but used often to be written “( )” (the brackets enclosing
the variable of quantification), and is sometimes written “

” (called a California
quantifier). But the symbols can be anything and in particular, as we shall exploit,
they can be digits also used to generate numerals.
We shall in this course always have full first-order classical (as opposed to intu-
itionistic) predicate logic with identity as our background logic. Quite apart from
the shapes of the symbols, there are choices to be made as to which are our prim-
itive propositional functions and quantifiers. In our background logic we shall take
as our primitives of propositional logic negation, conjunction, disjunction, and im-
plication, for which the symbols will be ∼, ∧, ∨, and ⊃, and both the universal
and existential quantifiers as primitive, written as ∀ and ∃. As you will know from
a previous logic course, we could take as primitive just negation with any one of
conjunction, disjunction, and implication and either the universal or the existential
quantifier, and in our official formal language we shall take as primitive ∼, ⊃, and ∀.
We shall operate with the langauge as if it also contained symbols for conjunction,
disjunction, and the existential quantifiers, but strictly (A ∨ B) will be an abbrevi-
ation for (∼ A ⊃ B), etc. There is also the connective “if and only if”, which we
will symbolize as ≡. Even in our background logic we will take (A ≡ B) to be an
abbreviation for ((A ⊃ B) ∧ (B ⊃ A)).
Stipulating that our background logic is first-order classical logic with identity,
means that we have a primitive symbol for identity, which we will take as the
common symbol for identity = (though others are sometimes used, e.g. ≈, ≏,
.
=),
but having a symbol and first-order axioms for identity is not sufficient to make a
logical system first-order logic with identity. Identity is a two place relation which
every object bears to itself and to no other object. The first property is easily
expressed: ∀v
1
v
1
= v
1
, but the second property is not easily expressed, in particular
not expressed by ∀v
1
∀v
2
(v
1
= v
2
⊃ v
1
= v
2
) (if v
1
is a different object from v
2
then
v
1
= v
2
).
terms in the language
formulas in the language
LECTURE 0 7
0.2 Interpretations of first-order formal languages
with identity; truth of a first-order formula in
an interpretation; logical validity and logical
consequence
0.3 A system of natural deduction for first-order
logic with identity
This system is based on natural deduction as developed by Gerhard Gentzen. In
Gentzen-style systems of natural deduction, deductions consist of branching trees.
For ease of formulation on the page, I give here a variant form of natural deduction in
which deductions are linear
1
. I shall call this system LND, standing either for Linear
Natural Deduction (but in that case be aware that this usage has nothing to do
with Girard’s usage in what he calls Linear Logic), or Lemmon Natural Deduction.
[Strictly I should call this system something like CLND, for Classical, as opposed
to Intuitionistic, Linear Natural Deduction, but since in this course our background
logic is always classical and not intuitionistic, I won’t mark the distinction.]
Deductions are generated by using the Rule of Assumption and fourteen Rules of
Inference. There are no axioms. The rules of inference come in pairs, an Introduction
and an Elimination rule for each one of the seven logical constants ∧, ∨, ⊃, ∼, ∀, ∃,
and =. We take (F ≡ G) to be defined as ((F ⊃ G) ∧ (G ⊃ F)). The first formula
of a Deduction is always an Assumption; later formulas can also be Assumptions.
An Introduction Rule for a given logical constant results in a formula that has that
logical constant as its main logical constant and is deduced from formulas that that
do not have that logical constant as their main constant (though that constant may
occur in a subformula). An Elimination Rule for a given logical constant deduces a
formula that does not have that logical constant as its main logical constant from a
formula that has that logical constant as its main logical constant.
A deduction consists of a numbered sequence of formulas, with each of which
is associated on the left a finite set of numbers (possibly empty) which are the
numbers of the formulas that constitute the assumptions on which the formula in
that line depends, and on the right notation indicating how and from what other
formulas that formula was derived, if it’s not an Assumption. Viewed vertically
rather than horizontally, a deduction consists of four columns. The middle two
columns are a numbered sequence of formulas; the third column is a sequence of
formulas and the second column is an enumeration of those formulas. To the left of
1
A linear formulation of natural deduction was given by E.J. Lemmon in Beginning Logic,
Nelson, London, 1965; the system here is, in essentials, as in that book.
LECTURE 0 8
each numbered formula are the numbers of the Assumptions, if any, on which that
formula depends. The entry in the right-hand column gives the basis on which the
formula in the third column at that line is introduced into the deduction, i.e. either
as an Assumption or by one of the Rules of Inference. An Assumption depends on
itself, so in an application of the rule of Assumption there is one number in the first
column which is the same number as in the second column. If the formula in the
third column of a given line is introduced by a Rule of Inference, the entry in the
fourth column for that line says what Rule of Inference has been used and numbers
of the formulas to which that Rule of Inference has been applied. The Assumptions
on which those formulas depend, as given by the numbers in the first column at
the lines for those formulas, are gathered together as the Assumptions on which
the formula that results from application of that Rule of Inference depends. Four
of the Rules of Inference, ⊃-Introduction, ∼-Introduction, ∨-Elimination, and ∃-
Elimination, discharge an Assumption, so it is possible to arrive at a formula which
depends on no assumptions. Such formulas are, by the Soundness of the Rules of
Inference, logically valid.
∧-Introduction
assumptions numbering formulas justifications
A (a) F [whatever]
B (b) G [whatever]
A ∪ B (c) (F ∧ G) (a) (b) ∧-Introduction
∧-Elimination
1
assumptions numbering formulas justifications
A (a) (F ∧ G) [whatever]
A (b) F (a) ∧-Elimination
1
∧-Elimination
2
assumptions numbering formulas justifications
A (a) (F ∧ G) [whatever]
A (b) G (a) ∧-Elimination
1
∨-Introduction
1
LECTURE 0 9
assumptions numbering formulas justifications
A (a) F [whatever]
A (b) (F ∨ G) (a) ∨-Introduction
1
∨-Introduction
2
assumptions numbering formulas justifications
A (a) G [whatever]
A (b) (F ∨ G) (a) ∨-Introduction
1
∨-Elimination
assumptions numbering formulas justifications
A (a) (F ∨ G) [whatever]
¦b¦ (b) F Assumption
B ∪ ¦b¦ (c) H [whatever]
¦d¦ (d) G Assumption
C ∪ ¦d¦ (e) H [whatever]
A ∪ B ∪ C (f) H (a)(c)(e) ∨-Elimination
⊃-Introduction
assumptions numbering formulas justifications
¦a¦ (a) F Assumption
A ∪ ¦a¦ (b) G [whatever]
A (c) (F ⊃ G) (a)(b) ⊃-Introduction
⊃-Elimination
assumptions numbering formulas justifications
A (a) (F ⊃ G) [whatever]
B (b) F [whatever]
A ∪ B (c) G (a)(b) ⊃-Elimination
The Introduction and Elimination rules for negation in this system of rules are
not as natural as those for the other logical connectives. A more natural formulation
of logic in natural deduction is to take negation to be defined in terms of a primitive
false sentence, sometimes symbolized as ⊥, by the equivalence ∼ ψ ≡ ψ ⊃⊥. How-
ever, in terms of formalizing informal maths talk it is more natural to take negation
LECTURE 0 10
as a primitive symbol, which we do here. It is subject to the following Introduction
and Elimination rules.
∼-Introduction
assumptions numbering formulas justifications
¦a ¦ (a) F Assumption
A ∪ ¦a¦ (b) (G∧ ∼ G) [whatever]
A (c) ∼ F (a)(b) ∼-Introduction
∼-Elimination
assumptions numbering formulas justifications
A (a) ∼∼ F [whatever]
A (b) F (a) ∼-Elimination
∀-Introduction
assumptions numbering formulas justifications
A (a) F(v
i
) [whatever]
A (b) ∀v
i
F(v
i
) (a) ∀-Introduction, if v
i
does not occur free
in any formula enumerated by A
∀-Elimination
assumptions numbering formulas justifications
A (a) ∀v
i
F(v
i
) [whatever]
A (b) F(t) (a) ∀-Elimination, for t any term free for v
i
in F(v
i
)
∃-Introduction
assumptions numbering formulas justifications
A (a) F(t) [whatever] for t any term free for v
i
in F(v
i
)
A (b) ∃v
i
F(v
i
) (a) ∃-Introduction
∃-Elimination
LECTURE 0 11
assumptions numbering formulas justifications
A (a) ∃v
i
F(v
i
) [whatever]
¦b¦ (b) F(v
i
) Assumption
B ∪ ¦b¦ (c) G [whatever]
A ∪ B (d) G (a)(b)(c) ∃-Elimination, if v
i
does not occur free
in any formula enumerated by B
General restriction on terms t in ∀-elimination and ∃-introduction: The term t
must not contain any free variable which in ψ(x) is quantified by a quantifier whose
scope in ψ(x) includes an occurrence of the variable x, i.e. the substitution of t
in ψ(x) must not result in ψ(t) having a different quantifier structure from that of
ψ(x).
=-Introduction
assumptions numbering formulas justifications
(a) t = t =-Introduction, for t any term
=-Elimination
assumptions numbering formulas justifications
A (a) t
1
= t
2
[whatever] for t
1
, t
2
any terms
B (b) F(t
1
) [whatever]
A ∪ B (c) F(t
2
) =-Elimination
Definition 1 For F a formula and Γ a set of formulas, we say that F is (logically)
derivable from Γ, notated Γ ⊢
LDN
F, or just Γ ⊢ F, if there is deduction from the
Rule of Assumption and the 14 Rules of Inferences of LDN in which the last line
has the numbered formula F and the assumptions on which F in that line depends
on are exactly the formulas of Γ.

LDN
(F∨ ∼ F)
(1) (1) ∼ (F∨ ∼ F) Assumption
(2) (2) F Assumption
(2) (3) (F∨ ∼ F) (2) ∨-Introduction
(1)(2) (4) ((F∨ ∼ F)∧ ∼ (F∨ ∼ F)) (1)(3) ∧-Introduction
(1) (5) ∼ F (2)(4) ∼-Introduction
LECTURE 0 12
(1) (6) (F∨ ∼ F) (5) ∨-Introduction
(1) (7) ((F∨ ∼ F)∧ ∼ (F∨ ∼ F)) (1)(6) ∧-Introduction
(8) ∼∼ (F∨ ∼ F) (1)(7) ∼-Introduction
(9) (F∨ ∼ F) (8) ∼-Elimination
A modification to make LND more readily usable: Any truth functional
tautology can be written down as a line of a derivation, resting on no Assumption,
with Tautology as the Justification, e.g.
(1) (F∨ ∼ F) Tautology
0.4 Prenex normal form
Lemma 1 (change of quantified variable) For any formula in a first-order lan-
gauge, any variable of quantification can be changed to any one of infinitely many
other variables while preserving the free variables of the formula, if there are any, in
a way that results in a formula that is logically equivalent to the original formula.
Remark. There are restrictions on logically equivalent changes of bound vari-
able, e.g. ∃v
1
∀v
2
v
1
= v
2
is not logically equivalent to ∃v
2
∀v
2
v
2
= v
2
(the latter is
logically valid while the former is not).
Lemma 2 (prenex equivalences) The following formulas are logically valid.
(ia) (∼ ∀v
i
F(v
i
) ≡ ∃v
i
∼ F(v
i
))
(ib) (∼ ∃v
i
F(v
i
) ≡ ∀v
i
∼ F(v
i
))
In the following the variable v
i
does not occur free in the formula G.
(iia) ((∀v
i
F(v
i
) ∨ G) ≡ ∀v
i
(F(v
i
) ∨ G))
(iia

) ((G∨ ∀v
i
F(v
i
)) ≡ ∀v
i
(G∨ F(v
i
)))
(iib) ((∃v
i
F(v
i
) ∨ G) ≡ ∃v
i
(F(v
i
) ∨ G))
(iib

) ((G∨ ∃v
i
F(v
i
)) ≡ ∃v
i
(G∨ F(v
i
)))
(iiia) ((∀v
i
F(v
i
) ∧ G) ≡ ∀v
i
(F(v
i
) ∧ G))
(iiia

) ((G∧ ∀v
i
F(v
i
)) ≡ ∀v
i
(G∧ F(v
i
)))
(iiib) ((∃v
i
F(v
i
) ∧ G) ≡ ∃v
i
(F(v
i
) ∧ G))
(iiib

) ((G∧ ∃v
i
F(v
i
)) ≡ ∃v
i
(G∧ F(v
i
)))
(iva) ((∀v
i
F(v
i
) ⊃ G) ≡ ∃v
i
(F(v
i
) ⊃ G))
(iva

) ((G ⊃ ∀v
i
F(v
i
)) ≡ ∀v
i
(G ⊃ F(v
i
)))
(ivb) ((∃v
i
F(v
i
) ⊃ G) ≡ ∀v
i
(F(v
i
) ⊃ G))
(ivb

) ((G ⊃ ∃v
i
F(v
i
)) ≡ ∃v
i
(G ⊃ F(v
i
)))
LECTURE 0 13
Proof.
Theorem 3 (Prenex Normal Form) Each formula F in a first-order language
with ∀ and ∃ is logically equivalent to a formula F

whose quantifiers all occur as a
string at the beginning of the formula, and such that F

contains exactly the same
free variables as F does.
Proof The proof is by induction on the number of quantifiers in F.
Base case: F is atomic. Then F has no quantifiers so, vacuously, all its quanti-
fiers, namely none, occur in a string at the beginning of the formula.
Induction steps: (i) F is of the form ∼ G. By induction hypothesis G has a
prenex normal form G

. If G

has no quantifiers, we are done, as in the base case. If
G

has quantifiers, they occur in a string at the beginning of the formula. Then by
logical equivalences (ia) and (ib) from Lemma 2, the negation sign can be pushed
past all the quantifiers, changing each to the other quantifier in the process.
(ii) F is of the form (G ⊃ H). By Induction Hypothesis G and H have logically
equivalent prenex normal forms G

and H

. Then F is logically equivalent to (G


H

). By Lemma 1 and (iva), (ivb), and (iva

), (iv

) of Lemma 2, the quantifiers of
G

and H

can be pulled out into prenx normal form, those from G

changing to
the other quantifier.
(iii) F is of the form (G ∧ H), (iv) F is of the form (G ∨ H). These two cases
are as (ii) except simpler, with no changes of quantifiers as the quantifiers are put
into prenex form. (v) F is of the form ∀v
i
G. Then by induction hypothesis G has
prenex normal form G

which has the same free variables as G has, particular v
i
(in the case when the quantification ∀v
i
of F is not vacuous. Then F is logically
equivalent to ∀v
i
G

, which is in prenex normal form. (vi) F is of the form ∃v
i
G.
Same argument as for (v).
Note that the Prenex Normal Form Theorem holds only on the assumption that
all domains of interpretation are non-empty. Otherwise we have, for example, that
for v
1
not free in G, and G true in the empty domain (e.g. G = ∀v
1
v
1
= v
1
, or
equally G = ∀v
1
∼ v
1
= v
1
),
(∀v
1
F(v
1
) ⊃ G) is true in the empty domain, since Gis true, but ∃v
1
(F(v
1
) ⊃ G))
is false, since every existentially quantified statement is false in the empty domain.
Note that prenex normal forms are not in general unique, e.g.
(∀v
1
F(v
1
) ⊃ ∀v
2
G(v
2
) has as prenex normal form both ∃v
1
∀v
2
(F(v
1
) ⊃ G(v
2
)) and
∀v
2
∃v
1
(F(v
1
) ⊃ G(v
2
)).
LECTURE 0 14
0.4.1 Model theory and proof theory
0.5 Completeness of a system of natural deduc-
tion with respect to first-order logical conse-
quence
0.5.1 Lindenbaum’s Lemma
0.5.2 The Completeness Theorem
The completeness theorem for a given formal system of first-order logic means that
a model theoretic argument for a logical consequence (as above) establishes the
existence of a formal derivation of that logical consequence in the system whose
completeness has been proved without actually finding a derivation.
While we talk about the completeness theorem for first order logic, there are ac-
tually many completeness theorems, one for each different complete formal system
for first-order logical consequence. On the other hand there is an intrinsic com-
pleteness theorem, namely completeness of a system which consists of exactly those
axioms and rules of inference needed for the proof of the completeness theorem.
Comparison between the completeness theorem for first-order logical consequence
and the incompleteness theorem for truth in arithmetic. Both these results are
due to G¨odel, in 1930 and 1931. (The first was his Ph.D. thesis, the second his
Habilitation thesis. They were written under the nominal supervision of the very
notable mathematician Hans Hahn, but G¨odel was essentially working on his own.)
0.5.3 The Compactness Theorem for first-order logic
0.5.4 The existence of non-standard models of the truths of
arithmetic
0.5.5 Completeness of other systems of derivation with re-
spect to first-order logical consequence
Definition 2 Let S
1
and S
2
be formal systems such that the language L
1
of S
1
is a
sub-language of the language L
2
of S
2
. We say that S
1
is a subsystem of S
2
if for
every formula φ in L
1
, if S
1
⊢ φ, then S
2
⊢ φ
Lemma 4 If S
1
is a complete system for first-order logical consequence and S
1
is a
subsystem of S
2
, then S
2
is complete for first-order logical consequence.
Lecture 1
Introduction: a weak form of
G¨ odel’s First Incompleteness
Theorem; the symbols and
expressions of a language for
arithmetic L
E
; G¨ odel numbering
of the expressions of L
E
(Tuesday, 12 October 2010)
1.1 Introduction: a weak form of G¨odel’s First
Incompleteness Theorem
The context of G¨odel’s discovery of the phenomenon of formal incompleteness is
David Hilbert’s programme for giving mathematics a secure foundation by estab-
lishing the consistency of systems formalizing it. In 1918 he declared that
we must make the concept of specifically mathematical proof itself into
an object of investigation (Hilbert [2]).
Hilbert formulated the distinction between finitary and infintary mathematics.
The paradigm of finitary mathematics is arithmetical calculation. Finitary mathe-
matics is mathematical bedrock, corresponding to observation statements in science.
A calculation such as 2
7
= 128 is finitary, but the claim that exponentiation to the
15
LECTURE 1 16
power 2 always yields a value, i.e. ∀x∃y(2
x
= y) is infinitary, and more generally,
quantification over the infinite domain of natural numbers is infinitary. However,
quantification over a bounded, i.e. initial segment of the natural numbers, which is
finite, belongs to finitary mathematics.
Hilbert’s deep insight was to recognize that the formal manipulation of all sym-
bols, not just the symbols for numbers, i.e. numerals and terms built up from
numerals and symbols for arithmetical operations, belongs to finitary mathematics.
In particular,
a formalized proof, like a numeral, is a concrete and surveyable object.
([3], p. 383 and also in [4], p. 471.)
Hilbert recognized two sorts of finitary statements, general and particular (though
he did not introduce terminology for this distinction). Particular finitary statements
are decided by computations, e.g. 7 5 = 35, and 2
10
= 1024 and truth functional
combinations of them (the truth values of such combinations being computable from
the truth values of the component statements). General finitary statements contain
free variables, and can be thought of as a template for particular finitary state-
ments that result by substitution of numerals for the free variables, for example
x +y = y +x, and (n > 2 ⊃ x
n
+y
n
= z
n
). On the other hand, ∀x∀y x +y = y +x
and ∀n∀x∀y∀z(n > 2 ⊃ x
n
+y
n
= z
n
) are infinitary.
For F(v
1
) a general finitary statement with free variable v
1
, bounded quantifica-
tion on the variable v
1
, which is finitary, is expressible using (apparently) unbounded
quantification by, in the case of universal quantification, ∀v
1
(v
1
≤ t ⊃ F(v
1
)), for
t a term in the language of arithmetic, which we abbreviate as (∀v
1
≤ t)F(v
1
),
and in the case of existential quantification, ∃v
1
(v
1
≤ t ∧ F(v
1
)), which we abbre-
viate (∃v
1
≤ v
2
)F(v
1
). For t a numerical term (a numeral or a composition of
arithmetical functions applied to numerals), (∃v
1
≤ t)F(v
1
) and (∀v
1
≤ t)F(v
1
) are
particular finitary statements if v
1
is the only free variable in F(v
1
). If t is a free
variable or a composition of arithmetical functions applied to one or more variables,
(∃v
1
≤ t)F(v
1
) and (∀v
1
≤ t)F(v
1
) are general finitary statements.
Hilbert noted that general finitary statements are not closed under negation, i.e.
the negation of a general finitary statement cannot be expressed as a general finitary
statement. For example, Fermat’s Last Theorem is expressible as a general finitary
statement, (n > 2 ⊃ x
n
+y
n
= z
n
), but to say (falsely) that Fermat’s Last Theorem
is false requires existential quantification, ∃n∃x∃y∃z(n > 2 ⊃ x
n
+ y
n
= z
n
). On
the other hand, the statement that a specific quadruple of numbers a, b, c, d is a
counterexample, i.e. (a > 2 ∧ a
n
+b
n
= c
n
), is a particular finitary statement.
Statements about particular formal proofs are, as Hilbert recognized, finitary
statements, e.g. that a particular formal proof is or is not a proof of a particular
statement.
LECTURE 1 17
Hilbert missed something about his insight which G¨odel realized, namely that
formal proofs can be literally identified with natural numbers, i.e. they could be
taken to be numerical expressions, rather than merely like them. As G¨odel put this
point in (1931),
Of course, for metamathematical considerations it does not matter what
objects are chosen as primitive signs, and we shall assign natural numbers
to this use, that is, we map the primitive signs one-to-one onto some
natural numbers.
Numbers assigned to formulas of a formal language in this way are called G¨odel
numbers.
Definition 3 We denote the G¨ odel number of a formula F by F.
To carry out the arithmetization of syntax, the system must be able to ”talk”
about numbers, i.e. there must be for each natural number a formal numeral in the
language of the system that denotes that numbers.
Definition 4 For formal languages that have a numeral for each natural number,
we denote by n the numeral for the natural number n.
Definition 5 Whenever in these notes I speak of a sentence in the specified language
of arithmetic as being true, I mean true in the usual (intended) structure consisting
of the domain of natural numbers with the usual arithmetical functions and relations
on the natural numbers (also known as the standard model).
What G¨odel showed is that the property of being (the G¨odel number of) a
provable formula is expressible within any system which can express basic arithmetic,
i.e. there is a formula Pr(v
1
) with one free variable, in the language of a formal
system for arithmetic, S, such that for every formula X in the language of S, S ⊢ X
if and only if Pr(X) is true. We shall establish the existence of such a formula for
a particular formal system of arithmetic in Lecture 4. G¨odel also showed that for any
formula with one free variable (in particular a formula that expresses the property
of being the G¨odel number of an unprovable formula) there is a diagonal sentence,
i.e for formula F(v
1
) there is a sentence D such that the equivalence (D ≡ F(D))
is true. We shall establish this result in Lecture 3.
From these results and on the assumption that everything provable in a given
system S is true (about the natural numbers) (a strong assumption, much stronger
than is needed to establish incompleteness, but it is illuminating to consider this
simple case), it is easy to see that for G such that (G ≡∼ Pr(G)), S G, and also
thereby that G is true, which implies, from the assumption that everything provable
in a system S is true, that S ∼ G.
LECTURE 1 18
Theorem 5 (weak form of G¨odel’s first incompleteness theorem) Let S be
a theory such that for each natural number n there is a numeral n in the language
of S, and assume that
(i) there is a sentence G such that the sentence (G ≡∼ Pr(G)) is true;
(ii) there is an assignment of numbers to the formulas of the language of S and
a formula Pr(v
1
) in the language of S such that for each formula X, S ⊢ X if
and only if the sentence Pr(X) is true, so in particular S ⊢ G if and only if the
sentence Pr(G) is true.
(iii) every theorem of S is true.
Then S G, G is true, and S ∼ G.
Proof. (1) Suppose that S ⊢ G. (2) Then by (ii), Pr(G) is true. (3) From
(2) and (i), G is false. (4) From (3) and (iii), S G. (5) Since (4) contradicts (1),
we have by reductio ad absurdum that S G (from assumptions (i), (ii), and (iii)).
(6) From (5) and (ii), Pr(G) is false. (7) From (6) and (i), G is true.
(8) From (7) and (iii), S ∼ G.
Remarks about this result:
(1) This is a weak version of the first G¨odel incompleteness theorem since, while
assumptions (i) and (ii) are provable for weak systems of arithmetic, assumption (iii),
soundness of the system with respect to truth in arithmetic, is a strong (highly non-
finitistic) assumption, and unprovability of the G¨odel sentence holds from the much
weaker assumption that S is consistent. G¨odel sketches the proof of this weak form
of the First Incompleteness Theorem in section 1 of his 1931 paper, and notes that
“The purpose of carrying out the above proof with full precision in what follows is,
among other things, to replace the second of the assumptions just mentioned [every
provable formula is true in the interpretation considered] by a purely formal and
much weaker one.” (van Heijenoort Sourcebook p. 599).
(2) In his introductory section G¨odel notes that this argument is “closely re-
lated” to the argument for the “Liar” paradox. The argument does not lead to a
contradiction since it starts from the assumption that G is provable, and so by RAA
establishes that G is not provable. Use of the Liar paradox also shows, as we shall
see, that unlike provability in a formal system, truth in a language of arithmetic
cannot be expressed in the language.
(3) We shall establish how much basic arithmetic is required for arithmetization
of syntax, i.e. to prove assumptions (i) and (ii), to be made precise by the notion
of Σ
0
-arithmetic, essentially just computations with addition and multiplication. In
particular we don’t need exponentiation. This shows that arithmetized syntax is a
proper sub-part of what Hilbert meant by finitist mathematics. Hilbert never gave
a precise characterization of finitist mathematics, but it is clear that it includes all
primitive recursive functions, so plus and times, but then also exponentiation and
LECTURE 1 19
so on. On the other had, both addition and multiplication are needed for incom-
pleteness: there is a complete theory of zero, successor, and addition (Presburger
Arithmetic).
(4) The independence of the G¨odel sentence from formal arithmetic was unprece-
dented. In the previous hundred years the independence of Euclid’s fifth postulate
from the other postulates of geometry had been established. G¨odel’s result differs
from this earlier one in two crucial ways. One was the technique used. The result
concerning the fifth postulate was established by the construction of models. The
G¨odel result is purely syntactic (exploiting Hilbert’s insight). The other difference
is even more fundamental. The fifth postulate is neither true nor false, per se. It
is true in Euclidean geometry and false in non-Euclidean geometries. The G¨odel
sentence is demonstrably true, though not demonstrable in the system for which it
is constructed.
Proposition 6 For any system S and sentence X, a proof that S X from the
assumption of consistency is best possible.
Proof. If S is inconsistent, it proves everything, so in particular S ⊢ X.
Remark. That the G¨odel sentence G for a system S is not refutable, i.e. ∼ G
is not provable, requires a stronger condition on S than consistency but a condition
much weaker than the soundness of S is sufficient.
1.2 The symbols and expressions of a language
for arithmetic L
E
A formal language is generated by combining symbols from a specified finite alpha-
bet. For our formal language for arithmetic, as in [7], this consists of the following
13 symbols:
0

( ) f ′ v ∼ ⊃ ∀ = ≤ ♯
These formal symbols will be used with the following intended meanings:
The symbol 0 denotes the natural number zero
1
.
The symbol

denotes the successor function
The symbols ( and ) are left and right brackets
The symbols f and v are for functions and variables, to which numerical sub-
scripts in tally notation, i.e. iterations of the subscript ′ are attached. The strings of
1
Note that in this sentence I am being casual about the distinction between use and mention.
That distinction is easily but cumbersomely dealt with by using quotation marks, in which case
this given sentence would read: “The symbol ’0’ denotes the natural number zero”, which is fine,
though fussy, but the next sentence becomes just about unreadable if use vs mention is spelled out
in this way.
LECTURE 1 20
symbols f

, f
′′
, f
′′′
will denote the functions addition, multiplication, and exponenti-
ation, respectively, which we will write informally as +, , and exp or x
y
in the usual
notation. There are an infinity of variables v

, v
′′
, v
′′′
, . . ., which we will usually write
as v
1
, v
2
, v
3
, . . .. If we want to signify a variable without specifying which variable it
is, we will write v
i
, v
j
etc or use informal variable letters x, y, z, u, v, w.
The symbol for the propositional connectives negation and implication are ∼
and ⊃. The symbol for the universal quantifier is ∀.
The symbols = and ≤ are for the two-place relations of equality and less than
or equals.
The symbol ♯ will be used to mark breaks between strings of symbols that are
terms and formulas of the language (to be defined in the next lecture) in sequences
of terms and formulas.
An expression in the language is (almost) any finite string of these symbols.
The set of expressions for the language L
E
is specified by the following recursive
definition.
Definition 6 (expressions) basis: Each one of the symbols 0

( ) f ′ v ∼ ⊃
∀ = ≤ ♯ is by itself an expression.
recursion: If E
i
and E
j
are expressions, and E
i
=

, then the result of writing E
i
directly followed by E
j
, which we call the concatenation of E
i
and E
j
and symbolize
as E
i
E
j
, is an expression.
Remark: It’s for a technical reason (to do with our choice of G¨odel numbering)
that we excluded from the class of expressions as here specified strings of more than
one symbol that begin with the symbol

.
1.3 G¨odel numbering of the expressions of L
E
.
We assign G¨odel numbers to the expressions of L
E
. This can be done in infinitely
many ways. The way we shall do it, following Smullyan following Quine, makes the
link between G¨odel numbering of expressions as strings of formal symbols particu-
larly transparent. G¨odel’s original method involved coding by exponents of prime
factors. On our method each number is the G¨odel number of an expression, while
on G¨odel’s method not every number is a G¨odel number. Having every number be
a G¨odel number makes the formulation of some results a little simpler but is not
essential.
Definition 7 (notation for expressions and G¨odel numbers) E
n
=
df
the ex-
pression with G¨ odel number n.
E =
df
the G¨ odel number of expression E.
LECTURE 1 21
Corollary 7 (of Definition 7) E
n
= n.
We are used to the idea that numbers are denoted by numerals and that numerals
are not the same thing as numbers. The Roman numerals for the first five non-zero
natural numbers are I, II, III, IV, V, while the Arabic numerals are 1, 2, 3, 4 , 5. The
crucial property of the Arabic numerals is that they are constructed on a place-value
system with a base of 10. That the system of numerals in common use is base 10 is
presumably down to the contingent fact (it could have been otherwise) that human
beings have 10 fingers. Any other number greater or equal to 2 gives a perfectly
good numeral system with that base. Base 2 is used in machine code for computers,
with 0 and 1 represented by current off and current on. The number we write as 15
in base 10 we write as 1111 in base 2 and as 13 in base 12. In the formal language for
arithmetic we shall be using the numerals for numbers use a tally notation, rather
than place values. The formal numeral for the number n is the expression 0
n

′ . . . ′
,
i.e. concatenation of the symbol 0 with n-many concatenations of the symbol

.
The following function plays a key role in our chosen system of G¨odel numbering.
Definition 8 (concatenation of base b numerals) For natural numbers m and
n, we denote by m∗
b
n the number designated by the base b numeral that results from
concatenating the base b numeral for m with the base b numeral for n.
Note that ∗
b
is a function mapping pairs of natural numbers to natural numbers,
and that natural numbers are not intrinsically in base b or any other base notation.
The role of b in this function is to specify the method of calculating this function.
Examples: For m = 673, n = 32 (written in base 10), m ∗
10
n = 67332 and
n ∗
10
m = 32673. For m = 59, n = 0, m∗
10
n = 590 and n ∗
10
m = 059 = 59.
Remark: As illustrated by these examples, ∗
b
is not commutative. It is also not
associative, e.g. (17∗
b
0)∗
b
59 = 17059 = 1759 = 17∗
b
(0∗
b
59). Non-associativity only
arises when the middle value is 0, but since we will include 0 as a G¨odel number we
cannot suppress parentheses in multiple computations with ∗
b
except by adopting a
convention for reinstating them; we adopt the common convention of association to
the left, i.e. x ∗
b
y ∗
b
z = (x ∗
b
y) ∗
b
z.
We assign G¨odel numbers to expressions by first stipulating the G¨odel numbers
of the symbols. We assign to these thirteen symbols the numbers denoted by the
thirteen digits of base 13 notation, where the digits for 10, 11, and 12 (as we write
them in base 10) are taken to be η, ǫ, and δ, respectively.
Definition 9 (assignment of G¨odel numbers to expressions) By recursion over
the recursive definition of expressions.
Base case: The assignment of numbers to symbols is specified by
LECTURE 1 22
0

( ) f ′ v ∼ ⊃ ∀ = ≤ ♯
1 0 2 3 4 5 6 7 8 9 η ǫ δ
Recursion: For expressions X and Y , XY = X ∗
13
Y
By these stipulations,

= 0,

. . .

= 0, and

∀ = 09 = 9 = ∀. It is in
order for each expression to have a unique G¨odel number that we stipulated above
that the class of expressions does not contain strings of more than one symbol that
begin with the prime symbol

.
There is a technical advantage in taking the base b to be a prime number but
it is by no means essential. We can use base 10 and the operation ∗
10
even with
thirteen symbols by, for example, assigning the thirteen symbols respectively the
following numbers (written in base 10):
0

( ) f ′ v ∼ ⊃ ∀ = ≤ ♯
1 0 2 3 4 5 6 7 89 899 8999 89999 899999
Of course on this assignment not every number is a G¨odel number. But we can
effectively tell the ones that are, i.e. we know that if an 8 or a 9 occurs its base 10
notation it must occur within a string of the form 89, 899, 8999, 89999, 899999, and
we know which symbol is coded by counting the number of 9s in that string.
Lecture 2
Terms and formulas of the
language L
E
; expressibility of
diagonal substitution in the
language L
E
(Wednesday, 13 October 2010)
2.1 Terms and formulas of the language L
E
2.1.1 Terms
Terms in the formal language L
E
are expressions that, on the intended interpretation
of L
E
as a language for arithmetic, denote a number if the term does not contain a
free variable, and if it contains one or more free variables, then the term that results
from substituting a numerals for each variable denotes a number.
Definition 10 (Variables) v

is a variable, and if the expression E is a variable
then the expression E

, i.e. the concatenation of E and the subscript symbol ‘

’, is
a variable.
Remark: Formal variables are expressions of the form v

, v
′′
, v
′′′
, . . .. We will
abbreviate the string of symbols consisting of the formal variable symbol v followed
by n subscripts as v
n
.
Further remark: Smullyan’s formal variables are of the form (v′ ), (v′′ ), (v′′′ ), . . .,
i.e. formal variables as we have defined them enclosed in brackets (p. 15). Unique
readability for terms within expressions does not require enclosing variables within
23
LECTURE 2 24
brackets in this way. The motivation for this unnecessary use of brackets might
be an artefact of what’s simple to write in LaTeX. Evidently Smullyan produces
e.g. the substring v′′ of his variable (v′′ ) by the LaTeX code v_{’’}. If this was
v
2
, then the seemingly natural way to write LaTeX code for the concatenation of
v
2
with

would be $v_{’’}^{\prime}$, but this produces v

′′ , in which the

sym-
bol occurs interposed over the string v′′ rather than concatenated at the end of
that string. On the other hand, $(v_{’’})^{\prime}$ produces (v′′ )

, in which

is correctly concatenated at the end. Nonetheless, more complicated LaTeX code
(also with compounded subscript command to lower the apostrophe further so it
looks more convincing as a subscript) produces the required concatenation with vari-
ables not enclosed in brackets, namely $v_{_{’’}}$\hspace{-.12ex}$^{\prime}$,
which compiles to the required concatenation of symbols v
′′

.
Definition 11 (Numerals) The symbol 0 is a numeral (which denotes the number
zero). If the expression E is a numeral then the expression E

is a numeral (which
denotes the successor of the number denoted by E, so the expressions 0, 0

, 0
′′
, 0
′′′
, . . .
are formal names of the natural numbers 0, 1, 2, 3, . . ., respectively).
Notation: For natural number n we write n for the numeral that denotes the
number n, e.g. 7 = 0
′′′′′′′
.
Corollary 8 (of Definition 11 and Notation) For any natural number n, n + 1
is n

, i.e. the numeral for the number n + 1 is the concatenation of the numeral for
the number n and the symbol

.
Definition 12 (Terms) Among expressions of L
E
, the class of terms is specified
by the following recursive definition:
Base clause: Each variable and each numeral is a term.
Induction clauses: If t is a term, then t

is a term. If t
1
and t
2
are terms,
then (t
1
f

t
2
), (t
1
f
′′
t
2
), and (t
1
f
′′′
t
2
) are terms. [Recall that the notations in L
E
for
addition, multiplication, and exponentiation are, respectively, f

, f
′′
, and f
′′′
.]
Definition 13 (expression E
1
occurs in expression E
2
) (i) Every expression oc-
curs in itself.
(ii) Expression E
1
occurs in expression E
2
if there exist expressions E
3
, E
4
such
that E
2
= E
3
E
1
E
4
or E
2
= E
3
E
1
or E
2
= E
1
E
4
Definition 14 (constant terms) A term in which no variable occurs is called a
constant term, or a closed term.
LECTURE 2 25
2.1.2 Formulas
Definition 15 (Atomic formulas) An atomic formula is any expression of the
form t
1
= t
2
or of the form t
1
≤ t
2
, where t
1
and t
2
are terms.
Definition 16 (Formulas) The class of formulas is specified by the following re-
cursive definition:
Base clause: Every atomic formula is a formula.
Induction clauses: If F and G are formulas, then ∼ F and (F ⊃ G) are formulas,
and for every variable v
i
, the expression ∀v
i
F is a formula. [Note that the formula
(F ⊃ G) is enclosed in brackets, but that the other two formation rules do not
introduce new brackets.]
We will use logical equivalences between conjunction, disjunction, and existential
quantification and expressions in terms of ∼, ⊃, and ∀ as abbreviations, i.e.
Definition 17 For formulas A and B,
(A ∧ B) =
df
∼ (A ⊃∼ B);
(A ∨ B) =
df
(∼ A ⊃ B);
∃v
i
A =
df
∼ ∀v
i
∼ A.
2.1.3 Free and bound variables; open formulas and sen-
tences
Definition 18 (a variable occurs free in a formula) (i) If F is an atomic for-
mula, all occurrences of variables in F are free.
(ii) For F and G formulas, if variable v
i
occurs free in F, then v
i
occurs free in
∼ F and in (F ⊃ G) and (G ⊃ F).
(iii) For any formula F, if v
i
occurs free in F, v
i
is free in ∀v
j
F iff j = i.
Definition 19 (bound variables) A variable occurs bound in a formula F iff it
occurs in F and does not occur free in F.
Definition 20 (open formulas) A formula with one or more free variables is an
open formula
Notation: We write F(v
i
) to signify a formula in which the variable v
i
occurs
free. Other unspecified variables may occur free as well unless we stipulate that
v
i
is the only variable free in F(v
i
). In the latter case we may say that F(v
i
) is a
one-place formula. Similarly we write F(v
i
1
, . . . v
i
k
) for a formula in which variables
v
i
1
, . . . v
i
k
occur free, possibly with other free variables unless we stipulate that these
are the only ones.
LECTURE 2 26
[Note that this convention on possible occurrence of free variables other than
those explicitly shown is different from Smullyan’s: “We write F(v
i
1
, . . . v
i
k
) for any
formula in which v
i
1
, . . . v
i
k
are the only free variables.” (p. 16). Since there are
situations, e.g. in stating the Induction axioms, in which we need to allow for the
possibility of other free variables than those explicitly shown, Smullyan’s convention
has in those situations to be violated, e.g. “F(v
1
) is to be any formula at all (it
may contain free variables other than v
1
)” (Smullyan, p. 29). It seems to me more
coherent for the convention to allow other variables, which also then allows us under
the same convention to stipulate in a given situation that there are no other free
variables.]
Definition 21 (closed formulas a.k.a. sentences) A formula with no free vari-
ables is a closed formula, also called a sentence.
Notation: Substitution of numerals for free variables in formulas: We write
F(n) to signify the result of substituting the numeral n for every free occurrence
of v
i
in the open formula F(v
i
). If v
i
is the only variable free in F(v
i
), then F(n)
is a sentence. For numbers n
1
, . . . , n
k
, F(n
1
, . . . , n
k
) signifies the result of substi-
tuting the numerals n
1
, . . . , n
k
for all free occurrences in F(v
i
1
, . . . v
i
k
) of v
i
1
, . . . v
i
k
,
respectively. If v
i
1
, . . . v
i
k
are the only variables that occur free in F(v
i
1
, . . . v
i
k
), then
F(n
1
, . . . , n
k
) is a sentence.
Definition 22 (regular open formulas) An open formula is said to be regular if
its k-many free variables are the first k variables.
For F(v
1
, . . . , v
k
) a regular open formula, the expression F(n
1
, . . . , n
k
) is unam-
biguous; we don’t need to stipulate which number is substituted for which variable.
Note that for F(v
1
) a formula with one free variable, F(n) is a sentence while
F(n) is not. The latter can be construed as shorthand for the statement that the
open formula F(v
1
) is satisfied by the number n, and that statement will be true if
and only the sentence F(n) is true.
2.2 Designation by terms in L
E
, truth of sentences
of L
E
, and expressibility of sets and relations
of natural numbers by formulas of L
E
2.2.1 Designation
On the intended interpretation of the formal language L
E
each constant term des-
ignates a particular natural number, by the following recursive specification:
LECTURE 2 27
(i) the numeral n designates the number n.
(ii) If constant term c designates n, then constant term c

designates the next
natural number after n; if constant terms c
1
and c
2
designate n
1
and n
2
respectively,
then (c
1
f′ c
2
) designates the sum of n
1
and n
2
, (c
1
f′′ c
2
) designates the product of n
1
and n
2
, and (c
1
f′′′ c
2
) designates n
1
raised to the power n
2
.
2.2.2 Truth
Truth for a sentence of L
E
in the structure of the natural numbers (the intended
interpretation) can be defined by recursion over the recursive generation of the
sentence in the usual way. For none of the results in this course do we require a
formal definition of truth, and I will take it as known informally what it means for
a formula in the language of arithmetic to be true in the structure of the natural
numbers.
2.2.3 Expressibility
Definition 23 (expressibility of relations) A formula F(v
1
, . . . , v
n
) in L
E
is said
to express a relation R ⊆ N
n
iff for every n-tuple < k
1
, . . . , k
n
> of natural numbers
the sentence F(k
1
, . . . , k
n
) is true iff < k
1
, . . . , k
n
>∈ R. In such case the relation
R is said to be expressible in L
E
.
Definition 24 (expressibility of functions) A function f(v
1
, . . . , v
n
) : N
n
→ N
is expressible in L
E
iff the relation f(v
1
, . . . , v
n
) = v
n+1
is expressible in L
E
.
Definition 25 (Arithmetical) A relation or a function is Arithmetical if it is
expressible by a formula in L
E
.
Definition 26 (arithmetical) A relation or a function is arithmetical [with a
lower-case “a”] iff it is expressible by a formula in L
E
in which the expression f′′′
(for exponentiation) does not occur.
2.3 Concatenation of numbers in a given base no-
tation is Arithmetical.
Lemma 9 For a fixed number b ≥ 2, the condition that v
1
is a power of b, which
expression we abbreviate as Pow
b
(v
1
), is Arithmetical.
Proof. Pow
b
(v
1
) iff ∃v
2
(v
1
= b
v
2
), or more formally
Pow
b
(v

) iff ∼ ∀v
′′
∼ v

= (bf
′′′
v
′′
).
LECTURE 2 28
Lemma 10 For ℓ
b
(n) the length of the base b notation for n, i.e. the number of
digits in the base b notation of n, the two-place relation b

b
(v
1
)
= v
2
is Arithmetical
Proof. This relation is expressed by the following condition on v
1
and v
2
.
((v
1
= 0∧v
2
= b)∨(v
1
= 0∧Pow
b
(v
2
)∧v
1
< v
2
∧∀v
3
((Pow
b
(v
3
)∧v
1
< v
3
) ⊃ v
2
≤ v
3
))).
This equivalence is seen as follows: If v
1
= 0, ℓ(v
1
) = 1 and b
1
= b. The first
disjunct takes care of this case. If v
1
= 0, then the length of v
1
(in base b notation) is
the least power of b > v
1
, e.g. ℓ
10
(935) = 3, and 10
3
= 1000 is the least power of 10
> 935. [Note that we need to treat these two cases separately, since ℓ(0) = 1, but—
writing µv
3
F(v
3
) for the least v
3
s.t. F(v
3
)—µv
3
(10
v
3
> 0) = 0 since 10
0
= 1 > 0.]
That v
2
is the least power of b greater than v
1
is expressed by the two conditions
that b
v
2
> v
1
, and for any v
3
such that b
v
3
> v
1
, v
2
≤ v
3
.
The above condition is expressible in L
E
by Lemma 9 and the fact that v
1
< v
2
is equivalent to (v
1
≤ v
2
∧ ∼ v
1
= v
2
).
Theorem 11 For any number b ≥ 2, the relation v
1

b
v
2
= v
3
is Arithmetical.
Proof. The relation v
1

b
v
2
= v
3
is expressed by the condition that
v
1
b

b
(v
2
)
+v
2
= v
3
.
For example, 1570 ∗
10
365 = 1570365 = 1570000 + 365 = 1570 10
3
+ 365 =
1570 10

10
(365)
+ 365.
This condition is equivalent to
∃v
4
(b

b
(v
2
)
= v
4
∧ ((v
1
v
4
) +v
2
) = v
3
).
By Lemma 10, this relation is Arithmetical.
2.4 Substitution and diagonal functions, and their
arithmetization
The operation of substituting a numeral for a free variable lies at the heart of
the construction of a ‘self referential’ arithmetical sentence (‘This sentence is not
provable in the given formal system’). To describe the formula F(n) obtained by
substituting the numeral n for the free variable v
i
in the formula F(v
i
) is complicated
(requiring recursion over the logical complexity of formulas). We follow Smullyan
in utilizing a trick due to Tarski by which we construct a formula F[n] which is not
the same formula as F(n) but is logically equivalent to it for which there is a simple
general description that does not depend on the logical complexity of the formula
F(v
i
).
LECTURE 2 29
Definition 27 (quasi-substitution) For F(v
1
) any formula of L
E
with one free
variable and n any numeral, F[n] =
df
∀v
1
(v
1
= n ⊃ F(v
1
)).
Quasi-substitution is logically equivalent to substitution, i.e.
Lemma 12 (∀v
1
(v
1
= n ⊃ F(v
1
)) ≡ F(n)) is logically valid.
Proof. (i) Suppose ∀v
1
(v
1
= n ⊃ F(v
1
)). Then by universal instantiation,
(n = n ⊃ F(n)). Since n = n is logically valid, by modus ponens, F(n).
(ii) Suppose F(n). Then by substitutivity of identity, (v
1
= n ⊃ F(v
1
)). By
universal generalization, ∀v
1
(v
1
= n ⊃ F(v
1
)).
Note We could also have defined F[n] as ∃v
1
(v
1
= n ∧ F(v
1
)) since:
Lemma 13 (∀v
1
(v
1
= n ⊃ F(v
1
)) ≡ ∃v
1
(v
1
= n ∧ F(v
1
))) is logically valid.
Proof. Exercise.
Our first step in the arithmetization of syntax, i.e. showing that syntactic oper-
ations on expressions of L
E
can be reflected into arithmetically definable operations
on their G¨odel numbers, is to show that the function
s(v
1
, v
2
) = ∀v
1
(v
1
= v
2
⊃ E
v
1
)
is Arithmetical.
The value of the function s(v
1
, v
2
) is the G¨odel number of a formula logically
equivalent to the substitution of the numeral of the number v
2
into the expression
E
v
1
when E
v
1
is a formula in which v
1
occurs free. Note that for some values of v
1
,
E
v
1
is a formula in which the variable v
1
occurs free, and for other values it is not.
Indeed for some values of v
1
, the expression whose G¨odel numbers if v
1
, i.e. E
v
1
,
will not be a formula, in which case s(v
1
, v
2
) is the G¨odel number of an expression
which is not a formula—a ‘don’t care’ case. If E
v
1
is a formula in which v
1
does not
occur free then s(v
1
, v
2
) is the G¨odel number of a formula that has nothing to do
with substitution of the numeral for v
2
into it—another ‘don’t care’ case.
Before we can establish that the two-place function (three-place relation) s(v
1
, v
2
) =
v
3
is Arithmetical, we must calculate the G¨odel numbers of the numerals.
Lemma 14 The G¨ odel number of the numeral v
2
is 13
v
2
(in our base-13 assignment
of G¨ odel numbers ).
Proof. The numeral v
2
is the expression 0
v
2

′ . . . ′
. The symbol 0 is assigned G¨odel
number 1, the symbol

is assigned the number 0, so that whole expression is, by
concatenation, assigned the number written in base-13 notation as 1
v
2

0 . . . 0, which
is 13
v
2
.
LECTURE 2 30
Theorem 15 The function s(v
1
, v
2
) = ∀v
1
(v
1
= v
2
⊃ E
v
1
) is Arithmetical.
Proof. Let k = ∀v
1
(v
1
=, a particular number whose base 13 notation—
given our assignment of base 13 digits to the symbols of our language—is 965265η
(or if we use the base 10 G¨odel numbering also given in Lecture 1, whose base
10 notation is 899652658999). Given that v
2
= 13
v
2
, ⊃ = 8, E
v
1
= v
1
,
) = 3, s(v
1
, v
2
) = v
3
iff ∃v
4
(v
4
= 13
v
2
∧ v
3
= k ∗ v
4
∗ 8 ∗ v
1
∗ 3), which by left-
hand association of ∗
13
and repeated use of Theorem 11 is Arithmetical, i.e. there
is a formula S(v
1
, v
2
, v
3
) such that for all numbers m, n, k, S(m, n, k) is true iff
s(m, n) = k. Expressing this argument from Theorem 11 more strictly, the formula
needs to be the following:
∃v
4
∃v
5
∃v
6
∃v
7
(v
4
= 13
v
2
∧ k ∗ v
4
= v
5
∧ v
5
∗ 8 = v
6
∧ v
6
∗ v
1
= v
7
∧ v
7
∗ 3 = v
3
)
Definition 28 (diagonal substitution) The diagonal substitution function d(v
1
)
is s(v
1
, v
1
), i.e. d(v
1
) =
df
∀v
1
(v
1
= v
1
⊃ E
v
1
).
Remark. Since by Definition 27, ∀v
1
(v
1
= x ⊃ E
v
1
) =
df
E
v
1
[v
1
], Definition 28
means that d(v1) = E
v
1
[v
1
].
Corollary 16 (corollary to the proof of Theorem 15) The relation d(v
1
) = v
2
is
Arithmetical.
Proof. Let D(v
1
, v
2
) be the formula that results by substituting v
1
for v
2
and
v
2
for v
3
in S(v
1
, v
2
, v
3
). For all numbers m, n, S(m, m, n) is true iff s(m, m) = n.
By the definition of d(v
1
), s(m, m) = d(m). So D(m, n) is true iff d(m) = n.
Lecture 3
The Diagonal Lemma;
expressibility of properties of
sequence numbers
(Tuesday, 19 October 2010)
3.1 The Diagonal Lemma
Theorem 17 (The Diagonal Lemma) For each one-place formula F(v
1
) in L
E
,
there exists a sentence C in L
E
such that the equivalence (C ≡ F(C)) is a true
sentence in L
E
.
Proof. (1) Let D(v
1
, v
2
) be the formula in L
E
from the proof of Corollary 16
that expresses the relation d(v
1
) = v
2
.
(2) Let k =
df
∀v
2
(D(v
1
, v
2
) ⊃ F(v
2
)).
(3) Let C =
df
∀v
1
(v
1
= k ⊃ ∀v
2
(D(v
1
, v
2
) ⊃ F(v
2
))).
(4) By Lemma 12, (C ≡ ∀v
2
(D(k, v
2
) ⊃ F(v
2
)).
(5) By (1), (∀v
2
(D(k, v
2
) ⊃ F(v
2
)) ≡ ∀v
2
(d(k) = v
2
⊃ F(v
2
))).
(6) By Lemma 12, (∀v
2
(d(k) = v
2
⊃ F((v
2
))) ≡ F(d(k))).
(7) By the chain of equivalences (4), (5), (6), (C ≡ F(d(k))).
(8) By (2) and (3) and Definition 28, C = d(k) (fuller explanation below).
(9) From (7) and (8) by substitutivity of identity, (C ≡ F(C)).
Explanation of step (8): k is the G¨odel number of a formula with one free variable,
C is the formula that results from the diagonal substitution of quasi-substituting
into that formula its own G¨odel number, and the one-place function d(v
1
) generates
from the G¨odel number of a formula with one free variable the G¨odel number of the
formula that results from diagonal substitution into that formula.
31
LECTURE 3 32
The proof of the Diagonal Lemma is by a kind of double substitution into the one-
place formula for which a diagonal sentence is being established, first the substitution
of the Arithmetical expression of the diagonal function, and then the substitution
of the numeral for the G¨odel number of the formula that results from that first
substitution. Both of these substitutions are quasi-substitutions in this construction.
It might make the idea of the proof more perspicuous if we consider how it goes
with actual substitutions if we have a term s(v
1
) in our language such that s(v
1
) =
E
v
1
(v
1
). (We could have such a language, at the cost of taking more functions as
primitive.)
Theorem 18 (variant diagonal lemma) Given a formula F(v
1
) with one free
variable in a language for arithmetic L that has a term s(v
1
) such that for each
number n, s(n) = E
n
(n), there is a sentence C in L such that (C ≡ F(C)) is
true.
Proof. Consider the formula F(s(v
1
)) formed by substituting the term s(v
1
) for
the free occurrences of v
1
in F(v
1
). (1) Let k = F(s(v
1
)). (2) Let C =
df
F(s(k)).
(3) Then s(k) = C. (4) The numeral s(k) designates the same number as is
designated by the terms(k), i.e. the equation s(k) = s(k) is true, so by substitutivity
of identity, (F(s(k)) ≡ F(s(k))). (5) Hence by (2) and (3), (C ≡ F(C)).
3.2 Expressibility of properties of sequence num-
bers in the language L
E
3.2.1 Properties of sequences of digits
The first tool we need in order to code sequences of numbers in L
E
is to show that
we can express in L
E
the relations that the base b notation of a number m begins,
or ends, or is part of the base b notation of a number n.
Definition 29 (x begins y) x begins y in base b notation iff the base b notation
of x is a (not necessarily proper) initial segment of the base b notation of y. We
write this as xB
b
y.
Examples. In base 10, 2 begins 20, but note that in base 13, 2 (as it is written
in base 10, and in base 13) does not begin 20, since 20 base 10 is written 17 base
13. Other examples (base 10): The numbers which written in base 10 are 7, 76, 760,
7600, 76007, 760074, and 7600748 all begin 7600748 in base 10. The last of these
examples points up the fact that every number begins itself, i.e. an initial segment
need not be a proper initial segment. Note that the number 0 does not begin any
number except itself, i.e. we don’t say that 0 begins 760748, even though 0760748
= 760748.
LECTURE 3 33
Definition 30 (x ends y) x ends y in base b notation, which we write as xE
b
y,
if the base b notation of x is an end segment (not necessarily proper) of the base b
notation of y.
Examples. In base 10, the following numbers all end 7600748: 7600748, 600748,
748, 48, 8.
Given the notion of one number beginning another in a given base representation
and the notion of one number ending another in a given base, we can define the
notion of a number being part of another in a given base in terms of these two
notions:
Definition 31 (x is part of y) x is part of y, in base b notation, which we write
as xP
b
y, if x ends some number that begins y.
Remark. Every number is a base b part of itself. Given a base b notation for
x, every proper sub-segment of the base b notation that does not begin with a 0 is
the base b notation of a number y that is a base b part of x. In base 10, the parts
of 2600748 are all the numbers that begin or that end it, and 60074, and all the
numbers that begin or end it, and 7.
Theorem 19 For any b ≥ 2 the following relations are Arithmetical: (1) xB
b
y, (2)
xE
b
y, (3) xP
b
y and, for any natural number n ≥ 2, (4) x
1

b
. . . ∗
b
x
n
P
b
y
Proof. We will prove a stronger result, which we need later, that these relations
not only are expressible in L
E
, but also that this can be done using only bounded
quantifiers, i.e. that these are finitary properties of numbers.
1. If 0 does not occur in the base b numeral for y, then x begins y just in case
there exists z such that x ∗
b
z = y. However, if a zero or a string of zeros occurs
in the base b numeral for y and the base b numeral for x is an initial segment of
the base b numeral of y which ends just before the 0 or string of 0s in the base b
numeral for y or includes some but not all of those 0s, then the numeral of x has
to be extended by the remaining 0s before it can be concatenated with a numeral
to result in the numeral for y. The extension of the base b numeral for x by the
required number of 0s is accomplished by multiplying x by b raised to the power of
how many 0s need to be appended. This condition can be expressed in terms of the
previously expressed notions Pow
b
(w) and x ∗
b
z = y, as follows:
xB
b
y iff (x = y ∨ (x = 0 ∧ (∃z ≤ y)(∃w ≤ y)(Pow
b
(w) ∧ (x w) ∗
b
z = y)))
The bounds on the quantifiers hold from the fact that if z is part of y, then
z ≤ y, and any number of the form 10...0 in base b with a string of 0s of a string of
0s in y is ≤ y.
2. xE
b
y iff (x = y ∨ (∃z ≤ y)(z ∗
b
x = y)
For this case we don’t have any complications from the occurrence of zeros.
3. xP
b
y iff (∃z ≤ y)(xE
b
z ∧ zB
b
y)
4. x
1

b
. . . ∗
b
x
n
P
b
y iff (∃z ≤ y)(x
1

b
. . . ∗
b
x
n
= z ∧ zP
b
y).
LECTURE 3 34
3.2.2 Sequence numbers
Treating sequences of expressions as expressions with G¨odel numbers is very conve-
nient. It requires making sequences of expressions into single expressions, which is
done by expanding the langauge of arithmetic by introducing the symbol ♯ to mark
out the beginning and end of a sequence of expressions and the boundary between
two successive expressions. Thus among the primitive symbols of L
E
(Lecture 1) is
♯, which did not enter into the rules for the formation of terms and formulas of L
E
.
The use of this symbol is to allow us to concatenate a finite sequence of formulas of
L
E
into a single expression of the language, by serving as a marker between different
formulas in a sequence of formulas. When that is done, a sequence of formulas, as
an expression, will have a G¨odel number.
Definition 32 (sequence number) x is a sequence number if it is the G¨ odel
number of an expression of the form ♯E
i
1
♯E
i
2
♯ . . . E
i
k
♯ in L
E
where each expression
E
i
j
does not contain the symbol ♯,
3.3 Coding of finite sequences of G¨odel numbers
Recall the assignment of the first 12 digits of base 13 representation of natural
numbers to the 12 symbols that enter into formulas of the language L
E
:
0

( ) f ′ v ∼ ⊃ ∀ = ≤
1 0 2 3 4 5 6 7 8 9 η ǫ
Thus if a number is the G¨odel number of a formula in L
A
on the particular
G¨odel numbering we have adopted, then the 13th digit, δ, will not occur in its base
13 representation. Call the class of such numbers N
δ
.
A formal proof is a finite sequence of formulas, so to code a proof by a number
it suffices to find a way of coding finite sequences of numbers in N
δ
. We code such a
sequence (a
1
, . . . , a
n
) by the number δ ∗
13
a
1

13
δ ∗
13
a
2

13
δ ∗
13
. . . ∗
13
δ ∗
13
a
n

13
δ. In
future I shall mostly suppress the explicit notation of base 13 (or more generally base
b for any b ≥ 2) concatenation and write v
1
= v
2
v
3
for v
1
= v
2

b
v
3
, i.e. symbolize
the concatenation relation by concatenation itself.
There are several points about the concatenation relation that need to be borne
in mind. (1) It is a three place relation and not a two-place function. (2) It is
a relation between numbers, and numbers are expressible in base b notation for all
b ≥ 2 but are not in base b notation. The situation is similar to what it is in number
theory generally. When we compute with natural numbers we do so using their base
10 notation (or in the case of computers, base 2 notation). But when we prove
something about numbers what we prove is proved using properties of numbers that
LECTURE 3 35
are not specific to decimal notation, even if what is being proved specifically refers
to decimal notation, as in “if the digits of a number in its base 10 notation add up
to 9 then the number is divisible by 9”, which is proved from general properties of
the congruence relation and the fact that 10 ≡ 1 (mod 9).
Proposition 20 (sequence numbers) A natural number n is a sequence number
iff n = δa
1
δa
2
δ . . . δa
n
δ for a
i
∈ N
δ
.
Proof. Immediate from the definition of sequence number above.
Proposition 21 The property of being a sequence number, Seq(v
1
), is Arithmetical,
i.e. it is expressible in L
E
.
Proof. The property of being a sequence number is expressible by the following
formula:
(δBv
1
∧ δEv
1
∧ δ = v
1
∧ ∼ δδPv
1
∧ (∀v
2
≤ v
1
)(δ0v
2
Pv
1
⊃ δBv
2
))
The first four conjuncts characterize the required occurrences of the digit δ in the
base 13 representation of v
1
. The last conjunct rules out the occurrence of a string
of zeros of length greater than one. The reason for this requirement is that sequence
numbers code sequences of numbers in N
δ
. Since 00 = 0, if, e.g., δ00δ were allowed
as a sequence number it would code the sequence (0) but which is also coded by
δ0δ. Since δ00δ = δ0δ, if we allowed, e.g. δ00δ as a sequence number, any sequence
which includes the number zero would have more than one (indeed infinitely many)
sequence numbers.
Definition 33 For v
1
a sequence number, we say that v
2
is in v
1
, symbolized as
v
2
∈ v
1
, iff v
2
is one of the numbers coded by v
1
Proposition 22 v
2
∈ v
1
is Arithmetical.
Proof. v
2
∈ v
1
iff (Seq(v
1
) ∧ δv
2
δPv
1
∧ ∼ δPv
2
). It is a necessary condition for
v
2
∈ v
1
that δv
2
δPv
1
but not sufficient since numbers of the form δa
1
δa
2
δ satisfy it;
the condition ∼ δPv
2
rules out those cases.
In expressing the condition that a number is the G¨odel number of a proof in
a formal system we need to be able to express the condition that one part of a
sequence number occurs earlier in the sequence than another. We do this as follows.
Definition 34 v
2

v
1
v
3
iff v
1
is the sequence number of a sequence in which v
2
and v
3
occur and the first occurrence of v
2
in the sequence is earlier that the first
occurrence of v
3
in the sequence.
LECTURE 3 36
Proposition 23 The three-place relation v
2

v
1
v
3
is Arithmetical.
Proof. v
2

v
1
v
3
iff (v
2
∈ v
1
∧ v
3
∈ v
1
∧ (∃v
4
≤ v
1
)(v
4
Bv
1
∧ v
2
∈ v
4
∧ ∼ v
3
∈ v
4
))
Note that the formulas v
2
∈ v
1
and v
3
∈ v
1
each contains the condition Seq(v
1
),
so we don’t need a separate conjunct Seq(v
1
).
Lecture 4
A formal system PA
E
for
arithmetic; an Arithmetical proof
predicate for PA
E
; a weak version
of G¨ odel’s first incompleteness
theorem for PA
E
(Wednesday, 20 October 2010)
4.1 A formal system PA
E
for arithmetic
We now begin the investigation of formal first-order axiomatic systems of arithmetic.
A theory is first-order if its formal language is first-order. A formal language is first-
order if its quantifiers range only over the objects in its domain of interpretation
and not over collections (pluralities) of those objects. A second-order language has
quantifiers that range over collections (pluralities) of objects (possibly also over
relations between objects). Formal systems of first-order logic are complete (which
G¨odel proved in 1930, in his doctoral thesis–the incompleteness theorem was his
Habilitation thesis). There is no complete axiomatization of full second-order logic.
Accordingly, where we are interested in properties of what can, or cannot, be proved
in formal systems, we shall be concerned only with first-order systems. In this course
we will investigate properties of a number of different formal systems for arithmetic.
All the formal systems we investigate in this course
will explicitly or by assumption contain a com-
37
LECTURE 4 38
plete axiomatization of classical first-order logic
with identity.
We now set out a formal axiomatic system for arithmetic, PA
E
in the language
L
E
. ‘PA’ stands for Peano Arithmetic, a standard misnomer
1
The subscript E
signifies that in this system exponentiation is taken as primitive, i.e. it is governed
by its own axioms. In Lecture 5 we shall see that exponentiation need not be taken
as a primitive and that via coding of ordered pairs of natural numbers the relation
x
y
= z can be expressed in terms of zero, successor, addition, and multiplication.
We shall follow Smullyan in the specification of PA
E
. The axioms are in four
groups, the first two of which, with two rules of inference, is a formal system of
first-order predicate logic, and the third and fourth groups are axioms specific to
arithmetic.
Group I are axiom schemata for propositional logic: These are the standard
axioms for propositional logic with ∼ and ⊃ as primitive. L
1
and L
2
are exactly the
axioms for ⊃ required to establish the Deduction Theorem (taken as proved in a
previous course). L
3
establishes the classical logic of ∼. These axioms are complete
for truth-functional validity.
Group II are axiom schemata for first-order predicate logic. The axiomatization
of first-order predicate logic by the Group II schemata is highly unnatural in terms
of establishing formulas as logically valid. Its virtue for us is that it is very easy to
arithmetize since it involves no substitution of terms for free variables, which more
natural axiomatizations of predicate logic with identity do. Note that L
6
strictly
should be written ∼ ∀v
i
∼ v
i
= t. To prove the following valid formulas from these
schemata is non-trivial: v
i
= v
i
, (v
i
= v
j
⊃ v
j
= v
i
), (∀v
i
F(v
i
) ⊃ F(t)) for t any
term of L
E
not containing a variable that is quantified in F(v
i
) within the scope
of which v
i
occurs. For proofs of these formulas from these schemata see Donald
Kalish and Richard Montague, “On Tarski’s formalization of predicate logic with
identity”, Archiv f¨ ur mathematische Logik und Grundlagenforschung 7 (1965), pp.
81-101, Lemmas 2, 3, and 8 on pp. 85-87.
Group III are axioms specific to each of the primitive non-logical notions of the
language
Group IV is all instances of (a version of) the axiom schema of induction. A
usual formulation of the induction schema is
(F(0) ⊃ (∀v
1
(F(v
1
) ⊃ F(v

1
)) ⊃ ∀v
1
F(v
1
)))
.
1
It was Dedekind who established the first axiomatization of arithmetic, in 1888, which Peano
took over in his publication a year later. Peano cites Dedekind 1888 as the source of his axioms.
It seems to have been Russell who introduced the misnomer Peano Arithmetic.
LECTURE 4 39
However, two formulas within this schema are generated by substitution, namely
F(0) and F(v

1
), and for ease of arithmetization we want to use quasi-substitution
instead of substitution. We can’t use quasi-substitution directly on F(v
1
) to express
F(v

1
), since F[v

1
] would be ∀v
1
(v
1
= v

1
⊃ F(v
1
)), in which no variable occurs
free, so not equivalent to F(v

1
). We could change the variable in the auxiliary
quantification, say to v
2
if v
2
does not occur free in F, i.e. ∀v
2
(v
2
= v

1
⊃ F(v
2
)),
which is logically equivalent to F(v

1
). But this involves substitution of the variable
v
2
for all free occurrences of v
1
in F, which would defeat the purpose of avoiding
substitution. However, we can use a quasi-substitution to obtain from F(v
1
) a
formula logically equivalent to F(v
i
), namely ∀v
1
(v
1
= v
i
⊃ F(v
1
)) where v
i
is any
variable that does not occur in F(v
1
). Then we can use quasi-substitution to obtain
a formula without any substitutions that is logically equivalent to F(v

1
), namely
∀v
i
(v
i
= v

1
⊃ ∀v
1
(v
1
= v
i
⊃ F(v
1
)). We abbreviate this formula as F[[v

1
]]. It’s
easily seen that F[[v

1
]] is logically equivalent to F(v

1
). This logical equivalence only
requires that v
i
does not occur free in F(v
1
), but the sufficient condition that it does
not occur at all in F(v
1
) is easier to express in arithmetized syntax, and that’s the
condition we take.
Definition 35 (the system PA
E
) The axioms and rules of inference of PA
E
are
the following:
A. Logical axioms and rules of inference
Group I propositional logic. All instances of the following schemata:
L
1
(F ⊃ (G ⊃ F))
L
2
(F ⊃ (G ⊃ H)) ⊃ ((F ⊃ G) ⊃ (F ⊃ H))
L
3
((∼ F ⊃∼ G) ⊃ (G ⊃ F))
Group II predicate logic. All instances of the following schemata:
L
4
(∀v
i
(F ⊃ G) ⊃ (∀v
i
F ⊃ ∀v
i
G))
L
5
(F ⊃ ∀v
i
F), provided v
i
does not occur in F.
L
6
∃v
i
(v
i
= t), provided v
i
does not occur in t.
L
7
(v
i
= t ⊃ (X
1
v
i
X
2
⊃ X
1
tX
2
)), where X
1
and X
2
are any expressions such
that X
1
v
i
X
2
) is an atomic formula and t is any term of L
E
.
Rules of inference
R
1
From F and (F ⊃ G), infer G. [Modus Ponens]
R
2
From F, infer ∀v
i
F. [Generalization]
B. Non-logical axioms
Group III axioms specific to each of the primitive non-logical notions of the
language
N
1
(v

1
= v

2
⊃ v
1
= v
2
)
N
2
∼ 0 = v

1
N
3
(v
1
+ 0) = v
1
N
4
(v
1
+v

2
) = (v
1
+v
2
)

LECTURE 4 40
N
5
(v
1
0) = 0
N
6
(v
1
v

2
) = ((v
1
v
2
) +v
1
)
N
7
(v
1
≤ 0 ≡ v
1
= 0)
N
8
(v
1
≤ v

2
≡ (v
1
≤ v
2
∨ v
1
= v

2
))
N
9
(v
1
≤ v
2
∨ v
2
≤ v
1
)
N
10
v
0
1
= 0

N
11
v
v

2
1
= (v
v
2
1
v
1
)
Group IV axiom schema of mathematical induction
N
12
(F[0] ⊃ (∀v
1
(F(v
1
) ⊃ F[[v

1
]]) ⊃ ∀v
1
F(v
1
))),
where, for v
i
any chosen variable that does not occur in F(v
1
), F[[v

1
]] is
∀v
i
(v
i
= v

1
⊃ ∀v
1
(v
1
= v
i
⊃ F(v
1
))). Recall that when we write a schematic formula
F(v
i
), unless we stipulate otherwise, variables other than v
i
may occur free in it.
These other free variables are referred to as parameters.
Definition 36 (proof ) A proof in PA
E
is a sequence of formulas each one of which
is either an axiom of PA
E
or follows from an earlier formula in the sequence by the
rule of Generalization or follows from two earlier formulas in the sequence by Modus
Ponens.
Definition 37 (provable) A formula F of L
E
is provable in PA
E
, symbolized as
PA
E
⊢ F, if there exists a proof in PA
E
of which F is a member.
4.2 An Arithmetical proof predicate for PA
E
Each numbered paragraph in this section is both a definition of a property or relation
of numbers, and a proposition that that property or relation is Arithmetical, the
proof of which is established by the formula that follows. Because it will be needed
for later results we prove, in all but cases (4) and (6), a stronger result than is needed
for the present theorem, namely that the expressing formulas from L
E
require only
bounded quantifiers. The quantifications in (4) and (6) can also be bounded, but
these cases are considerably more complicated. The property of being the G¨odel
number of a provable formula requires an unbounded existential quantifier.
Both for ease of reading and of typesetting in the following I will, after the first
case, abbreviate the abbreviation ∗
13
for base 13 concatenation by using concatena-
tion itself, i.e. for v
i

13
v
j
I will write v
i
v
j
, and for 4 ∗
13
5, which is an abbreviation
for 0
′′′′

13
0
′′′′′
and could also be written f ∗
13


, I will write 45.
(1) V ar(v
1
): E
v
1
is a variable, an expression of the form v
′′
···

, i.e. the variable
symbol followed by a finite string of subscript symbols. Recall that the G¨odel
number of the subscript symbol, on our chosen G¨odel numbering, is 5, and that of
the variable symbol is 6.
LECTURE 4 41
(∃v
2
≤ v
1
)((∀v
3
≤ v
1
)(v
3
P
13
v
2
⊃ 0
′′′′′
P
13
v
3
) ∧ v
1
= 0
′′′′′′

13
v
2
)
In this formula I write out the formal numerals 0
′′′′′
and 0
′′′′′′
rather than ab-
breviating them as 5 and 6, respectively, to bring out the fact that the relation
0
′′′′′
P
13
v
3
) is between numbers, in this case between the number 5 (as we write it
in base 10, and also, as it happens, in base 13 notation) and some other number
and not a relationship between numerals, though the relation between numbers is
determined to hold or not by going from the number to its, in this case, base 13
representation. The number required to exist by the quantification over the variable
v
2
is the G¨odel number of a string of subscripts, by the condition that the subscript
symbol is a part of every part of that expression. All of which is to way that when
the formula V ar(v
1
) is written in the primitive notation of L
E
, the numbers in it
will be expressed by numerals, i.e. 0
′′′′′
for 5 and 0
′′′′′′
for 6, and similarly in the rest
of the formulas expressing arithmetized syntax.
(2) Num(v
1
): E
v
1
is a numeral, i.e. an expression of the form 0
′...′
Pow
13
(v
1
)
(3) Seqt(v
1
): E
v
1
is a formation sequence for terms, i.e. a sequence of expressions
each one of which is either a variable or a numeral or the result of applying one of
the four functions of successor, addition, multiplication, or exponentiation to an
expression or expressions occurring earlier in the sequence, i.e. of the form t

or
(t
1
f

t
2
) or (t
1
f
′′
t
2
) or (t
1
f
′′′
t
2
).
(Seq(v
1
) ∧(∀v
2
≤ v
1
)(v
2
∈ v
1
⊃ (V ar(v
2
) ∨Num(v
2
) ∨(∃v
3
≤ v
1
)(v
3

v
1
v
2
∧v
2
=
v
3
0)∨(∃v
3
≤ v
1
)(∃v
4
≤ v
1
)(v
3

v
1
v
2
∧v
4

v
1
v
2
∧(v
2
= 2v
3
45v
4
3∨v
2
= 2v
3
455v
4
3∨v
2
=
2v
3
4555v
4
3)))))
(4) Tm(v
1
): E
v
1
is a term.
∃v
2
(Seqt(v
2
) ∧ v
1
∈ v
2
)
Note: The formula Seqt(v
2
) in (4) is obtained from the formula Seqt(v
1
) in (3)
by changing the free variable from v
1
to v
2
. In changing the free variable in this
way corresponding changes of bound variables in Seqt(v
1
) must be made so that v
1
is free for v
2
in a logically equivalent transform of Seqt(v
1
), e.g.
(Seq(v
1
) ∧(∀v
5
≤ v
1
)(v
5
∈ v
1
⊃ (V ar(v
5
) ∨Num(v
5
) ∨(∃v
3
≤ v
1
)(v
3

v
1
v
5
∧v
5
=
v
3
0)∨(∃v
3
≤ v
1
)(∃v
4
≤ v
1
)(v
3

v
1
v
5
∧v
4

v
1
v
5
∧(v
5
= 2v
3
45v
4
3∨v
5
= 2v
3
455v
4
3∨v
5
=
2v
3
4555v
4
3)))))
If we had given the formula in (4) as the logically equivalent formula ∃v
5
(Seqt(v
5
)∧
v
1
∈ v
2
), the only change needed to obtain Seqt(v
5
) from Seqt(v
1
) is to replace all
occurrences of v
1
by v
5
.
Note: The formula in (4) above contains an initial unbounded existential quan-
tifier. This quantifier can be bounded by the correlate in arithmetized syntax of
Problem 2 on Problem sheet 1, i.e. decidability of whether or not an expression is
LECTURE 4 42
a term, but it’s a delicate question in which languages, i.e. with what primitives,
that bound can or cannot be expressed.
(5) AF(v
1
): E
v
1
is an atomic formula, i.e. of the form t
1
= t
2
or t
1
≤ t
2
for t
1
, t
2
terms.
(∃v
2
≤ v
1
)(∃v
3
≤ v
1
)(Tm(v
2
) ∧ Tm(v
3
) ∧ (v
1
= v
2
ηv
3
∨ v
1
= v
2
ǫv
3
))
(6) Seqf(v
1
): E
v
1
is a formation sequence for formulas, i.e. a finite sequence of
expressions each one of which is either an atomic formula or of the form ∼ E for E
occurring earlier in the sequence or of the form (E
i
⊃ E
j
) for E
i
and E
j
occurring
earlier in the sequence or of the form ∀v
i
E for v
i
any variable and E occurring earlier
in the sequence.
(Seq(v
1
)∧(∀v
2
≤ v
1
)(v
2
∈ v
1
⊃ (AF(v
2
)∨(∃v
3
≤ v
1
)(v
3

v
1
v
2
∧v
2
= 7v
3
)∨(∃v
3

v
1
)(∃v
4
≤ v
1
)(v
3

v
1
v
2
∧ v
4

v
1
v
2
∧ v
2
= 2v
3
8v
4
3) ∨ (∃v
3
≤ v
1
)(∃v
4
≤ v
1
)(v
3

v
1
v
2
∧ V ar(v
4
) ∧ v
2
= 9v
4
v
3
))))
(7) Fm(v
1
): E
v
1
is a formula.
∃v
2
(Seqf(v
2
∧ v
1
∈ v
2
))
Note: The remark as at (4) above applies here also. We know by Problem 2
on Problem sheet 1, that we can determine by a finite search whether an expression
is a formula, but it’s a delicate matter to determine in exactly what language of
arithmetic, i.e. with what primitives, this numerical quantifier can be bounded by
a term of the language.
(8) Ax(v
1
): E
v
1
is an axiom of PA
E
. There are seven schemata of logical axioms
L
1
− L
7
and eleven axioms of arithmetic N
1
− N
11
plus one axiom schemata of
arithmetic N
12
(Induction). We need formulas L
i
(v
1
) such that L
i
(v
1
) iff (E
v
1
is an
axiom of form L
i
) and N
i
(v
1
) such that N
i
(v
1
) iff (E
v
1
is an axiom of form N
i
). The
property of numbers Ax(v
1
) is expressed by (L
1
(v
1
) ∨ . . . ∨ L
7
(v
1
) ∨ N
1
(v
1
) ∨ . . . ∨
N
12
(v
1
)). Finding N
12
(v
1
) is problem 5 on Exercise sheet 2. I will treat a couple of
cases from each of the other groups of axioms.
Logical axioms:
Group I
L
1
(v
1
): (∃v
2
≤ v
1
)(∃v
3
≤ v
1
)(Fm(v
2
) ∧ Fm(v
3
) ∧ v
1
= 2v
2
82v
3
8v
2
33)
L
3
(v
1
): (∃v
2
≤ v
1
)(∃v
3
≤ v
1
)(Fm(v
2
) ∧ Fm(v
3
) ∧ x = 227v
2
87v
3
382v
3
8v
2
33)
Group II
L
4
(v
1
): (∃v
2
≤ v
1
)(∃v
3
≤ v
1
)(∃v
4
≤ v
1
)(Fm(v
2
) ∧ Fm(v
3
) ∧ V ar(v
4
) ∧ v
1
=
29v
4
2v
2
8v
3
3829v
4
v
2
89v
4
v
3
33)
L
7
(v
1
): (∃v
2
≤ v
1
)(∃v
3
≤ v
1
)(∃v
4
≤ v
1
)(∃v
5
≤ v
1
)(∃v
6
≤ v
1
)(∃v
7
≤ v
1
)(V ar(v
2
)∧
Tm(v
3
) ∧ v
6
= v
4
v
2
v
5
∧ AF(v
6
) ∧ v
7
= v
4
v
3
v
5
∧ v
1
= 2v
2
ηv
3
82v
6
8v
7
33)
Group III
N
1
(v
1
): E
v
1
is the axiom N
1
, which in primitive notation is
((v′ )

= (v′′ )

⊃ (v′ ) = (v′′ )).
LECTURE 4 43
v
1
= 226530η26553082653η265533.
N
7
(v
1
): E
v
1
is the axiom N
7
. To compute the G¨odel number of N
7
we must
write it in primitive notation. This requires expressing ≡ in terms of ∼ and ⊃,
given by the truth-functional equivalences ((A ≡ B) ≡ (A ⊃ B) ∧ (B ⊃ B)) and
((C ≡ D) ≡∼ (C ⊃∼ D)), which yields ((A ≡ B) ≡∼ ((A ⊃ B) ⊃∼ (B ⊃ B))).
So N
7
=∼ (((v′ ≤ 0 ⊃ (v′ ) = 0) ⊃∼ ((v′ = 0 ⊃ (v′ ≤ 0))
v
1
= 7222653ǫ182653η138722653η182653ǫ133
(9) Prf
PA
E
(v
1
): E
v
1
is a proof in PA
E
, i.e. a sequence of formulas each one of
which is either an axiom of PA
E
, or is the result of applying R
1
[Modus Ponens]
to two formulas occurring earlier in the sequence, or is the result of applying R
2
[Generalization] to a formula occurring earlier in the sequence.
(Seq(v
1
) ∧ (∀v
2
≤ v
1
)(v
2
∈ v
1
⊃ (Ax(v
2
) ∨ (∃v
3
≤ v
1
)(∃v
4
≤ v
1
)(v
3

v
1
v
2
∧ v
4

v
1
v
2
∧ v
4
= 2v
3
8v
2
3) ∨ (∃v
3
≤ v
1
)(∃v
4
≤ v
1
)(V ar(v
3
) ∧ v
4

v
1
v
2
∧ v
2
= 9v
3
v
4
)))))
(10) Prov
PA
E
(v
1
, v
2
): E
v
1
is proved by E
v
2
.
Prov
PA
E
(v
1
, v
2
) ≡ (Prf
PA
E
(v
2
) ∧ v
1
∈ v
2
))
(11) Pr
PA
E
(v
1
): E
v
1
is provable
∃v
2
Prov
PA
E
(v
1
, v
2
).
Theorem 24 (Arithmetical proof relation) The two place relation between num-
bers m and n given by the condition that n is the G¨ odel number of a proof in PA
E
of the formula whose G¨ odel number is m is expressible in L
E
.
Proof. The formula Prov
PA
E
(v
1
, v
2
) in (10) expresses “E
v
2
is a proof of E
v
1
”.
This is evident from this formula and (1) - (9).
Corollary 25 (Arithmetical proof predicate) The property of a number that it
is the G¨ odel number of a formula provable in PA
E
is expressible in L
E
,
Proof. PA
E
⊢ E
n
if and only if ∃v
2
Prov
PA
E
(n, v
2
) is true.
Remark: As the construction of the proof predicate for PA
E
shows, arithmeti-
zation by assignment of digits to symbols and of concatenation of corresponding se-
quences of digits (numbers in the given base notation) to concatenation of sequences
of symbols (expressions) makes the correspondence between formal expressions and
numbers, which Hilbert recognized (see quotation in the first lecture) completely
direct. What G¨odel achieved, going beyond Hilbert’s insight, was to show that the
formal syntax of strings of symbols by which a formal system of proof is established
corresponds exactly with arithmetically definable properties of the corresponding
numbers.
LECTURE 4 44
4.3 An inefficient and a weak version of G¨odel’s
First Incompleteness Theorem for PA
E
Theorem 26 (an inefficient proof of G¨odel’s First Incompleteness Theorem)
If every sentence provable in PA
E
is true, there exists a sentence which is true but
unprovable in PA
E
and whose negation is unprovable.
Proof. Having shown that the property of being the G¨odel number of a theorem
of PA
E
is Arithmetical (expressible in L
E
), the existence of a true sentence in the
language of arithmetic not provable in PA
E
follows immediately from Tarski’s theo-
rem, that the set of G¨odel numbers of true sentences in the language of arithmetic is
not arithmetical (Exercise sheet 1 problem 4) and the hypothesis that every sentence
provable in PA
E
is true: By the fact that the G¨odel numbers of provable formulas
is arithmetical, we have that this set is not identical with the set of G¨odel numbers
of true sentences. If every sentence provable in PA
E
is true and the set of provable
sentences does not coincide with the set of true sentences, there must exist a true
sentence which is not provable.
This argument via Tarski’s theorem is highly inefficient since it fails to generate
a particular true sentence that is unprovable in the given system while at the same
time it requires the construction of an arithmetized proof predicate, from which we
can by the Diagonal Lemma, itself used in the proof of Tarski’s theorem, obtain a
particular sentence, namely a diagonal sentence for ∼ Pr
PA
E
(v
1
) which we can show,
on the hypothesis that every sentence provable in PA
E
is true, is unprovable and
(thereby) true, and so whose negation is false and so by the hypothesis unprovable.
We have thus
Theorem 27 (weak version of G¨odel’s First Incompleteness Theorem for PA
E
)
There is a sentence G in L
E
such that, if every sentence provable in PA
E
is true,
PA
E
G, G is true, and PA
E
∼ G.
Proof. By Corollary 25, there is a formula in L
E
, Pr
PA
E
(v
1
), that expresses the
property of being the G¨odel number of a formulas derivable in PA
E
. By the Diagonal
Lemma (Theorem 17) there is a sentence G such that (G ≡∼ Pr
PA
E
(G))). This
establishes conditions (1) and (2) of Theorem 5. Hence on the assumption that
every sentence provable in PA
E
is true (condition (3)), Theorem 5 establishes that
PA
E
G, G is true, and PA
E
∼ G.
Lecture 5
The system PA with zero,
successor, addition, multiplication,
and ≤ as primitive; Σ
0
- and
Σ
1
-formulas; a Σ
0
-coding of finite
sets of ordered pairs; the relation
x
y
= z is ∆
1
-expressible in the
language of PA
.
(Tuesday, 26 October 2010)
5.1 The system PA with zero, successor, addition,
multiplication, and ≤ as primitive
We now drop exponentiation as a primitive of the language of arithmetic and drop
the axioms governing exponentiation from our first-order system of arithmetic PA
E
.
The resulting system is (with variant formulations) the standard first-order axiom
system for arithmetic, known as Peano Arithmetic, labeled PA. The mathematical
fact that justifies the move from PA
E
to PA is that the three-place relation x
y
= z
is expressible in the language of PA. We shall establish not only that exponentia-
tion is expressible in the language of PA, but that it is expressible in a way that
characterizes the total general recursive functions.
45
LECTURE 5 46
Definition 38 (the system PA) The language L for PA is obtained from the lan-
guage L
E
for PA
E
by dropping the condition in the definition of terms for L
E
, Def-
inition 12, that if t
1
and t
2
are terms, then so is (t
1
f′′′ t
2
), which correspondingly
removes from formulas of the language L
E
any expressions that contain the expres-
sion f′′′ (without having to make any change to Definition 16). The axioms of PA
are obtained from those of PA
E
by dropping axioms N
10
and N
11
(which are not
formulas in the language L of PA).
By simple modification of the construction of an Arithmetical proof predicate
for PA
E
in Section 4.2, we obtain an Arithmetical proof predicate for PA.
Theorem 28 (proof predicate for PA in L
E
) There are formulas Prf
E
PA
(v
1
),
Prov
E
PA
(v
1
, v
2
), and Pr
E
PA
(v
1
) in the language L
E
that express the property of being
the G¨ odel number of a proof sequence for PA, the relation of being the G¨ odel number
of a formula that occurs in the proof sequence coded by a given number, and the
property of being the G¨ odel number of a theorem of PA.
Proof. We do this by by modifying the constructions for PA
E
in Theorem 24 and
Corollary 25. We drop the disjunct corresponding to term formation by the function
expression f
′′′
, i.e. v
2
= 2v
3
4555v
4
3, so that Seqt(v
1
) is (Seq(v
1
) ∧ (∀v
2
≤ v
1
)(v
2

v
1
⊃ (V ar(v
2
) ∨ Num(v
2
) ∨ (∃v
3
≤ v
1
)(v
3

v
1
v
2
∧ v
2
= v
3
0) ∨ (∃v
3
≤ v
1
)(∃v
4

v
1
)(v
3

v
1
v
2
∧v
4

v
1
v
2
∧(v
2
= 2v
3
45v
4
3 ∨v
2
= 2v
3
455v
4
3))))). In the formula Ax(v
1
)
that expresses “E
v
1
is an axiom of PA
E
” (p. 39) the disjuncts N
10
(v
1
) and N
11
(v
1
) are
dropped. With these modifications, the formulas for Prf
PA
E
(v
1
), Prov
PA
E
(v
1
, v
2
),
and Pr
PR
E
(v
1
) on p. 40 are transformed to corresponding Arithmetical formulas
Prf
E
PA
(v
1
), Prov
E
PA
(v
1
, v
2
), and Pr
E
PA
(v
1
).
5.2 Σ
0
-formulas
Definition 39 (bounded quantifiers) For v
i
any variable, and n any numeral,
quantification in either of the forms ∀v
i
(v
i
≤ n ⊃ F) or ∃v
i
(v
i
≤ n ∧ F) is called
bounded quantification, abbreviated as (∀v
i
≤ n)F and (∃v
i
≤ n)F, respectively.
Also, for v
i
any variable, v
j
any variable such that i = j, and F any formula,
quantification in either of the forms ∀v
i
(v
i
≤ v
j
⊃ F) and ∃v
i
(v
i
≤ v
j
∧F) are called
bounded quantifiers, abbreviated as (∀v
i
≤ v
j
)F and (∃v
i
≤ v
j
)F, respectively.
Remark. The restriction that the variable v
j
be distinct from the variable v
j
when the bound on the quantification is a variable is essential, since ∀v
i
(v
i
≤ v
i
⊃ F)
is logically equivalent to ∀v
i
F, which is unbounded quantification. Also, note that
bounded existential quantifications are, in primitive notation, formulas of the form
∼ ∀v
i
∼ (v
i
≤ n ∧ F) and ∼ ∀v
i
∼ (v
i
≤ v
j
∧ F).
LECTURE 5 47
Definition 40 (Σ
0
-formulas) .
(a) Every atomic formula of the language L of PA is Σ
0
, i.e. for any terms t
1
and t
2
, t
1
= t
2
and t
1
≤ t
2
are Σ
0
.
(b) If F is Σ
0
, then ∼ F is Σ
0
.
If F and G are Σ
0
then (F ⊃ G) is Σ
0
.
(c) If F is Σ
0
, v
i
is any variable, and t is either a variable distinct from v
i
or a
numeral, then ∀v
i
(v
i
≤ t ⊃ F) is Σ
0
.
Corollary 29 For F any Σ
0
-formula, v
i
any variable, and t either a variable dif-
ferent from v
i
or a numeral, ∃v
i
(v
i
≤ t ∧ F) is Σ
0
.
Proof. By (b), ∼ F(v
i
) is Σ
0
. Then by (c), ∀v
i
(v
i
≤ t ⊃∼ F) is Σ
0
. Hence by
(b) again, ∼ ∀v
i
(v
i
≤ t ⊃∼ F) is Σ
0
. ∼ ∀v
i
(v
i
≤ t ⊃∼ F) is logically equivalent to
∼ ∀v
i
∼ (v
i
≤ t ∧ F), which is abbreviated as ∃v
i
(v
i
≤ t ∧ F), which is to say that
∃v
i
(v
i
≤ t ∧ F) is Σ
0
.
Proposition 30 (Decidability of Σ
0
-sentences) We can effectively decide (com-
pute) the truth or falsity of each Σ
0
-sentence, i.e. closed Σ
0
-formula.
Proof. By induction over the recursive definition of Σ
0
-formulas, corresponding
to the clauses (a), (b), and (c) of the definition:
(a) A closed term is computable to a numeral, and an equation between a term
and a numeral is decidable.
(b) Given the truth value of F and G, we can compute the truth value of ∼ F
and of (F ⊃ G).
(c) Suppose v
i
is not free in F. Suppose F is true. Then (H ⊃ F) is true for any
H, so ∀v
i
(v
i
≤ n ⊃ F) is true. Suppose F is false. Since 0 ≤ n is true, (0 ≤ n ⊃ F)
is false, so ∀v
i
(v
i
≤ n ⊃ F) is false.
Suppose v
i
occurs free in F. Then ∀v
i
(v
i
≤ n ⊃ F(v
i
)) is equivalent to (F(0) ∧
. . . ∧F(n)). By induction hypothesis each conjunct is decidable, so the conjunction
is decidable.
For reasons that will be apparent shortly, we also label Σ
0
-formulas as Π
0
and
as ∆
0
.
5.3 Σ
1
- and Π
1
-formulas; Σ
1
- Π
1
- and ∆
1
-relations
Definition 41 (Σ
1
-formula) A Σ
1
formula is any formula of the form ∃v
i
F where
F is a Σ
0
-formula.
LECTURE 5 48
Definition 42 (Σ
1
-relation) A relation R ⊆ N
k
is Σ
1
iff there is a Σ
1
-formula
G(v
1
, . . . , v
k
) that expresses R, i.e. such that for each k-tuple (n
1
, . . . , n
k
), G(n
1
, . . . n
k
)
is true iff (n
1
, . . . , n
k
) ∈ R.
Note that a Σ
1
-formula begins with one unbounded existential quantifier. (It
may contain other quantifiers so long as they are bounded.)
Proposition 31 Every Σ
0
-formula is logically, and hence provably, equivalent to a
Σ
1
-formula.
Proof. For v
i
not free in F, (F ≡ ∃v
i
F) is logically valid, and hence provable
in every system complete with respect to first-order logical validity.
Definition 43 (Π
1
-formula) A Π
1
formula is any formula of the form ∀v
i
F where
F is a Σ
0
-formula.
Definition 44 (Π
1
-relation) A relation R ⊆ N
k
is Π
1
iff there is a Π
1
-formula
G(v
1
, . . . , v
k
) that expresses R, i.e. such that for each k-tuple (n
1
, . . . , n
k
), G(n
1
, . . . n
k
)
is true iff (n
1
, . . . , n
k
) ∈ R.
Lemma 32 (Σ
1
and Π
1
are dual to each other) The negation of a Σ
1
-formula
is Π
1
, and the negation of Π
1
-formula is Σ
1
.
Proof. By logic and the fact that the negation of a Σ
0
-formula is Σ
0
.
Definition 45 (∆
1
relations) A relation is ∆
1
if and only if it is both Σ
1
and Π
1
.
Remark. Strictly, there is no such thing as a ∆
1
-formula, i.e. ∆
1
is not a syn-
tactic form. However, we shall sometimes label as ∆
1
a formula which is equivalent
both to a Σ
1
-formula and to a Π
1
-formula.
Corollary 33 A relation is ∆
1
iff it is Σ
1
and its complement is also Σ
1
.
Proof. Immediate from Definition 45 and Lemma 32.
Remark. The definitions of Σ
1
and Π
1
-formulas, and by the Remark above
also ∆
1
-formulas, include the case of formulas with no free variables, i.e. sentences.
The definitions of Σ
1
, Π
1
, and ∆
1
-relations include the case of 0-ary relations, i.e.
propositions, and 1-ary relations, i.e. sets.
Example of a ∆
1
-proposition: The proposition expressed by the equivalent
sentences ∃v
1
(v
1
= n ∧ F(v
1
)) and ∀v
i
(v
i
= n ⊃ F(v
i
)), where v
1
is the only free
variable in a Σ
0
-formula F(v
1
). However, as we have seen, these sentences are
logically equivalent to F(n), which is Σ
0
for F(v
1
) Σ
0
. It is an important fact that
there are ∆
1
relations, including 0-ary relations, that are ∆
1
and not Σ
0
, but this is
not a fact we can prove on the basis of results so far obtained.
LECTURE 5 49
Definition 46 (a function is Σ
1
, Π
1
, ∆
1
) For f an n-ary function from N
n
to
N, f is Σ
1
iff the n+1-ary relation f(v
1
, . . . , v
n
) = v
n+1
is Σ
1
. Similarly for Π
i
and

1
.
Lemma 34 If a total function f : N
n
→ N is Σ
1
, then it is ∆
1
.
Proof. We show that the relation f(v
1
, . . . , v
n
) = v
n+1
is also Σ
1
, from which
it follows that the relation f(v
1
, . . . , v
n
) = v
n+1
is ∆
1
. By the hypothesis that f is Σ
1
,
there is a Σ
0
-formula F(v
1
, . . . , v
n
, v
n+1
, v
n+2
) such that ∃v
n+2
F(v
1
, . . . , v
n
, v
n+1
, v
n+2
)
expresses the relation f(v
1
, . . . , v
n
) = v
n+1
. Then, since f is total so that, for every
v
1
, . . . , v
n
, there is v
n+1
such that f(v
1
, . . . , v
n
) = v
n+1
, the relation f(v
1
, . . . , v
n
) =
v
n+1
is expressed by the condition ∃v
n+3
∃v
n+2
(F(v
1
, . . . , v
n
, v
n+3
, v
n+2
)∧ ∼ v
n+3
=
v
n+1
). This condition is equivalently expressed by
∃v
n+4
(∃v
n+3
≤ v
n+4
)(∃v
n+2
≤ v
n+4
)(F(v
1
, . . . , v
n
, v
n+3
, v
n+2
)∧ ∼ v
n+3
= v
n+1
)
By Definition 40,
(∃v
n+3
≤ v
n+4
)(∃v
n+2
≤ v
n+4
)(F(v
1
, . . . , v
n
, v
n+3
, v
n+2
)∧ ∼ v
n+3
= v
n+1
)
is a Σ
0
-formula. Hence the preceding formula is Σ
1
.
5.4 Arithmetization of syntax in the language of
PA
We have already seen that for any base b ≥ 2, concatenation to base b, x∗
b
y = z, is
expressible in L
E
. We now show that for base p for p a prime number, concatenation
to base p is expressible in L, and indeed that it is Σ
0
-expressible in this language.
This result is based on an observation by John Myhill (see Smullyan, p. 43) that
the property of a number x that it is a power of a given prime p can be expressed
without using the exponentiation function, since x is a power of the given prime p
if and only if every proper divisor of x is divisible by p.
Lemma 35 For every prime number p the following conditions are Σ
0
.
1. x [ y —x divides y.
2. Pow
p
(x) —x is a power of p.
3. p

p
(x)
= y —y is the smallest positive power of p greater than x.
Proof.
1. ((∃z ≤ y)(x z = y)∧ ∼ x = 0). Note that by this condition every non-zero
number divides 0 (which reflect the usual practice in number theory that 0 is a
LECTURE 5 50
member of every principal ideal on the natural numbers) and 0 does not divide 0
by the second conjunct, nor any non-zero number by the first conjunct (cf problem
3(a) on Problem sheet 1).
2. (∀z ≤ x)((z [ x ∧ z = 1) ⊃ p [ z).
3. (Pow
p
(y) ∧ y > x ∧ y > 1 ∧ (∀z < y) ∼ (Pow
p
(z) ∧ z > x ∧ z > 1).
Lemma 36 (base p concatenation is Σ
0
) For any prime p, the relation x∗
p
y =
z is Σ
0
.
Proof.
x ∗
p
y = z is Σ
0
iff p

p
(y)
+y = z iff (∃v
1
≤ z)(v
1
= p

p
(y)
∧ ((x v
1
) +y) = z).
The result follows by part 3 of Lemma 35.
Lemma 37 The Arithmetical relation xP
b
y (‘x is part of y’) for base b a prime
number is Σ
0
.
Proof. By Theorem 19, Lemma 35 and Lemma 36.
5.5 A Σ
0
-coding of finite sets of ordered pairs of
numbers
We now establish a result that exponentiation can be expressed in the language of
PA, which does take exponentiation as primitive. The key to this results is a Σ
0
-
coding of finite set of ordered pairs of numbers. This results was proved by G¨odel in
his original paper, for his G¨odel numbering using the Chinese Remainder Theorem.
Given our different G¨odel numbering result, we don’t need the Chinese Remainder
Theorem for this result.
Theorem 38 (Σ
0
-coding of finite sets of ordered pairs) There is a Σ
0
-formula
K(v
1
, v
2
, v
3
) such that
1. For any finite set of ordered pairs of natural numbers (a
1
, b
1
), (a
2
, b
2
), . . . , (a
r
, b
r
),
there is a number k such that for any numbers m and n, K(m, n, k) holds if and
only if (m, n) is one of the pairs (a
1
, b
1
), (a
2
, b
2
), . . . , (a
r
, b
r
).
2. For any numbers v
1
, v
2
, v
3
, if K(v
1
, v
2
, v
3
) holds, then v
1
≤ v
3
and v
2
≤ v
3
.
Proof.
We need to describe two things (dependently related to each other): a process
whereby a set of ordered pairs of numbers (a
1
, b
1
), (a
2
, b
2
), . . . , (a
r
, b
r
) is coded by a
number k, and a formula K(v
1
, v
2
, v
3
) whereby, given a code number k, the set of
ordered pairs is decoded.
LECTURE 5 51
(1) Coding. By a frame number we shall mean a number whose numeral in a
specified base, in our case 13, is of the form 2t2 where t is a string of 1s (this idea goes
back to W.V. Quine, “Concatenation as a basis for arithmetic”, Journal of Symbolic
Logic 11 (1946), pp. 105-114; see Smullyan p. 45). The condition that the numeral
for x in base b representation is a string of 1s, which we shall abbreviate as 1
b
(x),
is expressible by the condition, (∀y ≤ x)(yP
b
x ⊃ 1P
b
y). Examples: 1
10
(111
10
), but
∼ 1
10
(111
13
) since 111
13
= 183
10
, which also means that 1
13
(183
10
).
Let θ be a finite sequence of ordered pairs, (a
1
, b
1
), (a
2
, b
2
), . . . , (a
r
, b
r
), and let
f be any frame number that has a longer string of 1s in it than the longest string
of 1s that occurs in any of the numbers a
1
, b
1
, . . . , a
r
, b
r
. The code k of the set of
ordered pairs is the number ffa
1
fb
1
ffa
2
fb
2
ff . . . ffa
r
fb
r
ff.
(2) Decoding We call x a maximal frame in y if x is a frame number, x is part
of y, and x is as long as any frame that is part of y. This relation is expressed by
the following formula, which we will label MF(x, y):
(xPy ∧ (∃z ≤ y)(1(z) ∧ x = 2z2∧ ∼ (∃w ≤ y)(1(w) ∧ 2zw2Py))).
By Lemmas 36 and 37, MF(x, y) is Σ
0
.
Note that if y has a frame number in it, it has a maximal frame number, since the
length of frame numbers in y is bounded by the length of y. Since frame numbers
whose numerals in the given base are of the same length are equal, any number
which contains a frame number contains a unique maximal frame number.
Having expressed the notion of a maximal frame, we are then able to define a
formula K(v
1
, v
2
, v
3
) with which to decode the ordered pairs coded by the process
specified in (1).
K(v
1
, v
2
, v
3
) =
df
(∃v
4
≤ v
3
)(MF(v
4
, v
3
) ∧ v
4
v
4
v
1
v
4
v
2
v
4
v
4
Pv
3
∧ ∼ v
4
Pv
1
∧ ∼ v
4
Pv
2
).
By Lemmas 36 and 37, the formula K(v
1
, v
2
, v
3
) is Σ
0
.
Let us suppose that the sequence of ordered pairs (a
1
, b
1
), (a
2
, b
2
), . . . , (a
r
, b
r
)
has been coded by the number k using the above procedure. The frame number f
chosen for the coding occurs in k will be a maximal frame number for k since the
string of 1s in f is longer than any strings of 1s from the a
i
and b
i
. As noted above,
it is unique. Having recovered f from k, we can then decode each of the pairs of
numbers coded in k.
Note that there is an error in Smullyan’s proof (p. 45), at the point when he
says, let f be any frame number that is longer than any frame which is part of
any of the numbers a
1
, b
1
, . . . , a
r
, b
r
. The problem is if one or more of the a
i
or b
i
is a string of 1s that is longer than any string of 1s occurring in a frame number
in one of the numbers being coded. Let c be the longest such number and let f
be the frame number specified by Smullyan. In this case the maximal frame in
ffa
1
fb
1
ffa
2
fb
2
ff . . . ffa
r
fb
r
ff is 2c2 and not f, and the decoding fails.
LECTURE 5 52
5.6 The relation x
y
= z is ∆
1
-expressible in the
language of PA
Theorem 39 (exponentiation is Σ
1
) The relation x
y
= z is Σ
1
.
Proof. The relation x
y
= z holds if and only if there is a set of ordered pairs
¦(0, 1), (1, x), (2, x
2
), . . . , (y, x
y
)¦ and (y, z) is a member of that set. Given the coding
and Σ
0
-decoding of finite sets of ordered pairs of numbers by Theorem 38, we can
express this by the formula ∃w(K(y, z, w) ∧ (∀u ≤ w)(∀v ≤ w)(K(u, v, w) ⊃ ((u =
0∧v = 1) ∨(∃r ≤ w)(∃s ≤ w)(K(r, s, w) ∧r

= u∧v = x s))). Since K(v
1
, v
2
, v
3
) is
Σ
0
, this whole formula is Σ
1
. Notice that for this result we need the second clause of
Theorem 38, i.e. that for any numbers v
1
, v
2
, v
3
, if K(v
1
, v
2
, v
3
) holds, then v
1
≤ v
3
and v
2
≤ v
3
in order to bound the universal quantifiers inside the formula, so that
it is, indeed, Σ
1
.
Corollary 40 (exponentiation is ∆
1
) The relation x
y
= z is ∆
1
.
Proof. Immediate from Theorem 39 by Lemma 34.
A stronger result than Corollary 40 is true, namely:
Theorem 41 (exponentiation is ∆
0
) The relation x
y
= z is ∆
0
.
Proof. The proof is too complicated to give here. See Petr H´ajek and Pavel
Pudl´ak, Metamathematics of First-Order Arithmetic, Springer, Berlin, 1993, pp.
299-303.
Theorem 41 and its proof are specific to exponentiation and do not generalize
to other functions. On the other hand, the result in Theorem 39 with Corollary 40
generalizes to the following important result:
Theorem 42 A total function is general recursive if and only if it is ∆
1
.
Proof. Later.
Lecture 6
Every Σ-formula is provably
equivalent to a Σ
1
-formula; the
arithmetized proof predicate for
PA is Σ
1
; the arithmetical
hierarchy
(Wednesday, 27 October 2010)
6.1 Σ-formulas
We now define a class of formulas each of which is provably equivalent in PA to a
Σ
1
-formula, but which are more flexibly expressed than is required in order to be a
Σ
1
-formula.
Definition 47 (Σ-formula) . Base: Every Σ
0
-formula is a Σ-formula.
Recursion: 1. If F is a Σ-formula, then for any variable v
i
, the formula ∃v
i
F is
a Σ-formula.
2. For any Σ-formulas F and G, (F ∨ G) and (F ∧ G) are Σ-formulas.
3. For any Σ
0
-formula F and Σ-formula G, (F ⊃ G) is a Σ-formula.
4. If F is a Σ-formula, then for any distinct variables v
i
and v
j
, (∃v
i
≤ v
j
)F and
(∀v
i
≤ v
j
)F are Σ-formulas, and for any number n, (∃v
i
≤ n)F and (∀v
i
≤ n)F are
Σ-formulas.
To show that every Σ-formula is provably equivalent in PA to a Σ
1
-formula we
prove the following five Lemmas, corresponding to the five clauses in the definition
of a Σ-formula.
53
LECTURE 6 54
Lemma 43 Every Σ
0
-formula is provably equivalent in PA to a Σ
1
-formula.
Proof. Any formula F(v
i
1
, . . . , v
i
k
) is logically equivalent to ∃v
i
k+1
F(v
i
1
, . . . , v
i
k
),
where v
i
k+1
does not occur free in F(v
i
1
, . . . , v
i
k
), so by the completeness of first-order
logic in PA, PA ⊢ (F(v
i
1
, . . . , v
i
k
) ≡ ∃v
i
k+1
F(v
i
1
, . . . , v
i
k
))
Lemma 44 If ∃v
i
F(v
i
) is a Σ
1
-formula, then ∃v
j
∃v
i
F(v
i
) is provably equivalent in
PA to a Σ
1
-formula
Proof. Exercise (Problem sheet 3 problem 2(a))
Lemma 45 Let ∃v
i
F(v
i
) and ∃v
j
G(v
j
) be Σ
1
-formulas. Then (∃v
i
F(v
i
)∨∃v
j
G(v
j
))
and (∃v
i
F(v
i
) ∧ ∃v
j
G(v
j
)) are each provably equivalent in PA to a Σ
1
-formula
Proof.
(a) ((∃v
i
F(v
i
) ∨ ∃v
j
G(v
j
)) ≡ ∃v
k
(F(v
k
) ∨ G(v
k
))) is logically valid (for v
k
any
variable substitutable into F and G) (Cf. Problem sheet 1 problem 5(a)), so provable
in PA.
(b) ((∃v
i
F(v
i
) ∧ ∃v
j
G(v
j
)) ≡ ∃v
i
∃v
j
(F(v
i
) ∧ G(v
j
))) is logically valid (on the
assumption, without loss of generality, that v
i
does not occur free in G and v
j
does
not occur free in F). Hence this equivalence is provable in PA. By Lemma 44,
⊢ (∃v
1
∃v
2
(F(v
i
) ∧ G(v
j
)) is provably equivalent in PA to a Σ
1
formula.
Lemma 46 For F a Σ
0
-formula and ∃v
i
H(v
i
) a Σ
1
-formula, (F ⊃ ∃v
i
H(v
i
)) is
provably equivalent in PA to a Σ
1
-formula.
Proof. (F ⊃ ∃v
i
H(v
i
)) ≡ ∃v
i
(F ⊃ H(v
i
)) is logically equivalent, so provably in
PA. We must assume, which we can do without loss of generality, that v
i
does not
occur free in F. The formula on the right is Σ
1
.
Lemma 47 For ∃v
i
R(v
i
, v
j
) a Σ
1
-formula, the following formulas are provably equiv-
alent in PA to Σ
1
-formulas.
(a) (∃v
j
≤ v
k
)∃v
i
R(v
i
, v
j
)
(b) (∀v
j
≤ v
k
)∃v
i
R(v
i
, v
j
)
and similarly for the formulas in (a) and (b) with a numeral in place of the
variable v
k
.
Proof.
(a) The formula v
j
≤ v
k
is Σ
0
by the definition of Σ
0
. Hence by Lemma 43 it is
provably equivalent to a Σ
1
formula. Hence by Lemma 45 (v
j
≤ v
k
∧∃v
i
R(v
i
, v
j
)) is
provably equivalent to a Σ
1
-formula. Hence by Lemma 44, ∃v
j
(v
j
≤ v
k
∧∃v
i
R(v
i
, v
j
)),
which we abbreviate as (∃v
j
≤ v
k
)∃v
i
R(v
i
, v
j
), is provably equivalent to a Σ
1
-
formula. This argument holds also for a numeral n in place of the variable v
k
.
(b) Exercise (Problem sheet 3 Problem 3(b)).
LECTURE 6 55
Theorem 48 (Σ equivalent to Σ
1
) Every Σ-formula is provably equivalent in PA
to a Σ
1
-formula.
Proof. By induction on the recursive definition of Σ-formulas, the base case and
each induction step established by one of Lemmas 43 - 47.
6.2 The arithmetized proof predicate for PA is Σ
1
Theorem 49 The arithmetized proof predicate for PA under our arithmetization of
syntax of PA is arithmetical, i.e. expressed in the language of PA.
Proof. By Theorem 28 and Lemmas 36 and 37.
A much stronger result than Theorem 49 is true and is needed for what is to
come, namely that the arithmetized proof predicate for PA is Σ
1
. The first step is
to show that the two-place formula (Prf
PA
(v
2
) ∧ v
1
∈ v
2
) is Σ
1
. With considerable
effort we can actually establish that it’s Σ
0
. We know from Problem 2 on Problem
sheet 1 that whether a string of symbols is a term or a formula is decidable, so we
know that there is a bound on those quantifiers, but giving an explicit formulation
of that bound in the language of PA is hard work which we can avoid. However, for
the overall result that the proof predicate for PA is Σ
1
it’s sufficient to show that
(Prf
PA
(v
2
) ∧ v
1
∈ v
2
) is Σ
1
.
Lemma 50 The formula (Prf
PA
(v
2
)∧v
1
∈ v
2
), i.e. (Seq(v
2
)∧(∀v
3
≤ v
2
)(v
3
∈ v
2

(Ax(v
3
) ∨(∃v
4
≤ v
2
)(∃v
5
≤ v
2
)(v
4

v
2
v
3
∧v
5

v
2
v
3
∧v
5
= 2v
4
8v
2
3) ∨(∃v
4
≤ v
2
)(∃v
5

v
2
)(V ar(v
4
) ∧ v
5

v
2
v
3
∧ v
3
= 9v
4
v
5
))) ∧ v
1
∈ v
2
) is equivalent to a Σ
1
-formula.
Proof. The key point is that the only place in the construction of Prf
PA
(v
2
) in
which we used an unbounded quantifier was in Tm(v
1
) (E
v
1
is a term) as ∃v
2
(Seqt(v
2
)∧
v
1
∈ v
2
) and Fm(v
1
) (E
v
1
is a formula) as ∃v
2
(Seqf(v
2
∧ v
1
∈ v
2
)). These occur
the formula Ax(v
1
) in L
1
− L
7
and L
12
(Induction axioms). In none of these oc-
currences is it in the antecedent of a conditional, and the occurrence of Ax(v
1
) in
(Prf
PA
(v
2
) ∧ v
1
∈ v
2
) is also not in the antecedent of a conditional. Hence in a
prenex normal form of (Prf
PA
(v
2
) ∧ v
1
∈ v
2
) these several existential quantifiers
come out as prenex existential quantifiers. By Lemma 44, this formula is Σ
1
.
Theorem 51 (proof predicate for PA is Σ
1
) The formula Pr
PA
(v
1
),
i.e. ∃v
2
(Prf
PA
(v
2
) ∧ v
1
∈ v
2
), is equivalent to a Σ
1
-formula.
Proof. By Lemma 50 and Lemma 44.
LECTURE 6 56
Corollary 52 ¦n : PA ⊢ E
n
¦ is Σ
1
.
Proof. By inspection of the formula ∃v
2
(Prf
PA
(v
2
) ∧ v
1
∈ v
2
).
Theorem 53 ¦n : PA ⊢ E
n
¦, which by Corollary 52 is Σ
1
, is not ∆
1
, if every
sentence provable in PA is true.
Proof. We require the result, which we will prove in Lecture 7, that PA is Σ
1
-
complete. Suppose the complement of ¦n : PA ⊢ E
n
¦ is expressed by a Σ
1
-formula,
call it NPr
PA
(v
1
). Then the G¨odel sentence G, such that (G ≡∼ Pr(G)) is
true, is equivalent to the Σ
1
-sentence NPR
PA
(G). By a simple modification of
Theorem 27 so that it applies to PA rather than PA
E
, G is true and, if every
sentence provable in PA is true, not provable in PA. But by the Σ
1
-completeness of
PA (to be proved in the next lecture), if G is true and equivalent to a Σ
1
-sentence,
then PA⊢ G, which contradicts the unprovability of G on the hypothesis that every
sentence provable in PA is true. (In Lecture 8 we shall prove this result on a much
weaker hypothesis.)
Proposition 54 ¦n : PA ⊢ E
n
[n]¦ is Σ
1
.
Proof. Exercise (Problem sheet 3 problem 1)
6.3 The arithmetical hierarchy
The kind of classifications we introduced in Lecture 5 with the notions of Σ
0
, Π
0
,

0
and Σ
1
, Π
1
, ∆
1
can be extended, by Theorem 3 (Prenex Normal Forms) and
a generalization of Theorem 48, to a hierarchy of all formulas in the language of
PA. This correspondingly defines a hierarchy of relations on natural numbers (the
arithmetical hierarchy).
Remark. A simple cardinality argument tells us that most relations on the
natural numbers, in particular the 1-ary relations, i.e. the sets of natural numbers,
are not in this hierarchy (there are uncountably many sets of natural numbers and
there are countably many formulas in the language of arithmetic).
Definition 48 (arithmetical hierarchy of formulas) (a) Σ
0
-formulas (=
df
Π
0
-
formulas) are as given by Definition 40.
(b) If F is Σ
n
, then ∀v
i
F is Π
n+1
.
(c) If F is Π
n
, then ∃v
i
F is Σ
n+1
.
There is a corresponding arithmetical hierarchy of sets and relations.
LECTURE 6 57
Definition 49 (arithmetical hierarchy of sets and relations) A relation on nat-
ural numbers is Σ
n
(or Π
n
) if and only if it is expressible by a Σ
n
-formula (respec-
tively a Π
n
-formula) in L.
Definition 50 If a relation is both Σ
n
and Π
n
, it is said to be ∆
n
.
In order to generalize Theorem 48, and for many other purposes, we require a
Σ
0
-pairing function.
Lemma 55 (Σ
0
pairing function) The function p(m, n) =
1
2
(m+n+1)(m+n)+
m is a bijection between the natural numbers and pairs of natural numbers which is
strictly increasing in both arguments, and it is Σ
0
.
Proof. Exercise (Problem sheet 3 problem 3).
Theorem 56 (1) for n > 0, formulas provably equivalent in PA to Σ
n
-formulas
are closed under existential quantification and formulas provably equivalent to
Π
n
-formulas are closed under universal quantification.
(2) formulas provably equivalent to Σ
n
-formulas, and formulas provably equivalent
to Π
n
-formulas, are both closed under conjunction and disjunction.
(b) Show that formulas provably ∆
n
are closed under conjunction, disjunction, and
negation.
Proof. By a single induction on n in the conjunction of the three statements.
(Problem sheet 3 problem 4).
Corollary 57 Every formula in L is equivalent to a Σ
n
or Π
n
formula for some n.
Proof. For a given formula, find a prenex normal form for it. By Theorem 56
(3), adjacent like quantifiers can be collapsed to a single quantifier.
Lecture 7
Σ
0
-completeness and
Σ
1
-completeness; weak systems of
arithmetic Q and R (without
induction); Σ
0
-completeness of
systems R, Q, and PA;
Σ
0
-soundness and Σ
1
-soundness
(Tuesday, 2 November 2010)
7.1 Σ
0
-completeness and Σ
1
-completeness
Definition 51 (Σ
0
-completeness) A system S is Σ
0
-complete iff for each true
Σ
0
-sentence X, S ⊢ X.
Definition 52 (Σ
1
-completeness) A system S is Σ
1
complete iff for each true
Σ
1
-sentence X, S ⊢ X.
Proposition 58 A system is Σ
1
-complete iff it is Σ
0
-complete.
Proof. Assume S is Σ
0
-complete and let X be a true Σ
1
-sentence, i.e. a sentence
of the form ∃v
i
F(v
i
) where F(v
i
) is a Σ
0
-formula. Since X is true, for some number
k, F(k) is a true Σ
0
-sentence. By Σ
0
-completeness of S, S ⊢ F(k). Then by
predicate logic in S, S ⊢ ∃v
i
F(v
i
).
58
LECTURE 7 59
Assume S is Σ
1
-complete and let X be a true Σ
0
-sentence. Then ∃v
1
(X∧v
1
= v
1
)
is a true Σ
1
-sentence, so S ⊢ ∃v
1
(X ∧ v
1
= v
1
). Since X is a sentence, v
1
does not
occur free in X, so (∃v
1
(X ∧ v
1
= v
1
) ⊃ (X ∧ ∃v
1
v
1
= v
1
)) is logically valid and
hence provable in S. So by propositional logic in S, S ⊢ X.
In this lecture we shall see that PA and much weaker subsystems of PA are
Σ
0
-complete.
Definition 53 (A system S correctly decides a sentence X) A system S cor-
rectly decides a sentence X iff either X is true and S ⊢ X, or X is false and S ⊢∼ X.
Lemma 59 A system S is Σ
0
-complete iff S correctly decides every Σ
0
-sentence.
Proof. Half of the condition that S correctly decides every Σ
0
-sentence is that
if X is a true Σ
0
-sentence then S ⊢ X, i.e. Σ
0
-completeness of S.
Conversely, suppose S is Σ
0
-complete, and X is any false Σ
0
-sentence. Then
∼ X is a true Σ
0
-sentence, so by Σ
0
-completeness, S ⊢∼ X, as required.
Proposition 60 The following two conditions on a system S together imply that S
correctly decides every Σ
0
-sentence.
C
1
. S correctly decides every atomic sentence.
C
2
. For any Σ
0
-formula F(v
i
) with v
i
the only free variable and for every number
n, if S ⊢ F(0), . . . , S ⊢ F(n), then S ⊢ (∀v
i
≤ n)F(v
i
).
Proof. By induction over the recursive definition of Σ
0
-formulas.
1. By C
1
S correctly decides all atomic Σ
0
-sentences.
2. By propositional logic in S, if X and Y are correctly decided by S, then ∼ X
and X ⊃ Y are correctly decided by S. [Exercise]
3. Any Σ
0
-sentence Z that is neither atomic nor of the form ∼ X or X ⊃ Y
must be of the form (∀v
i
≤ n)F(v
i
) where F(v
i
) is a Σ
0
-formula of lower degree
than Z and contains v
i
as its only free variable.
Suppose (∀v
i
≤ n)F(v
i
) is true. That means that each of the sentences F(0), . . . , F(n)
is true. Then by induction hypothesis, S ⊢ F(0), . . . , S ⊢ F(n). Then by condition
C
2
, S ⊢ (∀v
i
≤ n)F(v
i
).
Suppose (∀v
i
≤ n)F(v
i
) is false. Then for some m ≤ n, F(m) is false. Then
by induction hypothesis we have that S ⊢∼ F(m). Since m ≤ n is a true atomic
Σ
0
-formula, by C
1
, S ⊢ m ≤ n. Then by 2. S ⊢∼ (m ≤ n ⊃ F(m)).
The following formula is logically valid:
(∀v
i
(v
i
≤ n ⊃ F(v
i
)) ⊃ (m ≤ n ⊃ F(m)))
and hence provable in S, so by propositional logic in S,
S ⊢∼ ∀v
i
(v
i
≤ n ⊃ F(v
i
)), which is to say S ⊢∼ (∀v
i
≤ n)F(v
i
).
LECTURE 7 60
Proposition 61 The following three conditions on a system S jointly imply that S
is Σ
0
-complete.
D
1
. All true atomic sentences are provable in S.
D
2
. For any distinct numbers m and n, S ⊢∼ m = n.
D
3
. For any variable v
i
and any number n, S ⊢ (v
i
≤ n ⊃ (v
i
= 0∨. . . ∨v
i
= n)).
Proof. We show that conditions D
1
, D
2
, D
3
imply conditions C
1
, C
2
of the
previous Proposition, which establishes that S correctly decides every Σ
0
-sentence,
and is thereby Σ
0
-complete.
Specifically, D
1
, D
2
, D
3
together imply C
1
, and D
3
implies C
2
.
1. To establish C
1
, i.e. that S correctly decides all atomic sentences.
We are given by D
1
that all true atomic sentences are provable in S. So it
remains to show that all false atomic sentences are refutable in S.
(i) If the false atomic sentence is of the form t
1
= t
2
, there are true atomic
sentences of the form t
1
= m and t
2
= n, where m = n. Then by D
1
, S ⊢ t
1
= m,
and S ⊢ t
2
= n, and by D
2
S ⊢∼ m = n. Then by propositional logic in S,
S ⊢∼ t
1
= t
2
.
(ii) If the false atomic sentence is of the form t
1
≤ t
2
, there are true atomic
sentences of the form t
1
= m and t
2
= n, where ∼ m ≤ n. From D
1
and the fact
that t
1
= m and t
2
= n are true, S ⊢ t
1
= m and S ⊢ t
2
= n. Since ∼ m ≤ n, all the
sentences m = 0, . . . , m = n are false. Hence by D
2
, S ⊢∼ m = 0, . . . , S ⊢∼ m = n.
Then by propositional logic in S, S ⊢∼ (m = 0 ∨ . . . ∨ m = n).
From D
3
, by Generalization and Instantiation,
S ⊢ (m ≤ n ⊃ (m = 0 ∨ . . . ∨ m = n)). Hence by propositional logic in S,
S ⊢∼ m ≤ n. Then by substitutivity of identity, ⊢∼ t
1
≤ t
2
.
2. To show that D
3
implies C
2
.
Suppose F(v
i
) is a Σ
0
formula with v
i
its only free variable, and that n is a
number such that S ⊢ F(0), . . . , S ⊢ F(n). Then by pure logic (with identity) in
S, S ⊢ (v
i
= 0 ⊃ F(v
i
)), . . . , S ⊢ (v
i
= n ⊃ F(v
i
)). Then by propositional logic in
S, S ⊢ ((v
i
= 0 ∨ . . . ∨ v
i
= n) ⊃ F(v
i
)). Then by D
3
and propositional logic in S
(transitivity of ⊃), S ⊢ (v
i
≤ n ⊃ F(v
i
)). Then by pure logic (Generalization) in S,
S ⊢ ∀v
i
(v
i
≤ n ⊃ F(v
i
)), i.e. S ⊢ (∀v
i
≤ n)F(v
i
).
7.2 Weak systems of arithmetic Q and R (without
induction)
Definition 54 (system Q) The system Q is obtained from the system PA by drop-
ping the axiom schema for induction, N
12
. Thus Q has only finitely many (nine)
nonlogical axioms:
N
1
(v

1
= v

2
⊃ v
1
= v
2
)
LECTURE 7 61
N
2
∼ v

1
= 0
N
3
v
1
+ 0 = v
1
N
4
v
1
+v

2
= (v
1
+v
2
)

.
N
5
v
1
0 = 0
N
6
v
1
v

2
= (v
1
v
2
) +v
1
N
7
(v
1
≤ 0 ≡ v
1
= 0)
N
8
(v
1
≤ v

2
≡ (v
1
≤ v
2
∨ v
1
= v

2
)).
N
9
(v
1
≤ v
2
∨ v
2
≤ v
1
).
The system Q is a variant of one due to Raphael Robinson. We will show that
all true Σ
0
-sentences are provable Q and so provable in PA since all the axioms of
Q are axioms of PA. We prove this by proving a yet stronger result, namely that an
even weaker system R, also due to Raphael Robinson, is Σ
0
-complete. Instead of
the (finitely) many recursion axioms of PA and Q, it has as axioms infinitely many
instances of computations of addition, multiplication, and inequality, in three axiom
schemata, and two axiom schemata expressing properties of ≤.
Definition 55 (system R) The axioms of R are all sentences sentences and for-
mulas of L generated from natural numbers m and n by the following axiom schemata:

1
m+n = m+n.

2
m n = m n.

3
∼ m = n where m = n.

4
(v
1
≤ n ⊃ (v
1
= 0 ∨ . . . ∨ v
1
= n)).

5
(v
1
≤ n ∨ n ≤ v
1
).
Note that by our notational conventions the symbol + in the formulation of Ω
1
is
used to abbreviate expressions in the formal language L and does not actually occur
in sentences of the form Ω
1
, and also that + abbreviates different formal expressions
in its two occurrences in the formulation of Ω
1
. The instances of Ω
1
are generated
from pairs of natural numbers, m and n, by writing an equation between the term
(mf

n) on the left and the term 0 with m + n many occurrences of the symbol

suffixed to it on the right. A corresponding remark holds concerning occurrences of
the dot symbol for multiplication in the above formulation of Ω
2
, e.g. an instance
of Ω
2
is (0
′′
f
′′
0
′′′
) = 0
′′′′′′
.
We now show that the system R proves the converse of Ω
4
.
Lemma 62 For each natural number n, Ω
5
⊢ n ≤ n.
Proof. By ∀-Intro and ∀-Elim on the instance of Ω
5
for n, Ω
5
⊢ (n ≤ n∨n ≤ n),
so by propositional logic, Ω
5
⊢ n ≤ n.
LECTURE 7 62
Lemma 63 For each number n, (R) ⊢ (v
1
= n ⊃ v
1
≤ n).
Proof.
(1) v
1
= n ⊃ (n ≤ n ⊃ v
1
≤ n) substitutivity of identity
(2) (2) v
1
= n Assumption
(2) (3) (n ≤ n ⊃ v
1
≤ n) (1) (2) ⊃-elimination
(4) n ≤ n Lemma 62
(2) (5) v
1
≤ n (2)(4) ⊃-elimination
(6) v
1
= n ⊃ v
1
≤ n (2)(5) ⊃-introduction
Lemma 64 For each number k, (R) ⊢ ((v
1
= 0 ∨ . . . ∨ v
1
= k) ⊃ v
1
≤ k

)
Proof. Let m be any number ≤ k
(1) k

= 0, . . . , k

= m Ω
3
(2) (k

= 0 ∧ . . . ∧ k

= m) (2) ∧-introduction
(3) ∼ (k

= 0 ∨ . . . ∨ k

= m) (3) DeMorgan law
(4) (k

≤ m ⊃ (k

= 0 ∨ . . . ∨ k

= m)) Ω
4
∀-introduction and elimination
(5) ∼ k

≤ m (3), (4) prop. logic
(6) (v
1
= m ⊃∼ k

≤ v
1
) (5) substitutivity of identity
(7) ((v
1
= 0 ∨ . . . ∨ v
1
= k) ⊃∼ k

≤ v
1
) (6), for each m ≤ k, by ∨-elimin
(8) (v
1
≤ k

∨ k

≤ v
1
) instance of Ω
5
(9) ((v
1
= 0 ∨ . . . ∨ v
1
= k) ⊃ v
1
≤ k

) (7), (8) prop. logic
Theorem 65 R ⊢ ((v
1
= 0 ∨ . . . ∨ v
1
= n) ⊃ v
1
≤ n).
Proof. The results follows by ∨-elimination from (n = 0 ∨ n = 0). The case for
n = 0 is an instance of Lemma 63, i.e. R ⊢ (v
1
= 0 ⊃ v
1
≤ 0). If n = 0 then there
is some number k such that n = k + 1. This case follows by ∨-elimination from
the disjunction ((v
1
= 0 ∨ . . . ∨ v
1
= k) ∨ v
1
= k

). We have v
1
≤ k

from the first
disjunct by Lemma 64 and v
1
≤ k

from the second disjunct by Lemma 63. So by
⊃-Intro, R ⊢ (((v
1
= 0 ∨ . . . ∨ v
1
= k) ∨ v
1
= k

) ⊃ v
1
≤ k

).
7.3 Σ
0
-completeness of systems R, Q, and PA
We will now show the R is Σ
0
-complete. To do this we need the following lemma.
LECTURE 7 63
Lemma 66 (evaluation of closed terms by R) For each closed term t in L,
there is a unique numeral n such that R ⊢ t = n.
Proof. If t is a closed term it is either a numeral, or there is a closed term t
1
such that t is the term t

1
, or there are closed terms t
1
and t
2
such that t is (t
1
f′ t
2
),
i.e. t
1
+t
2
, or t is (t
1
f′′ t
2
), i.e. t
1
t
2
. If t is a numeral n then R ⊢ t = n by reflexivity
of identity. If t is of the form t

1
for some term t
1
, then by induction hypothesis there
is a numeral n such that R ⊢ t
1
= n. Then by the logic of identity, R ⊢ t

1
= n

.
But n

is n + 1, which is to say that R ⊢ t = n + 1. If t is of the form t
1
+ t
2
,
then by induction hypothesis there are numbers m
1
and m
2
such that R ⊢ t
1
= m
1
and R ⊢ t
2
= m
2
. Let m
1
+ m
2
= k. Then by Ω
1
, R ⊢ m
1
+ m
2
= k. Then by
substitutivity of identity,R ⊢ t
1
+ t
2
= k. The argument is the same for t of the
form t
1
t
2
.
The uniqueness of n for a given term t follows by transitivity of identity.
Proposition 67 The system R is Σ
0
-complete.
Proof. We establish this result by showing that R satisfies the conditions D
1
D
2
D
3
of Proposition 61.
D
1
. (i) Suppose t
1
and t
2
are terms such that t
1
= t
2
is a true sentence. Since
t
1
= t
2
is a sentence, the terms t
1
and t
2
contain no variables. Hence by Lemma 66
there are numbers m
1
and m
2
such that R ⊢ t
1
= m
1
and R ⊢ t
2
= m
2
. Since t
1
= t
2
is true, m
1
= m
2
, so m
1
and m
2
are the same numeral, so by reflexivity of identity
in R, R ⊢ m
1
= m
2
. Hence by transitivity of identity in R, R ⊢ t
1
= t
2
.
(ii) Suppose t
1
and t
2
are terms such that t
1
≤ t
2
is a true sentence. Then t
1
and t
2
have no free variables, so by Lemma 66 there are natural numbers m
1
and
m
2
such that t
1
= m
1
and t
2
= m
2
are true sentences, and so by (i) provable in S.
Since t
1
≤ t
2
is true, m
1
≤ m
2
. By the logic of identity, R ⊢ m
1
= m
1
and so by
propositional logic, R ⊢ (m
1
= 0 ∨. . . ∨m
1
= m
1
∨. . . ∨m
1
= m
2
). By Theorem 65,
with ∀-introduction and ∀-elimination, and Modus ponens, R ⊢ m
1
≤ m
2
. Then by
substitutivity of identity, R ⊢ t
1
≤ t
2
.
D
2
. This is the schema Ω
3
.
D
3
. This is a direct consequence of Ω
4
by ∀-Intro and ∀-Elim.
Proposition 68 R is a subsystem of Q.
Proof. We must show that each axiom of R, i.e. every instance of Ω
1
, Ω
2
, Ω
3
,

4
, is provable in Q.

1
: X is of the form m + n = m+n for natural numbers m and n. We argue
by induction on n in the statement Q ⊢ m+n = m+n.
Note that while Q does not contain induction, use of induction for this argument
is valid since we are arguing about Q and not in Q.
LECTURE 7 64
n = 0. By N
3
, ∀-I and ∀-E, Q ⊢ m+0 = m. Since m = m+0, m and m+ 0 are
the same term (formal numeral), so Q ⊢ m+ 0 = m+ 0.
Induction step.
(1) Q ⊢ m+n = m+n Induction hypothesis
(2) Q ⊢ m+n

= (m+n)

N
4
, ∀-I, ∀-E
(3) Q ⊢ m+n

= (m+n)

(1), (2), substitutivity of = in Q
(4) Q ⊢ m+n + 1 = ((m+n) + 1) (3), Corollary 8
(5) ((m+n) + 1) = (m+ (n + 1)) truth of arithmetic
(6) (m+n) + 1) and (m+ (n + 1)) are the same term (5)
(7) Q ⊢ m+n + 1 = m+ (n + 1) (4), (6)

2
: We show by induction on n that Q ⊢ m n = m n.
n = 0 By N
5
, ∀-Intro, ∀-Elim, Q ⊢ m 0 = 0. Since 0 = m 0, 0 and m 0 are
the same term. So Q ⊢ m 0 = m 0.
Induction step.
(1) Q ⊢ m n = m n Induction hypothesis
(2) Q ⊢ m n

= m n +m N
6
, ∀-I, ∀-E
(3) Q ⊢ m n

= m n +m (1), (2), substitutivity of = in Q
(4) Q ⊢ m n + 1 = m n +m (3), Corollary 8
(5) Q ⊢ m n + 1 = m n +m (4), previous case for Ω
1
(6) m n +m = m (n + 1) truth of arithmetic
(7) m n +m and m (n + 1) are the same term (6)
(8) Q ⊢ m n + 1 = m (n + 1) (5), (7)

3
: To show that for every m and n such that m = n, Q ⊢∼ m = n. Suppose
that m = n. Without loss of generality, we may suppose that m > n, since by logic
of identity in Q, if Q ⊢∼ m = n, then Q ⊢∼ n = m. We argue by cases.
n = 0. Then since m = 0, there is a number k such that k + 1 = m. By
Corollary 8, m is k

. By N
2
, ∀-Intro, ∀-Elim, Q ⊢∼ k

= 0, i.e. Q ⊢∼ m = 0.
n = 0. Since m > n, there is a non-zero number d such that m = d + n. By
taking d in place of m in the argument for the previous case, Q ⊢∼ d = 0. By
contraposition of N
1
, ∀-Intro, ∀-Elim, Q ⊢ (∼ d = 0 ⊃∼ d

= 0

), so by Modus
ponens, Q ⊢∼ d

= 0

. By n many applications of this argument,
Q ⊢∼ d
n

′ . . . ′
= 0
n

′ . . . ′
, i.e. Q ⊢∼ m = n.

4
: We show by induction on n that for each n, Q ⊢ (v
1
≤ n ⊃ (v
1
= 0∨. . .∨v
1
=
n)).
LECTURE 7 65
n = 0. By ∧-Elim from N
7
, Q ⊢ (v
1
≤ 0 ⊃ v
1
= 0).
Assume, as induction hyposthesis, that Q ⊢ (v
1
≤ n ⊃ (v
1
= 0 ∨ . . . ∨ v
1
= n)).
By ∧-Elim, ∀-Intro, ∀-Elim from N
8
, Q ⊢ (v
1
≤ n

⊃ (v
1
≤ n ∨ v
1
= n

)). Then
by ∨-Elim, Q ⊢ (v
1
≤ n

⊃ ((v
1
= 0 ∨ . . . ∨ v
1
= n) ∨ v
1
= n

)), i.e. by Corollary 8,
Q ⊢ (v
1
≤ n + 1 ⊃ (v
1
= 0 ∨ . . . ∨ v
1
= n ∨ v
1
= n + 1)).

5
: N
9
⊢ Ω
5
by ∀-Intro and ∀-Elim.
Proposition 69 Q is Σ
0
-complete.
Proof. By Propositions 67 and 68.
Theorem 70 PA is Σ
0
-complete.
Proof. By Proposition 69 and the fact that PA is an extension of Q.
7.4 Σ
0
-soundness and Σ
1
-soundness
Definition 56 (Σ
0
-soundness) A system S is Σ
0
-sound if and only if for every
Σ
0
-sentence X, if S ⊢ X, then X is true (in the structure of the natural numbers).
Definition 57 (Σ
1
-soundness) A system S is Σ
1
-sound if and only if for every
Σ
1
-sentence X, if S ⊢ X, then X is true (in the structure of the natural numbers).
Proposition 71 If a consistent system is Σ
0
-complete, it is Σ
0
-sound.
Proof. Let S be a Σ
0
-complete system and let X be a false Σ
0
-sentence such
that S ⊢ X. Since X is Σ
0
and false, ∼ X is Σ
0
and true. Hence by Σ
0
-completeness
of S, S ⊢∼ X. But this means that S is inconsistent, contrary to hypothesis.
Remark. Though a system is Σ
0
-complete if and only if it is Σ
1
-complete
(Proposition 58), there is no result corresponding to Proposition 71 that holds for
Σ
1
-completeness. As we shall see in Lecture 8, there are Σ
1
-complete systems that
are not Σ
1
-sound. In any case it is clear that the proof of Proposition 71 does not
extend to the case of Σ
1
-completeness since in general the negation of a Σ
1
-sentence
is not Σ
1
.
Lecture 8
The notions of consistency,
ω-consistency and 1-consistency;
incompleteness from the
assumption of 1-consistency; truth
of the G¨ odel sentence;
ω-incompleteness.
(Wednesday, 3 November 2010)
8.1 The notions of consistency, ω-consistency and
1-consistency.
Definition 58 (consistency) A system S is consistent if there is no formula X
in the language of S such that S ⊢ X and S ⊢∼ X.
Proposition 72 A system S containing propositional logic is consistent if and only
if there is formula Y in the language of S such that S Y .
Proof. (i) Left to right: We prove the contrapositive. Suppose for every X,
S ⊢ X. Then in particular for any formula Y , S ⊢ Y and S ⊢∼ Y .
(ii) Right to left: We prove the contrapositive. Suppose S is inconsistent, i.e.
there is a formula Y such that S ⊢ Y and X ⊢∼ Y . Then by ∧-introduction (as a
derived rule if not a primitive rule of S), S ⊢ (Y ∧ ∼ Y ). Propositional logic proves
66
LECTURE 8 67
((Y ∧ ∼ Y ) ⊃ Z) for every formula Z. So by Modus Ponens in S, S ⊢ Z for every
formula Z.
Remark. Proposition 72 shows that we could equivalently have defined consis-
tency by:
Definition 59 (alternative definition of consistency) S is consistent if there
is a formula X such that S X.
Definition 60 (ω-consistency) A system S in a language that contains a closed
term n, i.e. a numeral, for each natural number, is said to be ω-consistent if and
only if there is no formula F(v
i
) with one free variable in L such that S ⊢ ∃v
i
F(v
i
)
and for each natural number n, S ⊢∼ F(n).
Proposition 73 If a system is ω-consistent, then it is consistent.
Proof. The contrapositive is immediate by ex falso quodlibet: if a system S
is inconsistent S proves every formula in the language of S; in particular for any
formula F(w) with one free variable, S ⊢ ∃wF(w), and for each n S ⊢∼ F(n), i.e.
S is ω-inconsistent.
The converse of Proposition 73 does not hold, i.e.
Proposition 74 There are consistent systems that are ω-inconsistent.
Proof. Exercise (problem 3(a) of Problem sheet 4).
Proposition 75 If a system is sound with respect to truth in arithmetic, then it is
ω-consistent.
Proof. Let S be a system whose language contains numerals for the natural
numbers and which is sound with respect to truth in arithmetic. Suppose S ⊢
∃wF(w). Then ∃wF(w) is true, i.e. there is a natural number n such that F(n) is
true, which is to say that ∼ F(n) is false. So S ∼ F(n), which is to say that S is
ω-consistent.
The converse holds only to a strictly limited extent, as detailed by the following
two propositions, and in general ω-consistency does not imply truth.
Proposition 76 If a system S is Σ
0
-complete and ω-consistent, it is Σ
2
-sound, i.e.
if sentence X is Σ
2
and S ⊢ X, then X is true.
Proof. Exercise (problem 2(b) of Problem sheet 4).
Remark This result is best possible, i.e.
LECTURE 8 68
Proposition 77 There is an ω-consistent system that proves a false Σ
3
-sentence.
Proof. Exercise (problem 4(c) of Problem sheet 4).
G¨odel introduced the notion of ω-consistency in order to prove the second half
of his First Incompleteness Theorem. It is a very much weaker hypothesis than that
the system is sound, i.e. that all theorems are true (which we saw already in Lec-
ture 1 is sufficient for the result), established by Proposition 77, which shows that
ω-consistency only implies a very limited amount of truth. Even so, ω-consistency
is a considerably stronger hypothesis than is necessary to established formal in-
completeness. From the fact that a proof predicate for a formal deductive system
with arithmetized syntax is Σ
1
, the First Incompleteness Theorem can be proved,
as we shall see, with just the assumption that there is no ω-inconsistency with a
Σ
1
-formula. Kreisel in 1957 [5] noted that the minimum case of ω-consistency, which
he labeled 1-consistency, is sufficient for the second half of G¨odel First Incomplete-
ness Theorem. This special case of ω-consistency perhaps strictly should be labeled
something like Σ
1
ω-consistency, but the label 1-consistency introduced by Kreisel
has the virtue of brevity and is standard in the literature.
While 1-consistency is both weaker and more natural than ω-consistency as a
hypothesis for proof of the second half of the First Incompleteness Theorem, we
shall see later (but only after we have proved the Second Incompleteness Theorem)
that 1-consistency is also stronger than necessary for this result.
Definition 61 (1-consistency) A system S in a language that contains a closed
term n, i.e. a numeral, for each natural number, is said to be 1-consistent if and
only if there is no Σ
1
-formula ∃v
i
F(v
i
) with one free variable in the language such
that S ⊢ ∃v
i
F(v
i
) and for each natural number n, S ⊢∼ F(n).
Theorem 78 For a Σ
0
-complete system, 1-consistency is equivalent to Σ
1
-soundness.
Proof. Let S be a Σ
0
-complete system.
(i) Suppose S is 1-consistent, let ∃v
i
F(v
i
) be a Σ
1
-sentence such that S ⊢
∃v
i
F(v
i
), and suppose ∃v
i
F(v
i
) is false. Then for each number n, ∼ F(n) is true.
Since ∃v
i
F(v
i
) is a Σ
1
-sentence, F(v
i
) is Σ
0
and since Σ
0
-formulas are closed under
negation, ∼ F(v
i
) is a Σ
0
-formula. Hence by Σ
0
-completeness of S, for each nat-
ural number n, S ⊢∼ F(n). This violates the hypothesized 1-consistency of S, so
∃v
i
F(v
i
) is true, i.e. S is Σ
1
-sound.
(ii) Suppose S is Σ
1
-sound, and suppose S ⊢ ∃v
i
F(v
i
). Then by Σ
1
-soundness of
S, there is a natural number k such that F(k) is a true Σ
0
-sentence, so ∼ F(k) is a
false Σ
0
-sentence which, by Lemma 43, is logically equivalent to a false Σ
1
-sentence.
Then by Σ
1
-soundness of S, S ∼ F(k). Hence S cannot prove a 1-inconsistency,
i.e. S is 1-consistent.
LECTURE 8 69
Note that in (ii) of the proof of Theorem 78 the hypothesis that S is Σ
0
-complete
is not needed. This part of the proof of Theorem 78 is a sharpening of the argument
for Proposition 75.
Corollary 79 Every 1-consistent system is consistent.
Proof. By the first half of Theorem 78 and the fact that a Σ
1
-sound system is
consistent (since there are sentences it doesn’t prove, namely false Σ
1
-sentences).
Or we can prove the contrapositive, as in the proof of Proposition 73, i.e. an
inconsistent theory proves everything so in particular a 1-inconsistency.
The proof of incompleteness from 1-consistency is strictly a stronger result than
proving incompleteness from ω-consistency, in that
Theorem 80 1-consistency is strictly weaker than ω-consistency.
Proof. Exercise problem 3(b) on Problem sheet 4.
However, G¨odel’s original proof made no use of the extra strength of ω-consistency
over 1-consistency, so the proof from 1-consistency is really an improvement in clar-
ity rather than in strength.
Definition 62 (2-consistency) A system S in a language that contains a closed
term n, i.e. a numeral, for each natural number, is said to be 2-consistent if and
only if there is no Σ
2
-formula ∃v
i
F(v
i
) with one free variable in the language such
that S ⊢ ∃v
i
F(v
i
) and for each natural number n, S ⊢∼ F(n).
Lemma 81 If a system is Σ
2
-sound, then it is Σ
1
-sound
Proof. The proof is by vacuous quantification. Let S ⊢ ∃v
1
F(v
1
) for F(v
1
) a
Σ
0
-sentence (i.e. no free variables). Then S ⊢ ∃v
1
∀v
2
F(v
1
), and so by Σ
2
-soundness,
∃v
1
∀v
2
F(v
1
) is true. Then ∃v
1
F(v
1
) is true.
Theorem 82 For a Σ
0
-complete system, 2-consistency is equivalent to Σ
2
-soundness.
Proof. (i) Suppose S is 2-consistent and suppose S ⊢ ∃v
1
∀v
2
F(v
1
, v
2
), where
F(v
1
, v
2
) is a Σ
0
-formula, and ∃v
1
∀v
2
F(v
1
, v
2
) is false, which is to say that for
each natural number n, ∃v
2
∼ F(n, v
2
) is a true Σ
1
-sentence. Then by the Σ
1
-
completeness of every Σ
0
-complete theory and predicate logic, for each natural num-
ber n, S ⊢∼ ∀v
2
F(n, v
2
). But then S is 2-inconsistent. So by RAA, ∃v
1
∀v
2
F(v
1
, v
2
)
is true.
(ii) Suppose S is Σ
2
-sound and suppose S ⊢ ∃v
1
∀v
2
F(v
1
, v
2
). Then ∃v
1
∀v
2
F(v
1
, v
2
)
is true, so for some number k, ∀v
2
F(k, v
2
) is true. Suppose S ⊢∼ ∀v
2
F(k, v
2
).
LECTURE 8 70
Then S ⊢ ∃v
2
∼ F(k, v
2
). Since S is Σ
2
-sound, by Lemma 81 it is Σ
1
-sound, so
∃v
2
∼ F(k, v
2
) is true. But this contradicts the truth of ∀v
2
F(k, v
2
), so by RAA,
S ∼ ∀v
2
F(k, v
2
). This means that S is 2-consistent.
Remark. We could define 3-consistency and n-consistency for larger n exactly
as for 1- and 2-consistency, but these notions are not natural in the way 1- and
2-consistency are since there is no equivalence with truth corresponding to Theo-
rems 78 and 82. Since ω-consistency implies n-consistency for each n and so in
particular 3-consistency, by Theorem ?? and Proposition ??, there is a 3-consistent
system that is not Σ
3
-sound.
8.2 Incompleteness of PA from the assumption of
1-consistency
We are now in a position to prove G¨odel’s First Incompleteness Theorem. This
theorem is significantly stronger than the version of incompleteness we have already
proved, Theorem 27, in that it proves this result from the hypothesis that PA is
1-consistent, which is strictly weaker than the hypothesis that PA is sound, i.e. that
everything PA proves is true (in the structure of the natural numbers). It is also a
purely syntactic (finitary) property as opposed to a semantic (infinitary) property.
(We replace G¨odel’s hypothesis of ω-consistency by the strictly weaker hypothesis
of 1-consistency, so the theorem is stronger, but the argument used is exactly the
same as for G¨odel’s result.)
Since we are no longer working with the notion of truth (in the structure of
the natural numbers), we cannot use the Diagonal Lemma, Theorem 17, which
establishes the truth of the diagonal equivalence. However, we use the construction
of the diagonal sentence for the one-place formula ∼ Pr
PA
(v
1
) from the proof of
the Diagonal Lemma to obtain the same G¨odel sentence for this theorem as for the
weaker one.
Theorem 83 (G¨ odel’s First Incompleteness Theorem for PA) There is a Π
1
-
sentence G such that
1. If PA is consistent, PA G, and
2. If PA is 1-consistent, PA ∼ G.
Proof. By Proposition 54, ¦n : PA ⊢ E
n
[n]¦ is Σ
1
. Let ∃v
2
A(v
1
, v
2
) be a Σ
1
-
formula that expresses ¦n : PA ⊢ E
n
[n]¦. Let a =
df
∀v
2
∼ A(v
1
, v
2
), and let
G =
df
∀v
2
∼ A(a, v
2
).
1. Suppose PA ⊢ G. By Lemma 12 and the completeness of first-order logic of
PA, PA ⊢ (∀v
2
∼ A(a, v
2
) ≡ ∀v
2
∼ A([a], v
2
)), so then PA ⊢ ∀v
2
∼ A([a], v
2
). Then
LECTURE 8 71
a ∈ ¦n : PA ⊢ E
n
[n]¦. Since ∃v
2
A(v
1
, v
2
) expresses ¦n : PA ⊢ E
n
[n]¦, ∃v
2
A(a, v
2
)
is true. Since PA is Σ
0
-complete and hence Σ
1
-complete (Proposition 58), PA ⊢
∃v
2
A(a, v
2
). This contradicts the assumption that PA is consistent. So PA G.
2. Suppose PA ⊢∼ G, i.e. PA ⊢ ∃v
2
A(a, v
2
). From the assumed 1-consistency
of PA, PA is Σ
1
-sound, by Theorem 78, so ∃v
2
A(a, v
2
) is true. Since ∃v
2
A(v
1
, v
2
)
expresses ¦n : PA ⊢ E
n
[n]¦, PA ⊢ E
a
[a], i.e. PA ⊢ G, which with the assumption
of this argument means that PA is inconsistent. But by the condition that PA is
1-consistent, PA is consistent (Corollary 79), so PA ∼ G.
8.3 Truth of the G¨odel sentence
By the bivalence of truth, one or other of G and∼ G is true in the structure of
the natural numbers. Though PA cannot decide G, i.e. does not prove G and does
not prove ∼ G, we have good reason to hold that G is true (in the structure of the
natural numbers) by the following considerations.
Lemma 84 G is true if and only if PA G.
Proof. By Proposition 54, there is a Σ
1
-formula, ∃v
2
A(v
1
, v
2
), that expresses
¦n : PA ⊢ E
n
[n]¦, i.e. PA ⊢ E
n
[n] if and only if ∃v
2
A(n, v
2
) is true. Hence for
a =
df
∀v
2
∼ A(v
i
, v
2
), and G =
df
∀v
2
∼ A(a, v
2
), PA ⊢ G if and only if ∃v
2
A(a, v
2
)
is true, which is to say that PA ⊢ G if and only if ∼ G is true. So by contraposition,
PA G if and only if G is true.
Theorem 85 If PA is consistent, G is true.
Proof. This is equivalent to the first half of Theorem 83, by Lemma 84.
Corollary 86 Theorem 85 gives a variant proof of the second half of Theorem 83,
the First Incompleteness Theorem.
Proof. Assume PA is 1-consistent. Then PA is consistent, so by Theorem 85,
G is true. On the assumption that PA is 1-consistent, PA is Σ
1
-sound. Then, since
∼ G is a false Σ
1
-sentence, PA ∼ G.
Remarks (1) By Theorem 85, we are justified in holding that the G¨odel sen-
tence for PA is true insofar as we are justified in our conviction (universal among all
mathematicians except a few with very quirky views) that PA is consistent. Identi-
fying the basis of our conviction that PA is consistent lies outside the scope of these
lectures.
(2) Formalizing the proof of Theorem 85 in PA shows that if PA is consistent
it cannot prove its own consistency, since otherwise it would prove G, which we
LECTURE 8 72
have just shown, by Theorem 83 it cannot if it is consistent. This is G¨odel’s Second
Incompleteness Theorem. Formalizing the proof of Theorem 85 in PA requires some
hard work, the hardest of which is formalizing the proof that PA is Σ
1
-complete
(provable Σ
1
-completeness), which I will do in Lecture 11.
(3) Having shown that G for PA can be seen to be true on the basis of the
accepted consistency of PA, it is important to realize that there is no weaker basis
on which to hold that G is true than that PA is consistent, i.e. G for PA cannot be
established as true on the basis of any considerations that do not also establish the
consistency of PA, by the converse of Theorem 85, which also holds.
Theorem 87 The truth of G implies the consistency of PA.
Proof. Suppose G, i.e. ∀v
2
∼ A(a, v
2
), is true. Then every number is not the
code of a proof of E
a
[a], which is to say that PA G. But if there is any sentence
that a system doesn’t prove then the system is consistent, which is to say that PA
is consistent.
There is another argument to show that G is true which is weaker than the ar-
gument for Theorem 85 because it requires a stronger hypothesis, but the argument
itself is of independent interest.
Theorem 88 If a system is Σ
0
-complete and does not prove ∃v
i
F(v
i
) for F(v
i
) a
Σ
0
-formula, then ∃v
i
F(v
i
) is false.
Proof. This theorem is the contrapositive of the implication that every Σ
0
-
complete system is Σ
1
-complete (Proposition 58), i.e. if ∃v
i
F(v
i
) is true, then for
some number k, F(k) is true and hence provable in any Σ
0
-complete system, which
then by predicate logic also proves ∃v
i
F(v
i
).
Theorem 88 tells us that G is true from the fact that PA ∼ G. However, we
established that PA ∼ G on the assumption that PA is 1-consistent, and as we
have seen, the truth of G follows from the fact that PA G, which requires only
the assumption that PA is consistent. So this argument is inefficient as a means to
establishing that G is true.
Proposition 89 For S any Σ
0
-complete, consistent, Σ
1
-axiomatized theory, the
G¨ odel sentence for S is true.
Proof. Let S be any Σ
0
-complete, consistent, arithmetized theory that has a
Σ
1
-predicate that expresses ¦n : S ⊢ E
n
¦. All of the results of the previous two
sections generalize from PA to S. In particular, for any such theory S, the G¨odel
sentence G for S is a true Π
1
-sentence and ∼ G is a false Σ
1
-sentence.
LECTURE 8 73
8.4 PA is ω-incomplete
Definition 63 (ω-completeness) A system S in a language containing a numeral
n for each natural number n is ω-complete if for every formula F(v
1
) in the language
of S such that for each natural number n S ⊢ F(n), S ⊢ ∀v
i
F(v
i
) If S is not ω-
complete we say that S is ω-incomplete.
Theorem 90 If PA is consistent, PA is ω-incomplete.
Proof. We have shown that if PA is consistent, PA G, i.e. PA ∀v
2
∼ A(a, v
2
).
We have also seen that if PA is consistent then G is true, i.e. for each natural number
n, ∼ A(a, n) is true. The sentences ∼ A(a, n) are Σ
0
. Hence by the Σ
0
-completeness
of PA, for each n, PA ⊢∼ A(a, n). Hence PA is ω-incomplete.
Lecture 9
Enumerability and the Separation
Lemma; incompleteness of PA
from the assumption of
consistency (Rosser’s Theorem);
weak and strong definability of a
function in a system; formal
provability of the Diagonal Lemma
(Tuesday, 9 November 2010)
We have noted that since an inconsistent system proves everything, consistency
of a system S is a necessary condition for S X, for any sentence X and in particular
for G the G¨odel sentence for S. We have also seen that consistency is a sufficient
condition for S G, where S is any system that can arithmetize its own syntax. The
condition of ω-consistency is, as we have seen, a stronger condition than consistency
(and a weaker condition than that every sentence provable in the system is true). We
saw that 1-consistency arises in a natural way as a condition sufficient to establish
that S ∼ G for G the G¨odel sentence for S, while consistency of S is sufficient to
establish that S G. There turns out to be a form of incompleteness, discovered by
J. Barkley Rosser (1936), that is symmetric with respect to negation. This result
is of definite interest, though it does not supersede G¨odel’s First Incompleteness
Theorem, i.e. it does not show that the G¨odel sentence for S is undecidable in S
on the assumption that S is simply consistent, and indeed to show that S ∼ G
requires a stronger hypothesis than just that S is consistent. Rosser constructed
74
LECTURE 9 75
a different sentence from the G¨odel sentence which is provably undecidable just on
the hypothesis of consistency of the system.
The Rosser Incompleteness Theorem can be proved from a separation property,
itself of independent interest and which we use also in proving that the diagonal
equivalence in the diagonal lemma is not only true, as we have seen, but also formally
provable.
9.1 Enumerability and the Separation Lemma
Definition 64 (enumeration of a relation by a formula in a theory) A k-place
relation R ⊆ N
k
is enumerated by a formula F(v
1
, . . . , v
k
, v
k+1
) in a system S if and
only if:
1. If 'n
1
, . . . , n
k
` ∈ R, then there exists a number m such that S ⊢ F(n
1
, . . . , n
k
, m).
(We say that m is a witness to the fact that 'n
1
, . . . , n
k
` ∈ R.)
2. If 'n
1
, . . . , n
k
` / ∈ R, then for every number m, S ⊢∼ F(n
1
, . . . , n
k
, m).
Definition 65 A formula F(v
1
, . . . , v
k
) separates a k-ary relation A from a k-ary
relation B in a system S if and only if for all (n
1
, . . . n
k
) ∈ A, S ⊢ F(n
1
, . . . , n
k
),
and for all (n
1
, . . . n
k
) ∈ B, S ⊢∼ F(n
1
, . . . , n
k
).
Lemma 91 (1) If F(v
1
, . . . , v
k
) separates A from B in S, then ∼ F(v
1
, . . . , v
k
)
separates B from A in S. (2) If F(v
1
, . . . , v
k
) separates A from B in S and S is
consistent, then F(v
1
, . . . , v
k
) does not separate B from A in S. (3) If F(v
1
, . . . , v
k
)
separates A from B in S and S is consistent, then A and B are disjoint. (4) If S is
inconsistent, then for any formula F(v
1
, . . . , v
k
) and any k-ary relations A and B,
F(v
1
, . . . , v
k
) separates A from B in S.
Proof. Exercise.
Theorem 92 (Separation Lemma) Let S be a system in which Ω
4
and Ω
5
hold,
and let A and B be disjoint k-ary relations enumerated in S by formulas F(v
1
, . . . , v
k
, x)
and G(v
1
, . . . , v
k
, x), respectively. Then the formula
∃x(F(v
1
, . . . , v
k
, x) ∧ (∀y ≤ x) ∼ G(v
1
, . . . , v
k
, y))
separates A from B in S.
Proof. In order to shorten formulas on the page I shall take A and B to be unary
relations. The proof for A and B as k-ary relations is just a notational variant of
this proof.
i) To show: if n ∈ A, then S ⊢ ∃x(F(n, x) ∧ (∀y ≤ x) ∼ G(n, y)).
LECTURE 9 76
(1) n ∈ A Assumption
(2) there exists k such that S ⊢ F(n, k) (1) and enumeration of A by F(v
1
, v
2
) in S
(3) n / ∈ B (1) and the hypothesis that A and B as disjoint
(4) for every m, S ⊢∼ G(n, m) (3) and enumeration of B by G(v
1
, v
2
) in S
(5) S ⊢ ((y = 0 ∨ . . . ∨ y = k) ⊃∼ G(n, y)) (4) and substitutivity of =, ∨-elim, ⊃-intro
(6) S ⊢ (y ≤ k ⊃∼ G(n, y)) (5), instance of Ω
4
and prop. logic
(7) S ⊢ ∀y(y ≤ k ⊃∼ G(n, y)) (6) by ∀-Intro
(8) S ⊢ (F(n, k) ∧ ∀y(y ≤ k ⊃∼ G(n, y))) (2) (7) ∧-Intro
(9) S ⊢ ∃x(F(n, k) ∧ ∀y(y ≤ k ⊃∼ G(n, y)))(8) ∃-intro
(10) S ⊢ ∃x(F(n, x) ∧ (∀y ≤ x) ∼ G(n, y)) (9) definition of (∀y ≤ x)
(ii) To show: if n ∈ B then S ⊢∼ ∃x(F(n, x) ∧ (∀y ≤ x) ∼ G(n, y)), which is
logically equivalent to S ⊢ ∀x(F(n, x) ⊃ (∃y ≤ x)G(n, y))
(1) n ∈ B Assumption
(2) there exists k such that S ⊢ G(n, k) (1) and enumeration of B by G(v
1
, v
2
) in S
(3) n / ∈ A (1) and the hypothesis that A and B as disjoint
(4) for every m, S ⊢∼ F(n, m) (3) and enumeration of A by F(v
1
, v
2
) in S
(5) S ⊢ ((y = 0 ∨ . . . ∨ y = k) ⊃∼ F(n, y)) (4) and substitutivity of =, ∨-elim, ⊃-intro
(6) S ⊢ (y ≤ k ⊃∼ F(n, y)) (5), instance of Ω
4
and prop. logic
(7) S ⊢ (F(n, y) ⊃∼ y ≤ k) (6) by prop logic (contraposition)
(8) S ⊢ (F(n, y) ⊃ k ≤ y) (7) by Ω
5
and prop logic
(9) S ⊢ (F(n, y) ⊃ (k ≤ y ∧ G(n, k))) (2) (8) propositional logic
(10) S ⊢ ∃y(F(n, x) ⊃ (y ≤ x ∧ G(n, y))) (9) ∃-Intro
(11) S ⊢ (F(n, x) ⊃ ∃y(y ≤ x ∧ G(n, y))) (10) predicate logic (anti-prenexing)
(12) S ⊢ ∀x(F(n, x) ⊃ ∃y(y ≤ x ∧ G(n, y))) (11) ∀-Intro
(13) S ⊢ ∀x(F(n, x) ⊃ (∃y ≤ x)G(n, y))) (12) definition of (∃y ≤ x)
Note that in the proof of the Separation Lemma, the argument for (ii) uses both

4
and Ω
5
while the argument for (i) uses just Ω
4
.
9.2 Incompleteness of PA from the assumption of
consistency (Rosser’s Theorem)
Theorem 93 If a formula H(v
1
) separates ¦n : S ⊢∼ E
n
[n]¦ from ¦n : S ⊢ E
n
[n]¦
in a consistent axiomatizable system S and S, then for h = H(v
1
), S H(h) and
S ∼ H(h).
LECTURE 9 77
Proof. (i) Suppose S ⊢ H(h). Then since S ⊢ (H(h) ≡ H[h]),
h ∈ ¦n : S ⊢ E
n
[n]¦. Then by the separation property of H(v
1
), S ⊢∼ H(h). This
contradicts the assumption S is consistent. So S H(h).
(ii) Suppose S ⊢∼ H(h). Then h ∈ ¦n : S ⊢∼ E
n
[n]¦. Then by the separation
property of H(v
1
), S ⊢ H(h). This contradicts the assumption that S is consistent.
So S ∼ H(h).
Lemma 94 If S be a consistent axiomatizable extension of R in which the formula
Pd(v
1
, v
2
) enumerates ¦n : S ⊢ E
n
[n]¦ and the formula Rd(v
1
, v
2
) enumerates
¦n : S ⊢∼ E
n
[n]¦, then the formula
∀v
1
(Pd(v
3
, v
1
) ⊃ (∃v
2
≤ v
1
)Rd(v
3
, v
2
))
separates ¦n : S ⊢∼ E
n
[n]¦ from ¦n : S ⊢ E
n
[n]¦.
Proof. By Theorem 92 and Lemma 91 (1).
Theorem 95 (Rosser’s Theorem) There is a sentence R such that if PA is con-
sistent, PA R and PA ∼ R.
Proof. By Theorem 93, Lemma 94, and the fact that the sets ¦n : PA ⊢ E
n
[n]¦
and ¦n : PA ⊢∼ E
n
[n]¦ are enumerable in PA.
9.3 Weak and strong definability of a function in
a system
Definition 66 (weak definability of a function in a system) A function f :
N
n
→ N is weakly definable in a system S iff there is a formula F(v
1
, . . . , v
n
, v
n+1
)
such that.
(1) If f(a
1
, . . . , a
n
) = b, then S ⊢ F(a
1
, . . . , a
n
, b).
(2) If f(a
1
, . . . , a
n
) = b, then S ⊢∼ F(a
1
, . . . , a
n
, b).
This notion is also sometimes called expressibility (e.g. Elliott Mendelson, In-
troduction to Mathematical Logic, 4th edn, Chapman and Hall, 1997, p. 170), or
numeralwise expressibility (e.g. Stephen Cole Kleene, Introduction to Metamathe-
matics, D. Van Nostrand, 1950, p. 195).
Definition 67 (strong definability of a function in a system) A function f :
N
n
→ N is strongly definable in a system S iff there is a formula G(v
1
, . . . , v
n
, v
n+1
)
such that if f(a
1
, . . . , a
n
) = b, then
S ⊢ (G(a
1
, . . . , a
n
, b) ∧ ∀v
n+1
(G(a
1
, . . . , a
n
, v
n+1
) ⊃ v
n+1
= b)).
LECTURE 9 78
It’s easy to show that strong definability implies weak definability (Proposi-
tion 96). The converse also holds but is considerably more complicated to prove
(Theorem 97).
Proposition 96 Let S be a system in which Ω
3
holds. If a function is strongly
definable in S then it is weakly definable in S.
Proof. If f : N
n
→ N is strongly defined in S by the formula F(v
1
, . . . , v
n
, v
n+1
),
then it is weakly defined in S by the same formula. Assume that f(a
1
, . . . , a
n
) = b.
Then by the first conjunct of the condition for strong definability, S ⊢ F(a
1
, . . . , a
n
, b),
which is condition (1) of weak definability. Suppose c = b, so f(a
1
, . . . , a
n
) = c. By
∀-elimination from the second conjunction of the condition for strong definability
in S, S ⊢ F(a
1
, . . . , a
n
, c) ⊃ c = b. If c = b, then by Ω
3
, S ⊢∼ c = b. So by
propositional logic in S, S ⊢∼ F(a
1
, . . . , a
n
, b), i.e. clause (2) of the definition of
weak definability of a function.
We will now show, just with the use of Ω
4
and Ω
5
, that any function weakly
definable in a system S is strongly definable in S. Weak definability of a function
f(x) = y in S is the condition that there is a formula F(v
1
, v
2
) that separates in S
the graph of f(x) = y from its complement. The first conjunct of the condition for
strong definability is the same as the first clause for weak definability. The second
conjunct expresses the functionality condition for a given argument of the function,
i.e. that only number that bears the defining relation to the given argument is
the value of the function for that argument. The way we do this is to define a
new formula from the formula that weakly defines the function with the additional
condition that the relationship between the argument and a number holds just in
case it’s the least number for which the weak definability relation holds.
Theorem 97 If S is an extension of ¦Ω
4
, Ω
5
¦, then any function weakly definable
in S is strongly definable in S.
Proof. To reduce clutter I give the proof for the case of a unary function, which
is also the case of immediate interest since the diagonal function is unary. The
argument for the general case is a notational variant.
Let F(x, y) be a formula that weakly defines f(x) in S, i.e.
(1) If f(a) = b, then S ⊢ F(a, b).
(2) If f(a) = b, then S ⊢∼ F(a, b).
Let G(x, y) be the formula (F(x, y)∧∀z(F(x, z) ⊃ y ≤ z)). We show that G(x, y)
strongly defines f(x) in S, i.e. if f(a) = b then S ⊢ (G(a, b) ∧ ∀y(G(a, y) ⊃ y = b)).
We show that S proves both conjuncts, on the assumption that f(a) = b.
(i) To establish the first conjunct, i.e. S ⊢ G(a, b), i.e.
S ⊢ (F(a, b) ∧ ∀z(F(a, z) ⊃ b ≤ z)):
LECTURE 9 79
(1) Since f(x) in weakly definable in S by F(v
1
, v
2
), S ⊢ F(a, b).
(2) To prove the second conjunct we establish (F(a, v
1
) ⊃ b ≤ v
1
) by ∨-
elimination from the instance of Ω
5
for b, i.e. (v
1
≤ b ∨ b ≤ v
1
).
(3) We have S ⊢ (b ≤ v
1
⊃ (F(a, v
1
) ⊃ b ≤ v
1
)), as an instance of L
3
. So it
remains to show: S ⊢ (v
1
≤ b ⊃ (F(a, v
1
) ⊃ b ≤ v
1
))).
(4) Suppose k < b. Then k = b, so by clause (2) of weak definability, S ⊢∼
F(a, k). Then by propositional logic S ⊢ (F(a, k) ⊃ b ≤ k). Then by substitutivity
of identity, S ⊢ (v
1
= k ⊃ (F(a, v
1
) ⊃ b ≤ v
1
))
(5) Suppose k = b. We know by Proposition 62 that Ω
5
⊢ b ≤ b. So by L
3
and Modus ponens, S ⊢ (F(a, b) ⊃ b ≤ b). Then by substitutivity of identity,
S ⊢ (v
1
= b ⊃ (F(a, v
1
) ⊃ b ≤ v
1
).
(6) By ∨-elimination from the cases v
1
= 0, . . . , v
1
= b established in (4) and (5),
S ⊢ ((v
1
= 0 ∨ . . . ∨ v
1
= b) ⊃ (F(a, v
1
) ⊃ b ≤ v
1
)).
(7) From (6) by Ω
4
and propositional logic, S ⊢ (v
1
≤ b ⊃ (F(a, v
1
) ⊃ b ≤ v
1
)).
(8) By ∨-elimination from Ω
5
with (3) and (7), S ⊢ (F(a, v
1
) ⊃ b ≤ v
1
).
(9) by ∀-Intro from (8), S ⊢ ∀y(F(a, y) ⊃ b ≤ y).
(ii) To establish the second conjunct, i.e. S ⊢ ∀y(G(a, y) ⊃ y = b):
(1) By ∧-elimination S ⊢ (G(a, y) ⊃ (∀z(F(a, z) ⊃ y ≤ z)).
(2) Since (∀z(F(a, z) ⊃ y ≤ z) ⊃ (F(a, b) ⊃ y ≤ b)) is logically valid, from (1) it
follows that S ⊢ (G(a, y) ⊃ (F(a, b) ⊃ y ≤ b)).
(3) By condition (1) of weak definability, S ⊢ F(a, b), so from (2), S ⊢ (G(a, y) ⊃
y ≤ b).
(4) We aim to show that S ⊢ (y ≤ b ⊃ (G(a, y) ⊃ y = b)), which with (3) implies
(ii).
(5) If k < b, then k = b, so f(a) = k, so by weak definability of f(x) by F(v
1
, v
2
),
S ⊢∼ F(a, k), so S ⊢∼ G(a, k). So by propositional logic S ⊢ (G(a, k) ⊃ k = b).
(6) From (5) by substitutivity of identity, for k < b, S ⊢ (y = k ⊃ (G(a, y) ⊃
y = b)
(7) By L
3
, S ⊢ (y = b ⊃ (G(a, y) ⊃ y = b)).
(8) By ∨-elimination from (6) and (7) and ⊃-introduction, S ⊢ ((y = 0∨. . .∨y =
b) ⊃ (G(a, y) ⊃ y = b)).
(9) From (8) by Ω
4
and propositional logic, S ⊢ (y ≤ b ⊃ (G(a, y) ⊃ y = b)).
(10) From (3) and (9) by propositional logic, S ⊢ (G(a, y) ⊃ y = b).
(11) From (10) by ∀-introduction, S ⊢ ∀y(G(a, y) ⊃ y = b).
9.4 Formal provability of the Diagonal Lemma
The Second Incompleteness Theorem is proved by proving within the system S the
first half of the First Incompleteness Theorem, “If S is consistent, then S G”,
where G is the G¨odel sentence for S.
LECTURE 9 80
We began our proof of the First Incompleteness Theorem by establishing the
Diagonal Lemma. In our original proof of the Diagonal Lemma we proved the
truth of the equivalence (C ≡ F(C)). In arithmetizing the First Incompleteness
Theorem for S in S we must show that S ⊢ (C ≡ F(C)).
We first show the following.
Lemma 98 For S any extensions of R, the diagonal function is strongly definable
in S.
Proof. (1) If the graph of a total function is Σ
1
, the complement of its graph
is also Σ
1
) (Lemma 34 in Lecture 5). (2) The diagonal function is total and its
graph is expressed by a Σ
1
-formula, ∃v
3
A(v
1
, v
2
, v
3
), i.e. d(a) = b iff ∃v
3
A(a, b, v
3
) is
true. (3) By (1) and (2), the complement of its graph is expressed by a Σ
1
-formula
∃v
3
B(v
1
, v
2
, v
3
), i.e. d(a) = b iff ∃v
3
B(a, b, v
3
) is true. (4) Suppose d(a) = b; then
by (2) ∃v
3
A(a, b, v
3
) is true, so there is a number m such that A(a, b, m) is true. (5)
By Σ
0
-completeness of S, since S is an extension of R, S ⊢ A(a, b, m). (6) Suppose
d(a) = b; then by (2), ∃v
3
A(a, b, v
3
) is false, so for every m, ∼ A(a, b, m) is true. (6)
By Σ
0
-completeness of S, since S is an extension of R, for each m, S ⊢∼ A(a, b, m).
(7) Hence the d(v
1
= v
2
is enumerated in S by A(v
1
, v
2
, v
3
). (8) Similarly, d(v
1
= v
2
is
enumerated in S by B(v
1
, v
2
, v
3
). (9) Hence by the Separation Lemma (Theorem 92,
the formula ∃x(A(v
1
, v
2
, x) ∧ (∀y ≤ x) ∼ B(v
1
, v
2
, y)) separates d(v
1
) = v
2
in S, i.e.
if d(a) = b, then S ⊢ ∃x(A(a, b, x) ∧ (∀y ≤ x) ∼ B(a, b, y)) and if d(a) = b, then
S ⊢∼ ∃x(A(a, b, x)∧(∀y ≤ x) ∼ B(a, b, y)), which is to say that d(v
1
) = v
2
is weakly
defined by ∃x(A(v
1
, v
2
, x)∧(∀y ≤ x) ∼ B(v
1
, v
2
, y)) in S. (10) Then by Theorem 97,
d(v
1
) = v
2
is strongly definable in S.
Theorem 99 (provable substitution) If a total function f(x) is strongly defin-
able in a system S, then for each formula G(v
1
), there is a formula H(v
1
) such that
for each number n, S ⊢ (H(n) ≡ G(f(n))).
Proof. Let F(v
1
, v
2
) be the formula that strongly defines f(x) in S. For given
formula G(v
1
) let H(v
1
) =
df
∃v
2
(F(v
1
, v
2
)∧G(v
2
)). Let n and m be such that f(n) =
m. We establish the provability of the two halves of the required biconditional as
follows:
(i) To show that S ⊢ (G(f(n)) ⊃ H(n)): Since F(v
1
, v
2
) strongly defines f(x), by
clause (1) S ⊢ F(n, m). Hence by propositional logic in S, S ⊢ (G(m ⊃ (F(n, m) ∧
G(m)). Then by ∃-introduction in S (note that in inferring ∃v
1
A(v
1
) from A(t), v
1
is
substituted for some but not necessarily all occurrences of t in A(t)), S ⊢ (G(m) ⊃
∃v
2
(F(n, v
2
) ∧ G(v
2
))).
(ii) To show that S ⊢ (H(n) ⊃ G(f(n))): By ∀-elimination from the second
conjunct of the condition for strong definability, S ⊢ (F(n, v
2
) ⊃ v
2
= m). Therefore
LECTURE 9 81
by propositional logic in S,
S ⊢ ((F(n, v
2
) ∧ G(v
2
)) ⊃ (v
2
= m∧ G(v
2
))). Since ((v
2
= m∧ G(v
2
)) ⊃ G(m)) is
logically valid (by substitutivity of identity), S ⊢ ((v
2
= m ∧ G(v
2
)) ⊃ G(m)), so
by propositional logic in S, ⊢ (F(n, v
2
) ∧ G(v
2
)) ⊃ G(m). Hence by ∀-introduction,
S ⊢ ∀v
2
(F(n, v
2
) ∧ G(v
2
) ⊃ G(m)), and so by anti-prenexing,
S ⊢ (∃v
2
(F(n, v
2
) ∧ G(v
2
)) ⊃ G(m))
Theorem 100 For S any extension of R, for each formula F(v
1
) in the language of
S, with v
1
its only free variable, there is a sentence C such that S ⊢ (C ≡ F(C)).
Proof. By Theorem 99, given F(v
1
), there is a formula H(v
1
) such that for
each n, S ⊢ (H(n) ≡ F(d(n))). In particular, for h = H(v
1
), S ⊢ (H(h) ≡
F(d(h))). By the construction of d(x), d(h) = H[h]. We have noted before that
for any formula F(v
1
), (F(n) ≡ F[n]) is logically valid, so S ⊢ (H(h) ≡ H[h]). So
S ⊢ (H[h] ≡ F(d(h))). Thus taking C =
df
H[h], we have S ⊢ (C ≡ F(C)), as
required.
Lecture 10
Arithemization of consistency;
provability predicates; G¨ odel’s
Second Incompleteness Theorem;
L¨ ob’s Theorem; analyzing and
strengthening the First
Incompleteness Theorem
(Wednesday, 10 November 2010)
10.1 Arithmetization of the statement that a sys-
tem S is consistent
G¨odel’s Second Incompleteness Theorem for a system S satisfying certain conditions
is the inference that if S is consistent, S cannot prove the consistency of S. Clearly
the condition that S is consistent is necessary, since if S is inconsistent it proves ev-
erything, including any sentence in the language of S that expresses the consistency
of S.
The first condition for G¨odel’s Second Incompleteness Theorem to hold for a
system is that the consistency of the system be expressible by a sentence in the
language of the system.
Definition 68 (definition of Con
S
) For Pr(v
1
) a formula that expresses ¦n : S ⊢
82
LECTURE 10 83
E
n
¦ and X any sentence in the language of S such that S X, we let Con
S
, the
formal expression in the language of S of the consistency of S be ∼ Pr(X).
The notation Con
S
does not notate relativity to the sentence X, nor relativity
to the chosen G¨odel numbering and arithmetization of syntax. I will say something
later about provable invariance of the G¨odel sentence and of Con
S
with respect to
G¨odel numbering and arithmetization of syntax. As to relativity of Con
S
to the
unprovable sentence X, we have the following
Lemma 101 (justifying the definition of Con
S
) A system S is consistent if and
only if ∼ Pr(X) is true, for Pr(v
1
) any formula in the language of S that ex-
presses ¦n : S ⊢ E
n
¦, and X any sentence in the langauge of S such that S X.
Proof. Immediate from Proposition 72 (in Lecture 8).
Remark. The relativity to X is sometimes dealt with by fixing on a particular X
such that S X if and only if S is consistent. This could, with a lot of work—
which we have now done—be the G¨odel sentence itself. But with no work, except
as required to show S ⊢∼ X (which can be very little, as in ∼ 0 = 0

, immediate
from axiom N
2
), we can take Con
S
to be ∼ Pr(X) for some particular such X,
and it is nearly standard to take Con
S
as ∼ Pr(0 = 0

).
The Second Incompleteness Theorem is established by formalizing in S the proof
of the first half of the First Incompleteness Theorem, i.e. if S is consistent, then
S G. Hence if S proves Con
S
, then S proves that S G. But since S ⊢ (G ≡∼
Pr(G), in that case S ⊢ G, which, if S is consistent, it doesn’t.
The process of formalization in S is intricate. By work of Paul Bernays and
Martin L¨ob, the requirements for formalization are reduced to three conditions on
the proof predicate for S. We shall establish the Second Incompleteness Theorem
from the assumption of these three conditions. The first of these three conditions
we have already established (for PA), the second is a problem on Problem sheet 5,
and the third we will establish in Lecture 11.
10.2 Provability predicates.
Definition 69 (provability predicate) A formula P(v
1
) is called a provability
predicate for a system S if for all sentences X and Y in L(S) the following three
conditions hold:
P
1
: If S ⊢ X, then S ⊢ P(X).
P
2
: S ⊢ (P((X ⊃ Y )) ⊃ (P(X) ⊃ P(Y )))
P
3
: S ⊢ (P(X) ⊃ P(P(X)))
LECTURE 10 84
Note that P
1
is a one-way implication and not a biconditional. The converse
implication is an instance of Σ
1
-soundness. The effect of this is that these conditions
on being a provability predicate do not require that a provability predicate expresses
¦n : S ⊢ E
n
¦. Rather, it can express a superset of that set, in particular, the formula
x = x is a provability predicate.
Being a provability predicate is not extensional, i.e. there are pairs of formulas,
even in the same class of the arithmetical hierarchy, that have the same extension
and one of them is a provability predicate and the other is not.
We established property P
1
(and its converse) for the arithmetized proof predi-
cate for PA by Theorem 28.
For Property P
2
we have
Theorem 102 For Pr(v
1
) the arithmetical proof predicate for PA constructed for
Theorem 28, and X and Y any sentences in the language of PA,
PA⊢ (Pr(X ⊃ Y ) ⊃ (Pr(X) ⊃ Pr(Y ))).
Proof. Exercise (Problem sheet 5)
Property P
3
is the arithmetization of P
1
, and follows from provable Σ
1
-completeness
of PA, which I shall prove in Lecture 11.
The following three properties of a provability predicate are immediately provable
from the three that define what it is to be a provability predicate.
Lemma 103 For P(v
1
) a provability predicate for a system S,
P
4
If S ⊢ (X ⊃ Y ), then S ⊢ (P(X) ⊃ P(Y )).
P
5
If S ⊢ (X ⊃ (Y ⊃ Z)), then S ⊢ (P(X) ⊃ (P(Y ) ⊃ P(Z))).
P
6
If S ⊢ (X ⊃ (P(X) ⊃ Y )), then S ⊢ (P(X) ⊃ P(Y )).
Proof.
P
4
: If S ⊢ (X ⊃ Y ), then by P
1
, S ⊢ P((X ⊃ Y )). Then by P
2
and Modus
ponens, S ⊢ (P(X) ⊃ P(Y )).
P
5
: By P
4
, S ⊢ (P(X) ⊃ P((Y ⊃ Z))). By P
2
, S ⊢ (P((Y ⊃ Z)) ⊃
(P(Y ) ⊃ P(Z))). Then by propositional logic, S ⊢ (P(X) ⊃ (P(Y ) ⊃
P(Z)).
[Notice that there is no corresponding result for S ⊢ ((X ⊃ Y ) ⊃ Z).]
P
6
: If S ⊢ (X ⊃ (P(X) ⊃ Y )), then by P
5
S ⊢ (P(X) ⊃ (P(P(X)) ⊃ P(Y ))). Then by L
2
and Modus ponens,
S ⊢ ((P(X) ⊃ P(P(X))) ⊃ (P(X) ⊃ P(Y ))) Then by P
3
and Modus
ponens, S ⊢ (P(X) ⊃ P(Y )).
LECTURE 10 85
10.3 G¨odel’s Second Incompleteness Theorem
Theorem 104 (G¨ odel’s Second Incompleteness Theorem) Let P(v
1
) be a prov-
ability predicate for a system S, and let G be a sentence in the language of S such
that S ⊢ (G ≡∼ P(G)). For X any sentence in the language of S, if S is consis-
tent, S ∼ P(X).
Proof.
(1) S ⊢ (G ≡∼ P(G)) Provable diagonal equivalence
(2) S ⊢ (G ⊃∼ P(G)) (1) ∧-Elim
(3) S ⊢ (G ⊃ (P(G) ⊃ X)) (2) Prop Logic
(4) S ⊢ (P(G) ⊃ P(X)) (3) P
6
(5) S ⊢ (∼ P(X) ⊃∼ P(G)) (4) contraposition
(6) S ⊢ (∼ P(G) ⊃ G) (1) ∧-Elim
(7) S ⊢ (∼ P(X) ⊃ G) (5) (6) Prop Logic
(8) S ⊢∼ P(X) Assumption
(9) S ⊢ G (7) (8) ⊃-Elim
(10) S ⊢ P(G) (9) P
1
(11) S ⊢∼ G (10)(2) Prop Logic
(12) S is inconsistent (9)(11)
(13) S ∼ P(X) (8)(12) Consistency of S RAA
Remark 1: Line (5) of this proof of the Second Incompleteness Theorem gives
a formal proof in S of the first half of the First Incompleteness Theorem for S, i.e.
if S is consistent, S G.
Remark 2: G¨odel’s Second Incompleteness Theorem is a generalization of the
first half of the First Incompleteness Theorem. The first half of the First Incom-
pleteness Theorem establishes, for G the G¨odel sentence S, that if S is consistent,
S G. By the diagonal equivalence, this is tantamount to S ∼ P(G). The Sec-
ond Incompleteness Theorem establishes that for every sentence X in the language
of S, if S is consistent, S ∼ P(X)). The heart of the matter, however, is that
the G¨odel sentence for S is provably in S equivalent to the consistency of S. One di-
rection of this equivalence is established at line (7) of the above proof of the Second
Incompleteness Theorem: If S is consistent, i.e. some sentence is unprovable, then
G holds. The converse of (7) for arbitrary X cannot be proved because it doesn’t
hold, if S is 1-consistent: If S ⊢ (G ⊃∼ P(X)) for X such that S ⊢ X, then since
by P
1
, S ⊢ P(X), S ⊢∼ G. However, if S ⊢∼ X, then indeed the implication
holds, i.e.
LECTURE 10 86
Proposition 105 For X any sentence such that S ⊢∼ X, S ⊢ (G ⊃∼ P(X)).
Proof. If S ⊢∼ X, by propositional logic S ⊢ (X ⊃ G). Then by P
4
, S ⊢
(P(X) ⊃ P(G)). By contraposition, S ⊢ (∼ P(G) ⊃∼ P(X)). Then by
the diagonal equivalence for G and transitivity of implication,
S ⊢ (G ⊃∼ P(X)).
Corollary 106 For X any sentence such that S ⊢∼ X, S ⊢ (G ≡∼ P(X)).
Proof. By Proposition 105 and line (7) of the proof of Theorem 104.
10.4 L¨ob’s Theorem
L¨ob’s Theorem is a deep result which characterizes the abstract properties of prov-
ability, i.e. it can be used as the fundamental axiom for a theory of provability,
as we shall see in Lectures 13 and 14. It is also a generalization of the Second
Incompleteness Theorem, though this was not immediately realized,. It arose in
response to an almost jokey question in the 1950s by Leon Henkin: Is the sen-
tence that asserts its own provability (there is such a sentence, by diagonalization,
i.e. S ⊢ (Pr(H) ≡ H)) provable (in which case it is true ) or unprovable (in
which case it is false)? What Martin L¨ob showed was that from just half of that
diagonal equivalence, i.e. S ⊢ (P(H) ⊃ H), it follows that S ⊢ H (so H is prov-
able and true), because for any sentence X and any provability predicate P(v
1
), if
S ⊢ (P(X) ⊃ X), then S ⊢ X. (The converse holds by propositional logic, from
L
1
and Modus ponens.)
Theorem 107 (L¨ ob’s theorem) Let S be a system in which the Diagonal Lemma
is provable, and let P(v
1
) be a provability predicate for S. For X any formula in the
language of S, if S ⊢ (P(X) ⊃ X), then S ⊢ X.
Proof. (1) Assume that S ⊢ (P(X) ⊃ X).
(2) Let L be a provably diagonal sentence for the predicate (P(v
1
) ⊃ X), i.e.
S ⊢ (L ≡ (P(L) ⊃ X).
(3) S ⊢ (L ⊃ ((P(L) ⊃ X), by ∧-Elimination from the diagonal equivalence
in (2).
(4) By P
6
from (3), S ⊢ (P(L) ⊃ P(X)).
(5) From (1) and (4) we have S ⊢ (P(L) ⊃ X)
(6) S ⊢ ((P(L) ⊃ X) ⊃ L), by ∧-Elimination from the diagonal equivalence
in (2).
(7) By Modus ponens from (5) and (6) S ⊢ L.
(8) Then from (7) by P
1
, S ⊢ P(L).
LECTURE 10 87
(9) S ⊢ X by (8) and (5).
Remark. A sentence of the form (P(X) ⊃ X) expresses the soundness with
respect to provability of X of the system for which P(v
1
) expresses provability, i.e.
it says that if X is provable, then X, i.e. X is true. L¨ob’s Theorem says that the
only such statements that can be proved in a system are the ones that hold trivially
by propositional logic, i.e. for which S ⊢ X.
Theorem 108 L¨ob’s Theorem is a generalization of G¨ odel’s Second Incompleteness
Theorem.
Proof. The Second Incompleteness Theorem proves L¨ob’s Theorem in those
cases where S ⊢∼ X, as follows: If S ⊢ ((P(X) ⊃ X)) and S ⊢∼ X, then
by propositional logic, S ⊢∼ P(X). Then by the contrapositive of the Second
Incompleteness Theorem, S is inconsistent, which is to say that S proves everything,
so in particular, S ⊢ X.
L¨ob’s Theorem goes beyond what is established by the Second Incomplete-
ness Theorem and propositional logic by establishing the implication from S ⊢
((P(X) ⊃ X)) to S ⊢ X for the case of X such that S X and S ∼ X.
Despite L¨ob’s Theorem being a generalization of the Second Incompleteness The-
orem, L¨ob’s Theorem can be proved from the Second Incompleteness Theorem,
though not uniformly but on a sentence by sentence basis. The situation is the
following:
Theorem 109 The Second Incompleteness Theorem for S ∪ ¦∼ X¦ implies that if
S ⊢ (P(X) ⊃ X), S ⊢ X.
Proof. Problem 4 on Problem sheet 5.
10.5 Analyzing and strengthening the First In-
completeness Theorem
The Second Incompleteness Theorem greatly increases our understanding of the
First Incompleteness Theorem. In discussing the truth value of the G¨odel sentence
for a system S, we noted that the truth of the G¨odel sentence for S is implied by and
implies the consistency of S. The Second Incompleteness Theorem establishes that
the equivalence of G and Con S is provable in S, so that with respect to S itself, we
see that the meaning of G is “S is consistent”. The Second Incompleteness Theorem
enables us to show that the second half of the First Incompleteness Theorem, that
S ∼ G, cannot be established just from the hypothesis that S is consistent, by
LECTURE 10 88
allowing us to establish the existence of a consistent theory S such that S ⊢∼
G
S
. It enables us also to find a non-trivial formulation of the minimum sufficient
condition for unprovability of the negation of the G¨odel sentence for system S,
namely consistency of S∪¦ConS¦, and to show that this condition is strictly weaker
than the 1-consistency of S.
10.5.1 S ∼ G
S
cannot be proved from the consistency of S
This result can be established by showing that any of the most usual examples
of consistent, 1-inconsistent theories, S ∪ ¦∼ G
S
¦ for consistent Σ
0
-complete Σ
1
-
axiomatizable theory S, proves the negation of its G¨odel sentence (not the same as
the G¨odel sentence for S)—but also (as we shall see below) not every 1-inconsistent
theory proves the negation of its G¨odel sentence.
Theorem 110 For P(v
1
) a provability predicate for a consistent theory S, let G be
a sentence in the language of S such that S ⊢ (G ≡∼ P(G)), let S

be the system
S∪¦∼ G¦, which is consistent, given the consistency of S, let P

(v
1
) be a provability
predicate for S

. Let G

be a sentence such that S

⊢ (G

≡∼ P

(G

)). Then
S

⊢∼ G

.
Proof. (1) Let X be such that S ⊢∼ X. (2) Then also (by thinning), S

⊢∼ X.
(3) S ⊢ (X ⊃ (∼ G ⊃ X)), by propositional logic, for any sentence X. Then (4)
S ⊢ (P(X) ⊃ P((∼ G ⊃ X), by property of a provability predicate, and so
by thinning, (5) S

⊢ (P(X) ⊃ P((∼ G ⊃ X). Then (6) S

⊢ (P(X) ⊃
P

(X)) since for all Y , S

⊢ (P((∼ G ⊃ Y )) ≡ P

(Y )). By the proof of the
Second Incompleteness Theorem, G is provably equivalent in S to ConS, and so also
by thinning in S

, i.e. (7) S

⊢ (G ≡∼ P(X)), and similarly (8) S

⊢ (G

≡∼
P

(X)). Therefore (9) S

⊢ (∼ G ⊃∼ G

). Hence (10) S

⊢∼ G

.
Remark. We know that PA ∼ G
PA
. Theorem 110 shows is that this fact can-
not be proved without appealing to a property of PA stronger than it’s consistency.
10.5.2 Strengthened second half of the First Incompleteness
Theorem
Theorem 111 (strengthened First Incompleteness Theorem) Let P(v
1
) be
a provability predicate for a system S, and let G be a sentence in the language
of S such that S ⊢ (G ≡∼ P(G)). Let X be any sentence such that S ⊢∼ X. Let
ConS stand for ∼ P(X). Then if S ∪ ¦ConS¦ is consistent, S ∼ G.
Proof. Suppose S ⊢∼ G. Then by Lemma ??, S ⊢∼ ConS. This contradicts
the hypothesized consistency of S ∪ ¦ConS¦. Therefore S ∼ G.
LECTURE 10 89
10.5.3 Consistency of S ∪ ¦ConS¦ is strictly weaker than 1-
consistency of S
We establish this result by constructing a theory S such that if PA∪¦ConPA¦ is
consistent, S is 1-inconsistent and S ∪ ¦ConS¦ is consistent.
Theorem 112 Let PA
+
=
df
PA ∪ ¦ConPA¦ and S =
df
PA ∪ ¦∼ ConPA
+
¦. If
PA
+
is consistent, then S is 1-inconsistent and S ∪ ¦ConS¦ is consistent.
Proof. (i) If PA
+
is consistent, then ConPA
+
is a true Π
1
-sentence, so
∼ ConPA
+
is a false Σ
1
-sentence. Then by Theorem 78, S is 1-inconsistent.
(ii) (1) Suppose S ∪ ¦ConS¦ is inconsistent.
(2) Then by propositional logic, S ⊢∼ ConS.
(3) Similarly, S is inconsistent if and only if PA ⊢ ConPA
+
, so
(4) S proves the inconsistency of S iff S ⊢ Pr
PA
(ConPA
+
), i.e. PA ∪ ¦∼
ConPA
+
¦ ⊢ Pr
PA
(ConPA
+
).
(5) Then by the Deduction Theorem, PA ⊢ (∼ ConPA
+
⊃ Pr
PA
(ConPA
+
)),
and so by the contrapositive of (5),
(6) PA ⊢ (∼ Pr
PA
(ConPA
+
) ⊃ ConPA
+
).
(7) By arithmetized Second Incompleteness Theorem, PA ⊢ (ConPA ⊃∼ Pr
PA
(ConPA)).
(8) By arithmetization of the logical fact that if PA (or any other theory) does not
prove the consistency of a theory then a fortiori it does not prove the consistency of
an extension of that theory, PA ⊢ (∼ Pr
PA
(ConPA)) ⊃∼ Pr
PA
(Con(PA ∪ ¦ConPA¦))).
(9) By (7) and (8), PA ⊢ (ConPA ⊃∼ Pr
PA
(Con(PA ∪ ¦ConPA¦))).
(10) By (6) and (9), PA ⊢ (ConPA ⊃ ConPA
+
).
(11) Then by Modus Ponens, PA∪¦ConPA¦ ⊢ ConPA
+
, i.e. PA
+
⊢ ConPA
+
.
(12) From the assumption that PA ∪ ¦ConPA¦ =
df
PA
+
is consistent, PA
+

ConPA
+
, i.e. Second Incompleteness Theorem for PA
+
.
(13) Then by (1), (11), (12) and RAA, S ∪ ¦ConS¦ is consistent.
Proposition 113 Theorem 111 is stronger than Theorem 83, and is best possible.
Proof. By Theorem 112, the hypothesis of Theorem 111 is strictly weaker
than the hypothesis of the second half of Theorem 83. It is best possible since if
S ∪ ¦ConS¦ is inconsistent, S ⊢∼ ConS. By Theorem ??, S ⊢ (∼ ConS ⊃∼ G),
so then S ⊢∼ G, i.e. if the hypothesis doesn’t hold, the result to be proved doesn’t
hold.
Remark. The hypothesis S ∪ ¦Con
S
¦ for the second half of the First Incom-
pleteness Theorem not only makes the theorem best possible but is the natural
hypothesis for this result. The two parts of the First Incompleteness Theorem for a
Σ
0
-complete theory S that is Σ
1
-axiomatizable are of different character. To show
LECTURE 10 90
that S G requires the purely formal hypothesis that S is consistent. The consis-
tency of S implies the truth of G, as noted in Theorem 85. This result holds for
every system for which a G¨odel sentence can be constructed. Consequently, whether
or not S ∼ G is equivalent to whether or not S is sound to the extent that it does
not prove this false sentence. We have seen that the first half of the First Incom-
pleteness Theorem establishes the consistency and unsoundness of S ∪ ¦∼ G¦, if S
is consistent, and we have also seen that S ∪ ¦∼ G
S
¦ ⊢∼ G
S∪{∼G
S
}
. Given that S
is consistent, Con
S
is true. Exactly the amount of soundness of S required for the
second half of the First Incompleteness Theorem is that S does not prove the false
Σ
1
-sentence ConS, i.e. S ∪ ¦Con
S
¦ is consistent. We have seen that a theory can
have this much soundness and still not be Σ
1
-sound (Theorem 112).
Lecture 11
Provable Σ
1
-completeness
(Tuesday, 16 November 2010)
Proposition 114 For Pr(v
1
) a formula in the language of PA that expresses ¦n :
PA ⊢ E
n
¦ and X any Σ
1
-sentence in the language of PA, the sentence (X ⊃
Pr(X)) is true.
Proof. (i) If X is true, then by Σ
1
-completeness of PA, PA ⊢ X, and since
Pr(v
1
) expresses ¦n : PA ⊢ E
n
¦, Pr(X) is true. So (X ⊃ Pr(X)) is true.
(ii) If X is false, then (X ⊃ Pr(X)) is true.
In this lecture we will establish that all these true sentences are provable in PA
(provable Σ
1
-completeness). Since for Y any sentence in the language of PA, the
sentence Pr(Y ) is Σ
1
, these sentences include all sentences of the form (Pr(Y ) ⊃
Pr(Pr(Y ))), i.e. P
3
, the third condition on a provability predicate.
Note that for X a Σ
1
-sentence, (X ⊃ Pr(X)) is ∆
2
. As we shall see, PA is not

2
-complete. So the provability of these sentences in PA is specific to provability
properties of Σ
1
-sentences and of arithmetization of provability in PA.
While provable Σ
1
-completeness is a deep theorem whose proof is complicated,
provable completeness for Σ
0
-sentences is very easy to show.
Proposition 115 For Pr(v
1
) a formula in the language of PA that expresses ¦n :
PA ⊢ E
n
¦ and X any Σ
0
-sentence in the language of PA, PA ⊢ (X ⊃ Pr(X)).
Proof. Argument (1): (i) If X is true, then by Σ
0
-completeness of PA, PA ⊢ X,
and since Pr(v
1
) is Σ
1
, and expresses ¦n : PA ⊢ E
n
¦, Pr(X) is a true Σ
1
-sentence.
Then by Σ
1
-completeness of PA, PA ⊢ Pr(X), so by propositional logic in PA,
PA ⊢ (X ⊃ Pr(X)).
91
LECTURE 11 92
(ii) If X is false, then ∼ X is a true Σ
0
-sentence, so by Σ
0
-completeness, PA ⊢∼
X, so by propositional logic in PA, PA ⊢ (X ⊃ Pr(X)).
Argument (2): If X is Σ
0
, then (X ⊃ Pr(X)) is Σ
1
, and hence provable in PA
by Σ
1
-completeness.
Remark. Neither of the two arguments for Proposition 115 can be extended to
the case of X a Σ
1
-sentence. For Argument (1), (i) holds for X a Σ
1
-sentence, but
(ii) fails since the negation of a Σ
1
-sentence is not, in general, a Σ
1
-sentence. For
Argument (2), we noted above that for X a Σ
1
-sentence, (X ⊃ Pr(X)) is ∆
2
,
and as we shall see later, PA is not ∆
2
-complete.
Remark about the strategy for the proof. The definition of what it is to be a
Σ
1
-formula, Definition 41, is explicit (rather than recursive), i.e. a Σ
1
formula is
any formula of the form ∃v
i
F where F is a Σ
0
-formula. —— So as in our proof
that R and thereby Q and PA are Σ
1
-complete (Propositions 58, 67, and 69, and
Theorem 70), the proof of provable Σ
1
-completeness has to go via a proof of provable
Σ
0
-completeness.
The definition of Σ
0
-formula, Definition 40, is recursive, so the proof of prov-
able Σ
0
-completeness must proceed by induction over the recursive definition of
Σ
0
-formulas. The sequence of formulas by which a Σ
0
-sentence is generated by this
recursion will in general contain free variables. Thus to have a strong enough induc-
tion hypothesis for the inductive argument, we must prove provable Σ
0
-completeness
for formulas that may contain free variables.
How to formulate this result for Σ
0
-formulas with free variables requires some
thought. It is not that PA ⊢ (X ⊃ Pr(X)) for all Σ
1
-formulas, including ones
with free variables. For example,
Proposition 116 If PA is 1-consistent, PA (v
1
+v
2
= v
3
⊃ Pr(v
1
+v
2
= v
3
)).
Proof. IF PA ⊢ (v
1
+ v
2
= v
3
⊃ Pr(v
1
+v
2
= v
3
)), then by instantiation,
PA ⊢ (1 + 2 = 3 ⊃ Pr(v
1
+v
2
= v
3
)). PA ⊢ 1 + 2 = 3, so by Modus ponens,
PA ⊢ Pr(v
1
+v
2
= v
3
). By arithmetization of the rule R
1
, i.e. Generalization,
PA ⊢ (Pr(X) ⊃ Pr(∀v
i
X)). Hence PA ⊢ Pr(∀v
1
∀v
2
∀v
3
v
1
+v
2
= v
3
). By
logic of identity in PA, PA ⊢ (∀v
1
∀v
2
∀v
3
v
1
+ v
2
= v
3
⊃ 1 + 0 = 0). Hence PA
⊢ (Pr(∀v
1
∀v
2
∀v
3
v
1
+v
2
= v
3
) ⊃ Pr(1 + 0 = 0)). Then PA ⊢ Pr(1 + 0 = 0),
which contradicts the 1-consistency of PA.
The result we need to establish is that PA proves that each true substitution
instance of a Σ
0
-formula is provable in PA. So we need to find a way to express the
condition that a substitution instance of a formula is provable.
To do this we need first to modify the definition we gave in Lecture 2 of quasi-
substitution (Definition 27), which was defined just for substitution on the free
LECTURE 11 93
variable v
1
, i.e. s(x, y) = ∀v
1
(v
1
= y ⊃ E
x
), to allow substitution on any specified
variable, i.e. s(x, y, z) = ∀v
z
(v
z
= y ⊃ E
x
)
Proposition 117 For the function s(x, y, z) = ∀v
z
(v
z
= y ⊃ E
x
) there is a
Σ
1
-formula S(x, y, z, w) in the language of PA such that for all natural numbers
n
1
, n
2
, n
3
, n
4
, S(n
1
, n
2
, n
3
, n
4
) is true if and only if s(n
1
, n
2
, n
3
) = n
4
.
Proof.
∀v
z
(v
z
= y ⊃ E
x
) =
∀v′
. . .

z
(v′
. . .

z
= y ⊃ E
x
) = 96 5 . . . 5

z
26 5 . . . 5

z
η ∗ 13
y
∗ 8x3.
The function f(z) = 5 . . . 5

z
is generated by the following primitive recursion:
f(0) = 5
f(n +1) = 5 13
n+1
+f(n), so by generalization of Theorem 39, f(x) = y is Σ
1
,
so the relation s(x, y, z) = z is Σ
1
.
Definition 70 (arithmetized proof predicate with free variables) For Pr(v
1
)
a formula in the language of PA that expresses ¦n : PA ⊢ E
n
¦ and F(v
k
1
, . . . , v
k
m
)
any formula in the language of PA with exactly the free variables shown, and k =
max¦k
1
, . . . , k
m
¦, Pr[F(v
k
1
, . . . , v
k
m
)](v
k
1
, . . . , v
k
m
) =
df
∃v
k+1
. . . ∃v
k+m
((S(F(v
k
1
, . . . , v
k
m
), v
k
1
, k
1
, v
k+1
) ∧ S(v
k+1
, v
k
2
, k
2
, v
k+2
) ∧ . . .
∧ S(v
k+m−1
, v
k
m
, k
m
, v
k+m
)) ∧ Pr(v
k+m
)).
Note that F(v
k
1
, . . . , v
k
m
) and Pr[F(v
k
1
, . . . , v
k
m
)](v
k
1
, . . . , v
k
m
) have the same
free variables.
Remark: Where a formula S(v
1
, v
2
) represents a total function s(v
1
) = v
2
in
a theory T, we can express the substitution F(s(v
1
)) either by ∀v
2
(S(v
1
, v
2
) ⊃ F(v
2
))
or by ∃v
2
(S(v
1
, v
2
)∧F(v
2
)), since (∀v
1
∃v
2
∀v
3
(S(v
1
, v
3
) ⊃ v
3
= v
2
) ⊃ (∀v
2
(S(v
1
, v
2
) ⊃
F(v
2
)) ≡ ∃v
2
(S(v
1
, v
2
) ∧F(v
2
)))) is logically valid. So strictly we could have defined
Pr[F(v
k
1
, . . . , v
k
m
)](v
k
1
, . . . , v
k
m
) as
∀v
k+1
. . . ∀v
k+m
((S(F(v
k
1
, . . . , v
k
m
), v
k
1
, k
1
, v
k+1
) ∧ S(v
k+1
, v
k
2
, k
2
, v
k+2
) ∧ . . .
∧ S(v
k+m−1
, v
k
m
, k
m
, v
k+m
)) ⊃ Pr(v
k+m
)). We do not use this latter formula as the
definition since, given that Pr(v
1
) is Σ
1
, this formula is Π
2
, whereas on the given
definition, Pr[F(v
k
1
, . . . , v
k
m
)](v
k
1
, . . . , v
k
m
) is, like Pr(v
1
), Σ
1
.
Further remark: The notation Pr[F(v
k
1
, . . . , v
k
m
)](v
k
1
, . . . , v
k
m
) stresses the
important point that it is the G¨odel number of the formula F(v
k
1
, . . . , v
k
m
), and
not the formula itself, that occurs in Pr[F(v
k
1
, . . . , v
k
m
)](v
k
1
, . . . , v
k
m
). Occa-
sionally, to avoid clutter, we will abbreviate this formula as Pr[F(v
k
1
, . . . , v
k
m
)],
which must be read bearing in mind that F(v
k
1
, . . . , v
k
m
) is not a sub-formula of
Pr[F(v
k
1
, . . . , v
k
m
)]. Note that this abbreviation cannot be used if we need to show
the result of making a substitution for a free variable of Pr[F(v
k
1
, . . . , v
k
m
)](v
k
1
, . . . , v
k
m
).
LECTURE 11 94
Proposition 118 For a formula F with no free variables,
PA ⊢ (Pr[F] ≡ Pr(F)).
Proof. For F with no free variables, consider Pr[F] with one vacuous quan-
tifier, e.g. ∀v
1
(S(F, v
1
, 0

, v
2
) ⊃ Pr(v
2
)). etc.
Proposition 119 For any natural numbers a
1
, . . . a
m
, Pr[F(v
k
1
, . . . , v
k
m
)](a
1
, . . . , a
m
)
is true if and only if Pr(F(a
1
, . . . , a
m
)) is true.
Proof. By the definition of Pr[F], and provable equivalence of F(a) and F[a].

We need to generalize P
1
and P
2
to allow for occurrence of free variables. The
generalization of P
1
expresses that if a formula with free variables is provable in PA,
then for each sentence that results from substituting numerals for the free variables
of that formula, PA proves the proof predicate for PA applied to the G¨odel number
of that sentence.
Theorem 120 (P

1
= P
1
generalized to allow free variables) For any formula
F(v
k
1
, . . . , v
k
m
) in the language of PA, if PA ⊢ F(v
k
1
, . . . , v
k
m
), then
PA ⊢ Pr[F(v
k
1
, . . . , v
k
m
)](v
k
1
, . . . , v
k
m
).
Proof. The proof is by induction on the number of free variables in F(v
k
1
, . . . , v
k
m
).
It consists of arithmetizing the following argument, which for simplicity I formu-
late for the case in which F has just one free variable.
If PA⊢ F(v
k
), then by logic in PA, PA⊢ ∀v
k
F(v
k
). Then by P
1
, PA⊢ Pr(∀v
k
F(v
k
)).
For each natural number n, (∀v
k
F(v
k
) ⊃ F[n]) is logically valid, so PA⊢ (∀v
k
F(v
k
) ⊃
F[n]). Then by P
1
, PA⊢ Pr((∀v
k
F(v
k
) ⊃ F[n])). Then by P
2
, PA⊢ (Pr(∀v
k
F(v
k
)) ⊃
Pr(F[n])). Then by Modus pones, PA ⊢ Pr(F[n]). We need to formalize this
argument in PA.
Theorem 121 (P

2
= P
2
generalized to allow free variables) For any formu-
las F(v
k
1
, . . . , v
k
m
) and G(v
r
1
, . . . , v
r
s
) in the language of PA,
PA⊢ (P[(F ⊃ G)](v
k
1
, . . . , v
k
m
, v
r
1
, . . . v
r
s
) ⊃ (P[F](v
k
1
, . . . , v
k
m
) ⊃ P[G](v
r
1
, . . . v
r
s
))).
Proof. Exercise.
Lemma 122 For all formulas F(v
j
1
, . . . , v
j
m
) and G(v
k
1
, . . . , v
k
n
), PA ⊢ (P[F](v
j
1
, . . . , v
j
m
) ⊃
(P[G](v
k
1
, . . . , v
k
n
) ⊃ P[(F ∧ G)])(v
j
1
, . . . , v
j
m
, v
k
1
, . . . , v
k
n
))
Proof Exercise.
LECTURE 11 95
Definition 71 A term t is free for variable v
i
in formula F(v
i
) if v
i
in F(v
i
) does
not occur within the scope of a quantifier whose variable of quantification is a free
variable in t.
Definition 72 For F(v
i
) a formula with free variable v
i
and t any term, F(v
i
/t)
is the result of substituting the term t for all occurrences of the variable v
i
in the
formula F(v
i
).
Lemma 123 Let F(v
i
) be a formula with free-variable v
i
. Let v
j
be a variable and
t a term both free for v
i
in F(v
i
). Then the following equivalence is logically valid:
(Pr[F(v
i
)](t) ≡ Pr[F(v
j
)](t)),
and correspondingly for F with any number of variables.
Proof. From the definition of Pr[F(v
i
)](v
i
), substitutivity of =, and logical
equivalence of F(v
i
) and F[v
i
].
Lemma 124 Let F(v
i
) be a formula with free-variable v
i
and let t(v
j
) be a term
free for v
i
in F(v
i
with variable v
j
distinct from v
i
. Then the following equivalence
is logically valid:
(Pr[F(v
i
/t(v
j
))](v
j
) ≡ Pr[F(v
i
)](t(v
j
))),
and correspondingly for F with any number of variables.
Proof. From the definition of Pr[F(v
i
)](v
i
), substitutivity of =, and logical
equivalence of F(v
i
) and F[v
i
].
Our proof in Lecture 7 that PA is Σ
0
-complete went by way of proving the very
strong result that the extremely weak system R is Σ
o
-complete and then showing
that R is a subsystem of PA. Proving Σ
0
-completeness requires heavy use of math-
ematical induction, so that proof cannot be formalized in R or Q. It could be
formalized in PA, but that would be a very roundabout way of proving the provable
Σ
0
-completeness of PA in PA. Rather what we want to do is formalize in PA a di-
rect proof of the Σ
0
-completeness of PA. I will illustrate an informal direct proof of
the Σ
0
-completeness of PA that we will be formalizing in PA, by going through an
informal proof for the case of an atomic Σ
0
-formula of the form v
1
+v
2
= v
3
.
Lemma 125 If a +b = c, then PA ⊢ a +b = c.
LECTURE 11 96
Proof. We argue by (informal) induction on the free variable b in the statement,
for all c, if a+b = c, then PA ⊢ a+b = c. (The universal quantifier on the variable c
is to strengthen the induction hypothesis.) The result then follows by ∀-Elimination
on the quantifier ‘for all c’.
b = 0: If a + 0 = c, then a = c. From Axiom N
3
by ∀-Intro and ∀-Elim, PA
⊢ a + 0 = a. If a = c, then PA ⊢ a = c. By the logic of identity in PA, PA
⊢ ((a = c ∧ a + 0 = a) ⊃ a + b = c). Hence if a + 0 = c, PA ⊢ a + b = c. This is
proved outright, in particular with no assumption in which the variable c is free, so
for all c, if a + 0 = c, PA ⊢ a +b = c.
Induction step. Suppose that for all c, if a + b = c, then PA ⊢ a + b = c, and
suppose a + b

= c. Then by the informal version of N
4
, (a + b)

= c, so there is a
number d such that d

= c. By logic of identity, (a +b)

= d

. Then by the informal
version of N
1
, a + b = d. Instantiating the quantifier ‘forall c’ in the induction
hypothesis with d, we have that PA ⊢ a + b = d. Then by logic of identity in PA,
PA ⊢ (a + b)

= d

. From the fact that d

= c, PA ⊢ d

= c. We know that d

and
d

are the same expression, so by logic of identity in PA, PA ⊢ (a + b)

= c. Then
by N
4
and logic of identity, PA ⊢ a + b

= c. So from the Induction Hypothesis we
have proved that if a + b

= c, PA ⊢ a + b

= c. Since c is not free in the Induction
Hypothesis, it follows that for all c, if a +b

= c, PA ⊢ a +b

= c.
We now turn to proof of the main theorem.
Theorem 126 (provable Σ
0
-completeness with free variables) For each Σ
0
-
formula F(v
k
1
, . . . , v
k
m
), PA ⊢ (F(v
k
1
, . . . , v
k
m
) ⊃ Pr[F(v
k
1
, . . . , v
k
m
)](v
k
1
, . . . , v
k
m
)).
Proof. By induction over the inductive definition of Σ
0
-formulas.
Base case:
F is an atomic formula, i.e. a formula of the form t
1
= t
2
or t
1
≤ t
2
for t
1
, t
2
terms. The proof of this case is by a double induction over the recursive definition
of terms. We will prove one of these cases. The others are similar.
PA ⊢ (v
1
+v
2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
3
)), i.e. PA ⊢ (v
1
+v
2
= v
3

∃v
4
∃v
5
∃v
6
(S(v
1
+v
2
= v
3
, v
1
, 0

, v
4
) ∧ S(v
4
, v
2
, 0
′′
, v
5
) ∧ S(v
5
, v
3
, 0
′′′
, v
6
) ∧ Pr(v
6
))
The following is an informal description of a formal proof within PA. The proof
is by induction on the variable v
2
, but rather than argue by induction on v
2
in the
formula (v
1
+v
2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
3
)), we argue by induction on v
2
in the formula ∀v
3
(v
1
+v
2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v
3
, v
2
)), in order to have a
stronger Induction Hypothesis.
Base case: v
2
= 0. We need to show that PA ⊢ ∀v
3
(v
1
+0 = v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, 0, v
3
)).
(1) (1) v
1
+ 0 = v
3
Assumption
(2) v
1
+ 0 = v
1
N
3
LECTURE 11 97
(1) (3) v
1
= v
3
(1) (2) substitutivity of =
(4) Pr[v
1
+ 0 = v
1
](v
1
) (2) P

1
(5) (Pr[v
1
+ 0 = v
1
](v
1
) ⊃ Pr[v
1
+v
2
= v
3
](v
1
, 0, v
1
)) Lemma 124
(6) Pr[v
1
+v
2
= v
3
](v
1
, 0, v
1
) (4) (5) ⊃-elimination
(1) (7) Pr[v
1
+v
2
= v
3
](v
1
, 0, v
3
) (6) (3) substitutivity of =
(8) (v
1
+ 0 = v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, 0, v
3
)) (1)(7) ⊃-intro
(9) ∀v
3
(v
1
+ 0 = v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, 0, v
3
)) (8) ∀-intro
Induction step:
We have to show, within PA, that there exists a derivation of
(∀v
3
(v
1
+v
2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
3
)) ⊃ ∀v
3
(v
1
+v

2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v

2
, v
3
))).
(1) (1) ∀v
3
(v
1
+v
2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
3
) Assumption I.H.
(2) (2) v
1
+v

2
= v
3
Assumption
(3) (v
1
+v

2
) = (v
1
+v
2
)

N
4
(2) (4) (v
1
+v
2
)

= v
3
(2)(3) logic of =
(2) (5) ∃v
4
(v

4
= v
3
) (4) ∃-Intro
(6) (6) v

4
= v
3
Assumption
(6)(2) (7) (v
1
+v
2
)

= v

4
(4)(6) subst of =
(6)(2) (8) v
1
+v
2
= v
4
(7) N
1
, logic
(1) (9) (v
1
+v
2
= v
4
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
4
)) (1) ∀-elimin
(1)(2)(6) (10) Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
4
) (8)(9) ⊃-elim
(11) ((v
1
+v
2
= v
4
∧ v
1
+v

2
= (v
1
+v
2
)

) ⊃ v
1
+v

2
= v

4
)
substitutivity of =
(12) Pr[((v
1
+v
2
= v
4
∧ v
1
+v

2
= (v
1
+v
2
)

) ⊃ v
1
+v

2
= v

4
)](v
1
, v
2
, v
4
)
(11) P

1
(13) ((Pr[v
1
+v
2
= v
4
](v
1
, v
2
, v
4
) ∧ Pr[v
1
+v

2
= (v
1
+v
2
)

](v
1
, v
2
)) ⊃
Pr[v
1
+v

2
= v

4
](v
1
, v
2
, v
4
)) (12) P

2
(14) (Pr[v
1
+v
2
= v
4
](v
1
, v
2
, v
4
) ≡ Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
4
))
Lemma 123
(15) (Pr[v
1
+v

2
= v

4
](v
1
, v
2
, v
4
) ≡ Pr[v
1
+v
2
= v
3
](v
1
, v

2
, v

4
))
Lemma 124
(16) ((Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
4
) ∧ Pr[v
1
+v

2
= (v
1
+v
2
)

](v
1
, v
2
)) ⊃
Pr[v
1
+v
2
= v
3
](v
1
, v

2
, v

4
)) (13)(14)(15) propositional logic
(17) Pr[v
1
+v

2
= (v
1
+v
2
)

](v
1
, v
2
) (3) P

1
(1)(2)(6) (18) Pr[v
1
+v
2
= v
3
](v
1
, v

2
, v

4
) (16)(10)(17) ⊃-elim
(1)(2)(6) (19) Pr[v
1
+v
2
= v
3
](v
1
, v

2
, v
3
) (18)(6) subst =
(1)(2) (20) Pr[v
1
+v
2
= v
3
](v
1
, v

2
, v
3
) (6)(19) ∃-elimination
(1) (21) (v
1
+v

2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v

2
, v
3
)) (2)(20) ⊃-intro
LECTURE 11 98
(1) (22) ∀v
3
(v
1
+v

2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v

2
, v
3
)) (21) ∀-intro (v
3
not free in (1))
(23) (∀v
3
(v
1
+v
2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
3
)) ⊃
∀v
3
(v
1
+v

2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v

2
, v
3
))) (1)(22) ⊃-intro
By the instance of N
12
for induction on v
2
in the formula ∀v
3
(v
1
+ v
2
= v
3

Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
3
)), we have proved in PA, ∀v
3
(v
1
+v
2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
3
)).
Then by one step of ∀-elimination, we have (v
1
+v
2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
3
)),
which was to be proved.
We now turn to the induction steps of the proof.
(i) F is ∼ G where G is Σ
0
.
Exercise.
(ii) F is (G∧ H), for G and H both Σ
0
-formulas.
(1) (G ⊃ Pr[G]) Induction hypothesis
(2) (H ⊃ Pr[H]) Induction hypothesis
(3) (G ⊃ (H ⊃ (G∧ H))) propositional logic
(4) Pr[(G ⊃ (H ⊃ (G∧ H)))] (3) Theorem 120
(5)(Pr[(G ⊃ (H ⊃ (G∧ H)))] ⊃ (Pr[G] ⊃ Pr[(H ⊃ (G∧ H))])) Theorem 121
(6) (Pr[(H ⊃ (G∧ H))] ⊃ (Pr[H] ⊃ Pr[(G∧ H)])) Theorem 121
(7) (Pr[G] ⊃ (Pr[H] ⊃ Pr[(G∧ H)])) (4) (5) (6) propositional logic
(8) ((G∧ H) ⊃ Pr[(G∧ H)]) (1) (2) (7) propositional logic
(iii) F is (∀v
1
≤ v
2
)G(v
1
), for G(v
1
) a Σ
0
-formula. This means that v
2
is free in
F. To simplify notation we shall take it that no other variables are free in F, i.e. v
1
is the only free variable in G(v
1
). We need to give a proof in PA of (F ⊃ Pr[F]),
i.e.
((∀v
1
≤ v
2
)G(v
1
) ⊃ ∀v
3
(S((∀v
1
≤ v
2
)G(v
1
), v
2
, 0
′′
, v
3
) ⊃ P(v
3
)).
The proof is by induction on the variable v
2
occurring free in the formula (F ⊃
Pr[F]).
Base case: We need to prove that ((∀v
1
≤ 0)G(v
1
) ⊃ ∀v
3
(S((∀v
1
≤ 0)G(v
1
), 0, 0
′′
, v
3
) ⊃
P(v
3
))
To simplify notation I shall write v
1
as x and v
2
as y.
(1) (G(x) ⊃ Pr[G(x)](x)) Induction hypothesis for the main induction
(2) (G(x) ⊃ Pr[G(x)](x))(0) (1) ∀-Intro, ∀-Elim
(3) (G(0) ⊃ Pr[G(x)(x)](0)) (2) defn of subst
LECTURE 11 99
(4) (Pr[G(x)](0) ≡ Pr[G(0)]) Lemma
(5) (G(0) ⊃ Pr[G(0)]) (3) (4) propositional logic
(6) (∀x ≤ 0)G(x) ≡ G(0)) provable in PA
(7) ((∀x ≤ 0)G(x) ⊃ Pr[(∀x ≤ 0)G(x)]) (5) (6) logical equivalences
Induction step: On the assumption ((∀x ≤ y)G(x) ⊃ Pr[(∀x ≤ y)G(x)](y)),
which is the induction hypothesis for this sub-induction, we need to establish
((∀x ≤ y

)G(x) ⊃ Pr[(∀x ≤ y

)G(x)](y

)).
(1) (1) ((∀x ≤ y)G(x) ⊃ Pr[(∀x ≤ y)G(x)](y)) Assumption IH for sub-induction
(2) (2) (∀x ≤ y

)G(x) Assumption
(2) (3) ∀x(x ≤ y

⊃ G(x)) (2) defn of (∀x ≤ y

)
(2) (4) ∀x((x ≤ y ∨ x = y

) ⊃ G(x)) (3) N
6
(2) (5) ∀x((x ≤ y ⊃ G(x)) ∧ (x = y

⊃ G(x)) (4) propositional logic
(2) (6) (∀x(x ≤ y ⊃ G(x)) ∧ ∀x(x = y

⊃ G(x))) (5) predicate logic
(2) (7) ((∀x ≤ y)G(x) ∧ G(y

)) (6) defn of (∀x ≤ y), logic
(8) (8) (G(x) ⊃ Pr[G(x)](x)) Assumption IH for main induction
(8) (9) (G(y

) ⊃ Pr[G(x)](y

)) (8) ∀-Intro, ∀-Elim
(1)(2)(8) (10) (Pr[(∀x ≤ y)G(x)](y) ∧ Pr[G(x)](y

)) (7)(1)(9) propositional logic
(1)(2)(8) (11) Pr[((∀x ≤ y)G(x) ∧ [G(y)])](x, y

) (10) Lemma 122
(1)(2)(8) (12) Pr[(∀x ≤ y

)G(x)] (11) N
8
and logic
(1)(8) (13) ((∀x ≤ y

)G(x) ⊃ Pr[(∀x ≤ y

)G(x)]) (2) (12) ⊃-Intro
Step (9) in the above derivation calls for comment. In deriving (9) from (8)
∀-Intro is applied to (8) and the status of (8), as an induction hypothesis, is that of
assumption. If the variable in (8) were the variable of induction this ∀-Intro would
be illegitimate. However, the induction of which (8) is an induction hypothesis is
over formulas, so the assumption is about the formula G(x), and not about x, i.e.
x is a free variable in (8) to which ∀-Intro may be applied.
Corollary 127 (provable Σ
1
-completeness with free variables) For each Σ
1
-
formula F(v
k
1
, . . . , v
k
m
), PA ⊢ (F(v
k
1
, . . . , v
k
m
) ⊃ Pr[F(v
k
1
, . . . , v
k
m
)](v
k
1
, . . . , v
k
m
).
Proof. We abbreviate v
k
1
, . . . , v
k
m
as v, to avoid clutter. That F(v) is a Σ
1
-
formula means there is a Σ
0
-formula G(v, v
i
) such that F(v) = ∃v
i
G(v, v
i
). The
following describes a derivation in PA.
(1) (G(v, v
i
) ⊃ Pr[G(v, v
i
)](v, v
i
) Theorem 126
(2) (G(v, v
i
) ⊃ ∃v
i
G(v, v
i
)) predicate logic
LECTURE 11 100
(3) (Pr[G(v, v
i
)](v, v
i
) ⊃ Pr[∃v
i
G(v, v
i
)](v)) ( 2) P

1
, P

2
(4) (G(v, v
i
) ⊃ Pr[∃v
i
G(v, v
i
)](v) (1) (3) propositional logic
(5) ∀v
i
(G(v, v
i
) ⊃ Pr[∃v
i
G(v, v
i
)](v) (4) ∀-Intro
(6) (∃v
i
G(v, v
i
) ⊃ Pr[∃v
i
G(v, v
i
)](v) (5) anti-prenexing
Lecture 12
The ω-rule and uniform reflection;
PA proves that PA proves every
instance of the G¨ odel sentence;
Π
1
-uniform reflection and
consistency; PA is Π
1
-conservative
over PA
Π
2
∪ ¦Con
PA
¦
(Wednesday, 17 November 2010)
12.1 The ω-rule and uniform reflection
We noted in Section 8.4 that the first half of the G¨odel Incompleteness Theorem for
a Σ
0
-complete system S establishes that if S is consistent then it is ω-incomplete,
i.e. we have, for each n, S ⊢∼ Prov(G, n), but S ∀v
2
∼ Prov(G, v
2
). If each
numerical instance of a formula F(v
i
) with one free variable is true, then ∀v
i
F(v
i
)
is true. Hence the following inference, called the ω-rule, is sound with respect to
truth in arithmetic:
F(0), F(1), . . . , F(n) . . .
∀yF(y)
Definition 73 PA
ω
=
df
PA + ω-rule.
101
LECTURE 12 102
Definition 74 PA
ω
⊢ X if and only if there is a tree of finite height with formulas
in the language of PA at each node and with X at the bottom node, such that for
each formula at a node without predecessor is a theorem of PA, and such that each
node is either a theorem of PA or there is one node directly above the node at which
X occurs such that for Y the formula at the predecessor node, PA ⊢ (Y ⊃ X), or
X is the result of an application of the ω-rule. By a tree consisting of a single node
(height 0), if PA ⊢ X, then PA
ω
⊢ X.
For G the G¨odel sentence for S, PA
ω
⊢ G. A derivation using the ω-rule has
infinitely many premisses and hence is an infinite object, unlike a formal proof, or
our usual idea of a informal proof, so derivation of G by the ω-rule cannot be said
to constitute a proof of G. Indeed derivability from the axioms of PA by the ω-rule
is tantamount to truth.
Proposition 128 If a sentence X in the language of arithemtic is true, PA
ω
⊢ X.
Proof. We argue by induction over the arithmetical hierarchy. We know by
Lecture 7 that if X is a true Σ
0
(= Π
0
) or Σ
1
-sentence, PA ⊢ X, and hence PA
ω
⊢ X.
Suppose X is a true Π
1
-sentence, i.e. of the form ∀v
1
F(v
1
) for F(v
1
) a Σ
0
-
formula. Then for each natural number n, PA ⊢ F(n), so by one application of the
ω-rule, PA
ω
⊢ ∀v
1
F(v
1
). Assume for Induction Hypothesis that the result holds
for Σ
n
and Π
n
-sentences. (i) Let X be a true Σ
n+1
-sentence ∃v
i
F(v
i
), where F(v
i
)
is a Π
n
formula. Then for some natural number m, F(m) is a true Π
n
-sentence.
Then by Induction Hypothesis PA
ω
⊢ F(m). Since PA ⊢ (F(m) ⊃ ∃v
i
F(v
i
)),
PA
ω
⊢ ∃v
i
F(v
i
). (ii) Let X be a true Π
n+1
-sentence ∀v
i
F(v
i
). Then for each
number n, F(n) is a true Σ
n
-sentence. Then by Induction Hypothesis, for each n,
PA
ω
⊢ F(n). Then by one application of the ω-rule, PA
ω
⊢ ∀v
i
F(v
i
).
The completeness of PA
ω
with respect to truth is entirely to do with the ω-rule,
and essentially nothing to do with the axioms for arithmetic of PA, as shown by the
fact that the corresponding system for R, i.e. R
ω
=
df
R + ω-rule, is also complete
with respect to truth.
Proposition 129 If a sentence X in the language of arithmetic is true, R
ω
⊢ X.
Proof. The only facts about PA used in the proof of Proposition 128 are that
PA is Σ
0
-complete and that PA ⊢ (F(m) ⊃ ∃v
i
F(v
i
)). Both these properties are
also facts about R.
There are constructive versions of the ω-rule, based on the fact that we can
state in a single sentence that all numerical instances of a given formula F(v
1
) are
provable, and by the arithmetization of syntax such single sentences can be expressed
LECTURE 12 103
in the language of arithmetic, namely as ∀v
1
Pr[F(v
1
)](v
1
), where Pr[F(v
1
)](v
1
)
is defined by Definition 70 in Lecture 11. We can then give finite expression to an
ω-rule by the sentence:
(∀v
1
Pr[F(v
1
)](v
1
) ⊃ ∀v
1
F(v
1
)).
Such sentences are highly sensitive to the axiomatic strength of the system to
which they are applied. They are called Reflection Principles (see Smorynski [6], p.
845). We use the following terminology:
Definition 75 For F(v
1
) a formula in the language of PA with one free variable, a
Uniform Reflection Principle is any sentence of the form
(∀v
1
Pr[F(v
1
)](v
1
) ⊃ ∀v
1
F(v
1
)).
PA extended by Uniform Reflection Principles for all one-place formulas is strictly
weaker than PA extended by the infinitary ω-rule. This is because PA + ω-rule =
true arithmetic, while PA + all Uniform Reflection Principles is axiomatic and hence
incomplete.
12.2 PA proves that PA proves every instance of
the G¨odel sentence
Theorem 130 PA⊢ ∀v
1
Pr[∼ Prov(G, v
1
)](v
1
)
Proof. The following derivation shows the existence of a formal proof in PA.
(1) (1) ∼ Pr[∼ Prov(G, v
1
)](v
1
) Assumption
(2) (0 = 0

⊃∼ Prov(G, v
1
)) PA ⊢∼ 0 = 0

(3) (Pr[0 = 0

] ⊃ Pr[∼ Prov(G, v
1
)]) (2) P

1
, P

2
, MP
(4) (∼ Pr[∼ Prov(G, v
1
)] ⊃∼ Pr[0 = 0

]) (3) contraposition
(1) (5) ∼ Pr[0 = 0

] (1) (4) MP
(6) (∼ Pr[0 = 0

] ⊃∼ ∃v
1
Prov(G, v
1
)) line (5) of proof of Theorem 104
and Proposition 118
(1) (7) ∼ ∃v
1
Prov(G, v
1
)) (5) (6) MP
(1) (8) ∀v
1
∼ Prov(G, v
1
)) (7) pred logic
(1) (9) ∼ Prov(G, v
1
)) (8) ∀-elim
(10) (∼ Prov(G, v
1
) ⊃ Pr[∼ Prov(G, v
1
)]) Theorem 126
(1) (11) Pr[∼ Prov(G, v
1
)] (9) (10) MP
(12) Pr[∼ Prov(G, v
1
)] (1)(11) reductio ad absurdum
(13) ∀v
1
Pr[∼ Prov(G, v
1
)] (12) ∀-Intro
LECTURE 12 104
Corollary 131 PA ∪¦(∀v
1
Pr[∼ Prov(G, v
1
)](v
1
) ⊃ ∀v
1
∼ Prov(G, v
1
))¦
⊢ G
Proof. By the Theorem, Modus ponens, and the fact that
PA ⊢ (G ≡ ∀v
1
∼ Prov(G, v
1
)).
12.3 Equivalence of Π
1
-Uniform Reflection and con-
sistency
Π
1
-Uniform Reflection for PA, i.e. the Uniform Reflection Principle restricted to Π
1
-
formulas, is provably equivalent in PA to Con
PA
, the formal consistency statement
for PA, ∼ Pr
PA
(0 = 0

).
Theorem 132 PA + Π
1
-uniform reflection ⊢∼ Pr
PA
(0 = 0

).
Proof. Take the instance of Π
1
-uniform reflection for the Π
1
-formula
∀v
1
(0 = 0

∧ v
1
= v
1
). This is provably equivalent to (Pr(0 = 0

) ⊃ 0 = 0

).
PA ∪¦(Pr(0 = 0

) ⊃ 0 = 0

)¦ ⊢∼ Pr(0 = 0

).
Theorem 133 For F(v
1
) any Π
1
-formula with one free variable,
PA ∪¦Con
PA
¦ ⊢ (∀v
1
Pr[F(v
1
)](v
1
) ⊃ ∀v
1
F(v
1
)).
Proof.
(1) (1) ∀v
1
Pr[F(v
1
)](v
1
) Assumption
(2) (2) ∼ F(v
1
) Assumption
(3) (∼ F(v
1
) ⊃ Pr[∼ F(v
1
)](v
1
) Corollary 127 since
∼ F(v
1
) is Σ
1
(2) (4) Pr[∼ F(v
1
)](v
1
) (2)(3) MP
(1) (5) Pr[F(v
1
)](v
1
) (1) ∀-elim
(1)(2) (6) Pr[(F(v
1
)∧ ∼ F(v
1
))](v
1
) Lemma 122
(7) ((F(v
1
)∧ ∼ F(v
1
)) ⊃ 0 = 0

) propositional logic
(8) Pr[(F(v
1
)∧ ∼ F(v
1
)) ⊃ 0 = 0

](v
1
) (7) P

1
(9) Pr[(F(v
1
)∧ ∼ F(v
1
))](v
1
) ⊃ Pr[0 = 0

] (8) P

2
(10) ∼ Pr[0 = 0

] ⊃∼ Pr[((F(v
1
)∧ ∼ F(v
1
))](v
1
) (9) contraposition
(11) ∼ Pr(0 = 0

) given
(12) ∼ Pr[0 = 0

] (11) Proposition 118
(13) ∼ Pr[((F(v
1
)∧ ∼ F(v
1
))](v
1
) (12) (10) ⊃-elimination
(1)(2) (14) ((6) ∧ (13)) (6) (13) ∧-intro
(1) (15) F(v
1
) (2) (6) RAA
LECTURE 12 105
(1) (16) ∀v
1
F(v
1
) (15) ∀-intro (v
1
not free in (1)
(17) (∀v
1
Pr[F(v
1
)](v
1
) ⊃ ∀v
i
F(v
1
) (1) (16) ⊃-intro
This proof covers the case where F(v
1
) is Σ
0
rather than Π
0
, in which case the
justification at line (3) is Theorem 126 rather than Corollary 127.
12.4 PA is Π
1
-conservative over PA
Π
2
∪ ¦Con
PA
¦
In the 1920s David Hilbert adumbrated a research programme which had at its
heart the project of giving proofs of the consistency of formal systems of infini-
tary mathematics in finitary mathematics. The motivation for this programme was
foundational and philosophical but Hilbert came to see that it also promised math-
ematical application in terms of establishing “conservative extension” results.
A theory S
2
is an extension of S
1
if S
2
proves everything that S
1
proves, and
S
2
is a conservative extension of S
1
if for every formula X in the language of S
1
, if
S
2
⊢ X, then S
1
⊢ X, i.e. S
2
proves nothing in the language of S
1
that S
1
doesn’t
already prove. More precisely and more generally,
Definition 76 For theories S
1
and S
2
formulated in the same language (or such that
the language of S
2
is an extension of the language of S
1
or such that the language
of S
1
can interpreted in the language of S
2
, but we shall not be concerned with these
more general cases), S
2
is an extension of S
1
if for each formula X in the language
of S
1
, if S
1
⊢ X, then S
2
⊢ X (or S
2
⊢ X

where X

is the translation of X into
the language of S
2
).
Definition 77 An extension S
2
of S
1
is conservative over S
1
with respect to a class
of formulas Γ if for each formula X in Γ, if S
2
⊢ X, then S
1
⊢ X.
A fundamental insight of Hilbert’s that lies at the heart of his programme of
proof theory is that if finitary mathematics can prove the consistency of infinitary
mathematics, then infinitary mathematics is a conservative extension of finitary
mathematics with respect to finitary mathematics. Hilbert sketches an argument
for this claim in [4], p. 474.
G¨odel’s Second Incompleteness Theorem shows that, insofar as infinitary math-
ematics is an extension of finitary mathematics, the consistency of infinitary math-
ematics cannot be proved within finitary mathematics. Nonetheless Hilbert’s ar-
gument adumbrates a correct mathematical theorem the main content of which is
the proof of Theorem 133 that consistency implies the arithmetized ω-rule (a.k.a.
the uniform reflection principle) for Π
1
-sentences. Hilbert formulates his argument
in terms of a particular Π
1
-sentence, Fermat’s last theorem: “Let us suppose, for
LECTURE 12 106
example, that we had found, for Fermat’s great theorem, a proof in which the [in-
finitary] logical function ǫ was used. We could then make a finitary proof out of it
in the following way.”
Leaving aside the question, in what minimal system can the consistency of PA
be proved, which is beyond the scope of this course (the answer is, very roughly,
constructive principles of abstract mathematics, rather than finitary principles of
concrete mathematics), a precise working out of the argument Hilbert sketched
requires that the proof of Theorem 133 be carried out in finitary mathematics.
Hilbert never formulated clearly what he meant by finitary mathematics, i.e. he
never gave a formal system of finitary mathematics, and I won’t enter here the
debate over what formal system should be taken to capture the intended notion of
finitary arithmetic. Rather, I will address the question, in how weak a subsystem of
PA can the argument for Theorem 133 be carried out?
The key point is that such a system must be strong enough to prove provable Σ
0
-
completeness of PA. Hilbert did not explicitly formulate provable Σ
0
-completeness,
but it is implicit in his argument, and implicitly he takes it to be a fact of finitary
mathematics. “Let us assume that numerals p, a, b, c (p > 2) satisfying Fermat’s
equation a
p
+b
p
= c
p
are given; then we could also obtain this equation as a provable
formula by giving the form of a proof to the procedure by which we ascertain that
the numbers a
p
+b
p
and c
p
coincide.”
Proving provable Σ
0
-completeness requires mathematical induction, so we may
take the question to be, how much induction, measured by complexity in the arith-
metical hierarchy of formulas to which induction must be applied, is needed for this
proof?
The minimum is Σ
1
-induction, i.e. axioms N
12
for Σ
1
-formulas, as in the step of
the proof in Lecture 11 in which we proved that
((∀v
1
≤ v
2
)G ⊃ ∀v
3
(S((∀v
1
≤ v
2
)G, v
2
, 0
′′
, v
3
) ⊃ P(v
3
)) by induction on the free
variable v
2
. However, we also used Π
2
-induction, in our proof in PA that (v
1
+
v
2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
3
)), since we proved this by induction on the
formula ∀v
3
(v
1
+ v
2
= v
3
⊃ Pr[v
1
+v
2
= v
3
](v
1
, v
2
, v
3
)). It might be that there
is some clever way to reconstruct that proof so that the universal quantification of
the induction formula is not needed. In any case, from what has been established,
we have the following theorem corresponding to Hilbert’s argument claiming that
infinitary mathematics is conservative over finitary mathematics with respect to
Π
1
-theorems.
Theorem 134 For X any Π
1
-sentence in the language of PA, if PA ⊢ X, then
PA
Π
2
∪ ¦Con
PA
¦ ⊢ X.
Proof. By Theorem 133 and analysis of the proof of Theorem 126.
Lecture 13
Provability logic: the system GL
(Tuesday, 23th November 2010)
A proof predicate Pr(v
1
) for a system S can be thought of as an operator on
sentences in the language of S, i.e. it generates a sentence from a sentence. To signify
this viewpoint we write 2A for Pr(A). In this notation the arithmetization of
L¨ob’s theorem (problem 1 on Problem sheet 7) is expressed as (2(2A ⊃ A) ⊃ 2A).
As we shall see, this formula provides the fundamental axiom by which to axiomatize
the logic of provability.
When provability logic first began to be developed, in the 1970s, there existed
already, for more than fifty years, systems of logic for a sentence operator 2A with
the intended meaning, “A is necessarily true”, and in the 1930s, in a brief note, G¨odel
had obtained results by interpreting the 2 operator of modal logic as provability.
Such systems are called modal logic since necessity concerns not only the truth of
sentences but also the kind or mode of their truth. Modal logic provided a framework
for setting up systems for provability logic (and also a semantics of possible worlds
with an accessibility relation between worlds by which to study properties of such
systems, which has been exploited in the study of provability logic, but we will
establish the results that concern us here purely syntactically, i.e. not using these
semantic techniques). It is important to realize that provability logic is not an
extension of the logic of necessity, for which (2X ⊃ X) is valid, in contrast to which
L¨ob’s theorem shows that if S X, then S (Pr(X) ⊃ X).
The system of provability logic was named GL by George Boolos, after G¨odel
and L¨ob. It consists of propositional logic plus the provability operator.
13.1 The language of GL
The primitive symbols of GL:
107
LECTURE 13 108
A sentence ⊥ (standing for a generic false sentence, e.g. 0 = 0

in the language
of arithmetic).
Infinitely many sentence letters p
i
, generated from the symbol ‘p’ and iteration
of the subscript symbol ‘′ ’.
The sentential connective ⊃.
The sentential operator 2.
Definition 78 (sentences of GL) By recursion:
base: ⊥ and all p
i
are sentences.
recursion: If X and Y are sentences, (X ⊃ Y ) is a sentence.
If X is a sentence, 2X is a sentence.
We shall write sentences in the language of GL using the following abbrevia-
tions which, on the intended meaning for ⊃ and ⊥, express negation, conjunction,
disjunction, and equivalence:
Definition 79 ∼ X =
df
(X ⊃⊥),
(X ∧ Y ) =
df
((X ⊃ (Y ⊃⊥)) ⊃⊥),
(X ∨ Y ) =
df
((X ⊃⊥) ⊃ Y )
(X ≡ Y ) =
df
(((X ⊃ Y ) ⊃ ((Y ⊃ X) ⊃⊥)) ⊃⊥)
13.2 The axioms and inference rules of GL
Definition 80 (axioms of GL) A1. (Tautologies) Every sentence in the language
of GL that is a truth functional tautology when ⊥ is assigned the truth value F
(falsity) and ⊃ is interpreted as the truth function ‘if . . . then . . . ’ is an axiom.
A2. (Distribution) For X and Y any sentences in the language of GL, (2(X ⊃
Y ) ⊃ (2X ⊃ 2Y )) is an axiom. (Corresponds to property P
2
of provability predi-
cates.)
A3. (Arithmetized L¨ob’s Theorem) For each sentence X in the language of GL,
(2(2X ⊃ X) ⊃ 2X) is an axiom.
Definition 81 (rules of inference of GL) R1. From sentences X and (X ⊃ Y ),
infer Y . (Modus ponens)
R2. From sentence X infer 2X. (Corresponds to property P
1
of provability
predicates. In modal logic this rule is known as Necessitation.)
There is no axiom schema corresponding to property P
3
for provability predicates
since, as we shall see by Theorem 147, (2X ⊃ 22X) is derivable from the axioms
and rules of inference specified for GL.
LECTURE 13 109
The axioms and inference rules of GL arise by abstraction from the arithmetized
proof predicate for PA. Conversely, the axioms and inference rules, and hence all
theorems of GL, translate into theorems of PA. This result means that GL is sound
with respect to interpretation in PA.
Definition 82 An interpretation

of GL in PA is given by translating the language
of GL into the language of PA by the following inductive definition:
Base:
(i) ⊥

= X for X a specified sentence in the language of PA such that PA ⊢∼ X;
(ii) p

i
= s
i
for i → s
i
a specified enumeration of the sentences in the language
of PA.
Induction:
(iii) (X ⊃ Y )

= (X

⊃ Y

);
(iv) (2X)

= Pr(X

), for Pr(v
1
) an arithmetized proof predicate for PA.
Note that an interpretation

in the above specification is determined by the
choice of a sentence X such PA ⊢∼ X and a specific enumeration of the sentences
in the language of PA. Also note that we can reformulate the language and logic
of PA by including the symbol ⊥ in the language of PA with corresponding axioms
for propositional logic so that ⊥ is treated as falsity, in which case an interpretation
of GL in PA is determined by the choice of enumeration of the sentences in the
language of PA.
Theorem 135 (soundness of GL with respect to interpretation in PA) For
every interpretation

from GL into PA as in Definition 82, if GL ⊢ X, then
PA ⊢ X

.
Proof. By induction over the recursive definition of theorems of GL.
Basis step, for X an axiom of GL:
(i) If X is a truth functional tautology, then X

is a truth functional tautology,
and hence provable in PA.
(ii) If X is a Distribution Axioms, then X

is an instance of P
2
and hence provable
in PA.
(iii) If X is a L¨ob’s Theorem axiom, then X

is an instance of arithmetized L¨ob’s
Theorem, which by problem 1 on Problem sheet 7, is provable in PA.
Induction step:
If PA ⊢ X

and PA ⊢ (X ⊃ Y )

, then since (X ⊃ Y )

= (X

⊃ Y

), PA ⊢ Y

.
If PA ⊢ X

, then by P
1
from arithemetization of the provability predicate for
PA, PA ⊢ Pr(X

), i.e. PA ⊢ (2X)

.
There is also a completeness theorem, due to Robert Solovay, for GL with respect
to interpretation in PA.
LECTURE 13 110
Theorem 136 (Solovay completeness theorem for GL) For X any sentence
in the language of GL, if for every interpretation

(as in Definition 82), PA ⊢ X

,
then GL ⊢ X; or equivalently, for X any sentence in the language of GL, if GL X,
then there is an interpretation

such that PA X

.
Proof. Beyond the scope of this course.
13.3 Some derivations in GL
Unarithmetized L¨ob’s Theorem holds for GL, i.e.
Lemma 137 If GL ⊢ (2X ⊃ X), then GL ⊢ X.
Proof. Suppose GL ⊢ (2X ⊃ X). Then by R1, GL ⊢ 2(2X ⊃ X). By A3,
GL ⊢ (2X(2 ⊃ X) ⊃ 2X), so by R1, GL ⊢ 2X. Then by the initial supposition
and R1, GL ⊢ X.
Lemma 138 If GL ⊢ (A ⊃ B), then GL ⊢ (2A ⊃ 2B).
Proof. This is a notational variant of P
4
and the proof of P
4
in Lemma 103
proves this result.
Next we note that GL is closed under truth functional consequence, i.e.
Theorem 139 If Y is a truth functional consequence of finitely many formulas
which are each provable in GL, then GL ⊢ Y .
Proof. If Y follows truth functionally from the finitely many formulas X
1
, . . . , X
r
,
then (X
1
⊃ (X
2
⊃ (. . . ⊃ (X
r
⊃ Y ) . . .))) is a tautology and hence an axiom of GL,
and GL ⊢ Y by r-many applications of Modus ponens starting with this axiom.
Remarks. In proofs that depend on this theorem I will say “by propositional
logic in GL” or just “by propositional logic”, rather than citing Theorem 139. By
the compactness theorem for truth functional logic, the above result holds for Y a
truth functional consequence of any set of provable formulas, not just finite sets of
formulas, but we have not need for this generalization.
Theorem 140 GL ⊢ (2(X ∧ Y ) ≡ (2X ∧ 2Y ))
LECTURE 13 111
Proof. (i) Both ((X ∧ Y ) ⊃ X) and ((X ∧ Y ) ⊃ Y ) are tautologies and hence
axioms of GL. By Lemma 138, GL ⊢ (2(X∧Y ) ⊃ 2X) and GL ⊢ (2(X∧Y ) ⊃ 2Y ).
Then by propositional logic in GL, GL ⊢ (2(X ∧ Y ) ⊃ (2X ∧ 2Y ))
(ii) Since (X ⊃ (Y ⊃ (X ∧ Y ))) is a tautology, GL ⊢ (X ⊃ (Y ⊃ (X ∧ Y ))).
Hence by Lemma 138, GL ⊢ (2X ⊃ 2(Y ⊃ (X ∧ Y ))). By A2, GL ⊢ (2(Y ⊃
(X ∧ Y ) ⊃ (2Y ⊃ 2(X ∧ Y )))). Then by propositional logic in GL, GL ⊢ (2X ⊃
(2Y ⊃ 2(X ∧ Y ))). The formulas ((2X ∧ 2Y ) ⊃ 2X) and ((2X ∧ 2Y ) ⊃ 2Y )
are tautologies and hence axioms of GL, so by propositional logic in GL, GL ⊢
((2X ∧ 2Y ) ⊃ 2(X ∧ Y )).
Proposition 141 GL ⊢ (2(X ≡ Y ) ⊃ (2X ≡ 2Y ))
Proof. By Definition 79, (X ≡ Y ) =
df
((X ⊃ Y ) ∧ (Y ⊃ X)), so by Theo-
rem 140, GL ⊢ (2(X ≡ Y ) ≡ (2(X ⊃ Y ) ∧ 2(Y ⊃ X))). By A
2
and Theorem 139,
GL ⊢ ((2(X ⊃ Y ) ∧ 2(Y ⊃ X)) ⊃ ((2X ⊃ 2Y ) ∧ (2Y ⊃ 2X))). Then by propo-
sitional logic in GL, GL ⊢ (2(X ≡ Y ) ⊃ ((2X ⊃ 2Y ) ∧ (2Y ⊃ 2X))), which by
Definition 79 is GL ⊢ (2(X ≡ Y ) ⊃ (2X ≡ 2Y )).
The converse of Proposition 141 does not hold.
Proposition 142 If PA is 1-consistent, GL ((2X ≡ 2Y ) ⊃ 2(X ≡ Y )).
Proof. PA ((Pr(G) ≡ Pr(0 = 0

)) ⊃ (Pr(G ≡ 0 = 0

))), since PA ⊢
(Pr(G) ≡ Pr(0 = 0

)), by Theorem ??, and PA (Pr(G ≡ 0 = 0

)) if PA is
1-consistent. Hence by Theorem 135, GL ((2X ≡ 2Y ) ⊃ 2(X ≡ Y )).
13.4 Closure of GL under substitution by prov-
ably equivalent formulas
The result of substituting a sentence A for sentential variable p in formula F, F
p
(A),
is defined by the following recursion.
Definition 83 (F
p
(A)) 1. If F = p, then F
p
(A) = A.
2. If F = q where q = p, then F
p
(A) = q.
3. If F =⊥, then F
p
(A) =⊥.
4. If F = (G ⊃ H), then F
p
(A) = (G
p
(A) ⊃ H
p
(A)).
5. If F = 2G, then F
p
(A) = 2(G
p
(A)).
Proposition 143 (closure of GL under substitution) For X any formula in
the language of GL, and F(p) any formula in the language of GL in which the
sentence letter p occurs, if GL ⊢ F(p), then GL ⊢ F
p
(X).
LECTURE 13 112
Proof. By induction on the length of a proof of F(p) in GL.
Base case: The proof is of length 1, i.e. F(p) is an axiom, in which case it is of
the form A1 (tautology), A2 (distribution), or A3 (arithmetized L¨ob’s Theorem).
Then F
p
(X) is an axiom, of the same form as F(p) is, so GL ⊢ F
p
(X).
Induction steps: (i) The proof of F(p) ends with an application of R1, i.e. GL ⊢
Y and GL ⊢ (Y ⊃ F(p)). By Induction Hypothesis, GL ⊢ Y
p
(X) and GL ⊢ (Y ⊃
F(p))
p
(X). Then by clause 4 of Definition 83, GL ⊢ (Y
p
(X) ⊃ F
p
(X)), so by R1,
GL ⊢ F
p
(X).
(ii) The proof of F(p) ends with an application of R2, i.e. F(p) is of the form
2G(p) and GL ⊢ G(p). By Induction Hypothesis, GL ⊢ G
p
(X). By R2, GL ⊢
2(G
p
(X)). Then by clause 5 of Definition 83, GL ⊢ F
p
(X).
Theorem 144 (provable equivalence of substitution of provable equivalents)
For all formulas A, B, and F and any sentence letter p, if GL ⊢ (A ≡ B), then GL
⊢ (F
p
(A) ≡ F
p
(B)).
Proof. We argue by induction over the inductive definition of formulas F.
If F = p, what is to be proved is that if GL ⊢ (A ≡ B), then GL ⊢ (A ≡ B),
which holds trivially.
If F = q for p = q, what is to be proved is GL ⊢ (A ≡ B), then GL ⊢ (q ≡ q).
The consequent is provable outright, so the implication holds.
If F =⊥, what is to be proved is GL ⊢ (A ≡ B), then GL ⊢ (⊥≡⊥), for which
again the consequent is provable outright.
If F = (G ⊃ H), then we have as induction hypotheses, if GL ⊢ (A ≡ B), then
GL ⊢ (G
p
(A) ≡ G
p
(B)), and if GL ⊢ (A ≡ B), then GL ⊢ (H
p
(A) ≡ H
p
(B)). By
Definition 83, what is to be proved is if GL ⊢ (A ≡ B), then GL ⊢ ((G
p
(A) ⊃
H
p
(A)) ≡ (G
p
(B) ⊃ H
p
(B))), which follows by propositional logic from the induc-
tion hypotheses.
If F = 2G, then we have by induction hypothesis that if GL ⊢ (A ≡ B), then
GL ⊢ (G
p
(A) ≡ G
p
(B)). Then by Lemma 138 and propositional logic in GL, GL
⊢ (2(G
p
(A)) ≡ 2(G
p
(B))), so by Definition 83, GL ⊢ ((2G)
p
(A) ≡ (2G)
p
(B)),
which is to say, GL ⊢ (F
p
(A) ≡ F
p
(B)).
The proof of Theorem 144 generalizes to establish provable equivalence of sub-
stitution on more than one sentence letter.
Theorem 145 (substitution on more than one sentence letter) For F any
formula with sentence letters p
k
1
, . . . , p
k
n
, and for pairs of formulas A
i
, B
i
, i =
1, . . . , n, such that GL ⊢ (A
i
≡ B
i
), GL ⊢ (F(A
1
, . . . , A
n
) ≡ F(B
1
, . . . , B
n
)), where
F(A
1
, . . . , A
n
) is the result of substituting A
i
for p
k
i
in F, and F(B
1
, . . . , B
n
) is the
result of substituting B
i
for p
k
i
in F.
LECTURE 13 113
Proof. Exactly the same proof structure as for Theorem 144, with just reformu-
lation of the induction hypothesis so that it’s for multiple substitutions, establishes
this result.
Theorem 144 immediately establishes that GL is closed under substitution of
provable equivalents.
Corollary 146 If GL ⊢ F
p
(A) and GL ⊢ (A ≡ B), then GL ⊢ F
p
(B).
Proof. From GL ⊢ (A ≡ B) and Theorem 144, we have by ∧-elimination, GL ⊢
(F
p
(A) ⊃ F
p
(B)), so by Modus ponens from GL ⊢ F
p
(A), we have GL ⊢ F
p
(B).
We are now able to show that GL proves P
3
.
Theorem 147 GL ⊢ (2X ⊃ 22X)
Proof. The formula (X ⊃ ((22X ∧ 2X) ⊃ (2X ∧ X)) is a tautology and
hence GL ⊢ (X ⊃ ((22X ∧ 2X) ⊃ (2X ∧ X)). Since, by Theorem 140, GL
⊢ ((22X ∧ 2X) ≡ 2(2X ∧ X))), by Corollary 146, taking F
p
as (X ⊃ (p ⊃
(2X ∧ X)), GL ⊢ (X ⊃ (2(2X ∧ X) ⊃ (2X ∧ X)). Then by Lemma 138 (P
4
),
GL ⊢ (2X ⊃ 2(2(2X ∧ X) ⊃ (2X ∧ X)). The formula (2(2(2X ∧ X) ⊃
(2X ∧ X)) ⊃ 2(2X ∧ X)) is an A3 axiom of GL. Then by propositional logic in
GL, GL ⊢ (2X ⊃ 2(2X ∧ X)). Then by Theorems 140 and propositional logic in
GL, GL ⊢ (2X ⊃ (22X ∧ 2X). Since ((22X ∧ 2X) ⊃ 22X) is a tautology, by
propositional logic in GL, GL ⊢ (2X ⊃ 22X).
13.5 Closure of GL under substitution of provably
equivalent formulas is provable in GL
We now show that the closure of GL under substitution of provably equivalent
formulas, Theorem 144, can be formalized in GL.
Theorem 148 (arithmetized substitution theorem) For all formulas A, B,
and F and propositional variable p, GL ⊢ (2(A ≡ B) ⊃ 2(F
p
(A) ≡ F
p
(B))).
Proof. The proof is by induction over the recursion that generates the formula
F.
If F = p, what is to be proved is GL ⊢ (2(A ≡ B) ⊃ 2(A ≡ B)), which is a
tautology and hence provable in GL.
If F = q for p = q, what is to be proved is GL ⊢ (2(A ≡ B) ⊃ 2(q ≡ q)). Since
(q ≡ q) is a tautology, GL ⊢ (q ≡ q), so by R
2
, GL ⊢ 2(q ≡ q). The result follows
by propositional logic in GL.
LECTURE 13 114
If F =⊥, what is to be proved is GL ⊢ (2(A ≡ B) ⊃ 2(⊥≡⊥)). The argument
is as for the preceding case.
If F = (G ⊃ H), we have as induction hypotheses, GL ⊢ (2(A ≡ B) ⊃
2(G
p
(A) ≡ G
p
(B))), and GL ⊢ (2(A ≡ B) ⊃ 2(H
p
(A) ≡ H
p
(B))). By propo-
sitional logic in GL, GL ⊢ (2(A ≡ B) ⊃ (2(G
p
(A) ≡ G
p
(B)) ∧ 2(H
p
(A) ≡
H
p
(B)))). Then by Theorem 140 and propositional logic in GL, GL ⊢ (2(A ≡
B) ⊃ 2((G
p
(A) ≡ G
p
(B)) ∧ (H
p
(A) ≡ H
p
(B)))). The following formula is a tau-
tology and so provable in GL (as an axiom): (((G
p
(A) ≡ G
p
(B)) ∧ (H
p
(A) ≡
H
p
(B))) ⊃ ((G
p
(A) ⊃ H
p
(A)) ≡ (G
p
(B) ⊃ H
p
(B)))). Then by Lemma 138
(P
4
), GL ⊢ (2((G
p
(A) ≡ G
p
(B)) ∧ (H
p
(A) ≡ H
p
(B))) ⊃ 2((G
p
(A) ⊃ H
p
(A)) ≡
(G
p
(B) ⊃ H
p
(B)))). From this result and the two steps earlier we have, by proposi-
tional logic in GL, GL ⊢ (2(A ≡ B) ⊃ 2((G
p
(A) ⊃ H
p
(A)) ≡ (G
p
(B) ⊃ H
p
(B)))),
which by Definition 83(4), is GL ⊢ (2(A ≡ B) ⊃ 2(F
p
(A) ≡ F
p
(B))).
If F = 2G, we have as induction hypothesis that GL ⊢ (2(A ≡ B) ⊃ 2(G
p
(A) ≡
G
p
(B))). Then by Proposition 141 and propositional logic in GL, GL ⊢ (2(A ≡
B) ⊃ (2(G
p
(A)) ≡ 2(G
p
(B)))). Then by Definition 83 (5.), GL ⊢ (2(A ≡ B) ⊃
(F
p
(A) ≡ F
p
(B))). Then by Lemma 138 (P
4
), GL ⊢ (22(A ≡ B) ⊃ 2(F
p
(A) ≡
F
p
(B))). By Theorem 147, GL ⊢ (2(A ≡ B) ⊃ 22(A ≡ B)), so by propositional
logic in GL, GL ⊢ (2(A ≡ B) ⊃ 2(F
p
(A) ≡ F
p
(B))).
13.6 Strengthened proof that the closure of GL
under substitution of provably equivalent for-
mulas is provable in GL
This strengthened proof makes use of the following technical definition.
Definition 84 ⊡X =
df
(2X ∧ X)
Lemma 149 GL ⊢ (⊡X ⊃ X)
Proof. ⊢ (⊡X ⊃ X) is a tautology.
Remark. Lemma 149 is a triviality but draws attention to a key property of ⊡
that holds for all formulas and which, by L¨ob’s Theorem for GL (Lemma 137), holds
for 2 only for formulas provable in GL. The constraint of L¨ob’s Theorem makes it
very difficult to derive an unboxed conclusion from a boxed premiss, i.e. of the form
2X. The situation is much more flexible if we are able to strengthen the premiss to
⊡X.
Lemma 150 For each formula X in the language of GL, the formulas 2X, 2⊡X,
and ⊡2X are provably equivalent in GL.
LECTURE 13 115
Proof. The provable equivalence of 2 ⊡ X and ⊡2X is, by Definition 84, an
instance of Theorem 140, namely GL ⊢ (2(2X ∧ X) ≡ (22X ∧ 2X)).
By Theorem 147 (P
3
) and propositional logic in GL, GL ⊢ (2X ≡ (22X∧2X)),
which by Definition 84 is GL ⊢ (2X ≡ ⊡2X).
Corollary 151 GL ⊢ (⊡X ⊃ 2⊡X)
Proof. By Lemma 149.
Remark. The converse implication is not provable in GL, since
GL (2X ⊃ X) unless GL ⊢ X (Lemma 137).
Lemma 152 If GL ⊢ (⊡X ⊃ Y ), then GL ⊢ (2X ⊃ 2Y ).
Proof. Assume GL ⊢ (⊡X ⊃ Y ). Then by Lemma 138, GL ⊢ (2 ⊡X ⊃ 2Y ).
Then by Lemma 150 and propositional logic in GL, GL ⊢ (2X ⊃ 2Y ).
Lemma 153 GL ⊢ (⊡X ⊃ Y ) if and only if GL ⊢ (⊡X ⊃ ⊡Y ).
Proof. (i) From Definition 84 by propositional logic in GL, GL ⊢ (⊡X ⊃ 2X).
Assume GL ⊢ (⊡X ⊃ Y ). Then by Lemma 152, GL ⊢ (2X ⊃ 2Y ). Then by
Theorem 139, GL ⊢ (⊡X ⊃ 2Y ). From the assumption and this last result by
propositional logic in GL, GL ⊢ (⊡X ⊃ (2Y ∧ Y )), i.e. GL ⊢ (⊡X ⊃ ⊡Y ).
(ii) Assume GL ⊢ (⊡X ⊃ ⊡Y ). By Lemma 149, GL ⊢ (⊡Y ⊃ Y ) and hence by
propositional logic in GL, GL ⊢ (⊡X ⊃ Y ).
Theorem 154 (strengthened arithmetized substitution theorem) For all for-
mulas A, B, and F and propositional variable p, GL ⊢ (⊡(A ≡ B) ⊃ (F
p
(A) ≡
F
p
(B))).
Proof. The proof is by induction over the recursion that generates the formula
F.
If F = p, what is to be proved is GL ⊢ (⊡(A ≡ B) ⊃ (A ≡ B)), which holds by
Definition 84 and propositional logic in GL.
If F = q for p = q, what is to be proved is GL ⊢ (⊡(A ≡ B) ⊃ (q ≡ q)). For any
formula H, (H ⊃ (q ⊃ q)) is a tautology and hence an axiom of GL, so in particular
GL ⊢ (⊡(A ≡ B) ⊃ (q ≡ q)).
If F =⊥, what is to be proved is GL ⊢ (⊡(A ≡ B) ⊃ (⊥≡⊥)). The argument is
as for the preceding case.
If F = (G ⊃ H), we have as induction hypotheses, GL ⊢ (⊡(A ≡ B) ⊃ (G
p
(A) ≡
G
p
(B))), and GL ⊢ (⊡(A ≡ B) ⊃ (H
p
(A) ≡ H
p
(B))). By Definition 83(4), what
is to be proved is GL ⊢ (⊡(A ≡ B) ⊃ ((G
p
(A) ⊃ H
p
(A)) ≡ (G
p
(B) ⊃ H
p
(B)))),
which follows from the induction hypotheses by propositional logic in GL.
LECTURE 13 116
If F = 2G, we have as induction hypothesis that GL ⊢ (⊡(A ≡ B) ⊃ (G
p
(A) ≡
G
p
(B))). Then by Lemma 138, GL ⊢ (2⊡(A ≡ B) ⊃ 2(G
p
(A) ≡ G
p
(B))). By The-
orem 140, Axioms A2, and propositional logic in GL, GL ⊢ (2(G
p
(A) ≡ G
p
(B)) ⊃
(2(G
p
(A)) ≡ 2(G
p
(B)))). From these two last results, by propositional logic in
GL and Defintion 83(5), GL ⊢ (2 ⊡ (A ≡ B) ⊃ ((2G
p
)(A) ≡ (2G
p
)(B))). Then
by Corollary 151 and propositional logic in GL, GL ⊢ (⊡(A ≡ B) ⊃ ((2G
p
)(A) ≡
(2G
p
)(B))).
Corollary 155 (variant proof of Theorem 148) For all formulas A, B, and F
and propositional variable p, GL ⊢ (2(A ≡ B) ⊃ 2(F
p
(A) ≡ F
p
(B))).
Proof. By Theorem 154 and Lemma 152.
Theorem 156 (Theorem 154 generalized to multiple substitutions) Let F(p
i
1
, . . . , p
i
m
)
be a formula with sentence letters p
i
1
, . . . , p
i
m
. Then
GL ⊢ (⊡(A
1
≡ B
1
) ∧ . . . ∧ ⊡(A
m
≡ B
m
)) ⊃ (F(A
1
, . . . , A
m
) ≡ F(B
1
, . . . , B
m
))).
Proof. Exercise.
Lecture 14
The fixed-point theorem for GL
(Wednesday, 24 November 2010)
14.1 The notion of a sentence letter modalized
in a sentence, and arithmetized substitution
for modalized sentences
Definition 85 Y is a subsentence of X is defined recursively by
Base case: X is a substence of X
Recursion clauses: If the sentence (Z ⊃ W) is a subsentence of X, then Z is a
subsentence of X and W is a subsentence of X.
If the sentence 2Z is a subsentence of X, then Z is a subsentence of X.
Definition 86 A sentence letter p is modalized in a sentence X iff every occurrence
of p in X is a subsentence of a subsentence of X of the form 2Y .
Examples of sentences in which the sentence letter p is modalized: (1) 2p, (2)
∼ 2p, (3) 2 ∼ p, (4) ∼ 2 ∼ p, (5) (2p ⊃ q) (6) 2(2p ⊃ p), (7) (2(2p ⊃ p) ⊃ 2p),
(8) 2(p ≡ (2p ⊃ q)), (9) (2(2p ⊃ q)∧ ∼ 2p), (10) ⊥, (11) (⊥⊃⊥), (12) 2 ⊥, (13)
q (where q = p), (14) 2q (in these latter cases p is modalized in the sentence since p
does not occur in the sentence and so every occurrence of p in the sentence is within
the scope of 2).
Examples of sentences not modalized in p: (1) p, (2) (p ⊃⊥), (3) (2p ⊃ p), (4)
(p ≡ (2p ⊃ q).
Definition 87 For X a sentence in which the sentence letter p occurs modalized,
a sentence D(p
k
1
, . . . , p
k
n
) with sentence letters p
k
1
, . . . , p
k
n
that do not occur in X
117
LECTURE 14 118
and sentences 2C
1
(p), . . . , 2C
n
(p) such that X(p) = D(2C
1
(p), . . . , 2C
n
(p)) (i.e.
the result of substituting each sentence 2C
i
(p) for all occurrences of the variable p
i
in the sentence D(p
k
1
, . . . , p
k
n
)) is called a decomposition of X with respect to p;
D(p
k
1
, . . . , p
k
n
) is called a decomposition sentence for X, and 2C
1
(p), . . . , 2C
n
(p)
are called components of X.
Lemma 157 If X is modalized in p, then there is a decomposition of X with respect
of p.
Proof. We give a constructive proof of this result, i.e. a method, or in fact
two different methods, which generate a decomposition of X for X any sentence
modalized in p.
Method 1 (top down): The whole sentence X cannot consist just of a sentence
letter p
i
or ⊥ since no such sentence is modalized in p. So there are two cases, either
X is of the form 2Y for some sentence Y , or X is of the form (Y ⊃ Z) for sentences
Y and Z.
(i) X is of the form 2Y . Then we are done, with the decomposition sentence for
X any sentence letter p
i
distinct from p and component 2Y .
(ii) X is of the form (Y ⊃ Z). Then p occurs in Y or in Z, or both, and
wherever it occurs it occurs modalized. If the sentence or sentences in which p
occurs modalized is/are of the form 2W then by (i) we are done. If not then it is
of the form (U ⊃ V ) and we repeat the argument. Since X is generated in finitely
many steps, this process comes to an end, and since p occurs modalized, it ends in
components, i.e. sentences of the form 2C(p) in which p occurs modalized.
Method 2 (bottom up): For each occurrence of p in X, find the innermost oc-
currence of 2 in whose scope that occurrence of p occurs. For each such occurrence
of 2, let C
i
(p) be the sentence to which that occurrence of 2 is prefixed. From the
resulting set of sentences 2C
i
(p), discard those sentences that are a subsentence of
any of the others in which that occurrence of p occurs. The remaining sentences
2C
i
(p) will be the components of the decomposition of X. In X replace each of
these component sentences by a distinct sentence letter not occurring in X, at each
occurrence in X of that component. This gives the decomposition sentence of X for
those components.
Examples. (2(2(2p ⊃ p) ⊃ p) ⊃ 2p)
Method 1: The sentence is an implication between boxed formulas, so the com-
ponents are 2(2(2p ⊃ p) and 2p and the decomposition formula is (p
i
⊃ p
j
).
Method 2: When applied to the three occurrences of p in the antecedent sentence
of this implication, the procedure results in sentence 2p for the first occurrence of p
(going from the left), 2(2p ⊃ p)) for the second occurrence of p, and 2(2(2p ⊃ p) ⊃
p) for the third occurrence of p, and for the single occurrence of p in the consequent
sentence, it results in 2p. From the sentences resulting from the occurrences of p in
LECTURE 14 119
the antecedent sentence, we discard 2p because 2p with that occurrence of p occurs
in 2(2p ⊃ p)) (and also in 2(2(2p ⊃ p) ⊃ p)), and we discard 2(2p ⊃ p)) since it
is a subsentence of 2(2(2p ⊃ p) ⊃ p) and the two occurrences of p in 2(2p ⊃ p))
occur in 2(2(2p ⊃ p) ⊃ p). So the components are 2(2(2p ⊃ p) ⊃ p) and 2p,
and the decomposition sentence for these components is (p
i
⊃ p
j
).
Note that for some sentences modalized in a sentence letter the two methods
results in different decompositions, e.g.
(2(2p ⊃ q)∧ ∼ 2p)
By Method 1:
D
2
(p
1
, p
2
) = (p
1
∧ ∼ p
2
), 2C
1
(p) = 2(2p ⊃ q), and 2C
2
(p) = 2p.
By Method 2:
D
1
(p
1
, q) = (2(p
1
⊃ q)∧ ∼ p
1
), and 2C(p) = 2p.
Some modalized sentences have more than two decompositions, e.g.
222p has three decompositions:
(i) D
1
= p
1
and 2C
1
(p) = 222p.
(ii) D
2
= 2p
1
and 2C
1
(p) = 22p.
(iii) D
3
= 22p
1
and 2C
1
(p) = 2p.
The decomposition of 222p that results from the proof of Lemma 157 is (iii).
The second example generalizes to show that
Proposition 158 For each n, there is a sentence, modalized in a sentence letter,
that has n-many decompositions.
Proof. For each n, 2. . . 2

n
p has decompositions
D
i
= 2. . . 2

n−i
p
1
with component 2. . . 2

i
p, for each i such that 1 ≤ i ≤ n.
For sentences in a which a sentence letter occurs modalized there is an arith-
metized substitution theorem which yields the conclusion of Theorem 154 on the
hypothesis of Theorem 148.
Theorem 159 (arithmetized substitution in modalized sentences) Let F(p)
be a sentence in which sentence letter p occurs modalized. Then
GL ⊢ (2(A ≡ B) ⊃ (F
p
(A) ≡ F
p
(B))).
Proof. By Lemma 157, there is a decomposition D(p
1
, . . . , p
m
), with sentence
letters p
1
, . . . , p
m
not in F(p) and components 2C
1
(p), . . . , 2C
m
(p). The following
is (a recipe for) a proof in GL.
LECTURE 14 120
(1) (2(A ≡ B) ⊃ 2(C
i
p
(A) ≡ C
i
p
(B))) Theorem 148
(2) (2(A ≡ B) ⊃ (2C
i
(A) ≡ 2C
i
(B))) (1) Proposition 141 prop logic
(3) (22(A ≡ B) ⊃ 2(2C
i
(A) ≡ 2C
i
(B))) (2) Lemma 138
(4) (2(A ≡ B) ⊃ 2(2C
i
(A) ≡ 2C
i
(B))) (3) Theorem 147 prop logic
(5) (2(A ≡ B) ⊃ ⊡(2C
i
(A) ≡ 2C
i
(B))) (2)(4)
(6) (((⊡(2C
1
(A) ≡ 2C
1
(B))) ∧ . . . ∧ ⊡(2C
m
(A) ≡ 2C
m
(B))) ⊃
(D(2C
1
(A), . . . , 2C
m
(A)) ≡ D(2C
1
(B), . . . , 2C
m
(B))))
Theorem 156
(7) (2(A ≡ B) ⊃ (F
p
(A) ≡ F
q
(B))) (5)(6) prop logic
14.2 The fixed point theorem for GL
Definition 88 (fixed point) For X a sentence in the language of GL that contains
the sentence letter p
i
, a sentence F that contains only sentence letters contained in
X and does not contain p
i
is a fixed point for X with respect of p
i
if and only if
GL ⊢ (F ≡ X
p
i
(F)).
We shall see that every sentence modalized in p
i
has a fixed point with respect
to p
i
. We establish this result first for the simplest case of a sentence modalized in
p
i
, i.e. of the form 2Y (p
i
).
Lemma 160 For p
i
any propositional variable and Y (p
i
) any sentence in the lan-
guage of GL in which p
i
occurs, the result of substituting (⊥⊃⊥) for each occurrence
of p
i
in 2Y (p
i
) is a fixed point for 2Y (p
i
).
Proof. In this proof I abbreviate (⊥⊃⊥) as ⊤, and write 2Y (⊤) for the result
of substituting (⊥⊃⊥) for each occurrence of p
i
in 2Y (p
i
). The following derivation
is a proof in GL.
(1) (2Y (⊤) ⊃ (⊤ ≡ 2Y (⊤)) tautology.
(2) 2(2Y (⊤) ⊃ (⊤ ≡ 2Y (⊤)) (1) P
1
.
(3) (22Y (⊤) ⊃ 2(⊤ ≡ 2Y (⊤))) (2) P
2
Modus ponens
(4) (2Y (⊤) ⊃ 2(⊤ ≡ 2Y (⊤))) (3) P
3
propositional logic.
(5) (2(⊤ ≡ 2Y (⊤)) ⊃ 2(Y (⊤) ≡ Y (2Y (⊤)))) Theorem 148.
(6) (2(Y (⊤) ≡ Y (2Y (⊤))) ⊃ (2Y (⊤) ≡ 2Y (2Y (⊤)))) Proposition 141.
(7) (2Y (⊤) ⊃ (2Y (⊤) ≡ 2Y (2Y (⊤)))) (4) (5) (6) transitivity of ⊃.
(8) (2Y (⊤) ⊃ 2Y (2Y (⊤))) (7) prop logic.
LECTURE 14 121
(9) (2Y (⊤) ⊃ ⊡(⊤ ≡ 2Y (⊤))) (1) (4) prop logic
(10) (2Y (⊤) ⊃ (Y (⊤) ≡ Y (2Y (⊤)))) (9), Theorem 154, prop logic
(11) (Y (2Y (⊤)) ⊃ (2Y (⊤) ⊃ Y (⊤))) (10) prop logic.
(12) (2Y (2Y (⊤)) ⊃ 2(2Y (⊤) ⊃ Y (⊤))) (11) Lemma 138 (P
4
).
(13) (2Y (2Y (⊤)) ⊃ 2Y (⊤)) (12) A
3
prop logic.
(14) (2Y (⊤) ≡ 2Y (2Y (⊤))) (8)(13)∧-Introduction.
Before proving the fixed point theorem itself we need another lemma and for
that lemma we need a definition:
Definition 89 For 1 ≤ i ≤ m, let A
i
(p
k
1
, . . . , p
k
m
), be m-many sentences each of
which includes among its sentence letters the m-many sentence letters p
k
1
, . . . , p
k
m
.
A system of simultaneous equivalences of the form
(p
k
i
≡ A
i
(p
k
1
, . . . , p
k
m
))
is solvable (in GL) iff there are sentences F
1
, . . . , F
m
not containing p
k
1
, . . . , p
k
m
such that for 1 ≤ i ≤ m,
GL ⊢ (F
i
≡ A
i
(F
1
, . . . , F
m
)).
Lemma 161 (solving systems of simultaneous equivalences) Every system of
m-many simultaneous equivalences of the form
(p
k
i
≡ 2C
i
(p
k
1
, . . . , p
k
m
))
is solvable.
Proof. The proof is by induction on m.
m = 1. This case is Lemma 160.
Suppose the Lemma holds for m. Let 2C
i
(p
k
1
, . . . , p
k
m
, p
k
m+1
) be a set of (m +
1)-many sentences with m + 1-many sentence letters p
k
1
, . . . , p
k
m
, p
k
m+1
. By the
Induction Hypothesis, there are sentences F
i
(p
k
m+1
), 1 ≤ i ≤ m, each of which does
not contain the sentence letters p
k
1
, . . . , p
k
m
and does contain the sentence letter
p
k
m+1
, such that for 1 ≤ i ≤ m,
(1) GL ⊢ (F
i
(p
k
m+1
) ≡ 2C
i
(F
1
(p
k
m+1
), . . . , F
m
(p
k
m+1
), p
k
m+1
)).
LECTURE 14 122
Now by Lemma 160 applied to the sentence that results from2C
m+1
(p
k
1
, . . . , p
k
m
, p
k
m+1
)
by substituting F
i
(p
k
m+1
) for p
k
i
for each i such that 1 ≤ i ≤ m, i.e.
2C
m+1
(F
1
(p
k
m+1
), . . . , F
m
(p
k
m+1
), p
k
m+1
), there is a sentence F
m+1
such that
(2) GL ⊢ F
m+1
≡ 2C
m+1
(F
1
(F
m+1
), . . . , F
m
(F
m+1
), F
m+1
)).
By substitution of F
m+1
into the provable equivalences (1), the sentences
F
1
(F
m+1
), . . . , F
m
(F
m+1
), F
m+1
solve the set of equivalences.
Theorem 162 (Fixed Point Theorem) For every sentence X in the language of
GL modalized in p
r
, there is a sentence F containing only sentence letters that occur
in X and not containing p
r
such that GL ⊢ (F ≡ X
p
r
(F)).
Proof. Let X be any sentence modalized in p
r
. Then by Lemma 157, X(p
r
)
has a decomposition D(p
k
1
, . . . , p
k
n
), with sentence letters p
k
1
, . . . , p
k
n
distinct from
p
r
, and components 2C
1
(p
r
), . . . , 2C
n
(p
r
), i.e. X(p
r
) = D(2C
1
(p
r
), . . . , 2C
n
(p
r
)),
where 2C
i
(p
r
) is substituted for p
k
i
in D(p
k
1
, . . . , p
k
n
).
By Lemma 161, the set of equivalences (p
k
i
≡ 2C
i
(D(p
k
1
, . . . , p
k
n
))) indexed by
i = 1, . . . , n, where the sentence D(p
k
1
, . . . , p
k
n
) is substituted for p
r
in 2C
i
(p
r
), is
solvable. Let F
1
, . . . , F
n
be a solution, i.e. for each i = 1, . . . , n,
GL ⊢ (F
i
≡ 2C
i
(D(F
1
, . . . , F
n
))). Then by Theorem 145,
GL ⊢ (D(F
1
, . . . , F
n
) ≡ D(2C
1
(D(F
1
, . . . , F
n
)), . . . , 2C
n
(D(F
1
, . . . , F
n
)))),
which shows that D(F
1
, . . . , F
n
) is a fixed point for X(p
r
) with respect to p
r
.
Remark. The sentence F gives the explicit, i.e. non self-referential, content of
a self-referential sentence p such that (p ≡ X(p)), for X modalized in p.
There are a number of different proofs of this theorem, most of which use Kripke
models for the modal operator (three such proofs are expounded by George Boolos in
his book The Logic of Provability, Cambridge University Press, 1993, pp. 104-123).
The purely syntactic proof given here follows Per Lindstrom, “Provability logic—
a short introduction”, Theoria 62 (1996), pp. 31-35; see also Craig Smorynski,
Self-Reference and Modal Logic, Springer, 1985, pp. 78-82.
The fixed points proved to exist by the Fixed Point Theorem are unique to within
provable equivalence.
Theorem 163 (provable equivalence of fixed points) Let X(p) be a sentence
in the language of GL in which the sentence letter p occurs modalized, and let q
be a sentence letter that does not occur in X(p). Abbreviating X
p
(q) as X(q), GL
⊢ ((2(p ≡ X(p)) ∧ 2(q ≡ X(q))) ⊃ 2(p ≡ q)).
LECTURE 14 123
Proof. By Theorem 159 (arithmetized substitution into modalized sentences),
taking A as p and B as q,
(2(p ≡ q) ⊃ (X(p) ≡ X(q))) is provable in GL. Then by propositional logic,
(((p ≡ X(p)) ∧ (q ≡ X(q))) ⊃ (2(p ≡ q) ⊃ (p ≡ q))). By Lemma 138 and
Theorem 140,
((2(p ≡ X(p)) ∧ 2(q ≡ X(q))) ⊃ 2(2(p ≡ q) ⊃ (p ≡ q))). Then by Axiom A
3
and propositional logic,
((2(p ≡ X(p)) ∧ 2(q ≡ X(q))) ⊃ 2(p ≡ q)).
The uniqueness of the fixed point to within provable equivalence turns essentially
on the sentence letter of the fixed point equivalence occurring modalized in the
sentence for which the existence of a fixed point follows. We can have fixed points for
non-modalized sentences that are not unique with respect to provable equivalence.
For example, take X(p) as ((r ⊃ r) ⊃ ((q ⊃ q) ⊃ p)). By Definition 88, a fixed point
for X(p) is a sentence X containing only sentence letters that occur in X and not
containing p such that GL ⊢ (F ≡ X
p
(F)). By this criterion, since GL ⊢ (q ≡ ((r ⊃
r) ⊃ ((q ⊃ q) ⊃ q))) and GL ⊢ (r ≡ ((r ⊃ r) ⊃ ((q ⊃ q) ⊃ r))), q and r are both
fixed points of ((r ⊃ r) ⊃ ((q ⊃ q) ⊃ p)). By R
2
, GL ⊢ 2(q ≡ ((r ⊃ r) ⊃ ((q ⊃ q) ⊃
q))) and GL ⊢ 2(r ≡ ((r ⊃ r) ⊃ ((q ⊃ q) ⊃ r))). Suppose GL ⊢ 2(q ≡ r). Then
by Proposition 143, GL ⊢ 2(⊥≡ (⊥⊃⊥)), in which case GL ⊢ 2 ⊥. But assuming
the PA is Σ
1
-sound, by Theorem 135, GL 2 ⊥. In which case GL 2(q ≡ r).
The previous example can be tweaked to give an example of sentence in which
p occurs not modalized that has a fixed point with respect to p which is unique
to within provable equivalence, e.g. ((q ⊃ q) ⊃ p). It is also worth noticing how
modalizing our original example to non-uniqueness to within provable equivalence
of a fixed point for a non-modalized sentence yield provable equivalence, e.g. ((r ⊃
r) ⊃ ((q ⊃ q) ⊃ 2p)). This sentence is provably equivalent to 2p, so all fixed points
for it are provably equivalent to the Henkin sentence.
Theorem 164 (Strengthened Fixed Point Theorem for GL) For every sen-
tence X in the language of GL modalized in p, there is a sentence F containing only
sentence letters that occur in X and not containing p such that
GL ⊢ (⊡(p ≡ X(p)) ≡ ⊡(p ≡ F)).
Proof. Theorems 162 and 163 establish this result.
Theorem 165 Theorem 164 implies Theorems 162 and 163.
Proof. Substituting F for p in Theorem 164 results in GL ⊢ (⊡(F ≡ X(F)) ≡
⊡(F ≡ F)) Since (F ≡ F) is a tautology, GL ⊢ (F ≡ F), so GL ⊢ 2(F ≡ F), so by
∧-Introduction, GL ⊢ ⊡(F ≡ F). Hence GL ⊢ ⊡(F ≡ X(F)), so by ∧-Elimination,
GL ⊢ (F ≡ X(F)).
Example. To find a fixed point for (∼ 2p ∧ ∼ 2 ∼ p).
Bibliography
[1] George Boolos, The Logic of Provability, Cambridge University Press, 1993.
[2] David Hilbert, “Axiomatische Denken”, Mathematische Annalen 78 (1918), pp.
405-415; English translation in William B. Ewald (ed),
[3] David Hilbert, “On the infinite” (1926); English translation in Jean van Hei-
jenoort (ed), (1927), p. 471.
[4] David Hilbert, “Die Grundlagen der Mathematik”, Abhandlungen aus dem
mathematischen Seminar der Hamburgeshcen Universit¨ at 6 (1928), English
translation by Stephan Bauer-Mengelberg, ‘The foundations of mathematics”
in Jean van Heijenoort (ed.) From Frege to G¨ odel; A Source Book in Mathe-
matical Logic 1879-1931, Harvard University Press, 1967.
[5] Georg Kreisel, “A refinement of ω-consistency” (Abstract), The Journal of Sym-
bolic Logic 22 (1957), pp. 108-109.
[6] Craig Smorynski, “The incompleteness theorems”, Jon Barwise (ed.), Handbook
of Mathematical Logic, Horth-Holland Publishing Company, 1977, pp. 821-865.
[7] Raymond M. Smullyan, G¨ odel’s Incompleteness Theorems, Oxford University
Press, 1992.
124

Contents
0 Background: first-order logic and formal systems 0.1 First-order formal languages with identity . . . . . . . . . . . . . . . 0.2 Interpretations of first-order formal languages with identity; truth of a first-order formula in an interpretation; logical validity and logical consequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.3 A system of natural deduction for first-order logic with identity . . . 0.4 Prenex normal form . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0.4.1 Model theory and proof theory . . . . . . . . . . . . . . . . . 0.5 Completeness of a system of natural deduction with respect to firstorder logical consequence . . . . . . . . . . . . . . . . . . . . . . . . . 0.5.1 Lindenbaum’s Lemma . . . . . . . . . . . . . . . . . . . . . . 0.5.2 The Completeness Theorem . . . . . . . . . . . . . . . . . . . 0.5.3 The Compactness Theorem for first-order logic . . . . . . . . . 0.5.4 The existence of non-standard models of the truths of arithmetic 0.5.5 Completeness of other systems of derivation with respect to first-order logical consequence . . . . . . . . . . . . . . . . . . 1 Introduction: a weak form of G¨del’s First Incompleteness Theoo rem; the symbols and expressions of a language for arithmetic LE ; G¨del numbering of the expressions of LE o 1.1 Introduction: a weak form of G¨del’s First Incompleteness Theorem o 1.2 The symbols and expressions of a language for arithmetic LE . . . . . 1.3 G¨del numbering of the expressions of LE . . . . . . . . . . . . . . . o 2 Terms and formulas of the language LE ; expressibility of diagonal substitution in the language LE 2.1 Terms and formulas of the language LE . . . . . . . . . . . . . . . . . 2.1.1 Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.2 Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.3 Free and bound variables; open formulas and sentences . . . . 2 2

4 4 9 11 11 11 11 11 11 11

12 12 16 17 20 20 20 22 22

1

CONTENTS 2.2 Designation by terms in LE , truth of sentences of LE , and expressibility of sets and relations of natural numbers by formulas of LE . 2.2.1 Designation . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Truth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.3 Expressibility . . . . . . . . . . . . . . . . . . . . . . . . . . Concatenation of numbers in a given base notation is Arithmetical. Substitution and diagonal functions, and their arithmetization . . .

2

2.3 2.4

. . . . . .

23 23 24 24 24 25 28 28 29 29 31 31

3 The Diagonal Lemma; expressibility of properties of sequence numbers 3.1 The Diagonal Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Expressibility of properties of sequence numbers in the language LE 3.2.1 Properties of sequences of digits . . . . . . . . . . . . . . . . . 3.2.2 Sequence numbers . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Coding of finite sequences of G¨del numbers . . . . . . . . . . . . . . o 4 A formal system PAE for arithmetic; an Arithmetical proof predicate for PAE ; a weak version of G¨del’s first incompleteness theoo rem for PAE 4.1 A formal system PAE for arithmetic . . . . . . . . . . . . . . . . . . . 4.2 An Arithmetical proof predicate for PAE . . . . . . . . . . . . . . . . 4.3 An inefficient and a weak version of G¨del’s First Incompleteness o Theorem for PAE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 The system PA with zero, successor, addition, multiplication, and ≤ as primitive; Σ0 - and Σ1 -formulas; a Σ0 -coding of finite sets of ordered pairs; the relation xy = z is ∆1 -expressible in the language of PA 5.1 The system PA with zero, successor, addition, multiplication, and ≤ as primitive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Σ0 -formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Σ1 - and Π1 -formulas; Σ1 - Π1 - and ∆1 -relations . . . . . . . . . . . . . 5.4 Arithmetization of syntax in the language of PA . . . . . . . . . . . . 5.5 A Σ0 -coding of finite sets of ordered pairs of numbers . . . . . . . . . 5.6 The relation xy = z is ∆1 -expressible in the language of PA . . . . . . 6 Every Σ-formula is provably equivalent to a Σ1 -formula; the arithmetized proof predicate for PA is Σ1 ; the arithmetical hierarchy 6.1 Σ-formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 The arithmetized proof predicate for PA is Σ1 . . . . . . . . . . . . . 6.3 The arithmetical hierarchy . . . . . . . . . . . . . . . . . . . . . . . .

34 34 37 41

42 42 43 44 46 47 49 50 50 52 53

84 10. . . . . . . . . . . . . . .2 Weak systems of arithmetic Q and R (without induction) . . . . . . . ω-incompleteness. . . . formal provability of the Diagonal Lemma 9. . . . . . . . . . 8.4 L¨b’s Theorem . . . . . . . . . . . . . . ω-consistency and 1-consistency.1 Enumerability and the Separation Lemma . . . . . . . . . . truth of the G¨del o sentence. L¨b’s Theorem. . 83 o 10.1 Σ0 -completeness and Σ1 -completeness . . . .2 Provability predicates. . 8 The notions of consistency. . . . analyzing and strengtho ening the First Incompleteness Theorem 79 10. . . . . . incompleteness of PA from the assumption of consistency (Rosser’s Theorem). . 3 55 55 57 59 62 63 63 67 68 70 71 72 73 74 76 10 Arithemization of consistency. 8. . . . . . . . . .4 Σ0 -soundness and Σ1 -soundness . Σ0 -soundness and Σ1 -soundness 7. . . . . . . . . . . . . . . . . . . .5. . .4 Formal provability of the Diagonal Lemma . 7. . provability predicates. . . . . . . . 9. . . . . . . . 7. . . . .3 Weak and strong definability of a function in a system . . .5 Analyzing and strengthening the First Incompleteness Theorem . . 85 10. . . . . . . . . . . . . . . . . . . 9. . Q. . . . . . . . . . . . .5. . . . . . . . . . . . . . .2 Incompleteness of PA from the assumption of 1-consistency . . . .2 Strengthened second half of the First Incompleteness Theorem 85 10. . . . .CONTENTS 7 Σ0 -completeness and Σ1 -completeness. .2 Incompleteness of PA from the assumption of consistency (Rosser’s Theorem) . . . . .1 Arithmetization of the statement that a system S is consistent . . . . .1 S ∼ GS cannot be proved from the consistency of S . . 8. . o 8. . . G¨del’s Seco ond Incompleteness Theorem. . . .4 PA is ω-incomplete . 79 10. .1 The notions of consistency. . . . . . . . weak systems of arithmetic Q and R (without induction). . . . . . . .3 Σ0 -completeness of systems R. . . . Q. . . .3 G¨del’s Second Incompleteness Theorem . . . . and PA. . . . . . . . . . . . . . . . . . . . . . . . . ω-consistency and 1-consistency. .3 Consistency of S∪{ConS} is strictly weaker than 1-consistency of S . . . 7. . . Σ0 -completeness of systems R. . . . . .5. . 82 o 10. . . . . . . . . and PA . . 9 Enumerability and the Separation Lemma. 9. . . . 80 10. . . . . . . . . incompleteness from the assumption of 1-consistency. . . . . . . . .3 Truth of the G¨del sentence . . . . . . 86 11 Provable Σ1 -completeness 88 . . . . . . weak and strong definability of a function in a system. . . . . . . . .

. . . . . . . 108 . . . . . and arithmetized substitution for modalized sentences . . . . . . . . . . . . . . . . . . . PA is Π1 -conservative over PAΠ2 ∪ {ConP A } 98 12. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 .1 The ω-rule and uniform reflection .3 Equivalence of Π1 -Uniform Reflection and consistency . . . . . PA proves that PA proves every instance of the G¨del sentence. . . . . . . . . . . 114 14. . . . . . . 13.CONTENTS 4 12 The ω-rule and uniform reflection. . .1 The language of GL . . . . . . . . . . . . . . . . . . .5 Closure of GL under substitution of provably equivalent formulas is provable in GL . . . . . .6 Strengthened proof that the closure of GL under substitution of provably equivalent formulas is provable in GL . . . . 110 . . 102 13 Provability logic: the system GL 13. . . . . . . . . . . . . . . . . . . . 98 12. . . .4 PA is Π1 -conservative over PAΠ2 ∪ {ConP A } . 13. . . . . 111 14 The fixed-point theorem for GL 114 14. . . . . . . . . . . 13.2 The fixed point theorem for GL . . . . . . . 13. 107 . . . .3 Some derivations in GL .2 PA proves that PA proves every instance of the G¨del sentence . . . . . 117 . 105 . . . . . . . . . .1 The notion of a sentence letter modalized in a sentence. .2 The axioms and inference rules of GL . . . . . 100 o 12. . 101 12. . . . 104 . . . . 13. . . Π1 -uniform reflection and consiso tency. . . .4 Closure of GL under substitution by provably equivalent formulas .

Some but not all of the expressions of the language constitute well-formed expressions.1 First-order formal languages with identity The notion of a formal language begins from the specification of a finite alphabet of symbols which are strung together (concatenated) to produce the expressions of the language. A sentence φ is a logical consequence of a set of sentences Γ. 0. symbolized by φ. The symbols of a first-order language are of three sorts.Lecture 0 Background: first-order logic and formal systems This unspoken lecture reviews essential results on first-order logic and formal systems as background to the development of G¨del’s incompleteness theorems. If Γ is empty. identity. and quantificational. logical. symbolized as Γ φ. and may include a third. Accordingly. according to our intended use of the formal language. o The notion of logical consequence is the starting point. The symbols of the alphabet of a formal language have no intrinsic meaning (they are purely formal symbols) but are chosen with the intention that they should be interpretable in certain ways (which in general does not rule out their being interpreted in other ways). function symbols and predicate or relation symbols. A rough characterization of this notion is the following. Functions and relations are of a particular ’arity’ 5 . we need to make precise the notion of a formal language and the notion of interpretation of a formal langauge. and syntactic. φ is logically valid. The logical symbols include two sorts: propositional. if and only if φ is true in every interpretation of the language of Γ ∪ {φ} in which all the sentences of Γ are true. non-logical. The non-logical are of two sorts. Γ φ reduces to the condition that φ is true in every interpretation of the language of φ. which is to say. Which expressions are well-formed is a matter of stipulation.

though there are some conventional choices. But the symbols can be anything and in particular. Identity is a two place relation which every object bears to itself and to no other object. and is sometimes written “ ” (called a California quantifier). conjunction. =). In our background logic we shall take as our primitives of propositional logic negation. Even in our background logic we will take (A ≡ B) to be an abbreviation for ((A ⊃ B) ∧ (B ⊃ A)). As you will know from a previous logic course. e. ≏. ternary. disjunction. but used often to be written “( )” (the brackets enclosing the variable of quantification). ⊃. but the second property is not easily expressed. written as ∀ and ∃. for which the symbols will be ∼. etc. ≈. common symbol for identity = (though others are sometimes used. We shall operate with the langauge as if it also contained symbols for conjunction.g. disjunction. e. there are choices to be made as to which are our primitive propositional functions and quantifiers. ∧. we could take as primitive just negation with any one of conjunction. means that we have a primitive symbol for identity. Quite apart from the shapes of the symbols. ∨.LECTURE 0 6 (unary. and implication and either the universal or the existential quantifier. they can be digits also used to generate numerals. and in our official formal language we shall take as primitive ∼.g. The first property is easily expressed: ∀v1 v1 = v1 . a zero-ary relation symbol represents a sentence. and ∀. disjunction. which we will take as the . terms in the language formulas in the language . binary. and so on). and both the universal and existential quantifiers as primitive. A zero-ary function symbol is a constant term. and the existential quantifiers. Stipulating that our background logic is first-order classical logic with identity. We shall in this course always have full first-order classical (as opposed to intuitionistic) predicate logic with identity as our background logic. which we will symbolize as ≡. but strictly (A ∨ B) will be an abbreviation for (∼ A ⊃ B). the universal quantifier is usually written “∀”. The shape of a symbol is completely arbitrary. and ⊃. in particular not expressed by ∀v1 ∀v2 (v1 = v2 ⊃ v1 = v2 ) (if v1 is a different object from v2 then v1 = v2 ). as we shall exploit. and implication. but having a symbol and first-order axioms for identity is not sufficient to make a logical system first-order logic with identity. There is also the connective “if and only if”.

I shall call this system LN D. but since in this course our background logic is always classical and not intuitionistic. in essentials. deductions consist of branching trees. with each of which is associated on the left a finite set of numbers (possibly empty) which are the numbers of the formulas that constitute the assumptions on which the formula in that line depends. truth of a first-order formula in an interpretation. ∀. a deduction consists of four columns. For ease of formulation on the page. the system here is. if it’s not an Assumption.J. the third column is a sequence of formulas and the second column is an enumeration of those formulas. The rules of inference come in pairs.] Deductions are generated by using the Rule of Assumption and fourteen Rules of Inference. and on the right notation indicating how and from what other formulas that formula was derived. standing either for Linear Natural Deduction (but in that case be aware that this usage has nothing to do with Girard’s usage in what he calls Linear Logic). I won’t mark the distinction. Viewed vertically rather than horizontally. 1 . ∃. or Lemmon Natural Deduction. To the left of A linear formulation of natural deduction was given by E.LECTURE 0 7 0.3 This system is based on natural deduction as developed by Gerhard Gentzen. Nelson.2 Interpretations of first-order formal languages with identity. There are no axioms. An Introduction Rule for a given logical constant results in a formula that has that logical constant as its main logical constant and is deduced from formulas that that do not have that logical constant as their main constant (though that constant may occur in a subformula). [Strictly I should call this system something like CLN D. The first formula of a Deduction is always an Assumption. as opposed to Intuitionistic. London. An Elimination Rule for a given logical constant deduces a formula that does not have that logical constant as its main logical constant from a formula that has that logical constant as its main logical constant. for Classical. logical validity and logical consequence A system of natural deduction for first-order logic with identity 0. an Introduction and an Elimination rule for each one of the seven logical constants ∧. In Gentzen-style systems of natural deduction. I give here a variant form of natural deduction in which deductions are linear1 . The middle two columns are a numbered sequence of formulas. We take (F ≡ G) to be defined as ((F ⊃ G) ∧ (G ⊃ F )). ∨. ⊃. and =. 1965. ∼. later formulas can also be Assumptions. Lemmon in Beginning Logic. as in that book. A deduction consists of a numbered sequence of formulas. Linear Natural Deduction.

so in an application of the rule of Assumption there is one number in the first column which is the same number as in the second column. The Assumptions on which those formulas depend. ∨-Elimination. The entry in the right-hand column gives the basis on which the formula in the third column at that line is introduced into the deduction. Four of the Rules of Inference. either as an Assumption or by one of the Rules of Inference. ⊃-Introduction. the entry in the fourth column for that line says what Rule of Inference has been used and numbers of the formulas to which that Rule of Inference has been applied.LECTURE 0 8 each numbered formula are the numbers of the Assumptions. on which that formula depends. logically valid. i. If the formula in the third column of a given line is introduced by a Rule of Inference. by the Soundness of the Rules of Inference. and ∃Elimination. as given by the numbers in the first column at the lines for those formulas. if any. An Assumption depends on itself. ∼-Introduction. discharge an Assumption. are gathered together as the Assumptions on which the formula that results from application of that Rule of Inference depends.e. so it is possible to arrive at a formula which depends on no assumptions. Such formulas are. ∧-Introduction assumptions A B A∪B numbering (a) (b) (c) formulas F G (F ∧ G) justifications [whatever] [whatever] (a) (b) ∧-Introduction ∧-Elimination1 assumptions A A numbering (a) (b) formulas (F ∧ G) F justifications [whatever] (a) ∧-Elimination1 ∧-Elimination2 assumptions A A numbering (a) (b) formulas (F ∧ G) G justifications [whatever] (a) ∧-Elimination1 ∨-Introduction1 .

A more natural formulation of logic in natural deduction is to take negation to be defined in terms of a primitive false sentence. sometimes symbolized as ⊥. in terms of formalizing informal maths talk it is more natural to take negation .LECTURE 0 assumptions A A numbering (a) (b) formulas F (F ∨ G) justifications [whatever] (a) ∨-Introduction1 9 ∨-Introduction2 assumptions A A numbering (a) (b) formulas G (F ∨ G) justifications [whatever] (a) ∨-Introduction1 ∨-Elimination assumptions A {b} B ∪ {b} {d} C ∪ {d} A∪B∪C numbering (a) (b) (c) (d) (e) (f) formulas (F ∨ G) F H G H H justifications [whatever] Assumption [whatever] Assumption [whatever] (a)(c)(e) ∨-Elimination ⊃-Introduction assumptions {a} A ∪ {a} A numbering (a) (b) (c) formulas F G (F ⊃ G) justifications Assumption [whatever] (a)(b) ⊃-Introduction ⊃-Elimination assumptions A B A∪B numbering (a) (b) (c) formulas (F ⊃ G) F G justifications [whatever] [whatever] (a)(b) ⊃-Elimination The Introduction and Elimination rules for negation in this system of rules are not as natural as those for the other logical connectives. by the equivalence ∼ ψ ≡ ψ ⊃⊥. However.

which we do here. if vi does not occur free in any formula enumerated by A ∀-Elimination assumptions A A numbering (a) (b) formulas ∀vi F (vi ) F (t) justifications [whatever] (a) ∀-Elimination. It is subject to the following Introduction and Elimination rules.LECTURE 0 10 as a primitive symbol. for t any term free for vi in F (vi ) ∃-Introduction assumptions A A numbering (a) (b) formulas F (t) ∃vi F (vi ) justifications [whatever] for t any term free for vi in F (vi ) (a) ∃-Introduction ∃-Elimination . ∼-Introduction assumptions {a } A ∪ {a} A numbering (a) (b) (c) formulas F (G∧ ∼ G) ∼F justifications Assumption [whatever] (a)(b) ∼-Introduction ∼-Elimination assumptions A A numbering (a) (b) formulas ∼∼ F F justifications [whatever] (a) ∼-Elimination ∀-Introduction assumptions A A numbering (a) (b) formulas F (vi ) ∀vi F (vi ) justifications [whatever] (a) ∀-Introduction.

we say that F is (logically) derivable from Γ.e. notated Γ ⊢LDN F . ⊢LDN (F ∨ ∼ F ) (1) (2) (2) (1)(2) (1) (1) (2) (3) (4) (5) ∼ (F ∨ ∼ F ) F (F ∨ ∼ F ) ((F ∨ ∼ F )∧ ∼ (F ∨ ∼ F )) ∼F Assumption Assumption (2) ∨-Introduction (1)(3) ∧-Introduction (2)(4) ∼-Introduction .LECTURE 0 assumptions A {b} B ∪ {b} A∪B numbering (a) (b) (c) (d) formulas ∃vi F (vi ) F (vi ) G G 11 justifications [whatever] Assumption [whatever] (a)(b)(c) ∃-Elimination. i. =-Introduction assumptions numbering (a) formulas t=t justifications =-Introduction. the substitution of t in ψ(x) must not result in ψ(t) having a different quantifier structure from that of ψ(x). if vi does not occur free in any formula enumerated by B General restriction on terms t in ∀-elimination and ∃-introduction: The term t must not contain any free variable which in ψ(x) is quantified by a quantifier whose scope in ψ(x) includes an occurrence of the variable x. if there is deduction from the Rule of Assumption and the 14 Rules of Inferences of LDN in which the last line has the numbered formula F and the assumptions on which F in that line depends on are exactly the formulas of Γ. or just Γ ⊢ F . t2 any terms [whatever] =-Elimination Definition 1 For F a formula and Γ a set of formulas. for t any term =-Elimination assumptions A B A∪B numbering (a) (b) (c) formulas t1 = t2 F (t1 ) F (t2 ) justifications [whatever] for t1 .

e. (iia) ((∀vi F (vi ) ∨ G) ≡ ∀vi (F (vi ) ∨ G)) (iia′ ) ((G ∨ ∀vi F (vi )) ≡ ∀vi (G ∨ F (vi ))) (iib) ((∃vi F (vi ) ∨ G) ≡ ∃vi (F (vi ) ∨ G)) (iib′ ) ((G ∨ ∃vi F (vi )) ≡ ∃vi (G ∨ F (vi ))) (iiia) ((∀vi F (vi ) ∧ G) ≡ ∀vi (F (vi ) ∧ G)) (iiia′ ) ((G ∧ ∀vi F (vi )) ≡ ∀vi (G ∧ F (vi ))) (iiib) ((∃vi F (vi ) ∧ G) ≡ ∃vi (F (vi ) ∧ G)) (iiib′ ) ((G ∧ ∃vi F (vi )) ≡ ∃vi (G ∧ F (vi ))) (iva) ((∀vi F (vi ) ⊃ G) ≡ ∃vi (F (vi ) ⊃ G)) (iva′ ) ((G ⊃ ∀vi F (vi )) ≡ ∀vi (G ⊃ F (vi ))) (ivb) ((∃vi F (vi ) ⊃ G) ≡ ∀vi (F (vi ) ⊃ G)) (ivb′ ) ((G ⊃ ∃vi F (vi )) ≡ ∃vi (G ⊃ F (vi ))) . if there are any. (ia) (∼ ∀vi F (vi ) ≡ ∃vi ∼ F (vi )) (ib) (∼ ∃vi F (vi ) ≡ ∀vi ∼ F (vi )) In the following the variable vi does not occur free in the formula G. in a way that results in a formula that is logically equivalent to the original formula. Remark. ∃v1 ∀v2 v1 = v2 is not logically equivalent to ∃v2 ∀v2 v2 = v2 (the latter is logically valid while the former is not). Lemma 2 (prenex equivalences) The following formulas are logically valid.g. resting on no Assumption. with Tautology as the Justification. There are restrictions on logically equivalent changes of bound variable.LECTURE 0 (1) (1) (6) (7) (8) (9) (F ∨ ∼ F ) ((F ∨ ∼ F )∧ ∼ (F ∨ ∼ F )) ∼∼ (F ∨ ∼ F ) (F ∨ ∼ F ) (5) ∨-Introduction (1)(6) ∧-Introduction (1)(7) ∼-Introduction (8) ∼-Elimination 12 A modification to make LND more readily usable: Any truth functional tautology can be written down as a line of a derivation. e.g. any variable of quantification can be changed to any one of infinitely many other variables while preserving the free variables of the formula.4 Prenex normal form Lemma 1 (change of quantified variable) For any formula in a first-order langauge. (1) (F ∨ ∼ F ) Tautology 0.

all its quantifiers. particular vi (in the case when the quantification ∀vi of F is not vacuous. (v) F is of the form ∀vi G. (ii) F is of the form (G ⊃ H). (iii) F is of the form (G ∧ H). that for v1 not free in G. the negation sign can be pushed past all the quantifiers. Then F is logically equivalent to ∀vi G∗ . for example. e. G = ∀v1 v1 = v1 . those from G∗ changing to the other quantifier. If G∗ has quantifiers. By Lemma 1 and (iva). Note that the Prenex Normal Form Theorem holds only on the assumption that all domains of interpretation are non-empty. which is in prenex normal form. (∀v1 F (v1 ) ⊃ G) is true in the empty domain. 13 Theorem 3 (Prenex Normal Form) Each formula F in a first-order language with ∀ and ∃ is logically equivalent to a formula F ∗ whose quantifiers all occur as a string at the beginning of the formula. and G true in the empty domain (e. and such that F ∗ contains exactly the same free variables as F does. (iv) F is of the form (G ∨ H). If G∗ has no quantifiers. changing each to the other quantifier in the process. Induction steps: (i) F is of the form ∼ G. Same argument as for (v). By induction hypothesis G has a prenex normal form G∗ . (∀v1 F (v1 ) ⊃ ∀v2 G(v2 ) has as prenex normal form both ∃v1 ∀v2 (F (v1 ) ⊃ G(v2 )) and ∀v2 ∃v1 (F (v1 ) ⊃ G(v2 )). namely none. and (iva′ ). Base case: F is atomic. By Induction Hypothesis G and H have logically equivalent prenex normal forms G∗ and H ∗ . Note that prenex normal forms are not in general unique. occur in a string at the beginning of the formula. with no changes of quantifiers as the quantifiers are put into prenex form. Then by logical equivalences (ia) and (ib) from Lemma 2. Otherwise we have. as in the base case.g. they occur in a string at the beginning of the formula. . (iv′ ) of Lemma 2. we are done. but ∃v1 (F (v1 ) ⊃ G)) is false. (ivb). Then F is logically equivalent to (G∗ ⊃ H ∗ ). or equally G = ∀v1 ∼ v1 = v1 ).g. Then F has no quantifiers so. since every existentially quantified statement is false in the empty domain. These two cases are as (ii) except simpler. the quantifiers of G∗ and H ∗ can be pulled out into prenx normal form. since G is true. vacuously. Then by induction hypothesis G has prenex normal form G∗ which has the same free variables as G has. Proof The proof is by induction on the number of quantifiers in F . (vi) F is of the form ∃vi G.LECTURE 0 Proof.

3 0.LECTURE 0 14 0. Comparison between the completeness theorem for first-order logical consequence and the incompleteness theorem for truth in arithmetic. We say that S1 is a subsystem of S2 if for every formula φ in L1 .5 The Compactness Theorem for first-order logic The existence of non-standard models of the truths of arithmetic Completeness of other systems of derivation with respect to first-order logical consequence Definition 2 Let S1 and S2 be formal systems such that the language L1 of S1 is a sub-language of the language L2 of S2 .5.1 Model theory and proof theory 0. in 1930 and 1931. On the other hand there is an intrinsic completeness theorem. the second his o Habilitation thesis. if S1 ⊢ φ.5. namely completeness of a system which consists of exactly those axioms and rules of inference needed for the proof of the completeness theorem. one for each different complete formal system for first-order logical consequence.5 Completeness of a system of natural deduction with respect to first-order logical consequence Lindenbaum’s Lemma The Completeness Theorem 0. thesis. While we talk about the completeness theorem for first order logic.5. They were written under the nominal supervision of the very notable mathematician Hans Hahn.4.1 0.5. Both these results are due to G¨del.) o 0. there are actually many completeness theorems.4 0. . (The first was his Ph.5. but G¨del was essentially working on his own. then S2 ⊢ φ Lemma 4 If S1 is a complete system for first-order logical consequence and S1 is a subsystem of S2 . then S2 is complete for first-order logical consequence.2 The completeness theorem for a given formal system of first-order logic means that a model theoretic argument for a logical consequence (as above) establishes the existence of a formal derivation of that logical consequence in the system whose completeness has been proved without actually finding a derivation.D.

In 1918 he declared that we must make the concept of specifically mathematical proof itself into an object of investigation (Hilbert [2]). corresponding to observation statements in science. 12 October 2010) 1.Lecture 1 Introduction: a weak form of G¨del’s First Incompleteness o Theorem. the symbols and expressions of a language for arithmetic LE . G¨del numbering o of the expressions of LE (Tuesday. Hilbert formulated the distinction between finitary and infintary mathematics. but the claim that exponentiation to the 15 . A calculation such as 27 = 128 is finitary. The paradigm of finitary mathematics is arithmetical calculation. Finitary mathematics is mathematical bedrock.1 Introduction: a weak form of G¨del’s First o Incompleteness Theorem The context of G¨del’s discovery of the phenomenon of formal incompleteness is o David Hilbert’s programme for giving mathematics a secure foundation by establishing the consistency of systems formalizing it.

On the other hand. the statement that a specific quadruple of numbers a. (∃v1 ≤ t)F (v1 ) and (∀v1 ≤ t)F (v1 ) are general finitary statements. For t a numerical term (a numeral or a composition of arithmetical functions applied to numerals). On the other hand. Particular finitary statements are decided by computations.LECTURE 1 16 power 2 always yields a value. In particular. belongs to finitary mathematics. the negation of a general finitary statement cannot be expressed as a general finitary statement. belongs to finitary mathematics. Hilbert noted that general finitary statements are not closed under negation. If t is a free variable or a composition of arithmetical functions applied to one or more variables. General finitary statements contain free variables. For example. not just the symbols for numbers. that a particular formal proof is or is not a proof of a particular statement. d is a counterexample. a formalized proof. but to say (falsely) that Fermat’s Last Theorem is false requires existential quantification. 7 × 5 = 35.e. like a numeral. e. initial segment of the natural numbers. p. and can be thought of as a template for particular finitary statements that result by substitution of numerals for the free variables. ∃n∃x∃y∃z(n > 2 ⊃ xn + y n = z n ). b. which we abbreviate (∃v1 ≤ v2 )F (v1 ). for t a term in the language of arithmetic. (∃v1 ≤ t)F (v1 ) and (∀v1 ≤ t)F (v1 ) are particular finitary statements if v1 is the only free variable in F (v1 ). . (a > 2 ∧ an + bn = cn ). is expressible using (apparently) unbounded quantification by. quantification over a bounded. for example x + y = y + x. i. in the case of universal quantification.g.g. quantification over the infinite domain of natural numbers is infinitary. ∀v1 (v1 ≤ t ⊃ F (v1 )). is a particular finitary statement. numerals and terms built up from numerals and symbols for arithmetical operations. 471.e. general and particular (though he did not introduce terminology for this distinction). e. bounded quantification on the variable v1 . i. ∀x∀y x + y = y + x and ∀n∀x∀y∀z(n > 2 ⊃ xn + y n = z n ) are infinitary. i. p. finitary statements. (n > 2 ⊃ xn + y n = z n ). is a concrete and surveyable object. c. and in the case of existential quantification. For F (v1 ) a general finitary statement with free variable v1 . and (n > 2 ⊃ xn + y n = z n ). However. i.e.e. ∀x∃y(2x = y) is infinitary. which is finitary. as Hilbert recognized. which is finite. and 210 = 1024 and truth functional combinations of them (the truth values of such combinations being computable from the truth values of the component statements). which we abbreviate as (∀v1 ≤ t)F (v1 ). Fermat’s Last Theorem is expressible as a general finitary statement. Statements about particular formal proofs are. and more generally.) Hilbert recognized two sorts of finitary statements. Hilbert’s deep insight was to recognize that the formal manipulation of all symbols. i. 383 and also in [4]. ([3].e. ∃v1 (v1 ≤ t ∧ F (v1 )).

from the assumption that everything provable in a system S is true. such that for every formula X in the language of S. they could be taken to be numerical expressions. i.e. From these results and on the assumption that everything provable in a given system S is true (about the natural numbers) (a strong assumption. there is a formula P r(v1 ) with one free variable. i. Definition 4 For formal languages that have a numeral for each natural number. We shall establish the existence of such a formula for a particular formal system of arithmetic in Lecture 4. for metamathematical considerations it does not matter what objects are chosen as primitive signs. As G¨del put this o point in (1931). which implies. the system must be able to ”talk” about numbers.e. but it is illuminating to consider this simple case). namely that o formal proofs can be literally identified with natural numbers. Numbers assigned to formulas of a formal language in this way are called G¨del o numbers. we denote by n the numeral for the natural number n. there must be for each natural number a formal numeral in the language of the system that denotes that numbers. I mean true in the usual (intended) structure consisting of the domain of natural numbers with the usual arithmetical functions and relations on the natural numbers (also known as the standard model).e. we map the primitive signs one-to-one onto some natural numbers. much stronger than is needed to establish incompleteness. o i. S. S G. rather than merely like them. Definition 3 We denote the G¨del number of a formula F by F . i. We shall establish this result in Lecture 3.e for formula F (v1 ) there is a sentence D such that the equivalence (D ≡ F ( D )) is true. Of course. . and we shall assign natural numbers to this use. What G¨del showed is that the property of being (the G¨del number of) a o o provable formula is expressible within any system which can express basic arithmetic. that S ∼ G. Definition 5 Whenever in these notes I speak of a sentence in the specified language of arithmetic as being true. it is easy to see that for G such that (G ≡∼ P r( G )). o To carry out the arithmetization of syntax. G¨del also showed that for any o formula with one free variable (in particular a formula that expresses the property of being the G¨del number of an unprovable formula) there is a diagonal sentence.LECTURE 1 17 Hilbert missed something about his insight which G¨del realized. that is. and also thereby that G is true. S ⊢ X if and only if P r( X ) is true. in the language of a formal system for arithmetic.

and so by RAA establishes that G is not provable. and unprovability of the G¨del sentence holds from the much o weaker assumption that S is consistent. (ii). to be made precise by the notion of Σ0 -arithmetic. and notes that “The purpose of carrying out the above proof with full precision in what follows is. soundness of the system with respect to truth in arithmetic. Remarks about this result: (1) This is a weak version of the first G¨del incompleteness theorem since.” (van Heijenoort Sourcebook p. In particular we don’t need exponentiation. S ∼ G. S G. while o assumptions (i) and (ii) are provable for weak systems of arithmetic. (7) From (6) and (i). G is false. G¨del sketches the proof of this weak form o of the First Incompleteness Theorem in section 1 of his 1931 paper. that unlike provability in a formal system. S ⊢ X if and only if the sentence P r( X ) is true. Then S G.LECTURE 1 18 Theorem 5 (weak form of G¨del’s first incompleteness theorem) Let S be o a theory such that for each natural number n there is a numeral n in the language of S. assumption (iii). The argument does not lead to a contradiction since it starts from the assumption that G is provable. and (iii)). (ii) there is an assignment of numbers to the formulas of the language of S and a formula P r(v1 ) in the language of S such that for each formula X. but then also exponentiation and . P r( G ) is false. (5) Since (4) contradicts (1). This shows that arithmetized syntax is a proper sub-part of what Hilbert meant by finitist mathematics. i. to prove assumptions (i) and (ii). P r( G ) is true. we have by reductio ad absurdum that S G (from assumptions (i). Hilbert never gave a precise characterization of finitist mathematics. 599). as we shall see. essentially just computations with addition and multiplication. truth in a language of arithmetic cannot be expressed in the language. G is true. so in particular S ⊢ G if and only if the sentence P r( G ) is true.e. and assume that (i) there is a sentence G such that the sentence (G ≡∼ P r( G )) is true. (iii) every theorem of S is true. is a strong (highly nonfinitistic) assumption. G is true. so plus and times. (2) In his introductory section G¨del notes that this argument is “closely reo lated” to the argument for the “Liar” paradox. Proof. to replace the second of the assumptions just mentioned [every provable formula is true in the interpretation considered] by a purely formal and much weaker one. (4) From (3) and (iii). (3) From (2) and (i). Use of the Liar paradox also shows. (6) From (5) and (ii). and S ∼ G. (3) We shall establish how much basic arithmetic is required for arithmetization of syntax. (2) Then by (ii). but it is clear that it includes all primitive recursive functions. among other things. (8) From (7) and (iii). (1) Suppose that S ⊢ G.

a proof that S assumption of consistency is best possible. X from the Proof. Remark. That the G¨del sentence G for a system S is not refutable.e. it proves everything. both addition and multiplication are needed for incompleteness: there is a complete theory of zero. so in particular S ⊢ X. to which numerical subscripts in tally notation. Proposition 6 For any system S and sentence X. For our formal language for arithmetic. In the previous hundred years the independence of Euclid’s fifth postulate from the other postulates of geometry had been established. If S is inconsistent. i. ∼ G o is not provable. successor. 1 . requires a stronger condition on S than consistency but a condition much weaker than the soundness of S is sufficient. but the next sentence becomes just about unreadable if use vs mention is spelled out in this way.LECTURE 1 19 so on. On the other had.2 The symbols and expressions of a language for arithmetic LE A formal language is generated by combining symbols from a specified finite alphabet. The result concerning the fifth postulate was established by the construction of models. That distinction is easily but cumbersomely dealt with by using quotation marks. per se. The other difference o is even more fundamental. this consists of the following 13 symbols: 0 ′ ( ) f ′ v ∼ ⊃ ∀ = ≤ ♯ These formal symbols will be used with the following intended meanings: The symbol 0 denotes the natural number zero1 . The symbol ′ denotes the successor function The symbols ( and ) are left and right brackets The symbols f and v are for functions and variables. though not demonstrable in the system for which it is constructed. G¨del’s result differs o from this earlier one in two crucial ways. which is fine.e. The fifth postulate is neither true nor false. (4) The independence of the G¨del sentence from formal arithmetic was unpreceo dented. iterations of the subscript ′ are attached. The G¨del o sentence is demonstrably true. One was the technique used. It is true in Euclidean geometry and false in non-Euclidean geometries. as in [7]. in which case this given sentence would read: “The symbol ’0’ denotes the natural number zero”. though fussy. The G¨del result is purely syntactic (exploiting Hilbert’s insight). i. 1. and addition (Presburger Arithmetic). The strings of Note that in this sentence I am being casual about the distinction between use and mention.

. u. o E =df the G¨del number of expression E. then the result of writing Ei directly followed by Ej . Definition 7 (notation for expressions and G¨del numbers) En =df the exo pression with G¨del number n. There are an infinity of variables v′ . and exponentiation. z. which we will usually write as v1 . o . The symbol ♯ will be used to mark breaks between strings of symbols that are terms and formulas of the language (to be defined in the next lecture) in sequences of terms and formulas. multiplication. f′′′ will denote the functions addition. which we call the concatenation of Ei and Ej and symbolize as Ei Ej . The symbols = and ≤ are for the two-place relations of equality and less than or equals. following Smullyan following Quine. which we will write informally as +. On our method each number is the G¨del number of an expression. An expression in the language is (almost) any finite string of these symbols. respectively. v2 . The set of expressions for the language LE is specified by the following recursive definition. . The symbol for the universal quantifier is ∀. v′′′ . This can be done in infinitely o many ways. . v3 . Having every number be o o a G¨del number makes the formulation of some results a little simpler but is not o essential. we will write vi .3 . Definition 6 (expressions) basis: Each one of the symbols 0 ′ ( ) f ′ v ∼ ⊃ ∀ = ≤ ♯ is by itself an expression. is an expression. and exp or xy in the usual notation. v. The symbol for the propositional connectives negation and implication are ∼ and ⊃. vj etc or use informal variable letters x. G¨del’s original method involved coding by exponents of prime o factors. The way we shall do it. while o on G¨del’s method not every number is a G¨del number. 1. and Ei = ′ . y. . .LECTURE 1 20 symbols f′ . ·. . If we want to signify a variable without specifying which variable it is. f′′ . . v′′ . makes the link between G¨del numbering of expressions as strings of formal symbols particuo larly transparent. w. Remark: It’s for a technical reason (to do with our choice of G¨del numbering) o that we excluded from the class of expressions as here specified strings of more than one symbol that begin with the symbol ′ . G¨del numbering of the expressions of LE o We assign G¨del numbers to the expressions of LE . recursion: If Ei and Ej are expressions..

n = 0. m ∗10 n = 590 and n ∗10 m = 059 = 59. 21 We are used to the idea that numbers are denoted by numerals and that numerals are not the same thing as numbers. 3. Non-associativity only arises when the middle value is 0. Remark : As illustrated by these examples. i. V. with 0 and 1 represented by current off and current on. The formal numeral for the number n is the expression 0′ . 11. ǫ. Examples: For m = 673. 5. Base 2 is used in machine code for computers. we denote by m ∗b n the number designated by the base b numeral that results from concatenating the base b numeral for m with the base b numeral for n. respectively. That the system of numerals in common use is base 10 is presumably down to the contingent fact (it could have been otherwise) that human beings have 10 fingers. Any other number greater or equal to 2 gives a perfectly good numeral system with that base. i. 4 . concatenation of the symbol 0 with n-many concatenations of the symbol ′ .LECTURE 1 Corollary 7 (of Definition 7) En = n. . Definition 9 (assignment of G¨del numbers to expressions) By recursion over o the recursive definition of expressions. In the formal language for arithmetic we shall be using the numerals for numbers use a tally notation. It is also not associative. 2. Base case: The assignment of numbers to symbols is specified by . The number we write as 15 in base 10 we write as 1111 in base 2 and as 13 in base 12. ∗b is not commutative. For m = 59. x ∗b y ∗b z = (x ∗b y) ∗b z.e. We assign to these thirteen symbols the numbers denoted by the thirteen digits of base 13 notation. Note that ∗b is a function mapping pairs of natural numbers to natural numbers. we adopt the common convention of association to the left. o Definition 8 (concatenation of base b numerals) For natural numbers m and n. The role of b in this function is to specify the method of calculating this function. IV. II. III. . and δ. We assign G¨del numbers to expressions by first stipulating the G¨del numbers o o of the symbols. (17∗b 0)∗b 59 = 17059 = 1759 = 17∗b (0∗b 59). and 12 (as we write them in base 10) are taken to be η.g.e. The crucial property of the Arabic numerals is that they are constructed on a place-value system with a base of 10. and that natural numbers are not intrinsically in base b or any other base notation. rather n than place values. e. m ∗10 n = 67332 and n ∗10 m = 32673. The Roman numerals for the first five non-zero natural numbers are I. but since we will include 0 as a G¨del number we o cannot suppress parentheses in multiple computations with ∗b except by adopting a convention for reinstating them. The following function plays a key role in our chosen system of G¨del numbering. where the digits for 10. ′ . n = 32 (written in base 10). while the Arabic numerals are 1.

′ . . We can use base 10 and the operation ∗10 even with thirteen symbols by. 8999. for example. we know that if an 8 or a 9 occurs its base 10 notation it must occur within a string of the form 89. ′ = 0.′ = 0. It is in order for each expression to have a unique G¨del number that we stipulated above o that the class of expressions does not contain strings of more than one symbol that begin with the prime symbol ′ . But we can o effectively tell the ones that are.e.LECTURE 1 0 1 ( 0 2 ′ 22 f 4 v 5 6 ∼ 7 ⊃ 8 ∀ 9 = η ≤ ǫ ♯ δ ) 3 ′ Recursion: For expressions X and Y . 899999. . assigning the thirteen symbols respectively the following numbers (written in base 10): 0 1 ( 0 2 ′ ) 3 f 4 v 5 6 ′ ∼ 7 ⊃ 89 ∀ = ≤ ♯ 899 8999 89999 899999 Of course on this assignment not every number is a G¨del number. There is a technical advantage in taking the base b to be a prime number but it is by no means essential. X Y = X ∗13 Y By these stipulations. and ′ ∀ = 09 = 9 = ∀ . 89999. and we know which symbol is coded by counting the number of 9s in that string. . i. 899.

. . v′′ . expressibility of diagonal substitution in the language LE (Wednesday. formal variables as we have defined them enclosed in brackets (p. . and if it contains one or more free variables. . Unique readability for terms within expressions does not require enclosing variables within 23 .1. (v′′ )..e. . is a variable. the concatenation of E and the subscript symbol ‘′ ’. and if the expression E is a variable then the expression E′ .e. 13 October 2010) 2. (v′′′ ). We will abbreviate the string of symbols consisting of the formal variable symbol v followed by n subscripts as vn . 15). Remark: Formal variables are expressions of the form v′ . denote a number if the term does not contain a free variable. i. v′′′ . Further remark: Smullyan’s formal variables are of the form (v′ ). Definition 10 (Variables) v′ is a variable. on the intended interpretation of LE as a language for arithmetic. then the term that results from substituting a numerals for each variable denotes a number. .1 Terms and formulas of the language LE Terms Terms in the formal language LE are expressions that. .1 2.Lecture 2 Terms and formulas of the language LE . i.

respectively). namely $v_{_{’’}}$\hspace{-. and f′′′ . . Nonetheless. then (t1 f′ t2 ). n + 1 is n′ . in which ′ is correctly concatenated at the end. (ii) Expression E1 occurs in expression E2 if there exist expressions E3 . Notation: For natural number n we write n for the numeral that denotes the number n. multiplication. If this was v2 . 0′′′ . Definition 11 (Numerals) The symbol 0 is a numeral (which denotes the number zero). E4 such that E2 = E3 E1 E4 or E2 = E3 E1 or E2 = E1 E4 Definition 14 (constant terms) A term in which no variable occurs is called a constant term. the numeral for the number n + 1 is the concatenation of the numeral for the number n and the symbol ′ . Evidently Smullyan produces e. . so the expressions 0. The motivation for this unnecessary use of brackets might be an artefact of what’s simple to write in LaTeX. 0′ .LECTURE 2 24 brackets in this way. or a closed term.. more complicated LaTeX code (also with compounded subscript command to lower the apostrophe further so it looks more convincing as a subscript) produces the required concatenation with variables not enclosed in brackets. . are formal names of the natural numbers 0.] Definition 13 (expression E1 occurs in expression E2 ) (i) Every expression occurs in itself.g. $(v_{’’})^{\prime}$ produces (v′′ )′ . [Recall that the notations in LE for addition. . 2. If the expression E is a numeral then the expression E ′ is a numeral (which denotes the successor of the number denoted by E.g. . 3. respectively.12ex}$^{\prime}$. f′ . and exponentiation are. (t1 f′′ t2 ). Definition 12 (Terms) Among expressions of LE . . 7 = 0′′′′′′′ . and (t1 f′′′ t2 ) are terms. which compiles to the required concatenation of symbols v′′ ′ . . On the other hand. i. Induction clauses: If t is a term. Corollary 8 (of Definition 11 and Notation) For any natural number n. then t′ is a term. the substring v′′ of his variable (v′′ ) by the LaTeX code v_{’’}.e. the class of terms is specified by the following recursive definition: Base clause: Each variable and each numeral is a term. in which the ′ symbol occurs interposed over the string v′′ rather than concatenated at the end of that string. f′′ . then the seemingly natural way to write LaTeX code for the concatenation of ′ v2 with ′ would be $v_{’’}^{\prime}$. If t1 and t2 are terms. 1. but this produces v′′ . 0′′ . e.

if vi occurs free in F . where t1 and t2 are terms. and existential quantification and expressions in terms of ∼. Definition 20 (open formulas) A formula with one or more free variables is an open formula Notation: We write F (vi ) to signify a formula in which the variable vi occurs free. (A ∨ B) =df (∼ A ⊃ B). . the expression ∀vi F is a formula. Other unspecified variables may occur free as well unless we stipulate that vi is the only variable free in F (vi ). Induction clauses: If F and G are formulas. [Note that the formula (F ⊃ G) is enclosed in brackets.2 Formulas Definition 15 (Atomic formulas) An atomic formula is any expression of the form t1 = t2 or of the form t1 ≤ t2 . Definition 16 (Formulas) The class of formulas is specified by the following recursive definition: Base clause: Every atomic formula is a formula. In the latter case we may say that F (vi ) is a one-place formula. vik ) for a formula in which variables vi1 . if variable vi occurs free in F . ∃vi A =df ∼ ∀vi ∼ A. .3 Free and bound variables.] We will use logical equivalences between conjunction. and for every variable vi . and ∀ as abbreviations. possibly with other free variables unless we stipulate that these are the only ones. then ∼ F and (F ⊃ G) are formulas. . then vi occurs free in ∼ F and in (F ⊃ G) and (G ⊃ F ). . Definition 17 For formulas A and B. vik occur free. 2.e. (ii) For F and G formulas. . . (A ∧ B) =df ∼ (A ⊃∼ B). ⊃.1. open formulas and sentences Definition 18 (a variable occurs free in a formula) (i) If F is an atomic formula.1. Definition 19 (bound variables) A variable occurs bound in a formula F iff it occurs in F and does not occur free in F . all occurrences of variables in F are free.LECTURE 2 25 2. . but that the other two formation rules do not introduce new brackets. Similarly we write F (vi1 . i. disjunction. vi is free in ∀vj F iff j = i. (iii) For any formula F .

] Definition 21 (closed formulas a. F (n) is a sentence while F (n) is not. . . “F (v1 ) is to be any formula at all (it may contain free variables other than v1 )” (Smullyan. . which also then allows us under the same convention to stipulate in a given situation that there are no other free variables.2 Designation by terms in LE . If vi is the only variable free in F (vi ). nk ) signifies the result of substituting the numerals n1 . nk ) is unambiguous. truth of sentences of LE . . . by the following recursive specification: . e. . vik . . . . we don’t need to stipulate which number is substituted for which variable. . 29).a. then F (n) is a sentence. nk ) is a sentence. . . .LECTURE 2 26 [Note that this convention on possible occurrence of free variables other than those explicitly shown is different from Smullyan’s: “We write F (vi1 . e. and that statement will be true if and only the sentence F (n) is true. .2. Note that for F (v1 ) a formula with one free variable. p.” (p. .k. . . 2. For F (v1 .1 On the intended interpretation of the formal language LE each constant term designates a particular natural number. vik ) for any formula in which vi1 . . also called a sentence. Notation: Substitution of numerals for free variables in formulas: We write F (n) to signify the result of substituting the numeral n for every free occurrence of vi in the open formula F (vi ). . vik ) of vi1 . . . Definition 22 (regular open formulas) An open formula is said to be regular if its k-many free variables are the first k variables. vk ) a regular open formula. nk for all free occurrences in F (vi1 . the expression F (n1 . . .g. and expressibility of sets and relations of natural numbers by formulas of LE Designation 2. F (n1 . .g. . vik are the only free variables. . . . The latter can be construed as shorthand for the statement that the open formula F (v1 ) is satisfied by the number n. . Smullyan’s convention has in those situations to be violated. . vik ). 16). . . . then F (n1 . . For numbers n1 . in stating the Induction axioms. . . vik are the only variables that occur free in F (vi1 . . If vi1 . sentences) A formula with no free variables is a closed formula. . It seems to me more coherent for the convention to allow other variables. . in which we need to allow for the possibility of other free variables than those explicitly shown. Since there are situations. respectively. nk . . . .

Definition 25 (Arithmetical) A relation or a function is Arithmetical if it is expressible by a formula in LE . . P owb (v1 ) iff ∃v2 (v1 = bv2 ).3 Expressibility Definition 23 (expressibility of relations) A formula F (v1 . vn ) = vn+1 is expressible in LE . (c1 f′′ c2 ) designates the product of n1 and n2 .2. . Lemma 9 For a fixed number b ≥ 2. vn ) in LE is said to express a relation R ⊆ Nn iff for every n-tuple < k1 . . is Arithmetical. which expression we abbreviate as P owb (v1 ). or more formally P owb (v′ ) iff ∼ ∀v′′ ∼ v′ = (bf′′′ v′′ ). k n ) is true iff < k1 . . and (c1 f′′′ c2 ) designates n1 raised to the power n2 . . . In such case the relation R is said to be expressible in LE . 2. Definition 24 (expressibility of functions) A function f (v1 . . vn ) : Nn → N is expressible in LE iff the relation f (v1 . . . . . . and I will take it as known informally what it means for a formula in the language of arithmetic to be true in the structure of the natural numbers. .LECTURE 2 27 (i) the numeral n designates the number n. . kn > of natural numbers the sentence F (k 1 . . . . Definition 26 (arithmetical) A relation or a function is arithmetical [with a lower-case “a”] iff it is expressible by a formula in LE in which the expression f′′′ (for exponentiation) does not occur. (ii) If constant term c designates n. . .2 Truth Truth for a sentence of LE in the structure of the natural numbers (the intended interpretation) can be defined by recursion over the recursive generation of the sentence in the usual way. then (c1 f′ c2 ) designates the sum of n1 and n2 .2. the condition that v1 is a power of b. . . 2. kn >∈ R. . 2.3 Concatenation of numbers in a given base notation is Arithmetical. For none of the results in this course do we require a formal definition of truth. . if constant terms c1 and c2 designate n1 and n2 respectively. . Proof. . then constant term c′ designates the next natural number after n.

The first disjunct takes care of this case. ((v1 = 0∧v2 = b)∨(v1 = 0∧P owb (v2 )∧v1 < v2 ∧∀v3 ((P owb (v3 )∧v1 < v3 ) ⊃ v2 ≤ v3 ))). then the length of v1 (in base b notation) is the least power of b > v1 . ℓ(v1 ) = 1 and b1 = b. This relation is expressed by the following condition on v1 and v2 . This equivalence is seen as follows: If v1 = 0. and their arithmetization The operation of substituting a numeral for a free variable lies at the heart of the construction of a ‘self referential’ arithmetical sentence (‘This sentence is not provable in the given formal system’). For example. This condition is equivalent to ∃v4 (bℓb (v2 ) = v4 ∧ ((v1 · v4 ) + v2 ) = v3 ). To describe the formula F (n) obtained by substituting the numeral n for the free variable vi in the formula F (vi ) is complicated (requiring recursion over the logical complexity of formulas). and for any v3 such that bv3 > v1 . .4 Substitution and diagonal functions. i. We follow Smullyan in utilizing a trick due to Tarski by which we construct a formula F [n] which is not the same formula as F (n) but is logically equivalent to it for which there is a simple general description that does not depend on the logical complexity of the formula F (vi ). e. the two-place relation bℓb (v1 ) = v2 is Arithmetical Proof. Theorem 11 For any number b ≥ 2.t.] That v2 is the least power of b greater than v1 is expressed by the two conditions that bv2 > v1 . By Lemma 10. 2. F (v3 )—µv3 (10v3 > 0) = 0 since 100 = 1 > 0. but— writing µv3 F (v3 ) for the least v3 s. 1570 ∗10 365 = 1570365 = 1570000 + 365 = 1570 · 103 + 365 = 1570 · 10ℓ10 (365) + 365. ℓ10 (935) = 3. If v1 = 0. The above condition is expressible in LE by Lemma 9 and the fact that v1 < v2 is equivalent to (v1 ≤ v2 ∧ ∼ v1 = v2 ). [Note that we need to treat these two cases separately.e. v2 ≤ v3 . the number of digits in the base b notation of n. Proof. since ℓ(0) = 1.g. the relation v1 ∗b v2 = v3 is Arithmetical. this relation is Arithmetical. The relation v1 ∗b v2 = v3 is expressed by the condition that v1 · bℓb (v2 ) + v2 = v3 .LECTURE 2 28 Lemma 10 For ℓb (n) the length of the base b notation for n. and 103 = 1000 is the least power of 10 > 935.

By universal generalization. so that whole expression is. Proof. v2 ) is the G¨del number of a formula logically o equivalent to the substitution of the numeral of the number v2 into the expression Ev1 when Ev1 is a formula in which v1 occurs free.e. Ev1 is a formula in which the variable v1 occurs free. 0. which is 13v2 . . we must calculate the G¨del numbers of the numerals.e. (n = n ⊃ F (n)). o v2 o Proof. F [n] =df ∀v1 (v1 = n ⊃ F (v1 )). . by v2 concatenation. the symbol ′ is assigned the number 0. Ev1 . is to show that the function o s(v1 . .LECTURE 2 29 Definition 27 (quasi-substitution) For F (v1 ) any formula of LE with one free variable and n any numeral. The symbol 0 is assigned G¨del number 1. the expression whose G¨del numbers if v1 . Before we can establish that the two-place function (three-place relation) s(v1 . The numeral v2 is the expression 0′ . showing that syntactic operations on expressions of LE can be reflected into arithmetically definable operations on their G¨del numbers. and for other values it is not. . ′ . i. v2 ) = ∀v1 (v1 = v2 ⊃ Ev1 ) is Arithmetical. (ii) Suppose F (n). Note that for some values of v1 . v2 ) is the G¨del number of an expression o which is not a formula—a ‘don’t care’ case. Since n = n is logically valid. v2 ) = v3 is Arithmetical. F (n). Indeed for some values of v1 . o Lemma 14 The G¨del number of the numeral v 2 is 13v2 (in our base-13 assignment o of G¨del numbers ). i. Then by substitutivity of identity. If Ev1 is a formula in which v1 does not occur free then s(v1 . Lemma 12 (∀v1 (v1 = n ⊃ F (v1 )) ≡ F (n)) is logically valid. in which case s(v1 . Proof. assigned the number written in base-13 notation as 1 0 . Then by universal instantiation. Exercise. i. o will not be a formula. ∀v1 (v1 = n ⊃ F (v1 )).e. The value of the function s(v1 . . Quasi-substitution is logically equivalent to substitution. Note We could also have defined F [n] as ∃v1 (v1 = n ∧ F (v1 )) since: Lemma 13 (∀v1 (v1 = n ⊃ F (v1 )) ≡ ∃v1 (v1 = n ∧ F (v1 ))) is logically valid. v2 ) is the G¨del number of a formula that has nothing to do o with substitution of the numeral for v2 into it—another ‘don’t care’ case. Our first step in the arithmetization of syntax. (v1 = n ⊃ F (v1 )). by modus ponens. (i) Suppose ∀v1 (v1 = n ⊃ F (v1 )).

Expressing this argument from Theorem 11 more strictly. i. whose base o 10 notation is 899652658999). n) = k. n. s(m. For all numbers m. i. a particular number whose base 13 notation— given our assignment of base 13 digits to the symbols of our language—is 965265η (or if we use the base 10 G¨del numbering also given in Lecture 1. v3 ). Remark. ) = 3. k. v2 . Ev1 = v1 . Since by Definition 27. Let D(v1 . . S(m. ⊃ = 8. Given that v2 = 13v2 . S(m.e. Definition 28 means that d(v1) = Ev1 [v1 ] . there is a formula S(v1 . v2 ) = ∀v1 (v1 = v2 ⊃ Ev1 ) is Arithmetical. v2 ) be the formula that results by substituting v1 for v2 and v2 for v3 in S(v1 . s(v1 . 30 Proof. m) = d(m). v1 ). n. Corollary 16 (corollary to the proof of Theorem 15) The relation d(v1 ) = v2 is Arithmetical. v2 .e. ∀v1 (v1 = x ⊃ Ev1 ) =df Ev1 [v1 ]. Proof. k) is true iff s(m. Let k = ∀v1 (v1 = . m. n) is true iff d(m) = n. v3 ) such that for all numbers m. d(v1 ) =df ∀v1 (v1 = v1 ⊃ Ev1 ) . So D(m. v2 ) = v3 iff ∃v4 (v4 = 13v2 ∧ v3 = k ∗ v4 ∗ 8 ∗ v1 ∗ 3). n) is true iff s(m. n.LECTURE 2 Theorem 15 The function s(v1 . m) = n. the formula needs to be the following: ∃v4 ∃v5 ∃v6 ∃v7 (v4 = 13v2 ∧ k ∗ v4 = v5 ∧ v5 ∗ 8 = v6 ∧ v6 ∗ v1 = v7 ∧ v7 ∗ 3 = v3 ) Definition 28 (diagonal substitution) The diagonal substitution function d(v1 ) is s(v1 . which by lefthand association of ∗13 and repeated use of Theorem 11 is Arithmetical. By the definition of d(v1 ).

(5) By (1). (4) By Lemma 12. (2) Let k =df ∀v2 (D(v1 . v2 ) ⊃ F (v2 )) ≡ ∀v2 (d(k) = v2 ⊃ F (v2 ))). (C ≡ ∀v2 (D(k. v2 ) ⊃ F (v2 )) . Proof. 31 . 19 October 2010) 3. (8) By (2) and (3) and Definition 28. (6) By Lemma 12. (∀v2 (D(k. (1) Let D(v1 . C = d(k) (fuller explanation below). (6). Explanation of step (8): k is the G¨del number of a formula with one free variable. (9) From (7) and (8) by substitutivity of identity. v2 ) ⊃ F (v2 )). v2 ) be the formula in LE from the proof of Corollary 16 that expresses the relation d(v1 ) = v2 .Lecture 3 The Diagonal Lemma. (5). v2 ) ⊃ F (v2 ))). (3) Let C =df ∀v1 (v1 = k ⊃ ∀v2 (D(v1 .1 The Diagonal Lemma Theorem 17 (The Diagonal Lemma) For each one-place formula F (v1 ) in LE . o C is the formula that results from the diagonal substitution of quasi-substituting into that formula its own G¨del number. (C ≡ F (d(k))). and the one-place function d(v1 ) generates o from the G¨del number of a formula with one free variable the G¨del number of the o o formula that results from diagonal substitution into that formula. (7) By the chain of equivalences (4). there exists a sentence C in LE such that the equivalence (C ≡ F ( C )) is a true sentence in LE . (∀v2 (d(k) = v2 ⊃ F ((v2 ))) ≡ F (d(k))). (C ≡ F ( C )). expressibility of properties of sequence numbers (Tuesday.

LECTURE 3

32

The proof of the Diagonal Lemma is by a kind of double substitution into the oneplace formula for which a diagonal sentence is being established, first the substitution of the Arithmetical expression of the diagonal function, and then the substitution of the numeral for the G¨del number of the formula that results from that first o substitution. Both of these substitutions are quasi-substitutions in this construction. It might make the idea of the proof more perspicuous if we consider how it goes with actual substitutions if we have a term s(v1 ) in our language such that s(v1 ) = Ev1 (v1 ) . (We could have such a language, at the cost of taking more functions as primitive.) Theorem 18 (variant diagonal lemma) Given a formula F (v1 ) with one free variable in a language for arithmetic L that has a term s(v1 ) such that for each number n, s(n) = En (n) , there is a sentence C in L such that (C ≡ F ( C )) is true. Proof. Consider the formula F (s(v1 )) formed by substituting the term s(v1 ) for the free occurrences of v1 in F (v1 ). (1) Let k = F (s(v1 )) . (2) Let C =df F (s(k)). (3) Then s(k) = C . (4) The numeral s(k) designates the same number as is designated by the term s(k), i.e. the equation s(k) = s(k) is true, so by substitutivity of identity, (F (s(k)) ≡ F (s(k))). (5) Hence by (2) and (3), (C ≡ F ( C )).

3.2
3.2.1

Expressibility of properties of sequence numbers in the language LE
Properties of sequences of digits

The first tool we need in order to code sequences of numbers in LE is to show that we can express in LE the relations that the base b notation of a number m begins, or ends, or is part of the base b notation of a number n. Definition 29 (x begins y) x begins y in base b notation iff the base b notation of x is a (not necessarily proper) initial segment of the base b notation of y. We write this as xBb y. Examples. In base 10, 2 begins 20, but note that in base 13, 2 (as it is written in base 10, and in base 13) does not begin 20, since 20 base 10 is written 17 base 13. Other examples (base 10): The numbers which written in base 10 are 7, 76, 760, 7600, 76007, 760074, and 7600748 all begin 7600748 in base 10. The last of these examples points up the fact that every number begins itself, i.e. an initial segment need not be a proper initial segment. Note that the number 0 does not begin any number except itself, i.e. we don’t say that 0 begins 760748, even though 0760748 = 760748.

LECTURE 3

33

Definition 30 (x ends y) x ends y in base b notation, which we write as xEb y, if the base b notation of x is an end segment (not necessarily proper) of the base b notation of y. Examples. In base 10, the following numbers all end 7600748: 7600748, 600748, 748, 48, 8. Given the notion of one number beginning another in a given base representation and the notion of one number ending another in a given base, we can define the notion of a number being part of another in a given base in terms of these two notions: Definition 31 (x is part of y) x is part of y, in base b notation, which we write as xPb y, if x ends some number that begins y. Remark. Every number is a base b part of itself. Given a base b notation for x, every proper sub-segment of the base b notation that does not begin with a 0 is the base b notation of a number y that is a base b part of x. In base 10, the parts of 2600748 are all the numbers that begin or that end it, and 60074, and all the numbers that begin or end it, and 7. Theorem 19 For any b ≥ 2 the following relations are Arithmetical: (1) xBb y, (2) xEb y, (3) xPb y and, for any natural number n ≥ 2, (4) x1 ∗b . . . ∗b xn Pb y Proof. We will prove a stronger result, which we need later, that these relations not only are expressible in LE , but also that this can be done using only bounded quantifiers, i.e. that these are finitary properties of numbers. 1. If 0 does not occur in the base b numeral for y, then x begins y just in case there exists z such that x ∗b z = y. However, if a zero or a string of zeros occurs in the base b numeral for y and the base b numeral for x is an initial segment of the base b numeral of y which ends just before the 0 or string of 0s in the base b numeral for y or includes some but not all of those 0s, then the numeral of x has to be extended by the remaining 0s before it can be concatenated with a numeral to result in the numeral for y. The extension of the base b numeral for x by the required number of 0s is accomplished by multiplying x by b raised to the power of how many 0s need to be appended. This condition can be expressed in terms of the previously expressed notions P owb (w) and x ∗b z = y, as follows: xBb y iff (x = y ∨ (x = 0 ∧ (∃z ≤ y)(∃w ≤ y)(P owb (w) ∧ (x · w) ∗b z = y))) The bounds on the quantifiers hold from the fact that if z is part of y, then z ≤ y, and any number of the form 10...0 in base b with a string of 0s of a string of 0s in y is ≤ y. 2. xEb y iff (x = y ∨ (∃z ≤ y)(z ∗b x = y) For this case we don’t have any complications from the occurrence of zeros. 3. xPb y iff (∃z ≤ y)(xEb z ∧ zBb y) 4. x1 ∗b . . . ∗b xn Pb y iff (∃z ≤ y)(x1 ∗b . . . ∗b xn = z ∧ zPb y).

LECTURE 3

34

3.2.2

Sequence numbers

Treating sequences of expressions as expressions with G¨del numbers is very conveo nient. It requires making sequences of expressions into single expressions, which is done by expanding the langauge of arithmetic by introducing the symbol ♯ to mark out the beginning and end of a sequence of expressions and the boundary between two successive expressions. Thus among the primitive symbols of LE (Lecture 1) is ♯, which did not enter into the rules for the formation of terms and formulas of LE . The use of this symbol is to allow us to concatenate a finite sequence of formulas of LE into a single expression of the language, by serving as a marker between different formulas in a sequence of formulas. When that is done, a sequence of formulas, as an expression, will have a G¨del number. o Definition 32 (sequence number) x is a sequence number if it is the G¨del o number of an expression of the form ♯Ei1 ♯Ei2 ♯ . . . Eik ♯ in LE where each expression Eij does not contain the symbol ♯,

3.3

Coding of finite sequences of G¨del numbers o

Recall the assignment of the first 12 digits of base 13 representation of natural numbers to the 12 symbols that enter into formulas of the language LE : 0 1 ( 0 2

) 3

f 4

′ v 5 6

∼ 7

⊃ 8

∀ 9

= η

≤ ǫ

Thus if a number is the G¨del number of a formula in LA on the particular o G¨del numbering we have adopted, then the 13th digit, δ, will not occur in its base o 13 representation. Call the class of such numbers Nδ . A formal proof is a finite sequence of formulas, so to code a proof by a number it suffices to find a way of coding finite sequences of numbers in Nδ . We code such a sequence (a1 , . . . , an ) by the number δ ∗13 a1 ∗13 δ ∗13 a2 ∗13 δ ∗13 . . . ∗13 δ ∗13 an ∗13 δ. In future I shall mostly suppress the explicit notation of base 13 (or more generally base b for any b ≥ 2) concatenation and write v1 = v2 v3 for v1 = v2 ∗b v3 , i.e. symbolize the concatenation relation by concatenation itself. There are several points about the concatenation relation that need to be borne in mind. (1) It is a three place relation and not a two-place function. (2) It is a relation between numbers, and numbers are expressible in base b notation for all b ≥ 2 but are not in base b notation. The situation is similar to what it is in number theory generally. When we compute with natural numbers we do so using their base 10 notation (or in the case of computers, base 2 notation). But when we prove something about numbers what we prove is proved using properties of numbers that

v2 ∈ v1 iff (Seq(v1 ) ∧ δv2 δP v1 ∧ ∼ δP v2 ). Proof. . any sequence which includes the number zero would have more than one (indeed infinitely many) sequence numbers. δan δ for ai ∈ Nδ .e. as in “if the digits of a number in its base 10 notation add up to 9 then the number is divisible by 9”. e.. Proposition 20 (sequence numbers) A natural number n is a sequence number iff n = δa1 δa2 δ . It is a necessary condition for v2 ∈ v1 that δv2 δP v1 but not sufficient since numbers of the form δa1 δa2 δ satisfy it. We do this as follows. we say that v2 is in v1 . . Seq(v1 ). which is proved from general properties of the congruence relation and the fact that 10 ≡ 1 (mod 9). Immediate from the definition of sequence number above. δ00δ were allowed as a sequence number it would code the sequence (0) but which is also coded by δ0δ. Definition 33 For v1 a sequence number. Proof.g.LECTURE 3 35 are not specific to decimal notation. if we allowed. The reason for this requirement is that sequence numbers code sequences of numbers in Nδ . e. Proposition 21 The property of being a sequence number. iff v2 is one of the numbers coded by v1 Proposition 22 v2 ∈ v1 is Arithmetical. if. symbolized as v2 ∈ v1 . In expressing the condition that a number is the G¨del number of a proof in o a formal system we need to be able to express the condition that one part of a sequence number occurs earlier in the sequence than another. . the condition ∼ δP v2 rules out those cases. Definition 34 v2 ≺ v3 iff v1 is the sequence number of a sequence in which v2 v1 and v3 occur and the first occurrence of v2 in the sequence is earlier that the first occurrence of v3 in the sequence. The last conjunct rules out the occurrence of a string of zeros of length greater than one. Proof. The property of being a sequence number is expressible by the following formula: (δBv1 ∧ δEv1 ∧ δ = v1 ∧ ∼ δδP v1 ∧ (∀v2 ≤ v1 )(δ0v2 P v1 ⊃ δBv2 )) The first four conjuncts characterize the required occurrences of the digit δ in the base 13 representation of v1 . it is expressible in LE . Since 00 = 0. δ00δ as a sequence number.g. Since δ00δ = δ0δ. i. even if what is being proved specifically refers to decimal notation. is Arithmetical.

v1 36 Proof. . so we don’t need a separate conjunct Seq(v1 ).LECTURE 3 Proposition 23 The three-place relation v2 ≺ v3 is Arithmetical. v2 ≺ v3 iff (v2 ∈ v1 ∧ v3 ∈ v1 ∧ (∃v4 ≤ v1 )(v4 Bv1 ∧ v2 ∈ v4 ∧ ∼ v3 ∈ v4 )) v1 Note that the formulas v2 ∈ v1 and v3 ∈ v1 each contains the condition Seq(v1 ).

A formal language is firstorder if its quantifiers range only over the objects in its domain of interpretation and not over collections (pluralities) of those objects. a weak version of G¨del’s first incompleteness o theorem for PAE (Wednesday. in his doctoral thesis–the incompleteness theorem was his o Habilitation thesis). 20 October 2010) 4. In this course we will investigate properties of a number of different formal systems for arithmetic. All the formal systems we investigate in this course will explicitly or by assumption contain a com37 . A second-order language has quantifiers that range over collections (pluralities) of objects (possibly also over relations between objects).Lecture 4 A formal system PAE for arithmetic. There is no complete axiomatization of full second-order logic.1 A formal system PAE for arithmetic We now begin the investigation of formal first-order axiomatic systems of arithmetic. an Arithmetical proof predicate for PAE . where we are interested in properties of what can. Formal systems of first-order logic are complete (which G¨del proved in 1930. A theory is first-order if its formal language is first-order. we shall be concerned only with first-order systems. or cannot. Accordingly. be proved in formal systems.

a standard misnomer1 The subscript E signifies that in this system exponentiation is taken as primitive. In Lecture 5 we shall see that exponentiation need not be taken as a primitive and that via coding of ordered pairs of natural numbers the relation xy = z can be expressed in terms of zero. L3 establishes the classical logic of ∼. i. (∀vi F (vi ) ⊃ F (t)) for t any term of LE not containing a variable that is quantified in F (vi ) within the scope of which vi occurs.LECTURE 4 38 plete axiomatization of classical first-order logic with identity.e. and the third and fourth groups are axioms specific to arithmetic. the first two of which. 85-87. it is governed by its own axioms. L1 and L2 are exactly the axioms for ⊃ required to establish the Deduction Theorem (taken as proved in a previous course). Group II are axiom schemata for first-order predicate logic. For proofs of these formulas from these schemata see Donald Kalish and Richard Montague. The axiomatization of first-order predicate logic by the Group II schemata is highly unnatural in terms of establishing formulas as logically valid. Archiv f¨r mathematische Logik und Grundlagenforschung 7 (1965). Group III are axioms specific to each of the primitive non-logical notions of the language Group IV is all instances of (a version of) the axiom schema of induction. and 8 on pp. is a formal system of first-order predicate logic. Note that L6 strictly should be written ∼ ∀vi ∼ vi = t. successor. A usual formulation of the induction schema is ′ (F (0) ⊃ (∀v1 (F (v1 ) ⊃ F (v1 )) ⊃ ∀v1 F (v1 ))) . It was Dedekind who established the first axiomatization of arithmetic. Peano cites Dedekind 1888 as the source of his axioms. We shall follow Smullyan in the specification of PAE . Lemmas 2. These axioms are complete for truth-functional validity. 1 . 3. u 81-101. which more natural axiomatizations of predicate logic with identity do. It seems to have been Russell who introduced the misnomer Peano Arithmetic. in 1888. We now set out a formal axiomatic system for arithmetic. pp. To prove the following valid formulas from these schemata is non-trivial: vi = vi . The axioms are in four groups. ‘PA’ stands for Peano Arithmetic. Its virtue for us is that it is very easy to arithmetize since it involves no substitution of terms for free variables. and multiplication. with two rules of inference. which Peano took over in his publication a year later. Group I are axiom schemata for propositional logic: These are the standard axioms for propositional logic with ∼ and ⊃ as primitive. (vi = vj ⊃ vj = vi ). PAE in the language LE . addition. “On Tarski’s formalization of predicate logic with identity”.

Definition 35 (the system PAE ) The axioms and rules of inference of PAE are the following: A. two formulas within this schema are generated by substitution. However. [Generalization] B. since F [v1 ] would be ∀v1 (v1 = v1 ⊃ F (v1 )). L7 (vi = t ⊃ (X1 vi X2 ⊃ X1 tX2 )).e. say to v2 if v2 does not occur free in F . namely ′ ′ ∀vi (vi = v1 ⊃ ∀v1 (v1 = vi ⊃ F (v1 )). provided vi does not occur in F . All instances of the following schemata: L4 (∀vi (F ⊃ G) ⊃ (∀vi F ⊃ ∀vi G)) L5 (F ⊃ ∀vi F ). we can use a quasi-substitution to obtain from F (v1 ) a formula logically equivalent to F (vi ). and that’s the condition we take. We could change the variable in the auxiliary ′ quantification. infer ∀vi F . ′ which is logically equivalent to F (v1 ). L6 ∃vi (vi = t). provided vi does not occur in t. Rules of inference R1 From F and (F ⊃ G). i. All instances of the following schemata: L1 (F ⊃ (G ⊃ F )) L2 (F ⊃ (G ⊃ H)) ⊃ ((F ⊃ G) ⊃ (F ⊃ H)) L3 ((∼ F ⊃∼ G) ⊃ (G ⊃ F )) Group II predicate logic. We abbreviate this formula as F [[v1 ]]. [Modus Ponens] R2 From F . Then we can use quasi-substitution to obtain ′ a formula without any substitutions that is logically equivalent to F (v1 ). namely ∀v1 (v1 = vi ⊃ F (v1 )) where vi is any variable that does not occur in F (v1 ). so not equivalent to F (v1 ). It’s ′ ′ easily seen that F [[v1 ]] is logically equivalent to F (v1 ). ∀v2 (v2 = v1 ⊃ F (v2 )). This logical equivalence only requires that vi does not occur free in F (v1 ). but the sufficient condition that it does not occur at all in F (v1 ) is easier to express in arithmetized syntax. We can’t use quasi-substitution directly on F (v1 ) to express ′ ′ ′ F (v1 ).LECTURE 4 39 However. infer G. in which no variable occurs ′ free. namely ′ F (0) and F (v1 ). which would defeat the purpose of avoiding substitution. and for ease of arithmetization we want to use quasi-substitution instead of substitution. where X1 and X2 are any expressions such that X1 vi X2 ) is an atomic formula and t is any term of LE . Logical axioms and rules of inference Group I propositional logic. But this involves substitution of the variable v2 for all free occurrences of v1 in F . Non-logical axioms Group III axioms specific to each of the primitive non-logical notions of the language ′ ′ N1 (v1 = v2 ⊃ v1 = v2 ) ′ N2 ∼ 0 = v1 N3 (v1 + 0) = v1 ′ N4 (v1 + v2 ) = (v1 + v2 )′ .

e. if there exists a proof in PAE of which F is a member. i. abbreviate the abbreviation ∗13 for base 13 concatenation by using concatenation itself. which is an abbreviation for 0′′′′ ∗13 0′′′′′ and could also be written f ∗13 ′ . in all but cases (4) and (6). The quantifications in (4) and (6) can also be bounded. variables other than vi may occur free in it. unless we stipulate otherwise. Recall that when we write a schematic formula F (vi ). is 5. Recall that the G¨del o number of the subscript symbol. (1) V ar(v1 ): Ev1 is a variable. ′ where. Both for ease of reading and of typesetting in the following I will. after the first case. the variable symbol followed by a finite string of subscript symbols. I will write 45. and that of o the variable symbol is 6. The property of being the G¨del o number of a provable formula requires an unbounded existential quantifier.e. namely that the expressing formulas from LE require only bounded quantifiers. for vi any chosen variable that does not occur in F (v1 ). Definition 37 (provable) A formula F of LE is provable in PAE . Definition 36 (proof ) A proof in PAE is a sequence of formulas each one of which is either an axiom of PAE or follows from an earlier formula in the sequence by the rule of Generalization or follows from two earlier formulas in the sequence by Modus Ponens. the proof of which is established by the formula that follows. on our chosen G¨del numbering. a stronger result than is needed for the present theorem. for vi ∗13 vj I will write vi vj . Because it will be needed for later results we prove. and for 4 ∗13 5. 4.2 An Arithmetical proof predicate for PAE Each numbered paragraph in this section is both a definition of a property or relation of numbers. i. symbolized as P AE ⊢ F . F [[v1 ]] is ′ ∀vi (vi = v1 ⊃ ∀v1 (v1 = vi ⊃ F (v1 ))). These other free variables are referred to as parameters. .LECTURE 4 40 N5 (v1 · 0) = 0 ′ N6 (v1 · v2 ) = ((v1 · v2 ) + v1 ) N7 (v1 ≤ 0 ≡ v1 = 0) ′ ′ N8 (v1 ≤ v2 ≡ (v1 ≤ v2 ∨ v1 = v2 )) N9 (v1 ≤ v2 ∨ v2 ≤ v1 ) 0 N10 v1 = 0′ ′ v2 v N11 v1 = (v12 · v1 ) Group IV axiom schema of mathematical induction ′ N12 (F [0] ⊃ (∀v1 (F (v1 ) ⊃ F [[v1 ]]) ⊃ ∀v1 F (v1 ))). an expression of the form v′′ ···′ . and a proposition that that property or relation is Arithmetical. but these cases are considerably more complicated.

(Seq(v1 ) ∧ (∀v2 ≤ v1 )(v2 ∈ v1 ⊃ (V ar(v2 ) ∨ N um(v2 ) ∨ (∃v3 ≤ v1 )(v3 ≺ v2 ∧ v2 = v1 v3 0)∨(∃v3 ≤ v1 )(∃v4 ≤ v1 )(v3 ≺ v2 ∧v4 ≺ v2 ∧(v2 = 2v3 45v4 3∨v2 = 2v3 455v4 3∨v2 = v1 v1 2v3 4555v4 3))))) (4) T m(v1 ): Ev1 is a term. and similarly in the rest of the formulas expressing arithmetized syntax. addition.. i.LECTURE 4 41 (∃v2 ≤ v1 )((∀v3 ≤ v1 )(v3 P13 v2 ⊃ 0′′′′′ P13 v3 ) ∧ v1 = 0′′′′′′ ∗13 v2 ) In this formula I write out the formal numerals 0′′′′′ and 0′′′′′′ rather than abbreviating them as 5 and 6. e. (Seq(v1 ) ∧ (∀v5 ≤ v1 )(v5 ∈ v1 ⊃ (V ar(v5 ) ∨ N um(v5 ) ∨ (∃v3 ≤ v1 )(v3 ≺ v5 ∧ v5 = v1 v3 0)∨(∃v3 ≤ v1 )(∃v4 ≤ v1 )(v3 ≺ v5 ∧v4 ≺ v5 ∧(v5 = 2v3 45v4 3∨v5 = 2v3 455v4 3∨v5 = v1 v1 2v3 4555v4 3))))) If we had given the formula in (4) as the logically equivalent formula ∃v5 (Seqt(v5 )∧ v1 ∈ v2 ). an expression of the form 0′.g.′ P ow13 (v1 ) (3) Seqt(v1 ): Ev1 is a formation sequence for terms. as it happens. a sequence of expressions each one of which is either a variable or a numeral or the result of applying one of the four functions of successor. the only change needed to obtain Seqt(v5 ) from Seqt(v1 ) is to replace all occurrences of v1 by v5 . by the condition that the subscript o symbol is a part of every part of that expression. In changing the free variable in this way corresponding changes of bound variables in Seqt(v1 ) must be made so that v1 is free for v2 in a logically equivalent transform of Seqt(v1 ). i. All of which is to way that when the formula V ar(v1 ) is written in the primitive notation of LE . and also. multiplication. respectively. decidability of whether or not an expression is . i. i. in base 13 notation) and some other number and not a relationship between numerals. The number required to exist by the quantification over the variable v2 is the G¨del number of a string of subscripts. base 13 representation. i.e. though the relation between numbers is determined to hold or not by going from the number to its. 0′′′′′ for 5 and 0′′′′′′ for 6. Note: The formula in (4) above contains an initial unbounded existential quantifier. ∃v2 (Seqt(v2 ) ∧ v1 ∈ v2 ) Note: The formula Seqt(v2 ) in (4) is obtained from the formula Seqt(v1 ) in (3) by changing the free variable from v1 to v2 . or exponentiation to an expression or expressions occurring earlier in the sequence. of the form t′ or (t1 f′ t2 ) or (t1 f′′ t2 ) or (t1 f′′′ t2 ).. to bring out the fact that the relation 0′′′′′ P13 v3 ) is between numbers.e. in this case. This quantifier can be bounded by the correlate in arithmetized syntax of Problem 2 on Problem sheet 1. in this case between the number 5 (as we write it in base 10.e.e. (2) N um(v1 ): Ev1 is a numeral. the numbers in it will be expressed by numerals.e.

e. . i. but it’s a delicate matter to determine in exactly what language of arithmetic. . (∃v2 ≤ v1 )(∃v3 ≤ v1 )(T m(v2 ) ∧ T m(v3 ) ∧ (v1 = v2 ηv3 ∨ v1 = v2 ǫv3 )) (6) Seqf (v1 ): Ev1 is a formation sequence for formulas. Finding N12 (v1 ) is problem 5 on Exercise sheet 2. I will treat a couple of cases from each of the other groups of axioms. . (5) AF (v1 ): Ev1 is an atomic formula. The property of numbers Ax(v1 ) is expressed by (L1 (v1 ) ∨ . that bound can or cannot be expressed. that we can determine by a finite search whether an expression is a formula.e. ∨ N12 (v1 )). (8) Ax(v1 ): Ev1 is an axiom of PAE . a finite sequence of expressions each one of which is either an atomic formula or of the form ∼ E for E occurring earlier in the sequence or of the form (Ei ⊃ Ej ) for Ei and Ej occurring earlier in the sequence or of the form ∀vi E for vi any variable and E occurring earlier in the sequence. but it’s a delicate question in which languages. with what primitives. ∃v2 (Seqf (v2 ∧ v1 ∈ v2 )) Note: The remark as at (4) above applies here also. i. . Logical axioms: Group I L1 (v1 ): (∃v2 ≤ v1 )(∃v3 ≤ v1 )(F m(v2 ) ∧ F m(v3 ) ∧ v1 = 2v2 82v3 8v2 33) L3 (v1 ): (∃v2 ≤ v1 )(∃v3 ≤ v1 )(F m(v2 ) ∧ F m(v3 ) ∧ x = 227v2 87v3 382v3 8v2 33) Group II L4 (v1 ): (∃v2 ≤ v1 )(∃v3 ≤ v1 )(∃v4 ≤ v1 )(F m(v2 ) ∧ F m(v3 ) ∧ V ar(v4 ) ∧ v1 = 29v4 2v2 8v3 3829v4 v2 89v4 v3 33) L7 (v1 ): (∃v2 ≤ v1 )(∃v3 ≤ v1 )(∃v4 ≤ v1 )(∃v5 ≤ v1 )(∃v6 ≤ v1 )(∃v7 ≤ v1 )(V ar(v2 )∧ T m(v3 ) ∧ v6 = v4 v2 v5 ∧ AF (v6 ) ∧ v7 = v4 v3 v5 ∧ v1 = 2v2 ηv3 82v6 8v7 33) Group III N1 (v1 ): Ev1 is the axiom N1 . . t2 terms. of the form t1 = t2 or t1 ≤ t2 for t1 .LECTURE 4 42 a term. We know by Problem 2 on Problem sheet 1. (Seq(v1 )∧(∀v2 ≤ v1 )(v2 ∈ v1 ⊃ (AF (v2 )∨(∃v3 ≤ v1 )(v3 ≺ v2 ∧v2 = 7v3 )∨(∃v3 ≤ v1 v1 )(∃v4 ≤ v1 )(v3 ≺ v2 ∧ v4 ≺ v2 ∧ v2 = 2v3 8v4 3) ∨ (∃v3 ≤ v1 )(∃v4 ≤ v1 )(v3 ≺ v1 v1 v1 v2 ∧ V ar(v4 ) ∧ v2 = 9v4 v3 )))) (7) F m(v1 ): Ev1 is a formula. with what primitives. There are seven schemata of logical axioms L1 − L7 and eleven axioms of arithmetic N1 − N11 plus one axiom schemata of arithmetic N12 (Induction). i. We need formulas Li (v1 ) such that Li (v1 ) iff (Ev1 is an axiom of form Li ) and Ni (v1 ) such that Ni (v1 ) iff (Ev1 is an axiom of form Ni ). this numerical quantifier can be bounded by a term of the language. i.e. ∨ L7 (v1 ) ∨ N1 (v1 ) ∨ . which in primitive notation is ((v′ )′ = (v′′ )′ ⊃ (v′ ) = (v′′ )).e.

Theorem 24 (Arithmetical proof relation) The two place relation between numbers m and n given by the condition that n is the G¨del number of a proof in PAE o of the formula whose G¨del number is m is expressible in LE . P rovP AE (v1 .LECTURE 4 43 v1 = 226530η26553082653η265533. given by the truth-functional equivalences ((A ≡ B) ≡ (A ⊃ B) ∧ (B ⊃ B)) and ((C ≡ D) ≡∼ (C ⊃∼ D)). N7 (v1 ): Ev1 is the axiom N7 . v2 ) is true. i. So N7 =∼ (((v′ ≤ 0 ⊃ (v′ ) = 0) ⊃∼ ((v′ = 0 ⊃ (v′ ≤ 0)) v1 = 7222653ǫ182653η138722653η182653ǫ133 (9) P rfP AE (v1 ): Ev1 is a proof in PAE . v2 ). The formula P rovP AE (v1 . which yields ((A ≡ B) ≡∼ ((A ⊃ B) ⊃∼ (B ⊃ B))).e. (Seq(v1 ) ∧ (∀v2 ≤ v1 )(v2 ∈ v1 ⊃ (Ax(v2 ) ∨ (∃v3 ≤ v1 )(∃v4 ≤ v1 )(v3 ≺ v2 ∧ v4 ≺ v1 v1 v2 ∧ v4 = 2v3 8v2 3) ∨ (∃v3 ≤ v1 )(∃v4 ≤ v1 )(V ar(v3 ) ∧ v4 ≺ v2 ∧ v2 = 9v3 v4 ))))) v1 (10) P rovP AE (v1 . v2 ): Ev1 is proved by Ev2 . To compute the G¨del number of N7 we must o write it in primitive notation. What G¨del achieved. which Hilbert recognized (see quotation in the first lecture) completely direct. v2 ) ≡ (P rfP AE (v2 ) ∧ v1 ∈ v2 )) (11) P rP AE (v1 ): Ev1 is provable ∃v2 P rovP AE (v1 .(9). Remark: As the construction of the proof predicate for PAE shows. o Proof. or is the result of applying R1 [Modus Ponens] to two formulas occurring earlier in the sequence. . arithmetization by assignment of digits to symbols and of concatenation of corresponding sequences of digits (numbers in the given base notation) to concatenation of sequences of symbols (expressions) makes the correspondence between formal expressions and numbers. a sequence of formulas each one of which is either an axiom of PAE . o Proof. v2 ) in (10) expresses “Ev2 is a proof of Ev1 ”. PAE ⊢ En if and only if ∃v2 P rovP AE (n. was to show that the o formal syntax of strings of symbols by which a formal system of proof is established corresponds exactly with arithmetically definable properties of the corresponding numbers. Corollary 25 (Arithmetical proof predicate) The property of a number that it is the G¨del number of a formula provable in PAE is expressible in LE . or is the result of applying R2 [Generalization] to a formula occurring earlier in the sequence. going beyond Hilbert’s insight. This requires expressing ≡ in terms of ∼ and ⊃. This is evident from this formula and (1) .

This establishes conditions (1) and (2) of Theorem 5. is unprovable and (thereby) true. . and PAE ∼ G. from which we can by the Diagonal Lemma. P rP AE (v1 ). that the set of G¨del numbers of true sentences in the language of arithmetic is o not arithmetical (Exercise sheet 1 problem 4) and the hypothesis that every sentence provable in PAE is true: By the fact that the G¨del numbers of provable formulas o is arithmetical. we have that this set is not identical with the set of G¨del numbers o of true sentences. that expresses the property of being the G¨del number of a formulas derivable in PAE . there is a formula in LE . By Corollary 25. namely a diagonal sentence for ∼ P rP AE (v1 ) which we can show. Having shown that the property of being the G¨del number of a theorem o of PAE is Arithmetical (expressible in LE ). This argument via Tarski’s theorem is highly inefficient since it fails to generate a particular true sentence that is unprovable in the given system while at the same time it requires the construction of an arithmetized proof predicate. there exists a sentence which is true but unprovable in P AE and whose negation is unprovable. and so whose negation is false and so by the hypothesis unprovable. PAE G. Proof. obtain a particular sentence. Hence on the assumption that every sentence provable in PAE is true (condition (3)). and PAE ∼ G. itself used in the proof of Tarski’s theorem. G is true.3 An inefficient and a weak version of G¨del’s o First Incompleteness Theorem for PAE Theorem 26 (an inefficient proof of G¨del’s First Incompleteness Theorem) o If every sentence provable in P AE is true. If every sentence provable in PAE is true and the set of provable sentences does not coincide with the set of true sentences. there must exist a true sentence which is not provable.LECTURE 4 44 4. on the hypothesis that every sentence provable in PAE is true. Proof. G is true. if every sentence provable in PAE is true. Theorem 5 establishes that PAE G. We have thus Theorem 27 (weak version of G¨del’s First Incompleteness Theorem for PAE ) o There is a sentence G in LE such that. the existence of a true sentence in the language of arithmetic not provable in PAE follows immediately from Tarski’s theorem. By the Diagonal o Lemma (Theorem 17) there is a sentence G such that (G ≡∼ P rP AE ( G ))).

addition. (Tuesday. 26 October 2010) 5.1 The system PA with zero. successor. Σ0. and ≤ as primitive We now drop exponentiation as a primitive of the language of arithmetic and drop the axioms governing exponentiation from our first-order system of arithmetic PAE . multiplication. the relation xy = z is ∆1-expressible in the language of PA . a Σ0-coding of finite sets of ordered pairs.and Σ1-formulas. labeled PA. and ≤ as primitive. The mathematical fact that justifies the move from PAE to PA is that the three-place relation xy = z is expressible in the language of PA. but that it is expressible in a way that characterizes the total general recursive functions. 45 . The resulting system is (with variant formulations) the standard first-order axiom system for arithmetic. addition. We shall establish not only that exponentiation is expressible in the language of PA.Lecture 5 The system PA with zero. multiplication. known as Peano Arithmetic. successor.

With these modifications. and n any numeral. respectively. i. v2 = 2v3 4555v4 3. and P rP A (v1 ). the relation of being the G¨del number o o of a formula that occurs in the proof sequence coded by a given number. respectively. for vi any variable. and the property of being the G¨del number of a theorem of PA. which is unbounded quantification. o Proof. Definition 12. We do this by by modifying the constructions for PAE in Theorem 24 and Corollary 25. 5. Also. formulas of the form ∼ ∀vi ∼ (vi ≤ n ∧ F ) and ∼ ∀vi ∼ (vi ≤ vj ∧ F ). In the formula Ax(v1 ) v1 v1 that expresses “Ev1 is an axiom of PAE ” (p. P rovP A (v1 . abbreviated as (∀vi ≤ vj )F and (∃vi ≤ vj )F . then so is (t1 f′′′ t2 ). v2 ). 39) the disjuncts N10 (v1 ) and N11 (v1 ) are dropped. E Theorem 28 (proof predicate for PA in LE ) There are formulas P rfP A (v1 ). which correspondingly removes from formulas of the language LE any expressions that contain the expression f′′′ (without having to make any change to Definition 16).2 Σ0-formulas Definition 39 (bounded quantifiers) For vi any variable. note that bounded existential quantifications are. and P rP RE (v1 ) on p. since ∀vi (vi ≤ vi ⊃ F ) is logically equivalent to ∀vi F . The restriction that the variable vj be distinct from the variable vj when the bound on the quantification is a variable is essential. 40 are transformed to corresponding Arithmetical formulas E E E P rfP A (v1 ). P rovP AE (v1 . abbreviated as (∀vi ≤ n)F and (∃vi ≤ n)F . and F any formula. we obtain an Arithmetical proof predicate for PA. Also. quantification in either of the forms ∀vi (vi ≤ n ⊃ F ) or ∃vi (vi ≤ n ∧ F ) is called bounded quantification. By simple modification of the construction of an Arithmetical proof predicate for PAE in Section 4. We drop the disjunct corresponding to term formation by the function expression f′′′ . in primitive notation.LECTURE 5 46 Definition 38 (the system PA) The language L for PA is obtained from the language LE for PAE by dropping the condition in the definition of terms for LE . the formulas for P rfP AE (v1 ). vj any variable such that i = j. v2 ).2. that if t1 and t2 are terms. v2 ). E E P rovP A (v1 .e. and P rP A (v1 ) in the language LE that express the property of being the G¨del number of a proof sequence for PA. Remark. quantification in either of the forms ∀vi (vi ≤ vj ⊃ F ) and ∃vi (vi ≤ vj ∧ F ) are called bounded quantifiers. so that Seqt(v1 ) is (Seq(v1 ) ∧ (∀v2 ≤ v1 )(v2 ∈ v1 ⊃ (V ar(v2 ) ∨ N um(v2 ) ∨ (∃v3 ≤ v1 )(v3 ≺ v2 ∧ v2 = v3 0) ∨ (∃v3 ≤ v1 )(∃v4 ≤ v1 v1 )(v3 ≺ v2 ∧ v4 ≺ v2 ∧ (v2 = 2v3 45v4 3 ∨ v2 = 2v3 455v4 3))))). The axioms of PA are obtained from those of PAE by dropping axioms N10 and N11 (which are not formulas in the language L of PA). .

Corollary 29 For F any Σ0 -formula. .and Π1-formulas. Suppose vi occurs free in F . which is abbreviated as ∃vi (vi ≤ t ∧ F ). Hence by (b) again. (b). Σ1. so ∀vi (vi ≤ n ⊃ F ) is true. .and ∆1-relations Definition 41 (Σ1 -formula) A Σ1 formula is any formula of the form ∃vi F where F is a Σ0 -formula.3 Σ1. vi is any variable. then ∼ F is Σ0 . which is to say that ∃vi (vi ≤ t ∧ F ) is Σ0 . we also label Σ0 -formulas as Π0 and as ∆0 . and an equation between a term and a numeral is decidable.LECTURE 5 47 Definition 40 (Σ0 -formulas) . vi any variable. If F and G are Σ0 then (F ⊃ G) is Σ0 . Since 0 ≤ n is true. corresponding to the clauses (a).Π1. Suppose F is true. so ∀vi (vi ≤ n ⊃ F ) is false. By (b). Proof. we can compute the truth value of ∼ F and of (F ⊃ G). Then by (c). Proposition 30 (Decidability of Σ0 -sentences) We can effectively decide (compute) the truth or falsity of each Σ0 -sentence. Then ∀vi (vi ≤ n ⊃ F (vi )) is equivalent to (F (0) ∧ . (0 ≤ n ⊃ F ) is false. ∀vi (vi ≤ t ⊃∼ F ) is Σ0 .e. ∼ F (vi ) is Σ0 . (a) Every atomic formula of the language L of PA is Σ0 . i. and (c) of the definition: (a) A closed term is computable to a numeral. Proof. Then (H ⊃ F ) is true for any H. for any terms t1 and t2 . By induction over the recursive definition of Σ0 -formulas. closed Σ0 -formula. so the conjunction is decidable. then ∀vi (vi ≤ t ⊃ F ) is Σ0 . (c) If F is Σ0 . and t is either a variable distinct from vi or a numeral. (b) Given the truth value of F and G. ∼ ∀vi (vi ≤ t ⊃∼ F ) is logically equivalent to ∼ ∀vi ∼ (vi ≤ t ∧ F ). (c) Suppose vi is not free in F . Suppose F is false. . i. ∃vi (vi ≤ t ∧ F ) is Σ0 . ∼ ∀vi (vi ≤ t ⊃∼ F ) is Σ0 . By induction hypothesis each conjunct is decidable. and t either a variable different from vi or a numeral. 5.e. t1 = t2 and t1 ≤ t2 are Σ0 . ∧ F (n)). For reasons that will be apparent shortly. (b) If F is Σ0 .

. Definition 43 (Π1 -formula) A Π1 formula is any formula of the form ∀vi F where F is a Σ0 -formula. For vi not free in F . Remark. ∆1 is not a syntactic form. these sentences are logically equivalent to F (n). i. i. . sentences. such that for each k-tuple (n1 . Proof. nk ) is true iff (n1 . Definition 45 (∆1 relations) A relation is ∆1 if and only if it is both Σ1 and Π1 . . vk ) that expresses R. that are ∆1 and not Σ0 . . as we have seen.LECTURE 5 48 Definition 42 (Σ1 -relation) A relation R ⊆ Nk is Σ1 iff there is a Σ1 -formula G(v1 . nk ). and by the Remark above also ∆1 -formulas. . nk ). .e.e. . such that for each k-tuple (n1 . we shall sometimes label as ∆1 a formula which is equivalent both to a Σ1 -formula and to a Π1 -formula. Proof. Corollary 33 A relation is ∆1 iff it is Σ1 and its complement is also Σ1 . . . propositions.) Proposition 31 Every Σ0 -formula is logically. and hence provable in every system complete with respect to first-order logical validity. . Remark. where v1 is the only free variable in a Σ0 -formula F (v1 ). Lemma 32 (Σ1 and Π1 are dual to each other) The negation of a Σ1 -formula is Π1 . and 1-ary relations. . However. Proof. . . . equivalent to a Σ1 -formula. . i. which is Σ0 for F (v1 ) Σ0 . . . . G(n1 . . nk ) is true iff (n1 .e. . Note that a Σ1 -formula begins with one unbounded existential quantifier. By logic and the fact that the negation of a Σ0 -formula is Σ0 . vk ) that expresses R. (F ≡ ∃vi F ) is logically valid. . (It may contain other quantifiers so long as they are bounded. . nk ) ∈ R. i. It is an important fact that there are ∆1 relations. nk ) ∈ R. and ∆1 -relations include the case of 0-ary relations. but this is not a fact we can prove on the basis of results so far obtained. include the case of formulas with no free variables. . However. Strictly. . there is no such thing as a ∆1 -formula. . Π1 . including 0-ary relations. . The definitions of Σ1 and Π1 -formulas. Definition 44 (Π1 -relation) A relation R ⊆ Nk is Π1 iff there is a Π1 -formula G(v1 . . sets. . Immediate from Definition 45 and Lemma 32.e. G(n1 . i. . and hence provably. and the negation of Π1 -formula is Σ1 .e. i. .e. The definitions of Σ1 . . Example of a ∆1 -proposition: The proposition expressed by the equivalent sentences ∃v1 (v1 = n ∧ F (v1 )) and ∀vi (vi = n ⊃ F (vi )).

vn . This condition is equivalently expressed by ∃vn+4 (∃vn+3 ≤ vn+4 )(∃vn+2 ≤ vn+4 )(F (v1 . . . P owp (x) —x is a power of p. . vn ) = vn+1 is expressed by the condition ∃vn+3 ∃vn+2 (F (v1 . . concatenation to base p is expressible in L. is expressible in LE . vn . 1. . . vn+3 . (∃vn+3 ≤ vn+4 )(∃vn+2 ≤ vn+4 )(F (v1 . . vn ) = vn+1 is also Σ1 . x ∗b y = z. Lemma 34 If a total function f : Nn → N is Σ1 . ∆1 ) For f an n-ary function from Nn to N. for every v1 . . . from which it follows that the relation f (v1 . vn ) = vn+1 . 43) that the property of a number x that it is a power of a given prime p can be expressed without using the exponentiation function. . Π1 . Then. Proof. Hence the preceding formula is Σ1 . vn+2 )∧ ∼ vn+3 = vn+1 ). . . vn+1 . . . vn+2 ) expresses the relation f (v1 . . . vn . vn ) = vn+1 is Σ1 . . 5. . vn+2 )∧ ∼ vn+3 = vn+1 ) By Definition 40. . ((∃z ≤ y)(x · z = y)∧ ∼ x = 0). . . then it is ∆1 .4 Arithmetization of syntax in the language of PA We have already seen that for any base b ≥ 2. . . pℓp (x) = y —y is the smallest positive power of p greater than x. . vn ) = vn+1 is ∆1 . vn . vn+2 ) such that ∃vn+2 F (v1 . . vn ) = vn+1 . Similarly for Πi and ∆1 . p. f is Σ1 iff the n + 1-ary relation f (v1 . vn . since f is total so that. By the hypothesis that f is Σ1 . . x | y —x divides y. . there is a Σ0 -formula F (v1 . 2. vn+1 . This result is based on an observation by John Myhill (see Smullyan. and indeed that it is Σ0 -expressible in this language. Lemma 35 For every prime number p the following conditions are Σ0 . 1. concatenation to base b. . vn . . . since x is a power of the given prime p if and only if every proper divisor of x is divisible by p. . We show that the relation f (v1 . Proof. vn+2 )∧ ∼ vn+3 = vn+1 ) is a Σ0 -formula. . 3. . Note that by this condition every non-zero number divides 0 (which reflect the usual practice in number theory that 0 is a . . . . . there is vn+1 such that f (v1 . vn+3 . . . . . . . . the relation f (v1 . . vn+3 . .LECTURE 5 49 Definition 46 (a function is Σ1 . We now show that for base p for p a prime number. . .

This results was proved by G¨del in o his original paper. there is a number k such that for any numbers m and n. Proof. (a2 . v3 ) holds. . (a2 . Proof. . . br ). 3. v3 ) such that 1. . v2 . . the set of ordered pairs is decoded. given a code number k. For any numbers v1 . if K(v1 . b1 ). . nor any non-zero number by the first conjunct (cf problem 3(a) on Problem sheet 1). Lemma 37 The Arithmetical relation xPb y (‘x is part of y’) for base b a prime number is Σ0 . (ar . we don’t need the Chinese Remainder o Theorem for this result. x ∗p y = z is Σ0 iff pℓp (y) + y = z iff (∃v1 ≤ z)(v1 = pℓp (y) ∧ ((x · v1 ) + y) = z). (P owp (y) ∧ y > x ∧ y > 1 ∧ (∀z < y) ∼ (P owp (z) ∧ z > x ∧ z > 1). 2. br ) is coded by a number k. which does take exponentiation as primitive. v3 . b1 ). and a formula K(v1 . k) holds if and only if (m. 5. Lemma 35 and Lemma 36. By Theorem 19. For any finite set of ordered pairs of natural numbers (a1 . n. K(m. v3 ) whereby. br ). b2 ). Lemma 36 (base p concatenation is Σ0 ) For any prime p. The result follows by part 3 of Lemma 35. Proof. v2 . Theorem 38 (Σ0 -coding of finite sets of ordered pairs) There is a Σ0 -formula K(v1 . The key to this results is a Σ0 coding of finite set of ordered pairs of numbers. 2. b2 ). b1 ). . v2 . . We need to describe two things (dependently related to each other): a process whereby a set of ordered pairs of numbers (a1 . . . b2 ). .LECTURE 5 50 member of every principal ideal on the natural numbers) and 0 does not divide 0 by the second conjunct. n) is one of the pairs (a1 . . v2 . . then v1 ≤ v3 and v2 ≤ v3 . (a2 . (ar .5 A Σ0-coding of finite sets of ordered pairs of numbers We now establish a result that exponentiation can be expressed in the language of PA. (ar . for his G¨del numbering using the Chinese Remainder Theorem. the relation x ∗p y = z is Σ0 . (∀z ≤ x)((z | x ∧ z = 1) ⊃ p | z). o Given our different G¨del numbering result.

v2 . . b2 ). Note that there is an error in Smullyan’s proof (p. As noted above. Quine. M F (x. In this case the maximal frame in f f a1 f b1 f f a2 f b2 f f . (∀y ≤ x)(yPb x ⊃ 1Pb y). Note that if y has a frame number in it. Let θ be a finite sequence of ordered pairs. y) is Σ0 . pp. b2 ). . . . (ar . By Lemmas 36 and 37. Examples: 110 (11110 ). K(v1 . which we will label M F (x. ar . b1 . Having expressed the notion of a maximal frame. f f ar f br f f . y): (xP y ∧ (∃z ≤ y)(1(z) ∧ x = 2z2∧ ∼ (∃w ≤ y)(1(w) ∧ 2zw2P y))). 45). The problem is if one or more of the ai or bi is a string of 1s that is longer than any string of 1s occurring in a frame number in one of the numbers being coded. “Concatenation as a basis for arithmetic”. we are then able to define a formula K(v1 . .V. br . which also means that 113 (18310 ). The frame number f chosen for the coding occurs in k will be a maximal frame number for k since the string of 1s in f is longer than any strings of 1s from the ai and bi . which we shall abbreviate as 1b (x). x is part of y. (ar . Let us suppose that the sequence of ordered pairs (a1 . . . since the length of frame numbers in y is bounded by the length of y. . b1 ). . in our case 13. let f be any frame number that is longer than any frame which is part of any of the numbers a1 . By a frame number we shall mean a number whose numeral in a specified base. f f ar f br f f is 2c2 and not f . . . . . Since frame numbers whose numerals in the given base are of the same length are equal. . see Smullyan p. v3 ) =df (∃v4 ≤ v3 )(M F (v4 . v3 ) ∧ v4 v4 v1 v4 v2 v4 v4 P v3 ∧ ∼ v4 P v1 ∧ ∼ v4 P v2 ). v3 ) is Σ0 . the formula K(v1 . we can then decode each of the pairs of numbers coded in k. (a2 . b1 . Having recovered f from k. any number which contains a frame number contains a unique maximal frame number. . The code k of the set of ordered pairs is the number f f a1 f b1 f f a2 f b2 f f . This relation is expressed by the following formula. v2 . Journal of Symbolic Logic 11 (1946). (a1 . v3 ) with which to decode the ordered pairs coded by the process specified in (1). (2) Decoding We call x a maximal frame in y if x is a frame number. . . it is unique. and let f be any frame number that has a longer string of 1s in it than the longest string of 1s that occurs in any of the numbers a1 . and x is as long as any frame that is part of y. b1 ). By Lemmas 36 and 37. (a2 .LECTURE 5 51 (1) Coding. and the decoding fails. br ). Let c be the longest such number and let f be the frame number specified by Smullyan. but ∼ 110 (11113 ) since 11113 = 18310 . 105-114. The condition that the numeral for x in base b representation is a string of 1s. br ) has been coded by the number k using the above procedure. is of the form 2t2 where t is a string of 1s (this idea goes back to W. is expressible by the condition. v2 . br . it has a maximal frame number. 45). at the point when he says. . . . . ar .

e. Given the coding and Σ0 -decoding of finite sets of ordered pairs of numbers by Theorem 38.LECTURE 5 52 5. Theorem 41 and its proof are specific to exponentiation and do not generalize to other functions. we can express this by the formula ∃w(K(y. Berlin. the result in Theorem 39 with Corollary 40 generalizes to the following important result: Theorem 42 A total function is general recursive if and only if it is ∆1 . z. x). s. v2 . if K(v1 . (y. v2 . v3 ) holds. so that it is. indeed. See Petr H´jek and Pavel a Pudl´k. Since K(v1 . The relation xy = z holds if and only if there is a set of ordered pairs {(0. (1. Σ1 . . a 299-303. Springer. v3 ) is Σ0 . x2 ). i. Later. namely: Theorem 41 (exponentiation is ∆0 ) The relation xy = z is ∆0 . Proof.6 The relation xy = z is ∆1-expressible in the language of PA Theorem 39 (exponentiation is Σ1 ) The relation xy = z is Σ1 . Proof. z) is a member of that set. Corollary 40 (exponentiation is ∆1 ) The relation xy = z is ∆1 . xy )} and (y. w) ⊃ ((u = 0 ∧ v = 1) ∨ (∃r ≤ w)(∃s ≤ w)(K(r. Metamathematics of First-Order Arithmetic. The proof is too complicated to give here. that for any numbers v1 . pp. . (2. 1). this whole formula is Σ1 . then v1 ≤ v3 and v2 ≤ v3 in order to bound the universal quantifiers inside the formula. Proof. w) ∧ (∀u ≤ w)(∀v ≤ w)(K(u. 1993. On the other hand. . . Proof. . v. v2 . Notice that for this result we need the second clause of Theorem 38. w) ∧ r′ = u ∧ v = x · s))). Immediate from Theorem 39 by Lemma 34. v3 . A stronger result than Corollary 40 is true.

then for any distinct variables vi and vj . 53 . but which are more flexibly expressed than is required in order to be a Σ1 -formula. If F is a Σ-formula. 27 October 2010) 6. (∃vi ≤ vj )F and (∀vi ≤ vj )F are Σ-formulas. the arithmetical hierarchy (Wednesday. corresponding to the five clauses in the definition of a Σ-formula. (F ⊃ G) is a Σ-formula. (∃vi ≤ n)F and (∀vi ≤ n)F are Σ-formulas. If F is a Σ-formula. (F ∨ G) and (F ∧ G) are Σ-formulas. and for any number n. Definition 47 (Σ-formula) . For any Σ0 -formula F and Σ-formula G.1 Σ-formulas We now define a class of formulas each of which is provably equivalent in PA to a Σ1 -formula. Base: Every Σ0 -formula is a Σ-formula. For any Σ-formulas F and G. then for any variable vi . Recursion: 1. the arithmetized proof predicate for PA is Σ1.Lecture 6 Every Σ-formula is provably equivalent to a Σ1-formula. 3. To show that every Σ-formula is provably equivalent in PA to a Σ1 -formula we prove the following five Lemmas. 4. 2. the formula ∃vi F is a Σ-formula.

. . vj ). . By Lemma 44. 54 Proof. This argument holds also for a numeral n in place of the variable vk . vj )) is provably equivalent to a Σ1 -formula. Hence by Lemma 43 it is provably equivalent to a Σ1 formula. vj ) (b) (∀vj ≤ vk )∃vi R(vi . . (F ⊃ ∃vi H(vi )) is provably equivalent in PA to a Σ1 -formula. Lemma 46 For F a Σ0 -formula and ∃vi H(vi ) a Σ1 -formula. . . Proof. (a) (∃vj ≤ vk )∃vi R(vi . . Hence by Lemma 45 (vj ≤ vk ∧ ∃vi R(vi . . vik )) Lemma 44 If ∃vi F (vi ) is a Σ1 -formula. The formula on the right is Σ1 . (a) The formula vj ≤ vk is Σ0 by the definition of Σ0 . Then (∃vi F (vi ) ∨ ∃vj G(vj )) and (∃vi F (vi ) ∧ ∃vj G(vj )) are each provably equivalent in PA to a Σ1 -formula Proof. vik ). is provably equivalent to a Σ1 formula. (b) ((∃vi F (vi ) ∧ ∃vj G(vj )) ≡ ∃vi ∃vj (F (vi ) ∧ G(vj ))) is logically valid (on the assumption. Lemma 47 For ∃vi R(vi . Exercise (Problem sheet 3 problem 2(a)) Lemma 45 Let ∃vi F (vi ) and ∃vj G(vj ) be Σ1 -formulas. . . . . so provable in PA. (b) Exercise (Problem sheet 3 Problem 3(b)). PA ⊢ (F (vi1 . so provably in PA. that vi does not occur free in G and vj does not occur free in F ). . . We must assume. Any formula F (vi1 . where vik+1 does not occur free in F (vi1 . ⊢ (∃v1 ∃v2 (F (vi ) ∧ G(vj )) is provably equivalent in PA to a Σ1 formula. . that vi does not occur free in F . vik ). . (a) ((∃vi F (vi ) ∨ ∃vj G(vj )) ≡ ∃vk (F (vk ) ∨ G(vk ))) is logically valid (for vk any variable substitutable into F and G) (Cf.LECTURE 6 Lemma 43 Every Σ0 -formula is provably equivalent in PA to a Σ1 -formula. without loss of generality. . vj )). Hence by Lemma 44. . ∃vj (vj ≤ vk ∧∃vi R(vi . vik ) ≡ ∃vik+1 F (vi1 . Hence this equivalence is provable in PA. which we can do without loss of generality. vj ) and similarly for the formulas in (a) and (b) with a numeral in place of the variable vk . . vj ) a Σ1 -formula. (F ⊃ ∃vi H(vi )) ≡ ∃vi (F ⊃ H(vi )) is logically equivalent. . vik ) is logically equivalent to ∃vik+1 F (vi1 . Problem sheet 1 problem 5(a)). so by the completeness of first-order logic in PA. then ∃vj ∃vi F (vi ) is provably equivalent in PA to a Σ1 -formula Proof. the following formulas are provably equivalent in PA to Σ1 -formulas. Proof. which we abbreviate as (∃vj ≤ vk )∃vi R(vi . .

i. . By Theorem 28 and Lemmas 36 and 37.47. We know from Problem 2 on Problem sheet 1 that whether a string of symbols is a term or a formula is decidable. Proof. (Seq(v2 )∧(∀v3 ≤ v2 )(v3 ∈ v2 ⊃ (Ax(v3 ) ∨ (∃v4 ≤ v2 )(∃v5 ≤ v2 )(v4 ≺ v3 ∧ v5 ≺ v3 ∧ v5 = 2v4 8v2 3) ∨ (∃v4 ≤ v2 )(∃v5 ≤ v2 v2 v2 )(V ar(v4 ) ∧ v5 ≺ v3 ∧ v3 = 9v4 v5 ))) ∧ v1 ∈ v2 ) is equivalent to a Σ1 -formula. this formula is Σ1 . These occur the formula Ax(v1 ) in L1 − L7 and L12 (Induction axioms). By induction on the recursive definition of Σ-formulas. Theorem 51 (proof predicate for PA is Σ1 ) The formula P rP A (v1 ).e. However. v2 Proof. By Lemma 50 and Lemma 44. In none of these occurrences is it in the antecedent of a conditional. Proof. i. expressed in the language of PA. ∃v2 (P rfP A (v2 ) ∧ v1 ∈ v2 ).e. By Lemma 44. but giving an explicit formulation of that bound in the language of PA is hard work which we can avoid. the base case and each induction step established by one of Lemmas 43 . With considerable effort we can actually establish that it’s Σ0 .LECTURE 6 55 Theorem 48 (Σ equivalent to Σ1 ) Every Σ-formula is provably equivalent in PA to a Σ1 -formula. namely that the arithmetized proof predicate for PA is Σ1 . A much stronger result than Theorem 49 is true and is needed for what is to come. 6. Hence in a prenex normal form of (P rfP A (v2 ) ∧ v1 ∈ v2 ) these several existential quantifiers come out as prenex existential quantifiers. and the occurrence of Ax(v1 ) in (P rfP A (v2 ) ∧ v1 ∈ v2 ) is also not in the antecedent of a conditional. for the overall result that the proof predicate for PA is Σ1 it’s sufficient to show that (P rfP A (v2 ) ∧ v1 ∈ v2 ) is Σ1 . is equivalent to a Σ1 -formula. The key point is that the only place in the construction of P rfP A (v2 ) in which we used an unbounded quantifier was in T m(v1 ) (Ev1 is a term) as ∃v2 (Seqt(v2 )∧ v1 ∈ v2 ) and F m(v1 ) (Ev1 is a formula) as ∃v2 (Seqf (v2 ∧ v1 ∈ v2 )). Proof. so we know that there is a bound on those quantifiers.2 The arithmetized proof predicate for PA is Σ1 Theorem 49 The arithmetized proof predicate for PA under our arithmetization of syntax of PA is arithmetical.e. Lemma 50 The formula (P rfP A (v2 )∧v1 ∈ v2 ). The first step is to show that the two-place formula (P rfP A (v2 ) ∧ v1 ∈ v2 ) is Σ1 . i.

. By a simple modification of Theorem 27 so that it applies to PA rather than PAE . not provable in PA. by Theorem 3 (Prenex Normal Forms) and a generalization of Theorem 48. Then the G¨del sentence G. if every sentence provable in PA is true. Suppose the complement of {n : P A ⊢ En } is expressed by a Σ1 -formula. (In Lecture 8 we shall prove this result on a much weaker hypothesis. A simple cardinality argument tells us that most relations on the natural numbers. We require the result. Π0 . which we will prove in Lecture 7. There is a corresponding arithmetical hierarchy of sets and relations. Proof. that PA is Σ1 complete. Remark. in particular the 1-ary relations.3 The arithmetical hierarchy The kind of classifications we introduced in Lecture 5 with the notions of Σ0 . which contradicts the unprovability of G on the hypothesis that every sentence provable in PA is true. the sets of natural numbers. Exercise (Problem sheet 3 problem 1) 6. Definition 48 (arithmetical hierarchy of formulas) (a) Σ0 -formulas (=df Π0 formulas) are as given by Definition 40. ∆1 can be extended. if every sentence provable in PA is true. call it N P rP A (v1 ). are not in this hierarchy (there are uncountably many sets of natural numbers and there are countably many formulas in the language of arithmetic). then PA⊢ G. Π1 . Proof. if G is true and equivalent to a Σ1 -sentence. such that (G ≡∼ P r( G )) is o true. is equivalent to the Σ1 -sentence N P RP A ( G ). then ∀vi F is Πn+1 . Proof. to a hierarchy of all formulas in the language of PA. 56 Theorem 53 {n : P A ⊢ En }.e. which by Corollary 52 is Σ1 . is not ∆1 . i.) Proposition 54 {n : PA ⊢ En [n]} is Σ1 . ∆0 and Σ1 . (b) If F is Σn . But by the Σ1 -completeness of PA (to be proved in the next lecture).LECTURE 6 Corollary 52 {n : P A ⊢ En } is Σ1 . then ∃vi F is Σn+1 . (c) If F is Πn . By inspection of the formula ∃v2 (P rfP A (v2 ) ∧ v1 ∈ v2 ). G is true and. This correspondingly defines a hierarchy of relations on natural numbers (the arithmetical hierarchy).

Proof. formulas provably equivalent in PA to Σn -formulas are closed under existential quantification and formulas provably equivalent to Πn -formulas are closed under universal quantification. Proof. Exercise (Problem sheet 3 problem 3). Proof. find a prenex normal form for it. and negation. n) = 2 (m + n + 1)(m + n) + m is a bijection between the natural numbers and pairs of natural numbers which is strictly increasing in both arguments. it is said to be ∆n . (2) formulas provably equivalent to Σn -formulas. Corollary 57 Every formula in L is equivalent to a Σn or Πn formula for some n. disjunction. 1 Lemma 55 (Σ0 pairing function) The function p(m. are both closed under conjunction and disjunction. we require a Σ0 -pairing function. and for many other purposes. For a given formula. and it is Σ0 . and formulas provably equivalent to Πn -formulas. Definition 50 If a relation is both Σn and Πn .LECTURE 6 57 Definition 49 (arithmetical hierarchy of sets and relations) A relation on natural numbers is Σn (or Πn ) if and only if it is expressible by a Σn -formula (respectively a Πn -formula) in L. Theorem 56 (1) for n > 0. (b) Show that formulas provably ∆n are closed under conjunction. (Problem sheet 3 problem 4). By Theorem 56 (3). By a single induction on n in the conjunction of the three statements. . In order to generalize Theorem 48. adjacent like quantifiers can be collapsed to a single quantifier.

1 Σ0-completeness and Σ1-completeness Definition 51 (Σ0 -completeness) A system S is Σ0 -complete iff for each true Σ0 -sentence X. S ⊢ X. weak systems of arithmetic Q and R (without induction). Σ0-completeness of systems R. F (k) is a true Σ0 -sentence. a sentence of the form ∃vi F (vi ) where F (vi ) is a Σ0 -formula. Σ0-soundness and Σ1-soundness (Tuesday. S ⊢ F (k). Proof. 2 November 2010) 7.Lecture 7 Σ0-completeness and Σ1-completeness. Then by predicate logic in S.e. By Σ0 -completeness of S. Assume S is Σ0 -complete and let X be a true Σ1 -sentence. Proposition 58 A system is Σ1 -complete iff it is Σ0 -complete. 58 . for some number k. S ⊢ X. i. Q. S ⊢ ∃vi F (vi ). and PA. Definition 52 (Σ1 -completeness) A system S is Σ1 complete iff for each true Σ1 -sentence X. Since X is true.

v1 does not occur free in X. so by Σ0 -completeness. S ⊢ F (0). Lemma 59 A system S is Σ0 -complete iff S correctly decides every Σ0 -sentence. Then for some m ≤ n. or X is false and S ⊢∼ X. . if X and Y are correctly decided by S. For any Σ0 -formula F (vi ) with vi the only free variable and for every number n. . . by C1 .LECTURE 7 59 Assume S is Σ1 -complete and let X be a true Σ0 -sentence. F (m) is false. Proposition 60 The following two conditions on a system S together imply that S correctly decides every Σ0 -sentence. 2. Then by condition C2 . C2 . Any Σ0 -sentence Z that is neither atomic nor of the form ∼ X or X ⊃ Y must be of the form (∀vi ≤ n)F (vi ) where F (vi ) is a Σ0 -formula of lower degree than Z and contains vi as its only free variable. . Then by induction hypothesis. In this lecture we shall see that PA and much weaker subsystems of PA are Σ0 -complete. Since X is a sentence. S ⊢ m ≤ n. then ∼ X and X ⊃ Y are correctly decided by S. Definition 53 (A system S correctly decides a sentence X) A system S correctly decides a sentence X iff either X is true and S ⊢ X. . so S ⊢ ∃v1 (X ∧ v1 = v1 ). [Exercise] 3. . S ⊢ (∀vi ≤ n)F (vi ). Suppose (∀vi ≤ n)F (vi ) is true. Conversely. . Then ∃v1 (X ∧v1 = v1 ) is a true Σ1 -sentence. By propositional logic in S. as required. i. F (n) is true. Σ0 -completeness of S. and X is any false Σ0 -sentence. then S ⊢ (∀vi ≤ n)F (vi ). Proof. S correctly decides every atomic sentence. Suppose (∀vi ≤ n)F (vi ) is false. . S ⊢ F (n). Since m ≤ n is a true atomic Σ0 -formula. . S ⊢∼ X. S ⊢ X. if S ⊢ F (0).e. C1 . By induction over the recursive definition of Σ0 -formulas. . Then by 2. S ⊢∼ (m ≤ n ⊃ F (m)). That means that each of the sentences F (0). so (∃v1 (X ∧ v1 = v1 ) ⊃ (X ∧ ∃v1 v1 = v1 )) is logically valid and hence provable in S. Then ∼ X is a true Σ0 -sentence. S ⊢ F (n). suppose S is Σ0 -complete. so by propositional logic in S. Half of the condition that S correctly decides every Σ0 -sentence is that if X is a true Σ0 -sentence then S ⊢ X. Proof. which is to say S ⊢∼ (∀vi ≤ n)F (vi ). Then by induction hypothesis we have that S ⊢∼ F (m). By C1 S correctly decides all atomic Σ0 -sentences. . . S ⊢∼ ∀vi (vi ≤ n ⊃ F (vi )). So by propositional logic in S. The following formula is logically valid: (∀vi (vi ≤ n ⊃ F (vi )) ⊃ (m ≤ n ⊃ F (m))) and hence provable in S. . 1.

∨ vi = n) ⊃ F (vi )). Hence by propositional logic in S. S ⊢∼ m = n. ⊢∼ t1 ≤ t2 .2 Weak systems of arithmetic Q and R (without induction) Definition 54 (system Q) The system Q is obtained from the system PA by dropping the axiom schema for induction. and is thereby Σ0 -complete. Thus Q has only finitely many (nine) nonlogical axioms: ′ ′ N1 (v1 = v2 ⊃ v1 = v2 ) . Then by pure logic (with identity) in S. . that S correctly decides all atomic sentences. To show that D3 implies C2 . S ⊢∼ m = 0. by Generalization and Instantiation. . D2 . To establish C1 . D1 . Then by propositional logic in S. For any variable vi and any number n. where ∼ m ≤ n. m = n are false. We show that conditions D1 . We are given by D1 that all true atomic sentences are provable in S. S ⊢ F (n). Specifically. and by D2 S ⊢∼ m = n.∨vi = n)). 1. S ⊢ ((vi = 0 ∨ . S ⊢ (vi ≤ n ⊃ (vi = 0∨. . Then by D1 . D3 . . . S ⊢ (∀vi ≤ n)F (vi ). where m = n. Suppose F (vi ) is a Σ0 formula with vi its only free variable. . . So it remains to show that all false atomic sentences are refutable in S. D3 together imply C1 . and that n is a number such that S ⊢ F (0). ∨ m = n)). S ⊢ t1 = m. S ⊢∼ (m = 0 ∨ . Proof. S ⊢∼ m = n. . (ii) If the false atomic sentence is of the form t1 ≤ t2 . Then by propositional logic in S. . D1 . S ⊢∼ t1 = t2 . From D3 . All true atomic sentences are provable in S. i. and S ⊢ t2 = n. 7. (i) If the false atomic sentence is of the form t1 = t2 . . S ⊢∼ m ≤ n. . there are true atomic sentences of the form t1 = m and t2 = n. . there are true atomic sentences of the form t1 = m and t2 = n. S ⊢ (vi ≤ n ⊃ F (vi )). . D2 . .LECTURE 7 60 Proposition 61 The following three conditions on a system S jointly imply that S is Σ0 -complete. . D2 . . . .e. 2. S ⊢ (vi = 0 ⊃ F (vi )).e. . all the sentences m = 0. Since ∼ m ≤ n. S ⊢ (vi = n ⊃ F (vi )). . which establishes that S correctly decides every Σ0 -sentence. i. S ⊢ ∀vi (vi ≤ n ⊃ F (vi )). . Then by pure logic (Generalization) in S. D3 imply conditions C1 . ∨ m = n). . Then by propositional logic in S. S ⊢ (m ≤ n ⊃ (m = 0 ∨ . From D1 and the fact that t1 = m and t2 = n are true. N12 . For any distinct numbers m and n. Then by substitutivity of identity. and D3 implies C2 . Then by D3 and propositional logic in S (transitivity of ⊃). C2 of the previous Proposition. . Hence by D2 . . S ⊢ t1 = m and S ⊢ t2 = n.

We prove this by proving a yet stronger result. 61 The system Q is a variant of one due to Raphael Robinson. an instance of Ω2 is (0′′ f′′ 0′′′ ) = 0′′′′′′ .LECTURE 7 N2 N3 N4 N5 N6 N7 N8 N9 ′ ∼ v1 = 0 v1 + 0 = v1 ′ v1 + v2 = (v1 + v2 )′ . Ω5 (v1 ≤ n ∨ n ≤ v1 ). Lemma 62 For each natural number n. . Ω4 (v1 ≤ n ⊃ (v1 = 0 ∨ . By ∀-Intro and ∀-Elim on the instance of Ω5 for n. We will show that all true Σ0 -sentences are provable Q and so provable in PA since all the axioms of Q are axioms of PA. it has as axioms infinitely many instances of computations of addition. The instances of Ω1 are generated from pairs of natural numbers. Ω5 ⊢ n ≤ n. and two axiom schemata expressing properties of ≤. We now show that the system R proves the converse of Ω4 . e. A corresponding remark holds concerning occurrences of the dot symbol for multiplication in the above formulation of Ω2 . and inequality. Instead of the (finitely) many recursion axioms of PA and Q. Ω3 ∼ m = n where m = n. Ω5 ⊢ (n ≤ n ∨ n ≤ n). ∨ v1 = n)). Ω2 m · n = m · n. . is Σ0 -complete. m and n. by writing an equation between the term (mf′ n) on the left and the term 0 with m + n many occurrences of the symbol ′ suffixed to it on the right. namely that an even weaker system R. so by propositional logic. Proof. Ω5 ⊢ n ≤ n. multiplication. v1 · 0 = 0 ′ v1 · v2 = (v1 · v2 ) + v1 (v1 ≤ 0 ≡ v1 = 0) ′ ′ (v1 ≤ v2 ≡ (v1 ≤ v2 ∨ v1 = v2 )). in three axiom schemata. Definition 55 (system R) The axioms of R are all sentences sentences and formulas of L generated from natural numbers m and n by the following axiom schemata: Ω1 m + n = m + n.g. also due to Raphael Robinson. . and also that + abbreviates different formal expressions in its two occurrences in the formulation of Ω1 . Note that by our notational conventions the symbol + in the formulation of Ω1 is used to abbreviate expressions in the formal language L and does not actually occur in sentences of the form Ω1 . (v1 ≤ v2 ∨ v2 ≤ v1 ).

If n = 0 then there is some number k such that n = k + 1. . So by ′ ′ ⊃-Intro. logic Theorem 65 R ⊢ ((v1 = 0 ∨ . (1) v1 = n ⊃ (n ≤ n ⊃ v1 ≤ n) (2) (2) v1 = n (2) (3) (n ≤ n ⊃ v1 ≤ n) (4) n ≤ n (2) (5) v1 ≤ n (6) v1 = n ⊃ v1 ≤ n substitutivity of identity Assumption (1) (2) ⊃-elimination Lemma 62 (2)(4) ⊃-elimination (2)(5) ⊃-introduction ′ 62 Lemma 64 For each number k. The case for n = 0 is an instance of Lemma 63. Let m be any number ≤ k (1) (2) (3) (4) (5) (6) (7) (8) (9) k = 0. To do this we need the following lemma. . (4) prop. ∨ v1 = k) ⊃∼ k ≤ v1 ) ′ ′ (v1 ≤ k ∨ k ≤ v1 ) ′ ((v1 = 0 ∨ . . i. .LECTURE 7 Lemma 63 For each number n. Proof. ∨ v1 = k) ⊃ v1 ≤ k ) Proof. . . .e. (R) ⊢ (v1 = n ⊃ v1 ≤ n). Q. . . . R ⊢ (v1 = 0 ⊃ v1 ≤ 0). . ∨ v1 = n) ⊃ v1 ≤ n). . . and PA We will now show the R is Σ0 -complete. . by ∨-elimin instance of Ω5 (7). k = m ′ ′ (k = 0 ∧ . 7. We have v1 ≤ k from the first ′ disjunct by Lemma 64 and v1 ≤ k from the second disjunct by Lemma 63. ∨ v1 = k) ∨ v1 = k ) ⊃ v1 ≤ k ). . ∧ k = m) ′ ′ ∼ (k = 0 ∨ . . R ⊢ (((v1 = 0 ∨ . ∨ v1 = k) ∨ v1 = k ). . . logic (5) substitutivity of identity (6). for each m ≤ k. .3 Σ0-completeness of systems R. (R) ⊢ ((v1 = 0 ∨ . This case follows by ∨-elimination from ′ ′ the disjunction ((v1 = 0 ∨ . ∨ k = m)) ′ ∼k ≤m ′ (v1 = m ⊃∼ k ≤ v1 ) ′ ((v1 = 0 ∨ . ∨ v1 = k) ⊃ v1 ≤ k ) ′ ′ Ω3 (2) ∧-introduction (3) DeMorgan law Ω4 ∀-introduction and elimination (3). . Proof. (8) prop. The results follows by ∨-elimination from (n = 0 ∨ n = 0). . . ∨ k = m) ′ ′ ′ (k ≤ m ⊃ (k = 0 ∨ . .

Let m1 + m2 = k. R ⊢ t1 = t2 . then by induction hypothesis there is a numeral n such that R ⊢ t1 = n. Since t1 = t2 is a sentence. The uniqueness of n for a given term t follows by transitivity of identity. there is a unique numeral n such that R ⊢ t = n. R ⊢ m1 + m2 = k. Proposition 68 R is a subsystem of Q. or t is (t1 f′′ t2 ). . use of induction for this argument is valid since we are arguing about Q and not in Q. We establish this result by showing that R satisfies the conditions D1 D2 D3 of Proposition 61. By the logic of identity. R ⊢ m1 = m1 and so by propositional logic. Proof. R ⊢ m1 = m2 . If t is a closed term it is either a numeral. so by reflexivity of identity in R. m1 ≤ m2 . and so by (i) provable in S. R ⊢ t′1 = n′ . (i) Suppose t1 and t2 are terms such that t1 = t2 is a true sentence. Then t1 and t2 have no free variables. t1 + t2 . . t1 · t2 . Then by Ω1 . R ⊢ (m1 = 0 ∨ . Ω1 : X is of the form m + n = m + n for natural numbers m and n.R ⊢ t1 + t2 = k. If t is of the form t′1 for some term t1 .e. Proof. i. Then by the logic of identity. then by induction hypothesis there are numbers m1 and m2 such that R ⊢ t1 = m1 and R ⊢ t2 = m2 . D2 . Since t1 = t2 is true. so m1 and m2 are the same numeral. ∨ m1 = m1 ∨ . The argument is the same for t of the form t1 · t2 . If t is a numeral n then R ⊢ t = n by reflexivity of identity. . (ii) Suppose t1 and t2 are terms such that t1 ≤ t2 is a true sentence. R ⊢ m1 ≤ m2 . R ⊢ t1 ≤ t2 . We must show that each axiom of R. Proof. is provable in Q. Proposition 67 The system R is Σ0 -complete. Hence by transitivity of identity in R. Ω3 . Ω4 . or there is a closed term t1 such that t is the term t′1 . i. By Theorem 65. Then by substitutivity of identity.e. Ω2 . i. This is the schema Ω3 . ∨ m1 = m2 ). We argue by induction on n in the statement Q ⊢ m + n = m + n. But n′ is n + 1. so by Lemma 66 there are natural numbers m1 and m2 such that t1 = m1 and t2 = m2 are true sentences. or there are closed terms t1 and t2 such that t is (t1 f′ t2 ). This is a direct consequence of Ω4 by ∀-Intro and ∀-Elim. m1 = m2 . Hence by Lemma 66 there are numbers m1 and m2 such that R ⊢ t1 = m1 and R ⊢ t2 = m2 .e. D3 . with ∀-introduction and ∀-elimination. If t is of the form t1 + t2 . . .LECTURE 7 63 Lemma 66 (evaluation of closed terms by R) For each closed term t in L. and Modus ponens. Since t1 ≤ t2 is true. which is to say that R ⊢ t = n + 1. Then by substitutivity of identity. Note that while Q does not contain induction. D1 . the terms t1 and t2 contain no variables. every instance of Ω1 .

LECTURE 7

64

n = 0. By N3 , ∀-I and ∀-E, Q ⊢ m + 0 = m. Since m = m + 0, m and m + 0 are the same term (formal numeral), so Q ⊢ m + 0 = m + 0. Induction step. (1) (2) (3) (4) (5) (6) (7) Q⊢m+n=m+n Q ⊢ m + n′ = (m + n)′ Q ⊢ m + n′ = (m + n)′ Q ⊢ m + n + 1 = ((m + n) + 1) ((m + n) + 1) = (m + (n + 1)) (m + n) + 1) and (m + (n + 1)) are the same term Q ⊢ m + n + 1 = m + (n + 1) Induction hypothesis N4 , ∀-I, ∀-E (1), (2), substitutivity of = in Q (3), Corollary 8 truth of arithmetic (5) (4), (6)

Ω2 : We show by induction on n that Q ⊢ m · n = m · n. n = 0 By N5 , ∀-Intro, ∀-Elim, Q ⊢ m · 0 = 0. Since 0 = m · 0, 0 and m · 0 are the same term. So Q ⊢ m · 0 = m · 0. Induction step. (1) (2) (3) (4) (5) (6) (7) (8) Q⊢m·n=m·n Q ⊢ m · n′ = m · n + m Q ⊢ m · n′ = m · n + m Q⊢m·n+1=m·n+m Q⊢m·n+1=m·n+m m · n + m = m · (n + 1) m · n + m and m · (n + 1) are the same term Q ⊢ m · n + 1 = m · (n + 1) Induction hypothesis N6 , ∀-I, ∀-E (1), (2), substitutivity of = in Q (3), Corollary 8 (4), previous case for Ω1 truth of arithmetic (6) (5), (7)

Ω3 : To show that for every m and n such that m = n, Q ⊢∼ m = n. Suppose that m = n. Without loss of generality, we may suppose that m > n, since by logic of identity in Q, if Q ⊢∼ m = n, then Q ⊢∼ n = m. We argue by cases. n = 0. Then since m = 0, there is a number k such that k + 1 = m. By ′ ′ Corollary 8, m is k . By N2 , ∀-Intro, ∀-Elim, Q ⊢∼ k = 0, i.e. Q ⊢∼ m = 0. n = 0. Since m > n, there is a non-zero number d such that m = d + n. By taking d in place of m in the argument for the previous case, Q ⊢∼ d = 0. By ′ contraposition of N1 , ∀-Intro, ∀-Elim, Q ⊢ (∼ d = 0 ⊃∼ d = 0′ ), so by Modus ′ ponens, Q ⊢∼ d = 0′ . By n many applications of this argument,
n n

= 0′ . . . ′ , i.e. Q ⊢∼ m = n. Q ⊢∼ d Ω4 : We show by induction on n that for each n, Q ⊢ (v1 ≤ n ⊃ (v1 = 0∨. . .∨v1 = n)).

′...′

LECTURE 7

65

n = 0. By ∧-Elim from N7 , Q ⊢ (v1 ≤ 0 ⊃ v1 = 0). Assume, as induction hyposthesis, that Q ⊢ (v1 ≤ n ⊃ (v1 = 0 ∨ . . . ∨ v1 = n)). By ∧-Elim, ∀-Intro, ∀-Elim from N8 , Q ⊢ (v1 ≤ n′ ⊃ (v1 ≤ n ∨ v1 = n′ )). Then by ∨-Elim, Q ⊢ (v1 ≤ n′ ⊃ ((v1 = 0 ∨ . . . ∨ v1 = n) ∨ v1 = n′ )), i.e. by Corollary 8, Q ⊢ (v1 ≤ n + 1 ⊃ (v1 = 0 ∨ . . . ∨ v1 = n ∨ v1 = n + 1)). Ω5 : N9 ⊢ Ω5 by ∀-Intro and ∀-Elim. Proposition 69 Q is Σ0 -complete. Proof. By Propositions 67 and 68. Theorem 70 PA is Σ0 -complete. Proof. By Proposition 69 and the fact that PA is an extension of Q.

7.4

Σ0-soundness and Σ1-soundness

Definition 56 (Σ0 -soundness) A system S is Σ0 -sound if and only if for every Σ0 -sentence X, if S ⊢ X, then X is true (in the structure of the natural numbers). Definition 57 (Σ1 -soundness) A system S is Σ1 -sound if and only if for every Σ1 -sentence X, if S ⊢ X, then X is true (in the structure of the natural numbers). Proposition 71 If a consistent system is Σ0 -complete, it is Σ0 -sound. Proof. Let S be a Σ0 -complete system and let X be a false Σ0 -sentence such that S ⊢ X. Since X is Σ0 and false, ∼ X is Σ0 and true. Hence by Σ0 -completeness of S, S ⊢∼ X. But this means that S is inconsistent, contrary to hypothesis. Remark. Though a system is Σ0 -complete if and only if it is Σ1 -complete (Proposition 58), there is no result corresponding to Proposition 71 that holds for Σ1 -completeness. As we shall see in Lecture 8, there are Σ1 -complete systems that are not Σ1 -sound. In any case it is clear that the proof of Proposition 71 does not extend to the case of Σ1 -completeness since in general the negation of a Σ1 -sentence is not Σ1 .

Lecture 8 The notions of consistency, ω-consistency and 1-consistency; incompleteness from the assumption of 1-consistency; truth of the G¨del sentence; o ω-incompleteness.
(Wednesday, 3 November 2010)

8.1

The notions of consistency, ω-consistency and 1-consistency.

Definition 58 (consistency) A system S is consistent if there is no formula X in the language of S such that S ⊢ X and S ⊢∼ X. Proposition 72 A system S containing propositional logic is consistent if and only if there is formula Y in the language of S such that S Y . Proof. (i) Left to right: We prove the contrapositive. Suppose for every X, S ⊢ X. Then in particular for any formula Y , S ⊢ Y and S ⊢∼ Y . (ii) Right to left: We prove the contrapositive. Suppose S is inconsistent, i.e. there is a formula Y such that S ⊢ Y and X ⊢∼ Y . Then by ∧-introduction (as a derived rule if not a primitive rule of S), S ⊢ (Y ∧ ∼ Y ). Propositional logic proves 66

e. Then ∃wF (w) is true. Definition 60 (ω-consistency) A system S in a language that contains a closed term n. then it is consistent. Suppose S ⊢ ∃wF (w). Proof. S ⊢ Z for every formula Z. i. which is to say that ∼ F (n) is false. for each natural number. and in general ω-consistency does not imply truth. . Remark. Exercise (problem 2(b) of Problem sheet 4). Let S be a system whose language contains numerals for the natural numbers and which is sound with respect to truth in arithmetic. S is ω-inconsistent. then X is true. in particular for any formula F (w) with one free variable. then it is ω-consistent. it is Σ2 -sound. Exercise (problem 3(a) of Problem sheet 4). a numeral. i. if sentence X is Σ2 and S ⊢ X. i. Proposition 75 If a system is sound with respect to truth in arithmetic. The converse of Proposition 73 does not hold. The contrapositive is immediate by ex falso quodlibet: if a system S is inconsistent S proves every formula in the language of S.e. i. Proof. and for each n S ⊢∼ F (n). is said to be ω-consistent if and only if there is no formula F (vi ) with one free variable in L such that S ⊢ ∃vi F (vi ) and for each natural number n. Proposition 72 shows that we could equivalently have defined consistency by: Definition 59 (alternative definition of consistency) S is consistent if there is a formula X such that S X. So by Modus Ponens in S. i.e. S ⊢ ∃wF (w).e. The converse holds only to a strictly limited extent. Proposition 74 There are consistent systems that are ω-inconsistent. i. Proof. Remark This result is best possible. as detailed by the following two propositions.LECTURE 8 67 ((Y ∧ ∼ Y ) ⊃ Z) for every formula Z. Proposition 76 If a system S is Σ0 -complete and ω-consistent. which is to say that S is ω-consistent. Proof. So S ∼ F (n). Proposition 73 If a system is ω-consistent.e. S ⊢∼ F (n).e. there is a natural number n such that F (n) is true.

S ∼ F (k). Then for each number n. so ∼ F (k) is a false Σ0 -sentence which. Proof. that all theorems are true (which we saw already in Lecture 1 is sufficient for the result). Hence by Σ0 -completeness of S. we shall see later (but only after we have proved the Second Incompleteness Theorem) that 1-consistency is also stronger than necessary for this result. Proof. (ii) Suppose S is Σ1 -sound. which shows that ω-consistency only implies a very limited amount of truth. with just the assumption that there is no ω-inconsistency with a Σ1 -formula. Since ∃vi F (vi ) is a Σ1 -sentence. G¨del introduced the notion of ω-consistency in order to prove the second half o of his First Incompleteness Theorem. i. and suppose ∃vi F (vi ) is false.e. by Lemma 43. (i) Suppose S is 1-consistent.e.LECTURE 8 68 Proposition 77 There is an ω-consistent system that proves a false Σ3 -sentence. Hence S cannot prove a 1-inconsistency. ∼ F (vi ) is a Σ0 -formula. the First Incompleteness Theorem can be proved. This special case of ω-consistency perhaps strictly should be labeled something like Σ1 ω-consistency. is logically equivalent to a false Σ1 -sentence. S ⊢∼ F (n). Definition 61 (1-consistency) A system S in a language that contains a closed term n. but the label 1-consistency introduced by Kreisel has the virtue of brevity and is standard in the literature. Kreisel in 1957 [5] noted that the minimum case of ω-consistency. F (vi ) is Σ0 and since Σ0 -formulas are closed under negation.e. This violates the hypothesized 1-consistency of S. Let S be a Σ0 -complete system.e. ∼ F (n) is true. is said to be 1-consistent if and only if there is no Σ1 -formula ∃vi F (vi ) with one free variable in the language such that S ⊢ ∃vi F (vi ) and for each natural number n. let ∃vi F (vi ) be a Σ1 -sentence such that S ⊢ ∃vi F (vi ). Even so. is sufficient for the second half of G¨del First Incompleteo ness Theorem. S ⊢∼ F (n). While 1-consistency is both weaker and more natural than ω-consistency as a hypothesis for proof of the second half of the First Incompleteness Theorem. S is 1-consistent. 1-consistency is equivalent to Σ1 -soundness. S is Σ1 -sound. as we shall see. for each natural number n. From the fact that a proof predicate for a formal deductive system with arithmetized syntax is Σ1 . Then by Σ1 -soundness of S. and suppose S ⊢ ∃vi F (vi ). which he labeled 1-consistency. for each natural number. so ∃vi F (vi ) is true. Exercise (problem 4(c) of Problem sheet 4). Theorem 78 For a Σ0 -complete system. Then by Σ1 -soundness of S. i. i. i. It is a very much weaker hypothesis than that the system is sound. ω-consistency is a considerably stronger hypothesis than is necessary to established formal incompleteness. established by Proposition 77. there is a natural number k such that F (k) is a true Σ0 -sentence. a numeral. .

This part of the proof of Theorem 78 is a sharpening of the argument for Proposition 75.e. The proof is by vacuous quantification. Proof. Then ∃v1 F (v1 ) is true. S ⊢∼ ∀v2 F (n. v2 ). Definition 62 (2-consistency) A system S in a language that contains a closed term n. 2-consistency is equivalent to Σ2 -soundness. where F (v1 . (i) Suppose S is 2-consistent and suppose S ⊢ ∃v1 ∀v2 F (v1 . in that Theorem 80 1-consistency is strictly weaker than ω-consistency. v2 ) is true. v2 ) is a true Σ1 -sentence. and ∃v1 ∀v2 F (v1 . then it is Σ1 -sound Proof. Then by the Σ1 completeness of every Σ0 -complete theory and predicate logic. Lemma 81 If a system is Σ2 -sound. . Or we can prove the contrapositive. v2 ) is true. ∃v1 ∀v2 F (v1 ) is true. Theorem 82 For a Σ0 -complete system. ∀v2 F (k. v2 ) is true. i. v2 ). an inconsistent theory proves everything so in particular a 1-inconsistency. Exercise problem 3(b) on Problem sheet 4. S ⊢∼ F (n). (ii) Suppose S is Σ2 -sound and suppose S ⊢ ∃v1 ∀v2 F (v1 . as in the proof of Proposition 73. v2 ) is a Σ0 -formula.e. Proof. Let S ⊢ ∃v1 F (v1 ) for F (v1 ) a Σ0 -sentence (i. v2 ). So by RAA. no free variables). By the first half of Theorem 78 and the fact that a Σ1 -sound system is consistent (since there are sentences it doesn’t prove. a numeral. Proof. But then S is 2-inconsistent.e. v2 ) is false. for each natural number. which is to say that for each natural number n. namely false Σ1 -sentences). Corollary 79 Every 1-consistent system is consistent. Suppose S ⊢∼ ∀v2 F (k. for each natural number n. so for some number k. The proof of incompleteness from 1-consistency is strictly a stronger result than proving incompleteness from ω-consistency. i. and so by Σ2 -soundness. G¨del’s original proof made no use of the extra strength of ω-consistency o over 1-consistency. Then ∃v1 ∀v2 F (v1 . Then S ⊢ ∃v1 ∀v2 F (v1 ). is said to be 2-consistent if and only if there is no Σ2 -formula ∃vi F (vi ) with one free variable in the language such that S ⊢ ∃vi F (vi ) and for each natural number n. However. so the proof from 1-consistency is really an improvement in clarity rather than in strength.LECTURE 8 69 Note that in (ii) of the proof of Theorem 78 the hypothesis that S is Σ0 -complete is not needed. ∃v1 ∀v2 F (v1 . ∃v2 ∼ F (n. v2 ).

Let ∃v2 A(v1 . v2 ) be a Σ1 formula that expresses {n : PA ⊢ En [n]}. Let a =df ∀v2 ∼ A(v1 . Remark. Proof.and 2-consistency. but the argument used is exactly the same as for G¨del’s result. By Lemma 12 and the completeness of first-order logic of PA. This o theorem is significantly stronger than the version of incompleteness we have already proved.) o Since we are no longer working with the notion of truth (in the structure of the natural numbers). that everything PA proves is true (in the structure of the natural numbers). Since ω-consistency implies n-consistency for each n and so in particular 3-consistency. This means that S is 2-consistent. we cannot use the Diagonal Lemma. by Lemma 81 it is Σ1 -sound. so the theorem is stronger. but these notions are not natural in the way 1. which is strictly weaker than the hypothesis that PA is sound. But this contradicts the truth of ∀v2 F (k. there is a 3-consistent system that is not Σ3 -sound. Theorem 17. PA G. {n : PA ⊢ En [n]} is Σ1 . v2 ) is true. PA ∼ G. (We replace G¨del’s hypothesis of ω-consistency by the strictly weaker hypothesis o of 1-consistency. If PA is 1-consistent. v2 ). v2 ) . we use the construction of the diagonal sentence for the one-place formula ∼ P rP A (v1 ) from the proof of the Diagonal Lemma to obtain the same G¨del sentence for this theorem as for the o weaker one. 1. S ∼ ∀v2 F (k. If PA is consistent. v2 ).e. We could define 3-consistency and n-consistency for larger n exactly as for 1. so by RAA. However. v2 ). Then . PA ⊢ (∀v2 ∼ A(a. Since S is Σ2 -sound. v2 ). and let G =df ∀v2 ∼ A(a. v2 ).and 2-consistency are since there is no equivalence with truth corresponding to Theorems 78 and 82. and 2. i. v2 ) ≡ ∀v2 ∼ A([a].2 Incompleteness of PA from the assumption of 1-consistency We are now in a position to prove G¨del’s First Incompleteness Theorem. Theorem 27. By Proposition 54. It is also a purely syntactic (finitary) property as opposed to a semantic (infinitary) property.LECTURE 8 70 Then S ⊢ ∃v2 ∼ F (k. so then PA ⊢ ∀v2 ∼ A([a]. v2 )). by Theorem ?? and Proposition ??. Suppose PA ⊢ G. so ∃v2 ∼ F (k. which establishes the truth of the diagonal equivalence. Theorem 83 (G¨del’s First Incompleteness Theorem for PA) There is a Π1 o sentence G such that 1. 8. in that it proves this result from the hypothesis that PA is 1-consistent.

2. v2 ). Hence for a =df ∀v2 ∼ A(vi . This contradicts the assumption that PA is consistent. (2) Formalizing the proof of Theorem 85 in PA shows that if PA is consistent it cannot prove its own consistency. i. which we . Though PA cannot decide G. i. v2 ) . Remarks (1) By Theorem 85. v2 ). But by the condition that PA is 1-consistent. so ∃v2 A(a. v2 ). there is a Σ1 -formula. On the assumption that PA is 1-consistent. PA ⊢ En [n] if and only if ∃v2 A(n.e. i. So PA G. Proof.e. v2 ) is true. since ∼ G is a false Σ1 -sentence. By Proposition 54.LECTURE 8 71 a ∈ {n : PA ⊢ En [n]}. v2 ) is true. Assume PA is 1-consistent. Lemma 84 G is true if and only if P A G. which with the assumption of this argument means that PA is inconsistent. ∃v2 A(a. Suppose PA ⊢∼ G. PA G if and only if G is true. Then PA is consistent. PA ∼ G. does not prove G and does not prove ∼ G. v2 ) expresses {n : PA ⊢ En [n]}. ∃v2 A(v1 . by Lemma 84. PA ⊢ Ea [a]. that expresses {n : PA ⊢ En [n]}. G is true. Theorem 85 If PA is consistent. This is equivalent to the first half of Theorem 83. we have good reason to hold that G is true (in the structure of the natural numbers) by the following considerations. which is to say that PA ⊢ G if and only if ∼ G is true. PA is Σ1 -sound.e. Since PA is Σ0 -complete and hence Σ1 -complete (Proposition 58). v2 ). G is true. Proof. since otherwise it would prove G. Since ∃v2 A(v1 . v2 ) is true. the First Incompleteness Theorem. and G =df ∀v2 ∼ A(a. we are justified in holding that the G¨del seno tence for PA is true insofar as we are justified in our conviction (universal among all mathematicians except a few with very quirky views) that PA is consistent. by Theorem 78. PA ⊢ G if and only if ∃v2 A(a. one or other of G and∼ G is true in the structure of the natural numbers. Corollary 86 Theorem 85 gives a variant proof of the second half of Theorem 83. PA ⊢ G. From the assumed 1-consistency of PA. Then.3 Truth of the G¨del sentence o By the bivalence of truth. Since ∃v2 A(v1 . PA ⊢ ∃v2 A(a. So by contraposition. PA is Σ1 -sound. PA ⊢ ∃v2 A(a. PA is consistent (Corollary 79). Proof.e. v2 ) is true. Identifying the basis of our conviction that PA is consistent lies outside the scope of these lectures. 8. so by Theorem 85. i. so PA ∼ G. v2 ) expresses {n : PA ⊢ En [n]}.

then for some number k. Formalizing the proof of Theorem 85 in PA requires some hard work. the truth of G follows from the fact that PA G. ∀v2 ∼ A(a. by Theorem 83 it cannot if it is consistent. Suppose G. i. F (k) is true and hence provable in any Σ0 -complete system. arithmetized theory that has a Σ1 -predicate that expresses {n : S ⊢ En }. which requires only the assumption that PA is consistent. the G¨del sentence for S is true. we established that PA ∼ G on the assumption that PA is 1-consistent. . So this argument is inefficient as a means to establishing that G is true. All of the results of the previous two sections generalize from PA to S. G for PA cannot be established as true on the basis of any considerations that do not also establish the consistency of PA. which I will do in Lecture 11.e. is true. Theorem 87 The truth of G implies the consistency of PA. This theorem is the contrapositive of the implication that every Σ0 complete system is Σ1 -complete (Proposition 58). by the converse of Theorem 85. which is to say that PA G. Proof. if ∃vi F (vi ) is true. which then by predicate logic also proves ∃vi F (vi ).e. and as we have seen. v2 ). There is another argument to show that G is true which is weaker than the argument for Theorem 85 because it requires a stronger hypothesis. Theorem 88 If a system is Σ0 -complete and does not prove ∃vi F (vi ) for F (vi ) a Σ0 -formula. which is to say that PA is consistent. o Proof.e. Theorem 88 tells us that G is true from the fact that PA ∼ G. Proof. for any such theory S. i. consistent. However. i. Then every number is not the code of a proof of Ea [a]. Let S be any Σ0 -complete. which also holds. it is important to realize that there is no weaker basis on which to hold that G is true than that PA is consistent. consistent. Σ1 -axiomatized theory. the G¨del o sentence G for S is a true Π1 -sentence and ∼ G is a false Σ1 -sentence.LECTURE 8 72 have just shown. but the argument itself is of independent interest. This is G¨del’s Second o Incompleteness Theorem. the hardest of which is formalizing the proof that PA is Σ1 -complete (provable Σ1 -completeness). (3) Having shown that G for PA can be seen to be true on the basis of the accepted consistency of PA. then ∃vi F (vi ) is false. But if there is any sentence that a system doesn’t prove then the system is consistent. In particular. Proposition 89 For S any Σ0 -complete.

4 PA is ω-incomplete Definition 63 (ω-completeness) A system S in a language containing a numeral n for each natural number n is ω-complete if for every formula F (v1 ) in the language of S such that for each natural number n S ⊢ F (n).LECTURE 8 73 8. We have also seen that if PA is consistent then G is true. Proof. .e. for each natural number n. i. i. v2 ). n) are Σ0 . Hence by the Σ0 -completeness of PA. We have shown that if PA is consistent. PA ∀v2 ∼ A(a. ∼ A(a. PA G. The sentences ∼ A(a. n). Theorem 90 If PA is consistent. S ⊢ ∀vi F (vi ) If S is not ωcomplete we say that S is ω-incomplete. n) is true. Hence PA is ω-incomplete. PA is ω-incomplete. for each n. PA ⊢∼ A(a.e.

though it does not supersede G¨del’s First Incompleteness o Theorem. where S is any system that can arithmetize its own syntax. discovered by J. This result is of definite interest. Rosser constructed 74 . 9 November 2010) We have noted that since an inconsistent system proves everything. for any sentence X and in particular for G the G¨del sentence for S.e. consistency of a system S is a necessary condition for S X. There turns out to be a form of incompleteness. The condition of ω-consistency is. We have also seen that consistency is a sufficient o condition for S G. incompleteness of PA from the assumption of consistency (Rosser’s Theorem).Lecture 9 Enumerability and the Separation Lemma. weak and strong definability of a function in a system. formal provability of the Diagonal Lemma (Tuesday. while consistency of S is sufficient to o establish that S G. and indeed to show that S ∼ G requires a stronger hypothesis than just that S is consistent. a stronger condition than consistency (and a weaker condition than that every sentence provable in the system is true). it does not show that the G¨del sentence for S is undecidable in S o on the assumption that S is simply consistent. that is symmetric with respect to negation. i. We saw that 1-consistency arises in a natural way as a condition sufficient to establish that S ∼ G for G the G¨del sentence for S. as we have seen. Barkley Rosser (1936).

. . m). . . . F (v1 . and let A and B be disjoint k-ary relations enumerated in S by formulas F (v1 . vk . vk ) separates A from B in S. . itself of independent interest and which we use also in proving that the diagonal equivalence in the diagonal lemma is not only true. . . . . but also formally provable. vk ) and any k-ary relations A and B. nk ). . . . . . . . . and for all (n1 . . . . vk ) separates a k-ary relation A from a k-ary relation B in a system S if and only if for all (n1 . vk . i) To show: if n ∈ A. then A and B are disjoint. . 9. If n1 . . . . . . . . . . then for every number m. S ⊢∼ F (n1 . . . vk . nk ) ∈ A. If n1 . vk ) separates B from A in S. . / Definition 65 A formula F (v1 . .1 Enumerability and the Separation Lemma Definition 64 (enumeration of a relation by a formula in a theory) A k-place relation R ⊆ Nk is enumerated by a formula F (v1 . The Rosser Incompleteness Theorem can be proved from a separation property. . The proof for A and B as k-ary relations is just a notational variant of this proof. m). . . . . . . Exercise. . then for any formula F (v1 . . respectively. x) ∧ (∀y ≤ x) ∼ G(v1 . (2) If F (v1 . Proof. x) and G(v1 . (We say that m is a witness to the fact that n1 . .) 2. . . Then the formula ∃x(F (v1 . . . vk . . . (3) If F (v1 . . . . nk ∈ R. then ∼ F (v1 . . . . . . . . x). vk . . then there exists a number m such that S ⊢ F (n1 . . . vk+1 ) in a system S if and only if: 1.LECTURE 9 75 a different sentence from the G¨del sentence which is provably undecidable just on o the hypothesis of consistency of the system. . Proof. . vk ) separates A from B in S and S is consistent. nk . x) ∧ (∀y ≤ x) ∼ G(n. Lemma 91 (1) If F (v1 . . nk ). (4) If S is inconsistent. nk ) ∈ B. . . . . . vk ) separates A from B in S. nk ∈ R. . vk ) does not separate B from A in S. S ⊢ F (n1 . . . . . . vk ) separates A from B in S and S is consistent. nk . nk ∈ R. . then S ⊢ ∃x(F (n. S ⊢∼ F (n1 . y)). Theorem 92 (Separation Lemma) Let S be a system in which Ω4 and Ω5 hold. as we have seen. In order to shorten formulas on the page I shall take A and B to be unary relations. . . . . . . . . y)) separates A from B in S. then F (v1 .

m) (5) S ⊢ ((y = 0 ∨ . x) ⊃ ∃y(y ≤ x ∧ G(n. y))) (13) S ⊢ ∀x(F (n. . k) ∧ ∀y(y ≤ k ⊃∼ G(n. v2 ) in S (4) and substitutivity of =. v2 ) in S (3) n ∈ B / (1) and the hypothesis that A and B as disjoint (3) and enumeration of B by G(v1 . k) ∧ ∀y(y ≤ k ⊃∼ G(n. ∨ y = k) ⊃∼ G(n. y))) (8) ∃-intro (10) S ⊢ ∃x(F (n. ⊃-intro (5). y)) (6) by ∀-Intro (7) S ⊢ ∀y(y ≤ k ⊃∼ G(n. ⊃-intro (5). S ⊢∼ G(n. which is logically equivalent to S ⊢ ∀x(F (n. y))) (12) S ⊢ ∀x(F (n. . y)). S ⊢∼ F (n. x) ∧ (∀y ≤ x) ∼ G(n. x) ⊃ (∃y ≤ x)G(n. y)) (9) definition of (∀y ≤ x) (ii) To show: if n ∈ B then S ⊢∼ ∃x(F (n. v2 ) in S (1) and the hypothesis that A and B as disjoint (3) and enumeration of A by F (v1 . k) (1) and enumeration of A by F (v1 .2 Incompleteness of PA from the assumption of consistency (Rosser’s Theorem) Theorem 93 If a formula H(v1 ) separates {n : S ⊢∼ En [n]} from {n : S ⊢ En [n]} in a consistent axiomatizable system S and S. x) ⊃ (y ≤ x ∧ G(n. x) ⊃ (∃y ≤ x)G(n. ∨ y = k) ⊃∼ F (n. y) ⊃ k ≤ y) (9) S ⊢ (F (n. ∨-elim. . ∨-elim. y))) (11) S ⊢ (F (n. y))) Assumption (1) and enumeration of B by G(v1 . y) ⊃∼ y ≤ k) (8) S ⊢ (F (n. . y)) (8) S ⊢ (F (n. x) ∧ (∀y ≤ x) ∼ G(n. instance of Ω4 and prop. y)) (1) n ∈ B (2) there exists k such that S ⊢ G(n. 9. S H(h) and S ∼ H(h). m) (5) S ⊢ ((y = 0 ∨ . logic (6) by prop logic (contraposition) (7) by Ω5 and prop logic (2) (8) propositional logic (9) ∃-Intro (10) predicate logic (anti-prenexing) (11) ∀-Intro (12) definition of (∃y ≤ x) Note that in the proof of the Separation Lemma. v2 ) in S (4) for every m. instance of Ω4 and prop. k))) (10) S ⊢ ∃y(F (n. the argument for (ii) uses both Ω4 and Ω5 while the argument for (i) uses just Ω4 . k) (3) n ∈ A / (4) for every m.LECTURE 9 76 (1) n ∈ A Assumption (2) there exists k such that S ⊢ F (n. then for h = H(v1 ) . . y)) (7) S ⊢ (F (n. x) ⊃ ∃y(y ≤ x ∧ G(n. y)) (4) and substitutivity of =. y)) (6) S ⊢ (y ≤ k ⊃∼ F (n. logic (6) S ⊢ (y ≤ k ⊃∼ G(n. y))) (2) (7) ∧-Intro (9) S ⊢ ∃x(F (n. y) ⊃ (k ≤ y ∧ G(n.

. vn+1 ) such that. . then the formula ∀v1 (P d(v3 . . . By Theorem 93. . v2 ) enumerates {n : S ⊢ En [n]} and the formula Rd(v1 . an . . 9. 4th edn. . Stephen Cole Kleene. . This notion is also sometimes called expressibility (e. (2) If f (a1 . Then by the separation property of H(v1 ). Proof. PA R and PA ∼ R. Lemma 94. Van Nostrand. an ) = b. By Theorem 92 and Lemma 91 (1). (ii) Suppose S ⊢∼ H(h). b). . D. Introduction to Metamathematics. This contradicts the assumption that S is consistent. v2 ) enumerates {n : S ⊢∼ En [n]}. . . v2 )) separates {n : S ⊢∼ En [n]} from {n : S ⊢ En [n]}. vn . . Proof. an . then S ⊢∼ F (a1 .3 Weak and strong definability of a function in a system Definition 66 (weak definability of a function in a system) A function f : Nn → N is weakly definable in a system S iff there is a formula F (v1 . S ⊢∼ H(h). .LECTURE 9 77 Proof. an ) = b. an . Then by the separation property of H(v1 ). an ) = b. 195). . . . . b) ∧ ∀vn+1 (G(a1 . vn+1 ) ⊃ vn+1 = b)). . Then h ∈ {n : S ⊢∼ En [n]}. an . Chapman and Hall. v1 ) ⊃ (∃v2 ≤ v1 )Rd(v3 . p. . 1950. . S ⊢ H(h). and the fact that the sets {n : PA ⊢ En [n]} and {n : PA ⊢∼ En [n]} are enumerable in PA. Theorem 95 (Rosser’s Theorem) There is a sentence R such that if PA is consistent. . vn . . . . 170). This contradicts the assumption S is consistent. Then since S ⊢ (H(h) ≡ H[h]). then S ⊢ (G(a1 . Introduction to Mathematical Logic. So S ∼ H(h). then S ⊢ F (a1 . . h ∈ {n : S ⊢ En [n]}. vn+1 ) such that if f (a1 . Elliott Mendelson. . Lemma 94 If S be a consistent axiomatizable extension of R in which the formula P d(v1 . . .g. So S H(h). . or numeralwise expressibility (e. . . 1997. (1) If f (a1 .g. b). . . . Definition 67 (strong definability of a function in a system) A function f : Nn → N is strongly definable in a system S iff there is a formula G(v1 . p. (i) Suppose S ⊢ H(h). . . .

. Then by the first conjunct of the condition for strong definability. that only number that bears the defining relation to the given argument is the value of the function for that argument.LECTURE 9 78 It’s easy to show that strong definability implies weak definability (Proposition 96). . b) ∧ ∀z(F (a. . Let F (x. . If f : Nn → N is strongly defined in S by the formula F (v1 . an ) = c. . y) be the formula (F (x. By ∀-elimination from the second conjunction of the condition for strong definability in S. . Proposition 96 Let S be a system in which Ω3 holds. S ⊢∼ c = b. i. So by propositional logic in S. Assume that f (a1 . an ) = b. . S ⊢ F (a1 . v2 ) that separates in S the graph of f (x) = y from its complement. The converse also holds but is considerably more complicated to prove (Theorem 97). Proof. then S ⊢∼ F (a. then it is weakly defined in S by the same formula.e. vn . an . S ⊢ (F (a. . b). We show that G(x. b) ∧ ∀y(G(a. . . . S ⊢ G(a. so f (a1 . The way we do this is to define a new formula from the formula that weakly defines the function with the additional condition that the relationship between the argument and a number holds just in case it’s the least number for which the weak definability relation holds. i. that any function weakly definable in a system S is strongly definable in S. . S ⊢∼ F (a1 . The argument for the general case is a notational variant. clause (2) of the definition of weak definability of a function. i. . . Theorem 97 If S is an extension of {Ω4 . To reduce clutter I give the proof for the case of a unary function. Weak definability of a function f (x) = y in S is the condition that there is a formula F (v1 . b). y) ⊃ y = b)). then any function weakly definable in S is strongly definable in S. We will now show. If a function is strongly definable in S then it is weakly definable in S. on the assumption that f (a) = b. (1) If f (a) = b.e. b). which is condition (1) of weak definability. . . b).e. . The first conjunct of the condition for strong definability is the same as the first clause for weak definability. y)∧∀z(F (x. We show that S proves both conjuncts. . an .e. Let G(x. an . . i. . b). . y) strongly defines f (x) in S. Proof. Ω5 }. then by Ω3 . i. z) ⊃ y ≤ z)).e. . . The second conjunct expresses the functionality condition for a given argument of the function. . Suppose c = b. (i) To establish the first conjunct. just with the use of Ω4 and Ω5 .e. then S ⊢ F (a. vn+1 ). if f (a) = b then S ⊢ (G(a. (2) If f (a) = b. i. If c = b. c) ⊃ c = b. which is also the case of immediate interest since the diagonal function is unary. z) ⊃ b ≤ z)): . S ⊢ F (a1 . y) be a formula that weakly defines f (x) in S.

y) ⊃ y = b) (7) By L3 . We know by Proposition 62 that Ω5 ⊢ b ≤ b. S ⊢ (v1 = k ⊃ (F (a. S ⊢ ∀y(F (a. (5) If k < b.e. S ⊢ (G(a. v2 ). S ⊢ (y = k ⊃ (G(a. (6) From (5) by substitutivity of identity. S ⊢ (F (a. for k < b. i. so f (a) = k. z) ⊃ y ≤ z) ⊃ (F (a. v1 ) ⊃ b ≤ v1 )) (5) Suppose k = b. v2 ). k). So it remains to show: S ⊢ (v1 ≤ b ⊃ (F (a. b).∨y = b) ⊃ (G(a. . y) ⊃ y = b). S ⊢ (G(a. so from (2). v1 ) ⊃ b ≤ v1 ) by ∨elimination from the instance of Ω5 for b. . then k = b. (2) Since (∀z(F (a. y) ⊃ y ≤ b). y) ⊃ y = b)). S ⊢ (y = b ⊃ (G(a. S ⊢ ((y = 0∨. k) ⊃ k = b). S ⊢ ∀y(G(a. S ⊢ (F (a. v1 = b established in (4) and (5). v1 ) ⊃ b ≤ v1 )). y) ⊃ y = b)). (2) To prove the second conjunct we establish (F (a. S ⊢ F (a. S ⊢ ∀y(G(a. so by clause (2) of weak definability.4 Formal provability of the Diagonal Lemma The Second Incompleteness Theorem is proved by proving within the system S the first half of the First Incompleteness Theorem. y) ⊃ (∀z(F (a. . . b). (7) From (6) by Ω4 and propositional logic. (4) We aim to show that S ⊢ (y ≤ b ⊃ (G(a. y) ⊃ y = b). Then by substitutivity of identity. v1 ) ⊃ b ≤ v1 )). Then by substitutivity of identity. k) ⊃ b ≤ k). (6) By ∨-elimination from the cases v1 = 0. then S G”. S ⊢∼ F (a. b) ⊃ b ≤ b). v1 ) ⊃ b ≤ v1 ). v1 ) ⊃ b ≤ v1 ). S ⊢ (v1 ≤ b ⊃ (F (a. S ⊢ (y ≤ b ⊃ (G(a. (11) From (10) by ∀-introduction. (8) By ∨-elimination from (6) and (7) and ⊃-introduction. v1 ) ⊃ b ≤ v1 )). (v1 ≤ b ∨ b ≤ v1 ).LECTURE 9 79 (1) Since f (x) in weakly definable in S by F (v1 . b) ⊃ y ≤ b)) is logically valid. k). (3) We have S ⊢ (b ≤ v1 ⊃ (F (a. “If S is consistent. y) ⊃ (F (a. where G is the G¨del sentence for S. (8) By ∨-elimination from Ω5 with (3) and (7). Then by propositional logic S ⊢ (F (a. S ⊢ F (a. which with (3) implies (ii). 9. (3) By condition (1) of weak definability. v1 ) ⊃ b ≤ v1 ))). b) ⊃ y ≤ b)). Then k = b. (9) by ∀-Intro from (8). (ii) To establish the second conjunct. y) ⊃ y = b): (1) By ∧-elimination S ⊢ (G(a. . o . y) ⊃ y = b)). . i. S ⊢ (v1 = b ⊃ (F (a. So by L3 and Modus ponens. (10) From (3) and (9) by propositional logic. from (1) it follows that S ⊢ (G(a. as an instance of L3 . z) ⊃ y ≤ z)). . . k). ∨ v1 = b) ⊃ (F (a. so S ⊢∼ G(a. (9) From (8) by Ω4 and propositional logic. So by propositional logic S ⊢ (G(a. y) ⊃ y = b)). (4) Suppose k < b.e. S ⊢ ((v1 = 0 ∨ . so by weak definability of f (x) by F (v1 . S ⊢∼ F (a. y) ⊃ b ≤ y).

Therefore . then by (2). S ⊢∼ A(a. b. Then by ∃-introduction in S (note that in inferring ∃v1 A(v1 ) from A(t). i. v2 . S ⊢ (H(n) ≡ G(f (n))). v2 ) ⊃ v2 = m). m). Proof. b. for each m.LECTURE 9 80 We began our proof of the First Incompleteness Theorem by establishing the Diagonal Lemma. y)) in S. so there is a number m such that A(a. We first show the following. there is a formula H(v1 ) such that for each number n. m). Theorem 99 (provable substitution) If a total function f (x) is strongly definable in a system S. (1) If the graph of a total function is Σ1 . (6) By Σ0 -completeness of S. d(a) = b iff ∃v3 A(a. v3 ). y)) separates d(v1 ) = v2 in S. b. ∼ A(a. v2 ) ∧ G(v2 ))). d(v1 ) = v2 is strongly definable in S. x) ∧ (∀y ≤ x) ∼ B(a. v2 . v2 . b. (3) By (1) and (2). We establish the provability of the two halves of the required biconditional as follows: (i) To show that S ⊢ (G(f (n)) ⊃ H(n)): Since F (v1 . v2 . the complement of its graph is expressed by a Σ1 -formula ∃v3 B(v1 . by clause (1) S ⊢ F (n. x)∧(∀y ≤ x) ∼ B(a. the diagonal function is strongly definable in S.e. For given formula G(v1 ) let H(v1 ) =df ∃v2 (F (v1 . so for every m. In our original proof of the Diagonal Lemma we proved the truth of the equivalence (C ≡ F ( C )). (10) Then by Theorem 97. S ⊢ A(a. ∃v3 A(v1 . x) ∧ (∀y ≤ x) ∼ B(v1 . y)). then by (2) ∃v3 A(a. the formula ∃x(A(v1 . v2 )∧G(v2 )). m). S ⊢ (G(m ⊃ (F (n. m) is true. which is to say that d(v1 ) = v2 is weakly defined by ∃x(A(v1 . m) is true. Lemma 98 For S any extensions of R. ∃v3 A(a. v2 . b. v2 ) be the formula that strongly defines f (x) in S. b. v3 ) is true. v3 ) is true. (5) By Σ0 -completeness of S. then S ⊢∼ ∃x(A(a. v2 . In arithmetizing the First Incompleteness Theorem for S in S we must show that S ⊢ (C ≡ F ( C )). (7) Hence the d(v1 = v2 is enumerated in S by A(v1 . v3 ). then for each formula G(v1 ). b. Let n and m be such that f (n) = m. y)) and if d(a) = b. Let F (v1 . then S ⊢ ∃x(A(a. S ⊢ (F (n. v2 . v2 ) strongly defines f (x). i. b. d(v1 = v2 is enumerated in S by B(v1 . v3 ). since S is an extension of R. Proof. (8) Similarly. (6) Suppose d(a) = b. d(a) = b iff ∃v3 B(a. b. since S is an extension of R. i. v3 ). (ii) To show that S ⊢ (H(n) ⊃ G(f (n))): By ∀-elimination from the second conjunct of the condition for strong definability. b. v3 ) is false. m) ∧ G(m)). v3 ) is true.e. if d(a) = b. x)∧(∀y ≤ x) ∼ B(v1 . S ⊢ (G(m) ⊃ ∃v2 (F (n. (9) Hence by the Separation Lemma (Theorem 92. (2) The diagonal function is total and its graph is expressed by a Σ1 -formula. Hence by propositional logic in S. the complement of its graph is also Σ1 ) (Lemma 34 in Lecture 5). b. b. (4) Suppose d(a) = b.e. v1 is substituted for some but not necessarily all occurrences of t in A(t)). v2 .

Since ((v2 = m ∧ G(v2 )) ⊃ G(m)) is logically valid (by substitutivity of identity). we have S ⊢ (C ≡ F ( C )). with v1 its only free variable. (F (n) ≡ F [n]) is logically valid. S ⊢ (H(h) ≡ F (d(h))). S ⊢ ((F (n. S ⊢ (∃v2 (F (n. for each formula F (v1 ) in the language of S. there is a formula H(v1 ) such that for each n. v2 ) ∧ G(v2 )) ⊃ (v2 = m ∧ G(v2 ))). In particular. Hence by ∀-introduction. Thus taking C =df H[h]. given F (v1 ). ⊢ (F (n. for h = H(v1 ) . By Theorem 99. . as required. Proof. So S ⊢ (H[h] ≡ F (d(h))). S ⊢ ∀v2 (F (n. S ⊢ ((v2 = m ∧ G(v2 )) ⊃ G(m)). d(h) = H[h] . there is a sentence C such that S ⊢ (C ≡ F ( C )).LECTURE 9 81 by propositional logic in S. and so by anti-prenexing. By the construction of d(x). so by propositional logic in S. v2 ) ∧ G(v2 )) ⊃ G(m)) Theorem 100 For S any extension of R. We have noted before that for any formula F (v1 ). v2 ) ∧ G(v2 )) ⊃ G(m). v2 ) ∧ G(v2 ) ⊃ G(m)). S ⊢ (H(n) ≡ F (d(n))). so S ⊢ (H(h) ≡ H[h]).

Lecture 10 Arithemization of consistency. analyzing and o strengthening the First Incompleteness Theorem (Wednesday. provability predicates. Definition 68 (definition of ConS ) For P r(v1 ) a formula that expresses {n : S ⊢ 82 . The first condition for G¨del’s Second Incompleteness Theorem to hold for a o system is that the consistency of the system be expressible by a sentence in the language of the system. Clearly the condition that S is consistent is necessary. S cannot prove the consistency of S. 10 November 2010) 10. since if S is inconsistent it proves everything. G¨del’s o Second Incompleteness Theorem. including any sentence in the language of S that expresses the consistency of S. L¨b’s Theorem.1 Arithmetization of the statement that a system S is consistent G¨del’s Second Incompleteness Theorem for a system S satisfying certain conditions o is the inference that if S is consistent.

e. This could. The process of formalization in S is intricate. P2 : S ⊢ (P ( (X ⊃ Y ) ) ⊃ (P ( X ) ⊃ P ( Y ))) P3 : S ⊢ (P ( X ) ⊃ P ( P ( X ) )) . The first of these three conditions we have already established (for PA). The Second Incompleteness Theorem is established by formalizing in S the proof of the first half of the First Incompleteness Theorem. the second is a problem on Problem sheet 5. Remark.LECTURE 10 83 En } and X any sentence in the language of S such that S X. the requirements for formalization are reduced to three conditions on o the proof predicate for S. as in ∼ 0 = 0′ . But with no work. for P r(v1 ) any formula in the language of S that expresses {n : S ⊢ En }. the formal expression in the language of S of the consistency of S be ∼ P r( X ). we have the following Lemma 101 (justifying the definition of ConS ) A system S is consistent if and only if ∼ P r( X ) is true. By work of Paul Bernays and Martin L¨b. Hence if S proves ConS . with a lot of work— which we have now done—be the G¨del sentence itself. and X any sentence in the langauge of S such that S X. and the third we will establish in Lecture 11. i. which. if S is consistent. Proof. then S proves that S G. and it is nearly standard to take ConS as ∼ P r( 0 = 0′ ). Immediate from Proposition 72 (in Lecture 8). immediate from axiom N2 ). except o as required to show S ⊢∼ X (which can be very little. in that case S ⊢ G. But since S ⊢ (G ≡∼ P r( G ). it doesn’t. then S ⊢ P ( X ). I will say something o later about provable invariance of the G¨del sentence and of ConS with respect to o G¨del numbering and arithmetization of syntax. then S G. nor relativity to the chosen G¨del numbering and arithmetization of syntax. We shall establish the Second Incompleteness Theorem from the assumption of these three conditions.2 Provability predicates. Definition 69 (provability predicate) A formula P (v1 ) is called a provability predicate for a system S if for all sentences X and Y in L(S) the following three conditions hold: P1 : If S ⊢ X. The relativity to X is sometimes dealt with by fixing on a particular X such that S X if and only if S is consistent. As to relativity of ConS to the o unprovable sentence X. if S is consistent. we let ConS . 10. The notation ConS does not notate relativity to the sentence X. we can take ConS to be ∼ P r( X ) for some particular such X.

The converse implication is an instance of Σ1 -soundness. the formula x = x is a provability predicate. Rather. i. there are pairs of formulas. . The effect of this is that these conditions on being a provability predicate do not require that a provability predicate expresses {n : S ⊢ En }. and follows from provable Σ1 -completeness of PA.] P6 : If S ⊢ (X ⊃ (P ( X ) ⊃ Y )). By P2 . then S ⊢ (P ( X ) ⊃ P ( Y )). The following three properties of a provability predicate are immediately provable from the three that define what it is to be a provability predicate. in particular. even in the same class of the arithmetical hierarchy. Then by P2 and Modus ponens.e. then by P1 . and X and Y any sentences in the language of PA.LECTURE 10 84 Note that P1 is a one-way implication and not a biconditional. P4 : If S ⊢ (X ⊃ Y ). Then by L2 and Modus ponens. S ⊢ ((P ( X ) ⊃ P ( P ( X ) )) ⊃ (P ( X ) ⊃ P ( Y ))) Then by P3 and Modus ponens. P4 If S ⊢ (X ⊃ Y ). S ⊢ P ( (X ⊃ Y )). S ⊢ (P ( (Y ⊃ Z) ) ⊃ (P ( Y ) ⊃ P ( Z ))). S ⊢ (P ( X ) ⊃ (P ( Y ) ⊃ P ( Z )). Proof. P6 If S ⊢ (X ⊃ (P ( X ) ⊃ Y )). For Property P2 we have Theorem 102 For P r(v1 ) the arithmetical proof predicate for PA constructed for Theorem 28. S ⊢ (P ( X ) ⊃ P ( (Y ⊃ Z) )). P5 If S ⊢ (X ⊃ (Y ⊃ Z)). then S ⊢ (P ( X ) ⊃ (P ( Y ) ⊃ P ( Z ))). Exercise (Problem sheet 5) Property P3 is the arithmetization of P1 . which I shall prove in Lecture 11. Then by propositional logic. it can express a superset of that set. S ⊢ (P ( X ) ⊃ P ( Y )). [Notice that there is no corresponding result for S ⊢ ((X ⊃ Y ) ⊃ Z). then by P5 S ⊢ (P ( X ) ⊃ (P ( P ( X ) ) ⊃ P ( Y ))). P5 : By P4 . PA⊢ (P r( X ⊃ Y ) ⊃ (P r( X ) ⊃ P r( Y ))). We established property P1 (and its converse) for the arithmetized proof predicate for PA by Theorem 28. S ⊢ (P ( X ) ⊃ P ( Y )). Lemma 103 For P (v1 ) a provability predicate for a system S. Proof. that have the same extension and one of them is a provability predicate and the other is not. Being a provability predicate is not extensional. then S ⊢ (P ( X ) ⊃ P ( Y )).

i. (1) S ⊢ (G ≡∼ P ( G )) (2) S ⊢ (G ⊃∼ P ( G )) (3) S ⊢ (G ⊃ (P ( G ) ⊃ X)) (4) S ⊢ (P ( G ) ⊃ P ( X )) (5) S ⊢ (∼ P ( X ) ⊃∼ P ( G )) (6) S ⊢ (∼ P ( G ) ⊃ G) (7) S ⊢ (∼ P ( X ) ⊃ G) (8) S ⊢∼ P ( X ) (9) S ⊢ G (10) S ⊢ P ( G ) (11) S ⊢∼ G (12) S is inconsistent (13) S ∼ P ( X ) Provable diagonal equivalence (1) ∧-Elim (2) Prop Logic (3) P6 (4) contraposition (1) ∧-Elim (5) (6) Prop Logic Assumption (7) (8) ⊃-Elim (9) P1 (10)(2) Prop Logic (9)(11) (8)(12) Consistency of S RAA Remark 1: Line (5) of this proof of the Second Incompleteness Theorem gives a formal proof in S of the first half of the First Incompleteness Theorem for S. The Second Incompleteness Theorem establishes that for every sentence X in the language of S.e. The heart of the matter. this is tantamount to S ∼ P ( G ). S G. if S is consistent.e.LECTURE 10 85 10. then indeed the implication holds. however. The converse of (7) for arbitrary X cannot be proved because it doesn’t hold. if S is 1-consistent: If S ⊢ (G ⊃∼ P ( X )) for X such that S ⊢ X. some sentence is unprovable.3 G¨del’s Second Incompleteness Theorem o Theorem 104 (G¨del’s Second Incompleteness Theorem) Let P (v1 ) be a provo ability predicate for a system S. and let G be a sentence in the language of S such that S ⊢ (G ≡∼ P ( G )). i. Proof. if S is consistent. . for G the G¨del sentence S. For X any sentence in the language of S. o S G. if S ⊢∼ X. By the diagonal equivalence. S ⊢ P ( X ). The first half of the First Incompleteness Theorem establishes. if S is consistent. is that the G¨del sentence for S is provably in S equivalent to the consistency of S. One dio rection of this equivalence is established at line (7) of the above proof of the Second Incompleteness Theorem: If S is consistent. S ∼ P ( X )). Remark 2: G¨del’s Second Incompleteness Theorem is a generalization of the o first half of the First Incompleteness Theorem. that if S is consistent. then G holds.e. S ∼ P ( X ). then since by P1 . i. S ⊢∼ G. However.

from L1 and Modus ponens.) Theorem 107 (L¨b’s theorem) Let S be a system in which the Diagonal Lemma o is provable.e.e. 10. S ⊢ P ( L ). Proof. i. i. i. For X any formula in the language of S. and let P (v1 ) be a provability predicate for S. by ∧-Elimination from the diagonal equivalence in (2). it can be used as the fundamental axiom for a theory of provability. Proof. (1) Assume that S ⊢ (P ( X ) ⊃ X). then S ⊢ X. S ⊢ (P r( H ) ≡ H)) provable (in which case it is true ) or unprovable (in which case it is false)? What Martin L¨b showed was that from just half of that o diagonal equivalence. If S ⊢∼ X. by ∧-Elimination from the diagonal equivalence in (2). (5) From (1) and (4) we have S ⊢ (P ( L ) ⊃ X) (6) S ⊢ ((P ( L ) ⊃ X) ⊃ L). S ⊢ (G ⊃∼ P ( X )). (3) S ⊢ (L ⊃ ((P ( L ) ⊃ X). (4) By P6 from (3). if S ⊢ (P ( X ) ⊃ X). (2) Let L be a provably diagonal sentence for the predicate (P (v1 ) ⊃ X). it follows that S ⊢ H (so H is provable and true). S ⊢ (P ( H ) ⊃ H). By Proposition 105 and line (7) of the proof of Theorem 104. It is also a generalization of the Second Incompleteness Theorem. (8) Then from (7) by P1 . It arose in response to an almost jokey question in the 1950s by Leon Henkin: Is the sentence that asserts its own provability (there is such a sentence. Then by P4 . as we shall see in Lectures 13 and 14. .e. S ⊢ (G ⊃∼ P ( X )).. Proof.e. by propositional logic S ⊢ (X ⊃ G). S ⊢ (P ( X ) ⊃ P ( G )). though this was not immediately realized. Then by the diagonal equivalence for G and transitivity of implication. S ⊢ (P ( L ) ⊃ P ( X )). (The converse holds by propositional logic. Corollary 106 For X any sentence such that S ⊢∼ X. S ⊢ (L ≡ (P ( L ) ⊃ X). (7) By Modus ponens from (5) and (6) S ⊢ L. if S ⊢ (P ( X ) ⊃ X).LECTURE 10 86 Proposition 105 For X any sentence such that S ⊢∼ X. i. by diagonalization.4 L¨b’s Theorem o L¨b’s Theorem is a deep result which characterizes the abstract properties of provo ability. then S ⊢ X. because for any sentence X and any provability predicate P (v1 ). By contraposition. S ⊢ (G ≡∼ P ( X )). S ⊢ (∼ P ( G ) ⊃∼ P ( X )).

In discussing the truth value of the G¨del sentence o for a system S. The Second Incompleteness Theorem proves L¨b’s Theorem in those o cases where S ⊢∼ X. that S ∼ G. Problem 4 on Problem sheet 5.5 Analyzing and strengthening the First Incompleteness Theorem The Second Incompleteness Theorem greatly increases our understanding of the First Incompleteness Theorem. which is to say that S proves everything. o though not uniformly but on a sentence by sentence basis. for which S ⊢ X.e. so that with respect to S itself. then by propositional logic. 87 Remark. The Second Incompleteness Theorem enables us to show that the second half of the First Incompleteness Theorem. Then by the contrapositive of the Second Incompleteness Theorem. Proof.LECTURE 10 (9) S ⊢ X by (8) and (5). Despite L¨b’s Theorem being a generalization of the Second Incompleteness Theo orem. S ⊢ X.e. S is inconsistent.e. i. we see that the meaning of G is “S is consistent”. i. A sentence of the form (P ( X ) ⊃ X) expresses the soundness with respect to provability of X of the system for which P (v1 ) expresses provability. L¨b’s Theorem says that the o only such statements that can be proved in a system are the ones that hold trivially by propositional logic. we noted that the truth of the G¨del sentence for S is implied by and o implies the consistency of S. S ⊢ X. The situation is the following: Theorem 109 The Second Incompleteness Theorem for S ∪ {∼ X} implies that if S ⊢ (P ( X ) ⊃ X). then X. cannot be established just from the hypothesis that S is consistent. X is true. S ⊢∼ P ( X ). it says that if X is provable. Theorem 108 L¨b’s Theorem is a generalization of G¨del’s Second Incompleteness o o Theorem. as follows: If S ⊢ ((P ( X ) ⊃ X)) and S ⊢∼ X. so in particular. L¨b’s Theorem goes beyond what is established by the Second Incompleteo ness Theorem and propositional logic by establishing the implication from S ⊢ ((P ( X ) ⊃ X)) to S ⊢ X for the case of X such that S X and S ∼ X. i. 10. L¨b’s Theorem can be proved from the Second Incompleteness Theorem. Proof. by . The Second Incompleteness Theorem establishes that the equivalence of G and Con S is provable in S.

(3) S ⊢ (X ⊃ (∼ G ⊃ X)).5. o namely consistency of S ∪{ConS}. Therefore (9) S ∗ ⊢ (∼ G ⊃∼ G∗ ). Then if S ∪ {ConS} is consistent. G is provably equivalent in S to ConS. Then by Lemma ??. Proof. (2) Then also (by thinning). This contradicts the hypothesized consistency of S ∪ {ConS}. Theorem 110 shows is that this fact cannot be proved without appealing to a property of PA stronger than it’s consistency. let P ∗ (v1 ) be a provability predicate for S ∗ . and let G be a sentence in the language of S such that S ⊢ (G ≡∼ P ( G )). let G be a sentence in the language of S such that S ⊢ (G ≡∼ P ( G )).1 S ∼ GS cannot be proved from the consistency of S This result can be established by showing that any of the most usual examples of consistent. and to show that this condition is strictly weaker than the 1-consistency of S. given the consistency of S. S ∪ {∼ GS } for consistent Σ0 -complete Σ1 axiomatizable theory S. for any sentence X. (7) S ∗ ⊢ (G ≡∼ P ( X )).LECTURE 10 88 allowing us to establish the existence of a consistent theory S such that S ⊢∼ GS . S ∗ ⊢ (P ( (∼ G ⊃ Y ) ) ≡ P ∗ ( Y )). let S ∗ be the system S ∪{∼ G}. Hence (10) S ∗ ⊢∼ G∗ . i. (1) Let X be such that S ⊢∼ X. by propositional logic. Proof. proves the negation of its G¨del sentence (not the same as o the G¨del sentence for S)—but also (as we shall see below) not every 1-inconsistent o theory proves the negation of its G¨del sentence.2 Strengthened second half of the First Incompleteness Theorem Theorem 111 (strengthened First Incompleteness Theorem) Let P (v1 ) be a provability predicate for a system S. Suppose S ⊢∼ G. Let X be any sentence such that S ⊢∼ X. by property of a provability predicate. It enables us also to find a non-trivial formulation of the minimum sufficient condition for unprovability of the negation of the G¨del sentence for system S. and similarly (8) S ∗ ⊢ (G∗ ≡∼ P ∗ ( X )). Let G∗ be a sentence such that S ∗ ⊢ (G∗ ≡∼ P ∗ ( G∗ )). 10. By the proof of the Second Incompleteness Theorem. Then S ∗ ⊢∼ G∗ . S ∗ ⊢∼ X. Then (4) S ⊢ (P ( X ) ⊃ P ( (∼ G ⊃ X) . Remark. o Theorem 110 For P (v1 ) a provability predicate for a consistent theory S. We know that P A ∼ GP A . Then (6) S ∗ ⊢ (P ( X ) ⊃ P ∗ ( X )) since for all Y . S ⊢∼ ConS. which is consistent. . 10.e. S ∼ G. and so also by thinning in S ∗ .5. Therefore S ∼ G. and so by thinning. Let ConS stand for ∼ P ( X ). 1-inconsistent theories. (5) S ∗ ⊢ (P ( X ) ⊃ P ( (∼ G ⊃ X) .

(5) Then by the Deduction Theorem. To show . (8) By arithmetization of the logical fact that if PA (or any other theory) does not prove the consistency of a theory then a fortiori it does not prove the consistency of an extension of that theory. the hypothesis of Theorem 111 is strictly weaker than the hypothesis of the second half of Theorem 83. P A+ ⊢ ConP A+ .5. if the hypothesis doesn’t hold. i. (10) By (6) and (9). P A ⊢ (ConP A ⊃∼ P rP A ( ConP A )). so ∼ ConP A+ is a false Σ1 -sentence. S ⊢∼ ConS. By Theorem ??. i. If P A+ is consistent. (13) Then by (1). P A+ ConP A+ .3 Consistency of S ∪ {ConS} is strictly weaker than 1consistency of S We establish this result by constructing a theory S such that if PA∪{ConP A} is consistent. S ⊢∼ ConS. (12) and RAA. Then by Theorem 78.e. the result to be proved doesn’t hold. (3) Similarly. (11). P A ⊢ (∼ P rP A ( ConP A )) ⊃∼ P rP A ( Con(P A ∪ {ConP A}) )).e. S is 1-inconsistent and S ∪ {ConS} is consistent. (ii) (1) Suppose S ∪ {ConS} is inconsistent. S ⊢ (∼ ConS ⊃∼ G). It is best possible since if S ∪ {ConS} is inconsistent. (12) From the assumption that P A ∪ {ConP A} =df P A+ is consistent. S is inconsistent if and only if P A ⊢ ConP A+ . (11) Then by Modus Ponens. Second Incompleteness Theorem for P A+ .e. S ∪ {ConS} is consistent. Theorem 112 Let P A+ =df P A ∪ {ConP A} and S =df P A ∪ {∼ ConP A+ }. (7) By arithmetized Second Incompleteness Theorem. so (4) S proves the inconsistency of S iff S ⊢ P rP A ( ConP A+ ). then S is 1-inconsistent and S ∪ {ConS} is consistent. then ConP A+ is a true Π1 -sentence. Proof.LECTURE 10 89 10. and is best possible. i. (2) Then by propositional logic. S is 1-inconsistent. and so by the contrapositive of (5). P A ∪ {ConP A} ⊢ ConP A+ . so then S ⊢∼ G. (6) P A ⊢ (∼ P rP A ( ConP A+ ) ⊃ ConP A+ ). P A ∪ {∼ ConP A+ } ⊢ P rP A ( ConP A+ ). (i) If P A+ is consistent. The hypothesis S ∪ {ConS } for the second half of the First Incompleteness Theorem not only makes the theorem best possible but is the natural hypothesis for this result. (9) By (7) and (8). By Theorem 112. P A ⊢ (ConP A ⊃∼ P rP A ( Con(P A ∪ {ConP A}) )). Proposition 113 Theorem 111 is stronger than Theorem 83. P A ⊢ (ConP A ⊃ ConP A+ ). The two parts of the First Incompleteness Theorem for a Σ0 -complete theory S that is Σ1 -axiomatizable are of different character. Proof.e. i. Remark. P A ⊢ (∼ ConP A+ ⊃ P rP A ( ConP A+ )).

ConS is true. . if S is consistent. i. whether o or not S ∼ G is equivalent to whether or not S is sound to the extent that it does not prove this false sentence. as noted in Theorem 85. Consequently. This result holds for every system for which a G¨del sentence can be constructed. We have seen that a theory can have this much soundness and still not be Σ1 -sound (Theorem 112). Given that S is consistent. Exactly the amount of soundness of S required for the second half of the First Incompleteness Theorem is that S does not prove the false Σ1 -sentence ConS.LECTURE 10 90 that S G requires the purely formal hypothesis that S is consistent. S ∪ {ConS } is consistent. We have seen that the first half of the First Incompleteness Theorem establishes the consistency and unsoundness of S ∪ {∼ G}. The consistency of S implies the truth of G. and we have also seen that S ∪ {∼ GS } ⊢∼ GS∪{∼GS } .e.

As we shall see. While provable Σ1 -completeness is a deep theorem whose proof is complicated. P3 . So (X ⊃ P r( X )) is true. Proof. PA is not ∆2 -complete. P r( X ) is a true Σ1 -sentence. P A ⊢ (X ⊃ P r( X )). PA ⊢ X. then by Σ0 -completeness of PA. Argument (1): (i) If X is true. So the provability of these sentences in PA is specific to provability properties of Σ1 -sentences and of arithmetization of provability in PA. Since for Y any sentence in the language of PA. and expresses {n : PA ⊢ En }. provable completeness for Σ0 -sentences is very easy to show. 16 November 2010) Proposition 114 For P r(v1 ) a formula in the language of PA that expresses {n : P A ⊢ En } and X any Σ1 -sentence in the language of PA. P r( X ) is true. so by propositional logic in PA. Proposition 115 For P r(v1 ) a formula in the language of PA that expresses {n : P A ⊢ En } and X any Σ0 -sentence in the language of PA. then (X ⊃ P r( X )) is true. In this lecture we will establish that all these true sentences are provable in PA (provable Σ1 -completeness). (X ⊃ P r( X )) is ∆2 . and since P r(v1 ) expresses {n : PA ⊢ En }. these sentences include all sentences of the form (P r( Y ) ⊃ P r( P r( Y ) )). the sentence (X ⊃ P r( X )) is true. and since P r(v1 ) is Σ1 . 91 . i.Lecture 11 Provable Σ1-completeness (Tuesday. Note that for X a Σ1 -sentence. the sentence P r( Y ) is Σ1 . the third condition on a provability predicate. Then by Σ1 -completeness of PA. (ii) If X is false. Proof. (i) If X is true. PA ⊢ X.e. then by Σ1 -completeness of PA. P A ⊢ P r( X ). P A ⊢ (X ⊃ P r( X )).

a Σ1 -sentence. i. Thus to have a strong enough induction hypothesis for the inductive argument. Remark about the strategy for the proof. we noted above that for X a Σ1 -sentence. It is not that P A ⊢ (X ⊃ P r( X )) for all Σ1 -formulas. For example. For Argument (2). So we need to find a way to express the condition that a substitution instance of a formula is provable. Remark. including ones with free variables. and 69. and Theorem 70). is explicit (rather than recursive). we must prove provable Σ0 -completeness for formulas that may contain free variables. PA ⊢ (P r( X ) ⊃ P r( ∀vi X )). which contradicts the 1-consistency of PA. so by Modus ponens. To do this we need first to modify the definition we gave in Lecture 2 of quasisubstitution (Definition 27). PA ⊢ 1 + 2 = 3. By logic of identity in PA. then (X ⊃ P r( X )) is Σ1 . but (ii) fails since the negation of a Σ1 -sentence is not.e. PA ⊢ (1 + 2 = 3 ⊃ P r( v1 + v2 = v3 )). PA ⊢ P r( v1 + v2 = v3 ). How to formulate this result for Σ0 -formulas with free variables requires some thought. i. For Argument (1). is recursive. and as we shall see later. Argument (2): If X is Σ0 . Then PA ⊢ P r( 1 + 0 = 0 ).e. Proof. Neither of the two arguments for Proposition 115 can be extended to the case of X a Σ1 -sentence. so by propositional logic in PA. so the proof of provable Σ0 -completeness must proceed by induction over the recursive definition of Σ0 -formulas. the proof of provable Σ1 -completeness has to go via a proof of provable Σ0 -completeness. so by Σ0 -completeness. The result we need to establish is that PA proves that each true substitution instance of a Σ0 -formula is provable in PA. —— So as in our proof that R and thereby Q and PA are Σ1 -complete (Propositions 58. The sequence of formulas by which a Σ0 -sentence is generated by this recursion will in general contain free variables. 67. By arithmetization of the rule R1 . Generalization. Definition 41. PA is not ∆2 -complete. in general. Hence PA ⊢ (P r( ∀v1 ∀v2 ∀v3 v1 + v2 = v3 ) ⊃ P r( 1 + 0 = 0 )). then by instantiation. The definition of Σ0 -formula. PA ⊢ (∀v1 ∀v2 ∀v3 v1 + v2 = v3 ⊃ 1 + 0 = 0). (i) holds for X a Σ1 -sentence. Definition 40. (X ⊃ P r( X )) is ∆2 . a Σ1 formula is any formula of the form ∃vi F where F is a Σ0 -formula. Hence PA ⊢ P r( ∀v1 ∀v2 ∀v3 v1 + v2 = v3 ).LECTURE 11 92 (ii) If X is false. then ∼ X is a true Σ0 -sentence. Proposition 116 If PA is 1-consistent. IF PA ⊢ (v1 + v2 = v3 ⊃ P r( v1 + v2 = v3 )). PA (v1 + v2 = v3 ⊃ P r( v1 + v2 = v3 )). The definition of what it is to be a Σ1 -formula. and hence provable in PA by Σ1 -completeness. P A ⊢ (X ⊃ P r( X )). which was defined just for substitution on the free . P A ⊢∼ X.

. km . v2 ) represents a total function s(v1 ) = v2 in a theory T . . y. . since (∀v1 ∃v2 ∀v3 (S(v1 . vkm ) as ∀vk+1 . . z. . so by generalization of Theorem 39. . vkm ) any formula in the language of PA with exactly the free variables shown. Further remark: The notation P r[ F (vk1 . . . . . . . Definition 70 (arithmetized proof predicate with free variables) For P r(v1 ) a formula in the language of PA that expresses {n : P A ⊢ En } and F (vk1 . 5 26 5 . . ∧ S(vk+m−1 . P r[ F (vk1 . . . P r[ F (vk1 . vkm . vkm ). we can express the substitution F (s(v1 )) either by ∀v2 (S(v1 . ∀vk+m ((S( F (vk1 . vkm ) ](vk1 . . . . n2 . Remark: Where a formula S(v1 . . k2 . . . . vk2 . .e. vkm ) ](vk1 . . So strictly we could have defined P r[ F (vk1 . . z z z z The function f (z) = 5 . vkm ) ](vk1 . . . . vk1 . km }. ∀vz (vz = y ⊃ Ex ) = ∀v′ . Note that this abbreviation cannot be used if we need to show the result of making a substitution for a free variable of P r[ F (vk1 . n2 . vkm ) ](vk1 . . n3 ) = n4 . . y. v2 )∧F (v2 )). k2 . . . . . k1 . v3 ) ⊃ v3 = v2 ) ⊃ (∀v2 (S(v1 . Σ1 . . . . . . vk+2 ) ∧ . n3 . vkm ) have the same free variables. i. so the relation s(x. vkm ). . .LECTURE 11 93 variable v1 . s(x. . z) = ∀vz (vz = y ⊃ Ex ) Proposition 117 For the function s(x. vkm ) is. vkm ) =df ∃vk+1 . vk+1 ) ∧ S(vk+1 . to allow substitution on any specified variable. . . . . . . . n3 . vkm ) is not a sub-formula of P r[F (vk1 . and o not the formula itself.e. . . y) = ∀v1 (v1 = y ⊃ Ex ) . and k = max{k1 . . given that P r(v1 ) is Σ1 . vkm )]. . . vk+m )) ⊃ P r(vk+m )). . . . . . . y. vkm )]. vkm ) . vk+1 ) ∧ S(vk+1 . v2 ) ⊃ F (v2 )) ≡ ∃v2 (S(v1 . . ∧ S(vk+m−1 . vk1 . . that occurs in P r[ F (vk1 . . k1 . . vkm ) . . w) in the language of PA such that for all natural numbers n1 . . this formula is Π2 . We do not use this latter formula as the definition since. . to avoid clutter. . . . . z) = ∀vz (vz = y ⊃ Ex ) there is a Σ1 -formula S(x. . . .′ = y ⊃ Ex ) = 96 5 . f (x) = y is Σ1 . . . 5 is generated by the following primitive recursion: z f (0) = 5 f (n + 1) = 5 · 13n+1 + f (n). . . . z) = z is Σ1 . . s(x. km . which must be read bearing in mind that F (vk1 . . . . . Occasionally. vkm . . .′ (v′ . n4 . . Note that F (vk1 . vkm ) ](vk1 . vkm ) ](vk1 . . . . . . vkm ) stresses the important point that it is the G¨del number of the formula F (vk1 . vk2 . n2 . . 5 η ∗ 13y ∗ 8x3. . . i. vkm ) and P r[ F (vk1 . v2 ) ⊃ F (v2 )) or by ∃v2 (S(v1 . . whereas on the given definition. vk+2 ) ∧ . S(n1 . . . . . . . . . . ∃vk+m ((S( F (vk1 . y. vkm ) ](vk1 . . . Proof. like P r(v1 ). vkm ). n4 ) is true if and only if s(n1 . . we will abbreviate this formula as P r[F (vk1 . v2 ) ∧ F (v2 )))) is logically valid. . . vk+m )) ∧ P r(vk+m )). .

. Exercise. PA ⊢ P r( F [n] ). .g. . Lemma 122 For all formulas F (vj1 . v2 ) ⊃ P r(v2 )). vkm ) and G(vr1 . vkm ). then for each sentence that results from substituting numerals for the free variables of that formula. . PA proves the proof predicate for PA applied to the G¨del number o of that sentence. . . . . By the definition of P r[ F ]. etc. . .LECTURE 11 Proposition 118 For a formula F with no free variables. We need to formalize this argument in PA. am . . vrs ) ⊃ (P [F ](vk1 . . v1 . then by logic in PA. . vkm ). . . am ) is true if and only if P r( F (a1 . 94 Proof. . . 0′ . vr1 . . . . . Then by P1 . . . P r[ F (vk1 . . vkm ) ](a1 . . vkn ) ⊃ P [ (F ∧ G) ])(vj1 . e. . It consists of arithmetizing the following argument. . . . For F with no free variables. PA ⊢ P r( (∀vk F (vk ) ⊃ F [n]) ). . ∗ Theorem 120 (P1 = P1 generalized to allow free variables) For any formula F (vk1 . . . . . . am ) ) is true. Proposition 119 For any natural numbers a1 . . vjm ) and G(vk1 . Then by Modus pones. . ∀v1 (S( F . . . . which for simplicity I formulate for the case in which F has just one free variable. . The proof is by induction on the number of free variables in F (vk1 . . . . . then PA ⊢ P r[ F (vk1 . . . Proof. Then by P2 . . . vkn )) Proof Exercise. . Proof. PA ⊢ (P r[ F ] ≡ P r( F )). vkm ) ⊃ P [G](vr1 . (∀vk F (vk ) ⊃ F [n]) is logically valid. . . . . vrs ) in the language of PA. . vrs ))). . . . . If PA⊢ F (vk ). PA ⊢ P r( ∀vk F (vk ) ). . For each natural number n. . . . . We need to generalize P1 and P2 to allow for occurrence of free variables. . The generalization of P1 expresses that if a formula with free variables is provable in PA. . and provable equivalence of F (a) and F [a]. . consider P r[ F ] with one vacuous quantifier. . if PA ⊢ F (vk1 . . P A ⊢ (P [ F ](vj1 . vkm ). . . . . . . . vjm . . so PA⊢ (∀vk F (vk ) ⊃ F [n]). vkm ) in the language of PA. . . vkn ). Then by P1 . PA ⊢ (P r( ∀vk F (vk ) ) ⊃ P r( F [n] )). vkm ) ](vk1 . ∗ Theorem 121 (P2 = P2 generalized to allow free variables) For any formulas F (vk1 . PA⊢ ∀vk F (vk ). Proof. . vkm . vk1 . . PA⊢ (P [(F ⊃ G)](vk1 . vjm ) ⊃ (P [ G ](vk1 . .

Our proof in Lecture 7 that PA is Σ0 -complete went by way of proving the very strong result that the extremely weak system R is Σo -complete and then showing that R is a subsystem of PA. but that would be a very roundabout way of proving the provable Σ0 -completeness of PA in PA. F (vi /t) is the result of substituting the term t for all occurrences of the variable vi in the formula F (vi ). Lemma 123 Let F (vi ) be a formula with free-variable vi . It could be formalized in PA. and correspondingly for F with any number of variables. Then the following equivalence is logically valid: (P r[ F (vi )](t) ≡ P r[ F (vj ) ](t)). Proving Σ0 -completeness requires heavy use of mathematical induction. so that proof cannot be formalized in R or Q. I will illustrate an informal direct proof of the Σ0 -completeness of PA that we will be formalizing in PA. then PA ⊢ a + b = c. Rather what we want to do is formalize in PA a direct proof of the Σ0 -completeness of PA. . Lemma 124 Let F (vi ) be a formula with free-variable vi and let t(vj ) be a term free for vi in F (vi with variable vj distinct from vi . substitutivity of =. Let vj be a variable and t a term both free for vi in F (vi ). Lemma 125 If a + b = c. From the definition of P r[ F (vi )](vi ).LECTURE 11 95 Definition 71 A term t is free for variable vi in formula F (vi ) if vi in F (vi ) does not occur within the scope of a quantifier whose variable of quantification is a free variable in t. Then the following equivalence is logically valid: (P r[ F (vi /t(vj )) ](vj ) ≡ P r[ F (vi ) ](t(vj ))). Definition 72 For F (vi ) a formula with free variable vi and t any term. and logical equivalence of F (vi ) and F [vi ]. substitutivity of =. by going through an informal proof for the case of an atomic Σ0 -formula of the form v1 + v2 = v3 . Proof. and logical equivalence of F (vi ) and F [vi ]. and correspondingly for F with any number of variables. From the definition of P r[ F (vi )](vi ). Proof.

. The proof of this case is by a double induction over the recursive definition of terms. Theorem 126 (provable Σ0 -completeness with free variables) For each Σ0 formula F (vk1 . PA ⊢ ((a = c ∧ a + 0 = a) ⊃ a + b = c). (The universal quantifier on the variable c is to strengthen the induction hypothesis. . PA ⊢ (F (vk1 . . v3 )). vkm ) ](vk1 . 0′ .e. . PA ⊢ a + b = c. PA ⊢ (a + b)′ = c. i. Hence if a + 0 = c. then PA ⊢ a + b = c. so for all c. b = 0: If a + 0 = c. v3 . so there is a number d such that d′ = c.e. Base case: v2 = 0. a + b = d. we argue by induction on v2 in the formula ∀v3 (v1 + v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . . it follows that for all c. t2 terms. Then by the informal version of N4 . then PA ⊢ a + b = c. Since c is not free in the Induction ′ Hypothesis. PA ⊢ a + b = c. vkm )). if a + b′ = c. . if a + 0 = c. and suppose a + b′ = c. v6 ) ∧ P r(v6 )) The following is an informal description of a formal proof within PA. 0. Proof. . then a = c. vkm ) ⊃ P r[ F (vk1 . . PA ⊢ (v1 + v2 = v3 ⊃ ∃v4 ∃v5 ∃v6 (S( v1 + v2 = v3 .) The result then follows by ∀-Elimination on the quantifier ‘for all c’. if a + b = c. Instantiating the quantifier ‘forall c’ in the induction hypothesis with d. By induction over the inductive definition of Σ0 -formulas. Then by the informal version of N1 .LECTURE 11 96 Proof. vkm ). . PA ⊢ a + b = c. We need to show that PA ⊢ ∀v3 (v1 +0 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . . if a + b = c. PA ⊢ d′ = c. If a = c. v5 ) ∧ S(v5 . PA ⊢ a + b = c. . 0′′ . then PA ⊢ a = c. v3 )). . v1 . We now turn to proof of the main theorem. Then ′ by N4 and logic of identity. 0′′′ . . So from the Induction Hypothesis we ′ have proved that if a + b′ = c. We argue by (informal) induction on the free variable b in the statement. but rather than argue by induction on v2 in the formula (v1 + v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . . This is proved outright. Base case: F is an atomic formula. . (a + b)′ = c. From Axiom N3 by ∀-Intro and ∀-Elim. PA ⊢ (v1 + v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . v3 . v2 . v2 . v3 )). We know that d and d′ are the same expression. The others are similar. The proof is by induction on the variable v2 . so by logic of identity in PA. i. Induction step. v2 . PA ⊢ a + b = c. for all c. in order to have a stronger Induction Hypothesis. . in particular with no assumption in which the variable c is free. By the logic of identity in PA. PA ⊢ a + 0 = a. (a + b)′ = d′ . From the fact that d′ = c. Then by logic of identity in PA. By logic of identity. ′ ′ PA ⊢ (a + b)′ = d . a formula of the form t1 = t2 or t1 ≤ t2 for t1 . we have that PA ⊢ a + b = d. We will prove one of these cases. v4 ) ∧ S(v4 . v2 )). (1) (1) v1 + 0 = v3 (2) v1 + 0 = v1 Assumption N3 . Suppose that for all c.

v2 . v1 ) (4) (5) ⊃-elimination (6) (3) substitutivity of = P r[ v1 + v2 = v3 ](v1 . v2 ) (3) P1 ′ ′ (16)(10)(17) ⊃-elim (1)(2)(6) (18) P r[ v1 + v2 = v3 ](v1 . 0. v4 )) (1) ∀-elimin (1)(2)(6) (10) P r[ v1 + v2 = v3 ](v1 . v2 . v2 . 0. v4 ) ∧ P r[ v1 + v2 = (v1 + v2 )′ ](v1 . v1 )) Lemma 124 P r[ v1 + v2 = v3 ](v1 . v4 ) ≡ P r[ v1 + v2 = v3 ](v1 . (1) ∀v3 (v1 + v2 = v3 ⊃ P r[v 1 + v2 = v3 ](v1 . v3 )) (8) ∀-intro Induction step: We have to show. v2 )) ⊃ ∗ ′ ′ (12) P2 P r[ v1 + v2 = v4 ](v1 . v4 )) Lemma 124 ′ (16) ((P r[ v1 + v2 = v3 ](v1 . v4 )) Lemma 123 ′ ′ ′ ′ (15) (P r[ v1 + v2 = v4 ](v1 . v2 . v2 . v2 . v3 ) (6)(19) ∃-elimination ′ ′ (2)(20) ⊃-intro (1) (21) (v1 + v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . v2 . v4 ) ′ (18)(6) subst = (1)(2)(6) (19) P r[ v1 + v2 = v3 ](v1 . v2 . v2 . v3 ) ′ (2) v1 + v2 = v3 Assumption ′ ′ (3) (v1 + v2 ) = (v1 + v2 ) N4 ′ (2) (4) (v1 + v2 ) = v3 (2)(3) logic of = ′ (2) (5) ∃v4 (v4 = v3 ) (4) ∃-Intro ′ (6) (6) v4 = v3 Assumption ′ (6)(2) (7) (v1 + v2 )′ = v4 (4)(6) subst of = (6)(2) (8) v1 + v2 = v4 (7) N1 . v2 . v2 . Assumption I. v2 . v2 )) ⊃ ′ ′ (13)(14)(15) propositional logic P r[ v1 + v2 = v3 ](v1 . v2 . v3 ) (1)(7) ⊃-intro (v1 + 0 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . v3 )) ⊃ ∀v3 (v1 +v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . within PA. v2 . v4 ) ∧ P r[ v1 + v2 = (v1 + v2 )′ ](v1 . 0. v2 .LECTURE 11 (1) (3) (4) (5) (6) (1) (7) (8) (9) 97 v1 = v3 (1) (2) substitutivity of = ∗ P r[ v1 + 0 = v1 ](v1 ) (2) P1 (P r[ v1 + 0 = v1 ](v1 ) ⊃ P r[ v1 + v2 = v3 ](v1 .H. v . v3 )) ∀v3 (v1 + 0 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . v3 )) (1) (2) . v3 ))). v3 ) ′ (1)(2) (20) P r[ v1 + v2 = v3 ](v1 . that there exists a derivation of ′ ′ (∀v3 (v1 +v2 = v3 ⊃ P r[v 1 + v2 = v3 ](v1 . 0. v2 . logic (1) (9) (v1 + v2 = v4 ⊃ P r[v 1 + v2 = v3 ](v1 . v2 . 0. v4 ) (8)(9) ⊃-elim ′ ′ ′ (11) ((v1 + v2 = v4 ∧ v1 + v2 = (v1 + v2 )′ ) ⊃ v1 + v2 = v4 ) substitutivity of = ′ ′ ) ⊃ v + v ′ = v ′ ) ](v . v4 ) ≡ P r[ v1 + v2 = v3 ](v1 . v ) (12) P r[ ((v1 + v2 = v4 ∧ v1 + v2 = (v1 + v2 ) 1 1 2 4 2 4 ∗ (11) P1 ′ (13) ((P r[ v1 + v2 = v4 ](v1 . v4 )) (14) (P r[ v1 + v2 = v4 ](v1 . v4 )) ∗ ′ (17) P r[ v1 + v2 = (v1 + v2 )′ ](v1 .

Then by one step of ∀-elimination. v3 )) ⊃ ′ ′ ∀v3 (v1 + v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . which was to be proved. i. ((∀v1 ≤ v2 )G(v1 ) ⊃ ∀v3 (S( (∀v1 ≤ v2 )G(v1 ) . for G(v1 ) a Σ0 -formula. v2 . ∀-Elim (2) defn of subst . we have proved in PA. (1) (G(x) ⊃ P r[ G(x) ](x)) (2) (G(x) ⊃ P r[G(x)](x))(0) (3) (G(0) ⊃ P r[G(x)(x)](0)) Induction hypothesis for the main induction (1) ∀-Intro. i.LECTURE 11 (1) 98 ′ ′ (22) ∀v3 (v1 + v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . v2 . We now turn to the induction steps of the proof. v2 . v3 ) ⊃ P (v3 )). v3 )). ∀v3 (v1 +v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . v1 is the only free variable in G(v1 ). v2 . Exercise. We need to give a proof in PA of (F ⊃ P r[F ]). we have (v1 +v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . v2 . (i) F is ∼ G where G is Σ0 . for G and H both Σ0 -formulas. (ii) F is (G ∧ H). v3 ) ⊃ P (v3 )) To simplify notation I shall write v1 as x and v2 as y. v2 . 0′′ . v3 )). v3 )) (21) ∀-intro (v3 not free in (1)) (23) (∀v3 (v1 + v2 = v3 ⊃ P r[v 1 + v2 = v3 ](v1 . v2 . The proof is by induction on the variable v2 occurring free in the formula (F ⊃ P r[F ]).e. v3 ))) (1)(22) ⊃-intro By the instance of N12 for induction on v2 in the formula ∀v3 (v1 + v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . This means that v2 is free in F . v3 )). Base case: We need to prove that ((∀v1 ≤ 0)G(v1 ) ⊃ ∀v3 (S( (∀v1 ≤ 0)G(v1 ) . (1) (G ⊃ P r[G]) (2) (H ⊃ P r[H]) (3) (G ⊃ (H ⊃ (G ∧ H))) (4) P r[(G ⊃ (H ⊃ (G ∧ H)))] (5)(P r[(G ⊃ (H ⊃ (G ∧ H)))] ⊃ (P r[G] ⊃ P r[(H ⊃ (G ∧ H))])) (6) (P r[(H ⊃ (G ∧ H))] ⊃ (P r[H] ⊃ P r[(G ∧ H)])) (7) (P r[G] ⊃ (P r[H] ⊃ P r[(G ∧ H)])) (8) ((G ∧ H) ⊃ P r[(G ∧ H)]) Induction hypothesis Induction hypothesis propositional logic (3) Theorem 120 Theorem 121 Theorem 121 (4) (5) (6) propositional logic (1) (2) (7) propositional logic (iii) F is (∀v1 ≤ v2 )G(v1 ). 0′′ . To simplify notation we shall take it that no other variables are free in F .e. 0.

. That F (v) is a Σ1 formula means there is a Σ0 -formula G(v. vi ) ⊃ P r[ G(v. i. the induction of which (8) is an induction hypothesis is over formulas. so the assumption is about the formula G(x). vi ) ](v. Corollary 127 (provable Σ1 -completeness with free variables) For each Σ1 formula F (vk1 . we need to establish ((∀x ≤ y ′ )G(x) ⊃ P r[ (∀x ≤ y ′ )G(x) ](y ′ )).e. ∀-Elim (7)(1)(9) propositional logic (10) Lemma 122 (11) N8 and logic (2) (12) ⊃-Intro Step (9) in the above derivation calls for comment. However. . to avoid clutter. If the variable in (8) were the variable of induction this ∀-Intro would be illegitimate. . We abbreviate vk1 . . which is the induction hypothesis for this sub-induction. . . . . . . x is a free variable in (8) to which ∀-Intro may be applied. (1) (2) (2) (2) (2) (2) (2) (8) (8) (1)(2)(8) (1)(2)(8) (1)(2)(8) (1)(8) (1) ((∀x ≤ y)G(x) ⊃ P r[ (∀x ≤ y)G(x) ](y)) (2) (∀x ≤ y ′ )G(x) (3) ∀x(x ≤ y ′ ⊃ G(x)) (4) ∀x((x ≤ y ∨ x = y ′ ) ⊃ G(x)) (5) ∀x((x ≤ y ⊃ G(x)) ∧ (x = y ′ ⊃ G(x)) (6) (∀x(x ≤ y ⊃ G(x)) ∧ ∀x(x = y ′ ⊃ G(x))) (7) ((∀x ≤ y)G(x) ∧ G(y ′ )) (8) (G(x) ⊃ P r[ G(x) ](x)) (9) (G(y ′ ) ⊃ P r[ G(x) ](y ′ )) (10) (P r[ (∀x ≤ y)G(x) ](y) ∧ P r[ G(x) ](y ′ )) (11) P r[ ((∀x ≤ y)G(x) ∧ [G(y)]) ](x. . . vi ) (2) (G(v. . vkm ). vkm ). logic Assumption IH for main induction (8) ∀-Intro. In deriving (9) from (8) ∀-Intro is applied to (8) and the status of (8). . vkm ) ](vk1 . . . and not about x. vi ). is that of assumption. . The following describes a derivation in PA. vkm ) ⊃ P r[ F (vk1 .LECTURE 11 (4) (5) (6) (7) (P r[G(x)](0) ≡ P r[G(0)]) (G(0) ⊃ P r[G(0)]) (∀x ≤ 0)G(x) ≡ G(0)) ((∀x ≤ 0)G(x) ⊃ P r[(∀x ≤ 0)G(x)]) Lemma (3) (4) propositional logic provable in PA (5) (6) logical equivalences 99 Induction step: On the assumption ((∀x ≤ y)G(x) ⊃ P r[ (∀x ≤ y)G(x) ](y)). Proof. PA ⊢ (F (vk1 . . . as an induction hypothesis. vi ) such that F (v) = ∃vi G(v. (1) (G(v. vi ) ⊃ ∃vi G(v. . vi )) Theorem 126 predicate logic . vkm as v. y ′ ) (12) P r[(∀x ≤ y ′ )G(x)] (13) ((∀x ≤ y ′ )G(x) ⊃ P r[(∀x ≤ y ′ )G(x)]) Assumption IH for sub-induction Assumption (2) defn of (∀x ≤ y ′ ) (3) N6 (4) propositional logic (5) predicate logic (6) defn of (∀x ≤ y).

vi ) ⊃ P r[ ∃vi G(v. P2 (1) (3) propositional logic (4) ∀-Intro (5) anti-prenexing . vi ) ⊃ P r[ ∃vi G(v. vi ) ](v)) (G(v. vi ) ](v) 100 ∗ ∗ ( 2) P1 . vi ) ](v) (∃vi G(v. vi ) ](v.LECTURE 11 (3) (4) (5) (6) (P r[ G(v. vi ) ⊃ P r[ ∃vi G(v. vi ) ](v) ∀vi (G(v. vi ) ⊃ P r[ ∃vi G(v.

Lecture 12 The ω-rule and uniform reflection; PA proves that PA proves every instance of the G¨del sentence; o Π1-uniform reflection and consistency; PA is Π1-conservative over PAΠ2 ∪ {ConP A}
(Wednesday, 17 November 2010)

12.1

The ω-rule and uniform reflection

We noted in Section 8.4 that the first half of the G¨del Incompleteness Theorem for o a Σ0 -complete system S establishes that if S is consistent then it is ω-incomplete, i.e. we have, for each n, S ⊢∼ P rov( G , n), but S ∀v2 ∼ P rov( G , v2 ). If each numerical instance of a formula F (vi ) with one free variable is true, then ∀vi F (vi ) is true. Hence the following inference, called the ω-rule, is sound with respect to truth in arithmetic: F (0), F (1), . . . , F (n) . . . ∀yF (y) Definition 73 PAω =df PA + ω-rule. 101

LECTURE 12

102

Definition 74 PAω ⊢ X if and only if there is a tree of finite height with formulas in the language of PA at each node and with X at the bottom node, such that for each formula at a node without predecessor is a theorem of PA, and such that each node is either a theorem of PA or there is one node directly above the node at which X occurs such that for Y the formula at the predecessor node, PA ⊢ (Y ⊃ X), or X is the result of an application of the ω-rule. By a tree consisting of a single node (height 0), if PA ⊢ X, then PAω ⊢ X. For G the G¨del sentence for S, PAω ⊢ G. A derivation using the ω-rule has o infinitely many premisses and hence is an infinite object, unlike a formal proof, or our usual idea of a informal proof, so derivation of G by the ω-rule cannot be said to constitute a proof of G. Indeed derivability from the axioms of PA by the ω-rule is tantamount to truth. Proposition 128 If a sentence X in the language of arithemtic is true, PAω ⊢ X. Proof. We argue by induction over the arithmetical hierarchy. We know by Lecture 7 that if X is a true Σ0 (= Π0 ) or Σ1 -sentence, PA ⊢ X, and hence PAω ⊢ X. Suppose X is a true Π1 -sentence, i.e. of the form ∀v1 F (v1 ) for F (v1 ) a Σ0 formula. Then for each natural number n, PA ⊢ F (n), so by one application of the ω-rule, PAω ⊢ ∀v1 F (v1 ). Assume for Induction Hypothesis that the result holds for Σn and Πn -sentences. (i) Let X be a true Σn+1 -sentence ∃vi F (vi ), where F (vi ) is a Πn formula. Then for some natural number m, F (m) is a true Πn -sentence. Then by Induction Hypothesis PAω ⊢ F (m). Since PA ⊢ (F (m) ⊃ ∃vi F (vi )), PAω ⊢ ∃vi F (vi ). (ii) Let X be a true Πn+1 -sentence ∀vi F (vi ). Then for each number n, F (n) is a true Σn -sentence. Then by Induction Hypothesis, for each n, PAω ⊢ F (n). Then by one application of the ω-rule, PAω ⊢ ∀vi F (vi ). The completeness of PAω with respect to truth is entirely to do with the ω-rule, and essentially nothing to do with the axioms for arithmetic of PA, as shown by the fact that the corresponding system for R, i.e. Rω =df R + ω-rule, is also complete with respect to truth. Proposition 129 If a sentence X in the language of arithmetic is true, Rω ⊢ X. Proof. The only facts about PA used in the proof of Proposition 128 are that PA is Σ0 -complete and that PA ⊢ (F (m) ⊃ ∃vi F (vi )). Both these properties are also facts about R. There are constructive versions of the ω-rule, based on the fact that we can state in a single sentence that all numerical instances of a given formula F (v1 ) are provable, and by the arithmetization of syntax such single sentences can be expressed

LECTURE 12

103

in the language of arithmetic, namely as ∀v1 P r[ F (v1 ) ](v1 ), where P r[ F (v1 ) ](v1 ) is defined by Definition 70 in Lecture 11. We can then give finite expression to an ω-rule by the sentence: (∀v1 P r[ F (v1 ) ](v1 ) ⊃ ∀v1 F (v1 )). Such sentences are highly sensitive to the axiomatic strength of the system to which they are applied. They are called Reflection Principles (see Smorynski [6], p. 845). We use the following terminology: Definition 75 For F (v1 ) a formula in the language of PA with one free variable, a Uniform Reflection Principle is any sentence of the form (∀v1 P r[ F (v1 ) ](v1 ) ⊃ ∀v1 F (v1 )). PA extended by Uniform Reflection Principles for all one-place formulas is strictly weaker than PA extended by the infinitary ω-rule. This is because PA + ω-rule = true arithmetic, while PA + all Uniform Reflection Principles is axiomatic and hence incomplete.

12.2

PA proves that PA proves every instance of the G¨del sentence o

Theorem 130 PA⊢ ∀v1 P r[ ∼ P rov( G , v1 ) ](v1 ) Proof. The following derivation shows the existence of a formal proof in PA. (1) (1) (2) (3) (4) (5) (6) ∼ P r[ ∼ P rov( G , v1 ) ](v1 ) (0 = 0′ ⊃∼ P rov( G , v1 )) (P r[0 = 0′ ] ⊃ P r[∼ P rov( G , v1 )]) (∼ P r[∼ P rov( G , v1 )] ⊃∼ P r[0 = 0′ ]) ∼ P r[0 = 0′ ] (∼ P r[0 = 0′ ] ⊃∼ ∃v1 P rov( G , v1 )) Assumption PA ⊢∼ 0 = 0′ ∗ ∗ (2) P1 , P2 , MP (3) contraposition (1) (4) MP line (5) of proof of Theorem 104 and Proposition 118 (5) (6) MP (7) pred logic (8) ∀-elim Theorem 126 (9) (10) MP (1)(11) reductio ad absurdum (12) ∀-Intro

(1)

(1) (1) (1) (1)

(7) ∼ ∃v1 P rov( G , v1 )) (8) ∀v1 ∼ P rov( G , v1 )) (9) ∼ P rov( G , v1 )) (10) (∼ P rov( G , v1 ) ⊃ P r[∼ P rov( G , v1 )]) (11) P r[∼ P rov( G , v1 )] (12) P r[∼ P rov( G , v1 )] (13) ∀v1 P r[∼ P rov( G , v1 )]

and the fact that PA ⊢ (G ≡ ∀v1 ∼ P rov( G . (1) (2) (1) ∀v1 P r[ F (v1 ) ](v1 ) (2) ∼ F (v1 ) (3) (∼ F (v1 ) ⊃ P r[ ∼ F (v1 ) ](v1 ) Assumption Assumption Corollary 127 since ∼ F (v1 ) is Σ1 (2)(3) MP (1) ∀-elim Lemma 122 propositional logic ∗ (7) P1 ∗ (8) P2 (9) contraposition given (11) Proposition 118 (12) (10) ⊃-elimination (6) (13) ∧-intro (2) (6) RAA (2) (4) P r[ ∼ F (v1 ) ](v1 ) (1) (5) P r[ F (v1 ) ](v1 ) (1)(2) (6) P r[ (F (v1 )∧ ∼ F (v1 )) ](v1 ) (7) ((F (v1 )∧ ∼ F (v1 )) ⊃ 0 = 0′ ) (8) P r[ (F (v1 )∧ ∼ F (v1 )) ⊃ 0 = 0′ ](v1 ) (9) P r[ (F (v1 )∧ ∼ F (v1 )) ](v1 ) ⊃ P r[ 0 = 0′ ] (10) ∼ P r[ 0 = 0′ ] ⊃∼ P r[( (F (v1 )∧ ∼ F (v1 )) ](v1 ) (11) ∼ P r( 0 = 0′ ) (12) ∼ P r[ 0 = 0′ ] (13) ∼ P r[( (F (v1 )∧ ∼ F (v1 )) ](v1 ) (1)(2) (14) ((6) ∧ (13)) (1) (15) F (v1 ) .e.3 Equivalence of Π1-Uniform Reflection and consistency Π1 -Uniform Reflection for PA. Theorem 133 For F (v1 ) any Π1 -formula with one free variable. PA ∪{ConP A } ⊢ (∀v1 P r[ F (v1 ) ](v1 ) ⊃ ∀v1 F (v1 )). PA ∪{(P r( 0 = 0′ ) ⊃ 0 = 0′ )} ⊢∼ P r( 0 = 0′ ). By the Theorem. Modus ponens. is provably equivalent in PA to ConP A . Take the instance of Π1 -uniform reflection for the Π1 -formula ∀v1 (0 = 0′ ∧ v1 = v1 ). Proof. the formal consistency statement for PA. ∼ P rP A ( 0 = 0′ ).LECTURE 12 104 Corollary 131 PA ∪{(∀v1 P r[ ∼ P rov( G . This is provably equivalent to (P r( 0 = 0′ ) ⊃ 0 = 0′ ). 12. Theorem 132 PA + Π1 -uniform reflection ⊢∼ P rP A ( 0 = 0′ ). Proof. i. v1 )). v1 ))} ⊢G Proof. the Uniform Reflection Principle restricted to Π1 formulas. v1 ) ](v1 ) ⊃ ∀v1 ∼ P rov( G .

then S1 ⊢ X. Nonetheless Hilbert’s argument adumbrates a correct mathematical theorem the main content of which is the proof of Theorem 133 that consistency implies the arithmetized ω-rule (a. Hilbert sketches an argument for this claim in [4]. 474. if S1 ⊢ X. More precisely and more generally. if S2 ⊢ X.4 PA is Π1-conservative over PAΠ2 ∪ {ConP A} In the 1920s David Hilbert adumbrated a research programme which had at its heart the project of giving proofs of the consistency of formal systems of infinitary mathematics in finitary mathematics. and S2 is a conservative extension of S1 if for every formula X in the language of S1 . i. Definition 76 For theories S1 and S2 formulated in the same language (or such that the language of S2 is an extension of the language of S1 or such that the language of S1 can interpreted in the language of S2 . but we shall not be concerned with these more general cases). in which case the justification at line (3) is Theorem 126 rather than Corollary 127. then S1 ⊢ X.e. the uniform reflection principle) for Π1 -sentences.a. the consistency of infinitary mathematics cannot be proved within finitary mathematics.k. The motivation for this programme was foundational and philosophical but Hilbert came to see that it also promised mathematical application in terms of establishing “conservative extension” results. p. S2 is an extension of S1 if for each formula X in the language of S1 . for . S2 proves nothing in the language of S1 that S1 doesn’t already prove. Definition 77 An extension S2 of S1 is conservative over S1 with respect to a class of formulas Γ if for each formula X in Γ. Fermat’s last theorem: “Let us suppose. then infinitary mathematics is a conservative extension of finitary mathematics with respect to finitary mathematics.LECTURE 12 (1) (16) ∀v1 F (v1 ) (17) (∀v1 P r[ F (v1 ) ](v1 ) ⊃ ∀vi F (v1 ) 105 (15) ∀-intro (v1 not free in (1) (1) (16) ⊃-intro This proof covers the case where F (v1 ) is Σ0 rather than Π0 . insofar as infinitary matho ematics is an extension of finitary mathematics. G¨del’s Second Incompleteness Theorem shows that. A theory S2 is an extension of S1 if S2 proves everything that S1 proves. Hilbert formulates his argument in terms of a particular Π1 -sentence. A fundamental insight of Hilbert’s that lies at the heart of his programme of proof theory is that if finitary mathematics can prove the consistency of infinitary mathematics. then S2 ⊢ X (or S2 ⊢ X ∗ where X ∗ is the translation of X into the language of S2 ). if S2 ⊢ X. 12.

for Fermat’s great theorem. b. axioms N12 for Σ1 -formulas. but it is implicit in his argument. as in the step of the proof in Lecture 11 in which we proved that ((∀v1 ≤ v2 )G ⊃ ∀v3 (S( (∀v1 ≤ v2 )G . measured by complexity in the arithmetical hierarchy of formulas to which induction must be applied. v2 . Rather. then P AΠ2 ∪ {ConP A } ⊢ X. Proof. v3 )). However. It might be that there is some clever way to reconstruct that proof so that the universal quantification of the induction formula is not needed. and I won’t enter here the debate over what formal system should be taken to capture the intended notion of finitary arithmetic. “Let us assume that numerals p. Hilbert never formulated clearly what he meant by finitary mathematics. a proof in which the [infinitary] logical function ǫ was used. very roughly. c (p > 2) satisfying Fermat’s equation ap +bp = cp are given. By Theorem 133 and analysis of the proof of Theorem 126. so we may take the question to be. that we had found. v2 . constructive principles of abstract mathematics. then we could also obtain this equation as a provable formula by giving the form of a proof to the procedure by which we ascertain that the numbers ap + bp and cp coincide. is needed for this proof? The minimum is Σ1 -induction. I will address the question.” Leaving aside the question. we have the following theorem corresponding to Hilbert’s argument claiming that infinitary mathematics is conservative over finitary mathematics with respect to Π1 -theorems. We could then make a finitary proof out of it in the following way. . which is beyond the scope of this course (the answer is. a.e. we also used Π2 -induction. in what minimal system can the consistency of PA be proved. rather than finitary principles of concrete mathematics). and implicitly he takes it to be a fact of finitary mathematics. v3 ) ⊃ P (v3 )) by induction on the free variable v2 . i. he never gave a formal system of finitary mathematics.LECTURE 12 106 example.e. In any case. Theorem 134 For X any Π1 -sentence in the language of PA. in our proof in PA that (v1 + v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . 0′′ . a precise working out of the argument Hilbert sketched requires that the proof of Theorem 133 be carried out in finitary mathematics. in how weak a subsystem of PA can the argument for Theorem 133 be carried out? The key point is that such a system must be strong enough to prove provable Σ0 completeness of PA. Hilbert did not explicitly formulate provable Σ0 -completeness. since we proved this by induction on the formula ∀v3 (v1 + v2 = v3 ⊃ P r[ v1 + v2 = v3 ](v1 . v3 )).” Proving provable Σ0 -completeness requires mathematical induction. i. v2 . if P A ⊢ X. from what has been established. how much induction.

but we will establish the results that concern us here purely syntactically. “A is necessarily true”. When provability logic first began to be developed. In this notation the arithmetization of L¨b’s theorem (problem 1 on Problem sheet 7) is expressed as (2(2A ⊃ A) ⊃ 2A). i. then S (P r( X ) ⊃ X). o The system of provability logic was named GL by George Boolos. there existed already. o As we shall see. after G¨del o and L¨b. i. not using these semantic techniques). Such systems are called modal logic since necessity concerns not only the truth of sentences but also the kind or mode of their truth. To signify this viewpoint we write 2A for P r( A ). and in the 1930s. It is important to realize that provability logic is not an extension of the logic of necessity. 23th November 2010) A proof predicate P r(v1 ) for a system S can be thought of as an operator on sentences in the language of S. this formula provides the fundamental axiom by which to axiomatize the logic of provability. in contrast to which L¨b’s theorem shows that if S X. in the 1970s. systems of logic for a sentence operator 2A with the intended meaning.e. for which (2X ⊃ X) is valid.Lecture 13 Provability logic: the system GL (Tuesday. it generates a sentence from a sentence. Modal logic provided a framework for setting up systems for provability logic (and also a semantics of possible worlds with an accessibility relation between worlds by which to study properties of such systems. which has been exploited in the study of provability logic.1 The language of GL The primitive symbols of GL: 107 .e. G¨del o had obtained results by interpreting the 2 operator of modal logic as provability. It consists of propositional logic plus the provability operator. for more than fifty years. o 13. in a brief note.

disjunction. Definition 78 (sentences of GL) By recursion: base: ⊥ and all pi are sentences. (Tautologies) Every sentence in the language of GL that is a truth functional tautology when ⊥ is assigned the truth value F (falsity) and ⊃ is interpreted as the truth function ‘if . on the intended meaning for ⊃ and ⊥. as we shall see by Theorem 147. (X ∧ Y ) =df ((X ⊃ (Y ⊃⊥)) ⊃⊥). (X ∨ Y ) =df ((X ⊃⊥) ⊃ Y ) (X ≡ Y ) =df (((X ⊃ Y ) ⊃ ((Y ⊃ X) ⊃⊥)) ⊃⊥) 13. .g. (Corresponds to property P2 of provability predicates. and equivalence: Definition 79 ∼ X =df (X ⊃⊥). . infer Y . recursion: If X and Y are sentences. . In modal logic this rule is known as Necessitation. 0 = 0′ in the language of arithmetic). The sentential connective ⊃. conjunction. Definition 81 (rules of inference of GL) R1. generated from the symbol ‘p’ and iteration of the subscript symbol ‘′ ’. o (2(2X ⊃ X) ⊃ 2X) is an axiom.LECTURE 13 108 A sentence ⊥ (standing for a generic false sentence.2 The axioms and inference rules of GL Definition 80 (axioms of GL) A1. (Arithmetized L¨b’s Theorem) For each sentence X in the language of GL. 2X is a sentence. Infinitely many sentence letters pi . (X ⊃ Y ) is a sentence.) A3. From sentence X infer 2X. If X is a sentence. express negation. (2(X ⊃ Y ) ⊃ (2X ⊃ 2Y )) is an axiom. The sentential operator 2. (Corresponds to property P1 of provability predicates. A2. then . (Distribution) For X and Y any sentences in the language of GL. . We shall write sentences in the language of GL using the following abbreviations which. From sentences X and (X ⊃ Y ).) There is no axiom schema corresponding to property P3 for provability predicates since. . e. (Modus ponens) R2. (2X ⊃ 22X) is derivable from the axioms and rules of inference specified for GL. ’ is an axiom.

due to Robert Solovay.e. . for P r(v1 ) an arithmetized proof predicate for PA. which by problem 1 on Problem sheet 7. Conversely. then by P1 from arithemetization of the provability predicate for PA. then X ∗ is an instance of P2 and hence provable in PA. then P A ⊢ X ∗. Theorem 135 (soundness of GL with respect to interpretation in PA) For every interpretation ∗ from GL into PA as in Definition 82. Induction step: If P A ⊢ X ∗ and P A ⊢ (X ⊃ Y )∗ . There is also a completeness theorem. (ii) p∗ = si for i → si a specified enumeration of the sentences in the language i of PA. (iii) If X is a L¨b’s Theorem axiom. Definition 82 An interpretation ∗ of GL in PA is given by translating the language of GL into the language of PA by the following inductive definition: Base: (i) ⊥∗ = X for X a specified sentence in the language of PA such that P A ⊢∼ X. P A ⊢ (2X)∗ . if GL ⊢ X. is provable in PA. This result means that GL is sound with respect to interpretation in PA. for GL with respect to interpretation in PA. and hence provable in PA. Basis step. P A ⊢ P r( X ∗ ). in which case an interpretation of GL in PA is determined by the choice of enumeration of the sentences in the language of PA.LECTURE 13 109 The axioms and inference rules of GL arise by abstraction from the arithmetized proof predicate for PA. for X an axiom of GL: (i) If X is a truth functional tautology. Also note that we can reformulate the language and logic of PA by including the symbol ⊥ in the language of PA with corresponding axioms for propositional logic so that ⊥ is treated as falsity. (iv) (2X)∗ = P r( X ∗ ). Induction: (iii) (X ⊃ Y )∗ = (X ∗ ⊃ Y ∗ ). (ii) If X is a Distribution Axioms. then since (X ⊃ Y )∗ = (X ∗ ⊃ Y ∗ ). i. P A ⊢ Y ∗ . translate into theorems of PA. and hence all theorems of GL. Note that an interpretation ∗ in the above specification is determined by the choice of a sentence X such P A ⊢∼ X and a specific enumeration of the sentences in the language of PA. Proof. then X ∗ is an instance of arithmetized L¨b’s o o Theorem. then X ∗ is a truth functional tautology. the axioms and inference rules. If P A ⊢ X ∗ . By induction over the recursive definition of theorems of GL.

e. . and GL ⊢ Y by r-many applications of Modus ponens starting with this axiom. Then by the initial supposition and R1. GL ⊢ 2X.))) is a tautology and hence an axiom of GL. i. Then by R1. Xr . Proof. Suppose GL ⊢ (2X ⊃ X). 13. In proofs that depend on this theorem I will say “by propositional logic in GL” or just “by propositional logic”. . Theorem 139 If Y is a truth functional consequence of finitely many formulas which are each provable in GL. Proof. P A ⊢ X ∗ . . then GL ⊢ X. if for every interpretation ∗ (as in Definition 82). not just finite sets of formulas. o Lemma 137 If GL ⊢ (2X ⊃ X). GL ⊢ X. ⊃ (Xr ⊃ Y ) .3 Some derivations in GL Unarithmetized L¨b’s Theorem holds for GL. then GL ⊢ (2A ⊃ 2B). the above result holds for Y a truth functional consequence of any set of provable formulas. By A3. then there is an interpretation ∗ such that P A X ∗ . for X any sentence in the language of GL. Theorem 140 GL ⊢ (2(X ∧ Y ) ≡ (2X ∧ 2Y )) . Proof. then GL ⊢ X. so by R1. or equivalently. then (X1 ⊃ (X2 ⊃ (.LECTURE 13 110 Theorem 136 (Solovay completeness theorem for GL) For X any sentence in the language of GL. Beyond the scope of this course. GL ⊢ 2(2X ⊃ X). Next we note that GL is closed under truth functional consequence. By the compactness theorem for truth functional logic.e. . . if GL X. . . If Y follows truth functionally from the finitely many formulas X1 . but we have not need for this generalization. Lemma 138 If GL ⊢ (A ⊃ B). Remarks. i. Proof. rather than citing Theorem 139. GL ⊢ (2X(2 ⊃ X) ⊃ 2X). then GL ⊢ Y . . This is a notational variant of P4 and the proof of P4 in Lemma 103 proves this result.

Then by propositional logic in GL. since P A ⊢ (P r( G ) ≡ P r( 0 = 0′ )). 13. 2. GL ⊢ ((2X ∧ 2Y ) ⊃ 2(X ∧ Y )). Hence by Lemma 138. Proposition 141 GL ⊢ (2(X ≡ Y ) ⊃ (2X ≡ 2Y )) Proof. Proof. By Definition 79. GL ((2X ≡ 2Y ) ⊃ 2(X ≡ Y )). The formulas ((2X ∧ 2Y ) ⊃ 2X) and ((2X ∧ 2Y ) ⊃ 2Y ) are tautologies and hence axioms of GL. GL ⊢ (2(Y ⊃ (X ∧ Y ) ⊃ (2Y ⊃ 2(X ∧ Y )))). Then by propositional logic in GL. 3. so by Theorem 140. Proposition 142 If PA is 1-consistent. which by Definition 79 is GL ⊢ (2(X ≡ Y ) ⊃ (2X ≡ 2Y )). then Fp (A) =⊥. and P A (P r( G ≡ 0 = 0′ )) if PA is 1-consistent. is defined by the following recursion. 4. Fp (A). If F =⊥. (X ≡ Y ) =df ((X ⊃ Y ) ∧ (Y ⊃ X)). GL ⊢ (2X ⊃ 2(Y ⊃ (X ∧ Y ))).4 Closure of GL under substitution by provably equivalent formulas The result of substituting a sentence A for sentential variable p in formula F . If F = q where q = p. GL ⊢ (2(X ∧ Y ) ⊃ (2X ∧ 2Y )) (ii) Since (X ⊃ (Y ⊃ (X ∧ Y ))) is a tautology. (i) Both ((X ∧ Y ) ⊃ X) and ((X ∧ Y ) ⊃ Y ) are tautologies and hence axioms of GL. P A ((P r( G ) ≡ P r( 0 = 0′ )) ⊃ (P r( G ≡ 0 = 0′ ))). and F (p) any formula in the language of GL in which the sentence letter p occurs. GL ⊢ (X ⊃ (Y ⊃ (X ∧ Y ))). GL ⊢ (2X ⊃ (2Y ⊃ 2(X ∧ Y ))). GL ⊢ (2(X ≡ Y ) ⊃ ((2X ⊃ 2Y ) ∧ (2Y ⊃ 2X))). If F = p. Hence by Theorem 135. then Fp (A) = (Gp (A) ⊃ Hp (A)). GL ⊢ (2(X ≡ Y ) ≡ (2(X ⊃ Y ) ∧ 2(Y ⊃ X))). GL ((2X ≡ 2Y ) ⊃ 2(X ≡ Y )). then Fp (A) = 2(Gp (A)). . by Theorem ??. GL ⊢ (2(X ∧Y ) ⊃ 2X) and GL ⊢ (2(X ∧Y ) ⊃ 2Y ). GL ⊢ ((2(X ⊃ Y ) ∧ 2(Y ⊃ X)) ⊃ ((2X ⊃ 2Y ) ∧ (2Y ⊃ 2X))). then Fp (A) = q. By Lemma 138. then Fp (A) = A. 5. if GL ⊢ F (p). Definition 83 (Fp (A)) 1. then GL ⊢ Fp (X). so by propositional logic in GL. Then by propositional logic in GL. If F = (G ⊃ H). The converse of Proposition 141 does not hold. By A2 and Theorem 139.LECTURE 13 111 Proof. Proposition 143 (closure of GL under substitution) For X any formula in the language of GL. If F = 2G. By A2.

We argue by induction over the inductive definition of formulas F . then GL ⊢ (⊥≡⊥). . GL ⊢ Y and GL ⊢ (Y ⊃ F (p)). so the implication holds. . . then GL ⊢ (Gp (A) ≡ Gp (B)). GL ⊢ (2(Gp (A)) ≡ 2(Gp (B))). what is to be proved is if GL ⊢ (A ≡ B). If F =⊥. then GL ⊢ (Gp (A) ≡ Gp (B)). . n. The proof of Theorem 144 generalizes to establish provable equivalence of substitution on more than one sentence letter. GL ⊢ (Yp (X) ⊃ Fp (X)). then we have by induction hypothesis that if GL ⊢ (A ≡ B). GL ⊢ 2(Gp (X)). If F = p. i. so GL ⊢ Fp (X). GL ⊢ Fp (X). or A3 (arithmetized L¨b’s Theorem). A2 (distribution). If F = q for p = q. so by Definition 83. An ) ≡ F (B1 . which holds trivially. B. such that GL ⊢ (Ai ≡ Bi ). GL ⊢ Gp (X). which follows by propositional logic from the induction hypotheses. By R2. If F = 2G. then we have as induction hypotheses. .e. . then GL ⊢ (A ≡ B). and F and any sentence letter p. (ii) The proof of F (p) ends with an application of R2. if GL ⊢ (A ≡ B). GL ⊢ (Fp (A) ≡ Fp (B)). if GL ⊢ (A ≡ B). pkn . Proof. and F (B1 . and if GL ⊢ (A ≡ B). Then by clause 4 of Definition 83.e. . . . . The consequent is provable outright. If F = (G ⊃ H). . Bn )). GL ⊢ ((2G)p (A) ≡ (2G)p (B)). Bi . GL ⊢ Fp (X). . F (p) is of the form 2G(p) and GL ⊢ G(p). . then GL ⊢ (Fp (A) ≡ Fp (B)). By Induction Hypothesis. . then GL ⊢ (q ≡ q). F (p) is an axiom. then GL ⊢ (Hp (A) ≡ Hp (B)). i = 1. . then GL ⊢ ((Gp (A) ⊃ Hp (A)) ≡ (Gp (B) ⊃ Hp (B))). Then by clause 5 of Definition 83. which is to say. By Induction Hypothesis. . Base case: The proof is of length 1. of the same form as F (p) is. . By induction on the length of a proof of F (p) in GL. so by R1. GL ⊢ (F (A1 . what is to be proved is GL ⊢ (A ≡ B). . i. what is to be proved is that if GL ⊢ (A ≡ B). i. Induction steps: (i) The proof of F (p) ends with an application of R1. . Bn ) is the result of substituting Bi for pki in F . . . . where F (A1 . By Definition 83. An ) is the result of substituting Ai for pki in F . in which case it is of the form A1 (tautology). and for pairs of formulas Ai . .e. for which again the consequent is provable outright.LECTURE 13 112 Proof. what is to be proved is GL ⊢ (A ≡ B). Theorem 144 (provable equivalence of substitution of provable equivalents) For all formulas A. o Then Fp (X) is an axiom. . . Then by Lemma 138 and propositional logic in GL. Theorem 145 (substitution on more than one sentence letter) For F any formula with sentence letters pk1 . GL ⊢ Yp (X) and GL ⊢ (Y ⊃ F (p))p (X).

Exactly the same proof structure as for Theorem 144. Then by propositional logic in GL. Proof. From GL ⊢ (A ≡ B) and Theorem 144. Since (q ≡ q) is a tautology. .5 Closure of GL under substitution of provably equivalent formulas is provable in GL We now show that the closure of GL under substitution of provably equivalent formulas. 13. The result follows by propositional logic in GL. by propositional logic in GL. GL ⊢ (q ≡ q). what is to be proved is GL ⊢ (2(A ≡ B) ⊃ 2(A ≡ B)). The formula (X ⊃ ((22X ∧ 2X) ⊃ (2X ∧ X)) is a tautology and hence GL ⊢ (X ⊃ ((22X ∧ 2X) ⊃ (2X ∧ X)). Theorem 144. GL ⊢ (Fp (A) ⊃ Fp (B)). GL ⊢ (2X ⊃ 22X). GL ⊢ (2(A ≡ B) ⊃ 2(Fp (A) ≡ Fp (B))).LECTURE 13 113 Proof. Proof. by Theorem 140. The proof is by induction over the recursion that generates the formula F. can be formalized in GL. Theorem 147 GL ⊢ (2X ⊃ 22X) Proof. and F and propositional variable p. If F = p. Then by Theorems 140 and propositional logic in GL. Since ((22X ∧ 2X) ⊃ 22X) is a tautology. what is to be proved is GL ⊢ (2(A ≡ B) ⊃ 2(q ≡ q)). then GL ⊢ Fp (B). Corollary 146 If GL ⊢ Fp (A) and GL ⊢ (A ≡ B). Theorem 144 immediately establishes that GL is closed under substitution of provable equivalents. taking Fp as (X ⊃ (p ⊃ (2X ∧ X)). we have GL ⊢ Fp (B). GL ⊢ (2X ⊃ 2(2X ∧ X)). we have by ∧-elimination. establishes this result. B. so by R2 . GL ⊢ (X ⊃ (2(2X ∧ X) ⊃ (2X ∧ X)). Then by Lemma 138 (P4 ). which is a tautology and hence provable in GL. The formula (2(2(2X ∧ X) ⊃ (2X ∧ X)) ⊃ 2(2X ∧ X)) is an A3 axiom of GL. GL ⊢ 2(q ≡ q). Theorem 148 (arithmetized substitution theorem) For all formulas A. We are now able to show that GL proves P3 . so by Modus ponens from GL ⊢ Fp (A). GL ⊢ ((22X ∧ 2X) ≡ 2(2X ∧ X))). If F = q for p = q. GL ⊢ (2X ⊃ 2(2(2X ∧ X) ⊃ (2X ∧ X)). by Corollary 146. GL ⊢ (2X ⊃ (22X ∧ 2X). Since. with just reformulation of the induction hypothesis so that it’s for multiple substitutions.

GL ⊢ (2(A ≡ B) ⊃ (2(Gp (A) ≡ Gp (B)) ∧ 2(Hp (A) ≡ Hp (B)))). Then by Lemma 138 (P4 ). we have as induction hypotheses. GL ⊢ (22(A ≡ B) ⊃ 2(Fp (A) ≡ Fp (B))). GL ⊢ (2((Gp (A) ≡ Gp (B)) ∧ (Hp (A) ≡ Hp (B))) ⊃ 2((Gp (A) ⊃ Hp (A)) ≡ (Gp (B) ⊃ Hp (B)))). Then by Proposition 141 and propositional logic in GL. GL ⊢ (2(A ≡ B) ⊃ 2((Gp (A) ≡ Gp (B)) ∧ (Hp (A) ≡ Hp (B)))). Lemma 150 For each formula X in the language of GL. and GL ⊢ (2(A ≡ B) ⊃ 2(Hp (A) ≡ Hp (B))). GL ⊢ (2(A ≡ B) ⊃ (Fp (A) ≡ Fp (B))). 13. The constraint of L¨b’s Theorem makes it o very difficult to derive an unboxed conclusion from a boxed premiss. The situation is much more flexible if we are able to strengthen the premiss to ⊡X.). by propositional logic in GL.6 Strengthened proof that the closure of GL under substitution of provably equivalent formulas is provable in GL This strengthened proof makes use of the following technical definition. GL ⊢ (2(A ≡ B) ⊃ 2(Gp (A) ≡ Gp (B))). Then by Definition 83 (5. The following formula is a tautology and so provable in GL (as an axiom): (((Gp (A) ≡ Gp (B)) ∧ (Hp (A) ≡ Hp (B))) ⊃ ((Gp (A) ⊃ Hp (A)) ≡ (Gp (B) ⊃ Hp (B)))). by L¨b’s Theorem for GL (Lemma 137).LECTURE 13 114 If F =⊥. and ⊡2X are provably equivalent in GL. is GL ⊢ (2(A ≡ B) ⊃ 2(Fp (A) ≡ Fp (B))). 2 ⊡ X. Definition 84 ⊡X =df (2X ∧ X) Lemma 149 GL ⊢ (⊡X ⊃ X) Proof. Lemma 149 is a triviality but draws attention to a key property of ⊡ that holds for all formulas and which. The argument is as for the preceding case. Remark. Then by Theorem 140 and propositional logic in GL. GL ⊢ (2(A ≡ B) ⊃ (2(Gp (A)) ≡ 2(Gp (B)))). From this result and the two steps earlier we have. If F = (G ⊃ H). GL ⊢ (2(A ≡ B) ⊃ 2((Gp (A) ⊃ Hp (A)) ≡ (Gp (B) ⊃ Hp (B)))). By propositional logic in GL. Then by Lemma 138 (P4 ). of the form 2X. If F = 2G. .e. GL ⊢ (2(A ≡ B) ⊃ 2(Fp (A) ≡ Fp (B))). ⊢ (⊡X ⊃ X) is a tautology. By Theorem 147. GL ⊢ (2(A ≡ B) ⊃ 22(A ≡ B)). we have as induction hypothesis that GL ⊢ (2(A ≡ B) ⊃ 2(Gp (A) ≡ Gp (B))). so by propositional logic in GL. which by Definition 83(4). holds o for 2 only for formulas provable in GL. what is to be proved is GL ⊢ (2(A ≡ B) ⊃ 2(⊥≡⊥)). the formulas 2X. i.

GL ⊢ (⊡Y ⊃ Y ) and hence by propositional logic in GL. i. which follows from the induction hypotheses by propositional logic in GL. By Definition 83(4). what is to be proved is GL ⊢ (⊡(A ≡ B) ⊃ ((Gp (A) ⊃ Hp (A)) ≡ (Gp (B) ⊃ Hp (B)))). . B. GL ⊢ (2X ≡ (22X ∧2X)). If F = q for p = q. GL ⊢ (⊡X ⊃ 2X). (H ⊃ (q ⊃ q)) is a tautology and hence an axiom of GL. The proof is by induction over the recursion that generates the formula F. Proof. GL ⊢ (⊡X ⊃ Y ). By Lemma 149. GL ⊢ (⊡(A ≡ B) ⊃ (Gp (A) ≡ Gp (B))). by Definition 84. which by Definition 84 is GL ⊢ (2X ≡ ⊡2X).e. Then by Lemma 152. GL ⊢ (⊡X ⊃ 2Y ). which holds by Definition 84 and propositional logic in GL. GL ⊢ (2 ⊡ X ⊃ 2Y ). an instance of Theorem 140. Corollary 151 GL ⊢ (⊡X ⊃ 2 ⊡ X) Proof. Then by Lemma 138. Assume GL ⊢ (⊡X ⊃ Y ). what is to be proved is GL ⊢ (⊡(A ≡ B) ⊃ (q ≡ q)). Lemma 152 If GL ⊢ (⊡X ⊃ Y ). Assume GL ⊢ (⊡X ⊃ Y ). From the assumption and this last result by propositional logic in GL. The provable equivalence of 2 ⊡ X and ⊡2X is. Remark. and GL ⊢ (⊡(A ≡ B) ⊃ (Hp (A) ≡ Hp (B))). Theorem 154 (strengthened arithmetized substitution theorem) For all formulas A. GL ⊢ (2X ⊃ 2Y ). If F = p. By Lemma 149. Lemma 153 GL ⊢ (⊡X ⊃ Y ) if and only if GL ⊢ (⊡X ⊃ ⊡Y ). GL ⊢ (⊡(A ≡ B) ⊃ (Fp (A) ≡ Fp (B))). The converse implication is not provable in GL. (ii) Assume GL ⊢ (⊡X ⊃ ⊡Y ). and F and propositional variable p. since GL (2X ⊃ X) unless GL ⊢ X (Lemma 137). GL ⊢ (⊡X ⊃ ⊡Y ). If F =⊥. GL ⊢ (2X ⊃ 2Y ). GL ⊢ (⊡X ⊃ (2Y ∧ Y )). so in particular GL ⊢ (⊡(A ≡ B) ⊃ (q ≡ q)). Then by Lemma 150 and propositional logic in GL. Proof. If F = (G ⊃ H). what is to be proved is GL ⊢ (⊡(A ≡ B) ⊃ (⊥≡⊥)). then GL ⊢ (2X ⊃ 2Y ). we have as induction hypotheses. Then by Theorem 139. The argument is as for the preceding case. namely GL ⊢ (2(2X ∧ X) ≡ (22X ∧ 2X)). For any formula H.LECTURE 13 115 Proof. Proof. what is to be proved is GL ⊢ (⊡(A ≡ B) ⊃ (A ≡ B)). By Theorem 147 (P3 ) and propositional logic in GL. (i) From Definition 84 by propositional logic in GL.

. . Axioms A2. . pim ) be a formula with sentence letters pi1 . GL ⊢ (2(Gp (A) ≡ Gp (B)) ⊃ (2(Gp (A)) ≡ 2(Gp (B)))). . we have as induction hypothesis that GL ⊢ (⊡(A ≡ B) ⊃ (Gp (A) ≡ Gp (B))). Proof. Am ) ≡ F (B1 . . and F and propositional variable p. pim . By Theorem 140. Then GL ⊢ (⊡(A1 ≡ B1 ) ∧ . . . . . . . Exercise. . . by propositional logic in GL and Defintion 83(5). . From these two last results. Theorem 156 (Theorem 154 generalized to multiple substitutions) Let F (pi1 . ∧ ⊡(Am ≡ Bm )) ⊃ (F (A1 . Proof. By Theorem 154 and Lemma 152. GL ⊢ (2 ⊡ (A ≡ B) ⊃ ((2Gp )(A) ≡ (2Gp )(B))). and propositional logic in GL. . Bm ))).LECTURE 13 116 If F = 2G. Corollary 155 (variant proof of Theorem 148) For all formulas A. B. GL ⊢ (⊡(A ≡ B) ⊃ ((2Gp )(A) ≡ (2Gp )(B))). . GL ⊢ (2(A ≡ B) ⊃ 2(Fp (A) ≡ Fp (B))). GL ⊢ (2⊡(A ≡ B) ⊃ 2(Gp (A) ≡ Gp (B))). Then by Lemma 138. . . Then by Corollary 151 and propositional logic in GL. .

. . Examples of sentences in which the sentence letter p is modalized: (1) 2p. (3) 2 ∼ p. Examples of sentences not modalized in p: (1) p. then Z is a subsentence of X. (2) (p ⊃⊥). Definition 86 A sentence letter p is modalized in a sentence X iff every occurrence of p in X is a subsentence of a subsentence of X of the form 2Y . a sentence D(pk1 . (8) 2(p ≡ (2p ⊃ q)). pkn that do not occur in X 117 . (9) (2(2p ⊃ q)∧ ∼ 2p). If the sentence 2Z is a subsentence of X. (5) (2p ⊃ q) (6) 2(2p ⊃ p). . Definition 87 For X a sentence in which the sentence letter p occurs modalized.Lecture 14 The fixed-point theorem for GL (Wednesday.1 The notion of a sentence letter modalized in a sentence. (12) 2 ⊥. (13) q (where q = p). and arithmetized substitution for modalized sentences Definition 85 Y is a subsentence of X is defined recursively by Base case: X is a substence of X Recursion clauses: If the sentence (Z ⊃ W ) is a subsentence of X. (4) (p ≡ (2p ⊃ q). . . 24 November 2010) 14. . (4) ∼ 2 ∼ p. . (14) 2q (in these latter cases p is modalized in the sentence since p does not occur in the sentence and so every occurrence of p in the sentence is within the scope of 2). (11) (⊥⊃⊥). then Z is a subsentence of X and W is a subsentence of X. (7) (2(2p ⊃ p) ⊃ 2p). (3) (2p ⊃ p). pkn ) with sentence letters pk1 . (10) ⊥. . (2) ∼ 2p.

and since p occurs modalized. This gives the decomposition sentence of X for those components. Then p occurs in Y or in Z. (ii) X is of the form (Y ⊃ Z).e. and 2C1 (p). and 2(2(2p ⊃ p) ⊃ p) for the third occurrence of p. Examples. find the innermost occurrence of 2 in whose scope that occurrence of p occurs. with the decomposition sentence for X any sentence letter pi distinct from p and component 2Y . The remaining sentences 2Ci (p) will be the components of the decomposition of X. . . a method. and wherever it occurs it occurs modalized. or X is of the form (Y ⊃ Z) for sentences Y and Z. . Method 1 (top down): The whole sentence X cannot consist just of a sentence letter pi or ⊥ since no such sentence is modalized in p. So there are two cases. . this process comes to an end. Then we are done. For each such occurrence of 2. let Ci (p) be the sentence to which that occurrence of 2 is prefixed. or both. . Method 2: When applied to the three occurrences of p in the antecedent sentence of this implication. Lemma 157 If X is modalized in p. 2Cn (p) are called components of X. . Method 2 (bottom up): For each occurrence of p in X. (i) X is of the form 2Y . the result of substituting each sentence 2Ci (p) for all occurrences of the variable pi in the sentence D(pk1 . the procedure results in sentence 2p for the first occurrence of p (going from the left). and for the single occurrence of p in the consequent sentence. . or in fact two different methods. If the sentence or sentences in which p occurs modalized is/are of the form 2W then by (i) we are done. it ends in components. 2(2p ⊃ p)) for the second occurrence of p. . . . . . sentences of the form 2C(p) in which p occurs modalized. i. In X replace each of these component sentences by a distinct sentence letter not occurring in X. We give a constructive proof of this result. pkn ) is called a decomposition sentence for X. . it results in 2p. From the sentences resulting from the occurrences of p in . D(pk1 . pkn )) is called a decomposition of X with respect to p. which generate a decomposition of X for X any sentence modalized in p. If not then it is of the form (U ⊃ V ) and we repeat the argument. . . . 2Cn (p) such that X(p) = D(2C1 (p).LECTURE 14 118 and sentences 2C1 (p). (2(2(2p ⊃ p) ⊃ p) ⊃ 2p) Method 1: The sentence is an implication between boxed formulas. From the resulting set of sentences 2Ci (p).e. Proof. discard those sentences that are a subsentence of any of the others in which that occurrence of p occurs. .e. either X is of the form 2Y for some sentence Y . . . at each occurrence in X of that component. . 2Cn (p)) (i. then there is a decomposition of X with respect of p. i. Since X is generated in finitely many steps. so the components are 2(2(2p ⊃ p) and 2p and the decomposition formula is (pi ⊃ pj ).

g. 2 p1 with component 2 . and the decomposition sentence for these components is (pi ⊃ pj ). and 2C2 (p) = 2p. .g. . (2(2p ⊃ q)∧ ∼ 2p) By Method 1: D2 (p1 . . (iii) D3 = 22p1 and 2C1 (p) = 2p. 2 p has decompositions Di = 2 . 2Cm (p). pm not in F (p) and components 2C1 (p). 222p has three decompositions: (i) D1 = p1 and 2C1 (p) = 222p. So the components are 2(2(2p ⊃ p) ⊃ p) and 2p. modalized in a sentence letter. Note that for some sentences modalized in a sentence letter the two methods results in different decompositions. . that has n-many decompositions. For each n. The decomposition of 222p that results from the proof of Lemma 157 is (iii). for each i such that 1 ≤ i ≤ n. pm ). 2 p. . . q) = (2(p1 ⊃ q)∧ ∼ p1 ). and 2C(p) = 2p. . Some modalized sentences have more than two decompositions. e. . with sentence letters p1 . Proof. The following is (a recipe for) a proof in GL. 2 . . .LECTURE 14 119 the antecedent sentence. . . . Theorem 159 (arithmetized substitution in modalized sentences) Let F (p) be a sentence in which sentence letter p occurs modalized. (ii) D2 = 2p1 and 2C1 (p) = 22p. Then GL ⊢ (2(A ≡ B) ⊃ (Fp (A) ≡ Fp (B))). there is a sentence. e. we discard 2p because 2p with that occurrence of p occurs in 2(2p ⊃ p)) (and also in 2(2(2p ⊃ p) ⊃ p)). . 2C1 (p) = 2(2p ⊃ q). p2 ) = (p1 ∧ ∼ p2 ). By Lemma 157. n−i i n For sentences in a which a sentence letter occurs modalized there is an arithmetized substitution theorem which yields the conclusion of Theorem 154 on the hypothesis of Theorem 148. and we discard 2(2p ⊃ p)) since it is a subsentence of 2(2(2p ⊃ p) ⊃ p) and the two occurrences of p in 2(2p ⊃ p)) occur in 2(2(2p ⊃ p) ⊃ p). . . By Method 2: D1 (p1 . there is a decomposition D(p1 . . . . Proof. The second example generalizes to show that Proposition 158 For each n.

Lemma 160 For pi any propositional variable and Y (pi ) any sentence in the language of GL in which pi occurs. . . We establish this result first for the simplest case of a sentence modalized in pi . (4) (5) (6) transitivity of ⊃. . (1) (2) (3) (4) (5) (6) (7) (8) (2Y (⊤) ⊃ (⊤ ≡ 2Y (⊤)) 2(2Y (⊤) ⊃ (⊤ ≡ 2Y (⊤)) (22Y (⊤) ⊃ 2(⊤ ≡ 2Y (⊤))) (2Y (⊤) ⊃ 2(⊤ ≡ 2Y (⊤))) (2(⊤ ≡ 2Y (⊤)) ⊃ 2(Y (⊤) ≡ Y (2Y (⊤)))) (2(Y (⊤) ≡ Y (2Y (⊤))) ⊃ (2Y (⊤) ≡ 2Y (2Y (⊤)))) (2Y (⊤) ⊃ (2Y (⊤) ≡ 2Y (2Y (⊤)))) (2Y (⊤) ⊃ 2Y (2Y (⊤))) tautology. Proposition 141. Proof. i. . (1) P1 . (7) prop logic. ∧ ⊡(2Cm (A) ≡ 2Cm (B))) ⊃ (D(2C1 (A). . 2Cm (B)))) Theorem 156 (2(A ≡ B) ⊃ (Fp (A) ≡ Fq (B))) (5)(6) prop logic 120 (7) 14. a sentence F that contains only sentence letters contained in X and does not contain pi is a fixed point for X with respect of pi if and only if GL ⊢ (F ≡ Xpi (F )). and write 2Y (⊤) for the result of substituting (⊥⊃⊥) for each occurrence of pi in 2Y (pi ). We shall see that every sentence modalized in pi has a fixed point with respect to pi . In this proof I abbreviate (⊥⊃⊥) as ⊤. .e. 2Cm (A)) ≡ D(2C1 (B). . of the form 2Y (pi ). . (2) P2 Modus ponens (3) P3 propositional logic. the result of substituting (⊥⊃⊥) for each occurrence of pi in 2Y (pi ) is a fixed point for 2Y (pi ). . The following derivation is a proof in GL. . Theorem 148.LECTURE 14 (1) (2) (3) (4) (5) (6) (2(A ≡ B) ⊃ 2(Cip (A) ≡ Cip (B))) Theorem 148 (2(A ≡ B) ⊃ (2Ci (A) ≡ 2Ci (B))) (1) Proposition 141 prop logic (22(A ≡ B) ⊃ 2(2Ci (A) ≡ 2Ci (B))) (2) Lemma 138 (2(A ≡ B) ⊃ 2(2Ci (A) ≡ 2Ci (B))) (3) Theorem 147 prop logic (2(A ≡ B) ⊃ ⊡(2Ci (A) ≡ 2Ci (B))) (2)(4) (((⊡(2C1 (A) ≡ 2C1 (B))) ∧ . .2 The fixed point theorem for GL Definition 88 (fixed point) For X a sentence in the language of GL that contains the sentence letter pi .

Fm (pkm+1 ). . The proof is by induction on m. Proof. . . . . . . . GL ⊢ (Fi ≡ Ai (F1 . . . pkm )) is solvable (in GL) iff there are sentences F1 . . Fm )). . . . . (14) (2Y (⊤) ≡ 2Y (2Y (⊤))) Before proving the fixed point theorem itself we need another lemma and for that lemma we need a definition: Definition 89 For 1 ≤ i ≤ m. (1) GL ⊢ (Fi (pkm+1 ) ≡ 2Ci (F1 (pkm+1 ). pkm such that for 1 ≤ i ≤ m. (8)(13)∧-Introduction. pkm . . . be m-many sentences each of which includes among its sentence letters the m-many sentence letters pk1 . . . . . pkm+1 . pkm and does contain the sentence letter pkm+1 . each of which does not contain the sentence letters pk1 . Let 2Ci (pk1 . By the Induction Hypothesis. pkm . (11) Lemma 138 (P4 ). . . . . . Lemma 161 (solving systems of simultaneous equivalences) Every system of m-many simultaneous equivalences of the form (pki ≡ 2Ci (pk1 . . such that for 1 ≤ i ≤ m. prop logic (10) prop logic. . . . pkm . m = 1. . pkm+1 )).LECTURE 14 (9) (10) (11) (12) (13) (2Y (⊤) ⊃ ⊡(⊤ ≡ 2Y (⊤))) (2Y (⊤) ⊃ (Y (⊤) ≡ Y (2Y (⊤)))) (Y (2Y (⊤)) ⊃ (2Y (⊤) ⊃ Y (⊤))) (2Y (2Y (⊤)) ⊃ 2(2Y (⊤) ⊃ Y (⊤))) (2Y (2Y (⊤)) ⊃ 2Y (⊤)) 121 (1) (4) prop logic (9). . . let Ai (pk1 . . . pkm+1 ) be a set of (m + 1)-many sentences with m + 1-many sentence letters pk1 . pkm ). . pkm )) is solvable. . Suppose the Lemma holds for m. . there are sentences Fi (pkm+1 ). . . . 1 ≤ i ≤ m. Theorem 154. A system of simultaneous equivalences of the form (pki ≡ Ai (pk1 . . (12) A3 prop logic. . Fm not containing pk1 . . This case is Lemma 160. .

. X(pr ) = D(2C1 (pr ). Fm+1 solve the set of equivalences. . 1993. . the set of equivalences (pki ≡ 2Ci (D(pk1 . Theoria 62 (1996). non self-referential. Fn ) is a fixed point for X(pr ) with respect to pr . where the sentence D(pk1 . 2Cn (pr ). . . . . . where 2Ci (pr ) is substituted for pki in D(pk1 . . . which shows that D(F1 . . . and components 2C1 (pr ). . Fn ))). pp. pp. pp. . GL ⊢ (D(F1 . . Fm (pkm+1 ). . 2Cn (pr )). . pkn ). Cambridge University Press.e. i. . . . pkn ) is substituted for pr in 2Ci (pr ). GL ⊢ (Fi ≡ 2Ci (D(F1 . . there is a sentence Fm+1 such that (2) GL ⊢ Fm+1 ≡ 2Cm+1 (F1 (Fm+1 ). Proof. for each i = 1. . Then by Lemma 157. . Remark. 31-35. . . . Fn )).LECTURE 14 122 Now by Lemma 160 applied to the sentence that results from 2Cm+1 (pk1 . 1985. The purely syntactic proof given here follows Per Lindstrom. There are a number of different proofs of this theorem.e. Fm (Fm+1 ). . . pkm+1 ). . . Springer. . pkm+1 ) by substituting Fi (pkm+1 ) for pki for each i such that 1 ≤ i ≤ m. Fm (Fm+1 ). Let F1 . 2Cm+1 (F1 (pkm+1 ). . . i. 104-123). n. i. Fn be a solution. . Theorem 162 (Fixed Point Theorem) For every sentence X in the language of GL modalized in pr . By substitution of Fm+1 into the provable equivalences (1). . there is a sentence F containing only sentence letters that occur in X and not containing pr such that GL ⊢ (F ≡ Xpr (F )). . . . . pkn ))) indexed by i = 1. . Let X be any sentence modalized in pr . . . the sentences F1 (Fm+1 ). pkn distinct from pr . . with sentence letters pk1 . . . Self-Reference and Modal Logic. By Lemma 161.e. for X modalized in p. “Provability logic— a short introduction”. Fm+1 )). most of which use Kripke models for the modal operator (three such proofs are expounded by George Boolos in his book The Logic of Provability. . . pkm . . see also Craig Smorynski. Fn )))). . . n. The fixed points proved to exist by the Fixed Point Theorem are unique to within provable equivalence. and let q be a sentence letter that does not occur in X(p). Theorem 163 (provable equivalence of fixed points) Let X(p) be a sentence in the language of GL in which the sentence letter p occurs modalized. . . 78-82. is solvable. . GL ⊢ ((2(p ≡ X(p)) ∧ 2(q ≡ X(q))) ⊃ 2(p ≡ q)). . . . . . . . . . . . 2Cn (D(F1 . . . . . X(pr ) has a decomposition D(pk1 . . .e. Fn ) ≡ D(2C1 (D(F1 . . . content of a self-referential sentence p such that (p ≡ X(p)). . . Abbreviating Xp (q) as X(q). i. Then by Theorem 145. . . . The sentence F gives the explicit. . . . pkn ). .

This sentence is provably equivalent to 2p. It is also worth noticing how modalizing our original example to non-uniqueness to within provable equivalence of a fixed point for a non-modalized sentence yield provable equivalence. Then by Proposition 143. GL ⊢ ⊡(F ≡ F ). Proof. e. The uniqueness of the fixed point to within provable equivalence turns essentially on the sentence letter of the fixed point equivalence occurring modalized in the sentence for which the existence of a fixed point follows. Then by Axiom A3 and propositional logic. Then by propositional logic. GL ⊢ 2(q ≡ ((r ⊃ r) ⊃ ((q ⊃ q) ⊃ q))) and GL ⊢ 2(r ≡ ((r ⊃ r) ⊃ ((q ⊃ q) ⊃ r))). so by ∧-Elimination. Theorem 165 Theorem 164 implies Theorems 162 and 163. By Definition 88. GL 2 ⊥. since GL ⊢ (q ≡ ((r ⊃ r) ⊃ ((q ⊃ q) ⊃ q))) and GL ⊢ (r ≡ ((r ⊃ r) ⊃ ((q ⊃ q) ⊃ r))). GL ⊢ (F ≡ X(F )). GL ⊢ 2(⊥≡ (⊥⊃⊥)). The previous example can be tweaked to give an example of sentence in which p occurs not modalized that has a fixed point with respect to p which is unique to within provable equivalence. in which case GL ⊢ 2 ⊥. take X(p) as ((r ⊃ r) ⊃ ((q ⊃ q) ⊃ p)). GL ⊢ (F ≡ F ). Suppose GL ⊢ 2(q ≡ r). taking A as p and B as q. q and r are both fixed points of ((r ⊃ r) ⊃ ((q ⊃ q) ⊃ p)). ((2(p ≡ X(p)) ∧ 2(q ≡ X(q))) ⊃ 2(2(p ≡ q) ⊃ (p ≡ q))). Proof. For example. To find a fixed point for (∼ 2p ∧ ∼ 2 ∼ p). there is a sentence F containing only sentence letters that occur in X and not containing p such that GL ⊢ (⊡(p ≡ X(p)) ≡ ⊡(p ≡ F )). By Lemma 138 and Theorem 140. By this criterion. In which case GL 2(q ≡ r). By Theorem 159 (arithmetized substitution into modalized sentences). . so GL ⊢ 2(F ≡ F ). Theorem 164 (Strengthened Fixed Point Theorem for GL) For every sentence X in the language of GL modalized in p. Theorems 162 and 163 establish this result. By R2 . We can have fixed points for non-modalized sentences that are not unique with respect to provable equivalence. by Theorem 135. (((p ≡ X(p)) ∧ (q ≡ X(q))) ⊃ (2(p ≡ q) ⊃ (p ≡ q))). ((2(p ≡ X(p)) ∧ 2(q ≡ X(q))) ⊃ 2(p ≡ q)). Example.g. Substituting F for p in Theorem 164 results in GL ⊢ (⊡(F ≡ X(F )) ≡ ⊡(F ≡ F )) Since (F ≡ F ) is a tautology.g. e. ((r ⊃ r) ⊃ ((q ⊃ q) ⊃ 2p)). Hence GL ⊢ ⊡(F ≡ X(F )). ((q ⊃ q) ⊃ p).LECTURE 14 123 Proof. (2(p ≡ q) ⊃ (X(p) ≡ X(q))) is provable in GL. a fixed point for X(p) is a sentence X containing only sentence letters that occur in X and not containing p such that GL ⊢ (F ≡ Xp (F )). But assuming the PA is Σ1 -sound. so by ∧-Introduction. so all fixed points for it are provably equivalent to the Henkin sentence.

English translation in Jean van Heijenoort (ed). 821-865. Oxford University o Press. Mathematische Annalen 78 (1918). G¨del’s Incompleteness Theorems. 108-109. A Source Book in Matheo matical Logic 1879-1931. [3] David Hilbert. The Journal of Symbolic Logic 22 (1957). 1992. [7] Raymond M. Harvard University Press.) From Frege to G¨del. [6] Craig Smorynski. “Die Grundlagen der Mathematik”.Bibliography [1] George Boolos. pp. [5] Georg Kreisel. (1927). p. ‘The foundations of mathematics” in Jean van Heijenoort (ed. Cambridge University Press. “The incompleteness theorems”. 405-415. 471. [2] David Hilbert. English translation in William B. pp. 1993. Horth-Holland Publishing Company. Handbook of Mathematical Logic.). “A refinement of ω-consistency” (Abstract). Jon Barwise (ed. [4] David Hilbert. “Axiomatische Denken”. Smullyan. The Logic of Provability. “On the infinite” (1926). Ewald (ed). pp. 1977. 1967. English a translation by Stephan Bauer-Mengelberg. 124 . Abhandlungen aus dem mathematischen Seminar der Hamburgeshcen Universit¨t 6 (1928).

Sign up to vote on this title
UsefulNot useful