Documente Academic
Documente Profesional
Documente Cultură
net/publication/31167882
CITATIONS READS
4 120
2 authors, including:
Mohammad Ardeshir
Sharif University of Technology
38 PUBLICATIONS 158 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Mohammad Ardeshir on 01 February 2017.
Abstract
We study Basic Arithmetic BA, which is the basic logic equivalent of Heyting Arithmetic
over intuitionistic logic, and of Peano Arithmetic over classical logic. It turns out that The
Friedman translation is applicable to BA. Using this translation, we prove that BA is closed
under a restricted form of the Markov rule. Moreover, it is proved that all nodes of a finite
Kripke model of BA are classical models of I∃+
1 , a fragment of Peano arithmetic with Induction
restricted to the formulas made up of ∃, ∧ and/or ∨. An interesting extension of BA, called
EBA, is the extension by the axiom > → ⊥ ⇒ ⊥. This extension behaves very like to HA.
We prove a D. de Joung style theorem for this extension.
Mathematics Subject Classification: Primary 03F30; secondary 03F50.
Keywords: Basic Logic, Basic Arithmetic, Completeness, Heyting Arithmetic, Kripke models.
1 Introduction
Basic Predicate Calculus BQC is introduced by W. Ruitenburg in [4] and [5]. He was motivated
by a revision of the Brouwer-Heyting-Kolmogorov proof interpretation (see [4].) BQC is a proper
subsystem of Intuitionistic Predicate Calculus IQC, in which the rule Modus Ponens is weakened.
BQC is strongly complete with respect to a generalized class of Kripke models, i.e., Kripke models
whose underlying set is not necessarily reflexive. Basic Arithmetic BA is the basic logic equivalent of
Heyting Arithmetic over intuitionistic logic and of Peano Arithmetic over classical logic. Ruitenburg
in [5] axiomatized BA and using Kripke model theory, proved that BA has disjunction and explicit
definability properties. We use his axiomatization for BA and for the sake of completeness, we
include in this paper some of his basic results.
In section 2, we will present axiomatizations of BQC, , BA, that is from [5], and then some
elementary basic facts of BA is recalled to learn how to live in BA. In section 2.2, we apply the
Friedman translation to BA. Using this translation, it is shown that BA is closed under a restricted
form of the Markov rule. Moreover, it is also proved that all nodes of a finite Kripke model of BA
are classical models of I∃+
1 , a fragment of Peano arithmetic with Induction restricted to the formulas
made up of ∃, ∧ and/or ∨.
In the last section of this paper we will consider an extension of basic logic augmented by the
axiom > → ⊥ ⇒ ⊥. It turns out that this logic behaves very similar to IQC. In particular, the
arithmetical extension of BA, called EBA, by this axiom is very like to HA, for example, it satisfies
a D. Joung style theorem, namely, the propositional part of that extension, EBPC is complete with
respect to EBA.
1
2 Basic Arithmetic
2.1 Axioms, Rules and some elementary facts
The language of BQC is a little different from the usual one for IQC. It was originally axiomatized
in sequent notation, i.e., A ⇒ B where A and B are formulas in the language {∨, ∧, →, ⊥, >, ∃, ∀}.
Since Modus Ponens is not a rule in BQC, the universally quantified formulas ∀x∀y A is different
from ∀xy A. The first one over BQC and IQC is equivalent to Ex → (Ey → A), where E is the
extent operator of Heyting and Scott (see [7]) and the second is equivalent to Ex ∧ Ey → A. So in
BQC, when we write ∀x (A → B), we mean x to be a finite sequence of variables once quantified.
Besides, a set of predicate and function symbols of possibly different finite arity, we include the
binary predicate “=” for equality. Terms, atomic formulas and formulas are defined as usual except
for universal quantification: if A and B are formulas, and x is a finite sequence of variables, the
∀x (A → B) is a formula. The concepts of free and bound variables are defined as usual. A sentence
is a formula with no free variable. An implication is a universal quantification ∀x (A → B) with
empty variables x. ¬A means A → ⊥. Given a sequence of variables x without repetitions, s[x/t]
and A[x/t] stand for, respectively, the term and formula that results from substituting the term t
for all free occurrences of the variables of x in the term s and the formula A. For details, see [5].
In the following rules a double horizontal line means that the rule is reversible.
1. A ⇒ A,
2. A ⇒ >,
3. ⊥ ⇒ A,
4. A ∧ (B ∨ C) ⇒ (A ∧ B) ∨ (A ∧ C),
5. A ∧ ∃x B ⇒ ∃ x(A ∧ B), where x is not free in A,
6. > ⇒ x = x,
7. x = y ∧ A ⇒ A[x/y], where A is atomic,
8. ∀x (A → B) ∧ ∀x (B → C) ⇒ ∀x (A → C),
9. ∀x (A → B) ∧ ∀x (A → C) ⇒ ∀x (A → B ∧ C),
10. ∀x (B → A) ∧ ∀x (C → A) ⇒ ∀x (B ∨ C → A),
11. ∀x (A → B) ⇒ ∀x (A[x/t] → B[x/t]), where no variable in the sequence of terms t is bound
by a quantifier of A or B,
12. ∀x (A → B) ⇒ ∀y (A → B), where no variable in y is free on the left hand side,
13. ∀yx (B → A) ⇒ ∀y (∃x B → A), where x is not free in A,
A⇒B A⇒C
14. A⇒B∧C ,
B⇒A C⇒A
15. B∨C ⇒A ,
16. A⇒B , where no variable in the sequence of terms t is bound by a quantifier in the
A[x/t] ⇒ B[x/t]
denominator,
17. B ⇒ A , where x is not free in A,
∃x B ⇒ A
18. A ∧ B ⇒ C , where no variable in x is free in A,
A ⇒ ∀x (B → C)
1. Sx = 0 ⇒ ⊥,
2. Sx = Sy ⇒ x = y,
3. x + 0 = x,
4. x + Sy = S(x + y),
5. x · 0 = 0,
6. x · Sy = x · y + x,
7. ∀yx (A → A[x/Sx]) ⇒ ∀yx (A[x/0] → A),
A ⇒ A[x/Sx]
8. .
A[x/0] ⇒ A
We often write A for > ⇒ A. Let Γ be a set of sequents and rules. We say Γ proves A ⇒ B,
written Γ ` A ⇒ B, when A ⇒ B can be obtained, after finitely many applications of the BQC
rules and the rules of Γ from the BQC axioms and the axioms of Γ . Similarly, Γ proves the rule
A1 ⇒ B1 , . . . , An ⇒ Bn
R= A⇒B ,
Proposition 2.1 (Soundness and Completeness) BA is sound and complete with respect to
Kripke models for the language L = {0, S, +, ·}.
Let L be the language of BA, i.e., {0, S, +, ·} which includes at least one constant symbol. So
the set T of closed terms of L is nonempty. Let S = T / ∼, where ∼ is the equivalence relation on T
defined by s ∼ t when Γ ` s = t. For each set of models {Ki }i∈I of BA we can construct two new
models, denoted by K• and K◦ , as follows. Both models are formed by taking the disjoint union of
the models {Ki }i∈I and then adding a new root k0 with domain D(k0 ) = S. In model K• the node
k0 is reflexive and in model K◦ , the node is irreflexive.
Let C be the class of rooted Kripke models of BA. We call Γ reflexively rooted if for each set
of models {Ki }i∈I ⊆ C of Γ , the model K• is also a model of Γ . Similarly an irreflexively rooted
theory is defined. The theory Γ is called fully rooted if both K• and K◦ are models of Γ .
A theory Γ has disjunction property, DP, if Γ ` A ∨ B implies Γ ` A or Γ ` B, for all sentences
A and B. A theory Γ has explicit definability property, or existence property, EP, if Γ ` ∃xA
implies Γ ` A[x/t] for some closed term t in L, for all sentences ∃xA.
Corollary 2.4 BA, HA = IQC + BA, and PA = CQC + BA prove same positive sentences.
Proof. We prove that if PA ` A, for some positive sentence A, then BA ` A. Other cases are easy
or similar.
Let PA ` A, for a positive sentence A. Let K be a model of BA. Add a new root k0 to K with
universe of natural numbers. The new model K0 is also a model of BA, by Propositions 2.2. k0
structure by itself is a model of PA, i.e., k0 |= A. So k0
A, and then BA |= A. So BA ` A, by
Theorem 2.1. Similar argument works for PA and HA. a
Proof. We prove only two cases to see how work with BA.
2. First we show that BA ` Sx+y = S(x+y). Let A(x, y) be the formula Sx+y = S(x+y). We
have BA ` ∀xy[A(x, y) → A(x, Sy)] ⇒ ∀xy[A(x, 0) → A(x, y)]. But BA ` ∀yx[Sx+y = S(x+y) →
Sx + Sy = S(x + Sy)]. That means BA ` ∀xy[A(x, y) → A(x, Sy)] Then BA ` ∀xy[A(x, 0) →
A(x, y)]. By faithfulness of BA, we have BA ` A(x, 0) ⇒ A(x, y). Since BA ` Sx + 0 = S(x + 0),
then BA ` A(x, y).
To prove (2), let A(x, y) be the formula x+y = y +x. We have BA ` ∀xy[A(x, y) → A(Sx, y)] ⇒
∀yx[A(0, y) → A(x, y)]. Note that by the above fact BA ` ∀xy[A(x, y) → A(Sx, y)]. So BA `
∀yx[A(0, y) → A(x, y)]. Again by faithfulness of BA, we have BA ` A(0, y) ⇒ A(x, y). But A(0, y)
is 0 + y = y + 0 and it is easy to show that BA ` A(o, y). So BA ` A(x, y).
14. By 8, BA ` (x = 0 ∨ x > 0) ∧ (y = 0 ∨ y > 0). Then BA ` ((x = 0 ∧ y = 0) ∨ (x > 0 ∧ y =
0) ∨ (x = 0 ∨ y > 0)) ∨ (x > 0 ∧ y > 0). Let A ≡ (x = 0 ∧ y = 0) ∨ (x > 0 ∧ y = 0) ∨ (x = 0 ∧ y > 0) and
B ≡ x > 0 ∧ y > 0. We have BA ` x · y = 0 ⇒ (x · y = 0) ∧ (A ∨ B). Then BA ` x · y = 0 ⇒ (x · y =
0 ∧ A) ∨ (x · y = 0 ∧ x · y > 0), since BA ` B ⇒ x · y > 0, by 13. So BA ` x · y = 0 ⇒ x · y = 0 ∧ A,
since BA ` x · y = 0 ∧ x · y > 0 ⇒ ⊥. Then BA ` x · y = 0 ⇒ A. Now, it is easy to see that
BA ` A ⇒ x = 0 ∨ y = 0. So BA ` x · y = 0 ⇒ x = 0 ∨ y = 0. a
Proposition 2.6 BA ` x < y ∨ x = y ∨ y < x.
Proof. Let A(x, y) be the formula x < y ∨x = y ∨y < x. We have BA ` ∀xy[A(x, y) → A(x, Sy)] ⇒
∀xy[A(x, 0) → A(x, y)]. By item 6 of Lemma 2.5, BA ` A(x, 0). We also have
1. BA ` x < y ⇒ x < Sy, since BA ` y < Sy,
2. BA ` x = y ⇒ x < Sy,
3. BA ` y < x ⇒ x = Sy ∨ Sy < x, by Lemma 2.5( 16).
So BA ` ∀xy[A(x, y) → A(x, Sy)]. Then BA ` ∀xy[> → A(x, y)]. By faithfulness of BA, we will
have BA ` A(x, y). a
It is worth mentioning that the cancellation law, x + y = x + z ⇒ y = z does not hold in BA. To
see this, let K be a Kripke model with just one irreflexive node with structure N∗ , with the universe
consisting of all natural numbers with a top nonstandard element a, i.e., n < a, for all n ∈ N . It
can easily be seen that K |= BA, and moreover, K |= ∃x∃y(x + Sy = x).
We end this section with two more properties of BA.
Proof. We use induction on x in the formula A(x) := ∃r∃s (x = y · s + r ∧ r < y). Clearly
BA ` A(x) ⇒ x + 1 = y · s + (r + 1). Since BA ` r < y ⇒ r + 1 < y ∨ r + 1 = y, then BA ` A(x) ⇒
(x+1 = y·(s+1)+0)∨((x+1 = y·s+(r+1))∧r+1 < y). So BA ` ∃r1 ∃s1 (x+1 = y·s1 +r1 ∧r1 < y).
Then BA ` A ⇒ A[x/Sx]. Clearly BA ` y 6= 0 ⇒ A[x/0]. Then BA ` y 6= 0 ⇒ A. a
Definition 2.10 Let A be any formula of BA. For an arbitrary formula, we define (·)A by
• P A = P ∨ A, for atomic P , and ⊥A = A,
• (B ◦ C)A = B A ◦ C A , for ◦ ∈ {∧, ∨, →},
• (Qx B)A = Qx B A , for Q ∈ {∀, ∃}, with the condition that where x occurs free in A, we first
replace all free occurrences of x in B by a variable not occurring free in A, and finally, for
sequents define
• (B ⇒ C)A = B A ⇒ C A .
Let Γ be a set of formulas and sequents. Define Γ A to be {B A : B ∈ Γ }.
Lemma 2.11 Let Γ A = {(B ⇒ C)A : B ⇒ C ∈ Γ }, and let ` be derivability in BQC. Then
• ` A ⇒ BA,
• if Γ ` B ⇒ C then Γ A ` B A ⇒ C A ,
• if BA ` B ⇒ C then BA ` B A ⇒ C A .
Proof.
• As usual, ` A ⇒ B A is proved by induction on B. The only interesting case is when B is
∀x (C → D). By induction, ` A ⇒ DA . Then ` A ∧ C A ⇒ DA . So ` A ⇒ ∀x (C → D)A .
• It is enough to check that all axioms and rules of BQC are closed under the translation. That
is easy.
• By second clause, if BA ` B ⇒ C then BAA ` B A ⇒ C A . So it is enough to show that BA
proves the translation of its axioms and rules. That is easy.
a
Proof. Suppose BA ` ¬¬∃x A(x), for a positive formula A(x). Then by Lemma 2.11, BA `
(¬¬∃x A(x))∃x A(x) . So BA ` ((∃x A(x) → ⊥) → ⊥)∃x A(x) , i.e., BA ` (∃x A(x) ∨ ∃x A(x) →
∃x A(x)) → ∃x A(x). Then BA ` (∃x A(x) → ∃xA(x)) → ∃x A(x). By faithfulness of BA, we have
BA ` ∃x A(x) → ∃x A(x) ⇒ ∃x A(x). Then BA ` ∃x A(x). a
Definition 2.13 Let K = (K, ≺, D,
), be a Kripke model for the language of BA, B a sentence
such that, for at least one node k ∈ K, B ∈ Lk and k 1 B. Then the model obtained by pruning
B-nodes from K shall be the model obtained from K by deleting all nodes that forces B. This model
will be denoted as KB = (K B , ≺B , DB ,
B ). Recall that AB , for formulas A and B, is the Friedman
translation of A by B.
In [2], the authors, by pruning, proved that “every finite Kripke model of HA is locally PA.”
A Kripke model K for the language of arithmetic is called locally PA if all its nodes are classical
models of PA. In a finite Kripke model of HA, the terminal nodes are trivially (classical) models
of PA. By pruning these terminal nodes, and possibly other classical nodes, we change the original
Kripke model to “smaller” model of HA. Then, by an induction based argument, one concludes
that all nodes of the pruned model are classical. The following simple proposition shows that even
a one-node Kripke model of BA may not be locally PA:
Proposition 2.16 There is a one node Kripke model of BA that is not locally PA.
Proof. Let I∃+ 1 be the fragment of PA with the Induction schema restricted to positive formulas.
Take a single irreflexive node Kripke model with the attached structure to be a classical model of
I∃+
1 . This is a Kripke model of BA. a
Proposition 2.17 Let K be a model of BA, and k ∈ K. Then for every positive sentence A, k
A
iff k A.
Let K be a non-single tree model of BA and let k be its root. If one can find a formula A such
that k
> → A but k 1 A, then one can prune the model to obtain KA , whose underlying frame
must be the singleton {k}. As we have seen, this node is a classical model of I∃+
1 . If for each formula
A we have “if k
> → A then k
A” then k is not a “proper” irreflexive node, so we call such
a node an “improper irreflexive node” through this section. Using induction, one can show that
“improper irreflexive nodes can be taken as reflexive nodes.”
Lemma 2.19 Let ◦ ∈ K be an improper irreflexive node. Define ≺0 :=≺ ∪{(◦, ◦)},
0 :=
and take
K0 to be the model (K, ≺0 , D,
0 ). Then for each k ∈ K and all formula A we have k
A iff k
0 A.
From here to the end of this section, by an “irreflexive node” we mean an “proper irreflexive
node.” Of course, this makes no loss of generality. From the discussion above, we have
Theorem 2.20 Every irreflexive node in a Kripke model of BA is a classical model of I∃+
1.
Proof. Let k be the irreflexive node we are speaking about and let K be the model. Consider the
truncated model Kk and let A be the formula such that k
> → A but k 1 A. Then the A-pruned
model (Kk )A satisfies the hypothesis of Theorem 2.18. a
Suppose that we have a (classical) prenex formula A with a positive open part. Replace each
subformula of the form ∀x B with ∀x (> → B) and call the obtained formula Ab .
Lemma 2.21 Let K be a Kripke model and let k ∈ K be a reflexive node. Then for a prenex
formula A with a positive open part, if k
Ab then k A.
Theorem 2.22 Let K be a finite (depth) Kripke model of BA. Then all its nodes are classical
models of I∃+
1 . Further, each reflexive node is a model of PA.
Proof. By Theorem 2.20, it is sufficient to prove that reflexive nodes are classical models of PA.
To do that, suppose that k be a reflexive node in K and consider the truncated model Kk . So we
may suppose that the root of our model is reflexive. Now, proceed by induction on the depth of the
model.
If K has maximal irreflexive nodes, then consider K>→⊥ . If all maximal nodes are irreflexive
then we are done.
If K has maximal reflexive nodes, then proceed as follows. Every consequence of PA can be
written as a prenex formula with a positive open part. If k satisfies PA then we have nothing to
prove. Otherwise, let A be a (prenex with positive open part) consequence of PA with k 2 A. Then
k 1 Ab , where all (reflexive) maximal nodes force Ab , and then we can use pruning with respect to
b
Ab . Now, (K>→⊥ )A is of lower depth. a
Note that, the tree need not to be finite, but it must be of finite depth. In case of finite trees,
one can also use induction on the number of nodes, which is a little easier to handle: either our
finite model has an irreflexive node or all its nodes are reflexive. In the first case use pruning as in
proof of Theorem 2.20, and in the second case, recall that the model is a finite model of HA. Using
the argument, we can extend the result to a wider class of models.
Proof. Proofs of all items are like the intuitionistic case. The key step stopping derivations of
the above sequents in BQC is the sequent schema > → ⊥ ⇒ ⊥. For example, let us look at some
items. For the first one, note that EBQC ` A ∧ ¬A ⇒ (> → A) ∧ (A → ⊥) ⇒ > → ⊥ ⇒ ⊥.
So EBQC ` A ⇒ ¬¬A. The second one, so called the Brouwer Theorem on logic, is derived from
the first item and the fact that BQC ` A → B ⇒ ¬B → ¬A. For the fifth clause, note that
BQC ` ∀x(A → B) ⇒ A[x/t] → B[x/t], and BQC ` A → B ⇒ ∀x(A → B), if x is not free on the
left hand side. The intuitionistic derivation of the last one holds here as well. a
Proof.
1. Proof is like in the intuitionistic case, by induction on the complexity of A. We may use
Lemma 3.1.
2. By first clause it is enough to show EBQC ` > → A ⇒ ¬¬A. We have EBQC ` (> →
A) ∧ (A → ⊥) ⇒ > → ⊥ ⇒ ⊥. So EBQC ` > → A ⇒ ¬¬A.
a
Definition 3.4 For a formula A in the language of CQC, define the Gödel translation of A (denoted
by Ag ) by the following clauses.
• Ag := ¬¬A for atomic A, and ⊥g := ¬¬⊥,
The role of ¬¬ in ∀x¬¬Ag is just keeping Ag in the language of EBQC. Furthermore, we can
translate ∀x (A1 → A2 ) to ∀x (Ag1 → Ag2 ). Note also that EBQC ` ¬¬⊥ ⇔ ⊥. So by > = ⊥ → ⊥,
we have EBQC ` >g ⇔ >. Similarly EBQC ` (¬A)g ⇔ ¬Ag .
Corollary 3.5 For every formula A, EBQC ` ¬¬Ag ⇔ Ag .
Proof. Note that for any formula A, Ag is a negative formula. Then use Lemma 3.3. a
Proof.
1. Clear.
2. The ‘if’ part is clear. The proof of the ‘only if’ part uses Induction on the the length of
derivation of Γ `c A. We may use Lemma 3.3.
a
Kripke semantics for BPC and BQC are introduced in [1] and [5]. Now we have:
Theorem 3.7 A Kripke model of BQC is a model of EBQC iff all its terminal nodes are reflexive.
Corollary 3.8 EBQC is sound and strongly complete with respect to the rooted Kripke models with
reflexive terminal nodes.
Corollary 3.9 EBPC is sound and complete with respect to the rooted finite Kripke models with
reflexive terminal nodes.
Proof. The ‘if’ part is trivial. For the converse, suppose that CPC ` A. Then, all terminal nodes of
any Kripke model of EBPC force A and so the root forces ¬¬A. By completeness, EBPC ` ¬¬A.
a
EBA is the extension of Basic Arithmetic, BA, by the axiom schema > → ⊥ ⇒ ⊥. Similar to
BA, it can easily be seen that:
Theorem 3.13 EBA has disjunction and existence properties.
It is worth knowing that EBA is not closed under the Friedman translation, since EBA 0 (> →
⊥)A ⇒ ⊥A , i.e., EBA 0 > → A ⇒ A. So the technique called “pruning”, [2] is not applicable for
EBA. However, we have:
Theorem 3.15 Let A(x) be a formula containing only x as a free variable and let EBA ` A(x) ∨
¬A(x). If EBA ` ¬¬∃xA(x) then EBA ` ∃xA(x).
Proof. Suppose EBA ` ¬¬∃xA(x). Then ∃xA(x) is true in one node model of EBA. So A(n), for
some n, is true in one node model of EBA. But EBA ` A(n) ∨ ¬A(n) and then by Theorem 3.13,
EBA ` A(n) or EBA ` ¬A(n). So EBA ` A(n), and therefore EBA ` ∃xA(x). a
Theorem 3.16 Let A(x, y) be a formula containing only x, y as free variables and let EBA `
A(x, y) ∨ ¬A(x, y). Then PA ` ∃yA(x, y) if and only if EBA ` ∃yA(x, y)g , where A(x, y)g is the
Gödel translation of A(x, y).
Proof. Let PA ` ∃yA(x, y). Then EBA ` ¬∃y¬A(x, y)g , by Theorem 3.14. Then EBA `
¬¬∃yA(x, y)g , by Lemma 3.1. Now, by Theorem 3.15, EBA ` ∃yA(x, y)g . a
Corollary 3.18 If (K, ≺) can be embedded in (K 0 , ≺0 ) then
0 can be defined such that for all k ∈ K
and each proposition A, k
A iff k
0 A.
This corollary by itself is interesting: suppose that we have a family of (finite) trees such that every
finite tree can be embedded in (at least) one of them. Then we have a completeness result with
respect to all Kripke models with frames in this family. One of these families which plays a crucial
role in our work, is the family of modified Jaskowski trees defined inductively as follows [8]: J1 is a
single node and Jn+1 is obtained by putting n + 1 copies of Jn together and adding a single node
below them. These frames have the following property:
Lemma 3.19 Let k1 , . . . , kn! be all terminal nodes of an EBPC model with frame Jn and suppose
that for all 1 ≤ i 6= j ≤ n there is a proposition Aij such that ki
Aij and kj 1 Aij . Then for each
0 0
k ∈ Jn there is a sentence Ak such that k
Ak iff k k .
V
Proof. For all i with 1 ≤ i ≤ n!, let Ai be j6=i ¬Aij , then ki
Ai and for all j 6= i, ki
¬Aj .
V 0 0
Define Ak to be ki k ¬Ai . Now, if k k , then there is a terminal node ki such that k ki but
0
k ki , so k 1 Ak . a
Now we have all necessary tools for our main theorem for EBA:
Theorem 3.20 Let A(p1 , . . . , pn ) be a proposition whose atoms are among p1 , . . . , pn . If EBPC 0
A(p1 , . . . , pn ) then there are sentences B1 , . . . , Bn (in the language of EBA) such that EBA 0
A(B1 , . . . , Bn ).
Proof. Since EBPC 0 A, there is a Kripke model of EBPC whose frame is Jm for some natural
number m. Let {A1 , . . . Am! } be an independent set over PA and take ωi to be a model of PA +
V
j6=i Aj + ¬Ai . Let K be a model of EBA whose frame is Jm in which its terminal nodes are ωi ’s
and other nodes are obtained by taking (suitable) Smorynski operations. Then, for each pair i, j
with 1 ≤ i 6= j ≤ m!, there is a sentence Aij such that ki
Aij and kj 1 Aij . Therefore, for each
0 0
node k there is a sentence Ak such that k
Ak if and only ifWk k.
For an atom pi , let Si := {k : k
pi } and define Bi := k∈Si Ak . Then, Si = {k : k
Bi }.
To complete our proof, it suffices to show that k
C(p1 , . . . , pn ) if and only if k
C(B1 , . . . , Bn )
for any proposition C whose atoms are among p1 , . . . , pn . The case for atomic C is obviously true,
since we made Bi ’s in this way, and cases for ∧ and ∨ are immediate consequences of the Induction
hypothesis. Let C = C1 → C2 .
0 0 0
If k
C(p1 , . . . , pn ) and k ≺ k and k
C1 (B1 , . . . , Bn ), then using the hypothesis k
0
C1 (p1 , . . . , pn ), so k
C2 (p1 , . . . , pn ) which, using the Induction hypothesis once more, shows that
0
k
C2 (B1 , . . . , Bn ), therefore k
C(B1 , . . . , Bn ). The same argument (with simultaneous change
of p1 , . . . , pn and B1 , . . . , Bn ) proves the converse. a
4 Concluding remarks
1. It is shown in [6] that the class of provably total functions of BA are exactly primitive recursive
functions in the following sense: if BA ` ∃yA(x, y), for quantifier free formula A, then there
is a primitive recursive function f such that N |= ∀xA(x, f (x)). It is not known if the class of
primitive functions are definable in BA.
View publication stats
2. Another natural question is if Theorem 3.20 holds for BA and BPC.
Acknowledgment. This research is partially supported by a grant from the Research Consul of
Sharif University of Technology.
References
[1] M. Ardeshir and W. Ruitenburg. Basic Propositional Calculus, I, Math. Log. Quart. 44(1998)
317-344.
[2] D. van Dalen, H. Mulder, E. C. W. Krabbe, and A. Visser. Finite Kripke models of HA are
locally PA, Notre Dame Journal of Formal Logic 27 (1986), 528-532.
[3] R. Kaye. Models of Peano Arithmetic, Clarendon Press. Oxford, 1991.
[4] W. Ruitenburg. Constructive logic and the paradoxes, Modern Logic, vol. 1 (1991), pp. 271-301.
[5] W. Ruitenburg. Basic Predicate Calculus, Notre Dame Journal of Formal Logic 39 (1998),
18-46.
[6] S. Salehi. Provably total functions of Basic Arithmetic, Mathematical Logic Quarterly, 49, No.
3, pp. 316-322, 2003.
[7] D. Scott. Identity and existence in intuitionistic logic, in Applications of Sheaves, Lectures
Notes in Mathematics, 753, edited by M. P. Fourman, C. J. Mulvey, and D. S. Scott, Springer-
Verlag, Berlin, 1979, pp. 660-96.
[8] C. Smorynski. Applications of Kripke models, in [9], 324-391.
[9] A Troelstra. Metamathematics of intuitionistic arithmetic and analysis, Lecture Notes
in Mathematics, Springer-Verlag, No. 344, 1973.
[10] A. S. Troelstra and D. van Dalen. Constructivism in Mathematics, An Introduction,
Vol. 1, North-Holland, 1988.