- •Contents
- •Preface to the Fifth Edition
- •1 Enumerability
- •1.1 Enumerability
- •1.2 Enumerable Sets
- •Problems
- •2 Diagonalization
- •Problems
- •3 Turing Computability
- •Problems
- •4 Uncomputability
- •4.1 The Halting Problem
- •4.2* The Productivity Function
- •Problems
- •5 Abacus Computability
- •5.1 Abacus Machines
- •5.2 Simulating Abacus Machines by Turing Machines
- •5.3 The Scope of Abacus Computability
- •Problems
- •6 Recursive Functions
- •6.1 Primitive Recursive Functions
- •6.2 Minimization
- •Problems
- •7 Recursive Sets and Relations
- •7.1 Recursive Relations
- •7.2 Semirecursive Relations
- •7.3* Further Examples
- •Problems
- •8.1 Coding Turing Computations
- •8.2 Universal Turing Machines
- •8.3∗ Recursively Enumerable Sets
- •Problems
- •9.1 First-Order Logic
- •9.2 Syntax
- •Problems
- •10.1 Semantics
- •10.2 Metalogical Notions
- •Problems
- •11 The Undecidability of First-Order Logic
- •11.1 Logic and Turing Machines
- •11.2 Logic and Primitive Recursive Functions
- •11.3 Lemma
- •Problems
- •12 Models
- •12.1 The Size and Number of Models
- •12.2 Equivalence Relations
- •Problems
- •13 The Existence of Models
- •13.1 Outline of the Proof
- •13.2 The First Stage of the Proof
- •13.3 The Second Stage of the Proof
- •13.4 The Third Stage of the Proof
- •13.5* Nonenumerable Languages
- •Problems
- •14 Proofs and Completeness
- •14.1 Sequent Calculus
- •14.2 Soundness and Completeness
- •14.3* Other Proof Procedures and Hilbert’s Thesis
- •Problems
- •15 Arithmetization
- •15.1 Arithmetization of Syntax
- •Problems
- •16 Representability of Recursive Functions
- •16.2 Minimal Arithmetic and Representability
- •16.3 Mathematical Induction
- •16.4* Robinson Arithmetic
- •Problems
- •17.1 The Diagonal Lemma and the Limitative Theorems
- •17.2 Undecidable Sentences
- •17.3* Undecidable Sentences without the Diagonal Lemma
- •Problems
- •18 The Unprovability of Consistency
- •Historical Remarks
- •19 Normal Forms
- •19.1 Disjunctive and Prenex Normal Forms
- •19.2 Skolem Normal Form
- •19.3 Herbrand’s Theorem
- •19.4 Eliminating Function Symbols and Identity
- •Problems
- •20 The Craig Interpolation Theorem
- •20.1 Craig’s Theorem and Its Proof
- •20.2 Robinson’s Joint Consistency Theorem
- •20.3 Beth’s Definability Theorem
- •Problems
- •21 Monadic and Dyadic Logic
- •21.1 Solvable and Unsolvable Decision Problems
- •21.2 Monadic Logic
- •21.3 Dyadic Logic
- •Problems
- •22 Second-Order Logic
- •Problems
- •23.2 Arithmetical Definability and Forcing
- •Problems
- •24 Decidability of Arithmetic without Multiplication
- •Problems
- •25 Nonstandard Models
- •25.1 Order in Nonstandard Models
- •25.2 Operations in Nonstandard Models
- •25.3 Nonstandard Models of Analysis
- •Problems
- •26 Ramsey’s Theorem
- •Problems
- •27 Modal Logic and Provability
- •27.1 Modal Logic
- •27.2 The Logic of Provability
- •27.3 The Fixed Point and Normal Form Theorems
- •Problems
- •Annotated Bibliography
- •General Reference Works
- •Textbooks and Monographs
- •By the Authors
- •Index
24
Decidability of Arithmetic without Multiplication
Arithmetic is not decidable: the set V of code numbers of sentences of the language L of arithmetic that are true in the standard interpretation is not recursive (nor even arithmetical). But for some sublanguages L* of L, if we consider the elements of V that are code numbers of sentences of L*, then the set V* of such elements is recursive: arithmetic without some of the symbols of its language is decidable. A striking case is Presburger arithmetic, or arithmetic without multiplication. The present chapter is entirely devoted to proving its decidability.
We have used (true) arithmetic to mean the set of sentences of the language of arithmetic L = {0, < , , +, ·} that are true in the standard interpretation N. By arithmetic without multiplication we mean the set of sentences of (true) arithmetic that do not contain the symbol ·. By arithmetic without addition we mean the set of sentences of (true) arithmetic that do not contain the symbols < , , +. In contrast to the undecidability of arithmetic stand Presburger’s theorem, to the effect that arithmetic without multiplication is decidable, and Skolem’s theorem, to the effect that arithmetic without addition is decidable. [Note in connection with the latter theorem that is easily definable in terms of < and that + is definable in terms of and ·, as follows:
x + y = z ↔ (x · z ) · (y · z ) = ((x · y ) · (z · z )) .
That is why < and have to be dropped along with +.] This chapter will be entirely devoted to proving the former theorem, by describing an effective procedure for determining whether or not a given sentence of the language of arithmetic not involving · is true in the standard interpretation.
We begin with a reduction of the problem. Let K be the language with constants 0 and 1, infinitely many one-place predicates D2, D3, D4, . . . , the two-place predicate < , and the two-place function symbols + and ---. Let M be the interpretation with domain the set of all integers (positive, zero, negative), and with the following denotations for the nonlogical symbols. 0, 1, < , +, --- will denote the usual zero and unity elements, order relation, and addition and subtraction operations on integers. Dn will denote the set of integers divisible without remainder by n.
295
296 DECIDABILITY OF ARITHMETIC WITHOUT MULTIPLICATION
Given a sentence S of L without ·, replace everywhere in it by +1, and replace every quantification x or x by a relativized quantification
x((x = 0 0 <x) → · · ·) x((x = 0 0 <x) & . . .).
to obtain a sentence S* of K . Then S will be true in N if and only if S* is true in M. Thus to prove Presburger’s theorem, it will be sufficient to describe an effective procedure for determining whether or not a given sentence of K is true in M.
For the remainder of this chapter, therefore, term or formula or sentence will always mean term or formula or sentence of K, while denotation or satisfaction or truth will always mean denotation or satisfaction or truth in M. We call two terms r and s coextensive if v1 . . . vn r = s is true, where the vi are all the variables occurring in r or s. We call two formulas F and G coextensive if v1. . . vn (F ↔ G) is true, where the vi are all the free variables occurring in F or G.
Given any closed term, we can effectively calculate its denotation. Given any atomic sentence, we can effectively determine its truth value; and we can therefore do the same for any quantifier-free sentence. We are going to show how one can effectively decide whether a given sentence S is true by showing how one can effectively associate to S a coextensive quantifier-free sentence T : once T is found, its truth value, which is also the truth value of S, can be effectively determined.
The method to be used for finding T , given S, is called elimination of quantifiers. It consists in showing how one can effectively associate to a quantifier-free formula F(x), which may contain other free variables besides x, and quantifier-free G such that G is coextensive with x F(x) and G contains no additional free variables beyond the free variables in x F(x). This shown, given S, we put it in prenex form, then replace each quantification x by x , and work from the inside outward, successively replacing existential quantifications of quantifier-free formulas by coextensive quantifier-free formulas with no additional free variables, until at last a sentence with no free variables, which is to say, a quantifier-free sentence T , is obtained.
So let F(x) be a quantifier-free formula. We obtain G, coextensive with x F(x) and containing no addition free variables beyond those in x F(x), by performing, in order, a sequence of 30 operations, each of which replaces a formula by a coextensive formula with no additional free variables.
In describing the operations to be gone through, we make use of certain notational conventions. When writing of a positive integer k and a term t we allow ourselves to write
−t |
instead of |
0 |
--- t |
k |
instead of |
1 |
+ 1 + · · · + 1 (k times) |
kt |
instead of |
t + t + · · · + t (k times) |
for instance. With such notation, the 30 operations are as follows:
(1)Put F into disjunctive normal form. (See section 19.1.) Thus we get a disjunction of conjunctions of atomic formulas of the forms r = s or r <s or Dm s (where r and s are terms) and negations of such.
DECIDABILITY OF ARITHMETIC WITHOUT MULTIPLICATION |
297 |
(2)Replace each formula of form r = s by (r < s + 1 & s < r + 1).
(3)Replace each formula of form r = s by (r < s s < r).
(4)Put the result back into disjunctive normal form.
(5)Replace each formula of form r < s by s < r + 1.
(6)Replace each formula of form Dm s by the disjunction of Dm (s + i) for all i with 0 < i < m. The result is coextensive with the original, because for any number
a, m divides exactly one of a, a + 1, a + 2, . . . , a + m − 1.
(7)Put the result back into disjunctive normal form.
(8)At this point we have a disjunction of conjunctions of atomic formulas of the forms r < s and Dm s. Replace each formula of form r < s by 0 < (s --- r).
(9)We say a term is in normal form if it has one of the five forms
kx, --- kx, kx + t, --- kx + t, or t, wherein t is a term not containing the variable x. For every term one can effectively find a coextensive term in normal form by ordinary algebraic operations, such as regrouping and reordering summands. Replace each term in the formula that is not in normal form by a coextensive one that is.
(10)Replace each formula of form
0 < ---kx, |
0 < kx + t, or 0 < ---kx + t |
by |
|
kx < 0, |
---t < kx, or kx < t |
as the case may be.
(11)At this point all atomic formulas with predicate < that contain the variable x have either the form t < kx or the form kx < t, where k is positive and t does not contain x. We call those of the former form lower inequalities and those of the latter form upper inequalties. Rearrange the order of conjuncts in each disjunct so that all lower inequalities occur on the left.
(12)Towards reducing the number of lower inequalities occurring in any disjunct, if a conjunction of form t1 < k1 x & t2 < k2 x occurs in a disjunct, replace it by the disjunction of the following three conjunctions:
(i)t1 < k1 x & k1t2 < k2t1
(ii)t1 < k1 x & k1t2 = k2t1
(iii)t2 < k2 x & k2t1 < k1t2.
To see that this substitution is justified (that is, to see that it produces a result coextensive with the original), note that exactly one of the second conjuncts in
(i)–(iii) must hold, and that (i) or (ii) holds, then so do k2t1 < k1k2 x and
k1t2 < k1k2 x, and hence so does t2 < k2 x; while similarly (iii) yields t1 < k1 x.
(13)Eliminate any occurrences of = introduced at the previous step by repeating steps
(2) and (4).
(14)The effect of the preceding three steps is to reduce by one the number of lower inequalities in any disjunct where there were more than one to begin with. (Note that k1t2 < k2t1, for instance, does not count as a lower inequality, since it does not
298 DECIDABILITY OF ARITHMETIC WITHOUT MULTIPLICATION
contain the variable x.) Repeat these three steps over and over until no disjunct has more than one lower inequality among its conjuncts.
(15)Carry out an analogous process for upper inequalities, until no disjunct has more than one lower or upper inequality among its conjuncts.
(16)Replace each formula of form
Dm (kx), |
Dm (---kx), |
Dm (kx + t), or Dm (---kx + t) |
by |
|
|
Dm (kx --- 0), |
Dm (kx --- 0), |
Dm (kx --- (---t)), or Dm (kx --- t) |
as the case may be. This step is justified because for any number a, m divides a if and only if m divides −a.
(17)At this point all atomic formulas with Dm and involving x have the form
Dm (kx --- t), where k is a positive integer. Replace any formula of this form by the disjunction of all conjunctions
Dm (kx --- i) & Dm (t --- i)
for 0 ≤ i < m. To see that this step is justified, note that m divides the difference of two numbers a and b if and only if a and b leave the same remainder on division by m, and that the remainder on dividing a (respectively, b) by m is the unique i with 0 ≤ i < m such that m divides a − i (respectively, b − i).
(18)Put the result back into disjunctive normal form.
(19)At this point all atomic formulas with Dm and involving x have the form
Dm (kx --- i), where k is a positive integer and 0 ≤ i < m. Replace any formula of this form with k >1 by the disjunction of the formulas Dm (x --- j) for all j with 0 ≤ j < m such that m divides k j − i. This step is justified because for any
number a, ka leaves a remainder of i on division by m if and only if kj does, where j is the remainder on dividing a by m.
(20)Put the result back into disjunctive normal form.
(21)At this point all atomic formulas with Dm and involving x have the form
Dm (x --- i), where i is a nonnegative integer. In any such case consider the prime decomposition of m; that is, write
m = p1e1 · · · pkek where p1 < p2 < · · · < pk and all ps are primes. If k >1, then let m1 = p1e1 , . . . , mk = pkek , and replace Dm (x --- i) by
Dm1 (x --- i) & . . . & Dmk (x --- i).
This step is justified because the product of two given numbers having no common factor (such as powers of distinct primes) divides a given number if and only if each of the two given numbers does.
(22)At this point all atomic formulas with Ds and involving x have the form
Dm (x − i), where i is a nonnegative integer, and m a power of a prime. If in a given disjunct there are two conjuncts Dm (x − i) and Dn (x − j) where m and n are powers of the same prime, say m = pd , n = pe, d ≤ e, then drop Dm (x --- i) in favor of Dn (i − j), which does not involve x. This step is justified because, since
DECIDABILITY OF ARITHMETIC WITHOUT MULTIPLICATION |
299 |
m divides n, for any number a, if a leaves remainder j on division by n, a will leave remainder i on division by m if and only if j does.
(23)Repeat the preceding step until for any two conjuncts Dm (x --- i) and Dn (x --- j) in a single disjunct, m and n are powers of distinct primes, and therefore have no common factors.
(24)Replace each Dm (x --- i) by Dm (x --- i*), where i* is the remainder on dividing i by m.
(25)Rewrite each disjunct so that all atomic formulas with with Ds and involving x are on the left.
(26)At this point each disjunct has the form
Dm1 (x --- i1) & . . . & Dmk (x --- ik ) & (other conjuncts)
where 0 ≤ i1 < m1, . . . , 0 ≤ ik < mk . Let m = m1 · · · · · mk . According to the Chinese remainder theorem (see Lemma 15.5), there exists a (unique) i with 0 ≤ i < m such that i leaves remainder i1 on division by m1, . . . , i leaves
remainder ik on division by mk . Replace the conjuncts involving Ds by the single formula Dm (x --- i).
(27)At this point we have a disjunction F1 · · · Fk each of whose disjuncts is a conjunction containing at most one lower inequality, at most one upper inequality, and at most one formula of form Dm (x − i). Rewrite x(F1 · · · Fk ) as
x F1 · · · x Fk .
(28)Within each disjunct xF, rewrite the conjunction F so that any and all conjuncts involving x occur on the left, and confine the quantifier to these conjuncts, of which there are at most three; if there are none, simply omit the quantifier.
(29)At this point, the only occurrences of x are in sentences of one of the seven types listed in Table 24-1. Replace these by the sentences listed on the right.
Table 24-1. Elimination of quantifiers
x s < j x |
0 < 1 |
x kx < t |
0 < 1 |
x Dm (x --- i) |
0 < 1 |
x(Dm (x --- i) & s < j x) |
0 < 1 |
x(Dm (x --- i) & kx < t) |
0 < 1 |
x(s < j x & kx < t) |
x(D jk (x --- 0) & ks < x & x < jt) |
x(Dm (x --- i) & s < j x & kx < t) x(D jkm (x --- jki) & ks < x & x < jt) |
|
|
|
This step is justified in the first five cases because in these cases the sentence on the right is automatically true. (In the fourth case this is because there are arbitrarily large integers leaving a prescribed remainder i on division by m, and similarly in the fifth case.) The sixth and seventh cases are similar to each other. We discuss the latter because it is slightly more complicated. First note that the sentence on the left,
(i) |
x(Dm (x − i) & s < j x & kx <t) |
is coextensive with |
|
(ii) |
x(D jkm ( jkx − jki) & ks < jkx & jkx < jt). |