A theorem on shortening the length of proof in formal systems of arithmetic

Abstract

This note is concerned with an aspect of the length of proof of formulas in recursively enumerable theories T adequate for recursive arithmetic. In particular, we consider the relative length of proof of formulas in the theories T and T(S), where F represents an r.e. set A in T and T(S) is the theory obtained from T by adjunction, as a new axiom, of a sentence S undecidable in T.Throughout the sequel T is a consistent, r.e. theory with standard formalization [7] in which all recursive functions of one variable are definable, and in which there is a binary formula x ≤ satisfying the well-known conditions [7]:Here is the constant term corresponding to the natural number n. Wn is the nth r.e. set in a standard enumeration of the r.e. sets. Also, we assume an a priori Gödel numbering of our formalism satisfying the usual conditions, so that all formulas are numbers ab initio.In the more common applications of the theorem below, if F is a k-ary formula of T, is a natural number that measures in some way the length of the shortest proof of in T.