A new reduction sequence for arithmetic

Abstract

The main type of result in proof theory having external applications is normalization theorems: any derivation can be reduced to normal form with the help of a finite number of standard transformations (reductions). An estimate for the convergence of the process of normalization gives an estimate for the growth of provable recursive functions, which recently found application to finitary combinatorics [J. Paris and L. Harrington, “A mathematical incompleteness in Peano arithmetic,” in: Handbook of Mathematical Logic (1978), pp. 1133–1142]. The definition of reduction for the predicate calculus and arithmetic is due to Gentzen, who proved a normalization theorem for derivations of atomic formulas and gave an estimate for convergence with the help of an assignment to derivations d of ordinals O(d)<ε0 such that O(d)>O(d) for reductions ¯d of the derivation d. Later convergence was proved for arbitrary derivations; however, the slight clarity in the choice of ¯d and the proof of decrease of ordinals impede understanding and exposition: both Takeuti (G. Takeuti, Proof Theory, North-Holland, 1975) and Schütte (Ref. Zh. Mat., 1978, 7A48K) restrict themselves to derivations of atomic formulas. In the presentpaper there is given a shorter and clearer construction of reduction ¯d and assignment of ordinals, allowing one also to simplify the proof of decrease of ordinals. The source of all the simplifications is the connection with the continuous operator of cut elimination from infinite derivations proposed by the author. In particular, the symbol Ok(d), in terms of which O(d) is defined, can be read as “ordinal height of the derivation, obtained after elimination of cuts of degree ≥ k from an infinite derivation in which d is represented in a standard way.” We consider the L-formulation, where there are rules of introduction of connectives in the antecedent and succedent. Extension to the natural formulation is obtained with the help of Prawitz translation.