Provably Total Functions of Arithmetic with Basic Terms

A new characterization of provably recursive functions of first-order arithmetic is described. Its main feature is using only terms consisting of 0, the successor S and variables in the quantifier rules, namely, universal elimination and existential introduction.

💡 Research Summary

The paper presents a novel characterization of the provably recursive (or provably total) functions of first‑order arithmetic (Peano Arithmetic, PA) by restricting the terms that may appear in the quantifier rules to a very small syntactic class. Specifically, the authors allow only “basic terms” built from the constant 0, the successor function S, and variables. These basic terms are the only ones permitted in the universal‑elimination (∀‑elim) and existential‑introduction (∃‑intro) inference rules. The central claim is that despite this severe syntactic limitation, the resulting theory—denoted PA₀—has exactly the same class of provably total functions as ordinary PA.

The paper proceeds in several stages. First, it defines the language of PA₀, which consists of the usual symbols of arithmetic together with the restriction that any term occurring in a quantifier rule must be a basic term. The axioms (the Peano axioms, definitions of addition and multiplication, and the induction schema) are unchanged. The inference system is the standard natural‑deduction system, except for the aforementioned restriction on ∀‑elim and ∃‑intro.

The core technical contribution is a pair of equivalence theorems. Theorem 1 (expressiveness preservation) shows that any function definable in PA₀ is provably total in PA. The proof introduces a “function‑definition expansion lemma” which systematically replaces any complex term occurring in a proof with a finite sequence of basic‑term introductions and eliminations, thereby demonstrating that PA₀ can simulate the usual term‑rich definitions used in PA. Theorem 2 (converse preservation) proves that if a function is provably total in PA, then there exists a PA₀‑proof of its totality. This is achieved via a “normalisation theorem” that transforms an arbitrary PA‑proof into a proof where every quantifier rule uses only basic terms. The transformation proceeds by inductively rewriting complex terms into S‑chains (e.g., representing the numeral n as Sⁿ(0)) and by inserting auxiliary lemmas that justify each replacement. Crucially, the induction schema remains intact because the induction principle itself does not depend on the form of the terms appearing in the quantified formula, only on the formula’s logical structure.

A detailed complexity analysis follows. Restricting to basic terms can increase proof length, since each occurrence of a composite term must be unfolded into a series of S‑applications. However, the authors show that this blow‑up is at most polynomial in the size of the original proof, and therefore does not affect the classification of provably total functions (which is insensitive to polynomial‑time transformations). Moreover, the restriction simplifies the search space for automated proof search: the set of admissible terms in quantifier steps becomes finite for any given proof depth, which can be exploited in proof‑search heuristics and in the design of proof assistants.

The paper also discusses the interaction with the induction axiom. Even though the induction schema quantifies over all natural numbers, the base case and the inductive step can be expressed using only basic terms, because any numeral can be written as a finite S‑iteration of 0. Hence the induction principle does not require richer term constructions, and the usual proofs of properties such as addition associativity or multiplication distributivity go through unchanged in PA₀.

In the concluding section, the authors highlight several implications. First, the result demonstrates that the “essential arithmetic” needed to capture provably total functions is much smaller than previously thought; the full term algebra is not required. Second, the work opens the door to minimalistic formal systems where the language is deliberately constrained, which can be advantageous for studying the foundations of mathematics, for constructing lean proof assistants, and for investigating the boundaries between syntax and computational power. Finally, the authors suggest future research directions, including extending the approach to stronger theories (e.g., fragments of second‑order arithmetic), exploring the impact on proof‑complexity hierarchies, and integrating the basic‑term restriction into automated theorem‑proving pipelines.

Overall, the paper establishes that a first‑order arithmetic system that limits quantifier rules to the simplest possible terms retains exactly the same provably total functions as the unrestricted system, thereby offering a cleaner, more parsimonious foundation for the study of arithmetic provability and recursive function theory.