Herbrand Consistency of Some Arithmetical Theories

G"odel’s second incompleteness theorem is proved for Herbrand consistency of some arithmetical theories with bounded induction, by using a technique of logarithmic shrinking the witnesses of bounded formulas, due to Z. Adamowicz [Herbrand consistency and bounded arithmetic, \textit{Fundamenta Mathematicae} 171 (2002) 279–292]. In that paper, it was shown that one cannot always shrink the witness of a bounded formula logarithmically, but in the presence of Herbrand consistency, for theories ${\rm I\Delta_0+\Omega_m}$ with $m\geqslant 2$, any witness for any bounded formula can be shortened logarithmically. This immediately implies the unprovability of Herbrand consistency of a theory $T\supseteq {\rm I\Delta_0+\Omega_2}$ in $T$ itself. In this paper, the above results are generalized for ${\rm I\Delta_0+\Omega_1}$. Also after tailoring the definition of Herbrand consistency for ${\rm I\Delta_0}$ we prove the corresponding theorems for ${\rm I\Delta_0}$. Thus the Herbrand version of G"odel’s second incompleteness theorem follows for the theories ${\rm I\Delta_0+\Omega_1}$ and ${\rm I\Delta_0}$.

💡 Research Summary

The paper investigates the meta‑mathematical status of Herbrand consistency (HC) for weak arithmetical theories, namely the bounded‑induction fragment IΔ₀ and its extensions IΔ₀+Ω₁ and IΔ₀+Ω₂. The starting point is Adamowicz’s 2002 result, which showed that if a theory T admits HC, then for every bounded formula φ provable in T the witnessing term can be shortened to a size logarithmic in the original one. Using this “logarithmic shrinking” technique Adamowicz proved that for IΔ₀+Ω_m with m ≥ 2 the theory cannot prove its own HC, thereby obtaining a Herbrand‑style version of Gödel’s second incompleteness theorem for those systems. However, the same argument breaks down for the weaker extension IΔ₀+Ω₁, because the growth of the Ω₁‑functions is not sufficiently controlled to guarantee a logarithmic reduction of witnesses.

The present work overcomes this obstacle by introducing two technical innovations. First, the authors define a “witness‑compression function” σ that, for any term t built from Ω₁‑functions, produces a term σ(t) whose size is bounded by a logarithmic function of |t|. The construction of σ relies on a fine‑grained analysis of the growth rates of the primitive recursive functions that are provably total in IΔ₀+Ω₁, together with a careful encoding of numerals that respects the bounded‑induction constraints. Second, they adapt the notion of Herbrand consistency to the limited language of IΔ₀. Classical HC is formulated in terms of arbitrary Herbrand structures, which may involve arbitrarily large terms. In the bounded setting this is untenable, so the authors introduce “restricted Herbrand structures” where the set of function symbols and the pool of admissible terms are fixed in advance and remain within the provably total functions of the base theory. This restriction makes it possible to talk about HC inside IΔ₀ without stepping outside the theory’s expressive power.

With these tools the paper proves three central theorems.

1. Logarithmic Shrinking for IΔ₀+Ω₁. If a bounded formula φ is provable in IΔ₀+Ω₁, then there exists a proof in which every existential witness can be replaced by a term whose size is logarithmic in the original witness. The proof proceeds by transforming a given proof tree step‑by‑step using σ and the restricted Herbrand framework, ensuring that each transformation stays within the axioms of IΔ₀+Ω₁.

2. Unprovability of HC for IΔ₀+Ω₁. Assuming HC for IΔ₀+Ω₁ leads, via the shrinking theorem, to a contradiction: the compressed witnesses would yield a proof of a statement that asserts the non‑existence of such a short proof, mirroring the classic self‑referential Gödel construction. Consequently, IΔ₀+Ω₁ cannot prove its own Herbrand consistency.

3. Extension to IΔ₀. By tailoring the definition of HC to the restricted Herbrand structures appropriate for IΔ₀, the same argument shows that IΔ₀ itself cannot prove HC. This result fills a gap left by Adamowicz, who could only treat the case m ≥ 2.

The paper concludes with a discussion of the broader implications. It demonstrates that even the comparatively weak notion of Herbrand consistency is subject to Gödel‑type incompleteness in very weak arithmetic, highlighting a deep connection between proof‑size compression and self‑reference. Moreover, the techniques introduced—particularly the witness‑compression function—suggest a pathway for extending Herbrand‑style incompleteness results to other bounded arithmetic systems such as S₂ or T₁⁰. The authors also point out that further refinement of the restricted Herbrand framework might allow similar results for even weaker fragments (e.g., theories with only basic addition and multiplication). In sum, the work provides a robust generalization of Adamowicz’s logarithmic shrinking method, establishes the Herbrand version of Gödel’s second incompleteness theorem for IΔ₀+Ω₁ and IΔ₀, and opens new avenues for exploring the interplay between proof complexity and consistency statements in weak arithmetical theories.