Herbrand Consistency of Some Finite Fragments of Bounded Arithmetical Theories

We formalize the notion of Herbrand Consistency in an appropriate way for bounded arithmetics, and show the existence of a finite fragment of ${\rm I\Delta_0}$ whose Herbrand Consistency is not provable in the thoery ${\rm I\Delta_0}$. We also show the existence of an ${\rm I\Delta_0}-$derivable $\Pi_1-$sentence such that ${\rm I\Delta_0}$ cannot prove its Herbrand Consistency.

💡 Research Summary

The paper “Herbrand Consistency of Some Finite Fragments of Bounded Arithmetical Theories” investigates the notion of Herbrand consistency (HCon) within the weak arithmetic system IΔ₀, which is the fragment of Peano Arithmetic whose induction is restricted to Δ₀‑formulas. The author first revisits the classical Herbrand‑Skolem theorem and points out that the usual arithmetization of propositional satisfiability does not fit well with the limited resources of IΔ₀. To overcome this, a new, more effective definition of “evaluation” on a finite set of ground (Skolem) terms is introduced.

A pre‑evaluation is a sequence that orders the terms and inserts symbols “≈” (equality) and “≺” (order) between successive terms. From this ordering, two binary relations ≈ₚ and ≺ₚ are defined on the term set. The relation ≈ₚ is shown to be an equivalence relation, while ≺ₚ is a total order compatible with ≈ₚ. An evaluation is a pre‑evaluation whose equivalence relation is a congruence with respect to all function symbols; in other words, substitution of equal terms preserves equality. This construction yields a concrete, computable way to check whether a finite term set admits a T‑evaluation for a given theory T.

Using this machinery, the paper proves two main theorems.

1. Existence of a finite fragment T of IΔ₀ whose Herbrand consistency is not provable in IΔ₀.
   The fragment T consists of a small, finite collection of axioms of IΔ₀ (including basic arithmetic axioms and a few additional Δ₀‑definable schemata). By applying the evaluation framework, the author shows that any IΔ₀‑proof of HCon(T) would yield a uniform bound k such that every T‑evaluation on the k‑th Skolem hull of any term set must satisfy all instances of the axioms. A diagonalisation argument then produces a term set for which no such evaluation exists, contradicting the assumed provability. Consequently, IΔ₀ ⊬ HCon(T).

2. Existence of an IΔ₀‑derivable Π₁‑sentence U whose Herbrand consistency is not provable in IΔ₀.
   The sentence U is chosen to be a Π₁‑statement expressing a basic arithmetic fact (for example, “for every n, n·n is defined”). The theory S = IΔ₀ + U is considered, and its Skolemized version Sₛₖ is examined. Lemma 2.15 (the “Herbrand universal formula” lemma) guarantees that if S ⊢ ∀x ψ(x) then there is a finite k such that any S‑evaluation on the k‑th Skolem hull must satisfy ψ(t) for every term t. By constructing a term set where ψ(t) fails, the author shows that no S‑evaluation can exist, which means HCon(U) cannot be proved in IΔ₀.

These results extend earlier work by Adamowicz‑Zbierski, Salehi, Willard, and Kolodziejczyk, who proved various unprovability statements for stronger extensions of IΔ₀ (e.g., IΔ₀ + Ω₁, IΔ₀ + Exp). The present paper removes the need for auxiliary axioms such as Ω₀ (totality of the squaring function) and works directly with the base theory IΔ₀. Moreover, the new evaluation definition, which bounds the Gödel codes of terms, makes the construction effective and suitable for formalization inside IΔ₀ itself.

The paper concludes with several directions for future research: (a) investigating whether even smaller finite fragments admit the same unprovability phenomenon, (b) relating Herbrand consistency to other weakened consistency notions such as Cut‑Free Consistency, and (c) exploring algorithmic implementations of the evaluation framework for automated proof search in weak arithmetic. Overall, the work demonstrates that even the weakened notion of Herbrand consistency remains sufficiently strong to separate IΔ₀ from its finite extensions, providing a clear illustration of the meta‑mathematical limits of bounded arithmetic.