Cut-Elimination and Proof Search for Bi-Intuitionistic Tense Logic

TL;DR

This paper introduces a new sequent calculus, LBiKt, for bi-intuitionistic tense logic, proves cut elimination, and develops a deep inference variant, DBiKt, to improve proof search, while establishing soundness and exploring extensions.

Contribution

It presents the LBiKt calculus for bi-intuitionistic tense logic, proves cut elimination, and introduces the deep inference calculus DBiKt for better proof search capabilities.

Findings

LBiKt is sound with respect to Kripke semantics.

Cut elimination is successfully established for LBiKt.

DBiKt enables more effective proof search through deep inference.

Abstract

We consider an extension of bi-intuitionistic logic with the traditional modalities from tense logic Kt. Proof theoretically, this extension is obtained simply by extending an existing sequent calculus for bi-intuitionistic logic with typical inference rules for the modalities used in display logics. As it turns out, the resulting calculus, LBiKt, seems to be more basic than most intuitionistic tense or modal logics considered in the literature, in particular, those studied by Ewald and Simpson, as it does not assume any a priori relationship between the diamond and the box modal operators. We recover Ewald's intuitionistic tense logic and Simpson's intuitionistic modal logic by modularly extending LBiKt with additional structural rules. The calculus LBiKt is formulated in a variant of display calculus, using a form of sequents called nested sequents. Cut elimination is proved for LBiKt, using a technique similar to that used in display calculi. As in display calculi, the inference rules of LBiKt are ``shallow'' rules, in the sense that they act on top-level formulae in a nested sequent. The calculus LBiKt is ill-suited for backward proof search due to the presence of certain structural rules called ``display postulates'' and the contraction rules on arbitrary structures. We show that these structural rules can be made redundant in another calculus, DBiKt, which uses deep inference, allowing one to apply inference rules at an arbitrary depth in a nested sequent. We prove the equivalence between LBiKt and DBiKt and outline a proof search strategy for DBiKt. We also give a Kripke semantics and prove that LBiKt is sound with respect to the semantics, but completeness is still an open problem. We then discuss various extensions of LBiKt.