On modal logic of deductive closure (Q2498907)

Let \(T\) be a fixed logical theory. For any set of formulas \(\Sigma\) in the language of \(T\), the deductive closure of a subset \(A\subseteq\Sigma\) with respect to (provability in) \(T\) in (the universe) \(\Sigma\) is defined to be the set \(\{\alpha\in\Sigma\mid A\vdash_T\alpha\}\). The paper under review studies the modal logic of this deductive closure, i.e., when propositional variables are interpreted as arbitrary subsets of \(\Sigma\) (and \(\bot\) is interpreted as \(\emptyset\)), Boolean connectives as usual Boolean operations on sets, and the BOX operator \textbf{B} as the above deductive closure (when \(\phi\) is interpreted as \(A\), then \textbf{B}\(\phi\) is interpreted as \(\{\alpha\in\Sigma\mid A\vdash_T\alpha\}\)). The (modal) logic of deductive closure of a theory \(T\) is, by definition, the set of (modal) formulas and rules which are valid for every set of formulas \(\Sigma\) and every interpretation. When \(T\) is just the classical propositional logic, the logic of deductive closure of \(T\) (called \({\mathcal D}\) in the paper) is axiomatized by the reflexivity (\(\phi\rightarrow{\mathbf B}\phi)\) and transitivity \(({\mathbf B}(\phi\vee{\mathbf B}\phi)\rightarrow{\mathbf B}\phi)\) axioms, and the monotonicity rule (if \(\phi\rightarrow\psi\) is provable, then so is \textbf{B}\(\phi\rightarrow\) \textbf{B}\(\psi\)). So, we see that \({\mathcal D}\) is so-called non-normal, i.e., the axiom \(\text{ K}\): \textbf{B}\((\phi\rightarrow\psi)\rightarrow (\)\textbf{B}\(\phi\rightarrow\) \textbf{B}\(\psi)\) is not valid in it. Kripke-style models are introduced for \({\mathcal D}\), and their soundness and completeness proved; as a corollary, \({\mathcal D}\) is decidable. It is noted that the same arguments can be carried out when \(T\) is taken to be the classical predicate logic. Characterizing the modal logic of deductive closure of other logical systems remains an open problem.