S4.2: Modal Logic Convergence & Forcing

• S4.2 is a modal logic that extends S4 with the convergence axiom, ensuring reconvergence in reflexive, transitive, and upward directed frames.
• It plays a central role in set-theoretic forcing and potentialism, linking abstract modal principles with concrete semantic models.
• Recent studies highlight its finite frame properties, complexity in small-variable fragments, and algebraic significance via weak excluded middle laws.

In modal logic, $\mathbf{S4.2}$ is a normal modal logic obtained by extending $\mathbf{S4}$ with the axiom of convergence, usually written as $\Diamond \Box p \to \Box \Diamond p$, and in closely related presentations as $\Box \Diamond p \to \Diamond \Box p$ over the relevant frame classes (Ya'ar, 2017). It is characterized semantically by reflexive, transitive, upward directed frames and, more sharply, by finite pre-Boolean algebras, and it recurs as the exact modal logic of forcing, of several potentialist systems, and of multiple newer semantics based on embeddability and homomorphisms (Kurahashi et al., 2023).

1. Axiomatic core and semantic characterizations

The Hilbert presentation of $\mathbf{S4.2}$ consists of propositional logic together with the normality axiom

$$\Box(p \rightarrow q)\rightarrow(\Box p\rightarrow \Box q),$$

the $\mathbf{S4}$ axioms

$$\Box p\rightarrow p,\qquad \Box p\rightarrow \Box\Box p,$$

and the additional “.2” axiom

$\Diamond\Box p\rightarrow \Box\Diamond p,$

closed under modus ponens and necessitation (Ya'ar, 2017). In bimodal forcing/provability presentations the same logic is given by propositional tautologies, $K$, $\mathbf{S4}$0, $\mathbf{S4}$1, and .2 for the forcing modality (Kurahashi et al., 2023).

Its Kripke semantics is described in several equivalent ways across the literature. One formulation characterizes $\mathbf{S4}$2-frames as those whose accessibility relation is reflexive, transitive, and upward directed (Kurahashi et al., 2023). Another uses finite pre-Boolean algebras: if one quotients a preorder by mutual accessibility, the resulting partial order is a Boolean algebra, and $\mathbf{S4}$3 is characterized by all finite frames of this kind (Ya'ar, 2017). In the multi-agent epistemic setting, the same logic is formalized over weakly-directed pre-orders, with the agent-indexed form of .2 written as

$\mathbf{S4}$4

and its soundness and completeness have been mechanized in Isabelle/HOL for countably many agents (Guzman et al., 2024).

The logic also has the finite frame property in the forcing-oriented presentation: a formula is provable in $\mathbf{S4}$5 iff it is valid in all finite rooted $\mathbf{S4}$6-frames, equivalently in all finite rooted pre-Boolean algebra frames (Kurahashi et al., 2023). This finite semantics underwrites most completeness and transfer arguments that identify $\mathbf{S4}$7 as the exact logic of a given modality.

2. Forcing, directedness, and potentialism

A central source of $\mathbf{S4}$8 is the modal logic of forcing. Interpreting $\mathbf{S4}$9 as “$\Diamond \Box p \to \Box \Diamond p$0 holds in all forcing extensions” and $\Diamond \Box p \to \Box \Diamond p$1 as “$\Diamond \Box p \to \Box \Diamond p$2 holds in some forcing extension,” Hamkins and Löwe showed that, assuming ZFC is consistent, the ZFC-provably valid principles of forcing are exactly $\Diamond \Box p \to \Box \Diamond p$3 (Kurahashi et al., 2023). The structural explanation is that forcing classes that are reflexive, transitive, and directed validate $\Diamond \Box p \to \Box \Diamond p$4, $\Diamond \Box p \to \Box \Diamond p$5, and .2, so $\Diamond \Box p \to \Box \Diamond p$6 is a general lower bound for such classes (Hamkins et al., 2012).

This perspective refines naturally to forcing subclasses. For $\Diamond \Box p \to \Box \Diamond p$7-centered forcing, the ZFC-provable principles are exactly $\Diamond \Box p \to \Box \Diamond p$8; the proof combines the lower bound from reflexivity, transitivity, and directedness with an upper bound obtained by realizing all finite pre-Boolean algebras using buttons, switches, $\Diamond \Box p \to \Box \Diamond p$9-switches, and ratchets (Ya'ar, 2017). The same paper and subsequent work show that countably closed forcing, $\Box \Diamond p \to \Diamond \Box p$0-closed forcing for absolutely definable regular $\Box \Diamond p \to \Diamond \Box p$1, CH-preserving forcing, $\Box \Diamond p \to \Diamond \Box p$2CH-preserving forcing, and GCH-preserving forcing likewise have exact modal logic $\Box \Diamond p \to \Diamond \Box p$3, whereas c.c.c. forcing, proper forcing, and related classes lie strictly below this threshold in the sense that their modal logics do not contain $\Box \Diamond p \to \Diamond \Box p$4 and are contained in $\Box \Diamond p \to \Diamond \Box p$5 (Hamkins et al., 2012).

Set-theoretic potentialism supplies another large family of $\Box \Diamond p \to \Diamond \Box p$6-phenomena. In transitive-set potentialism and in several countable-model systems, directedness of the accessibility relation yields $\Box \Diamond p \to \Diamond \Box p$7 as the lower bound, with buttons and dials showing optimality at suitable worlds (Hamkins et al., 2017). In a choiceless large-cardinal setting based on Berkeley cardinals, the global propositional modal assertions valid at every world are exactly those of $\Box \Diamond p \to \Diamond \Box p$8; here the accessibility relation is defined using preservation of elementary self-embeddings, and .2 is obtained from an amalgamation argument using Berkeley embeddings (Cutolo et al., 2020).

These results make $\Box \Diamond p \to \Diamond \Box p$9 the canonical modal logic of “directed potentiality”: the system is stronger than plain $\mathbf{S4.2}$0 because distinct possibilities can be merged, but weaker than $\mathbf{S4.2}$1 unless the world already satisfies a maximality principle.

3. Epistemic, bimodal, and bundled interpretations

In epistemic logic, $\mathbf{S4.2}$2 appears as the knowledge fragment of Stalnaker’s system for knowledge and belief. The fragment uses formulas built from propositional variables and agent-indexed operators $\mathbf{S4.2}$3, with .2 written as

$\mathbf{S4.2}$4

and it is sound and complete for weakly-directed pre-orders (Guzman et al., 2024). This gives an epistemic reading of convergence: if some world compatible with an agent’s knowledge makes $\mathbf{S4.2}$5 known, then the agent already knows that $\mathbf{S4.2}$6 is compatible with what is known.

A different but related role is played by $\mathbf{S4.2}$7 in the bimodal logic $\mathbf{S4.2}$8, which combines provability logic $\mathbf{S4.2}$9 with forcing. In that setting, the forcing modality is exactly $$\Box(p \rightarrow q)\rightarrow(\Box p\rightarrow \Box q),$$0, the provability modality is $$\Box(p \rightarrow q)\rightarrow(\Box p\rightarrow \Box q),$$1, and the interaction axioms connect the two without altering their unimodal fragments (Kurahashi et al., 2023). The resulting systems $$\Box(p \rightarrow q)\rightarrow(\Box p\rightarrow \Box q),$$2 and $$\Box(p \rightarrow q)\rightarrow(\Box p\rightarrow \Box q),$$3 are conservative over $$\Box(p \rightarrow q)\rightarrow(\Box p\rightarrow \Box q),$$4 in the forcing-only language, so the forcing component remains precisely $$\Box(p \rightarrow q)\rightarrow(\Box p\rightarrow \Box q),$$5 even inside the richer bimodal semantics (Kurahashi et al., 2023).

More recently, bundled-modal semantics has used $$\Box(p \rightarrow q)\rightarrow(\Box p\rightarrow \Box q),$$6 as a base frame condition for composite modalities. A case study axiomatizes the bundle

$$\Box(p \rightarrow q)\rightarrow(\Box p\rightarrow \Box q),$$7

over $$\Box(p \rightarrow q)\rightarrow(\Box p\rightarrow \Box q),$$8-models, interpreting the corresponding unary operator as $$\Box(p \rightarrow q)\rightarrow(\Box p\rightarrow \Box q),$$9, that is, belief without knowledge (Ding et al., 27 Mar 2026). The framework provides convex neighborhood semantics, a bundle-specific bisimulation, and a complete axiomatization over $\mathbf{S4}$0-frames, showing that convergence is compatible with richer unary modalities encoded directly in the semantics (Ding et al., 27 Mar 2026).

4. Algebraic significance and weak excluded middle

The algebraic role of $\mathbf{S4}$1 is isolated through the notion of a weak excluded middle law (WEML). For algebraizable deductive systems, the existence of a WEML implies that every relatively subdirectly irreducible member of the equivalent quasivariety has a greatest proper congruence, and the converse holds under an inconsistency lemma (Lávička et al., 2021). This makes WEML a structural strengthening of inconsistency lemmas rather than merely a syntactic schema.

Applied to normal modal logics, the decisive result is that if $\mathbf{S4}$2 is a normal extension of $\mathbf{S4}$3, then the global consequence relation $\mathbf{S4}$4 has a WEML iff

$\mathbf{S4}$5

that is, iff $\mathbf{S4}$6 extends $\mathbf{S4}$7 (Lávička et al., 2021). In this sense, $\mathbf{S4}$8 is exactly the threshold among $\mathbf{S4}$9-extensions at which global consequence acquires a weak excluded middle law.

The same paper places $$\Box p\rightarrow p,\qquad \Box p\rightarrow \Box\Box p,$$0 in direct analogy with $$\Box p\rightarrow p,\qquad \Box p\rightarrow \Box\Box p,$$1, the weak excluded middle extension of intuitionistic logic: a super-intuitionistic logic has a WEML iff it extends $$\Box p\rightarrow p,\qquad \Box p\rightarrow \Box\Box p,$$2, while a normal extension of $$\Box p\rightarrow p,\qquad \Box p\rightarrow \Box\Box p,$$3 has a global WEML iff it extends $$\Box p\rightarrow p,\qquad \Box p\rightarrow \Box\Box p,$$4 (Lávička et al., 2021). This parallel aligns frame-theoretic directedness, modal convergence, and algebraic greatest-proper-congruence behavior.

5. Finite-variable complexity

The small-variable complexity of $$\Box p\rightarrow p,\qquad \Box p\rightarrow \Box\Box p,$$5 has recently been sharpened. In the language of modal logics with the axiom of convergence, $$\Box p\rightarrow p,\qquad \Box p\rightarrow \Box\Box p,$$6 and $$\Box p\rightarrow p,\qquad \Box p\rightarrow \Box\Box p,$$7 are PSPACE-complete already in the fragment with two propositional variables, while $$\Box p\rightarrow p,\qquad \Box p\rightarrow \Box\Box p,$$8 and $$\Box p\rightarrow p,\qquad \Box p\rightarrow \Box\Box p,$$9 are PSPACE-complete in the one-variable fragment (Rybakov et al., 16 Jul 2025). The paper also proves more general hardness results for every modal logic between $\Diamond\Box p\rightarrow \Box\Diamond p,$0 and $\Diamond\Box p\rightarrow \Box\Diamond p,$1, and for every logic between $\Diamond\Box p\rightarrow \Box\Diamond p,$2 and $\Diamond\Box p\rightarrow \Box\Diamond p,$3 (Rybakov et al., 16 Jul 2025).

For $\Diamond\Box p\rightarrow \Box\Diamond p,$4, the convergence axiom therefore does not collapse complexity when the propositional vocabulary is small: even with only two variables, satisfiability remains PSPACE-complete (Rybakov et al., 16 Jul 2025). At the same time, the one-variable fragment of $\Diamond\Box p\rightarrow \Box\Diamond p,$5 remains open in that analysis, which distinguishes it from $\Diamond\Box p\rightarrow \Box\Diamond p,$6 and $\Diamond\Box p\rightarrow \Box\Diamond p,$7, where one variable already suffices for PSPACE-completeness (Rybakov et al., 16 Jul 2025).

6. New semantic domains and terminological variation

Recent work has shown that $\Diamond\Box p\rightarrow \Box\Diamond p,$8 is not confined to set-theoretic or epistemic semantics. In modal group theory with embeddability as accessibility, the formulaic propositional modal validities of groups under embeddings are precisely $\Diamond\Box p\rightarrow \Box\Diamond p,$9 (Wołoszyn, 13 May 2026). The lower bound comes from amalgamation by free products with amalgamation, which provides the directedness required for .2, and the upper bound comes from independent buttons and dials built from torsion and center-size assertions (Wołoszyn, 13 May 2026).

A related homomorphism semantics produces a different split. If possibility is interpreted by the existence of a homomorphism out of a group, then sentential validities are exactly $K$0, the trivial group has exact parameter-validities $K$1, and uniformly prime-indivisible groups have exact parameter-validities $K$2 (Wołoszyn, 14 May 2026). Here the .2 behavior is supplied by amalgamability over parameter images, while the failure of full $K$3 with parameters is witnessed by button-like collapse statements such as $K$4 (Wołoszyn, 14 May 2026).

The label “S4.2” also appears in unrelated literatures. In structured state space modeling, it is used informally to refer to later S4-style sequence models with safer parameterizations and initializations, notably S4D and related variants, but that usage is not a term defined explicitly in the original S4 paper (Gu et al., 2021). This suggests a terminological overlap rather than a conceptual connection: the modal-logic $K$5 and the sequence-model “S4.2” belong to distinct technical traditions (Gu et al., 2021).

Across these settings, the persistent theme is that $K$6 captures a middle level of modal strength: more structured than arbitrary reflexive-transitive possibility, because accessible branches can be reconverged, but not so saturated that every possible necessity is already actual.