Topological Semantics

Origin. Tarski and McKinsey (1944). Modal logic via topology. Interior operator = necessity. Open sets = possible. Foundation for spatial modal logic.

Models. Topological spaces. Interior for □, closure for ◇. Opens give truth. S4 = all topological spaces.

Formalism.

Topological space: (X, τ) where τ = open sets. Closed under ∪, finite ∩. ∅, X ∈ τ.

Interior operator: int(A) = largest open subset of A. □φ true at x iff x ∈ int(⟦φ⟧).

Closure operator: cl(A) = smallest closed superset. ◇φ true at x iff x ∈ cl(⟦φ⟧).

Kuratowski axioms: int(A ∩ B) = int(A) ∩ int(B) int(A) ⊆ A int(int(A)) = int(A) int(X) = X

S4 correspondence: □φ → φ (T): int(A) ⊆ A □φ → □□φ (4): int(int(A)) = int(A) S4 complete for topological semantics.

S5 and locally indiscrete spaces: S5 = the topologies in which every open set is closed. Equivalently the partition topologies; the indiscrete one is the extreme case. Then □φ is a union of blocks, and 5 holds. The discrete topology, where every set is clopen, collapses □ entirely: □φ ↔ φ.

Alexandroff spaces: Every intersection of opens is open. Corresponds to preorders. Kripke frames special case.

Derived set: d(A) = limit points of A. Modal logic of derivative. Different from interior.

Symbols.

Symbol	Unicode	Name	Meaning
int	—	Interior	Largest open
cl	—	Closure	Smallest closed
τ	U+03C4	Topology	Open sets
⊆	U+2286	Subset	Containment
□	U+25A1	Box	Necessity, interpreted by int
◇	U+25C7	Diamond	Possibility, interpreted by cl
X	—	Space	Underlying point set
⟦φ⟧	—	Truth set	Points where φ holds
d	—	Derived set	Set of limit points of A
∪	U+222A	Union	Arbitrary unions of opens are open
∩	U+2229	Intersection	Finite intersections of opens are open
∅	U+2205	Empty set	Open in every topology
∈	U+2208	Membership	x ∈ int(⟦φ⟧) gives □φ at x
⟦	U+27E6	Semantic bracket, left	Opening of ⟦φ⟧
⟧	U+27E7	Semantic bracket, right	Closing of ⟦φ⟧
ℝ	U+211D	Real line	Validates exactly S4
ℚ	U+211A	Rationals	Validates exactly S4
→	U+2192	Material conditional	Form of the T and 4 axioms
↔	U+2194	Biconditional	Collapse □φ ↔ φ on discrete spaces
∧	U+2227	Conjunction	Antecedent of the wK4 axiom

Metatheory. McKinsey and Tarski (1944) proved S4 sound and complete for the interior semantics over all topological spaces, and complete already for any dense-in-itself metric space, so ℝ, ℚ and the Cantor set each validate exactly S4 and nothing more. Algebraically the same result says every closure algebra is representable as the closure algebra of a space, which is why S4 and interior algebras are two descriptions of one structure. Relational semantics is the special case of Alexandrov spaces, those closed under arbitrary intersections of opens, which correspond exactly to preorders; general topologies are strictly more permissive, yet the validities coincide, so S4 satisfiability remains PSPACE-complete (Ladner, 1977). Reading the modality by the derived set d rather than by interior gives a different logic: over all spaces d validates exactly wK4, that is K plus □φ ∧ φ → □□φ, with 4 itself valid only on T_D spaces (Esakia), and GL is the derivational logic of the scattered spaces.

Applies to. Spatial reasoning. Mereotopology. Point-free topology. Modal logic foundations.

Limitations. Under the interior reading T and 4 hold on every space, so only extensions of S4 have a topological semantics of this kind: K, D and B do not, and GL needs the quite different derived-set semantics on scattered spaces. Interior distributes over finite but not arbitrary intersection, so the match with Kripke frames holds only for Alexandrov spaces; elsewhere the box of an infinite conjunction is not the intersection of the boxes. Negative introspection collapses the whole structure — S5 forces every open set to be closed, so the space is a partition and the topology does no work, which makes S4 the strongest logic for which the semantics is informative. Interpretation is over points, so nothing is said directly about regions, and mereotopology requires a separate region-based apparatus rather than an extension of this one. The completeness proofs are for the propositional language; quantified topological modal logic is not covered by them.