Jónsson–Tarski Duality

Origin. Bjarni Jónsson and Alfred Tarski, "Boolean algebras with operators, Parts I and II" (American Journal of Mathematics, 1951–52) — written before Kripke semantics existed, and containing it. Goldblatt's "Metamathematics of modal logic" (1976) and Thomason's work made the connection explicit; Blackburn, de Rijke, and Venema's Modal Logic (2001) is where the field learned to teach it.

Models. Stone duality plus one operator. A Boolean algebra with a normal additive operator is dual to a Stone space with a relation — which is to say, to a Kripke frame carrying a topology. The relation is recovered from the algebra as inclusion among ultrafilters modulo the operator; the algebra is recovered from the frame as its complex algebra. Modal logic's two semantics are one construction read in two directions, and Jónsson and Tarski had the theorem fifteen years before anyone read it that way.

Formalism.

Boolean algebra with operators (BAO): (B, ∧, ∨, ¬, 0, 1, f) with f normal (f(0) = 0) and additive (f(a ∨ b) = f(a) ∨ f(b)). The modal ◇ is exactly such an f.

Complex algebra of a frame: For a frame F = (W, R), take B = ℘(W) with ◇X = {w : ∃v (wRv and v ∈ X)} This is F⁺, the complex algebra. Additivity is automatic; normality is ◇∅ = ∅.

Ultrafilter frame of an algebra: For a BAO B, take W = ultrafilters, and U R V iff {◇a : a ∈ V} ⊆ U This is B₊, the ultrafilter (canonical) frame.

The canonical embedding: B ↪ (B₊)⁺ Every BAO embeds in the complex algebra of its ultrafilter frame — the Jónsson–Tarski theorem. The image is not all of it: (B₊)⁺ is the canonical extension, and it is strictly larger.

Descriptive general frames: Frames alone are not dual to algebras: ℘(W) is too big. A general frame (W, R, A) carries an admissible set algebra A ⊆ ℘(W). Descriptive = differentiated, tight, and compact. Then: BAO ≃ DGF^op, an equivalence.

Why canonical models work: The canonical model of a normal modal logic is the ultrafilter frame of its Lindenbaum algebra. Completeness proofs are the Jónsson–Tarski construction, run without knowing it.

Canonicity: A formula is canonical if it is preserved from B to (B₊)⁺. Canonical formulas axiomatize complete logics. Sahlqvist formulas are canonical — that is the algebraic content of Sahlqvist's theorem.

Symbols.

Metatheory. The representation theorem of Jónsson and Tarski (1951) states that every Boolean algebra with normal additive operators embeds into a complete and atomic such algebra — the complex algebra of its ultrafilter frame — and that the operators extend uniquely to completely additive ones on the extension; this is the canonical extension, and the ultrafilter frame is literally the canonical model of the corresponding logic. The correspondence sharpens to a dual equivalence of categories between BAOs with homomorphisms and descriptive general frames with bounded morphisms. Under it, the lattice of normal modal logics is dually isomorphic to the lattice of varieties of modal algebras, so Birkhoff's variety theorem yields the Goldblatt–Thomason theorem as its frame-side reading. Algebraic completeness is unconditional — every normal modal logic is complete for its Lindenbaum–Tarski algebra — which is exactly why Kripke incompleteness (Thomason 1974, Fine 1974) is possible without any semantics failing. Canonicity is the algebraic property of a variety being closed under canonical extensions; Fine (1975) proved that a logic determined by an elementary frame class is canonical, and Goldblatt, Hodkinson and Venema (2004) refuted the converse.

Applies to. Every completeness proof in modal logic. Correspondence theory and canonicity. Coalgebraic logic, which generalizes the construction past Boolean algebras. Description logics and dynamic logics, whose semantics are BAOs with several operators. The relation between Algebraic and Modal in this collection, which this theorem is.

Limitations. Kripke frames are not dual to modal algebras at all: complex algebras are complete and atomic while arbitrary BAOs are neither, so the frame side must carry an admissible set algebra and satisfy differentiation, tightness and compactness, and dropping any of these breaks the equivalence. The construction is non-constructive — ultrafilter frames require the ultrafilter lemma, so the duality is unavailable in ZF alone — and for infinite B the embedding into (B₊)⁺ is proper, the extension outgrowing B and giving no effective handle on the frame; only in the finite case does the algebra coincide with its canonical extension. Canonicity has no syntactic characterization beyond Sahlqvist's sufficient condition, and Chagrova (1991) proved it undecidable whether a given modal formula has a first-order frame correspondent. Frame validity quantifies over every valuation, hence over every subset of W, so it remains a second-order notion that the duality reorganizes rather than reduces. Normality and additivity of the operator are necessary: non-normal modal logics fall outside entirely and require neighbourhood or coalgebraic duality instead.

© 2026 Lingenic LLC