Weak MSO: Automata and Expressiveness Modulo Bisimilarity

We prove that the bisimulation-invariant fragment of weak monadic second-order logic (WMSO) is equivalent to the fragment of the modal $\mu$-calculus where the application of the least fixpoint operator $\mu p.\varphi$ is restricted to formulas $\varphi$ that are continuous in $p$. Our proof is automata-theoretic in nature; in particular, we introduce a class of automata characterizing the expressive power of WMSO over tree models of arbitrary branching degree. The transition map of these automata is defined in terms of a logic $\mathrm{FOE}_1^\infty$ that is the extension of first-order logic with a generalized quantifier $\exists^\infty$, where $\exists^\infty x. \phi$ means that there are infinitely many objects satisfying $\phi$. An important part of our work consists of a model-theoretic analysis of $\mathrm{FOE}_1^\infty$.

💡 Research Summary

The paper investigates the bisimulation‑invariant fragment of weak monadic second‑order logic (WMSO) and shows that it coincides precisely with a restricted fragment of the modal μ‑calculus, denoted µcML, in which the least‑fixpoint operator μ p.φ may be applied only when φ is continuous in the variable p. Continuity here means Scott‑continuity: the semantic function induced by φ on subsets of a domain is determined by its behaviour on finite subsets. This restriction strengthens the usual positivity condition required for μ‑formulas in the full μ‑calculus.

To establish the equivalence, the authors develop an automata‑theoretic framework tailored to WMSO on trees of arbitrary branching degree. Traditional MSO‑automata use a one‑step language FOE₁ (first‑order logic with equality) for transition formulas. For WMSO, the authors replace this by FOE₁^∞, which extends FOE₁ with the generalized quantifier ∃^∞ (“there exist infinitely many”). However, using the full FOE₁^∞ would make the automata too expressive; they would exceed WMSO’s power. Consequently, two constraints are imposed on the automata:

1. Weakness – all states belonging to the same strongly connected component (SCC) share the same parity priority. This mirrors the classic notion of weak parity automata.

2. Continuity – if a state a has an odd (resp. even) priority, then for every colour c the transition formula Δ(a,c) must be continuous (resp. co‑continuous) with respect to any other state a′ in the same SCC. Continuity of a FOE₁^∞‑sentence in a predicate a′ is defined model‑theoretically and can be captured syntactically.

Automata satisfying both constraints form the class Aut_cw(FOE₁^∞). The authors prove a tight correspondence:

• WMSO ↔ Aut_cw(FOE₁^∞) – every WMSO formula can be effectively translated into an equivalent automaton in this class, and conversely each such automaton can be turned back into a WMSO formula. Hence WMSO is exactly captured by these weak‑continuous automata over all tree models.

The next step connects these automata to the μ‑calculus fragment. By providing a translation from FOE₁^∞ to FO₁ (first‑order logic without equality) that preserves continuity, the authors obtain a mapping from Aut_cw(FOE₁^∞) to Aut_cw(FO₁). The latter class is shown to be precisely the automata-theoretic counterpart of µcML, i.e., the μ‑calculus where every μ‑binding respects continuity. Moreover, the translation respects ω‑unravelings of transition systems: an automaton A accepts a system T iff its translated version A· accepts the ω‑unraveling T^ω. This guarantees that the translation preserves bisimulation invariance.

Putting the pieces together yields the main theorem:

Theorem (informal). µcML ≡ WMSO /↔ (over the class of all labelled transition systems).

In other words, a property of transition systems is definable in WMSO and invariant under bisimulation exactly when it is definable in the continuous‑μ fragment of the modal μ‑calculus. The result refines earlier characterizations: while full MSO corresponds to the unrestricted μ‑calculus, and AFMC (alternation‑free μ‑calculus) corresponds to WMSO on finitely branching trees, the present work identifies the precise fragment for arbitrary branching degree.

A substantial auxiliary contribution is the model‑theoretic analysis of FOE₁^∞. The paper supplies normal forms for its sentences and syntactic characterizations of the monotone and (co‑)continuous fragments with respect to individual monadic predicates. These results make the continuity condition decidable at the syntactic level, enabling effective construction of the required automata.

Overall, the paper advances the understanding of the expressive power of weak second‑order logic in a bisimulation‑sensitive setting, bridges it to a well‑behaved fragment of the modal μ‑calculus, and enriches the toolbox of automata‑theoretic techniques for logics on infinite trees.