Lattices of Logical Fragments over Words

This paper introduces an abstract notion of fragments of monadic second-order logic. This concept is based on purely syntactic closure properties. We show that over finite words, every logical fragment defines a lattice of languages with certain closure properties. Among these closure properties are residuals and inverse C-morphisms. Here, depending on certain closure properties of the fragment, C is the family of arbitrary, non-erasing, length-preserving, length-multiplying, or length-reducing morphisms. In particular, definability in a certain fragment can often be characterized in terms of the syntactic morphism. This work extends a result of Straubing in which he investigated certain restrictions of first-order logic formulae. In contrast to Straubing’s model-theoretic approach, our notion of a logical fragment is purely syntactic and it does not rely on Ehrenfeucht-Fraisse games. As motivating examples, we present (1) a fragment which captures the stutter-invariant part of piecewise-testable languages and (2) an acyclic fragment of Sigma_2. As it turns out, the latter has the same expressive power as two-variable first-order logic FO^2.

💡 Research Summary

The paper introduces an abstract notion of “fragments” of monadic second‑order (MSO) logic, defined solely by syntactic closure properties rather than semantic or game‑theoretic criteria. A fragment is a set of formulas closed under Boolean connectives, quantifier introduction, and certain variable‑binding operations. The authors show that for finite words every such fragment induces a lattice of languages: the class of languages definable by the fragment is closed under finite unions, intersections, and, crucially, under taking residuals (left and right quotients) and under inverse morphisms belonging to a family C. The family C depends on the fragment’s syntactic restrictions and may consist of arbitrary morphisms, non‑erasing morphisms, length‑preserving morphisms, length‑multiplying morphisms, or length‑reducing morphisms.

A central technical contribution is the connection between these closure properties and the syntactic morphism of a language. If a fragment is closed under inverse C‑morphisms, then any language definable in the fragment can be characterized by its syntactic morphism: the language is a union of syntactic congruence classes that are stable under the corresponding C‑morphisms. This generalizes a result of Straubing, who studied specific restrictions of first‑order logic using model‑theoretic tools and Ehrenfeucht‑Fraïssé games. Here, the authors obtain the same kind of algebraic characterizations without invoking game arguments, relying only on the purely syntactic definition of fragments.

To illustrate the framework, two concrete fragments are presented. The first captures the stutter‑invariant part of piecewise‑testable languages. “Stutter” refers to the insertion or deletion of consecutive identical letters; the fragment defines exactly those piecewise‑testable languages that remain unchanged under such operations. The second fragment is an acyclic fragment of Σ₂ (the second level of the quantifier alternation hierarchy). Although Σ₂ allows formulas with an existential block followed by a universal block, the acyclic restriction forces the quantifier dependency graph to be a DAG. The authors prove that this fragment has precisely the expressive power of two‑variable first‑order logic (FO²). Consequently, the well‑known equivalence between FO² and a certain subclass of Σ₂ is recovered in a purely syntactic manner.

The paper proceeds as follows. After formalizing fragments and the associated closure conditions, the authors develop the lattice theory: they prove that residuals and inverse C‑morphisms preserve definability within any fragment satisfying the corresponding syntactic constraints. They then establish the algebraic characterization via syntactic morphisms, showing that the syntactic monoid of a language encodes exactly the fragment’s expressive power when the fragment is closed under the appropriate class of morphisms. The two motivating examples are worked out in detail, demonstrating how the abstract theory yields concrete characterizations for well‑studied language families.

In summary, the work provides a unified, syntactic framework for relating logical fragments, language lattices, and algebraic recognizers. By abstracting away from model‑theoretic arguments, it offers a versatile tool for analyzing the expressive power of restricted logics over words, and it opens the door to systematic exploration of new fragments based on their syntactic closure properties.