Automaton-based Characterisations of First Order Logic over Infinite Trees

Abstract: We study the expressive power of First-Order Logic (\FO) over (unordered) infinite trees, with the aim of identifying robust characterisations in terms of branching-time specification formalisms. While such correspondences are well understood in the linear-time setting, the branching-time case presents well-known structural challenges. To this end, we introduce two classes of hesitant tree automata and show that they capture precisely the expressive power of two branching-time temporal logics, namely \PolPCTL and \CTLsf, both of which have been previously shown to be equivalent to \FO over infinite trees. These results provide uniform automata-theoretic characterisations and yield a natural normal form for the latter in terms of a new fragment of \CTLs called \PolCTLs. As a consequence, we identify a fundamental limitation of \FO in this setting: along each branch, it can express only properties that are either safety or co-safety, thereby revealing a sharp expressive boundary for first-order definability over infinite trees.

• The paper introduces two novel classes of hesitant tree automata that exactly capture FO expressiveness over infinite, unordered trees.
• It demonstrates that FO on infinite trees is limited to safety or co-safety properties, linking branch behaviours with temporal logic modalities.
• The work establishes equivalences between these automata and tailored temporal logics (PastCTL± and CTL*₍f₎), offering new tools for formal verification.

Automaton-based Characterisations of First-Order Logic over Infinite Trees

Introduction and Motivation

The expressive boundaries of First-Order Logic (FO) over infinite trees remain a nuanced topic in logic, automata theory, and formal verification. While automata-theoretic and logical characterisations are well-understood in linear-time (word) settings, the branching-time (tree) setting presents unresolved challenges. In particular, the lack of effective procedures to determine FO-definability for regular tree languages underscores the need for structural correspondences. This paper addresses a substantial gap by providing uniform automaton-theoretic characterizations of FO over infinite, unordered, unranked, leafless trees, strictly via new classes of hesitant tree automata that align precisely with the expressive reach of certain branching-time temporal logics.

Background: Logical and Automata-Theoretic Landscape

Linear vs. Branching Time

• Linear-time: The equivalences between FO, LTL, monadic second-order logic (MSO), star-free, and $\omega$-regular languages are foundational; aperiodicity and counter-freeness provide robust algebraic and automata-theoretic criteria for FO-definability.
• Branching-time: Despite progress in logical and algebraic fragments, the full automaton-based characterisation of FO over infinite trees—especially unordered and unranked trees with the descendant relation—remains elusive.

FO, Path Logics, and Temporal Logics

• MSO over infinite trees is captured by parity tree automata; for FO, only partial algebraic and automata-theoretic characterizations were known.
• Prior work connects temporal logics (notably CTL$^*$, PastCTL, and their graded/counting extensions) to monadic path logics and fragments of chain logics, but the precise automaton characterisation matching FO was unavailable.

Main Contributions

1. New Classes of Hesitant Tree Automata

The authors introduce two novel classes of automata: two-way linear polarised hesitant tree automata and counter-free visible polarised hesitant tree automata (cfHTA$^-_\text{vis}$). These automata are shown to characterise FO on infinite trees via equivalences to temporal logics PastCTL$_\pm$ and CTL$^*_{f}$ respectively, both of which are proven FO-equivalent in this setting.

Two-Way Linear Polarised Hesitant Tree Automata (2W-LPHTA)

• Automata are linear (singleton state components), polarised (acceptance bifurcates along existential/universal lines), and equipped with a two-way movement (moving along parent/child).
• Shown equivalent to PastCTL$_\pm$, a fragment of computation tree logic with polarised path quantification and past/future modalities that aligns with FO.

Counter-Free Visible Polarised Hesitant Tree Automata (cfHTA$^-_\text{vis}$)

• One-way automata with components interpretable as counter-free (aperiodic) $\omega$-word automata, subject to a visibility property on branch annotations.
• Shown equivalent to the CTL$^*_{f}$ normal form, a syntactically restricted fragment of CTL$^*$ with finite-path quantification, which is proven to capture FO on infinite trees.

2. Automata-Temporal Logic-Fo Correspondence

• Expressive Equivalence: Both automata correspond in expressiveness to their matching temporal logics; both past and future temporal branching-time logics PastCTL$^*$0 and CTL$^*$1 are FO-complete on infinite unranked unordered trees.
• Normal Forms and Polarisation: The automata-theoretic analysis yields natural normal forms for these logics and suggests a polarisation property: along branches, existential quantification aligns with co-safety (liveness-like) properties, while universal quantification aligns with safety.

3. Limitation of FO: Polarisation and Branch Definability

• Fundamental barrier: FO on infinite trees can express only properties that, on every path, are either safety or co-safety. Thus, FO cannot express properties requiring true alternation between safety and liveness along the same branch.
• This is formalised as a sharp expressive separation: “On each branch, FO definability is restricted to safety or co-safety expressible properties.”

4. Characterisations for Safety/Co-safety Fragments of LTL

• The authors provide direct automaton-based characterisations for the safety and co-safety fragments of linear temporal logic (LTL), demonstrating that counter-free looping universal B\"uchi automata capture SafeLTL and the dual holds for co-safeLTL with co-B\"uchi acceptance. This closes a gap in the literature by making the word case correspond exactly to the tree case automatonically.

Technical Core

Temporal Logics Equivalent to FO

• PastCTL$^*$2: Polarised fragment with path quantification, future/past modalities, and graded (counting) extensions.
  • Only safety or co-safety properties can be defined per branch.
  • Universality and existentiality are chained syntactically to these properties.
• CTL$^*$3: CTL$^*$4 variant with finite-path quantification.
  • Syntactically identical to CTL$^*$5 except for path quantifier semantics.
• Equivalence: Both PastCTL$^*$6 and CTL$^*$7 are shown expressively equivalent to FO over infinite trees with descendant relation.

Automaton Theoretic Constructions

• Alternating graded B\"uchi tree automata with polarised hesitant partitions.
• For PastCTL$^*$8, two-way, linearisation (singleton components) is crucial; for CTL$^*$9, one-way, visibility on components (branch annotations) and counter-freeness are imposed.
• Correctness and completeness are established by tight, mutual inductive translations between automata and their logical fragments.

Automata-Logic Expressive Hierarchy

• Strict inclusions: The logical and automata fragments exhibit strict inclusions, mirroring the established strictness results between CTL, CTL$^-_\text{vis}$0, PastCTL.
• Incomparability: Some logical logics (e.g., PastCTL$^-_\text{vis}$1 and CTL) are shown to be incomparable in their definability over trees.

Key Claims and Results

• Equivalence Theorems: Both 2W-LPHTA and cfHTA$^-_\text{vis}$2 precisely capture FO-definable languages over infinite trees (2604.26364).
• Expressive Limitation: FO cannot define properties that on some branches require mutual alternation of safety and liveness "in the FO sense".
• Automaton Hierarchy: Linear hesitant automata (CTL-like) are strictly less expressive than two-way linear hesitant ones (PastCTL-like); the tree setting forces two-way or more global reasoning to fully capture FO ([Section 6], [Section 7]).

Implications and Future Work

Theoretical Implications

• Automata/Logic Robustness: The results situate the boundary of FO definability in a precise automata-theoretic regime, sharpening the understanding of the logic-automata correspondence beyond the word case.
• Normal Forms and Polarisation: The emergence of polarised normal forms for both the automaton and logical sides suggests general principles likely relevant to further studies in tree logics, modal $^-_\text{vis}$3-calculus fragments, and variants of path logic.
• Limitation of FO: The inability of FO to express "non-polarised" properties along paths impacts definability and the decision procedures for model checking and synthesis in verification contexts.

Practical and Algorithmic Directions

• Membership Testing: While the algebraic membership problem for FO-definable regular tree languages remains open (e.g., for deterministic automata), the automaton characterisations here may provide a conceptual road to algorithms or approximations, especially given the tight constraints (counter-freeness, visibility, polarisation).
• Tool Support: Translating tree properties specified in FO or temporal logics to automata of the types described may facilitate implementation and verification in system design tools, provided the characterisations can be made effectively constructive.

Open Problems and Prospects

• Finite Tree Extensions: The characterisations for finite trees, especially for reaching deterministic automata or establishing analogs over finite/hybrid structures, remain unresolved.
• Automata Class Transformations: Effective, direct transformations between the one-way visible cfHTA and two-way linear polarised automata (without detouring through logic) are still absent and pose intriguing combinatorial challenges.
• Logical Fragments: Precisely isolating the expressive power and boundaries of PastCTL and its relation to subfragments of MSO, as well as further normal forms for FO on trees, remains a fertile direction.

Conclusion

This work gives two automaton-theoretic characterisations—via two-way linear polarised hesitant tree automata and one-way counter-free visible hesitant tree automata—of the expressive power of first-order logic over infinite unranked, unordered, leafless trees. It establishes that FO on such trees is limited, on every branch, to properties that are either safety or co-safety. The results unify previously disparate logical and automata frameworks and advance the understanding of the FO–automata boundary for infinite trees, thus providing foundational tools for subsequent investigation in automata theory, logical definability, and temporal logic. The open problems highlighted—particularly regarding effective membership algorithms and the extension to finite trees—should stimulate significant further research in the area.