Model Checking Positive Equality-free FO: Boolean Structures and Digraphs of Size Three

We study the model checking problem, for fixed structures A, over positive equality-free first-order logic – a natural generalisation of the non-uniform quantified constraint satisfaction problem QCSP(A). We prove a complete complexity classification for this problem when A ranges over 1.) boolean structures and 2.) digraphs of size (less than or equal to) three. The former class displays dichotomy between Logspace and Pspace-complete, while the latter class displays tetrachotomy between Logspace, NP-complete, co-NP-complete and Pspace-complete.

💡 Research Summary

The paper investigates the model‑checking problem for a fixed finite structure A when the input sentences belong to positive equality‑free first‑order logic (denoted P‑FO). In P‑FO only conjunction, disjunction, existential and universal quantifiers are allowed; negation and the equality predicate are forbidden. This restriction makes the language a natural extension of the non‑uniform quantified constraint satisfaction problem QCSP(A), because QCSP(A) corresponds to the fragment where only universal quantifiers appear in front of a conjunction of atomic formulas. By allowing arbitrary interleavings of ∀ and ∃ while still forbidding negation, P‑FO is strictly more expressive, and the complexity of its model‑checking problem can be higher.

The authors present complete complexity classifications for two families of structures:

1. Boolean structures (i.e., structures whose domain consists of two elements).
2. Directed graphs (digraphs) with at most three vertices.

The main technical tool is the algebraic notion of surjective hyper‑endomorphisms (shes). A she is a multivalued, surjective map from the domain to itself that preserves all relations of the structure. Two special kinds of shes are distinguished:

• ∀‑shes (universal‑preserving shes) – they allow every universally quantified variable to be collapsed to a single element without changing the truth of any sentence.
• ∃‑shes (existential‑preserving shes) – they enable a similar collapse for existentially quantified variables.

There is a Galois connection between the set of relations definable in P‑FO over A and the set of shes that A admits. Consequently, the presence or absence of these two types of shes determines how much quantifier alternation can be eliminated, which in turn dictates the computational difficulty of MC(A).

Boolean Structures – Dichotomy

For any Boolean structure A the authors prove:

• Logspace: If A admits a ∀‑she (or both a ∀‑she and an ∃‑she), then every ∀‑quantifier can be replaced by a constant, and the remaining formula can be evaluated by a deterministic log‑space algorithm. This covers, for example, the two‑element total order, the Boolean algebra with both constants, and any structure where every element is a universal absorber.

• PSPACE‑complete: If A does not admit a ∀‑she (regardless of whether an ∃‑she exists), then quantifier alternation cannot be eliminated. The authors give a reduction from quantified Boolean formulas (QBF) to the P‑FO model‑checking problem, showing that MC(A) is PSPACE‑hard, and they also show containment in PSPACE by a straightforward recursive evaluation. Hence the problem is PSPACE‑complete.

Thus the Boolean case exhibits a clean dichotomy between Logspace and PSPACE‑complete, mirroring the well‑known dichotomy for QCSP on two‑element domains but with a different boundary because equality is absent.

Digraphs of Size ≤ 3 – Tetrachotomy

The second part of the paper performs an exhaustive analysis of all non‑isomorphic digraphs on at most three vertices (including possible self‑loops). For each digraph D the authors determine which shes it admits and then classify MC(D) into one of four complexity classes:

1. Logspace – D admits a ∀‑she (or both a ∀‑she and an ∃‑she). Typical examples are the complete digraph K₃, any digraph where every vertex has a self‑loop, or the directed 3‑cycle with self‑loops. In these cases all universal quantifiers can be collapsed, and the remaining existential part can be evaluated in deterministic log‑space.

2. NP‑complete – D admits an ∃‑she but no ∀‑she. Here universal quantifiers cannot be eliminated, but existential quantifiers can be reduced to a CSP‑like core. The authors construct a polynomial‑time many‑one reduction from SAT to MC(D) by encoding each clause as a positive existential conjunct, proving NP‑hardness; membership in NP follows from guessing assignments for the existential variables.

3. co‑NP‑complete – D admits a ∀‑she but no ∃‑she. The situation is dual to the NP case: universal quantifiers can be collapsed, leaving a formula that essentially asks whether every assignment to the remaining variables satisfies a positive conjunctive condition. The authors reduce UNSAT to MC(D), establishing co‑NP‑hardness, and show co‑NP membership by universal verification of all assignments.

4. PSPACE‑complete – D admits neither a ∀‑she nor an ∃‑she. In this most expressive setting the formula retains full alternation of quantifiers, and the authors adapt the classic QBF reduction to the equality‑free positive fragment, proving PSPACE‑hardness. A straightforward recursive evaluation yields PSPACE containment.

The paper supplies explicit reductions for each hardness claim, often by “gadget” constructions that embed Boolean variables and clauses into the edge relation of the digraph. For the NP and co‑NP cases, the reductions are carefully designed to respect the positivity constraint (no negation) and the absence of equality, which requires encoding variable equality through shared vertices or self‑loops rather than an explicit equality predicate.

Methodological Contributions

• Algebraic Characterisation – By introducing shes and establishing their Galois connection with definable relations, the authors provide a clean algebraic lens for understanding the expressive power of P‑FO over a fixed structure. This mirrors the role of polymorphisms in CSP and QCSP classifications but is tailored to the positivity and equality‑free setting.

• Complete Exhaustive Classification – The analysis of all three‑vertex digraphs is exhaustive; the authors enumerate each isomorphism class, determine its she‑profile, and then place it in the appropriate complexity class. This systematic approach leaves no gaps and demonstrates that the tetrachotomy is tight.

• Bridging QCSP and P‑FO – The results show that allowing existential quantifiers to appear arbitrarily (while still forbidding negation) yields a richer landscape than QCSP, which only permits a universal prefix. Nevertheless, the same algebraic dichotomies (presence of certain polymorphisms/she) continue to dictate tractability versus hardness.

Conclusions and Outlook

The paper establishes that for Boolean structures the model‑checking problem for positive equality‑free FO collapses to either Logspace or PSPACE‑complete, and for digraphs with up to three vertices it falls into exactly one of Logspace, NP‑complete, co‑NP‑complete, or PSPACE‑complete. The classification hinges on the existence of ∀‑shes and ∃‑shes, providing a clean algebraic criterion.

Future work suggested by the authors includes extending the tetrachotomy to larger digraphs (four or more vertices), investigating the impact of re‑introducing equality, and exploring connections with other algebraic frameworks such as clones and polymorphism algebras. Moreover, the techniques may be applicable to other fragments of first‑order logic that restrict negation or equality, potentially leading to broader dichotomy or trichotomy theorems in descriptive complexity.