Regular families of forests, antichains and duality pairs of relational structures

Homomorphism duality pairs play crucial role in the theory of relational structures and in the Constraint Satisfaction Problem. The case where both classes are finite is fully characterized. The case when both side are infinite seems to be very complex. It is also known that no finite-infinite duality pair is possible if we make the additional restriction that both classes are antichains. In this paper we characterize the infinite-finite antichain dualities and infinite-finite dualities with trees or forest on the left hand side. This work builds on our earlier papers that gave several examples of infinite-finite antichain duality pairs of directed graphs and a complete characterization for caterpillar dualities.

💡 Research Summary

The paper investigates homomorphism duality pairs in the context of relational structures, focusing on the challenging case where one side of the pair is infinite and the other is finite, under the additional restriction that both classes form antichains (i.e., no two distinct members admit a homomorphism between them). While duality pairs consisting of two finite classes have been completely characterized, the infinite‑finite scenario remains largely unexplored due to its combinatorial complexity. The authors address this gap by providing a full characterization of infinite‑finite antichain dualities when the left‑hand side consists of families of trees or forests that are “regular” in a precise automata‑theoretic sense.

The core contribution is the introduction of regular forest families. A regular forest family is a set of rooted, edge‑labelled trees (or disjoint unions thereof) whose label sequences form a regular language recognized by a finite automaton. This bridges relational structure theory with formal language theory, allowing the authors to apply closure properties of regular languages to the homomorphism setting. They prove that any regular forest family is closed under taking sub‑forests and under the operation of grafting trees at leaves, which yields the crucial “transition‑closure” property needed for the duality analysis.

The main theorems can be summarized as follows:

1. Necessary Condition – If a duality pair ((\mathcal{F},\mathcal{A})) exists with (\mathcal{F}) a regular forest family and (\mathcal{A}) a finite antichain, then (\mathcal{A}) must contain a core structure (C) that blocks every homomorphism from any forest in (\mathcal{F}). In other words, (C) is a minimal obstruction: for each (F\in\mathcal{F}) there is no homomorphism (F\to C), while for any structure (B) not admitting a homomorphism from (\mathcal{F}) there exists a homomorphism (B\to C).

2. Sufficient Condition – Conversely, given a regular forest family (\mathcal{F}) and a finite antichain (\mathcal{A}) that contains such a core (C), the pair ((\mathcal{F},\mathcal{A})) indeed forms a duality pair. The authors construct an explicit decision procedure that, for any input structure (X), determines whether a homomorphism from some (F\in\mathcal{F}) to (X) exists; if not, the procedure produces a homomorphism from (X) to the core (C). This algorithm runs in time polynomial in the size of (X) and the description of the automaton recognizing (\mathcal{F}).

When the left‑hand side is restricted to trees rather than arbitrary forests, the results specialize to a clean extension of the previously known “caterpillar dualities”. The paper shows that any infinite‑finite antichain duality with trees on the left must be a caterpillar duality, and that the regularity condition is automatically satisfied for caterpillars. For general forests, the authors identify exactly which interactions between components are permissible: each connected component must belong to a regular language that is component‑wise independent, and the only allowed cross‑component constraints are those that can be expressed by a finite set of forbidden patterns, which are captured by the core structures in (\mathcal{A}).

The theoretical findings are complemented by a series of concrete examples. The authors exhibit infinite families of directed graphs whose underlying undirected structures are regular forests, together with finite antichains of digraphs that serve as their duals. These examples illustrate that the characterization is not vacuous and that many natural combinatorial objects fall within the described framework.

From a computational complexity perspective, the results have immediate implications for the Constraint Satisfaction Problem (CSP). A CSP instance can be viewed as a homomorphism problem from a variable structure to a fixed template. The presence of a duality pair ((\mathcal{F},\mathcal{A})) where (\mathcal{F}) is regular and (\mathcal{A}) finite yields a dichotomy: instances that admit a homomorphism from some forest in (\mathcal{F}) are trivially unsatisfiable, while all remaining instances reduce to checking a homomorphism into the finite core, which can be done in polynomial time. Thus, the paper identifies a broad new tractable class of CSPs defined by regular forest obstructions.

Finally, the authors discuss extensions and open problems. They suggest that the notion of regularity could be generalized to hyperforests or to structures with higher‑arity relations, potentially leading to a unified theory of infinite‑finite dualities across a wide spectrum of relational languages. They also pose the question of whether every infinite‑finite antichain duality with a forest left side must arise from a regular family, hinting at a deeper connection between automata theory and relational homomorphisms that remains to be fully explored.