On tractability and congruence distributivity

Constraint languages that arise from finite algebras have recently been the object of study, especially in connection with the Dichotomy Conjecture of Feder and Vardi. An important class of algebras are those that generate congruence distributive varieties and included among this class are lattices, and more generally, those algebras that have near-unanimity term operations. An algebra will generate a congruence distributive variety if and only if it has a sequence of ternary term operations, called Jonsson terms, that satisfy certain equations. We prove that constraint languages consisting of relations that are invariant under a short sequence of Jonsson terms are tractable by showing that such languages have bounded relational width.

💡 Research Summary

This paper investigates the algorithmic tractability of constraint satisfaction problems (CSPs) whose constraint languages are derived from finite algebras that generate congruence‑distributive varieties. The authors focus on algebras that admit a short sequence of Jónsson terms—ternary term operations (t_0,\dots,t_n) satisfying the classic Jónsson identities (e.g., (t_0(x,y,z)=x), (t_n(x,y,z)=z), and for each (i) the equations (t_i(x,x,y)=t_{i+1}(x,y,y))). Such a sequence exists if and only if the algebra’s variety is congruence‑distributive, a class that includes lattices and, more generally, algebras with near‑unanimity (NU) terms.

The main contribution is a proof that any constraint language (\Gamma) whose relations are invariant under a bounded‑length Jónsson chain has bounded relational width. In CSP terminology, relational width is the smallest integer (k) such that enforcing ((k,k-1))-consistency (or more generally ((k,l))-consistency) suffices to decide satisfiability. The authors show that if an algebra admits (n) Jónsson terms, then (\Gamma) has width at most (n+1) (or a constant depending only on the length of the chain). Consequently, CSP instances over (\Gamma) can be solved in polynomial time by a simple local‑consistency algorithm, without resorting to more sophisticated global reasoning.

The proof proceeds in two stages. First, the invariance under Jónsson terms imposes a strong algebraic structure on each relation: applying any (t_i) to tuples from the relation yields another tuple of the same relation. This property enables the authors to decompose each relation into a product of multi‑congruence blocks, essentially partitioning the domain into a bounded number of equivalence classes that interact in a highly regular way. The second stage constructs a consistency‑propagation procedure that iteratively refines these blocks. By repeatedly applying ((k,l))-consistency with (k=n+1), the algorithm eliminates infeasible assignments while preserving all solutions. The authors prove that the refinement process stabilizes after a polynomial number of steps and that any non‑empty block assignment at termination extends to a global solution. The argument relies on the Jónsson identities to guarantee that local compatibility propagates globally.

A particularly important corollary concerns algebras with NU terms. Near‑unanimity operations correspond to very short Jónsson chains (typically (n=2) or (3)), which yields relational width 3 or less. Hence CSPs over languages invariant under an NU term are solved by the classic ((2,3))-consistency algorithm, recovering and strengthening known tractability results for lattices, semilattices, and other distributive structures.

The paper situates its findings within the broader context of the Feder–Vardi Dichotomy Conjecture, which predicts that every finite‑domain CSP is either in P or NP‑complete. Prior algebraic characterisations identified the presence of a Taylor term as a sufficient condition for tractability, but they did not provide explicit width bounds. By focusing on the stronger congruence‑distributive condition, this work not only confirms tractability but also quantifies it via bounded width, thereby bridging the gap between abstract algebraic criteria and concrete algorithmic guarantees.

In the concluding discussion, the authors suggest several avenues for future research: extending the bounded‑width analysis to varieties that are merely congruence‑modular, exploring the exact relationship between the length of a Jónsson chain and the optimal width bound, and investigating whether similar techniques can yield width bounds for other well‑studied polymorphism classes (e.g., Maltsev or majority operations). Overall, the paper delivers a clear and technically robust connection between Jónsson term sequences, congruence‑distributivity, and polynomial‑time solvability of CSPs, enriching the algebraic toolkit for tackling the dichotomy problem.