The Isomorphism Problem On Classes of Automatic Structures

Automatic structures are finitely presented structures where the universe and all relations can be recognized by finite automata. It is known that the isomorphism problem for automatic structures is complete for $\Sigma^1_1$; the first existential level of the analytical hierarchy. Several new results on isomorphism problems for automatic structures are shown in this paper: (i) The isomorphism problem for automatic equivalence relations is complete for $\Pi^0_1$ (first universal level of the arithmetical hierarchy). (ii) The isomorphism problem for automatic trees of height $n \geq 2$ is $\Pi^0_{2n-3}$-complete. (iii) The isomorphism problem for automatic linear orders is not arithmetical. This solves some open questions of Khoussainov, Rubin, and Stephan.

💡 Research Summary

The paper investigates the isomorphism problem for several natural subclasses of automatic structures, establishing precise complexity classifications that sharpen the previously known Σ¹₁‑completeness for the general case. An automatic structure is one whose domain and all relations are recognizable by finite automata; this finitary presentation makes it a fertile ground for studying algorithmic properties of infinite structures. The authors focus on three families: automatic equivalence relations, automatic trees of bounded height, and automatic linear orders.

Result 1 – Automatic equivalence relations.
The authors show that deciding whether two automatic equivalence relations are isomorphic is Π⁰₁‑complete. The proof proceeds by encoding the multiset of class sizes of an automatic equivalence relation as an infinite sequence of natural numbers. Two such relations are isomorphic exactly when, for every integer k, the number of classes of size k coincides in both structures. This universal quantification over ℕ yields a Π⁰₁ condition, and a reduction from the canonical Π⁰₁‑complete problem (e.g., totality of a Turing machine) establishes hardness.

Result 2 – Automatic trees of height n (n ≥ 2).
For each fixed height n, the isomorphism problem for automatic trees of that height is Π⁰_{2n‑3}‑complete. The authors construct a hierarchical encoding: the branching pattern at level i (1 ≤ i ≤ n) is described by a regular language that can be checked by a Π⁰_{2i‑1} predicate. By inductively combining the predicates for all levels, the overall isomorphism condition becomes a Π⁰_{2n‑3} statement. Hardness is proved by a many‑one reduction from the canonical Π⁰_{2n‑3}‑complete problem, namely the universal quantifier alternation over a Σ⁰_{2n‑4} predicate. This result demonstrates a linear increase of descriptive complexity with the tree height, a phenomenon not observed for other automatic classes.

Result 3 – Automatic linear orders.
The most striking finding is that the isomorphism problem for automatic linear orders is not arithmetical; in fact it is Σ¹₁‑complete. The authors design a uniform encoding that maps any analytic set A ⊆ ℕ (given by a Σ¹₁ formula) to an automatic linear order L_A such that A contains a number x iff L_A has a specific order‑type component (e.g., a copy of ω·x). Consequently, two automatic linear orders are isomorphic precisely when the corresponding analytic sets coincide, yielding a Σ¹₁‑hardness reduction. Membership in Σ¹₁ follows from the observation that isomorphism can be expressed as “there exists a bijection definable by a synchronous two‑tape automaton that preserves the order,” a Σ¹₁ condition.

These three theorems answer open questions raised by Khoussainov, Rubin, and Stephan concerning the exact complexity of isomorphism for these subclasses. The paper also outlines the technical toolkit employed: reductions from classical decision problems, careful use of regular language representations for hierarchical structures, and the exploitation of Cantor‑Bendixson rank arguments for linear orders.

Beyond the immediate results, the work suggests several avenues for future research. One direction is to explore whether similar height‑dependent hierarchies arise for other automatic graph families, such as bounded‑degree graphs or automatic partial orders. Another is to compare automatic structures with those presented by more powerful devices (push‑down automata, higher‑order automata) to see how the isomorphism complexity shifts. Finally, the authors hint at the possibility of developing practical algorithms for the low‑complexity cases (e.g., Π⁰₁ for equivalence relations) while acknowledging the inherent infeasibility for the Σ¹₁‑complete linear orders.

In sum, the paper delivers a comprehensive and nuanced classification of the isomorphism problem across several key automatic structure classes, deepening our understanding of the interplay between automata‑theoretic presentations and descriptive set‑theoretic complexity.