On Decidability Properties of One-Dimensional Cellular Automata

In a recent paper Sutner proved that the first-order theory of the phase-space $\mathcal{S}\mathcal{A}=(Q^\mathbb{Z}, \longrightarrow)$ of a one-dimensional cellular automaton $\mathcal{A}$ whose configurations are elements of $Q^\mathbb{Z}$, for a finite set of states $Q$, and where $\longrightarrow$ is the “next configuration relation”, is decidable. He asked whether this result could be extended to a more expressive logic. We prove in this paper that this is actuallly the case. We first show that, for each one-dimensional cellular automaton $\mathcal{A}$, the phase-space $\mathcal{S}\mathcal{A}$ is an omega-automatic structure. Then, applying recent results of Kuske and Lohrey on omega-automatic structures, it follows that the first-order theory, extended with some counting and cardinality quantifiers, of the structure $\mathcal{S}_\mathcal{A}$, is decidable. We give some examples of new decidable properties for one-dimensional cellular automata. In the case of surjective cellular automata, some more efficient algorithms can be deduced from results of Kuske and Lohrey on structures of bounded degree. On the other hand we show that the case of cellular automata give new results on automatic graphs.

💡 Research Summary

The paper investigates the logical properties of the phase‑space of one‑dimensional cellular automata (CA) and shows that a much richer logical language than plain first‑order (FO) remains decidable. The authors begin by formalising the phase‑space 𝒮_𝔄 = (Q^ℤ, →) of a CA 𝔄, where Q is a finite set of states, configurations are bi‑infinite words over Q, and → denotes the one‑step “next‑configuration” relation. By viewing each configuration as an ω‑word, they construct a Büchi automaton that recognises the binary relation →: the automaton reads two ω‑words in parallel, locally checks the CA’s transition rule (which depends only on a fixed neighbourhood), and accepts exactly when the second word is the image of the first under the CA rule. This construction proves that 𝒮_𝔄 is an ω‑automatic structure, i.e. its domain and all basic relations are recognised by ω‑automata.

Having established ω‑automaticity, the authors invoke recent results by Kuske and Lohrey on the logical theory of ω‑automatic structures. Those results state that for any ω‑automatic structure, the FO theory extended with counting quantifiers (∃≥k, ∃≤k) and infinite‑cardinality quantifiers (∃^∞, ∀^∞) is decidable. Consequently, any property of the CA phase‑space that can be expressed with these quantifiers—such as “there exist infinitely many cells in state 1”, “exactly three occurrences of a given pattern appear”, or “the number of cells satisfying a predicate is even”—can be algorithmically decided. The paper provides explicit decision procedures based on the underlying Büchi automata and analyses their theoretical complexity.

The authors further specialise to surjective (onto) cellular automata. Surjectivity guarantees that the transition graph has bounded out‑degree, which in turn yields a bounded degree ω‑automatic structure. For such structures, Kuske and Lohrey proved that the extended FO theory can be decided in lower complexity classes (e.g., PSPACE). Leveraging this, the paper presents more efficient algorithms for surjective CA, outlines their implementation details, and demonstrates how the bounded‑degree property leads to practical model‑checking tools for a wide class of dynamical properties.

Finally, the work connects cellular automata to the theory of automatic graphs. Traditional automatic graphs are defined via finite‑word automata; by showing that CA phase‑spaces are ω‑automatic graphs, the authors extend the automatic‑graph framework to infinite‑word settings. This yields new decidability results for graph‑theoretic questions (reachability, connectivity, regular subgraph detection) when the underlying graph originates from a CA. The paper argues that this bridge opens fresh research avenues both in the analysis of dynamical systems and in the study of automatic structures.

In summary, the paper makes four principal contributions: (1) a uniform proof that every one‑dimensional CA yields an ω‑automatic phase‑space; (2) the transfer of Kuske‑Lohrey’s decidability results to this setting, thereby establishing decidability of FO extended with counting and cardinality quantifiers for CA; (3) refined, lower‑complexity decision procedures for surjective CA based on bounded‑degree techniques; and (4) the identification of CA phase‑spaces as a new class of ω‑automatic graphs, providing novel decidability insights for automatic‑graph theory. These results significantly broaden the logical toolkit available for reasoning about cellular automata and related infinite‑state systems.