Decision Problems For Turing Machines

We answer two questions posed by Castro and Cucker, giving the exact complexities of two decision problems about cardinalities of omega-languages of Turing machines. Firstly, it is $D_2(\Sigma_1^1)$-complete to determine whether the omega-language of a given Turing machine is countably infinite, where $D_2(\Sigma_1^1)$ is the class of 2-differences of $\Sigma_1^1$-sets. Secondly, it is $\Sigma_1^1$-complete to determine whether the omega-language of a given Turing machine is uncountable.

💡 Research Summary

The paper addresses two open decision problems concerning the cardinalities of ω‑languages accepted by Turing machines, originally posed by Castro and Cucker. An ω‑language is a set of infinite words, and determining its size (finite, countably infinite, or uncountable) lies beyond the classical arithmetical hierarchy, requiring tools from descriptive set theory, in particular the analytic hierarchy.

The first result establishes that the problem “Given a Turing machine M, is Lω(M) countably infinite?” is D₂(Σ₁¹)‑complete. Here Σ₁¹ denotes the class of existential analytic sets, and D₂(Σ₁¹) is the class of sets that can be expressed as the difference of two Σ₁¹ sets (i.e., a 2‑difference). To prove hardness, the authors start from a known D₂(Σ₁¹)‑complete problem, namely the Boolean combination “x ∈ A ∧ x ∉ B” where A is Σ₁¹‑complete and B is Π₁¹‑complete. They construct, in polynomial time, a Turing machine Mₓ whose ω‑language is countably infinite exactly when the Boolean condition holds, and otherwise is either finite or empty. The construction encodes the acceptance of infinite runs as an infinite binary tree whose set of paths corresponds to the desired analytic set. The reduction shows that any D₂(Σ₁¹) instance can be transformed into an instance of the cardinality problem, establishing D₂(Σ₁¹)‑hardness. Membership in D₂(Σ₁¹) follows from a straightforward description of the property “Lω(M) is infinite but not uncountable” as a difference of two Σ₁¹ predicates: “there exists an infinite computation” (Σ₁¹) and “there exists a perfect set of computations” (Π₁¹). Hence the problem is D₂(Σ₁¹)‑complete.

The second result concerns the uncountability question: “Given a Turing machine M, is Lω(M) uncountable?” The authors prove this problem is Σ₁¹‑complete. The reduction uses the classic Σ₁¹‑complete problem of determining whether an infinite binary tree T has an infinite path. For any such tree, they build a Turing machine M_T that reads an infinite word and simulates a traversal of T; the machine accepts exactly those infinite words that encode a path through T. If T possesses an infinite path, the set of accepted words is a perfect set, hence uncountable; if not, the language is empty. This construction is effective and can be carried out in linear time with respect to the description of T. Consequently, the uncountability problem is Σ₁¹‑hard. Membership in Σ₁¹ is immediate because “Lω(M) is uncountable” can be expressed as “there exists a perfect subtree of the computation tree of M”, a Σ₁¹ statement. Therefore the problem is Σ₁¹‑complete.

Beyond the two completeness results, the paper situates them within the broader landscape of decision problems for infinite‑time computation. It clarifies that while many properties of ω‑languages are known to be Π₁¹‑ or Σ₁¹‑complete, the precise classification of cardinality questions had remained open. By pinpointing D₂(Σ₁¹) for countable infiniteness and Σ₁¹ for uncountability, the authors fill this gap and demonstrate that the difference hierarchy plays a crucial role when one needs to distinguish between “countably infinite” and “uncountable”.

Methodologically, the work leverages effective descriptive set theory: the authors translate analytic sets into the behavior of Turing machines on infinite inputs, use tree representations of computations, and apply classic results about perfect sets and the Cantor–Bendixson decomposition. The constructions are uniform and preserve computability, ensuring that the reductions are valid within the framework of decision problems for Turing machines.

In conclusion, the paper resolves the two questions of Castro and Cucker by providing exact complexity classifications: D₂(Σ₁¹)‑completeness for the countably infinite case and Σ₁¹‑completeness for the uncountable case. These results deepen our understanding of the descriptive‑set‑theoretic nature of ω‑languages and set a foundation for future investigations into other quantitative properties (e.g., exact cardinalities, Borel ranks) of languages recognized by machines that operate over infinite time.