Computable analysis on the space of marked groups (2111.01179v5)

Abstract: We investigate decision problems for finitely generated groups described by word problem algorithms. This is equivalent to studying groups described by computable labelled Cayley graphs. We relate these problems to the study of computable analysis on the space of marked groups, and point out several results of computable analysis that can be directly applied to obtain group theoretical results. Those results, used in conjunction with the version of Higman's Embedding Theorem that preserves solvability of the word problem, provide powerful tools to build finitely presented groups with solvable word problem but with various undecidable properties. Interestingly, the only results of computable analysis that can be applied together with Higman's Embedding Theorem are those of what is known as ``Markov's school of computable analysis''. We thus focus on articles which predate most of what is seen as the modern corpus of research in computable analysis. The presented results show that these older results are still very relevant. We also place group properties in the first levels of the arithmetical hierarchy for decision problems about groups given by word problem algorithms, and show that on many group properties usually considered, this effective hierarchy corresponds sharply to the Borel hierarchy. Finally, we prove that the space of marked groups is a Polish space that is not \emph{effectively Polish}. Because of this, many of the most important results of Markovian computable analysis cannot be applied to the space of marked groups. This includes the Kreisel-Lacombe-Schoenfield-Ceitin Theorem and a theorem of Moschovakis. The space of marked groups constitutes the first natural example of a Polish space that is not effectively Polish.