On aperiodic relational morphisms. (Q2505011)

This paper provides a refreshing new perspective of various attempts of the school of John Rhodes to prove that there is an algorithm to compute the so-called group complexity of a finite semigroup. The unifying starting point is a simple and well-known aperiodicity lemma, which characterizes aperiodic relational morphisms, that is relational morphisms between finite semigroups for which the inverse images of aperiodic semigroups are also aperiodic. Among the many results in the paper, of which some are too technical to state here in detail, the following deserve special mention. The following notation for pseudovarieties of finite semigroups is convenient for their statement: \(\mathbf G\) for groups, \(\mathbf H\) for an arbitrary subpseudovariety of \(\mathbf G\), \(\mathbf G_p\) for \(p\)-groups, \(\mathbf A\) for aperiodic semigroups, \(\mathbf{Sl}\) for semilattices, \(\mathbf D\) for semigroups in which idempotents are right zeros, and \(\mathbf{LV}\) for the pseudovariety of all semigroups whose submonoids belong to the pseudovariety \(\mathbf V\). (1)~A regular finite semigroup \(S\) belongs to the join \(\mathbf A\vee\mathbf H\) if and only if \(S\) belongs to the semidirect product \(\mathbf A*\mathbf H\) and Green's relation \(\mathcal H\) is a congruence on~\(S\). (2)~If \(\mathbf H\) is a non-trivial proper subpseudovariety of \(\mathbf G\), then there is some monoid of partial permutations in \((\mathbf A*\mathbf G)\setminus(\mathbf A*\mathbf H)\) whose subgroups belong to~\(\mathbf H\). (3)~If \(\mathbf G_p*\mathbf H=\mathbf H\) then membership of finite semigroups in \(\mathbf A*\mathbf H\) is decidable if and only if membership of finite inverse monoids in \(\mathbf{Sl}*\mathbf H\) is decidable. (4)~A coordinate-free improved version of the Presentation Lemma of \textit{J. Rhodes} [J. Pure Appl. Algebra 101, 245-289 (1995; Zbl 0836.20081)]. (5)~An application of (4) to the characterization of finite semigroups with at most two \(\mathcal J\)-classes of complexity~2, thus providing a new proof of a result of \textit{B. Tilson} [Adv. Math. 11, 157-266 (1973; Zbl 0269.20047)]. (6)~A simplified, coordinate-free, proof of J. Rhodes' result [see first citation above] that the complexity 1 pseudovariety \(\mathbf A*\mathbf G*\mathbf A\) is not local, neither in Eilenberg's sense that it satisfies the equation \(\mathbf{LV}=\mathbf V\), nor in Tilson's sense that it is a solution of the equation \(\mathbf{LV}=\mathbf V*\mathbf D\).