Arbitrary vs. regular semigroups (Q798783)

This paper builds on the article reviewed above in which it was proved that any semigroup S is the morphic image of an unambiguous semigroup \(\hat S^+_ A\) (meaning that \(\hat S^+_ A\) has an unambiguous \({\mathcal L}\)- and \({\mathcal R}\)-order) and this morphism preserves finiteness and other structural features. The aim of this paper is to show that any unambiguous semigroup S can be embedded in a regular semigroup \((S)_{reg}\) such that finiteness is preserved along with the group and ideal structure. The two papers yield the conclusion that any semigroup divides, in a special way, a regular semigroup with similar ideal structure and the same subgroups. Given S (not necessarily unambiguous) \((S)_{reg}\) is defined by imposing relations on \(S\cup\bar S\cup\{0\}\) (where the members of the union are pairwise disjoint and \(\bar S=\{\bar s|\) \(s\in S\}):\) (1) \(s_ 1s_ 2=s_ 3\) if \(s_ 1\cdot s_ 2=s_ 3\) in S, (2) \(\bar s_ 1\bar s_ 2=\bar s_ 3\) if \(s_ 2\cdot s_ 1=s_ 3\) in \(S_ 1\), (3) \(OO=O\), (4) \(Os=sO=O\) for all \(s\in S\), (5) \(s\bar ss=s\), \(\bar ss\bar s=\bar s\), for any \(s\in S\) (6L) \(s_ 1\bar s_ 2=0\)if \({\mathcal L}_{s_ 1}\) is incomparable to \({\mathcal L}_{s_ 2}\) in the \({\mathcal L}\)-order of S, (6R) \(\bar s_ 1s_ 2=0\) if \({\mathcal R}_{s_ 1}\) is incomparable to \({\mathcal R}_{s_ 2}\) in the \({\mathcal R}\)-order of S. Axioms (6R) and (6L) are crucial in allowing members of \((S)_{reg}\) to be expressed in a (not necessarily unique) normal form. This in turn is used to establish the regularity of \((S)_{reg}\). Unambiguity is important in ensuring that these relations do not introduce any new relations within S (i.e. S is embedded in \((S)_{reg})\). The main theorem then says, in part, for unambiguous S, (a) The \({\mathcal L}\)- (\({\mathcal R}\) or \({\mathcal J})\)-order of \((S)_{reg}\) restricted to S, is the \({\mathcal L}\)- (\({\mathcal R}\) or \({\mathcal J})\)-order of S. (b) Every \({\mathcal D}\)- class (\({\mathcal J}\)-class) of \((S)_{reg}\), except \{0\}, contains one and only one \({\mathcal D}\)-class (\({\mathcal J}\)-class) of S. (c) Every group of \((S)_{reg}\) is a subgroup of S or a Schützenberger group of a non- regular \({\mathcal D}\)-class of S. (d) If every non-regular \({\mathcal H}\)-class of S is a singleton, then every subgroup of \((S)_{reg}\) is isomorphic to a subgroup of S. (e) If S is finite, then \((S)_{reg}\) is finite. If S is finite-\({\mathcal J}\)-above, then \((S)_{reg}\) is finite-\({\mathcal J}\)- above except at zero. If S is infinite then \(| S| =| (S)_{reg}|.\) The concept of unambiguity corresponds in the author's thinking to the ''uniqueness of the past'' in which the semigroup S is envisaged as a semigroup of transformations evolving in time, and indeed the \((- )_{reg}\) construction as applied to the free semigroup \(A^+\) is closely related to so-called two-way automata (see \textit{S. Eilenberg}, Automata, Languages and Machines, Vol. A (1974; Zbl 0317.94045), p. 282ff).