Almost perfect restriction semigroups (Q891457)

A restriction semigroup \(S\) is a semigroup with two unary operations \(^+\) and \(^\ast\) that satisfy the identities \[ x=x^+x,\quad (x^+y)^+=x^+y^+=y^+x^+,\quad xy^+=(xy)^+x \] and the duals for \(^\ast\), together with \[ (x^+)^\ast=x^+\text{ and }(x^\ast)^+=x^\ast. \] This class of semigroups has been object of study by various authors in different contexts. The author's main aim is a ``McAlister-type'' theorem. An important figure of this paper is that the author takes great care to tie his results to results by other researchers on the field, such as \textit{C. Cornock} and \textit{V. Gould} [J. Pure Appl. Algebra 216, No. 4, 935--949 (2012; Zbl 1258.20047)], \textit{J. Fountain} and \textit{G. M. S. Gomes} [Glasg. Math. J. 35, No. 3, 293--306 (1993; Zbl 0802.20051)], \textit{M. B. Szendrei} [Commun. Algebra 41, No. 4, 1458--1483 (2013; Zbl 1275.20064)] or \textit{G. Kudryavtseva} [J. Algebra 429, 342--370 (2015; Zbl 1321.20052)]. Let \(P_S:=\{a^+:a\in S\}=\{a^\ast:a\in S\}\). The least congruence on \(S\) that identifies the elements of \(P_S\) is denoted by \(\sigma\). The semigroup \(S\) is said to be \textit{perfect} if the set product \(a\sigma .b\sigma\) equals \((ab)\sigma\), for any \(a,b\in S\). When \(a\sigma b\) together with \(a^+=b^+\) imply \(a=b\) and, dually, if (\(a\sigma b\) and \(a^\ast=b^\ast\)) then \(a=b\), the semigroup \(S\) is called \textit{proper}. The author introduces the notion of an almost perfect restriction semigroup: one that is both perfect and proper. The McAlister-type theory that is developed concerns not the proper case, as studied by others, but the almost perfect case. As usual, an almost perfect restriction semigroup \(\hat S\) is said to be a \textit{cover} of \(S\) if there is a \((2,1,1)\)-morphism from \(\hat S\) onto \(S\) that separates the elements of \(P\hat S\). The monoid case is also discussed. A restriction monoid \(M\) is said to be \textit{perfect} if each \(\sigma\)-class has a greatest element for the partial order \(\geq\), which is defined by \(a\geq b\) if and only if \(a=eb\), for some \(e\in P_M\). Observe that in a monoid one requires \(1^\ast=1^+=1\). The definition of a \(W\)-semidirect product \(W(T,Y)\) applied here is inspired by that introduced by Fountain and Gomes [loc. cit.] and later developed by Szendrei [loc. cit.]. Here, \(Y\) is a semilattice acted upon by a monoid \(T\) (on the right) by isomorphisms between ideals of \(Y\). So, for any \(t\in T\), there is an isomorphism \[ t:\Delta t\to\nabla t,\quad e\mapsto e^t,\text{ where }\Delta t\text{ and }\nabla t\text{ are ideals of }Y. \] Let \[ W(T, Y):=\{(t,e^t)\in T\times Y:e\in\Delta t\}. \] With the operations defined by \[ \begin{aligned} (t,e^t)(u,f^u)& =(tu,(e^t f)^u),\\ (t,e^t)^+ & =(1,e),\\ (t, e^t)^\ast & =(1,e^t),\end{aligned} \] the set \(W(T,Y)\) becomes an almost perfect restriction semigroup. It is proved that any restriction semigroup \(S\) has an almost perfect restriction cover \(\hat S\) and that every almost perfect restriction semigroup is \((2,1,1)\)-isomorphic to a \(W\)-semidirect product \(W(T,Y)\). The monoid case is also discussed. As stated by the author, the main results are described in Theorem 1.1, however much more can be found in this paper. We point out some references, given by the author, that are most related to this review.