Bipartite graphs and completely 0-simple semigroups. (Q415549): Difference between revisions

With any completely 0-simple semigroup \(S\), represented as a Rees matrix semigroup \(\mathcal M^0(I,G,\Lambda,P)\), the author associates the bipartite graph \(\Gamma(S)\), the vertex set of which is \(I\cup\Lambda\), the edge set consisting of the pairs \((i,\lambda)\) for which \(p_{\lambda i}\neq 0\). Conversely, with any bipartite graph \(\Gamma=I\cup\Lambda\) he associates the `sandwich semigroup' \(\Pi_1(\Gamma,I,\Lambda)^0\), the nonzero elements of which are paths in \(\Gamma\) that start in \(I\) and end in \(\Lambda\). It is a 0-direct sum of idempotent generated completely 0-simple semigroups, corresponding to the connected components of \(\Gamma\); its nonzero \(\mathcal H\)-classes are the fundamental groups of those components. Starting from \(S\), as above, one obtains a natural covering of the idempotent generated part of \(S\) in this fashion; starting with \(\Gamma\) a complete bipartite graph, one obtains a free idempotent generated completely simple semigroup. In the second part of the paper, this construction is extended to a more general context. Use is made of a broad generalization of \textit{R. L. Graham}'s normalization [Math. Syst. Theory 2, 325-339 (1968; Zbl 0177.03103)] of the matrix \(P\) of a Rees matrix semigroup. The author has used these constructions to study free objects in Rees-Sushkevich varieties [``Regular principal factors in free objects in Rees-Sushkevich varieties'', Semigroup Forum 86, No. 1, 162-182 (2013), \url{DOI: 10.1007/s00233-012-9398-y}].