Nonnegative matrix semigroups with finite diagonals (Q627943)

If \(\mathcal G\) is an irreducible group of complex matrices (meaning that there is no nontrivial subspace which is simultaneously invariant under all the members of \(\mathcal G\), considered as linear operators) and if \(\{\mathrm{tr}\, A: A \in \mathcal G\}\) is a finite set, then \(\mathcal G\) itself is finite. \textit{J. Okniński} [Proposition 4.9, Semigroups of matrices. Singapore: World Scientific (1998; Zbl 0911.20042)] generalized this to semigroups of matrices. This was further extended by \textit{H. Radjavi} and \textit{P. Rosenthal} [Semigroup Forum 76, No.~1, 25--31 (2008; Zbl 1140.15014)], where it was demonstrated that the result holds if the trace functional is replaced by any linear functional. The objects of interest in the paper under review are semigroups of matrices with nonnegative entries. Here, irreducibility should be replaced by indecomposability. In this framework, recently it was proved that if \(\mathcal S\) is an indecomposable semigroup of nonnegative matrices and if for some nonzero positive linear functional \(\phi\), the set \(\{\phi(S): S \in \mathcal S\}\) is bounded, then \(\mathcal S\) itself is bounded [\textit{H. Gessesse, A. I. Popov, H. Radjavi, E. Spinu, A. Tcaciuc} and \textit{V. G. Troitsky}, Proposition 8, Positivity 14, No.~3, 383--394 (2010; Zbl 1204.15041)]. The authors of the paper under review consider the following question: If all diagonal entries in an indecomposable nonnegative semigroup come from a fixed finite set, is the semigroup itself finite? In spite of the answer for the general question being in the negative, in this interesting paper, the authors identify two important cases under which they obtain affirmative results: that of a self-adjoint semigroup (\(\mathcal S\) self-adjoint means that the implication \(S \in \mathcal S \Rightarrow S^* \in \mathcal S\) holds) and that of constant-rank semigroups.