A generalization of a theorem of von Neumann (Q7014469)

A well-known result by J. von Neumann states that for every \(n\times n\) doubly substochastic matrix \(A= [a_{i,j}]\) there exists a doubly stochastic matrix \(D=[d_{i,j}]\) such that \(a_{i,j} \leq d_{i,j}, i, j= 1, \cdots, n, \) see [\textit{A. W. Marshall} et al., Inequalities: theory of majorization and its applications. 2nd edition. New York, NY: Springer (2011; Zbl 1219.26003)]. This fact is denoted by \(A\leq D.\) In such a case \(A\) is said to be increasable. Notice that the matrix \(D\) is not necessarily unique. \N\NOn the other hand, if \(I \subset \mathbb{N}\) and \(\mathcal{A}\) is a non-empty set of doubly substochastic matrices of size \(I\times I,\) then a stochastic matrix \(D\) of size \(I\times I\) is said to be a cover of \(\mathcal{A}\) if \(A\leq D\) for every \(A\in \mathcal{A}.\) The matrix \(U= [u_{i.j}]\) such that \(u_{i,j}=\sup_{A\in \mathcal{A}} a_{i,j}\) is denoted by \(\sup(\mathcal{A}).\) Notice that, in general, \(\sup(\mathcal{A})\) is not a doubly substochastic matrix.\N\NIn the paper under review it is proved that a non-empty set \(\mathcal{A}\) of doubly substochastic matrices of size \(I\times I\) has a cover if and only if \(\sup(\mathcal{A})\) is an increasable doubly substochastic matrix. Moreover, necessary and sufficient conditions for the existence of an infinite number of covers are given.In particular, if \(I\) is a finite set, then there exist infinitely many covers if and only if there are at least two rows and columns such that the summations of their respective entries are less than 1.\N\NA characterization of the uniqueness of a cover of \(\mathcal{A}\) is also presented. In particular, if \(I\) is a finite set, then \(\mathcal{A}\) has a cover and it is unique if and only if \(\sup(\mathcal{A})\) is doubly substochastic matrix and it has at most one row (or column) such that the sum of its entries is less than 1.\N\NFinally, an approach to the cover problem in terms of a combined linear system of equations and inequalities is analyzed.