Positive, path product, and inverse \(M\)-matrices (Q869901)

Inverses of \(M\)-matrices (called \(IM\)-matrices) have the following property, first noted by Willoughby and developed further by the authors: an \(n\times n\) \(IM\)-matrix \(A= (a_{ij})\) with \(n\geq 3\) satisfies \((a_{ij} a_{jk})(a_{jj})^{-1}\leq a_{ik}\) for all distinct indices \(i\), \(j\), \(k\) with \(1\leq i\), \(j,k\leq n\). These are called the path product (PP) conditions, and a nonnegative matrix with positive diagonal entries which satisfies these is called a PP-matrix. If there is strict inequality whenever \(i=k\), then it is an SPP-matrix. The authors noted in an earlier paper that \(IM\) implies SPP. The reverse holds for \(n\leq 3\), but not necessarily for larger \(n\). In this paper the authors show that any square, positive matrix can be made SPP by predictable additions to the diagonal and that any (normalized) SPP-matrix can be made \(IM\) by additions to the diagonal that are bounded in terms of \(n\). A determinantal inequality involving principal minors is derived for normalized \(IM\)-matrices.