Interval hulls of \(N\)-matrices and almost \(P\)-matrices (Q2656637)

If \(A=[a_{ij}]\) and \(B=[b_{ij}]\) lie in \(\mathbb{R}^{n\times n}\) then the interval hull \(\mathbb{I}(A,B) \) consists of all \(C\in \mathbb{R}^{n\times n} \) whose \((i,j)\)th entry lies between \(a_{ij}\) and \(b_{ij}\) for each \((i,j) \). A matrix \(A\) is called an \(N\)-matrix if all its principal minors are negative. Similarly, \(A\) is called a \(P\)-matrix if its principal minors are positive; it is an almost \(P\)-matrix if its proper principal minors are positive, but \(\det A\) itself is negative. These classes of matrices can be characterized by sign non-reversal properties: \(A\) is a \(P\)-matrix if and only if for a vector \(x\) the product of the \(i\)th components \( x_{i}(Ax)_{i}\leq 0\) for all \(i\) implies \(x=0\); and \(A\) is an \(N\)-matrix with \(A<0\) if and only if \(x_{i}(Ax)_{i}\leq 0\) for all \(i\) implies \(x\geq 0\) or \(x\leq 0\) (see [\textit{D. Gale} and \textit{H. Nikaidô}, Math. Ann. 159, 81--93 (1965; Zbl 0158.04903)] and [\textit{T. Parthasarathy} and \textit{G. Ravindran}, Linear Algebra Appl. 139, 89--102 (1990; Zbl 0719.15010)]). The main object of the current paper is to give criteria for \(\mathbb{I} (A,B)\) consisting entirely of one of these classes. A result is the following. For any \(A,B\in \mathbb{R}^{n\times n}\) and any diagonal matrix \( D_{z}\) whose diagonal \(z\) consists of \(\pm 1\) entries, define \(I_{z}:=\frac{1 }{2}(A+B)-\frac{1}{2}D_{z}\left\vert A-B\right\vert D_{z}\) (the entries of \( \left\vert A-B\right\vert \) are the absolute values of the entries of \(A-B\) ). According to Theorem 3.6 in the paper, if the diagonal entries of \(A\) and \(B\) are strictly negative and each \(I_{z}\) with \(D_{z}\neq \pm I\) is an \(N\)-matrix with all entries being negative, then each matrix in \(\mathbb{I}(A,B)\) is an \(N\)-matrix. The sign non-reversal properties play a critical part in the proof.