Monotonicity, path product matrices, and principal submatrices (Q2310402)

Let \(A=[a_{ij}]\) be a real \(n\times n\) matrix. Then \(A\) is called monotone if \(A\) is invertible and \(A^{-1}\geq0\) (entrywise); equivalently, for real \(n\)-vectors, \(Ax_{2}\geq Ax_{1}\) implies \(x_{2}\geq x_{1}\). Beginning from [\textit{A. V. Gavrilov}, Comput. Math. Math. Phys. 41, No. 9, 1237--1238 (2001; Zbl 1024.15009); translation from Zh. Vychisl. Mat. Mat. Fiz. 41, No. 9, 1301--1302 (2001)], the authors of the present paper investigate monotonicity relationships between a matrix and its principal submatrices. The following classes of matrices are considered: (i) \(A\) is a \(P\)-matrix if all of its principal minors are positive; (ii) \(A\) is an \(M\)-matrix if \(A\) is monotone with entries on the diagonal \(>0\) and all other entries \(\leq0\); (iii) \(A\) is a path product matrix (\(PP\)) if \(A\geq0\) with all \(a_{ii}>0\) and (*) \(a_{ij}a_{kk}\geq a_{ik}a_{kj}\) for all \(i,j\) and \(k\); (iv) for each \(m\) with \(2\leq m<n\) the matrix \(A\) is called \(m\)-monotone if every \(m\times m\) principal submatrix is monotone. It is known that the inverse of an \(M\)-matrix is always a \(PP\)-matrix but not conversely, and it can be shown that if \(A\) is a \(P\)-matrix such that (*) holds then \(A\) is an invertible \(PP\)-matrix. The main theorems of this paper are the following ones: Theorem. If \(A\) is an \(m\)-monotone \(P\)-matrix for some \(m\) with \(2\leq m\leq n\), then \(A\) is monotone. Theorem. If \(A\) is invertible then \(A\) is doubly monotone (both monotone and \((n-1)\)-monotone) if and only if \(A^{-1}\) is a \(PP\)-matrix. Connections among the various classes of matrices listed above follow as corollaries to these theorems.