Inequalities for \(M\)-matrices and inverse \(M\)-matrices (Q996314): Difference between revisions

An \(n\times n\) real matrix \(A\) with nonpositive off-diagonal entries is called an \(M\)-matrix if it is nonsingular and its inverse is nonnegative. The matrix \(A\) is said to be an inverse \(M\)-matrix if its inverse is an \(M\)-matrix. \textit{T. Ando} [Linear Multilinear Algebra 8, 291--316 (1979; Zbl 0452.15020)] has shown that if \(A=(a_{ij})\), \(B=(b_{ij})\) are \(M\)-matrices of order \(n\), then \[ \det(A\circ B)+\det A \cdot\det B \geq(\det A) \prod_{i=1}^nb_{ii} + (\det B) \prod_{i=1}^na_{ii}, \] where \(\circ\) denotes the Hadamard product. This inequality has been strengthened by the author in [Linear Algebra Appl. 368, 99--106 (2003; Zbl 1026.15014)]. In the paper under review the author extends his inequality to the case where one of the matrices is allowed to be an inverse \(M\)-matrix. Namely, the author proves that if \(A\) is an \(M\)-matrix, \(B\) an inverse \(M\)-matrix, and \(i_1,\ldots,i_n\) is any permutation of \(\{1,2,\ldots,n\}\), then \[ \det(A\circ B) \geq(\det AB) \prod_{s=2}^n\left(\frac{a_{i_si_s} \det A[i_1,i_2,\ldots,i_{s-1}]} {\det A[i_1,i_2,\ldots,i_{s-1},i_s]}+ \frac{b_{i_si_s} \det B[i_1,i_2,\ldots,i_{s-1}]} {\det B[i_1,i_2,\ldots,i_{s-1},i_s]} -1\right), \] where for any set \(\alpha\), \(A[\alpha]\) denotes the principal submatrix of \(A\) in the rows and columns indexed by the elements of \(\alpha\). The author also shows that if \(A\) is an \(n\times n\) \(M\)-matrix or inverse \(M\)-matrix, then \(\det A=\prod a_{ii}\) if and only if \(A\) is essentially triangular; and that if \(i_1,\ldots,i_n\) is any permutation of \(\{1,2,\ldots,n\}\), then \[ \det A\leq(\prod_{i=1}^n a_{ii}) \prod_{s=2}^n\left(1-\frac{| a_{i_1i_2}\cdots a_{i_{s-1}i_{s}}a_{i_si_1}| } {a_{i_1i_1}a_{i_2i_2}\cdots a_{i_si_s}}\right). \]