Combinatorial aspects of generalized complementary basic matrices (Q2440532): Difference between revisions

This paper deals with combinatorial aspects of some properties of the generalized complementary basic matrices. These include inheritance for alternating sign matrices, spectral and sign nonsingularity properties. Let \(A_1, A_2, \dots, A_s\) be square matrices of respective size \(k_1, k_2, \dots, k_s\), \(k_i \geq 2\) for all \(i\). The authors denote \(n=\sum_{i=1}^s k_i-s+1\), and construct the block diagonal matrices \(G_1, G_2, \dots, G_s\) as follows: \[ G_1=\left[ \begin{matrix} A_1 & 0 \\ 0 & I_{n-k_1} \end{matrix} \right], \qquad G_2= \left[ \begin{matrix} I_{k_1-1} & 0 & 0 \\ 0 & A_2 & 0 \\ 0 & 0 & I_{n-k_1-k_2+1} \end{matrix} \right], \quad \cdots \] \[ G_{s-1}= \left[ \begin{matrix} I_{n-k_{s-1}-k_s+1} & 0 & 0 \\ 0 & A_{s-1} & 0 \\ 0 & 0 & I_{k_s-1} \end{matrix} \right], \quad G_s=\left[ \begin{matrix} I_{n-k_s} & 0 \\ 0 & A_s \end{matrix} \right]. \] Then, for any permutation \((i_1,i_2,\dots,i_s)\) of \((1,2,\dots,s)\), they consider the product \(\prod G_k:=G_{i_1}G_{i_2} \cdots G_{i_s}\). These products are called generalized complementary basic matrices (\(GCB\)-matrices), the diagonal blocks \(A_k\) are called distinguished blocks and \(G_k\) generators of \(\prod G_k\). In this paper, the authors extend some properties of the \(GCB\)-matrices, emphasizing the case of two generators since in most cases it allows to continue to the general case. In a first result, they prove that if each \(A_k\) in a \(GCB\)-matrix is a \(P\)-matrix, then \(\prod G_k\) is a \(P\)-matrix. On the other hand, an alternating sign matrix (\(ASM\)) is an \(n \times n\) \((0,+1,-1)\)-matrix without zero rows and columns, such that the +1s and -1s alternate in each row and column, beginning and ending with a +1. In the context of \(GCB\)-matrices, the authors prove that if each distinguished block \(A_k\) is an \(ASM\), then \(\prod G_k\) is an \(ASM\). An \(n \times n\) sign pattern \(A\) is said to be sign nonsingular (\(SNS\)) if every matrix \(B \in Q(A)=\{ B: \operatorname{sgn}(B)=A \}\) is nonsingular. Analogously, \(A\) is said to be strongly sign nonsingular (\(S^2NS\)), if \(A\) is \(SNS\) and for all \(B \in Q(A)\), \(B^{-1}\) is in the same sign pattern class. A square sign pattern \(A\) that does not have a zero row or column is sign potentially orthogonal (\(SPO\)) if every pair of rows and every pair of columns allows orthogonality. The authors prove different properties of these concepts related to \(GCB\)-matrices. Finally, some spectral properties, irreducibility, positive definite, \(\dots\), of \(GCB\)-matrices are discussed.