Rectangular \(L\)-matrices (Q1316185)

An \(L\)-matrix is an \(m\times m\) (0,1,--1)-matrix \(A\) such that every \(m \times n\) real matrix with the same sign pattern as \(A\) has linearly independent rows. \(A\) is called \(L\)-decomposable if there are permutation matrices \(P,Q\) such that \(A=P{X_ 1 \brack Y X_ 2} Q\) where \(X_ 1,X_ 2\) are nonvacuous \(L\)-matrices. The purpose of this paper is to introduce a totally \(L\)-matrix (i.e. an \(m\times n\) \(L\)-matrix provided every submatrix of \(A\) of order \(m\) is a square \(L\)-matrix) and a barely \(L\)-matrix (i.e. an \(m \times n\) \(L\)-matrix provided each of its \(m \times (n-1)\) submatrices is not an \(L\)-matrix), and then to show some results on these two classes mainly: (i) An \(L\)-matrix is an \(L\)-indecomposable, barely \(L\)-matrix if and only if for each column \(i\) there is a strictsigning \(DA\) of \(A\) whose only unisigned column is column \(i\). (ii) An \(m \times n\) totally \(L\)-matrix with \(m \geq 2\) satisfies \(n \leq m+2\). (iii) For \(m\geq 2\), there are three classes of \(m\times n\) totally \(L\)-matrices, i.e. square matrices, \(S^*\)-matrices, and totally \(L\)-matrices with \(n=m+2\). (iv) In contrast to general \(L\)-matrices, a barely \(L\)-matrix has a unique set of \(L\)- indecomposable components.