On Minimally Non-firm Binary Matrices

Abstract

For a binary matrix ${\textbf {X}}$, the Boolean rank $br({\textbf {X}})$ is the smallest integer k for which ${\textbf {X}}$ equals the Boolean sum of k rank-1 binary matrices, and the isolation number $i({\textbf {X}})$ is the maximum number of 1s no two of which are in a same row, column and a $2\times 2$ submatrix of all 1s. In this paper, we continue Lubiw’s study of firm matrices. ${\textbf {X}}$ is said to be firm if $i({\textbf {X}})=br({\textbf {X}})$ and this equality holds for all its submatrices. We show that the stronger concept of superfirmness of ${\textbf {X}}$ is equivalent to having no odd holes in the rectangle cover graph of ${\textbf {X}}$, the graph in which $br({\textbf {X}})$ and $i({\textbf {X}})$ translate to the clique cover and the independence number, respectively. A binary matrix is minimally non-firm if it is not firm but all of its proper submatrices are. We introduce two matrix operations that lead to generalised binary matrices and use these operations to derive four infinite classes of minimally non-firm matrices. We hope that our work may pave the way towards a complete characterisation of firm matrices via forbidden submatrices.

Supported by The Alan Turing Institute, London, UK