An integer programming problem and rank decomposition of block upper triangular matrices (Q1968759)

The background for this paper is formed by articles which deal with sums of idempotent elements in Banach algebras. For the general background the immediate motivation for this paper comes from a paper by \textit{H. Bart}, \textit{T. Ehrhardt} and \textit{B. Silbermann} [Sums of idempotents and logarithmic residues in matrix algebras. Operator Theory: Advances and Applications (to appear)] where logarithmic residues of analytic matrix valued functions are identified as sums of idempotent matrices. The rank decomposition theorem is the topic of the present paper. A necessary and sufficient condition is given for blank upper rectangular matrices satisfying certain rank constraints. The condition is formulated in terms of the ranks of certain submatrices of \(A\). The proof goes by reduction to an integer programming problem. This integer programming problem has a totally unimodular constraint matrix which makes it possible to utilize Farkas' lemma. In describing the contents of the paper, the authors outline the main steps. In Section 2, they use column operations to bring \(A\) into a simple ``block column reduced'' form. The process resembles that of reducing a matrix to column echelon form. In Section 3, the problem of decomposing \(A\) (assumed to be in ``block column form'') is reduced to an integer programming problem involving \({1\over 2} k (k+1)\) equations and \({1\over 6} k (k+1)(k+ 2)\) unknowns where \(A\) can be partitioned as a block matrix \(A= (A_{ij})^k_{i,j= 1}\).