On condensed forms for partially commuting matrices (Q1971044)

Let \(A\) and \(B\) be \(n\times n\) matrices over \(\mathbb{C}\). The authors call \(A\) and \(B\) partially commuting if they have a common eigenvector; equivalently, there is a nonzero subspace of \(\mathbb{C}^{n}\) which is invariant under \(A\) and \(B\) and on which these matrices commute. The main theorem (Theorem 3) describes a reduction (called the condensed form) of a pair of partially commuting matrices \(A\) and \(B\). It states that there exists an invertible matrix \(P\) such that \(P^{-1}AP\) and \(P^{-1}BP\) are both in block upper trianglar form with, say \(k\), diagonal blocks where each of the first \(k-1\) diagonal blocks of \(P^{-1}AP\) commutes with the corresponding block of \(P^{-1}BP\) and the final pair of blocks either commute or are not partially commuting. Moreover, the integer \(k\) and the dimensions of the blocks are uniquely determined by \(A\) and \(B\), and the matrix \(P\) can be computed rationally (by field operations) from \(A\) and \(B\). Refinements of this result are given under various additional hypotheses. For example, when \(AB-BA\) has rank \(1\) the triangular matrices in the condensed form have two diagonal blocks with both pairs of diagonal blocks commuting, and this condensed form can be used to bound the dimension of the algebra generated by \(A\) and \(B\) [compare with \textit{T. J. Laffey}, Linear Multilinear Algebra 6, 269-306 (1978; Zbl 0399.15008)].