Deflation for block eigenvalues of block partitioned matrices with an application to matrix polynomials of commuting matrices (Q5948892)

The most immediate generalization of complex polynomials are matrix polynomials where the coefficients and the variable are commuting matrices over \(\mathbb{C}\). In the paper a numerical method of solving the related generalized polynomial algebraic equation is presented. It uses, basically, a block version of the Wielandt deflation procedure, applied to the the block companion matrix of the generalized polynomial in question. The procedure preserves a certain L-shaped structure of the sequence of deflated matrices. In essence, this algorithm extends the computation of a matrix root (called ``solvent'') presented by \textit{J. E. Dennis, J. P. Traub} and \textit{R. P. Weber} [On the matrix polynomial lambda-matrix and block eigenvalue problems, Tech. Rep. 71-109, Computer Science Department, Cornell Univ., Ithaca, NY and Carnegie-Mellon Univ., Pittsburgh, PA (1971)] by providing all of them.