A ``shift-and-deflate'' technique for quadratic matrix polynomials (Q1940341)

The author proposes, analyzes and implements a two stage (shift and deflate) algorithm to solve the following problem. Given two eigenvalues \(\lambda _1\) and \(\lambda _2\) of an \(n\times n\) quadratic matrix polynomial \(A(z)\) and a right eigenvector corresponding to \(\lambda _1\) and a left eigenvector corresponding to \(\lambda _2\), find an \((n-1)\times (n-1)\) quadratic matrix polynomial \(B(z)\) whose eigenvalues are all those of \(A(z)\) except \(\lambda _1\) and \(\lambda _2\). When A(z) is \(*-\)palindromic, the method can ensure that \(B(z)\) shares this property. Some open questions are also mentioned.