Stability of matrix polynomials in one and several variables (Q6040647)

The authors consider matrix polynomials \[ P(\lambda)=\lambda^dA_d+\lambda^{d-1}A_{d-1}+\cdots+A_0 \] and the problem of localisation of their eigenvalues, i.e., zeros of the function \(\lambda\mapsto\det P(\lambda)\). If a polynomial \(P(\lambda)\) has no eigenvalues in a set \(D\), then it is said to be \emph{stable with respect to \(D\)}. The issue of stability of matrix polynomials is widely studied. In the present paper the authors define a ``relaxed'' condition similar to stability. Namely, a matrix polynomial \(P(\lambda)\) is called \textit{hyperstable with respect to set \(D\)} if for all \(x\in\mathbb C^n\setminus\left\{0\right\}\) there exists \(y\in\mathbb C^n\setminus\left\{0\right\}\) such that for all \(\mu\in D:\; y^*P(\mu)x\neq 0\). The authors present some properties connected with this notion. In particular, if \(D\) is an open or closed subset of the complex plane and \(P(\lambda)\) is either strictly equivalent to an upper-triangular matrix polynomial or of the form \(p(\lambda)A+q(\lambda)B\), where \(p\), \(q\) are scalar polynomials, then \(P(\lambda)\) is hyperstable with respect to \(D\) if and only if it stable with respect to \(D\). Another important result of this article is a variation of the classical Gauss-Lucas theorem: Theorem. If \(D\subseteq\mathbb C\) is a nonempty open or closed set such that \(\mathbb C\setminus D\) is convex, \(P(\lambda)\) is a matrix polynomial that is hyperstable with respect to \(D\) andthe entries of \(P'(\lambda)\) are linearly independent polynomials, then \(P'(\lambda)\) is also hyperstable with respect to \(D\). The last part of the work focuses on the multivariate matrix polynomials. The authors extend the definitions of stability and hyperstability for this case.