Pseudospectra, critical points and multiple eigenvalues of matrix polynomials (Q999799)

Let \(\mathbb{C}^{n\times n}\) be the space of \(n\times n\) matrices over \(\mathbb{C}\), the field of complex numbers. Given a norm \(|| . ||\) on \(\mathbb{C}^{n\times n}\), the \(\epsilon\)-pseudospectrum \(\Lambda _\epsilon (A)\) of an \(A\in \mathbb{C}^{n\times n}\) is defined by \(\Lambda _\epsilon (A):=\bigcup_{||\Delta A||\leq \epsilon}\Lambda (A+\Delta A)\) where \(\Lambda (A):=\{\lambda \in \mathbb{C}:\text{rank} (A-\lambda I)<n\}\) is the spectrum of \(A\). Let \(\mathbb{L}_m(\mathbb{C}^{n\times n})\) be the vector space of \(n\times n\) matrix polynomials of degree at most \(m\). Then for a given norm \(||| . |||\) on \(\mathbb{L}_m(\mathbb{C}^{n\times n})\), define the \(\epsilon\)-pseudospectrum \(\Lambda _\epsilon (L)\) of a matrix polynomial \(L\in \mathbb{L}_m(\mathbb{C}^{n\times n})\) by \(\Lambda _\epsilon (L):=\bigcup_{|||\Delta L|||\leq \epsilon}\Lambda (L+\Delta L)\) where \(\Lambda (L)\) is the spectrum of \(L\) and \(\Delta L\in \mathbb{L}_m(\mathbb{C}^{n\times n})\). The authors prove that the pseudospectra of matrix polynomials introduced by \textit{N. J. Higham} and \textit{F. Tisseur} [Linear Algebra Appl. 351--352, 435--453 (2002; Zbl 1004.65046)] correspond to the pseudospectra defined above when \(\mathbb{L}_m(\mathbb{C}^{n\times n})\) is equipped with Hölder's \(\infty\)-norm. Their unified treatment shows that the pseudospectra of matrix polynomials can be defined and analyzed in the same manner as that of matrices. The authors also define backward errors of approximate eigenvalues of matrix polynomials in the normed space \(\mathbb{L}_m(\mathbb{C}^{n\times n})\) and analyze their subharmonicity and nonconstancy on open subsets of the complex plane. They analyze critical points of the backward error functions and show that a critical point is a multiple eigenvalue of an appropriately perturbed matrix polynomial. The authors show that the common boundary points of components of pseudospectra are critical points of the backward error function. The authors also show some consequences of their analysis.