Eigenvalue estimates for the resolvent of a non-normal matrix (Q2018374)

Let \(A\in\mathbb{C}^{n\times n}\), let \(\zeta\in\mathbb{C}\), and denote by \(\sigma(A)\) the spectrum of \(A\). Given a norm \(\|\cdot\|\) on \(\mathbb{C}^{n\times n}\), the problem is to find a good upper bound for \(\|(\zeta-A)^{-1}\|\) a good upper bound depending only on \(\zeta\), \(n\) and~\(\sigma(A)\). In other words, the problem is to find a good function~\(\Phi\) such that \(\|(\zeta-A)^{-1}\|\leq\Phi(\zeta,n,\sigma(A))\) for all \(\zeta\in\mathbb{C}\setminus\sigma(A)\). First, let \(\|\cdot\|\) be the spectral norm. Assuming that \(\|A\|\leq 1\), the author finds the optimal \(\Phi\) and gives explicitly the matrices with equality. This recovers and generalizes results by \textit{E. B. Davies} and \textit{B. Simon} [J. Approx. Theory 141, No. 2, 189--213 (2006; Zbl 1127.47037)]. Second, let \(\|\cdot\|\) be arbitrary. Assuming that \(\sup_{k\geq 0}\|A^k\|\leq C\) for some constant~\(C\), he finds the strongest~\(\Phi\) so far. Finally, he applies these results in studying the sensitivity of stationary states of a classical or quantum Markov chain with respect to perturbations of the transition matrix.