On the variation of the spectrum of a normal matrix (Q1816939)

Let \(A\) and \(B\) be \(n\times n\) matrices with spectral \(\{\lambda_i\}\) and \(\{\mu_i\}\) respectively. The author shows that, if \(A\) is normal, there is a permutation \(\pi\) such that \[ \sqrt{\sum|\mu_{\pi(j)}- \lambda_j|^2}\leq \sqrt{n} |B-A|_F. \] Thus \(\max|\mu_{\pi(j)}- \lambda_j|\leq \sqrt{n}|B-A|_F\), and this improves a bound of \textit{R. Bhatia} [Perturbation bounds for matrix eigenvalues (1987; Zbl 0696.15013), Theorem 24.1].