Hoffman-Wielandt type inequality for block companion matrices of certain matrix polynomials (Q6064604)

If \(A\) and \(B\) are two \(n\times n\) normal matrices with eigenvalues \(\lambda_1\), \(\ldots\), \(\lambda_n\) and \(\mu_1\), \(\ldots\), \(\mu_n\), respectively, then for some permutation \(\pi\) of the set \(\left\{1,\ldots,n\right\}\) we have \[ \sum_{i=1}^n|\lambda_i-\mu_{\pi(i)}|^2\leqslant||A-B||^2_F,\tag{HW} \] that is the Hoffman-Wielandt inequality. In [\textit{Kh. D. Ikramov} and \textit{Yu. R. Nesterenko}, Dokl. Math. 80, No. 1, 536--540 (2009; Zbl 1189.15021); translation from Dokl. Akad. Nauk, Ross. Akad. Nauk 427, No. 4, 447--451 (2009)] one can find generalized versions of the above theorem. In particular, if \(A\) is diagonalizable and \(X\) denotes a nonsingular matrix whose columns are the eigenvectors of \(A\), then for some \(\pi\) we have \[ \sum_{i=1}^n|\lambda_i-\mu_{\pi(i)}|^2\leqslant\kappa^2(X)||A-B||^2_F,\tag{HW'} \] where \(\kappa(X)\) is the spectral condition number of \(X\). In the present paper the authors study the inequalities (HW) and (HW') for block companion matrices of matrix polynomials whose coefficients are either unitary or doubly stochastic matrices. The main results concern some quadratic matrix polynomials whose block companion matrices turn out to be diagonalizable. Therefore, for such matrices (HW') holds. From this results the authors also derive some bounds for condition numbers.