Palindromic quadratization and structure-preserving algorithm for palindromic matrix polynomials of even degree (Q635809): Difference between revisions

The authors consider \((*,\varepsilon)\)-palindromic matrix polynomials of even degree \(2d\), \({\mathcal P}(\lambda)=\sum_{k=0}^{d-1}\lambda^{2d-k}A_{d-k}^*+\lambda^dA_0+\varepsilon\sum_{k=1}^d\lambda^{d-k}A_k\), where \(d\geq2\), \(\varepsilon=\pm1\) and *\,=\,H (Hermitian) or T (transpose), \(A_k\in{\mathbb C}^{n\times n}\) \((k=0,1,\ldots,d)\) and \(A_0^*=\varepsilon A_0\). They propose a palindromic quadratization to transform \({\mathcal P}(\lambda)\) with \((*,\varepsilon)\neq(T,-1)\) to a \((*,\varepsilon)\)-palindromic quadratic pencil, instead of the orthodox palindromic linearization approach. The structure-preserving algorithm for solving such an eigenvalue problem based on the \(({\mathcal S}+{\mathcal S}^{-1})\)-transform and \textit{R. V. Patel}'s algorithm [Linear Algebra Appl. 188--189, 591--611 (1993; Zbl 0782.65057)] can then be applied. This approach reduces the relative residuals of approximate eigenpairs with much cheaper computational cost.