Incomplete double-cone factorizations of centrosymmetric matrices arising in spectral methods (Q6200844)

This interesting paper studies the so called class of centrosymmetric matrices \((a_{ij})\in \mathbb R^{n\times n}\) which are symmetric about their centers: That is, \(a_{i,j}=a_{n+1-i,n+1-j}\) for \(1\leq i,j\leq n\). Centrosymmetric matrices occur in many applications for example, partial differential equations, signal processing, several Markov processes and many more. Two examples of centrosymmetric matrices are symmetric Toeplitz and symmetric circulant matrices and there are several well known numerical methods for solving problems involving these matrices in the literature. For example, various preconditioners have been used in the literature for studying different symmetric systems involving Toeplitz matrices. In the paper under review, the authors are interested in studying centrosymmetric matrices which arise in various spectral methods for solving some families of partial differential equations. The authors develop structure-preserving incomplete LU type factorizations in order to precondition centrosymmetric matrices and to then use these factorizations to solve numerically some nearly centrosymmetric and centrosymmetric linear systems arising from spectral methods for certain partial differential equations. The algorithms developed build, in part, on direct solution techniques previously developed for these types of linear systems, featuring double-cone factorizations. The authors illustrate their results on certain discretizations of some model problems involving Poisson, diffusion, Helmholtz, and biharmonic equations in one, two, and three dimensions. The paper is well written with a good set of references.