Iterative methods for the computation of a few eigenvalues of a large symmetric matrix (Q1923866): Difference between revisions

New iteration methods are proposed to compute a few extreme and nonextreme eigenvalues and associated eigenvectors of a large sparse symmetric matrix \(A\). These methods do not require the factorization of matrices of the form \(A-xI\). They use the recursion formulas of the implicitly restarted Lanczos method given by \textit{D. Calvetti}, \textit{L. Reichel} and \textit{D. C. Sorensen} [ETNA, Electron. Trans. Numer. Anal. 2, 1-21 (1994; Zbl 0809.65030)]. The recursive formulas are applied to introduce an accelerating potential. For chosing the sequence of Krylov subspaces in the computation and for chosing the shifts as Leja points the corresponding algorithms are proposed. Numerical examples include the computation of 3 smallest eigenvalues of the matrix \(A = \text{diag} (1,2,3,\dots,n)\), the minimum energy equilibrium configuration of liquid crystals and 3 other smallest eigenvalue computations.