Volume 41, pp. 93-108, 2014.

The block preconditioned steepest descent iteration for elliptic operator eigenvalue problems

Klaus Neymeyr and Ming Zhou

Abstract

The block preconditioned steepest descent iteration is an iterative eigensolver for subspace eigenvalue and eigenvector computations. An important area of application of the method is the approximate solution of mesh eigenproblems for self-adjoint elliptic partial differential operators. The subspace iteration allows to compute some of the smallest eigenvalues together with the associated invariant subspaces simultaneously. The building blocks of the iteration are the computation of the preconditioned residual subspace for the current iteration subspace and the application of the Rayleigh-Ritz method in order to extract an improved subspace iterate. The convergence analysis of this iteration provides new sharp estimates for the Ritz values. It is based on the analysis of the vectorial preconditioned steepest descent iteration which appeared in [SIAM J. Numer. Anal., 50 (2012), pp. 3188–3207]. Numerical experiments using a finite element discretization of the Laplacian with up to $5\cdot 10^7$ degrees of freedom and with multigrid preconditioning demonstrate the near-optimal complexity of the method.

Full Text (PDF) [800 KB], BibTeX

Key words

subspace iteration, steepest descent/ascent, Rayleigh-Ritz procedure, elliptic eigenvalue problem, multigrid, preconditioning

AMS subject classifications

65N12, 65N22, 65N25, 65N30

Links to the cited ETNA articles

[9]Vol. 7 (1998), pp. 104-123 Andrew V. Knyazev: Preconditioned eigensolvers - an oxymoron?

ETNA articles which cite this article

Vol. 46 (2017), pp. 424-446 Ming Zhou and Klaus Neymeyr: Sharp Ritz value estimates for restarted Krylov subspace iterations
Vol. 51 (2019), pp. 331-362 Marija Miloloža Pandur: Preconditioned gradient iterations for the eigenproblem of definite matrix pairs
Vol. 58 (2023), pp. 597-620 Ming Zhou and Klaus Neymeyr: Convergence rates of individual Ritz values in block preconditioned gradient-type eigensolvers

< Back