Parallel Solvers for Large Eigenvalue Problems Originating from Maxwell's Equations

Stefan Adam and Peter Arbenz and Roman Geus

Abstract
We present experiments with various solvers for large sparse matrixeigenvalue problems. These problems occur in the computation of afew of the lowest frequencies of standing electromagnetic waves incavities that have been discretized by the finite element method.The solvers investigated are (1) subspace iteration, (2) blockLanczos algorithm, (3) implicitly restarted Lanczos algorithm and(4) Jacobi-Davidson algorithm. The experiments have been conductedon the HP Exemplar multiprocessor computer.
Contact
Dr. Peter Arbenz
Institute of Scientific Computing,Swiss Federal Institute of Technology,ETH Zentrum IFW C25.1,CH-8092 Zurich,Switzerland,,
arbenz@inf.ethz.ch