Domain Decomposition Methods for Elliptic
Equations
with Multiple Stochastic Coefficients
Chao Jin
Department of Applied
Mathematics
University of Colorado at Boulder
Boulder, CO
80309
Xiao-Chuan Cai
Department of Computer Science
University of
Colorado at Boulder
Boulder, CO 80309
Abstract
Many physical problems can be modeled by partial differential equations with stochastic coefficients. Different coefficients in the same problem may be perturbed by different random functions. In this talk, we discuss multilevel Schwarz type domain decomposition preconditioned Krylov subspace methods for the numerical solution of second order elliptic equations with multiple stochastic coefficients. We show that the Karhunen-Loeve expansion and double orthogonal polynomials can be used to reformulate the stochastic problem into a large number of uncoupled deterministic equations. We also discuss a PETSc based parallel implementation of a recycling Krylov subspace method and show numerical results obtained on machines with a large number of processors.