Optimized Schwarz Methods without Overlap for the Helmholtz Equation

DOI10.1137/S1064827501387012MaRDI QIDQ4785911zbMATH OpenOpenAlexFDO

Authors Martin J. Gander, Frédéric Magoulès, Frédéric Nataf

Publication date 5 January 2003

Published in SIAM Journal on Scientific Computing (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1137/s1064827501387012, https://archive-ouverte.unige.ch/unige:6291

acoustics domain decomposition Krylov methods numerical examples optimized Schwarz methods preconditioner

Laplace operator, Helmholtz equation (reduced wave equation), Poisson equation (35J05) Iterative numerical methods for linear systems (65F10) Numerical computation of matrix norms, conditioning, scaling (65F35) Stability and convergence of numerical methods for boundary value problems involving PDEs (65N12) Finite element, Rayleigh-Ritz and Galerkin methods for boundary value problems involving PDEs (65N30) Multigrid methods; domain decomposition for boundary value problems involving PDEs (65N55)

Review text

The classical Schwarz method has been applied primarily to boundary value problems for elliptic partial differential operators. In this paper, the method is applied to hyperbolic operators. The focus of the work is on determination of boundary conditions for the artificial boundary between domains that maximize the convergence rate of the resulting system. These conditions are called ``transmission conditions. The authors present their results applied to problems defined on the whole Cartesian plane divided into left and right half-planes and separated by the \(y\)-axis. Numerical examples are presented in which the method is applied in bounded regions. The authors first consider transmission conditions of the form \(\partial_x+{\mathcal S}\), where \(\mathcal S\) represents a tangential operator. Based on representing solutions as exponentials and using expressions for the symbol of the operators, they prove that the tangential operators can be chosen so that the Schwarz iteration converges in just two iterations. Unfortunately, the operators so specified have symbols with square roots in them and are not local. They go on to show that differential operators in the tangential space can be specified based on maximizing the convergence rate of the induced iterative method. While the iteration will sometimes have a single mode with convergence rate equal to one, the use of a Krylov method preconditioned by the Schwarz method will nonetheless converge. Asymptotic estimates are presented that yield convergence rates of the order \(1-O(h^{1/2})\), with \(h\) representing mesh size, and that converge as \(1-O(h^{1/4})\) for the propagating modes. Finally, an implementation of the method in a finite element context is discussed and the resulting discrete system is shown to have a solution. Numerical experiments are also presented, one on the unit square with several combinations of parameters and another consisting of the passenger compartment of a Volvo sedan.

Recommendations

Cited in

(only showing first 100 items - show all)

This page was built for publication: Optimized Schwarz Methods without Overlap for the Helmholtz Equation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4785911)