A parallel domain decomposition method for the Helmholtz equation in layered media

Abstract
An efficient domain decomposition method and its parallel implementation for the solution of the Helmholtz equation in three-dimensional layered media are considered. A modified trilinear finite element discretization scheme is applied to the equation system leading to fourth-order phase accuracy and thereby reducing the pollution error considerably. The resulting linear system is solved with the GMRES method using a multiplicative nonoverlapping domain decomposition preconditioner with layers defining the subdomains. This right preconditioner is constructed by embedding each layer into a rectangular domain and by employing a fast direct solver. Due to the construction of the preconditioner the iterations can be reduced to a subspace corresponding to the interfaces between the layers. Numerical experiments with several test cases demonstrate the effectiveness and scalability of the proposed method and ability to solve large-scale problems with up to billions of unknowns.
Main Authors
Format
Articles Research article
Published
2019
Series
Subjects
Publication in research information system
Publisher
Society for Industrial and Applied Mathematics
The permanent address of the publication
https://urn.fi/URN:NBN:fi:jyu-202402272159Use this for linking
Review status
Peer reviewed
ISSN
1064-8275
DOI
https://doi.org/10.1137/18M1230906
Language
English
Published in
SIAM Journal on Scientific Computing
Citation
  • Heikkola, E., Ito, K., & Toivanen, J. (2019). A parallel domain decomposition method for the Helmholtz equation in layered media. SIAM Journal on Scientific Computing, 41(5), C505-C521. https://doi.org/10.1137/18M1230906
License
In CopyrightOpen Access
Funder(s)
Research Council of Finland
Funding program(s)
Akatemiahanke, SA
Academy Project, AoF
Research Council of Finland
Additional information about funding
This work was supported by the Academy of Finland under project 295897.
Copyright© 2019 Society for Industrial and Applied Mathematics

Share