{"title":"Fast direct solution of FEM systems using overlapped localizing modes on a shifted grid","authors":"Xin Xu, R. Adams","doi":"10.1109/APS.2011.5997231","DOIUrl":null,"url":null,"abstract":"Overlapped, localizing local-global solution (OL-LOGOS) modes have been proposed to develop fast direct solvers for low frequency electromagnetic wave problems. The efficiencies of the resulting OL-LOGOS factorization algorithms have been demonstrated for the matrix equations associated with dense three-dimensional integral equations and sparse two-dimensional partial differential equations. In both cases, approximately O(N log N) time and O(N) memory complexities have been observed. In this work, the OL-LOGOS method is applied to three-dimensional scalar FEM systems. In order to improve the factorization speed and reduce memory costs for FEM applications, a pre-factorization permutation step is incorporated into the OL-LOGOS factorization algorithm. Numerical results demonstrate factorization and memory complexities of approximately O(N log N) and O(N) as the problem size grows.","PeriodicalId":6449,"journal":{"name":"2011 IEEE International Symposium on Antennas and Propagation (APSURSI)","volume":"45 1","pages":"3265-3267"},"PeriodicalIF":0.0000,"publicationDate":"2011-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2011 IEEE International Symposium on Antennas and Propagation (APSURSI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/APS.2011.5997231","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Overlapped, localizing local-global solution (OL-LOGOS) modes have been proposed to develop fast direct solvers for low frequency electromagnetic wave problems. The efficiencies of the resulting OL-LOGOS factorization algorithms have been demonstrated for the matrix equations associated with dense three-dimensional integral equations and sparse two-dimensional partial differential equations. In both cases, approximately O(N log N) time and O(N) memory complexities have been observed. In this work, the OL-LOGOS method is applied to three-dimensional scalar FEM systems. In order to improve the factorization speed and reduce memory costs for FEM applications, a pre-factorization permutation step is incorporated into the OL-LOGOS factorization algorithm. Numerical results demonstrate factorization and memory complexities of approximately O(N log N) and O(N) as the problem size grows.