CISE Postdoctoral Research Associates in Experimental Computer Science: Parallel Hierarchical Methods for Computational Electromagnetics (CCR-NSG; ACIR)

Iowa State University, Ames IA

Investigators

Abstract

0103748 Aluru, Srinivas. Iowa State University CISE Postdoctoral Associates in Experimental Computer Science: Parallel Hierarchical Methods for Computational Electromagnetics The goal of this research is to develop parallel algorithms and build parallel software for the solution of a wide variety of problems involving the computational analysis of electromagnetic scattering. Specific problems of interest are: 1) electromagnetic scattering from quasi-planar surfaces, 2) dielectric random rough surfaces, 3) multiregion surfaces, 4) three-dimensional problems where the fields are characterized by frequency and 5) three-dimensional problems where the field behavior is dependent on time history. The unifying theme in addressing these problems will be the use of the hierarchical Fast Multipole Method and its variants. A major goal of the research is to develop the capability to solve highly non-uniform problems efficiently. Emphasis will be placed on the development of distribution-independent algorithms, i.e., provably efficient algorithms for which the run-time is independent of the distribution without making any assumptions on either the range of distributions or the limited precision of computer arithmetic. The postdoctoral research associate will develop and validate software employing these algorithms in close cooperation with experts in electromagnetics at Iowa State University. Validation of the results will be carried out via comparisons against experimental data as well as numerical results obtained from slower, established solvers. The associate will perform experimental evaluation of the performance of the software using conventional parallel computers and high-performance clusters. 0103748 Aluru, Srinivas. Iowa State University CISE Postdoctoral Associates in Experimental Computer Science: Parallel Hierarchical Methods for Computational Electromagnetics The goal of this research is to develop parallel algorithms and build parallel software for the solution of a wide variety of problems involving the computational analysis of electromagnetic scattering. Specific problems of interest are: 1) electromagnetic scattering from quasi-planar surfaces, 2) dielectric random rough surfaces, 3) multiregion surfaces, 4) three-dimensional problems where the fields are characterized by frequency and 5) three-dimensional problems where the field behavior is dependent on time history. The unifying theme in addressing these problems will be the use of the hierarchical Fast Multipole Method and its variants. A major goal of the research is to develop the capability to solve highly non-uniform problems efficiently. Emphasis will be placed on the development of distribution-independent algorithms, i.e., provably efficient algorithms for which the run-time is independent of the distribution without making any assumptions on either the range of distributions or the limited precision of computer arithmetic. The postdoctoral research associate will develop and validate software employing these algorithms in close cooperation with experts in electromagnetics at Iowa State University. Validation of the results will be carried out via comparisons against experimental data as well as numerical results obtained from slower, established solvers. The associate will perform experimental evaluation of the performance of the software using conventional parallel computers and high-performance clusters.