Scientific Computing Research Environments for the Mathematical Sciences (SCREMS)

University Of Delaware, Newark DE

Investigators

Abstract

The Department of Mathematical Sciences at the University of Delaware will purchase a distributed memory parallel computer which will be dedicated to the support of research in the mathematical sciences. It will be used for several research projects, including: Ordering and coarsening dynamics for an fcc binary alloy (Braun); Domain decomposition for modal computations (Driscoll); Electromagnetic scattering (Monk); Vortex computations of vortex dipole evolution, dipole pair collisions and simulations of synthetic jets (Rossi). These projects all involve the numerical solution of complex partial differential equations in applied mathematics. Realistic solutions use algorithms which have large memory and time requirements. The proposed Beowulf cluster of 32 processors linked with a dedicated Myrinet will allow us to investigate these problems efficiently using parallel algorithms. The broader impacts of our proposal are twofold. First, we will generate new results of interest to environmental and materials scientists, fluid dynamicists, electrical engineers and the defense industry. Second, through involvement of graduate and undergraduate students in the proposed projects, we will prepare a new generation of applied mathematicians skilled in parallel computation.