Speaker: Julien Langou , Computer Science, University of Tennessee
Numerical methods for solving large eigenvalue problems arising in nanoscience
Abstract: As recently as five years ago, a respectable theoretical challenge in the area of quantum nanostructures was the prediction of energy levels and structural properties of ~100-atom clusters. While there certainly remain some important unsolved computational problems in this size regime, recent experimental advances in nanoscience and nanotechnology now pose staggeringly more difficult challenges to theory and simulation efforts. While the number of atoms increases, the size of the problem to solve also increases and then computation issues becomes one of the bottleneck. This problem is addressed in this talk and in particular we focus on the computation of a few of the eigenvalues of a large Hermitian matrix.
Short Bio: Julien Langou graduated from SupAero (Toulouse, France), one of the leading engineer schools in France ("Grandes Ecoles"), in 1999 and his (Master) degree is mainly on Propulsion. Then he got a PhD. in Applied Mathematics done at CERFACS, the subject of the thesis is "Solving Large Linear Systems with Multiple Right-Hand Sides". The thesis was funded by a grant from EADS, an industrial partner. Since Otcober 2003, Julien Langou is a Senior Research Associate fellow in the Computer Science Department of University of Tennessee at Knoxville. He is working in the Innovative COmputing Laboratory headed by Jack Dongarra. His position is funded by a grant from a DOE project on nanoscience.
Host: Billy Stewart, Computer Science, NCSU