Tuesday, April 20, 2010
101 Stanley Thomas Hall
Tulane University (Uptown)
Refreshments will be served
Juan Meza, Department Head, Senior Scientist, High Performance Computing Research, Lawrence Berkeley National Laboratory
New Optimization Methods for Solving the Kohn-Sham Equations
Nanostructures have been proposed for many applications including solar cells for renewable energy, biomedical imaging, and other novel materials. To fully explore these ideas however, requires ab initio materials simulations. Today, most of these codes are based on Density Functional Theory (DFT), which can be used for computing the ground state energy and the corresponding single particle wave functions associated with a many-electron atomistic system. At the heart of these codes, one typically finds a Self Consistent Field (SCF) iteration. In this talk, we propose a different approach based on new optimization methods that minimize the Kohn-Sham total energy directly. I will discuss the similarities between our new approach and SCF and show how the SCF iteration can fail in certain cases. A trust region technique is introduced as a way to restrict the update of the wave functions within a small neighborhood of an approximate solution. Numerical examples demonstrate that the combination of these approaches is more efficient than SCF.
Center for Computational Science, Stanley Thomas Hall 402, New Orleans, LA 70118 firstname.lastname@example.org