HPC for GW-BSE Simulations

Jul 1, 2019
Electronic Structure of a benzen on a graphene Sheet

GW-BSE simulations allows to obtain avery detailled and accurate calculation of the excitation spectrum of small molecular systems. However this approach requires to solve large-scale eigenvalue problems that becomes rapidly untracktable using conventional methods.

Using Votca-XTP, a recent engine for GW-BSE simulations we have implemented Matrix-Free Jacobi-Davidson eigenvalue solvers to accelerate the simulations of small molecule and solve the memory bottleneck associated with the calculations of large systems. Porting these methods to GPU give an additional computational benefits making Votca-XTP a valuable tool for the future of electronic-structure calculations

Timing and Memory consumption of the Davidson solver compared to the DSYVEX solver from LAPACK. The Davidson solver can be either much faster than DSYVEX or consume orders of magnitude less of memory depending on the type of applications
Timing and Memory consumption of the Davidson solver compared to the DSYVEX solver from LAPACK. The Davidson solver can be either much faster than DSYVEX or consume orders of magnitude less of memory depending on the type of applications

Votca-xtp

Quantum Chemistry GW-BSE VOTCA-XTP
Section Head
Section Head
Natural Science and Engineering

Advancing Academic Research Through Software Development

Related