It is well known that trial wavefunction quality often makes or breaks the performance of ground state Quantum Monte Carlo (QMC) techniques [1]. Much of our intuition about how trial wavefunctions influence QMC methods, however, has been predicated upon our experience using them in real space methods, such as Diffusion Monte Carlo. In this talk, I will illustrate how trial wavefunctions impact the performance of a second-quantized QMC method, Auxiliary Field Quantum Monte Carlo (AFQMC) [2]. I will first detail my group’s recent efforts to develop a high-performance, fully open source AFQMC software package for widespread community use and discuss its performance on a number of strongly-correlated solids. I will then discuss a new stochastic approach for generating trial wavefunctions suited for strongly correlated materials and demonstrate how this approach can significantly improve the quality of AFQMC results for the Hubbard Model and small molecules for a modest cost. Ideas discussed in this talk will shed light upon the quality of wavefunctions that will be needed to study strongly-correlated materials using AFQMC in the future. References [1] M.A. Morales, J. McMinis, B. K. Clark, J. Kim, and G.E. Scuseria, J.Chem. Theory. Comput. 8 (7) (2012). [2] S. Zhang and H. Krakauer, Phys. Rev. Lett. 90, 136401 (2003).

Obtaining trial wavefunctions on the cheap: A stochastic approach to multideterminant wavefunctions in auxiliary field quantum Monte Carlo

251st National Meeting (2016)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.