Magnitude of pseudopotential localization errors in fixed node diffusion quantum Monte Carlo

Research output: Contribution to journalArticlepeer-review

17 Scopus citations

Abstract

Growth in computational resources has lead to the application of real space diffusion quantum Monte Carlo to increasingly heavy elements. Although generally assumed to be small, we find that when using standard techniques, the pseudopotential localization error can be large, on the order of an electron volt for an isolated cerium atom. We formally show that the localization error can be reduced to zero with improvements to the Jastrow factor alone, and we define a metric of Jastrow sensitivity that may be useful in the design of pseudopotentials. We employ an extrapolation scheme to extract the bare fixed node energy and estimate the localization error in both the locality approximation and the T-moves schemes for the Ce atom in charge states 3+ and 4+. The locality approximation exhibits the lowest Jastrow sensitivity and generally smaller localization errors than T-moves although the locality approximation energy approaches the localization free limit from above/below for the 3+/4+ charge state. We find that energy minimized Jastrow factors including three-body electron-electron-ion terms are the most effective at reducing the localization error for both the locality approximation and T-moves for the case of the Ce atom. Less complex or variance minimized Jastrows are generally less effective. Our results suggest that further improvements to Jastrow factors and trial wavefunction forms may be needed to reduce localization errors to chemical accuracy when medium core pseudopotentials are applied to heavy elements such as Ce.

Original languageEnglish
Article number244101
JournalJournal of Chemical Physics
Volume146
Issue number24
DOIs
StatePublished - Jun 28 2017

Funding

This research was sponsored by the Laboratory Directed Research and Development Program of Oak Ridge National Laboratory, managed by UT-Battelle, LLC, for the U.S. Department of Energy. This research used resources of the Oak Ridge Leadership Computing Facility at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725

FundersFunder number
U.S. Department of Energy
Office of ScienceDE-AC05-00OR22725
Oak Ridge National Laboratory

    Fingerprint

    Dive into the research topics of 'Magnitude of pseudopotential localization errors in fixed node diffusion quantum Monte Carlo'. Together they form a unique fingerprint.

    Cite this