Decorrelation of samples in Quantum Monte Carlo calculations and applications to metallic nanoclusters

ORAL

Abstract

We discuss decorrelation of samples in Quantum Monte Carlo (QMC) ground-state energy calculations for large lithium and water nanoclusters and show how accurate results can be obtained without the need for decorrelating samples. The scaling of the integrated autocorrelation time $\tau$ is analyzed as a function of nanocluster size. $\tau$ is found to scale quadratically in Li nanoclusters, which adds a quadratic factor to the scaling of the total computation time in this metallic case, a factor which does not appear in computations of non-metallic H$_{2}$O nanoclusters. We choose nanoclusters which are relatively large in the context of QMC to demonstrate the application of these techniques - lithium nanoclusters with up to 64 atoms and water nanoclusters with up to 20 molecules.

*Work supported in part by the USDOE.

Authors

  • Daniel Nissenbaum

    • Northeastern University
  • Bernardo Barbiellini

    • Northeastern U.
    • Northeastern University
  • Arun Bansil

    • Northeastern University
    • Northeastern U.
    • Northeastern University, Boston, MA 02115