Efficient sampling for ab initio Monte Carlo simulation of molecular clusters using an interpolated potential energy surface

An approach is developed to enhance sampling for ab initio Monte Carlo and ab initio path integral Monte Carlo calculations of molecular clusters by utilizing an approximate potential as a guide to move in the configuration space more efficiently. The interpolated potential energy obtained by the moving least-squares method is used as an approximate potential, and this scheme is applied to a water molecule and small protonated water clusters (H(3)O(+), H(5)O(2+)). It is found that the statistical errors are reduced by almost a factor of 3 in most calculations, which translates into a reduction in the computational cost by an order of magnitude. We also provide an automatic scheme where the ab initio data obtained during the simulation is added to the reference data set of interpolation dynamically, which further speeds up the convergence. (C) 2009 American Institute of Physics. [DOI: 10.1063/1.3055910]