作者: Simone Chiesa , D. M. Ceperley , Shiwei Zhang
DOI: 10.1103/PHYSREVLETT.94.036404
关键词: Monte Carlo molecular modeling 、 Statistical physics 、 Estimator 、 Quantum Monte Carlo 、 Wave function 、 Dynamic Monte Carlo method 、 Variance (accounting) 、 Efficient estimator 、 Monte Carlo method 、 Physics
摘要: Computation of ionic forces using quantum Monte Carlo methods has long been a challenge. We introduce simple procedure, based on known properties physical electronic densities, to make the variance Hellmann-Feynman estimator finite. obtain very accurate geometries for molecules H2, LiH, CH4, NH3, H2O, and HF, with Slater-Jastrow trial wave function. Harmonic frequencies diatomics are also in good agreement experiment. An antithetical sampling method is discussed additional reduction variance. The optimization molecular crystal structures ab initio dynamics simulations among most significant achievements singleparticle theories. These accomplishments were both possible thanks possibility readily computing ions within framework BornOppenheimer approximation. approximate treatment electron interactions typical these approaches can, however, lead quantitatively, sometimes qualitatively, wrong results. This fact, together favorable scaling computational cost respect number particles, spurred development stochastic techniques, i.e., (QMC) methods. Despite higher accuracy achievable many properties, lack an efficient prevented, until recently [1–3], use QMC predict even simplest geometry. chief problem have (MC) force sufficiently small For example, allelectron calculations, straightforward application MC infinite can be easily seen from definition force. nucleus chargeZ at origin, written, its variance, as function