Quantum Monte Carlo for High Accuracy ab initio Calculations

Quantum Monte Carlo (QMC) is among the most accurate ab initio Quantum Chemistry methods available. Furthermore, in stark contrast with comparable methods, it is trivially parallelizable, it requires only a negligible amount of memory, and it’s computational cost scales as only O(N^3). Unfortunately, as a stochastic method, it requires a sizeable number of state space evaluations. We have been developing QMcBeaver, a very computationally efficient program to run these calculations, and we report here preliminary results. We show how they compare with exact results, other published QMC results, and coupled cluster results and show that our code is very competitive. To accomplish this, we worked to stabilize the calculation and minimize the noise in several ways. First, we fit Slater Type Orbitals to arbitrary Hartree-Fock Gaussian Type Orbitals near nuclei. Second, not only do we use an efficient initialization scheme, we apply additional criteria for identifying and eliminating poor Markov chains. Third, we rigorously control Monte Carlo branching according to a list of heuristics. Fourth, for statistical purposes. We summarize progress made with QMcBeaver and show how it can be used to obtain highly accurate results.