Unbiased Diffusion Monte Carlo for non local operators

Abstract

We propose a new mathematically exact method for computing unbiased Diffusion Monte Carlo (DMC) estimates of non-local operators. We demonstrate that the current state-of- the-art technique, Forward Walking, is only exact for local quantities and fails to yield unbiased results for the non-local components of reduced density matrices (RDMs). Our method significantly outperforms Forward Walking, as shown in two systems: in the symmetric Hubbard dimer it yields a pure 1RDM; while in the Helium atom it will give an unbiased 1RDM in the limits of zero time step and infinite walkers.

14 pages, 5 figures