Advancing reverse Monte Carlo structure refinements to the nanoscale

Over the past decade, the RMCProfile software package has evolved into a powerful computational framework for atomistic structural refinements using a reverse Monte Carlo (RMC) algorithm and multiple types of experimental data. However, realizing the full potential of this method, which can provide a consistent description of atomic arrangements over several length scales, requires a computational speed much higher than that permitted by the current software. This problem has been addressed via substantial optimization and development of RMCProfile, including the introduction of the new parallel-chains RMC algorithm. The computing speed of this software has been increased by nearly two orders of magnitude, as demonstrated using the refinements of a simulated structure with two distinct correlation lengths for the atomic displacements. The new developments provide a path for achieving even faster performance as more advanced computing hardware becomes available. This version of RMCProfile permits refinements of atomic configurations of the order of 500 000 atoms (compared to the current limit of 20 000), which sample interatomic distances up to 10 nm (versus 3 nm currently). Accurate, computationally efficient corrections of the calculated X-ray and neutron total scattering data have been developed to account for the effects of instrumental resolution. These corrections are applied in both reciprocal and real spaces, thereby enabling RMC fitting of an atomic pair distribution function, which is obtained as the Fourier transform of the total-scattering intensity, over the entire nanoscale distance range accessible experimentally.