Assessing the Reliability of Minimally Constrained Reverse Monte Carlo Simulations for Model Metallic Liquids

TL;DR

This study evaluates the reliability of reverse Monte Carlo simulations constrained only by total pair correlation functions in modeling atomic structures of metallic liquids, revealing limitations in reproducing local structural details.

Contribution

It provides a systematic assessment of RMC's accuracy in modeling metallic liquids using minimal constraints, highlighting its limitations.

Findings

RMC often does not match MD configurations in local structural distributions.

Using only TPCFs as constraints limits the accuracy of RMC in reproducing atomic structures.

Caution is advised when interpreting RMC results constrained solely by pair correlation functions.

Abstract

Molecular dynamics simulations using semi-empirical potentials are examined for three liquids to check the reliability of reverse Monte Carlo (RMC) simulations to reproduce atomic configurations when only total pair correlation functions (TPCF) are used as constraints. The local structures are determined from a Voronoi tessellation of the ensemble and compared with the structures obtained by RMC in terms of asphericity, volume, coordination number, Voronoi index, and nearest-neighbor distance. It is found that in general the distributions generated from RMC do not match the MD configurations, using the $L^1$ (taxicab) distance as a metric, although in some cases a measure of central tendency for the distribution did match. Since only TPCFs are typically used to constrain the RMC simulations of experimental data, this study establishes the limits on what can be learned by this analysis. It indicates that caution should be used when interpreting RMC-generated structures using few constraints since many structural quantities are not reproduced well.