Comparison of reverse Monte Carlo algorithm variants via the accuracy of their three-particle correlation functions and configurational temperatures

Configurations obtained by reverse Monte Carlo algorithm variants on a one-component Lennard-Jones fluid were compared by analysing properties calculated from the three-particle correlation functions. The variants were quantified by their accuracy, even when the results were not distinguishable at the level of standard cosine and neighbour distributions. In addition, the concept of configurational temperatures. developed recently was applied as a diagnostic to check the correctness of the potential energy surface models. The best method is the simultaneous fit on the structure factor at the power of two with averaged measurement error and on the pair-correlation function at the power of four with r dependent measurement error. The same orders of accuracy were obtained by comparing pair potentials derived by a recently suggested method that combines the Born-Green-Yvon equation and the reverse Monte Carlo simulation.