Iterative predictor-corrector method for extraction of the pair interaction from structural data for dense classical liquids

Abstract

First, we demonstrate that careful simulations of fluids have enough accuracy that the resulting radial distribution function can be used to test inversion methods. Second, we introduce a method which allows extraction of the pair interaction starting from structural data for simple liquids even under triple-point condition. The method is an iterative predictor-corrector method in which the predictor is the modified hypernetted-chain equation and the corrector is simulation. We have verified the convergence of the method for the Lennard-Jones fluid and for a model potential for aluminum. We find that other methods of inversion give unreliable results. As a first application of our method we have inverted the experimental structural data of Na at 100 °C.