Efficient determination of equilibrium state points of Yukawa fluids through the combined features of the first peak and first valley of the radial distribution function

Determining the equilibrium thermodynamic state points of Yukawa fluids from local structural features of the radial distribution function (RDF) is often limited by the large number of equilibrium configurations required to obtain sufficiently stable features. In the present work, the limitation is addressed by introducing a more efficient strategy that uses the displacement of the first peak rpeak∗ and the amplitude of the first valley gvalley. Empirical correlations between these two structural features and the state parameters of inverse temperature and density are established, and an artificial neural network (ANN) model is further developed. The combination of rpeak∗ and gvalley provides input structural features for accurate state-point prediction in the Yukawa fluids considered here. The required number of equilibrium configurations is reduced to 3500, leading to a reduction of about 65% in RDF calculation time. The empirical correlations and ANN are established for a three-dimensional truncated repulsive Yukawa system with A=100 and κ0=2 over β∗∈[0.1,1.3] and ρ∗∈[0.5,1.2]. Additional simulations at κ0=1 and 4 support the usefulness of the selected descriptor pair with separately calibrated empirical correlations. The reported ANN accuracy and approximately 65% reduction in RDF calculation time are validated for the reference κ0=2 system. Applications beyond the investigated conditions require further validation.