Force field optimization and solid–liquid equilibrium predictions of methane and noble gases by molecular dynamics simulations

Accurate prediction of solid–liquid phase equilibrium through molecular dynamics simulation remains a significant challenge. Within the framework of reference state method, sensitivity analysis was conducted to evaluate the effect of reference temperature selection on melting point prediction. The predicted melting point slightly decreases as the reference temperature gradually deviates from the target value. Considering the limitations of traditional force fields (FFs) typically fitted with thermodynamic properties of fluid phases, the classical TraPPE FF was evaluated by calculating different thermodynamic properties at supercooled liquid phase and melting points along the solid–liquid coexistence curve. Despite high accuracy for liquid-phase thermodynamic properties, the TraPPE FF shows significant deviations for melting point predictions. Furthermore, though a dimensionless conversion of the solid–liquid coexistence points, an empirical correlation function accurately relating the coexistence pressure and the coexistence temperature is proposed. Using this correlation function as a constrain, an optimized FF set (ε = 0.256 kcal/mol and σ = 3.466 Å) is efficiently determined by Levenberg-Marquardt algorithm, exhibiting an average absolute deviation of merely 1.49 % and a maximum deviation of 2.42 %. Outside the temperature range of parameter fitting, the good extrapolation prediction capability (ARD = 1.76 %) of the optimized FF is also validated. Besides, the proposed correlation model shows good applicability to four noble gases with deviations in melting point predictions all falling below 1.7 K.