Thermodynamic properties of liquid mixtures: Benchmarking the transferability of general purpose and quantum-mechanically derived force fields

Molecular dynamics simulations are used to investigate key thermodynamic properties of selected pure liquids and their binary mixtures, including excess volume and excess enthalpy. Since the predictive accuracy of these properties depends critically on the force field (FF), two parameterization strategies were assessed for benzene, pyridine, ethanol, and propanol: (i) the general-purpose OPLS model, and (ii) system-specific quantum-mechanically derived force fields (QMD-FFs), parameterized exclusively from purposefully computed ab initio data. For pure liquids, OPLS systematically reproduces densities and vaporization enthalpies with high accuracy, whereas the performance of QMD-FFs is more system dependent and closely reflects the ability of the underlying electronic-structure method to describe intermolecular interactions. In the systems examined here, QMD-FFs may either overestimate or underestimate cohesion, as illustrated by the different behavior of benzene and ethanol. The thermal expansion coefficient and isothermal compressibility also exhibit marked compound dependence for both approaches, with no uniformly superior strategy. The investigated models are thereafter extended to the six binary mixtures, exploiting standard combination rules for cross interactions. OPLS again provides an overall reliable description, often approaching a nearly quantitative agreement with experiment. Conversely, although QMD-FFs capture several thermodynamic trends upon mixing, they show limited transferability from the pure-component data. Although the choice of mixing rule affects the magnitude of these deviations, it does not fully compensate for deficiencies in the underlying parameterization. These results highlight the need for more robust QMD-FF parameterization strategies for multi-component complex systems.