It has been increasingly common to investigate dynamic and thermodynamic properties of green solvents at atomistic scales with molecular simulation. These designed solvents such as ionic liquids are often highly charged species, which pose a problem for molecular modelling with classical fixed-charge force fields. Simulation outcomes with atomic charges derived from ab initio calculations often show significant deviations from the experimental reference, and charge scaling is widely applied as a simple yet efficient solution to achieve satisfactory experiment-simulation accordance. Although the massive emphasis on the reproduction of bulk properties such as density of the liquids as charge scaling criteria, a more crucial thermodynamic observable is the solvation behavior of external agents in these green solvents. Astonishingly, our recent large-scale benchmark calculation on solvation thermodynamics suggests that the solvation-free-energy-derived scaling factor is generally slightly larger (~0.1) than the bulk-property-derived estimate. This phenomenon is rather not unexpected, as the density-matching estimate only considers the solvent-solvent interaction (overfitting), while accurate calculations of solvation free energies require balanced descriptions of solute-solvent and solvent-solvent interactions. A more interesting observation is that different solute-solvent pairs exhibit different responses to the variation of the scaling parameter, which arises from the competing electrostatic and vdW contributions. Another perspective provided in our previous extensive benchmark is about the suitability of general-purpose force fields for bonded and vdW interactions in the modelling of ionic liquids. The bond stretching and angle bending terms in pre-fitted GAFF derivatives are often problematic, while the torsional potential shows satisfactory reproduction of ab initio results. In the current work, we expect to accumulate more experiences from large-scale fast-growth solvation free energy simulations in ionic liquids with compositions different from our previous benchmark. The obtained new results are combined with our previous dataset to form a large solvation set (and also partition or water-ionic-liquids transfer set), from which a universally applicable charge scaling factor with at least half-optimal performance is derived. Aside from the force-field issue, another extremely important modelling detail considered in the current work is the finite-size effect. It is observed that the finite-size artifacts in solvation thermodynamics are much more severe than mass density, which emphasizes the use of a sufficiently large ionic solvent box in molecular simulations of ionic liquids derivatives. Finally, combining the extensive computational perspectives accumulated in our series works, general guidelines for molecular modelling of ionic liquids with fixed-charge force fields are summarized.
More analyses about molecular ESP are added to consolidate the charge generation protocol. The main text is shortened.