Molecular Modeling of Ionic Liquids: Force‐Field Validation and Thermodynamic Perspective from Large‐Scale Fast‐Growth Solvation Free Energy Calculations

Abstract

In molecular modelling of novel solvents such as ionic liquids, it is common to scale atomic charges to improve the experiment-simulation agreement for some selected properties. As these liquids are designed to solvate solutes, whether the solvation thermodynamics could be correctly described is of utmost importance. Therefore, we present a comprehensive large-scale calculation of solvation free energies via nonequilibrium fast-switching simulations for a spectrum of molecules in ionic liquids, the atomic charges of which are scaled to maximize the prediction-experiment correlation. Further, the density-derived choice is compared with the solvation-thermodynamics-derived one. When the scaling factor is decreased, the density exhibits a monotonically decreasing behavior due to weaker inter-molecular interactions produced by scaled atomic charges. However, solvation free energies do not show consistent monotonic behaviors, which are caused by competing electrostatic and vdW responses to the scaling-parameter variation. More intriguingly, the solvation-free-energy-derived scaling factor is generally slightly higher than the density-derived one. We further calculate partition coefficients ortransfer free energies of solutes from water to ionic liquids to provide another thermodynamic perspective of charge scaling. Another central result is the detailed evaluation of the widely used force fields for bonded and vdW terms, i.e., the GAFF derivatives.

Molecular Modelling of Ionic Liquids: Situations When Charge Scaling Seems Insufficient

“…From hundreds of solvation/partition data, we summarize that the scaling factor of 0.8 serves as a near-optimal solution generally applicable to most ILs [ 59 , 60 ], which is somehow consistent with electrostatic potential (ESP) analysis in isolated and small clusters [ 61 ]. Ion-pair-specific tuning of this parameter could still lead to marginal improvements, and the best solvation/partition-derived charge-scaling parameter is observed to be slightly larger than the density-derived solution [ 59 , 60 ]. Although this slight increase in the scaling factor could degrade the reproduction of the bulk property, it effectively improves the accuracy of solvation and partition thermodynamics and does not overfit a given observable.…”

“…However, in our series of papers benchmarking the charge-scaling issue with large-scaling solvation free-energy calculations and also unbiased simulations of the bulk solvent, this density-matching procedure is found to degrade the reproduction of solvation free energies and water–ILs transfer free-energies [ 59 , 60 ], which is attributed to the overfitting of solvent–solvent interactions for the density-derived solution, while accurate calculation of solvation and partition thermodynamics requires balanced descriptions of solute–solvent and solvent–solvent interactions. Unlike bulk properties (e.g., mass density) that are monotonic with respect to the charge-scaling factor, the non-monotonic dependence of solvation/partition thermodynamics due to the competing electrostatic and vdW contributions of molecular solvation further complicates the situation [ 59 , 60 ]. From hundreds of solvation/partition data, we summarize that the scaling factor of 0.8 serves as a near-optimal solution generally applicable to most ILs [ 59 , 60 ], which is somehow consistent with electrostatic potential (ESP) analysis in isolated and small clusters [ 61 ].…”

“…Unlike bulk properties (e.g., mass density) that are monotonic with respect to the charge-scaling factor, the non-monotonic dependence of solvation/partition thermodynamics due to the competing electrostatic and vdW contributions of molecular solvation further complicates the situation [ 59 , 60 ]. From hundreds of solvation/partition data, we summarize that the scaling factor of 0.8 serves as a near-optimal solution generally applicable to most ILs [ 59 , 60 ], which is somehow consistent with electrostatic potential (ESP) analysis in isolated and small clusters [ 61 ]. Ion-pair-specific tuning of this parameter could still lead to marginal improvements, and the best solvation/partition-derived charge-scaling parameter is observed to be slightly larger than the density-derived solution [ 59 , 60 ].…”

“…Thus, such treatment is considered to produce balanced descriptions of solvent–solvent and solute–solvent interactions. The large-scale benchmark on typical ILs and solute molecules of diverse structural and chemical features ensures the general applicability of the accumulated insights [ 59 , 60 ]. Direct application of the summarized modelling protocol to species significantly different from the benchmark set is attempted in the following work, specifically ILs involving Trihexyltetradecylphosphonium, L-Lactate, and (1S)-(+)-10-camphorsulfonate [ 62 ].…”

“…The 1-alkylquinuclidinium cations are often associated with large organic fluorinated anions to form liquids at ambient conditions. A good candidate for this purpose is bis(trifluoromethylsulfonyl)imide [NTF], which is an anion studied extensively in our previous works [ 59 , 60 ]. The melting points of ionic solvents formed by 1-alkylquinuclidinium and NTF are critically influenced by the length of the aliphatic chain.…”