Extension of the TraPPE Force Field for Battery Electrolyte Solvents

Optimizing electrolyte formulations is key to improving
performance
of Li-/Na-ion batteries, where transport properties (diffusion coefficient,
viscosity) and permittivity need to be predicted as functions of temperature,
salt concentration and solvent composition. More efficient and reliable
simulation models are urgently needed, owing to the high cost of experimental
methods and the lack of united-atom molecular dynamics force fields
validated for electrolyte solvents. Here the computationally efficient
TraPPE united-atom force field is extended to be compatible with carbonate
solvents, optimizing the charges and dihedral potential. Computing
the properties of electrolyte solvents, ethylene carbonate (EC), propylene
carbonate (PC), dimethyl carbonate (DMC), diethyl carbonate (DEC),
and dimethoxyethane (DME), we observe that the average absolute errors
in the density, self-diffusion coefficient, permittivity, viscosity,
and surface tension are approximately 15% of the corresponding experimental
values. Results compare favorably to all-atom CHARMM and OPLS-AA force
fields, offering computational performance improvement of at least
80%. We further use TraPPE to predict the structure and properties
of LiPF6 salt in these solvents and their mixtures. EC
and PC form complete solvation shells around Li+ ions,
while the salt in DMC forms chain-like structures. In the poorest
solvent, DME, LiPF6 forms globular clusters despite DME’s
higher permittivity than DMC.