Coarse-Grained Simulation of PEO/LiTFSI Electrolytes with Assistance of Bayesian Optimization

PEO-Li+ electrolytes are believed to be one of the most promising solid polymer electrolytes for advanced iontronics and high-specific energy storages. A combination of the Kremer–Grest polymer model with the 1/r4-form solvation potential is attractive for coarse-grained simulation of PEO-Li+ electrolytes. Here, we present a method to estimate the solvation parameters from the experimental data by using LiTFSI as the Li salt. The EO-Li+ solvation strength (SEO+) is chosen to reflect the real value of EO-Li+ separation (∼2.47 Å). The undetermined parameters are determined with the assistance of Bayesian optimization (BO) to match the experimental observations. With the optimized parameters, Li+ and LiTFSI– can be well dissolved in the PEO host, which is confirmed by radial distribution functions. The average coordination number (CN) for Li+–EO complexes is 4.4–4.5. Due to the dynamic nature of Li+–EO coordination, the proportions of different CNs fluctuated with time. The rational value SEO– is determined as 9ϵ by comparison with experimental or theoretical data. With SEO– = 9ϵ, the force fields can reproduce the ion-transport properties of PEO16-LiTFSI, in terms of self-diffusion coefficient, cation transference number, and ionic conductivity. The force fields are also assessed by reproducing the dependence of ion and EO transports on the ionic concentration and molecular weight of PEO. These results indicate that the structure and dynamics of PEO/LiTFSI electrolytes with [Li+]/[EO] ≤ 1/12 can be well approximated using the force–field parameters.