Key points are not available for this paper at this time.
Abstract Artificial intelligence/machine learning (AI/ML) applied to battery research is considered to be a powerful tool for accelerating the research cycle. However, the development of appropriate materials descriptors is often the first hurdle toward implementing meaningful and accurate AI/ML. Currently, rational solvent selection remains a significant challenge in electrolyte development and is still based on experiments. The dielectric constant (ε) and donor number (DN) in electrolyte design are insufficient. Finding theoretically computable solvent descriptors for evaluating Li + solvation is a significant step toward accelerating electrolyte development. Here, based on the electrostatic interaction between Li + and solvent, the electrostatic potential (ESP) of electrolyte solvent is calculated by density functional theory calculations and reveals significant regularity. ESP as a direct and simple solvent descriptor for conveniently designing electrolytes is proposed. The lowest negative electrostatic potential (ESP min ) ensures the nucleophilic capacity of the solvating solvent and the weak ESP min means decreased solvation energy. Weak ESP min and strong highest positive electrostatic potential (ESP max ) are the main characteristics of non‐solvating antisolvents. Using the plot of ESP min – ESP max strong solvating solvent, weakly solvating solvent, or antisolvent are identified that have been used in electrolyte engineering. This solvent descriptor can boost AI/ML to develop high performance electrolytes.
Wu et al. (Fri,) studied this question.