Ionic liquids are good candidates as the main component of safe electrolytes for high-energy lithium-ion batteries. The identification of a reliable algorithm to estimate the electrochemical stability of ionic liquids can greatly speed up the discovery of suitable anions able to sustain high potentials. In this work, we critically assess the linear dependence of the anodic limit from the HOMO level of 27 anions, whose performances have been experimentally investigated in the previous literature. A limited r Pearson's value of ≈0.7 is found even with the most computationally demanding DFT functionals. A different model considering vertical transitions in a vacuum between the charged state and the neutral molecule is also exploited. In this case, the best-performing functional (M08-HX) provides a Mean Squared Error (MSE) of 1.61 V2 on the 27 anions here considered. The ions which give the largest deviations are those with a large value of the solvation energy, and therefore, an empirical model that linearly combines the anodic limit calculated by vertical transitions in a vacuum and in a medium with a weight dependent on the solvation energy is proposed for the first time. This empirical method can decrease the MSE to 1.29 V2 but still provides an r Pearson's value of ≈0.72.
Keywords: DFT calculations; QSAR; anodic stability; ionic liquids; machine learning.