Ab Initio Calculations of the Redox Potentials of Additives for Lithium-Ion Batteries and Their Prediction through Machine Learning

Ab
initio molecular orbital calculations were carried out to examine
the redox potentials of 149 electrolyte additives for lithium-ion
batteries. These potentials were employed to construct regression
models using a machine learning approach. We chose simple descriptors
based on the chemical structure of the additive molecules. The descriptors
predicted the redox potentials well, in particular, the oxidation
potentials. We found that the redox potentials can be explained by
a small number of features, which improve the interpretability of
the results and seem to be related to the amplitude of the eigenstate
of the frontier orbitals.