A High-Speed Screening Method by Combining a High-Throughput Method and a Machine-Learning Algorithm for Developing Novel Organic Electrolytes in Rechargeable Batteries

Abstract

A high-speed screening method was established by combining a high-throughput quantum simulation (HT) with a machine learning algorithm (ML), which is named as MAssive Molecular Map BUilder (MAMMBU). For the applications to high-voltage rechargeable batteries, the MAMMBU was applied to screen organic electrolytes based on redox potential calculations. The HT-method consists of the automatic input generator from existing organic compound database, error correction, and job scheduler. In the ML-method, the computational results of redox potentials for organic compounds obtained from HT were used as training and test data for an artificial neural network. This MAMMBU was targeted for building redox potentials map based on whole organic compounds database, such as PubChem, and finding novel organic electrolytes with the new types of cores, substituents, and functional groups. This will be possible by remarkable reducing time to predict redox potentials from several hours for quantum calculations to several seconds for the ML-method in MAMMBU.