Finding a novel electrolyte solution of lithium-ion batteries using an autonomous search system based on ensemble optimization

For an autonomous search for electrolyte solutions in electrochemistry, we performed a search task using a parallel Bayesian optimization method with the best composition with maximum lithium (Li) ion conductivity in a mixture of four lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) solutions: propylene carbonate (PC), γ-butyrolactone (GBL), tetraethylene glycol diethyl ether (TGDE), and trimethyl phosphate (TMP). The maximum ionic conductivity was achieved using a mixed solution of 98 vol% GBL and 2 vol% TGDE and showed a high Li ionic conductivity in manual measurements. Furthermore, molecular dynamics simulations of a similar electrolyte implied that the small solvation shell of Li-TGDE caused an increase in the ionic conductivity. The reason for this increase is that the dilute TGDE molecules in the GBL solution can be associated with dilute Li ions. • An original electrolyte optimizer for electrolyte improved throughput by a factor of 10. • Parallel use of different Bayesian algorithm improved optimization stability. • The best composition among PC-GBPL-TMP-TGDE for LiTFSI solution was autonomously determined. • The results of autonomous testing system was validated by manual experiments. • MD simulations illustrated the small Li-ion solvation structure of Li-TGDE-GBL.