Data-Driven Automated Robotic Experiments Accelerate Discovery of Multi-Components Electrolyte for Rechargeable Lithium-Metal Batteries

Materials informatics (MI), a data-driven approach for materials discovery, has attracted significant recent attention in the field of rechargeable batteries. Instead of relying on the experience and intuition of researchers for exploring new materials, the MI approach employs data science techniques that can, in principle, reduce the time and cost of the discovery of new materials with superior battery performance. In fact, virtual screening of compounds has been a particularly active research field; new solid-state materials (primarily crystalline) have been successfully discovered using this strategy by applying computational techniques to conveniently predict desirable properties. By contrast, the MI-based approach has not been frequently employed for liquid-state materials (primarily liquid electrolytes). This is primarily because of the difficulties involved in obtaining sufficient datasets for applying MI techniques. The properties of the SEI film formed at the electrode/electrolyte interface and those of the bulk electrolyte, such as Li-ion conductivity and electrochemical stability, must be considered in the material design of liquid electrolytes. However, the complicated mechanism of the formation of the SEI film can induce significant computational costs for preparing the numerous datasets required to apply MI techniques. An effective approach for MI-driven materials discovery for liquid electrolytes involves the use of an automated robotic high-throughput experimental setup. Although high-throughput experiments have been effectively used to acquire large datasets, these methods have not yet been extensively applied for MI-driven materials discovery of liquid electrolytes, owing to difficulties in performing automated sequential electrochemical analysis in conjunction with big data processing.<br/>An automated robotic experimental system was recently developed by our group for discovering new multicomponent electrolytes for Li-metal-based batteries (ref). Although superior searching throughput more than 400 samples per day was achieved, adequate experimental design is essential to realize high-throughput exploration of electrolyte composition from large searching space. For example, when considering a combination of selecting 5 types from 30 types of chemicals, the candidates are up to 305 (2.43 × 10⁷). Thus, it is not realistic to comprehensively evaluate all the possible combination even using such high throughput experimental setup. Therefore, a specific electrolyte composition that realises a superior battery performance must be determined with only a limited number of experimental trials. In the present study, the effectiveness of a data-driven high-throughput screening method was investigated to discover multicomponent electrolyte additives for lithium metal batteries. Established machine-learning methodologies using Bayesian optimisation were employed to solve combinatorial optimisation problems for analysing datasets obtained from the automated robotic experiments, thereby minimising the number of trials required to identify the ideal electrolyte composition. As a result, we identified the specific electrolyte composition that enhanced the discharge/charge performance of the lithium metal batteries. Studies empowered by data-driven high-throughput screening methods offer new opportunities for efficiently identifying electrolyte compositions and accelerating the development of next-generation rechargeable batteries.<br/><u><b>Refereces:</b></u><br/>S. Matsuda, K. Nishioka, S. Nakanishi, Scientific Reports, 2019, 9, 6211