Towards an Artificial Intelligence-Driven Automated Workflow for Battery Material Synthesis

The realization of breakthrough energy storage technologies depends on design of new battery materials, informed by high-throughput synthesis and characterization pipelines. Here we present an artificial intelligence (AI)-driven workflow that harnesses the predictive capabilities of machine learning to optimize electrolyte solutions for improved battery materials design. In this closed-loop automated workflow, a robotic synthesis platform is integrated with two diagnostics systems (UV-vis / electrochemical impedance [EIS]) using an analysis and experimental design feedback loop. An analysis program, written specifically for this work, reads diagnostic spectroscopy data for a plate of samples and studies absorbance or impedance spectra. Standards are used to fit an absorbance vs. concentration curve, to identify the range where the model has high confidence in its predictive capability, and this is used to predict concentrations for all samples. These predicted concentrations are fed into a machine learning optimizer that automatically determines dilution amounts for the next plate of samples. The automated workflow continues, with the robot initiating the next round of sample preparation, synthesis, and diagnostics. Using AI-driven analysis platforms such as this, we can guide synthesis and characterization of electrolytes to improve design and performance and streamline our path to game-changing research breakthroughs. PNNL-SA-185959