Jackie Teddy Yik: A Self-Driving Lab for Battery Electrolyte Design

Abstract

As the transition to renewable energy accelerates, development of efficient, safe, and cost-effective energy storage systems become increasingly critical. Lithium-ion batteries dominate the current market, but their technological maturity motivates exploration of alternative chemistries, like aqueous zinc batteries, which offer safety and cost advantages. However, significant challenges remain before such systems can reach practical, industrial deployment. At the research level, battery testing is often time- and resource-intensive, with long testing protocols alongside process variability from factors such as environmental conditions, component quality, and assembly precision. Self-driving labs (coupling automation and data-driven experimentation) offer a pathway to improved reproducibility, higher throughput, and deeper insight into these electrochemical systems. This thesis presents ODACell, a self-driving lab designed for automated coin cell assembly, in-line battery cycling, and closed-loop electrolyte optimization. The initial ODACell prototype is described and evaluated using an aqueous lithium battery model system, revealing poor reproducibility that motivated hardware and workflow improvements: integration of computer vision, environmental enclosure, and increased space efficiency. The upgraded ODACell was applied to multiple electrolyte systems using Bayesian optimization as the experiment planner. Co-solvent effects (from acetonitrile, dimethyl sulfoxide, trimethyl phosphate, and water) on Coulombic efficiency (CE) were explored with ODACell, confirming the detrimental impact of water. In another study with conventional carbonate-based electrolytes, formation cycle CE was mapped for varying additive concentrations (of 1,3,2-dioxathiolane 2,2-dioxide, prop-1-ene-1,3-sultone, tris(trimethylsilyl) phosphite, and vinylene carbonate); Bayesian optimization combined with statistical analysis identified first-cycle benefits from specific additives while confirming known resistance increases from vinylene carbonate. Finally, ODACell was demonstrated on a zincsystem, highlighting its adaptability across chemistries. Optimization of zinc salts (ZnSO4 and ZnCl2) and additives (thiourea, ammonium phosphate monobasic, and sodium dodecyl sulfate) was supported by a separate advanced characterization technique to provide physiochemical explanations to observed CE effects. Overall, this thesis demonstrates the capabilities and challenges of ODACell as a self-driving lab for liquid electrolyte optimization in coin cell batteries.