Abstract: Redox flow batteries (RFBs) are a promising technology for stationary energy storage applications due to their flexible design, easy scalability, and low cost. Further improvements in energy density of RFBs requires redox-active material (redoxmer) designs with wider redox potential window, higher solubility, long-term stability, and self-reporting functionality. To accelerate the discovery of redoxmers with desired properties, state-of-the-art active learning methods based on Bayesian optimization (BO) are useful.
In this talk, I will discuss two recent applications of our active learning framework to accelerate the discovery of redoxmer molecules with targeted properties. In the first application, the active learning model was utilized to identify desired homobenzylic ether (HBE) molecules with targeted oxidation potential. The BO successfully identified 42 desired HBEs from a search space of 112,000 candidates after evaluating only 100 molecules using density functional theory (DFT) simulations. In second application, our active learning method was extended to discover redoxmers which satisfy multiple desired properties using multi-objective Bayesian optimization (MBO). With MBO, the Pareto-optimal candidates were identified at least 15 times more efficiently compared to brute force or random selection approach. We anticipate that this active learning technique is general and can be utilized for the discovery of any class of functional materials that satisfies multiple desired property criteria.
Bio: Dr. Garvit Agarwal is a postdoctoral appointee in Materials Science Division (MSD) at Argonne National Laboratory. He is currently working as a part of the Joint Center for Energy Storage Research (JCESR), a DOE energy innovation hub led by Argonne. He received his Ph.D. in Materials Science and Engineering from University of Connecticut in 2019.