Fuel cells have emerged as promising candidates to revolutionize the global clean energy infrastructure, but their adoption is hindered by the prohibitive costs and operational considerations of leading Proton Exchange Membrane (PEM) fuel cells. Anion Exchange Membrane (AEM) fuel cells offer a cost-effective alternative, but conflicting property requirements have posed significant materials design challenges. We employ ML techniques to accelerate the design of robust polymers for AEMs with optimal combinations of ionic conductivity and mechanical stability and present a framework that can be expanded to screen candidates across a range of critical sustainability-related properties (glass transition, gas permeability, chemical stability). We collect relevant AEM data from literature, train ML models to predict hydroxide conductivity and maximum water uptake and generate over 9 million novel copolymer candidates. We identified 12 promising copolymers exhibiting superior predicted hydroxide conductivity (> 100 mS/cm) and controlled maximum water uptake (< 30%). The computation-aided design of these novel high-performance copolymers for resilient AEM fuel cells can enable a transformative shift in the utilization of AEM fuel cells and other membrane-dependent sustainable energy solutions across a diverse set of industries.

(a) Schematic of a typical Anion Exchange Membrane (AEM) fuel cell and (b) requirement of expensive materials in PEM vs. AEM fuel cells, as a result of the acidic and alkaline environments, respectively.

(a) OH- conductivity vs maximum water uptake for the training set and candidate set. The red box indicates the ideal combination of properties as set by our initial screening criteria, 12 candidate polymers are predicted to fall within this region. (b) Three best-performing candidate polymers from the ideal region with corresponding monomer composition, temperature, prediction and uncertainties.

Shivank Shukla

Appearances