Catalyzing Anhydrous Proton Conduction: A Computational Workflow for Fuel Cell Materials Design

The development of materials for anhydrous proton conduction is crucial for enhancing proton exchange membrane fuel cell performance, reducing costs, and expanding operating conditions. We present a computational workflow using deep learning potentials (DPs), graph lattice models, deep learning charge density predictors, and a reactive active learning scheme to design fuel cell membrane materials. Our approach highlights that graphanol (hydroxylated graphane) conducts protons anhydrously with low diffusion barriers. Proton self-diffusion coefficients of graphanol were calculated with accurate and efficient DPs to estimate overall diffusion barriers. The sensitivity of intrinsic energy barriers to the overall diffusion barrier for proton conduction was assessed using GLMs. We discovered transient hydrogen-bonded structures (Grotthuss chains) governing proton conduction. Long-range transport occurs through the formation of new Grotthuss chains via hydroxyl group rotation. Thus, the overall diffusion barrier consists of a convolution of the intrinsic proton hopping barrier and the intrinsic hydroxyl rotation barrier. This work yielded design rules for developing advanced proton-conducting materials. Finally, we demonstrate how an efficient reactive active learning method can be used to study chemical reactions involving degradation pathways. This method is highly versatile and can be applied to many other systems to develop DPs capable of accounting for chemical reactions.