DFT, ML and GNN-based insights into fluxional nanoclusters for fuel cell application

Abstract

Electrochemical energy conversion technology has emerged as the most promising avenue to address the growing environmental problems and global energy crisis in an eco-friendly, carbon-neutral cycle [1]. In this context, the development of high-performance electrocatalysts for devices such as regenerative fuel cells, metal-air batteries, and proton exchange membrane fuel cells (PEMFCs) [2], remains a critical challenge despite their strong potential for clean energy conversion. In particular, the sluggish kinetics and high reaction overpotential values associated with the oxygen reduction reaction (ORR) at the cathode significantly limit the efficiency of PEMFCs and impede their large-scale commercialization. Moreover, the expensive Pt loading required and its susceptibility to poisoning and degradation under prolonged operating conditions further reduce cathodic efficiency, underscoring the need for alternative catalytic materials that minimize the use of precious metals. Subnanometer clusters have recently emerges as an important class of heterogeneous electrocatalysts due to their exceptional catalytic performance and high atomic utilization efficiency [3]. At finite temperatures, these clusters exhibit a relatively flat potential energy surface (PES), which renders them highly dynamic and results in non-Arrhenius behavior. The presence of multiple undercoordinated atomic sites leads to pronounced fluxionality, giving rise to non-monotonic size-dependent activity trends that distinguish subnanoclusters from bulk and nanoparticle catalysts [4]. Recent studies have shown that clusters in the non-scalable regime have complex isomeric distribution and electronic structures, leading to metastability-triggered reactivity and the need for ensemble representation to accurately describe the overall activity [5].