Researchers used machine learning to map how a six-metal high-entropy alloy resists carbon monoxide poisoning in fuel cell anodes. The study, published in Ionics, focused on PtRuNiCoFeMo and its behavior during alkaline hydrogen oxidation. CO is a trace contaminant in reformed hydrogen that can degrade platinum catalysts, so this matters for making fuel cells more practical. The team combined geometric descriptors, sampling, density functional theory, and a graph neural network to predict CO adsorption energies across 120,000 surface sites. The result shows the alloy broadens the distribution of binding energies rather than uniformly weakening CO adsorption. That statistical picture helps explain CO tolerance and could guide design of better anode catalysts. This is early-stage computational work, not a commercial product yet, but it gives catalyst researchers a clearer target for developing cheaper, more durable fuel cell components.
