Neural network potential calculations for melamine adsorption onto Pt (111) compared with density functional theory

The investigation of Pt-adsorbed melamine is important in elucidating the effect of molecular decoration on the enhancement of catalytic performance for fuel cells, and is an interesting system in which covalent stabilization between the surface and the molecule competes with resonance destabilization by the coordination. The present work discusses whether graph neural network potentials can predict the adsorption structure with the competition. The most stable structure predicted by preferred potential was consistent with that by density functional theory but underestimated resonance destabilization. Graphical Abstract