Capturing the Complexities of Catalyst–Support Interactions with the Help of Machine Learning

The structure of metal nanoparticles is central to their catalytic activity, but metal-support interactions are difficult to model via quantum-mechanical calculations. Using a machine-learned potential to model supported silver nanoparticles, it has been shown that the idealized nanoparticle shapes commonly invoked in the literature do not reflect experiments for diameters below 8 nm, as reported by Maxson and Szilvási.