(560if) Generalized Adsorption Models on Metal Nanoparticles

Computational catalyst design has attracted tremendous interest over the last years as a means of identifying active and selective catalysts without the need of trial-and-error experimentation in the lab. However, there is still a scarcity of generalizable, physically-grounded models capable of both rapidly and accurately capturing the adsorption behavior of reactants and intermediates on the catalyst surface. With adsorption being central to catalyst design efforts, it is crucial to develop models that capture adsorption as a function of metal nanoparticle size, shape and metal composition. Using a combination of symbolic regression and statistical methods, we develop a novel first-principles-based adsorption model which is parameterized by both experimental literature data, and DFT calculations on very simple systems. Our model accurately captures a wealth of reported adsorption data and predicts adsorption energies of many intermediates onto nanoparticles of any size, shape and metal composition, in agreement with first principles calculations.