Anharmonic Correction to the Adsorption Free Energy of Oxygen-Containing Intermediates on Pt(111) by Machine-Learned Force Field-Based Thermodynamic Integrations

Accurate and efficient descriptions of adsorption free energy are of fundamental importance to bringing theoretical findings closer to experiments. The present study introduced a robust combination of machine learning and (un)constrained molecular dynamics-based λ-path thermodynamic integration methods to accurately estimate the anharmonic correction to the
free energy of adsorbing systems and, thus, the adsorption free energy. The approach was demonstrated for the adsorptions of the oxygenated species exhibiting different anharmonic behaviors, namely, atomic oxygen (O), hydroxyl (OH), and hydroperoxyl (OOH), on the Pt(111) surface. The corrected adsorption free energies reveal the significant influence of anharmonic effects, which could account for up to 39% of the entropy loss estimated by the conventional harmonic approximation, even at near room temperature. This highlights the limitations of the current harmonic approximation and underscores the need to account for anharmonicity for better descriptions of the adsorption processes. The present method paves the way for the accurate adsorption of free energy calculations to be performed routinely toward more powerful theoretical predictions of heterogeneously catalyzed reactions.