On the Use of Dispersion Corrections by Default in Computational Catalysis

Abstract

Accurate adsorption energy predictions are central in computational catalysis. The need for higher accuracy without prohibitive computational costs has led to the development of correction schemes for density functional theory (DFT) predictions. For instance, dispersion corrections are nowadays routinely applied, even when long-range interactions are expected to be negligible. To evaluate this practice, we assess the effects of D3 dispersion corrections on CO adsorption over transition metals using the Perdew‒Burke‒Ernzerhof (PBE) and revised PBE (RPBE) exchange-correlation functionals. These corrections systematically stabilize the CO adsorption energies, which generally increases deviations from experimental data. Furthermore, depending on the damping function used (zero or Becke-Johnson damping), D3 corrections alter the gas-phase thermochemistry. However, applying both gas- and adsorbed-phase corrections yields adsorption energies that closely match experimental values. Consequently, the application of dispersion corrections by default may compromise the predictive capability of DFT-based models.