Acceleration vs Accuracy: Influence of Basis Set Quality on the Mechanism and Dynamics Predicted by Ab Initio Molecular Dynamics

Ab initio molecular dynamics (AIMD) is an indispensable tool for understanding the mechanistic details of externally-energy mediated chemical reactions. In this work, we show that the predicted thermodynamic and catalytic properties of certain reactions using AIMD simulations critically depend on the quality of the employed basis set. To this end, we have examined the reactants and products of the water-gas shift reaction (viz., CO, CO₂, H₂, and H₂O) and studied their interaction with the ZnO(101̄0) surface using density functional theory (DFT) and Born Oppenheimer Molecular Dynamics (BOMD) simulations. By merely increasing the quality of the basis, from double zeta (commonly used in most calculations of these systems) to triple zeta, we surprisingly find that the reaction outcome of an H₂O molecule colliding with a ZnO surface pre-covered with carbon monoxide gives qualitatively different results. These surprising results are shown to be robust with similar trends that are also obtained with other software packages. Furthermore, we show that the calculated adsorption energies can vary by as much as 380 meV (which is an order of magnitude larger than room temperature) by simply changing the basis set. Using electron density difference maps, we present mechanistic insight into the origin of these changes. Finally, we propose a simple diagnostic test that uses a single-point binding energy calculation to estimate the impact of basis-set quality, which can be used before carrying out a computationally-expensive BOMD simulation.