Modeling Water Adsorption on Rutile (110) Using van der Waals Density Functional and DFT plus U Methods

We study the energetics and structure of water absorption on the ideal rutile TiO2 (110) surface using dispersion-corrected periodic density functional theory (DFT) calculations and on-site Coulomb potential (DFT+U) corrections. Conventional (PBE) and self-consistent dispersion-corrected DFT methods (vdw-DF1 and vdw-DF2) both suggest that molecular adsorption of intact water molecules on the rutile (110) surface is increasingly preferred with increasing simulation slab thickness. However, empirical dispersion corrections indicate a mix of molecular and dissociated water may coexist at room temperature, with less dependence on slab thickness. This same behavior is seen for DFT+U with U = 3 eV in combination with or without self-consistent dispersion corrected DFT. We find that the preference for the occurrence of dissociated water increases with increasing U. When compared with experimental bond-length data for the adsorbed water species, none of the methods and slab thicknesses correctly predict all bond lengths simultaneously. However, of the methods that energetically favor coexisting associated and dissociated water species on the surface, the three-layer slab with conventional DFT (PBE) and the empirically dispersion-corrected DFT methods come closest to correctly reproducing all of the experimentally observed bond lengths. We conclude that the current level of DFT is insufficient to definitively distinguish between the fully associated and partially dissociated states of water adsorbed on the pristine rutile (110) surface, due to the very small (similar to 0.1 eV) total energy differences between these states.