The energetics of oxide surfaces by quantum Monte Carlo

Density functional theory (DFT) is widely used in surface science, but for some properties the predictions depend strongly on the approximation used for exchange-correlation energy. We note recent suggestions that the widely used generalized gradient approximation (GGA) is inferior to the local density approximation (LDA) for the surface formation energy sigma of both transition metals and oxides. We report quantum Monte Carlo calculations of s for the MgO(001) surface which support the accuracy of LDA for this case, and indicate that GGA is too low by similar to 30%. We point out the potentially important implications of this result for nanoscience modelling