The accurate knowledge of the potential energy surface describing the interaction of molecules with surfaces is of primary importance when studying catalytic reactions. While density functional theory is computationally efficient, it suffers from the possibly incorrect description of the exchange-correlation potential and therefore does not allow accurate predictions of reaction rates. Diffusion Monte Carlo may provide a viable way to obtain more accurate results for selected energies along the potential energy landscape, but the applicability of QMC is not yet well established for such systems. It is therefore important to study the benchmark system H2 on Cu(111). The transition metal surface, however, makes matters even more complicated : Transition metals are generally not well described in DFT. This can lead to large fixed node errors. Furthermore, their large number of electrons necessitates the use of pseudo-potentials, which can lead to errors due to the locality approximation and the simplified interaction. In order to assess the achievable accuracy of QMC for reactions of molecules on transition-metal-surfaces, I studied the interaction energy of transition metal molecules with diffusion Monte Carlo, focusing on the influence of the DFT wavefunction used as trial wave function and on the locality error.
Université Paul Sabatier - Bat. 3R1b4 - 118 route de Narbonne 31062 Toulouse Cedex 09, France