Ab initio surface energetics: beyond chemical accuracy

Density functional theory (DFT) is the work–horse of modern materials modeling techniques, but scattered evidence indicates it often fails for important surface properties. This thesis investigates how DFT estimates of the surface energy (σ) and molecular adsorption energies of ionic systems are affected by the choice of exchange–correlation (xc) functional. Accurate diffusion Monte–Carlo (DMC) and quantum chemistry (QC) calculations are presented for these quantities showing marked improvement over DFT and agreement of much better than chemical accuracy. DFT estimates of σ are presented for the (001) surfaces of LiH, LiF, NaF and MgO. Five xc functionals, LDA, PBE, RPBE, Wu–Cohen and PW91 are used. A clear xc functional bias is demonstrated with σLDA > σWC > σPBE > σPW91 > σRPBE. To improve the picture detailed pseudopotential DMC calculations are presented for LiH and LiF. The lattice parameters and cohesive energies agree with experiment to better than 0.2 % and 30 meV respectively. For LiH novel all–electron DMC calculations are also presented showing significant improvement over pseudopotential DMC. Accurate all–electron Hartree–Fock calculations of σ for LiH(001) and LiF(001) are presented along with calculations of the LiF bulk using specially adapted Gaussian basis–sets. Combined with existing QC correlation estimates the bulk and surface properties of LiH and LiF show excellent agreement to both experiment and DMC and allow a longstanding disagreement between two experimental estimates for σLiF to be resolved. Finally the adsorption energy curve for water on LiH(001) is obtained by both DMC and incremental QC techniques leading to agreement of better than 10 meV. DFT and dispersion corrected DFT estimates are also presented highlighting the large xc functional dependence. Thus we demonstrate that is possible and necessary to obtain agreement between higher levels of theory and produce benchmark values beyond DFT