Duy Lee, Thanks for your reply. I also suspected that CO could be the problem, but I don't see any obvious issues. I put one CO molecule in a big square box (ibrav=1,celldim(1)=40.0), evaluated the energy at the gamma point only, and didn't use Marzari-Vanderbilt screening. All other parameters (ecutwfc,ecutrho,conv_thr) were kept at the values from the slab calculation. The box size seems saturated: increasing it from 20 to 40 au changed the energy by 7*10^(-5) rydberg, which is well below the error in CO-Au binding energy. The C-O bond length was 1.14 Angstrom, which is fairly reasonable for PBE. Any other comments would be appreciated.
Best, - BGJ
