Dear Prof. Marzari, Thanks for your reply. I will use the sampling you suggested.
With regards to "right/wrong", after some literature review i would say the results are rather correct. The interlayer distance change during surface relaxation could indeed be non-monotonous, as was shown, for example, here Huang, G. (2008). Electronic structures, surface phonons, and electron-phonon interactions of Al(100) and Al(111) thin films from density functional perturbation theory. Physical Review B, 78(21), 214514. doi:10.1103/PhysRevB.78.214514 In any case, i will play more with this simple toy-problem. Sincerely, Denis On 06.05.2013, at 19:25, Nicola Marzari <nicola.marzari at epfl.ch> wrote: > > > Hi Denis, > > not sure anything looks wrong, a part that your smearing is quite small, > and you might want to test that and the k-points used. > > In any case, you could try to leave all atoms free to relax, and study > interlayer relaxations as a function of thickness, and see how the outer > interlayer relaxations converge to their true value as a function of > thickness (and progressively the inner relaxations as well). > > Last - no need to sampe in the x direction - your system > should not have any band dispersion there (if it has, there is not > enough vacuum). > > Hence, 1 16 16 0 1 1 is a very good sampling (even 1 8 8 0 1 1 would > be great, provided you use maybe a more reasonable smearing of .01-.02). > > > nicola
