> Thanks for the reply. Taking the limit of what you've said, Andrei - > if you want a really slow dispersion in the k_z direction, it would > presumably be best to work with a thin (i.e zero thickness) slab.
Not really. The dispersion in z-direction has nothing to do with the thickness of your slab, it depends on the periodicity in the z-direction, i.e. interaction of your slab with its replica. With thick enough vacuum layer in whatever code, the k_z-dispersion becomes negligible, so one point along z suffices. The thickness of the slab (i.e. number of layers) determines how many bands you have. In the limit of infinitely thick slab, you'll have bands smoothly covering whole areas. A thin slab simply has less bands; whether it is "better" for your purposes depends on what you are actually calculating. > would mean the dispersion in the k_z direction should be infinitely > slow - right? As I said above, this has nothing to do with the thickness of your slab. > This is in fact the sort of thing I've been doing, using just the > surface layer to 'eyeball' the 'surface' bands, then comparing this > with the dispersions taken from the slab calculation. I think this is not very useful idea. You should decide whether you want to calculate a slab or a single layer; these are different things. If you take slab, and want to figure out what is the dispersion of surface states, you need to identify them among all other bands, using something like "fat bands" analysis (bands decorated by selected orbital contributions). > The problem is > the bands tend to distort and shift in energy as the bulk is > re-attached, of course they will do it! What did you expect? Best regards Andrei Postnikov
