[gpaw-users] Fwd: slab-calculations

[Jussi Enkovaara]

On 2011-10-27 10:03, Peter Deák wrote:
>
> One more layman's question: is a muffin-tin like approach for defining the grid principally hopeless?

Hi,
I assume you mean something like LAPW or LMTO where one uses different basis 
functions within atomic spheres and interstitial region. In principle,
I think it should be possible to use similar approach with real-space grids,
i.e. atomic-like functions within atomic spheres and uniform grids outside,
however I do not think that anyone has really worked out the formalism.

Also, PAW method can be seen as muffin-tin like approach with fixed basis functions 
(projectors etc. constructed at setup generation) in atomic
(=muffin-tin) spheres, represented in radial grids.

One should also note that muffin-tin like approach does not solve the problem of 
wasting grid-points/planewaves in the vacuum region, the required resolution is 
still determined mostly by the regions between atoms. I think that proper approach 
to variable resolution would be with finite elements or wavelets, but they 
introduce naturaly intricancies (in impelementation, numerics, etc.) of their own.

As Jens pointed out, also LCAO provides variable resolution, the accuracy of 
calculation can be just more difficult to control.

Best regards,
Jussi