The ABINIT Users Mailing List ( CLOSED )

[verstraete@pcpm.ucl.ac.be]

Hello Jan.

The spikes in the DOS are a side-effect of the tetrahedron method. If a
tetrahedron lands with identical eigenvalues at all apexes, we insert a
gaussian, and if the DOS grid is too fine, the gaussian is very high (the 
DOS isn't wrong, though).

One important point, already mentioned in the forum, is that the results
are better with unshifted grids (at least 1 shift is 0 0 0). More special
kpoints are included, the degeneracies are better represented, and you
have fewer or no spikes. The other option is to make the grid finer and 
finer (although it's kind of counter-productive - the whole point of using 
the tetrahedron method is getting a better DOS out of fewer kpoints).

Regarding the Fermi level, the tetrahedron method is NOT used to determine 
it. It could be more precise, but as things stand E_f comes from an 
integration on the smeared DOS, as usual. I'm quite surprised you find 
that E_f is shifted for Si (of all systems!).

Matthieu



On Sun, 31 Aug 2003, Jan Gryko wrote:

> 
>  Hello:
> 
>  Using  prtdos 2, the density of states is calculated on a very fine 
>  energy grid, 0.00005 Hartree.  Probably because this, very unusual
>  "spikes" show up in DOS.  For example, for silicon or germanium, 
>  these spikes are quite big even for 10 x 10 x 10 grid, and much 
>  bigger for a 8 x 8 x 8 grid.  In addition, the Fermi level for Si is 
>  shifted deep into the valence band (for prtdos = 1, and prtdos =2).
>  For germanium, the Fermi level is OK.  For both elements, 
>  Troullier-Martins pseudopotenytials have been used. 
>  Did someone have this problem? 
> 
>  With best regards,
>  Jan Gryko
>  Jacksonville, Alabama
> 
> 
> 

-- 
===================================================================
Matthieu Verstraete               mailto:verstraete@pcpm.ucl.ac.be 
PCPM, Boltzmann, pl. Croix du Sud, 1            tel: 010/ 47 86 81
B-1348 Louvain-la-Neuve  Belgium                fax: 010/ 47 34 52