The ABINIT Users Mailing List ( CLOSED )

[PGanesh <pganesh@ciw.edu>]

In the description of 'prtdos' the last paragraph seems to suggest that 
the DOS for PAW potentials may not be correct: (pasted below)

"Note that (for the moment), in case of PAW, the prtdos=3 choice does 
not give the "true" DOS. Only the pseudized part of the wave functions 
is used (which is not norm-conserving). The "true" density should be 
obtained from the "true" wave functions (adding a spherical part). 
Results in PAW for prtdos=3 may be wrong. "

So does that mean that using the input file of Takeshi-san we won't get 
the "true" DOS ? What does "Adding the spherical part" mean?

Thanks,
Ganesh


Matthieu Verstraete wrote:
> Peakyness either appears if tsmear is too small (not your case - you 
> use the tetrahedron method) or if for some reason there are tetrahedra 
> of kpoints with the same eigenvalue at 3 or 4 apexes (apices?). In 
> that case the linear interpolation of the energy gives a constant in 
> the whole tetrahedron and this gives a peak in the DOS. Normally you 
> always have a dispersion, and increasing the number of kpoints will 
> make this appear (and the peaks disappear). In practice the delta peak 
> is replaced by a thin Lorentzian. High symmetry will make more of 
> these degenerate cases appear.
>
> There is by now little you do not know about the implementation of 
> prtdos 2 and 3, boys and girls!
>
> Matthieu