The ABINIT Users Mailing List ( CLOSED )

["Javad hashemifar" <jhashemifar@gmail.com>]

Dear Matthieu

I benefit from your kind explanations especially about core correction
techniques.
But still I am not sure about my results. I am not worry about the
wiggles of DOS but the main problem is the nontrivial difference
between up and down splittings (exchange splitting) of Abinit and
Wien2k. I plotted the DOSs again to make this problem more visible
(please see attachment). As it is clear, the splitting of the main up
and down peaks (~0.1 Ryd below the Fermi level) in Wien2k (~0.03 Ryd)
is considerably larger than Abinit (~0.01).
Is there any way to remove this difference or I should consider it as
a trivial difference and simply neglect it.

Thanks in advance.
Javad Hashemifar



On 12/7/06, Matthieu Verstraete <mjv500@york.ac.uk> wrote:
> On the contrary! This is really a very good result for the
> pseudopotentials. The core corrections is obviously necessary in an
> element with high lying core states, but the results are as close as you
> can expect them in the pseudopotential approximation.
>
> If you are concerned about the wiggles in the abinit DOS, they come from
> the kpoint sampling and the way the DOS is calculated. Use prtdos 2 to get
> a smoother results with the tetrahedron method (which is certainly what is
> used in wien).
>
> Your last phrases were a bit worrying, it sounded like you were changing
> convergence parameters to get the DOS closer to the wien one. You must not
> do this! You should converge the kpoints, ecut, etc... until the results
> do not change anymore. They are not free parameters.
>
> The psp core radii and NLCC parameters are a bit trickier, but normally
> fitting the psp for the free atom should be enough to give you a nice
> transferable potential (well, not always strictly true). As you can see,
> changing the NLCC parameter does change the DOS, but mainly in the
> unoccupied states. There is a general recipe for setting the rnlc to the
> radius at which the valence density becomes larger than the core density,
> but this is a bit arbitrary. In many cases, people do what you have done
> here, and use the rnlc which best reproduces an all electron calculation.
>
> Matthieu
>
> --
> ================================================================
> Dr. Matthieu Verstraete                 mailto:mjv500@york.ac.uk
> Dept. of Physics, University of York,     tel: +44 1904 43 22 08
> Heslington, YO10 5DD York, United Kingdom fax: +44 1904 43 22 14


--
======================================
Seyed Javad  Hashemifar,    Ph.D.
[current:]
Tel:+49-203-3794743        Fax:+49-203-3794742
Fachbereich Physik (AG Kratzer), Universitat Duisburg-Essen
47048 Dusiburg, Germany
[permanent:]
Tel:  +98-311-3912375       Fax: +98-311-3912376
Physics Department, Isfahan University of Technology
84156 Isfahan, Iran
---------------------------------------------------------------------------

Attachment: dos-abinit2.ps
Description: PostScript document