The ABINIT Users Mailing List ( CLOSED )

["Anglade Pierre-Matthieu" <anglade@gmail.com>]

1) You can hardly compare calculations with different occopt also
unless both are semiconductor or insulator with a band gap wider than
your tsmear.
1.1) The criterion to be able to make a comparison is that the tsmear
has a small influence on your total energy. The influence of tsmear
defining the accuracy of your comparison. If you demonstrate a low
enough  dependance of total energy toward tsmear for all relevant
systems you can go on with the comparisons of total energy between
systems with occopt 1 and 4.

2) Choose occopt according to the computed band gap: occopt is mainly
used to allow smearing and then easier integration in the FBZ. If your
system is simulated metallic you  need a smearing (unless you have
very large parallel machines)  and then you must use metallic occopt
and tsmear.
2.1) When you expect a semiconductor, Use metallic occopt only if you
witness a convergence that is too slow with respect to the number of
k-point. Otherwise stick with occopt 1.
2.2)  I think the DOS is enough as long as your k-point sampling is fine.
2.3)  Usually people look at the lines between high symmetry k-point.

On 8/30/07, vincent.chevrier@dahn.phys.dal.ca
<vincent.chevrier@dahn.phys.dal.ca> wrote:
> Hi,
> I'm studying the different phases in the Li-Si binary.
> The phases have the following experimental band gaps:
>
> Si      1.17 eV
> Li12Si7 0.6 eV
> Li7Si3  0.08 eV
> Li13Si4 ?
> Li22Si5 ?
> Li      Metal
>
> I am studying these phases with GGA pseudopotentials and I know I should 
> expect calculated band gaps to be smaller and therefore Li7Si3, Li13Si4, 
> Li22Si5 to likely have no gap.
>
> I ran calculations on all the mixed phases with occopt 4 and tsmear 0.04, 
> and the results were reasonably good but I didn't look into the question of 
> band gaps at the time.
>
> I have two main questions (with sub questions!):
>
> 1) For my research I need to compare the energies of the different phases 
> and from Pierre-Matthieu Anglade's notes 
> (http://anglade.googlepages.com/anglade-lecture-notes-on-abinit.pdf) I have 
> learned that calculations with different tsmear cannot be compared. What 
> about calculations with different occopt (1 and 4) be compared?
>         1.1) If I can compare them are there any extra considerations for 
> tsmear?
>
> 2) Do I choose occopt based on the experimental or calculated band gap?
>         2.1) If I look for a band gap using calculated data, do I use 
> occopt 1 or occopt 4 for calculations?
>         2.2) Is a DOS output (prtdos 1) enough or do I have to make a band 
> structure diagram?
>         2.3) If I have to consider individual bands, can I only look at 
> high symmetry k-points? (from http://www.cryst.ehu.es/cryst/get_kvec.html)
>
> Thanks for your help!
>
> Vincent Chevrier
> PhD Candidate
> Physics Department
> Dalhousie University
>


-- 
Pierre-Matthieu Anglade