On Tue 08 Mar 2011, Marci wrote:
> Hi Roberto,
> 
> Why do you use that huge smearing (3000K)? Are you sure you need it?

Hi Marton,

I used a high (actually 3000K = 0.25 eV, not so high compared to the 
gap=1.1eV) temperature in order to be able to overcome eventual local 
potential minima during the relaxation.

> Given the small gap, your system behaves like a metal right now.

The gap is 1.1 eV, why you say it's small?

> If I were you I would try to reduce it to let's say 30K so there are no
> more fractional occupations.

I followed you suggestion and it worked (attached). Thank you very much.
I didn't considered that parameter. But also I really didn't expected such 
a dependence: the fermi level was at half gap!
But in principle you're right: some of the conduction bands become occupied 
and the sum on the transitions is performed between occupied-unoccupied 
bands, and therefore at the end also "intraband" transitions become 
considered. Right?

Roberto

> Marton
> 
> --
> PhD student
> Department of Atomic Physics
> Budapest University of Technology and Economics
> 
> On Tue, Mar 8, 2011 at 11:45 AM, Roberto Guerra <[email protected]> 
wrote:
> > Hello,
> > 
> > noone of the developers is interested in such a fundamental issue? It
> > may be a bug...
> > 
> > RG
> > 
> > On Mon 07 Mar 2011, you wrote:
> >> Dear all,
> >> 
> >> I repeated the calculation using a 2x2x2 grid (input file attached).
> >> The result is the same. This is not surprising, since the problem is
> >> not related, in my opinion, to the calculation accuracy (also, the
> >> calculation at the gamma point should be reasonable for an amorphous
> >> system).
> >> 
> >> I verified that the Fermi energy lies inside the gap (bands
> >> attached).
> >> 
> >> The epsimg is larger than zero at E<Egap. That's not possible.
> >> 
> >> 
> >> Roberto
> >> 
> >> In date 5/3/11 23:52:32, Alexander Vozny wrote:
> >> > Roberto,
> >> > are you sure your Fermi level is in the gap?
> >> > If yes, then I would agree with Karim that most likely your k-grid
> >> > for DOS is different from k-grid for optical calculation.
> >> > Alexander.
> >> 
> >> In date 4/3/11 18:40:44, karim rezouali wrote:
> >> > Roberto,
> >> > 
> >> > The band structure appears not to be totally correct. Seeing your
> >> > input file systlab.fdf, I realized that this curve is obtained
> >> > under the gamma point.approximation
> >> > 
> >> > Try to sample the Brillouin zone as follows:
> >> > %block kgrid_Monkhorst_Pack
> >> > 2  0  0   0.0
> >> > 0  2  0   0.0
> >> > 0  0  2   0.0
> >> > %endblock kgrid_Monkhorst_Pack
> >> > 
> >> > 
> >> > Optical calculations:
> >> > 
> >> > %block Optical.Mesh
> >> > 
> >> >  2  2  2
> >> > 
> >> > %endblock Optical.Mesh
> >> > 
> >> > These changes will make the calculation slightly heavier but that's
> >> > the price to pay for something more accurate.

<<attachment: epsimg-good.png>>

Responder a