The ABINIT Users Mailing List ( CLOSED )

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

Hi,

Band structure depend on:
-XC functional (with Abinit you set that with "ixc")
- it can depend on pseudopotentials (yet it shall not if the
pseudopotentials are corrects)
-it depends on the quality of the convergence of your calculation.

On this last point I have a few comments (may be because you have just
omitted to tell us about those point in your mail; but it is worth
checking if you haven't already done so).
You may try:
- to set a finer K-point mesh
- to be sure that 50 Ha of cutof fis enough for the pseudopotentials
you use (I would guess Cd pseudopotential to be quite hard)
- to check the band convergency during your  SCF iterations:
  a) Be sure that you include enough bands. In your case you let
Abinit use the default. Although it should be enough in your case, you
can try to add a few. The algorithm that check the convergence of the
non self consistent iteration do not take care of the two last bands
and maybe your conduction bands are not converged enough ?
  b) Some systems converge badly. You may have to increase your
convergence criterion in order to get small values (1e-4 or less; 1e-8
is overkiling ) for both what Abinit calls Vres2 and residm. If one of
them is big at the end of your SCF calculation it may be the source of
your problem.
- choose a different XC functional that gives results closer to litterature
-Also it is important to be certain that the k-lines under examination
are those that include the max and min of the valence and conduction
bands.

Hoping one of this comments helps.

regards

PMA



On 9/26/07, Hua Bao <hbao@purdue.edu> wrote:
> Dear Abinit-user,
>
> I am trying to do a CdTe band structure calculation using ABINIT. I
> followed the literature Physica B 373 (2006) 124-130. They used "
> first-principles pseudopotential method within the density functional
> formalism and the local density approximation using the Coperley-
> Alder functional for exchange-correlation energy, and standard norm-
> conserving pseudopotential of Troullier-Martins in non-relativistic
> form". Also, they use a mesh cutoff of 100 Ry and 108 k points. They
> are using a different codes.
>
> Follow them, I calculated the band structure of CdTe with the similar
> parameters. However, I got a huge underestimate of band gap. The
> literature gives 1.88 eV, I got only 0.45 eV in the ABINIT
> calculation, experimental result is 1.56 eV. I even tried GW method.
> It still got 0.97 eV which is still much less than experimental
> result (some others using GW got 1.88 eV).
>
> My question is how does this happen. Is band structure result codes-
> dependent? Is there anyway to improve the result for band structure?
>
> BTW, I am using 48cd.pspnc and 52te.pspnc from the ABINIT website.
> Here is the input file for band structure calculation.
> #Data set
> ndtset 2
>
> #Definition of the unit cell
> acell  6.480 6.480 6.480  angstrom
> rprim 0 .5 .5
>        .5 0 .5
>        .5 .5 0
> #Definition of the atom types
> ntypat 2
> znucl 48 52
>
> #coordinates
> xred 0 0 0
>       0.25 0.25 0.25
> #Definition of the atoms
> natom 2
> typat 1 2
>
> ecut 50.0         # Maximal kinetic energy cut-off, in Hartree
> #Functional
>
>
> #Data 1
> #Definition of the k-point grid
> ngkpt1 6 6 6
> kptopt 1
> prtden1 1
> toldfe1 1.0d-6
> nshiftk1 4
> shiftk1  0.5 0.5 0.5  # These shifts will be the same for all grids
>           0.5 0.0 0.0
>           0.0 0.5 0.0
>           0.0 0.0 0.5
>
> #Dataset 2
> iscf2 -2
> getden2 -1
> kptopt2 -3
> nband2  18
> prtwf2 0
> prteig2 1
> ndivk2 20 20 20
> kptbounds2      0.5  0.0  0.0 # L point
>                  0.0  0.0  0.0 # Gamma point
>                  0.0  0.5  0.5 # X point
>                  1.0  1.0  1.0 # Gamma point in another cell
> tolwfr2 1.0d-12
> enunit2 1
>
> #Definition of the SCF procedure
> nstep 100          # Maximal number of SCF cycles
> ixc 2  (I also tried 1).
>
>
>
> Hua Bao
> Graduate Research Assistant
> School of Mechanical Engineering,
> Purdue University
> O: 765-494-6944
>
>
>
>


-- 
Pierre-Matthieu Anglade