[Wannier] About Born Charge

[Malgorzata Wierzbowska]

On Wed, 6 Dec 2006, lan haiping wrote:

   Hi,

> Dear Developers ,
>
> I am wondering about born charge calculation by relative displacement
> Wannier centers.
> I tried to repeat this calculation for anatase TiO2 system. Since there are
> 6 atoms in unit cell, i just displace one of O sites about 0.1% of  cell
> lattice along x direction.  I then come to a problem about kpoint setting in
> 'nscf' calculation of the displacing  case.
> With kmesh.sh scripts, i set 4x4x4 kmesh for 'nscf' calculation. When i
> tried to run pw2wannier90.x code, calculation failed due to different
> kpoints between result of 'nscf' calculation and wannier90 's kmesh setting.

   The k-mesh in nscf for later use in wannier calculations
   has to be given in the whole Brillouin zone and it needs to be uniform.

> For my case , there are 64 kpoints i set for 'nscf' calculation.Due to
> symmetry , the number of kpoints output file gave is 164 kpoints.
> I think there must be some aspects of this calculation scheme i donot catch
> much, or understand wrong. Would you please give me some comments ?
>

   You should not use the symmetry in this case. You cannot use the option
  "authomatic" k-point generation. Instead, in nscf, you should give
  the list of all k-points.

  I know this takes longer for the calculation but at this stage it is done
  like this because we had no time yet to do it better.

  About the number of k-points: It depends on how big is your cell in the
  reciprocal space. If you have some time, you could use 2*2*2=8
  and 4*4*4=64 and 8*8*8=512 and tell us what is the difference
  at the end :-)
  If your cell is cubic you use number of k-points n1=n2=n3, if your cell
  is alongated you should find  such grid with which the distances
  between k-points (in reciprocal space) are almost equal.

  Gosia