[Wannier] Wien2k2Wannier90

Jonathan Yates jry20 at cam.ac.uk
Wed Sep 17 18:42:47 CEST 2008

On Wed, 17 Sep 2008, fatemeh.mirjani wrote:

> I have written an interface code between Wien2k & Wannier90. I tested my 
> code for Silicon (4valance bands), but unfortunately the bandstructure 
> of Wien2k2Wannier90 doesn't coincide to the bandstructure of 
> PW2Wannier90. The total shape of band-structure from Wien2k to Wannier90 
> matches with the shape of band-structure from PWscf to Wannier90, but it 
> oscillates a lot! Generally speaking, at the kpoints with higu symmetry 
> and the first and the end of bandstructure there is agreement with PWscf 
> but in other points there are a lot of oscillations.

  Remember that in plotting the bandstructure with wannier90 you are doing 
an interpolation. The original points (ie those you computed using the 
ab-initio code) will be reproduced exactly - I suspect these happen to 
coincide with the high-symmetry points you mention. But all other points 
are interpolated.
  The fact that you get lots of oscillations suggests to me that the 
Hamiltonian does not decay rapidly in your wannier basis - eg because your 
wannier functions are not well localised. Indeed you say:

> (the amount of spread in seedname.wout is very large and it cannot be 
> improved by changing the centers and increasing the number of kpoints.)

  That's not a good sign. For silicon you should be able to get four well 
localised, bond centred wannier functions. The spread should be very close 
to the value in the original MV paper (or as given by pwscf+w90). You need 
to solve this problem before you can expect to get a good band structure.


Dr Jonathan Yates         |    Theory of Condensed Matter Group
Corpus Christi College    |    Cavendish Laboratory
Cambridge, CB2 1RH, UK    |    Cambridge, CB3 OHE, UK
email jry20 at cam.ac.uk     |    Tel +44 (0)1223 337461

More information about the Wannier mailing list