[Wannier] problem with slab calculation

[Jonathan Yates]

On Tue, 3 Apr 2007, Marc de Wergifosse wrote:

> I'm using wannier90 and pwscf for calculation on a silicon slab.
> I need a big sampling of k points to get the Hamiltonian on the Wannier
> functions with a good precision.
> I have done my nscf calculation on a grid of 24 24 1 k points, which I get
> with the script kmesh.pl.
> When I want to do the -pp operation with wannier90 on the same grid, I
> obtain this error :

> Exiting.......
> kmesh_get_automatic: Unable to satisfy B1 with any of the first 12 shells

Hi Marc,

  There is a problem finding the b-vectors to satisfy Eqn B1 of the 
Marzari, Vanderbilt paper (finite difference formula). You've found that 
our routine to choose those vectors is not as robust as it should be!

  Wannier90 tries to choose the b-vectors from amongst the first 12 nearest 
neighbour shells of kpoints. In your case I suspect the first 12 shells 
all lie in the plane of the slab; we'd have to look further out to find 
the shell corresponding to the direction perpendicular to the slab.

  Could you send me the *.win file, so I can check this is indeed the 
problem. I'll send you back a patched version of the code, and we'll 
include the changes into the next release.

  Thanks
    Jonathan



-- 
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