[Wannier] Different num_wann setting for pw2wannier90.x and wannier90.x....

[Jonathan Yates]

On Mon, 16 Jun 2008, Han Hsu wrote:

> I am dealing with an insulator. Its valence band contains 46 bands. Among
> these bands, 22 of them can be excluded, only the reminaing 24 bands are
> needed for MLWFs.
>
> When I ran pw2wannier90.x to calculate Amn, Mmn, and Unk, I did not realize
> only 24 bands are necessary, so I set num_wann=46 in my *.win file, and got
> 46X46 Amn and Mmn.
>
> Now I realize I only need 24 bands. Do I have to rerun pw2wannier90.x to
> get 24X24 Amn and Mmn? Or can I directly run wannier90.x by setting
> num_wann=24 with the previously calculated 46X46 Amn and Mmn?

Han,

  If the 22 bands you want to exclude form an isolated set _below_ (or 
above) the remaining valence bands then you can just do the following:

  Set num_bands=46 and num_wann=24
  Set the outer window for disentanglement to include only the upper 24 
bands (ie put dis_min_win in the gap between the lower 22 and upper 24).

  This should cause w90 to try disentanglement - which will converge 
instantly as it has no work to do. This is identical to using 
exclude_bands.
  Any problems, just report back.


  If the 22 bands happen to form an isolated set within the original 24 - 
eg as for the 3d bands in ZnSe - then the only option is to use 
exclude_bands. No fundamental reason for this, it could be coded into w90.

  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