Dear Zhaoming Fu,

For the benefit of other wien2wannier users, I am taking this to the
Wien2k mailing list.  I hope you do not mind.

On 12/13/2015 10:43 AM, 付召明 wrote:
> For a degenerate energy level,  take t2g (including three orbitals:
> dxy,dyz and dxz) as an example,  Can we get three Wannier functions
> corresponding to dxy,dyz and dxz localized orbitals by wannier90 program?
> I feel it can not give them. According to the defination of wannier
> functions and the initializing steps of Wannier90 calculations, it seems
> that one energy band can only give one wannier function though this band
> include three
> degenerate states at each K point. I list the detail problems in the
> attachment.

On general grounds, where you say you have “one energy band that
includes three degenerate states at each k-point”, I would see three
bands that happen to be degenerate in this region of k-space.  However
you want to call it, what matters is the number of Bloch and Wannier
states.  The “normal” Wannier transformation gives you an equal number
of Wannier and Bloch states.  With disentanglement, you can also have
fewer Wannier than Bloch states.

Now, some comments on your concrete example, as far as I can tell what
is going on from the DOS.  It seems that the states are approximately
but not entirely degenerate.

By “disentanglement window”, I assume you mean what Wannier90 calls the
frozen window.  It seems that above -1 eV or so, you target bands are
not in fact entangled, so you could actually increase your frozen window
to [-1, 1] eV.  This may or may not help with the problem of “wrong” dxz
and dyz states.

Some other tips (which apply to many situations):

1.  Seeing that your Wannier centers are “wrong”, make sure that your
initial projections in case.inwf are correct.

2.  In case.wout, check that first Disentanglement and then
Wannierization are both converged.

3.  Also check the Wannier-interpolated band structure in case_bands.dat.

4.  Look at the band structure from Wien2k, not just the DOS.
Personally, I find a color-coded band structure most helpful (you can
use e.g. for that).  Often, this
brings to light relevant details which are otherwise overlooked.


Elias Assmann

Wien2Wannier: maximally localized Wannier functions
              from linearized augmented plane waves


Attachment: signature.asc
Description: OpenPGP digital signature

Wien mailing list

Reply via email to