[Wannier] xaxis

[Gianluca Giovannetti]

Dear Jonathan (and All),

as i understand the flag xaxis helps to converge the WFs function to the
correct one but it doesn`t change the local and internal  frame which is set
from PW calculations.

I had a look to the paper suggested.
I get Wannier functions which have correct shape and they are the same
starting from different starting guess.
probably i`m in the correct minimum.

Then the only solution is to rotate the unit cell in PW calculations such to
have Wannier functions of Fe states with internal reference frame directed
along the Fe-Fe bonds (then pointing along o1-----2o bonds) from:

Y
|    ---------------------
|    |   o1          +  |
|    |                    |
|    |                    |
|    |   -           2o |
|    ---------------------
|
|------------------------------ X

to (rotating the unit cell of 45 degree):


Y
|
|                            +
|
|                  o1               2o
|
|                            -
|
|------------------------------ X

X=1,0,0
Y=0,1,0
(Z=0,0,1 is pointing out of the plane)

Then i expect the Hlda(R)_n,m is written in the correct local frame.
Is this way correct?

thank you.

Gianluca



On Sat, Feb 13, 2010 at 12:38 AM, Jonathan Yates <
jonathan.yates at materials.ox.ac.uk> wrote:

>
> On 12 Feb 2010, at 15:46, Gianluca Giovannetti wrote:
> > however i want to have printed the H(R) in the file "_hr.dat" such
> > to have Wannier functions of Fe states with internal reference frame
> > directed along the Fe-Fe bonds (then pointing along o1-----2o bonds).
> > Formally, if we want to maintain that e_g orbitals are x^2 - y^2 and
> > 3z^2 - r^2 and t_2g are xy, yz, zx then the above choice of
> > coordinate system gives
> > the correct e_g - t_2g splitting coming from the Se crystal field
> > (e_g go lower in energy and t_2g higher).
> >
> > As my hamiltonian must be written in this set, my idea would be to
> > set x'=1,-1,0 in the file ".win".
>
>
> The global x,y,z axes are set by your pw code. Wannier90 can use a
> projection of the bloch states onto a local function times spherical
> harmonics as an initial guesses for the MLWF. For example a p_z
> function. You might want the local coordinate system of this function
> to be different from the global x,y,z - in otherwords to have the p_z
> point in an arbitary direction. For this reason we allowed the
> possibility to specify the orientation for each projection.
>  So what you did to align your starting guess functions is sound.
>
> However, these functions are just starting guesses. The MLWF should be
> independent of these functions. No symmetry is imposed during the
> minimisation. Maybe the true MLWF do have the symmetry you want - in
> any case you can't automatically enforce that they have this symmetry.
>  In practice the landscape can be complex, and you might find picking
> a 'good' starting guess helps to get to the true minimum. (if two
> different initial projections lead to different sets of WF, you can at
> least compare the spreads).
>
> It might be worth looking at
> Phys. Rev. B 65, 184422 (2002)
> Maximally localized Wannier functions in antiferromagnetic MnO within
> the FLAPW formalism
>  In this situation the minimizing the spread was not sufficient to
> uniquely define the MLWF - and a further step was used to obtain MLWF
> with the desired symmetry.
>
>
>  Jonathan
>
>
> --
> Department of Materials, University of Oxford, Parks Road, Oxford, OX1
> 3PH, UK
> tel: +44 (0)1865 612797                http://users.ox.ac.uk/~oums0549/<http://users.ox.ac.uk/%7Eoums0549/>
>
> _______________________________________________
> Wannier mailing list
> Wannier at quantum-espresso.org
> http://www.democritos.it/mailman/listinfo/wannier
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20100213/a4f34a16/attachment.html>