[Wien] wien2wannier error

[Elias Assmann]

On 06/13/2013 07:31 AM, wasim raja Mondal wrote:
> (1) For the generating of *.nkp file, after running Write_win, one has
> to run wannier90.x - -pp subdir which will creat the *.nnkp file if you
> have used prepare_w2w case subdir. This is not mentioned in the user
> guide. only wannier90.x will not work. This is a small thingh. But it
> creates lot of tension in the first time user mind like me.

What user guide are you reading?  The one I downloaded from 
<http://www.wien2k.at/reg_user/unsupported/wien2wannier/wien2wannier_userguide.pdf> 
says (p. 3, under "Initialization (init_w2w)"):

wannier90.x(preliminary run): wannier90 requires the input data to be 
given on a special k-mesh which includes information with respect to 
nearest-neighbor k-points. A preliminary call of wannier90 with the 
option -pp stores this mesh to the file case.nnkp.


> (2) Before running init_w2w, one hast to do subdir.sym file. In the file
> one has to set number of symmetry operation is 1 and first operation has
> to to be identity. This is discussed by phillip in the wien2k forum and
> in the UG also. I think this symmetry is strongly related with k-points
> generation. I may be wrong.

I assume you mean the "case.ksym" file.  This is generated (rather, you 
are asked to generate it) in init_w2w.


> (3) I am felling that we are generating k-points in the init_w2w, which
> is showing the number of k-points for LAPW computation. There is
> previous generated k-points also which is called k-points for band
> structure calculation. The number k-points for LAPW should be less than
> the number of k-point for band structure k-point. In this point I may be
> completely wrong. I want some experts comment on that.

The Wannier projection is done on a full-BZ unshifted k-mesh as 
generated in init_w2w.  This k-mesh can be rather sparse (8x8x8 already 
gives well-converged results for SrVO3, for example), but since no 
symmetries are considered, you still tend to end up with a lot of 
k-points to run lapw1 on compared to your standard Wien2k run.

write_win also reads the case.klist_band file, tries to sift out the 
"special" points (it takes those that have labels) and writes those to 
case.win (section kpoint_path).  Wannier90 will then (if bands_plot is 
True) make a band structure for a BZ path generated from those points. 
The discretization of that path is controlled by the variable 
bands_num_points.


> (4)If third point is right, than one has to increase the number of
> k-point in the band structure . In this point I am struggling at the
> moment. For example I have run SrVO3 example in wien2k. In the scf I
> have used 8000 k-points. In the band structure calculation I have
> generated k-points in the simple cubic not with xcrysden. In the
> init_w2w run, it is showing 10 k-points for band structure calculation.
> Now the point is that how can I increase the k-point for band structure
> calculation?

Note the last thing I said above -- the number of k-points in 
case_band.dat is independent of all the "klist" files; it is instead 
controlled by bands_num_points in case.win.


HTH,

	Elias