[Wannier] Format of UNK files

[Robert Peyton Cline]

Hi Hung,

Thanks a bunch for checking that.  I did indeed exclude higher-energy bands
in one case, and in the other case, I set an outer window that yielded the
same ndimwin as the excluded-band case.  So the bands that I actually used
to construct my wannier orbitals were the same between the calculations (no
core states were excluded).  So, that makes sense why the plotting routines
didn't crash on me.  I had a hunch that was the case, but I just wanted to
be sure.

I'll go ahead and regenerate the files and redo some of the plots just to
be safe, but this makes sense now.

Best,
Peyton Cline

On Tue, Jul 2, 2019 at 5:55 PM Hung Pham <phamx494 at umn.edu> wrote:

> Hello Peyton,
>
> I checked the plot.F90 which is used to generate MLWFs from UNK. There
> will be a problem if UNK files from the two calculations have different
> grid size and No. of k-point.
> The number of bands in UNK (i.e., *nbnd*) and the number of the required
> band = No. of MLWFs (i.e., *num_bands*) can be different. Meaning *nbnd* is
> not necessarily equal to *num_bands*
> The plot.F90 will read the first *num_bands* th band when constructing
> MLWFs.
> With that being said, I guess in your case, you excluded some upper bands
> hence the first *num_bands* th band in both UNK files from two setup are
> actually the same. I think if you exclude some core bands then the
> generated MLWFs from different setup should be different.
>
> I think if you're not sure which UNK file you have in hand then regenerate
> them would be a good idea. Otherwise, you can read the No. of band in UNK
> like the script I sent and check (still not 100 % sure which bands you have
> in UNK file). But the problem would be two setup with the same No. of bands
> (by excluding the same number of different bands) then I don't know how you
> can be sure which one should be from which setup.
>
> maybe other people here have a better practice.
>
> Hung
>
> On Tue, Jul 2, 2019 at 6:34 PM Robert Peyton Cline <
> Robert.Cline at colorado.edu> wrote:
>
>> Hi Hung,
>>
>> Ok, thanks for the info and the link.  However, this raises another
>> question for me if you have the time.  I have 2 different projection setups
>> of the same parent VASP job, and additionally, one of these setups sets the
>> exclude_band tag, while the other one does not.  I generated all the
>> wannier.* files from the same WAVECAR file using IALGO = 2 in VASP, so all
>> the information about the overlaps and the eigenvalues are the same between
>> the two setups (except the effect of the exclude_bands tag, of course).  I
>> mistakenly mixed up the UNK files between the runs, so I would have guessed
>> there should have been some clash with the UNK file format considering I
>> either set or did not set exclude_bands.  However, when I got my plotting
>> results back, everything appeared to run OK, and my Im/Re ratios were
>> good.  So my question is - is there some safety protocol in place that
>> ensures this behavior?  I would have figured my jobs would have crashed
>> because of the mismatch, but that's not the case.  Any help would be
>> appreciated.
>>
>> Thanks,
>> Peyton Cline
>>
>> On Tue, Jul 2, 2019 at 5:25 PM Hung Pham <phamx494 at umn.edu> wrote:
>>
>>> Hi,
>>>
>>> Indeed, the number of bands in UNK file depends on the 'exclude_bands'.
>>> I wrote a python function to read the UNK files generated by VASP. It is
>>> a part of my MCU code to analyze VASP output and so.
>>> But if you are only interested in the UNK file reader then it can be
>>> accessed here:
>>> https://github.com/hungpham2017/mcu/blob/master/mcu/vasp/utils.py#L108
>>>
>>> Hung
>>>
>>>
>>> On Tue, Jul 2, 2019 at 6:11 PM Robert Peyton Cline <
>>> Robert.Cline at colorado.edu> wrote:
>>>
>>>> Dear Wannier developers,
>>>>
>>>> I have a basic question about the format of the UNK* files that I can't
>>>> uncover in the source code.  Are the UNK* files written differently
>>>> depending on whether the 'exclude_bands' tag is set in the wannier90.win
>>>> file?  It would make sense to me if they were, considering the other files
>>>> (*amn, *mmn, and *eig) are written without the bands declared by
>>>> 'exclude_bands', but I just find get a clear answer anywhere and I can't
>>>> check the files themselves since they are written in binary.
>>>>
>>>> Many thanks,
>>>> Peyton Cline
>>>> 5th year PhD student
>>>> Eaves Group
>>>> CU-Boulder
>>>> _______________________________________________
>>>> Wannier mailing list
>>>> Wannier at lists.quantum-espresso.org
>>>> https://lists.quantum-espresso.org/mailman/listinfo/wannier
>>>>
>>>
>>>
>>> --
>>>
>>> Hung Q. Pham
>>> Gagliardi Group
>>> Office: Smith 101
>>> Email: phamx494 at umn.edu
>>>
>>> Department of Chemistry
>>> University of Minnesota - Twin Cities, Minneapolis, MN 55455
>>>
>>
>
> --
>
> Hung Q. Pham
> Gagliardi Group
> Office: Smith 101
> Email: phamx494 at umn.edu
>
> Department of Chemistry
> University of Minnesota - Twin Cities, Minneapolis, MN 55455
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20190702/f8ca48e1/attachment.html>