[Wannier] pw2wannier90 runtime error at file = 'sn.nnkp'

Shu-Ting Pi pipidog at gmail.com
Wed Mar 9 20:09:10 CET 2016


Hello CS,

I checked your band structure, i think it is very good enough. NCPP is
usually less accurate than PAW, but your bands fit very well. When it comes
to wannierize, there will be much larger errors. So the slightly difference
between your NCPP and PAW in high energy really doesn't matter.  You should
just focus on bands around Ef.

If u really want improve your NCPP calculation, u can try to increase ecut
in your NCPP calculation. Typically, PAW requires around 350-450eV. USPP
around 450-550eV. NCPP around 550 - 650 eV are reasonable choices. But it
is not guaranteed.

Best,
Shu-Ting
On Mar 9, 2016 9:51 AM, "Cuauhtemoc Salazar" <salazar at physics.utoronto.ca>
wrote:

> Dear All,
>
> I’d like to tell that the error I reported in this thread is solved.
>
> There was a problem in my .win file:  num_bands was set equal to the
> number of bands in the abinitio nscf run,  but it should be set to the
> actual number of bands passed to wannier90.x. Since I am using
> exclude_bands, the correct value for num_bands is nbnd_nscf minus the total
> number of excluded bands.
>
> The wannierization now runs till completion. I might keep posting more
> questions about this calculation, but I think I should close the topic of
> this thread.
>
> Dear Shu-Ting, I still wonder if you, or someone else, would have any
> comments on whether my simple “bandstructure benchmark" of your NCPP would
> be satisfactory (attached figure in my previous post on this thread).
>
> Thanks a lot,
> Temok
>
>
>
> On Mar 9, 2016, at 05:26, Cuauhtemoc Salazar <salazar at physics.utoronto.ca>
> wrote:
>
> Dear Shu-Ting and General Audience,
>
> Shu-Ting, thank you very much for your comments and for sharing with me
> your NCPP file, it is very kind of you. Using your file, I have run the
> relaxation and plane wave convergence. In the attached figure I compare
> both the PAW and  NCPP band structures. They have a good match at low
> energies around the Fermi level, which is what I am interested in, and both
> give band gaps similar to what is reported in the literature. However,
> there are differences at 2eV, and beyond. I do not expect perfect match,
> since they come from different pseudo potentials, but I wonder if this
> ‘’bandstructure check’’ is good enough to trust your NCPP file, given that
> I am only interested in energies around +/- 1eV around the Fermi level.
>
> In regard to the purpose of my Wannier calculation, I wish to obtain a few
> MLWF’s around the band gap, and then construct a Wannier-tight-binding
> model for optical properties. The idea is to compare this Wannier model
> with another one we already have. Then as a first step I should obtain the
> .xsf files and this is where I am stuck. Using the NCPP scheme, I get the
> following error
>     from pw2wannier90 : error #         1
>     write_unk not implemented with ncls
>
> From other posts in this forum, it seems that plotting the MLWFs is NOT
> implemented for SOC (correct?), but it is indeed possible to compute the
> MLWF’s, as it is done in example17 ( I ran it successfully).
>
> Following this example17, I tried to get the MLWF and their spreads (write_unk
> = .false., wannier_plot = false, plot_bands=false), but then I get this
> other error:
>
> Found a mismatch in sn.eig
> Wanted band  : 21 found band  : 1
> Wanted kpoint: 1 found kpoint: 2
>
> A common cause of this error is using the wrong
> number of bands. Check your input files.
> If your pseudopotentials have shallow core states remember
> to account for these electrons.
>
>  Exiting.......
>  param_read: mismatch in sn.eig
>
>
> I inspected the input files but can’t figure out what is causing this
> mismatch (nbnd in pwscf and num_bands in .win are the same). Would this
> be related to the fact that the NCPP includes those d-electrons?
>
> Thanks,
> Temok
>
> PhD Candidate
> University of Toronto
>
>
>
>
>
>
>
>
>
>
> <bandsPAW-zoom_wide.pdf>
>
> On Mar 8, 2016, at 01:00, Shu-Ting Pi <pipidog at gmail.com> wrote:
>
> Hello Cuauhtemoc,
>
> You didn't provide enough information to figure out your problems. I just
> list a few suggestions.
>
> To perform a SOC wanniersation using QE, beware of the following things:
>
> 1. You MUST use norm-conserving pseudopotential (NCPP) in QE / W90
> inferface.
>     NCPP with SOC is not provided in most cases. So you have to generate
> by yourself.
>     I guess you are not familiar with pseudopotential generation, so I
> have generated
>     one (using the optimized norm conserving Vanderbilt pseudopotential)
> for you. I'm
>     not sure if it works. You should have a benchmark calculation.
>
>     If you still want to use QE as your major platform, make sure you have
> your own
>     tool to generate NCPP with SOC.
>
> 2. You can also use other free ab initio software to perform wannier
> calculation, e.g.
>      abinit.  They usually provide PAW compatible w90 interface. However,
> as far as
>      I know, only QE is always interfaced with the latest w90 (currently
> v2.0)  which
>      means you will not be able to get several properties, e.g. spin, in
> wannier calculation
>      if you use other ab initio softwares.
>
>      Therefore, I will still recommend you to learn QE, but you will need
> to generate
>       NCPP by yourself. Google it.
>
> 3. For most systems, you should assign the projection instead of random
> projection.
>     Random project only works for simple systems. I don't think your case
> work for that.
>
> 4. Reference ex.17 in the tutorial to make sure you did each step
> correctly.
>
> Hopefully it helps !
>
> ----
> Shu-Ting Pi
> Postdoctoral Researcher
> University of California, Irvine.
>
>
> On Mon, Mar 7, 2016 at 9:14 PM, Cuauhtemoc Salazar <
> salazar at physics.utoronto.ca> wrote:
>
>> Dear All,
>>
>> I am facing the following error
>>          At line 886 of file pw2wannier90.f90 (unit = 99, file =
>> 'sn.nnkp')
>>          Fortran runtime error: Bad integer for item 1 in list input
>>
>> I am using
>>      PW2WANNIER v.5.3.0 (svn rev. 11974)
>>      PWSCF             v.5.3.0 (svn rev. 11974)
>>      Wannier90         Release: 2.0.1 (2nd April 2015)
>>
>> and Wannier90 got build automatically during the build of the espresso
>> suite.
>>
>> I wish to compute
>> — MLWF’s and
>> — the wannier-Interpolated bandstructure
>> for stanene, a monolayer of Sn atoms, including spin-orbit effects.
>> Although the band gap of this system is at K, the gap at Gamma depends
>> significantly on whether or not the d-electrons of Sn are included as
>> valence electrons (i.e., Z valence =14). I could not find norm-conserving
>> pseudo potentials with these characteristics, hence I am using PAW.
>>
>> As an initial step, using espresso’s pwscf, I got well converged
>> structural and plane wave parameters, and a bandstructure in agreement with
>> the literature.
>>
>> The scf and nscf input files have
>>         lspinorb = true
>>         noncolin = true
>>
>> In regard to wannier90, besides the basic set up, the .win file includes:
>>         spinors = true
>>         guiding_centres = true
>>         wannier_plot = true
>>         bands_plot = true
>>
>>         Begin Projections
>>         random
>>         End Projections
>> (There are two atoms per unit cell)
>>
>> Similar errors were reported in the forum due to a mismatch between the
>> homogeneous k-grid used in the nscf and the .win files, but in my case this
>> is OK, they are the same (generated with w90/utility/kmesh.pl)
>>
>> The pw2wann file is simply:
>>         &inputpp
>>           prefix='sn'
>>           seedname = 'sn'
>>           write_unk = true
>>           outdir='./work/'
>>         /
>>
>> Could this error be because I am either adding or missing a keyword in my
>> input files?
>>
>> I would appreciate your advice on this,
>> With regards,
>> Temok
>>
>> PhD Candidate
>> University of Toronto
>>
>>
>>
>> _______________________________________________
>> Wannier mailing list
>> Wannier at quantum-espresso.org
>> http://mailman.qe-forge.org/cgi-bin/mailman/listinfo/wannier
>>
>
> <Sn_ONCV_PBE-1.1_r.upf>
>
>
> _______________________________________________
> Wannier mailing list
> Wannier at quantum-espresso.org
> http://mailman.qe-forge.org/cgi-bin/mailman/listinfo/wannier
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20160309/94f18b64/attachment.html>


More information about the Wannier mailing list