The ABINIT Users Mailing List ( CLOSED )

[Michel Côté <michel.cote@umontreal.ca>]

Hi George,

I had a quick at the cut3d program and as far as I can tell I did not see
anything wrong. The program gets the value of the istwfk variable from the
header.

I will try to reproduce your problem in the next few days and let you know
what I find.

Michel


Le 03/05/06 08:21, « Shu Heng Pau » <gpau@MIT.EDU> a écrit :

> Hi Xavier,
> 
> I don't think it is due to cut3d as I have written a separate Matlab
> code which
> transform between the real and the fourier spaces.   My code gave the same
> results as cut3d.
> 
> However, within abinit, it is most likely consistent --- abinit might have
> computed the wavefunction in the symmetry format, and used it in the 
> symmetry
> format, although the istwfk=1 is specified.  I based my conclusion on the 
> same
> total energies that are obtained. However, cut3d must have assumed that it 
> is
> istwfk = 1 and thus the resulting wavefunction in the real space is wrong.
> 
> In addition, it is not clear whithin ABINIT if it is always the case, i.e.
> symmetry form is always used regardless of istwfk; this is because for the
> limited test cases that I have done, the error only occurs when the nband > 
> 1
> and only for the highest band.
> 
> Thanks for your attention,
> George
> 
> Quoting Xavier Gonze <gonze@pcpm.ucl.ac.be>:
> 
>> Dear George,
>> 
>> Thanks for your investigations.
>> Perhaps the bug is only in the cut3d treatment of the istwfk=1 case ?
>> 
>> Xavier
>> 
>> On 03 May 2006, at 02:20, Shu Heng Pau wrote:
>> 
>>> Hi,
>>> 
>>> I think the option istwfk = 1 is faulty.  If I insist on not using the
>>> time-reversal symmetry for k = [0 0 0]  (thus istwfk = 1), I
>>> occasionally end
>>> up getting a complex wavefunction.  When I set istwfk = 2, I always
>>> obtain the
>>> correct output --- a real wavefunction.
>>> 
>>> Thus, unless the use of istwfk =1 for k=[0 0 0] is illegal, I think
>>> there is a
>>> bug.
>>> 
>>> Thanks,
>>> George
>>> 
>>> 
>>> Quoting Shu Heng Pau <gpau@MIT.EDU>:
>>> 
>>>> Hi,
>>>> 
>>>> Thank you, Matthieu.  I believe I have done all the points
>>>> mentioned below and
>>>> am still getting a complex wavefunction:
>>>> 
>>>> 1. Gamma point used: included the keyword kpt 0 0 0.
>>>> 2. The phase is not constant: imag(psi)/real(psi) is not constant.
>>>> 3. istwfk is set to 1.
>>>> 4. and xred is set to  -0.125 -0.125 -0.125; 0.125  0.125  0.125.
>>>> 
>>>> For the cut3d, I have used option 4 and the output for the 4th psi
>>>> can be found
>>>> here:
>>>> 
>>>> http://web.mit.edu/~gpau/Public/abinit/
>>>> 
>>>> In the above folder, I have also included my input files, my
>>>> pseudopotential
>>>> file and the output file.  Note that I have excluded the nonlocal
>>>> part of the
>>>> pseudopotential and also core correction.  I have previously done
>>>> a simulation
>>>> for a single atom with simple cubic structure --- both the psi
>>>> obtained are
>>>> real though.
>>>> 
>>>> Is there any other reasons that may lead to a complex wavefunction?
>>>> 
>>>> Thanks,
>>>> George
>>>> 
>>>> Quoting Matthieu Verstraete <mjv500@york.ac.uk>:
>>>> 
>>>>> 
>>>>> Hi -
>>>>> 
>>>>> The default behavior of abinit is to use a shift of 0.5 0.5 0.5
>>>>> for the kpoints. Check the values in the kpt array of your
>>>>> output. In your case this still brings you to the "zone boundary"
>>>>> where the wavefunctions can still be chosen as real. The
>>>>> operative word is "can": as you said yourself it can have an
>>>>> arbitrary phase.
>>>>> 
>>>>> In our case the psi are stocked in reciprocal space, where this
>>>>> implies 1/2 the coefficients can be obtained from the other 1/2
>>>>> using symmetries (k -> -k+G), but for k/=Gamma I'm not sure this
>>>>> is equivalent to making the choice "psi real". So
>>>>> 
>>>>> 1) make sure you really are using Gamma and not L
>>>>> 
>>>>> 2) look at the value of istwfk which states how the coeffs are
>>>>> stocked and recreated using symmetries. If you set istwfk to 1
>>>>> and still have your problem, and are sure you're using gamma
>>>>> _and_ that it's not just a phase shift, then we may have a bug...
>>>>> Possibly in the way the wfk are output. You used cut3d? Which
>>>>> mode, etc?
>>>>> 
>>>>> 3) Oh, yeah, it may be linked to your choice of unit cell: yours
>>>>> doesn't explicitly include inversion symmetry. Use
>>>>> 
>>>>> xred -0.125 -0.125 -0.125
>>>>>       0.125  0.125  0.125
>>>>> 
>>>>> to be sure. Or try a simpler system (1 atom/UC).
>>>>> 
>>>>> Matthieu
>>>>> 
>>>>> -- 
>>>>> ================================================================
>>>>> Dr. Matthieu Verstraete                 mailto:mjv500@york.ac.uk
>>>>> Dept. of Physics, University of York,     tel: +44 1904 43 22 08
>>>>> Heslington, YO10 5DD York, United Kingdom fax: +44 1904 43 22 14
>>>>> 
>>>> 
>>>> 
>>>> 
>>> 
>>> 
>>> 
>> 
>> 
> 
> 


-- 
***************************************************************
  Michel Cote                          tel: +1 (514) 343-5628
  Professeur agrégé                   fax: +1 (514) 343-2071
  Département de physique
  Université de Montréal
  C.P. 6128, succ. Centre-ville
  Montréal (Québec) H3C 3J7          Michel.Cote@umontreal.ca
  Canada                   www.phys.umontreal.ca/~michel_cote
***************************************************************