The ABINIT Users Mailing List ( CLOSED )

[Jorge Iniguez <jiniguez@nist.gov>]

Hi Artem,

Thanks for your suggestion! I increased the cutoff from 1000 eV to 1500
eV, which should be enough to see some improvement. The 1000 eV results
were:

  Phonon wavevector (reduced coordinates) :  0.00000  0.00000  0.00000
 Phonon energies in Hartree :
   8.954702E-04  1.138484E-03  1.197535E-03  2.116134E-03  3.353738E-03
   3.641500E-03  4.411735E-03  4.837036E-03  5.507493E-03
 Phonon frequencies in cm-1    :
-  1.965330E+02  2.498684E+02  2.628285E+02  4.644378E+02  7.360604E+02
-  7.992168E+02  9.682638E+02  1.061607E+03  1.208755E+03

and the 1500 eV results are:

  Phonon wavevector (reduced coordinates) :  0.00000  0.00000  0.00000
 Phonon energies in Hartree :
   8.957446E-04  1.138428E-03  1.197495E-03  2.116817E-03  3.353848E-03
   3.643423E-03  4.411971E-03  4.837699E-03  5.506935E-03
 Phonon frequencies in cm-1    :
-  1.965932E+02  2.498561E+02  2.628197E+02  4.645877E+02  7.360846E+02
-  7.996390E+02  9.683157E+02  1.061752E+03  1.208633E+03

As you see, things have not changed much, not even for the acoustic modes.

But note something important: It would have been a surprise that the lack
of degeneracies had to do with the cutoff.  I ask abinit to calculate the
phonons by constructing the dynamical matrix out of an irreducible set of
symmetry independent perturbations.  Hence, the symmetries should be exact
by construction, no matter what cutoff I use.

Actually, the above test indicates this probable bug (in the construction
of the dynamical matrix from the minimal set of RF calculations) may be
responsible for the nonzero acoustic modes too, since increasing the
cutoff does not seem to help.

Cheers,
Jorge


On Fri, 8 Aug 2003, Artem R. Oganov wrote:

> Dear Jorge,
>
> It seems to me that you are using a too small Ecut. Have you tried to
> increase it and see what happens? This would probably give you
> nearly-zero acoustic frequencies, but also could resolve your problem:
> ABINIT always gives some small deviations from the exact degeneracy
> (consider this as a numerical error), and I would expect these
> deviations to become smaller as you increase Ecut - in other words, in a
> crude calculation the degeneracy may be hidden by  the errors due to a
> small number of plane waves. If you try your calculation with a bigger
> Ecut - please let me know what you see.
>
> With my best regards,
>
> Artem
> ************************
> Dr. Artem R. Oganov
> Lab. of Crystallography
> Department of Materials
> ETH Zurich
> CH-8092 Zurich
> Switzerland
> Tel.: +41-1-632-3752; Fax: +41-1-632-1133
> E-mail: a.oganov@mat.ethz.ch
> Website: http://olivine.ethz.ch/~artem
> ************************
>
>
> -----Original Message-----
> From: Jorge Iniguez [mailto:jiniguez@nist.gov]
> Sent: Thursday, August 07, 2003 7:35 PM
> To: forum@abinit.org
> Subject: [abinit-forum] phonons of hexagonal systems: probable bug
>
> Dear abinitioners,
>
> We have found what seems to be bug in the calculation of phonons of
> hexagonal systems. Input and output files go attached.
>
> We have studied MgB2, which has 3 atoms per unit cell and space group
> P6_3/mmm.  At the Gamma point, this system should have four different
> optical phonon frequencies, two of them being doubly degenerate.  When
> we
> try to calculate them, abinit correctly recognizes four (as opposed to
> nine) irreducible perturbations:
>
>  The list of irreducible perturbations for this q vector is:
>     1)  idir= 1 ipert= 1
>     2)  idir= 3 ipert= 1
>     3)  idir= 1 ipert= 2
>     4)  idir= 3 ipert= 2
>
> However, the final result is a list of frequencies...
>
>   Phonon wavevector (reduced coordinates) :  0.00000  0.00000  0.00000
>  Phonon energies in Hartree :
>    8.954702E-04  1.138484E-03  1.197535E-03  2.116134E-03  3.353738E-03
>    3.641500E-03  4.411735E-03  4.837036E-03  5.507493E-03
>  Phonon frequencies in cm-1    :
> -  1.965330E+02  2.498684E+02  2.628285E+02  4.644378E+02  7.360604E+02
> -  7.992168E+02  9.682638E+02  1.061607E+03  1.208755E+03
>
> .... that does not show the expected degeneracies!!  I could only briefly
> look for the bug, but did not succeed, so I turn to the experts out
> there :)
>
> Cheers,
> Jorge
>
> PS: I guess that the nonzero acoustic frequencies are to be attributed
> to
> the problem described in the FAQ...
>
>
> ------------------------------------------------------------------------
> ------
>    Jorge Iniguez             NIST Center for Neutron Research
>                              100 Bureau Drive, Mail Stop 8562
>                              Gaithersburg, MD 20899-8562
>                              Phone: 1-301-975-8367 Fax: 1-301-921-9847
>                              E-mail address: jiniguez@nist.gov
>                              http://www.ncnr.nist.gov/staff/jorge
> ------------------------------------------------------------------------
> ------
>
>
>