The ABINIT Users Mailing List ( CLOSED )

["Yong Liu" <yongliu.yl@gmail.com>]

Dear Ganesh,

     Do you use the experimental lattice constants or theoretical ones? You'd better fully relaxe the unit cell before perform the RF calculations, since the residual stress may introduce noises. And you can use anaddb to obtain the better results by imposing ASR. Aloutgh I am not the expert in your problem, I think the residual criterion is good enough. Moreover you can try tolvrs 1.0e-12 instead of toldfe 1.0e-20 in the GS dataset.

 

2008/1/2, P. Ganesh <pganesh@ciw.edu>:

Dear All,

Wish you all a very Happy New Year.

I  am trying to compute the phonon DOS of a 10 atom primitive cell
(insulator material).  I am using Abinit 5.4.2.

I find that the accoustic phonons at Gamma (the first 3 numbers in the
freq. list) are far away from 'zero'.  They are all around ~30 cm-1.  I
have used a 6x6x6 kpt. mesh and 65 Ha cutoff.  I also use 'E' filed
perturbation to get the Lo-To splitting.  Apart from the common input
the special inputs for my 4 datasets are below.  (The runs are done
separately and not from same input file)  The nband has been set exactly
to the number of occupied bands in all calculations.

Dataset-1: (GS calculation)

toldfe 1.0d-20

Dataset-2: (non-self consistent calculation)

   iscf   -3         # Need this non-self-consistent option for d/dk
kptopt   2          # Modify default to use time-reversal symmetry
tolwfr   1.0d-22    # Use wave function residual criterion instead
getwfk 1
getden 1

Dataset-3: (ddk calculation at G)
nqpt   1
qpt           0.00000000E+00  0.00000000E+00  0.00000000E+00
   iscf   -3         # Need this non-self-consistent option for d/dk
kptopt   2          # Modify default to use time-reversal symmetry
rfphon   0          # Cancel default
rfelfd   2          # Calculate d/dk wave function only
tolwfr   1.0d-22    # Use wave function residual criterion instead
getwfk 1
rfdir 1 1 1

Dataset-4: (phonon perturbation at G)
nqpt   1
qpt           0.00000000E+00  0.00000000E+00  0.00000000E+00
  getwfk   1          # Use GS wave functions from dataset1
  getddk 1
  kptopt   2          # Need full k-point set for finite-Q response
  rfphon   1          # Do phonon response
  rfatpol   1 10        # Treat displacements of all atoms
  rfdir   1 1 1      # Do all directions (symmetry will be used)
  tolvrs   1.0d-22     # This default is active for sets 3-10
  rfelfd 3
  prtwf 0

I would like to know why the phonon freq. are around 30cm-1 and not
zero.  If it is the kpt. or ecut. that I should increase, which one
matters more?  Is my convergence criteria good enough to get phonons ?


Thanks,
Ganesh

-- 
Best regards.

-------------------------------------------------------------------------------------
Liu, Yong, Ph D
Institute of Inorganic and nonmetal Materials,
Department of Materials Science and Engineering,
Zhejiang University,PR China