Phonon Dispersion Problem

[w008egs]

I am attempting to calculate the phonon dispersion for Silicon, but I am receiving some errors which I am unable to fix. My search through the forum has not yeilded a solution so any help would be appreciated. Here is what I have tried:
Initial Calculation
First I relaxed the cell's shape, and volume and the ionic positions with the following input files:
INCAR

Code: Select all

  SYSTEM = Diamond Structure Silicon
  ISMEAR = 0  ! Gaussian smearing
  SIGMA = 0.1  ! Width of smearing
  PREC = Accurate ! Use the high end value of ECUT from the potential file
  IBRION = 5 ! use the conjugate gradient algorithm
  ISIF = 3   ! Relax volume
  NSW = 15   ! Number of steps in volume optimization
  EDIFF = 0.1E-04  ! Stop SCF cycle when energy change is less than EDIFF
  EDIFFG = -0.01  ! Stop ionic relaxation when energy change is less than EDIFFG

POSCAR

Code: Select all

Diamond Structure Silicon
  5.430  ! Lattice Constant Kittel pg 17
  0.0 0.5 0.5  ! Lattice vector  a(1)
  0.5 0.0 0.5  ! Lattice vector  a(2)
  0.5 0.5 0.0  ! Lattice vector  a(3)
  2  ! Number of atoms for each species
Direct
 -0.125 -0.125 -0.125  ! Reduced coordinates atom 1
  0.125  0.125  0.125  ! Reduced coordinates atom 2

KPOINTS

Code: Select all

KPOINTS
  0  ! Evenly spaced grid
MONKHORST PACK
  21 21 21  ! Large Monkhorst Pack grid
  0  0  0   ! Center grid about gamma

This calculation runs with no errors and provides a CHGCAR file which I can use to restart a phonon dispersion calculation. The trouble begins when I try to start a new calculation using the CHGCAR file.

Phonon Dispersion Calculation
Next I altered the POSCAR file to reflect the new relaxed cell geometry. Then I altered the KPOINTS file (as per the guide) and the INCAR file as follows:
INCAR

Code: Select all

  SYSTEM = Diamond Structure Silicon
  ISMEAR = 0  ! Gaussian smearing
  SIGMA = 0.1  ! Width of smearing
  PREC = Accurate ! Use the high end value of ECUT from the potential file
  IBRION = 5 ! use the conjugate gradient algorithm
  NFREE = 2    ! central differences
  POTIM = .015  ! 
  NSW = 1  ! ionic steps > 0
  ICHARG = 11  ! NON SELF-CONSISTENT CALCULATION
  LMAXMIX = 2  ! s or p element

KPOINTS

Code: Select all

KPOINTS
  20  ! 20 intersections between symmetry points
Line-mode
rec
  0.5   0.5   0.5    ! L
  0.0   0.0   0.0    ! Gamma

  0.0   0.0   0.0    ! Gamma
  0.375 0.375 0.750  ! K

  0.375 0.375 0.750  ! K
  0.5   0.0   0.5    ! X

  0.5   0.0   0.5    ! X
  1.0   1.0   1.0    ! Gamma

When I try to start this calculation I get the following error riddled output:

Code: Select all

 running on    4 nodes
 distr:  one band on    1 nodes,    4 groups
 vasp.5.2.2 15Apr09 complex 
 POSCAR found :  1 types and    2 ions
 LDA part: xc-table for Ceperly-Alder, standard interpolation
 error in IBZKPT_HF: two k-points are equivalent          20          21
  this will cause problems in the HF routine

 ----------------------------------------------------------------------------- 
|                                                                             |
|           W    W    AA    RRRRR   N    N  II  N    N   GGGG   !!!           |
|           W    W   A  A   R    R  NN   N  II  NN   N  G    G  !!!           |
|           W    W  A    A  R    R  N N  N  II  N N  N  G       !!!           |
|           W WW W  AAAAAA  RRRRR   N  N N  II  N  N N  G  GGG   !            |
|           WW  WW  A    A  R   R   N   NN  II  N   NN  G    G                |
|           W    W  A    A  R    R  N    N  II  N    N   GGGG   !!!           |
|                                                                             |
|      Your generating k-point grid is not commensurate to the symmetry       |
|      of the lattice.  This can cause   slow convergence with respect        |
|      to k-points for HF type calculations                                   |
|      suggested SOLUTIONS:                                                   |
|       ) if not already the case, use automatic k-point generation           |
|       ) shift your grid to Gamma (G) (e.g. required for hex or fcc lattice) |
|                                                                             |
 ----------------------------------------------------------------------------- 

 error in IBZKPT_HF: two k-points are equivalent          20          21
  this will cause problems in the HF routine
 error in IBZKPT_HF: two k-points are equivalent          20          21
  this will cause problems in the HF routine
 error in IBZKPT_HF: two k-points are equivalent          20          21
  this will cause problems in the HF routine

What mistakes have I made?

[Danny]

In your initial calculation/relaxation, you should set IBRION=2 I think IBRION=5 is for phononcalculations.
Second, I'm not sure you can actaully do phonon calculations along high symmetry lines...only Gamma...or am I incorrect?

Danny

[forsdan]

You can't perform phonon calculations along different phonon wavevectors q in VASP, just the gamma point. The KPOINTS file is only for the electronic k-points in the Brillouin zone. So the line-mode option in combination with a non-selfconsistent run is used to obtain the electronic bandstructure only. For the initial relaxation part to obtain the optimized geometry you should also use IBRION = 1, 2 or 3 as Danny points out.

In order to calculate the modes at the gamma point for the phonon spectrum, once you have your geometry, you should perform a selfconsistent run with IBRION equal to 5,6,7 or 8. The last two uses DFPT while the first two uses finite differences to obtain the force constants. Make sure the k-point grid covers the entire Brilliouin zone, i.e. that you should use the same KPOINTS file as for your initial relaxation.

As a side note: When you perform the relaxation of the volume in the first step by using ISIF = 3 you must consider the presence of Pulay stresses. So typically you will have to increase the cutoff compared to the default value with at least 30%.

Cheers,
/Dan


<span class='smallblacktext'>[ Edited Mon May 03 2010, 07:08PM ]</span>