Accurate force-constants by using density functional perturbation theory
Posted: Tue Dec 22, 2009 4:45 pm
Dear all,
I'm performing convergence tests for the interatomic force-constants by using density functional perturbation theory (IBRION =
. The aim in the end is to obtain accurate phonon spectrums for a couple of systems. The calculations are running smooth and I get reasonble results. However, I have some thoughts about my procedure.
1. I have checked the convergence of the frequencies with respect to the k-point grid, ENCUT, EDIFF and also checked if ADDGRID = .TRUE. influences the results. Based on these tests I set the INCAR as below. Have I missed any important parameter for further accuracy? (The POTIM tag naturally doesn't affect anything in this case.)
NOTE: I will increase the supercell size as soon as I get all basic parameters in order.
2. So far I have only used even Monkhorst-Pack grids for the k-point mesh. Do I need to include the gamma point in this case (bcc cell)? I can't see a direct reason why it should be necessary.
3. By using the phonopy software I can obtain the phonon band structure and the phonon DOS from the Hessian matrix as obtain from VASP. It seems to work very well. However, I still wonder if there is any tools included in VASP that can do the same thing? Or alternatively, does anyone know if P4VASP include this feature? I checked the corresponding homepage but all links are currently dead so I can't install it and check.
Best regards,
/Dan Fors
INCAR:
=====
SYSTEM = Fe Bulk (FM)
ISTART = 0
ICHARG = 2
EDIFF = 1E-8
IBRION = 8
ENCUT = 700.0000 eV
ENAUG = 700.0000 eV
PREC = Accurate
ADDGRID = .TRUE.
LREAL = .FALSE.
ISPIN = 2
NELM = 400
NELMDL = -8
LMAXMIX = 4
ALGO = Normal
ISMEAR = 1
SIGMA = 0.1
MAGMOM = 2*4
LWAVE = .FALSE.
LCHARG = .FALSE.
<span class='smallblacktext'>[ Edited ]</span>
I'm performing convergence tests for the interatomic force-constants by using density functional perturbation theory (IBRION =
. The aim in the end is to obtain accurate phonon spectrums for a couple of systems. The calculations are running smooth and I get reasonble results. However, I have some thoughts about my procedure.1. I have checked the convergence of the frequencies with respect to the k-point grid, ENCUT, EDIFF and also checked if ADDGRID = .TRUE. influences the results. Based on these tests I set the INCAR as below. Have I missed any important parameter for further accuracy? (The POTIM tag naturally doesn't affect anything in this case.)
NOTE: I will increase the supercell size as soon as I get all basic parameters in order.
2. So far I have only used even Monkhorst-Pack grids for the k-point mesh. Do I need to include the gamma point in this case (bcc cell)? I can't see a direct reason why it should be necessary.
3. By using the phonopy software I can obtain the phonon band structure and the phonon DOS from the Hessian matrix as obtain from VASP. It seems to work very well. However, I still wonder if there is any tools included in VASP that can do the same thing? Or alternatively, does anyone know if P4VASP include this feature? I checked the corresponding homepage but all links are currently dead so I can't install it and check.
Best regards,
/Dan Fors
INCAR:
=====
SYSTEM = Fe Bulk (FM)
ISTART = 0
ICHARG = 2
EDIFF = 1E-8
IBRION = 8
ENCUT = 700.0000 eV
ENAUG = 700.0000 eV
PREC = Accurate
ADDGRID = .TRUE.
LREAL = .FALSE.
ISPIN = 2
NELM = 400
NELMDL = -8
LMAXMIX = 4
ALGO = Normal
ISMEAR = 1
SIGMA = 0.1
MAGMOM = 2*4
LWAVE = .FALSE.
LCHARG = .FALSE.
<span class='smallblacktext'>[ Edited ]</span>