NEB with ISIF=3,4 etc

Vasp transition state theory tools

Moderator: moderators

raspberry
Posts: 15
Joined: Mon Oct 24, 2011 6:18 pm

NEB with ISIF=3,4 etc

Post by raspberry »

hi,
can we run NEB calculations with ISIF=4 in VASP (that is relaxing ions and allowing cell shape to change at constant volume) or ISIF=3 (allowing ions to move, and cell shape and volume to change as well)?
If it is possible, is there any flag I should be paying attention to?
I would appreciate your response.
xph
Posts: 39
Joined: Tue Mar 13, 2012 9:33 pm

Re: NEB with ISIF=3,4 etc

Post by xph »

Yes, you can use the solid state NEB by setting LNEBCELL = True, ISIF=3, and IOPT =3
There are details discussed here:
viewtopic.php?f=2&t=1136
raspberry
Posts: 15
Joined: Mon Oct 24, 2011 6:18 pm

Re: NEB with ISIF=3,4 etc

Post by raspberry »

Do I have to use a specific version of VTST?
Also, I am a bit confused about LNEBCELL flag. Could you please define it for me?
Thanks.
xph
Posts: 39
Joined: Tue Mar 13, 2012 9:33 pm

Re: NEB with ISIF=3,4 etc

Post by xph »

You may need to update to the latest version of VTSTCODE.

LNEBCELL is the flag to turn on solid state NEB, which includes the cell variables in the reaction coordinates.
The paper on the SSNEB method can be found here:
http://theory.cm.utexas.edu/henkelman/p ... 074103.pdf
raspberry
Posts: 15
Joined: Mon Oct 24, 2011 6:18 pm

Re: NEB with ISIF=3,4 etc

Post by raspberry »

Thank you for replying to my question.
When running NEB with ISIF=3, is it necessary to set the PSTRESS flag (in what units)?
If not, is PSTRESS=0 the default value?
Thanks.
xph
Posts: 39
Joined: Tue Mar 13, 2012 9:33 pm

Re: NEB with ISIF=3,4 etc

Post by xph »

Yes, the default is PSTRESS = 0, in the unit of 0.1GPa (kB) . It is not necessary to set PSTRESS.
raspberry
Posts: 15
Joined: Mon Oct 24, 2011 6:18 pm

Re: NEB with ISIF=3,4 etc

Post by raspberry »

hi,
I was trying to run NEB with ISIF=3 flag for a relatively simple system with 2 atoms in the unit cell. The job went through 3 ionic iterations, but then it stopped giving the following is the error message:
internal ERROR: DEPLE: IRDMAX must be increased to 275
internal ERROR: DEPLE: IRDMAX must be increased to 402
internal ERROR: DEPLE: IRDMAX must be increased to 275
internal ERROR: DEPLE: IRDMAX must be increased to 405
internal ERROR: DEPLE: IRDMAX must be increased to 407
internal ERROR: DEPLE: IRDMAX must be increased to 405
internal ERROR: DEPLE: IRDMAX must be increased to 407

Could you please comment on what this error message means?
I would greatly appreciate it.
Thanks.
xph
Posts: 39
Joined: Tue Mar 13, 2012 9:33 pm

Re: NEB with ISIF=3,4 etc

Post by xph »

Can you post your INCAR so that we can look into it ?
raspberry
Posts: 15
Joined: Mon Oct 24, 2011 6:18 pm

Re: NEB with ISIF=3,4 etc

Post by raspberry »

hi,
Below is the INCAR. I greatly appreciate your help.
============================================
SYSTEM = R.S.
NWRITE = 1
ISTART = 0

ENCUT = 400.0 eV
PREC = HIGH
EDIFF = 1E-05
EDIFFG = -1E-03

LCORR = T
IBRION = 3
NSW = 100
ISYM = 0

ISMEAR = 0; SIGMA = .01
IALGO = 38
LDIAG = T
LREAL = F

LNEBCELL = T
ISIF=3
IOPT =3
NPAR = 2
IMAGES = 7
SPRING = -5
LCLIMB = T

NBANDS = 16
LWAVE = F
=======================================
xph
Posts: 39
Joined: Tue Mar 13, 2012 9:33 pm

Re: NEB with ISIF=3,4 etc

Post by xph »

Seems like that you missed POTIM = 0.0
But I am not sure whether the error come from this. Also try to set a larger ENCUT (1.3*ENMAX).
raspberry
Posts: 15
Joined: Mon Oct 24, 2011 6:18 pm

Re: NEB with ISIF=3,4 etc

Post by raspberry »

hi,
Thank you for your quick response. It seems like the POTIM =0.0 flag helped to get rid of the "internal ERROR: DEPLE: IRDMAX" problem. However, I am not sure if my job is correct for two reasons: First, in the CONTCAR files after the two lines giving atoms' direct coordinates, I find some strange characters (see below):

-0.0000000026441520 0.0000000011814896 0.0000000053589062
0.5000000026441521 0.4999999988185105 0.4999999946410940

NaN NaN NaN
NaN NaN NaN

The "converged" images are very similar to the "intial" (created) images. Second, I tried to run my calculation with PSTRESS flag set to 100. Again, the "NaN" characters appear, but also the energies are exactly identical to the ones I found for the PSTRESS=0 run, which does not make sense I suppose.
I would appreciate if you would comment on what could be wrong.
Thank you.

Here is my INCAR file again. I am using vasp 5.12.2 and my system is relatively simple with 2 atoms per unit cell undergoing a phase transition.
SYSTEM = R.S.
NWRITE = 1
ISTART = 0
ENCUT = 400.0 eV
PREC = HIGH
EDIFF = 1E-05
EDIFFG = -1E-03
LCORR = T
IBRION = 3
NSW = 100
ISYM = 0
ISMEAR = 0; SIGMA = .01
IALGO = 38
LDIAG = T
LREAL = F
LNEBCELL = T
ISIF=3
POTIM = 0.0
IOPT = 3
NPAR = 2
IMAGES = 7
SPRING = -5
LCLIMB = T
NBANDS = 16
LWAVE = F
xph
Posts: 39
Joined: Tue Mar 13, 2012 9:33 pm

Re: NEB with ISIF=3,4 etc

Post by xph »

Those should be the velocities of the atoms. It doesn't matter in SSNEB calculation.
There is no surprise that the converged structures are similar to the linear interpolated ones, since only two atoms are included in the supercell. I think the barrier should be lower than the initial guess. You can try nebconverge.pl to track the forces and energies for a SSNEB run.
raspberry
Posts: 15
Joined: Mon Oct 24, 2011 6:18 pm

Re: NEB with ISIF=3,4 etc

Post by raspberry »

Hello,
thanks for your reply. I appreciate your time.
However, I don't think my calculations are correct. I just checked that the energies I get from the SS-NEB run are identical to single point (no relaxation) energies for my images, which implies that no shape/volume optimization was performed during the ss-neb run. The geometries I get from the ss-neb run are identical to my original images. Also, as I noted before, imposing external stress (PSTRESS) did not change the energies. Lastly, the nebconverge.pl script does not yielding the following message: "mv: cannot stat `vaspout1.eps': No such file or directory".
Can you please comment on how to check whether the ss-neb is actually running.
Thank you in advance.
xph
Posts: 39
Joined: Tue Mar 13, 2012 9:33 pm

Re: NEB with ISIF=3,4 etc

Post by xph »

Is the OUTCAR in each image folder finished?
raspberry
Posts: 15
Joined: Mon Oct 24, 2011 6:18 pm

Re: NEB with ISIF=3,4 etc

Post by raspberry »

Yes, it is.
Post Reply