Page 1 of 1

How to perform ss-neb

Posted: Sat Mar 15, 2014 9:14 am
by Jiang12
Hi dear friends

I am doing a neb calculation of the process from bcc structure to hcp structure. I have some questions for this.

1. when I use nebmake.pl to generate the images, a warning appears, WARNING: LATTICE CONSTANTS ARE NOT THE SAME. The shape and volume of the initial
unit cell and the final unit cell are different, how to make sure that the mass center is the same? Are there some things I should pay attention to?

2. If I set LNEBCELL=TRUE, IOPT=3(This is not a very accurate one) to perform the calculation, should I perform another calculation by setting LNEBCELL=False, IOPT=1 or 2? IOPT=1 and IOPT=2, which is better?

Thank you very much

Re: How to perform ss-neb

Posted: Sat Mar 15, 2014 2:55 pm
by graeme
1) The first warning is to let you know that the initial and final cells are different, and that the ss-neb is the appropriate method to use.

2) This question is all mixed up. First, for a changing cell, use LNEBCELL=TRUE. Then, you are free to choice your optimizer of choice. IOPT=3 is conservative (slow, but will systematically converge). IOPT=1 is aggressive (fastest, but less stable away from the minimum). Once any calculation has converged, you do not need to repeat it with other optimizer settings. You may wish to reconverge with more kpoints or a high energy cutoff, and to make sure that the energy cutoff is consistent with your changing cell, but not because any optimizer is more accurate than any other.

Re: How to perform ss-neb

Posted: Sun Mar 16, 2014 2:45 am
by Jiang12
[quote="graeme"]1) The first warning is to let you know that the initial and final cells are different, and that the ss-neb is the appropriate method to use.

2) This question is all mixed up. First, for a changing cell, use LNEBCELL=TRUE. Then, you are free to choice your optimizer of choice. IOPT=3 is conservative (slow, but will systematically converge). IOPT=1 is aggressive (fastest, but less stable away from the minimum). Once any calculation has converged, you do not need to repeat it with other optimizer settings. You may wish to reconverge with more kpoints or a high energy cutoff, and to make sure that the energy cutoff is consistent with your changing cell, but not because any optimizer is more accurate than any other.[/quote]

Thank you very much. I understand. I have two other questions.
1. Is the following INCAR appropriate ?
PREC = Normal
ENCUT = 410
IBRION = 3
ISIF = 3
POTIM = 0
NSW = 100
ALGO = Normal I(blocked Davidson)
NELM = 100
NELMIN = 2
EDIFF = 1.0e-05
EDIFFG = -0.01
ISPIN = 1
INIWAV = 1
ISTART = 0
ICHARG = 2
LWAVE = .FALSE.
LCHARG = .FALSE.
ADDGRID = .FALSE.
ISMEAR = 1
SIGMA = 0.2
LREAL = .FALSE.
IMAGES = 4
SPRING = -5
LCLIMB = False
IOPT = 3
ICHAIN = 0
LNEBCELL= TRUE
ISYM = 0

Concerning the parameter LCLIMB, if the minimum energy path is monotonous, so there is not a saddle point, does it mean that it is not necessary to use the climbing image?
2. In the vasp guide, there is one phrase "If all degrees of freedom are allowed to relax (isolated molecules, no surface, etc.), make sure that the sum of all positions is the same for each cell.". If my system is a 2D-surface, should I do something to make sure that the sum of all positions is the same for each cell? I do not quite understand " the sum of all positions is the same", I feel that it is impossible to ensure it.