CI-NEB calculation can't converge, the structure is damaged

[zhengjifine]

I performed a cl-neb calculation, the system always cannot converge, and the structure is damaged dramatically after about 40 steps. the barrier reached about 4.0eV, INCAR is following:

PREC = Normal
ENCUT = 400.000
IBRION = 1 ( for IBRION = 2, the results is almost the same )
NSW = 100
ISIF = 2
NELMIN = 2
EDIFF = 1.0e-05
EDIFFG = -0.05
VOSKOWN = 1
NBLOCK = 1
NELM = 300
ALGO = Normal (blocked Davidson)
ISPIN = 1
INIWAV = 1
ISTART = 0
ICHARG = 2
IMAGES = 8
SPRING = -5
LCLIMB = .TRUE.
LWAVE = .FALSE.
LCHARG = .FALSE.
ADDGRID = .FALSE.
ISMEAR = 1
SIGMA = 0.2
LREAL = Auto
RWIGS = 1.25 0.73 0.75

so what's wrong with my calculation?

[graeme]

Start by looking at similar posts on the forum. I have posted many times about how to converge a NEB, starting with careful optimization using IBRION=3 and a small POTIM, and then switching to IBRION=1 when the band is close to convergence. When you really get stuck, you can post your files -- it is impossible to see what is going wrong from an INCAR file.

[zhengjifine]

thanks very much for the reply.I will try it. I performed a cl-neb calculation that a interstitial N atom moves to the nearest interstitial site. the results are unreasonable.I add the file at the end.

[graeme]

It looks like exactly the case I warned you about. The initial band has very high forces. You need to do a careful initial relaxation of the band, e.g. using IBRION=3 and POTIM=0.01 until the forces drop below about 1 eV/Ang. This may only take 20 iterations. Then you can switch to a more aggressive optimizer. You can't start from such a high energy path using IBRION=1.

Also, it doesn't look like your endpoint structures are relaxed.

This is basic stuff that I have written about many times on this forum; please read it.

[zhengjifine]

Dear professorgraeme. How to calculate the NEB of a N atom from insterstitial site to the nearest insterstitial site in the ab plane.
I try to the energy of barrier when N atoms diffuse in the ZnO bulk. Firstly, I found that the barrier of N is about 0.4 eV from a interstitial site to neighboring interstitial site along c axis. However, when N transit from a interstitial site to the neighboring interstitial site in ab plane,the energy is negative, meaning a potential well. what's wrong with the latter or how to explain this situation?
thank you very much！

[graeme]

Two suggestions:

1) Set ISYM=0 for all simulations; your endpoints have a different symmetry from the band images. Your endpoints might not even be minima; check by relaxing them without symmetry.

2) Relax the intermediate image, 03, in your first path to see if there is a lower energy structure than your endpoint.