CINEB Diverges in specific images

[yudeng]

Hello Prof. Graeme,

I am trying to use CINEB for calculation of the energy barrier for O2 decomposation on La-doped-MoS2 slab.
I first use NEB for preconvergence with IOPT=2, Attached NEB files.
However, during NEB calculation, oxygen will rotate strangely on the slab and produce a large force.
How would you suggest to approach this problem? How do I set SPRING parameters？
Thank you Professor!

I am using these INCAR tags:
SYSTEM = MoS2
ISTART = 0
ICHARG = 2
PREC = N

NBANDS = 300
ISPIN = 2
MAGMOM = 47*0 1 2*1
IVDW = 12
LORBIT = 12

LREAL = A
ISIF = 2
NSW = 600
ENCUT = 450 eV
NELM = 300
EDIFF = 1E-05
LCHARG = .F.
LWAVE = .F.
ISMEAR = 0
SIGMA = 0.1
ALGO = Fast

AMIX = 0.1
BMIX = 0.0001
ISYM = 0
NPAR = 2

IBRION = 3
POTIM = 0
IOPT = 2
ICHAIN = 0
LCLIMB = .F.
SPRING = -5
IMAGES = 6
EDIFFG = -0.05

[yudeng]

I also tested tags with IOPT = 1, 3, 7 and different SPRING, but the results still did not converge.

[graeme]

Look at my standard go-to calculation using IOP=3 TIMESTEP=0.1 with 8 images. It seems to be converging, although with an intermediate minimum.

[yudeng]

graeme wrote:
> Look at my standard go-to calculation using IOP=3 TIMESTEP=0.1 with 8
> images. It seems to be converging, although with an intermediate minimum.

Thank you Professor!
I examined your calculations and I have three questions.
1. dist.pl shows that the distance between the initial and the final structure is only 3.37, why the calculation needs 8 images? Is it necessary to do NEB pre-convergence before CINEB？
2. I find the position of O2 molecule fluctuates on the slab. Why is this phenomenon?
3. For the dissociation of the molecules with strong binding, such as O2 molecules, do we need to increase the value of the spring constant tag?

Thank you again Professor!

[graeme]

Since this path has an intermediate minimum, it is really 2 paths, so having 8 images is resolving each with only 3-4.

You do not need to do NEB before CINEB.

I expect that the fluctuations of O2 are related to the intermediate minimum. If you minimize that and resolve the two reactions separately, I expect that each will be a fairly simple process.

Finally, no, leave the spring constant alone; there is no need to change it.

[yudeng]

Many thanks for your time on this!

Good luck!