No barrier along MEP from NEB

[woaibama]

Dear Dr. Henkelman,
I was trying to use NEB and dimer to find the MEP and TST for a reaction on MgO(100) surface. However, after the NEB converges to the force below -0.1 eV/A, the energy keeps decreasing along the MEP and the tangent force on each image is also all positive. The initial state and the final state have been optimized to the force below 0.05 eV/A. It seems there's no barrier for this reaction, which however is not reasonable. Is there something wrong with my INCAR settings as shown below? Can you please help me to figure out why? Thanks!

Electronic structure
PREC = NORMAL
ADDGRID = .TRUE.
LREAL = Auto
ENCUT = 400
ISYM = 0
ISPIN = 2
Electronic minimization
IALGO = 48
EDIFF = 1e-4
NELM = 1000
MAXMIX = 80
BMIX = 4
ISMEAR = 0
SIGMA = 0.2
Density functional
GGA = PE
Ionic relaxation
ISIF = 0
EDIFFG = -0.1
NSW = 1000
IBRION = 1
POTIM = 0.1
Job Params
LWAVE = .FALSE.
NEB
ICHAIN = 0
IMAGES = 8
ISPRING = 1
LCLIMB = .FALSE.
LDNEB = .TRUE.
Parallelization
NSIM = 4
NPAR = 4
LPLANE = .TRUE.
LSCALU = .FASLE
Other parameters
IDIPOL = 3

[graeme]

Well, I would have to see the calculation to get to the bottom of this (and will if you attach it to a post), but I can describe a few issues which can cause this. I see that you have symmetry off, which is good; make sure that is is also off for the endpoints, so that you know the endpoint has not converged to a symmetric state which is not a minimum. Second, if you have a low barrier near one of the endpoints, it may not be resolved with the resolution of your band. Turning on the climbing image can help with this because it will push one of the images up to the saddle. Increasing resolution be adding more images is also possible, but since you already have 8 images I think the climbing image is a much better option. Finally, you can use the neb2dim script to switch between an NEB calculation and a single-image dimer search, which like the climbing image, will only converge to the saddle.

[woaibama]

Dear Dr. Henkelman,

Thanks a lot for your quick reply. I think this problem is solved. After plotting the spline.dat energy along the reaction coordinate, I found a small peak which looks like a reasonable TST. The image right before this peak seems slightly more stable than the initial reactant state, thus the energy profile looks all goes down. The energy barrier is so small that it can not be seen until the forces of NEB are converged to a value small enough.

Thanks again for your help.