SSNEB for phase transition in Oxide

[Neutrino]

I have an issue with SSNB to study a martensitic phase transition in a metal oxide and any tips would be very appreciated.
In all cases I did linear interpolation between the first and final states. First I tried to do energy cutoff convergence on 5 intermediate images. The lower energy cuttoffs (<450eV) converge to a saddle point, while the higher ones keep oscillating (every now and then the max force on an atom in an intermediate image shoots up to 50+ ev/A and then go down after few ionic steps).
So I tried these high energy cuttoff on larger number of images (11), and again I observe these oscillations in the forces.
Any suggestions?

The VTST version I use is VTST: version 3.0b, (11/11/12), and here is a typical INCAR:

SSNEB Metal Oxide
ISTART=0 ; ICHARG = 2
ISMEAR = 0 ; SIGMA =0.05
ENCUT= 600
IBRION=3
POTIM=0
IOPT=3 ; ISIF=3; NSW=5000
EDIFF = 1E-06
EDIFFG = -0.01
LWAVE = .FALSE.
LCHARG= .FALSE.
PREC=Accurate
IMAGES=5
SPRING=-5
LCLIMB=.TRUE.
NPAR=1
LNEBCELL=.TRUE.
TIMESTEP=0.04

[Neutrino]

A related question, is ISYM=0 recommended for this type of calculations?

Thank you!

[graeme]

First, yes, I suggest always using ISYM=0 for NEB calculations unless you really know what you are doing.

Poor convergence with a high energy cutoff sounds strange. Check to make sure that the electronic iterations are converging. Then, I would try reducing the TIMESTEP variable to see if you can converge (even slowly) with the high cutoff.

That's all I can think of; we would have to look at the output for other suggestions.

[Neutrino]

Thank you for your help! It was indeed the electronic convergence that lead the high energy cutoff calculations to show this wild behavior. I started from scratch using cutoff 600 eV , EDIFF = 1E-08, and TIMESTEP=0.03. This calculation while is still running, seems to be ok.