NEB convergence

[mukul]

Dear Forum,

I doing some vacancy diffusion calculation if bcc Fe. My NEB part of the INCAR is the following:

===========================
SYSTEM = Single vacancy in bcc Fe

ISTART = 0
INIWAV = 1
ICHARG = 2

# Relaxation ======
EDIFF = 1.e-4 ! Energy Convergence
EDIFFG = -0.01 ! Force Convergence

# NEB section

IBRION = 2
NSW = 40
POTIM = 0.5

ICHAIN = 0
IMAGES = 5
SPRING = -5
LCLIMB = .TRUE.

# DOS related ======
PREC = medium
ISMEAR = 0 ! Gaussian = 0
SIGMA = 0.05 ! Broadning
LORBIT = 11

# Magnetism section ======
ISPIN = 2
VOSKOWN = 1

ALGO = Fast
LDIAG = .TRUE.
NELM = 40
NELMIN = 4

# Out-Put ======

LREAL = Auto
ISYM = 0
LCHARG = .FALSE.
LELF = .FALSE.
LWAVE = .TRUE. ! No wavefunction

==================================

I have no problem with running the NEB calculation properly with this INCAR.
My question is how do I know that - whether the NEB calculation has been converged or not.

If I look at the REAL part of "NEB: projections on to tangent (spring, REAL) " in OUTCAR -

After 100 NEB steps I am getting-

# distance barrier force(REAL)

0 0.000000 0.000000 0.000000
1 0.401172 0.257654 -1.164694
2 0.800653 0.663588 -0.558699
3 1.197449 0.745532 0.012815 (Transition state)
4 1.619693 0.685321 0.489240
5 2.039776 0.279412 1.177896
6 2.459090 0.003660 0.000000

These REAL forces are only changing in the 3rd decimal place, but all the images except the TS has a substantial force on it. However, the energy of the images (and thus the NEB barrier) are NOT changing (only in the 2nd/3rd decimal place).

Do I only need to care about the force on the TS here? How do I make sure that my NEB calculation has converged?

Thanks,
Mukul Kabir

[mukul]

I am looking a learned answer to this. Any help would be greatly appreciated.

[alex]

You have to carry out a frequency analysis. The atom in the TS has one mode with an imaginary wavenumber.

alex