Convergence Problem in NEB calculations.

[magnon963]

Dear Admin and VASP usres,

My NEB calculations are not converging. I'm moving adatom over graphene sheet. I am considering only straight line paths connecting 2 nearest neighbour hexagonal sites. Distance between 2 hexagonal site is around 2.5 A. I am taking 5 images in between initial and final positions. My INCAR file:

System =Graphene sheets 65 atoms in a cell

Start parameter for this run:
ISTART = 0 job : 0-new 1-cont 2-samecut
ICHARG = 2 charge : 1-file 2-atom 10-const
INIWAV = 1 electr : 0-lowe 1-rand

Electronic Relaxation 1
ENCUT = 550 eV
IALGO = 48
NELM = 220
NELMIN = 0
NELMDL = -8 # of ELM STEP m
EDIFF = 1E-04 stopping cretiria for ELM
GGA = PE
LDIAG = .T.
MAGMOM = 64*1.0 1*4.1
LREAL = AUTO

LMAXMIX = 4
INIMIX = 0
AMIX = 0.1
BMIX = 0.00001
AMIX_MAG= 0.4
BMIX_MAG= 0.000001
MAXMIX = -8

Ionic Relaxation
NSW = 180 number of steps for IOM #
EDIFFG = -0.05 stopping-criterion for IOM
ISIF = 2
ISYM = 2
ISPIN = 2
LCORR = .TRUE.
POTIM = 0.2
VOSKOWN = 1

NEB related tags
IMAGES = 5
ICHAIN = 0
SPRING = -5
IBRION = 1
LCLIMB = .TRUE.
LSCALAPACK = .FALSE.

NPAR = 25
NSIM = 1
LPLANE = .TRUE.

DOS related values:
SIGMA = 0.02
ISMEAR = 2 broad. in eV, -4-tet -1-fermi 0 gaus

NBAND = 150

-----------------------------------------------------------------------------------------------------------

I usually use Chemcraft to find images, so the distance between images is only approximately same.

Is it necessary to have distance between images strictly same??. This NEB calculation has not converged even after 400 ionic steps.

Thanks for ur help.

regards,
PS

[graeme]

You do not need to have the spacing between NEB image precisely the same. Spring forces between the images will make them equally spaced at convergence.

It is hard to know what might be going wrong with your calculation without more information. Perhaps try starting with IBRION=3, POTIM=0.1 and make sure that the forces drop steadily. When you are satisfied that things are converging, you can switch to a more aggressive optimizer, such as IBRION=1, for more rapid convergence.

[magnon963]

Thanks a lot Graeme, I'm trying it and it seems that forces are coming down.

best regards,
Pooja