UT theoretical chemistry code forum

Hi All,

I am trying to perform a CI-NEB calculation for a proton transport reaction that believed to be associated with a extremely low barrier of about 0.08 eV.

Issue# : The images that are next to the end points are lower in energy as the NEB converges. It has already been discussed in the forum that this sort of behavior would likely arise if the parameters (KPOINTS, ENCUT, PREC, etc) used in the geometry minimization and CINEB are inconsistent or if the end points are not properly minimized. In this case, the end points are converged with the force convergence of EDIFFG = -0.00040 which is many orders stricter than the NEB force convergence EDIFFG = -0.05. The inputs are shown far below.

It would be great, if someone can find any issue with my setup or any pointers on how to get around with this issue.
Thanks !

The convergence using nebef.pl is as follows:
0 0.000369 -941.408700 0.000000
1 0.017650 -941.408800 -0.000100
2 0.032270 -941.408800 -0.000100
3 0.036451 -941.379000 0.029700
4 0.035822 -941.343700 0.065000
5 0.027859 -941.330900 0.077800
6 0.020152 -941.330700 0.078000
7 0.025605 -941.324300 0.084400
8 0.017760 -941.329500 0.079200
9 0.047168 -941.354900 0.053800
10 0.068932 -941.388700 0.020000
11 0.040333 -941.406000 0.002700
12 0.017457 -941.412500 -0.003800
13 0.000363 -941.412400 -0.003700

#---------------------------------------NEB input script ----------------------------------------------
SYSTEM = Graphene-proton
PREC = High
ENCUT = 520
LPLANE = .TRUE.
NCORE = 20
ISPIN = 1
NELECT = 392

# Electronic Relaxation
NELM = 80
EDIFF = 1E-09
NELMIN = 5
LREAL = Auto

NSW = 1000
IBRION = 1
ISIF = 2
ISYM = 0
POTIM = 0.10
EDIFFG = -0.05

TEBEG = 800
TEND = 800
SMASS =-3

Write flags
LWAVE = .FALSE.
LCHARG = .FALSE.

DOS related values:
ISMEAR = 0
SIGMA = 0.05

NEB simulation
LSCALAPACK = .FALSE.
IMAGES = 12
SPRING = -5
ICHAIN = 0
LTANGENT = .TRUE.
LCLIMB = .TRUE.
ICHAIN = 0

#---------------------------------------Geometry Minimization----------------------------------------------
SYSTEM = Graphene-proton
PREC = High
ENCUT = 520
LPLANE = .TRUE.
NCORE = 20
ISPIN = 1
NELECT = 392

# Electronic Relaxation
NELM = 80
EDIFF = 1E-09
NELMIN = 5
LREAL = Auto

NSW = 1000
IBRION = 1
ISIF = 2
ISYM = 0
POTIM = 0.010
EDIFFG = -0.00040

TEBEG = 800
TEND = 800
SMASS =-3

Write flags
LWAVE = .FALSE.
LCHARG = .FALSE.
NWRITE = 1

DOS related values:
ISMEAR = 0
SIGMA = 0.05

Goodness, if we are able to worry about the accuracy of our DFT calculations to 0.1 meV, we're in pretty good shape!

Anyway, if I did have to worry about this problem, it would be that the endpoint energies are very close to the neighboring images. I suspect that the atoms in the cell are translating at this point. Small changes in the positions of the atoms with respect to the FFT grid can give rise to changes in energy of this magnitude. Using LREAL=.False. may help. Freezing an atom could also prevent translation.