from converged neb to dimer: ts goes away....

[kai]

Hi, all,

I have such a problem to examine the barrier corresponds to the barrier of flipping the Si-dimer on the famous Si(001) surface. I started by climbing neb, using 3 images, gamma point k, and the calculation successfully converged to every image below 0.02 ev/A. TS is image #2

As a next step, i tweaked up k-mesh to 221, and i turned to the dimer method. I started from the structure of previous neb ts structure, generated MODECAR from image #1 and #3. However, the calculation appears to drift away from TS.

DIMCAR:
Step Force Torque Energy Curvature Angle
1 0.79196 2.22529 -689.47803 -0.05269 8.39019
1 0.79196 1.79098 -689.47803 -0.21251 12.66320
1 0.79196 1.93659 -689.47803 -0.34641 11.15903
1 0.79196 1.82265 -689.47803 -0.46111 9.41741
2 0.56822 2.53633 -689.49585 0.50137 6.08050
2 0.56822 2.38090 -689.49585 0.48074 7.26693
2 0.56822 2.37506 -689.49585 0.48513 5.86436
2 0.56822 2.46124 -689.49585 0.52810 7.57540
3 0.70682 3.14154 -689.46733 0.77514 -0.35409
3 0.70682 3.13991 -689.46733 0.76921 4.51139
3 0.70682 3.05952 -689.46733 0.79272 1.42769
3 0.70682 3.02414 -689.46733 0.77780 1.48496
4 0.84691 3.18173 -689.43448 0.99776 -4.59753
4 0.84691 3.08622 -689.43448 1.03419 -0.22461
4 0.84691 3.09120 -689.43448 1.03370 0.38756
4 0.84691 3.08001 -689.43448 1.03427 0.07787
5 1.07428 3.47539 -689.39200 0.71152 -1.29994
5 1.07428 3.51904 -689.39200 0.72328 0.13260
5 1.07428 3.51275 -689.39200 0.72327 -0.02025
5 1.07428 3.51361 -689.39200 0.72330 0.00983
6 1.33913 3.14813 -689.33800 0.97536 0.00712
6 1.33913 3.14720 -689.33800 0.97688 -0.07852
6 1.33913 3.14807 -689.33800 0.97717 0.01788
6 1.33913 3.14787 -689.33800 0.97713 -0.00317
7 1.62116 3.46599 -689.27034 1.35962 4.20846
7 1.62116 3.51514 -689.27034 1.31415 -0.11275
7 1.62116 3.51437 -689.27034 1.31286 0.01526
7 1.62116 3.51444 -689.27034 1.31303 -0.00691
8 1.90995 3.31805 -689.18560 1.68274 2.84761
8 1.90995 3.39782 -689.18560 1.66275 -0.26209
8 1.90995 3.39241 -689.18560 1.66156 0.06151
8 1.90995 3.39363 -689.18560 1.66176 -0.01510
9 2.20850 3.98320 -689.08104 1.82547 -0.35611
9 2.20850 3.97275 -689.08104 1.78972 0.07866
9 2.20850 3.97419 -689.08104 1.78975 -0.03946
9 2.20850 3.97315 -689.08104 1.78980 0.02738
10 2.52954 3.63465 -688.95385 1.89024 0.97218
10 2.52954 3.64857 -688.95385 1.87923 -0.27431
10 2.52954 3.64379 -688.95385 1.87989 0.11248
10 2.52954 3.64567 -688.95385 1.87961 -0.04681


I suspect i did something wrong with the incar setting for the dim run, as below

INCAR_DIM

PREC = Normal
ENCUT = 400.000
IBRION = 3 # NEB/dimer needs 1 or 3, not 2
NSW = 100
ISIF = 2
NELMIN = 6
EDIFF = 1.0e-07
EDIFFG = -0.02
VOSKOWN = 1
NBLOCK = 1
NWRITE = 1
NELM = 60
ALGO = Normal (blocked Davidson)
ISPIN = 1 # no spin polarized calculation
INIWAV = 1
ISTART = 0
ICHARG = 2
LWAVE = .FALSE.
LCHARG = .FALSE.
ADDGRID = .FALSE.
ISMEAR = 0 # recommended for semiconductor
SIGMA = 0.05 # recommended for semiconductor
LREAL = A # Manual recommends A for mixed ions or more than 20 atoms
POTIM = 0

# NEB/dimer
#ICHAIN = 2 # 0 switches on NBE;
#IMAGES = 3 # number of images, exclude is and fs
#SPRING = -5
#LCLIMB = .TRUE.

# DIMER PARAMETERS
ICHAIN=2 # swithes on dimer
DdR=0.005
DRotMax=4
DFNMin=0.01
DFNMax=1.0

# optimizer
IOPT = 2 # cg
MAXMOVE = 0.1
#TIMESTEP = 0.1



Could anyone give some help on what i did wrong?

I attach neb-gama and dim-k221 calculatiions here

[kai]

further update,

a contiuned neb at k221 successfully converged the ts to 0.02 ev/A....

[graeme]

I've been having trouble with the k221 dimer calculation using the gamma point saddle from the NEB. I'll try to figure out what's going on.

Nice to hear that the k221 NEB worked.