saddle point of cNEB

[hexiujie]

I have finished a cNEB calculation. But the mep.eps shows that there is no image sits right at the saddle point.Why？Please give me some suggestion, thank you very much!the INCAR is as follows：
PREC = Accurate
ISTART = 0
ICHARG = 2
ENCUT = 400
IMAGES = 8
SPRING = -5
ICHAIN = 0
LCLIMB = T
IOPT = 2
NELM = 80
NELMIN = 2
NELMDL = -8
EDIFF = 1.E-06
GGA = 91
VOSKOWN = 1
LREAL = A
ALGO = Fast
EDIFFG = -0.03
NSW = 500
IBRION = 3
ISIF = 2
POTIM = 0
ISMEAR = 0
SIGMA = 0.05
LPLANE = T
NPAR = 8
ISYM = 0
NSIM = 4
LWAVE = F
LCHARG = F

[graeme]

We would need to see some of the output to know why. Is the run converged?

And, are you sure that you have our code linked into vasp? If not, the climbing image tag will be ignored.

[hexiujie]

The results of nebef.pl are as follows:
0 0.00000000 -168.19069600 0.00000000
1 0.01305000 -168.15477200 0.03592400
2 0.02422600 -167.25659400 0.93410200
3 0.01487900 -166.64542600 1.54527000
4 0.00908800 -166.62037000 1.57032600
5 0.00092800 -166.60048900 1.59020700
6 0.00477000 -166.62851900 1.56217700
7 0.02577400 -166.97955000 1.21114600
8 0.01483000 -168.15475000 0.03594600
9 0.00000000 -168.19087100 -0.00017500
So it is really converaged and when I grep for VTST in the OUTCAR to see if the code has been compiled into the binary. I really see something like:
VTST: version 2.03d, (02/18/09)
Why there is no image at the saddle point? Thanks!

[graeme]

So isn't image 5 right at the saddle? It is the highest energy image and the force on the image is low.

[hexiujie]

part of the spline.dat are as follows：
5 1.59045
5.05 1.59024
5.1 1.58982
5.15 1.58921
…… ……
6.4 1.59179
6.45 1.59252
6.5 1.59078
…… ……
so, 1.59045 is not the maximum. And image 5 is not the saddle point. Why? Thank you very much!

[graeme]

Ok, I think I'm starting to understand. So you are saying that the spline interpolation is showing a second saddle along the path.

Please post the contents of your neb.dat file, or the entire calculation, so I can recreate the spline and calculate the extrema found.

[hexiujie]

The contents of the neb.dat file are as follows：
0 0.00000 0.00000 0.00000 0
1 0.57878 0.03575 0.01385 1
2 1.15495 0.9684 -2.87329 2
3 1.72417 1.54421 -0.09036 3
4 2.29537 1.57097 -0.03536 4
5 2.86048 1.59045 0.00321 5
6 3.54456 1.56574 0.03835 6
7 4.23201 1.31292 1.80079 7
8 4.91565 0.03651 0.02382 8
9 5.59881 -0.000175 0.00000 9
Additionally, The initial state is graphene with a H atom adsorbing on one of the C atom, the final state is graphene with a H atom adsorbing on another C atom. Thank you very much！You are so kind！

[hexiujie]

By the way, the contents of exts.dat file are as follows：
Extrema 1 found at image 0 with energy: 0
Extrema 2 found at image 0.9664 with energy: 0.03588
Extrema 3 found at image 1.00345 with energy: 0.03573
Extrema 4 found at image 4.97809 with energy: 1.59047
Extrema 5 found at image 6.02624 with energy: 1.5654
Extrema 6 found at image 6.44144 with energy: 1.59255
Extrema 7 found at image 9 with energy: -1.75E-4
And in the INCAR file, IMAGES=8.

[graeme]

The calculation looks ok; the saddle is at the highest energy image number 5.

If you run the nebspline.pl script and look at the interpolation in mep.eps, you can see where the second high-energy extrema is being introduced between images 6 and 7. This is most likely caused by a small overall translation in your system between these images. An artificially longer distance between these images would cause the interpolation, which is matching the forces (slopes) at the endpoints, to overshoot.

You could either just ignore the interpolation, or align each structure along the neb so that the C atoms coincide as closely as possible. This alignment will change the image spacing and should give you a more accurate interpolation.