NEB Not Locating Transition Point

[cswang]

Sorry for the repost, but I think my last one was confusing, and I've redone my job to make it clearer for others to follow.

Essentially, I am trying to run a CO oxidation reaction on Pt embedded in single vacancy graphene (from this paper http://pubs.rsc.org/-/content/articlepd ... c2cp41441d). However, when I run NEB, I encounter the following issues:
- the images spread away from the transition point and so do not find a meaningful energy barrier
- the molecules involved perform a bunch of random spinning that effectively waste the images

I am attaching a job. I would be incredibly appreciative of anyone who could help me. Thanks so much. (Graeme, I'm sorry if I'm pestering you...)

[graeme]

The endpoints were not well-minimized. The attached NEB calculation from converged endpoints shows a barrier of 0.63 eV (run2 subdirectory). This is not an example of an efficient calculation; rather, it uses a conservative optimizer to find a reasonable path. I keep saying this on this forum: you should use inexpensive settings to find transition states. There was no need for a 3x3x1 k-point mesh for this calculation. My calculation was done with an order of magnitude less cpu time than the one posted, which did not converge. You can easily take this calculation and reconverge it with more accurate settings, now that you have a reasonable transition state.

[cswang]

Thanks so much for the help. Sorry for the trouble, but I am learning a lot from you!

A few questions for clarification:

- Did run2 reach convergence? (I'm not complaining, it's closer along than my job is obviously, but I'm a little confused if I should continue the job and finish it before increasing to the 3x3x1 KPOINTS (or if I should even use higher KPOINTS at all).)

- What's the difference between run1 and run2? The INCAR and KPOINTS are the same and the NSW is set high enough that I don't think you ever hit the step limit. Did you just check run1 and then use its CONTCAR for run2 just to check progress midway?

- Are you saying I needed to use a stricter convergence on the initial and final image, but a looser one for the NEB calculation itself? (Originally, I had better converged initial and final images but was advised by someone else to rotate the molecules along a more reasonable reaction pathway...)

- Was it mostly the KPOINT setting that was holding me up? The INCAR settings seemed largely the same as my own.

- What is the Pt_NEB file folder and subdirectories for?

[cswang]

Hi Graeme, I didn't know what to do so I continued your job (same settings and everything) on Friday, and it's been running all weekend without reaching convergence.

Using nebef.pl, I see some of the forces converging, but some are increasing:

0 0.028045 -680.293900 0.000000
1 0.019058 -680.296900 -0.003000
2 0.064905 -680.291900 0.002000
3 0.179782 -680.116600 0.177300
4 0.033045 -679.684000 0.609900
5 0.078006 -679.797000 0.496900
6 0.054415 -682.226300 -1.932400
7 0.006119 -683.744500 -3.450600
8 0.008350 -683.994700 -3.700800
9 0.000000 0.000000 680.293900


And using the nebbarrier.pl, I find that the barrier is getting larger (and moving away from the reported value):

0 0.000000 0.000000 0.000000 0
1 0.803311 -0.002971 0.001858 1
2 1.604370 0.001795 -0.026296 2
3 2.422417 0.173848 -0.690243 3
4 3.226271 0.609949 0.042116 4
5 4.281157 0.497569 0.356189 5
6 5.313782 -1.932565 7.124959 6
7 6.346395 -3.450779 0.562014 7
8 7.380025 -3.700821 0.052017 8
9 8.414957 680.293939 0.000000 9

I guess I"m just wondering if I'm doing something wrong or if I should just leave everything alone...

Thanks a lot!