Question about NEB

[leon_qian]

Dear Sir,
I am currently learning to use the CI-NEB method, through which I simulated CO oxidation to CO2. I just interpolated only one image between the stable initial and final states with the "nebmake.pl" in my calculation. However, I found that the structure in the image may not be quite proper. For example, in the image, the C-O distance in CO molecule was too short(<1 angstrom) at first, and the calculation failed to converge. Then, I slightly modified the coordinates of C and O atoms in POSCAR of the image to prolong C-O distance and it converged at last. So, I would like to ask a few questions:
1. Am I right? Is the modification correct? If it is right, what's the guideline of the modification?
2. If I'm wrong, what should I do?
3. When I increased the number of images to continue, after interpolation with "nebmake.pl", the C-O distances in CO molecule were still very short in each image, and the force were also very large. To help the calculation converge, I modified the coordinates, but it did not work. Then, what should I do?

I'm looking forwards any comments. Thanks!

Best regards,
Leon

[graeme]

What ever number of images you choose, you should converge the band so that the forces are low (<0.1 eV/Ang). It doesn't make any sense to draw conclusions from un-converged images.

If you have a complex pathway, a single image may not converge. By this, I mean that the force stays high and the image geometry keeps changing. In this case, you may need to use 3 or even 5 images for a curved path. If you are using the climbing image, check that the highest energy image converges. At that point, you have a saddle and are done.

And let the optimizers relax the geometries instead of trying to do it by hand.

[leon_qian]

Dear Sir,

Thanks for your quick reply. It was really helpful to me. I will take your advice. At the present time, I may know how to proceed.
Thank you very much!