the distance between images along the CNEB band

[lgxyz]

hi,
In my CNEB calculation experience, I found that if I used few images (e.g. 4 images) and the distance between images is large (always exceed 1.0 anstrg), and the calculations always failed. If 8 images were used, the distance between images is about 0.5 anstrg, and the calculations will converge at the most time. Thus, to get a converged CNEB calculation, we must lower the distancce between images as much as possible. Is this necessary? Pls give suggestions! Thanks a lot!

[graeme]

It is not really an issue with distance. There must be enough images to resolve curvature along the path. For a linear reaction pathway, a single climbing image will converge, irrespective of the distance between images. To the extent that the path curves, there must be enough images for the tangent along the path near the transition state to fall along the negative curvature mode at the saddle.

[lgxyz]

OK! Thanks vry much!