UT theoretical chemistry code forum

Hi,
I wonder is that possible to set up an array of spring constants for an NEB calculation? i.e. for NEB with images of unequal distances.
Sometimes this feature is very useful. For example, in calculating a methane activation, the methane molecule may move in a long distance before 'actually' reacts and raise the energy. In this case, I would like to use fewer images to sample the 'move' of methane, but use a higher resolution for the real C-H breaking part.
I can use ASE (atomic simulation environment) to do this type of NEB calculation, but sometimes a native VASP implementation would be more convenient because of the parallelization issue.

Thanks in advance.
Best,
Geng

Many years ago we did play around with variable spring constants. Specifically, we had a spring constant that scaled with energy, so that the images near the saddle had a higher resolution. Another strategy we tried was a spring constant that scaled with curvature along the path. With the climbing image, however, these tricks didn't seem as valuable, since there just needs to be sufficient image density to resolve the saddle. That said, it would be very easy to try other ideas. Perhaps based upon testing with ASE, if there were a compelling strategy, we could implement it in the vtstcode.

Dear Prof. Henkelman,
Thanks for your reply.
If I understand your comments correctly, you are referring to an automatic updating of the spring constants during optimizations.
I was trying to figure out if it is possible to set up different K, which are inversely proportional to inter-image distances. Once the spring constants are set, they are not updated during the optimization.
I do not have a rigorous test right now, but in practice, I found the unequal Springs are very helpful.
1) I can manually insert a few images between the original path, only for the highest energy images.
2) I can use a guessed TS to create the initial path between IS and FS. For example, I can add a guessed M between the IS and FS. The interpolation of IS and M may not have the same distance as that from interpolation between M and FS. Hence the unequal Ks are required to do the NEB.

Best,
Geng