UT theoretical chemistry code forum

Dear sir :
Recently during my NEB calculation,I am confused a lot by the limitation of the current NEB that the two end points must have the same cell constant and can not be changed during the calculation. I want to ask that when will the public can get the version that with the variable cell shap,that is somrthing about the parameter NEBCELL .Also I find that some paper have done such calculation.for eaample,'nature materials | VOL 4 | FEBRUARY 2005 '.
Best wishes
Ross

It's a good suggestion; we're working on it.

Dear sir:
Sometimes if we use ISIF=3 in the INCAR of VASP with the NEB now available , the cell shape of the IMAGE changes, is this activated energy reasonable?
By the way, sir,according to your work,when will the public will get the script of NEB with variable cell shape?
ross

If you let the cell relax along the band, make sure that the changes in the cell are continuous. If you see the cell geometry change abruptly, it means that the cell degrees of freedom are participating in the reaction. In this case the barrier is not meaningful.

We should have the neb working with the cell geometry over the summer.

Dear all,

I'm in the progress to perform CI-NEB calculations where the initial and final state have rather different shape and volumes, if they are relaxed using ISIF = 3. So in connection to the above discussion I am wondering that if the cell geometry change abruptly along the band using ISIF = 3, would it still be possible to at least estimate upper and lower limits for the diffusion barrier from constant cell calculations? I can always postpone the calculations until the NEBCELL option is available, but I'm interested if it's conceptual possible (a priori) to choose a set of calculations that will do this. Naively I'm thinking about performing two sets of calculations where one simulation is performed using the cell volume and shape from the initial state for all images and one simulation where the cell volume and shape from the final state is used for all images. The atom positions in the final (initial) state must then of course first be re-optimized using ISIF = 2 before starting the NEB method. However, I can't find any motivation to why this would give limiting values, it would just give a approximate size of the barrier, so I'm might be out on thin ice.

Does anyone have any experience from estimating diffusion barriers where the initial and final cell volume and shape differs from constant cell calculations? I'll have a look at the "Nature materials | VOL 4 | FEBRUARY 2005" paper, and if anyone know any additional papers I would appreciate if you could let me know.

Cheers,
/Dan

I think you are right that a relaxed cell NEB will give you a lower bound to the barrier and any frozen box calculation will give you an upper bound. Perhaps using a linearly interpolated box would be as good as using either endpoint, because then you at least have the correct energies of the endpoints and can extract a meaningful barrier.

There are a couple of papers that I know of which talk about processes where the cell changes shape. There is a 2005 PNAS paper by Caspersen and Carter, and a more recent paper about transformation in Fe from the group of Duane Johnson.