zero diffusion barrier using NEB

[Tiger-paw]

Hi folks,

I need your comment/advice on a problem of zero diffusion barrier by NEB.

I understand from some earlier discussions that to prep for NEB, one need first fully relax the two end points/structures, from which a type of problem may be resulted:

For example, for a solute Y atom exchange with a 1st-NN (nearest neighboring) vacancy in a bcc lattice. After relaxation, I found that the solute atom in the two fully-relaxed end structures are very close to each other, say ~0.8 A. And perhaps due to this fact, I obtained almost zero energy variation along the band.

This should not be caused by computational errors, since using the same procedures, I have well reproduced some other solute-vacancy exchange barriers in the same lattice.

Does it mean NEB failed for this case, where the two end points have very close structure?

Any advice/comment would be very appreciated.

[admin]

no, the NEB should work in such a case as well (to check please have a look at the final ionic configurations of each of the images).
However, I wonder whether the final and the initial state are so close together. Have you relaxed both structures in 2 independent ionic scf runs?

[Tiger-paw]

Thanks for your instant reply.

I have fully relaxed the two end images.

The solute-Vac pair are originally sitting by 1/2[111] distance in the large bcc supercell before relaxation. To be more specific, in the starting image, before relaxation, the vacancy is originally at a corner site, 000, and the solute at the cubic center 0.5,0.5,0.5. After relaxation, the solute finally sits at about 2/3*1/2[111] distance from the corner 000. While in the ending image, before relaxation, the vacancy is at the cubic center and the solute at 000. After relaxation, the solute finally sits at about 1/3*1/2[111] distance from the corner 000. Therefore, the solutes are actually separated only by (1/3)*(SQRT(3)/2)*a = ~0.8 A in two end images. Consequently, for the following NEB calculation which I inserted evenly 4 intermediate images, I ended up with almost zero total energy variation among all images.

This problem might not be explained by computational errors, since using the same method and input parameters, I have well reproduced Fe-Vac exchange. Do you have any comment/ suggestion for this result ?

Thanks!

[Tiger-paw]

I have rechecked my calculations. I have a thought: this atom is a very large atom indeed. Furthermore it cannot really dissolve in Fe (the heat of solution is highly positive), so it can move a lot to vacancy in Fe during relaxation, making the atom-vacancy exchange barrier to be almost zero by NEB.

I am wondering if this is the reason for the problem, i.e. the foreign atom is actually not solvable.