non-symmetric CI-NEB profile

[randome]

[attachment=0]ci-neb.jpg[/attachment]Dear graeme and others:
I want to obtain the migration energy of Zr mediated by Al-vacancy in Al matrix with vasp5.3 and vtst 3.0c. However, the CI-NEB profile looks strange to me: the migration barrier of Zr is very high and the CI-NEB profile is non-symmetric. In my opinion, the CI-NEB profile should be symmetric since the local enviroment of Zr is the same befor and after its migration. One of its CI-NEB migration profile look like:

0 0.000000 0.000000 0.000000 0
1 0.567843 0.439777 -0.649155 1
2 1.135446 0.909418 -0.937834 2
3 1.702760 2.624547 -0.001909 3
4 2.270045 1.746428 0.936656 4
5 2.838103 0.700414 0.644985 5
6 3.406742 -0.000274 0.000000 6

[attachment=0]ci-neb.jpg[/attachment]

The 64-atom fcc supercell was used in my calculaitons. The problem still exists even I enlarge the superll to 108-atom or use vtst 3.0d. The whole files were packed and attached.
Any comments are highly appreciated!
Thanks in advanced!

randome

[graeme]

Try setting ISYM=0. For testing, you can use a 2x2x2 k-point mesh, PREC=normal and ENCUT=250.
Once you have a converged band, you can reconverge with your high precision settings.

There may be another subtle issue, but start with ISYM, which must be turned off when looking at a diffusion pathway.

[randome]

Dear graeme, thanks for you kind reply!
As your suggestions, I set ISYM=0 and used the following low precision setting: 2x2x2 k-point mesh, PREC=normal, ENCUT=250 eV and EDIFFG=-0.1. However, the CI-NEB profile is still non-symmetric and looks more strange, as the following figure illustrated:
[attachment=1]Zr-ci-neb2.jpg[/attachment]
However, the diffusion profile of Al mediated by Al-vacancy are symmetric and consists well other calculations. So how can I figure out the strange energy profile? The whole calculation files are packed and attached. Thanks again!

[graeme]

Alright, I think I've got to the bottom of it. I think there is some issue with problems converging the electronic structure of Zr. Setting IALGO=38 seems to make things behave properly. Incidentally, this has nothing to do with the NEB, but in my attached calculation, you can see a nice smooth symmetric barrier of just under 1 eV.

mep.png (8.64 KiB) Viewed 11334 times

[randome]

Dear graeme, very thanks for your help! I also obtained the nice symmetric barrier profile of Zr migration mediated by Al-cacancy with IALGO=38. The non-symmetric profile is indeedly related to the electronic converging problem of Zr. Thanks again.