Page 1 of 1

NEB Convergence

Posted: Wed Oct 02, 2013 5:45 am
by thomasd
Hello, I am trying to calculate the migration energy of a He atom from a tetrahedral to an octahedral interstitial location in a Y2Ti2O7 system. I am using the Nudged Elastic Band method, but I am unable to get the system to converge.

Also, in calculating the initial and final positions, I have simply calculated the path for the intermediate images and my understanding is that the NEB method will determine the minimum energy path from there. However, I am concerned with some things in the way that I have done this. The cell geometry in each of the interstitial locations is different after full relaxation of a single He atom in each location (for initial and final locations). When calculating the path from the initial to the final location, does the change in the cell geometry need to be accounted for, or can I simply use the same cell geometry for all images? I know the fully relaxed configurations of both the initial and final positions, but keeping the cell shape and volume the same does not allow for the relaxed position of the final position to be in its fully relaxed configuration.

I have attached a test I have run.

Thank you very much.

Re: NEB Convergence

Posted: Thu Nov 21, 2013 6:44 am
by graeme
You do not have our vtst code linking into your vasp binary. To calculate pathways with different cell geometries, you should use our code ( http://theory.cm.utexas.edu/vtsttools/downloads/ ) and include the LNEBCELL=.TRUE. flag to use the generalized solid-state NEB.

Re: NEB Convergence

Posted: Tue Jan 28, 2014 9:02 pm
by thomasd
Hello, I have gotten my NEB tests to converge. This time I am running a test for BCC Fe with a single Helium interstitial migrating from a tetrahedral position to another tetrahedral interstitial position. My results are not agreeing with prior calculations of the migration barrier in the literature. However, the problem seems to be in the fixed endpoints, as these two points are the outliers on the plot of the migration barrier. I am wondering if, in the initial relaxations of these two points, we should use the same setting for relaxation as what is used in the NEB calculation, that is with ISIF = 2. If I remove the two endpoints from the graph. the barrier is very near what it should be which is between .06 and .08 eV, making me think that the NEB calculation is working fine, but that the relaxation of the endpoints is not exactly correct.

Would you have any suggestions on why these two endpoints do not seem to be correct? Or do you see anything else that may be wrong with my NEB calculation?

Attached is the NEB calculation.

Any help would be greatly appreciated,
Thomas Danielson

Re: NEB Convergence

Posted: Tue Jan 28, 2014 10:39 pm
by graeme
The endpoint calculations need to have the same parameters as the NEB calculation.

It looks to me that you used different versions of vasp, a different number of kpoints, and a different projection scheme.

Using ISYM=0 may fix the difference in the k-point sampling, which is the most likely cause of the difference in energy that you are seeing. But anyway, just use the same binary, KPOINT, POTCAR, and INCAR files (expect for the NEB-specific tags).