Using Vasp as calculator in ASE

[karol_kawka]

Hello,

I'm trying to use VASP as a calculator in ASE to obtain minimum‑energy paths with the so‑called NEB method.
Previously I used another DFT code in ASE (Siesta) and it worked well. I’d like to compare the results from the two programs, but I’m having trouble doing the same with VASP.

It seems that during a single‑point calculation VASP produces very large forces in the first optimisation step, which show up as asterisks (*) for certain atoms. For example, the output from the second image (out of seven) has this issue—only img0 and img1 ran to completion. Is there a tag or setting that can prevent this error? This is how the NEB routine in ASE launches VASP; if there’s no workaround, I may have to switch to VASP’s own NEB.

[christopher_sheldon1]

Hello Karol,

Thank you for your question. It's not immediately clear from the image what is happening. What is the system that you are studying? I wonder if those particular lines are for a specific element or atoms that lie very closely together. This can sometimes cause problems with the images generated using NEB.

If you could upload the POSCAR, POTCAR, KPOINTS, INCAR, and OUTCAR, then it would help me to make more sense of the issue.

Best wishes,

Chris

[karol_kawka]

I uploaded files from image 2 of the procedure.

The issue is that ASE starts from the initial atomic positions and interpolates the path to the final positions. In my case, there are 7 images in total (from 0 to 6, so the uploaded one is the 3rd structure). As you can see, VASP completed the job, but the forces at the end of the OUTCAR file are shown as asterisks, so ASE couldn’t read them to determine the next step in the optimization path.

[christopher_sheldon1]

Hi Karol,

Thank you for the input and output files. The forces are far too large and meaningless. The energy is also positive (check your OSZICAR file to see that each elementary step is positive). Does this occur for all of your images?

I also notice that there seem to be some atoms missing at the edge of the cell (when viewing a 3x3x1 supercell). I suspect that this will cause big issues (though the positive energy is still unexplained).

Best wishes,

Chris

[christopher_sheldon1]

Hi Karol,

I've found the issue. Some of your ions are in the same positions. You can see this by looking at the nearest neighbor table in the OUTCAR file:

ion position nearest neighbor table
1 0.222 0.222 0.125- 165 1.94 164 1.94 163 1.94 180 1.95
2 0.111 1.000 0.125- 306 1.94 163 1.94 201 1.94 178 1.95
3 0.333 0.111 0.125- 57 0.01 164 1.93 217 1.94 201 1.95 232 1.96
4 0.332 0.221 0.290- 164 1.94 168 1.94 221 1.94 237 1.95 220 1.98
...
57 0.333 0.111 0.124- 3 0.01 164 1.94 217 1.94 201 1.94 232 1.95

E.g. 3 and 57 are 0.01 away in direct units. This creates the anomalously large forces. If you remove the duplicates, then the calculation will run normally. I suggest that you rebuild your POSCAR, though, to ensure that you have all the ions you want. My guess would be that, since all the ions are free to move, some have crossed the periodic boundary of the cell when generating the NEB images. These ions will be on both sides of the cell and so their "average" positions will be at the centre of the cell, where ions already are, resulting in two ions at the same place. Freezing ions at the edge of the cell may be a solution.

I'll close this topic on Friday, unless you have any further comments or questions?

Best wishes,

Chris