Hi, I'm getting the following error/bug
internal error in: nonlr.F at line: 969
RSPHER: internal ERROR: 33 -43 2 1
I'm using version 6.5.0
This is the setting files + some of the output files includding the error message in "output.log" file:
I hope you can help me with this issue.
Dear Marco,
Thank you for your report. I will try to repeat the calculation to work out what is going on. Your system is quite large. If you have a minimal reproducible example, this would help to solve the problem quicker.
Best wishes,
Chris
Dear Chris,
Thank you for helping me!
Unfortunately, I don't have a smaller version of this system/simulation. However, I have the same system using gold instead of copper as the metal atom. In this case, the simulation works. If it's helpful for you, I could share the files.
Best regards,
Marco A.
Hi Marco,
An example using copper would be helpful. Could you include the OUTCAR along with the input files?
Many thanks,
Chris
Hi Chris,
This is the example using copper. This system is the same, but changing the metal atom.
Best regards,
Marco A. Villena
Thank you, Marco. I'll compare them once my calculation using the Au atom has finished and get back to you then.
Best,
Chris
Hi Marco,
Sorry for the slow reply, I was testing a few different things. What you see is an issue with the Brent algorithm in VASP. In some difficult cases, it fails. We are working to improve this.
As a workaround, you can try to tighten the energetic convergence criterion EDIFF=1E-5 or 1E-6. This will ensure that the calculation does not terminate so abruptly.
I think it would also be worth considering whether the dispersion correction that you want to use is suitable. The D3 functional tends to overbind for metal ions. When I switched off the dispersion correction, the calculation converged in 39 steps. IVDW=202 (the many-body dispersion, MBD), will take more time but would likely improve the situation. dDsC (IVDW=4) is also a possibility for metals.
Let me know if this helps or if you've got any follow-up questions.
Best wishes,
Chris
Hi Chris,
Thank you for your suggestions. I’ll try to decrease the EDIFF.
In this situation, I’m using D3 because the bottom part of the system is a 2D material, so I think that I should keep it. However, any suggestions in this sense are welcome.
Best regards,
Marco A.
Okay, glad I could help. I should make clear that all the dispersion corrections (e.g., dDsC, MBD) work for non-metals, so they would be appropriate for your system.
Let me know if EDIFF makes the difference.
Best wishes,
Chris