Dear VASP developers,
I've searched the forum trying to find a similar error and the closest I've been able to find is treated in https://www.vasp.at/forum/viewtopic.php ... ror#p31885, although I reckon the origin of the issue is different in this case. That's the reason why I'm opening this thread.
I believe this issue (internal error in: nonlr.F at line: 969) can be attributed to the use of KERNEL_TRUNCATION/LTRUNCATE to appropriately treat electrostatics on a system with a net charge in VASP 6.5.0, which is the most recent version available in my HPC cluster. Please find attached the files that represent a minimum implementation of the system I'm trying to model (POSCAR, INCAR, KPOINTS, POTCAR). I also attach the OUTCAR and the standard output in case you find them useful before re-running the systems on your own.
For the sake of argument, this bug occurs when doing an ionic relaxation of a charged molecule (anion, in this case, although also shows up when dealing with cations) in gas phase. The first SCF calculation is completed and then this bug is encountered and the simulation stops. The behaviour seems to be a bit unpredictable, as if I restart the simulation sometimes it goes beyond the first SCF calculation... but eventually this issue is encountered sooner or later.
You might find interesting that this bug doesn't happen when modelling the interactions of said molecule with another element (e.g., interacting with a finite graphene surrogate), although the use of KERNEL_TRUNCATION/LTRUNCATE seems to hamper the convergence of the electronic cycle – no convergence found after a few ionic steps after 150 electronic iterations, besides dramatically slowing down the calculations. My understanding is that both of these "problems" are somewhat expected.
Happy to follow up on this issue if needed and looking forward to hearing from you about it.
Best,
Carlos
