Page 1 of 1
Relaxation for noncollinear magnetization with Hubbard U crashes
Posted: Mon Aug 14, 2023 11:14 pm
by vahid_askarpour
Dear VASP Community,
Attached please see the files I used to relax an antiferromagnetic structure with Hubbard U with VASP-5.4.4. The code crashes with no clear message. Any help would be appreciated.
Thanks,
Vahid
Re: Relaxation for noncollinear magnetization with Hubbard U crashes
Posted: Wed Aug 16, 2023 10:40 am
by manuel_engel1
Dear Vahid,
I tried to reproduce the error in a limited environment without success. This issue could be related to multiple things. While I investigate the issue, could you please answer the following questions:
1) Are you perhaps running out of memory? How much RAM do you have available on that machine?
2) Can you reproduce the error by rerunning the calculation? If yes, does it also appear when you reduce the number of k-points?
3) The backtrace suggests that there might be an issue related to MPI. Do you have access to a VASP version that uses a different MPI library?
Re: Relaxation for noncollinear magnetization with Hubbard U crashes
Posted: Wed Aug 16, 2023 9:06 pm
by vahid_askarpour
Dear Manuel,
The 48-core nodes on our Cedar cluster has a total of 192000M of memory. There is another version of VASP on our cluster compiled with intelmpi. I tried this version and it has not crashed yet using the same K-grid. So this might be an OpenMPI issue.
I will continue to use the impi version and if it does crash, I will report back.
Thank you,
Vahid
Re: Relaxation for noncollinear magnetization with Hubbard U crashes
Posted: Thu Aug 17, 2023 7:45 am
by manuel_engel1
I tend to agree that this is likely related to the MPI library. The available memory is more than sufficient, so that is not a factor.
So far, I have been unable to reproduce the error with the system you submitted. If you experience further crashes, please, report back and we will have to look at it in more detail.