Page 1 of 1
Convergence problem
Posted: Mon Dec 11, 2023 1:22 am
by alok_shukla1
Dear Experts,
I am trying to relax one 2D heterostructure with GGA+U+SOC in AFM and FM configurations. After doing the CONTCAR to POSCAR conversion, I rerun my calculations 12 times (wall times of each run = 120hrs.), but still not able to achieve convergence. Here I am attaching my initial POSCAR and final CONTCAR after the 12th run. Please have a look. Please let me know if some changes in the INCAR can solve my convergence problem issue.
Re: Convergence problem
Posted: Mon Dec 11, 2023 8:56 am
by jonathan_lahnsteiner2
Dear alok_shukla1,
I was checking the files that you sent. There seems to be no problem
with your electronic minimization procedure. What is suspicious nevertheless,
your OSZICAR and your OUTCAR files look like they were terminated within
one minimization step. Could it be that you reached some time limit
on your cluster which terminated your job before finishing?
You can set the time limit in your slurm job script with
For more information on the time keyword for slurm you can check
https://slurm.schedmd.com/sbatch.html
I hope this is of help
All the Jonathan
Re: Convergence problem
Posted: Mon Dec 11, 2023 9:41 am
by alok_shukla1
Dear jonathan_lahnsteiner2,
Yes, the job is terminated because of the cluster time limit. I rerun the calculation by doing CONTCAR to POSCAR, and this I did around 12 times, but still, I am not able to get the convergence. The wall time of my cluster is 120 hrs, after that, I need to rerun my calculations. The thing that is bothering me is that the system is not converging after so many reruns, and I am still in the rerun process.
Re: Convergence problem
Posted: Mon Dec 11, 2023 9:56 am
by jonathan_lahnsteiner2
Dear alok_shukla1,
You have the tag ICHARG=2 in your INCAR file which initializes your charges by:
Code: Select all
Take superposition of atomic charge densities.
and you set ISTART = 0 which in the wiki is described as
Code: Select all
Start job: "begin from scratch". Initialize the orbitals according to the flag INIWAV
.
And the default value for INIWAV = 1. According to the vasp wiki INIWAV = 1 has the following functionality
Code: Select all
Fill the Kohn-Sham–orbital arrays with random numbers. It is definitely the safest fool-proof switch..
So in principle, you always start your electronic minimization from scratch. You should
check the vasp wiki for the tags:
ICHARG
https://www.vasp.at/wiki/index.php/ICHARG
ISTART
https://www.vasp.at/wiki/index.php/ISTART
You should read your charge from the CHGCAR file or construct it from the WAVECAR file, if you want to continue an electronic minimization. And as mentioned
already maybe it is also advisable to set a higher time limit for your job. Then you don't have to bother with continuation runs.
All the best Jonathan
Re: Convergence problem
Posted: Mon Dec 11, 2023 10:26 am
by alok_shukla1
Dear jonathan_lahnsteiner2,
I'll add all your suggestions to my INCAR. I'll update here if everything goes fine.