Page 2 of 2

Re: SCF convergence for a big system

Posted: Mon Nov 16, 2020 8:16 am
by henrique_miranda
look into the ISMEAR options (1 and -1 should be good choices for metals)
I did not find in your attached files a calculation done with ISMEAR=-1 or ISMEAR=1.
Did you try it?

Re: SCF convergence for a big system

Posted: Fri Nov 20, 2020 12:44 pm
by SKM
Thanks for the reply.
attaching here the test runs with ISMEAR=1 and -1.
at ISMEAR= -1 it got converged at coarse convergence criteria.
at ISMEAR= 1, not converged within same wall time as that given for ISMEAR = -1, as wall time consumed.
attaching the tar files of two runs done in continued runs. Please review the same.
Kindly what next and how to proceed from here. My ultimate aim is to get HSE06 band structures. and few more bigger supercells of same system is to be done if this is succesfull.

it seems there file size limit to upload. i am attaching in two separate messages, please

regards

Re: SCF convergence for a big system

Posted: Fri Nov 20, 2020 12:46 pm
by SKM
attaching the continued set of files here, please

regards

Re: SCF convergence for a big system

Posted: Mon Nov 23, 2020 6:54 pm
by henrique_miranda
Kindly what next and how to proceed from here. My ultimate aim is to get HSE06 band structures. and few more bigger supercells of same system is to be done if this is succesfull.
So, when using ISMEAR=-1 you get convergence in the electronic loops.
Now you can try to obtain a relaxed structure by adding the tags related to ionic relaxation (IBRION,ISIF,NSW).
Once you obtain a relaxed cell, and I think now it will work, you can use the new POSCAR to compute the electronic band-structure.

Re: SCF convergence for a big system

Posted: Tue Nov 24, 2020 5:05 am
by SKM
Hi Admin
Thanks for the reply. !!

I still need to do ENCUT, KPOINTS convergence test before adding additional tags for full ionic convergence?

I Am waiting for the reply for the post on convergence test issue of same system at forum/viewtopic.php?f=3&t=3005#p19679
after the reply i will proceed as per your suggestion. Thank you again.

Regards

Re: SCF convergence for a big system

Posted: Sun Nov 29, 2020 6:59 am
by SKM
Hi
Further to my reply below, i did the following tags changed. Rest all tags, same as before.

1. set PERC=Accurate (earlier it was 'Normal')
2. set EDIFF = 1E-07, and EDIFFG = 1E-06 (earlier they were 1E-05 and 1E-04 respectively)
3. started convergence test fro ENCUT. So changed ENCUT from400 to 615 eV, in steps and beyond 616 eV, it gives REAL_OPTLAY:internal error
now kindly clarify the queries below:

1. Why after a very specific value this REAL_OPTLAY:internal error starting (i tested with incremental values of 1eV, i.e. 616, 617 etc. it worked upto 616 but not after that.
2. Even if the run is successful upto 616eV, the ENCUT Vs Energy graph for convergence test is not as expected (look wired). (attached the picture). kindly suggest what went wrong.
3. I still need to do KPOINTS convergence after this ENCUT convergence. and so far i did not add the other tags such as IBRION, ISIF, NSW. If i add them after the above convergence test, Do i need to do both ENCUT and KPOINTS convergence again?

Kind regards

Re: SCF convergence for a big system

Posted: Sun Feb 07, 2021 9:59 am
by SKM
hi
in absence of any other clue to such ENCUT convergence failure, I stick to ENCUT 500 and proceeded. and the SCF convergence done. Later i did PBE band structure, which again lead to confusion. I will start another thread as its for Band structure, for this same system