SCF convergence problem in vacancy defect modelling

Total energy, geometry optimization, DFT+U, spin....

Moderator: bguster

Locked
mmaeda
Posts: 3
Joined: Wed Dec 07, 2016 4:49 am

SCF convergence problem in vacancy defect modelling

Post by mmaeda » Thu Dec 08, 2016 12:26 pm

Dear Abinit users and developers,

I am modelling a spinel metal oxide system with O vacancy as a cubic supercell (55 atoms) by PAW and GGA+U; but struggling with SCF convergence. Any suggestion to fix the problem?

Some input parameters:
ecut 500 eV
pawecutdg 1600 eV
ixc 11
occopt 7
diemac 12.0
ionmov 2
optcell 1
dilatmx 1.05
ecutsm 0.5
tsmear 0.01
strprecon 0.1

Thanks,
Maki

ebousquet
Posts: 469
Joined: Tue Apr 19, 2011 11:13 am
Location: University of Liege, Belgium

Re: SCF convergence problem in vacancy defect modelling

Post by ebousquet » Mon Dec 12, 2016 11:46 am

Dear Maki,
How bad is your convergence? Does it succeed to converge up to a given precision and then it is stuck at this precision or is it completely bad from the beginning or does it converge and then starts to diverge, do you have an error/wraning message, etc? An output file will be welcome or at least the "grep ETOT" of it to reply to you.
I would already advise you to first relax the internal coordinates by keeping the cell parameters unrelaxed and then relax them once the internal atomic coordinates will be relaxed, that might help the algorithm in finding the minimum.
Best wishes,
Eric

mmaeda
Posts: 3
Joined: Wed Dec 07, 2016 4:49 am

Re: SCF convergence problem in vacancy defect modelling

Post by mmaeda » Mon Dec 26, 2016 9:05 am

Hi Eric,

Thanks for your advice. I have tried to converge the system with ion position optimization (without cell parameter optimization). But the problem of SCF convergence has not been solved. The fisrt SCF cycle slowly converged, but remained the same order of magnitude in absolute energy change in the system and did not reach the convergence criteria. I tried to use several different iscf parameters to see if SCF converges. But none of them solved the problem, but in some cases SCF actually diverged instead. Any suggestion?

Thanks,
Maki

ilukacevic
Posts: 271
Joined: Sat Jan 16, 2010 12:05 pm
Location: Dept. of Physics, University J. J. Strossmayer, Osijek, Croatia
Contact:

Re: SCF convergence problem in vacancy defect modelling

Post by ilukacevic » Tue Dec 27, 2016 9:56 am

Dear Maki,

Did you try increasing some of these variables: npulayit, nnsclo or nline in combination with iscf 7 or 17?

Best regards,

Igor

ebousquet
Posts: 469
Joined: Tue Apr 19, 2011 11:13 am
Location: University of Liege, Belgium

Re: SCF convergence problem in vacancy defect modelling

Post by ebousquet » Fri Jan 13, 2017 11:45 am

Dear Maki,
Did you try to simply reduce the mixing parameter diemix?
All the best,
Eric

Locked