Page 1 of 1

system very near to the ground state

Posted: Wed May 01, 2019 5:11 pm
by mmg016
Dear Prof. Henkelman,

I have been simulating a system for a long time now. Each run of NSW=500 steps finished without reaching the accuracy set (EDIFFG = -0.01 and EDIFF = 1.0E-04) and I had to continue from the last run for more than 2 times.

As you can see from the attached vef.pl, the system seems to be very near to the ground state, but it is approaching the ground state with the required accuracy extremely slowly.

I have tried various optimizers, including both the native vasp optimizers as well as vtst optimizers. The current vaspout.eps plot is from a native vasp Quasi-Newton (IBRION=1) optimizer.

Could you please advise me a suitable optimizer (either from native vasp or a vtst optimizer) under this scenario which will help me reach the required accuracy very quickly.

Best regards,
mmg016

Re: system very near to the ground state

Posted: Wed May 01, 2019 6:51 pm
by graeme
I would have to see more about the calculation to try to understand why the energy is dropping slowly. For example, are there soft modes in the system such as rotations or translations along which the energy changes slowly? Also, seeing the INCAR and OUTCAR would help. If you want to make a tar.gz of the run, or just the input files, I can look into this further.

Re: system very near to the ground state

Posted: Thu May 02, 2019 7:47 am
by mmg016
Dear Prof. Henkelman,

Please find attached a tar.gz archive of the whole run.

One more run of NSW=500 is complete without reaching accuracy.

I would greatly appreciate your valuable insights on my system.

Best regards,
mmg016

Re: system very near to the ground state

Posted: Wed May 08, 2019 2:10 am
by graeme
Well, I see what you mean. I can reproduce the problem but I'm not sure what's going on.

I'll keep poking at it.

Re: system very near to the ground state

Posted: Thu May 09, 2019 12:19 pm
by mmg016
Thank you Prof. Henkelman, for your time and efforts.