Dear Professor,
I am working on a Mg, Mn, Ti and S system. I am having convergence issues. Could you please help me solve this? I have attached relevant files.
Thanks
CI-NEB not converging
Moderator: moderators
-
- Posts: 2
- Joined: Thu Sep 26, 2019 1:29 am
CI-NEB not converging
- Attachments
-
- CINEB.rar
- (20.58 MiB) Downloaded 1773 times
Re: CI-NEB not converging
Start by further minimizing your initial and final state structures - they do not appear stable to me.
-
- Posts: 2
- Joined: Thu Sep 26, 2019 1:29 am
Re: CI-NEB not converging
Dear Professor,
I rerelaxed the Inital and Final images with EDIFF = 1E-5 and EDIFFG = 1E-6 and then ran the NEB calculations but I am still having convergence issues. I have attached relevant files. Could you please advice me on how to proceed further?
[attachment=0]Step4_POTIM_Relax2.rar[/attachment]
I rerelaxed the Inital and Final images with EDIFF = 1E-5 and EDIFFG = 1E-6 and then ran the NEB calculations but I am still having convergence issues. I have attached relevant files. Could you please advice me on how to proceed further?
[attachment=0]Step4_POTIM_Relax2.rar[/attachment]
- Attachments
-
- Step4_POTIM_Relax2.rar
- (20.78 MiB) Downloaded 825 times
Re: CI-NEB not converging
Your endpoints should be relaxed with the same settings as your band. You endpoints appear to be spin polarized with +U and your band has neither.
Re: CI-NEB not converging
This may help. In the attached, I reconverged your endpoints (min1a.tar.gz and min2a.tar.gz) and then did an NEB calculation between them. I used more reasonable settings (regular potentials, ENCUT~300eV, KPOINTS=2x2x2, prec=Normal, LREAL=Auto) and was able to complete these calculations in a couple of hours.
I have said this many many times on this forum, but I will repeat here that it is much better to use a minimal set of parameters that will qualitatively reproduce the geometry. It is then trivial to reconverge your calcualtions with more accurate settings. In your case, you used +U and spin polarization for the endpoints with (in my view) far too precise settings of LREAL=False, prec=Accurate, _pv/_sv potentials, KPOINTS=5x5x5, ENCUT=500, but then turned off +U and spin polarization for your NEB with only 3 images. No. The way to go is to use modest settings to find reaction mechanisms and geometries and then increase parameters to efficiently determine the precision of your calculations.
I have said this many many times on this forum, but I will repeat here that it is much better to use a minimal set of parameters that will qualitatively reproduce the geometry. It is then trivial to reconverge your calcualtions with more accurate settings. In your case, you used +U and spin polarization for the endpoints with (in my view) far too precise settings of LREAL=False, prec=Accurate, _pv/_sv potentials, KPOINTS=5x5x5, ENCUT=500, but then turned off +U and spin polarization for your NEB with only 3 images. No. The way to go is to use modest settings to find reaction mechanisms and geometries and then increase parameters to efficiently determine the precision of your calculations.
- Attachments
-
- neba.tar.gz
- (9.51 MiB) Downloaded 1779 times
-
- min2a.tar.gz
- (1.12 MiB) Downloaded 823 times
-
- min1a.tar.gz
- (902.83 KiB) Downloaded 808 times