I am performing relaxing a 2D hereostructure. I found that the force doesn't converge as the criteria demands when VASP claims the converge by `reached required accuracy - stopping structural energy minimisation` in the OUTCAR. To be specific, I set `EDIFFG = -1E-2` in the INCAR, but the TOTAL-FORCE certainly doesn't meet the demands (see the attached).
I have tried loose the `EDIFF` tag, but that would incur the error stating:
`ZBRENT: fatal error in bracketing
please rerun with smaller EDIFF, or copy CONTCAR
to POSCAR and continue`.
I have attached the files below which are runned under vasp-640 in my case.
Any help is warmly welcomed!
Could you show the POSCAR and the OUTCAR ?
Also, are you sure the code is stopping because of EDIFFG and not because of NSW (i.e. exceeding the max amount of ionic loops, 200 in your case) ?
To get around the "ZBRENT: fatal error", switch from IBRION =2 (conjugate gradient algorithm) to IBRION =1 (RMM-DIIS).
Rémi Pasquier Yes, the last iteration is `Iteration 14( 3)`, which is less than the NSW.
And the convergence information from OUTCAR is
```
reached required accuracy - stopping structural energy minimisation
```
I guess vasp stops iteration because it dicides that energy convergence is met. And I am wondering if the convergence is met either energy criterion is met or the force criterion is met. If this is true, what is the default energy criterion?
I've taken a look at your POSCAR and I've noticed you are fixing atoms.
That's the reason, the EDIFFG condition does not apply for any fixed direction. In your POSCAR, atoms are only allowed to relax along the z direction where the condition is fulfilled as you can see on your screenshot.
Gabriel Vinicius Thank you sir, I have tried your suggestion and it just saved my day!
- Badges
- Science topic