Weird convergence issue involving band occupation problem

Question on input files/tags, interpreting output, etc.

Please check whether the answer to your question is given in the VASP online manual or has been discussed in this forum previously!

Moderators: Global Moderator, Moderator

Post Reply
Message
Author
meng-fu_chen
Newbie
Newbie
Posts: 4
Joined: Wed Sep 01, 2021 2:43 am

Weird convergence issue involving band occupation problem

#1 Post by meng-fu_chen » Fri Dec 24, 2021 1:52 pm

Hi everyone:

I am facing a problem when doing a simple optimization run of an ion pair in a large unit cell. I am using VASP 5.4.4 (intel parallel)

As can be seen in the log file in the attached folder, the optimization is normal at first but during the 10th ionic step, it suddenly fails to converge and there is an error about "Your highest band is occupied at some k-points." When I check the OUTCAR, the band occupancy is strange: the highest band number has strongly negative energy and is occupied.

Increasing NBANDS as suggested by the warning doesn't solve this problem.
Also, this optimization includes implicit solvation using VASPsol. When the calculation is done in a vacuum, it works fine.

A more detailed description of the problem and other relevant info is included in the pdf file in the folder attached.
The INCAR, POSCAR, POTCAR, KPOINTS are also in the folder attached.

Many thanks to anyone who can help with this problem.
You do not have the required permissions to view the files attached to this post.

fabien_tran1
Global Moderator
Global Moderator
Posts: 316
Joined: Mon Sep 13, 2021 11:02 am

Re: Weird convergence issue involving band occupation problem

#2 Post by fabien_tran1 » Fri Dec 24, 2021 5:09 pm

Hi,

First, a few questions regarding the geometry: Did you monitor it during the optimization? Is the problem occuring for any starting geometry? Do you think that the starting geometry is reasonable? Problems may occur in particular if two atoms are too close to each other.

meng-fu_chen
Newbie
Newbie
Posts: 4
Joined: Wed Sep 01, 2021 2:43 am

Re: Weird convergence issue involving band occupation problem

#3 Post by meng-fu_chen » Fri Dec 24, 2021 5:16 pm

Hi:

The starting geometry is from a previous optimization done in vacuum. I believe that the distance between the atoms is reasonable and shouldn't be the main source of the problem. I have actually tried several initial geometries but all suffer from the same problem.

fabien_tran1
Global Moderator
Global Moderator
Posts: 316
Joined: Mon Sep 13, 2021 11:02 am

Re: Weird convergence issue involving band occupation problem

#4 Post by fabien_tran1 » Mon Jan 03, 2022 9:41 pm

Hi,

That is not clear what is the source of the problem. As you described in your PDF file, the problems start at some geometry.

In fact, from OUTCAR we can see that during the SCF at this geometry, everything is fine until the 10th iteration, where
eigenvalues EBANDS = -514.64659208
Solvation Ediel_sol = -22.16881041
and nonsense values start at the 11th iteration, where
eigenvalues EBANDS = -4522.75798903
Solvation Ediel_sol = 675.67159974
(the other components of the energy are still reasonable at this 11th iteration).

Thus, we can guess that it is a failure of either the solvation model (VASPsol) or the electronic minimizer (ALGO). Besides ALGO=VeryFast and Normal that don't work, could you try other minimizers? At the moment, I can not test myself since I have no installation of VASPsol. I have only tested without solvation model, where there is no problem.

Post Reply