Does the ionic convergence depend on NSW?

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

Locked
Message
Author
naoum_bacalis
Newbie
Newbie
Posts: 2
Joined: Mon Nov 16, 2020 6:48 pm

Does the ionic convergence depend on NSW?

#1 Post by naoum_bacalis » Sat Dec 12, 2020 7:29 pm

Dear Forum;

In a relaxation run, for a clean W(001) surface under an external
electrostatic field, the program stopped after 20 ionic steps giving:

"ZBRENT: fatal error in bracketing
please rerun with smaller EDIFF, or copy CONTCAR to POSCAR and continue"

After such an error the program doesn't update the wavefunctions and
charge density. Therefore, I re-submitted the job with NSW= 19 (using
the *same* input files) in order to generate new WAVECAR and CHGCAR and
then copy CONTCAR to POSCAR and continue.

However, the calculation converged (!) after 18 ionic steps. How it can
be?

As you'll see in the attached plot the total energies, for the two runs,
are the same in the first five ionic steps and then differ.

Thank you, looking forward to your answer at your earliest convenience.

Best Regards,

Naoum Bacalis
You do not have the required permissions to view the files attached to this post.

martin.schlipf
Global Moderator
Global Moderator
Posts: 458
Joined: Fri Nov 08, 2019 7:18 am

Re: Does the ionic convergence depend on NSW?

#2 Post by martin.schlipf » Mon Dec 14, 2020 8:43 am

If you run VASP twice from the same input files, you should get the same result. There are some really minor caveats regarding parallelization, which may introduce some noise leading to divergence for very long MD simulations. Otherwise, we use a deterministic sequence of "random numbers" so that the results are the same.

If you observe that a calculation changes the result based on rerunning the same calculation changing just NSW, most likely you overlooked something which lead to different restarting behavior. Carefully compare the initial part of the two respective OUTCAR files and typically you will spot something where the code prints something different. If you can really reproduce the behavior that two calculations from scratch return different results based on NSW, I would like to see the whole input and know the version of Vasp you are using.

Finally, you do not need a WAVECAR or CHGCAR to restart. You could have copied CONTCAR to POSCAR and started again, which would simply create a new "initial" charge density.

Locked