My Community

Dear VASP Team,

In my previous post, I was trying to use KERNEL_TRUNCATION:

https://vasp.at/forum/viewtopic.php?p=30167

However, there was a bug, which has apparently been fixed in VASP 6.5.1. My question is: are there still any considerations regarding vacuum size when using this tag? For instance, is there still a minimum vacuum requirement that must be met regardless of using this tag? Specifically, should I still ensure at least 15 Å of vacuum when using DFT-D4?

Thanks!

You still need to converge with respect to vacuum size but the convergence should be much faster. For comparison, previously, you needed to include enough vacuum that the interaction between the periodic replicas decays; with the truncation method, you need to include enough vacuum that the exponential tail of the orbitals decays. That should be much smaller.

In your particular case, however, you need to be aware that DFT-D4 does not consider the truncation. Hence, you still need to include sufficient vacuum such that the vdW interaction decays to 0 between the periodic replica.

Dear Martin,

My apologies for getting back to you a bit late — our clusters were down until yesterday.

I have been testing KERNEL_TRUNCATION, and it seems to work fine now without causing issues related to vacuum size for a 2D material, unlike what I experienced previously. However, I ran into a different problem:

1) When I use ISIF = 2, everything runs smoothly, and the system relaxes properly toward minimum energy using IBRION=3.

2) But when I switch to ISIF = 3 with LATTICE_CONSTRAINTS = .TRUE. .TRUE. .FALSE. and IBRION=2, the energy skyrockets uncontrollably during the second step of the ion/volume/shape relaxation. I tested this with different materials and observed the same behavior. But as soon as I turn off KERNEL_TRUNCATION, the relaxation proceeds smoothly again.

Do you think this could be a bug, or am I perhaps misusing KERNEL_TRUNCATION in combination with LATTICE_CONSTRAINTS and ISIF = 3 and IBRION=2?

I would really appreciate your thoughts or suggestions on this.

Best regards,

Reza.

Could you post an example that demonstrates this, please? Both KERNEL_TRUNCATION and LATTICE_CONSTRAINTS are relatively new features, so it is possible that they are not compatible yet.

Hi Martin,

I ran two examples (with and without kernel truncation) in the attached file: test_kernel_truncation.zip

When using kernel truncation, the entire system blew up, and I got bunch of bug warnings in the error log file.

I haven’t tested other cases yet, but I assume kernel truncation might also cause issues with different ISIF values, MD runs (IBRION = 0).

Let me know if you need any additional information from my side.

Best regards,

Reza.

Thank you for this report, I could reproduce the issue and will now inquire what is happening.

PS: It turns out the stress tensor is not implemented for KERNEL_TRUNCATION so at this stage the feature is incompatible. I have added this to the known issues