LATTICE_CONSTRAINTS & IBRION = 3

[reza_namakian1]

Dear VASP Team,

I already asked this question to Manuel in this post: https://www.vasp.at/forum/viewtopic.php?t=20031, but the thread was locked before I got a chance to ask these follow-ups, so I’m asking them here.

1) Occasionally, I’ve found that IBRION = 3 is notably faster than IBRION = 2 for the material systems I’m interested in. One major drawback, though, is that IBRION = 3 can’t be used with NEB calculations, which makes it a bit painfully slow for 2D materials when using IBRION = 2 with LATTICE_CONSTRAINTS tag.

2) You guys have no plans to add any new minimizer, like implementing the latest version of the FIRE minimizer as in "min_style fire" in LAMMPS (https://docs.lammps.org/min_style.html), right?

Thanks!

[manuel_engel1]

Hi again,

It's perfectly fine to start a new topic here. We strive to keep the forum organized, which sometimes means locking threads that have already received answers.

I understand that using IBRION=3 would be favorable in your context since you already raised interest in this feature in an earlier post. While we recognize its potential benefits and will keep it in mind for future development, we currently don't have any concrete plans or timelines for its implementation.

Regarding new minimizers, I'll consult with my colleagues who work on that part of the code and let you know if there's any additional information.

Thanks for your understanding and for contributing to the discussion.

Best regards

[manuel_engel1]

Here is a quick follow up, as promised, regarding the development of new minimizers. This is certainly an interesting proposal and one that can come with potential benefits. However, we do not currently have any plans of implementing new minimizers for ionic updates directly in VASP.

Some related information: it is possible to run VASP together with LAMMPS via a new interface if you already have a pre-trained machine-learned force field (MLFF). This approach is however not available for standard DFT.

[reza_namakian1]

Thanks, Manuel, for the follow-up!

I'm glad you guys took an interest in the new minimizers, and I hope you have the time and patience to add them. Just one last thing to mention—ABC-FIRE looks the most promising for minimization based on my experience: https://docs.lammps.org/min_modify.html, https://www.sciencedirect.com/science/a ... 5622006899.

Thanks also for the hint on using MLFF—that's a relief to have MLFF available in LAMMPS.

Best,

Reza.