[DISCUSSION] Validating INCAR parameters for external VASP calculations

[matthewkuner]

Hi all,

I'm currently working on validating the INCAR parameters used in VASP calculations, with the eventual goal of building an automated pipeline for external data ingestion into the MP DB (crowd-sourced MP). I was hoping to get your opinions on if/how we should restrict parameters.

This will be a long post.

Currently, the focus is only on ingesting static and relaxation calculations using GGA / GGA+U / R2SCAN.

Below is a list of INCAR parameters that I believe are in any way relevant. I may have left some out, so please mention any that you think could lead to erroneous results being submitted--we want to account for every possible way to mess up VASP calculations.

The current position for how to restrict a given parameter is written below each entry.

@tschaume @janosh @munrojm @yang-ruoxi (and anyone else who is knowledgeable), any help / input on how these parameters should be restricted would be greatly appreciated!

Please see this Google Doc so we can more easily discuss this: https://docs.google.com/document/d/1Xunv1TiqHsa0s0KNuC-GA6vh71PGEukPyupe4W2IMb0/edit?usp=sharing
If you don't have a UC Berkeley email, you'll have to request access.

[janosh]

Wow, impressive list! 😄

I know a bunch of people who might want to weigh in here. Pinging @JaGeo @arosen93 @utf @mkhorton @rkingsbury. If you have time take a look if your favorite INCAR param is already in this list and constrained as you'd expect.

[Andrew-S-Rosen]

This probably deserves a meeting :) Happy to chat though!

Also, yes, if it was not previously clear I have an unhealthy obsession with knowing the ins and outs of the VASP INCAR flags and their subtleties. Opinions abound.

[JaGeo]

I think this should be a meeting or we work on a shared md file where we all add.our opinions. Otherwise, Impossible.

[matthewkuner]

@JaGeo I agree. We could do both, perhaps? I.e. look into these params on our own time so we are more prepared to discuss this during a meeting.

[janosh]

@matthewkuner suggested a Google Doc which would prob make collaboration easiest (even during a meeting). Also great at keeping edit history. @matthewkuner If you already have one, you could link it here and give everyone write access.

I suggested to start an issue first so we have sth less ephemeral to refer to later.

[matthewkuner]

@janosh I don't have one currently, but will make one and share it here before the end of the day.

[matthewkuner]

Hello @tschaume @janosh @munrojm @yang-ruoxi @JaGeo @arosen93 @utf @mkhorton @rkingsbury. Please see this Google Doc so we can more easily discuss this: https://docs.google.com/document/d/1Xunv1TiqHsa0s0KNuC-GA6vh71PGEukPyupe4W2IMb0/edit?usp=sharing
If you don't have a UC Berkeley email, you'll have to request access.

[Andrew-S-Rosen]

@matthewkuner: Great! Let's do a Google Doc and a meeting, which we can transition to a Markdown file later for longevity's sake.

[janosh]

Btw, Google Docs has a markdown mode so we can just copy-paste later.

[Andrew-S-Rosen]

Joke comment: What about the undocumented LMUSIC INCAR flag, the VASP flag that somehow (?) plays a little tune when you run the calculation? There are some goodies in the source code... 😉

[janosh]

I demand we unequivocally refuse unmelodic calcs!

[JaGeo]

It would be good to set a deadline and a meeting right away.

[janosh]

I just sent out a Zoom invite for next Thu Jul 13 @ 11 - 12 PT, 20 - 21 German time. Let me know if y'all can make that. Happy to shift around as necessary.

[JaGeo]

Thanks for doing so and also for adapting the timing. I usually start working at 6:30 am and an 8-9 pm meeting is very late for me (and out of our working times).