NSW=1 --> NSW=0 in static, Cf in POTCARs

[Andrew-S-Rosen]

Quick Hello

Hi @utf! I'm opening this very minor PR in part to say hello and that I think atomate2 looks fantastic! I am really excited to start using it. Also a few small things.

Summary of Proposed Changes

• I added Cf: Cf to the list of default pseudopotentials since it's available with the VASP PAW_PBE_54 potentials. It probably won't be relevant to anyone, but why not aim for completeness. You got all the other ones!
• I changed NSW=1 to NSW=0 in your get_incar_updates() for StaticSetGenerator(). Was there a specific reason to use NSW=1 for the static calculation? One of my concerns is that other codes might look at NSW and think that it's doing a relaxation even though it's a static calculation.
  • Take the DriftErrorHandler() in Custodian, for example. It checks to see if EDIFFG >= 0 or NSW == 0. If this condition is met, it doesn't run the drift error checker; otherwise, it does. However, here, NSW=1 would make Custodian think it's a relaxation. Granted, Custodian is unlikely to actually do anything here, but it's more the premise that this might influence other codes in unexpected ways.

[utf]

Hi @arosen93, great to have someone looking over atomate2. It is obviously still forming so any and all suggestions/PR are very welcome.

Thanks for catching the Cf pseudo potential. As for NSW = 1 this was a on purpose, as certain calculations require it (e.g., DFPT LEPSILON needs NSW = 1 and won't work with NSW = 0). That said, this does cause an incompatibility with custodian and that seems more difficult to fix than just setting NSW = 1 for certain calculation types. (I think that is still a bug in custodian and should probably be fixed).

Before I merge, can you add NSW = 1 to the custom lepsilon settings in StaticSetGenerator?

[Andrew-S-Rosen]

@utf, interesting! I didn't know that about DFPT. I see the thought process there, although my personal opinion is that it's better to set NSW=1 on an as-needed basis. Just changed the lepsilon settings.

[Andrew-S-Rosen]

I figured that might break something along the way. Sorry for not running tests locally. I'll get the testing stuff set up tomorrow and can fix then. :)

[utf]

Ah, the test data is now inconsistent. Should be as simple as editing the INCAR in the Si_bandstructure/static/inputs/INCAR file.

Also, to be consistent, we should probably also adjust NSW in the HSEStaticSetGenerator.

If I have a chance, I can take a look at this later.

[Andrew-S-Rosen]

Hi again @utf. That should do it now. Tests passed locally.

I had to change the INCARs in the following:

Si_hse_band_structure/hse_static
Si_band_structure/static

[utf]

For the lint error. The easiest thing to do is to run pre-commit install when in the root atomate2 directory. Then, anytime you make a commit, any staged files will get linted automatically.

[Andrew-S-Rosen]

@utf: I guess I must have run the wrong pytest locally before. Clearly, there are some errors!

I wanted to run a few things by you here. Which INCARs should have NSW = 1? Is it just the Si_dielectric job? There are a bunch of test files with NSW = 1 in the INCARs right now, which is causing the pytest errors, e.g. the Si_elastic, Si_hse_optics, and Si_optics jobs. Is it okay if I change NSW to 0 in these INCARs? I'm looking back at Atomate1, and it seems like they have NSW = 0, but I've never run elastic or optics jobs before, so I didn't want to proceed without asking about this.