GULP Parser does not thoroughly parse pGFN-FF force field
Closed this issue · 7 comments
Issue
This is a very niche error that I don't expect to be addressed for the JOSS review but wanted to report anyway.
In GULP >= 6.1, they introduced a new class of force fields called (p)GFN-FF. Calculations of this type do not appear to be parsed correctly in NOMAD just yet.
Input:
from nomad.client import parse
archive = parse('gulp.got')
print(archive[0].run[0].calculation[0].energy.total.value)
print(archive[0].run[0].method[0].force_field)Returns:
None
ForceField()The archive[0].metadata.domain also returns dft, which seems incorrect since this is a force field.
Test file, which you can keep or do as you wish: gulp.got.txt
Details of Build
System: Ubuntu via WSL
conda create --name joss_test python=3.9
conda activate joss_test
pip install -e .[parsing] # from https://github.com/arosen93/nomad/tree/developNote: I am using a fork of the repo because #44 is preventing me from running the parsers unless I hard-code where the tools.json file is located.
It seems that for this particular calculation, the pGFNFF force field was used and is not currently parsed. I will try to add a support for this. However, since I think this is a specific kind of force field, I would simply store the parameters as code-specific metadata, e.g. x_gulp_pffnff_accuary (@JFRudzinski do you think we can generalize this?). metadata.domain is something that we need deprecate, this was initially intended to differentiate between the dft and experimental data but I think it has lost its meaning. I will open an issue to address this in the gitlab project.
I don't know what the exact situation is for the GULP or other non-Lammps, non-Gromacs parsers. But generally I think it is safe to say that we do not really have support for force field metadata (or at least we cannot promise that the working parsers are also properly storing the force field).
In the future, the plan is to support full force field storage through collaboration with the OpenKim project. However, they are just now developing support for bonded force fields, so it will still be some time.
Another option in the future will be to store your FF via a custom MD upload to NOMAD. I am working on a schema and parser for this now, but again it will be some time before this is ready.
In the mean time, I agree with you @ladinesa that we should try to grab any easily accessible force field parameters and throw them into code-specific metadata, but I would not spend very much time on this, since IMO it will likely never be used in this way. Rather, I think the most important thing is to just make sure that the parser doesn't break when, e.g., new FFs are added as in this case.
In terms of tagging these as FF or MD calculation, this has really only been done for Gromacs, Lammps, and FHIaims parsers that I know of. @ladinesa we should start to compile a list of all the parsers that support MD simulations and then start to go through systematically and add support.
@JFRudzinski thanks for the input. can you perhaps open an issue where we can address your last point?
implemented parsing of pGFN-FF in nomad-coe/atomistic-parsers#54