Charge units for create_hoomd_forcefield about mbuild HOT 5 CLOSED

Hi @scmartin, I believe this relate to the unit system being written to the hoomd snapshot. At the moment, the create_hoomd_snaphot() and create_hood_forcefield() is working on a parmed Structure object, which utilize the unit of kCal and Angstrom, and hence permittivity e0 is in e^2 / ((kcal/mol) * Angstrom) with the value of 2.396452e-04 (please refer to this code excerpt https://github.com/mosdef-hub/mbuild/blob/main/mbuild/formats/hoomd_snapshot.py#L419-L426).

We are working on developing a different backend, called GMSO, which will allow more control over the unit aspect during conversion. The package is still under development and there is no direct path to go from gmso to hoomd forcefield (there is currently a WIP PR).

from mbuild.

So does this mean that the units that create_hoomd_forcefield() outputs are in the units used by the parmed object by default? I just saw that the option ref_mass, ref_energy, and ref_distance arguments. In the current version, I take it I need to pass these arguments the proper conversion factors if I want to convert back to the units used by Foyer. Am I understanding this correctly?

This unit conversion seems extremely opaque in the documentation. It wasn't until I saw the ref_[dimension] optional arguments that I realized the output units were changed from the units used by Foyer.

from mbuild.

Hello @scmartin. You're right, the unit conversions that are happening are not made clear. After applying a forcefield using Foyer, the object returned is now a Parmed structure instead of an mBuild compound, and the units of distance and energy are converted from nm and kJ/mol to Angstrom and kcal/mol. So, the conversion is technically happening before calling create_hoomd_forcefield().

You should be able to the use ref_ parameters to convert back to nm, and kJ/mol, and as @daico007 pointed out, I believe the correct conversion of e0 is being performed on line 420 of hoomd_snapshot.py.

I'll start a PR that adds a better explanation to the documentation. Also, I noticed some inconsistencies of the doc strings for the reference value parameters in create_hoomd_forcefield and to_hoomdsnapshot in hoomd_snapshot.py.

from mbuild.

@daico007 and @chrisjonesBSU Thank you both for clarifying this behavior. I'm still trying to get a feel for the interactions of Foyer and mBuild (for example, calling the apply method on an mBuild compound changing the underlying mBuild object) but you have completely answered my original question, so I'm going to close this issue.

from mbuild.

You're welcome @scmartin, the unit issue is something we have been trying to address with the new backend data structure (GMSO), and hopefully when that library is ready for production, it will eliminate such confusion with unit that you're experiencing

from mbuild.