Reference units different between hoomd_simulation.py, gsdwriter.py and hoomdxml.py

[chrisjonesBSU]

This is very similar to the changes made in #463

From what I understand, the idea in #463, is that for the sake of consistency, the units used by the reference value parameters in write_hoomdxml should be same units that are used in mbuild and foyer (nm, kJ/mol, amu).

Since parmed is actually being used behind the scenes, the reference values are converted from nm and kJ/mol into angstroms and kcal/mol. #463 added units to the reference value parameter doc strings and took care of the conversions required.

However, this is not how the reference values are being handled in create_hoomd_simulation and write_gsd (I'm not familiar with the lammps or par writers). These functions still require the user to enter their reference values in parmed units (kcal/mol, angstroms, Da) so that they are consistent with the parametrized parmed structure. There isn't any conversion being done, the values are used as entered by the user.

I agree with the method we're using in write_hoomdxml. To me, its a lot more intuitive to enter reference value as they appear in the forcefield files, and in the case of foyer that's nm and kJ/mol, so we should probably update the other relevant writers to handle to use the same process.

I can write up the PR for it, I just wanted to start some discussion since messing around with units can result in breaking changes to any existing workflows. Note that if auto_scale = True for any of these 3 writers, then this isn't an issue. Also, I just realized that write_gsd doesn't have an auto scale option, maybe that could be added with this PR as well.

This came up when talking with Mike earlier about the changes made in #798
#798

Also, if these changes are made, I assume they should also be made in #698 @mikemhenry

[jennyfothergill]

@chrisjonesBSU was this issue fixed by #798 or #698?

[chrisjonesBSU]

@chrisjonesBSU was this issue fixed by #798 or #698?

No it doesn't look like it was.

[slenka93]

Hi,

I wanted to clarify the units of mbuild gsdwriter.

I am using mbuild to build a coarse grain molecule and then mbuild fill box to pack molecules in a box. The position of atoms are defined such as the distance is in nm and my box length is in units of nm. After preparation of a box of the coarse grain molecule, i am using foyer and openmm to define the atom, bond, angle and dihedral types and charge of the coarse grain molecule. Distance is in nm, angles is in radian and energies are in kj/mol (I already have the values of equilibrium bond, angle...)

After this I want to save a gsd file to use it for hoomd runs. For this I am using mbuilds gsdwriter.
from mbuild.formats.gsdwriter import write_gsd
write_gsd(struc, 'box.gsd', ref_distance=1, ref_energy=1)

Using this set of commands changes the distance in angstroms and energy units I am not sure even though all the parameters in my files are nm and kj/mol.

[daico007]

Hi @slenka93, I believe this is because when we perform the atomtyping using foyer, we actually have to convert our object to a parmed Structure to be able to store information related to potential parameters, which have a different set of default units than ours. The default unit of parmed is Angstrom and kcal/mol, so if you observed the unit discrepancy at the final gsd file, can adjust the ref_distance and ref_energy to revert back to the your original unit. We are working on a new backend that would have a better representation of units called GMSO, which will be used to replaced parmed in the near future.

[slenka93]

Hi @daico007, thanks for the clarification.
Does the charge also converts to AKMA units (angstrom, kcal/mol) or is it the same as (nm, kj/mol)? Also is there any account of 4pipermittivity in the charges like gmso?

[daico007]

Hi @slenka93, the charge will also be in electron charge (which i think is also the case in the in the original XML). The gsdwriter.py does account for the 4 * pi * permittivity (see https://github.com/mosdef-hub/mbuild/blob/main/mbuild/formats/gsdwriter.py#L135-L142)

[slenka93]

Thanks @daico007!

[CalCraven]

GSD writer has been transitioned to GMSO in PR 680.