mBuild+Foyer+Hoomd+Freud+multi-state LJ

[ahy3nz]

• Recreating LJ RDFs from Moore et al. (2014).
• mbuild + foyer to pack boxes of LJ argon and parametrize according to ffxml in this PR
• Hoomd simulation from parmed structure
• Freud to analyze hoomd simulation

[jennyfothergill]

Hi, I hope this is the right kind of feedback:

I was able to run this notebook with the following environment: conda create -n mosdef_hoomd -y -c conda-forge -c mosdef -c omnia 'python=3.7' mbuild foyer unyt hoomd matplotlib freud gsd openbabel jupyterlab I still needed omnia for fftw3f and openmm.

Everything in this notebook ran for me and was fairly straightforward to understand. The comparison to a published figure in a very lightweight notebook is really clean!

The next bits maybe aren't related to this notebook in particular, so please let me know if these are in the wrong place:

I really like how the units are more exposed--that's always been a stumbling block for me when loading files into mbuild. Sometimes lengths were scaled without me knowing, so I like the change to use unyt. Will unyt also be used in the write_gsd and create_hoomd_simulation functions? Right now there is ref_distance and ref_energy, but maybe using units with these would be more clear?

I know for some of our larger simulations applying the force field to a filled box (structure = ff.apply(filled_box)) might be slower than applying the forcefield to one compound and then filling the box. Having the option to either do type then fill or fill then type might be nice.

[jennyfothergill]

Actually, as a suggestion--it'd be really cool if there were a binder link or an environment yaml for these workflows. It'd make it so much easier for folks to try them.