[WIP] OverdampedLangevin simulator and PotentialFunction interaction

[axsk]

I would really like to have these functions (And eventually a GirsanovOverdampedLangevin simulator).

This is still work in progress.

[codecov]

Codecov Report

Attention: 2 lines in your changes are missing coverage. Please review.

Comparison is base (715c075) 72.37% compared to head (b8b0796) 72.38%.
Report is 12 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #150      +/-   ##
==========================================
+ Coverage   72.37%   72.38%   +0.01%     
==========================================
  Files          35       35              
  Lines        5212     5232      +20     
==========================================
+ Hits         3772     3787      +15     
- Misses       1440     1445       +5     

Files	Coverage Δ
src/Molly.jl	100.00% <ø> (ø)
src/energy.jl	64.84% <ø> (ø)
src/force.jl	65.30% <ø> (ø)
src/types.jl	72.37% <100.00%> (-0.08%)	⬇️
src/simulators.jl	96.10% <90.47%> (-0.33%)	⬇️

... and 2 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jgreener64]

Looks good, I'm not sure the potential function stuff is worth putting in the main Molly codebase though.

Would be good to add OverdampedLangevin as a simulator in the "Langevin splitting" test set, or maybe there is a better way to test some property resulting from an overdamped Langevin simulation.

There was a discussion about how different atomic masses affect the time step, is this implementation consistent with the conclusions of that discussion?

[jgreener64]

Is there a canonical paper to reference here?

[axsk]

The Euler Maruyama scheme is pretty standard, even Wikipedia forgoes the reference..
Considering the limit underdamped Langevin $\rightarrow$ overdamped Langevin and the resulting unit considerations I would point to the "Free energy computations" book by Lelievre, Rousset and Stoltz, but its merely hinted at there..

[axsk]

Looks good, I'm not sure the potential function stuff is worth putting in the main Molly codebase though.

I removed that part

Would be good to add OverdampedLangevin as a simulator in the "Langevin splitting" test set, or maybe there is a better way to test some property resulting from an overdamped Langevin simulation.

It is actually simulating another dynamics then (underdamped) Langevin. The two would conincide only in the limit friction $\rightarrow \infty$
I added a line to the Lennard-Jones test case to at least see that it runs without error.

There was a discussion about how different atomic masses affect the time step, is this implementation consistent with the conclusions of that discussion?

This is now consistent with these considerations. The units are accounted for by taking the (finite) friction into account.

[jgreener64]

Great, thanks.