Boltzmann Constant in LAMMPS 'real' units does not work about molly.jl

[jgreener64]

Can you paste code giving the error? As I recall we convert k to make kT the same units as energy_units, with any molar term removed.

[jgreener64]

Internally where is kB being used that the unit conversion is important? I didn't look that closely.

Off the top of my head it is things like velocity generation and temperature. There might be a more principled way to do it.

More generally k is only supposed to be supplied to the constructor if it is numerically different to the real Boltzmann constant, which is apparently a useful feature in some cases. In most cases, including here, you can just omit k and it gets converted appropriately.

[ejmeitz]

Will do later, if you construct a System with those arguments it should give the error though.

[jgreener64]

I get an error with all 3 arguments, but it works without specifying k. In this case sys.k ends up as 3.299830305927343e-27 kcal K^-1 which is equal to k / Unitful.Na.

Alternatively passing in k=3.299491e-27u"kcal * K^-1" also works.

[ejmeitz]

Yeah I got the same results, I was not expecting k to be converted to away from being molar when the energy I specify is molar. At least for my use case kB needed to be molar for the math to work out.

Internally where is kB being used that the unit conversion is important? I didn't look that closely.

Also is there a reason to not support passing kB as molar?