Comments on N-body problem about diffeqphysics.jl

[Mikhail-Vaganov]

Thanks! Very good comments! I will surely take them into account and change some parts of my project. But before that, I would rethink the comments more thoroughly...

[dextorious]

Just a quick remark on those earlier comments - the Andersen and Berendsen thermostats are in fact fundamentally different in that the Andersen thermostat is local (which means, among other things, that it destroys momentum correlations and is therefore unsuitable for any simulation interested in transport properties - diffusion coefficients, viscosities, etc.), while the Berendsen thermostat (despite having other disadvantages like not quite corresponding to the NVT ensemble) is global and therefore of more general applicability. The main point here is that they both have radically different uses.

More generally, Nose-Hoover variants seem to be the standard at the moment for any simulations involving transport properties and other non-equilibrium phenomena, whereas Langevin type integrators (including various geodesic derivatives) are popular in biological applications where they can increase the timestep by an order of magnitude or so.