Thermodynamic ensemble of Stars instead of molecules If we take an enormous amount of molecules (an ensemble), the laws of statistical thermodynamics become valid: we can use them to make predictions of the macroscopic behavior of the ensemble.
I was thinking, a galaxy has an enormous amount of stars and planets. If we consider them to be particles (in a galactic ensemble, made of stars instead of molecules), then we should be able to use the same laws of statistical thermodynamics as before.
We will be replacing microscopic with astronomic dynamics, to make predictions about how the galaxy behaves. We could have analogous versions of temperature, entropy, partition functions, etc. But for a star-gas instead of an oxygen-gas.
Has this model been tried? Could it provide new insight on astronomy?
 A: This question is discussed in section 4.10 of Modern Classical Physics (p.203) by Kip Thorne, where statistical mechanics is used for the description of gravitational systems such as galaxies. It is indeed possible to calculate for instance the entropy of a galaxy, unfortunately it is not possible to calculate any of its (currently) observable properties from thermodynamic extremum principles. This is because galaxies are not close to their equilibrium state.
The non-equilibrium state of galaxies is due to their long thermalization-time: stars undergo very little star-star collisions where momentum transfer can occur. This may be a good thing, as the predicted equilibrium state of a galaxy maximizes entropy by shooting a small fraction of its stars to infinity with the bulk of stars slowing down and collapsing into a black hole. Furthermore this equilibrium state is particular to a galaxy which does not have any interaction with a thermalized bath of stars, which could dramatically change this prediction.
In my opinion this example shows that some information can be gathered from treating star-ensembles with statistical physics tools, but one should be very careful with interpreting the results.
