After reading this question What's the intuitive reason that phase space flow is incompressible in Classical Mechanics but compressible in Quantum Mechanics? and some papers of Steuernagel’s group, like “The Quantum Liouville Equation is non-Liouvillian”
I started to think about the relation between Wigner flows and entropy.
I imagined an ensemble of Wigner quasi-probability distributions over the phase space of a closed system.
The ensemble has the property that the sum of W(x, p) over all its elements always equals some constant for each point [x, p] inside some shell of the phase space: the ensemble as a whole has constant density over the shell.
With “shell” I mean the set of all points that would imply total energy between some E1 and E2.
My question is: can you choose an ensemble and a Hamiltonian (keeping the system closed) in such a way that this property is respected at time 0 for some shell but is no longer true at some later time for that same shell?
This would be possible if divergence was not a function of the quasi-probability distribution, but it is not clear to me how changes in quasi-probability distribution affect divergence. For example: how does a rotation of the distribution around the centre of phase space affect the divergence field? Are there any systems for which the divergence field does not simply rotate by the same angle?
If you add the fact that for the ensemble as a whole probability on shells is roughly conserved over time (I am not completely sure about this, but it seems sensible, especially for a thick shell) I think an ensemble evolving in such a way would violate the second law of thermodynamics since it would move away from a uniform density distribution, that is the one with maximum entropy.
I don’t see how a non-Liouvillian flow can possibly not imply violation of the second law at least in some cases.
Do you think Wigner flows being non-Liouvillian could be somehow connected to this line of research?
Here they claim that local violations of the second law are possible.