H-theorems from macroscopic autonomous equations (Q2506289)

The authors give mathematically precise and physically reasonable formulations of the Boltzmann entropy and the Second Law. They consider a family of dynamical systems \((\Omega^N, U_t^N, \rho^N)\), where \(\Omega^N\) is a microscopic phase space and \(\rho^N\) is a probability measure on \(\Omega^N\) invariant under the dynamics \(U_t^N: \Omega^N \to \Omega^N\). The integer \(N\) is thought of as the degree of freedom. They also introduce a metric space \(({\mathcal F}, d)\), which is regarded as the space of macroscopic states and maps \(m^N:\Omega^N \to {\mathcal F}\). A version of the Boltzmann entropy is defined by \[ H(m) = \lim_{\delta\downarrow 0}\lim_{N\uparrow +\infty} \tfrac{1}{N}\log\rho^N(\{d(m^N(\,\cdot\,), m)\leq \delta\}) \; \text{ for } \; {m \in \mathcal F}, \] if it exists. Then under an assumption of dynamical autonomy in a weak sense for a map \( \phi_t:{\mathcal F} \to {\mathcal F}\) and the dynamics \(U_t^N\), the Second Law is shown, i.e., \( H(\phi_t(m))\) is increasing in \(t\). The case for finite but large \(N\) is also considered.