Microscopic Dynamical Entropy I: Quantifying Hamiltonian Irreversibility in Large and Small Systems
Abstract
We introduce a Microscopic Dynamical Entropy (MDE) for Hamiltonian systems, defined with respect to a chosen partition of degrees of freedom into a system X and its environment Y. The construction is based on the conditional phase-space volume (CPV), or conditional Boltzmann entropy, associated with the unmonitored degrees of freedom Y. The MDE is a microscopically defined entropy functional of the marginal distribution ρX(t), obtained by discarding conditional microscopic information associated with Y from the Gibbs entropy of the joint XY system, while retaining exact Hamiltonian dynamics. This construction clarifies the microscopic origin of thermal entropy. The dependence of MDE solely on ρX(t) is consistent with the thermodynamic assumption that the entropy increment of a heat bath Y depends on its heat content and temperature, not on details of its probability distribution. Indeed, the MDE recovers dS = dQ/T connecting entropy increments to heat flow between system and environment. More generally, it provides a consistent description of irreversible relaxation under exact Hamiltonian dynamics, while permitting transient entropy decreases in small systems and in spin-echo type protocols. Under time-scale separation between X and Y, the MDE becomes strictly monotonic in time, recovering the familiar structure of irreversible thermodynamics. The MDE can foreshadow thermodynamics even in a small isolated Hamiltonian system, if X is a well chosen subset of its degrees of freedom. An example is the centre of mass of interacting particles confined to a box. Even for as few as N=10 particles, the MDE increases during relaxation towards a maximum at equilibrium, with increasing monotonicity at larger N. Taken together, our results show that the MDE offers a microscopic interpretation of nonequilibrium thermal entropy and its time dependence within exact Hamiltonian dynamics.
Turn this paper into a full lesson
ArcXiv compiles a staged curriculum from this paper: 8-12 lessons across beginner → advanced, synthesised section guides, visuals, flashcards, a quiz, exercises, and on-demand deep dives per section. Grounded in the abstract, never invented.