Stochastic evolution on a manifold of states (Q1083131)

There are physical systems large enough so that a statistical description is needed but not necessarily so large that fluctuations may be ignored. We introduce a general framework for such systems based on assuming that the system's evolution defines a Markov process occurring in a manifold of physical states. A single system is described by its state-manifold M and three fields thereon: the entropy-scalar s, the diffusion tensor \(K^{ab}\), and the vector of non-diffusive evolution \(u^ a.\) The second law of thermodynamics holds if and only if \(u^ a\) is a symmetry of s. Detailed balance holds if and only if \(u^ a\) vanishes. In the macroscopic limit where fluctuations disappear one obtains a general framework for irreversible thermodynamics (of isolated systems); and in that limit the equations of motion are derived from a variational principle. The meaning of the concepts introduced is illustrated by means of an analysis in some detail of Brownian motion and the ''Langevin process''. An application to black hole evolution is also discussed.