Poincaré on Gibbs and on Probability in Statistical Mechanics

Abstract

This paper reviews a paper from 1906 by J. Henri Poincaré on statistical mechanics with a background in his earlier work and notable connections to J. Willard Gibbs. Poincaré's paper presents important ideas that are still relevant for understanding the need for probability in statistical mechanics. Poincaré understands the foundations of statistical mechanics as a many-body problem in analytical mechanics (reflecting his 1890 monograph on The Three-Body Problem and the Equations of Dynamics) and possibly influenced by Gibbs independent development published in chapters in his 1902 book, Elementary Principles in Statistical Mechanics. This dynamical systems approach of Poincaré and Gibbs provides great flexibility including applications to many systems besides gasses. This foundation benefits from close connections to Poincaré's earlier work. Notably, Poincaré had shown (e.g. in his study of non-linear oscillators) that Hamiltonian dynamical systems display sensitivity to initial conditions separating stable and unstable trajectories. In the first context it precludes proving the stability of orbits in the solar system, here it compels the use of ensembles of systems for which the probability is ontic and frequentist and does not have an a priori value. Poincaré's key concepts relating to uncertain initial conditions, and fine- and coarse-grained entropy are presented for the readers' consideration. Poincaré and Gibbs clearly both wanted to say something about irreversibility, but came up short.