H-theorem
The H-theorem, proved by Ludwig Boltzmann in 1872, states that the quantity H = ∫ f log f dv — a measure of the entropy of a gas described by the Boltzmann equation — decreases monotonically with time until it reaches a minimum at equilibrium. The theorem is not a consequence of the fundamental laws of mechanics, which are time-reversible, but of the molecular chaos assumption: the claim that colliding particles are uncorrelated before collision. This is where the arrow of time enters statistical mechanics — not through the equations, but through the boundary conditions we impose on them.
The H-theorem was the source of intense controversy in the nineteenth century. Critics, notably Loschmidt and Zermelo, pointed out that time-reversible mechanics cannot yield irreversible conclusions without additional assumptions. Boltzmann's response — that the irreversibility is statistical, not mechanical — was correct but incomplete. The deeper issue is that the H-theorem is not a theorem about molecules; it is a theorem about coarse-graining, about the information loss that occurs when we replace the exact N-particle distribution with a one-particle approximation. The H-function measures the distance between our approximate description and the true equilibrium, and its decrease is a measure of our own epistemic convergence, not the universe's temporal direction.
The H-theorem generalizes beyond kinetic theory. In dynamical systems theory, any system that coarse-grains its phase space loses information, and this loss can be quantified by a Lyapunov function analogous to H. In information theory, the theorem is a continuous-time version of the data processing inequality: processing cannot increase information. The H-theorem is thus a universal statement about the consequences of approximation, not a special result about gases.
The H-theorem is not physics. It is epistemology disguised as physics. It tells us that when we simplify, we lose, and that the direction of this loss is what we call time.