International Symposium Honoring Mathematical Physicist Jean-Pierre Vigier
Information, Entropy, and the Classical Ideal Gas
David Sands, Jeremy Dunning-Davies
Abstract
The physical basis of the canonical and grand canonical distributions is questioned. In particular, we question the usual methods by which these distributions are derived, namely that fluctuations in entropy around energy and particle number are assumed to occur when the entropy depends only on variables which cannot themselves fluctuate. We show, starting from the Maxwellian velocity distribution, that the probability that a classical ideal gas at a fixed temperature occupies a given energy state corresponds not to the canonical ensemble of classical statistical mechanics but to the Gamma distribution. Computer simulations of a hard-sphere fluid demonstrate the principles. The analysis is extended to open systems in which the number of particles fluctuates and we show that for a system connected to a particle reservoir the Poisson distribution governs the probability of finding a given number of particles. The resulting probability distributions are used to calculate the Shannon information entropy which is then compared with the thermodynamic entropy. We argue that information theoretic entropy and thermodynamic entropy, whilst related, are not necessarily identical and that the information entropy contains non-thermodynamic components.