Information Flow and Computation in the Maxwell Demon Problem

Abstract

In this paper we examine the Maxwell Demon problem from an information theoretic and computational point-of-view. In particular we calculate the required capacity of a communication channel that transports information to and from the Demon. Equivalently, this is the number of bits required to store the information on a computer tape. We show that, in a simple model for the Maxwell Demon, the entropy of the universe increases by at least an amount eta=0.83999552 bits per particle in going from unsorted to sorted particles and by an amount eta*=2.37314 in going from one sorted state to another sorted state.