Shortened abstract: Microcanonical equilibrium macrostates are characterized as the solutions of a constrained minimization problem, while canonical equilibrium macrostates are characterized as the solutions of a related, unconstrained minimization problem. In Ellis, Haven, and Turkington (J. Stat. Phys. 101, 999, 2000) the problem of ensemble equivalence was completely solved at two separate, but related levels: the level of equilibrium macrostates, which focuses on relationships between the corresponding sets of equilibrium macrostates, and the thermodynamic level, which focuses on when the microcanonical entropy $s$ can be expressed as the Legendre-Fenchel transform of the canonical free energy. The present paper extends the results of Ellis et al. significantly by addressing the following motivational question. Given that the microcanonical ensemble can be nonequivalent with the canonical ensemble, is it possible to replace the canonical ensemble with a generalized canonical ensemble that is equivalent with the microcanonical ensemble? The generalized canonical ensemble that we consider is obtained from the standard canonical ensemble by adding an exponential factor involving a continuous function $g$ of the Hamiltonian. As in the paper by Ellis et al., we analyze the equivalence of the two ensembles at both the level of equilibrium macrostates and the thermodynamic level. A neat but not quite precise statement of the main result in the present paper is that the microcanonical and generalized canonical ensembles are equivalent at the level of equilibrium macrostates if and only if they are equivalent at the thermodynamic level, which is the case if and only if the generalized microcanonical entropy $s-g$ is concave.
Read full abstract