Misusing the entropy maximization in the jungle of generalized entropies

TL;DR

The paper demonstrates that the maximum entropy principle only justifies factorized equilibrium distributions for additive entropies, and not for generalized entropies like Tsallis and Rényi, challenging their thermodynamic consistency.

Contribution

It clarifies the conditions under which maximum entropy yields valid partition functions, highlighting limitations for non-additive generalized entropies.

Findings

Normalized equilibrium distributions require additive slope of entropy.

Tsallis and Rényi entropies do not produce factorized canonical distributions.

Maximum entropy principle is not universally applicable for generalized entropies.

Abstract

It is well-known that the partition function can consistently be factorized from the canonical equilibrium distribution obtained through the maximization of the Shannon entropy. We show that such a normalized and factorized equilibrium distribution is warranted if and only if the entropy measure $I \{(p_i)\}$ has an additive slope i.e. $\partial I \{(p_i)\} / \partial p_i$ when the ordinary linear averaging scheme is used. Therefore, we conclude that the maximum entropy principle of Jaynes should not be used for the justification of the partition functions and the concomitant thermodynamic observables for generalized entropies with non-additive slope. Finally, Tsallis and R\'enyi entropies are shown not to yield such factorized canonical-like distributions.