Paper detail

A comprehensive classification of complex statistical systems and an ab-initio derivation of their entropy and distribution functions

To characterize strongly interacting statistical systems within a thermodynamical framework - complex systems in particular - it might be necessary to introduce generalized entropies, $S_g$. A series of such entropies have been proposed in the past, mainly to accommodate important empirical distribution functions to a maximum ignorance principle. Until now the understanding of the fundamental origin of these entropies and its deeper relations to complex systems is limited. Here we explore this questions from first principles. We start by observing that the 4th Khinchin axiom (separability axiom) is violated by strongly interacting systems in general and ask about the consequences of violating the 4th axiom while assuming the first three Khinchin axioms (K1-K3) to hold and $S_g=\sum_ig(p_i)$. We prove by simple scaling arguments that under these requirements {\em each} statistical system is uniquely characterized by a distinct pair of scaling exponents $(c,d)$ in the large size limit. The exponents define equivalence classes for all interacting and non interacting systems. This allows to derive a unique entropy, $S_{c,d}\propto \sum_i Γ(d+1, 1- c \ln p_i)$, which covers all entropies which respect K1-K3 and can be written as $S_g=\sum_ig(p_i)$. Known entropies can now be classified within these equivalence classes. The corresponding distribution functions are special forms of Lambert-$W$ exponentials containing as special cases Boltzmann, stretched exponential and Tsallis distributions (power-laws) -- all widely abundant in nature. This is, to our knowledge, the first {\em ab initio} justification for the existence of generalized entropies. Even though here we assume $S_g=\sum_ig(p_i)$, we show that more general entropic forms can be classified along the same lines.