Title: A non-isomorphic decomposition analysis of percolation, correlated systems, and renormalization groups

Abstract: Any system with internal correlations (e.g., interactions), such as spin glasses, random media, and brains, can be studied as a network where nodes are elements and edges denote correlated interactions. As internal correlations increase, correlated behaviours originate and propagate among elements to create percolation phenomena. In this work, we study percolation in correlated systems by proposing a framework referred to as the non-isomorphic decomposition analysis. This framework analyzes the evolution of the space of non-isomorphic percolation configuration classes governed by internal correlations, where a metric called as the automorphism entropy is proposed to measure the reduction of information (i.e., freedom degrees) in the space. A special case of our theory under the mean-field approximation is elaborated, where all concepts can be explicitly calculated based on group theory and percolation theory. Our framework has been applied to studying absorbing phase transition, spreading process, and synchronization, whose results reveal the shared properties of percolation and dynamics in different correlated systems. First, the percolation process driven by internal correlations is intrinsically an information loss process in the probability space of non-isomorphic percolation configuration classes. The metric of information loss degree, automorphism entropy, serves as a sensitive indicator of percolation transition and related processes in diverse systems. Second, the information loss intensity depends on the strength of internal correlations. Sufficiently strong correlations ensure slight information losses and make specific freedom degrees omissible, which supports the implementation of renormalization groups. Sufficiently weak correlations, on the contrary, imply significant information losses and make renormalization groups less effective for the correlated system.