INFORMATION CASCADE WITH MARGINAL STABILITY IN A NETWORK OF CHAOTIC ELEMENTS

A newly discovered cascade process of clusters is studied in a network of chaotic elements. It is shown that the splitting of clusters and the synchronization of elements are balanced in a class of partially ordered states, where marginal stability is sustained over an interval of the bifurcation parameters. The partition information creation in bit space shows an avalanche process of information, which leads to the anomalous behavior of power spectra, roughly fitted by a power law form of the wavenumber. Lyapunov spectra have accumulation at null exponents, analogous with those studied in fluid turbulence models.