Linear response, susceptibility and resonances in chaotic toy models

Abstract

We consider simple examples illustrating some new features of the linear response theory developed by Ruelle for dissipative and chaotic systems [{\em J. of Stat. Phys.} {\bf 95} (1999) 393]. In this theory the concepts of linear response, susceptibility and resonance, which are familiar to physicists, have been revisited due to the dynamical contraction of the whole phase space onto attractors. In particular the standard framework of the "fluctuation-dissipation" theorem breaks down and new resonances can show up oustside the powerspectrum. In previous papers we proposed and used new numerical methods to demonstrate the presence of the new resonances predicted by Ruelle in a model of chaotic neural network. In this article we deal with simpler models which can be worked out analytically in order to gain more insights into the genesis of the ``stable'' resonances and their consequences on the linear response of the system. We consider a class of 2-dimensional time-discrete maps describing simple rotator models with a contracting radial dynamics onto the unit circle and a chaotic angular dynamics $θ_{t+1} = 2 θ_t (\mod 2π)$. A generalisation of this system to a network of interconnected rotators is also analysed and related with our previous studies \cite{CS1,CS2}. These models permit us to classify the different types of resonances in the susceptibility and to discuss in particular the relation between the relaxation time of the system to equilibrium with the {\em mixing} time given by the decay of the correlation functions. Also it enables one to propose some general mechanisms responsible for the creation of stable resonances with arbitrary frequencies, widths, and dependency on the pair of perturbed/observed variables.