Quadratic response and speed of convergence of invariant measures in the zero-noise limit (Q1983314)

The authors study statistical properties of expanding maps of the circle perturbed by additive noise. In particular, let \(S_0\) be a deterministic map and \(\delta\) a perturbation parameter. Let \(\mu_\delta\) be the stationary measures associated to the random dynamical system \(S_{\delta}\), obtained by adding at each step to the deterministic dynamics a random independent and identically distributed perturbation of size \(\delta\). One would like to give a meaning to \[ \mu_{\delta} = \mu_0 + \delta \dot{\mu} + \frac{1}{2} \delta^2 \ddot{\mu} + o(\delta^{2}). \] The core result is an explicit formula for the first and second term in the expansion above for sufficiently smooth expanding maps in Sobolev spaces. In particular the element \(\ddot{\mu}\) can be described explicitly in terms of invariants of the noise and of the unperturbed dynamics. The key technique relies on constructing the transfer operator associated to the deterministic dynamics and realizing that the annealed operator of \(S_{\delta}\) can be seen as the convolution of the deterministic operator and a noise kernel. The analysis is performed on chain of spaces of the type \(W^{k,1}\), i.e., Sobolev spaces of functions with \(k\)-derivatives in \(L^1\). For less regular maps, namely, piecewise expanding maps, such expansion does not hold, yet the authors can describe quantitatively the speed of convergence to the zero-noise limit when it exists. For general results on quadratic response see also [the first author and \textit{J. Sedro}, Chaos 30, No. 2, 023113, 15 p. (2020; Zbl 1434.37032)].