Large deviations and the distribution of pre-images of rational maps (Q679397)

Let \(T\) be a rational function viewed as a self mapping of degree \(d\) of the complex sphere, and let \(J\) be the Julia set attached to \(T\). Due to a result of Lyubich, for each point \(x \in J\) and as \(n \to \infty\), the evenly distributed weight on the sets \(S_n(x)=\{y:T^n y=x\}\) converges to the unique measure \(\mu_0\) of maximal entropy (with respect to the weak* topology). The pressure \(P(f)\) of a continuous mapping \(f:J\to{\mathbb R}\) is defined to be \(P(f) = \sup \{ h(\nu)+\int f d\nu : \nu\) is a \(T\)-invariant probability measure\}, where \(h(\nu)\) denotes the entropy of \(T\) with respect to \(\nu\). If \(P(f) > \sup f\), then due to a result of Denker and Urbanski \(f\) has a unique equilibrium state \(\mu\). The authors prove the following theorem which generalizes Lyubich' result. Theorem. Let \(f^n(y) = f(y) + \cdots + f(T^{n-1}y)\) and \(\mu_{y,n} = \frac{1}{n}(\delta_y + \cdots + \delta_{T^{n-1}y})\). Then there exists some constants \(C>0\) and \(0<\eta<1\) such that for every weak* open neighborhood \(U\) of \(\mu\) the inequality \[ \sum_{T^ny=x, \mu_{y,n} \not\in U} e^{f^n(y)} \Bigl/ \sum_{T^ny=x} e^{f^n(y)} \leq C \eta^n \] holds. In other words, the weighted proportion of the measures \(\mu_{y,n} \not\in U\) converge to zero exponentially fast.