Large deviation principle for Benedicks-Carleson quadratic maps (Q690735)

The authors consider the family of quadratic maps \(f_a(x)=1-ax^2\), \(0<a\leq 2\) and \(x\in [-1,1]\). From the works of \textit{M. V. Jakobson} [Commun. Math. Phys. 81, 39--88 (1981; Zbl 0497.58017)] and \textit{M. Benedicks} and \textit{L. Carleson} [Ann. Math. (2) 122, 1--25 (1985; Zbl 0597.58016); ibid. (2) 133, No. 1, 73--169 (1991; Zbl 0724.58042)], it is well known that there exists a set (of positive Lebesgue measure) of values of the parameter \(a\), close to \(2\), for which the corresponding \(f_a\) admits an invariant probability measure \(\mu\) absolutely continuous with respect to the Lebesgue measure. Consequently, for Lebesgue-almost every point \(x\in[-1,1]\) it holds that \(\delta_{x}^n:=\frac{1}{n}\sum_{j=0}^{n-1}\delta_{f_{a}^j(x)}\) tends in the weak topology to \(\mu\) as \(n\) goes to infinity. The paper under review deals with the problem of analyzing the situation (large deviation principle) when \(\delta_{x}^n\) stays away from \(\mu\). If a set of appropriate conditions (A1)-(A4) is imposed, namely,\newline (A1) \(a\) is sufficiently close to \(2\),\newline (A2) \(|Df_a^n(f_a(0))|\geq e^{n\frac{9}{10}\log 2}\) for every \(n\geq 0\),\newline (A3) \(|f_a^n(0)|\geq e^{-\frac{1}{100}\sqrt{n}}\) for every \(n\geq 1\),\newline (A4) \(f_a\) is topologically mixing on \([f_a^2(0),f_a(0)],\)\newline then it is possible to establish a main theorem concerning ``a full large deviation principle''. The statement of this theorem is too involved to be described here. It is interesting to mention that, in the words of the authors, their ``theorem is the first full large deviations result for a positive measure set of quadratic maps, despite a large number of papers over the past thirty years dedicated to stochastic properties of dynamics in one-dimensional maps.''