Deviations of a random walk in a random scenery with stretched exponential tails (Q2490072)

Consider a \(d\)-dimensional random walk in a random scenery \(Z_n =\sum_{k=0}^{n-1} Y_{S_k}\), \(n\geq 1\), where \((S_k; k\in \mathbb N)\) is a random walk in \(\mathbb Z^d\) starting at the origin and \((Y_x;x\in\mathbb Z^d)\) is an i.i.d. scenery, independent of the walk. The assumptions on the scenery are that \(Y_0\) is centered with a finite second moment and has a stretched exponential tail. In particular, \(Y_0\) does not have any positive exponential moments. Under these assumptions, it is known that for \(d\geq 1\), there is a sequence \((a_n)\) such that \(Z_n/na_n\) converges in distribution towards some non-degenerate random variable. For \(d\geq3\), \(a_n =n^{-1/2}\). The goal of this paper is to obtain precise logarithmic asymptotics for \( \mathbb{P}(Z_n > n t_n)\) for some sequences of positive numbers \((t_n)\) such that \(\lim_n t_n/a_n = \infty\). The first step of the proof is a generalization of a result of \textit{A. V. Nagaev} [Teor. Veroyatn. Primen. 14, 203-216 (1969; Zbl 0181.45004) and ibid. 14, 51-63 (1969; Zbl 0172.21901)]. Then, using moderate deviations for the local time of the random walk at the origin, the proof can be completed.