Moderate deviations for a random walk in random scenery (Q952824)

The cumulative scenery process of the random walk in random scenery \(\{X_n; n \geq 0\}\) is considered. Here \(X_n=\sum_{1\leq k\leq n} \xi(S_k)\) where \(\{S_n; n \geq 0\}\) is a symmetric and aperiodic walk on the lattice \(\mathbb{Z}^d\) with nondegenerate covariance matrix and \(\{\xi(z):z \in \mathbb{Z}^d\}\) is a scenery given by real-valued random variables, which are identically distributed, independent of each other and the walk and satisfy Cramér's condition. The authors consider moderate deviations of \(\{X_n; n \geq 0\}\) extending from the central limit scaling up to the point where either the deviation speed or the rate function starts to depend on the actual distribution of the scenery. For the case \(d \geq 4\) a precise asymptotics for the probability of a moderate deviation is obtained which extends a classical central limit theorem of Kesten and Spitzer. For \(d \geq 3\), an important ingredient in the proofs are new concentration inequalities for self-intersection local times, whilst for \(d \geq 2\) a recent moderate deviation result for self-intersection local times is used which is due to Bass, Chen and Rosen.