Large deviations for unbounded additive functionals of a Markov process with discrete time (noncompact case) (Q1344658)

Exponential tightness and the fact that it implies the large deviation principle (LDP) for subsequences are used to prove LDP of average for unbounded additive functionals \(g(\xi_ k)\) of an ergodic Markov process \(\xi_ k\) with real line as its state space. The proof is based on the LDP of empirical measures \({1 \over n} \sum^ n_{k=1} \delta_{ \xi_ k}\), due to Donsker and Varadhan. Examples in Section 3 of the paper apply the main theorem to Markov processes defined by nonlinear recursions. Explicit sufficient growth conditions for \(g (\cdot)\) are given.