Rate of escape and central limit theorem for the supercritical Lamperti problem (Q2638362)

The authors study the limit behavior for certain supercritical Lamperti problems. More precisely, let \(X:=(X_t)_{t\geq0}\) be a time-homogeneous Markov process on \([0,\infty[\) with repelling bound \(0\) such that the increment moment functions \[ \mu_k:=E((X_{t+1}-X_t)^k|X_t=x) \] satisfy \(\mu_1(x)\sim x^{-\beta}\) for \(\beta\in]0,1[\) and possibly in addition, \(\mu_2(x)-\mu_1(x)^2\to \sigma^2>0\) for \(x\to\infty\). The authors then prove that \(X\) is transient, that \(X_t/t^{1/(1+\beta)}\) tends a.s.~to some explicit constant \(\lambda\), and that \[ (X_t-\lambda t^{1/(1+\beta)})/\sqrt{t} \] tends to a centered normal distribution with variance \(\sigma^2(1+\beta)/(1+3\beta)\). These interesting results are proved even in a more general setting, but under some technical conditions like existence of certain moments. Moreover, some of the results also cover the case \(\beta=1\). Finally, applications to birth-and-death random walks and multi-dimensional non-homogeneous random walks are given. The LLN of the paper was known in some special cases, while the CLT seems to be completely new.