The Key Renewal Theorem for a Transient Markov Chain.pdf

a r X i v : 0 7 1 1 .2 1 6 9 v 1 [ m a t h .P R ] 1 4 N o v 2 0 0 7 The Key Renewal Theorem for a Transient Markov Chain Dmitry Korshunov1 We consider a time-homogeneous Markov chain Xn, n ≥ 0, valued in R. Sup- pose that this chain is transient, that is, Xn generates a σ-finite renewal mea- sure. We prove the key renewal theorem under condition that this chain has asymptotically homogeneous at infinity jumps and asymptotically positive drift. KEY WORDS: transient Markov chain, renewal kernel, renewal measure, key renewal theorem, Green function. Let ξ1, ξ2, . . . be independent identically distributed random variables with a common distribution F on R. Put Sn = ξ1+ · · ·+ξn, S0 = 0, and consider the renewal measure generated by sums: U(B) ≡ E ∞∑ n=0 I{Sn ∈ B} = ∞∑ n=0 F ?n(B). If F is non-lattice then the celebrated key renewal theorem states that, for every fixed h 0, U(x, x+ h] → h Eξ1 as x → ∞, (1) provided Eξ1 is finite and positive (see, for example, Feller and Orey [8], Feller [7, Ch. XI], Woodroofe [25, Appendix]); F is called lattice if it is concentrated on some lattice {ka, k ∈ Z} with a 0. If F is lattice the same is true when h is a multiple of the span a. It is proved in Wang and Woodroofe [23] and in Borovkov and Foss [5, Theorem 2.6] that (1) holds uniformly over certain classes of distributions F . Some extensions of the key renewal theorem to the nonidentically dis- tributed case are considered by Williamson [24], Maejima [18]. Another extension is aimed to include random walks perturbed by both a slowly changing sequence and a stationary one, Zn = Sn + ηn + ζn say; see, for example, Lai and Siegmund [16, 17], Woodroofe [25], Horva?th [11], Zhang [26], Kim and Woodroofe [13]. All these extensions deal with perturbations depending on time rather than on state space; in most cases summands are independent and have finite variance. Many papers (see, for example, Kesten [12], Athreya, McDonald and Ney [2], Nummelin [20], Alsmeyer [1], Klu?ppe