Cutpoints of (1,2) and (2,1) random walks on the lattice of positive half line (Q6204787)

The authors consider \((2,1)\) and \((1,2)\) random walks in spatially inhomogeneous environments on the lattice of positive half line. The transition probabilities of the random walks at site \(n\) are assumed to be asymptotically constant as \(n \to\infty\). The notion of \((1,2)\) random walk means that the left-oriented jumps are always of size 1 and the right-oriented jumps are either of size 1 or of size 2. The notion of \((2,1)\) random walk is introduced similarly. The purpose is to give criteria for finiteness of the number of cutpoints for both \((2,1)\) and \((1,2)\) random walks and then to study the asymptotics of the number of cutpoints in \([0, n]\) as \(n \to\infty\). Roughly speaking, if the walk never returns to \([0, x]\) after its first entry into \([x + 1,\infty)\), then \(x\) is a cutpoint. Clearly, in the recurrent case, there is no cutpoint on the path of the walk and in the transient case, if the walk runs to infinity more quickly, there are more cutpoints. Such \((1,2)\) and \((2,1)\) random walks are more complicated than the nearest-neighbor random walks because the escape probability of the walk from an interval is written in terms of the products of inhomogeneous 2-by-2 nonnegative matrices, whose entries are hard to evaluate. To overcome such difficulties, the author's idea is to estimate the entries of the products of matrices in terms of the products of tails of some related continued fractions. Then, based on some delicate analysis of the continued fraction and the products of matrices, they find the asymptotics of the escape probabilities and the hitting probabilities of the random walk, which is crucial to give criteria for the finiteness of the number of cutpoints. Finally, sufficient and necessary condition for the finiteness of the number of cutpoints is obtained.