Persistent random walks. II. Functional scaling limits (Q1741872)

A persistent random walk \((S_n)_{n\geq0}\) on \(\mathbb Z\) has jumps of unit size and the probability of jumping up=\(u\) or down=\(d\) depends on the time the walker already spent in the current direction. It can be written as \(S_n=\sum_{k=0}^n X_k\), where \((X_k)\) is a Markov chain with variable length on \(\{-1,+1\}\) induced by a certain probabilized context tree. The paper is a continuation of Part I [\textit{P. Cénac} et al., J. Theor. Probab. 31, No. 1, 232--243 (2018; Zbl 1390.60165)] in which recurrence and transience properties of a persistent random walk (PRW) were investigated. The movement of the PRW can be completely described by the distributions of the lengths \(\tau^u,\,\tau^d\) of a typical rise or descent of the walk. The authors introduce a mean drift function \[m_S=\lim_{t\to\infty}\frac{\Theta_u(t)-\Theta_d(t)}{\Theta_u(t)+\Theta_d(t)},\] where \(\Theta_\ell(t)=\mathbb E[\tau^\ell\wedge t]\) for \(\ell\in\{u,d\}\) and assume that \(m_S\in(-1,1)\) is not extremal and \((1-m_S)\tau^u+(1+m_S)\tau^d\) belongs to the domain of attraction of an \(\alpha\)-stable distribution with \(\alpha\in(0,2]\). If \(\tau^u\) and \(\tau^d\) are identically distributed, the PRW is a directionally reinforced random walk (DRRW) and thus a PRW can be considered as an anisotropic version of a DRRW. The main result of the paper states that for a certain normalization function \(\lambda\), which is regularly varying of index \(\min\{1, 1/\alpha\}\), a functional limit theorem holds \[\frac{S_{\lfloor ct\rfloor}-m_S ct}{\lambda(c)}\to Z_\alpha(t)\] as \(c\to\infty\) in the \(J_1\)-topology on the space of càdlàg functions and the limiting process \(\{Z_\alpha(t)\}_{t\geq0}\) is either Brownian motion if \(\alpha=2\), a strictly \(\alpha\)-stable Lévy process with certain skewness if \(\alpha\in(1,2)\), a symmetric Cauchy process if \(\alpha=1\), or an arcsin Lamperti anomalous diffusion if \(\alpha\in(0,1)\).