Attraction time for strongly reinforced walks (Q983889)

Let \(\mathcal{G}\) be a locally finite connected graph with the edge set \(E(\mathcal{G})\) and the vertex set \(V(\mathcal{G})\). The \textit{edge-reinforced random walk} (ERRW) is defined as follows. Consider a \textit{reinforcement weight function} \(w: \{0,1,\dots \}\to (0,\infty)\). The particle being at the vertex \(v\in V(\mathcal{G})\) in the next step jumps to a vertex \(u \in V(\mathcal{G})\) where \(u\) is a neighbor of \(v\) (adjacent to \(v\)); the probability of such jump is \(w\)-proportional to the number of previous traversals of the edge \(\{v,u\}\). The weight \(w\) is assumed \textit{strong}, i.e. \(\sum_{k}1/w(k)< \infty\). Due to \textit{T. Selke} [Markov Proc. Rel. Fields 14, 291--308 (2008; Zbl 1154.82011)] one has \({\mathbf P}(\)the walk ultimately traverses a single edge\()=1\) if and only if \(w\) is strong. Denote by \(I_n\) the random position of the ERRW at time \(n\). The random time of attraction \(T:=\inf\{k\geq 0: \{I_n,I_{n+1}\}= \{I_{n+1},I_{n+2}\}\; \text{for any}\; n\geq k\}\). The authors study the asymptotical behavior of the tail distribution of \(T\). The exact asymptotics (up to a multiplicative constant) are given if \(\mathcal{G}\) has two edges. Some extensions are provided for finite and infinite graphs with bounded degree. It is proved that if \(w(k)=k^{\rho}, \rho > 1\), then universally over finite graphs the expectation \({\mathbf E}T\) is infinite if and only if \(\rho \leq 1+ \frac{1+\sqrt{5}}{2}\).