A central limit theorem for biased random walks on Galton-Watson trees (Q2480817): Difference between revisions

Let \({\mathcal T}\) be a rooted Galton-Watson tree with offspring distribution \(\{p_k\}\). That is, the number of offspring \(d_v\) of vertices \(v\in {\mathcal T}\) are i.i.d. random variables, with \(P(d_v=k)=p_k\). The authors assume that \(p_0=0, \;\;m:=\sum k p_k>1\) and the existence of some \(\beta>1\) such that \(\sum \beta^k p_k<\infty\). In the paper the \(\lambda -\) biased random walks on the tree \({\mathcal T}\) are considered. These are Markov chains \(\{X_n\}_{n\geq 0}\) with \(X_0=o\) and transition probabilities \(P(X_{n+1}=\omega | X_n=v)= \lambda/(\lambda+d_v)\) (resp. \(1/(\lambda+d_v)\)) if \(v\) is an offspring of \(\omega\) (resp. if \(\omega\) is an offspring of \(v\)). It is known that this walk has an a.s. constant speed \(\text{ v}=\lim_n | X_n| /n\) (where \(| X_n| \) is the distance of \(X_n\) from the root). For all \(\lambda\leq m\) a quenched CLT for \(| X_n| -n\text{ v}\) is proved. For \(\lambda=m\) it is proved that for almost every \({\mathcal T}\), the ratio \(| X_{[nt]}| / \sqrt{n}\) converges in law as \(n\to\infty\) to a deterministic multiple of the absolute value of a Brownian motion.