Abstract:Erdős and Révész initiated the study of favorite sites by considering the one-dimensional simple random walk. We investigate in this paper the same problem for a class of null-recurrent randomly biased walks on a supercritical Gaton-Watson tree. We prove that there is some parameter $\kappa \in (1, \infty]$ such that the set of the favorite sites of the biased walk is almost surely bounded in the case $\kappa \in (2, \infty]$, tight in the case $\kappa=2$, and oscillates between a neighborhood of the root and the boundary of the range in the case $\kappa \in (1, 2)$. Moreover, our results yield a complete answer to the cardinality of the set of favorite sites in the case $\kappa \in (2, \infty]$. The proof relies on the exploration of the Markov property of the local times process with respect to the space variable and on a precise tail estimate on the maximum of local times, using a change of measure for multi-type Galton-Watson trees.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the behavior of "favorite sites" of a class of randomly biased walks on supercritical Galton - Watson trees. Specifically, the author studies the properties of the set of favorite sites of such random walks, in particular whether these sets are almost surely bounded, compact, or oscillate over time. The main contribution of the paper is to prove that there exists a parameter \(\kappa\in(1,\infty]\) such that when \(\kappa\in(2,\infty]\), the set of favorite sites is almost surely bounded; when \(\kappa = 2\), the set of favorite sites is compact; and when \(\kappa\in(1,2)\), the set of favorite sites may almost surely move to infinity. In addition, the paper also gives a complete answer to the cardinality of the set of favorite sites when \(\kappa\in(2,\infty]\).
### Main Conclusions
1. **When \(\kappa\in(2,\infty]\)**:
- The set of favorite sites \(F(n)\) is almost surely bounded, that is, there exists an almost surely finite set \(M\) such that for large enough \(n\), \(F(n)\subset M\).
2. **When \(\kappa = 2\)**:
- The set of favorite sites \(F(n)\) is compact, that is, \(\sup_{x\in F(n)}|x|\) is compact.
3. **When \(\kappa\in(1,2)\)**:
- For any \(\epsilon> 0\),
\[
\limsup_{n\rightarrow\infty}n^{-\frac{\kappa - 1}{\kappa+\epsilon}}\inf_{x\in F(n)}|x|=\infty,\quad P^*\text{-a.s.}
\]
\[
\liminf_{n\rightarrow\infty}\sup_{x\in F(n)}|x|<\infty,\quad P^*\text{-a.s.}
\]
### Proof Strategies
- **Markov Property of the Spatial Variable of the Local Time Process**: By exploring the Markov property of the spatial variable of the local time process, the author introduces the edge local time \(L_n(x)\) and a sequence of stopping times \(T_n\).
- **Tail Distribution Estimation**: One of the key steps is to estimate the tail distribution of the maximum of the maximum edge local time. Specifically, as \(r\rightarrow\infty\),
\[
P\left(\max_{x\in T}L_{T_1}(x)\geq r\right)\asymp
\begin{cases}
r^{-1}, & \text{if }1 < \kappa < 2,\\
r^{-1}(\log r)^{-1/2}, & \text{if }\kappa = 2,\\
r^{-\kappa/2}, & \text{if }2 < \kappa < \infty.
\end{cases}
\]
- **Application of Extreme Value Theory**: Using extreme value theory and combining with the properties of local time, the behavior of favorite sites is derived.
### Related Theorems
- **Theorem 1.1**: Describes the properties of the set of favorite sites for different values of \(\kappa\).
- **Theorem 1.5**: Provides the tail distribution estimate of the maximum of the maximum edge local time.
### Applications
- **Corollary 1.3**: When \(\kappa\in(2,\infty]\), the cardinality of the set of favorite sites is eventually no more than 3, and this upper bound is optimal.
- **Corollary 1.4**