Abstract:We survey distributional properties of $\mathbb{R}^d$-valued cocycles of finite measure preserving ergodic transformations (or, equivalently, of stationary random walks in $\mathbb{R}^d$) which determine recurrence or transience.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is about the recurrence conditions of **random walks (recurrence)**. Specifically, the author has studied how to judge the recurrence properties of the related random walks when only the distribution of the random variable \(Y_n\) is known.
### Main Problems and Background
1. **Recurrence of Random Walks**:
- The paper mainly explores the recurrence properties of \(\mathbb{R}^d\)-valued ergodic stationary random processes \((X_n, n\geq0)\) and their related random walks \((Y_n = X_0+\cdots+X_{n - 1}, n\geq1)\).
- The author focuses on how to infer the recurrence of these random walks only through the distribution of the random variable \(Y_n\).
2. **Application of Ergodic Theory**:
- The author uses the methods of ergodic theory to provide some sufficient conditions for the recurrence of this type of random walks without making assumptions about the independence or moments of the process.
### Key Concepts
- **Cocycle**: Given a measure - preserving ergodic transformation \(T\), and a Borel mapping \(f:X\rightarrow\mathbb{R}^d\), a cocycle \(f(n, x)\) is defined as:
\[
f(n, x)=\begin{cases}
f(T^{n - 1}x)+\cdots+f(x)&\text{if }n > 0,\\
0&\text{if }n = 0,\\
- f(-n, T^n x)&\text{if }n < 0.
\end{cases}
\]
- **Recurrence**: A cocycle \(f\) is recurrent if for almost all \(x\in X\),
\[
\liminf_{n\rightarrow\infty}\|f(n, x)\| = 0.
\]
- **Coboundary**: If there exists a Borel mapping \(b:X\rightarrow\mathbb{R}^d\) such that
\[
f(x)=b(Tx)-b(x),
\]
then \(f\) is called a coboundary.
- **Cohomologous**: Two Borel mappings \(f\) and \(f'\) are cohomologous if there exists a Borel mapping \(b:X\rightarrow\mathbb{R}^d\) such that
\[
f'(x)=b(Tx)+f(x)-b(x).
\]
### Main Results
- **Theorem 3**: Let \(T\) be an ergodic measure - preserving automorphism on a standard probability space, and \(f:X\rightarrow\mathbb{R}\) be an integrable Borel mapping. Then \(f\) is recurrent if and only if \(\int f d\mu=0\).
- **Theorem 4**: If a Borel mapping \(f:X\rightarrow\mathbb{R}\) satisfies the weak law of large numbers (i.e., \(\lim_{n\rightarrow\infty}\frac{f(n,\cdot)}{n}=0\)), then \(f\) is recurrent.
- **Theorem 9**: If there exist \(\epsilon, K>0\) such that for every \(n\geq1\), \(\mu(\{x\in X:|f(n, x)|\leq K\})>\epsilon\), then \(R(f)\neq\emptyset\).
- **Theorem 12**: Let \(T\) be a standard probability