Remarks on central limit theorems for the number of percolation clusters (Q2495264)

Let \(G=(V,E)\) be an infinite connected graph. Each bond \(e\in E\) is independently declared to be open (resp. closed) with probability \(p\) (resp. \(1-p\)). By \(P_p\) the Bernoulli measure is denoted. Fix an increasing sequence \(\{B(n)\}\) of finite regions containing the origin. Denote \[ K_n=\sum_{x\in B(n)}\frac{1}{| C_n(x)| }, \quad \widetilde K_n=\sum_{x\in B(n)}\frac{1}{| C_n(x)| }1_{\{| C_n(x)| >1\}}, \] where \(C_n(x)=\{y\in B(n): \) there is an open path in \(B(n)\) from \(x\) to \(y\}\). The central limit theorem is said to hold for \(\{f_n\}\) if \[ \frac{f_n-E_pf_n}{\sqrt{\text{var}_pf_n}}\to N(0,1)\quad \text{as} \quad n\to \infty, \] where the convergence is meant in law and \(N(0,1)\) denotes the standard normal distribution. Here \(E_p\) (resp. \(\text{var}_p\)) denotes the expectation (resp. variance) relative to \(P_p\). The Bernoulli bond percolation problem on Bethe and Sierpinski carpet lattices is considered respectively, and the central limit theorems for \(K_n\) and \(\widetilde K_n\) with respect to these percolations are proved. The authors use the argument given in [\textit{Y. Zhang}, J. Theor. Probab. 14, 165--187 (2001; Zbl 0974.60098)].