The limiting distribution and error terms for return times of dynamical systems (Q1433441): Difference between revisions

The authors study the limiting distribution of return times of mixing dynamical systems, and also obtain error estimates. First, they study properties of 0-1 valued random variables. Let \(\{\eta_v\colon v=1,\ldots,N\}\) be 0-1 valued random variables with probability \(\mu\), and put \(\zeta=\sum_{v=1}\eta_v\). They study the probability of \({\mathcal N}_k=\{y: \zeta(y)=k\}\), that is, the probability that \(\zeta\) arrives at \(k\). Under several conditions, this probability is almost Poisson distributed: \[ \left|\mu({\mathcal N}_k)-{t^ke^{-t}\over k!}\right| \leq C_1\varepsilon e^{2\alpha't}t, \] where \(\varepsilon\) and \(\alpha'\) are given in conditions. Details and more precise results are given in theorem 1. Let \((\Omega,\mu,T)\) be a dynamical system, and \(\mathcal A\) be a measurable partition. Assume that \(\mathcal A\) is generating and denote \[ {\mathcal A}^n=\bigvee_{j=0}^{n-1}T^{-j}{\mathcal A},\qquad {\mathcal A}^*=\bigcup_n{\mathcal A}^n. \] For a measurable set \(A\), denote by \(\chi_A\) the characteristic function of \(A\). Define the number of visits to \(A\) within the normalized time \(t\) by \[ \xi_A=\sum_{j=1}^{[t/\mu(\chi_A)]}\chi_A\circ T^j. \] Let \[ \begin{aligned} \tau_A(x)&=\min\{k\geq1\colon T^kx\in A\},\\ \tau(A)&= \min_{x\in A}\tau_A(x), \end{aligned} \] and the restriction of the probability measure \(\mu\) to \(n\)-cylinder \(A_n\) is given by \[ \mu_n(B)={\mu(B\cap A_n)\over \mu(A_n)}. \] The authors give several estimates on return times under various conditions (theorem 13--theorem 22), using theorem 1. For example, Theorem~13. Let \(\mu\) be a \(\phi\)-mixing probability measure for the transformation \(T:\Omega\to\Omega\) so that \(\limsup_{v\to\infty}\phi(v)v^p<\infty\) for some positive \(p\). Then there exist constants \(C_7,C_7'\) so that for all \(A_n\in{\mathcal A}^n\) and all \(t,r\) for which \[ {(r+t+1)^2\over t+1}n\mu(A_m)^{p/(p+1)}\leq C_7' \] one has \[ \left|\mu ({\mathcal N}_n^r-{t^r\over r!}e^{-t}\right| \leq n^{q^*}C_7\mu(A_m)^{p/(1+p)} \begin{cases} (r+t)^{2\over r!}(4t+1)^{r-1}e^{4t}&r\geq1,\cr e^{4t}(t+1)&r=0,\end{cases} \] where \(m\) is such that \(m\leq\tau(A_m)\) and \(A_n\subset A_m\in{\mathcal A}^m\) and where I) \(q^*=0\), if \(\mu_*=\mu\) for the distribution of entries, II) \(q^*=1\), if \(\mu_*=\mu_n\) for the distribution of return times where we also require that \(\sum_v\phi(v)<\infty\).