given probability space $ (\Omega, T, \mu), \mu$ is ergodic and $ T $ is invertible ( can regard $T$ as two sides shift)

define random $\beta$-transformations: random variable $\beta:\Omega \to (1,\infty)$. random $\beta$-transformations $ f_{\omega}: S^1\to S^1: x\to \beta_{\omega}\cdot x \mod(1) $.( it is in fact random Lasota Yorke maps)

let $ f_{\omega}^n=f_{T^{n-1}\omega} \circ f_{T^{n-2}\omega}\circ \circ \circ f_{T^{1}\omega}\circ f_{\omega}$, $ \phi \in BV[0,1]$.

consider the limit behavior of $ \frac{\sum_{i \le n} \phi \circ f_{\omega}^i}{n^{1/2}}$, can we prove it converges(in distribution) to Gaussian distribution a.s. $ \omega$ ( it is called quenched Central Limit Theorem)?

if $ \inf_{\omega}{\beta_{\omega}} >1$, then it is uniformly hyberbolic, the quenched Central Limit Theorem has been understood very well.

The idea of proof is as follow: we can find global invariant density $h_{\omega }$ ( i.e. $ h_{T \omega}= P_{\omega} h_{\omega}, P_{\omega} \text{ is transfer operator of } f_{\omega}$) from skew product, then we can prove for any $ s>0, \exists C_{\omega, s}, \text{ s.t.} ||h_{T^n \omega}||_{BV} \le C_{\omega,s} \cdot e^{sn}$ ( see Buzzi 1999). If $ \inf_{\omega}{\beta_{\omega}} >1$, I can further show $ \sup_{n,\omega} ||h_{T^n \omega}||_{BV} < \infty $. To prove Central Limit Theorem a.s $ \omega$, decay of correlation a.s $ \omega$ plays an important role, and uniformly bounded invariant density helps us to give a good control of decay of correlation independent of $ \omega$.

However if $ \inf_{\omega}{\beta_{\omega}}=1$, we do not have uniformaly bounded invariant density :$ \sup_{n,\omega} ||h_{T^n \omega}||_{BV} = \infty $, this makes it hard to control Decay of Correlation due to random term $ \omega$. In this case, quench Central Limit Theorem is still true?

Many Thanks!

Your Answer


By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.