Bourgain's "Return Times Theorem" establishes that if $(\Omega_{j},\mathcal{F}_{j},\mathbb{P}_{j},T_{j})$ ($j=1,2$) are measure-preserving Dynamical systems (i.e. $(\Omega_{j},\mathcal{F}_{j},\mathbb{P}_{j})$ are probability spaces and $T_{j}:\Omega_{j}\to\Omega_{j}$ are measure-preserving transformations), then for every $p_{1}>1$ and every $X\in L^{p_{1}}_{\mathbb{P}_{1}}$, there exists $\Omega_{1,X}$ (depending only on $X$) with $\mathbb{P}_{1}\Omega_{1,X}=1$ such that, if $p_{2}=p_{1}/(p_{1}-1)$ is the conjugate of $p_{1}$, then for every $Y\in L^{p_{2}}_{\mathbb{P}_{2}}$ and every $(\omega_{1},\omega_{2})\in \Omega_{1,X}\times\Omega_{2}$ $$\frac{1}{n}\sum_{k=0}^{n-1}T_{1}^{k}X(\omega_{1})T_{2}^{k}Y(\omega_{2}) \,\,\,\,\,\,\,\,(1)$$ is convergent, where $T^{k}_{j}Z=Z \circ T^{k}_{j}$ is the $k-$th iterate of the "Koopman" operator $Z\mapsto Z\circ T_{j}$ in $L^{p_{j}}_{\mathbb{P}_{j}}$ (the reason for this pedantic presentation will be cleared below). See for instance this entry of Tao's blog and the references therein for a more detailed discussion.

As indicated by Tao in his post, the Return Times Theorem comes as an answer to a natural question arising from the "automatic" extension of Birkhoff's Ergodic theorem to double recurrence.

Now, Birkhoff's ergodic theorem is a particular instance of the following satement: if $T$ is a positive contraction in $L^p_{\mathbb{P}}$ for every $p\geq 1$ (a positive Dunford-Schwartz operator), then for every $p\geq 1$ and every $Z\in L^{p}_{\mathbb{P}}$ $$\frac{1}{n}\sum_{k=0}^{n-1}T^{k}Z$$ converges $\mathbb{P}-$a.s. The proof of this theorem can be found for instance in the recent book by Eisner et.al.

Question: Is there a version of the Return Times Theorem for (positive) Dunford-Schwartz operators?

In other words. Can we prove the existence of $\Omega_{1,X}$ as above such that (1) holds in $\Omega_{1,X}\times\Omega_{2}$ if we replace $T_{1}$ and $T_{2}$ by positive Dunford-Schwartz operators in the corresponding $L^{p}-$spaces?

Motivation: (Besides natural curiosity) I want to prove that for a certain Dunford-Schwartz operator $T$ and a given $Z\in L^{2}_{\mathbb{P}}$, the averages $$\frac{1}{n}\sum_{k=0}^{n-1}|T^{k}Z|^{2} \,\,\,\,(2)$$ are $\mathbb{P}-$a.s. convergent. This claim would follow from an affirmative answer to the question by considering $T_{1}=T_{2}=T$ and $Z=X=\bar{Y}$ (the complex conjugate of $Y$).

As a matter of fact, (2) can be proved, using results by Assani, under the assumptions that $T$ and $T^{*}$ preserve the constant function $\mathbf{1}$ (which is actually verified by my particular operator) and that $Z\in L^{p}_{\mathbb{P}}$ for some $p>2$, but I haven't found so far an argument for the case $p=2$ (if you can give one please do so).

Thank you!


1 Answer 1


I've heard the following from Michael Lin: both Koopman operators in the return times theorem can be replaced by Dunford--Schwartz operators, see

Çömez, Doğan; Lin, Michael; Olsen, James, Weighted ergodic theorems for mean ergodic (L_{1})-contractions, Trans. Am. Math. Soc. 350, No. 1, 101-117 (1998). ZBL0888.47006.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.