4
$\begingroup$

I suppose that what I look for is known, but I can't find it.

Let $\left\lbrace I_n=[a_n,b_n)\right\rbrace$ and $\left\lbrace J_n=[b_n,c]\right\rbrace$ ($n\in\mathbb{N}$) be two countable families of intervals in the unit circle $S^1$. Notice that $I_n$ and $J_n$ are adjacent for every $n$, and that the extremum $c$ is fixed. Assume that the total length of every pair goes to 0, that is $$\lim_{n\to\infty}|c-a_n|= 0.$$ Assume also that the length of the "left" interval is a higher order infinitesimal than the length of the "right" one, that is $$\lim_{n\to\infty}\frac{|b_n-a_n|}{|c-b_n|}= 0.$$

Now let $R:S^1\to S^1$ be an irrational rotation. I say that $x\in S^1$ ultimately first visits $J_n$ if, for every sufficiently large $n$, $$\min\left\lbrace k:R^k(x)\in J_n\right\rbrace<\min\left\lbrace k:R^k(x)\in I_n\right\rbrace.$$ Now my question: is it possible to take the families $\{I_n\}$ and $\{J_n\}$ such that every point in $S^1$ ultimately first visits $J_n$?

$\endgroup$

1 Answer 1

2
$\begingroup$

No, it is not possible. In the following I will use $I_n=(a_n,b_n)$ instead of $[a_n,b_n)$ (this is not a problem, you can just increase $a_n$ a bit so that the statement with $I_n=(a_n,b_n)$ is stronger).

For fixed $n$, we say that $x$ first visits $J_n$ if $\min\left\lbrace k:R^k(x)\in J_n\right\rbrace<\min\left\lbrace k:R^k(x)\in I_n\right\rbrace$. So the set of points that first visit $J_n$ is $A_n:=\bigcup_{m=1}^\infty\left(R^{-m}J_n\setminus\bigcup_{k=1}^{m-1}R^{-k}I_n\right)$.

Then the set of points that ultimately first visit $J_n$ is $X:=\bigcup_{N=1}^\infty\bigcap_{n=N}^\infty A_n$. What we want is $X=\mathbb{S}^1$, or equivalently, $\varnothing=\mathbb{S}^1\setminus X=\bigcap_{N=1}^\infty\bigcup_{n=N}^\infty(\mathbb{S}^1\setminus A_n)$. Moreover, $\mathbb{S}^1\setminus A_n$ is the set of points that first visit $I_n$, that is, $B_n:=\bigcup_{m=1}^\infty\left(R^{-m}I_n\setminus\bigcup_{k=1}^{m-1}R^{-k}J_n\right)$.

So, we want $\bigcap_{N=1}^\infty\bigcup_{n=N}^\infty B_n=\varnothing$. However, for each $N$, the set $\bigcup_{n=N}^\infty B_n$ is open and dense: it is open because $B_n$ is open for all $n$. To check that it is dense, we will prove that for every $\varepsilon>0$ there is some $n$ such that $B_n$ is $\varepsilon$-dense in $\mathbb{S}^1$. Indeed, let $z$ be the complex number such that $R(x)=zx$ for all $x$. Now let $k$ be such that $\{z,z^2,\dots,z^{k}\}$ is $\varepsilon$-dense in $\mathbb{S}^1$, and let $n$ be so big that the distance between any two points of $\{z,z^2,\dots,z^{k}\}$ is bigger than $|c-a_n|$. This implies that $R^{-a}I_n$ does not intersect $R^{-b}J_n$ if $a,b<k$ (this is obvious if $a=b$, and if not we use that $d(z^{-a},z^{-b})>|c-a_n|$, so as $I_n$ and $J_n$ are contained in an interval of length $|c-a_n|$, $z^{-a}I_n$ doesn't intersect $z^{-b}J_n$). So $B_n$ contains $\bigcup_{m=1}^k\left(R^{-m}I_n\setminus\bigcup_{k=1}^{m-1}R^{-k}J_n\right)=\bigcup_{m=1}^kR^{-m}I_n$, which is $\varepsilon$-dense in $\mathbb{S}^1$ as we wanted.

So as $\bigcup_{n=N}^\infty B_n$ is open and dense for all $N$, by Baire's theorem $\bigcap_{N=1}^\infty\bigcup_{n=N}^\infty B_n$ is nonempty.

$\endgroup$
16
  • $\begingroup$ Still didn't have time to look closely, but two things: 1. $B_n:=A_n$ must be just $B_n:=$... right? 2. the conclusion is a bit suspicious, as by your reasoning $\bigcap_{N=1}^\infty\bigcup_{n=N}^\infty B_n$ is not only nonempty, but even comeager. That's strange, no? $\endgroup$ Commented Nov 25, 2022 at 18:07
  • $\begingroup$ You are right about the typo. I guess it's strange? Honestly I didn't think about it that much, since for example meager sets can have full measure $\endgroup$
    – Saúl RM
    Commented Nov 25, 2022 at 18:12
  • $\begingroup$ Yeah but why the set of points that first visit the small one should be topologically larger than the set of those visiting first the large one? Anyway, I'll check it later, thanks. $\endgroup$ Commented Nov 25, 2022 at 18:19
  • $\begingroup$ I wouldn't describe it as "larger". I haven't even used that the $I_n$ are smaller than the $J_n$ when $n$ goes to infinity, in fact you can change $I_n$ to be the interior of $J_n$ and change $J_n$ be the closure of $I_n$ and the result still holds $\endgroup$
    – Saúl RM
    Commented Nov 25, 2022 at 18:26
  • 2
    $\begingroup$ Got it! I am interested in the non disjoint case because there I think it's purely a measure theoretic argument. Will try to write details. $\endgroup$ Commented Nov 25, 2022 at 20:02

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .