6

I know that the set of points of the form $(e^{2\pi i t},e^{2\pi i a t })$ where $t \in \mathbb{R}$ is dense in the torus $S^1 \times S^1$ when $a \in \mathbb{R}$ is irrational. (This problem is asked here.)

I also know that the set of points of the form $e^{2\pi i a n}$ where $n \in \mathbb{Z}$ is dense in $S^1$, when $a$ is irrational. (This problem is asked here.)

I believe it is true that the set of points of the form $$ (e^{2\pi i an},e^{2\pi i b n }) $$ is also dense in the torus, where $n \in \mathbb{Z}$ and $a,b$ are irrationals such that $a/b$ is also irrational. But I am unable to show this, and would appreciate any hint on how to proceed.


I tried viewing the torus as the unit square $[0,1] \times [0,1]$ with edges identified appropriately. So, I need to show that the set of points $$(an \pmod 1, bn \pmod 1)$$ is dense in the unit square.

In showing that the line $(e^{2\pi i t},e^{2\pi i a t })$ is dense in the torus, it suffices to show that it intersects every ball with centre on the vertical edge $\{ 0 \} \times [0,1]$. However, to apply the same idea here, I need to find an integer $n$ such that $an \pmod 1$ and $bn \pmod 1$ are both arbitrarily close to $0$.

I can find integers $n_1$ and $n_2$ such that $an_1 \pmod 1$ and $bn_2 \pmod 1$ are both arbitrarily close to $0$, by the same argument used to show that $e^{2 \pi i a n}$ is dense in $S^1$. But I'm unable to arrive at a single integer that does the job for $a$ and $b$ simultaneously.

I'm possibly missing only some small idea. Can anyone help me?

  • This is an interesting question. It's not immediately obvious to me that your belief is true. Seems that there conceivably could be a line of irrational slope such that evenly spaced points on the line form some pattern that's not dense? So I guess I'd back up and look for counterexamples. – Neal Nov 03 '18 at 15:48
  • 2
    This is a beautiful question, but any hint is likely to give the whole game away. How about this: what are the closed subgroups of the torus? – kimchi lover Nov 03 '18 at 19:28
  • @kimchilover thank you! That's a lovely hint, I will come back to it first thing in the morning tomorrow. :) –  Nov 03 '18 at 19:36
  • @kimchilover I've posted an answer based on your hint, can you let me know if my approach is what you had in mind? :) –  Nov 05 '18 at 07:31

1 Answers1

4

Based on @kimchilover's hint, the problem can be solved in the following manner.

The closed subgroups of the torus belong to the following isomorphism classes: $S^1 \times S^1$, $S^1 \times \mathbb{Z}_m$, $\mathbb{Z}_n \times S^1$ or $\mathbb{Z}_n \times \mathbb{Z}_m$, for any $n,m \in \mathbb{N}$. Let $$ A = \{ (e^{2\pi i a n},e^{2 \pi i b n}) : n \in \mathbb{Z} \}. $$ Note that the closure of a subgroup is also a subgroup, so $\bar{A}$ must be of one of the above types.

Since $A$ is infinite, $\bar{A} \not\cong \mathbb{Z}_n \times \mathbb{Z}_m$.

Since $a/b$ is irrational, $\bar{A} \not\cong S^1$. So, we have ruled out the cases that $\bar{A}$ is isomorphic to $S^1 \times \{e\}$ or $\{e \} \times S^1$, where $\{ e \} = \mathbb{Z}_1 = \mathbb{Z}/1\mathbb{Z}$.

Let $m > 1$ and suppose that $\varphi : \bar{A} \to S^1 \times \mathbb{Z}_m$ is an isomorphism. Then $\varphi |_A : A \to S^1 \times \mathbb{Z}_m$ is an injective homomorphism. Since $\mathbb{Z}_m$ has the discrete topology, it must be that $\varphi |_A$ is surjective onto the second component $\mathbb{Z}_m$ for $\varphi$ to be an isomorphism. Hence, $A$ must have some torsion, which is a contradiction. We can similarly rule out the cases that $\bar{A}$ is isomorphic to $\mathbb{Z}_n \times S^1$ for any $n > 1$.

Thus, $\bar{A} = S^1 \times S^1$.

  • 2
    Looks good to me. When you posted your question it was hard to gauge how much you know, and hence what sort of hint would help best. My favorite proof of the result would show that the empirical measure of the first $n$ points converges weak * to Lebesgue measure, ruling out the existence of an unvisited ball. – kimchi lover Nov 05 '18 at 12:08
  • @kimchilover I would love to see that proof. Do you have a reference that I can look up? Or if you can type it up then I would be happy to accept that instead of accepting a self-answer. :) –  Nov 05 '18 at 12:19
  • 2
    See https://en.wikipedia.org/wiki/Kronecker%27s_theorem and the links https://en.wikipedia.org/wiki/Equidistributed_sequence#Weyl's_criterion , and so on. The book by Kuipers & Niederreiter is a modern classic reference to this subject. – kimchi lover Nov 05 '18 at 12:33
  • The conditions on $a$ and $b$ are not quite strong enough: rather, it is necessary that $a$ and $b$ are rationally independent, that is, $ma + nb \notin \mathbb{Z}$ for all integers $m$ and $n$, not both $0$. See https://math.stackexchange.com/questions/1266904/torus-translation-is-ergodic-if-and-only-if-the-components-of-the-translation-ve – Mirlan Jan 21 '24 at 01:59