Confusion over the metric in the product of two topological groups

40 Views Asked by At

$\newcommand{\m}{\,\operatorname{mod}}$I cite this text, page $38$ - although they reference results from earlier on in the text:

Some definitions:

A topological dynamical system $(K;\phi)$ is a non-empty compact Hausdorff topological space $K$ equipped with a map $\phi:K\to K$ which is continuous in $K$'s topology.

Let $K=[0,1)$ be endowed with the metric $d(x,y)=|e^{2\pi i\cdot x}-e^{2\pi i\cdot y}|=4\sin^2\left(\frac{x-y}{2}\right)$; then $K$ is a compact Hausdorff metric space. For an $x\in\Bbb R$ we write $x\m1=x-\lfloor x\rfloor$. Given $\alpha\in K$, let $\phi(x)=x+\alpha\m1$ be the (continuous) dynamic, and let this topological dynamical system be denoted by $([0,1);\alpha)$.

Let $\Bbb T=\{z\in\Bbb C:|z|=1\}$, and take $a\in\Bbb T$, defining $\phi(z)=a\cdot z$ to obtain an invertible dynamic system, denoted $(\Bbb T;a)$.

$([0,1);\alpha)$ is isomorphic to $(\Bbb T;a)$ when $a=e^{2\pi i\cdot\alpha}$ under the isomorphism $\psi:[0,1)\to\Bbb T,\,x\mapsto e^{2\pi i\cdot x}$.

A group extension of a dynamic system $(K;\phi)$ by a compact topological group $G$, along $\Psi:K\to G$ continuous, is the product topological space $K\times G$ endowed with the dynamic $\psi(x,g)=(\phi(x),\Psi(x)\cdot g)$.

A point $x\in K$ is said to be recurrent if for every open neighbourhood $U$ of $x$, there is $m\in\Bbb N_1$ such that $\phi^m(x)\in U$.

Theorem proved in the text:

Let $(K;\phi)$ be a topological system, $G$ a compact group and $(H;\psi)$ the group extension of $K$ by $G$ along some $\Psi$. If a point $x_0\in K$ is recurrent, then $(x_0,g)$ is recurrent in $H$ for all $g\in G$.

They use this to show that:

Let $\alpha\in\Bbb R,\epsilon\gt0$ be given. Then $\exists n\in\Bbb N_1,m\in\Bbb Z$ such that: $$|n^2\alpha-m|\lt\epsilon$$

Their proof goes along the lines of:

By isomorphy with $(\Bbb T;a)$, the point $x=0$ is recurrent in $([0,1);\alpha)$ (proof is discussed earlier in the text, proposition $3.12$). Consider the group extension of this system, as a group with addition modulo $1$, with itself, along $\Psi:[0,1)\to[0,1),\,x\mapsto2x+\alpha\m1$, giving $H=[0,1)\times[0,1),\,\psi(x,y)=(x+\alpha,y+2x+\alpha)\m1$ as the extended system. As $0$ is recurrent in the original system, $(0,g)$ is recurrent in $H$ for any $g\in[0,1)$ - in particular, $(0,0)$ is recurrent in $(H;\psi)$. Consider the orbit of $(0,0)$ - by induction: $$\psi^n(0,0)=(n\alpha,n^2\alpha)$$

This following part is the part I don't understand:

The recurrence of $(0,0)$ implies that for any $\epsilon\gt0$, there exists $n\in\Bbb N_1$ with $d(0,n^2\alpha-\lfloor n^2\alpha\rfloor)\lt\epsilon\quad\blacksquare$

How exactly am I supposed to understand $d$ here? Recurrence of this point implies that if I pick an open neighbourhood, which in the product metric may be chosen as a ball $B=\{(x,y)\in[0,1)\times[0,1):d((x,y),(0,0))\lt\epsilon\}$, i.e. $\sqrt{d(x,0)^2+d(y,0)^2}\lt\epsilon$ since I am inferring a product metric here. The recurrence of $(0,0)$ implies there is $n$ so that $d((n\alpha,n^2\alpha)\m1,(0,0))\lt\epsilon$, or: $$\sqrt{d(n\alpha-\lfloor n\alpha\rfloor,0)^2+d(n^2\alpha-\lfloor n^2\alpha\rfloor,0)^2}\lt\epsilon$$

Where I am lazily implying by $d$ either the product metric or the metric given at the start, $d(x,y)=|e^{2\pi i\cdot x}-e^{2\pi i\cdot y}|$. The text tells me that this metric is continuous w.r.t. the usual metric on $\Bbb R$, so $d(\cdots)\lt\epsilon\implies|\cdots|\lt\epsilon$, I think.

I can't understand how their conclusion follows. What am I missing? I've been quite confused by this text so far, as someone who doesn't really study topology.

1

There are 1 best solutions below

0
On

The answer (I think) is that the product topology on metric spaces may be induced by $d_p=\max\{d_1,d_2\}$, and so $d_p\lt\epsilon\implies d_1,d_2\lt\epsilon$, and therefore we have two results:

$$|n\alpha-\lfloor n\alpha\rfloor|\lt\epsilon,|n^2\alpha-\lfloor n^2\alpha\rfloor|\lt\epsilon$$

For some $n\in\Bbb N_1$.