I wonder if anyone can give me sensible proof.

The Cantor set $C$ is defined by \begin{align*} C_0&=[0,1],\\ C_1&=[0,\tfrac13]\cup [\tfrac23,1],\\ C_2&=[0,\tfrac19]\cup [\tfrac29,\tfrac13]\cup [\tfrac23,\tfrac79]\cup [\tfrac89,1],\\[-6pt] &\ \,\vdots\\[-12pt] C&=\bigcap_{n=0}^\infty C_n. \end{align*}

Observe that $$C=\left\{ x\in [0,1] \quad | \quad x=\sum_{k=1}^\infty\frac{2\epsilon_k}{3^k} \quad \text{where} \quad \epsilon_k \in \{0,1\}^{\mathbb N}\right\}.$$

Consider the ternary dilation $x\longmapsto T(x)=3x \quad \text{mod} \, \, 1 $.

Question: Show that there exists $x_0 \in [0 ; 1)$ with the closure of its forward orbit $\{T^n(x_0) : n \in \mathbb N \}$ are identical to $C$.

\begin{align*} T(x)=& 3x - \lfloor 3x \rfloor \\ T(T(x))= & T(3x - \lfloor 3x \rfloor) \\ = & 3(3x - \lfloor 3x \rfloor) - \lfloor 3(3x - \lfloor 3x \rfloor)\rfloor \\ = & 3^2 x - \lfloor 3^2 x \rfloor\\ \vdots \\ T^n(x) = & 3^n x - \lfloor 3^n x \rfloor. \end{align*}

So, by induction, \begin{align*} \overline{\{T^n(x_0)\}}=& \overline{\bigcup_{k=1}^{n}\{3^n - \lfloor 3^n x\rfloor}\} \\ & = \bigcup_{k=1}^{n}\{\overline{3^n - \lfloor 3^n x\rfloor}\}. \end{align*}

Another way consists to do some calculation from \begin{align*} x=& \sum_{k=1}^\infty\frac{2\epsilon_k}{3^k} & \\ =& 2 \epsilon_1 + \sum_{k=}^\infty\frac{2\epsilon_{k+1}}{3^k} & \\ =& \sum_{k=1}^\infty\frac{2\epsilon_k}{3^k} & \text{mod} \, \, 1 \end{align*} Thus $T^n(x)= \sum_{k=1}^\infty\frac{2\epsilon_{k + n}}{3^k}$, hence $|T^n(x)- x| = \sum_{k=1}^\infty\frac{2(\epsilon_{k+n} - \epsilon_k)}{3^k}\leq \sum_{k=1}^\infty\frac{2}{3^k}=1$ So it isn't clear that the claim is true.

Any help is welcome.


Consider the set $W$ of all finite words in the alphabet $\{0,1\}$. This is a countable set, so we may write $W=\{w_n: n\in \mathbb N\}$.

Now consider the infinite word $w$ obtained by concatenating the $w_n$ one after another.

If the digits of $w$ are used in the ternary expansion of $x$, then the orbit of $x$ under $T$ will be dense.

Notice that the collected works of Shakespeare, if encoded in a binary fashion, will appear as part of $w$, and so will all of the questions and answers in MSE.

EDIT: Let me try to be a bit more specific regarding the choice of the point with a dense orbit.

Take the universal word $w$ constructed above as the concatenation of all finite words $w_n$, and write $$ w=\alpha _1\alpha _2\alpha _3\ldots , $$ with each $\alpha _i$ in $\{0, 1\}$. Defining $$ x_0=\sum_{k=1}^\infty\frac{2\alpha _k}{3^k}, $$ I claim that $\{T^n(x_0):n\in {\bf N}\}$ is dense in the Cantor set $C$. To see this, let $x$ be any point in $C$, and write $$ x=\sum_{k=1}^\infty\frac{2\epsilon _k}{3^k}, $$ with each $\epsilon _i$ in $\{0, 1\}$. Given any $\varepsilon >0$ let us therefore find some $n$ such that $|T^n(x_0)-x|<\varepsilon $.

We first choose $N$ such that the truncated sum $$ x'=\sum_{k=1}^N\frac{2\epsilon _k}{3^k} $$ satisfies $|x-x'|<\varepsilon /2$. By increasing $N$, if necessary, we may suppose that $3^{-N}<\varepsilon /2$.

Next, consider the finite word $\epsilon _1\epsilon _2\epsilon _3\ldots \epsilon _N$. Since all finite words appear in the set $W$ defined above, and since $w$ is the concatenetion of all words in $W$, there must be some $n$ such that $$ \alpha _{n+1}=\epsilon _1, \quad \alpha _{n+2}=\epsilon _2, \quad \cdots , \quad \alpha _{n+N}=\epsilon _N. $$ As noted by the OP, to apply $T$ to any element in the Cantor set has the same effect as shifting its ternary digits. Therefore $$ T^n(x_0) = \frac{2\alpha _{n+1}}{3} + \frac{2\alpha _{n+2}}{3^2} + \cdots + \frac{2\alpha _{n+N}}{3^N} + \sum_{k=N+1}^\infty \frac{2\epsilon _{n+k}}{3^k} = $$$$ = x' + \sum_{k=N+1}^\infty \frac{2\epsilon _{n+k}}{3^k}. $$ This implies that $|T^n(x_0) -x'|<3^{-N}$, so $$ |T^n(x_0) -x|\leq |T^n(x_0) -x'| + |x'-x| < \varepsilon /2+\varepsilon /2=\varepsilon . $$

  • $\begingroup$ Although the information you provide is valuable to me, I believe you have not responded to my question. "Show that there exists $x_0 \in [0; 1)$ for which the closure of its forward orbit is equals $C_3$." $\endgroup$ – Furdzik Dec 31 '20 at 19:51
  • $\begingroup$ I meant to describe the point $x$ whose orbit is dense in the third paragraph of my answer. $\endgroup$ – Ruy Dec 31 '20 at 20:41
  • $\begingroup$ Thanks for your help and support. $\endgroup$ – Furdzik Jan 1 at 1:23
  • $\begingroup$ You are welcome! $\endgroup$ – Ruy Jan 1 at 2:09

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.