Here is Prob. 17 in the Exercises after Chapter 2 in the book Principles of Mathematical Analysis by Walter Rudin, 3rd edition.

Let $E$ be the set of all $x \in [0,1]$ whose decimal expansion contains only the digits $4$ and $7$. Is $E$ countable? Is $E$ dense in $[0,1]$? Is $E$ compact? Is $E$ perfect?

My effort:

The set $E$ is not countable. The proof is essentially the same as that for showing that the set of all the binary sequences is uncountable. Am I right?

The set $E$ is not dense in $[0, 1]$. The smallest element in $E$ is $$x_\min \colon= \frac{4}{10} + \frac{4}{10^2} + \frac{4}{10^3} + \ldots = \frac{4}{9},$$ and the largest element in $E$ is $$x_\max \colon= \frac{7}{10} + \frac{7}{10^2} + \frac{7}{10^3} + \ldots = \frac{7}{9}.$$ Thus, the set $E$ is (strictly) contained in the closed interval $[\frac{4}{9}, \frac{7}{9}]$. So, the element $\frac{1}{5}$ of $[0,1]$, for example, does not lie in the closure of $E$. Am I right?

The set $E$ is clearly bounded. So, for compactness, it suffices to show that $E$ is closed. [It does not matter if $E$ is closed in $[0,1]$ or $\mathbb{R}$, as the former is a closed set in the latter.] So we show that the complement of $E$ in $[0,1]$ is open in $[0,1]$.

Let $x \colon= \sum_{n=1}^\infty \frac{d_n}{10^n}$ be an arbitrary element of $[0,1]-E$, where each $d_n \in \{ 0, 1, 2, \ldots, 9 \}$. Then there is a positive integer $n$ such that $d_n \not\in \{4, 7\}$. Let $N$ be the least such positive integer, and let $\delta$ be a real number such that $$0< \delta < \frac{\min\left( \vert d_N - 4 \vert, \vert d_N - 7 \vert \right)}{10^{N+2}}. $$ Let $y \colon= \sum_{n=1}^\infty \frac{e_n}{10^n}$ be an element of $E$, where each $e_n$ is either $4$ or $7$. Let's also assume that $e_n = d_n$ for all $n \in \{1, \ldots, N-1\}$. What next? How to show that $y$ fails to be within $\delta$ of $x$?

For showing that $E$ is perfect, we need to show that $E$ is closed and that each element of $E$ is a limit point of $E$.

Let $x \colon= \sum_{n=1}^\infty \frac{d_n}{10^n}$ be an arbitrary element of $E$, where each $d_n$ is either $4$ or $7$. Let $\delta > 0$. Then there exists a smallest positive integer $N$ such that $$\frac{3}{10^N} < \delta.$$

Let $y \colon= \sum_{n=1}^\infty \frac{d_n^\prime}{10^n}$, where each $d_n^\prime$ is either $4$ or $7$, be the element of $E$ such that $$d_n^\prime = \begin{cases} d_n \ \mbox{ if } \ n \in \mathbb{N} \ \mbox{ and } n \neq N; \\ 4 \ \mbox{ if } \ n = N \ \mbox{ and } d_N = 7; \\ 7 \ \mbox{ if } \ n = N \ \mbox{ and } d_N = 4. \end{cases} $$ Then $$0< \vert x -y \vert < \delta.$$ This shows that each element $x$ of $E$ is also a limit point of $E$. Am I right?

  • $\begingroup$ Very nice work. +1. Only one thing: you define $\;N\;$ as "the least such positive integer". What if only there are zeros? $\endgroup$
    – DonAntonio
    Mar 27, 2016 at 8:17
  • $\begingroup$ @Joanpemo I didn't get your comment. So can you please elaborate? $\endgroup$ Apr 3, 2016 at 7:01

3 Answers 3


Your arguments for uncountable and not dense look good to me. As far as showing $[0,1]-E$ is open, I think for simplicity since you know that $d_N\not\in\{4,7\}$, and we're dealing with integers, take $$\delta<\frac{1}{10^{N+2}}.$$ Then if $y$ is such that $|x-y|<\delta$ you know that $y$ has to agree with $x$ at $d_N$, and thus $y\not \in E$.

I believe that your argument for $E$ being perfect is correct as well :)

  • $\begingroup$ Sir please can you explain, "if $y$ such that $|x-y|<\delta$ then $y$ has to agree with $x$ at $d_N$ why? and then how $y$ not in $E$ $\endgroup$ Sep 8, 2019 at 5:20

As you mentioned, for any number in [0,1] - E there is some decimal place which is not 4 or 7. The closest number in E would have a 4 or 7 in that place. Maybe this is not mathematical enough, but could I say that if the two numbers differ at decimal place N, and the next decimal place for one of the numbers is either 4 or 7, then the difference between the 2 numbers must be at least $2/10^{n+1}$? Therefore the set of numbers $\notin E$ is open.

  • $\begingroup$ I can think of the following argument. Assume without loss of generality that $x > y$. Then $d_N \geq e_N$. So, $$\left\vert x - y \right\vert = x - y = \frac{ d_N - e_N}{10^N} + \sum_{n=N+1}^\infty \frac{d_n - e_n }{10^n}.$$ What next? $\endgroup$ Apr 3, 2016 at 7:15

Concerning the compact argument :

With $x := \sum \frac{d_{n}}{10^n}$, let's say that $d_{N} \notin \left\{4, 7\right\}$. Thus $x \in B(x, 10^{-N}) \subseteq \mathbb{R} - E$, where $B(x, r)$ define the open ball centered in $x$ of radius $r$.

Since $d_{N} \notin \left\{4, 7\right\}$, for all $$y = \sum_{n=1}^{n=N} \frac{d_{n}}{10^{n}} + \sum_{n>N} \frac{d'_{n}}{10^{n}} \ \ \ (*)$$ We have got : $$ \left| x-y \right| \leqslant 10^{-N} $$

And (almost) reciprocaly, for all $y$ such that $ \left| x-y \right| < 10^{-N} $, it can be written in the $(*)$ way. Then all such $y$ are in $E^{C}$ (since $d_{N} \notin \left\{4,7\right\}$).

Then, by the open ball characterization of open set in metric spaces, we can conclude.

It seems sufficient to show that $E^{C}$ is open in $\mathbb{R}$.

The rest of your solution looks good.

N.B :
- We may have had some problem for $x = 0 \in [0,1]$ and $x = 1 \in [0,1]$, so it's easier to take $x \in \mathbb{R}$ and thus to show that $\mathbb{R} - E$ is open.
- We have got $\left| x-y \right| \leqslant 10^{-N}$ and then $\left| x-y \right| < 10^{-N}$ because of $x$ and $y$ like $x=0,19999...$ and $y=0,10000...$

  • $\begingroup$ can you please explain in a bit of detail how $B(x, 10^{-N} ) \subseteq [0,1]-E$? Suppose that $y = \sum_{n=1}^\infty \frac{e_n}{10^n}$ is in this ball. Then ...? $\endgroup$ Apr 10, 2016 at 12:57
  • $\begingroup$ I added some precisions on the post. $\endgroup$
    – Johan
    Apr 10, 2016 at 18:16

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.