1

Prove that the set $A$, defined as follows, is dense in $\mathbb{R}$.

$$A = \left\{ \frac{m}{p^n}~|~m \in \mathbb{Z}, n \in \mathbb{N}, p \in \mathbb{N} ~~\text{with} ~~p>1 \right\}. $$

Remark: $A$ is dense in $\mathbb{R}$ if any open interval $(a, b)$ has an element of $A$.

Observation: In the textbook where this problem is from, the convention is that $0 \notin \mathbb{N}$.

Attempt:

Let $b - a = \varepsilon$, with $\varepsilon > 0$. We have to demonstrate the existence of some $\frac{m}{p^n}$ such that $\frac{m}{p^n} < \varepsilon ~~\forall~~ \varepsilon$. So using this inequality, we have

$$ \frac{m}{\varepsilon} < p^n \tag{1} $$

$$ \left( \frac{m}{\varepsilon} \right)^{1/n} < p \tag {2} $$

Now take any $\varepsilon$ arbitrarily small. By setting $p > \left( \frac{m}{\varepsilon} \right)^{1/n}$, we have that $\frac{m}{p^n} < \varepsilon $, and therefore $\frac{m}{p^n} \in (a, b). $

Thoughts:

I feel that what I'm doing might be wrong because I just solved for one variable, $p$ in this case, then back-substituted, but $m$ and $n$ are also variables.

Sigma
  • 1,836
  • 15
  • 35

3 Answers3

1

Fix any $p \ge 2$. Then if the set $A_p=\{\frac{m}{p^n}; n \in \mathbb{N}; m \in \mathbb{Z}\}$ is dense, then so is $A$, as $A_p \subset A$. So now we show that $A_p$ is dense. To this end, it suffices to show the existence of some $m \in \mathbb{Z}$ and $n \in \mathbb{N}$ such that $a < \frac{m}{p^n} < b$. [So this is where I think your proof is convoluted OP, why does $\frac{m}{p^n} < b-a$ imply an element in $A$ i.e., of the form $\frac{m'}{p^{n'}}$ between $a$ and $b$?]

There is indeed the existence of some $m \in \mathbb{Z}$ and $n \in \mathbb{N}$ such that $a < \frac{m}{p^n} < b$. Indeed, let $n \in \mathbb{N}$ be such that $2p^{-n} < (b-a)$. [There exists such an $n$ as $p^n$ goes to infinity as $n$ does, it follows that $p^{-n}$ goes to 0 as $n$ goes to infinity. So for any $b$ and $a$ satisfying $b-a > 0$ there indeed exists an $n$ s.t. $2p^{-n} < (b-a)$.] Next, write $a=A/p^n$ and $b=B/p^n$ for some real $A$ and $B$, where $n$ is picked so that $2p^{-n} < (b-a)$. Then $(b-a)p^n = B-A > 2$. So let $m$ be an integer satisfying $A < m <B$; there is indeed such an integer $m$ because $B-A > 2$, then $\frac{m}{p^n} \in (a,b)$.

Mike
  • 20,434
1

Just because $\frac{m}{p^n}$ is less than $\varepsilon$, that does not mean that $\frac{m}{p^n}$ is inside $(a,b)$. That only tells you that $\frac{m}{p^n}$ is inside $(0,\varepsilon)$ (assuming $m$ positive).

You should choose $p$ and $n$ such that $\frac{1}{p^n}<\varepsilon$. There are many solutions to this part, since any $n$ will give you a $p$ in a way similar to your attempt.

Then you argue that some integer multiple of $\frac{1}{p^n}$ lands inside $(a,b)$. There are many ways to do this, but one is to note that the collection of all $\left(\frac{m-1}{p^n},\frac{m}{p^n}\right]$ cover $\mathbb{R}$. (For clarity, I mean for fixed $p$ and $n$, and letting $m$ vary through $\mathbb{Z}$.) So one interval of the form $\left(\frac{m-1}{p^n},\frac{m}{p^n}\right]$ intersects with $(a,b)$. Since $(a,b)$ is longer, actually at least two adjacent $\left(\frac{m-1}{p^n},\frac{m}{p^n}\right]$ intersect with $(a,b)$, and therefore for one of those two values of $m$, $\frac{m}{p^n}$ is inside $(a,b)$.

2'5 9'2
  • 54,717
  • I'm wondering if I can fix $p = 2$ and still have $A$ dense in $\mathbb{R}$. Then the proof would look simpler and the statement would be more general. – Sigma Sep 29 '19 at 00:05
  • 1
    @VictorS. Yes, $\left{\frac{m}{2^n}\right}$ is also dense in $\mathbb{R}$. In fact it is the set of all numbers whose binary expansion terminates. Take any real number's binary expansion, and you can get something close by truncating. Just truncate far enough down to be within $\varepsilon$ of that number. – 2'5 9'2 Sep 29 '19 at 00:32
1

I think it would be more interesting to fix $p$, because as it is, the question is almost trivial since $A$ contains $\mathbb Q$.

Indeed let have $\dfrac uv\in\mathbb Q$ then

  • if $v=1$ take $m=u,\ n=0,\ p=2$
  • if $v>1$ take $m=u,\ n=1,\ p=v$

On the other hand if you fix $p$ then you can show that $A$ is an additive subgroup of $\mathbb R$.

Let assume $n_1\le n_2\quad$ then $\quad\dfrac{m_1}{p^{n_1}}+\dfrac{m_2}{p^{n_2}}=\dfrac{m_1p^{n_2-n_1}+m_2}{p^{n_2}}=\dfrac{m_3}{p^{n_2}}$.

Now you can use that additive subgroups of $\mathbb R$ are either $\alpha\mathbb Z$ or dense : Subgroup of $\mathbb{R}$ either dense or has a least positive element?

Now since $\dfrac 1{p^n}\in A$ and $\lim\limits_{n\to\infty}\frac 1{p^n}=0$ there is no least positive element, so $A$ is dense.

zwim
  • 28,563
  • $\mathbb{Q}$ is not contained in $A$ because if you fix $n = 1$, you have $m/p$ but $p \geq 2$ – Sigma Sep 29 '19 at 00:11
  • sure, but $n$ is also a free variable. – zwim Sep 29 '19 at 00:13
  • but $n \geq 1 \forall ~ n \in \mathbb{N}$, at least that is the convention of the textbook. So you can't set $n = 0$ and get $1$ in the denominator. – Sigma Sep 29 '19 at 00:14
  • 1
    ok, $\mathbb Q\setminus\mathbb Z$ then, but it changes nothing about density. – zwim Sep 29 '19 at 00:18