0

Given $x \in \mathbb{R}$ and $n \in \mathbb{N}$ there exists $k \in \mathbb{Z}$ such that $\left|2^{n} x-k\right|<1$ (this has been proven on a previous question).

Deduce from this that $\left\{\frac{k}{2^{n}}: k \in \mathbb{Z} \quad n \in N\right\}$ is dense in $\mathbb{R}$

I have tried using the Archimedean property of Real numbers but I do not know where to go from there. I am aware that my final objective is to show that $x, y \in \mathbb{R} \quad x<y$ s.t.$\quad x<\frac{k}{2^{n}}<y$

Erik Dz
  • 51

4 Answers4

2

For the given two numbers $x<y$, let $\epsilon=y-x$.

According to Archimedean property, there exist natural number k such that $\frac{\epsilon}2\cdot k>1$. $2^k>k$ is so trivial.

Then, $2^ky-2^kx>2$, so there exists an natural number $n$ such that $2^kx<n<2^ky$.

So $x<\frac n{2^k}<y$. The proof ends.

  • thank you for your answer, but why did you are Epsilon/2 on the Archimedean property step? – Erik Dz Oct 25 '21 at 19:37
  • @ErikDz I misjudged while writing the answer that an exception would occur when those two numbers $2^kx, 2^ky$ were both integers. In fact, you don't have to do that. I can see all the mistakes after posting them LOL. –  Oct 25 '21 at 19:40
1

HINT: Then given $x \in \mathbb{R}$ and $n \in \mathbb{N}$, there exists a $k \in \mathbb{Z}$ such that $|x-\frac{k}{2^n}| \le 2^{-n}$. So for any $x \in \mathbb{R}$, you can make $n$ arbitrarily large.

So now let $x_1$ and $x_2$ be 2 elements in $\mathbb{R}$ satisfying $x_1 \not = x_2$. It suffices to show that there is an element in the set $\{\frac{k}{2^n}; k \in \mathbb{Z}; n \in \mathbb{N}\}$ between $x_1$ and $x_2$. Now let $x=\frac{x_1+x_2}{2}$. By the above paragraph, for every $n \in \mathbb{N}$, there exists a $k \in \mathbb{Z}$ such that $|x-\frac{k}{2^n}| \le 2^{-n}$. What if you were to use the above paragraph, and make $n$ large enough so that $2^{-n} < \frac{|x_2-x_1|}{4}$? Then you note that there is a $k$ such that $|x-\frac{k}{2^n}| \le \frac{|x_2-x_1|}{4}$, which gives $\frac{k}{2^n}$ must fall between $x_1$ and $x_2$...

Mike
  • 23,484
  • Hi @Mike, thank you for your answer. I have a doubt; why are you dividing $x$ by 2 when you want to compare it to $2^{-n}$? – Erik Dz Oct 25 '21 at 19:49
  • The idea is that for any distinct $x_1,x_2 \in \mathbb{R}$, you want to find an element in the set $A \doteq {\frac{k}{2^n}; k \in \mathbb{Z}; n \in \mathbb{N}}$ that is between $x_1$ and $x_2$. This will happen if you find an element in the set $A$ that is "close enough" to [informally] the halfway point $x$ between $x_1$ and $x_2$. In particular, no more than distance $\frac{|x_2-x_1|}{4}$ away from $x$ will suffice as "close enough". Equivalently, it suffices to show that there is an element in the set $A$ that is no more than $\frac{|x_2-x_1|}{4}$ away from $x$. – Mike Oct 25 '21 at 21:01
  • Note that $x=\frac{x_1+x_2}{2}$ i.e., $x$ is the average between $x_1$ and $x_2$, is halfway between $x_1$ and $x_2$, and the distance from $x$ to $x_1$ is $\frac{|x_2-x_1|}{2}$, and the distance from $x$ to $x_2$ is the same $\frac{|x_2-x_1|}{2}$. So any $a \in \mathbb{R}$ satisfying $|x-a| < \frac{|x_2-x_1|}{4}$ must be between $x_1$ and $x_2$. So if there is an $a \in A \doteq {\frac{k}{2^n}; k \in \mathbb{Z}; n \in \mathbb{N}}$ satisfying $|x-a| < \frac{|x_2-x_1|}{4}$, then this $a$ must be between $x_1$ and $x_2$. – Mike Oct 25 '21 at 21:09
1

You mentioned that it suffices to prove the following proposition I'll call (P): $$\text{(P)}\qquad\text{if $x<y$, there exist $k,n$ such that $x<\frac{k}{2^n}<y$.}$$ If (P) holds, then certainly the dyadic rationals are dense in $\mathbb R$. Here is another property (Q) that implies the dyadic rationals are dense, and follows more immediately from what you have already shown: $$\text{(Q)}\qquad\text{for each $\varepsilon>0$, and each $x\in\mathbb R$, there exist $k,n$ such that $|x-\frac k{2^n}|<\varepsilon$.}$$ Property (Q) says that any interval centered at $x$ contains a dyadic rational, which is a way of stating density of the collection of dyadic rationals that generalizes more easily to other dimensions/spaces. You can prove (Q) immediately from what you already know by using the archimedean property of $\mathbb R$ to find $n$ such that $2^n>1/\varepsilon$.

You can also prove (Q) implies (P) (in fact, they are equivalent). To do this, you can apply (Q) to the midpoint of $x$ and $y$ and an appropriate choice of $\varepsilon$.

Alex Ortiz
  • 26,211
0

An alternative without using the given statement:

Note that $0$ is a limit point of the given set (because the sequence $(x_n)$ where $x_n=\frac 1{2^n}\in$ the given set, converges to $0$).

Let $a<b$ be arbitrarily chosen. Let $\epsilon =b-a$. For large enough $n$, we have $\frac 1{2^n}\lt \epsilon \tag 1$

Define $M:=\{m\in \mathbb Z: \frac m{2^n}\le a\}$. Note that $M$ is non-empty and is bounded above by $\epsilon$ so must have a supremum $p\in \mathbb Z$. Hence, $\frac {p}{2^n}\le a$ and $\frac {p+1}{2^n}\gt a$

Claim: $\frac {p+1}{2^n}\lt b$.

Proof: Suppose on the contrary that, $\frac {p+1}{2^n}\ge b$, it follows that $\frac {p+1}{2^n}-\frac {p}{2^n}\ge b-a\implies \frac 1{2^n}\ge \epsilon$, which contradicts $(1)$. So $\frac{p+1}{2^n}\in (a,b)$. This completes the proof.

Koro
  • 11,766