8
$\begingroup$

Let $(r_n)_{n \ge 1}$ be an enumeration of the rationals. Consider the union $A := \cup_n (r_n-\frac{1}{n^2},r_n+\frac{1}{n^2})$. It is unclear a-priori whether $A$ covers the real line, since although the rationals are dense in the reals, the $\frac{1}{n^2}$'s might shrink too fast. However, using measure theory, it is very easy to see $A$ does not cover much: indeed, $m(A) \le \sum_n m((r_n-\frac{1}{n^2},r_n+\frac{1}{n^2})) = \sum_n \frac{2}{n^2} = \frac{\pi^2}{3}$.

Since this argument relies much on the convergence of $\sum_n \frac{1}{n^2}$, I am wondering whether $B := \cup_n (r_n-\frac{1}{n},r_n+\frac{1}{n})$ covers the whole real line, or what portion of it? Does the amount covered depend on the enumeration we choose?

$\endgroup$
  • 1
    $\begingroup$ I am quite sure you can find some enumeration of rationals such that $B$ does not cover the whole real line. The interesting question is whether $B$ may cover the real line for some enumeration. $\endgroup$ – Crostul Dec 6 '16 at 8:34
  • $\begingroup$ Yea, you could probably take an enumeration where for every $n$ not a perfect square, $r_n \in [0,1]$ and for every $n = m^2$, $r_n \not \in [0,1]$, so that we could apply the first argument to outside $[0,1]$. $\endgroup$ – mathworker21 Dec 6 '16 at 8:47
  • $\begingroup$ You can enumerate $\Bbb Q$ in such a way that $ B=\Bbb R.$ You can also enumerate $\Bbb Q$ in such a way that $ B$ has finite measure $\endgroup$ – DanielWainfleet Nov 4 '18 at 8:32
  • $\begingroup$ @DanielWainfleet How do you know you can enumerate $\mathbb{Q}$ in such a way that $B = \mathbb{R}$? $\endgroup$ – mathworker21 Nov 4 '18 at 8:33
  • $\begingroup$ @mathworker21. I have posted an answer to that part of the Q. $\endgroup$ – DanielWainfleet Nov 4 '18 at 10:39
1
$\begingroup$

Let $q_3\in \Bbb Q \cap (5/6,1).$

Recursively, for $j\in \Bbb Z^+$ let $q_{3(j+1)} \in \Bbb Q \cap (q_{3j}+\frac {1}{3j}-\frac {1}{6(j+1)}, q_{3j}+\frac {1}{3j}).$

Then $0<q_{3i}< q_{3j}$ when $i<j,$ and we have $\cup_{j=1}^n(-\frac {1}{3j}+q_{3j},\frac {1}{3j}+q_{3j}\supset [1,1+ \sum_{j=1}^n\frac {1}{6j}).$

So $\cup_{j\in \Bbb Z^+}(-\frac {1}{3j}+q_{3j},\frac {1}{3j}+q_{3j})\supset [1,\infty).$

Similarly we can find a discrete $\{q_{3j-1}:j\in \Bbb Z^+\}\subset \Bbb Q$ such that $q_{3j+2}<q_{3j-1}<0$ and $\cup_{j\in \Bbb Z^+}(-\frac {1}{3j-1}+q_{3j-1},\frac {1}{3j-1}+q_{3j-1})\supset (-\infty,-1].$

Let $q_1=0.$

Since the set $S=\{q_1\}\cup \{q_{3j}:j\in \Bbb Z^+\}\cup \{q_{3j-1}:j\in \Bbb Z^+\}$ is discrete, the set $\Bbb Q$ \ $S$ is infinite so we can enumerate $\Bbb Q$ \ $S=\{q_{3j+1}:j\in \Bbb Z^+\}.$

And we have $\cup_{j\in \Bbb Z^+}(-1/j+q_j,1/j+q_j)=\Bbb R.$

We can also enumerate $\Bbb Q$ in a different way, to make $\cup_{n\in \Bbb N}(-1/n+q_n,1/n+q_n)$ a set of finite measure.

$\endgroup$
  • $\begingroup$ Sorry for late response. I don't think you necessarily enumerated all of the rationals. For example, we can have $C = \{1\}, A = \{2,4,6,8,\dots\}, B = \{3,5,7,9,\dots\}$ in which case you can't get the rationals not covered by $A$ or $B$ by rationals coming from $C$. How would you fix this? I don't think you can just throw in the rest of the rationals wherever you please $\endgroup$ – mathworker21 Mar 3 at 17:51
  • $\begingroup$ I will look into this shortly. $\endgroup$ – DanielWainfleet Mar 4 at 10:38
  • $\begingroup$ time is not this short $\endgroup$ – mathworker21 Mar 25 at 7:03
0
$\begingroup$

Okay, so let's first cover the case, that $r_n$ does not have to enumerate all the rational numbers. Then you can make an enumeration of the rationals between (0,1) and B will obviously not cover the reals.

Now you can replace the elements $r_{n^2} $ with an enumeration of all the rationals which don't cover the reals because of your first argument. You will have a few duplicates in your sequence though. But you can fix that by skipping all the rationals in (0,1) which are not enumerated by a squared natural number.

This argument also works for balls with radius $\frac{1}{log(n)}$ or worse. You just replace the $exp(n^2)$ elements instead

For a sequence that covers the reals, take $r_n=\frac{1}{2n}$ then you will cover all the positive reals. Halving/quartering the distance frees up subsequences which you can use to cover the negative reals or all the rationals if you want.

(Okay I will never again type an answer on a phone...)

$\endgroup$

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.