Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Measure theory guarantees that every Lebesgue finite measurable set $E$ has a closed subset $F$ such that $m(E \backslash F)<\epsilon$ for small $\epsilon$.

But today I saw in some text that for all small $\epsilon >0$ there exist a subset of $\mathbb{I}\cap [0,1]$ (irrationals in [0,1]) that is closed in $\mathbb{R}$ and has Lebesgue measure less than $\epsilon$? Note the set is required to be closed in $\mathbb{R}$, not in $\mathbb{I}\cap [0,1]$. Can someone give me an example / a proof that such thing does not exist?

share|cite|improve this question
What is $\mathbb{I}$? – Chris Eagle Jun 19 '12 at 10:28
@ChrisEagle Irrationals. I have fixed it. – Polymorpher Jun 19 '12 at 10:29
Such a thing does indeed exist. See this answer, for example. – Dejan Govc Jun 19 '12 at 10:44
@DejanGovc Thanks! Can you make it an answer so that people can see this question is answered? – Polymorpher Jun 19 '12 at 11:09
In what text did you see this and what page? – T. Eskin Jun 19 '12 at 11:55
up vote 4 down vote accepted

In fact, we can choose a set which have exactly measure $\varepsilon$.

For a fixed $\delta>0$, consider the set $S_{\delta}:=\bigcup_{n\in\mathbb N}(q_n-2^{-n}\delta,q_n+\delta 2^{-n})$, where $\{q_n,n\in\Bbb N\}$ is an enumeration of the rationals of $[0,1]$. Then $S_{\delta}$ is open and dense in $[0,1]$, since it contains all the rationals of this interval. The maps $f\colon\delta\mapsto \lambda(S_{\delta}\cap (0,1))$ is Lipschitz-continuous. Indeed, if $\delta_1\leq\delta_2$, we have \begin{align*}f(\delta_2)-f(\delta_1)&=\lambda(S_{\delta_2}\setminus S_{\delta_1})\\\ &\leq \lambda\left((0,1)\cap \bigcup_{n=0}^{+\infty}(q_n-2^{-n}\delta_2,q_n+\delta_2 2^{-n})\setminus (q_n-2^{-n}\delta_1,q_n+\delta_1 2^{-n})\right)\\\ &\leq \sum_{n=0}^{+\infty}\lambda((q_n-2^{-n}\delta_2,q_n+\delta_2 2^{-n})\setminus (q_n-2^{-n}\delta_1,q_n+\delta_1 2^{-n}))\\\ &=2(\delta_2-\delta_1)\sum_{n=0}^{+\infty}2^{-n}. \end{align*} Now we use the intermediate value theorem to pick $\delta$ such that $\lambda(S_{\delta}\cap (0,1))=1-\varepsilon$, and we consider the complement in $[0,1]$ of $S_{\delta}$.

share|cite|improve this answer
And with a bit more work we can find a more "constructive" construction of such a set (at least, one that doesn't rely on the intermediate value theorem). Let $U_0 = \emptyset$, and $U_n = U_{n-1} \cup (q_n - r, q_n + r)$ where $r$ is chosen so that $\lambda(U_n) = (1-2^{-n})(1-\epsilon)$. Since $U_{n-1}$ is a finite union of intervals, $\lambda(U_n)$ is a piecewise-linear function, so this can be done explicitly. – Robert Israel Jun 24 '12 at 6:58

Let $\epsilon > 0$. Enumerate the rationals in $[0,1]$ with the sequence $\{q_n\}_{n=1}^\infty$. Now for each $n$, choose an open interval $I_n\subseteq [0,1]$ containing $q_n$ with length less than $\epsilon/2^n$ If $q_n = 0$ or $q_n = 1$, choose a half-open interval. Said half-open intervals are still relatively open in $[0,1]$.

Put $G = \bigcup_n I_n$. The Lebesgue measure of $G$ is at most $\epsilon$. Now put $F = [0,1] - G.$

The set $G$ is open. The set $F$ is closed. And its measure is at least $1 - \epsilon$. Since $G$ contains all of the rationals in $[0,1]$, $F$ contains only irrationals.

share|cite|improve this answer
Sorry I messed up a little bit in my question. What I really meant was for all small $\epsilon$, the subset can be made to have measure less than $\epsilon$. The measure of F being constructed is bounded below by $\epsilon$, not above. – Polymorpher Jun 19 '12 at 11:45

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.