The problem is as follows:

Given a Lebesgue-measurable set $M\subset\mathbb{R^n}$ find a compact $C_\epsilon$ and an open $A_\epsilon$, $C_\epsilon\subset M\subset A_\epsilon$, such that $\lambda(A_\epsilon)-\lambda(C_\epsilon)<\epsilon$.

$\lambda$ is the Lebesgue-measure, in particular, if $M$ is not bounded, $k\in\mathbb{N}$, \begin{equation} \lambda(M)=\lim_{k\to+\infty}\lambda(M\cap[-k,k]^n). \end{equation}

Now, if $M$ is bounded I can find $A_\epsilon$ and $C_\epsilon$ with the required properties.

if $M$ is not bounded but $\lambda(M)<\infty$, how can I find the open and the compact?

Furthermore, if $\lambda(M)=\infty$, I think we can not found a compact such that the difference between its measure and the measure of $M$ is small, is it correct?


If $M$ is unbounded and $\lambda(M)<\infty$, start by taking a $k>0$ such $\lambda(M)-\lambda\bigl(M\cap[-k,k]\bigr)<\frac\varepsilon2$.

If $\lambda(M)=\infty$, then you are right: there is no such compact set.

  • $\begingroup$ thanks for your answer. I thought about two proofs.... $\endgroup$ – Jack J. May 23 '18 at 19:42
  • 1
    $\begingroup$ @JackJ. They look fine. $\endgroup$ – José Carlos Santos May 23 '18 at 21:37

@Josè Carlos Santos: thanks for your answer. I thought about two proofs:

1) For all $k\in\mathbb{N}$, let a compact $C_k\subset M\cap [-k,k]^n$ such that $\lambda(M\cap [-k,k]^n\setminus C_k)<\frac{\epsilon}{2^{k+1}}$.

Since, $\lim_{k\to\infty}\lambda(M\cap[-k,k]^n)=\lambda(M)$, $\exists\bar{k}\in\mathbb{N}$ such that for $k>\bar{k}$ \begin{equation} \lambda(M)-\lambda(M\cap[-k,k]^n)<\frac{\epsilon}{2}. \end{equation}

Now, for $k>\bar{k}$ we consider \begin{equation} \begin{split} \lambda(M)-\lambda(C_k) &= \lambda(M)-\lambda(M\cap[-k,k]^n)+\lambda(M\cap[-k,k]^n)-\lambda(C_k) \\ & = \lambda(M)-\lambda(M\cap[-k,k]^n)+\lambda(M\cap[-k,k]^n\setminus C_k) \\ &< \frac{\epsilon}{2}+\sum_{k=1}^{\infty}\frac{\epsilon}{2^{k+1}}=\epsilon \end{split} \end{equation}

2) Since, $\lim_{k\to\infty}\lambda(M\cap[-k,k]^n)=\lambda(M)$, $\exists\bar{k}\in\mathbb{N}$ such that for $k>\bar{k}$ \begin{equation} \lambda(M)-\lambda(M\cap[-k,k]^n)<\frac{\epsilon}{2}. \end{equation} At this point just take $M\cap[-k,k]^n$, for same $k>\bar{k}$ and to observe that $M\cap[-k,k]^n$ is bounded.


Thanks for the attention!


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.