Let $m^*$ be Lebesgue outer measure (also called exterior measure) on $\mathbb{R}^n$. Suppose $E$ is a subset of $\mathbb{R}^n$ with $m^*(E) < \infty$. Let $\mathcal{O}_m$ be the open set: \begin{align*} \mathcal{O}_m &= \left\{ x : d(x, E) < \frac{1}{m} \right\} \\ \end{align*} Part (a): show that if $E$ is compact, then: \begin{align*} m^*(E) &= \lim\limits_{m \to \infty} m^*(\mathcal{O}_m) \\ \end{align*} Part (b): Give examples to show that this property may not hold for cases where $E$ is closed and unbounded or $E$ is open and bounded.

My work:

$\mathcal{O}_m$ is an open ball covering of $E$ with ball radius $\epsilon = 1/m$

The precise definition of Lebesgue outer measure is:

\begin{align*} m^*(E) &= \inf \sum\limits_{j=1}^{\infty} |Q_j| \\ \end{align*}

taken over all countable coverings of closed cubes: $E \subseteq \cup_{j=1}^\infty Q_j$

I suspect I can work out a proof for (a) with the Vitali covering lemma.

I'm mainly looking for help on part (b). I can't think of any irregular open and bounded sets that would break this property. By definition, all open sets are Lebesgue measurable. I also can't think of any closed but unbounded set that would break this property.


Example when $E$ is closed:

Let $E = [1,1+1/2] \cup [2,2+1/4] \cup [3,3+1/8] \cup \cdots = \bigcup_{k=1}^\infty [k,k+2^{-k}]$.

Then $m(E) = 1$, but $m(O_m) = \infty$ for all $m$.

Example when $E$ is open:

Let $\{r_k\}$ denote an enumeration of the rational numbers in $(0,1)$.

Fix $\epsilon > 0$ and for each $k$ choose $\delta_k$ so that the interval $I_k = (r_k - \delta_k,r_k+\delta_k)$ satifies

  • $I_k \subset (0,1)$, and
  • $\ell(I_k) < \dfrac{\epsilon}{2^k}$.

One way to do this is to take $\delta_k = \min\{ \epsilon/2^{k+1}, \mathrm{dist}(x,\{0,1\}))$.

Let $E = \cup_k I_k$. Then $E$ is open and $m(E) \le \sum \ell(I_k) < \epsilon$.

On the other hand, since $\{r_k\} \subset E$ each set $O_m$ contains $[0,1]$ so that $m(O_m) \ge 1$ for all $m$.

  • $\begingroup$ wow, thanks!!!! $\endgroup$ – clay Feb 11 '19 at 22:38

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.