Given $X=\{(x,y)\mid x\in \mathbb{Q}, y\in \mathbb{R}\}$. I want to show that $\lambda^*(X)=0$, where $\lambda^*$ is the Lebesgue outer measure.

My attempt:

Let the enumeration of $\mathbb{Q}$ be $\{x_1,x_2,x_3,...\}$. Hence, we have $$ X=\bigcup_{n=1}^{\infty}X_n, $$ where $X_n=\{(x_n,y)\mid y\in \mathbb{R}\}$. We show that for every $n$, we have $\lambda^*(X_n)=0$, so that by the countably subadditivity of Lebesgue outer measure, we have $\lambda^*(X)=0$.
Let $n\geq 1$. Take $\epsilon>0$. For each $k$, consider the open cell $$ I_k=\left(x_n-\frac{\epsilon}{4},x_n+\frac{\epsilon}{4}\right)\times \left(x_k-\frac{1}{2^{k+1}},x_k+\frac{1}{2^{k+1}}\right). $$

If $(x_n,y)\in X_n$, then $y$ is in a small open interval $\big(x_k-\frac{1}{2^{k+1}},x_k+\frac{1}{2^{k+1}}\big)$ for some $k$, so that $(x_n,y)\in I_k$. This means that $X_n \subseteq I_1 \cup I_2 \cup I_3 \cup \ldots$.

Observe that, if $l(I_n)$ is the length of $I_n$, we have $$ \sum_{k=1}^{\infty} l(I_k)=\sum_{k=1}^{\infty}2(\frac{\epsilon}{4})2(\frac{1}{2^{k+1}})=\frac{\epsilon}{2}<\epsilon. $$ Thus, $\lambda^*(X_n)=0$ for every $n$, and hence $\lambda^*(X)=0$.

Is this proof correct? I am not sure about the reasoning on $(x_n,y)$ part. Thank you.

  • $\begingroup$ "...then $y$ is in a small open interval $(x_k-\frac{1}{2^{k+1}},x_k+\frac{1}{2^{k+1}})$ for some $k$..." - This is not true. The sum of the lengths of these intervals is $\sum_{k=1}^{\infty}1/2^k = 1$, so there's no way they can cover all of $\mathbb R$. $\endgroup$ – Bungo Sep 9 at 7:25

$(x_n,y) \in X_n$ does not tell you anything about $y$. It can be any real number. So your assertion that $y \in (x_k-\frac 1 {2^{k+1}},x_k+\frac 1 {2^{k+1}})$ for some $k$ is not true. The union of these intervals does not exhaust the real line because their union has finite measure.

For a correct proof you can use the intervals $(x_n-\frac {\epsilon} {2^{k}},x_n+\frac {\epsilon} {2^{k}}) \times (x_k-1,x_k+1)$.

  • $\begingroup$ For each fixed $n$, your proposed covering has the same positive measure, so taking the union over all $n$ you'll get infinity. The width of the first interval in the product needs to depend on both $n$ and $k$. I think replacing $\epsilon / 2^k$ with $\epsilon / 2^{k+n}$ would do the trick. $\endgroup$ – Bungo Sep 9 at 7:34
  • $\begingroup$ @Bungo OP has already mentioned that he is using sigma subadditivity so it is enough to show that $\lamda^{*}(X_n)=0$ for each fixed $n$. From that point on $n$ is fixed and there is no need to take union over $n$ in the inequalities. $\endgroup$ – Kabo Murphy Sep 9 at 7:38
  • $\begingroup$ OK, fair enough. $\endgroup$ – Bungo Sep 9 at 7:41
  • $\begingroup$ @KaviRamaMurthy I see. Thank you, sir. I think I use the density of the real line in the wrong way. I shouldn't be able to guarantee that any irrational is in such small $I_k$ with rational end points. Correct? $\endgroup$ – 21understanding Sep 9 at 9:01
  • $\begingroup$ Yes. If you take small intervals around all the rationals their union may still be a small set in measure theoretic sense, though this union is dense. $\endgroup$ – Kabo Murphy Sep 9 at 9:03

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.