6
$\begingroup$

Let's view $\mathbb{R}$ as a vector space over $\mathbb{Q}$, and pick some basis $(v_\alpha)_{0 \leq \alpha < \mathfrak{c}}$ of it. We can then consider the subspace $L$ spanned by $(v_\alpha)_{0 < \alpha < \mathfrak{c}}$ (ie leaving out one vector from the basis). Given the horrible way we have built $L$, I don't suppose there is much a priory reason for $L$ to be measurable. However, I am wondering whether we can say something about the outer measure of $L$.

$\endgroup$
4
  • 1
    $\begingroup$ This is the Vitali set, which is not Lebesgue measurable, and in fact has infinite outer measure and zero inner measure $\endgroup$ Apr 8, 2019 at 15:54
  • 1
    $\begingroup$ Vitali sets can be found in arbitrarily small intervals, if my memory serves me right. $\endgroup$
    – Asaf Karagila
    Apr 8, 2019 at 16:02
  • $\begingroup$ @AsafKaragila Correct; in fact a non-Lebesgue measurable set can be found even in any Lebesgue set of positive measure. $\endgroup$ Apr 8, 2019 at 16:05
  • $\begingroup$ @Pietro: Just to clarify on your comment which could be read to imply somehow that Vitali sets are the only examples of non-measurable sets, which is of course not true. We have many different examples of non-measurable sets! $\endgroup$
    – Asaf Karagila
    Apr 8, 2019 at 17:15

1 Answer 1

8
$\begingroup$

This is the (a version of the) Vitali set, which is not Lebesgue measurable. A quick reason is:

$\mathbb{R}= (v_0\mathbb{Q})\oplus_\mathbb{Q} V$ shows that $\mathbb{R}$ is a countable union of translates of $V$, so $V$ cannot be a Lebesgue set of measure zero, but must have positive outer measure. To show that the outer measure of $V$ is actually $+\infty$, note that, being $V$ a $\mathbb{Q}$-linear subspace, $2V=V$ so that its Lebesgue outer measure is $\lambda^*(V)=2\lambda^*(V)$, which has to be $+\infty$ because it is not $0$.

On the other hand, for any Lebesgue measurable set of positive measure $S$, according to Steinhaus property, $S-S$ is a nbd of $0$. Therefore $V=V-V$, which is nowhere dense, contains no measurable set of positive measure.

$\endgroup$
8
  • $\begingroup$ It's not a Vitali set, since it is closed under sums and products by a rational scalar, and a Vitali set is not (a priori) closed under such sum and products. Moreover, Vitali sets can be taken from arbitrarily small intervals so they can have arbitrarily small outer measure. $\endgroup$
    – Asaf Karagila
    Apr 8, 2019 at 16:05
  • $\begingroup$ That's depend on your definition of "Vitali set"; of course there are non measurable sets even into any measurable set with positive measure. A standard way to exhibit a non-Lebesgue measurable set is, a rational hyperplane of $\mathbb{R}$, which is essentially the original quotient construction. $\endgroup$ Apr 8, 2019 at 16:18
  • 2
    $\begingroup$ Nobody imposed a set of representatives to be bounded. I am simply claiming that there is a bounded one. And I'd be happy to see any reference to Vitali set meaning anything other than the set of representatives. $\endgroup$
    – Asaf Karagila
    Apr 8, 2019 at 16:42
  • 4
    $\begingroup$ @AsafKaragila If the omitted basis vector, called $v_0$ in the problem, happens to be a rational number, then the subspace spanned by the other vectors is a set of representatives of $\mathcal R/\mathcal Q$, a Vitali set. If the omitted basis vector is something else, then you get a set of representatives of $\mathcal R/(v_0\mathcal Q)$, which I think still qualifies as a version of the Vitali set. Shrink it by a factor $v_0$ and it becomes a genuine Vitali set again. $\endgroup$ Apr 8, 2019 at 17:12
  • 1
    $\begingroup$ @Skeeve since you made me think, I got a simpler reason (added) $\endgroup$ Apr 8, 2019 at 18:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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