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

Prompted by this question I was looking for $A \subset (0,1)$ such that for any interval $(a,b)\subset (0,1), A \cap (a,b)$ and $A^c \cap (a,b)$ are both uncountable. One such $A$ is the set of all numbers that have a finite number of $1$'s in their base $3$ expansion. As no choice was used in the construction, it should be Lebesgue measurable, but I can't prove it. How is it proved?

share|cite|improve this question
up vote 5 down vote accepted

For each $n$, let $A_n$ be the subset of elements of $A$ that have at least a $1$ in the first $n$ digits of their ternary expansion, but no $1$s after the $n$th digit. Your set $A$ is equal to the union of the $A_n$s and the Cantor set. Each $A_n$ is a finite union of scaled translates of the Cantor set (take $3^{-n}C$, where $C$ is the Cantor set, to get all numbers that have ternary expansion with no $1$s and that have $0$ in the first $n$ positions; then you can "translate" by adding a suitable number that has a tail of $0$s and appropriate $1$s in the appropriate coordinates).

So each $A_n$ is a finite union of Lebesgue-measurable sets (scaled Cantor sets are Lebesgue-measurable, and translates of a Lebesgue-measurable set are Lebesgue-measurable), hence Lebesgue measurable. $A$ is a countable union of Lebesgue measurable sets, hence Lebesgue measurable.

Added. As Andres Caicedo points out in the comments below, the argument above shows that the set $A$ is in fact not merely Lebesgue measurable, but Borel.

share|cite|improve this answer
This would prove that $A$ is null, as we might expect. – Ross Millikan Apr 4 '11 at 0:54
Indeed it would. – Arturo Magidin Apr 4 '11 at 0:58
@Ross, note that the description Arturo gives shows that, in fact, $A$ is Borel (this is much stronger than simply being Lebesgue measurable). Just about any set that you explicitly describe can be easily shown Borel by unraveling the given description: In this case, if only finitely many 1s are used, then your set is a countable union (depending on where the last 1 appears), etc. – Andrés E. Caicedo Apr 4 '11 at 1:39
@Andres: Thanks. Good point. – Ross Millikan Apr 4 '11 at 2:02

The number $x_n$ in the $n^\text{th}$ place after the radix point in the ternary expansion of $x$ is a measurable function of $x$. This is true because the floor function is Borel measurable, and $x_n=\left\lfloor 3\cdot\left(3^{n-1}x-\lfloor3^{n-1}x\rfloor\right)\right\rfloor$. The set $\{x:x_n\neq 1\}$ is measurable for each $n$, and so therefore is the set $$\bigcup_{n=1}^\infty\bigcap_{k=n}^\infty \{x:x_k\neq 1\}.$$

Certain null sets of the second category give further examples of this. The complement of such a set can be contructed by taking a countable union of closed sets with empty interior whose complements have progressively smaller measure (e.g., using fat Cantor sets).

share|cite|improve this answer
Following on Andres' comment in Arturo's answer, you have also proved the set Borel, right? I only see countably many of the allowable operations here. – Ross Millikan Apr 4 '11 at 2:05
Yes, the functions involved are Borel measurable (floor is Borel, continuous functions are Borel, and compositions of Borel functions are Borel), and the inverse image of the open set $\mathbb{R}\setminus\{1\}$ under a Borel map is Borel. The Borel sets form a $\sigma$-algebra, so the countable union of countable intersections of Borel sets is Borel. (The set is even an $F_\sigma$, but that isn't as clear in my answer as in Arturo's.) The construction of null sets of the second category I mentioned also yields Borel sets ($G_\delta$s, even). – Jonas Meyer Apr 4 '11 at 2:12

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.