Are there sets A and B such that any $x \in \mathbb{R}$ can be decomposed as $x = a+b$, $a \in A$ and $b \in B$, where the Lebesgue measure of $A$ and of $B$ is null.

There is an indication that this should follow from the fact that $C_q = \{z \in[0,1]; z_i\in\{0,2,\cdots,q-1\}\}$ has null measure for every $q \in \mathbb{N}$, where $z_i$ are the numbers in the q-adic expansion of $z$, that is $$z=\sum_{i=1}^{\infty}\frac{z_i}{q^i}$$

  • $\begingroup$ This was asked recently on this site. Anyway, let $A=\bigcup_{n\in\mathbb Z}(C+2n) $ and $B=C $, where $C $ is the Cantor set, and use that $C+C=[0,2] $. $\endgroup$ Aug 31, 2017 at 15:09
  • $\begingroup$ Where was it asked recently? Can you please link it? $\endgroup$ Aug 31, 2017 at 16:15
  • $\begingroup$ I finally found the question I had in mind. It is not the same, but one of the answers addresses your question. See here. $\endgroup$ Aug 31, 2017 at 16:43

2 Answers 2


Yes. The following example was given here by Davide Giraudo.

If $A$ is the set of real numbers such that in their proper binary expansion, the even terms are $0$, and $B$ the same with odd numbers, then $A$ and $B$ have measure $0$ but their sum is the whole real line

  • $\begingroup$ I do not understand the definition of sets $A$ and $B$. Can you give an example? $\endgroup$
    – Idonknow
    Aug 31, 2017 at 13:15
  • $\begingroup$ Can you give an indication of the proof of this fact? $\endgroup$ Aug 31, 2017 at 13:38
  • $\begingroup$ For $n>0$ let $D(2n)$ be the set of finite binary sequences of length $2n.$ For $z\in \mathbb Z$ and $d\in D(n)$ let $R(z,d)$ be the set of $x\in [z,z+1)$ that have $d$ as their first $2n$ binary digits to the right of the decimal point. For distinct $d,d'\in D(2n)$ the sets $R(z,d'),R(z,d')$ are disjoint and each is a translate of the other so each has measure $4^{-n}.$ CONTINUED IN NEXT COMMENT. $\endgroup$ Aug 31, 2017 at 17:27
  • $\begingroup$ CONTINUED. Let $E(2n)$ be the set of members of $D(2n)$ with $0$ in all even places. $E(2n)$ has $2^n$ members. Now $A\cap [z,z+1))\subset \cup_{d\in E(2n)}R(z,d)$ and the measure $m(\cup_{d\in E(2n)}R(z,d))=2^n4^{-n}=2^{-n}.$ So $m(A\cap [z,z+1)\leq 2^{-n}$ for all $n\in \mathbb N.$ $\endgroup$ Aug 31, 2017 at 17:35
  • 3
    $\begingroup$ Intuitively the probability that a sequence of $2n$ coin-tosses will come up heads on every even-numbered toss is $2^{-n}$ so the probability that any $x\in [z,z+1) $ belongs to A should be less than $2^{-n}$ for all $n.$..... I am taking $A$ as those reals that have $0$ in every even binary digit to the right of the decimal point. $\endgroup$ Aug 31, 2017 at 17:40

Consider the classical cantor set $C$, then $$C-C = \{x-y\mid x,y\in C\} = [-1,1]$$

  • $\begingroup$ Sure, this is very easy to see. But I want it to form the Real Line, not the unitary interval. $\endgroup$ Aug 31, 2017 at 13:39
  • 2
    $\begingroup$ @FranciscoMaion You can extend $C$ to all of $\mathbb{R}$ by making it periodic with period 1. Then after extension, $C-C = \mathbb{R}$. $\endgroup$
    – pisco
    Aug 31, 2017 at 15:53

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.