Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

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

In a previous post I asked the following question:

Let $\{b_n\}_{n=1}^\infty$ be a dense subset of $\mathbb{R}$ and let $D \subseteq \mathbb{R}$ be a measurable set such that $m(D \triangle (D + b_n))=0$ for all $n \in \mathbb{N}$ (here, the $\triangle$ denotes the symmetric difference of the two sets, $D+ b_n = \{d + b_n : d \in D\}$, and $m$ stands for the Lebesgue measure). Prove that $m(D)=0$ or $m(D^c)=0$ (here $D^c$ is the complement of $D$ in $\mathbb{R}$).

Robert Israel gave hints to one neat strategy for proving the above result. But now I want to see whether anyone visiting can prove the above result without using the Lebesgue density theorem. In particular, a fellow graduate student in my program suggested that the result follows via contradiction by making smart uses of an interesting result I asked about earlier today:

Let $A$ be Lebesgue measurable, with $m(A)>0$ (here $m$ denotes the Lebesgue measure). Then for any $0<\rho<1$, there exists an open interval $I$ such that $m(A \cap I)> \rho \cdot m(I)$.

While I am curious to see alternative strategies for proof, I would be particularly interested to see a proof using this last result, if anyone's game to try!

share|cite|improve this question

The assumption gives that $$\chi_D(x)=\chi_D(x+b_n)\mbox{ for almost every }x\tag{1}$$ (because the integral of the non-negative measurable function $|\chi_D-\chi_{D+b_n}|$ is $0$).

Assume that $\mu(D)\gt 0$. Our goal is to prove that $\mu(D^c)=0$. Fix $\rho\in (0,1)$. There is an interval $I$ such that $\lambda(D\cap I)\gt \rho\lambda(I)$. From (1), we get $\lambda((D+b_n)\cap I)\gt \rho\lambda(I)$, and by density of $\{b_n,n\geqslant 1\}$, that for each $t\in\mathbb R$, $\lambda((D+t)\cap I)\geqslant \rho\lambda(I)$. Taking $N:=\lfloor\frac 1{\lambda(I)}\rfloor$ and using the fact that $\lambda(D\cap (k\lambda(I),(k+1)\lambda(I))\geqslant\rho\lambda(I)$ we get that $\lambda(D\cap (0,1))\geqslant \rho$. Indeed, $$\lambda(D\cap (0,1))\geqslant \lambda\left(\bigsqcup_{k=0}^ND\cap (k\lambda(I),(k+1)\lambda(I)\right)\geqslant (N+1)\rho\lambda(I)\geqslant \rho.$$

As $\rho$ was arbitrary, we obtain that $\lambda(D\cap (0,1))=1$. A similar reasoning gives $\lambda(D\cap (n,n+1))=1$ for all $n\in\mathbb Z$, hence $\lambda(D^c\cap (n,n+1))=0$ for $n$ integer.

share|cite|improve this answer
Almost everything is clear. What I don't understand is "Taking $N:=\lfloor\frac 1{\lambda(I)}\rfloor$ and using the fact that $\lambda(D\cap (k\lambda(I),(k+1)\lambda(I))\geqslant\rho\lambda(I)$". a) you don't use $N$ later. b) from what follows the "using the fact" inequality? Many thanks. – vesszabo Oct 1 '13 at 9:54
$N$ didn't appeared because I didn't detailed enough. I've edited. – Davide Giraudo Oct 1 '13 at 9:58
(This is a comment by a new user @Eduardo, which was put as an answer) Beware with the answer by Davide Giraudo, it is flawed. He is assuming that $(0,1)\supset\bigsqcup_{k=0}^N(k\lambda(I),(k+1)\lambda(I))$, which makes no sense. – John Ma Jun 17 '15 at 10:23
@Eduardo Can you tell me why it does not make sense? Maybe the union should be stopped at $N-1$ instead of $N$, but I think it is correct. – Davide Giraudo Jun 17 '15 at 11:58

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.