Let $(U_n)_n$ be an arbitrary sequence of open subsets of the unit disk $D(0,1)\subseteq \mathbb{R}^2$ s.t. $\sum_{n=0}^\infty \lambda(U_n)=\infty$ (where $\lambda$ is the Lebesgue measure). Does there exist a sequence $(q_n)_n$ in $\mathbb{R}^2$ s.t. $D(0,1) \subseteq \bigcup_{n=0}^\infty (q_n+U_n)$?

With the notation $q_n+U_n$, I mean $$q_n+U_n:=\{x\in \mathbb{R}^2|x-q_n\in U_n\}$$

EDIT: Fedor Petrov was quick to find an easy answer to this one and I'm forced to accept it. His method doesn't hold up though if I additionally demand that all the $U_n$ are convex. So, submissions with a take on such a related question are still welcome (although I'll not be able to reward your submission with a "accepted answer" badge)

UPDATE: Acting on popular request, I've reposted the revised question over here.

  • 3
    $\begingroup$ It looks reasonable to ask a separate question for convex sets. $\endgroup$ Nov 10, 2019 at 8:30
  • 1
    $\begingroup$ By Fritz John's theorem, the question involving convex sets is the same as requiring that all of the $U_n$ be open ellipses (or rectangular boxes - not necessarily coordinate aligned) $\endgroup$ Nov 10, 2019 at 14:04
  • $\begingroup$ . . . and then you can assume for each $\epsilon \gt 0$ that they're aligned within $\epsilon$ if that helps. $\endgroup$ Nov 10, 2019 at 15:40
  • $\begingroup$ @FedorPetrov: there is now another question circulating mathoverflow.net/questions/345706/… $\endgroup$
    – 5th decile
    Nov 10, 2019 at 18:17

1 Answer 1


No even in dimension 1 (and multiplying the example for $\mathbb{R}$ by the small segment you get a counterexample in $\mathbb{R}^2$).

Take the set $A_n\subset \mathbb{R}$ defined as $\bigcup_{k\in \mathbb{Z}} (2k\cdot 10^{-n},(2k+1)\cdot 10^{-n})$. I claim that there exists no finite family of translates $\bigcup_{n=1}^\infty (A_n+q_n)$ which covers $[0,1]$. Indeed, we may recursively find a nested family segments $[0,1]\supset \Delta_1\supset \Delta_2\ldots$ such that length of $\Delta_i$ equals $10^{-i}$ and $\Delta_i\cap (A_i+q_i)=\emptyset$. The intersection of $\Delta_i$'s is not covered by our translates.

Now if $U_n=[0,1]\cap A_n$, the measures of $U_n$ are bounded from below.


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.