Is the following proof correct?

Proposition: if we have a sequence of set $U_i$ such as $\bigcup_{i\in \mathbb{N}} U_i=[a,b]$ then there exist a $i$ such as $U_i$ is dense somewhere in $[a,b]$

Proof: If for all $i, U_i$ is nowhere dense, we have $\forall ]u,v[$

$$\forall i, \exists c \in ]u,v[, \exists h > 0,\quad U_i \cap ]c-h,c+h[ =\emptyset\text{ and } ]c-h,c+h[ \subset ]u,v[$$

Let's define

$$\left\lbrace \begin{array} .E_{c,h}^i= ]c-h,c+h[ & \text{ if } & U_i \cap ]c-h,c+h[ =\emptyset \\ E_{c,h}^i= \emptyset &\text{ if } & U_i \cap ]c-h,c+h[ \neq \emptyset \end{array}\right.$$ and $$E_i = \bigcup_{h>0} \bigcup_{c\in ]u,v[} E^i_{c,h}$$

$E_i$ is open (as a reunion of open set), and dense in $]u,v[$. Indeed, we have as

$$\forall \epsilon >0 \forall x\in ]u,v[, \exists c \in ]x-\epsilon,x+\epsilon[,\exists h >0,\quad U_i \cap ]c-h,c+h[ = \emptyset$$

$$\forall \epsilon >0 \forall x\in ]u,v[, \exists ]c-h,c+h[ \subset U_i^C,\quad]c-h,c+h[\subset ]x-\epsilon,x+\epsilon[$$

$$\forall \epsilon >0 \forall x\in ]u,v[, \exists c\in E_i,\quad c\in ]x-\epsilon,x+\epsilon[$$

So $E_i$ is an open dense set for all i. By Baire theorem, we get that $\bigcap_{i\in\mathbb{N}} E_i$ is dense in ]u,v[. As we have $\forall i, E_i\subset U_i^C$, we get that $\bigcap_{i\in\mathbb{N}} E_i \subset \bigcap_{i\in\mathbb{N}} B^C_i$ and $\bigcap_{i\in\mathbb{N}} B_i^C$ is dense in $]u,v[$. But we also have

$$\bigcap_{i\in\mathbb{N}} \left(B_i^C \cap ]u,v[\right) = ]u,v[ \cap \left( \bigcup_{i\in\mathbb{N}} B_i \right)^C = ]u,v[^C \cap ]u,v[= \emptyset$$


  • $\begingroup$ I think you can just say that by Baire's Category Theorem, it is impossible that all $U_i$ are nowhere dense. Thus, some $U_j$ satisfies $\overline{U_j}^\circ \neq \emptyset$. $\endgroup$ – PhoemueX Feb 23 '15 at 13:38
  • $\begingroup$ Oh, yes, I didn't have this version of Baire's Theorem in mind. Thanks you ! $\endgroup$ – Tryss Feb 23 '15 at 14:46

There is an easier proof (also by contradiction). Let us assume that [a,b] is the union of countable number of nowhere dense sets. By a version of Baire theorem, the countable union of nowhere dense sets cannot be complete. Interval [a,b] is a complete space. Contradiction.

P.S. More exactly, the sets which can be represented as a countable union of nowhere dense sets, are sometimes called the sets of the 1st category. Other sets are 2nd category. A theorem says that complete spaces are the sets of 2nd category.


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.