Let $\langle\mathbf{R}^n,\mathscr{O}\rangle$ be the $n$-dimensional Euclidean space. Define $\mathbf{Q}\subseteq\mathcal{P}(\mathscr{O})$ to consist of all sets $\mathsf{Q}$ which simultanously satisfy the following four conditions:

  1. $\mathsf{Q}\neq\emptyset$ and $\mathsf{Q}\neq\{\emptyset\}$
  2. $\forall_{U\in\mathsf{Q}}\exists_{V\in\mathsf{Q}}\,\mathrm{Cl}\,{V}\subseteq U$
  3. $\forall_{U,V\in\mathsf{Q}}\,(U\neq V\rightarrow \mathrm{Cl}\,V\subseteq U\vee \mathrm{Cl}\,U\subseteq V)$
  4. $\forall_{A,B\in\mathscr{O}}\,((\forall_{U\in\mathsf{Q}}\,A\cap U\neq\emptyset\neq B\cap U)\rightarrow\mathrm{Cl}\,A\cap \mathrm{Cl}\,B\neq\emptyset)$

So every element of the family $\mathbf{Q}$ is a set in which well-inside relation ($\mathrm{Cl}\,U\subseteq V$) is connected (in the sense of order) and which is decreasing with respect to this relation, but also satisfies the requirement according to which all open sets with non-empty intersections with all elements from the set must have a common adherent point.

Is it true that for every $\mathsf{Q}\in\mathbf{Q}$ there is a point $p\in\mathbf{R}^n$ such that $\bigcap\mathsf{Q}=\{p\}$? I will be grateful for your help.

  • $\begingroup$ It seems that $\mathsf{Q}=\{\emptyset\}$ has all your properties, but not your goal property. $\endgroup$ Commented Mar 26, 2016 at 18:29
  • $\begingroup$ Apart from that, I am confused by property 4. How can you hope to achieve this? If $\mathsf{Q}$ has a nonempty open set $U$ in it, then we can find two disjoint balls $A$ and $B$ inside $U$, with disjoint closures. So property 4 seems impossible except in the trivial case of my previous comment. Have I misunderstood? $\endgroup$ Commented Mar 26, 2016 at 18:31
  • $\begingroup$ @JoelDavidHamkins Thanks for pointing the singleton of $\emptyset$ - I corrected the question and ruled it out explicitly. As for the second part of your question. The only family with which you could obtain what you write about is $\{\mathbf{R}^n\}$, since this is the only clopen non-empty set in the space, and thus can satisfy 2. But the intention of 4 is to rule this case and guarantee that sets are <<shrinking>> around some precise location in the space. François's answer proves that 4 is a good embodiment of the intention. Is it OK, or have I misunderstood your question? $\endgroup$ Commented Mar 26, 2016 at 19:09
  • $\begingroup$ Ah, I had mis-read the scope of the second quantifier in statement 4, which makes it trivial as I had indicated. I suggest editing the parentheses to indicate that the implication is not under the scope of the quantifier. $\endgroup$ Commented Mar 26, 2016 at 19:18
  • $\begingroup$ @JoelDavidHamkins But the parentheses seem to be unambiguous, aren't they? $\endgroup$ Commented Mar 26, 2016 at 23:31

1 Answer 1


Yes. (Assuming $\varnothing$ isn't allowed in your $Q$'s, as Joel pointed out.)

First note that for such $Q$, $\bigcap Q$ cannot contain two points. Indeed, if $a,b \in \bigcap Q$ are distinct, we can pick open neghborhoods $a \in A$, $b \in B$ such that $\operatorname{Cl} A \cap \operatorname{Cl} B = \varnothing$ and these two open sets $A$, $B$ contradict condition 4. So it suffices to rule out the case that $\bigcap Q = \varnothing$.

Suppose, for the sake of contradiction, that $\bigcap Q = \varnothing$. Pick a cofinal sequence $U_0 \supset U_1 \supset \cdots$ of elements of $Q$. Define $V_n = U_n \setminus \operatorname{Cl} U_{n+1}$ and consider $A = \bigcup_{n=0}^\infty V_{4n}$ and $B = \bigcup_{n=0}^\infty V_{4n+2}$. Note that $\operatorname{Cl} A = \bigcup_{n=0}^\infty \operatorname{Cl} V_{4n}$ and $\operatorname{Cl} B = \bigcup_{n=0}^\infty \operatorname{Cl} V_{4n+2}$ because $\bigcap_{n=0}^\infty \operatorname{Cl} U_n = \varnothing$. Thus $\operatorname{Cl} A \cap \operatorname{Cl} B = \varnothing$ and yet $A \cap U \neq \varnothing \neq B \cap U$ for any $U \in Q$, which contradicts condition 4.

  • $\begingroup$ François, your proof demonstrates that it is enough to assume that the space is Urysohn and has a countable dense subset. Am I missing something? $\endgroup$ Commented Mar 26, 2016 at 22:08
  • $\begingroup$ @MadHatter: That's right. Countable dense subset is a little much, I think the countable chain condition (ccc) is enough. $\endgroup$ Commented Mar 26, 2016 at 22:26
  • $\begingroup$ François, some second thoughts. In order to apply ccc one must know that when $U\neq V$ are from $Q$, then (in case $\operatorname{Cl} U\subseteq V$), $V\setminus U$ contains a non-empty open set. This of course follows from $V\setminus \operatorname{Cl} U$ being non-empty. Yet I fail to see how to get it from Urysohn property. (The same applies to separability). $\endgroup$ Commented Mar 26, 2016 at 23:56
  • 1
    $\begingroup$ @MadHatter: The key is that $\operatorname{Cl}(\bigcup_{n=m}^\infty V_n) \subseteq \operatorname{Cl} U_m$. So $\operatorname{Cl} A \subseteq \operatorname{Cl} V_0 \cup \cdots \cup \operatorname{Cl} V_{4n} \cup \operatorname{Cl} U_{4n+4}$ and take the limit as $n \to \infty$. $\endgroup$ Commented Apr 19, 2016 at 23:39
  • 1
    $\begingroup$ François, I get the full picture now. By the way, to my taste this is a beautiful proof. Thank you very much for the hint and the help. $\endgroup$ Commented Apr 21, 2016 at 8:35

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.