Prove that if $X$ is Hausdorff and $\mathfrak{C}$ is a nonempty chain of compact and connected subsets of $X$, then $\bigcap \mathfrak{C}$ is compact and connected.

Here are the definitions which were provided:

A family $\mathfrak{F}$ of sets is a chain if for all $A,B \in \mathfrak{F}$ either $A \subset B$, or $B \subset A$. (If $\mathfrak{C}$ is a nonempty chain of nonempty compact sets, then $\bigcap \mathfrak{C} \neq \emptyset$.)

A space is Hausdorff (or $T_2$) if for all distinct $a, b \in X$ there are $U, V \in \mathcal{T}$ so that $a \in U$, $b \in V$ and $U \cap V = \emptyset$. Here, $\mathcal{T}$ is the topology on $X$.

My confusion starts in what seems to me like a contradiction between definitions. We are assuming that $X$ is Hausdorff, so there is a separation of $X$ so there are disjoint $U, V \in \mathcal{T}$ with $a \in U$ and $b \in V$. But we are also given that there is a chain of compact and connected sets $\mathfrak{C}$, which by definition, $U \subset V$ or $V \subset U$. But they are supposed to be disjoint according to the Hausdorff property.

Can somebody please explain this to me and assist me in solving this problem with a proper proof? Many thanks in advance for your time and patience, it is greatly appreciated.

  • $\begingroup$ The fact that distinct points are contained in disjoint open sets has nothing to do with the chain of closed, connected sets that you’re given by hypothesis. For instance, $\Bbb R$ is Hausdorff, and $\{[-x,x]:x>0\}$ is a chain of closed, connected subsets of $\Bbb R$. There is nothing in the hypothesis to suggest that the open sets of $X$ form a chain. $\endgroup$ Apr 16 '15 at 20:21
  • $\begingroup$ The fact that $U,V$ are disjoint open non-empty does not imply that $X$ is disconnected unless $X=U\cup V$, so you need not confuse T$_2$ (points could be separated by disjoint neighborhoods) with $X$ being disconnected ($X$ could be partitioned into disjoint nonempty open sets). $\endgroup$
    – Mirko
    Apr 16 '15 at 21:48

It seems the following.

Fix any element $K\in\mathfrak{C}$. Then the set $C=\bigcap \mathfrak{C}$ is compact as an intersection $\bigcap\{ K\cap L:L\in\mathfrak{C}\}$ of a family $ \{ K\cap L:L\in\mathfrak{C}\}$ of closed subsets of a compact set $K$.

Assume that the set $C$ is not connected. This means that $C$ can be represented as a union $C_1\cup C_2$ of two its disjoint non-empty clopen (that is, closed and open) subsets. Since $C_1$ and $C_2$ are disjoint compact subsets of a Hausdorff space $X$, it is well known and easy to prove that they can be separated by disjoint open neighborhoods, that is there exist two disjoint open subsets $U_1$ and $U_2$ of the space $X$ such that $C_1\subset U_1$ and $C_2\subset U_2$. Put $U=U_1\cup U_2$.

Now we are going to find a set $L\in\frak C$ such that $L\subset U$. If $K\subset U$, we are done. In the opposite case consider a family ${\frak C}’=\{(K\cap L)\setminus U:L\in\frak C\}$. Since $\bigcap{\frak C}=C\subset U$, $\bigcap \mathfrak{C}'=\varnothing$. Since $\frak C’$ is a family of compact subsets of the compact space $K$, it contains an empty element $(K\cap L)\setminus U$. Since the family $\frak C $ is a chain, this implies that there exists a set $L\in\frak C$ such that $L\subset U$. Then $(L\cap U_1)\cup (L\cap U_2)$ is a partition of the set $L$ into two its disjoint clopen subsets. Each of these sets is non-empty. Indeed, since $\frak C$ is a chain, $L\supset \bigcap \mathfrak{C}=C$. Then $L\cap U_i\supset C\cap U_i\ne\varnothing$. So the set $L$ is not connected, a contradiction.


To start, notice that the intersection of any chain of nonempty compact sets in a Hausdorff space must be nonempty (by the finite intersection property for closed sets). Let $D = \cap \mathfrak{C}$. Then $D$ is nonempty by the preceding statement.

To prove that $D$ is compact, choose some $C_0 \in \mathfrak{C}$. Since $X$ is Hausdorff, all compact subsets of $X$ are closed in $X$, and thus all of the sets in $\mathfrak{C}$ are closed in $X$. It follows that $D=\cap \mathfrak{C}$ is closed in $X$ (since intersections of closed sets are closed). Thus $\cap \mathfrak{C}$ is closed in $C_0$, so since closed subsets of compact spaces are compact, it follows that $D$ is compact.

To prove that $D$ is connected, argue by contradiction. Suppose that $D$ is not connected, so we can find closed subsets $A,B \subset X$ such that $D \subset A \cup B$ and $A \cap D \neq \emptyset$ and $B \cap D \neq \emptyset$ and $A \cap B \cap D = \emptyset$. Notice that for any $C \in \mathfrak{C}$, the set $A \cap B \cap C$ is compact because it is a closed subset of the compact set $C$. Therefore, there exists some $C_1 \in \mathfrak{C}$ such that $A \cap B \cap C_1 = \emptyset$, otherwise $\{ A \cap B \cap C\; | \; C \in \mathfrak{C} \}$ would form a chain of nonempty compact sets, and thus $\bigcap_{C \in \mathfrak{C}} A \cap B \cap C= A \cap B \cap D$ would be nonempty (which we know is false). Also notice that $A \cap C_1$ and $B \cap C_1$ are both nonempty, since they contain the nonempty sets $A \cap D$ and $B \cap D$, respectively. Thus $A \cap C_1$ and $B \cap C_1$ form a separation of the set $C_1$ which contradicts the fact that $C_1$ is connected.

  • $\begingroup$ You want to prove that $D$ is compact, so you can't use its compactness. But it's much easier: every member of $\mathfrak{C}$ is closed (the space is Hausdorff), so their intersection $D$ is closed; then $D$ is a closed subset of a compact set, hence compact. In the proof of connectedness, you state that $U\cap V=\emptyset$, but all you can say is that $U\cap V\cap D=\emptyset$. $\endgroup$
    – egreg
    Apr 16 '15 at 21:46
  • $\begingroup$ @egreg: Yes, my post was circular (I was very careless), but I have edited it, and I think that it works now. Can someone verify if it is correct now? $\endgroup$
    – shalop
    Apr 17 '15 at 4:12
  • $\begingroup$ Seems good now! $\endgroup$
    – egreg
    Apr 17 '15 at 8:48
  • $\begingroup$ I confirm, quite nice proof, clean concise and clear. $\endgroup$
    – am70
    Feb 28 '20 at 17:20

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.