Proving that the closure of a connected set in a metric space is connected.

This was a proof that was done in class and I'm sure that it is wrong

Proof: Let $(X, d)$ be a metric space. Suppose that $Y \subseteq X$ is connected. We show that $\overline{Y}$ is also connected.

Suppose that $(\overline{Y}, d)$ was disconnected, then there exist open sets $U, V$ of $\overline{Y}$ such that $U \cup V = \overline{Y}$ and $U \cap V = \emptyset$ and $U \neq \emptyset$ and $V \neq \emptyset$.

Since $Y \subseteq \bar{Y}$ we thus have $U \cap Y$ and $V \cap Y$ to be clopen in $Y$ and $(U \cap Y) \cup (V \cap Y) = Y$. Assume that $U \cap Y \neq \emptyset$ then if $V \cap Y \neq \emptyset$ we have a contradiction.

So if $U \cap Y \neq \emptyset$ then we must have $V \cap Y = \emptyset$ and $U \cap Y = Y$. Since $V \cap Y =\emptyset$ we must have that $V$ contains only limit points of $Y$.

But then for any $x \in V$ we have $B(x, r) \cap U \neq \emptyset$ for any $r > 0$. So any ball $B(x, r)$ contains points outside of $V$ contradicting the fact that $V$ is open in $(\bar{Y}, d)$. $\square$

I'm a bit confused by the last two lines. $V$ being open in $\bar{Y}$ means $V = \bar{Y} \cap A$ for some open $A$ in $(X, d)$, I don't see why any open ball $B(x, r)$ containing points outside of $V$ would contradict the fact that $V$ is open in $(\bar{Y}, d)$ (It sure would contradict the fact that $V$ is open in $(X, d)$ though)

For example take $(\mathbb{R}, d)$ with the usual metric, and consider $\bar{Y} = [0, 1]$, and the open set $V = [0, \frac{1}{2})$ in $\bar{Y}$ then $B(0, r) = (-r, r)$ contains points outside of $V$ so by the above logic $V$ cannot be open in $\bar{Y}$ (which is wrong because $V$ is certainly open in $\bar{Y}$)

Am I correct in saying that there is an error in this proof?

  • $\begingroup$ I think the intention is that $B(x,r)$ contains points in $\bar{Y} \setminus V$ (since they are in $U$), giving the contradiction. $\endgroup$ – B. Mehta May 14 '18 at 19:34
  • $\begingroup$ The ball $B(x,r)$ is taken in $\overline{Y}$, not in $X$. This is not explicit but that's the only way. You can even forget about $Y$ and $X$ at all and restate the thesis as: if $X$ has a dense connected subset then it is connected. $\endgroup$ – freakish May 14 '18 at 19:39
  • $\begingroup$ @freakish When you say the ball $B(x, r)$ is taken in $\bar{Y}$" and not in $X$ do you mean that we should take it as $B(x, r) \cap \bar{Y} \neq \emptyset$? Then any neighbourhood (in $\bar{Y}$) of $x \in V$ contains points of $U$ which are outside $V$ giving us the contradiction? $\endgroup$ – Perturbative May 14 '18 at 20:02
  • $\begingroup$ @Perturbative Yes, exactly. $\endgroup$ – freakish May 14 '18 at 20:21

The proof you quote makes it needlessly complicated: we know $U \cap Y = Y$, $Y \cap V = \emptyset$.

continuing: $V$ being open in $\overline{Y}$ means that there exists $V'$ open in $X$ with $V' \cap \overline{Y} = V$.

Note that $Y \cap V =\emptyset$ implies $Y \subseteq X\setminus V'$ (otherwise for some $y$, $y \in Y \cap V' \subseteq \overline{Y} \cap V' = V$, which is not the case), and so $\overline{Y} \subseteq X\setminus V'$, (as the latter set is closed) and so $V' \cap \overline{Y} = V = \emptyset$. Contradiction and done.

No need to mention balls at all; this works in any topological space.


The proof is valid but writing $B(x, r) \cap U \neq \emptyset$ is a bit confusing. To clarify the argument, I would replace the last paragraph of the proof with this:

But then for any $x \in V$ we have $B(x, r) \cap Y \neq \emptyset$ for any $r > 0$. So any ball $B(x, r)$ contains points outside of $V$ since $Y \subset U$ and $U \cap V = \emptyset$. But this is absurd since $V$ is open in $(\bar{Y}, d)$.

We also get a contradiction if we assume that $V \cap Y \neq \emptyset$, so it is not possible to disconnect $\bar{Y}$ with two open sets $U$ and $V$. $\square$

Note: We make use of the fact that any point $w$ in an open set $W$ of a metric space belongs to an open ball $B(w, r)$ that is wholly contained in $W$.

So to show $\bar{Y}$ is connected the above examines opens sets. We can argue by focusing on just the closure operator used in topology.

In what follows $X$ is assumed to be a disconnected space with closed sets $K$ and $L$ providing a non-trivial decomposition.

Now any subspace of $T$ of $X$ with points in both $K$ and $L$ is manifestly disconnected along with its closure $\bar{T}$.

If $S$ is a subspace of $X$ that is contained in, say $K$, then $\bar{S} \subset K$.

Suppose we can write $X = \bar{Y}$. Then $Y$ can't be connected. If it was connected it would have to be wholly contained in either $K$ or $L$. In one case we would have that $X = K$ and in other case $X = L\,$. But either one of these identities contradicts the assumption that we started with a non-trivial decomposition of $X$.


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.