I want to show the following:

Let $X$ be a metric space. Show that every compact subset $Y$ of $X$ is closed.

The idea is to show that $X\setminus Y$ is open. So, for any $x \in X\setminus Y$, I cover up the compact set $Y$ with balls defined as $B_{\epsilon}(y_0):=\{y\in Y \mid d(y,y_0)<\frac{d(x,y_0)}{2}\}$. Then $\bigcup_{i\in I}B_{\epsilon}(y_0^i)=Y$. Due the compactness of $Y$ there exists finitely many balls such that: $\bigcup_{i = 1}^n B_{\epsilon}(y_0^i)$. Now it is clear that $B_{\epsilon}(x)\subseteq X\setminus Y$.

Is this correct? And I don't like the style of the notations. Maybe someone can give me some advice?


Well, the notation isn't formally correct. It's ok to take $B_ε(y_0) := \{y ∈ Y \mid d(y, y_0) < \frac{d(x, y_0)}{2} \}$. But then $ε$ stands for nothing and is just part of the name so it doesn't make sense when you take the final $ε$ in the last step. On the other hand the definition corresponds to the definition of open ball with radius of $\frac{d(x, y_0)}{2}$ so probably for each $y ∈ Y$ you choose radius $ε_y := \frac{d(x, y)}{2}$ and then take open ball $B_{ε_y}(y)$.

Also $I$ doesn't cover anything, it is just index set, which in our case has no well-defined meaning. What is $y_0^i$? You can actually index by the points of $Y$ so your cover is $\{B_{ε_y}(y): y ∈ Y\}$. Then there is finite $F ⊆ Y$ such that $\{B_{ε_y}(y): y ∈ F\}$ still covers $Y$. Then we take $ε := \min\{ε_y: y ∈ F\}$ and now $B_ε(x) ⊆ X \setminus Y$.

  • $\begingroup$ Thanks for the answer, that helped a lot. One question is left. $\epsilon$ is not 0, only because we have a finite set of balls, which covers our compact set, right? If we do not have compactness, then $\epsilon$ could be zero, right? $\endgroup$ – Marm Jun 26 '14 at 15:25
  • 1
    $\begingroup$ @Balla Yes you are right, but formally it is more like $ε > 0$ because it is mininum: it is one of those $ε_y$'s which are positive. If the set was infinite it may have not a minimum. And if you took infimum instead, then it could be zero as you say. $\endgroup$ – user87690 Jun 26 '14 at 15:40

Disregard this if you haven't covered sequential compactness.

In the scope of metric spaces, compactness via open cover and sequential compactness are equivalent.

To prove that $Y$ is closed, consider a sequence $(x_n)\in Y^\mathbb N$ such that $(x_n)$ converges to some $l \in X$.

Your goal is to prove $l\in Y$.

Since $Y$ is compact, there is some subsequence $(x_{n_k})$ that converges to some $L\in Y$

But any subsequence of a convergent sequence converges to the limit of the original sequence.

Hence $l=L$ and $l\in Y$

  • $\begingroup$ thanks for your answer, but i dont want to prove it in this way :). So without sequential compactness $\endgroup$ – Marm Jun 26 '14 at 15:21

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.