My question is related with the understanding of open covering definition of compactness which can be stated as: A metric space $X$ is called compact iff each of its open cover has a finite subcover. For example $[0,1)$ is not compact. I understand what open covers are; for example set $A = \{(-1,1)\}$ may be one of open coverings of $[0,1)$. But I don't understand how to check whether set A admits a finite subcover or not? I tried many times but I don't know where I am lacking? Please help me with this.

Edit 1: Let $O$ be the set of open intervals of the form $(-1/2, 1-1/t)$. $O$ is an open cover but it does not admit a finite subcover. My question is related with understanding of how can we say $O$ doesn't admit a finite subcover.

Edit 2: I find it difficult to apply the open covering definition of compactness. If someone can explain this definition with examples it would be good for me.


  • $\begingroup$ I can't quite tell what you're asking. Some thoughts: The $A$ you give is already finite. Do you want an example of an open covering of $[0, 1)$ that has no finite subcover? $\endgroup$ Jun 2 '12 at 4:17
  • 2
    $\begingroup$ Exactly! Take a finite subcover $O_f$. Then there is a largest $m$ such that $(-\frac{1}{2},1-\frac{1}{m})$ is in $O_f$. But then the point $1-\frac{1}{2m}$ is not covered by any element of $O_f$. $\endgroup$ Jun 2 '12 at 4:32
  • 1
    $\begingroup$ We are just on the line, nothing fancy. Think geometrically. The distance from $1$ of any point in $-1/2,1-1/t)$ is greater than $1/t$. Let $m$ be the largest number such that $(-1/2,1/m)$ is in our supposed finite subcover. The point $1-1/(2m)$ is distance $1/(2m)$ from $1$, which is closer than any point in the subcover. Actually, already $1-1/m$ is (barely) too close to $1$ to be covered. Sketch a picture, pick a finite number of $t$ like $5$, $8$, and $17$. Find a point not covered. $\endgroup$ Jun 2 '12 at 5:14
  • 1
    $\begingroup$ Good. You shouldn't apologize, these things are hard at the beginning. But remember that there are simple pictures associated with most of these ideas. We use the pictures to write down a formal argument, because our mental picture could be wrong. There are for example some continuous functions that are weirder than anything one could imagine drawing. $\endgroup$ Jun 2 '12 at 5:36
  • 1
    $\begingroup$ That's right, none of them can be covered. $\endgroup$ Jun 2 '12 at 6:01

Here's the first task.

Let $U_n = (0, 1 - \frac1n)$ for $n \in \mathbf N$. Show that the open covering $\{U_n\}$ of $(0, 1)$ has no finite subcover.

Some steps you could follow:

  1. A finite subcover is of the form $\{U_n\}_{n \in S}$ for some finite subset $S$ of $\mathbf N$.
  2. If $S$ is non-empty then let $N$ be the largest element of $S$. Then $U_n \subset U_N$ for all $n \in S$, so it is enough to show that $U_N$ does not contain all of $(0, 1)$.
  3. $1 - \frac1N$ is an element of $(0, 1) \setminus U_N$.

I can't tell you how to understand the definition in a few words, but here are some thoughts. It's often enough to understand a special case of something, while keeping in mind that exceptions exist. Here, the familiar case is that of $\mathbf R$. More generally, in a metric space compactness is equivalent to sequential compactness. This might be a more visceral notion, and it might help to see the proof of equivalence in section 2 of this handout of Brian Conrad's.

For me, this justifies thinking of non-compact sets as those in which points can run off, either to a point outside of the set or "to infinity". See also this (closed) question at MO. Try thinking about why you can't construct an example like the one you gave for $[0, 1]$.

As opposed to applying the open covering definition directly, it's often enough to remember the fundamental fact that a subset of $\mathbf R^n$ is compact if and only if it is closed and bounded, together with standard theorems on compactness.

  • $\begingroup$ Please check whether I am right or not. If yes then i got the concept. I took U = (0, 1- 1/2n) as open covering of (0,1). Now took a finite subcover say A ={(0, 1-1/n)} OF (0,1). But there is largest m such that point 1-1/2m is not covered by A. $\endgroup$
    – Srijan
    Jun 2 '12 at 4:57
  • $\begingroup$ @srijan I can't quite follow the notation. What I wrote above makes some sense to me. Does it look alright to you? $\endgroup$ Jun 2 '12 at 5:39

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.