1. Limited time only! Sign up for a free 30min personal tutor trial with Chegg Tutors
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Homework Help: Compact subset of open set?

  1. Feb 6, 2014 #1
    1. The problem statement, all variables and given/known data

    Let [itex]K \subset \mathbb{R^n}[/itex] be compact and [itex]U[/itex] an open subset containing [itex]K[/itex]. Verify that there exists [itex]r > 0[/itex] such that [itex]B_r{u} \subset U[/itex] for all [itex]u \in K [/itex].

    2. Relevant equations

    Every open cover of compact set has finite subcover.

    3. The attempt at a solution

    I tried to cover my [itex]K[/itex] with open balls therefore there should be finitely many open balls (because [itex]K[/itex] is compact). If I choose [itex]r' = min(r_1,r_2,...,r_n)[/itex] ([itex]r_i[/itex] being the radius of the ball) then every element in [itex]K[/itex] has that required ball-neighbourhood. Because [itex]U[/itex] is open then [itex]B_{r'}{u} \subset U[/itex]. Is this correct?
  2. jcsd
  3. Feb 6, 2014 #2


    User Avatar
    Science Advisor
    Homework Helper
    Gold Member

    That doesn't prove it. Even if you have a finite subcover of open balls, then that subcover might extend beyond U. You need to find a way to make your subcover be subsets of U.

    ... also, you need the r to work for every u in K, not just a finite subset.

    ... maybe using open covers is not the answer? You are working in $$\mathbb{R}^n$$
    Last edited: Feb 6, 2014
  4. Feb 6, 2014 #3

    Ok, so my reasoning does not work. How about the fact that since [itex]U[/itex] is open, then every element [itex]x \in U[/itex] has an open ball-neigbourhood [itex]B_r{x} \subset U[/itex]. Then every element in [itex]K[/itex] has that neigbourhood also. Take all these neighbourhoods and then this is an open cover of [itex]K[/itex]. Because [itex]K[/itex] is compact then it has to have finite subcover which has finitely many of these open balls. Then this subcover has to be subset of [itex]U[/itex] and if I take minimum of the [itex]r[/itex]:s, then every element [itex]u \in K[/itex] has that required ball-neighbourhood.

    I do not know why [itex]\mathbb{R^n}[/itex] is relevant here.. Am I supposed to use some metric?
  5. Feb 6, 2014 #4


    User Avatar
    Science Advisor
    Homework Helper
    Gold Member

    Take a ball, radius r, centred on k in U. Let x be in this ball. Then the ball, radius r, centred on x may not be in U.

    In general, any finite subcover will give you a finite set of balls, but that isn't easily going to show that a ball of radius r centred on any point in K is in U.

    Instead, I'd think about K being closed and bounded.
  6. Feb 6, 2014 #5
    I feel I'm completely lost now.. Of cource since [itex]K[/itex] is compact then it is closed and bounded. Because it is bounded then for every [itex]u \in K[/itex] [itex]\|u\| \leq M[/itex].

    I really need some hint now..
  7. Feb 6, 2014 #6
  8. Feb 6, 2014 #7


    User Avatar
    Science Advisor
    Homework Helper
    Gold Member

    Here's an outline of a basic proof:

    Assume not, then you get a sequence in K and a sequence not in U that get arbitrarily close.

    As K is compact, the sequence in K has a subsequence which converges to k in K.

    And, it's easy to show that the corresponding subsequence of points not in U also converges to k.

    But, as U is open, this is a contradiction (as k has a neighbourhood in U).

    Can you do your own formal proof based on that?
  9. Feb 6, 2014 #8
    Well, only difficulty I see in the proof is that how to show the corresponding subsequence of points not in U also converges to k. Of cource if points get arbitarily close, so the distance between points is less than r > 0.. ?
  10. Feb 6, 2014 #9
    I think I got it now.

    So let's assume that there exist no such r. Therefore there is sequence [itex](x_n) \subset \mathbb{R^n}\setminus U[/itex] such that [itex]\|{x_n - k_{n_j}}\| < \frac{\epsilon}{2}[/itex]. Now [itex]k_{n_j}[/itex] is a converging subsequence in [itex]K[/itex].

    Because [itex]K[/itex] is compact there exists subsequence [itex]k_{n_j}[/itex] which converges to some [itex]k[/itex], so [itex]\|{k_{n_j} - k}\| < \frac{\epsilon}{2} [/itex]

    Now consider [itex]\|{x_n - k}\| =\|{x_n - k_{n_j} + k_{n_j} - k}\| \leq \|{x_n - k_{n_j}}\| + \|{k_{n_j} - k}\| < \frac{\epsilon}{2} + \frac{\epsilon}{2} = \epsilon [/itex]

    So [itex](x_n)[/itex] converges too to [itex]k[/itex]. This is impossible, since [itex]U[/itex] is open and therefore every element in [itex]U[/itex] has open ball-neigbourhood that is a subset of [itex]U[/itex].

    Therefore there is such r.

    Q.E.D ?
    Last edited: Feb 6, 2014
  11. Feb 6, 2014 #10


    User Avatar
    Science Advisor
    Homework Helper

    I don't see how those inequalities with a fixed ##\epsilon## are telling you much. Why don't you start by assuming something like ##\|{x_{n} - k_{n}}\| < \frac{1}{n}##?
  12. Feb 7, 2014 #11
    You could also try out the Lebesgue's number lemma . But I think you will have to prove that lemma for sake of completion. Of course , proof is available online and using the Lemma almost solves the question. http://en.wikipedia.org/wiki/Lebesgue's_number_lemma
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook

Have something to add?
Draft saved Draft deleted