While trying to prove that a set is sequentially compact, I was suggested to prove by contradiction -- this is how it went, at least part of it:

Definition. We say that a set $A$ is sequentially compact if every sequence $(x^m)$ with $\{x^m\}\subseteq A$ has a sequence $(x^{m_k})$ that $x^{m_k}\rightarrow x\in A$.

Theorem. Being $(M,d)$ a metric space, a set $A\subseteq M$ is compact only and only if it is also sequentially compact.

part of what you should assume for the demonstration. Suppose initially that $A$ is a compact set. Suppose too that the sequence $(x^m)$ is such that $\{x^m\}\subseteq A$ and $(x^m)$ does not possess any sequence that converges to any element of $A$. As we know that $A$ is a closed set, that implies that $(x^m)$ does not possess any convergent sequence in $(M,d)$.

Actual Demonstration (part of it [the part I cannot understand]). Begin trying to define $\{x^m\}$ as a closed set. Take $y\in M-\{x^m\}$ and suppose $\{x^m\}$ as a not closed set (implying $M-\{x^m\}$ as a not open set). Suppose that for every ε>0, and for every N, $B(y,ε)∩({(x^m):m≥N})≠∅$. For $k=1$, take $x^{m_1}\in\{x^m\}\cap B(y,1)\neq\emptyset$. Take $x^{m_2}\in\{x^m\}\cap B,(y,1/2)\neq\emptyset$, such that $m_2>m_1$. To see that this is possible, define $\delta:=\min(x^m,y)$. By hypothesis, $\{x^m\}\cap B,(y,\delta)\cap B(y,1/2)\neq\emptyset$; so we can take $x^{m_2}\in\{x^m\}\cap B,(y,\delta)\cap B(y,1/2)\neq\emptyset$. We can guarantee that $m_2>m_1$. Proceeding inductively in such manner, we can obtain a subsequence $(x^{m_k})$ of $(x^m)$ such that, for every $k$, $d(x^{m_k}, y)<1/k$. So $x^{m_k}\rightarrow y$. That is a contradiction with the statement $(x^m)$ does not have any converging subsequences. We conclude that $\{x^m\}$ is a closed set.

Here it is what I haven't figured out: I don't know if I understood the whole demonstration correctly (or copied correctly from the board in my class), but how can we prove that $\{x^m\}$ is closed if a subsequence $(x^{m_k})$ of $(x^m)$ converges to a point outside of $\{x^m\}$, since $y\in M-\{x^m\}$? The trick was in assuming that $\{x^m\}$ didn't have any converging sequences in (M,d) and proving a contradiction in the demonstration. But how does that make $\{x^m\}$ a closed set? I still cannot see what I am missing. Can someone point if I am understanding something wrong or if there is any correction to be made in this demonstration?

Notation: I used $\{x^m\}$ as the set of the sequence $(x^m)$. If there are any questions about notation, please ask.

Observations (edited): I'm having some trouble in writing in this system -- I welcome any help. My question changed a bit now because of further development of the discussion of this demonstration with friends.

  • $\begingroup$ 1) Is $x^m$ supposed to be an element of $A$ ? Then you soulf write $x^m \in A$ instead of $x^m \subset A$. 2) What do you mean by 'define $x^m$ as a closed set' ? 3) What is $M\{x^m\}$ ? $\endgroup$ – user10676 May 5 '13 at 9:34
  • 1
    $\begingroup$ There is a big issue with what you have copied. The sentence 'Suppose that for every $\varepsilon>0$...' should be replaced by something like 'Claim: $\forall \varepsilon>0, \forall N, B(y,\varepsilon) \cap \{ x^m: m \geq N \} \neq \emptyset$' (I let you figure out why it is true). $\endgroup$ – user10676 May 5 '13 at 12:47
  • $\begingroup$ I am sorry! something about the notation came out wrong. I think the system here took away the "\" between {x^m} and M. Writing in another way: M-{x^m}. So y belongs to complement of {x^m}. For the other questions: x^m is an element of A, but I was trying to define the set of the elements of the sequence (x^m) to be contained in A. Hence the notation {x^m}. As for {x^m} being a closed set, the demonstrations tries to prove that. And I think you're right, user10676, we have to define an element N for which the statement starts to be true -- I will edit it right now. $\endgroup$ – John Doe May 5 '13 at 15:57

I think $M\{x^m\}$ is supposed to be the margin of $\{x^m\}$, i.e. $$ M(X) := \overline{X} \setminus X \text{.} $$ where $\overline{A}$ is the closure of $A$.

The proof then shows that for every $y \in M(\{x^m\})$ there's a sub-sequence of $(x^m)$ which converges to $y$. Since $(x^m)$ is not supposed to have any convergent sub-sequences, it follows that $M\{x^m\}$ is empty, and thus that $\overline{\{x^m\}} = \{x^m\}$, i.e. that ${x^m}$ is closed.

You do the rest with Heine-Borel. If $A$ is compact then $A$ is closed and bounded, hence $\{x^m\}$ is bounded. But by the above$\{x^m\}$ is also closed, hence also compact. But if $\{x^m\}$ contains no convergent subsequence, you can create a covering of $\{x^m\}$ with open balls (centers as the $x^m$, you have pick a suitable $\epsilon$) with no open subcover, which is a contradiction to $\{x^m\}$ compact.

  • $\begingroup$ fgp, there was a notation error. I meant to write M-{x^m} or the complement of {x^m}. My trouble is with understanding this bit: why can we have $(x^{m_k}\)→y$ when y belongs to the complement of {x^m}? But thank you for your help anyhow. $\endgroup$ – John Doe May 5 '13 at 16:38

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.