2.9 Sequential Compactness
There is a second description of compactness, phrased in sequences rather than covers, and in metric spaces it says exactly the same thing. Sequences are usually the easier of the two to work with.
Definition 2.56. A metric space \(X\) is sequentially compact if every sequence in \(X\) has a subsequence converging to a point of \(X\).
Lemma 2.57 (Lebesgue number lemma). Let \(X\) be sequentially compact and let \(\xi \) be an open covering of \(X\). Then there is a number \(\delta >0\), called a Lebesgue number for the cover, such that every subset of \(X\) of diameter less than \(\delta \) is contained in a single member of \(\xi \).
Proof. Suppose no such \(\delta \) exists. Then for each \(n\) the number \(\frac 1n\) fails, so there is a set \(A_n\subset X\) with \(\diam (A_n)<\frac 1n\) lying inside no member of \(\xi \). Choose \(x_n\in A_n\).
By sequential compactness some subsequence \(x_{n_k}\rightarrow x\). Now \(x\) lies in some \(G\in \xi \), and \(G\) is open, so \(B(x;\varepsilon )\subset G\) for some \(\varepsilon >0\). Choose \(k\) so large that \[d(x_{n_k},x)<\frac {\varepsilon }{2}\qquad \text {and}\qquad \frac {1}{n_k}<\frac {\varepsilon }{2}.\] Every point of \(A_{n_k}\) is within \(\frac {1}{n_k}\) of \(x_{n_k}\), hence within \(\varepsilon \) of \(x\), so \[A_{n_k}\subset B(x;\varepsilon )\subset G ,\] contradicting the choice of \(A_{n_k}\) as lying inside no member of \(\xi \). □
Theorem 2.58. For a metric space \(X\) the following are equivalent.
- (i)
- \(X\) is compact.
- (ii)
- \(X\) is sequentially compact.
- (iii)
- \(X\) is complete and totally bounded.
Proof.
(i) \(\implies \) (ii)
Let \((x_n)\) be a sequence in \(X\) with no convergent subsequence. Then no point \(x\) of \(X\) is a limit of a subsequence, so each \(x\) has a ball \(B(x;r_x)\) containing \(x_n\) for only finitely many indices \(n\). These balls form an open cover; a finite subcover \(B(y_1;r_1),\dots ,B(y_m;r_m)\) would then contain \(x_n\) for only finitely many \(n\) in total, yet it must contain every term of the sequence. That is impossible, so some subsequence converges.
(ii) \(\implies \) (iii)
Total boundedness is the theorem of the previous subsection: every sequence in a sequentially compact space has a convergent, hence Cauchy, subsequence, and that is equivalent to total boundedness. For completeness, let \((x_n)\) be Cauchy. By hypothesis it has a subsequence converging to some \(x\). A Cauchy sequence with a convergent subsequence converges to the same limit, since given \(\varepsilon >0\) one may choose \(N\) with \(d(x_m,x_n)<\frac {\varepsilon }{2}\) for \(m,n\geq N\) and then a term \(x_{n_k}\) of the subsequence with \(n_k\geq N\) and \(d(x_{n_k},x)<\frac {\varepsilon }{2}\); the triangle inequality gives \(d(x_n,x)<\varepsilon \) for all \(n\geq N\).
(iii) \(\implies \) (ii)
Let \((x_n)\) be a sequence. Total boundedness gives a Cauchy subsequence, and completeness makes it converge.
(ii) \(\implies \) (i)
Let \(\xi \) be an open cover of \(X\) and let \(\delta >0\) be a Lebesgue number for it. By the implication already proved, \(X\) is totally bounded, so it can be covered by finitely many sets \(E_1,\dots ,E_n\) each of diameter less than \(\delta \). By the choice of \(\delta \) each \(E_k\) lies in some \(G_k\in \xi \), and then \[X = \bigcup _{k=1}^{n}E_k \subset \bigcup _{k=1}^{n}G_k .\] So \(\{G_1,\dots ,G_n\}\) is a finite subcover. □
Note. The equivalence is a metric-space phenomenon. In general topological spaces compactness and sequential compactness are different conditions, neither implying the other. Everything in this course takes place in metric spaces, so the two may be used interchangeably — but it is worth knowing that the licence to do so has a reason behind it.
Solution. Let \(F\) be closed in the compact space \(X\), and let \((x_n)\) be a sequence in \(F\). Regarded as a sequence in \(X\) it has a subsequence converging to some \(x\in X\), by sequential compactness. Since \(F\) is closed and every term of the subsequence lies in \(F\), the limit \(x\) lies in \(F\). So every sequence in \(F\) has a subsequence converging in \(F\), and \(F\) is sequentially compact, hence compact.
The cover argument is just as short, and either is acceptable; the point of having both descriptions is that one of them is usually the easier.
Questions on this section
Stuck on something here? Ask below and it stays attached to this topic.