This problem is from our analysis midterm exam.

Let $X$ be the set of all sequences $\{a_n\}$ with $\sup_n|a_n|\leq 1$, and the metric on $X$ is given as $$d(a,b)=\sup_{n}\left|\frac{a_n-b_n}{n}\right|$$ where $a=\{a_n\}$ and $b=\{b_n\}.$ Using the open cover definition of compactness, prove that $X$ is compact under this metric.

I already know that this metric induces the product topology of $[-1,1]^\mathbb{N}$, and so the compactness follows from Tychonoff's theorem. However, this problem requires me to prove the compactness using the open cover directly(since this is an undergraduate analysis exam). How should I do?

Any hints and advice are welcome!


Here is a proof using sequences:

Let $A_s=(\{a_{n,s}\})_{s\in \Bbb N}$ be any sequence in $X .$

Take a sub-sequence $S_1$ of $\Bbb N$ such that the sequence $(a_{1,s})_{s\in S_1}$ converges to some $b_1\in [-1,1].$ Take $m_1\in S_1.$ Recursively,for $k\in \Bbb N$ take $S_{k+1}$ to be a sub-sequence of $S_k$ such that $(a_{n+1,s})_{s\in S_{k+1}}$ converges to $b_{k+1}.$ Then take $m_{k+1}\in S_{k+1}$ such that $m_{k+1}>m_k.$

Now $(m_k)_{k\ge n}$ is a sub-sequence of $S_n$ for each $n.$ So $(a_{n,m_k})_{k\in \Bbb N}$ converges to $b_n$ for each $n.$ Let $B=\{b_n\}\in X.$

For $\epsilon >0$ take $m\in \Bbb N$ with $1/m</2\epsilon,$ and then take $k^*\in \Bbb N,$ large enough that $\forall n\le m \,\forall k\ge k^*\,(|a_{n,m_k}-b_n|<\epsilon).$ Then $$\forall k\ge k^* \,(d(A_{m_k}, B)<\epsilon).$$

So $X$ is sequentially compact. For a metric space this implies being compact.

APPENDIX. Let $(X,d)$ be any sequentially compact metric space. Let $C$ be an open cover of $X$. For each $n\in \Bbb N$ let $D_n$ be the set of open balls $b$ of radius $1/n$ such that $b\subset c$ for some $c\in C. $

(i-a). If some $D_n$ is a cover of $X$ and $D_n$ has a finite sub-cover $E$ then for each $e\in E$ let $e\subset f(e)\in C.$ Then $\{f(e):e\in E\}$ is a finite subset of $C$ and is a cover of $X$.

(i-b). If some $D_n$ is a cover of $X$ but $D_n$ has no finite sub-cover then for $j\in \Bbb N$ let $b_j=B_d(x_j,1/n)\in D_n$ such that $x_j\not \in \cup_{k<j} b_k.$ Then $d(b_k,b_j)\ge 1/n$ for all distinct $j,k\in \Bbb N$ so the sequence $(b_j)_{j\in \Bbb N}$ has no convergent sub-sequence, a contradiction.

(ii). If no $D_n$ is a cover of $X,$ let $x_n\in X$ \ $\cup D_n$ and let $(x_{n_i})_{i\in \Bbb N}$ be a convergent sub-sequence of $(x_n)_{n\in \Bbb N},$ with limit point $x.$ There exists $c\in C$ and $m\in \Bbb N$ such that $B_d(x,1/m)\subset c$ and there exists $i\in \Bbb N$ such that $n_i>2m$ and $d(x,x_{n_i})<1/2m.$ But then $B_d(x_{n_i},1/n_i)\subset B_d(x,1/m)\subset c\in C,$ implying $x_{n_i}\in \cup D_{n_i},$ a contradiction.

Therefore a sequentially compact metric space is compact.

Remarks.(1). From the Appendix we also conclude that if $(X,d)$ is a compact metric space and $C$ is an open cover of $X$ then for some $n$ and some finite $E\subset D_n$ we have $X=\cup E,$ which is the Lebesgue Covering Lemma.

(2). The $\in$-order topology on the cardinal ordinal $\omega_1$ is sequentially compact but not compact, but this space is not metrizable.


It's easy to construct a continuous surjection from $2^{\mathbb N}=\{0,1\}^{\mathbb N}$ to $[0,1]$. (Hint: $ [0,1]=\{\sum_{k \in N} \frac{b_{k}}{2^{k}} \text { where } b_{k} \in\{0,1\}\} $ ). Then there is a continuous surjection from $(2^\mathbb N)^{\mathbb N}$ to $[0,1]^\mathbb N$. Notice $(2^\mathbb N)^{\mathbb N}$ is homeomorphic to $2^{\mathbb N \times\mathbb N}$, which is homeomorphic to $2^{\mathbb N}$ (as there is a bijection from $\mathbb N \times\mathbb N$ to ${\mathbb N}$). Now it suffices to prove $2^\mathbb N$ is compact, as the continuous image of a compact set is compact. Let $C=\{\sum_{k \in N} \frac{a_{k}}{3^{k}}, \text { where } a_{k} \in\{0,2\}\}$. Again, it's easy to constuct a homeomorphism from $C$ to $2^\mathbb N$. But what is $C$? It's the Cantor set, being closed and bounded, is compact. Thus $2^\mathbb N$ is compact. Hence $[0,1]^\mathbb N$, being the continuous image of a compact set, is compact.

Remark: @DanielWainfleet proof is easier (+1!). But my proof doesn't use the axiom of choice. So I think it's worth leaving it here.

  • $\begingroup$ I'm afraid the fact that $(2^{\mathbb{N}})^\mathbb{N}$ is homeomorphic to $2^{\mathbb{N}\times\mathbb{N}}$ does not appear in the analysis course. $\endgroup$
    – bellcircle
    Apr 15 '19 at 11:18
  • $\begingroup$ @bellcircle It's easily proven by definition $\endgroup$ Apr 15 '19 at 11:22

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.