Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am thinking about this question for a while. Is any infinite set $X$ on the line separable? I think the answer is yes, but to do so, I need to show there exists a countable subset $A$ in $X$ such that $X$ is in the closure of $A$.

Can anyone outline the proof for me? Thanks.

share|cite|improve this question
By "the line", do you mean the "real line". And do you put on this set $X$ the topology given by $\mathbb R$? – Davide Giraudo Dec 7 '11 at 15:52
While it is not very relevant, it may still be of some interest to people reading this thread: without the axiom of choice it is consistent that there is a subset of real numbers which is not separable. – Asaf Karagila Dec 8 '11 at 0:40
@Asaf: in my opinion your comment is totally irrelevant but somewhat interesting. :) – Pete L. Clark Dec 8 '11 at 2:53

Yes, assuming by "the line" you mean "the real line $\mathbb{R}$ equipped with the Euclidean topology".

Here is a broader take on this: compare with second countability. Facts:

1) Every second countable topological space is separable.
2) The converse need not hold, but every metrizable separable space (like $\mathbb{R}$!) is second countable.
3) Every subspace of a second countable space is second countable: just restrict the base.

Thus every subspace of a metrizable separable space is separable.

Added in response to the OP's comment: Hmm. In the departments with which I am familiar, students take undergraduate general topology before they take graduate real analysis. I am not completely confident in the ability of someone to succeed at the latter without having at least some mastery of the former.

Anyway, you asked me if there is a simpler explanation. Well, "simpler" is subjective, so in my, the explanation I gave above is the one which is simplest to me. But I can explain it in a way which doesn't use second countability: here goes.

For $k \in \mathbb{Z}$ and $n \in \mathbb{Z}^+$, let $I_{n,k} = [\frac{k}{2^n},\frac{k+1}{2^n})$.

Fix $n \in \mathbb{Z}^+$ and consider the partition $\{ I_{n,k} \}_{k \in \mathbb{Z}}$ of $\mathbb{R}$. For each $k \in \mathbb{Z}$ such that $I_{n,k} \cap X$ is nonempty, choose one point $a_{n,k}$ of $X$. Let $A$ be the subset of $X$ consisting of these points $a_{n,k}$. Then $A$ is countable. Moreover, for every $x \in X$, we may construct a sequence in $A$ converging to $x$ as follows: let $x_n$ be the unique element $a_{n,k}$ $A$ lying in the same interval $I_{n,k}$ as $x$.

share|cite|improve this answer
I do mean the real line with This topology. I will have to learn the second countability, although it's beyond my current knowledge. Can you explain it in a it simpler way, perhaps something covered in a first graduate course in real analysis? – Reggie Dec 7 '11 at 16:13

This is just the special case of Pete’s answer that you need for your problem, with a few more details.

There are countably many rationals, so there are countably many open intervals with rational endpoints; list them as $\{I_n:n\in\mathbb{N}\}$. Let $X$ be any subset of $\mathbb{R}$. For each $n\in\mathbb{N}$ there are two possibilities:

  1. $I_n\cap X=\varnothing$: In this case do nothing.
  2. $I_n\cap X\ne\varnothing$: Let $x_n$ be any point of $I_n\cap X$.

Now let $D=\{x_n:I_n\cap X\ne\varnothing\}$. Clearly $D$ is countable, and I claim that $D$ is a dense subset of $X$, i.e., such that $X=\operatorname{cl}_X D$.

To see this, let $x$ be any point of $X$, and let $(x-\epsilon,x+\epsilon)$ be an open interval centred at $x$. The rationals are dense in $\mathbb{R}$, so there are rationals $p\in(x-\epsilon,x)$ and $q\in(x,x+\epsilon)$. Then $(p,q)$ is an open interval with rational endpoints, so $(p,q)=I_n$ for some $n\in\mathbb{N}$. Moreover, $x\in I_n\cap X$, so $I_n\cap X\ne\varnothing$, and according to $(2)$ above there is a point $x_n\in D\cap I_n$. But $I_n\subseteq(x-\epsilon,x+\epsilon)$, so $x_n\in (x-\epsilon,x+\epsilon)\cap D$. In other words, for each $\epsilon>0$ the $\epsilon$-nbhd of $x$ contains a point of $D$, and therefore $x\in\operatorname{cl}_X D$, as desired.

To connect this with second countability: the family $\{I_n:n\in\mathbb{N}\}$ is a countable base for the Euclidean topology of $\mathbb{R}$. A base for the topology is just a collection $\mathscr{B}$ of open sets such that every open set is a union of some subcollection of $\mathscr{B}$; a space is second countable if its topology has a countable base. If $X$ is any subset of $\mathbb{R}$, $\{X\cap I_n:n\in\mathbb{N}\}$ is a countable base for the subspace topology on $X$, so second countability is hereditary: if a space is second countable, so are all of its subspaces. And the trick that I used above can pretty clearly be applied to any second countable space to get a countable dense subset $-$ which is why every second countable space is separable, and even hereditarily separable.

share|cite|improve this answer
$@$Brian: +1. I think this will be helpful to the OP. – Pete L. Clark Dec 8 '11 at 0:34

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.