Take the 2-minute tour ×
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.

I'm trying to come up with a simple example of a second countable Hausdorff space that is not metrisable. The most promising I've come up with so far is the Box topology:

The following is a basis for the Box topology on $\mathbb R^{\mathbb N}$: $\{ \prod_{n \in \mathbb N} O_n \mid O_n \text{ open in } \mathbb R \}$.

Unfortunately, I have been unable to either put a suitable metric on it or show that it's not metrisable. The metrics that have come to mind so far are:

(i) $d_n (x,y) = (\sum_k |x_k - y_k|^n)^{\frac{1}{n}}$, inducing the product topology and

(ii) the discrete metric inducing the discrete topology


Question 1: What other metrics (equivalent to the above or not) exist that I have not thought of?

Question 2: If it turned out that this is not metrisable: how does one show that a space is not metrisable? (in general, or if there is not general recipe, in this case)

Question 3: Can you come up with a simpler example than the Box topology?

Many thanks for the help.

share|improve this question
1  
What about $\sup_n |x_n - y_n|$? –  Ilya Jan 15 '13 at 18:56
3  
For examples, try using this website. –  David Mitra Jan 15 '13 at 18:57
1  
@Ilya, you need to bound that metric. –  JSchlather Jan 15 '13 at 18:59
1  
@MattN.: oh, as Jacob advised instead of $|\cdot|$ please take $\frac{|\cdot|}{1+|\cdot|}$. And don't forget to check whether this metric indeed induces the very same topology –  Ilya Jan 15 '13 at 19:01
5  
The box topology is not first countable, so it is not metrizable. –  Martin Jan 15 '13 at 19:02
show 13 more comments

3 Answers

up vote 7 down vote accepted

A second countable Hausdorff space is metrizable iff it is regular, so you needn’t even worry about metrics: you just need a non-regular Hausdorff space with a countable base. One easy example is to let $X=\Bbb Q^2$ with the following topology. Points $\langle p,q\rangle$ with $q\ne 0$ are isolated. A basic open nbhd of $\langle p,0\rangle$ is any set of the form

$$\big\{\langle p,0\rangle\big\}\cup\left((p-\epsilon,p+\epsilon)\times\Big((-\epsilon,\epsilon)\setminus\{0\}\Big)\right)\;.$$

In other words, points on the $x$-axis have nbhds consisting of themselves and the intersection of an ordinary Euclidean nbhd with the complement of the $x$-axis.

Added: Box products are not the way to go. Suppose that there are distinct points $x,y\in X$ such that $D=\{x,y\}$ is a discrete subspace of $X$. Then the box product $\square X^\omega$ of $\omega$ copies of $X$ contains $\square D^\omega$ as a discrete subspace of cardinality $2^\omega$ and therefore cannot be second countable.

Added2: Let $S=\{0\}\cup\left\{2^{-n}:n\in\Bbb N\right\}$ as a subspace of $\Bbb R$. Let $Y=\Bbb N\times S$, let $p$ be any point not in $Y$, and let $X=\{p\}\cup Y$. $Y$ is an open subspace of $X$. For $n\in\Bbb N$ let

$$B_n(p)=\{p\}\cup\big\{\langle k,s\rangle\in Y:k\ge n\text{ and }s\ne 0\big\}\;,$$

and take $\{B_n(p):n\in\Bbb N\}$ as a local base at $p$. $X$ is countable and first countable, so $X$ is second countable, and $X$ is clearly Hausdorff. However, $\Bbb N\times\{0\}$ is a closed subset of $X$ that cannot be separated from $p$ by disjoint open sets, so $X$ is not $T_3$.

share|improve this answer
    
Thank you! You're very helpful. I will accept it tomorrow after re-reading it. Now I'm too tired. I hope you don't mind. (I already upvoted) –  Matt N. Jan 15 '13 at 19:35
    
@Matt: No problem. I might even think of another good example. :-) –  Brian M. Scott Jan 15 '13 at 19:38
add comment

Another example is the $R_K$ topology. If $K$ denotes the set of all numbers of the form $\frac{1}{n}$ for $n \in \mathbb{Z}_+$, it is the topology generated by all open intervals $(a.b)$ together with those of form $(a,b) - K$. So it's in spirit related to Brian Scott's example (actually, 2 of them), but - for all I know - one of the 2 or 3 standard easy examples to look at for unusual behavior (eg, as Munkres introduces point set topology, it keeps being discussed (text and exercises); it is also in Counterexamples in Topology)).

As in BMS's examples, the $R_K$ topology is not regular but 2nd countable, and so cannot be metrizable. To see this, note:

(a) Not regular:

The set $K$ is closed, and it cannot be separated from $0$: assuming two open sets $U$ (containing $0$) and $V$ (containing $K$) separate them, then there is an open neighborhood around $0$ in $U$ which must be of form $(a,b) - K$ (or it intersects $K$); choose $n$ large enough that $\frac{1}{n} \in (a,b)$, and a neighborhood around $\frac{1}{n}$ in $V$ which must be of form $(c,d)$, chosen such that $c > 0$, and $(c,d) \subset (a,b)$. Then it is easy to see (eg, draw the picture of the above) that there is a point in $U \cap V$ ("between $c$ and $\frac{1}{n}$").

(b) 2nd countable:

This is rather obvious (for the same reason that the standard topology is 2nd countable)

share|improve this answer
add comment

You might want to check out Example 2 in Topology (2nd ed.) by Munkres on p. 133. It shows that an uncountable product of $\mathbb{R}$ cannot be metrizable since it fails the sequence lemma given on p. 130 of this textbook.

share|improve this answer
add comment

Your Answer

 
discard

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.