15
$\begingroup$

Off the top of my head, I can't think of a non-separable compact space.

Can you provide a good example?

$\endgroup$
3
  • 2
    $\begingroup$ That's because any compact metric space is separable. So any example has to be non metrizable $\endgroup$ Commented May 7, 2014 at 15:03
  • 3
    $\begingroup$ Check dantopology.wordpress.com/2012/10/13/alexandroff-double-circle $\endgroup$ Commented May 7, 2014 at 15:07
  • $\begingroup$ Thanks Léo & Giuseppe. The Alexandroff double circle is non-separable and compact. $\endgroup$ Commented May 7, 2014 at 15:25

5 Answers 5

9
$\begingroup$

$\pi$-Base, an online version of the reference chart from Steen and Seebach's Counterexamples in Topology, gives the following examples of compact, non-separable spaces. You can view the search result for more details on these spaces.

Alexandroff Square

Closed Ordinal Space $[0, \Omega]$

Concentric Circles

Either-Or Topology

Lexicographic Ordering on the Unit Square

The Extended Long Line

Tychonoff Plank

Uncountable Excluded Point Topology

Uncountable Fort Space

Uncountable Modified Fort Space

$\endgroup$
2
  • $\begingroup$ Awesome list, Austin! $\endgroup$ Commented May 8, 2014 at 3:58
  • $\begingroup$ The long line is my favourite example of this phenomenon! $\endgroup$ Commented Dec 15, 2018 at 3:07
7
$\begingroup$

A not so simple, but curious example is the Alexandroff Double Circle:

Consider $C_i=\{(x,y) \in \mathbb{R}^2; x^2+y^2=i\}$ for $i=1,2$. Let $X=C_1 \cup C_2$ and $f:C_1 \to C_2$ the radial homeomorphism. Define a topology on $X$ as follows: the points in $C_2$ are all isolated; for each $x \in C_1$ and $n \in \mathbb{N}$ let $O(x,n)$ the arc in $C_1$ centered at $x$ and with length $\frac{1}{n}$. Now take $B(x,n)=O(x,n) \cup f(O(x,n)-\{x\})$ as open neighborhood. The picture below illustrate such $B(x,n)$.

enter image description here

This way we form a basis for a topology on $X$. $X$ provided with this topology is what we call the Alexandroff Double Circle.

Now, $X$ is not separable because all points in $C_2$ are isolated, hence no countable subset of $C_2$ is dense in $C_2$. For compactness, take an open cover $\{A_{\alpha}\}$ of $X$ which consists of basis neighborhoods. Then it is also an open cover of $C_1$. But notice that the topology of $C_1$ as a subspace of $X$ is exactly the usual one. So it is compact and we can find a finite subcollection $\{A_1,\dots,A_k\} \subset \{A_\alpha\}$ that covers $C_1$. Now, by the nature of our neighborhhods, $X-(A_1\cup \dots \cup A_k)$ is a finite set, so we take an extra $A_{\alpha}$ for each of these points, obtaining a finite subcover.

I don't know if this space is useful for anything other than counterexamples. This and a lot of other crazy spaces can be found on "Counterexamples in topology", by L. A. Steen and J. A. Seebach.

$\endgroup$
4
$\begingroup$

This I think would be the simplest example. Let $(Y,\tau_d)$ be an uncountable discrete space. Consider the space $X=Y\cup \{x\}$ with topology $\tau=\tau_d \cup \{X\}$. $(X,\tau)$ is clearly compact and, since all points in $Y\subset X$ are isolated, not separable.

The space I described is $\text{T}_0$ but not $\text{T}_1$. @Léo 's answer gives an example of a compact Hausdorff space that is not separable.

$\endgroup$
0
3
$\begingroup$

A space $X$ which is product of a collection of compact Hausdorff spaces $\{C_j\}_{j \in J}$ such that the cardinality $|J|$ of the collection is very large. $X$ is compact by Tychonoff's theorem. If $X$ is separable, i.e. if there is a countable dense subset $A$, then every point of $X$ is the limit of a sequence of points in $A$. The number of such sequences is at most the cardinality of the reals, and each sequence has at most one limit because $X$ is Hausdorff, and so $X$ has cardinality at most equal to the cardinality of the reals. But you can make $X$ have larger cardinality by choosing $|J|$ sufficiently large.

Edit: As pointed out in the comment of @Henno Brendsma, I should not have used convergence of sequences since that carries an additional assumption that each point has a countable neighborhood basis. Instead one can just use convergence of nets. The cardinality of all nets in a countable subset $A$ is bounded, so the cardinality of the set of limits of convergent nets in $A$ is bounded, and every point in the closure of $A$ is the limit of a convergent net in $A$.

$\endgroup$
2
  • 6
    $\begingroup$ It's not true that in every point of $X$ is the limit of points from a dense subset. This is true for first countable spaces, but large products of $[0,1]$, say, won't be first countable. In fact $[0,1]^I$ is separable iff $|I| \le |\mathbb{R}|$. So your idea does work, but for a different reason. $\endgroup$ Commented May 7, 2014 at 16:13
  • $\begingroup$ The cardinality of all subnets of a countable set is not bounded. You need to use filters to get the upper bound of $2^{2^{\aleph_0}}$ for the cardinality of a separable space (because a filter on $X$ is an element of $\mathcal{P}(\mathcal{P}(X))$). This is a known area of deficiency of Moore-Smith convergence. $\endgroup$ Commented Nov 11, 2019 at 10:20
0
$\begingroup$

Based on this question I asked recently, the one-point compactification of a non-separable metric space.

I specifically asked about the space of cadlag real-valued functions on $[0,1]$ under the supremum metric, but there are lots of other non-separable metric spaces out there, and other compactifications where the same argument applies.

The metric space is an open subset of its one-point compactification, and is not separable, so the one-point compactification is also not separable.

$\endgroup$

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.