The question is in the title:

Q1: Is there a topological space $X$ containing a copy of the real line and having the property that all the nonempty open subsets of $X$ are homeomorphic?

Let us say that $X$ is a homeomorphic open set space, or a hoss for short, if all the nonempty open subsets of $X$ are homeomorphic. Such spaces were asked about here, and N. de Rancourt's answer shows that if $D$ is infinite and discrete then $D^\omega$ is a hoss. It follows that every ``ultrametrizable'' space embeds in a hoss. (A space is called ultrametrizable if it is homeomorphic to an ultrametric space. Spaces of the form $D^\omega$ are themselves ultrametrizable, and every other ultrametrizable space embeds in one of this form.) This is just about all I know about hosses and spaces that embed in them -- any other information or insight is welcome.

Familiar examples of hosses include the space $\mathbb Q$ of rational numbers and the space $\mathbb R \setminus \mathbb Q$ of irrational numbers.


User bof has answered my question by finding a $T_1$ space $X$ containing the real line, and with the property that all nonempty open subsets of $X$ are homeomorphic. However, I still wonder if there is a Hausdorff space with this property:

Q2: Is there a Hausdorff space $X$ containing a copy of the real line and having the property that all the nonempty open subsets of $X$ are homeomorphic?

  • $\begingroup$ The other question you referred to asked about Hausdorff spaces, but you just wrote "topological space". Do you want a Hausdorff space, or will you accept a $\text T_1$-space satisfying your stated conditions? $\endgroup$ – bof Jul 30 '18 at 9:47
  • 1
    $\begingroup$ It isn't clear to me how to get any space satisfying the stated conditions, so I did not specify any further conditions, and would be interested to see a non-Hausdorff example. That being said, I would be even more interested to see a Hausdorff example. $\endgroup$ – Will Brian Jul 30 '18 at 10:44
  • 2
    $\begingroup$ @bof: I don't think it's a dumb question, but no, $\mathbb R \times \mathbb Q$ is not a hoss because it is not homeomorphic to $\mathbb R \times \mathbb Q$ minus a point. The following (admittedly contrived) property distinguishes the two spaces (true in the first, false in the second): Suppose $\langle x_n \rangle_n$ is a sequence converging to $x$ and $\langle y_n \rangle_n$ is a sequence converging to $y$; if $x_n$ and $y_n$ are connected by an arc for every $n$, then so are $x$ and $y$. $\endgroup$ – Will Brian Jul 31 '18 at 10:02
  • $\begingroup$ Could there be a subspace of the hyperreals that contains $\mathbb R$ and has similar properties to $\mathbb R \setminus \mathbb Q$? $\endgroup$ – R. van Dobben de Bruyn Jul 31 '18 at 17:54

A metrizable example can be constructed as follows. In the plane consider the subset $$\Xi:=\big\{(x,\tfrac{2k+1}{2^n}):k,n\in\mathbb Z,\;x\in\mathbb R\setminus \tfrac1{2^n}\mathbb Z\big\}.$$

It is clear that $\Xi$ contain (countably many) topological copies of the real line.

There are at least two ways of proving that any non-empty open subset of $\Xi$ is homeomorphic to $\Xi$. One is more geometric and is due to Volodymyr Mykhaylyuk (from Chernivtsi). He observed that for any open set $U\subset \Xi$ and any connected component $C$ of $U$ the interval $C$ has a base of clopen neighborhoods homeomorphic to the strip $\Xi\cap(\mathbb R\times(-\sqrt{2},\sqrt{2}))$ in $\Xi$. Then $U$ can be decomposed into countably many pairwise homeomorphic clopen sets and the same can be done with the space $\Xi$.

Another way is more global. Just to prove a characterization theorem for the space $\Xi$:

Theorem. A topological space $X$ is homeomorphic to the space $\Xi$ if and only if

1) $X$ is a metrizable space;

2) for any point $x\in X$ the connected component $C_x$ containing $x$ is homeomorphic to the real line;

3) the family $\mathcal C=\{C_x\}_{x\in X}$ of connected components of $X$ is countable;

4) for any point $x\in X$ and a neighborhood $O_x\subset X$ of $x$ there exists a non-empty set $V=\bigcup\{C\in \mathcal C:C\cap V\ne\emptyset\}$ in $O_x\setminus C_x$ such that $V$ is clopen in $X\setminus C_x$, $C_x\cup V$ is a neighborhood of $x$.

The proof of this characterization uses the back-and-forth argument: We enumerate the connected components and at the $n$-th step construct clopen neighborhoods of $n$-th intervals and establish the combinatorial correspondence between these neighborhoods. I will write down the details in a preprint (with many authors with whom I discussed this problem being on a Summer School in Carpathian mountains).

The characterization theorem implies the following corollary:

Corollary. Let $(C_n)_{n\in\omega}$ be a family of parwise disjoint curves in $\mathbb R^d$ such that each $C_n$ is homeomorphic to $\mathbb R$, is closed and nowhere dense in the union $X:=\bigcup_{n\in\omega}C_n$ and $\sum_{n=1}^\infty lenth(C_n)<\infty$. Then the space $X$ is homeomorphic to $\Xi$.

  • $\begingroup$ Thanks for this, Taras -- it is a clever idea. I will have to think a bit to convince myself that a back-and-forth argument will show that any nonempty subset of $\Xi$ is homeomorphic to $\Xi$. (I trust you're probably right, but I want to think through some of the details for myself.) $\endgroup$ – Will Brian Aug 2 '18 at 23:24

Proposition. Any topological space $X$ can be embedded in a topological space $Y$ with the property that every nonempty open subspace of $Y$ is homeomorphic to $Y.$ Moreover, if $X$ is a
$\text T_1$-space, it is possible to take $Y$ a $\text T_1$-space as well.

Proof. Let $\kappa$ be an infinite cardinal which is greater than or equal to the number of nonhomeomorphic open subspaces of $X,$ and let $\lambda=\kappa^+.$ (E.g., if $X=\mathbb R,$ take $\kappa=\aleph_0$ and $\lambda=\aleph_1.$)

Let $I$ be an index set with $|I|=\lambda.$ Choose disjoint topological spaces $U_i\ (i\in I)$ so that each $U_i$ is homeomorphic to some nonempty open subspace of $X$ and, for each nonempty open subspace $U$ of $X,$ $|\{i\in I:U_i\text{ is homeomorphic to }U\}|=\lambda.$

Let $Y=\bigcup_{i\in I}U_i$ have the following topology: A nonempty set $W\subseteq Y$ is open just in case $W\cap U_i$ is open in $U_i$ for all $i\in I,$ and $|\{i\in I:U_i\not\subseteq W\}|\le\kappa.$

  • 2
    $\begingroup$ In other words, the proposition says there is a Bonanza of Hosses. :-) $\endgroup$ – Todd Trimble Jul 30 '18 at 14:49

Another example of a $T_1$-space containing the real line and having all non-empty open sets homeomorphic can be constructed using some standard facts of Infinite-Dimensional Topology.

Namely it is known that each closed locally compact subset $Z$ of the Hilbert space $\ell_2$ is a $Z$-set in $\ell_2$, which implies that the complement $\ell_2\setminus Z$ is homeomorphic to $\ell_2$, being a contractible $\ell_2$-manifold.

Then the Hilbert space $\ell_2$ endowed with the topology $\tau$ consisting of all complements to closed locally compact sets has the required property: it contains a topological copy of the real line and has all non-empty open sets homeomorphic.


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.