7
$\begingroup$

Is there an infinite Hausdorff space $(X,\tau)$ with the following property?

If $x\neq y \in X$ and $f:\{x,y\}\to X$ is a map, then there is exactly one continuous function $f': X\to X$ such that $f'|_{\{x,y\}} = f$.

$\endgroup$
2
  • 2
    $\begingroup$ A few observations: (1) By uniqueness, the maps $f'$ are forced to be either homeomorphisms or constant maps, leading to the question whether spaces exists admitting only maps of this form. (2) If $X$ is (locally?) compact then $X$ cannot have non-trivial path-components by an application of Tietze's Extension Theorem. (3) The space $X$ will satisfy curious properties such as $\operatorname{Hom}(X \times X,X) \cong \operatorname{Hom}(X \sqcup X,X)$. $\endgroup$ Sep 12, 2018 at 15:09
  • $\begingroup$ @SofieVerbeek: Concerning your first observation, let me point out that such spaces do exist: see mathoverflow.net/questions/294511/…. $\endgroup$
    – Will Brian
    Sep 12, 2018 at 16:16

1 Answer 1

5
$\begingroup$

Using the technique of van Mill it is possible to prove the following

Theorem. There exists a subset $Z$ in the complex plane $\mathbb C$ such that

1) for any complex numbers $x,a,b,c,d\in Z$ with $a\ne b$ and $c\ne d$ we have $h_{a,b,c,d}(Z)\subset Z$ where $h_{a,b,c,d}(z):=c+\frac{d-c}{b-a}(z-a)$;

2) for any homeomorphism $h:Z\to Z$ there exist unique complex numbers $a,b,c,d\in Z$ with $a\ne b$ and $c\ne d$ such that $h=h_{a,b,c,d}$;

3) Under CH (more generally, under $\acute{\mathfrak n}=\mathfrak c$) for any non-constant continuous map $f:Z\to Z$ there exist unique points $a,b,c,d\in Z$ with $a\ne b$ and $c\ne d$ such that $f=h_{a,b,c,d}$.

Here by $\acute{\mathfrak n}$ denotes the smallest cardinality of an infinite cover of $[0,1]$ by pairwise disjoint closed sets. By the classical Sierpi$\acute{\mathfrak n}$ski Theorem, $\aleph_1\le \acute{\mathfrak n}\le\mathfrak c$. More information on the small uncountable cardinal $\acute{\mathfrak n}$ can be found in this MO-post.

Remark. Under $\acute{\mathfrak n}=\mathfrak c$, the space $Z$ from Theorem has the property required in the question of Dominic van der Zypen: for any distinct points $a,b\in Z$ and point $c,d\in Z$ the map $h_{a,b,c,d}$ is a unique continuous self-map $f:Z\to Z$ such that $f(a)=c$ and $f(b)=d$.

$\endgroup$
3
  • 1
    $\begingroup$ I love how there's CH lurking around the corner! $\endgroup$ Sep 13, 2018 at 7:19
  • 1
    $\begingroup$ Incredible, that you came up with this answer, and as @SofieVerbeek says, I would never have thought that CH comes into the picture! $\endgroup$ Sep 13, 2018 at 7:31
  • $\begingroup$ @SofieVerbeek I admit that there exists also a ZFC-example, constructed with the help of the Cook continuum, but the construction has to be more complicated: inductively we should glue Cook continua by pairs of points and then iterate this construction, etc. $\endgroup$ Sep 13, 2018 at 7:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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