15
$\begingroup$

The well-known Hahn-Mazurkiewicz theorem characterizes those nonempty Hausdorff spaces $X$ that admit a continuous surjection $\alpha: [0, 1] \to X$ from the closed unit interval: it is necessary and sufficient that $X$ be a compact, connected, locally connected metrizable space.

My Googling skills did not enable me to locate a nice way of characterizing Hausdorff spaces $X$ that admit a continuous surjection $(0, 1) \to X$ from an open interval. Is there one?

$\endgroup$
5
  • $\begingroup$ Two thoughts: (1) every space that is the image of $[0,1]$ is the image of $(0,1)$; (2) we can replace $(0,1)$ with $\mathbb{R}$. I wonder it could be as simple as asking that $X$ be a path-connected countable union of spaces satisfying Hahn-Mazurkowicz. $\endgroup$
    – Jeff Strom
    May 25, 2014 at 1:01
  • 1
    $\begingroup$ One such $X$ is the Warsaw circle, which may indicate the difficulty of getting a similarly nice characterization. $\endgroup$ May 25, 2014 at 1:01
  • $\begingroup$ @JeffStrom I wondered the same thing. $\endgroup$
    – Todd Trimble
    May 25, 2014 at 1:01
  • 1
    $\begingroup$ @JeffStrom: Any path-connected countable union of Hahn-Mazurkowicz spaces can certainly be covered by a map from $(0,1)$. Simply cover each one of the Hahn-Mazurkowicz subspaces one by one, interspersed with paths between their endpoints. $\endgroup$ May 25, 2014 at 1:04
  • $\begingroup$ @EricWofsey: I was busy typing this up below. $\endgroup$
    – Jeff Strom
    May 25, 2014 at 1:12

1 Answer 1

9
$\begingroup$

Call $A$ an HM-space if there is a continuous surjection $I\to A$ (where $I$ is the closed interval $[0,1]$). Note that if $A$ is an HM-space, then it is path-connected.

Theorem: If $X$ is path-connected, then the following are equivalent:

  1. $X = \bigcup_{n=1}^\infty A_n$ where each $A_n$ is an HM-space
  2. there is a continuous surjection $\mathbb{R} \to X$
  3. there is a continuous surjection $(0,\infty) \to X$
  4. there is a continuous surjection $[0,\infty) \to X$.

Proof: Clearly (2) and (3) are equivalent; let's prove (3) implies (1). If $f:(0,\infty) \to X$ is a continuous surjection, let $A_n = f([n-1,n])$. Then $A_n$ is evidently an HM-space and $X$ is the indicated countable union.

Now we show (1) implies (4). if $X = \bigcup_{n=1}^\infty A_n$ with each $A_n$ an HM-space, then we can choose surjective paths $\alpha_n: [2(n-1),2n-1]\to A_n$; write $x_n = \alpha_n(0)$ and $y_n= \alpha_n(1)$. Since $X$ is path-connected, we can find paths $\beta_n:[2n-1,2n]\to X$ from $y_n$ to $x_{n+1}$. Concatenating these paths gives the desired surjection $0,\infty) \to X$.

Finally, since there is a continuous surjection $\mathbb{R}\to [0,\infty )$, (4) implies (2).$\quad \square$

Just musing: there is a partial order on spaces defined by $X< Y$ if there is a continuous surjection $Y \to X$. Hahn-Mazurkowicz classifies $\{ X \mid X < I\}$. The equivalence of (2), (3) and (4) shows that half-open intervals are "equivalent" to open ones (i.e., both $(0,1)< [0,1)$ and $[0,1) < (0,1)$); and closed ones are different, being compact. Finally, note that a Peano space (i.e., HM-space) $X$ is "equivalent" in this sense to $I$ if and only if it is the domain of a nonconstant continuous function $f:X\to \mathbb{R}$.

$\endgroup$
7
  • 1
    $\begingroup$ Well, that was easy. Thanks! That might be the best one could hope for, but I'll wait a while to see if there are other responses before accepting. $\endgroup$
    – Todd Trimble
    May 25, 2014 at 1:26
  • 5
    $\begingroup$ Incidentally, I think the usual term for HM-space is Peano space. $\endgroup$
    – Todd Trimble
    May 25, 2014 at 1:29
  • $\begingroup$ Just to clarify, $I$ here denotes the closed unit interval? $\endgroup$ May 25, 2014 at 1:39
  • 1
    $\begingroup$ @NateEldredge That's right. $\endgroup$
    – Todd Trimble
    May 25, 2014 at 1:45
  • 5
    $\begingroup$ This characterization appears in arxiv.org/pdf/1404.5876.pdf (see Proposition 3.2). The author calls these spaces $\sigma$-Peano spaces. $\endgroup$ May 28, 2014 at 14:45

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.