3
$\begingroup$

Let $$ \mathbb D=\{z\in\mathbb C\ |\ |z|<1\} $$ be the open unit disk in $\mathbb C$. It is well known that an open (nonempty) set $U\subseteq\mathbb C$ is simply connected if and only if it is homeomorphic to the unit disk. One can show that this is equivalent to the fact that there is a continuous injective function $f:\mathbb D\to\mathbb C$ such that $f(\mathbb D)=U$.

Therefore, my question is: If $U$ is a (nonempty) open and connected subset of $\mathbb C$, is there always a continuous (but not necessarily injective) function $f:\mathbb D\to\mathbb C$ such that $f(\mathbb D)=U$? It sufficices to assume $U\subseteq\mathbb D$.

If not, how do images of such functions look like?

I know that any compact connected and locally connected subset of $\mathbb C$ is the image of a continuous function defined on $[0,1]$ and hence also the image of a continuous function on $\overline{\mathbb D}$. However, the closure of an open and connected set does not have to be locally connected.

I believe that the answer to my first question is negative (although I hope that it is not ;) ).

Any help is highly appreciated. Thank you very much in advance!

$\endgroup$
2
2
$\begingroup$

The Hahn-Mazurkiewicz theorem, as pointed out by Mathlover in the comments, is enough to show my educated guess actually works. In particular, it implies that closed balls in $\Bbb{C}$ are the continuous image of a compact interval. We will need a lemma:

Lemma $\quad$ Suppose $\{C_n\}_{n=0}^\infty$ is a countable collection of subspaces of a topological space $X$ (e.g. $\Bbb{C}$) which are continuous images of a compact interval and $$C := \bigcup_{n=0}^\infty C_n$$ is path-connected. Then $C$ is the continuous image of $[0, \infty)$.

Proof. Consider the intervals $I_n := [2n, 2n + 1]$ and $J_n := [2n + 1, 2n + 2]$ for $n \in \Bbb{N} \cup \{0\}$. Let $\mathfrak{i}_n : I_n \to C_n$ be surjective and continuous, and let $\mathfrak{j}_n : J_n \to C$ be a continuous function such that $\mathfrak{j}_n(2n + 1) = \mathfrak{i}_n(2n + 1)$ and $\mathfrak{j}_n(2n + 2) = \mathfrak{i}_{n+1}(2n + 2)$. Then, the map $$\phi : [0, \infty) = \bigcup_{n=0}^\infty (I_n \cup J_n) \to C$$ defined by $$\phi(x) = \begin{cases} \mathfrak{i}_n(x) & \text{if } \exists n \in \Bbb{N} \cup \{0\} : 2n \le x < 2n + 1 \\ \mathfrak{j}_n(x) & \text{if } \exists n \in \Bbb{N} \cup \{0\} : 2n + 1 \le x < 2n + 2. \end{cases}$$ This function is made up of countably infinitely many continuous pieces, joined at common points, making the function continuous. It's clearly surjective (even just restricting to $\bigcup I_n$), hence $C$ is the continuous image of $[0, \infty)$ under $\phi$. $\square$

Note that every ball in $\Bbb{C}$ is a countable union of closed balls, and recall that open subsets of $\Bbb{C}$ are a countable union of open balls. Also, in a locally path-connected space, open and connected implies path-connected. Thus, $U$ is the path-connected countable union of closed balls, so by the lemma, it is a continuous image of $[0, \infty)$.

So, to wrap up the proof, as in my comment, simply project the unit disk of $\Bbb{C}$ onto $(-1, 1)$, which is homeomorphic to $\Bbb{R}$. You can simply keep the map constant for points in $(-\infty, 0]$, and then use $[0, \infty)$ to map continuously onto $U$.

$\endgroup$
1
  • $\begingroup$ This is amazing, thank you! $\endgroup$
    – sranthrop
    Feb 14 at 13:45
2
$\begingroup$

One can prove a stronger claim: given a nonempty open connected $U\subseteq \mathbb{C}$, there's a holomorphic $\varphi:\mathbb{D}\to \mathbb{C}$ such that $\varphi(\mathbb{D})=U$.

First, let $U\neq \mathbb{C},\neq \mathbb{C}-\{z_0\}$. Thanks to the Riemann uniformization theorem, every such open connected subset of $\mathbb{C}$ is hyperbolic, i.e. has $\mathbb{D}$ as its universal cover. Thus there exists a surjective holomorphic map $\mathbb{D}\to U$. Note that there's a surjective holomorphic map $(\exp(z)+z_0)$ from $\mathbb{C}$ to $\mathbb{C}-\{z_0\}$, and thus it suffices to prove the result for $\mathbb{C}$. Now, let $f(z)=z^3-1$. It is easy to see that $f:\mathbb{C}-\{1,\exp(2\pi i/3)\}\twoheadrightarrow\mathbb{C}$. Precomposing with the covering map $\pi:\mathbb{D}\to \mathbb{C}-\{1,\exp(2\pi i/3)\}$, we get the claim.

One can also explore "how much" this function fails to be injective. Indeed, the uniformization theorem implies that, for $U\neq\mathbb{C},U\neq \mathbb{C}-\{z_0\}$, we have $U\simeq \mathbb{D}/\Gamma$, where $\Gamma$ is a discrete subgroup of $Aut(\mathbb{D})$ (complex automorphisms) isomorphic to $\pi_1(U)$. Thus two elements $z_1,z_2$ have the same image iff $\exists \gamma\in \Gamma:\gamma(z_1)=z_2$ (note that with $U$ simply connected we get injectivity).

$\endgroup$
1
  • 1
    $\begingroup$ Beautiful answer, thank you very much! $\endgroup$
    – sranthrop
    Feb 14 at 13:46

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.