12
$\begingroup$

Are there a closed subspace $C$ of $\mathbb{R}^2$ and a continuous, bijective function $f:C\to\mathbb{R}^2$ that is not open?

I cannot think about a continuous and bijective function $f:C\to\mathbb{R}^2$. My best candidate was the set $C=[0,+\infty)\times[0,+\infty)$ and the function $$f(s,t)=\Bigg(s\cos\bigg(2\pi\frac{t}{t+1}\bigg),s\sin\bigg(2\pi\frac{t}{t+1}\bigg)\Bigg),$$ but this is not injective in $\{0\}\times[0,+\infty)$.

The only hint that I can think about is that $C$ cannot be compact, because $f$ is bijective. Therefore, $C$ cannot be bounded.

An idea that I've been trying is to work in the sphere $S^2$. There, we can think of $C$ as a subset of $S^2$ with N=(0,0,1) as a limit point such that $C\cup\{N\}$ is compact.

Then the function $f$ takes $C$ onto $S^2-\{N\}$. But in this way I only see open functions.

Any suggestions?

$\endgroup$
  • $\begingroup$ I guess not open into $f(C)$. I post an example next. $\endgroup$ – Jesus RS Mar 27 '15 at 8:46
  • $\begingroup$ To be more precise, since $f:C\to\mathbb{R}^2$ has to be bijective, we have to find $f$ that is not open into $f(C)=\mathbb{R}^2$, don't forgetting that $C$ is closed. $\endgroup$ – Chilote Mar 27 '15 at 13:28
5
$\begingroup$

Let $C\subset\mathbb R^2$ be the line $y=0$, and $$f(x,0)=\big(\frac{x^3}{1+x^4},\frac{x}{1+x^4}\big).$$ Then $f(C)$ is a lemniscate with two branches crossing at the origin $(0,0)=f(0,0)$.

lemniscate

The mapping $f:C\to f(C)$ is not open, because an interval around $x=0$ in $C$ maps into a single branch (the vertical one) of $f(C)$, missing the other.

This idea can be used further, as in the following pictures.

bijection

One starts with a closed band $A\subset\mathbb R^2$ and modifies it homeomorphically to have a band with only to segments in the boundary. Then bends it a la lemniscate, which gives something like a disc with two holes, which is homeomorphic to a disc minus two points, then to the plane minus two points. Finally the non open bijection $A\to\mathbb R^2\setminus\{\alpha,\beta\}$ extends to a bijection $C\to\mathbb R^2$: $C$ is $A$ plus two separated points $a,b$ and the extension, continuous but not open, maps $a\mapsto\alpha, b\mapsto\beta$.

The next pictures are a variation of the above argument that accompasses @Chilote smart answer:

enter image description here

Here we see the same topologist strategy to produce a bijection $A\to\mathbb R^2\setminus\{\alpha\}$.

Finally, a remark to add context to the problem. Suppose we can define a homeomorphism $f:C\to\mathbb R^n$ from a closed subset $C\subset\mathbb R^n$ onto $\mathbb R^n$. Then it extends to the Alexandroff one point compactifications $C^*\sim{\mathbb R^n}^*$. Now, that of $\mathbb R^n$ is the sphere $\mathbb S^n$ and that of $C$ is its closure in $\mathbb S^n$. But a sphere can't be homeo to a proper subset, hence the starting $f$ cannot exist. Consequently, the problem is

Is there a continuous bijection $C\to \mathbb R^2$?

Indeed, none that exists can be open, because then it would be a homeo.

$\endgroup$
  • $\begingroup$ Thanks for the example, but I am looking for a function such that $f(C)=\mathbb{R}^2$. That is why I ask $f:C\to\mathbb{R}^2$ to be bijective. $\endgroup$ – Chilote Mar 27 '15 at 13:18
  • 1
    $\begingroup$ I think that the same idea works, but explicit equations are out of rich. In any case it's very interesting. I'm trying to make some pictures (for a topologist would be a proof) and edit my answer with them. $\endgroup$ – Jesus RS Mar 27 '15 at 16:32
1
$\begingroup$

The answer of @Jesus RS is brilliant! It shows the more important point of the problem: $C$ does not have to be connected.

If we have a closed subspace $C$ of $\mathbb{R}^2$ and a continuos and bijective function $g:C\to\mathbb{R}^2\setminus\{p\}$, where $p\in\mathbb{R}^2$, then we just have to consider a point in the complement of $C$, say $q$, and the function $$f:C\cup\{q\}\to\mathbb{R}^2$$ \begin{equation} f(z)= \begin{cases} &g(z)&,&x\neq q\\ &p&,&x=q \end{cases} \end{equation} will be continuous and bijective, but is not open because $\{q\}$ is open in $C\cup\{q\}$ but $f(\{q\})=\{p\}$ is not open in $\mathbb{R}^2$.

Now it is very easy to show an explicit function satisfying these conditions. The function $f:\big([0,+\infty)\times\mathbb{R}\big)\cup\{(-1,0)\}\to\mathbb{R}^2$ defined by \begin{equation} f(x,y)= \begin{cases} &\bigg(e^y\cos\big(\frac{2\pi x}{x+1}\big),e^y\sin\big(\frac{2\pi x}{x+1}\big)\bigg)&,&(x,y)\neq (-1,0)\\ &(0,0)&,&(x,y)=(-1,0) \end{cases} \end{equation} is continuous, bijective and not open, where the domain is a closed subspace of $\mathbb{R}^2$.

$\endgroup$
  • $\begingroup$ Nice example/explanation. $\endgroup$ – Mike F Mar 27 '15 at 20:57
  • 1
    $\begingroup$ Note that, even discounting the behavior at the isolated point $(-1,0) \in C$, we have problems with open-ness. For instance, $[0,1) \times \mathbb{R}$ is open in $C$, but its image by $f$ is not open in $\mathbb{R}^2$. $\endgroup$ – Mike F Mar 27 '15 at 21:03
  • 1
    $\begingroup$ But the question remains whether $f$ is open if $C$ is connected. $\endgroup$ – Jesus RS Mar 27 '15 at 23:42

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.