If $f:[0,1] \to \mathbb{R}^n$ is any continuous map, then the image $f([0,1])$ is a compact, path-connected set, which is easy to show using some elementary topology.

My question is the converse:

Namely, if $K \subset \mathbb{R}^n$ is compact and path-connected, then does there exist a continuous map $f:[0,1] \to \mathbb{R}^n$ such that $K = f([0,1])$?

My attempt at the problem:

I have a hunch that it might be true.

For any $k \in \mathbb{N}$, there exists a continuous surjection $f:[0,1] \to [0,1]^k$, which can be realized with a space filling curve. Therefore, any finite-dimensional cube can be realized as the continuous image of $[0,1]$.

Let $K \subset \mathbb{R}^n$ be compact and path-connected. My friend suggested that it would suffice to show that $K$ has the structure of a CW-complex with a finite number of cells, and then use the fact that any finite-dimensional cube is realized as the continuous image of $[0,1]$. However, I don't know if this is true.

Edit: It turns out that the answer is even more interesting than I anticipated, and is provided by the HM theorem: http://en.wikipedia.org/wiki/Space-filling_curve#The_Hahn.E2.80.93Mazurkiewicz_theorem

  • 2
    $\begingroup$ This is false. Take the closure of the topologist's sine curve with the ends joined up by a curve. $\endgroup$ Apr 17, 2015 at 5:54
  • $\begingroup$ @Shalop can you share the source of the problem if you dont mind ;I am also looking for some problems in this subject $\endgroup$
    – Learnmore
    Apr 17, 2015 at 6:06
  • $\begingroup$ @learnmore: A few years ago Wendelin Werner and associates proved that the frontier of the trace of a Brownian Motion has Hausdorff dimension $\frac{4}{3}$. Then they concluded that therefore the trace contains paths of dimension $\frac{4}{3}$. But a "path" is something which can be parametrized by a continuous map from $[0,1] \to \mathbb{R}$, and how do we know that the Brownian frontier is such a set? I thought of the above problem while thinking about how to prove this. $\endgroup$
    – shalop
    Apr 17, 2015 at 6:13
  • $\begingroup$ @Shalop I mean that set but with an arc attaching to a point on the sine curve and a point on the y-axis, thus making it path-connected $\endgroup$ Apr 17, 2015 at 6:16
  • 2
    $\begingroup$ See also Hahn-Mazurkiewicz theorem and this post on MO. $\endgroup$ Apr 17, 2015 at 7:37

2 Answers 2


Here's a non-trivial example: the closure of the topologist's sine curve with the ends joined up.

If you want an explicit representation, take $$ \{(x,y) : 0< x\leq 1, y = \sin(1/x) \} \bigcup \{(0,y): -1\leq y\leq 1\} \bigcup \{(x,0): -1\leq x\leq 0\} \bigcup \{(-1,y): -2\leq y\leq 0 \} \bigcup\{(x,-2): -1\leq x \leq 1\} \bigcup \{(1,y): -2\leq y\leq \sin(1) \}. $$

  • $\begingroup$ Maybe the post would look a bit better if you used picture from Wikipedia or one of many other pictures which can be found online. In fact, in this post there also a picture where the ends are joined. $\endgroup$ Apr 17, 2015 at 7:30
  • $\begingroup$ As I have learned here, this space is also called Warsaw circle. $\endgroup$ Apr 17, 2015 at 7:33

Trivially, no. Let $K=\emptyset$. The empty set is closed as $\mathbb{R}^n$ is open, certainly bounded. And given any two points $x,y \in \emptyset$, there is a path connecting them. :)

  • $\begingroup$ Great answer, I think maybe to make the question more interesting, it should add the non-empty hypothesis so this answer doesn't apply. Although of course, I don't want to diminish your correct contribution. $\endgroup$ Apr 12, 2017 at 0:37

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.