Suppose $X$, and $Y$ are topological spaces, and that they are homeomorphic. If $X$ is a metric space must $Y$ be as well?

Here is my thought process.

If $X$ is a metric space then there exists some distance function on $X$ given by $d:X\times X\rightarrow\Bbb R$. Let $\varphi:X\rightarrow Y$ be a homeomorphism between $X$ and $Y$.

Then for any distinct points $y_1,y_2\in Y$ we can determine a "distance" between them by $d(\varphi^{-1}(y_1),\varphi^{-1}(y_2))\in\Bbb R$

To be a metric space however, I need some way of combining this idea of passing the points back to $X$ from $Y$ in a nice way, and then applying the distance function on $X$. Would it be as simple as defining $d'(y_1,y_2) = d(\varphi^{-1}(y_1),\varphi^{-1}(y_2))$?

If not is there a way to do that? Is this enough to say that $Y$ is a metric space? Or am I just out of my mind?

  • 3
    $\begingroup$ I don't really think this question makes sense. However, a space that is homeomorphic to a metric space is called a metrisable space. $\endgroup$ – Lord Shark the Unknown Sep 11 '18 at 21:03

If $(X, d_X)$ is a metric space then the homeomorphic $Y$ is a so-called metrisable space: a space that can be given a metric that induces its topology. A metric space comes with a pre-given metric (and has an induced topology by this metric), and is a different type of structure.

The metric is indeed a transportation of the given one on $X$: $\phi: X \to Y$ (the homeomorphism) has a well defined continuous inverse (I'll call it $\psi: Y \to X$ so that $\psi \circ \phi = 1_X$ and $\phi \circ \psi = 1_Y$), and we can define $$d_Y: Y \times Y \to \mathbb{R} \text{ by: } d_Y(y_1,y_2) = d_X(\psi(y_1), \psi(y_2))$$

One easily checks the axioms for a metric for $d_Y$ and by bijectiveness and the definitions we get that $\phi[B_{d_X}(x, r)] = B_{d_Y}(\phi(x), r)$ for all $x \in X, r>0$ etc. so that balls in the $d_X$ metric get mapped to balls in the $d_Y$-metric showing (with a little thought) that indeed open balls under $d_Y$ are open and form a base for the topology of $Y$, so that $Y$ is metrisable.


You construction can be done regardless of $\varphi$ being a homeomorphism: for any bijection you will have a metric structure on $Y$. What might be interesting to verify is if the topology induced by the metric $d' = d(\varphi^{-1}(-),\varphi^{-1}(-))$ coincides with the original topology on $Y$. Let's describe the open balls of $d'$: let $y \in Y$ and $\varepsilon > 0$ . Now,

$$ B'_\varepsilon(y) := \{x:d'(x,y) < \varepsilon\} = \{x : d(\varphi^{-1}(x), \varphi^{-1}(y)) < \varepsilon\} = \varphi(B_\varepsilon(\varphi^{-1}(y))). $$

Since $B_\varepsilon(\varphi^{-1}(y))$ is a ball in $X$ and $\varphi$ is homeo, $B'_\varepsilon(y)$ will be open on $Y$ with the original topology. In the same fashion, if $U \subseteq Y$ is open for the original topology,

$$ U = \varphi(\varphi^{-1}(U)) = \varphi(\bigcup_{i \in I}B_{r_i}(x_i)) = \bigcup_{i \in I}\varphi(B_{r_i}(x_i)) = \bigcup_{i \in I}B'_{r_i}(\phi(x_i)) $$

and so $U$ is open for the metric topology. Here we use that $\varphi^{-1}(U)$ is an open set of $X$ and so it is a union of open balls.


I would say is as simple as that, since $\varphi$ is a homeomorphism it seems straightforward to check the metric space axioms, you may want to be carefull with the triangular inequality though, but at a glimpse, it is straightforward. Regards.


Given any sets A and B and a bijection $\varphi$ between them, any structure in one set induces an isomorhpic structure in the other; if A has a metric, or a topology, or an additive structure, etc., then a corresponding structure can be defined on B. For instance, if we define $b_1+b_2$ as being $\varphi(\varphi^{-1}(b_1)+\varphi^{-1}(b_2))$, then $\varphi$ will be a group isomorphism with respect to addition in A and addition in B.

Strictly speaking, a metric space is not a set; it's a tuple $(S,d)$ where $S$ is a set and $d$ is a distance function. So neither $X$ nor $Y$ are metric spaces, but $X$ along with its distance function is a metric space, and $Y$ along with the metric induced by $\varphi$ is a metric space. The only question I see that is remotely non-trival is whether the topology in $Y$ induced by the metric in $Y$ induced by the metric in $X$ is the same as the original topology in $Y$. Since the topology in $Y$ is isomorphic to the topology in $X$, and the topology in $X$ is presumably the topology induced by its metric, the two topologies in $Y$ would indeed be the same.


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.