12
$\begingroup$

A map is called proper if the pre-image of a compact set is again compact.

In the Differential Forms in Algebraic Topology by Bott and Tu, they remark that the image of a proper map $f: \mathbb{R}^n \to \mathbb R^m$ is closed, adding the comment "(why?)".

I can think of a simple proof in this case for continuous $f$:

If the image is not closed, there is a point $p$ that does not belong to it and a sequence $p_n \in f(\mathbb R^n)$ with $p_n \to p$. Since $f$ is proper $f^{-1}(\overline {B_\delta(p)})$ is compact for any $\delta$. Let $x_n$ be any point in $f^{-1}(p_n)$ and wlog $x_n \in f^{-1}(\overline{B_\delta(p)})$. Since in $\mathbb{R}^n$ compact and sequentially compact are equivalent, there exists a convergent subsequence $x_{n_k}$ of $x_n$. From continuity of $f$: $f(x_{n_k}) \to f(x)$ for some $x$. But $f(x_{n_k})=p_{n_k} \to p$ which is not supposed to be in the image and this gives a contradiction.

My problem is that this proof is too specific to $\mathbb{R}^n$ and uses arguments from basic analysis rather than general topology.

So the question is for what spaces does it hold that the image of a proper map is closed, how does the proof work, and is it necessary to pre-suppose continuity?

$\endgroup$
  • 3
    $\begingroup$ Often map already implies continuity. I'd check the text for this. $\endgroup$ – Henno Brandsma Jan 8 '16 at 12:14
29
$\begingroup$

First of all the definition of a proper map assumes continuity by convention (I have not come across texts that say otherwise)

Secondly, here is a more general result -

Lemma : Let $f:X\rightarrow Y$ be a proper map between topological spaces $X$ and $Y$ and let $Y$ be locally compact and Hausdorff. Then $f$ is a closed map.

Proof : Let $C$ be a closed subset of $X$. We need to show that $f(C)$ is closed in $Y$ , or equivalently that $Y\setminus f(C)$ is open.

Let $y\in Y\setminus f(C)$. Then $y$ has an open neighbourhood $V$ with compact closure. Then $f^{-1}(\bar{V})$ is compact.

Let $E=C\cap f^{-1}(\bar{V})$ . Then clearly $E$ is compact and hence so is $f(E)$. Since $Y$ is Hausdorff $f(E)$ is closed.

Let $U=V\setminus f(E)$. Then $U$ is an open neighbourhood of $y$ and is disjoint from $f(C)$.

Thus $Y\setminus f(C)$ is open. $\square$

I hope this helps.


EDIT: To clarify the statement $U$ is disjoint from $f(C)$ -

Suppose $z\in U\cap f(C)$ Then there exists a $c\in C$ such that $z=f(c)$. This means $c\in f^{-1}(U)\subseteq f^{-1}(V)\subseteq f^{-1}(\bar V)$. So $c\in C\cap f^{-1}(\bar V)=E$. So $z=f(c)\in f(E)$ which is a contradiction as $z\in U$.

$\endgroup$
  • $\begingroup$ Why is $U$ disjoint from $f(C)$? From your definition it is clear that $E \subseteq C$. So $f(E) \subseteq f(C)$. Hence $V \setminus f(C) \subseteq V \setminus f(E) = U$. If the containment is proper then $U$ may contain some element of $f(C)$. Who knows that? Isn't it so @R_D? $\endgroup$ – Dbchatto67 Jan 25 '19 at 10:26
  • $\begingroup$ Is it fine now? @Dbchatto67 $\endgroup$ – R_D Jan 26 '19 at 3:34
  • $\begingroup$ Yeah @R_D it's now absolutely fine. Thanks so much. $\endgroup$ – Dbchatto67 Jan 26 '19 at 6:04
10
$\begingroup$

One may generalize the result in R_D's answer even further:

A proper map $f:X\to Y$ to a compactly generated Hausdorff space is a closed map (A space $Y$ is called compactly generated if any subset $A$ of $Y$ is closed when $A\cap K$ is closed in $K$ for each compact $K\subseteq Y$).
Proof: Let $C\subseteq X$ be closed, and let $K$ be a compact subspace of $Y$. Then $f^{-1}(K)$ is compact, and so is $f^{-1}(K)\cap C =: B$. Then $f(B)=K\cap f(C)$ is compact, and as $Y$ is Hausdorff, $f(B)$ is closed. Since $Y$ is compactly generated, $f(C)$ is closed in $Y$.

A locally compact space $Y$ is compactly generated: If $A\subset Y$ intersects each compact set in a closed set, and if $y\notin A$, then $A$ intersects the compact neighborhood $K$ of $y$ in a closed set $C$. Now $K\setminus C$ is a neighborhood of $y$ disjoint from $A$, hence $A$ is closed.

$\endgroup$

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.