Lemma 5.17.1 (Tube lemma). Let $X$ and $Y$ be topological spaces. Let $A \subset X$ and $B \subset Y$ be quasi-compact subsets. Let $A \times B \subset W \subset X \times Y$ with $W$ open in $X \times Y$. Then there exists opens $A \subset U \subset X$ and $B \subset V \subset Y$ such that $U \times V \subset W$.

## 5.17 Characterizing proper maps

We include a section discussing the notion of a proper map in usual topology. We define a continuous map of topological spaces to be *proper* if it is universally closed and separated. Although this matches well with the definition of a proper morphism in algebraic geometry, this is different from the definition in Bourbaki. With our definition of a proper map of topological spaces, the proper base change theorem (Cohomology, Theorem 20.18.2) holds without any further assumptions. Furthermore, given a morphism $f : X \to Y$ of finite type schemes over $\mathbf{C}$ one has: $f$ is proper as a morphism of schemes if and only if the continuous map $f : X(\mathbf{C}) \to Y(\mathbf{C})$ on $\mathbf{C}$-points with the classical topology is proper. This is explained in [Exp. XII, Prop. 3.2(v), SGA1] which also has a footnote pointing out that they take properness in topology to be Bourbaki's notion with separatedness added on.

We find it useful to have names for three distinct concepts: separated, universally closed, and both of those together (i.e., properness). For a continuous map $f : X \to Y$ of locally compact Hausdorff spaces the word “proper” has long been used for the notion “$f^{-1}($compact$) =$ compact” and this is equivalent to universal closedness for such nice spaces. In fact, we will see the preimage condition formulated for clarity using the word “quasi-compact” is equivalent to universal closedness in general, if one includes the assumption of the map being closed. See also [Exercises 22-26 in Chapter II, LangReal] but beware that Lang uses “proper” as a synonym for “universally closed”, like Bourbaki does.

**Proof.**
For every $a \in A$ and $b \in B$ there exist opens $U_{(a, b)}$ of $X$ and $V_{(a, b)}$ of $Y$ such that $(a, b) \in U_{(a, b)} \times V_{(a, b)} \subset W$. Fix $b$ and we see there exist a finite number $a_1, \ldots , a_ n$ such that $A \subset U_{(a_1, b)} \cup \ldots \cup U_{(a_ n, b)}$. Hence

Thus for every $b \in B$ there exists opens $U_ b \subset X$ and $V_ b \subset Y$ such that $A \times \{ b\} \subset U_ b \times V_ b \subset W$. As above there exist a finite number $b_1, \ldots , b_ m$ such that $B \subset V_{b_1} \cup \ldots \cup V_{b_ m}$. Then we win because $A \times B \subset (U_{b_1} \cap \ldots \cap U_{b_ m}) \times (V_{b_1} \cup \ldots \cup V_{b_ m})$. $\square$

The notation in the following definition may be slightly different from what you are used to.

Definition 5.17.2. Let $f : X\to Y$ be a continuous map between topological spaces.

We say that the map $f$ is

*closed*if the image of every closed subset is closed.We say that the map $f$ is

*Bourbaki-proper*^{1}if the map $Z \times X\to Z \times Y$ is closed for any topological space $Z$.We say that the map $f$ is

*quasi-proper*if the inverse image $f^{-1}(V)$ of every quasi-compact subset $V \subset Y$ is quasi-compact.We say that $f$ is

*universally closed*if the map $f': Z \times _ Y X \to Z$ is closed for any continuous map $g: Z \to Y$.We say that $f$ is

*proper*if $f$ is separated and universally closed.

The following lemma is useful later.

Lemma 5.17.3. A topological space $X$ is quasi-compact if and only if the projection map $Z \times X \to Z$ is closed for any topological space $Z$.

**Proof.**
(See also remark below.) If $X$ is not quasi-compact, there exists an open covering $X = \bigcup _{i \in I} U_ i$ such that no finite number of $U_ i$ cover $X$. Let $Z$ be the subset of the power set $\mathcal{P}(I)$ of $I$ consisting of $I$ and all nonempty finite subsets of $I$. Define a topology on $Z$ with as a basis for the topology the following sets:

All subsets of $Z\setminus \{ I\} $.

For every finite subset $K$ of $I$ the set $U_ K := \{ J\subset I \mid J \in Z, \ K\subset J \} )$.

It is left to the reader to verify this is the basis for a topology. Consider the subset of $Z \times X$ defined by the formula

If $(J, x) \not\in M$, then $x \in U_ i$ for some $i \in J$. Hence $U_{\{ i\} } \times U_ i \subset Z \times X$ is an open subset containing $(J, x)$ and not intersecting $M$. Hence $M$ is closed. The projection of $M$ to $Z$ is $Z-\{ I\} $ which is not closed. Hence $Z \times X \to Z$ is not closed.

Assume $X$ is quasi-compact. Let $Z$ be a topological space. Let $M \subset Z \times X$ be closed. Let $z \in Z$ be a point which is not in $\text{pr}_1(M)$. By the Tube Lemma 5.17.1 there exists an open $U \subset Z$ such that $U \times X$ is contained in the complement of $M$. Hence $\text{pr}_1(M)$ is closed. $\square$

Remark 5.17.4. Lemma 5.17.3 is a combination of [I, p. 75, Lemme 1, Bourbaki] and [I, p. 76, Corollaire 1, Bourbaki].

Theorem 5.17.5. Let $f: X\to Y$ be a continuous map between topological spaces. The following conditions are equivalent:

The map $f$ is quasi-proper and closed.

The map $f$ is Bourbaki-proper.

The map $f$ is universally closed.

The map $f$ is closed and $f^{-1}(y)$ is quasi-compact for any $y\in Y$.

**Proof.**
(See also the remark below.) If the map $f$ satisfies (1), it automatically satisfies (4) because any single point is quasi-compact.

Assume map $f$ satisfies (4). We will prove it is universally closed, i.e., (3) holds. Let $g : Z \to Y$ be a continuous map of topological spaces and consider the diagram

During the proof we will use that $Z \times _ Y X \to Z \times X$ is a homeomorphism onto its image, i.e., that we may identify $Z \times _ Y X$ with the corresponding subset of $Z \times X$ with the induced topology. The image of $f' : Z \times _ Y X \to Z$ is $\mathop{\mathrm{Im}}(f') = \{ z : g(z) \in f(X)\} $. Because $f(X)$ is closed, we see that $\mathop{\mathrm{Im}}(f')$ is a closed subspace of $Z$. Consider a closed subset $P \subset Z \times _ Y X$. Let $z \in Z$, $z \not\in f'(P)$. If $z \not\in \mathop{\mathrm{Im}}(f')$, then $Z \setminus \mathop{\mathrm{Im}}(f')$ is an open neighbourhood which avoids $f'(P)$. If $z$ is in $\mathop{\mathrm{Im}}(f')$ then $(f')^{-1}\{ z\} = \{ z\} \times f^{-1}\{ g(z)\} $ and $f^{-1}\{ g(z)\} $ is quasi-compact by assumption. Because $P$ is a closed subset of $Z \times _ Y X$, we have a closed $P'$ of $Z \times X$ such that $P = P' \cap Z \times _ Y X$. Since $(f')^{-1}\{ z\} $ is a subset of $P^ c = P'^ c \cup (Z \times _ Y X)^ c$, and since $(f')^{-1}\{ z\} $ is disjoint from $(Z \times _ Y X)^ c$ we see that $(f')^{-1}\{ z\} $ is contained in $P'^ c$. We may apply the Tube Lemma 5.17.1 to $(f')^{-1}\{ z\} = \{ z\} \times f^{-1}\{ g(z)\} \subset (P')^ c \subset Z \times X$. This gives $V \times U$ containing $(f')^{-1}\{ z\} $ where $U$ and $V$ are open sets in $X$ and $Z$ respectively and $V \times U$ has empty intersection with $P'$. Then the set $V \cap g^{-1}(Y-f(U^ c))$ is open in $Z$ since $f$ is closed, contains $z$, and has empty intersection with the image of $P$. Thus $f'(P)$ is closed. In other words, the map $f$ is universally closed.

The implication (3) $\Rightarrow $ (2) is trivial. Namely, given any topological space $Z$ consider the projection morphism $g : Z \times Y \to Y$. Then it is easy to see that $f'$ is the map $Z \times X \to Z \times Y$, in other words that $(Z \times Y) \times _ Y X = Z \times X$. (This identification is a purely categorical property having nothing to do with topological spaces per se.)

Assume $f$ satisfies (2). We will prove it satisfies (1). Note that $f$ is closed as $f$ can be identified with the map $\{ pt\} \times X \to \{ pt\} \times Y$ which is assumed closed. Choose any quasi-compact subset $K \subset Y$. Let $Z$ be any topological space. Because $Z \times X \to Z \times Y$ is closed we see the map $Z \times f^{-1}(K) \to Z \times K$ is closed (if $T$ is closed in $Z \times f^{-1}(K)$, write $T = Z \times f^{-1}(K) \cap T'$ for some closed $T' \subset Z \times X$). Because $K$ is quasi-compact, $K \times Z\to Z$ is closed by Lemma 5.17.3. Hence the composition $Z \times f^{-1}(K)\to Z \times K \to Z$ is closed and therefore $f^{-1}(K)$ must be quasi-compact by Lemma 5.17.3 again. $\square$

Remark 5.17.6. Here are some references to the literature. In [I, p. 75, Theorem 1, Bourbaki] you can find: (2) $\Leftrightarrow $ (4). In [I, p. 77, Proposition 6, Bourbaki] you can find: (2) $\Rightarrow $ (1). Of course, trivially we have (1) $\Rightarrow $ (4). Thus (1), (2) and (4) are equivalent. The equivalence of (3) and (4) is [Chapter II, Exercise 25, LangReal].

Lemma 5.17.7. Let $f : X \to Y$ be a continuous map of topological spaces. If $X$ is quasi-compact and $Y$ is Hausdorff, then $f$ is universally closed.

**Proof.**
Since every point of $Y$ is closed, we see from Lemma 5.12.3 that the closed subset $f^{-1}(y)$ of $X$ is quasi-compact for all $y \in Y$. Thus, by Theorem 5.17.5 it suffices to show that $f$ is closed. If $E \subset X$ is closed, then it is quasi-compact (Lemma 5.12.3), hence $f(E) \subset Y$ is quasi-compact (Lemma 5.12.7), hence $f(E)$ is closed in $Y$ (Lemma 5.12.4).
$\square$

Lemma 5.17.8. Let $f : X \to Y$ be a continuous map of topological spaces. If $f$ is bijective, $X$ is quasi-compact, and $Y$ is Hausdorff, then $f$ is a homeomorphism.

**Proof.**
It suffices to prove $f$ is closed, because this implies that $f^{-1}$ is continuous. If $T \subset X$ is closed, then $T$ is quasi-compact by Lemma 5.12.3, hence $f(T)$ is quasi-compact by Lemma 5.12.7, hence $f(T)$ is closed by Lemma 5.12.4.
$\square$

## Post a comment

Your email address will not be published. Required fields are marked.

In your comment you can use Markdown and LaTeX style mathematics (enclose it like `$\pi$`

). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.

## Comments (2)

Comment #465 by Hanno on

Comment #483 by Johan on