7

Let $X$ and $Y$ be topological spaces such that $Y$ is compact, and let $f \colon X \to Y$ be a closed, surjective, and continuous map such that, for each $y \in Y$, the inverse image $f^{-1} ( \ \{y \} \ )$ is compact.

Then how to show that $X$ is compact also?

My effort: (A Mere Write-Up / Presentation Of The Proof Based On The Answers Below)

Let $\mathscr{A}$ be an open covering of $X$. Then, for each $y \in Y$, the collection $\mathscr{A}$ is a covering of $f^{-1} ( \ \{y\} \ )$ by sets open in $X$, and since $f^{-1} ( \ \{y\} \ )$ is compact, some finite subcollection of $\mathscr{A}$, say $\mathscr{A}_y$, also covers $f^{-1} ( \ \{y\} \ )$. That is, for each $y \in Y$, there exists a finite subcollection $\mathscr{A}_y$ of $\mathscr{A}$ such that $$ f^{-1} ( \ \{ y \} \ ) \subset \bigcup_{A \in \mathscr{A}_y } A. \tag{0} $$ For each $y \in Y$, let us put $$ U_y \colon= \bigcup_{A \in \mathscr{A}_y } A, \qquad \mbox{ and } \qquad C_y \colon= X \setminus U_y. \tag{A} $$ Then $U_y$ is an open set in $X$ and $C_y$ is a closed set; moreover we also have $$ f^{-1} ( \ \{ y \} \ ) \subset U_y, \tag{1} $$ and so $$ f^{-1} ( \ \{ y \} \ ) \cap C_y = \emptyset. \tag{2} $$ Now as $f \colon X \to Y$ is a closed map and as $C_y$ is a closed set in $X$, so the image set $f \left( C_y \right)$ is a closed set in $Y$. Let us put $$ W_y \colon= Y \setminus f \left( C_y \right). \tag{B} $$ Then $W_y$ is an open set in $Y$.

Now if $x \in f^{-1}\left( W_y \right)$, then by definition $x \in X$ is such that $f(x) \in W_y$, which implies that $f(x) \not\in f \left(C_y \right)$, [Refer to (B) above.] and so $x \not\in C_y$, which in turn implies that $x \in U_y$. [Refer to (A) above.] Therefore we have $$ f^{-1} \left( W_y \right) \subset U_y. \tag{3} $$

As $f$ is surjective, so, for each $y \in Y$, the set $f^{-}( \ \{ y \} \ )$ is non-empty, that is, there exists at least one $x_y \in X$ for which $$ y = f \left( x_y \right). \tag{4} $$ Any such $x_y \in f^{-1}( \ \{ y \} \ )$ of course, and so any such $x_y \not\in C_y$ because of (2) above.

So if $y \in f \left( C_y \right)$, then $y = f(v)$ for some element $v \in C_y$, which would imply that $v \in f^{-1}( \ \{ y \} \ ) \cap C_y$, which contradicts (2) above. Thus $y \not\in f \left( C_y \right)$, and therefore $ y \in W_y$. [Refer to (B) above.]

Thus we have shown that, for each $y \in Y$, there exists an open set $U_y$ in $X$ and there exists an open set $W_y$ in $Y$ such that $$ y \in W_y \ \qquad \mbox{ and } \qquad f^{-1} \left( W_y \right) \subset U_y; \tag{5} $$ [Refer to (3) above.] moreover the set $U_y$ is the union of some finite subcollection $\mathscr{A}_y$ of the open covering $\mathscr{A}$ of $X$. [Refer to (0) above.]

In this way, we obtain an open covering of $Y$, which is as follows: $$ \left\{ \ W_y \ \colon \ y \in Y \ \right\}. $$ And, since $Y$ is compact, this open covering has a finite subcollection that also covers $Y$; let $$ \left\{ \ W_{y_1}, \ldots, W_{y_n} \ \right\} $$ be this finite subcollection. Then we have $$ Y = \bigcup_{j=1}^n W_{y_j}. \tag{6} $$

Now we show that $$ X = \bigcup_{j=1}^n U_{y_j}. \tag{7} $$ Let $x$ be an arbitrary point of $X$. Let $y \colon= f(x)$. Then $y \in Y$, and so by (6) above we have $y \in W_{y_k}$ for at least one $k = 1, \ldots, n$; that is, $f(x) \in W_{y_k}$; this implies that $x \in f^{-1} \left( W_{y_k} \right)$ and hence $x \in U_{y_k}$, because of (3) above; therefore $$ x \in \bigcup_{j=1}^n U_{y_j}. $$ Thus one of the inclusions required for (7) above to hold holds; the other inclusion follows from the fact that, by our construction of these sets, each set $U_{y_j}$, for $j = 1, \ldots, n$, being a union of subsets of $X$, is itself a subset of $X$. [Refer to (0) and (1) above.]

Finally, as $X$ is a union of the collection $\left\{ \ U_{y_1}, \ldots, U_{y_n} \ \right\}$ consisting of finitely many sets and as each set $U_{y_j}$, for $j = 1, \ldots, n$, in this collection is itself a union of a finite subcollection $\mathscr{A}_{y_j}$ of our original open covering $\mathscr{A}$ of $X$ [Refer to (A) above.], so we can conclude that $X$ is eventually the union of the finite subcollection $\mathscr{A}_{y_1} \cup \ldots \cup \mathscr{A}_{y_n}$ of $\mathscr{A}$.

Since $\mathscr{A}$ is an arbitrary open covering of $X$, we can conclude that $X$ is compact.

Is my write-up correct? If so, then is it clear and easy enough to understand? If not, then where are the problems?

  • @Arthur can you please review my post once again? I've just tried to write-up a proof based on the answers, including yours? Is my write-up any clearer? – Saaqib Mahmood Sep 27 '18 at 06:23

3 Answers3

2

Continuing from your effort: Let $C(y)$ be the closed set of $X$ that is not covered by $\mathscr A(y)$. We know that $f(C(y))\subseteq Y$ is closed, which means that the complement $U(y)\subseteq Y$ of that again is open.

$\mathscr A(y)$ cover the fiber of $y$, so no point of $f^{-1}(\{y\})$ is in $C(y)$, and therefore $y \in U(y)$, so the $U(y)$ cover $Y$. Since $Y$ is compact, it can be covered by finitely many such $U(y)$, let's say $U(y_i), 1 \leq i \leq n$ for some $n$. I claim that the corresponding $\mathscr A(y_i)$ for all $i$ cover $X$.

To prove it, take an $x_0 \in X$. The point $f(x_0)$ is in some $U(y_j)$. That means that $f(x_0) \notin f(C(y_j))$, which again means that $x_0 \notin C(y_j)$, which again means that $x_0$ is inside one of the open sets in $\mathscr A(y_j)$.

Arthur
  • 204,511
  • do we have to suppose that $f$ is surjective? Have we actually used this assumption in our proof? – Saaqib Mahmood May 20 '15 at 18:07
  • No, I don't think we need it. Note that if $f$ is not surjective, then every $U(y)$ contains the complement of $f(X)$. Also, if we don't have surjectivity, we can just reduce to $Y'=f(X)$, which is closed and therefore compact. So it's probably there as a bonus that you would get almost for free anyway. – Arthur May 20 '15 at 23:35
  • have we made use of even the continuity of $f$ in this proof? – Saaqib Mahmood Jul 04 '15 at 20:32
  • @SaaqibMahmuud I can't find that anywhere, but in topology, all maps are continuous, always (unless there's some exercise showing you something that goes wrong if it's not continuous). All the theory builds on the assumption that all functions are continuous, so there might be some little detail where the assumption is used in my proof, but I can't find it. – Arthur Jul 05 '15 at 08:10
2

Let $\{ U_i \}_{i \in I}$ be an open cover of $X$. Then this is also an open cover of $f^{-1}(\{y\})$ for each $y \in Y$. Since $f^{-1}(\{y\})$ is compact, there exists a finite subset $I_y \subset I$, such that $$ f^{-1}(\{y\}) \subset \bigcup_{i \in I_y} U_i =: U_y \; $$ Since $U_y$ is open, $X \backslash U_y$ is closed subset of $X$, and since $f$ is a closed map, $f(X \backslash U_y)$ is a closed subset of $Y$. Note that $y \not\in f(X \backslash U_y)$. We define $W_y := Y \backslash f(X \backslash U_y)$. Now we see that $W_y$ is open in $Y$ and $y \in W_y$, so $W_y$ is an open neighbourhood of $y$. This means that $\{ W_y \}_{y \in Y}$ is an open cover of $Y$, and since $Y$ is compact, there exists a finite subset $\{ y_1, \ldots, y_m \} \subset Y$, such that $$ Y = \bigcup_{j=1}^m W_{y_j} \; .$$ We note that $$ f^{-1}(W_{y_j}) = f^{-1}( Y \backslash f(X \backslash U_{y_j})) \subset X \backslash f^{-1}(f(X \backslash U_{y_j})) \subset X \backslash (X \backslash U_{y_j}) = U_{y_j}$$ for each $j \in \{1, \ldots, m\}$, and from that it follows that $$ X = f^{-1}(Y) = \bigcup_{j=1}^m f^{-1}(W_{y_j}) \subset \bigcup_{j=1}^m U_{y_j} = \bigcup_{j=1}^m \bigcup_{i \in I_{y_j}} U_i \; , $$ so we have found a finite subcover of $\{ U_i \}_{i \in I}$, which means, that $X$ is compact.

Please check all these steps carefully, I'm not completely sure, if everything is working.

Bach
  • 5,874
  • 2
  • 22
  • 44
aexl
  • 2,152
1

Let $\{U_\alpha\}_{\alpha\in \Gamma}$ be an open cover of $X$. Then for each $y\in Y$, $f^{-1}(y)\subset \cup_{\alpha \in \Gamma}U_{\alpha}$. Since $f^{-1}(y)$ is compact, for each $y\in Y$, there exists a finite subcollection $\{U^y_{j}\}_{j=1}^n$ of $\{U_\alpha\}_{\alpha\in \Gamma}$ such that $f^{-1}(y)\subset\cup_{j=1}^nU^y_{j}$. So, $\cap_{j=1}^n(X-U^y_j)\subset X-f^{-1}(y)$. Notice that the left hand side of the previous expression is closed, so applying $f$ to both sides, the image remains closed. So using the fact that $f$ is surjective, we get, $f(\cap_{j=1}^n(X-U^y_j))\subset Y-y$. Therefore, $y\in Y-f(\cap_{j=1}^n(X-U^y_j))$, and $Y-f(\cap_{j=1}^n(X-U^y_j))$ is open. Now, taking union over $y\in Y$, we get, $Y=\cup_{y\in Y}(Y-f(\cap_{j=1}^n(X-U^y_j)))$. Now, since $Y$ is compact, there exists a finite collection of $y\in Y$ such that $Y=\cup_{k=1}^l(Y-f(\cap_{j=1}^n(X-U^{y_k}_j)))$. This implies $$ \begin{align} X = f^{-1}(Y) &=f^{-1}(\cup_{k=1}^l(Y-f(\cap_{j=1}^n(X-U^{y_k}_j)))) \\ &= \cup_{k=1}^l(f^{-1}(Y-f(\cap_{j=1}^n(X-U^{y_k}_j)))) \\ &= \cup_{k=1}^l(X-f^{-1}(f(\cap_{j=1}^n(X-U^{y_k}_j)))) \\ & \subset \cup_{k=1}^l(X-\cap_{j=1}^n(X-U^{y_k}_j)) \\ &=\cup_{k=1}^l\cup_{j=1}^nU^{y_k}_j. \end{align} $$ So, we've found a finite subcover, so $X$ is compact.

mathguy
  • 56
  • I hope you've not made any error. Can you please check your answer again just in case? – Saaqib Mahmood May 20 '15 at 17:35
  • does $f$ have to be surjective? I mean have you made use of surjectivity of $f$ in your proof? – Saaqib Mahmood May 20 '15 at 18:06
  • 2
    I've checked it, I don't think you need surjectivity, actually, since $f(f^{-1}(Y-y))\subset Y-y$ by default. Also, $f^{-1}(Y-y)=X-f^{-1}(y)$ is true even without $f$ being surjective. – mathguy May 20 '15 at 19:48
  • I've just rewritten your proof in an edit to my original post. So could you please take time reading through my edited post and comment on how correct and clear my presentation is? – Saaqib Mahmood Sep 27 '18 at 06:35