I'm dealing with a puzzling problem and hope some of you can help me.

Let be $(G,\cdot ) $ be a Hausdorff-group and let $A,B \subseteq G$. Show that if $A$ is closed and $B$ is compact, then $AB$ is closed.

Due to I'm considering a topological Hausdorff-group, we have two continous maps: $\psi: G \times G \rightarrow G, (x,y) \mapsto xy$ and $ \phi: G \times G \rightarrow G, (x,y) \mapsto x{y}^{-1} $

My idea was to show that $(AB)^{c}$ is open instead.

First we observe, that if $A$ is closed, then $A^{c}$ is open and due to $\phi$ is continous $\phi^{-1}(A^{c})$ is also open.
Now let's choose an arbitrary $x \in (AB)^{c}$. We see that $\forall y \in B: \phi(x,y)=xy^{-1} \in A^{c}$. (because if it wasn't it would be in $A$ we could conclude $\psi(xy^{-1},y)=xy^{-1}y \in AB$, what would be opposed to the chosen x)

So we know that $\phi(x,y)=xy^{-1} \in A^{c} \Rightarrow (x,y) \in \phi^{-1}(A^{c})$, which is open.

Then we find neighbourhoods $U$ from $x$ and $V$ from $y$ such that: $\forall y \in B \exists U_{y} \exists V : (x,y) \in U_{y} \times V \subseteq \phi^{-1}(A^{c})$

If I could manage to show, that $U:= \cap_{y \in B} U_{y}$ is open (I don't know this, because it might be an infinite intersection) i think my proof is complete.

Because then I can show that $U \subseteq (AB)^{c}$, so that for arbitrary $x$ $(AB)^{c}$ is a neighbourhood of x. This means that $(AB)^{c}$ is open.

Can anybody give me a hint how I can show, that this intersection is finite? I suppose it must follow from the compactness of B, because I haven't used it yet, but I have no clue how?

Thank you!

  • 1
    $\begingroup$ See math.stackexchange.com/questions/71983/… $\endgroup$ – peter a g Nov 7 '17 at 20:46
  • $\begingroup$ Hi! Thanks! I already know this thread, but I don't fully understand the way given there, so I tried to find my own solution. There are a couple of steps, which need to be made clearer, but I didn't overcome it. Maybe you have some further explainations for me? Or maybe another hint how this intersection can be finite? Thank you! $\endgroup$ – pcalc Nov 7 '17 at 22:14

It is basically the same proof as in a sequential space.

Let $(a_i, b_i) \in A \times B$ be a net with $a_i b_i \to c \in G$. Then, there is a subnet $b_{i_k}$ with $b_{i_k} \to b \in B$. Due to continuity the limit $$ a := \lim_k a_{i_k} = \lim_k (a_{i_k} b_{i_k}) b_{i_k}^{-1} = cb^{-1} \in A $$ exists and it follows $$ c = ab \in AB. $$

  • $\begingroup$ Hi! Thanks four your respond! I don't really see how this solves my problem with the finite intersection. Maybe I've overseen something? $\endgroup$ – pcalc Nov 7 '17 at 22:15
  • $\begingroup$ @pcalc in your proof: the system $\{V_y \mid y\in B\}$ is an open cover of $B$. Now use compactness to select a finite sub cover. $\endgroup$ – user251257 Nov 7 '17 at 23:25
  • $\begingroup$ Yes, that's true! Due to the compactness of $B$ one can follow that there exists a finite subcover made up of some $V_{y}$. I'm not sure how I can follow that there also exists a finite number of $U_{y}$. Does the line $\forall y \in B \exists U_{y} \exists V_{y}: (x,y) \in U_{y} \times V_{y} \subseteq \phi^{-1}(A^{c})$ still hold, if I take a finite number of $U_{y}$ ? $\endgroup$ – pcalc Nov 8 '17 at 8:31
  • $\begingroup$ @pcalc the $U_y$ and $V_y$ come in pairs... if you only need finitely many $V_y$'s, you only need finitely many $U_y$'s. $\endgroup$ – user251257 Nov 8 '17 at 16:19
  • $\begingroup$ I think i got it! Thank you! $\endgroup$ – pcalc Nov 10 '17 at 17:10

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.