Let $(X, \tau)$ be a topological space, and let $\overline{ (-) } : \mathcal{P}(X) \to \mathcal{P}(X)$ be defined as

$$ \overline{S} := \{x : \exists(x_\alpha)_{\alpha \in \Lambda} \subseteq S \text{ net s.t. } x_\alpha \to x\}. $$

I'm trying to prove that $\overline{A \cup B} \subseteq \overline{A} \cup \overline{B}$, since I've already seen the other inclusion.

Let $x \in \overline{A \cup B}$ so that we have $(x_\alpha)_{\alpha \in \Lambda} \subseteq A \cup B$ with $x_\alpha \to x$. My idea was to adapt a similar proof for sequences, where one can assume there are infinite terms of a sequence in one of the sets, and take an appropriate subsequence. I have first proved that $x \in A$ or $x \in B$ for $\Lambda$ finite, and now am tackling the infinite case.

Concretely, now we can assume that infinitely elements of (the image of) the net are on $A$. I want to prove that $\Gamma = \{\alpha \in \Lambda : x_\alpha \in A \}$ and the inclusion $\Gamma \hookrightarrow \Lambda$ will give a subnet of the original, thus concluding that $x \in \overline{A}$. However, that does not seem immediate (or correct, even) if we do not assume $\Gamma$ cofinal, which will not happen always.

Would you mind providing any hints on whether I am on the right track, and if so on how can I finish my argument? Thanks in advance.


The best idea is to assume $x \in \overline{A \cup B}$ giving us a net $n: \Lambda \to X$ such that $n \to x$ and for all $\alpha \in \Lambda$ we have $n(\alpha) \in A \cup B$, and then define $\Lambda(A) = \{ \alpha \in \Lambda: n(\alpha) \in A\}$ and likewise $\Lambda(B)$ and by definition we have that $\Lambda = \Lambda(A) \cup \Lambda(B)$, (not disjoint necessarily, of course, but that's OK).

Can it be true that neither is cofinal in $\Lambda$? (Hint : no) and then you're done (a cofinal subset gives another (sub)net using inclusion, as you suggested ) What does it mean to be not cofinal?? Get a contradiction from the definition of a directed set.

  • $\begingroup$ This is enlightening! I will read this carefully and rewrite my proof when I'm less tired, and accept the answer accordingly. Thanks so much for the insight :) $\endgroup$ – Guido A. Sep 8 '18 at 8:51

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.