-1

The union of a collection of connected subspace of $X$ that have a point in common is connected.

The following attempt is inspired by https://math.stackexchange.com/a/4384059/861687 post.

My attempt: let $f: \bigcup_{\alpha \in J} A_\alpha \to \{ 0,1\}$ be continuous map. By theorem 18.2, map $f|_{A_\alpha}:A_\alpha \to \{0,1\}$ defined by $f|_{A_\alpha}(x)=f(x),\forall x\in A_{\alpha}$ is continuous for each $\alpha \in J$. Since $A_\alpha$ is connected, we have $f|_{A_\alpha}(A_\alpha)=f(A_\alpha)=\{ i_\alpha \}$, $i_\alpha \in \{ 0,1\}$ and $i_\alpha$ depends on each $\alpha$. Since $\bigcap_{\alpha \in J} A_\alpha \neq \phi$, $\exists p\in \bigcap A_\alpha$. So $p\in A_\alpha, \forall \alpha \in J$. $p\in \bigcup A_\alpha$. Suppose $f(p)=i$, for some $i\in \{0,1\}$. Then $f(p)=i=i_\alpha ,\forall \alpha \in J$. Thus $i=i_\alpha, \forall \alpha \in J$. Let $x\in \bigcup A_\alpha$. Then $x\in A_\beta$, for some $\beta \in J$. So $f(x)=i_\beta =i$. Thus $f(x)=i, \forall x\in \bigcup A_\alpha$. Hence $f$ Is constant and $\bigcup_{\alpha \in J} A_\alpha$ is connected. Is this proof correct?

In this proof, we don’t make use of lemma 23.2 and exercise 1 of section 16 like Munkres’ proof.

Edit: Proof of equivalent definition of connected space: https://courses-archive.maths.ox.ac.uk/node/view_material/50743, proposition 1.77, page number 17. In lecture notes, it’s given $(1)\Rightarrow (2) \Rightarrow (3) \Rightarrow (1)$. One can also use $(1)\Leftrightarrow (2) \Leftrightarrow (3)$. Munkres have already proved $(1) \Leftrightarrow (2)$ and $(2) \Rightarrow (3)$ is given in lecture notes. Claim: $(3)\Rightarrow (2)$. Proof: Assume towards contradiction, i.e. $\exists P,Q\in \mathcal{T}_X$ such that $P,Q\neq \phi$, $P\cap Q=\phi$ and $P\cup Q=X$. Let $f: X\to \{0,1\}$ map defined by $f(P)=0$ and $f(Q)=1$. $f$ is a well defined map. It is easy to check $f^{-1}(\{0\})=P$ and $f^{-1}(\{1\})=Q$. Note $\mathcal{T}_Y =\{ \phi ,\{0\},\{1\}, \{0,1\} \}$. $\forall V\in \mathcal{T}_Y, f^{-1}(V)\in \mathcal{T}_X$. So $f$ is continuous but not constant. Which contradicts our initial assumption.

user264745
  • 4,595
  • 1
    Are we all supposed to know what are lemma 23.2 and exercise 1 of section 16? – José Carlos Santos Feb 18 '22 at 11:50
  • @JoséCarlosSantos No. I wrote that statement, to compare between two different proofs. Using different(from which one used to) definition of connected space one can get desired result easily. – user264745 Feb 18 '22 at 11:53
  • Instead of writing $\phi$, you should use \varnothing or \emptyset to denote the empty set: $\varnothing$ and $\emptyset$ respectively – FShrike Feb 18 '22 at 11:55
  • You can however show this theorem for when all the sets have pairwise nonempty intersection, which is different to and a weaker assumption than the intersection of them all being nonempty – FShrike Feb 18 '22 at 11:56
  • @FShrike I’m used to write $\phi$. – user264745 Feb 18 '22 at 11:56
  • You shoudn't use results as lemma 23.2 or prop 18.... The question has to be understood withouth having the book in front of us. Also $\phi$ is a realy bad notation for emptyset, although it look similar you sould change it. – Marcos Feb 18 '22 at 11:57
  • @Marcos That will make post extremely long, believe me. – user264745 Feb 18 '22 at 11:59
  • But you dont have to write all the previous result, you can assume that we all know it but instead of writing lemma 23.2 you can say by a result that says that the union of open sets is union then... But, there is no problem if your post is long, as long as you show effort and work it is nice. – Marcos Feb 18 '22 at 12:06
  • @Marcos IMO length of post do affects no. of people show interest to a post. For instance https://math.stackexchange.com/q/4367787/861687 , https://math.stackexchange.com/q/4369022/861687 etc. – user264745 Feb 18 '22 at 12:10

2 Answers2

1

This is not a full verification of your proof as your post mentions other results and doesn't complete the argument: $f$ constant implies what, exactly? Other than that however, your argument seems correct to me.

I leave this here as a demonstration of how it can be simpler, with no other theorems referenced.

$\newcommand{\C}{\mathcal{C}}\newcommand{\O}{\mathcal{O}}$A simple proof I produced myself for the same exercise (with weaker assumptions - it could well be the case that the entire intersection is empty):

Let $\{\C_\lambda:\lambda\in\Lambda\}$ be a nonempty collection of nonempty connected subsets of a topological space $X$, with each member having a common point with every other member, pairwise. Then $\bigcup_{\lambda\in\Lambda}\C_\lambda$ is a connected subspace of $X$.

Proof:

The case $|\Lambda|=1$ is trivial, so take at least $2$ elements therein.

Suppose $\O_1,\O_2$ disconnect $\bigcup_{\lambda\in\Lambda}\C_{\lambda}$. There must exist at least one $\lambda\in\Lambda$ for which $\O_1\cap\C_{\lambda}\neq\emptyset$. If $\O_2\cap\C_{\lambda}\neq\emptyset$, then $\O_1,\O_2$ would disconnect $\C_{\lambda}$ which is impossible, so take $\O_2\cap\C_\lambda=\emptyset$ and thus $\C_\lambda\subseteq\O_1$.

There exists $\lambda'\in\Lambda$ with $\lambda'\neq\lambda$ such that $\C_\lambda\cap\C_{\lambda'}\neq\emptyset$ and $\O_2\cap\C_{\lambda'}\neq\emptyset$ by assumption and by $|\Lambda|\gt1$. But as $\C_\lambda\subseteq\O_1$, we have $\C_{\lambda'}\cap\C_\lambda\neq\emptyset$ implies $\O_1\cap\C_{\lambda'}\neq\emptyset$. Then $\O_1,\O_2$ disconnect $\C_{\lambda'}$ which is again impossible.

Therefore no disconnection of nonempty disjoint open sets exists of $\bigcup_{\lambda\in\Lambda}\C_\lambda\,\blacksquare$

FShrike
  • 46,840
  • 3
  • 35
  • 94
  • Thank you for the answer. You could have also have used $f:\bigcup A_i \to {0,1 }$ approach. By weaker assumption, I assume $\bigcap A_i \neq \emptyset \Rightarrow A_i \cap A_j \neq \emptyset ,\forall i,j\in I$. So what you have proved above implies theorem 23.3. – user264745 Feb 19 '22 at 12:07
  • 1
    @user264745 Yes, my proof implies yours, but not the other way round – FShrike Feb 19 '22 at 12:09
1

Yes, that proof is correct, using the alternate characterisation using functions into $\{0,1\}$ we discussed earlier.

You could also start with $p$ and note that for each $\alpha$ we must have by constantness that $f\restriction_{A_\alpha} \equiv f(p)$ from the start. It then immediately follows that $f \equiv f(p)$ and we're done.

Henno Brandsma
  • 250,824