Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I have been working on the following problems, the first $3$ are from Dugundji's book, page $117$. I was wondering if someone can please check them or suggest easier proofs.

1) Let $\{A_{i}: i \in I\}$ be a family of connected subsets of $X$ and assume that there exists a connected set $A$ with $A \cap A_{i} \neq \emptyset$ for each $A_{i}$. Show $A \cup \bigcup_{i \in I} A_{i}$ is connected.

2) Let $\{A_{i}: i \in I\}$ be any family of connected sets. Assume that any two of them have non-empty intersection. Prove $\bigcup_{i \in I} A_{i}$ is connected.

3) Let $Y$ be a space and let $A \subset Y$ be any subset. Let $C \subset Y$ be connected such that $C$ intersects $A$ and $Y \setminus A$. Then it can be shown that $C$ contain points of $\textrm{bd}(A)$, the boundary of $A$.

Question: In $\mathbb{R}^{3}$, why is $A=\{(x,y,0) : x^{2}+y^{2} \leq 1\}$ and $C=\{(0,0,z): |z| \leq 1\}$ not a counterexample?

4) A doubt: $\mathbb{Q}$ is not locally connected.

My work:

1) For each $i \in I$ let $C_{i} = A \cup A_{i}$ then by assumption $C_{i}$ is connected. Now note that $A \subset \bigcap_{i \in I} C_{i}$. So $A \neq \emptyset$ for otherwise $A \cap A_{i} = \emptyset$, which contradicts the assumption. Thus $\bigcap_{i \in I} C_{i} \neq \emptyset$. Therefore $\bigcup_{i \in I} C_{i}$ is connected. But $\bigcup_{i \in I} C_{i} = A \cup \bigcup_{i \in I} A_{i}$.

2) Let $f: \bigcup_{i \in I} A_{i} \rightarrow \{0,1\}$ be a cts map. Now let $\alpha_{0} \in I$ be fixed and pick $\beta \in I$ such that $\beta \neq \alpha_{0}$. Then the restriction of $f$ to $A_{\beta}$ is constant (since $A_{\beta}$ is connected). So let $z_{\beta}=f(A_{\beta})$. By assumption $A_{\alpha_{0}} \cap A_{\beta}$ is non-empty so let $x$ be in the intersection. Then $f(x) \in f(A_{\alpha_{0}})$ and $f(x) in A_{\beta}$. Therefore $f(x)=z_{\beta}$ so $f$ is constant.

3) Is it because $C$ does not intersects the complement of $A$ or why?

4) Can we proceed as follows? suppose $\mathbb{Q}$ is locally connected and consider $(a,b) \cap \mathbb{Q}$ this is open in $\mathbb{Q}$. Now assuming $\mathbb{Q}$ is locally connected this would imply each connected component of $U$ is open. But $\mathbb{Q}$ is totally disconnected and $\mathbb{Q}$ is not discrete, so this is impossible.

share|cite|improve this question
In your counterexample for question 3, $A$ has empty interior so $\mathsf{bd}(A)=A$. And if you were to just look at a disk with a line through the center, then the point where the disk intersected the line would also be a boundary point. – JSchlather Jul 6 '11 at 19:56
@Jacob Schlather: Can you please explain why it has empty interior? – user10 Jul 6 '11 at 20:13
Sure in $\mathbb R^3$ an open set must contain an open ball around each of it's points. But if you move $\epsilon$ away from any point on the disk in the $z$ direction, then you're not on the disk anymore. Another way is to note that the complement is dense in $\mathbb R^3$, so it must have empty interior. – JSchlather Jul 6 '11 at 20:16
up vote 2 down vote accepted

(1) is basically correct but a little clumsy. First, it isn't by assumption that $C_i$ is connected: it's by virtue of Theorem V.1.5 and the fact that $A \cap A_i \ne \emptyset$. Secondly, 'So' is inappropriate in your third sentence: it implies that what follows -- $A \ne \emptyset$ -- is a consequence of what you'd previously said, which isn't the case. Better: Clearly $A \ne \emptyset$, since $A$ has non-empty intersection with each $A_i$, so $\bigcap_{i \in I} C_i \ne \emptyset$.

(2) Basically fine. The wording could be improved a bit, and there's one bit of notational sloppiness that I really don't like: $f(A_\beta)$ is a set, so it equals $\{z_\beta\}$, not $z_\beta$.

(3) Use Dugundji's hint that $Y = \operatorname{Int}(A) \cup \operatorname{Fr}(A) \cup \operatorname{Int}(Y \setminus A)$ to conclude that if $C \cap \operatorname{Fr}A = \emptyset$, then $C \subseteq \operatorname{Int}(A) \cup \operatorname{Int}(Y \setminus A)$. In this case $C$ must intersect both $\operatorname{Int}(A)$ and $\operatorname{Int}(Y \setminus A)$; why, and why does this give you a 'disconnection' of $C$? (Jacob has already answered the other part of the question in his comment.)

(4) is fine; you're using Theorem V.4.2 and Ex. 1 on p. 111.

share|cite|improve this answer

1,2 and 4 look fine, assuming that for 4 you can assume $\mathbb Q$ is totally disconnected (this isn't hard to prove regardless). A nicer way of looking at that question (and indeed what you did) is that a locally connected space is totally disconnected if and only if it is has the discrete topology.

For question 3 I think the easiest way to go about it is to suppose that $\mathsf{bd}(A) \cap C$ is empty. Then what can you say about $\mathsf{Int}(A)$ and $Y \setminus \overline{A}$ with regards to $C$?

share|cite|improve this answer

Questions (1)--(3) can all be handled by appealing to the topological definition of "connected"... That is, if I want to prove $X$ is connected, then I assume $X \subset U \cup V$ for open sets $U$, $V$ such that $U \cap V = \emptyset$. If I can logically show that $X$ must be contained only in $U$ or only in $V$, then I have proven $X$ connected. This type of proof often starts by observing some given connected set $C$ is in $U \cup V$, and therefore we know that $C$ lies only in $U$ (we don't have to consider $C$ lies only in $V$. Why not?).

For your question (3), the example you have given is not complete. I don't know what $Y$ is. Whatever $Y$ is, it must properly contain $A$ (so that $Y \setminus A$ is non-empty), so let's just assume $Y = \mathbb{R}^3$ for the moment. Then the hypotheses of (3) are satisfied. The conclusion is also satisfied, since $C$ contains the point $(0,0,0)$. This is a boundary point of $A$, since $A \subset \mathbb{R}^3$! (It would not be a boundary point if we only worked in $\mathbb{R}^2$).

For (4), I'd argue from definitions again. We say that $X$ is locally connected at $x$ if for every open set $V$ containing $x$ there exists a connected, open set $U$ with $x \in U \subseteq V$. $\mathbb{Q}$ is not locally connected at $0$ since for $V = (-1,1)$, any open set $U$ containing $0$ such that $U \subseteq V$ consists of unions of open intervals within $(-1,1)$. Just consider the open interval $(a,b) \subseteq U$ containing $0$. There must be a negative rational and a positive rational. Can you find an irrational number lying between these two? Such an irrational number would be a "hole" that can be used to define a separation of the interval $(a,b)$.

share|cite|improve this answer
@Jacob Schlather But totally disconnected is not equivalent to discrete, as the example $\mathbb{Q}$ shows. If $\mathbb{Q}$ were discrete, then singletons $\{a\}$ would be open sets! -------OOPS, just re-read the comment and saw "locally connected" before that if-and-only-if. You're correct :) – Shaun Ault Jul 6 '11 at 20:28

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.