4
$\begingroup$

Problem: Let $U\subseteq \mathbb{R}^2$ be any open set in the Euclidean plane $(\mathbb{R}^2$ with the standard topology) and consider $(U,T)$ as a topological space with the subspace topology inherited from the plane. Prove or disprove that $(U^*, T^*)$ is path-connected (where $(X^*,T^*$) denotes the one-point compactification/Alexandroff Extension of a topological space $(X,T)$).

I sketched out a proof as follows: Let $U$ be any open set in $\mathbb{R}^2$. Then we know that we can write $U= \cup_{\alpha}B_\alpha$ where $B_\alpha\subseteq U$ is an open ball. Then, \begin{align*} U^*&=(\cup_\alpha B_\alpha)^*\\ &=(\cup_\alpha B_\alpha)\cup \{\infty\}\\ &=\cup_\alpha (B_\alpha\cup \{\infty\}) \end{align*} Since $B_\alpha\cup\{\infty\}$ is homeomorphic to $S^2$ and $S^2$ is path-connected we have that $B_\alpha \cup \{\infty\}$ is path-connected for all $\alpha$ (path-connectedness is preserved under homeomorphisms).

Then, I want to use a result I've proved in the past: "Prove that if $A_i\subset X$ is path-connected for every $i\in I$, and $A_i\cap A_j\neq \emptyset$ for all $i,j\in I$, then $A=\cup_{i\in I} A_i$ is path-connected. " Applying this result then gives that $A^*$ is path-connected.

My concern with this proof is if I skipped too many steps for the homeomorphism part. However, is this sketch in the right direction?

$\endgroup$
7
  • 8
    $\begingroup$ $B_\alpha\cup\{\infty\}$ very rarely is path connected. For example it never is when $U=\mathbb{R}^2$ to begin with. It is path connected only when the ball touches the boundary of $U$, which rarely happens in practice. $\endgroup$ Commented 20 hours ago
  • 1
    $\begingroup$ One idea is to show that every point in $U$ can be connected to boundary. $\endgroup$ Commented 20 hours ago
  • 1
    $\begingroup$ Let $S^2\approx\mathbb R^2\cup\{\infty\}$ be the OPC of $\mathbb R^2$. $U$ is open in $S^2$. Let $A=S^2\setminus U$. Then $A$ is closed in $S^2$. By this theorem, the OPC of $U$ is homeomorphic $S^2/A$ ($A$ quashed to a point). Now take a path between any two points of $U$ in $S^2$, and map it under the quotient map to $S^2/A$. $\endgroup$ Commented 20 hours ago
  • $\begingroup$ @Chad: your claim about OPC of $U$ is wrong, think about the annuals as your open subset. $\endgroup$ Commented 15 hours ago
  • $\begingroup$ @MoisheKohan: $S^2/A$ is $S^2$ with $A$ collapsed to a point (the extra point $\{\infty\}$). What did it read like I meant? $\endgroup$ Commented 12 hours ago

2 Answers 2

Reset to default
4
$\begingroup$

For $U = \mathbb R^2$ it is trivial. So consider a proper subset $U$ of $\mathbb R^2$. To prove that $U^*$ is path connected, it suffices to show that for each $x \in U$ there exists a path $v :[0,1] \to U^*$ from $x$ to $\infty$.

Choose $y \in \mathbb R^2 \setminus U$ and a path $u : [0,1] \to \mathbb R^2$ such that $u(0) = x$ and $u(1) = y$. The set $u^{-1}(\mathbb R^2 \setminus U)$ is closed in $[0,1]$, hence compact with a minimum $m > 0$. By definition $u([0,m)) \subset U$ and $u(m) \in \mathbb R^2 \setminus U$.

We may assume w.l.o.g. that $m = 1$. Define $$v : [0,1] \to U^*, v(t) = u(t) \text{ for } t < 1, v(1) = \infty .$$ This map is continuous at all $\tau < 1$ because $v \mid_{[0,1)} = i \circ \tilde u$ with inclusion $i : U \to U^*$ and restriction $\tilde u : [0,1) \to U$ of $u$.

Continuity at $1$:

Let $V$ be an open neighborhood of $\infty$ in $U^*$. This means that $C = U^* \setminus V$ is a compact subset of $U$, thus a closed subset of $\mathbb R^2$. Therefore $u^{-1}(C)$ is closed in $[0,1]$ and does not contain $1$ (because $u(1) \in \mathbb R^2 \setminus U$ and $C \subset U$).

Therefore $W = [0,1] \setminus u^{-1}(C)$ is an open neighborhood of $1$ in $[0,1]$. To show that $v(W) \subset V$, it suffices to verify that $v(t) = u(t) \in V$ for $t \in W \setminus \{1\}$.

$t \in W$ means $t \notin u^{-1}(C)$, that is $u(t) \notin C$ . Therefore $v(t) = u(t) \in U^* \setminus C = V$.

By the way, all that was needed in the above proof is that $U$ is a proper open subset of a path connected Hausdorff space $X$.

$\endgroup$
3
$\begingroup$

You cannot expect that each $B_\alpha \cup \{\infty\}$ is path connected. As an example take $U = \mathbb R^n$. Let $B_n$ be the open ball with center $0$ and radius $n$. Then $U = \bigcup B_n$. Identifying $U^*$ with $S^2$ (where $\infty$ is the north pole $N$), you see that no $B_n \cup \{\infty\}$ is not path connected.

So what can we do?

Identifying $\mathbb R^2$ with $S^2 \setminus \{N\}$, we see that $A = S^2 \setminus U$ is a compact non-empty subset of $S^2$. Consider the quotient map $p : S^2 \to S^2/A$. It restricts to a continuous bijection $p' : U \to p(U) = (S^2/A) \setminus \{*\}$, where $*$ is the point obtained by collapsing $A$. Note that $\{*\}$ is closed in $S^2/A$ since $p^{-1}(\{*\}) = A$. Hence $p(U)$ is an open subspace of the compact space $S^2/A$.

We have $p^{-1}(p(U)) = U$, thus $U$ is a saturated subset of $S^2$ and $p'$ is a quotient map. Hence $p'$ is an open map (because it is bijective) and therefore a homeomorphism.

Let us show that $S^2/A$ "is" the one-point compactification of $U$. Indeed, define $$\phi : U^* \to S^2/A, \phi(x)= \begin{cases} p'(x) & x \in U \\ * & x = \infty \end{cases}$$ which is clearly a bijection. We claim that $\phi$ is a homeomorphism.

  1. $\phi$ is continuous.

It is clearly continuous in all $x \in U$ since $p(U)$ is open is $S^2/A$. Let us show continuity in $\infty$. Let $V$ be an open neighborhood of $*$ in $S^2/A$. Its complement $V' = (S^2/A) \setminus V$ is a compact subset of $(S^2/A) \setminus \{*\}$, thus $W'= (p')^{-1}(V)$ is a compact subset of $U$ and $W = U^* \setminus W'$ is an open neigborhood of $\infty$ in $U^*$. By construction, $\phi(W) = V$.

  1. $\phi^{-1}$ is continuous.

The proof is analogous to that of 1. Simply use $$\phi^{-1}(y)= \begin{cases} (p')^{-1}(y) & y \in (S^2/A) \setminus \{*\} \\ \infty & y = * \end{cases}$$

It therefore suffices to prove that $S^2/A$ is path connected. Let us show that for each $y \in (S^2/A) \setminus \{*\}$ there exists a path from $y$ to $*$.

Let $x = (p')^{-1}(y) \in U$. Choose a path $\gamma : I \to S^2$ such that $\gamma(0) = x$ and $\gamma(1) = N$. Then $\hat \gamma = p \circ \gamma : I \to S^2/A$ satisfies $\hat \gamma(0) = y$ and $\hat \gamma(1) = *$ (since $N \in A$).

Remark.

The proof can easily be generalized to open subsets of $\mathbb R^n$.

$\endgroup$
4
  • $\begingroup$ I actually do not see why each $B_n\cup \{\infty\}$ is not path connected. I thought that the one-point compactification of any open ball in $\mathbb{R}^n$ is homeomorphic to $S^n$ (exercise 2.7.8 in Gamelin's intro to topology). Then we can use the fact that path connectedness is a topological property. Why doesn't this work? $\endgroup$ Commented 12 hours ago
  • 2
    $\begingroup$ Being a bit hand-wavey, the $\infty$ you get from compactifying one of your $B_n$s is not the same $\infty$ you get from compactifying all of $\mathbb R^n$. To be a bit more rigorous, the inclusion map of $B$ into $R^n$ doesn't play well with "take one point compactification". $\endgroup$ Commented 12 hours ago
  • 3
    $\begingroup$ @Dooley: The one point compactification of any open ball $B \subset \mathbb R^n$ IS path connected. However, that fact is entirely irrelevant to understanding the one point compactification of $\mathbb R^n$, because the one point compactification of $B$ is not topologically embedded in the one point compactification of $\mathbb R^n$. $\endgroup$ Commented 12 hours ago
  • 1
    $\begingroup$ @Dooley - Because I like to try to find the most precise point where an argument goes wrong, I'll point out that when, in your comment, you write "$B_n \cup \{\infty\}$", that is not the OPC of $B_n$, (nor is it homeomorphic to its OPC). "$\infty$" is initially treated as a free symbol - we assume it is not an element of our original space, so we can use it to designate the new point we are adding. But once you've created the OPC of $\mathbb R^n$ you've "bound" that symbol, and you can't reuse it when creating the OPC of your $B$. $\endgroup$ Commented 10 hours ago

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.