4
$\begingroup$

Given a compact and connected set $A$ in $\mathbb{R}^2$. Can it be locally path-connected at only one point? And can it be locally path-connected at every point except one?

My thought

First I thought of the topological sine curve, which is compact and connected but not locally path-connected. But it doesn't solve the problems. Then I tried to construct a space that is locally path-connected at every point except one as $$ A=\{(t\cos x,t\sin x):\cos x\in \mathbb{Q} , t\in[0,1]\} $$ But this is not compact.

Do such spaces exist? Any hints would be highly appreciated.

$\endgroup$
5
  • $\begingroup$ Idea for loc. path connected at only one point: Hawaiian earring with rational radii $\endgroup$ – G. Chiusole Apr 6 '20 at 8:34
  • $\begingroup$ Cone over Cantor set is compact, path connected and locally path connected only at a single point (the top vertex). But I'm not sure about locally path connected except a single point. $\endgroup$ – freakish Apr 6 '20 at 8:36
  • $\begingroup$ @G.Chiusole Hawaiian earring is locally path connected at every point, regardless of radii. $\endgroup$ – freakish Apr 6 '20 at 8:37
  • $\begingroup$ I meant a union over circles in $\mathbb{R}^2$ with radius $q$ and centered at $q/2$ for all $q \in \mathbb{Q}$, so in particular at $(1,0)$ (and all others except $(0,0)$) this should not be path connected. But I just realized that this space is far from compact. $\endgroup$ – G. Chiusole Apr 6 '20 at 8:43
  • $\begingroup$ A trivial example for the first is the singleton: It is obviously compact, connected and path connected, and since it consists of only one point, it trivially is only locally path connected at one point. $\endgroup$ – celtschk Apr 6 '20 at 10:53
1
$\begingroup$

freakish's example checks out, so credit to him. I'll just elaborate on his comment.

Let $C \hookrightarrow \{0\} \times \mathbb{R} \subseteq \mathbb{R}^2$ the ternary cantor set. and let $\mathcal{C}(C)$ be the cone over it. That is

$$\mathcal{C}(C) = (C \times I)/C \times \{1\} \subseteq \mathbb{R}^2~~.$$

Then $\mathcal{C}(C)$ is the quotient of a compact Hausdorff space by collapsing a closed subspace and thus also compact (and Hausdorff). The cone over any space is contractible (to the tip) and thus connected, but $\mathcal{C}(C)$ is locally path connected only at the tip $t := \pi( C \times \{1\})$ with $\pi$ being the projection $C \times I \rightarrow \mathcal{C}(C)$.

Locally path connected at the tip: Let $U$ be a neighborhood of $t \in \mathcal{C}(C)$. Then since $\mathcal{C}(C)$ carries the subspace topology so there is some $\varepsilon > 0$ s.t. $B_{\varepsilon}(t) \cap \mathcal{C}(C) \subseteq U$. This neighborhood is contactible via the contraction on $\mathcal{C}(C)$ restricted to $B_{\varepsilon}(t) \cap \mathcal{C}(C) \subseteq U$.

Nowhere else locally path connected: Let $x = (x_1, x_2) \in C \times I$ with $x_1 \neq 1$ and let $\varepsilon > 0$ s.t. $1 \not\in (x_1 - \varepsilon, x_1 + \varepsilon)$. Then consider the open neighborhood $(x_1 - \varepsilon, x_1 + \varepsilon) \times C$ of $x$. Since $\{ 1 \} \times C \cap (x_1 - \varepsilon, x_1 + \varepsilon) \times C = \emptyset$, applying $\pi$ to this set yields a homeomorphism of an open subset of $\mathcal{C}(C)$ and $(x_1 - \varepsilon, x_1 + \varepsilon) \times C$. The latter is not locally path connected since $C$ is nowhere locally path connected.

$\endgroup$
0
$\begingroup$

At least subspaces reduced to a single point satisfy the requirements of the question.

$\endgroup$
2
  • $\begingroup$ As far as I am concerned, if $A$ is locally path connected at $a$, it means that for any open subset containing $a$, there is a smaller open set containing $a$, which is path-connected in the subspace topology. I can't see how to show that for any open subset containing $b$, there is a smaller open set containing $b$ which is path-connected in the subspace topology. Do I miss some point? $\endgroup$ – Chiquita Apr 6 '20 at 7:34
  • $\begingroup$ You're right and I'll modify my answer accordingly. However, a subspace reduced to a single point still answer your requirements. $\endgroup$ – mathcounterexamples.net Apr 6 '20 at 7:43

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.