3
$\begingroup$

EDIT The question was originally stated in a general setting, but now it is in terms of plane continua as these are the cases I care about.


A $continuum$ is a compact, connected metric space.

By a $plane\ continuum$ I mean a continuum which is homeomorphic to a subset of $\mathbb{R}^2$.

Suppose $X$ is a locally path connected plane continuum. (Note that if a space is locally path connected, it is path connected iff it is connected.) Also, suppose $X$ has covering dimension $1$. I have conjectured that every subcontinuum of $X$ must also be locally path connected.

Each of my proof-attempts have been rather fragmented so it's difficult to write anything coherent, but here's my general approach:


Let $X$ be a locally path connected one dimensional plane continuum. Suppose $S\subset X$ is a non-locally path connected continuum. Choose $s\in S$ so that every open set around $s$ is not path connected. Now, using the local path connectivity of $X$, choose a path connected open set $M$ in $X$ with $s \in M$ so that $M\cap S = N$ is connected. (I haven't been able to prove that choosing an $M$ so that $M\cap S$ is connected is possible.)

Now choose $t \in N$. Note that there is no path from $t$ to $s$ in $N$. However, there is a path from $t$ to $s$ in $M$. I should be able to use this to reach a contradiction along the lines of "$X$ must have covering dimension greater than $1$". However, I'm struggling to formalise anything as I'm not sure how to restrict from the possibility that $X$ has many "loops". Note that I also haven't used the fact that $X$ is a plane continuum. Any help will be appreciated!

tl;dr

Is every subcontinuum of a $1$ dimensional locally path connected plane continuum locally path connected?

$\endgroup$
  • $\begingroup$ This is probably true in the plane. You may want to start there. $\endgroup$ – Forever Mozart Apr 20 '17 at 23:39
  • $\begingroup$ That is actually the only case I'm interested in so I may as well change my question. I haven't used any plane-specific properties in my proof attempts yet. $\endgroup$ – Harambe Apr 21 '17 at 5:01
  • 2
    $\begingroup$ Isn't the Sierpinski carpet a counterexample? It has a subcontinuum $([0,1] \times C) \cup (C \times [0,1])$ where $C$ is the ternary Cantor set and this is definitely not locally connected. $\endgroup$ – Niels J. Diepeveen Apr 21 '17 at 9:58
3
$\begingroup$

The Menger curve $M$ is a counter example. The reason is that $M$ is universal in the sense that for every 1-dimensional metrizable compact space $C$, there is an embedding $C\to M$. This was proven by Menger himself. (I will find a more accessible reference when I have more time. This might be in Engelking's book "General Topology" or/and in Kuratowski's "Topology". You are welcome to check if you have either one.) Menger curve is also connected and locally path-connected. Now, take your favorite non-locally connected continuum (say, the infinite comb) and embed it into $M$.

Edit: Actually, it is easy to see that already the Sierpinski carpet contains an infinite comb and, hence, is a counter example. Namely, suppose we are using the standard ternary Sierpinski carpet $S$ (contained in the unit square $[0,1]\times [0,1]$). Let $C$ be the ternary Cantor set in the x-axis. Then for each $c\in C$, $c\times [0,1]\subset S$. The union (an infinite "comb") $$ C\times [0,1]\cup [0,1]\times \{0\}\subset S $$ is connected, compact, but not locally connected.

On the positive side, every 1-dimensional nonseparating planar Peano continuum satisfies your property, because it is a dendroid.

$\endgroup$
  • $\begingroup$ Engelking Dimension Theory is the ref for this. $\endgroup$ – John Samples Jun 14 '18 at 23:40
  • $\begingroup$ Also, you mean dendrite, not dendroid. Dendroids are ferocious creatures that we know basically nothing about, still haha $\endgroup$ – John Samples Jun 14 '18 at 23:41

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.