6

before I begin, I would like to provide some definitions and theorems.

Definition. A topological space $(X, \tau_X)$ is a continuum, if $X$ is a non-empty, metric, compact and connected space.

Definition. Let $X$ be a continuum and define $E(X)=\{p \in X: ord_X(p)=1\}$, $O(X)=\{p \in X: ord_X(p)=2\}$ and $R(X)=\{p \in X: ord_X(p) \geq 3\}$. $E(X)$, $O(X)$ and $R(X)$ denote the set of endpoints, ordinary points and branch points of $X$, respectively.

Notation. If $X$ is a topological space and $A\subseteq X$, then we denoted the boundary of $A$ in $X$ by $fr_X(A)$.

Definition. Let $(X, \tau_X)$ be a topological space, let $p \in X$, and let $\kappa$ be a cardinal number. We say that the order of $p$ in $X$, denoted by $\operatorname{ord}_X(p)$, is equal to $\kappa$ if the following conditions are satisfied:

  1. The order of $p$ in $X$ is less than or equal to $\kappa$; that is, for every open subset $U$ of $X$ containing $p$, there exists an open subset $V$ of $X$ such that $p \in V \subseteq U$ and $|fr_X(V)| \leq \kappa$.
  2. $\kappa$ is the smallest cardinal number that satisfies condition (1); that is, for every cardinal number $\alpha < \kappa$, there exists an open subset $U_\alpha$ of $X$ containing $p$ such that for every open subset $W$ of $X$ with $p \in W \subseteq U_\alpha$, it holds that $|fr_X(W)| > \alpha$.

Definition. An arc, is every topological space which is homeomorphic to the interval $[a,b]$. An arc with endpoints $a,b$ is denoted by $ab$.

Definition. Let $X$ be a topological space and $ab$ an arc, we said that $ab$ is a free arc if $ab \setminus \{a,b\}$ is open in $X$.

I would like to prove: Let $X$ be a continuum and $A$ an arc. $(A \setminus E(A)) \cap R(X)= \emptyset$ then $A$ is an free arc in $X$.

My attempt: In order to prove it, all I have been able to do is: Suppose that $(A \setminus E(A)) \cap R(X)$ is empty. We want to prove that $A$ is a free arc. Since, for each $y \in A \setminus E(A)$, $\operatorname{ord}_A(y) = 2$, using our hypothesis we have that $\operatorname{ord}_X(y) = 2$ for each $y \in A \setminus E(A)$. Let $p \in A \setminus E(A)$. We know that $A \setminus E(A)$ is an open set in $A$, so there is an open set $V$ in $X$ such that $p \in V \cap A \subseteq A \setminus E(A)$. Also, we have that there is an open subset $V_0$ of $X$ such that $p \in V_0 \subseteq V$ and $\mid \operatorname{fr}_X(V_0) \mid \leq 2$. But how can we find an open set $B$ in $X$ such that $p \in B \subseteq A \setminus E(A)$?

I have been working in this result for a long time, but I have not been able to prove it or disprove it. I hope can you help me.

Aldo
  • 179
  • 1
  • 14

1 Answers1

3

Suppose $p \in A \setminus E(A)$ and choose open $U$ such that $p \in U \subseteq cl(U) \subseteq X \setminus \{a,b\}$ and $\mid \operatorname{fr}_X(U) \mid \leq 2$. $U$ must have two boundary points within $A \setminus E(A)$, say $c$ and $d$. $p$ must be an ordinary point of the arc $cd$.

Suppose $q \in U \setminus A$. Cover $cd$ with open sets which have only two boundary points and which do not contain $a$, $b$ or $q$. (Their boundaries will all be within $A$.) Take a finite subcovering and call its union $V$. The boundary of $V$ is contained in the union of the boundaries of the members of the subcovering, and so is contained in $A \setminus cd$. So $q \in cl(U) \setminus V = U \setminus cl(V)$, which is clopen, contradicting $X$ is connected.Hence there cannot be such a $q$, $U \setminus A$ is empty and $U \subset A \setminus E(A)$. Therefore $A \setminus E(A)$ is open.

  • It's not clear to me why $U$ must have two boundary points within $A\setminus E(A)$ – Aldo Aug 29 '25 at 19:11
  • As $U \cap A$ is a non-empty open subset of $A$ not containing the endpoints, it must have at least two boundary points in $A$ (in the subspace topology). But these are also boundary points of $U$ in $X$, which only has two. Therefore they must both be in $A$. However, as I had already realised, that didn't exclude them being $a$ or $b$. I've edited the answer to require $U$ is chosen so that $cl(U)$ doesn't include those points, which ensures the two boundary points are in $A \setminus E(A)$. – David Hartley Aug 30 '25 at 02:33
  • Now, I don't see why $V$ has two points in its boundary. $\operatorname{fr}X(V) = \overline{V} \setminus V$, but $V$ is a finite union of open sets, so $\operatorname{fr}_X(V) = \left(\bigcup{i=1}^k \overline{U_i}^X\right) \setminus \left(\bigcup_{i=1}^k U_i \right) \subseteq \bigcup_{i=1}^k \operatorname{fr}_X(U_i)$ and it is bounded by $2k$ for some $k$. I would like to prove it was bounded by $2$ and conclude its boundary has two points with a similar argument as at the beginning. – Aldo Aug 30 '25 at 20:07
  • Sorry, it was late at night and I was a bit careless. You are quite right that $V$ can have more than two boundary points. We can make it have only two by taking a minimal subcovering, but that is not necessary. The important thing is that its boundary is within $A \setminus cd$ and so $U \setminus V$ is clopen. I've edited the answer accordingly. – David Hartley Aug 30 '25 at 22:48
  • I've proved that $\overline{U}\setminus V = U \setminus V$. In order to prove $U \setminus \overline{V}= U \setminus V$ I used the fact that $U \cap fr_X(V) = \emptyset$. But that comes from $cd = A\cap U$. There is way to prove the last equality without appealing the fact that $U\cap A$ is an open connected space in an arc $A$ whose boundary has only two points? – Aldo Aug 31 '25 at 18:08
  • $U \cap A$ is an open subset of $A$ which has only two boundary points which are not the endpoints of $A$. It must be $cd \setminus {c,d}$. I was implicitly assuming that throughout. But it might be unnecessary. Doesn't the proof still work if we form $V$ from a covering of $overline{U \cap A}$? The boundary points will all be in $A \setminus overline{U \cap A}$, and so not in $U$. – David Hartley Sep 01 '25 at 00:57
  • That should have been $\overline{U \cap A}$ and $A \setminus \overline{U \cap A}$. – David Hartley Sep 01 '25 at 01:08
  • I'd like to thank you a lot. An interesting question is whether we can omit the hypothesis that $X$ is connected. But that’s another story. For now, thank you again. – Aldo Sep 03 '25 at 01:37
  • 1
    No.Just add a sequence of isolated points converging on a point inside the arc. E.g. $[0,1] \times {0} \cup {(1/2,1/n)}| n=1,2,. . }$. – David Hartley Sep 03 '25 at 09:05
  • You can omit metrisable and compact. They are not used in the proof. You can also replace connected with locally connected. – David Hartley Sep 03 '25 at 16:05
  • 1
    An arc $A$ in a space $X$ is free if, and only if, each point of $A$ other than the end-points has order 2 in $X$ and has a connected neighbourhood in X. – David Hartley Sep 04 '25 at 00:35
  • That is a fancy characterization. I've been able to prove that $A$ is arc-free in $X$ (where $X$ is a connected $T_3$ space) if and only if $A \cap R(X) = \emptyset$, but yours is better and more precise. – Aldo Sep 04 '25 at 02:26