Let me define a continuum $X$ to be nice if for each pair $a,b\in X$ of distinct points, there are subcontinua $M,N\subseteq X$ such that: $a\in M\setminus N$; $b\in N\setminus M$; and $M\cup N=X$. (This is a very strong form of aposyndesis.) If a continuum is both locally connected (base consisting of connected open sets) and colocally connected (base consisting of open sets with connected complements), then it is clearly nice. My question is: does this property have an established name in the continuum-theoretic literature?
1 Answers
Yes, this is called freely decomposable and actually it's equivalent to aposyndesis as a global property of metric continua. Your "colocally connected" is a stronger form of a property called semi-locally connected (slc), where each point has a neighborhood base whose elements have only finitely many components in their complements. The slc property is also equivalent to freely decomposable as a global property.
Let $x, y \in X$ be distinct, where $X$ is a continuum.
Assume $X$ is freely decomposable, i.e. there are subcontinua $A \cup B = X$ with $x \in A \setminus B$ and $y \in B \setminus A$. If $x \notin \text{int}(A)$ then every neighborhood of $x$ intersects $X \setminus A \subset B \implies x \in \overline{B} = B$, a contradiction. Thus $A$ is a neighborhood of $x$ and $A \subset X \setminus \lbrace y \rbrace$, so that $X$ is aposyndetic at $x$ wrt $y$. But $x, y$ were arbitrary, so $X$ is aposyndetic.
Now assume $X$ is aposyndetic and $U$ is a neighborhood of $x$. Since $X$ is aposyndetic at $y$ wrt $x$ for every $y \notin U$ there is a neighborhood continuum $K_y$ of $y$ not containing $x$, and $X \setminus U$ is contained in $\cup K_y$. Since $X \setminus U$ is compact it's contained in finitely many such continua, say $K_1, K_2, \dots, K_n$. Then $V = X \setminus (\cup K_j)$ is a neighborhood of $x$ with $V \subset U$ and $X \setminus V = K_1 \cup \cdots \cup K_n$ has at most $n$ components. Therefore $X$ is slc at $x$ since $U$ was an arbitrary neighborhood of $x$. But $x$ was arbitrary, so $X$ is slc.
(note that the above implies: If $X$ is aposyndetic at $y$ wrt $x$ for every $y \in X \setminus \lbrace x \rbrace$ then $X$ is slc at $x$. As local properties slc is not equivalent to aposyndesis, even for planar curves)
Now assume that $X$ is slc and we show that $X$ is freely decomposable (this is the hardest part). Let $U$ be a neighborhood of $x$ with $y \notin \overline{U}$ such that $X \setminus U$ has only finitely many components $C_1, C_2, \dots, C_k$ with $y \in C_1$. Since $y \notin \overline{U}$ we have $y \in \text{int}(C_1)$. Note that since $U$ was open each $C_j$ is closed.
Since $X$ is slc at $y$ pick an open neighborhood $V \subset C_1$ of $y$ such that $X \setminus V$ has only finitely many components $D_1, D_2, \dots, D_n$ with $x \in D_1$ (similarly, $x \in \text{int}(D_1)$ and the $D_j$'s are closed). Then $\mathcal{A} = \lbrace C_1, \dots, C_k, D_1, \dots, D_n \rbrace$ is a finite cover of $X$ by closed, connected sets.
Then $C_1 \cup D_2, \cup \cdots \cup D_n$ is a neighborhood continuum of $y$ since $C_1$ is a neighborhood and $C_1 \cup D_i$ is connected for any $i$ by the boundary bumping theorem. Similarly, $D_1 \cup C_2 \cup \cdots \cup C_n = D_1$ since $U$ (and thus $U \cup C_2 \cup \cdots \cup C_n$ by bbp) is contained in $D_1$ since $D_1$ is a component of $X \setminus V$. Then these are a decomposition of $X$ between $x$ and $y$ as desired.
The above proofs work fine for Hausdorff continua. Below is the triangle basin $T$. $T$ is locally connected, and thus aposyndetic, at $v$, but is not slc at $v$. It's slc at $x$ but not aposyndetic at $x$ with respect to $v$.
- 203
-
With your suggestion, I found the theorem that aposyndetic = freely decomposable (even for non-metric continua) in F.B. Jones' 1941 paper, "Aposyndetic continua and certain boundary problems." – jinpa Feb 13 '21 at 15:08
-
Yeah, I wasn't sure about Hausdorff continua so I didn't want to assume. But I think it's just bbp and extension theorem stuff for the proof, so makes sense. – John Samples Feb 14 '21 at 00:02
-
Jones' proof is a little tricky; you can tease out the following elementary lemma that makes it work: Let $X$ be a connected space, with $U,V$ disjoint open subsets, $H$ a component of $X\setminus V$ containing $U$, and $K$ a component of $X\setminus U$ containing $V$. Then $H\cup K=X$. – jinpa Feb 14 '21 at 15:11
-
I have the proof given as freely decomposable $\implies$ aposyndetic $\implies$ slc with the final slc $\implies$ freely decomposable being the only tricky one. It doesn't use that theorem, it uses some finite ascent argument. The proof I'm looking at will work fine for Hausdorff continua as well. Anyway, I think I answered your question, so please feel free to give it the check mark. – John Samples Feb 14 '21 at 18:17
-
Check mark freely given. Could you tell me how you found this proof? It seems that when I google "freely decomposable" (or look it up on MathSci Net) I get either mappings or Jones' 1941 paper. – jinpa Feb 15 '21 at 22:26
-
Ok, I added the proof. – John Samples Feb 16 '21 at 00:48
-
Thanks for that argument! The proof I had in mind is more elementary and works for arbitrary connected topological spaces: you define aposyndesis in the usual way, only you replace "subcontinuum" with "closed connected subset." Likewise with free decomposability. You don't even need Hausdorff. (Of course, aposyndesis in this sense implies $T_1$.) – jinpa Feb 17 '21 at 10:28
