6
$\begingroup$

Question: Let $X$ be a continuum and $p \in X$. Under what conditions does there exist a compactification $\gamma (X-p)$ with $\gamma (X-p) - (X-p)$ connected and nondegenerate?

Throughout, $X$ is a continuum. That is to say, a compact connected Hausdorff space. (I'd be interested on any info for non-connected spaces too!)

The story is we want to remove a point $p \in X$ and replace it with a small subcontinuum. Formally we want a space $Y$ and continuous map $Y \to X$ such that all fibres are nontrivial save one, and the nontrivial fibre $F$ is a nondegenerate subcontinuum with void interior. The question is trivial if we allow $F$ to have interior as we can always take a copy of $[0,1]$ and glue it to $p$ at an endpoint.

When $X$ is metric this can always be done. For example we can resolve $X$ at the point $p$ onto the arc: consider the following subspace of $X \times [-1,1]$.

$\displaystyle Y = \Big \{ \Big (x, \sin \Big (\frac{1}{d(x,p)}\Big ) \Big): x \in X-p \Big\} \bigcup \Big ( \{p\} \times [-1,1]\Big )$.

Then the projection map onto the $x$-coordinate satisfies the above condition.

More generally we can resolve onto any locally connected continuum $F$, by recalling $F$ is the image of an arc and constructing a map from $[0,1)$ that visits each point cofinally many times.

When $X$ is not metric we cannot always do this. For example consider the long arc $[0,\omega_1]$. Removing the endpoint $p$ we get the long line whose Stone Čech compactification is known to be the same as the one-point compactification. Since the remainder of the Stone Čech maps onto the remainder of every compactification, we cannot replace $p$ with a nondegenerate continuum, as then that continuum would be the image of a point.

It is known exactly when $X-p$ admits a nontrivial remainder. Namely when $X-p$ has two disjoint closed non-compact subsets. Has it been studied whether there are any conditions for when a nontrivial connected remainder exists?

I know Smith has given a condition for when the Stone Čech remainder is connected (when $p$ fails to be a local cut point) but this is of no immediate use. For example if $X$ is a circle $(X-p)^*$ has exactly two components but the above shows we can always wrap $X-p$ around an arc. In this case we get something like a Warsaw circle.

Edit: You might try to generalise the resolution approach to something involving a gauge $\mathcal G $ of pseudometrics on $X$. But this throws up problems if $\rho(p,x)=0$ for some $x \ne p$ and $\rho \in \mathcal G$. The obvious solution is to replace $\mathcal G$ with a gauge that is positive definite at $p$ meaning $\rho(p,x)=0 \implies p=x$. But this cannot in general be done. In fact if a single $\rho$ is positive definite then $\{p\}$ is immediately a $G_\delta$ set which is not a guarantee for non-metric spaces.

$\endgroup$
5
  • $\begingroup$ As I understand the question, we can write $K=K(X,p)$ as the Stone-Cech remainder of $X\smallsetminus\{p\}$, and then the question boils down to understand whether $K$ admits a connected Hausdorff quotient (as a topological space) not reduced to a singleton. $\endgroup$
    – YCor
    Mar 12, 2018 at 19:42
  • $\begingroup$ $X\setminus \{p\}$ having disjoint closed sets limiting to $p$ is not enough; $X$ could be the one-point compactification of the double long-line ($X$ is like a circle). $X\setminus \{\infty\}$ has only two compactifications (one-point and two-point). $\endgroup$ Mar 12, 2018 at 20:39
  • 2
    $\begingroup$ If there are (A) infinitely many pairwise disjoint closed sets limiting to p, then the remainder of β(X∖{p}) should contain a copy of ω∗ which is certainly big enough to map onto nontrivial continua in place of p. Moreover, if (B) there is an unbounded continuous function f:X∖{p}→[0,∞), then X∖{p} has the arc as its remainder. I do not know if (A) and (B) are equivalent in this context. I would conjecture that every continuum remainder is degenerate if and only if $X\setminus \{p\}$ is pseudocompact. $\endgroup$ Mar 12, 2018 at 20:55
  • $\begingroup$ A separate stand-alone self-contained self-explanatory short formulation of the Question would be helpful. $\endgroup$
    – Wlod AA
    Mar 12, 2018 at 22:42
  • $\begingroup$ Actually I think it would be of benefit to isolate the question: which compact Hausdorff spaces admit a Hausdorff quotient topological space that is connected and not reduced to a singleton. $\endgroup$
    – YCor
    Mar 14, 2018 at 10:15

1 Answer 1

2
$\begingroup$

Given a compact Hausdorff space $X$ (not necessarily connected) and a point $p \in X$, the following are equivalent:

  1. $X - p$ has a compactification with non-trivial connected remainder.

  2. There is a continuous surjection $q: (X - p)^* \to I$, where $(X-p)^*$ denotes the Stone-Čech remainder of $X-p$ and $I = [0,1]$.

  3. $X-p$ has a compactification $\gamma(X-p)$ with remainder homeomorphic to $I$.

  4. There is a continuous $f: X-p \to I$ such that every point of $I$ is a cluster point of $f$ at $p$. (in other words, $\overline{f[U-p]} = I$ for every neighbourhood $U$ of $p$).

Sketch of a proof:

$(1. \implies 2.)$ $(X-p)^*$ can be mapped onto every other remainder and every non-degenerate continuum can be mapped onto $I$.

$(2. \implies 3.)$ Put $\gamma(X-p) = I \cup_q \beta(X-p)$. Since $X-p$ is open in $\beta(X-p)$, the inclusion $X-p \to \gamma(X-p)$ is a homeomorphic embedding.

$(3. \implies 4.)$ The identity mapping of $I$ has a continuous extension to $f: \gamma(X-p) \to I$. The desired property follows from the density of $X-p$ in $\gamma(X-p)$ and the correspondence between the neighbourhoods of $I \subset \gamma(X-p)$ and those of $p \in X$.

$(4. \implies 3. \implies 1.)$ The closure in $X \times I$ of the graph of $f$ is such a compactification. Clearly $I$ is connected.


Additional information on condition 1:

A compact Hausdorff space $K$ can be mapped continuously onto $I$ if and only if $K$ is not scattered. For sufficiency of this condition we need, by the Tietze-Urysohn extension theorem, only to see that there is some closed subset that can be mapped onto $I$. If $K$ is not totally disconnected, we can take this subset to be any nontrivial component. If $K$ is totally disconnected, then we can take its perfect kernel, which is zero-dimensional, compact and dense in itself and by repeated bisection find a mapping onto $\{0,1\}^\omega$, which maps onto $I$ in a straightforward way.

For necessity, observe that any continous surjection $q: K \to I$ is proper, which implies that $K$ has a closed subset $L$ such that $q|_L$ is irreducible. Since $q|_L$ is also closed, it must map every isolated point of $L$ to an isolated point of $I$, which has no isolated points. Hence $L$ is dense in itself and $K$ is not scattered.


Additional information on condition 4:

A easy sufficient, but not necessary, condition for the existence of such an $f$ is that $X-p$ is not pseudocompact. To see this, suppose $g$ is an unbounded nonnegative function on $X-p$. Then $g[X-p]$ contains an unbounded subset $W$ with order type $\omega$. Since $W$ is discrete and closed in $[0,\infty)$ there is a continuous $h: [0,\infty) \to I$ such that $h[W]$ is dense in $I$. Putting $f = h \circ g$ gives the desired result, since if $U$ is a neighbourhood of $p$, $f[U-p]$ contains all but finitely many points of $h[W]$.

An example of a pseudocompact locally compact Hausdorff space with a connected Stone-Čech remainder is easily constructed by taking the product of the long line with a non-degenerate continuum.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.