7

Suppose $X$ is a topological space such that the open sets form a family totally ordered by inclusion. Equivalently, the closed sets form a family totally ordered by inclusion. Such a space is also called hereditarily connected as an equivalent characterization is that all subspaces are connected. A simple example would be the Sierpinski space.

Question 1: Is every nonempty hereditarily connected space contractible?

That is, does there exist a homotopy between the identity on $X$ and a constant map?

If that is true, one could also wonder about a stronger result:

Question 2: Is every nonempty ultraconnected space contractible?

(A space is ultraconnected if any two nonempty closed sets meet.)

PatrickR
  • 7,165

2 Answers2

5

Here is a proof that all nonempty hereditarily connected spaces are contractible:

Per Geoffrey’s suggestion, I’m proving the following more general result:

Theorem: Let $X$ be a space which has a point $x_0 \in X$ satisfying the property that it is comparable to every $x \in X$ in the specialization preorder, i.e., for any $x \in X$, we have either $x_0 \in \overline{\{x\}}$ or $x \in \overline{\{x_0\}}$ (or both). Then $X$ is contractible.

Proof: We claim the following map $f: X \times [0, 1] \to X$ is a contraction:

$$f(x, t) = \begin{cases} x &, \text{ if }x_0 \in \overline{\{x\}} \text{ and } t < 1\\ x &, \text{ if }x_0 \notin \overline{\{x\}} \text{ and }t = 0\\ x_0 &, \text{ otherwise} \end{cases}$$

It is clear that $f(x, 0) = x$ and $f(x, 1) = x_0$ for all $x$, so it suffices to show $f$ is continuous. Let $K \subset X$ be closed. There are two cases:

  1. $x_0 \in K$. Now, if $x_0 \notin \overline{\{x\}}$, then by assumption on $x_0$, we have $x \in \overline{\{x_0\}} \subset K$. That is, any $x \in X$ s.t. $x_0 \notin \overline{\{x\}}$ is in $K$. From this we easily see that,

$$f^{-1}(K)^c = f^{-1}(K^c) = K^c \times [0, 1)$$

$\;\;\;\;\;\;$The latter set is open, so $f^{-1}(K)$ is closed.

  1. $x_0 \notin K$. Then, if $x_0 \in \overline{\{x\}}$, we must have $x \notin K$, as otherwise we would have $x_0 \in \overline{\{x\}} \subset K$. That is, any $x \in X$ s.t. $x_0 \in \overline{\{x\}}$ is outside $K$. From this we easily see that,

$$f^{-1}(K) = K \times \{0\}$$

$\;\;\;\;\;\;$So, $f^{-1}(K)$ is closed.

This shows $f$ is continuous. Thus, $X$ is contractible. $\square$

Corollary: If $X$ is nonempty and hereditarily connected, then it is contractible.

Proof: In a hereditarily connected space, every two points are comparable in the specialization preorder. Indeed, $\overline{\{x\}}$ and $\overline{\{y\}}$ are both closed, so either $\overline{\{x\}} \subset \overline{\{y\}}$, in which case $x \in \overline{\{y\}}$; or $\overline{\{y\}} \subset \overline{\{x\}}$, in which case $y \in \overline{\{x\}}$. $\square$

David Gao
  • 22,850
  • 9
  • 28
  • 48
  • 2
    I found a counterexample for ultraconnected spaces. It ended up being a quite long proof, and I've already nearly finished typing it up before my browser suddenly crashed... So, I now need a few hours to type it up again. Before that, I made some minor edits to this answer to remove the sentence saying I'm not sure about ultraconnected spaces, as well as to improve the typesetting a bit. (I'm writing a different answer instead of editing this one for ultraconnected spaces... That way, MSE will save my progress even if my browser crashes again.) – David Gao Nov 20 '24 at 06:14
  • Awesome. Looking forward to it (and sorry about the crash). – PatrickR Nov 20 '24 at 06:26
  • @PatrickR It was quite an awful experience to see a few hours of work vanish with a crash... But oh well, at least the second time of typing this takes less time than the first (and than I have expected). It's done now. – David Gao Nov 20 '24 at 08:08
3

Here is an example of a nonempty ultraconnected space that is not contractible:

We say a partial function $f$ from $\mathbb{Z}$ to $\{0, 1\}$ is admissible if it satisfies all of the following three conditions,

  1. There exists $n \in \mathbb{Z}$ s.t. the domain of $f$ is $\mathbb{Z}_{\leq n}$. We call $n$ the length of $f$, which shall be denoted by $l(f)$.
  2. There exists $n \in \mathbb{Z}$ s.t. $f(m) = 0$ for all $m \leq n$.
  3. Either $\text{range}(f) = \{0\}$ and $l(f) \leq 0$; or there exists $n < 0$ s.t. $f(n) = 1$, and furthermore if $n_0 = \min\{n < 0: f(n) = 1\}$, then $l(f) \leq -n_0$.

Let $X$ be the space of all admissible partial functions from $\mathbb{Z}$ to $\{0, 1\}$. We note that $X$ is countable.

For partial functions $f, g$ from $\mathbb{Z}$ to $\{0, 1\}$, we shall say $g$ is a truncation of $f$, denoted by $g \leq f$, if there exists $n \in \mathbb{Z}$ s.t. $g = f|_{\text{dom}(f) \cap \mathbb{Z}_{\leq n}}$. We observe that truncations of admissible partial functions are admissible. Furthermore, $\leq$ is a partial order on $X$ and makes $X$ into a meet-semilattice. For $f, g \in X$, we shall write $f \wedge g$ for the meet of $f$ and $g$ under $\leq$.

Now, we topologize $X$ as follows: for each $x \in X$, let,

$$T_x = \{y \in X: y \leq x\}$$

And, for each $n \in \mathbb{Z}$, let,

$$K_n = \{y \in X: l(y) \leq n\}$$

We then equip $X$ with the topology generated by the closed subbasis,

$$\{T_x: x \in X\} \cup \{K_n: n \in \mathbb{Z}\}$$

Lemma 1: $\overline{\{x\}} = T_x$ for all $x \in X$.

Proof: $T_x$ is closed and contains $x$, so $\overline{\{x\}} \subset T_x$. Conversely, if $x \in T_y$, then $x \leq y$ so $T_x \subset T_y$. If $x \in K_n$, then $l(x) \leq n$ so $T_x \subset K_n$. From this, it is easy to conclude that $T_x \subset \overline{\{x\}}$. $\square$

Corollary 1: $X$ is ultraconnected.

Proof: It suffices to show that, for any $x, y \in X$, $\overline{\{x\}} \cap \overline{\{y\}} \neq \varnothing$. Indeed, we have $\overline{\{x\}} \cap \overline{\{y\}} = T_x \cap T_y \ni x \wedge y$. $\square$

For any $x \in X$, it is easy to see that there are only finitely many $y \in X$ s.t. $y \geq x$. Thus,

$$E_x = \bigcup_{y: y \geq x} T_y$$

is closed.

Lemma 2: Any path $f: [0, 1] \to X$ has finite range. Furthermore, $\text{range}(f)$ has a minimum under $\leq$.

Proof: Note that $\text{range}(f)$ is compact. Thus, as $n$ decreases, $K_n \cap \text{range}(f)$ form a decreasing sequence of closed subsets of a compact space whose intersection is empty. Thus, there exists $n \in \mathbb{Z}$ s.t. $K_n \cap \text{range}(f) = \varnothing$, i.e., $\text{range}(f) \subset K_n^c$. On the other hand, $K_n^c$ can be written as a countable disjoint union of (relatively) closed sets:

$$K_n^c = \bigsqcup_{x: l(x) = n + 1} (E_x \cap K_n^c)$$

So, $[0, 1]$ is a countable disjoint union of closed sets $f^{-1}(E_x \cap K_n^c)$ where $x$ ranges over all elements of $X$ of length $n + 1$. By Sierpiński's theorem, exactly one of these closed sets is nonempty, i.e., there exists $x \in X$ with $l(x) = n + 1$ and $\text{range}(f) \subset E_x \cap K_n^c$. The latter set is easily seen to be finite.

Now, as $\text{range}(f)$ is finite, we may let $\{x_1, \cdots, x_n\}$ be the elements of $\text{range}(f)$ that are minimal under $\leq$. But then $\text{range}(f)$ can be written as a finite disjoint union of nonempty (relatively) closed sets,

$$\text{range}(f) = \bigsqcup_{i=1}^n (E_{x_i} \cap \text{range}(f))$$

But $\text{range}(f)$ is connected, so $n = 1$, i.e., $\text{range}(f)$ has a minimum. $\square$

Corollary 2: If $f: [0, 1] \to X$ is a path from $x \in X$ to $y \in X$, then there exists $t \in [0, 1]$ s.t. $f(t) \leq x \wedge y$.

Proof: By Lemma 2, $\min(\text{range}(f))$ exists, so there exists $t \in [0, 1]$ s.t.,

$$f(t) = \min(\text{range}(f)) \leq f(0) \wedge f(1) = x \wedge y$$

$\square$

Now, let us assume $X$ is contractible and $f: X \times [0, 1] \to X$ is a contraction. Let $x_0 \in X$ be the point $f$ contracts to, i.e., $f(x, 1) = x_0$ for all $x \in X$. For any $y \in X$, let $\Pi_y = f^{-1}(T_y) \subset X \times [0, 1]$, which is closed. Let $p: X \times [0, 1] \to X$ be the projection onto the first coordinate.

Lemma 3: For any $y \in X$, $p(\Pi_y)$ contains elements of arbitrarily large lengths.

Proof: For any $n \in \mathbb{Z}$, it is easy to show that there exists $x \in X$ s.t. $l(x) \geq n$ and $x \wedge x_0 \leq y$. (Indeed, such an $x$ can be chosen by modifying a $0$ at a sufficiently small index in $x_0 \wedge y$ to $1$, then adding sufficiently many numbers at larger indices to reach the desired length.) Then $t \mapsto f(x, t)$ is a path from $x$ to $x_0$, so by Corollary 2, there exists $t \in [0, 1]$ s.t. $f(x, t) \leq x \wedge x_0 \leq y$. Whence, $(x, t) \in \Pi_y$ and $x \in p(\Pi_y)$. $\square$

Fix $y \in X$. We note that the following is a closed subbasis for $X \times [0, 1]$:

$$\{T_x \times [0, 1]: x \in X\} \cup \{K_n \times [0, 1]: n \in \mathbb{Z}\} \cup \{X \times L: L \subset [0, 1] \text{ closed}\}$$

We note that $\{K_n: n \in \mathbb{Z}\}$ is linearly ordered by inclusion. Thus, as $\Pi_y$ is closed, it can be written as,

$$\Pi_y = \bigcap_{i \in I}\left(\left[\bigcup_{j \in J_i} (T_{x_j} \times [0, 1])\right] \cup (K_{n_i} \times [0, 1]) \cup (X \times L_i)\right)$$

where $I$ is a possibly infinite index set; for each $i \in I$, $J_i$ is a finite index set; for each $j \in J_i$ where $i \in I$, $x_j \in X$; for each $i \in I$, $n_i \in \mathbb{Z} \cup \{-\infty\}$ (here, we abuse the notation slightly and let $K_{-\infty} = \varnothing$ in order to accommodate the case where this term does not appear in the union); and for each $i \in I$, $L_i \subset [0, 1]$ is closed. Hence, for each $i \in I$, by Lemma 3, the following set has elements of arbitrarily large lengths.

$$p\left(\left[\bigcup_{j \in J_i} (T_{x_j} \times [0, 1])\right] \cup (K_{n_i} \times [0, 1]) \cup (X \times L_i)\right) = \begin{cases} \left(\bigcup_{j \in J_i} T_{x_j}\right) \cup K_{n_i} &, \text{ if }L_i = \varnothing\\ X &, \text{ if }L_i \neq \varnothing\end{cases}$$

Since $J_i$ is finite, it is easy to see that elements of $\left(\bigcup_{j \in J_i} T_{x_j}\right) \cup K_{n_i}$ have a uniform bound on lengths, so $L_i \neq \varnothing$.

Now, consider a nonempty finite subset $I_0 \subset I$ and take the intersection $U$ of,

$$\left[\bigcup_{j \in J_i} (T_{x_j} \times [0, 1])\right] \cup (K_{n_i} \times [0, 1]) \cup (X \times L_i)$$

where $i$ ranges over $I_0$. We note that the intersection of (nonzero) finitely many sets of the form $T_x$ is again of the form $T_x$; the intersection of (nonzero) finitely many $T_x$ and (nonzero) finitely many $K_n$ is of the form $T_x$; and the intersection of (nonzero) finitely many sets of the form $K_n$ is of the form $K_n$. From this, we see that $U$ can be written as,

$$\left[\bigcup_{j \in J} (T_{x_j} \times L'_j)\right] \cup \left[\bigcup_{j \in J'} (K_{n_j} \times L''_j)\right] \cup (X \times L)$$

where $J, J'$ are finite index sets; for each $j \in J$, $L'_j \subset [0, 1]$ is closed; for each $j \in J'$, $n_j \in \mathbb{Z} \cup \{-\infty\}$ and $L''_j \subset [0, 1]$ is closed; and $L \subset [0, 1]$ is closed. Furthermore, $L = \bigcap_{i \in I_0} L_i$. As $\Pi_y \subset U$, $p(U)$ has elements of arbitrarily large lengths, so the same argument shows $L \neq \varnothing$. That is,

$$\bigcap_{i \in I_0} L_i \neq \varnothing, \forall I_0 \subset I \text{ nonempty finite}$$

As $[0, 1]$ is compact, this implies $\bigcap_{i \in I} L_i \neq \varnothing$. Since $X \times \left(\bigcap_{i \in I} L_i\right) \subset \Pi_y$, we have the following set,

$$S_y = \{t \in [0, 1]: X \times \{t\} \subset \Pi_y\} = \bigcap_{x \in X} \{t \in [0, 1]: f(x, t) \in T_y\}$$

is nonempty closed. Note that as $T_{y_1} \cap T_{y_2} = T_{y_1 \wedge y_2}$, we have $S_{y_1} \cap S_{y_2} = S_{y_1 \wedge y_2}$. Thus, $\{S_y\}_{y \in X}$ is a downward directed collection of nonempty closed subsets of $[0, 1]$. By compactness, we then have $\bigcap_{y \in X} S_y \neq \varnothing$. But then, for any $t \in \bigcap_{y \in X} S_y$ and $x \in X$, we have,

$$f(x, t) \in \bigcap_{y \in X} T_y = \varnothing$$

This is a contradiction. Hence,

Theorem 1: $X$ is not contractible. Thus, there exists a nonempty ultraconnected space that is not contractible.


Here are two additional remarks regarding $X$:

Remark 1: Since $x \in \overline{\{y\}} = T_y$ iff $x \leq y$, the specialization preorder on $X$ coincide with $X$.

Remark 2: Here are some additional properties that $X$ enjoys:

  • $X$ is $T_0$ but not $T_1$. Indeed, if $x \in \overline{\{y\}} = T_y$ and $y \in \overline{\{x\}} = T_x$, then $x \leq y$ and $y \leq x$, so $x = y$ and thus $X$ is $T_0$. For any $x \in X$, $\overline{\{x\}} = T_x$ contains infinitely many elements distinct from $x$, so no point in $X$ is closed and $X$ is not $T_1$.
  • $X$ is second countable and thus also first countable. Indeed, the topology on $X$ is generated by a countable closed subbasis, whence also a countable open basis.
  • $X$ is not compact, in fact, not even countably compact. Indeed, as $n$ decreases, $K_n$ is a decreasing sequence of nonempty closed sets whose intersection is empty, contradicting countable compactness.
  • However, $X$ is locally compact, in the sense that every point $x \in X$ has a local basis of compact neighborhoods. Indeed, for any $n \in \mathbb{Z}$, we claim that any subset $L \subset K_n^c$ is compact. (In other words, $L \subset X$ is compact if there is a uniform lower bound on lengths of elements of $L$. This is clearly an if and only if, as the proof that $X$ is not compact can be easily adapted to show that any $L \subset X$ with no uniform lower bound on lengths of elements is not compact. This latter fact was also used in the proof of Lemma 2.) Since $K_n^c$ is open for all $n \in \mathbb{Z}$ and $\bigcup_{n \in \mathbb{Z}} K_n^c = X$, this is clearly enough to imply local compactness. By Alexander subbase theorem, it suffices to show that, if,

$$L \cap \left(\bigcap_{j \in J} T_{x_j}\right) \cap \left(\bigcap_{j’ \in J’} K_{n_{j’}}\right) = \varnothing$$

$\;\;\;\;\;\;$where $J, J’$ are possibly infinite index sets; for each $j \in J$, $x_j \in X$; and for each $j’ \in J’$, $n_{j’} \in \mathbb{Z}$; then there exists finite $J_0 \subset J$ and $J’_0 \subset J’$ s.t.,

$$L \cap \left(\bigcap_{j \in J_0} T_{x_j}\right) \cap \left(\bigcap_{j’ \in J’_0} K_{n_{j’}}\right) = \varnothing$$

$\;\;\;\;\;\;$There are several cases to consider:

$\;\;\;\;\;\;$ 1. If $\{n_{j’}: j’ \in J’\}$ is not bounded below, then there exists $j’ \in J’$ s.t. $n_{j’} \leq n$. But then $\varnothing = K_n^c \cap K_{n_{j’}} \supset L \cap K_{n_{j’}}$.

$\;\;\;\;\;\;$ 2. If $J$ is nonempty, consider the downward directed net,

$$\left\{\bigwedge_{j \in J_0} x_j: J_0 \subset J \text{ nonempty finite}\right\}$$

$\;\;\;\;\;\;\;\;\;\;$in $X$. If this net has no minimum, then there exists a sequence $J_0^{(k)}$ of nonempty finite subsets of $J$ s.t., $\bigwedge_{j \in J_0^{(k)}} x_j$ is strictly decreasing as $k \to \infty$. In particular, there exists some $k$ s.t. $l\left(\bigwedge_{j \in J_0^{(k)}} x_j\right) \leq n$. But then,

$$L \cap \left(\bigcap_{j \in J_0^{(k)}} T_{x_j}\right) \subset K_n^c \cap T_{\bigwedge_{j \in J_0^{(k)}} x_j} = \varnothing$$

$\;\;\;\;\;\;$ 3. If both $J$ and $J’$ are empty, then there is nothing to prove.

$\;\;\;\;\;\;$ 4. If $J$ is nonempty and $\left\{\bigwedge_{j \in J_0} x_j: J_0 \subset J \text{ nonempty finite}\right\}$ has a minimum, then there exists $J_0 \subset J$ nonempty finite s.t. $\bigwedge_{j_0 \in J_0} x_{j_0} \leq x_j$ for all $j \in J$, so,

$$\bigcap_{j \in J_0} T_{x_j} = T_{\bigwedge_{j \in J} x_j} \subset \bigcap_{j \in J} T_{x_j} \subset \bigcap_{j \in J_0} T_{x_j}$$

$\;\;\;\;\;\;\;\;\;\;$Now, if $J’$ is empty, then,

$$\varnothing = L \cap \left(\bigcap_{j \in J} T_{x_j}\right) \cap \left(\bigcap_{j’ \in J’} K_{n_{j’}}\right) = L \cap \left(\bigcap_{j \in J_0} T_{x_j}\right)$$

$\;\;\;\;\;\;\;\;\;\;$If, instead, $J’$ is nonempty, but $\{n_{j’}: j’ \in J’\}$ is bounded below, then $\min\{n_{j’}: j’ \in J’\}$ exists and there is some $j’_0 \in J’$ s.t. $n_{j’_0} = \min\{n_{j’}: j’ \in J’\}$. In this case, we have,

$$\bigcap_{j’ \in J’} K_{n_{j’}} = K_{n_{j’_0}}$$

$\;\;\;\;\;\;\;\;\;\;$So,

$$\varnothing = L \cap \left(\bigcap_{j \in J} T_{x_j}\right) \cap \left(\bigcap_{j’ \in J’} K_{n_{j’}}\right) = L \cap \left(\bigcap_{j \in J_0} T_{x_j}\right) \cap K_{n_{j’_0}}$$

$\;\;\;\;\;\;\;\;\;\;$Finally, if $J$ is empty, but $J’$ is nonempty and $\{n_{j’}: j’ \in J’\}$ is bounded below, then,

$$\varnothing = L \cap \left(\bigcap_{j \in J} T_{x_j}\right) \cap \left(\bigcap_{j’ \in J’} K_{n_{j’}}\right) = L \cap K_{n_{j’_0}}$$

$\;\;\;\;\;\;\;\;\;\;$where, again, $j’_0 \in J’$ is s.t. $n_{j’_0} = \min\{n_{j’}: j’ \in J’\}$.

  • $X$ is path connected. Indeed, all ultraconncted spaces are path connected. See, for example, here.
  • In fact, we even have $X$ is simply connected. It suffices to show any loop $f: S^1 \to X$ is null-homotopic. By Lemma 2, $\text{range}(f)$ has a minimum under $\leq$. Per Remark 1, $\text{range}(f)$ thus has a minimum under the specialization preorder. By the Theorem in the other answer of mine to this question (see also here), $\text{range}(f)$ is contractible. From there it is easy to see $f$ is indeed null-homotopic.
  • Even better, $X$ is weakly contractible, i.e., all its homotopy groups vanish. Indeed, as Sierpiński's theorem holds for any compact connected Hausdorff space, it is easy to see that Lemma 2 still holds when $[0, 1]$ is replaced by any compact connected Hausdorff space, in particular $S^n$ for any $n \geq 1$. Then the same proof that $X$ is simply connected shows $\pi_n(X) = 0$ for all $n \geq 1$.
  • $X$ is hyperconnected (also known as irreducible), i.e., any two nonempty open subsets of $X$ intersect. Given an open basis for the topology, in order to verify hyperconnectedness, it clearly suffices to show any two nonempty basic open sets intersect. Dualizing to closed basis, it suffices to show the union of any two basic closed proper subsets of $X$ cannot be $X$. Since the topology on $X$ is defined using a closed subbasis, we see that it suffices to show any finite union of elements of the closed subbasis is not the entire space. This is true because any finite union of $T_x$ and $K_n$ has a uniform upper bound on lengths of elements.
  • $X$ is meager, i.e., it is a countable union of nowhere dense sets. Indeed, for each $n \in \mathbb{Z}$, $K_n$ is closed while $K_n^c$ is nonempty open. Since $X$ is hyperconnected, if $K_n$ has interior, then $K_n \cap K_n^c$ would be nonempty, which is a contradiction. Thus, $K_n$ is nowhere dense for all $n$. Now, just observe that $X = \bigcup_{n \in \mathbb{Z}} K_n$.
  • $X$ is not Alexandrov. Indeed, fix $n \in \mathbb{Z}$ and $x \in X$ with $l(x) = n$. If $X$ were Alexandrov, then,

$$U = \bigcup_{y \in X: l(y) = n, y \neq x} T_y$$

$\;\;\;\;\;\;$would be closed and does not contain $x$. However, it is not hard to see that sets of the form,

$$K_{n-1}^c \cap \left(\bigcap_{i=1}^k T_{x_i}^c\right)$$

$\;\;\;\;\;\;$where $x_i \not\geq x$, form an open neighborhood basis of $x$. However, any set of the above form intersects $U$. Indeed, an element of the intersection can be constructed by modifying a $0$ at a sufficiently small index in $x \wedge \left(\bigwedge_{i=1}^k x_i\right)$ to $1$, then adding sufficiently many numbers at larger indices so that it has length exactly $n$. Thus, $x$ is in the closure of, but not in, $U$, so $U$ is not closed, a contradiction.

David Gao
  • 22,850
  • 9
  • 28
  • 48
  • 3
    Impressive! Will read it in detail later today. – PatrickR Nov 20 '24 at 22:49
  • In the proof of Lemma 2, if $x\in X$ has length $n+1$, isn't the case that $E_x\subseteq K_n^c$, so no need to take intersection with the complement of $K_n$? – PatrickR Nov 21 '24 at 07:25
  • 2
    @PatrickR No. You might have misunderstood the definition. $E_x$ is the union of $T_y$ for all $y \geq x$. That means it includes both extensions of $x$, as well as truncations of $x$. Intersecting with $K_n^c$ removes all proper truncations of $x$. – David Gao Nov 21 '24 at 07:28
  • 2
    Hopefully that’s the final edit - I just added one more property, namely, $X$ is weakly contractible. As long as I don’t find any more property I’d like to add to the list, that should be all. – David Gao Nov 21 '24 at 23:30
  • 2
    Great. I'm still working on it. Have to get into the meat of things after Lemma 3 :) – PatrickR Nov 22 '24 at 06:45
  • 2
    @PatrickR Yeah, it is a quite long and technical argument. The part after Lemma 3 and the proof that $X$ is locally compact in Remark 2 were definitely the hardest and most notationally messy parts to write. – David Gao Nov 22 '24 at 07:28
  • 2
    Beautiful argument (+1). I am curious, was the inspiration for this example some known kind of construction? – PatrickR Nov 24 '24 at 04:36
  • 2
    @PatrickR Very loosely, it was inspired by the diamond poset space. I considered a construction using a closed subbasis instead of an open subbasis to make it easier to ensure ultraconnectedness. This requires a meet-semilattice, and the easiest way seems to be using a binary tree. To ensure there’s no focal point (which would imply contractibility), the binary tree cannot have a root. Everything sort of grew from there, adding technical conditions to ensure the space is not contractible. – David Gao Nov 24 '24 at 05:09
  • 1
    Thank you for your more than brilliant answers, David. I hope that your browser doesn't crash again. – Almanzoris Nov 25 '24 at 07:11
  • 1
    @Almanzoris Thanks! I hope so as well, but at least I've learned the lesson to periodically make copies when writing a long answer. – David Gao Nov 25 '24 at 07:20