6

The motivation is this question. I was quite optimistic in the formalization I provided of the question, it would indeed be the case that such a maximal sub topology exists. Now, I learn that it's less clear than I initially thought (after seeing comment and answers provided).

I first bring in a definition:

Stable topology $\tau$ under $U$: A topology $\tau$ is stable under a non-empty open subset $U$, if $U$ is an open set in the topology $\tau$ , and, there exists a maximal sub topology of $\tau$ not containing $U$.

This leads to a shorter formulation of my question,

Is $\mathbb{R}$ stable under removal of an epsilon interval about a point?

  • Uhm, maximal in the sense finer than all or, there isn't one finer than it ? – Frousse Apr 23 '23 at 15:42
  • I thought they would be same @Frousse hmm – tryst with freedom Apr 23 '23 at 15:58
  • Nope. I'll edit to include an answer to the other interpretation (this time it's a yes) – Frousse Apr 23 '23 at 15:59
  • 1
    I think you misunderstood why @AnneBauval recommended you ask a new question. My answer in the linked question already showed that there isn't a unique maximal, ie one finer than all, but it was unclear whether there is one such that there isn't one that is finer. – ronno Apr 24 '23 at 11:54
  • Ah okay, I see what you mean. I understood the abstract of your answer, if not the exact details. It had slipped my mind when making this question, still, thank you @ronno – tryst with freedom Apr 24 '23 at 14:35

1 Answers1

4

Let $\tau$ be the standard topology on $\mathbb R$ and let $U=(a,b)$ a nonempty open interval.

There are two notions of maximal:

Answer 1: If Maximal sub-topology means: finer than all other sub-topologies stable under $U$.

We'll construct two topologies $\tau_a$ and $\tau_b$ such that $U$ is not an open set in either, but such that the join $\tau_a \vee \tau_b$ is the usual topology $\tau$ on $\mathbb R$. This would answer your question in the negative.

Choose $p,r$ distinct points in $(a,b)$. The idea is to have $\tau_a$ the simplest topology where $p$ is topologically indistinguishable from $a$, and similarly $\tau_b$ the simplest topology in which $r$ is indistinguishable from $b$. So, for example, taking the kolmogorov quotient of $\tau_a$, glues $p$ and $a$.

Here's an explicit description of the neighborhood system of $\tau_a$: Every $x\in \mathbb R\backslash\{p,a\}$ has the same neighborhoods as in $\tau$, and for $y\in\{p,a\}$, the neighborhoods of $y$ are the sets $N_a\cup N_p$ where $N_a$ is a neighborhood of $a$ and $N_p$ a neighborhood of $p$ in $\tau$.

$U\not\in\tau_a$ because $U$ is not in the neighborhood of $p$. Indeed, any open set containing $p$ must also contains $a$.

A similar construction for $\tau_b$ and $r$.

It's not hard to see that $\tau_a\vee\tau_b=\tau$. Indeed, $\tau=\{A\cap B,\ A\in\tau_a,\ B\in\tau_b\}$.

Answer 2: If Maximal sub topology means: is contained in no other sub topology stable under $U$.

This time the answer is positive. The idea is to construct the simplest topology $\tau'_a$ such that $p\in U$ is 'asymmetrically glued' to $a$.

An explicit description of the neighborhood system of $\tau'$ is as follows: For $x\in\mathbb R\backslash\{p\}$, the neighborhoods of $x$ are the same as in $\tau$. The neighborhoods of $p$ are the sets $N_a\cup N_p$ where $N_a$ is a neighborhood of $a$ and $N_p$ a neighborhood of $p$ in $\tau$.

Again $U\not\in\tau'$ because it's not a neighborhood of $p$. Any open set containing $p$ must also contain $a$.

On the other hand, extending $\tau'$ by an open set $V\in\tau\backslash \tau'$ retrieves $\tau$. Indeed there's an open interval $I\subset V$ such that $p\in I$ and $a\not\in I$. And any open sub-interval of $I$ is of the form $V\cap A$ with $A\in\tau'$. So we get back all neighborhoods of $p$.

Remark: In the first answer, two points $p,r\in U$ are used to keep things nice and symmetric. But, following the second construction, one point is in fact sufficient. Also, $a$ and $b$ could be replaced with any points not in $U$.

Frousse
  • 434
  • +1 This is similar but more efficient than my answer to the linked question, but I think OP is confused about why they were recommended to ask a new question in the first place. – ronno Apr 24 '23 at 11:51
  • @ronno To be fair, Op's new question technically isn't ambiguous. A maximal element of a poset usually means 'smaller than none' . I should've seen that or read the linked post more carefully. – Frousse Apr 24 '23 at 13:06
  • Which is why I didn't think to edit the question when it was initially posted. Since the OP was explicitly confused about the distinction in the comments, I think this is valid as an answer, even if redundant for myself personally. – ronno Apr 24 '23 at 13:12