3

I'm trying to prove, as a book excercise, the following statement:

A neighbourhood space $(X, \mathcal{N})$ is a topological space iff each neighbourhood filter $\mathcal{N}(x)$ has a filter base consisting of open sets.

I'm using the following definitions and theorems:

  1. A neighbourhood structure $\mathcal{N}$ on a set $X$ is an assignment to each $x \in X$ of a filter $\mathcal{N}(x)$ on $X$ all of whose elements contain the point $x$. The pair $(X,\mathcal{N})$ is called a neighbourhood space and the filter $\mathcal{N}(x)$ is called the neighbourhood filter of the point $x \in X$.
  2. A topological space is a neighbourhood space $(X,\mathcal{N})$ in which, for all $x \in X$ and for all $N \in \mathcal{N}(x)$, there exists $N^\star \in \mathcal{N}(x)$ such that, for all $y \in N^\star$, $N \in \mathcal{N}(y)$.
  3. A non-empty collection $\mathcal{B}$ of subsets of $X$ is a base for a specific filter $\mathcal{F}$ on $X$ iff (i) $\mathcal{B} \subseteq \mathcal{F}$ and (ii) if $A \in \mathcal{F}$, there exists $B \in \mathcal{B}$ such that $B \subseteq A$.

I was able to prove that a neighbourhood space satisfying the condition is a topological space: as all neighbourhood filters have an open base, there exists $B$ open and $N$ neighbourhood of $x \in X$ such that $B \subseteq N$. As $B$ is open, it contains a neigbourhood for each of its points, therefore $N$ is a neigbourhood of all points in $B$ because filters are upward closed, and also $B$ is a neighbourhood of $x$, so the definition of a topological space is met.

However, I'm having trouble with the other implication. Let $x \in X$ and $N \in \mathcal{N}(x)$. I know that $\exists N^\star \in \mathcal{N}(x)$ such that $\forall y \in N^\star : N \in \mathcal{N}(y)$, and want to see that there exists an open neighbourhood $B \subseteq N$; that would mean there exists a collection of open neighbourhoods $\mathcal{B}$ finer than $\mathcal{N}(x)$, and that would mean $\mathcal{B}$ is a filter base of open sets. However I can't find a way of arriving at that statement.

Am I on the right track, or is there something I'm overlooking/making some mistake?

GBRGR
  • 33
  • I'm assuming you're using axioms very similar to https://en.wikipedia.org/wiki/Topological_space#Neighbourhoods_definition ? – Henno Brandsma Feb 27 '16 at 06:00
  • It would help if you stated the exact axioms for your neighbourhood space, as well as the proof in the other direction. – Henno Brandsma Feb 27 '16 at 08:06
  • @HennoBrandsma Yes, the definitions are quite similar. I've added the axioms and the proof in the other direction to the post as you asked. – GBRGR Feb 27 '16 at 21:41

1 Answers1

1

Suppose $(X,\mathcal{N})$ is a topological space, so it satisfies the second axiom in your question. This is quite an important axiom as it ties the different neighbourhood filters (for different points) together. Otherwise we could just arbitarily assign filters for each point and we'd get no real spatial "structure".

Suppose $x \in X$ and $N \in \mathcal{N}(x)$. Define $$N^\circ = \{y \in X: N \in \mathcal{N}(y)\}\text{.}$$

This is (intuitively, and later more formally) the interior of $N$ in this topology, as it will turn out. In words, all $y$ that $N$ is a neighbourhood of.

  1. $x \in N^\circ$, which is trivial, as this is how $N$ was chosen in the first place.

  2. $N^\circ \subseteq N$, for pick $y \in N^\circ$, then $N \in \mathcal{N}(y)$ which means, by part of axiom 1, that $y \in N$.

  3. Suppose that $y \in N^\circ$, so $N \in \mathcal{N}(y)$. Then apply axiom 2. to this $y$ and $N$ to get $N^\ast \in \mathcal{N}(y)$ such that for all $z \in N^\ast$ we have $N \in \mathcal{N}(z)$. Note that this means that all these $z$ are in $N^\circ$ by definition, and so $N^\ast \subseteq N^\circ$ (we won't have equality in general, but we don't need it). But as $N^\ast \in \mathcal{N}(y)$, as $\mathcal{N}(y)$ is a filter, so upwards closed, $N^\circ \in \mathcal{N}(y)$ as well. So note what we have shown: we start with any $y \in N^\circ$ and show $N^\circ$ is a neighbourhood of $y$. So $N^\circ$ is open by the usual definition of openness in neighbourhood spaces.

So for every $N \in \mathcal{N}(x)$ we have found an open subset $N^\circ$ of $N$ that is open and a neighbourhood of $x$ as well (as $x \in N^\circ$ and $N^\circ$ is open!). So the open neighbourhoods of $x$ form an open filter base at $x$. We can just define $\mathcal{B}(x) = \{N^\circ: N \in \mathcal{N}(x)\}$ to make it explicit.

Henno Brandsma
  • 242,131
  • Is $M$ equal to $\operatorname{Int}_\mathcal{T}(N)$? In that case, I think I've understood what you meant. With the definitions I was using, I'd say the collection of interiors of members of $\mathcal{N}(x)$ is an open filter base and then $(X, \mathcal{N})$ is a topological space. – GBRGR Feb 27 '16 at 21:52
  • Sorry, I meant to say "... is an open filter base and then topological spaces meet the condition." – GBRGR Feb 27 '16 at 22:20
  • @GGR I did a total rewrite fitting your axioms and showing all details. $N^\ast$ need not be open, e.g. for the usual topology and $x=1$ we could have $N = (0,2)$ and $N^\ast = [\frac{1}{2},\frac{3}{2}]$ and we would be OK, but $N^\ast$ is then not open. It is, as I show, part of the interior, not equal to it. – Henno Brandsma Feb 28 '16 at 11:53
  • Excellent! That's exactly what I was missing: applying the second axiom twice. Thanks! – GBRGR Feb 28 '16 at 15:55
  • @GGR I don't apply it twice. Just at another point. But it was a fun argument. Glad to help! – Henno Brandsma Feb 28 '16 at 16:00
  • Yes, I hadn't understood correctly. I had missed that you never require $N^\circ$ to be a neighbourhood of $x$ before showing that it is open. I thought your definition of $N^\circ$ was analogous to obtaining some neighbourhood of $x$ through the second axiom, but now I see that it is not. Thanks again! – GBRGR Mar 31 '16 at 16:18