Let $X$ be a set on which we wish to define a topology. Let $\mathcal{A}$ be a subset of the power set of $X$, i.e. elements of $\mathcal{A}$ are subsets of $X$. This question arose when studying the Zariski topology, so I'm approaching it by closed subsets, not open ones. Consider the intersection of all topologies for which elements of $\mathcal{A}$ are all closed : it is itself a topology. Equip $X$ with that topology and call $\mathcal{T}$ the set of $X$'s closed elements.

From this Wikipedia page I get the impression that it is easy to explicitly describe $\mathcal{T}$ and to do so I introduce the notation $\mathcal{P}_f(\mathcal{A})$ to denote the finite elements of the power set of $\mathcal{A}$. Then, my understanding is that :

$$\mathcal{T} = \lbrace{\bigcap_{\omega \in \Omega} \bigcup_{F \in \omega} F \quad | \quad \Omega \subset \mathcal{P}_f(\mathcal{A}) \quad \rbrace} $$

I tried to prove this, convinced that it would be straightforward but I'm actually really stuck. To make notation simpler, notice that a supposedly closed subset of $X$ is entirely determined by $\Omega$ so I defined :

$$ \Phi : \mathcal{P}(\mathcal{P}_f(\mathcal{A})) \rightarrow \mathcal{P}(X) $$ by $$ \Phi(\Omega)=\bigcap_{\omega \in \Omega} \bigcup_{F \in \omega} F$$

My claim now simply turns into $\quad\mathcal{T}=\text{Im}\Phi$.

It is clear that $\text{Im}\Phi \subset \mathcal{T}$ from the properties of closed subsets.

I need to prove that also $\mathcal{T} \subset \text{Im}\Phi$, for which it suffices to prove that $\text{Im}\Phi$ is a topology (or rather, the set of closed sets of a topology).

  • $X=\Phi(\emptyset)$ and $\emptyset=\Phi(\lbrace \emptyset \rbrace)$ so $X$ and $\emptyset$ are given as closed subsets by $\Phi$
  • The intersection property should be easy to prove
  • Let $I \subset \mathcal{P}(\mathcal{P}_f(\mathcal{A}))$ be finite. We need to prove that $\quad \bigcup_{\Omega \in I} \Phi(\Omega)\in \text{Im}\Phi$ that is exactly : $$ \bigcup_{\Omega \in I}\bigcap_{\omega \in \Omega} \bigcup_{F \in \omega}F $$ can be written in the form $$\bigcap_{\omega \in \Omega'} \bigcup_{F \in \omega} F $$ and I've drawn diagrams and all sorts but I can't swap that left-hand side cup with the cap.
  • $\begingroup$ The Wikipedia page in english: Subbase. Generally, if you are in a wikipedia page in some language, you can find a link for that page in some other languages by choosing the language in the left pane. $\endgroup$ – amrsa Apr 28 '17 at 15:25

We need to prove that $H=\bigcup_{\Omega\in I}\bigcap_{\omega\in\Omega}\bigcup_{F\in\omega}F$ can be written in the form $\bigcap_{\omega\in\Omega'}\bigcup_{F\in\omega}F$ for some $\Omega'\in\mathcal{P}(\mathcal{P}_f(\mathcal{A}))$.

Denote $\mathcal{F}_I$ the family of all functions $\varphi\colon I\to\mathcal{P}_f(\mathcal{A})$ satisfying the property $(\forall \Omega\in I)\,\varphi(\Omega)\in\Omega$. For every $\varphi\in\mathcal{F}_I$ denote $\omega_\varphi=\bigcup_{\Omega\in I}\varphi(\Omega)$. Since $I$ is finite and $\varphi(\Omega)$ is a finite subset of $\mathcal{A}$ for every $\Omega\in I$, we have $\omega_\varphi\in\mathcal{P}_f(\mathcal{A})$. Put $\Omega'=\{\omega_\varphi:\varphi\in\mathcal{F}_I\}$.

Now, we have $x\in X\setminus H$ iff $$x\in\bigcap_{\Omega\in I}\bigcup_{\omega\in\Omega}\bigcap_{F\in\omega}(X\setminus F)$$ iff $$(\forall\Omega\in I)(\exists\omega\in\Omega)\ x\in\bigcap_{F\in\omega}(X\setminus F)$$ iff $$(\exists\varphi\in\mathcal{F}_I)(\forall\Omega\in I)\ x\in\bigcap_{F\in\varphi(\Omega)}(X\setminus F)$$ iff $$x\in\bigcup_{\varphi\in\mathcal{F}_I}\bigcap_{\Omega\in I}\bigcap_{F\in\varphi(\Omega)}(X\setminus F),$$ hence $$H=\bigcap_{\varphi\in\mathcal{F}_I}\bigcup_{\Omega\in I}\bigcup_{F\in\varphi(\Omega)}F=\bigcap_{\omega\in\Omega'}\bigcup_{F\in\omega}F.$$

  • $\begingroup$ You did really well on taking it exactly where I'd left it. Now since I could not have come up with this myself, I'd like to know if you think there's a simpler approach to this. Because it seems to me that anyone who has been introduced to general topology would want to know if a generated topology can be conveniently described (just as with $\sigma$-algebras, for which I was disadvised from trying). I'm surprised at how far this goes into power sets of power sets and all that $\endgroup$ – James Well Apr 28 '17 at 21:44
  • $\begingroup$ I think that steps I did are unavoidable. Rewriting an intersection of unions by a union of intersections using a family of functions is a standard trick, here we also had a good luck that it preserved the finiteness. I believe that you would have a better chance to come up with it, if you started with open sets instead of closed. And yes, good skills in set theory are needed for topology. $\endgroup$ – Peter Elias Apr 28 '17 at 22:50
  • $\begingroup$ Given what I'd done I never doubted that your steps were necessary to finish it off. I was questioning my own start actually, but with what you just said I guess not. $\endgroup$ – James Well Apr 28 '17 at 23:18
  • $\begingroup$ I think that what you did was optimal. $\endgroup$ – Peter Elias Apr 28 '17 at 23:36
  • 1
    $\begingroup$ I chose that direction because the step involving switching the order of quantification explains how each function $\varphi$ encodes the choice of $\omega\in\Omega$ in dependence on given $\Omega\in I$. It would be possible to go the other way but then I would have to reason negatively: ‘if there is $\Omega$ working for each $\omega\in\Omega$ then there is no way to make a choice of $\omega\in\Omega$ in dependence on a given $\Omega$ that is breaking each $\Omega$', or something like that. $\endgroup$ – Peter Elias Apr 30 '17 at 1:45

In this answer I describe a proposition for the open sets.

This can of course easily be dualised, using de Morgan and complements.

If we have a collection of sets $\mathcal{C}$ that we want to be closed, the minimal collection of a topology (described by closed sets) is

$$\hat{\mathcal{C}} = \left(\mathcal{C}^{\cup ,< \infty}\right)^{\cap}$$

and the open sets are their complements. The notation is analogous to the one used in the linked answer. It boils down to the same thing as your $\mathcal{T}$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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