Let $\mathcal{B}$ a filter basis on a topological space $X$ and define $\Lambda=\{(a,A):a\in A\in\mathcal{B}\}$. Define the net $x:\Lambda\rightarrow X$ by $x(a,A)=a$.

I need to prove that the filter generated by $\mathcal{B}$ is an ultrafilter $\Leftrightarrow (x_{\lambda})_{\lambda\in{\Lambda}}$ is an universal net.

What I know:

  • A filter $\mathcal{F}$ in $X$ (assuming $\leq$ as the partial order of $\Lambda$) is a subset of $\Lambda$ not empty satisfying:

$\qquad(1)\,\forall\, x,y\in\mathcal{F},\,\exists\, z\in\mathcal{F}$ such that $z\leq x$ and $z\leq y$

$\qquad(2)\,\forall\,x\in\mathcal{F}$ and $\forall\,y\in\Lambda,\quad x\leq y\implies y\in\mathcal{F}$

  • An ultrafilter is a maximal filter, that is, does not exists filter $\mathcal{F'}$ in $X$ with $\mathcal{F}\subset\mathcal{F'}$ (strict inclusion)

  • A universal net is a net such that, for any $A\subset X$, the net is eventually in $A$ or eventually in $X\setminus A$

I write these definitions and simply nothing comes to me. Can someone help?

  • $\begingroup$ A filter on $X$ in this context is a non-empty collection of subsets of $X$ such that $\emptyset \notin \mathcal{F}$, $\mathcal{F}$ is closed under finite intersections ($A, B \in \mathcal{F} \implies A \cap B \in \mathcal{F}$ ) and under supersets (i.e. $A \in \mathcal{F} \land A \subseteq B \implies B \in \mathcal{F}$). $\endgroup$ – Henno Brandsma Jun 13 '19 at 15:55
  • $\begingroup$ I define the direction (not partial order) $\le$ on $\Lambda$ in my answer. $\endgroup$ – Henno Brandsma Jun 13 '19 at 15:57
  • $\begingroup$ A non-empty collection of subsets $\mathcal{B}$ of $X$ is called a filter base when $\emptyset \notin \mathcal{B}$, $\forall B_1,B_2 \in \mathcal{B}: \exists B_3 \in \mathcal{B}: B_3 \subseteq B_1 \cap B_2$. The filter $\mathcal{F}$ generated by $\mathcal{B}$ is $\{A \subseteq X: \exists B \in \mathcal{B}: B \subseteq A \}$ and one easily checks this is a filter in the sense of my other comment. $\endgroup$ – Henno Brandsma Jun 13 '19 at 16:05
  • $\begingroup$ You're confusing the notion of a filter (base) on $X$ with a more general one, of which the definition I have is a specialised version), namely that of a filter base in a poset (or sometimes even a lattice). For topological convergence we use the filters of subsets as I defined in the other comment. $\endgroup$ – Henno Brandsma Jun 13 '19 at 21:27
  • $\begingroup$ You actually don't define a partial order on $\Lambda$..... $\endgroup$ – Henno Brandsma Jun 15 '19 at 23:03

To complete the definition of $\Lambda$, we have to define the direction relation $\le$:

$$\forall B_1,B_2 \mathcal{B}, \forall b_1 \in B, \forall b_2 \in B:\, (b_1,B_1) \le (b_2, B_2) \text{ iff } B_2 \subseteq B_1\tag{1}$$

and it's easy to check that this is a well-defined direction relation (not a partial order as two elements can be $\le$ each other and not equal; no antisymmetry) by the definition of a filter base: if $(b_1,B_1), (b_2,B_2) \in \Lambda$ we know that $B_1,B_2 \in \mathcal{B}$ so that there exists a $B_3 \in \mathcal{B}$ such that $B_3 \subseteq B_1 \cap B_2$ (definition of filter base) and pick $b_3 \in B_3$ (as all members of $\mathcal{B}$ are non-empty) and note that $(b_1,B_1) \le (b_3,B_3)$ and also $(b_2,B_2) \le (b_3,B_3)$ by definition (1). The fact that $\le$ is reflexive and transitive is very easy to check.


If $\mathcal{B}$ is a filterbase and $\mathcal{F}$ is its generated filter and $\Lambda$, $\le$ and $x$ are as defined in the OP and as above, then $$\forall A \subseteq X: x \text{ is eventually in } A \iff A \in \mathcal{F}$$

proof: $\Rightarrow$: As $x$ is eventually in $A$ there is some $(b_0,B_0) \in \Lambda$ such that $$\forall (b,B) \in \Lambda: (b,B) \ge (b_0,B_0) \implies x(b,B) \in A$$

Now, if $p \in B_0$ is arbitary, by definition $(1)$ (where the point part of the pair is irrelevant to the $\le$ relation) we have that $(p,B_0) \ge (b_0,B_0)$ so $p=x(p,B_0) \in A$. As $p$ was arbitary, we have shown $B_0 \subseteq A$ and so $A$ is in the filter $\mathcal{F}$ generated by $\mathcal{B}$, as $B_0 \in \mathcal{B}$ by definition.

$\Leftarrow$: $A \in \mathcal{F}$ so there is some $B_0 \in \mathcal{B}$ with $B_0 \subseteq A$. Now pick any $b_0 \in B_0$ so that $(b_0,B_0) \in \Lambda$ and if $(b,B) \ge (b_0,B_0)$ we know that $B \subseteq B_0$ by $(1)$ and also that $x(b,B) = b \in B \subseteq B_0 \subseteq A$, so $(b_0,B_0)$ shows that $x$ is eventually in $A$. QED.

Now the proof is finished by knowing that $\mathcal{F}$ is an ultrafilter on $X$ iff $$\forall A \subseteq X: A \in \mathcal{F} \lor X\setminus A \in \mathcal{F}$$

as we see here, e.g.

| cite | improve this answer | |

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.