3
$\begingroup$

This is a very basic question but am having trouble showing it so will really appreciate any help with this, thanks in advance.

Let $(X_i,\mathscr{T}_i)$, $i=1,...,n$ for some $n\in\{2,3...\}$ and $X=\prod\limits_{i=1}^nX_i$. Then the box topology $\mathscr{T}^*$ on $X$ is generated by the basis $\mathscr{B}^*:=\{\prod\limits_{i=1}^n U_i|U_i\in\mathscr{T}_i\}$ and the product topology $\mathscr{T}'$ is generated by the subbasis $\mathscr{S}':=\{\pi^{-1}_i(U_i)|U_i\in\mathscr{T}_i\}$. Now let $\mathscr{B}'$ be the basis of $\mathscr{T}'$, so the set of of finite intersections of $\mathscr{S}'$. So it seems to me that $\mathscr{B}^*\subset\mathscr{B}'$, (as $V_i\times ...\times V_n=\bigcap\limits_{i=1}^n\pi_i^{-1}(V_i)$) so $\mathscr{T}^*\subset\mathscr{T}'$ but I cannot see why the reverse inclusion is true. Also why does this inclusion fail for infinite products as I read that the box topology is finer.

So any assistance and corrections of any mistakes will be greatly appreciated. Thanks.

  • 0
    The finite products of open sets form by definition a generating set for both topologies. They are product open because they are finite intersection of subbase elements. And all such finite intersections are esentially of this form. That's all really.2017-01-18
  • 0
    Thank you for comment, but I am still not sure, so can you forgive my slowness and elaborate a bit more. So if $V_{i,k}\in\mathscr{T}_i$, then $\pi_i^{-1}(\cap_k V_{i,k})=\cap_k\pi^{-1}_i(V_{i,k})$, and as topologies are closed under finite intersections, wlog we can take a generic element of $\mathscr{T}'$ to be $\cap\limits_{i=1}^n\pi^{-1}_i(U_i)=\prod\limits_{i=1}^n U_i$, $U_i\in\mathscr{T}_i$. So is this correct and shows that $\mathscr{B}'\subset\mathscr{B}^*$? So does this holds for both the finite and infinte products? Is there a quicker and neater way to show this?2017-01-19
  • 0
    Sorry I was unable to edit the comment so the tex that didn't texify should read $\cap_{i=1}^n\pi^{-1}_i(U_i)=\prod\limits_{i=1}^n U_i$2017-01-19
  • 0
    Also can you explain why the box topology is finer for an infinite product? Is it because that if $|\mathcal{I}|\geq\aleph_0$, then as $\prod\limits_{i\in\mathcal{I}}U_i=\cap_{i\in\mathcal{I}}\pi^{-1}_i(U_i)$ so it is not the case that $\mathscr{B}*\subset\mathscr{B}’$ as a intersection of an infinite family is not necessarily in $\mathscr{B}’$? Also can you give me an example or a reference/link to one where the box topology is finer than the product one? Thanks again for your input and time.2017-01-19

1 Answers 1

1

Let $S$ be a finite intersection from the subbase $\mathcal{S}'$, so there is some $N$ and $I = i_1, \ldots, i_N \in \{1, \ldots, n\}$ and non-empty open sets $U_{i_1} \subseteq X_{i_1}, \ldots, U_{i_N} \subseteq X_{i_1}$ such that $S = \cap_{j=1}^N \pi_{i_j}^{-1}[U_{i_j}]$. We can collect sets from the same space together and some indices might not appear, so set $J_i = \{j \in \{1,\ldots, M\}: i_j = i \}$ for $i \in \{1,\ldots,n\}$ and $K = \{i \in \{1,\ldots, n\}: J_i =\emptyset\}$, all indices of spaces that are not used. Then

$$S =\cap_{i \notin K} (\pi_{i}^{-1}[\cap_{j \in J_i} U_{i_j}]) = \prod_{i=1}^n U_i \text{, where } U_i = X_i \text{ for } i \in K \text{ and } U_i = \cap_{j \in J_i} U_{i_j} \text{ for } i \notin K$$

so $S \in \mathcal{B}^\ast$. And indeed, as you said, $S = \prod_i U_i \in \mathcal{B}^\ast$ then $S = \cap_{i=1}^n \pi_i^{-1}[U_i]$ which is in the base generated by $\mathcal{S}'$. So the standard base for the box topology, is exactly the same as the base generated by the subbase for the product topology, so they generate the same topology.

EDIT: made more slick: Let $\mathcal{B}'$ be the base generated by $\mathcal{S}'$. Note that the box base $\mathcal{B}^\ast$ is closed under finite intersections as

$$\prod_i U_i \cap \prod_i V_i = \prod_i (U_i \cap V_i)$$

and also note that $(\pi_k)^{-1}[U] = \prod_i U_i \in \mathcal{B}^\ast$, where $U_i = X_i$ for $i \neq k$ and $U_k = U$, all of which are open if $U$ is.

The last remarks hows that $\mathcal{S}' \subseteq \mathcal{B}^\ast$ and as the latter is closed under finite intersections, $\mathcal{B}' \subseteq \mathcal{B}^\ast$. And the old

$$\prod_i U_i = \cap_i \pi_i^{-1}[U_i] \in \mathcal{B}'$$

shows that $\mathcal{B}^\ast \subseteq \mathcal{B}'$, hence equality of the bases is ensured, and the equality of the topologies as well.

For infinite products, the product topology is strictly coarser (for non-trivial spaces): because any proper box, $U = \prod_{i \in I} U_i$, where all $\emptyset \neq U_i \nsubseteq X_i$ are open, is not open in the product topology: if it were, pick $p_i \in U_i$, so that $p =(p_i) \in U$, and suppose there are finitely many $i_1, \ldots, i_N$ and open sets $O_{i_1},\ldots, O_{i_N} $ such that $p \in S = \cap_{i=1}^N (\pi_{i_j})^{-1}[O_{i_j}] \subseteq U$. But $I$ being infinite, we have $j \in I$ and $j \notin \{i_1,\ldots,i_N\}$, and pick $q_j \in X_j \setminus U_j$. Then $p'$ defined by $p'_i = p_i , i \in I\setminus \{j\}, p'_j = q_j$ obeys $p' \notin U$, but $p' \in S$ which cannot be. The intuition is easy: $S$ only sets conditions on finitely many coordinates (as a finite intersection of single "conditions"), while the box needs control over all infinitely many, so the esentially finite boxes of the product topology always stick out of the infinite boxes at teh missing coordinates. The above makes this more formal.

  • 0
    Thank you, this cleared it all up.2017-01-19
  • 0
    @AB see edit for slickification2017-01-20