This is a verbatim repost of this question by Jianing Song. A few months ago I placed a bounty on the question but there were no answers, so I am reposting it here.

Let $X$ be a nontrivial topological space, $I$ be a infinite set, we can endow $X^I$ (the set of all functions $I\to X$) with either the product topology or the box topology. We know that the box topology is strictly finer than product topology, but since a topology can be homeomorphic to a strictly finer topology, can the product topology be homeomorphic to the box topology anyway?

EDIT: Here I interpret the term "nontrivial" by that the topology endowed on $X$ is not the trivial topology. I do not ask that $X$ is Hausdorff, but I think restricting the problem to Hausdorff spaces is also a question worth asking.

  • 2
    $\begingroup$ In case it is useful, this is equivalent to proving that for non trivial spaces $X,Y$ we cannot have that $X^I\cong(Y^J)_{box}$ for some infinite cardinals $I,J$: if so, then $Z:=X^I$ would satisfy $Z^\mathbb{N}\cong(Z^\mathbb{N})_{box}$ $\endgroup$
    – Saúl RM
    Oct 2, 2022 at 18:02
  • 1
    $\begingroup$ Just shooting from the hip. In the Hausdorff case, you can assume that you have a two point discrete space--then consider compactness. Otherwise, look at the two point indiscrete case. $\endgroup$
    – terceira
    Oct 3, 2022 at 5:30
  • 1
    $\begingroup$ I would guess this is what OP meant by "nontrival topological space," but I guess it's not clear. $\endgroup$ Oct 6, 2022 at 16:23
  • 2
    $\begingroup$ @TarasBanakh I believe the antidiscrete topology is also commonly referred to as the trivial topology, so yes we do not include that in our example. I will write that in the post just to be clear. $\endgroup$ Oct 6, 2022 at 23:11
  • 7
    $\begingroup$ In case anyone needs a reminder, or to restrict the search for an example, it is shown in theorem 1.3 of Scott Williams's chapter “Box Product” in the Handbook of Set-Theoretic Topology (Kunen & Vaughan eds., 1984), that the box product of an infinite family of infinite non-discrete Hausdorff completely regular topological spaces is never any of the following: (i) locally compact, (ii) separable, (iii) connected or locally connected, (iv) first-countable or (v) perfect (as in: “every closed set is $G_\delta$”). $\endgroup$
    – Gro-Tsen
    Oct 7, 2022 at 7:46

1 Answer 1


This answer was supposed to contain a counterexample, but there was a mistake in it. In any case I think the following class of topological spaces may be useful to find a counterexample:

Suppose that we have a set $X$ and a collection $\mathcal{A}$ of subsets of $X$ closed under finite unions. Then we can define a topology on $2^X$ by the following basis of clopens: let $A\in\mathcal{A}$ and let $g:A\to\{0,1\}$ be a function. Then we define $C_{A,g}=\{f:X\to\{0,1\};f\text{ is an extension of }g\}$. Then $\{C_{A,g};A\in\mathcal{A},g:A\to\{0,1\}\}$ is a basis for a topology on the set $2^X$.

For example, $2^\mathbb{N}$ is obtained by setting $X=\mathbb{N}$ and $\mathcal{A}=\{$ finite subsets of $\mathbb{N}\}$. Similarly, $(2^\mathbb{N})^\mathbb{N}_\text{box}$ can be obtained by setting $X=\mathbb{N}^2$, and $\mathcal{A}=\{A\subseteq\mathbb{N}^2;A\cap(\mathbb{N}\times\{n\})\text{ is finite for every }n\in\mathbb{N}\}$, and $(((2^\mathbb{N}_\text{box})^\mathbb{N})^\mathbb{N}_\text{box})^\mathbb{N}$ can be expressed by setting $X=\mathbb{N}^4$ and saying that a set $A$ is in $\mathcal{A}$ if $\{a_1\in\mathbb{N};(a_1,a_2,a_3,a_4)\in A\}$ is bounded and for every pair $(x_1,x_2)$, the set $\{a_3\in\mathbb{N};(x_1,x_2,a_3,a_4)\in A\}$ is bounded.

In general, if we let $2^X_\mathcal{A}$ be the space we defined, then for any infinite set $I$:

  • $(2^X_\mathcal{A})^I_\text{box}\cong 2^Y_\mathcal{B}$, where $Y=X\times I$ and $\mathcal{B}=\{B\subseteq X\times I;B\cap (X\times\{i\})\in\mathcal{A}\;\forall i\in I\}$. (Here we identified $X\times\{i\}$ with $X$)

  • $(2^X_\mathcal{A})^I\cong 2^Y_\mathcal{B}$, where $Y=X\times I$ and $\mathcal{B}=\{B\subseteq X\times I;B\cap (X\times\{i\})\in\mathcal{A}\;\forall i, B\cap (X\times\{i\})=\varnothing\text{ for all $i$ except finite}\}$.

Moreover, suppose we have two spaces $2^X_\mathcal{A}$ and $2^Y_\mathcal{B}$ and we have a bijection $\phi:X\to Y$ which induces a bijection $\mathcal{A}\to\mathcal{B}$. Then the function $F:2^X_\mathcal{A}\to2^Y_\mathcal{B};F(f)=f\circ\phi^{-1}$ is a homeomorphism, just by how we have defined the topologies.

  • $\begingroup$ Thanks for the answer. Do you mind elaborating on why the function $F$ at the end works? $\endgroup$ Oct 8, 2022 at 11:35
  • $\begingroup$ I was planning on writing the details of that today and I have noticed a mistake in my reasoning, $F$ seems to take $\mathcal{D}$ to $\mathcal{C}$ but not the other way around, so it does not work :( sorry. I will leave the first part of the answer in case it is useful to anyone to operate with products and box products $\endgroup$
    – Saúl RM
    Oct 8, 2022 at 12:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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