6
$\begingroup$

What, from a categorical rather than topological point of view, are the interesting properties of the category of compact Hausdorff spaces?

In particular, is it the case that every monomorphism is actually the inclusion of a subspace, ie with the subspace topology?

Also, what is the relationship of this category to that of sets?

[This question previously read: Why in the category of compact Hausdorff spaces the subspaces are equivalent to subobjects? That is, if $Z\to X$ is an injective map of compact Hausdorff spaces, why does $Z$ automatically have the subspace topology from $X$?]

$\endgroup$
5
  • 5
    $\begingroup$ This seems like a basic question in category theory or general topology. Can you try to explain a bit the context, so that we can better judge if it is of research level and maybe give useful answers (as opposed to "just unwrap the definition")? $\endgroup$ Jan 27, 2021 at 17:33
  • $\begingroup$ The interesting thing about this fact is that it's not true in the category of all topological spaces. A key fact to use is the fact that every continuous bijection between compact Hausdorff spaces is a homeomorphism. $\endgroup$
    – Tim Campion
    Jan 27, 2021 at 18:56
  • 1
    $\begingroup$ I agree that the context is lacking, but I vote against closing the question given Paul Taylor's answer. $\endgroup$ Jan 27, 2021 at 20:18
  • 1
    $\begingroup$ I have half-written a detailed answer, which I will post here if the question is reopened, or on MSE if the question is moved there and someone points me to it. $\endgroup$ Jan 28, 2021 at 16:23
  • $\begingroup$ I agree with @DmitriPavlov (and implicitly with Neil) and have voted to reopen; the current version of the question seems to me to be a great improvement on the original one. $\endgroup$
    – Yemon Choi
    Jan 28, 2021 at 21:42

3 Answers 3

11
$\begingroup$

I don't see what the question is here, but instead of getting into an argument about closing the question, I'll tell you something about the category of compact Hausdorff spaces.

It's pretopos, which roughly means that it has the finitary properties of the category of sets.

For example, all monos are regular, so there is an unambiguous notion of subobject, namely closed subspace.

However, whilst the category of compact Hausdorff spaces also has infinitary limits and colimits, the latter are not stable under pullback as they are for sets.

$\endgroup$
12
  • 3
    $\begingroup$ Let me add that it is also monadic over sets, and in particular monomorphisms can be tested by looking at the underlying sets - this explains your characterizations of monos as inclusions of closed subspaces $\endgroup$ Jan 27, 2021 at 21:37
  • 3
    $\begingroup$ @MaximeRamzi Yes, that's Manes' Theorem. It uses ultrafilters, as I recall. I wonder whether there is a constructive or point-free way of seeing this. $\endgroup$ Jan 27, 2021 at 21:43
  • 1
    $\begingroup$ ¨That's a great question - I confess I don't know the answer, however the following paragraph from the nLab seems to indicate that there is : ncatlab.org/nlab/show/Hausdorff+space#monadicity (see the second paragraph). However it doesn't provide a proof of that statement $\endgroup$ Jan 27, 2021 at 21:51
  • 1
    $\begingroup$ Although Manes' theorem has an explicit proof using ultrafilters, it can also be proven via the monadicity theorem without explicitly invoking ultrafilters. Of course, to apply the monadicity theorem, one must show that the free compact hausdorff space on a set exists, and this is of course the space of ultrafilters, but this can also be deduced from some more general construction of the Stone-Cech compactification, or from the adjoint functor theorem. Of course, none of this is constructive or point-free. $\endgroup$
    – Tim Campion
    Jan 27, 2021 at 22:37
  • 2
    $\begingroup$ Actually -- the open/closed sets of the space of ultrafilters are in some sense "more concrete" than the ultrafilters themselves. In fact, I'm pretty sure the frame of open sets in $\beta(X)$ is (isomorphic to) the set of ideals in the powerset lattice $P(X)$, ordered by inclusion: an ideal $I$ corresponds to the set of ultrafilters $F$ such that there exists $A \in I$ with $A \in F$. Dually, a filter corresponds to the closed set of ultrafilters refining it. So it would make a lot of sense for the "locale of ultrafilters" to behave as well as or better than the space of ultrafilters. $\endgroup$
    – Tim Campion
    Jan 27, 2021 at 23:01
6
$\begingroup$

Let $\mathcal{C}$ be a category, and let $X$ be an object of $\mathcal{C}$. A subobject of $X$ is by definition an equivalence class of pairs $(Y,i)$, where $Y$ is another object and $i\colon Y\to X$ is a monomorphism. Here $(Y_0,i_0)$ and $(Y_1,i_1)$ are considered equivalent if there is an isomorphism $f\colon Y_0\to Y_1$ with $i_1f=i_0$.

Now take $\mathcal{C}$ to be the category of compact Hausdorff spaces. The claim is that the subobjects of $X$ biject with the closed subsets.

The first point is to understand the monomorphisms in $\mathcal{C}$. By definition, a monomorphism is a continuous map $i\colon Y\to X$ of compact Hausdorff spaces such that $i_*\colon\mathcal{C}(T,Y)\to\mathcal{C}(T,X)$ is injective for all $T\in\mathcal{C}$. If $i$ itself is injective, then this clearly holds. Conversely, if $i$ is a monomorphism, we can take $T$ to be a one-point space, and we find that $i$ must be injective.

Next, it is a standard lemma that a subset of a compact Hausdorff space is closed iff it is compact in the subspace topology. If $i\colon Y\to X$ is a continuous map between compact Hausdorff spaces, then it certainly sends compact subsets to compact subsets, so it also sends closed subsets to closed subsets. Using this, we see that if $i$ is injective, then $i(Y)$ is closed, and if we give $i(Y)$ the subspace topology, then the map $i\colon Y\to i(Y)$ is a homeomorphism. Thus, in the class of subobjects, we have $[Y,i]=[i(Y),\text{inc}]$. This makes it clear that the subobjects of $X$ biject with the closed subsets.

As the question has been edited to ask about general categorical properties of $\mathcal{C}$, I will mention some more:

  1. Manes' Theorem: $\mathcal{C}$ is equivalent to the category of algebras for the Stone-Cech monad. Many properties stated below follow immediately from this, but they can also be proved more directly.
  2. $\mathcal{C}$ has all set-indexed limits and colimits.
  3. The monomorphisms are precisely the injective morphisms, and they are all equalisers.
  4. The epimorphisms are precisely the surjective morphisms, and they are all coequalisers.
  5. If $E\subseteq X\times X$ is an equivalence relation that is also closed as a subset of $X\times X$, then $X/E$ is a compact Hausdorff space, and is the coequaliser in $\mathcal{C}$ of the two evident maps $E\to X$.
  6. The injective objects are precisely the retracts of powers of the unit interval.
  7. An object $X$ is projective iff it is extremally disconnected, i.e. the closure of every open set is open.
  8. The dual category $\mathcal{C}^{\text{op}}$ is equivalent to a certain full subcategory of rings. The relevant rings are ordered, but we do not need to consider that as an extra ingredient, because it is determined by the ring structure: we have $a\leq b$ iff $b-a$ is a square. The relevant rings are also topologised, but we do not need to consider that as an extra ingredient, because it is determined by the order. Because of this, each of the relevant rings has a unique $\mathbb{R}$-algebra structure.

Many of these facts can be found in Johnstone's book "Stone Spaces".

$\endgroup$
1
  • 1
    $\begingroup$ I think putting Manes' theorem at the top of the list is spot-on. It follows immediately that the category is complete and cocomplete and exact (and balanced, from which working out what subobjects and such are is straightforward). Since the ultrafilter monad preserves finite coproducts (indeed, the ultrafilter functor / monad is co-uiversal with this property), it follows that you have a pretopos. I'd add that the dual category is not just a subcategory of rings, but precisely the category of commutative, unital C^* algebras, and is in particular locally $\aleph_1$-presentable. $\endgroup$
    – Tim Campion
    Feb 3, 2021 at 22:11
1
$\begingroup$

I'm not completely sure how germane to the present discussion this is, but I have always found the title of this paper by Herrlich and Strecker a bit intriguing: $\mathrm{Algebra}\cap\mathrm{Topology}=\mathrm{Compactness}$.

$\endgroup$

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.