7
$\begingroup$

The forgetful functor $\mathbf{CompHaus} \rightarrow \mathbf{Top}$ from compact Hausdorff spaces to topological spaces famously has a left-adjoint, the Stone-Cech compactification.

Question. Does the forgetful functor $$\mathbf{Comp} \rightarrow \mathbf{Top}$$ from compact spaces to topological spaces have a left-adjoint?

This should be well-known, but I haven't been able to find any relevant information.

$\endgroup$
  • $\begingroup$ Remark. It seems unlikely, if such a functor exists, that it will preserve Hausdorffness. I'm completely okay with this. $\endgroup$ – goblin Jul 10 '17 at 18:06
  • $\begingroup$ CompHaus to Top indeed has a left adjoint, so why is this not a left adjoint for Comp to Top? What goes wrong? $\endgroup$ – Henno Brandsma Jul 10 '17 at 19:34
  • $\begingroup$ @HennoBrandsma By full-faithfulness of the inclusion functor $i$, the counit $L i X \to X$ would have to be an isomorphism. But if $L$ lands us in compact Hausdorff spaces, then each such $X$ would be compact Hausdorff. $\endgroup$ – Todd Trimble Jul 10 '17 at 19:41
13
$\begingroup$

No, it does not. If it did, then $\mathbf{Comp}$ would be a reflective subcategory of the total category $\mathbf{Top}$, and hence would be total itself. Now, total categories admit all small limits, and there is the rub: $\mathbf{Comp}$ does not have all small limits. In fact, $\mathbf{Comp}$ does not have equalizers (of course it admits small products, by the Tychonoff theorem).

Here is a simple example to show why not. Consider Sierpinski space $\mathbf{2} = \{0, 1\}$ where the point $1$ is open. This is of course compact. Take two continuous maps $f, g: [0, 1] \rightrightarrows \mathbf{2}$ where $f^{-1}(1) = [0, 1/2)$ and $g^{-1}(1) = [0, 1]$. Suppose there is an equalizer $i: E \to [0, 1]$ of $f, g$ in $\mathbf{Comp}$. Since $\hom(1, -): \mathbf{Comp} \to \mathbf{Set}$ would preserve this equalizer, we see the underlying set of $E$ would have to be $[0, 1/2)$. In order for $i: E \to [0, 1]$ to be continuous, the topology on $E = [0, 1/2)$ would have to be equal to or finer than the subspace topology on $[0, 1/2)$. But since the subspace topology on $[0, 1/2)$ is not compact, no finer topology on $[0, 1/2)$ can be either; contradiction.

$\endgroup$
  • $\begingroup$ I think something like this argument shows that the category of compact locales isn't complete either, but I haven't nailed down a proof of this. $\endgroup$ – Todd Trimble Jul 10 '17 at 19:37

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.