Let $\textbf{CartSp}$ be the category of spaces of the from $\mathbb{R}^n$ with smooth maps between them. This is a site with respect to (differentially) good open covers, so consider the Grothendieck topos of sheaves of sets on this site.

Now $\textbf{CartSp}$ is not just a site but also the syntactic category of the Lawvere theory of smooth algebras, namely those $\mathbb{R}$-algebras $A$ for which every smooth map $\mathbb{R}^n \to \mathbb{R}^m$ lifts to a map $A^n \to A^m$ in a compatible way.

That is to say, a smooth algebra is a product preserving copresheaf on $\textbf{CartSp}$ and $\textbf{Sh}(\textbf{CartSp})$ is the classifying topos of smooth algebras.

Now let $X$ be an object of $\textbf{Sh}(\textbf{CartSp})$. Then there is a naturally defined sheaf of smooth algebras$$\mathcal{O}_X: \textbf{CartSp} \to \textbf{Sh}(\textbf{CartSp}) \to \textbf{Sh}(\textbf{CartSp})/X$$which is supposed to be regarded as the structure sheaf.

So the outlook of this is that the subcategory of concrete and locally representable sheaves on $\textbf{CartSp}$ is equivalent to the category of smooth manifolds.

The trouble I am having is figuring out how to think of these as smooth manifolds. It would seem that such a sheaf $X$ really does not have the structure of a smooth manifold since it does not come with a structure sheaf.

Evidently I should instead be thinking of the slice topos $\textbf{Sh}(\textbf{CartSp})/X$ it induces, since this comes with a structure sheaf as I described earlier.

But I can not seem to figure out how to view $\textbf{Sh}(\textbf{CartSp})/X$ as a "space" in its own right. What are its open sets?

The motivation for this question is that I am trying to learn étale machinery from an abstract nonsense perspective but for smooth manifolds instead of schemes.

  • 1
    $\begingroup$ I think you might be trying too many things as the same time. Are you comfortable with the notion of (Grothendieck) topos as a space-like thing? The "open sets" of a topos are the subobjects of the terminal object. This makes a lot more sense if you think of the topos associated to a topological space. $\endgroup$ Feb 1, 2017 at 18:05
  • 1
    $\begingroup$ @DenisNardin I am familiar with space-like Grothendieck toposes. I'm good with open sets being subobjects of the terminal object, but I would like this to be restated in terms of probes since that would (hopefully) be more spatially intuitive $\endgroup$
    – KeD
    Feb 1, 2017 at 20:49

1 Answer 1


You want to think about about the objects $X$ as you do the Yoneda image $h_M = Hom(-,M)$ of a manifold $M$, i.e., as sheaves on the site $\bf{CartSp}$

${\bf Sh}({\bf CartSp})/X$ is the "big little topos." Look at what it is--it is an over category: it's objects are (all the) arrows $U \to X$ with $U$ ranging in the big topos. This is essentially all of the information of how its peers may map into $X$.

A standard way to present manifold is via atlases/charts. We see a cover $\sqcup h_{U_i} \to X$ (or just a chart $h_{U_i} \to X$) are important objects in ${\bf Sh}({\bf CartSp})/X$. (Here if one wants, can take $U_i = \mathbb{R}^n \in Ob({\bf CartSp})$ [equal!].)


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.