8
$\begingroup$

Suppose that $\mathscr{C}$ is a (possibly higher) category and $J$ is a Grothendieck topology which is not subcanonical. Denote the composite

$$\mathscr{C} \hookrightarrow \mathbf{Set}^{\mathscr{C}^{op}} \stackrel{a}{\longrightarrow} \mathbf{Sh}\left(\mathscr{C}\right)$$

by $\theta$, where $a$ is the sheafification functor. Let $W$ denote the class of morphisms in $\mathscr{C}$ which become isomorphisms after applying $\theta.$ Is the essential image of $\theta$ equivalent to the category $\mathscr{C}\left[W^{-1}\right]$ obtained from $\mathscr{C}$ by formally inverting the morphisms $W$?

Question: There are certainly examples where this is true, which I provide below. Is this always the case? If not, what further conditions are needed for it to be true?

Here are two examples where this is true:

Example 1: Let $\mathscr{C}=\mathbf{Haus}$ be the category of Hausdorff topological spaces. Let $J$ be the Grothendieck topology whose covering families for a space $X$ consist of subspace inclusions $$\left(V_\alpha \hookrightarrow X\right)_{\alpha \in A}$$ such that for each compact subspace $K$ of $X$ there exists a finite subset $A(K) \subset A$ such that $$\left(V_\alpha\cap K \hookrightarrow K\right)_{\alpha \in A(K)}$$ can be refined by an open cover of $K$. The essential image of $\theta$ in this case is equivalent to the category of compactly generated Hausdorff spaces, and $W$ consists of all those maps $$X \to Y$$ such that for each compact Hausdorff space $C$, the induced map $$Hom(C,Y) \to Hom(C,X)$$ is an isomorphism. Formally inverting such morphisms yields a category equivalent to compactly generated Hausdorff spaces.

Example 2: Let $\mathscr{C}$ be the $\left(2,1\right)$-category of Lie groupoids and smooth functors and natural transormations. Let $i$ denote the full and faithful inclusion $$i:\mathit{Mfd} \hookrightarrow \mathit{LieGpd}.$$ The adjoint pair $i^* \dashv i_*$ exhibits $\mathbf{Fun}\left(\mathit{Mfd}^{op},\mathbf{Gpd}\right)$ as a left-exact localization of $\mathbf{Fun}\left(\mathit{LieGpd}^{op},\mathbf{Gpd}\right).$ Moreover, stacks of groupoids $$\mathbf{St}\left(\mathit{Mfd}\right)$$ is a left-exact localization of $\mathbf{Fun}\left(\mathit{Mfd}^{op},\mathbf{Gpd}\right),$ so by composition is also a left-exact localization of $\mathbf{Fun}\left(\mathit{LieGpd}^{op},\mathbf{Gpd}\right).$ Therefore, there exists a unique Grothendieck topology $J$ on $\mathit{LieGpd}$ such that $$\mathbf{St}\left(\mathit{LieGpd},J\right) \simeq \mathbf{St}\left(\mathit{Mfd}\right).$$ This Grothendieck topology is not subcanonical because the assignment $$M \mapsto Hom\left(M,\mathcal{G}_1\right) \rightrightarrows Hom\left(M,\mathcal{G}_0\right)$$ does not satisfy descent unless the Lie groupoid $\mathcal{G}$ is really a manifold. $\theta$ in this case is the functor assigning a Lie groupoid $\mathcal{G}$ its stack of principal bundles. The essential image is the $\left(2,1\right)$-category of differentiable stacks, and $W$ consists of Morita equivalences. It is well known that the bicategory of fractions of Lie groupoids with respect to Morita equivalences is equivalent to differentiable stacks.

$\endgroup$
1
  • 2
    $\begingroup$ If the underlying category of the site is a groupoid then this always fails. (For example, take a non-trivial group considered as a groupoid; then the Grothendieck topology generated by the empty sieve is not subcanonical.) $\endgroup$
    – Zhen Lin
    Jul 17, 2014 at 14:03

1 Answer 1

4
$\begingroup$

I don't know whether this includes your two examples, but I believe one situation where this is true is if the topology $J$ is supercanonical, i.e. includes the canonical topology. For example, if $C$ is itself a topos, then supercanonical topologies on $C$ are the same as subtoposes of $C$, which can be obtained from $C$ by inverting the local isomorphisms.

$\endgroup$

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.