The Stacks project

Lemma 8.11.8. Let $p : \mathcal{S} \to \mathcal{C}$ be a gerbe over a site $\mathcal{C}$. Assume that for all $U \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{C})$ and $x \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{S}_ U)$ the sheaf of groups $\mathit{Aut}(x) = \mathit{Isom}(x, x)$ on $\mathcal{C}/U$ is abelian. Then there exist

  1. a sheaf $\mathcal{G}$ of abelian groups on $\mathcal{C}$,

  2. for every $U \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{C})$ and every $x \in \mathop{\mathrm{Ob}}\nolimits (\mathcal{S}_ U)$ an isomorphism $\mathcal{G}|_ U \to \mathit{Aut}(x)$

such that for every $U$ and every morphism $\varphi : x \to y$ in $\mathcal{S}_ U$ the diagram

\[ \xymatrix{ \mathcal{G}|_ U \ar[d] \ar@{=}[rr] & & \mathcal{G}|_ U \ar[d] \\ \mathit{Aut}(x) \ar[rr]^{\alpha \mapsto \varphi \circ \alpha \circ \varphi ^{-1}} & & \mathit{Aut}(y) } \]

is commutative.

Proof. Let $x, y$ be two objects of $\mathcal{S}$ with $U = p(x) = p(y)$.

If there is a morphism $\varphi : x \to y$ over $U$, then it is an isomorphism and then we indeed get an isomorphism $\mathit{Aut}(x) \to \mathit{Aut}(y)$ sending $\alpha $ to $\varphi \circ \alpha \circ \varphi ^{-1}$. Moreover, since we are assuming $\mathit{Aut}(x)$ is commutative, this isomorphism is independent of the choice of $\varphi $ by a simple computation: namely, if $\psi $ is a second such map, then

\[ \varphi \circ \alpha \circ \varphi ^{-1} = \psi \circ \psi ^{-1} \circ \varphi \circ \alpha \circ \varphi ^{-1} = \psi \circ \alpha \circ \psi ^{-1} \circ \varphi \circ \varphi ^{-1} = \psi \circ \alpha \circ \psi ^{-1} \]

The upshot is a canonical isomorphism of sheaves $\mathit{Aut}(x) \to \mathit{Aut}(y)$. Furthermore, if there is a third object $z$ and a morphism $y \to z$ (and hence also a morphism $x \to z$), then the canonical isomorphisms $\mathit{Aut}(x) \to \mathit{Aut}(y)$, $\mathit{Aut}(y) \to \mathit{Aut}(z)$, and $\mathit{Aut}(x) \to \mathit{Aut}(z)$ are compatible in the sense that

\[ \xymatrix{ \mathit{Aut}(x) \ar[rd] \ar[rr] & & \mathit{Aut}(z) \\ & \mathit{Aut}(y) \ar[ru] } \]

commutes.

If there is no morphism from $x$ to $y$ over $U$, then we can choose a covering $\{ U_ i \to U\} $ such that there exist morphisms $x|_{U_ i} \to y|_{U_ i}$. This gives canonical isomorphisms

\[ \mathit{Aut}(x)|_{U_ i} \longrightarrow \mathit{Aut}(y)|_{U_ i} \]

which agree over $U_ i \times _ U U_ j$ (by canonicity). By glueing of sheaves (Sites, Lemma 7.26.1) we get a unique isomorphism $\mathit{Aut}(x) \to \mathit{Aut}(y)$ whose restriction to any $U_ i$ is the canonical isomorphism of the previous paragraph. Similarly to the above these canonical isomorphisms satisfy a compatibility if we have a third object over $U$.

What if the fibre category of $\mathcal{S}$ over $U$ is empty? Well, in this case we can find a covering $\{ U_ i \to U\} $ and objects $x_ i$ of $\mathcal{S}$ over $U_ i$. Then we set $\mathcal{G}_ i = \mathit{Aut}(x_ i)$. By the above we obtain canonical isomorphisms

\[ \varphi _{ij} : \mathcal{G}_ i|_{U_ i \times _ U U_ j} \longrightarrow \mathcal{G}_ j|_{U_ i \times _ U U_ j} \]

whose restrictions to $U_ i \times _ U U_ j \times _ U U_ k$ satisfy the cocycle condition explained in Sites, Section 7.26. By Sites, Lemma 7.26.4 we obtain a sheaf $\mathcal{G}$ over $U$ whose restriction to $U_ i$ gives $\mathcal{G}_ i$ in a manner compatible with the glueing maps $\varphi _{ij}$.

If $\mathcal{C}$ has a final object $U$, then this finishes the proof as we can take $\mathcal{G}$ equal to the sheaf we just constructed. In the general case we need to verify that the sheaves $\mathcal{G}$ constructed over varying $U$ are compatible in a canonical manner. This is omitted. $\square$


Comments (0)

There are also:

  • 2 comment(s) on Section 8.11: Gerbes

Post a comment

Your email address will not be published. Required fields are marked.

In your comment you can use Markdown and LaTeX style mathematics (enclose it like $\pi$). A preview option is available if you wish to see how it works out (just click on the eye in the toolbar).

Unfortunately JavaScript is disabled in your browser, so the comment preview function will not work.

All contributions are licensed under the GNU Free Documentation License.




In order to prevent bots from posting comments, we would like you to prove that you are human. You can do this by filling in the name of the current tag in the following input field. As a reminder, this is tag 0CJY. Beware of the difference between the letter 'O' and the digit '0'.