Let $G$ be an algebraic group over a field $k$, and let $BG$ be its classifying space. Let $X$ be a stack over $k$ (e.g. $X$ could be the Picard stack $Pic(S)$, for some scheme $S$). I'm trying to understand what it means to have a morphism $$BG \to X$$ of stacks. Since $BG=[pt/G]$, my guess is that a map $BG \to X$ is just an object $x \in X(k)$ together with a homomorphism of groups $G(k) \to Aut(x)$. Is this correct? This doesn't seem to take into account any of the geometric structure of $G$, so I wouldn't be surprised if this is isn't right.

What is the right way to think of maps out of a classifying stack?


You're almost there! The problem is that, as you've surmised, the group $\mathrm{Aut}(x)$ does not capture enough of the geometric structure of $G$. But that's easily solved:

For every $x\in X$ we can define a sheaf of groups $\underline{\mathrm{Aut}}(x)$ such that for every $k$-scheme $S$ $$\underline{\mathrm{Aut}}(x)(S):=\mathrm{Aut}_{X(S)}(x_S)\,.$$ where $x_S$ is the image of $x$ under the functor $X(k)→X(S)$ induced by the map of schemes $S\to \mathrm{Spec}\,k$. That this is a sheaf follows from the definition of stack ("descent for morphisms"), the details might depend on your preferred way of phrasing the definition.

Then, for every sheaf of groups $G$ we can define a stack $BG$, such that for every $k$-scheme $S$ $BG(S)$ is the groupoid of $G_S$-torsors over $S$ (this recovers the notion when $G$ is a group scheme). Note that $BG(k)$ has a distinguished object $\ast$ (corresponding to the trivial $G$-torsor over $k$) such that there's a canonical isomorphism of sheaves of groups $$\underline{\mathrm{Aut}}(\ast)\cong G\,.$$

Then the groupoid of maps $BG\to X$ is the groupoid of pairs $(x\in X(k),\varphi:G\to \underline{\mathrm{Aut}}(x))$, where $x$ is an object of $X(k)$ and $\varphi$ is a map of sheaves of groups. The morphisms in the groupoid are the morphisms in $X(k)$ that conjugate the maps from $G$.

The map in one direction is easy: it sends $F:BG\to X$ to the pair $(F(\ast),G\cong \underline{\mathrm{Aut}}(\ast)\to \underline{\mathrm{Aut}}(F(\ast)))$.

The map in the other direction is slightly trickier: suppose I have a pair $(x,\varphi)$. Then for every $S$ I have to give a functor $BG(S)\to X(S)$. Essentially this sends a $G_S$-torsor $T$ to $T\times_{G_S} x_S$, where $G_S$ acts on $x_S$ via $\varphi$. Making this precise is a bit annoying and hence I shall leave it as an exercise.

Really, the easy way of conceptualizing this (and arguably proving it) is to notice that $BG$ is the stackification of the prestack $BG^{pre}$ sending $S$ to the groupoid with one objects and $G(S)$ automorphisms. Then we can compute the groupoid of maps $BG\to X$ as the groupoid of maps $BG^{pre}\to X$, and the latter is easily seen to coincide with my description above.

For the homotopy theorists that might be listening: yes the above could be simply summarized by saying that the functor from groups to pointed groupoids sending $G$ to $BG$ is the left adjoint to the functor sending $X$ to $\Omega X$.

All I described here works in an arbitrary topos: I did not use that the sheaf of groups $G$ came from a group scheme.

| cite | improve this answer | |

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.