4
$\begingroup$

Here, let $\mathbb{Z}$ be the group scheme whose functor of points is the constant functor which takes a connected affine scheme to the group $\mathbb{Z}$. I'm having a bit of trouble understanding the stack $B\mathbb{Z}$, in particular quasicoherent sheaves on it.

Here's how I want to approach $BG$ for $G$ say an affine algebraic group. One can take a smooth atlas $f: * \rightarrow BG$ and apply Barr-Beck to find that the category of coalgebras over the comonad $f^* f_*$ is equivalent to the category of quasicoherent sheaves on $BG$. One can use smooth base change ($f$ is smooth and thus flat, also separated and finite type, everything is noetherian I think) and see that this comonad is the same as $g_* g^*$ for $g: G \rightarrow *$. Then a coalgebra is a vector space $V$ (let's say everything is over a field of char 0) which is a comodule for $\mathcal{O}_G$ (ring of global functions). If $V^*$ is finite-dimensional then we can take symmetric powers and spectrum and get an algebraic action $G \times V^* \rightarrow V^*$ -- here we need that $G$ is affine. Thus a coherent sheaf on $BG$ is a vector space with a rational action of $G$. (Is the above correct? I understand there's also some argument involving Hopf duals, but I'm not sure there's a good analogue for the group ring for $G$ not finite.)

For $G$ not affine (e.g. $G = \mathbb{Z}$, or maybe $G$ an ind-scheme like $\hat{\mathbb{G}}_a$), do we have a way to understand (coherent sheaves on) these objects?

Edit: Extra question. In the above, I had to assume that $V$ as finite-dimensional, in order for a coaction $k[V] \rightarrow k[V] \otimes k[G]$ to give a rational/algebraic action. If $V$ is infinite-dimensional (i.e. quasicoherent sheaf instead of coherent sheaf), I'm also not really sure how to proceed.

$\endgroup$
  • 1
    $\begingroup$ I would guess that quasicoherent sheaves on $B G$ for any discrete group $G$ is just $G$-modules. $\endgroup$ – Zhen Lin Dec 1 '14 at 8:46
4
$\begingroup$

If $p : G \to S$ is any smooth group scheme over some base $S$, then $BG:=B(G/S)$ is an algebraic stack over $S$ with a presentation $S \to BG$ (corresponding to the trivial $G$-bundle). (Actually we only need $G \to S$ to be flat, then $BG$ will be an algebraic stack in the sense of Goerss-Naumann.) A quasi-coherent sheaf on $BG$ thus consists of a quasi-coherent sheaf $M$ on $S$ together with an isomorphism $p_1^* M \cong p_2^* M$ on $S \times_{BG} S$ which satisfies the cocycle condition. Using the isomorphism $S \times_{BG} S \cong G$, we get an isomorphism $p^* M \cong p^* M$. This is the same as a $G$-action on $M$, which is a natural family of group homomorphisms $G(T) \to \mathrm{Aut}_{\mathcal{O}_T}(M_T)$ for $S$-schemes $T$. So we get $\mathsf{Qcoh}(BG)=\mathsf{Rep}_S(G)$.

If $\Gamma$ is a discrete group, we may look at its corresponding group scheme $G=\coprod_{g \in \Gamma} S$. Its functor of points is the sheaf associated to the constant presheaf with value $\Gamma$. Therefore, an action of $G$ on a quasi-coherent module $M$ on $S$ is just an action of $\Gamma$ in the usual sense, i.e. a group homomorphism $\Gamma \to \mathrm{Aut}_{\mathcal{O}_S}(M)$. For $\Gamma=\mathbb{Z}$ this action is just an automorphism of $M$.

Hence, quasi-coherent sheaves on $B (\coprod_\mathbb{Z} S)$ are just quasi-coherent sheaves on $S$ equipped with an automorphism of $S$.

$\endgroup$
  • $\begingroup$ Some details are missing, I will try to include them later. $\endgroup$ – Martin Brandenburg Dec 1 '14 at 10:55
  • $\begingroup$ Ah, this makes sense. What I was missing is how to think of a $G$-action by functor of points. Thanks! $\endgroup$ – user148177 Dec 1 '14 at 11:10
  • $\begingroup$ Perhaps someone can explain (I forgot how to do this) why the isomorphism $p_1^* M \cong p_2^* M$ corresponds to a $G$-action on $M$? $\endgroup$ – Martin Brandenburg Dec 1 '14 at 18:54
  • $\begingroup$ I'd guess you can do it by functor of points definition for quasicoherent sheaves, i.e. a quasicoherent sheaf on $G$ is a functorial assignment to each affine $T$ and $\eta \in G(T)$ a $T$-module (morally, the pullback to $T$, you can say this more cleanly but I forget the terminology). So a map in $G(T)$ gives you an isomorphism of the pullback of $M$ to $T$, i.e. a $T$-automorphism of $M_T$, which is a $G$-action on $M$ $\endgroup$ – user148177 Dec 1 '14 at 20:44
  • $\begingroup$ Yes, this is a very good explanation! And the cocycle condition will (hopefully) correspond to the property that $G(T) \to \mathrm{Aut}(M_T)$ is a group homomorphism. $\endgroup$ – Martin Brandenburg Dec 1 '14 at 21:28

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.