Definitions: One says that a map $\pi\colon\mathcal X\to X$ from an algebraic stack to an algebraic space is a good moduli space if $\pi$ is cohomologically affine and universal for maps to schemes. (A good (but not equivalent) way to think about a stack whose good moduli space is a single point is that for any two points, the closure of their orbits intersect.)

One says that a stack $\mathcal X$ is a quotient stack if it is the quotient of an algebraic space by a subgroup of GL_n (see for instance EHKV, which also gives a criterion for a stack to be a quotient stack in terms of vector bundles).

One says that a stack has the resolution property if every coherent sheaf is a quotient of some vector bundle; Totaro's paper The resolution property for schemes and stacks relates the property of being a quotient stack to the resolution property.

Question: Let $\mathcal X$ be a stack with a good moduli space $\mathcal X \to X$ such that X is a geometric point (i.e., X = Spec k, where k is a separably closed field). Suppose further that the stabilizers are affine linearly reductive groups. Is $\mathcal X$ a quotient stack?

(See this answer for the definition of stabilizer of a point of a stack that isn't a quotient stack.)


  1. The condition on stabilizers excludes things like BE with E an elliptic curve.

  2. The condition that k is separably closed excludes non-trivial gerbes.

  3. I'd be just as happy with an answer to "Does the resolution property hold for $\mathcal X$ ?".

  • 5
    $\begingroup$ This is an interesting question. Here is a specific instance: let $n$ be a positive integer, and let $\mathcal X_n$ be the stack over $\mathbb C$ of nodal proper curves of genus 0 with at most $n$ nodes. One shows that it satisfies the conditions above. It is a quotient stack for $n \leq 1$, but I don't know the answer for any $n \geq 2$. $\endgroup$ – Angelo Feb 28 '11 at 7:03
  • 2
    $\begingroup$ The stack $\mathcal{X}_n$ is not a global quotient stack for $n\geq 2$ (Kresch: Flattening stratification and the stack of partial stabilisations of prestable curves). But, as noted by David Zureick-Brown in arxiv:1208.2882, $\mathcal{X}_n\to \mathrm{Spec}(\mathbb{C})$ is not a good moduli space so this does not answer the question. $\endgroup$ – David Rydh Feb 5 '13 at 19:10

This is answered in general by Theorem 13.1 in our paper The étale local structure of algebraic stacks (arXiv:1912.06162). Let $\mathcal{X}$ be a stack with a good moduli space $X$ such that

  1. $\mathcal{X}$ has affine stabilizers,
  2. $\mathcal{X}$ has separated diagonal, and
  3. $\mathcal{X}$ is of finite presentation over an algebraic space.

Then $\mathcal{X}$ has the resolution property étale-locally (and even Nisnevich-locally) on $X$. In particular, if $X$ is the spectrum of a field or a henselian local ring, then $\mathcal{X}$ has the resolution property.

Conditions 1 and 2 are necessary. Note that having the resolution property locally on $X$ implies that $\mathcal{X}$ has affine diagonal.

In general $\mathcal{X}$ does not have the resolution property Zariski-locally on $X$. There is an example in SGA 3, Exp X, §1.6 (cf. Remark 2.5 in our paper), of a 2-dimensional torus $G$ over the nodal cubic curve $C$ such that $G$ is not locally isotrivial. This means that $G$ cannot be embedded in $\mathrm{GL}_N$ for any $N$, not even Zariski-locally on $C$. It follows that $\mathcal{X}=BG$ does not have the resolution property Zariski-locally on its good moduli space $X=C$.


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.