MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Suppose I have a certain (contravariant) moduli functor $M:Schemes \to Groupoids$ that is represented by a quotient stack $[X//G]$ where $X$ is a scheme and $G$ a linearly reductive group. Roughly speaking $[X//G]$ is the fine moduli space for $M$; hence it carries a universal family and there's a correspondence between families of objects of $M$ over a scheme $S$ and maps $S\to [X//G]$. Now, I claim that also $X$ carries a universal family, obtained via pull-back from $[X//G]$, and it is constant along the fibers of the quotient map $X \to [X//G]$.

If $S$ is once again a family of objects of $M$, what can one say of maps from $S\to X$? That the family over $S$ induces a map to $X$ that is unique up to $G$-action? That such a map is unique as long as $X\to [X//G]$ has a section?

Basically, by the existence of the universal family over $X$ a map $S\to X$ induces a family over $S$ but not in a unique way. So, in the other direction, given a family over $S$ this induces an orbit under $G$ of maps, but there's no canonical choice of a particular map in this orbit.

share|cite|improve this question
up vote 3 down vote accepted

First a couple corrections. If $M$ is represented by quotient stack, it better be a contravariant functor, and moreover, it should probably take values in groupoids, not set. Anyway, here's what going on:

$X$ does not carry a universal family, but it carries a "locally" universal family $v \in M\left(X\right)$, in the sense that, if $S$ is a scheme, and $x \in M\left(S\right)$ then you can find a cover (say in the etale topology, if that's where your stack is) $\left(f_\alpha:S_\alpha \to S\right)$ such that each $f_\alpha^*\left(x\right) \in M\left(S_\alpha\right)$ arises, up to isomorphism as $g_\alpha^*\left(v\right)$ for some $g_\alpha:S_\alpha \to X.$

I haven't used the group $G$ yet. Lets do that. A morphism $S \to M$ corresponds to a choice of a $G$-torsor $P \to S$ and a $G$-equivariant map $P \to X.$ So the map $S \to M$ need not factor through $X,$ unless the torsor is trivial, but it always factors through etale-locally.

share|cite|improve this answer
Ok let's say that I composed with a forgetful functor $f: Groupoids \to Sets$.... :) Just kidding, it is a good remark, I edited. That's more or less what I felt: that I can lift the map $S$-globally iff the torsor $X \to [X//G]$ is trivial over the image of $S$. – IMeasy Apr 11 '13 at 15:13
But I still don't understand one thing. By pulling back the universal family from $[X//G]$ to $X$, don't I get a "universal" $G$-invariant family on $X$? – IMeasy Apr 11 '13 at 15:18
What do you mean by "universal"? To me, universal means that every family is a pullback of it, which is not true here, but is locally- that's why I said locally universal. If instead, you mean canonical, then sure. – David Carchedi Apr 11 '13 at 15:57
By the way, in some more detail, there is a canonical map $$\left[X//G\right] \to \left[pt//G\right]=BG$$, which is faithful. This is how you get a $G$-torsor out of a map $$S \to \left[X//G\right]$$ (it is classified by the composition into $\left[pt//G\right].$) I should be a bit careful, since I am used to working with topological stacks, but I think everything should work. – David Carchedi Apr 11 '13 at 16:01
Yes, I should have said canonical. Universal has a categorical meaning which is not true here. Thank you! – IMeasy Apr 11 '13 at 16:16

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.