7

If $G$ is a group scheme over $S$ acting on an $S$-scheme $X$, I'd like to understand the algebra of invariants $(\mathcal{O}_X)^G$. Specifically, I'd like to understand its relation to invariants $(\mathcal{O}_X)^{G(S)}$.

To simplify notation, say everything is affine: $G = \operatorname{Spec}R$, $X = \operatorname{Spec}A$, and $S = \operatorname{Spec}k$, where $k$ is an arbitrary ring (not necessarily a field). If it helps we can assume $G$ is smooth. We work in the category of $k$-schemes.

The action is given by a map $\sigma : G\times X\rightarrow X$. Let $p : G\times X\rightarrow X$ be the projection map. Then there is a natural bijection $A = \operatorname{Hom}(X,\mathbb{A}^1)$, and by definition the subalgebra of invariants $A^G$ is the set of $f\in A$ whose corresponding map $F : X\rightarrow\mathbb{A}^1$ satisfies $$F\circ\sigma = F\circ p$$ Via $\sigma$, the group $G(k)$ acts on $X(k)$, and for any $k$-scheme $T$, $G(k)$ maps to $G(T)$ and hence also acts on $X(T)$, so $G(k)$ acts on $X$. Thus, we may also consider the ring of invariants $A^{G(k)}$. Certainly we have $$A^G\subset A^{G(k)}$$ My main question is: What is the clearest way to express this relationship? I'm looking for a statement of the form $f\in A$ is $G$-invariant if and only if it is fixed by $G(k)$ and some other conditions.

I think one can say that $$A^G = \{f\in A| f\otimes_k 1\in A\otimes_k B \text{ is fixed by $G(B)$ for every $k$-algebra $B$}\}$$ Is this correct? Is it possible to further restrict the class of $B$'s that you have to consider? Are there other ways of thinking about this?

RobPratt
  • 5,169
stupid_question_bot
  • 7,151
  • 1
  • 12
  • 40
  • 1
    A simple and unsatisfactory one: if $k$ is not an algebraic extension of a finite field, and $G$ is smooth, then $G(k)$ is Zariski dense, so $A^{G(k)} = A^G$. – LSpice Oct 27 '20 at 12:51
  • 1
    I agree that the key thing to say is that, if $G(k)$ is Zariski dense in $G$, then $A^{G(k)} = A^G$. But I don't think your finite field criterion is the right one. For example, let $G = \mu_3$, the group of $3$-rd roots of unity, and let $k = \mathbb{R}$. @LSpice – David E Speyer Oct 27 '20 at 15:20
  • @DavidESpeyer is right; I definitely meant smooth connected, and there is a slight possibility I meant reductive. – LSpice Oct 27 '20 at 15:52
  • 1
    Ah, found it. If (1) $k$ is infinite, (2) $G$ is connected and either (3a) $G$ is reductive or (3b) $k$ is perfect, then $G(k)$ is Zariski dense in $G$. https://mathoverflow.net/q/56192/297 – David E Speyer Oct 27 '20 at 16:36
  • 1
    While I'm here, I just wanted to say that this is not at all a stupid question, and neither are any of your others! – David E Speyer Jan 14 '21 at 15:48
  • @RobPratt, I was just deliberating over whether "does" should be "do" in the title. I eventually decided that it's about the ring of invariants, not about the individual invariants, and so left it. But I think your edit does make the title read more smoothly. :-) – LSpice Jan 14 '21 at 17:52
  • If everything is affine and $G$ is smooth, then it suffices to take for $B=R$. – Wilberd van der Kallen Jan 15 '21 at 08:11

1 Answers1

5

Combining LSpice's (1 2) and my (1 2) comments into an answer: If $G(k)$ is Zariski dense in $G$, then $A^{G(k)} = A^G$. It is very common that $G(k)$ is Zariski dense in $G$: This happens whenever (1) $k$ is infinite and (2) $G$ is connected and either (3a) $G$ is reductive or else (3b) $k$ is perfect. See Density question in algebraic group.

LSpice
  • 11,423
David E Speyer
  • 150,821