Let $ G $ be a linear algebraic group. Is it true that a subgroup $ H $ of $ G $ is Zariski closed if and only if there exists a representation $ \pi: G \to \mathrm{GL}(V) $ and a vector $ v \in V $ such that the stabilizer $ G_v:=\{g \in G: \pi(g)v=v \} $ is equal to $ H $?

I think one implication is clear since $ G_v $ is certainly a subgroup and the equation $ \pi(g)v=v $ is polynomial in the matrix entries. Thus any stabilizer must be Zariski closed.

I am not sure of the reverse implication. Is it really true that every Zariski closed subgroup of $ G $ arises as the stabilizer of some vector in some representation?

  • $\begingroup$ One could still ask the question: when does a Zariski-closed subgroup $H$ appear as point stabilizer in one representation? This is true if $H$ has no nontrivial multiplicative character. Note that $H$ is not assumed connected and the question does not a priori boil down to this case (e.g., I don't know if when $H$ is finite, $H$ can always appear as a point stabilizer). $\endgroup$
    – YCor
    Nov 28 '21 at 15:19

Chevalley's theorem (see Theorem 4.19 in Milne) is very close to this. It says

Let $G$ be a linear algebraic group and let $H$ be a Zariski closed subgroup. Then there is a representation $V$ of $G$ and a one dimensional subspace $L$ of $V$ such that $H$ is the stabilizer of $\mathbb{P}(L)$ in the action of $G$ on $\mathbb{P}(V)$.

So this is close to what you asked for, but uses projective rather than linear representations.

To see that you can't get exactly what you asked for, work over a field of characteristic zero, take $G = \text{SL}_2$ and let $H$ be the Borel subgroup $B:=\left[ \begin{smallmatrix} \ast & \ast \\ 0 & \ast \end{smallmatrix} \right]$. The classification of $\text{SL}_2$-representations is well known, and we see that $V^B = V^{\text{SL}_2}$ for any $\text{SL}_2$-representation $V$.

  • $\begingroup$ Is there any way to parlay this fact into an example of a homogeneous space $ G/H $ which cannot be a linear group orbit $ G/G_v $ for any representation? $\endgroup$ Nov 27 '21 at 14:46
  • 6
    $\begingroup$ Yes, exactly. $SL_2/B$ is $\mathbb{P}^1$, which has no nonconstant global functions, hence cannot be embedded into any vector space. $\endgroup$ Nov 27 '21 at 14:57

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.