1
$\begingroup$

Let $X$ be a smooth projective variety over $\mathbb{C}$, $G$ a reductive group, and $P \to X$ a $G$-torsor. Let $B \subset G$ be a Borel subgroup. Is the associated $G/B$ fibre bundle $$ Y=G/B \times_{G} P $$ projective?

$\endgroup$

1 Answer 1

5
$\begingroup$

A similar, but shorter answer. Choose a $G$-equivariant projective embedding $G/B \to \mathbb{P}(V)$. It gives a closed embedding $$ G/B \times_G P \hookrightarrow \mathbb{P}(V) \times_G P $$ over $X$ whose target is obviously projective, hence so is the source.

Edit. Let me explain how the projectivity of $\mathbb{P}(V)$ can be deduced.

First, assume that $V$ is a representation of $G$. Then $\mathbb{P}(V)$ is the projectivization over $X$ of the vector bundle $V \times_G P$, and a projectivization of a vector bundle over a projective variety is projective.

Now, assume that $V$ is only a projective representation of $G$. Then $\mathrm{Sym}^d(V)$ is a representation of $G$ for some positive $d$, hence the relative Veronese embedding realizes the target as a closed subset of the projectivization of a vector bundle associated with $\mathrm{Sym}^d(V)$, hence it is projective.

Edit 2. Let me explain why a symmetric power of a projective representation is a linear representation.

An action of $G$ on $\mathbb{P}(V)$ is given by a homomorphism $$ G \to \mathrm{Aut}(\mathbb{P}(V)) = \mathrm{PGL}(V). $$ Consider the central extension $$ 1 \to \mu_N \to \mathrm{SL}(V) \to \mathrm{PGL}(V) \to 1 $$ and its pullback $$ 1 \to \mu_N \to \tilde{G} \to G \to 1, $$ where $N = \dim(V)$. Then the action of $G$ on $\mathbb{P}(V)$ lifts to an action of $\tilde{G}$ on $V$ such that the subgroup $\mu_N$ acts by a character. Clearly, the action of $\mu_N$ on $\mathrm{Sym}^N(V)$ is trivial, hence the action of $\tilde{G}$ on $\mathbb{P}(\mathrm{Sym}^N(V))$ factors through $G$. This shows that $\mathrm{Sym}^N(V)$ is a linear representation of $G$.

$\endgroup$
5
  • 1
    $\begingroup$ I must be missing something. Why is the target ``obviously'' projective? $\endgroup$ Commented May 22 at 19:34
  • 1
    $\begingroup$ @LaurentMoret-Bailly: Let me be a bit more precise. First, assume that $V$ is a representation of $G$. Then the target is the projectivization over $X$ of the vector bundle $V \times_G P$, and a projectivization of a vector bundle over a projective variety is projective. Now, assume that $V$ is only a projective representation of $G$. Then $\mathrm{Sym}^d(V)$ is a representation for some positive $d$, hence the relative Veronese embedding realizes the target as a closed subset of a projectivization of a vector bundle, hence it is projective. $\endgroup$
    – Sasha
    Commented May 22 at 20:12
  • $\begingroup$ @Sasha Thanks for this clarification. Could you maybe say a sentence about why for every projective representation $V$, $Sym^d(V)$ is a representation for some positive $d$? $\endgroup$ Commented May 27 at 0:19
  • $\begingroup$ @onefishtwofish: See Edit 2 in my answer. $\endgroup$
    – Sasha
    Commented May 27 at 13:29
  • $\begingroup$ @Sasha Thank you for taking your time to clarify everything! $\endgroup$ Commented May 27 at 14:22

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .