1
$\begingroup$

Let $\pi: X \to \mathbb{P}^n$ be a conic bundle over an (algebraically closed) field $k$. Let $g \in Aut(X)$ so that $g$ preserves the fibres of $\pi$. Clearly $g$ lives inside $PGL_3(k(\mathbb{P}^n))$ since it it determined by the action on the generic fibre. I think that in fact, $g$ must live inside $PGL_3(k)$. Is this true? If not, is it true if we further assume that $g$ has finite order?

My attempt is to look at the morphism $\mathbb{P}^n \to \mathbb{P}^8$ assigning to each point $x \in \mathbb{P}^n$ of the base, the corresponding matrix $g_x$ acting on the conic $\pi^{-1}(x)$. Then if the image has dimension greater than $0$ (I.E. $g$ is not from $PGL_3(k)$) it must intersect the divisor $D \subset \mathbb{P}^8$ cut out by the determinant equation, so that $g$ has vanishing determinant somewhere. Is this sufficient? I was worried about the existence of a family of projective matrices over $\mathbb{P}^n$ which are invertible everywhere.

Improve this question
$\endgroup$
4
  • 3
    $\begingroup$ The fiber of a conic bundle at the generic point $\eta $ is not in general isomorphic to $\mathbb{P}^1_{k(\eta )}$, hence its automorphism group is not $\operatorname{PGL}(2,k(\eta )) $. $\endgroup$
    – abx
    Jun 28 at 10:01
  • 2
    $\begingroup$ For a proper, geometrically connected, smooth curve of arithmetic genus $0$, the least degree of a very ample invertible sheaf is at most $2$, and the dual of the dualizing sheaf is a very ample invertible sheaf of degree $2$ whose associated closed immersion is in $\mathbb{P}^2$. Thus, the automorphism group is a closed subgroup of $\text{Aut}(\mathbb{P}^2)$, but this is $\textbf{PGL}_3$, not $\textbf{PGL}_2$. Perhaps that is your mistake. $\endgroup$
    – Jason Starr
    Jun 28 at 10:53
  • $\begingroup$ I'm sorry, I meant to use $PGL_3$ instead. Also, if the discriminant curve is smooth and sufficiently high genus, it looks like this group is either trivial or $C_2$. Is that correct? $\endgroup$
    – TCiur
    Jun 29 at 5:55
  • $\begingroup$ Even if the discriminant is smooth of high degree, there could still be nontrivial automorphisms: think about blowing up the restriction over a high degree hypersurface of a (constant) cross-section of a product conic bundle. If you add the hypothesis that the family is minimal, that will probably reduce the automorphism group to a finite group (by forcing the global sections of the tangent bundle to vanish). $\endgroup$
    – Jason Starr
    Jun 30 at 10:48

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.