Vector bundles on some non-projective surfaces - MathOverflow most recent 30 from http://mathoverflow.net2013-05-20T04:44:34Zhttp://mathoverflow.net/feeds/question/43700http://www.creativecommons.org/licenses/by-nc/2.5/rdfhttp://mathoverflow.net/questions/43700/vector-bundles-on-some-non-projective-surfacesVector bundles on some non-projective surfacesAlexander Braverman2010-10-26T17:27:19Z2010-10-26T17:27:19Z
<p>Let $X$ be a smooth projective curve over a field $k$ and let $L$ be a line bundle on $X$.
I will denote by $S$ the total space of $L$ -- this is a smooth surface over $k$ containing
$X$ (as the zero section).
Let $S^0$ be the complement to the zero section.
Let also $G$ be some reductive group (you can assume that $G$ is $GL(n)$ or $SL(n)$ if you
wish). I need some finiteness results about moduli spaces of $G$-bundles on $S$ and $S^0$. Specifically, I would
like to prove (or disprove) the following statements:</p>
<p>1) Assume that $deg(L)<0$. Consider all $G$-bundles $F$ on $S$ whose restriction to $X$
is fixed (i.e. it is isomorphic to a fixed $G$-bundle on $X$). Is it true
that when $k$ is finite, the number of isomorphism classes of such bundles is finite?
For general $k$, is it true that this (non-algebraic) stack can be covered by a scheme of finite type over $k$?
I think that it is easy to prove this when we replace $S$ by the formal neighbourhood
of $X$ in $S$ (by deformation theory), but is it true for $S$ itself?</p>
<p>2) Consider now $G$-bundles on $S^0$ and assume that $deg(L)\neq 0$ (note that $S^0$ doesn't change when we replace $L$ by
$L^{-1}$, so you may as well assume that $deg(L)<0$).
In the case when $k$ is finite I would like
to prove the number of isomorphism classes of such bundles is finite.
This is easy if $G=GL(1)$ (i.e. when we talk about line bundles); also it
is not difficult when $X={\mathbb P}^1$ (in this case the number of isomorphism
classes is finite for any $k$ -- not necessarily finite). For general
$k$ there should probably be a statement that the (non-algebraic) stack $Bun_G(S^0)$ can be covered by a scheme of finite type over $k$ (but again, I just need the statement for the finite field).</p>