3
$\begingroup$

Given an very ample line bundle $L$ on a projective variety we embed it into a projective space such that pullback of $\mathcal{O}(1)$ is $L$. Then we can identify the two. Consider the full sub-category of vector bundles of the form $\bigoplus \mathcal{O}(n_i)$ where $n_i\geq 0$ (this is a finite direct sum, so these are finite rank vector bundles), obviously we can lift each object from the variety to the ambient projective space. For any variety $X$ we denote this category by $\mathcal{C}_X$.

Now consider a projective variety $X$ a subvariety $Z$ (assume $Z$ is a hyperplane section). Is it true that we can lift the morphisms between two objects $\mathcal{C}_Z$ to a neighborhood of $Z$ in $X$?(or formal completion of $Z$ along $X$). If so is it possible to give a functorial lift from the category of vector bundles $\mathcal{C}_Z$ on $Z$ to the formal completion $X_Z$? (or some neighborhood of $Z$ along $X$)

So the question is whether there is a functor from $\mathcal{C}_Z$ to $\mathcal{C}_{X_Z}$ denoted by $i$, such that $res\circ i=id$ where $res$ is the restriction functor from $\mathcal{C}_{X_Z}$ to $\mathcal{C}_Z$? If such a functor exists (maybe in certain cases) when is it an equivalence of categories?

Since the morphisms between two objects of $\mathcal{C}_Z$ splits into direct sum of global section of line bundles in the form $\mathcal{O}(n)$ we can deduce that problem is equivalent to asking when the ring $R_Z=\bigoplus_{n\geq 0}\Gamma(Z, \mathcal{O}_Z(n))$ split injects into the ring $R_{X_Z}=\bigoplus_{n\geq 0}\Gamma(X_Z, \mathcal{O}_{X_Z}(n))$? and when they are isomorphic. (This might also depend on the choice of $\mathcal{O}(1)$)

If these are not true in general I'd like to know cases that they can be true.

$\endgroup$
7
  • $\begingroup$ Unfortunately this is not possible in general: The only bundles you can get from restrictions from the ambient projective space have K-theory classes a linear combination of the classes of $\mathcal{O}$ and $L$. $\endgroup$
    – dhy
    Jan 9, 2021 at 23:39
  • $\begingroup$ I think you forgot about the powers of $L$. That is the category I am asking. There is no problem in lifting the bundles themselves, all of them are lift-able by the definition. $\endgroup$
    – user127776
    Jan 10, 2021 at 0:05
  • 1
    $\begingroup$ Indeed, I misread the question and thought you were lifting the category of all bundles. But now I think there is a different issue. You write "the same is true for the morphisms..." but that amounts to choosing a section of $\Gamma(\mathcal{O}(m-n))\rightarrow\Gamma_C(\mathcal{O}(m-n)).$ This map is not surjective in general so I am skeptical of the claim that there is always a lifting of the morphisms. Am I missing something? $\endgroup$
    – dhy
    Jan 10, 2021 at 0:43
  • $\begingroup$ You are right, but isn't it true that at least you can lift on some neighborhood of the curve? $\endgroup$
    – user127776
    Jan 10, 2021 at 1:17
  • $\begingroup$ Do your morphisms have to have constant rank or are arbitrary sheaf maps allowed? $\endgroup$
    – Ben C
    Feb 13, 2021 at 1:54

0

Your Answer

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