2
$\begingroup$

Let $j: X \to Y$ be a closed embedding. Let $I_{X/Y}$ be the ideal sheaf of this closed embedding. Then there is a exact sequence

$$ I_{X/Y} \to \mathcal{O}_Y \to j_{*}\mathcal{O}_X \to 0$$

One use this exact sequence for computation of $H^i(Y, I_{X/Y})$.

In general it is easy to compute $H^i(Y, \mathcal{O}_Y)$ and $H^i(Y, j_{*}\mathcal{O}_X)$. But how does one determine the maps between $H^i(Y, \mathcal{O}_Y)$ and $H^i(Y, j_{*}\mathcal{O}_X)$? Are there any references where one can find examples of this kind of computation?

$\endgroup$
  • 3
    $\begingroup$ Are you aware that $H^i(Y,j_*\mathcal{O}_X)$ is nothing but $H^i(X,\mathcal{O}_X)$?? $\endgroup$ – abx Nov 1 '18 at 5:00
  • 1
    $\begingroup$ Yes, I'm aware of that. $\endgroup$ – john Nov 1 '18 at 13:43
  • $\begingroup$ If $Y$ and $X$ are smooth and projective, you can use Hodge theory to identify your morphism to the (conjugate of) the restriction $H^0(Y,\Omega ^i_Y)\rightarrow H^0(X,\Omega ^i_X)$, which may be easier to analyze. $\endgroup$ – abx Nov 1 '18 at 19:49
  • $\begingroup$ The precise description of the maximal rank conjecture for curves sheds light on why this may be hard, even though the context is a little different (see arxiv.org/pdf/1711.04906.pdf for a statement of the conjecture). $\endgroup$ – Anwesh Ray Nov 1 '18 at 19:59
4
$\begingroup$

The computation of the restriction morphism $H^i(Y,O_Y) \to H^i(X,O_X)$ is usually non-trivial (unless you know for instance an explicit locally free resolution of the ideal sheaf of $X$ in $Y$).

On $H^0$ this is quite easy, of course.

For $H^1$ one can act as follows: each class in $H^1(Y,O_Y)$ can be represented as an extension $$ 0 \to O_Y \to E \to O_Y \to 0; $$ restricting it to $X$ one obtains an exact sequence $$ 0 \to O_X \to E\vert_X \to O_X \to 0; $$ the restriction map takes the class of the first to the class of the second.

On higher $H^i$ one can use Yoneda representations too, but it becomes more complicated.

$\endgroup$

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.