1
$\begingroup$

I have some questions of same flavour about two following constructions in Daniel Huybrechts's notes on K3 surfaces.

Construction 1: Kummer surface (Example 1.3 (iii), page 8) Let $k$ be a field of $char(k)\neq 2$ and let $A$ be an abelian surface over $k$. The natural involution $\iota: A \to A, x \to −x$, has the $16$ two-torsion points as fixed points. Let $\widetilde{A}$ be the blowup in these $16$ fixed points. It's easy to see that the involution $\iota$ lifts to $\widetilde{A}$. The quotient $\pi: \widetilde{A}\to X:=\widetilde{A}/ \iota$ is a ramified double covering of degree two.

Claim I: $\pi_* \mathcal{O}_{\widetilde{A}}= \mathcal{O}_X \oplus \mathcal{L}^{*}$
where $\mathcal{L}$ is square root of line bundle $\mathcal{O}(\sum_{i=1}^{16} E_i)$ - ie $\mathcal{L}^{\otimes 2}= \mathcal{O}(\sum_{i=1}^{16} E_i)$ - where $E_i$ are the exceptional divisors of the fixed points.

Construction 2: (Example 1.3 (iv), page 9) Consider a double covering $p: Y \to \mathbb{P}^2$ brached along a curve $C \subset \mathbb{P}^2$ of degree six.

Claim II: $ p_* \mathcal{O}_{Y}= \mathcal{O}_{\mathbb{P}^2} \oplus \mathcal{O}(-3)$.

Question: How to derive the two claimed formulas for the pushforwards of the structure sheaves? The second construction looks more "generic", in the sense one may pose more general question:

Assume more generally $f:X \to Y$ is branched covering of surfaces of degree $d$ brached in a curve $C$ of degree $s$. Is there a formula for pushforward sheaf $f_*\mathcal{O}_X$ on terms of $\mathcal{O}_Y$ and sheaves related somehow to branch locus $C$ known?

Motivation: There is a well known formula for dualizing sheaves for branched coverings. But I nowhere found a formula for pushforward of structure sheaf even if it looks more "natural" that such should exist.

What, if we generalize by dropping the surface assumption and replace $C$ by a smooth divisor.

$\endgroup$
2
  • 3
    $\begingroup$ This type of computation is specific to cyclic covers. This is explained in detail in Lazarsfeld's Positivity in algebraic geometry I, Remark 4.1.7. Since both your examples have degree 2, they are necessarily cyclic (in particular Galois, albeit ramified). $\endgroup$ Commented Dec 31, 2023 at 3:27
  • 1
    $\begingroup$ You might have a look at this paper to see that there cannot be a "formula" for $f_*\mathcal{O}_X$. $\endgroup$
    – abx
    Commented Dec 31, 2023 at 14:14

0

You must log in to answer this question.