Connection between polarization of K3 surface, as a morphism of Hodge structures, and polarization of K3 surface as an ample line bundle

95 Views Asked by At

I'm currently reading through Daniel Huybrecht's Lectures on K3 Surfaces and have come across (what seems at first) two different interpretations of a polarization. At the end of chapter 2 (§2.4 existence of K3 surfaces), Huybrechts defines a polarized K3 surface to be:

A projective K3 surface together with an ample line bundle L such that L is primitive, i.e. indivisible in $Pic(X)$, with $(L)^2 = 2d$.

Roughly seven pages later, Huybrechts defines a polarization of a Hodge structure (of weight $n$) to be a morphism of Hodge structures $\psi : V \otimes V \to \mathbb{Q}(-n)$ (also of weight $n$) such that $(v, w) \mapsto \psi(v, Cw)$ is a positive definite, symmetric bilinear form.

I understand the significance of these polarizations as it relates to whether a K3 surface is in fact an algebraic K3 surface, able to be embedded in $\mathbb{P}^N$, or a complex K3 surface. At least this is my impression from looking at these notes on abelian varieties and Lefschetz Theorem by Brian Conrad. I imagine the prototypical example for a K3 surfaces $X$ is something like:

  • Let the Hodge structure on $X$ be $V^* = H^*(X ; \mathbb{Z})$, $V^{p, q} = \bigoplus_{p + q = n} H^{p,q}(X; \mathbb{C}) = \bigoplus_{p+q = n} H^q(X, \Omega^p)$.

  • Let the polarization be the standard Hodge-Riemann pairing $Q(\alpha, \beta) = \langle L^{n - k} \alpha, \omega \rangle = \int_X \omega^{n - x} \wedge \alpha \wedge \beta$ where $\omega$ Kahler form. I'm guessing this is effectively the Riemann form $H$ one wants to consider for the Appell-Humbert theorem regarding $\mathcal{L}(H, \alpha) = \{ \textrm{something},\ \textrm{something},\ \textrm{theta}\ \textrm{functions}\}$

$\dots$ (don't understand this stuff)

  • $\mathcal{L}^{\otimes 3}$ is very ample.

$\dots$ (where I'm trying to go)

  • The moduli functor of K3 surfaces is the map $ T \mapsto \{ (f: X \to T, L) \} $ with $L$ primitive ample line bundle $L_X$ (s.t. $(L_X)^2 = 2d$) and we say $(f, L) \sim (f', L')$ iff $\exists$ isomorphism of $T$-schemes $\psi : X \to X'$ and another line bundle $L_0$ on $T$ with $\psi^* L' \cong L \otimes f^* L_0 $.

I'm having trouble making the jump to the language of line bundles here, which is proving to be a difficulty in later chapters. I understand why we want to call two K3 surfaces isomorphic if their Hermitian forms $Q(\alpha, \beta)$ are isomorphic (similar to an isometry in Riemannian geometry pulling back the bilinear form $g$). So defining the Moduli space in terms of isomorphism classes of $(X, Q)$ would make sense, but I just don't see the connection to ample line bundles $L$ here. Sorry for any gaps in my understanding here, I just began reading these notes earlier this quarter and am trying to finish all the exercises in Vakil at the same time.