Modular invariants for lattice polarized \(K3\) surfaces (Q2469312): Difference between revisions

Let \(M=H\oplus E_8^2\) be the unique even unimodular lattice of signature \((1,17)\) where \(H\) is the standard rank \(2\)-hyperbolic lattice and \(E_8\) is the unique even, negative-definite, and unimodular lattice of rank \(8\). Let \(X\) be a \(K3\) surface with an \(M\)-polarization \(i: M\hookrightarrow NS(X)\), which is a primitive lattice embedding such that \(i(M)\) contains a pseudo-ample class. Theorem 1. Let \((X,i)\) be an \(M\)-polarized \(K3\) surface. (a) The surface \(X\) possesses a canonical involution \(\beta\) defining a Shioda-Inose structure. (b) The minimal resolution of \(X/\beta\) is a new \(K3\) surface \(Y\) endowed with a canonical Kummer surface structure. This structure realizes \(Y\) as the Kummer surface \(\text{Km}(E_1\times E_2)\) associated to an abelian surface \(A\) canonically represented as a Cartesian product of two elliptic curves: \(A=E_1\times E_2\). The elliptic curves \(E_1\) and \(E_2\) are unique up to permutation. (c) The construction induces a canonical Hodge isomorphism between the \(M\)-polarized Hodge structure of \(X\) and the natural \(H\)-polarized Hodge structure of the abelian surface \(A=E_1\times E_2\). It is the standard fact from the Hodge theory that the polarized Hodge structure on an \(M\)-polarized \(K3\) surface \((X,i)\) is identical with the polarized Hodge structure on an abelian surface \(E_1\times E_2\). The main point of Theorem 1 is to transform this Hodge-theoretic correspondence into a geometric relationship. This is done in the section 3. Then applications of the geometric transform obtained in Theorem 1 are discussed. For instance, a complete classification of the special class of \(M\)-polarized \(K3\) surfaces is obtained: they are completely classified by the two modular invariants \(\pi,\sigma\in{\mathbb C}\), which are defined only Hodge-theoretically. The special class of \(M\)-polarized \(K3\) surfaces considered here is the Inose surface, \(X(a,b)\). It is the two-parameter family of \(M\)-polarized \(K3\) surfaces obtained by taking minimal resolution of the projective quartic in \({\mathbb P}^3\) associated with the special equation \[ y^2zw-4x^3z+3axzw^2-{1\over{2}}(z^2w^2+w^4)+bzw^3=0,\, a\,b\in{\mathbb C}. \] Theorem 2. Let \(E_1\) and \(E_2\) be two elliptic curves associated to \(X(a,b)\) by Theorem 1. Let \(J(E_1)\) and \(J(E_2)\) be the \(J\)-invariants of \(E_1\) and \(E_2\), respectively. Then \(J(E_1)\) and \(J(E_2)\) are the two solutions of the quadratic equation \[ x^2-(a^3-b^2+1)x+a^3=0. \] The modular invariants of the Inose surface \(X(a,b)\) are given by \[ \pi=a^3,\,\, \sigma=a^3-b^2+1. \] Proof of Theorem 2 is given by very complicated explicit calculations of the periods of the quartic equation. This is due to the fact that the modular invariants are defined only Hodge-theoretically, and accordingly, the classification result is a consequence of the appropriate version of the global Torelli theorem.