Computing modular correspondences for abelian varieties (Q661914)

The notion of a modular correspodence of elliptic curves can be generalized to abelian varieties of dimension \(g\) in the followng way: Let \(k\) be a field of charactristic \(\neq 2\) and for \(n\in\mathbb{N}\) denote \(\overline n= (n,\dots, n)\in\mathbb{N}^g\). An abelian variety with an \(\overline n\)-marking over \(k\) is a triple \((A,L,\Theta_{\overline n})\) with an abelian variety \(A\) of dimension \(g\) over \(k\), \(L\) and \(\Theta_{\overline n}\) an ample symmetric line bundle and a symmetric theta structure of type \(\overline n\). The theta structure induces a decomposition \(K(L)= K_1(L)\times K_2(L)\) of the kernel of the polarization \(L\) into maximal isotropic subgroups. Now suppose \(\ell\) and \(n\) are relatively prime. For an abelian variety \((A,L,\theta_{\overline{\ell n}})\) with \(\overline{\ell n}\)-marking define \(B:= A/K_1(L)[\ell]\) and \(C:= A/K_2(L)[\ell]\). This induces a map \(\Phi_\ell:{\mathcal M}_{\overline{\ell n}}\to{\mathcal M}_{\overline n}\times{\mathcal M}_{\overline n}\) of the corresponding moduli spaces. It is called a modular correspondence. For \(g=1\) this coincides with the usual definition. The main result of the paper is a system of equations for the image of \(\Phi_\ell\) in the projective coordiate system provided by the theta constants as well as an efficient algorithm to compute the images of a point in \({\mathcal M}_{\overline n}\) under the correspondence. This can be used to speed up the initial phase of a point counting algorithm.