Definability of restricted theta functions and families of abelian varieties (Q1949160)

In [Ann. Math. (2) 173, No. 3, 1779--1840 (2011; Zbl 1243.14022)], \textit{J. Pila} gave an unconditional proof of the André-Oort conjecture for the Shimura variety \(\mathbb C^n\). One of the central ingredients to this proof is that the Weierstrass \(\wp\) function is, in a suitable sense, definable in the o-minimal structure \(\mathbb R_{\mathrm{an}, \mathrm{exp}}\). The present paper proves the analogue of this ingredient for Siegel modular varieties \(\text{Sp}(2g, \mathbb Z) \backslash \mathbb H_g\), \(g \geq 1\). Note however that to prove the André-Oort conjecture in this generality along the same lines, several other ingredients are still missing. More precisely, fix \(g \geq 1\) and a diagonal matrix \(D = \text{Diag}(d_1, \dots, d_g)\) with \(d_1 \mid d_2 \mid \dots \mid d_g\) non-negative integers. Then all the complex abelian varieties of polarization type \(D\) are of the form \(\mathcal{E}_\tau^D := \mathbb C^g/(\tau \mathbb Z^g + D \mathbb Z^g)\), where \(\tau\) runs over the set \(\mathbb H_g\) of symmetric \((g \times g)\)-matrices with a positive definite imaginary part. Write \(h_\tau^D : \mathbb C^g \to \mathbb P^k (\mathbb C)\) for a suitable holomorphic map inducing an embedding \(\mathcal{E}_\tau^D \to \mathbb P^k (\mathbb C)\) (where \(k\) only depends on \(D\)). Morally, the main result of the paper (Theorem~1.2) is that the map \(\mathbb H_g \times \mathbb C^g \to \mathbb P^k (\mathbb C), (\tau, z) \mapsto h_\tau^D(z)\) is definable in \(\mathbb R_{\mathrm{an}, \mathrm{exp}}\). Formulated like this, this is not possible for two reasons: \(h_\tau^D\) is periodic for each fixed \(\tau\), and, moreover, \(\tau \mapsto \mathcal{E}_\tau^D\) is periodic in the sense that \(\mathcal{E}_\tau^D\) and \(\mathcal{E}_{\tau'}^D\) are isomorphic as polarized abelian varieties iff \(\tau\) and \(\tau'\) lie in the same orbit of the action of \(G_D\) on \(\mathbb H_g\), where \(G_D\) is a discrete subgroup of \(\text{Sp}(2g, \mathbb Q)\). A more precise formulation of Theorem~1.2 is that the restriction of \((\tau, z) \mapsto h_\tau^D(z)\) to a subset \(U\) of \(\mathbb H_g \times \mathbb C^g\) is definable in \(\mathbb R_{\mathrm{an}, \mathrm{exp}}\), where the projection of \(U\) to \(\mathbb H_g\) contains a fundamental domain of the action of \(G_D\) on \(\mathbb H_g\), and for each \(\tau\) in this projection, the corresponding fiber of \(U\) contains a fundamental domain of the map \(\mathbb C^g \to \mathcal{E}_\tau^D\). The paper also contains several related results that might be of independent interest. In particular, one of the main ingredients to the proof of Theorem~1.2 is the definability of certain Riemann theta functions \(\vartheta[{a \atop b}](z, \tau)\).