The locus of real multiplication and the Schottky locus (Q2450160)

For \(g \geq 1\) we denote by \(\mathcal A_g\) the moduli space of principally polarized abelian varieties of dimension \(g\), by \(\mathcal M_g\) the moduli space of curves of genus \(g\), and by \(t: \mathcal{M}_g \to \mathcal{A}_g\) the Torelli map. The Schottky locus is the image of \(t\). As is well known, \(t(\mathcal{M}_g) = \mathcal{A}_g\) if and only if \(g \leq 3\). A web of conjectures (see [\textit{B. Moonen} and \textit{F. Oort}, Advanced Lectures in Mathematics (ALM) 25, 549--594 (2013; Zbl 1322.14065)]) suggests that for \(g \geq 8\), there are no special (Shimura) subvarieties \(Z\) of \(\mathcal A_g\) such that the generic point of \(Z\) lies in the image of \(t\). In support of these conjectures, \textit{J. de Jong} and \textit{S.-W. Zhang} [Centro di Ricerca Matematica Ennio De Giorgi (CRM) Series (Nuova Serie) 4, 165--172 (2007; Zbl 1138.14027)] showed that for \(g > 4\), a Hilbert modular variety parametrizing abelian varieties with real multiplication is not contained in the closure of \(t(\mathcal{M}_g)\). The paper under review extends this result to the case \(g=4\). The proof is based on the study of the boundary of a Hilbert modular variety in the Deligne-Mumford compactification.