Positive sheaves of differentials coming from coarse moduli spaces (Q424835)

Let \(f^\circ : X^\circ \rightarrow Y^\circ\) be a family of canonically polarised complex manifolds over a smooth base, i.e. a smooth morphism between quasi-projective complex manifolds such that the fibres are projective manifolds with ample canonical bundle. The family induces a natural map \(\mu: Y^\circ \rightarrow \mathcal M\) to a moduli space \(\mathcal M\) (of canonically polarised manifolds with fixed Hilbert polynomial) and we define the variation of the family \(f^\circ\) as the dimension of the image of this map. A series of recent papers by \textit{S. Kebekus} and \textit{S. J. Kovács} [Invent. Math. 172, No. 3, 657--682 (2008; Zbl 1140.14031); Duke Math. J. 155, No. 1, 1--33 (2010; Zbl 1208.14027)] which studies the relation between the moduli map \(\mu\) and the birational geometry of \(Y^\circ\) uses as a fundamental tool the ``Viehweg-Zuo sheaf'': given a compactification \(Y^\circ \subset Y\) such that \(D:=Y \setminus Y^\circ\) is a simple normal crossings divisor, \textit{E. Viehweg} and \textit{K. Zuo} [Bauer, Ingrid (ed.) et al., Complex geometry. Berlin: Springer. 279--328 (2002; Zbl 1006.14004)] show that there exists an integer \(m>0\) and an invertible sheaf \(A \subset S^m \Omega_Y(\log D)\) whose Kodaira dimension \(\kappa(A)\) is at least the variation of the family. In the paper under review the authors refine the theorem of Viehweg and Zuo by showing that the invertible sheaf \(A\) comes (at least generically) from the moduli space \(\mathcal M\). More precisely let \(B \subset \Omega_Y(\log D)\) be the subsheaf obtained by saturating the image of the differential \(d \mu: \mu^* \Omega_{\mathcal M} \rightarrow \Omega_{Y^\circ}\). Then the invertible sheaf \(A\) is contained in \(S^m B \subset S^m \Omega_Y(\log D)\). The proof is based on a careful analysis of the construction by Viehweg and Zuo. One consequence of the main theorem is a very short proof of a result \textit{F. Campana} [J. Inst. Math. Jussieu 10, No. 4, 809--934 (2011; Zbl 1236.14039)]: if \(Y^\circ\) is a surface such that the pair \((Y, D)\) is special in the sense of Campana, the variation of the family \(f^\circ : X^\circ \rightarrow Y^\circ\) is zero. Note that in their recent paper [Math. Z. 269, No. 3--4, 847--878 (2011; Zbl 1238.14024)], the authors proved a similar statement if \(Y\) is a threefold.