Families of hypersurfaces of large degree (Q2428722)

\textit{S. Lang} [Bull. Am. Math. Soc., New Ser. 14, 159--205 (1986; Zbl 0602.14019)] formulated the following function-field version of the Mordell conjecture: if \(\pi: \mathcal X \to Y\) is a projective surjective morphism of complex algebraic varieties whose generic fiber is of general type, then if \(\pi\) is not birationally isotrivial, there is a proper subscheme of \(\mathcal X\) that contains the images of all sections of \(\pi\). \textit{Yu. I. Manin} [Izv. Akad. Nauk SSSR, Ser. Mat. 27, 1395--1440 (1963; Zbl 0166.16901)] and \textit{H. Grauert} [Publ. Math., Inst. Hautes Étud. Sci. 25, 363--381 (1965; Zbl 0137.40503)] solved this for families of curves, and this paper deals with a certain higher-dimensional case: Main Theorem. Let \(C\) be a curve of genus at least \(2\) and let \(\lambda\) be a divisor with a large degree. For \(n\geq 3\) and \(d\) sufficiently large, let \(\mathcal X\) be the family of hypersurfaces of \(\mathbb P^{n}\) given by a section of an ample line bundle \(L = \lambda \boxtimes \mathcal O_{\mathbb P^{n}}(d)\) on \(C\times \mathbb P^{n}\). Then for a general such \(\mathcal X/C\), Lang's conjecture holds. Here, `general' means that the family is chosen outside a proper algebraic subset of the parameter space, thus automatically removing the birationally-isotrivial case. Grauert's proof and the proof for the family whose smooth members are hyperbolic [\textit{J. Noguchi}, Math. Ann. 258, 207--212 (1981; Zbl 0459.14002)] both use first-order differential equations, but this paper follows [\textit{J.-P. Demailly}, Proc. Symp. Pure Math. 62(pt.2), 285--360 (1997; Zbl 0919.32014)] and shows that there is a nontrivial differential equation of order \(n\) that is satisfied by the images of sections of such a family. Then adapting techniques similar to \textit{Y.-T. Siu} [in: The legacy of Niels Henrik Abel. Berlin: Springer. 543--566 (2004; Zbl 1076.32011)] and [\textit{S. Divero}, \textit{J. Merker} and \textit{E. Rousseau}, Invent. Math. 180, No. 1, 161--223 (2010; Zbl 1192.32014)], the Zariski-non-density of images of sections follows by constructing vector fields which isolate the coefficients of the differential equations. To obtain the desired differential equation, the author constructs a line bundle \(L\) on the jet space \(\mathcal X_{n}\) for the sections so that {\parindent=6mm \begin{itemize}\item[(1)] \(L = A - B\), where \(A\) and \(B\) are nef such that \(A^D > D A^{D-1}\cdot B\) (here \(D = \dim \mathcal X_{n}\)). \item[(2)] for any section \(s\) of \(\mathcal X\to C\), its lift to \(n\)-jet \(s_{n}: C\to \mathcal X_{n}\) pulls \(L\) back to a negative-degree divisor. \end{itemize}} Then it follows from algebraic Morse inequalities that \(L\) is big, and for any global section \(\sigma\) of a power of \(L\), \(s_{n}^*\sigma = 0\). Thus, the image of \(s_{n}\) for any section of \(\mathcal X\to C\) lies inside the zero of \(\sigma\), giving us the differential equation. The author constructs very specific \(A\) and \(B\), with \(B\) containing the pullback of \(\lambda\) as a subdivisor. To check nefness, he uses C. Voisin's idea that the nef cone and the pseudo-effective cone agree for a very general family. With this setup, (2) holds for \(\deg \lambda \gg 0\). The intersection numbers are estimated explicitly using Segre classes, and (1) is verified for \(d\gg 0\) and \(\deg(\mathcal X\to \mathbb P^n) \gg 0\). As a corollary of this method (it works for a finite cover of \(C\)), the author obtains a height inequality for sections whose \(n\)-jets do not lie inside a proper subvariety. This is a higher-dimensional analog of Proposition 3.1 in [\textit{P. Vojta}, Compos. Math. 78, No. 1, 29--36 (1991; Zbl 0731.14015)]. Moreover, in the appendix, the author shows by explicit calculations that one cannot find a differential equation of order \(\leq 2\) satisfied by the images of sections of such a family of surfaces (\(n=3\)). This demonstrates the optimality of order-\(n\), at least via this holomorphic Morse inequality method.