The Fano normal function (Q715172)

The paper under review deals with the geometry of smooth cubic threefolds, especially the study of \textit{Fano cycles} and their variations. Let us briefly recall its definition. For a smooth cubic threefold \(V\), denote by \(F\) its Fano surface of lines contained in \(V\). Then \(F\) is embedded, via the Albanese map, into the 5-dimensional abelian variety \(A:=\mathrm{Alb}(F)\), which is also identified with the intermediate Jacobian \(JV\) of \(V\). The image defines a 2-dimensional algebraic cycle \(F\in \mathrm{CH}_2(A)\). (Strictly speaking one has to fix a base point in \(F\), but let us ignore this here.) Composing with the involution \(-\mathrm{id}_{A}\), one has similarly \(F^-\in \mathrm{CH}_2(A)\). The \textit{Fano cycle} is by definition the cycle \(F-F^-\in \mathrm{CH}_2(A)\). The first main result of the paper is that for a generic cubic threefold \(V\), there is no \(f: W\to JV\) from a 4-dimensional variety \(W\) who has a homologically trivial cycle \(Z\in \mathrm{CH}_2(W)\) such that \(f_*(Z)\) is Abel-Jacobi equivalent to \(F-F^-\) in \(JV\). The second main result is a new proof of the result in [\textit{G. van der Geer} and \textit{A. Kouvidakis}, Doc. Math., J. DMV 15, 747--763 (2010; Zbl 1202.14009)] saying that for a generic cubic threefold \(V\), its Fano cycle \(F-F^-\in \mathrm{CH}^3(A)\) is not algebraically equivalent to zero. Their approach uses the theory of variation of Hodge structures. As the cohomology class of the Fano cycle is trivial, one can consider its Abel-Jacobi invariant \(\nu\in J^5(A)\). Corresponding to the Lefschetz decomposition of \(H^5(A)\) into primitive parts, we have \(J^5(A)=J(P^5(A)(2))\times J(P^3(A)(1))\times J(H^1(A))\), hence the decomposition \(\nu=\nu^1+\nu^3+\nu^5\), where \(J\) is the functor that associates an effective weight \(-1\) (pure) Hodge structure its Jacobian, \(P^3\) and \(P^5\) are primitive cohomology groups of \(V\) of degree 3 and 5 respectively and \(\nu^i\) is the component of \(\nu\) in \(J(P^i(A))\). As is remarked by the authors (Proposition 5.4), \(v^3=0\) and \(\nu^5\) is independent of the choice of the base point of \(F\). They make a detailed computation of the corresponding infinitesimal invariants \(\delta\nu\). Consider a family \(\mathcal V\to B\) parametrised by an open subset of the moduli space of smooth cubic threefolds, all the above objects vary in families \(\mathcal F\to B\), \(\mathcal A\to B\), \(\mathcal J^5\to B\), \(\mathcal P^5\to B\) etc. The Abel-Jacobi invariants of Fano cycles on fibres then give rise to a normal function of \(\mathcal J^5\to B\) or rather of \(\mathcal P^5\to B\). They prove actually a stronger result: this normal function is not a torsion section. Finally, the computation of the infinitesimal invariants shows also that the value of normal function associated to the Fano cycles determines the curve \(D(V)\subset F(V)\) parametrising the double lines in \(V\), which in turn determines the cubic threefold itself. One deduces therefore a Torelli type theorem.