Cycles in Jacobians: infinitesimal results (Q274713)

Let \(C\) a smooth projective curve of genus \(g \geq 3\), and \(p\) a point of it, and let us \(W_{k,p}\) the image of the symmetric product \(C^{(k)}\) inside the Jacobian \(J(C)\) via the Abel-Jacobi map. If we consider the involution \((-1): J(C) \to J(C)\), we can consider the cycle-theoretic pushforward \(W^-_{k,p}=(-1)_* W_{k,p}\). The difference \(W_{k,p}-W^-_{k,p}\), the Ceresa Cycle, is homologically trivial. Is a classical result of \textit{G. Ceresa}, Ann. Math., II. Ser. 117, 285-291 (1983; Zbl 0538.14024)] that, for \(C\) general, the Ceresa cycle is not algebraically trivial for \(k \leq g-2\). The author here presents a result on the algebraic triviality of a curve of genus \(g\geq 4\) varying in moduli. More precisely, he considers a family of smooth curves \(\pi: \mathcal{C} \to B\) with a section \(\sigma\) and associated Jacobian fibrations \(j(\pi): \mathcal{J}^p(\mathcal{C}) \to B\). We can consider the normal function \(v_{k,\sigma}(t)= AJ(W_{k, \sigma(t)}-W^-_{k, \sigma(t)})\), and in turn this proves to encode most information. Indeed the author is able to prove that the Ceresa cycle is not algebraically trivial in family for \(k=2, \ldots, g-2\) and moreover that we can reconstruct the curve \(C\) by its infinitesimal invariant \(\delta v_{2,\sigma}(0)\). Moreover in the end the author, generalizing a construction of \textit{A. Collino} [J. Algebr. Geom. 6, No. 3, 393--415 (1997; Zbl 0895.19001)], defines a \(K\)-theoretic counterpart of the 2-Ceresa cycle in the (general hyperelliptic) \(g=3\) case, showing that is indecomposable.