No label defined (Q4344784)

Let \(A\) be an abelian variety over \(\mathbb{Q}\). The Mordell-Weil theorem states that \(A(\mathbb{Q})\) is a finitely generated abelian group. One very interesting problem is to study how its rank varies as \(A\) varies. In this paper the author generalizes his previous results [\textit{P. Michel}, Monatsh. Math. 120, 127-136 (1995; Zbl 0869.11052)] on the average rank of families of elliptic curves to abelian varieties. He starts with a family of abelian varieties defined over \(\mathbb{Q}\) and parametrized by a non-empty affine subset \( U_{\mathbb{Q}}\) of the projective line, say \(U_{\mathbb{Q}}=\)Spec\((\mathbb{Q}[T,1/P(T)])\), where \(P(T)\in \mathbb{Z}[T]\) denotes a square free polynomial whose coefficients are each coprime to two. In this situation there exists an abelian scheme \(f:\mathcal{A}\rightarrow \) \(U_{\mathbb{Q}}\) of dimension \(g\), and the extension of scalars to \(\mathbb{C}\) defines a family of abelian varieties over \(U_{\mathbb{C}}=\mathbb{P}_{\mathbb{C}}^{1}-(\{z_{j} ; P(z_{j})=0\} \cup \{\infty \})\). There exists a sheaf \(\mathcal{F}=R^{1}f_{*}\mathbb{Q} _{\ell }\) over \(U_{\mathbb{C}}\) which is smooth of rank \(2g\), where \(\ell\) denotes a prime number. Fixing a base point \(z_{0}\in U_{\mathbb{C}}\), \(\mathcal{F}\) corresponds bijectively to a representation \(\rho _{z_{0}}\) (called monodromy) of the topological fundamental group \(\pi _{1}(U_{\mathbb{C}},z_{0})\) over \(\mathcal{F} _{z_{0}}=H^{1}(\mathcal{A}_{z_{0}},\mathbb{Q}_{\ell })\cong \mathbb{Q}_{\ell }^{2g}.\) The author's main result is: under the hypothesis that the Taniyama-Shimura and the Birch and Swinnerton-Dyer conjectures hold for the abelian varieties of the family \(f:\mathcal{A}\rightarrow U_{\mathbb{Q}}\) as well as the Generalized Riemann Hypothesis, he proves that if \(\rho _{z_{0}}\) is irreducible then \[ \sum_{|t|\leq T, P(t)\neq 0}\text{rank}(\mathcal{A}_{t}(\mathbb{Q}))\leq \frac{ \mathcal{L}(T)}{\log (T)}(1+o(1))+(2g(d-1)-\kappa +o(1))2T, \] where \(\kappa \in \mathbb{R}\) satisfying \(|\kappa |\leq 2g\) is determined by the image of \(\pi _{1}(U_{\mathbb{C}},z_{0})\) by \(\rho _{z_{0}}\), \(\mathcal{L} (T)=\sum_{|t|\leq T, P(t)\neq 0}\log (N_{\mathcal{A}_{t}})\leq (2g(d-1)+o(1))2T\log T\) denotes the logarithmic conductor, \(N_{\mathcal{A} _{t}}\) the conductor of the fiber at \(t\in U_{\mathbb{Q}}\) of \(f\) and \(d=\deg (P(T))+1\). The present source of examples of families of abelian varieties satisfying the irreducibility hypothesis of \(\rho _{z_{0}}\) come from families on Jacobian varieties of curves over a surface. More precisely, let \(f:\mathcal{ C}\rightarrow U_{\mathbb{C}}\) be a proper flat smooth morphism of smooth connected curves of genus \(g\geq 1\). In this case \(f\) has a section \(s\) which defines the unity elements of the Jacobian varieties of the curves of the family \(f\). In certain cases where one disposes of explicit equations for \(\mathcal{C}\), \textit{N. Katz} [Prog. Math. 12, 171-195 (1981; Zbl 0475.14025)] gives a sufficient condition for the irreducibility of \( \rho _{z_{0}}\). Namely, suppose that \(F(X)\in \mathbb{Z}[X]\) is a polynomial of degree at least \(3\) and the two variable polynomial \( F(X,Z)=(F(X)-F(Z))/(X-Z)\) is irreducible over \(\mathbb{C}\). Let \(y^{2}=F(x)+1\) be a family of hyperelliptic curves defined over \(U=\{t\in \mathbb{C}\); \(F(X)+t \) has \(d\) distinct roots in \(\mathbb{C}\}.\) Then for every \(z_{0}\in U\) the monodromy representation \(\rho _{z_{0}}\) is irreducible. However, the author works in the opposite direction, since the proof of the irreducibility of \(\rho _{z_{0}}\) depends on estimates of exponential sums over finite fields. Let \(\mathcal{X}\) be a smooth projective surface defined over \(\mathbb{Q}\) which is geometrically connected. For \(i=0,\cdots ,4\) let \( b_{i}(\mathcal{X},\ell )=\dim _{\mathbb{Q}_{\ell }}H_{\text{et}}^{i}(\mathcal{X} _{\overline{\mathbb{Q}}},\mathbb{Q}_{\ell })\), its \(i\)-th Betti number, where \(H_{ \text{et}}^{i}\) denotes the étale cohomology groups. In this set-up \( b_{i}(\mathcal{X},\ell )=b_{4-i}(\mathcal{X},\ell )\) and \(b_{0}(\mathcal{X} ,\ell )=0\). The author assumes that \(b_{1}(\mathcal{X},\ell )=0\) and considers a family \((\mathcal{X}_{t})_{t\in \mathbb{P}_{\mathbb{Q}}^{1}}\) of curves obtained by a Lefschetz pencil on \(\mathcal{X}\) defined over \(\mathbb{Q}\). Let \(U_{\mathbb{Q}}\subset \mathbb{P}_{\mathbb{Q}}^{1}\) be the open subset where the curves \(\mathcal{X}_{t}\) are smooth\(.\) Suppose that the common genera \(g\) of these curves is at least \(1\). Then \(\rho _{z_{0}}\) is irreducible. Furthermore, assuming once again that the Taniyama-Shimura and the Birch and Swinnerton-Dyer conjectures hold for the Jacobian varieties \(J_{\mathcal{X} _{t}}\) of the curves of the family \((\mathcal{X}_{t})_{t\in \mathbb{Z}}\), as well as the Generalized Riemann Hypothesis, he shows that as \(T\rightarrow +\infty \) the following inequality holds \[ \sum_{|t|\leq T, t\in U_{\mathbb{C}}}\text{rank}(J_{\mathcal{X}_{t}}(\mathbb{Q} ))\leq (2d-4g+\delta '-\delta -d_{2}'-1/2+o(1))2T, \] where \(d=\#(\mathbb{P}_{\mathbb{C}}^{1}-U_{\mathbb{C}})\) does not depend on the Lefschetz pencil (just on the projective embedding of \(\mathcal{X}\)), \(\delta \) is the cardinality of the intersection \(\Delta \) of the ax of the pencil with \(\mathcal{X}\), \(\delta '\) is the number of orbits under the action Gal\((\overline{\mathbb{Q}}/\mathbb{Q)}\) on \(\Delta \) and \(d_{2}'\) is the number of singular fibers of the pencil having two irreducible components. This bound is the analogue of the main result of [\textit{P. Michel}, Monatsh. Math. (loc. cit.)]. Finally, at the end of the paper the author gives a beautiful unconditional version of his main result for Lefschetz's pencils defined over the function field \(\mathbb{F}_{q}(X)\) of a smooth projective irreducible curve \(X\) defined over \(\mathbb{F}_{q}\).