A sharp Bogomolov-type bound (Q374038)

Let \(A\) be an abelian variety with a symmetric ample line bundle \(\mathcal L\). Denote by \(\hat{h}_{\mathcal L}\) the associated Néron-Tate height. The essential minimum \(\mu_{\mathcal L}(Y)\) of a subvariety \(Y\) of \(A\) is defined as \[ \mu_{\mathcal L}(Y)=\sup\{\theta \mid \overline{Y(\theta)}\neq Y\}, \] where \(Y(\theta)=\{x\in Y(\overline{ \mathbb Q}) \mid \hat{h}_{\mathcal L}(x)\leq \theta\}\) and \( \overline{Y(\theta)}\) is its Zariski closure. A theorem of \textit{E. Ullmo} [Ann. Math. (2) 147, No. 1, 167--179 (1998; Zbl 0934.14013)] and \textit{S.-W. Zhang} [Ann. Math. (2) 147, No. 1, 159--165 (1998; Zbl 0991.11034)], previously known as Bogomolov's Conjecture, states that \(\mu_{\mathcal L}(Y)\) is strictly positive if and only if \(Y\) is nontorsion, where one says that a subvariety is torsion if it is the union of components of algebraic subgroups. A subvariety is called translate if it is the union of translates of algebraic subgroups. A subvariety is called transverse if it is not contained in any translate other than \(A\). An effective Bogomolov bound is a statement according to which, for any \(\eta>0\), there exists an effectively computable constant \(C(A,\mathcal L,\eta)\) such that, for any transverse subvariety \(Y\), \[ \mu_{\mathcal L}(Y)\geq C(A,\mathcal L,\eta) \frac{1}{ (\deg_{\mathcal L} Y)^{ \frac{1}{\mathrm{codim}\, Y}+\eta} } \cdotp \] Such a bound is known in a certain number of cases [\textit{A. Galateau}, Comment. Math. Helv. 85, No. 4, 775--812 (2010; Zbl 1250.11071)] (Galateau's result includes the case where the abelian variety has a positive density of ordinary primes), [\textit{M. H. Baker} and \textit{J. H. Silverman}, Math. Res. Lett. 11, No. 2--3, 377--396 (2004; Zbl 1060.11041)]; [\textit{J.-P. Serre}, Abelian \(l\)-adic representations and elliptic curves. Wellesley, MA: A K Peters (1998; Zbl 0902.14016)]; [\textit{P. Deligne, J. S. Milne, A. Ogus} and \textit{K.-y. Shih}, Hodge cycles, motives, and Shimura varieties. Berlin etc.: Springer-Verlag (1982; Zbl 0465.00010)] and is conjectured by Serre to be always true \textit{R. Pink} [J. Reine Angew. Math. 495, 187--237 (1998; Zbl 0920.14006)]. The main result of the paper under review refines such an estimate as follows. Assume that an effective Bogomolov bound holds. Then there exists an effectively computable constant \(C'(A,\mathcal L,\eta)\) with the following property. Let \(H\subset A\) be a translate. Then, for any subvariety \(Y\) contained in \(H\) and not contained in any translate strictly contained in \(H\), \[ \mu_{\mathcal L}(Y)\geq C'(A,\mathcal L,\eta)\frac{ (\deg_{\mathcal L} H)^{\frac{1}{\mathrm{codim}_H Y}-\eta}}{ (\deg_{\mathcal L} Y)^{ \frac{1}{\mathrm{codim}_H Y}+\eta} } \cdotp \] This bound is a new step in the direction of an abelian analogue by David and Philippon of a toric conjecture of Amoroso and David and has applications in the framework of anomalous intersections.