The Bogomolov conjecture for totally degenerate abelian varieties (Q2385043)

Let \(K\) be a number field, let \(A\) be an abelian variety over \(K\), and let \(X\) be a closed subvariety of \(A_{\overline K}\). Assume that \(X\) is not a translate of an abelian subvariety by a torsion point. Let \(L\) be an ample symmetric line bundle on \(A\), and let \(\widehat h_L\colon A(\overline K)\to\mathbb R\) be the associated Néron-Tate canonical height. A form of Bogomolov's conjecture asserts that there is an \(\epsilon>0\) for which the set \(X(\epsilon):=\{P\in X(\overline K):\widehat h_L(P)\leq\epsilon\}\) is not Zariski-dense in \(X\). This was proved by \textit{S. Zhang} [Ann. Math. (2) 147, 159--165 (1998; Zbl 0991.11034)], building on work of E. Ullmo. Let now \(K\) be a function field; i.e., \(K=k(B)\), where the constant field \(k\) is algebraically closed and \(B\) is an integral projective variety over \(k\), regular in codimension 1. With \(X\), \(A\), and \(L\) defined similarly to the preceding paragraph, the canonical height \(\widehat h\) is again defined, but Bogomolov's conjecture does not carry over directly. For example, it is false in the split case. It is conjectured to hold when the \(K/k\)-trace of \(A\) is trivial, though. This paper proves the Bogomolov conjecture in the case where \(A\) is \textit{totally degenerate} at some place \(v\) of \(K\). This is defined as follows. Let \(v\) be a place of \(K\) (coming from a prime divisor on \(B\)), let \(\mathbb K_v\) be the completion of the algebraic closure of the completion of \(K\) at \(v\), let \(A_v=A\times_K\mathbb K_v\), and let \(A_v^{\text{an}}\) denote the \textit{Berkovich analytic space} associated to \(A_v\). Then \(A\) is \textit{totally degenerate} at \(v\) if analytically \(A_v^{\text{an}}\cong\bigl(\mathbb G_{\text{m}}^n\bigr)_v^{\text{an}}/M\) for a lattice \(M\) in \(\mathbb G_{\text{m}}^n(\mathbb K_v)\). The latter is defined to be a discrete subgroup that maps isomorphically onto a complete lattice in \(\mathbb R^n\) by the (continuous) map \[ \text{val}\colon\bigl(\mathbb G_{\text{m}}^n\bigr)_v^{\text{an}}\to\mathbb R^n,\qquad p\mapsto(p(x_1),\dots,p(x_n)) \] where \(x_1,\dots,x_n\) are fixed coordinates on \(\mathbb G_{\text{m}}^n\). For example, the Tate elliptic curve is totally degenerate. (Recall that in a Berkovich space a point \(p\) is defined as a seminorm on the affinoid ring of some open affinoid neighborhood.) The proof is an adaptation of Zhang's proof, replacing Zhang's equidistribution result at an archimedean place with a non-archimedean equidistribution result, phrased using Berkovich theory and tropical geometry. This result relies on work from the author's previous paper [\textit{W. Gubler}, Invent. Math. 169, 321--376 (2007; Zbl 1153.14036)].