Degeneration of the archimedean height pairing of algebraically trivial cycles
Zhelun Chen
TL;DR
The paper investigates the limiting behavior of the archimedean height pairing for homologically trivial cycles in a degenerating family and proves that, for algebraically trivial cycles, the limit is controlled by the geometric height pairing on the generic fiber provided Griffiths's incidence equivalence conjecture holds. It builds a bridge between archimedean heights and non-archimedean geometric height pairings via a network of regular homomorphisms, higher Picard varieties, and relative algebraic intermediate Jacobians, culminating in an isomorphism between the relevant height line bundles and Lear extensions. A key outcome is a precise comparison between geometric Beilinson–Bloch height pairings and their Néron–Tate/Néron-model counterparts, formalized through Poincaré biextensions and $G_m$-biextensions, and conditional on Griffiths's conjecture, establishing the desired equivalence of line bundles on the base. The results provide a geometric viewpoint on Beilinson–Bloch height positivity and connect to the positivity results of Brosnan–Pearlstein, with implications for the study of height pairings over one-variable complex function fields.
Abstract
We consider the limiting behaviour of the archimedean height pairing for homologically trivial algebraic cycles in a degenerating one-parameter family of smooth projective complex varieties. We conjecture that the limit is controlled by the non-archimedean geometric height pairing of the cycles on the generic fiber and verify this for algebraically trivial cycles, assuming a conjecture of Griffiths on incidence equivalence. Our work offers a more geometric understanding of a related asymptotic result of Brosnan--Pearlstein and suggests a new perspective on the positivity of the Beilinson--Bloch height pairing over a one-variable complex function field.
