On the torsion values for sections of an elliptic scheme (Q2065867)

Let \[ \mathcal E \to C \] be a complex elliptic surface equipped with a section \(O\). Assume that the fibration is not isotrivial, i.e. the j-invariant of the generic fibre is non-constant. Denote by \(B\) the a subset of the places \(v\in C\) where the fibre above \(v\) is a smooth elliptic curve. Consider another section \(P\neq O\) such that \(P\) is not a torsion section (relative to the group structure on the generic fibre with \(O\) serving as neutral element). The paper under review is centered around the general question on which (smooth) fibers \(P\) specializes to a torsion point, i.e.\ it asks for what the authors call the torsion points of \(P\): \[ T_P = \{v\in B; \;\; P_v\in\mathop{Tor} \mathcal E_v\}. \] In fact, the set \(T_P\) is known to be large and small at the same time. Here `large' first of all means `infinite'. This property can be derived as a consequence of Siegel's theorem for integral points over function fields. Using the Betti map (which is also instrumental for the present paper), one can also prove that \(T_P\) is dense in the complex topology of \(B\). In the setting where everything is defined over some number field, there are even results towards the Galois equidistribution of the set \(T_P\) by \textit{L. DeMarco} and \textit{N. M. Mavraki} [Am. J. Math. 142, No. 2, 443--473 (2020; Zbl 1461.11102)]. On the other hand, the set \(T_P\) being `small' refers to sparsity along the lines of Silverman-Tate's specialization theorem. It should thus be clear that the set \(T_P\) is natural and interesting. The main concern of the present paper lies in those points which appear with multiplicity in \(T_P\). Geometrically, this means that the curves on \(\mathcal E\) corresponding to \(O\) and \(P\) intersect tangentially (as opposed to transverse intersection). The main theorem of the authors is that not only the set of multiple points in \(T_P\) is finite, but the same holds true for the union over all non-torsion sections in any finitely generated group \(\Gamma\) of sections. We note that the geometric interpretation opens the way to (complex) surface theory and surface geography which explains why similar results have been obtained independently, by different methods and with different aims, by \textit{D. Ulmer} and \textit{G. Urzúa} [Int. Math. Res. Not. 2021, No. 6, 4768--4802 (2021; Zbl 1477.14058); Sel. Math., New Ser. 28, No. 2, Paper No. 25, 36 p. (2022; Zbl 1492.14067)]. The paper under review also develops a new integral formula for the canonical height of a section. This uses again the Betti map, and in particular, the property that it is definable (proved in [\textit{G. Jones} and \textit{H. Schmidt}, Math. Ann. 379, No. 1--2, 825--864 (2021; Zbl 1457.14124)]) to control the behaviour of the sections at the bad fibres. The result can also be derived from [\textit{L. DeMarco} and \textit{N. M. Mavraki}, Am. J. Math. 142, No. 2, 443--473 (2020; Zbl 1461.11102)] as already cited above; the different approaches are compared in detail. Moreover the paper gives applications to diophantine approximation and (quasi-)integral points (i.e. with `small denominators'). Their proof involves an effctive version of Roth's theorem over function fields, adapting a method of \textit{J. T. Y. Wang} [Rocky Mt. J. Math. 26, No. 3, 1225--1234 (1996; Zbl 0942.11034)]. The authors also exhibit an extended explicit example illustrating the methods and techniques.