Arakelov geometry on degenerating curves (Q2121377)

In 1984, the author of the present article introduced the delta-invariant \(\delta_F(C)\) of a compact connected Riemann surface \(C\) [\textit{G.~Faltings}, Ann. Math. 119, No. 2, 387--424 (1984; Zbl 0559.14005)]. It appears as an archimedean contribution in the Noether formula for arithmetic surfaces. This formula suggests that the delta-invariant should be interpreted as an archimedean analogue of the discriminant (i.e.~the number of singular points in the geometric fiber at a prime \(p\)). Phrased differently, one could think of \(\delta_F(C)\) as a measure of the ``distance'' from the moduli point \([C]\) in the moduli space of curves towards the boundary. In view of this interpretation, the author raised the question to determine the asymptotic behavior of the invariant towards the boundary. Intimately related is the question about the asymptotic behavior of the Arakelov metric and the Arakelov Green's functions. In the present article, the author gives an answer to these questions: Consider a family \(C\) of semistable curves over a complex analytic base \(S\) with smooth generic fiber. Then asymptotic formulae for the Arakelov metric, for Arakelov's Green functions and for the delta-invariant are given. These involve local discriminants of the singularities and the dual graph of the semistable degeneration. For families over a one dimensional base, asymptotic formulae appear in the work of \textit{R.~de Jong} [J. Differential Geom III, No. 2, 241--301 (2019; Zbl 1436.14047)]. A main achievement of this paper is the generalization to a base of arbitrary dimension. Also \textit{J.~Jorgenson} [Duke Math. J. 61, No. 1, 221--254 (1990; Zbl 0746.30032)] and \textit{R.~Wentworth} [Comm. Math. Phys. 137, No. 3, 427--459 (1991; Zbl 0820.14017)] have considered the problem over higher dimensional bases. However, the main results here differ from those in that they are more geometric and do not use theta functions. For example, the result describing the asymptotic behavior of Green's functions is as follows. Consider a fiber \(C_s\). This is a union of curves \(C_{\nu,s}\) with discs removed, parametrised by the vertices of the dual graph \(\Gamma\) describing the degeneration, and annuli parametrised by the edges \(e\) of \(\Gamma\). On each \(C_{\nu,s}\), the Green's function differs up to uniformly (in \(s\)) bounded terms from a well-behaved function by a constant depending on \(\Gamma\). On the annuli one has linear interpolation except for some explicit correction terms.