The asymptotics of the Arakelov-Green's function and Faltings' delta invariant (Q1814250)

The Arakelov metric on a Riemann surface can be described as follows. Embed the surface \(M\) in its Jacobian \(J(M)\), and restrict the flat metric on \(J(M)\) to \(M\). This is the canonical metric on \(M\), and the Arakelov metric can be defined by prescribing that its curvature be proportional to the Kähler form of the canonical metric. Considering this metric on \(M\), the author considers the asymptotic behavior as \(M\) degenerates, acquiring a node. From the introduction: ``In our case, the starting point for the analysis of the Arakelov-Green's function is an expression in terms of a theta function and theta divisor: \[ \log G(x,y) = {1 \over h!} \int_{\Theta + x - y} \nu^{n - 1} \log \| \theta \| + A. \] This formula is due to Bost, and in section 5 we give a proof based on the bosonisation formula. With the Green's function in this form, the problem is essentially transferred to the Jacobian variety, where explicit formulae for the degeneration are available.'' This analysis is used to study Faltings' delta function. The formula derived for \(M\) degenerating into \(M_ 1 \cup M_ 2\) intersecting at a point is: \[ \lim_{t \to 0} \left[ \delta (M_ t) + 4{h_ 1h_ 2 \over h} \log | t | \right] = \delta (M_ 1) + \delta (M_ 2). \] (A similar, somewhat more complicated formula is given if \(M\) stays irreducible but acquires a node.) \textit{J. Jorgenson} [Duke Math. J. 61, No. 1, 221-254 (1990; Zbl 0746.30032)] also studied this situation. His result was that the left- hand side of the above expression is finite and bounded. He wrote [loc. cit.]: ``This limit, however, depends on the choices of local coordinates \(z\) on \(M_ t\).'' The author, on the other hand, writes: ``Thus, if for the local coordinates \(z_ 1\), \(z_ 2\) on the surfaces \(M_ 1\), \(M_ 2\) we define the parameter \(\tau = t \sqrt {g_{z_ 1 \overline z_ 1} (0)} \sqrt {g_{z_ 2 \overline z_ 2} (0)}\) (with obvious notation) then expansions to the first order in \(\tau\) are independent of the initial choise of coordinates'', an observation he attributes to Scott Wolpert. In this respect the result here is slightly more complete than that in the paper cited above.