On the variation of a series on Teichmüller space (Q952981)

In the celebrated paper [Invent. Math. 132, No. 3, 607--632 (1998; Zbl 0916.30039)], the author proved that the series \[ \sum_{\gamma} \frac{2}{1+\mathrm{exp}(\ell_\gamma)}=1 \] where \(\gamma\) ranges over the simple closed geodesics of a one-punctured hyperbolic torus. In this paper, the author addresses a question of T. Jørgensen and he proves that another similar series is constant on the Teichmüller space of tori with one geodesic boundary component, namely \(\mathcal{P}=3\pi/2\), where \[ \mathcal{P}:=\sum_{\gamma} \mathrm{arctan}\frac{\mathrm{cosh}(\ell_\delta/4)}{\mathrm{sinh}(\ell_\gamma/2)} \] and \(\ell_\delta\) is the length of the boundary geodesic. The series is strictly related to \textit{M. Mirzakhani}'s identities [Invent. Math. 167, 179--222 (2007; Zbl 1125.30039)]. The proof is indirect as the author produces another divergent series \(\mathcal{Q}\) (defined using intersection angles between pairs of geodesics) such that its formal differential \(d\mathcal{Q}\) absolutely converges to \(d\mathcal{P}\) and then works with \(\mathcal{Q}\). He uses variational formulae by \textit{S. Kerckhoff} [Ann. Math. (2) 117, 235--265 (1983; Zbl 0528.57008)] and \textit{S. Wolpert} [Ann. Math. (2) 115, 501--528 (1982; Zbl 0496.30039)]. The methods are different than in the author's Inventiones paper: he explains where the previous strategy fails.