A Torelli type theorem for the moduli space of rank two connections on a curve (Q2575107)

Let \(X\) be a smooth projective curve over \({\mathbb C}\), of genus \(g\geq 4\) and let \(x_0\in X\) be a fixed point. Let \({\mathcal M}_X\) denote the moduli space of the pairs \((E,D)\) where \(E\) is a rank 2 vector bundle on \(X\) with \({\bigwedge}^2E\simeq {\mathcal O}_X(x_0)\) and \(D : E\rightarrow E\otimes {\Omega}^1_X\otimes {\mathcal O}_X(x_0)\) is a logarithmic connection with residue at \(x_0\) (which belongs to \(\text{End}(E_{x_0})\)) equal to \(-\text{Id}/2\) and inducing on \({\bigwedge}^2E\) the logarithmic connection on \({\mathcal O}_X(x_0)\) defined by the de Rham differential. For the construction of \({\mathcal M}_X\) as an irreducible quasi-projective variety see \textit{N. Nitsure} [J. Am. Math. Soc. 6, 597--609 (1993; Zbl 0807.14007)]. In this particular case, \({\mathcal M}_X\) has dimension \(6g-6\) and is smooth. The authors show that \({\mathcal M}_X\) determines \(X\) up to isomorphism. Their idea is to reduce the proof to the classical Torelli theorem. They consider the open subset \({\mathcal M}_X^0\) of \({\mathcal M}_X\) corresponding to the pairs \((E,D)\) with \(E\) stable. \({\mathcal M}_X^0\) is an affine fibration over the moduli space \({\mathcal N}_X\) of stable rank 2 vector bundles on \(X\) with determinant \({\mathcal O}_X(x_0)\). \({\mathcal M}_X\setminus {\mathcal M}_X^0\) has codimension \(\geq 3\), hence \(\text{H}^3({\mathcal M}_X,{\mathbb Z})\simeq \text{H}^3({\mathcal M}_X^0,{\mathbb Z})\simeq \text{H}^3({\mathcal N}_X,{\mathbb Z})\). Using the isomorphism \(\text{H}^3({\mathcal N}_X,{\mathbb Z})\simeq \text{H}^1(X,{\mathbb Z})\) established by \textit{D. Mumford} and \textit{P. E. Newstead} [Am. J. Math. 90, 1200--1208 (1968; Zbl 0174.52902)] one deduces an isomorphism of Hodge structures \(\text{H}^3({\mathcal M}_X)\simeq \text{H}^1(X)\) which implies that the intermediate Jacobian \(\text{J}^2({\mathcal M}_X)\) is isomorphic to the Jacobian \(\text{J}(X)\). Then (and this is the main part of the proof) the authors show how to recuperate the canonical principal polarization of \(\text{J}(X)\) from \({\mathcal M}_X\).