Embedding of a compactified Jacobian and theta divisors (Q1785556)

Let \(Y\) be an integral curve with \(m\) nodes (ordinary double points) over \(\mathbb C\), and \(p:X\longrightarrow Y\) be the normalization. Let \(\overline J(Y)\) denotes the compactified Jacobian of rank one torsion-free sheaves of degree \(-1\) on \(Y\), and \(h: \widetilde J(Y) \longrightarrow\overline J(Y)\) be the normalization. One can identify \(\widetilde J(Y)\) with the moduli space of generalized parabolic line bundles on \(X\) [\textit{U. Bhosle}, Ark. Mat. 30, No. 2, 187--215 (1992; Zbl 0773.14006)]. Let \(U_Y(r,d)\) be the moduli space of semi-stable torsion-free sheaves of rank \(r\) and degree \(d\) on \(Y\), and let \(P(r,d)\) denotes the moduli space of \(\alpha\)-semi-stable \((0\leq\alpha < 1)\) generalized parabolic bundles (GPBs in short) of rank \(r\) and degree \(d\) on \(X\) [loc. cit.]. Assume \(\alpha\) is close to \(1\) (e.g. \(1-\frac{1}{r(r-1)}<\alpha<1\)). Then there is a surjective birational morphism \(f: P(r,d) \longrightarrow U_Y(r,d)\). Let \(E_0\) be a stable vector bundle on \(Y\) of rank \(r\) degree \(d+r, -r < d \leq 0\) such that \(p^*E_0\) is stable, and let \(\overline{E_0'}\) (\(E_0'= p^*E_0\)) be a corresponding GPB over \(X\). Consider the morphisms \(\alpha_Y: \overline{J}(Y) \longrightarrow U_Y(r,d)\) given by \(L \mapsto L\otimes E_0\) and \(\widetilde{\alpha}_Y: \widetilde{J}(Y) \longrightarrow P(r,d)\) given by \(\overline{L}\mapsto \overline{L}\otimes \overline{E_0'}\) (this is a tensor product of two GPBs). In this article, the author shows that, if \(E_0'=p^*E_0\) has no line bundle direct summands, then both \(\alpha_Y\) and \(\widetilde{\alpha}_Y\) are embedddings. Let \(\Theta_{\overline{J}(Y)}, \Theta_{U_Y(r,d)}\) be the theta divisors on \(\overline{J}(Y)\) and \(U_Y(r,d)\) respectively. Set \(\Theta_{\widetilde{J}(Y)}=h^*\Theta_{\overline{J}(Y)}\) and \(\Theta_{P(r,d)}=f^*\Theta_{U_Y(r,d)}\). With the assumption that \(p^*E_0\) has no line bundle direct summands, the author obtains several relations between the theta divisors on pull backs under the morphisms \(\alpha_Y\) and \(\widetilde{\alpha}_Y\). Let \(\text{gcd}(r,d)=1\), and \(E_{r,d}\) be the Picard bundle on \(U_Y(r,d), r \geq 2\) and \(d\geq (2g-2)r\). Choose a stable vector bundle \(E_0\) on \(Y\) such that \(\alpha_Y:\overline{J}(Y) \longrightarrow U_Y(r,d), L \mapsto L\otimes E_0\) is an embedding. The author proves that the restriction of \(E_{r,d}\) to \(\overline{J}(Y)\) is stable with respect to any theta divisor \(\Theta_{\overline{J}(Y)}\) if \(d > r(2g-1)\) and semi-stable if \(d=r(2g-1)\). Similar results for \(r=1\) were obtained in [\textit{U. N. Bhosle} and \textit{A. J. Parameswaran}, Int. Math. Res. Not. 2014, No. 15, 4241--4290 (2014; Zbl 1303.14042)]. The author also shows that if the genus of \(X\) is at least two, then any general stable bundle of rank \(\leq 2\) (of any degree) has natural cohomology; also provided sufficient conditions on vanishing of first and second cohomology of a general stable bundle of rank \(r\geq 3\). In the case of smooth curves, similar results were obtained in [\textit{Y. Li}, Int. J. Math. 2, No. 5, 525--550 (1991; Zbl 0751.14019)]. Reviewer's Comment: In the course of the proof of Theorem 1.1, the following fact is used: \(\mathcal{E}nd^0(p^*E_0)\), the trace zero endomorphisms, has no line bundle direct summands (Proposition 3.1). It's not clear that the assumption of Theorem 1.1 is sufficient to conclude this. But the author proves that for a general stable bundle \(E_0\), this condition is satisfied (Proposition 2.9).