Existence of semistable sheaves on Hirzebruch surfaces (Q2656148)

The moduli spaces \(M_{X,H}(\mathbf{v})\) of \(S\)-equivalence classes of Gieseker semistable sheaves on a smooth complex projective surface play a significant role in several areas in mathematics including algebraic geometry, symplectic geometry, mirror symmetry and mathematical physics. Probably the most basic question is the existence (or, geography) problem: determine when \(M_{X,H}(\mathbf{v})\) is empty or not. When \(H\) is ample and if there is an \(H\)-Gieseker semistable sheaf with Chern character \(\mathbf{v}\), a famous Bogomolov's inequality implies that the discriminant \(\delta(\mathbf{v}) \ge 0\) is nonnegative. On the other hand, if \(\delta(\mathbf{v})\) is sufficiently large, there is an \(H\)-Gieseker semistable sheaf with Chern character \(\mathbf{v}\) [\textit{K. G. O'Grady}, Invent. Math. 123, No. 1, 141--207 (1996; Zbl 0869.14005)] and the moduli space behaves nice as expected. Unfortunately, a complete classification of such \(H\)-semistable Chern characters is known only for a few simple cases such as \(\mathbb{P}^2\), \(K3\) surfaces, abelian surfaces, anticanonically polarized del Pezzo surfaces. The paper under review studies the problem for Hirzebruch surfaces \(\mathbb{F}_e := \mathbb{P}(\mathcal O_{\mathbb{P}^1} \oplus \mathcal O_{\mathbb{P}^1}(e))\) for any integer \(e \ge 0\). Let \(F\) be the fiber to \(\mathbb{P}^1\) and let \(E\) denote a section so that \(E^2 = -e\). Note that any ample \(\mathbb{R}\)-divisor \(H\) is a multiple of \(H_m := E + (e+m)F\) for some positive real number \(m>0\). Thus it is sufficient to study \(H_m\)-semistable sheaves on \(\mathbb{F}_e\): whether they exist or not, and how to construct them if they exist. The paper under review seems to follow a similar story as the work of Drézet and Le Potier on stable Chern characters on \(\mathbb{P}^2\) [\textit{J.-M. Drézet} and \textit{J. Le Potier}, Ann. Sci. Éc. Norm. Supér. (4) 18, 193--243 (1985; Zbl 0586.14007)], however, the authors point out that exceptional bundles do not control the complete classification in some polarizations. First, the authors showed a sharp Bogomolov inequality by a unique real-valued function \(\delta_m^{\mu -s} (\nu)\) of the total slope \(\nu\) (see Theorem 1.1 and Theorem 9.7) for \(\mu_{H_m}\)-stable sheaves. When \(e=0\) or \(e=1\) and \(H\) is (a multiple of) the anticanoncal polarization, this function can be computed by exceptional bundles as similar as Rudakov's work, cf. [\textit{A. N. Rudakov}, J. Reine Angew. Math. 453, 113--135 (1994; Zbl 0805.14023); \textit{A. N. Rudakov}, J. Math. Sci., Tokyo 3, No. 3, 495--532 (1996; Zbl 0883.14021)]. On the other hand, if \(H\) is different from the anticanonical polarization, one has to consider orthogonal pairs of Kronecker modules which may provide length \(2\) generic Harder-Narasimhan filtrations where no factor is semiexceptional, see Section 1.7 and Section 10. Then, to construct a (Gieseker) semistable sheaf, the authors considered an irreducible family of sheaves, namely, the stack of \(F\)-prioritary sheaves \(\mathcal{P}_F(\mathbf{v})\). It is clear that if \(M_{H_{m}}(\mathbf{v})\) is nonempty, then \(\mathcal{P}_F(\mathbf{v})\) has open dense substacks \[ M_{H_m}(\mathbf{v}) \subset \mathcal P_{H_{\lceil m \rceil + 1}} (\mathbf{v}) \subset \mathcal P_{F}(\mathbf{v}). \] Hence, the existence problem reduces to two smaller question: the nonemptiness of \(\mathcal P_{H_{\lceil m \rceil + 1}} (\mathbf{v})\), and the \(H_m\)-semistability of the general sheaf in \(\mathcal P_{H_{\lceil m \rceil + 1}} (\mathbf{v})\). For the nonemptiness, see Theorem 1.2 and Section 4. For the \(H_m\)-semistability, the authors studied the Harder-Narasimhan filtration of the general sheaf. Also note that the authors provide a method to approximate \(\delta_m^{\mu -s} (\nu)\) and to compute it as a limit, see Section 5. The authors first made the explicit classification for \(\mathbb{F}_0\) and \(\mathbb{F}_1\) (see Section 6-10), and then they show that the linear map \[ \begin{array}{rrcl} \pi : & K(\mathbb{F}_e) \otimes \mathbb{Q} & \longrightarrow & K(\mathbb{F}_{e-2}) \otimes \mathbb{Q} \\ & (r, aE+bF, d) & \mapsto & (r, aE^{\prime} + (b-a) F^{\prime}, d) \end{array} \] reduces the case over any Hirzebruch surface \(\mathbb{F}_e\) to either one of \(\mathbb{F}_0\) or \(\mathbb{F}_1\) (see Section 11).