Harder-Narasimhan filtration for rank 2 tensors and stable coverings (Q321218): Difference between revisions

The paper under review addresses the unstable case in the GIT approach to the moduli of rank-2 tensors on a smooth projective variety \(X\), i.e.~ pairs \((E, \phi : E^{\otimes s} \to M)\) where \(E\) is a coherent torsion-free sheaf of rank \(2\) on \(X\) and \(M\) is a line bundle on \(X.\) In this setting, a Harder-Narasimhan filtration of an unstable pair cannot be defined naïvely, as one cannot speak of quotient objects. \textit{T. Gómez} et al. proved in [Rev. Mat. Complut. 28, No. 1, 169--190 (2015; Zbl 1327.14059)] that the Harder-Narasimhan filtration of an unstable coherent sheaf \(E\) coincides with a filtration induced by a 1-parameter subgroup that maximally destabilizes \(E\); the latter arises from work of \textit{G. R. Kempf} [Ann. Math. (2) 108, 299--316 (1978; Zbl 0406.14031)] and is therefore called the \textit{Kempf filtration}. The main result of the paper under review (Theorem 4.2) ensures a well-defined Kempf filtration for rank-2 tensors, which in turn is used to define a reasonable Harder-Narasimhan filtration (Definition 7.3). For sufficiently large \(m,\) there is a ``maximally unstable'' flag of subspaces of \(H^{0}(E(m))\) whose images under the evaluation map \(H^{0}(E(m)) \otimes \mathcal{O}(-m) \to E\) together with restrictions of \(\phi\) form the so-called \textit{\(m\)-Kempf filtration} of the pair \((E,\phi).\) Theorem 4.2 produces an \(m'\) such that the \(m\)- and \(m'\)-Kempf filtrations coincide for \(m \geq m'\); the Kempf filtration of \((E,\phi)\) can then be defined as its \(m\)-Kempf filtration for \(m \gg 0.\) As an application of the previous results, the author gives a characterization (Theorem 8.6) of an unstable rank-2 tensor over a smooth projective curve in terms of the intersection theory of the associated ruled surface.