A GIT interpretation of the Harder-Narasimhan filtration (Q2256771)

Let \(E\) be a coherent unstable torsion free sheaf on a smooth complex projective variety \(X\). Let \(V=H^0(E(m))\) for \(m\in \mathbb Z\). For big \(m\) one obtains the quotient \(V\otimes \mathcal O_X(-m)\to E\) and hence \(E\) defines a GIT unstable point in the corresponding Quot scheme. As shown by \textit{G. R. Kempf} in [Ann. Math. (2) 108, 299--316 (1978; Zbl 0406.14031)] there is a maximally destabilizing \(1\)-parameter subgroup giving a weighted filtration of \(V\) and thus inducing a filtration of \(E\), called the \(m\)-Kempf filtration of \(E\). The authors show that the \(m\)-Kempf filtrations stabilize for big \(m\), which defines a filtration called the Kempf filtration of \(E\). The main result of the paper states that the Kempf filtration of \(E\) coincides with the Harder-Narasimhan filtration of \(E\). The paper consists of six sections. Section 1 is an introduction. Section 2 introduces the theorem of Kempf. In Section 3 some machinery related to convex cones is developed. It is used in Section 4, where the properties of the Kempf filtrations are studied. The authors demonstrate in Section 5 that the \(m\)-Kempf filtrations stabilize. This allows to introduce for a torsion free sheaf \(E\) on \(X\) its Kempf filtration. The main result of the paper is proven in Section 6.