Autoequivalences of derived categories via geometric invariant theory (Q323728)

The paper under review is devoted to the study of autoequivalences of the derived category of bounded complexes of coherent sheaves on a variety coming from a variation of a geometric invariant theory (GIT) quotient. More precisely, let \(X\) be a smooth quasiprojective variety and consider the action of a reductive group \(G\) on \(X\). Given a \(G\)-ample equivariant line bundle \(L\), we denote by \(X^{ss}\) the open semistable locus defined by GIT. A variation of GIT quotient (VGIT) is given by the birational transformation \(X^{ss}_-/G \dashrightarrow X^{ss}_+/G\) induced by the variation of \(L\). The authors restrict their investigation to a kind of VGIT called balanced wall crossing. In the first part they prove that the window shift autoequivalence \(\Phi_w\) of \(\text{D}^b(X^{ss}_-/G)\) is a twist corresponding to the spherical functor \(f_w:\text{D}^b(Z/L) \rightarrow \text{D}^b(X^{ss}_-/G)\); here \(Z/L\) denotes the critical locus of VGIT and \(w \in \mathbb{Z}\). Actually, they prove a more general result, stating that if \(\mathcal{C}\) is a pre-triangulated dg category with a semiorthogonal decomposition \(\mathcal{C}=\langle \mathcal{A},\mathcal{B} \rangle=\langle \mathcal{B},\mathcal{A}' \rangle=\langle \mathcal{A'},\mathcal{B'} \rangle=\langle \mathcal{B}',\mathcal{A}\rangle\), then the autoequivalence of \(\mathcal{B}\) induced by mutation is the twist corresponding to a spherical functor \(S: \mathcal{A} \rightarrow \mathcal{B}\) (Theorem 3.11). They also prove a converse of this statement (Theorem 3.15). In the case of a balanced wall crossing, this applies to the subcategory \(\mathcal{C}_w\) of \(\text{D}^b(X/G)\) defined in terms of grade restriction rules (see Section 2); the resulting autoequivalence agrees with \(\Phi_w\) (Proposition 3.4). A summary of these results is presented in Theorem 1.1. In the second part they define the fractional grade restriction rule, in order to construct additional equivalences according to the prediction coming from mirror symmetry. They show that if \(\text{D}^b(Z/L)_w\) has a full exceptional collection \(\langle E_i \rangle_{i=0}^N\), then the objects \(f_w(E_i)\) are spherical and \(\Phi_w\) is the composition of the corresponding twists (Corollary 4.13). In Section 4.2 the authors suggest an interpretation of this factorization theorem in terms of monodromy of the quantum connection in a neighborhood of a partial large volume limit. More generally, let \(S: \mathcal{E} \rightarrow \mathcal{G}\) be a spherical functor of dg categories with a semiorthogonal decomposition \(\mathcal{E}=\langle \mathcal{A}, \mathcal{B}\rangle=\langle F_S(\mathcal{B}),\mathcal{A}\rangle\), where \(F_S\) is the cotwist autoequivalence of \(\mathcal{E}\) induced by \(S\). Then the restrictions of \(S\) to \(\mathcal{A}\) and \(\mathcal{B}\) are spherical and the twist functor associated to \(S\) is the composition of twists corresponding to these restrictions (Theorem 4.14). The second part is summarized in Theorem 1.2.