Derived autoequivalences of generalised Kummer varieties (Q499870)

The bounded derived category of coherent sheaves \(\mathcal{D}(X):=D^b(\mathrm{Coh}(X))\) on a smooth projective variety \(X\) is nowadays one of the most studied invariants in algebraic geometry. One of its main features is the existence of analogues of what is known as correspondences on the level of cohomology or Grothendieck groups. Concretely, given an object \(\mathcal{P}\in \mathcal{D}(X\times Y)\) on the product of two varieties, there is the induced functor \[ \mathsf{FM}_{\mathcal{P}}:=R\mathrm{pr}_{Y*}(\mathcal P\otimes ^L\mathrm{pr}_X^*(-)): \mathcal {D}(X)\to \mathcal{D}(Y) \] called the Fourier-Mukai transform along the kernel \(\mathcal{P}\). In particular, if \(Y\) is a fine moduli space of sheaves on \(X\), one can study the Fourier-Mukai transform along the universal family. The hope is that this functor gives a better understanding of the relation between \(X\) and the associated moduli space \(Y\). \textit{N. Addington} [Algebr. Geom. 3, No. 2, 223--260 (2016; Zbl 1372.14009)] treated the case that \(X\) is a \(K3\) surface and \(Y=X^{[n]}\) is the Hilbert scheme of \(n\) points on \(X\) with \(n\geq 2\). Here, the Hilbert scheme is identified with the moduli space \(X^{[n]}\cong \mathcal{M}_X(1,0,n)\) of ideal sheaves of length \(n\) subschemes of \(X\). It turns out that the Fourier Mukai transform \(\mathrm{FM}_{\mathcal{I}_{\mathcal{Z}}}:\mathcal{D}(X)\to \mathcal{D}(X^{[n]})\) along the universal ideal sheaf is a \(\mathbb P^{n-1}\) functor and, consequently, induces an autoequivalence of \(\mathcal D(X^{[n]})\). The notion of a \(\mathbb P\)-functor was introduced in [loc. cit.] as a common generalisation of the \(\mathbb P\)-objects of Huybrechts-Thomas and split spherical functors (spherical functors were introduced in different ways and generalities by various authors; see Seidel-Thomas, Toda, Horja, Anno, Rouquier, Anno-Logvinenko). Concretely, let \(F: \mathcal{D}(X)\to \mathcal{D}(Y)\) be a Fourier-Mukai transform with left and right adjoints \(R\), \(L: \mathcal D(Y)\to \mathcal D(X)\) and let \(H\in \mathrm{Aut}(\mathcal{D}(X))\) be an exact autoequivalence. Then \(F\) is called a \(\mathbb P^n\)-functor with \(\mathbb P\)-cotwist \(H\) if there are isomorphisms of functors \(RF\cong \mathrm{id}\oplus H\oplus H^2\oplus \dots\oplus H^n\) and \(R\cong H^n L\) and the adjoint pair \((F,R)\) satisfies a technical condition concerning its monad structure. Every \(\mathbb P\)-functor induces an autoequivalence \(P_F\in \mathrm{Aut}(\mathcal D(Y))\), called \(\mathbb P\)-twist, of the target category. The main reason for the interest in \(\mathbb P\)-functors is the fact that \(P_F\) is often a non-standard autoequivalence. This means that it is not contained in the subgroup of \(\mathrm{Aut}(\mathcal{D}(Y))\) spanned by the obvious autoequivalences; namely push-forwards (and pull-backs) along automorphisms of the variety \(Y\), tensor products by line bundles, and degree shifts. In the paper under review, an analogue of Addington's result is proved for the second series of irreducible holomorphic symplectic moduli spaces of sheaves on surfaces; namely the generalised Kummer varieties. Given an abelian surface \(A\) and a positive integer \(n\in\mathbb N\), the generalised Kummer variety \(K_n A\subset A^{[n+1]}\) is the zero fibre of the morphism \(m: A^{[n+1]}\to A\) which sends the ideal sheaf of a subscheme \(Z\subset A\) to the sum of the weighted support of \(Z\). Concretely, the author proves that the Fourier-Mukai transform \(F_K: \mathcal{D}(A)\to \mathcal{D}(K_n A)\) along the universal ideal sheaf is a \(\mathbb P^{n-1}\)-functor with \(\mathbb P\)-cotwist \(H=[-2]\) for \(n\geq 2\). The case \(n=1\), where \(K_1A\) is a \(K3\) surface, was treated later by the author and the reviewer [Nagoya Math. J. 219, 1--8 (2015; Zbl 1342.14037)]. The paper is organised as follows: The first section gives a reminder of the definition of \(\mathbb P\)-functors and the induced twist autoequivalence along with some basic examples. In Section 2, the geometry of nested Hilbert schemes, which are natural subvarieties of the products \(A^{[n]}\times A^{[n+1]}\), is discussed. In Section 3, the composition of the Fourier-Mukai transform \(F: \mathcal{D}(A)\to \mathcal{D}(A^{[n+1]})\) along the universal ideal sheaf of the Hilbert scheme of points on the abelian surface with its right adjoint is computed. Section 4 provides the proof of the main result. Very roughly, considering the restriction of the nested Hilbert schemes to the Kummer varieties allows to pull-back Addington's computations from the Hilbert scheme case to the case of the generalised Kummer variety. One main ingredient of the proof is (one part of) the following statement which is of independent interest: The pull-back \(m^\ast:\mathcal{D}(A)\to \mathcal{D}(A^{[n+1]})\) along the summation morphism is a \(\mathbb P^n\)-functor with \(\mathbb P\)-cotwist \([-2]\). The proof of this statement is postponed to Section 5. Section 6 presents an alternative proof of the main result of the paper (as well as of Addington's result concerning the Hilbert schemes of points on \(K3\) surfaces) using the derived McKay correspondence. This section also contains some results concerning the McKay correspondence for generalised Kummer varieties which are interesting in its own. The final section contains some speculations on generalisations to other moduli spaces of sheaves on abelian and \(K3\) surfaces as well as to general hyper-Kähler manifolds.