Fourier-Mukai transformations on \(K3\) surfaces with \(\rho = 1\) and Atkin-Lehner involutions (Q404558)

Given a complex smooth projective variety \(X\), its (bounded) derived category \(D(X):=D^b(\mathrm{Coh}(X)\) is widely recognised as an interesting invariant. Two guiding questions in the study of the derived categories are: {\parindent=6mm \begin{itemize}\item[1.] Determine the set \(\mathrm{FM}(X)\) of isomorphism classes of Fourier-Mukai partners of \(X\), that is varieties \(Y\) with \(D(X)\cong D(Y)\). \item[2.] Describe the group \(\mathrm{Aut}(D(X))\) of autoequivalences of \(D(X)\). \end{itemize}} Both of these questions are usually very difficult, and answers, even partial ones, are known only for special classes of varieties. One of these classes is formed by the \(K3\) surfaces. By results of Mukai and Orlov, it is known that every Fourier-Mukai partner of a \(K3\) surface \(X\) is again a \(K3\) surface and a moduli space of stable sheaves on \(X\). Furthermore, two \(K3\) surfaces \(X\) and \(Y\) are Fourier-Mukai partners if and only if there is a Hodge isometry \(H^*(X, \mathbb Z)\cong H^*(Y, \mathbb Z)\) between the Mukai lattices; this is known as the derived Torelli theorem. Furthermore, every autoequivalence canonically induces an isometry of the Mukai lattice which means that there is a homomorphism of groups \(\rho: \mathrm{Aut}(D(X))\to \mathrm O(H^*(X,\mathbb Z))\). The image of this homomorphism was determined by work of Hosono, Lian, Oguiso and Yau [\textit{S. Hosono} et al., J. Algebr. Geom. 13, No. 3, 513--545 (2004; Zbl 1070.14042)] (see also Ploog's PhD thesis) and Huybrechts, Macrì, and Stellari [\textit{D. Huybrechts} et al., Duke Math. J. 149, No. 3, 461--507 (2009; Zbl 1237.18008)]. It is the index two subgroup of Hodge isometries preserving a natural orientation. For the kernel of \(\rho\), there is a conjectural description due to Bridgeland involving stability conditions and period domains. This conjecture was proved for (non-algebraic) \(K3\) surfaces of Picard rank 0 by \textit{D. Huybrechts} et al. [Compos. Math. 144, No. 1, 134--162 (2008; Zbl 1152.14037)] as well as for \(K3\) surfaces of Picard rank 1 by \textit{A. Bayer} and \textit{T. Bridgeland} [``Derived automorphism groups of \(K3\) surfaces of Picard rank 1'', \url{arXiv:1310.8266}]. \textit{S. Hosono} et al. [Contemp. Math. 322, 43--55 (2003; Zbl 1058.14056)] studied Fourier-Mukai partners of \(K3\) surfaces of Picard rank 1 in greater detail. In particular they provided the formula \(|\mathrm{FM}(X)|=2^{\nu(d)-1}\) for the Fourier-Mukai number \(|\mathrm{FM}(X)|\) in terms of the degree of \(X\), i.e.\ the self-intersection number \(2d=L_X^2\) of an ample generator of the Néron-Severi group. Here, \(\nu(d)\) is the number of different prime factors of \(d\). One can see that \(|\mathrm{FM}(X)|\) coincides with the index \([\mathrm{AL}_d: \mathrm{Fr}_d]\) of the Fricke group inside the Atkin-Leher group. The paper under review provides a natural explanation of this coincidence by showing that every element of the Atkin-Lehner group is naturally induced by Fourier-Mukai equivalence \(D(X)\mathop{\rightarrow}\limits^{\cong} D(Y)\) and is contained in the Fricke subgroup if and only if it is induced by an autoequivalence. This agrees with conjectural expectations of [loc. cit.]. More concretely, the author constructs a natural functor \(M: \mathcal{FM}_X\to \mathrm{AL}_d\). Here, \(\mathcal{FM}_X\) is the groupoid whose objects are the Fourier-Mukai partners of \(X\). The morphisms are given by the Fourier-Mukai equivalences between them. Then, he shows that the functor is surjective on the level of morphisms, and the image of \(\mathrm{Aut}_{\mathcal{FM}}(X,X)\) is \(\mathrm{Fr}_d\). Let us explain this in some more detail. The Fricke and Atkin-Lehner groups \(\mathrm{Fr}_d\leq \mathrm{AL}_d\leq \mathrm{PSL}_2(\mathbb R)\) are subgroups of the projective special linear group given by unions of certain cosets of the congruence subgroup \(\Gamma_0(d)\leq \mathrm{PSL}_2(\mathbb R)\). Recall that \(\mathrm{PSL}_2(\mathbb R)\) can be identified with the group of automorphisms of the upper half plane \(\mathbb H\) via Möbius transformations. Every Fourier-Mukai equivalence \(\Phi: D(X)\mathop{\rightarrow}\limits^{\cong} D(Y)\) induces an isometry \(\Phi_*: \mathcal N(X)\to \mathcal N(Y)\) between the numerical Grothendieck groups \(\mathcal N(X)=H^0(X,\mathbb Z)\oplus \mathrm{NS}(X)\oplus H^4(X,\mathbb Z)\). The special feature of the Picard rank 1 case that comes into play is that there are canonical identifications \(\mathcal N(X)\cong N_d\cong \mathcal N(Y)\) with the abstract lattice \(N_d=\mathbb Z e_0\oplus \mathbb Z \ell \oplus \mathbb Z e_4\) with \(e_0e_4=-1=e_4e_0\), \(\ell^2=2d\), and all other pairings being zero. These identifications are given by choosing as base vectors of \(\mathcal N(X)\) and \(\mathcal N(Y)\) the class of the whole surface, the class of an ample generator of the Néron-Severi group, and the class of a point. This identification allows to canonically associate to every Fourier-Mukai equivalence (not just to autoequivalences) an isometry of \(N_d\). Let \(\mathfrak D^+(X)\) be the connected component of \(\{[v]\in\mathbb P(\mathcal N(X)\otimes \mathbb C)\mid v^2=0\,,\, v\bar v>0\}\) which contains \([\exp(i\omega)]\) for \(\omega\) an ample divisor. The isometry \(\Phi_*: \mathcal N(X)\to \mathcal N(Y)\) induces a bijection \(\mathfrak D^+(X)\to \mathfrak D^+(Y)\) since it is orientation preserving. Furthermore, the identification \(\mathcal N(X)\cong N_d\cong \mathcal N(Y)\) induces a canonical identification \(\mathfrak D^+(X)\cong \mathbb H\cong \mathfrak D^+(Y)\). Hence, we get an induced \(\Phi_*\in \mathrm{Aut}(\mathbb H)\cong \mathrm{PSL}_2(\mathbb R)\). One can see from a result of \textit{I. V. Dolgachev} [J. Math. Sci., New York 81, No. 3, 2599--2630 (1996; Zbl 0890.14024)] that actually \(\Phi_*\in \mathrm{AL}_d\leq \mathrm{PSL}_2(\mathbb R)\). In summary, \(\Phi\mapsto \Phi_*\) defines the functor \(M: \mathcal{FM}_X\to \mathrm{AL}_d\). Very roughly, the paper is organised as follows. Section 2 is essentially a very convenient review of the necessary results from the literature. These results allow the author to construct the functor \(M\) and prove its surjectivity in Section 3.