Groups generated by spherical twists on K3 surfaces and full exceptional collections on Fano threefolds

Technical Summary: Groups Generated by Spherical Twists on K3 Surfaces and Full Exceptional Collections on Fano Threefolds

Problem Statement
The paper addresses two interconnected problems in algebraic geometry and derived categories:

1. The structural description of the subgroup $G \subset \text{Aut}(\mathcal{D}^b(Y))$ generated by all spherical twists on a smooth K3 surface $Y$ of Picard rank 1. Specifically, the authors seek to determine whether $G$ is a free group, identify its free generators, and establish conditions under which $G$ is finitely generated.
2. The verification of the first three-dimensional case of the Bondal-Polishchuk conjecture regarding the transitivity of the braid group action on full exceptional collections. The conjecture posits that the braid group $B_n$ acts transitively on the set of full exceptional collections of length $n$ in a triangulated category. While counterexamples exist in general, the status for derived categories of smooth projective varieties remains open in higher dimensions.

Methodology
The authors employ a synthesis of derived category theory, stability conditions, and the theory of Fuchsian groups acting on the hyperbolic plane.

1. Stability Conditions and Geometry of $\text{Stab}(Y)$:
   Building on the foundational work of Bridgeland and Bayer-Bridgeland, the authors utilize the contractibility of the distinguished component $\text{Stab}^\circ(Y)$ of the space of stability conditions on a K3 surface $Y$ of Picard rank 1. They leverage the exact sequence:
   $$1 \to \text{Aut}_0 \mathcal{D}^b(Y) \to \text{Aut} \mathcal{D}^b(Y) \to \text{Aut}^+ H^*(Y, \mathbb{Z}) \to 1$$
   where $\text{Aut}^+ H^*(Y, \mathbb{Z})$ is the group of Hodge isometries preserving orientation. The kernel $\text{Aut}_0$ is identified with the fundamental group of a specific component of the stability space.

2. Reduction to Fuchsian Groups:
   The authors map the problem to the geometry of the upper half-plane $\mathbb{H}$. They identify the quotient of the stability space (modulo shifts and automorphisms) with an orbifold Riemann surface $\mathbb{H}/\Gamma$, where $\Gamma$ is a Fuchsian group (specifically related to the Atkin-Lehner group $\Gamma_0^+(\delta)$).

   • Spherical objects in $\mathcal{D}^b(Y)$ correspond to "roots" (vectors of square $-2$) in the Mukai lattice.
   • Spherical twists correspond to reflections across hyperplanes orthogonal to these roots.
   • The group $G$ generated by spherical twists is analyzed via its image $\bar{G}$ in the group of spherical reflections acting on $\mathbb{H}$.

3. Topological Analysis of Orbifolds:
   Using the theory of orbifold fundamental groups, the authors analyze the quotient $\mathbb{H}_0 / \bar{G}$, where $\mathbb{H}_0$ is the hyperbolic plane with punctures corresponding to the roots. They demonstrate that $\mathbb{H}_0 / \bar{G}$ is a non-compact Riemann surface. Consequently, the fundamental group $\pi_1(\mathbb{H}_0 / \bar{G})$, which is isomorphic to $\bar{G}$, is a free group.

4. Application to Fano Threefolds:
   To address the Bondal-Polishchuk conjecture, the authors consider a Fano threefold $X$ with a full exceptional collection of length 4. They restrict exceptional objects from $X$ to a generic anticanonical divisor $Y \subset X$, which is a K3 surface of Picard rank 1.

   • They establish a map $\phi$ from exceptional collections on $X$ to tuples of spherical twists on $Y$.
   • They prove that $\phi$ intertwines the mutation action of the braid group $B_4$ on collections in $X$ with the Hurwitz action of $B_4$ on the group $G^4$.
   • Since $G$ is shown to be a free group of rank 4 in these specific cases, the transitivity and freeness of the braid group action on $G^4$ (a known result for free groups) imply the corresponding properties for the exceptional collections on $X$.

Key Contributions and Results

1. Structure of the Spherical Twist Group:

   • Theorem 1.3: For a K3 surface $Y$ of Picard rank 1, the subgroup $G \subset \text{Aut}(\mathcal{D}^b(Y))$ generated by all spherical twists is a free group. The free generators are explicitly identified as spherical twists associated with a specific set of spherical vector bundles (orbit representatives of the action of $G$ on the set of spherical objects).
   • Theorem 1.4: The group $G$ is finitely generated if and only if the degree $d = 2\delta$ of the K3 surface satisfies $\delta \in \{1, 2, 3, 5, 7, 11, 17, 19, 23, 29, 31, 41, 47, 59, 71\}$. For degrees outside this set (e.g., $d=8$), the group is infinitely generated.
   • The authors provide a precise recipe to find these generators using fundamental domains for the action of the corresponding reflection groups on the hyperbolic plane.

2. Resolution of the Bondal-Polishchuk Conjecture (3D Case):

   • Theorem 1.5: For any Fano threefold $X$ of Picard rank 1 admitting a full exceptional collection of four vector bundles (specifically $X = \mathbb{P}^3, Q^3, V_5, V_{22}$), the action of the braid group $B_4 \ltimes \mathbb{Z}^4$ (mutations and shifts) on the set of full exceptional collections is transitive and free.
   • Corollary 8.7: Furthermore, every full exceptional collection on such an $X$ consists entirely of shifts of vector bundles and is strong.

3. Classification of Spherical Objects:
   The description of $G$ yields a precise classification of spherical objects in $\mathcal{D}^b(Y)$ up to the action of the group generated by spherical twists and shifts.

Significance
The paper claims significance in two primary areas:

• Derived Categories of K3 Surfaces: It provides a complete structural description of the group generated by spherical twists, a fundamental object in the study of autoequivalences of K3 surfaces. It resolves the question of finite generation, showing it depends strictly on the arithmetic properties of the degree of the surface.
• Exceptional Collections on Fano Threefolds: It settles the first non-trivial dimension (dimension 3) of the Bondal-Polishchuk conjecture. By reducing the problem to the Hurwitz action on a free group, the authors bypass the difficulties of directly analyzing exceptional objects in dimension 3, instead leveraging the well-understood geometry of the associated K3 surface.

The authors note that while their results confirm transitivity for these specific Fano threefolds, they do not claim to classify all exceptional vector bundles on these varieties, nor do they rule out the existence of exceptional objects that are not shifts of vector bundles (though they prove that full collections must consist of such shifts). The work relies heavily on the contractibility of the stability space for K3 surfaces of Picard rank 1, a result established by Bayer and Bridgeland.