Documentation

DerivedAlgGeo.CategoryTheory.Triangulated.SphericalTwist.Autoequivalence

T_E is an autoequivalence: a supplied statement, and what follows from it #

SphericalTwistData (in Definition.lean) supplies the functor T_E. That it is an autoequivalence is the Seidel--Thomas theorem, and it is stated here, never proved and never asserted.

Main definitions and results #

What an inhabitant claims #

SphericalTwistData records no more than a cone-like triangulated functor with the right K₀ shadow: nothing ties its copower and ev to Hom^•(E,-) ⊗ E and the evaluation map. So Seidel--Thomas supports an inhabitant only when d is the genuine twist datum, as the dg realizations of ObjectTwistData.lean are. Padding a genuine datum by the triangle P F ⟶ 0 ⟶ (P F)⟦1⟧ with P = 𝟭 ⊞ ⟦1⟧, so that [P F] = 0, preserves every field, including the class of the copower, but adds the summands F⟦1⟧ ⊕ F⟦2⟧ to T F, which an autoequivalence cannot have. An inhabitant is a claim about the specific d.

What the hypothesis is, and is not #

The hypothesis h is IsSphericalObject k 2 E of Basic.lean: the Ext profile alone, with no Serre-functor clause and none of the finiteness clauses of Seidel--Thomas' Definition 2.9. This is the clause the K₀ consequences need (χ(E,E) = 2), and on a K3, where ω_X ≅ O_X makes the second clause of Definition 1.1(a) automatic, it is sphericity. It is not sufficient for the theorem in general: AlgebraicGeometry/Surface/Spherical.lean records a surface with the same Ext profile on which T_E is not an autoequivalence. So the statement must be supplied only where the full hypothesis of Seidel--Thomas holds, and its inhabitant is a claim about that case.

What the isomorphism does not record #

iso is a natural isomorphism of underlying functors. Seidel--Thomas work with exact functors, that is functors with their shift isomorphisms, up to graded natural isomorphism, and compatibility with the two shift isomorphisms is not recorded here. So Φ's CommShift datum and T's are not asserted to agree. The statement is thereby weaker than the literature's, and no weaker than what is consumed: K₀.map_congr needs only the natural isomorphism.

The order-two statement of StabilityAction.lean remains about the lattice and not the functor: T_E has infinite order.

References #

Seidel--Thomas, arXiv:math/0001043v2: Theorem 1.2 (first part) says T_E is an exact self-equivalence of Dᵇ(X) for a spherical E on a smooth complex projective X. Proposition 2.10 is the abstract form, for an n-spherical E in the sense of their Definition 2.9 (a bounded complex of injectives with finite-dimensional Homs, the Ext profile and a nondegenerate pairing); Definition 2.14 transfers that notion to Dᵇ(S'), and Lemma 3.1 identifies it on a smooth projective X with E ⊗ ω_X ≅ E. Huybrechts, Fourier--Mukai transforms in algebraic geometry, Proposition 8.6, is the same statement.

The Seidel--Thomas autoequivalence theorem for T_E (their Proposition 2.10), as a supplied statement: d.T is naturally isomorphic to the functor of a triangulated autoequivalence, whose inverse and instances TriEquiv carries. Never inhabited here.

The parameter h is the Ext profile only and is not sufficient for the theorem (AlgebraicGeometry/Surface/Spherical.lean records a counterexample), so an inhabitant may be supplied only where E is spherical in Seidel--Thomas' full sense and d is the genuine twist. h is consumed only by toTwistShaped, for χ(E,E) = 2.

Instances For

    The supplied twist is twist-shaped. Φ ≅ T and K₀.map T = twistK₀ give K₀.map Φ = twistK₀, and χ(E,E) = 2 is chiK₀_of_self_eq_two.

    Equations
    Instances For
      @[simp]

      The autoequivalence of the constructed TwistShaped is the supplied one.

      The supplied twist acts on stability conditions through the reflection. The charge of T_E • σ is the charge of σ precomposed with the Mukai reflection ρ_{v(E)}; this is TwistShaped.act_Z at the twist-shaped pair that toTwistShaped produces.