Documentation

DerivedAlgGeo.CategoryTheory.Triangulated.SphericalTwist.StabilityAction

A twist-shaped autoequivalence acts on stability conditions by the reflection #

This connects the lattice half of the spherical-twist lane to the stability machinery: an autoequivalence whose action on K₀ is the twist τ_E becomes a group element acting on WithClassMap C v, and that action is transport along the Mukai reflection ρ_{v(E)}.

It is the precise form of "the spherical twist acts on Stab(X) through the reflection" that is available before the twist is constructed.

Nothing here constructs a twist #

TwistShaped is supplied data and is named to say so. It carries an autoequivalence together with the hypothesis that its K₀ action is twistK₀. That is the property T_E will have once it exists; building T_E is a separate lane, and no inhabitant of TwistShaped is produced here.

Two things not to conclude #

The quotient is not Aut(D). AutPairQuot quotients by a bare natural isomorphism of underlying functors, leaving the CommShift datum unconstrained, so it is a priori coarser than exact autoequivalences up to isomorphism of exact functors. Slicing/Quotient.lean says this explicitly. No statement here is about Aut(Dᵇ(X)).

The order-two statement is about the lattice, not the functor. reflect_reflect makes lam an involution, so the class of a twist-shaped pair squares to one whose lattice part is the identity. It does not follow that T_E has order two in any automorphism group — it does not; T_E has infinite order. τ_E being an involution on K₀ says nothing about the functor.

Why [IsTriangulated C] is in the variable block #

GroupAction.AutPair is declared under a section carrying it, so it does not arrive from the sphericity hypotheses: chiK₀ and twistK₀ need only the pretriangulated structure, and EulerForm.lean's variable list omits it. Leaving it out makes AutPair fail to elaborate here for a reason that looks unrelated to the mathematics.

An autoequivalence shaped like the twist at E.

Supplied, not constructed: the K₀ action is asserted to be twistK₀. Φ is a TriEquiv rather than a bare C ≌ C precisely so that its additivity, shift-compatibility and triangulatedness travel as instances, which is what makes K₀.map Φ.e.functor well-formed.

Instances For

    The inverse acts by the twist too.

    Not free, and not waved at. K₀.map_congr on the unit isomorphism upgrades "Φ⁻¹ ∘ Φ is naturally isomorphic to the identity" into an equality of maps on K₀; composing that with map_eq and involutivity of twistK₀ identifies the inverse's action.

    The Mukai vector of E is spherical, from chi_self through the realization.

    The lattice automorphism attached to a twist-shaped pair: the Mukai reflection in v(E).

    Equations
    Instances For

      The compatibility AutPair asks for. v ∘ K₀(Φ⁻¹) = ρ ∘ v, which is map_inverse_eq followed by map_twistK₀.

      The group element.

      Equations
      Instances For

        The action is transport along the reflection. The existing actStabAut is reused, not restated; this records that the group element built here is the one that transport describes.

        The order-two statement, at the level the quotient supports.

        The lattice part of the square is the identity, by reflect_reflect. Read the module docstring before generalising this: it is a statement about lam, not about Φ, and T_E does not have order two in any automorphism group.