Documentation

DerivedAlgGeo.LinearAlgebra.Matrix.GeneralLinearGroup.UniversalCover.Basic

The lifted general linear group as compatible pairs #

Bridgeland presents the universal cover of GL⁺(2, ℝ) concretely, as pairs (T, f) with T ∈ GL⁺(2, ℝ) and f a normalized shift, subject to T and f inducing the same map on the circle of phases.

The circle is ℝ / 2ℤ, embedded in ℝ² by φ ↦ (cos πφ, sin πφ). The factor π (not 2π) is Bridgeland's phase convention: a semistable object of phase φ has central charge on the ray at angle πφ, so φ ↦ φ + 1 is the antipodal map — the shift functor [1]. rayVec_add_one below is exactly that statement, and it is the reason NormalizedShift demands f (φ + 1) = f φ + 1.

"Same map on the circle" is then: T sends the ray through rayVec φ to the ray through rayVec (f φ).

What is proved across this development #

This file proves that the compatible pairs form a group under componentwise multiplication, and that the type is nonempty (compat_one is the witness). The companion files prove the covering-space properties:

Thus GLTilde.universalCoverData packages the surjective covering-map and simple-connectedness properties, while exact_deckHom_toMatHom packages the extension with fibre ℤ. UniversalCover/TopologicalGroup.lean proves continuity of multiplication and inversion and installs IsTopologicalGroup GLTilde.

The action on stability conditions is downstream and stays with stability: StabilityCondition/Symmetry/GLTilde/Action/PreStability.lean and .../Action/Stability.lean use this group to define and prove it.

The phase circle #

The unit vector at phase φ: the ray at angle π φ.

The π, rather than 2π, is Bridgeland's phase convention — see the module docstring.

Equations
Instances For

    Advancing the phase by one is the antipodal map.

    This is the compatibility between NormalizedShift's +1-equivariance and the shift functor [1]: adding 1 to a phase negates the central charge.

    Phase vectors are never zero, so they really do determine rays.

    Rays #

    v lies on the same ray as w: a positive multiple of it.

    Deliberately not Mathlib's SameRay, which admits the zero vector. Every vector here is a rayVec, hence nonzero, and the strict version keeps the closure proofs free of degenerate cases.

    Equations
    Instances For

      Compatible pairs #

      T and f induce the same map on the circle of phases: T carries the ray at phase φ to the ray at phase f φ.

      Equations
      • One or more equations did not get rendered due to their size.
      Instances For

        The group #

        G̃L⁺(2, ℝ) in Bridgeland's concrete presentation: a matrix of positive determinant together with a phase relabelling that agrees with it on the circle.

        The covering-map and simple-connectedness properties are proved in the companion topology files and packaged by GLTilde.universalCoverData.

        Instances For

          compat is a Prop, so the two data fields determine the pair.

          Named ext': Lean auto-generates GLTilde.ext for the structure.

          @[implicit_reducible]
          Equations
          • One or more equations did not get rendered due to their size.

          Projection to the linear part. A group hom by construction.

          Equations
          • One or more equations did not get rendered due to their size.
          Instances For

            Projection to the phase relabelling. A group hom by construction.

            Step 3 composes the Slicing action with this.

            Equations
            • One or more equations did not get rendered due to their size.
            Instances For