Documentation

DerivedAlgGeo.AlgebraicGeometry.Numerical.RiemannRoch.K3

K3 surfaces, numerically #

A K3 surface is a smooth projective surface with trivial canonical bundle and no first cohomology. Numerically that is exactly two conditions on the Todd class:

IsK3 asserts those two and nothing else, so every consequence below is a consequence of those two facts about the Todd class, not of anything else one knows about K3 surfaces.

Main results #

The second identity is what makes the Mukai lattice and the discriminant two views of one object: v² ≥ −2 (the condition for a moduli space to be nonempty) and ∫Δ ≥ 0 (Bogomolov) differ by exactly 2r².

Not proved here #

That the axioms have a K3 model: a concrete NumericalVarietyData 2 A N with A = ℚ[t]/(t³), ∫t² = 2d, satisfying IsK3. Only the dimension-zero witness in Numerical/Models/DimensionZero/Point.lean exists so far, so the statements below are conditional on a NumericalVarietyData 2 A N existing at all. Building that model is the next task; it needs the internality of the grading, i.e. linear independence of 1, t, t².

The numerical signature of a K3 surface: trivial canonical class and χ(O_X) = 2.

Instances For
    theorem AlgebraicGeometry.Numerical.K3.chi_eq {A : Type u} {N : Type v} [CommRing A] [Algebra ℚ A] [AddCommGroup N] (V : NumericalVarietyData 2 A N) (hHRR : V.SatisfiesHRR) (hK3 : IsK3 V) (E : N) :
    ↑(V.chi E) = 2 * ↑(V.rank E) + V.ring.degree (V.chComp E 2)

    Riemann–Roch on a K3 surface: χ(E) = 2·rank(E) + ∫_X ch₂(E).

    The c₁-term of the surface formula drops out because td₁ = 0; this is the whole of what triviality of the canonical class buys.

    noncomputable def AlgebraicGeometry.Numerical.K3.mukaiS {A : Type u} {N : Type v} [CommRing A] [Algebra ℚ A] [AddCommGroup N] (V : NumericalVarietyData 2 A N) (E : N) :

    The third component of the Mukai vector v(E) = (r, c₁, s), integrated: s = rank(E) + ∫_X ch₂(E).

    For a K3, v(E) = ch(E)·√td(X) with √td(X) = 1 + [pt], so the top component of v is ch₂(E) + rank(E)·[pt]; mukaiS is its degree.

    Equations
    Instances For
      theorem AlgebraicGeometry.Numerical.K3.chi_eq_rank_add_mukaiS {A : Type u} {N : Type v} [CommRing A] [Algebra ℚ A] [AddCommGroup N] (V : NumericalVarietyData 2 A N) (hHRR : V.SatisfiesHRR) (hK3 : IsK3 V) (E : N) :
      ↑(V.chi E) = ↑(V.rank E) + mukaiS V E

      Riemann–Roch in Mukai coordinates: χ(E) = r + s.

      noncomputable def AlgebraicGeometry.Numerical.K3.mukaiSelfPairing {A : Type u} {N : Type v} [CommRing A] [Algebra ℚ A] [AddCommGroup N] (V : NumericalVarietyData 2 A N) (E : N) :

      The Mukai self-pairing ⟨v(E), v(E)⟩ = c₁² − 2rs.

      Equations
      Instances For

        The Mukai pairing and the Bogomolov–Gieseker discriminant differ by 2r²:

        ⟨v(E), v(E)⟩ = ∫_X Δ(E) − 2·rank(E)².

        So v² ≥ −2 and ∫Δ ≥ 0 are statements about the same quantity, shifted.

        IsK3 is omitted deliberately: this identity is arithmetic relating two definitions and holds on any surface. Only the reading of (r, c₁, s) as a Mukai vector is K3-specific.

        A rank-zero object has ⟨v, v⟩ = ∫_X c₁². Also independent of IsK3.