Documentation

DerivedAlgGeo.AlgebraicGeometry.Numerical.Models.Surface.RankOne

The intersection ring of a Picard-rank-one surface #

Every smooth projective surface with Pic X = ℤ·H has the same numerical intersection ring up to the one number ∫_X H²: it is ℚ[t]/(t³) with t = H, graded by codimension, and the degree map is pinned by where it sends H².

So the ring, its grading, the Chern character in the basis 1, H, H², and the degree map on the normal form are all shared. What distinguishes one surface from another is only the Todd class, and that is what the individual model files supply:

The degree is parametrised by h2 = ∫_X H² rather than by a surface, so a new Picard-rank-one model costs a Todd class and nothing else.

Models/MonogenicRing.lean performs this same construction in arbitrary dimension, and the threefold and fourfold models are built on it. This file is not rebuilt on top of that one: see the implementation notes there for why.

The ring ℚ[t]/(t³) #

The defining relation: nothing survives in codimension three on a surface.

Equations
Instances For
    @[reducible, inline]

    The numerical intersection ring A^•(X)_ℚ = ℚ[t]/(t³).

    Equations
    Instances For

      H³ = 0: the relation, in the ring.

      The degree map, pinned by ∫_X H² = h2 #

      ∫_X H² = h2: the degree of the surface.

      ∫_X on the normal form a + b·H + c·H². Only the top coefficient survives.

      @[reducible]

      The numerical intersection ring of a Picard-rank-one surface with ∫_X H² = h2.

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

        The Chern character #

        All three models use the same N. K3.lean and Abelian.lean share k3ChCoeff; ProjectivePlane.lean reparametrises with its own p2ChCoeff, because ch₂ is a half-integer on ℙ² — see Models/Surface/ProjectivePlane.lean. Coefficients are therefore rational rather than integral.

        @[reducible, inline]

        A class is recorded by the three rational coefficients of ch = a + b·H + c·H².

        Equations
        Instances For

          Shared Chern-character coordinates for the rank-one surface models: the class (r, c, s) has ch = r + c·H + s·H².

          The historical name is retained even though the construction is not K3-specific.

          Equations
          Instances For
            theorem AlgebraicGeometry.Numerical.Examples.surfaceCh_add (chCoeff : SurfaceNum → ℕ → ℚ) (hadd : ∀ (E F : SurfaceNum) (i : ℕ), chCoeff (E + F) i = chCoeff E i + chCoeff F i) (E F : SurfaceNum) (i : ℕ) :
            surfaceCh chCoeff (E + F) i = surfaceCh chCoeff E i + surfaceCh chCoeff F i
            theorem AlgebraicGeometry.Numerical.Examples.surfaceCh_sum (chCoeff : SurfaceNum → ℕ → ℚ) (E : SurfaceNum) :
            ∑ i ∈ Finset.range (2 + 1), surfaceCh chCoeff E i = (algebraMap ℚ SurfaceRing) (chCoeff E 0) + (algebraMap ℚ SurfaceRing) (chCoeff E 1) * H + (algebraMap ℚ SurfaceRing) (chCoeff E 2) * H ^ 2

            The product ch(E)·td(X), reduced. Everything of codimension above two is killed by H³ = H⁴ = 0, leaving the codimension-two coefficient a·t₂ + b·t₁ + c.

            This is the computational core shared by every Picard-rank-one model.