Documentation

DerivedAlgGeo.AlgebraicGeometry.Numerical.Models.Threefold.LinearSection

Linear-section coordinates on a Picard-rank-one threefold #

Both threefold models in this directory — ℙ³ and the quintic Calabi–Yau — record a class by its multiplicities against the flag of structure sheaves of linear sections

[O_X], [O_{X ∩ H}], [O_{X ∩ H²}], [O_pt].

Those coordinates are shared, and so are the resulting Chern-character coefficients: the Koszul resolution of a k-fold hyperplane section gives ch(O_{X ∩ Hᵏ}) = (1 − e^{−H})ᵏ, which through codimension three is

The point class is the only place the degree enters, and it is why the coefficients are parametrised by d rather than fixed.

Why these coordinates #

The alternative — recording (r, c₁, ch₂, ch₃) directly — cannot be a lattice: ch₂ is a half-integer and ch₃ a sixth-integer, so χ would not land in ℤ. Exactly the same problem forced the (r, c, v) coordinates in Models/Surface/ProjectivePlane.lean. Linear sections solve it structurally instead of by a substitution chosen to work: every generator is an actual coherent sheaf, so its χ is an integer by construction, and the integrality of χ on the whole lattice is then automatic rather than something the model has to arrange.

Main definitions #

Main results #

References #

@[reducible, inline]

A class on a Picard-rank-one threefold, in linear-section coordinates: (a, b, c, e) stands for a·[O_X] + b·[O_{X∩H}] + c·[O_{X∩H²}] + e·[O_pt].

Equations
Instances For

    The Chern-character coefficients of a class in linear-section coordinates on a threefold of degree ∫_X H³ = d. The coefficient of Hⁱ is entry i; see the module docstring for where each number comes from.

    Equations
    Instances For

      The rank of a class in linear-section coordinates is its [O_X] multiplicity: the other three generators are supported in positive codimension.

      Equations
      • One or more equations did not get rendered due to their size.
      Instances For
        theorem AlgebraicGeometry.Numerical.Examples.threefoldChi_sum (d : ℚ) (t : ℕ → ℚ) (E : ThreefoldNum) :
        ∑ i ∈ Finset.range (3 + 1), threefoldChCoeff d E i * t (3 - i) = ↑(E 0) * t 3 + ↑(E 1) * t 2 + (-↑(E 1) / 2 + ↑(E 2)) * t 1 + (↑(E 1) / 6 - ↑(E 2) + ↑(E 3) / d) * t 0

        The Riemann–Roch sum of rankOneNumericalVariety, with Finset.range 4 expanded and the 3 − i reductions performed.

        Each model then has a goal in the four Todd coefficients and nothing else, which is what makes a new threefold cost a table of numbers rather than a proof.