Documentation

HironakaExamples.MaximalContact.Example82

Kollár's Example 82: going down is not onto, and a general H need not give an equivalence #

[Kol07, Example 82] illustrates the two defects of going down along a hypersurface of maximal contact. First part: for I = (xy − zⁿ) and H = (x = 0), (H, I|_H) ≅ (𝔸², (zⁿ)) has a one-dimensional locus of order n, while I has an isolated point of order 2, so a blow-up of order ≥ n of (H, I|_H, 2) centred at another point of that line has no counterpart of order 2 on (X, I); for the general H_g = (x − y = 0), (H_g, I|_{H_g}) ≅ (𝔸², (x² − zⁿ)) has the same isolated point of order 2. Second part: for I = (x³ + xy⁵ + z⁴) and the general H = (x + u₁xy³ + u₂y⁴ + u₃z² = 0), after two chart blow-ups the transform of I has order 2 while its restriction to the transform of H still has order 3.

All statements are computations in polynomial rings, in the vocabulary of Hironaka/Resolution/Algebraic/Balanced/:

Two generic tools for orders at a point #

theorem Hironaka.Examples.pderiv_mem_pow_of_mem_pow_succ {σ : Type u_1} {R : Type u_2} [CommRing R] (i : σ) {N : Ideal (MvPolynomial σ R)} {a : ℕ} {g : MvPolynomial σ R} (hg : g ∈ N ^ (a + 1)) :

The Leibniz rule for a partial derivative: g ∈ N^{a+1} implies ∂ᵢ g ∈ N^a.

theorem Hironaka.Examples.eval_eq_zero_of_mem_span {σ : Type u_1} {R : Type u_2} [CommRing R] {S : Set (MvPolynomial σ R)} (a : σ → R) (hS : ∀ s ∈ S, (MvPolynomial.eval a) s = 0) {g : MvPolynomial σ R} (hg : g ∈ Ideal.span S) :

Evaluation at a point kills every polynomial of an ideal whose generators vanish there.

Example 82, first part, in ℚ[x, y, z] #

x ∈ 𝔪₀ = (x, y, z). HironakaExamples/MaximalContact/Example14.lean, which names the variables x₁, x₂, y, uses this statement and the next two for its own generators.

theorem Hironaka.Examples.cosupport_example82 (n : ℕ) (hn : 2 ≤ n) (a b c : ℚ) (h : ∀ g ∈ Ideal.derivative ℚ (Ideal.span {MvPolynomial.X 0 * MvPolynomial.X 1 - MvPolynomial.X 2 ^ n}), (MvPolynomial.eval ![a, b, c]) g = 0) :
a = 0 ∧ b = 0 ∧ c = 0

The locus of order 2, V(D(I)), has the origin as its only ℚ-point.

(zⁿ) has order ≥ n at every ℚ-point of the line z = 0: the one-dimensional locus of order n on H.

I has order < 2 at (0, 1, 0) ∈ H, a point of the line of order n on H.

D(x² − zⁿ) in ℚ[x, z]: the partial derivatives of the generator.

D(x² − zⁿ) = (x, z^{n−1}) in ℚ[x, z].

theorem Hironaka.Examples.cosupport_example82_Hg (n : ℕ) (hn : 2 ≤ n) (a c : ℚ) (h : ∀ g ∈ Ideal.derivative ℚ (Ideal.span {MvPolynomial.X 0 ^ 2 - MvPolynomial.X 1 ^ n}), (MvPolynomial.eval ![a, c]) g = 0) :
a = 0 ∧ c = 0

The locus of order 2 of I|_{H_g} has the origin as its only ℚ-point.

Example 82, second part, over a nontrivial commutative ℚ-algebra #

theorem Hironaka.Examples.mul_mem_pow_add {A : Type u_2} [CommRing A] {N : Ideal A} {a b : ℕ} {g g' : A} (hg : g ∈ N ^ a) (hg' : g' ∈ N ^ b) :
g * g' ∈ N ^ (a + b)

A product of elements of N^a and N^b lies in N^{a+b}.

theorem Hironaka.Examples.mem_pow_of_mem_pow_of_le {A : Type u_2} [CommRing A] {N : Ideal A} {k l : ℕ} (h : l ≤ k) {g : A} (hg : g ∈ N ^ k) :
g ∈ N ^ l

N^k ⊆ N^l for l ≤ k.

A nontrivial ℚ-algebra has characteristic zero.

h ∈ D²(f) = MC(I). Over the ℚ-algebra R the partial derivatives are ℚ-derivations, and 6x = ∂ₓ∂ₓ f, 5y⁴ = ∂_y∂ₓ f, 20xy³ = ∂_y∂_y f, 12z² = ∂_z∂_z f lie in D²(f); dividing by the nonzero rationals gives the four monomials of h.

The elimination identity: (1 + u₁y₂³)³ · f₂ = g + h₂ · q with g free of x₂, so f₂ restricted to h₂ = 0 is g up to the unit (1 + u₁y₂³)³.

Kollár's I₂|_{H₂} ⊂ (y₂³, y₂²z₂⁴): g ∈ (y₂³, y₂²z₂⁴).

Kollár's "its restriction to the birational transform H₂ of H still has order 3": g has order ≥ 3, and exactly 3 when u₂ is a unit (∂_y³ g at the origin is −6u₂).