Documentation

Hironaka.Resolution.Algebraic.Tuning.Corollary101

Tuning of ideals, II: Kollár's Corollary 101 #

Kollár's Corollary 101 [Kol07, Corollary 101] (Tuning of ideals, II): for m = max-ord I ≥ 1 and s = r · lcm(2, …, m) with r ≥ m − 1, the tuned ideal W_s(I) is MC-invariant and D-balanced, and a smooth blow-up sequence is of order ≥ m starting with (X, I, m, E) iff it is of order ≥ s starting with (X, W_s(I), s, E). "Everything follows from (99) and (100), except for the role played by E."

m = max-ord I bounds the order of every stalk (the hypothesis of the coordinate lemmas on Proposition 99 at each point).

The tuned ideal is MC-invariant, D-balanced, of maximal order s #

theorem AlgebraicGeometry.Scheme.IdealSheafData.isMCInvariant_W {k : Type u} [Field k] [CharZero k] {X : Scheme} (f : X ⟶ Spec ↧k) (n : ℕ) [SmoothOfRelativeDimension n f] (I : X.IdealSheafData) (m : ℕ) (hI : I.maxOrd = ↑m) (hm : 1 ≤ m) {s : ℕ} (hs : 1 ≤ s) :
IsMCInvariant f (W f I m s) s

[Kol07, Corollary 101; Proposition 99 (5)]: for m = max-ord I ≥ 1 and s ≥ 1, W_s(I) is MC-invariant with respect to s, that is MC_s(W_s(I)) · D(W_s(I)) ⊆ W_s(I). Proved in coordinates at the stalk of every closed point (RegularCoords.isMCInvariant_W), through the closed-point reduction and the stalk bridges.

theorem AlgebraicGeometry.Scheme.IdealSheafData.isDBalanced_W {k : Type u} [Field k] [CharZero k] {X : Scheme} (f : X ⟶ Spec ↧k) (n : ℕ) [SmoothOfRelativeDimension n f] (I : X.IdealSheafData) (m : ℕ) (hI : I.maxOrd = ↑m) (hm : 1 ≤ m) {r : ℕ} (hr : m - 1 ≤ r) :

[Kol07, Corollary 101; Proposition 99 (8)]: for m = max-ord I ≥ 1 and r ≥ m − 1, W_{r·lcm(2,…,m)}(I) is D-balanced with respect to r · lcm(2, …, m). Proved in coordinates at the stalk of every closed point (RegularCoords.isDBalanced_W).

theorem AlgebraicGeometry.Scheme.IdealSheafData.ord_W_of_isClosed {k : Type u} [Field k] [CharZero k] {X : Scheme} (f : X ⟶ Spec ↧k) (n : ℕ) [SmoothOfRelativeDimension n f] (I : X.IdealSheafData) (m : ℕ) (hI : I.maxOrd = ↑m) (hm : 1 ≤ m) (s : ℕ) {p : ↥X} (hp : IsClosed {p}) :
(W f I m s).ord p ≤ ↑s ∧ (I.ord p = ↑m → (W f I m s).ord p = ↑s)

The order of the tuned ideal at a closed point: at most s, and exactly s where I has order m (RegularCoords.ord_W; where ord I < m, W_s(I)_p = 𝒪_p by RegularCoords.W_eq_top_of_ord_lt).

theorem AlgebraicGeometry.Scheme.IdealSheafData.maxOrd_W {k : Type u} [Field k] [CharZero k] {X : Scheme} (f : X ⟶ Spec ↧k) (n : ℕ) [SmoothOfRelativeDimension n f] (I : X.IdealSheafData) (m : ℕ) (hI : I.maxOrd = ↑m) (hm : 1 ≤ m) (s : ℕ) :
(W f I m s).maxOrd = ↑s

For m = max-ord I ≥ 1, max-ord W_s(I) = s (the analogue of [Kol07, Theorem 54.2 (i)] for W_s). The bound at any point comes from a closed point it specializes to; the value is attained at a closed specialization of a point of order m, which exists because a supremum of naturals bounded by m and equal to m ≥ 1 is attained.

The role of E #

The role of E in the proof of [Kol07, Corollary 101]: a smooth blow-up sequence is of order ≥ m starting with (X, I, m, E) iff it is of order ≥ m starting with (X, I, m) (the empty family) and every centre has simple normal crossings with the total transform E_i. The direction → drops E to the sub-family of exceptional divisors (IsSubfamilyVia), the direction ← recombines.

The reduction to Theorem 100 and Corollary 101 #

The reduction of [Kol07, Corollary 101] to Theorem 100: for m ∣ s and s ≥ 1, Theorem 100 at the upper bound J = W_{ms'}(I) (tuning_iff_W) with s = m s'.

theorem Hironaka.Sequence.tuning_iff_E {X : AlgebraicGeometry.Scheme} {k : Type u} [Field k] (f : X ⟶ AlgebraicGeometry.Spec ↧k) [CharZero k] (n : ℕ) [AlgebraicGeometry.SmoothOfRelativeDimension n f] (S : X.BlowUpSequence) (I : X.IdealSheafData) (E : X.DivisorFamily) (m : ℕ) (hI : I.maxOrd = ↑m) (hm : 1 ≤ m) {r : ℕ} (_hr : m - 1 ≤ r) (hr1 : 1 ≤ r) :

Kollár's Corollary 101 [Kol07, Corollary 101], the equivalence for every divisor family: s = r · lcm(2, …, m) is divisible by m (dvd_Lcm) and, under the additional hypothesis r ≥ 1, positive, so the reduction to Theorem 100 applies. Kollár's hypothesis r ≥ m − 1 is part of his statement and is kept so that the corollary reads as printed (the D-balanced clause of corollary101 needs it, isDBalanced_W); the equivalence does not need it, because Theorem 100 asks only that s be positive and divisible by m. The hypothesis r ≥ 1 is not in Kollár's statement: it excludes only m = 1, r = 0, where s = 0, W_0(I) = 𝒪_X and the printed equivalence fails (every smooth blow-up sequence is of order ≥ 0 for (𝒪_X, 0)).

Kollár's Corollary 101 [Kol07, Corollary 101] (Tuning of ideals, II) as printed: W_s(I) is MC-invariant and D-balanced with respect to s = r · lcm(2, …, m), and a smooth blow-up sequence is of order ≥ m starting with (X, I, m, E) iff it is of order ≥ s starting with (X, W_s(I), s, E) — with the additional hypothesis r ≥ 1, which excludes only m = 1, r = 0 (there s = 0 and W_0(I) = 𝒪_X, and the equivalence fails: every smooth blow-up sequence is of order ≥ 0 for (𝒪_X, 0)). Kollár's hypothesis "E snc" is part of his statement — a triple of [Kol07, Notation 64] — and is kept so that the corollary reads as printed; the proof does not need it, because tuning changes only the ideal: the three clauses hold for every divisor family, the order-≥ m condition involving E only through the snc clauses on the centres, which are the same on both sides of the equivalence.