Documentation

Hironaka.Resolution.Algebraic.Hir64.Corollary1

Hironaka's Corollary 1: simultaneous trivialization of a finite system of ideal sheaves #

Corollary 1 of Main Theorem II [Hir64, Corollary 1, pp. 143–144]: given nonzero ideal sheaves J_1, …, J_e on a non-singular X, there is a finite succession of monoidal transformations with non-singular irreducible centers such that (1) every weak transform J_j(i) has constant positive order along the center D(i), and (2) at every point of the end result at least one J_j(r) is the unit ideal. Hironaka calls this the trivialization of a system of coherent sheaves of ideals.

The proof, following Hironaka's. Let J := ∏ J_j, T_0 := ⋂ |J_j| and, if T_0 ≠ ∅, d := max {ν(J_x) : x ∈ T_0} (finite since J is nonzero everywhere on a Noetherian scheme). Hironaka's loci are T := {ν(J) ≥ d}, S := T ∖ T_0 and T ∖ S = T ∩ T_0. The last is closed (trivializationLoci_closed); S is closed too, Hironaka's "we can prove that T, S and T − S are all closed", proved here by upper semicontinuity and additivity of the orders (isClosed_trivialLocus): around a point q ∈ T ∖ S, where ν(J_q) = d and every ν((J_j)_q) ≥ 1, the open set {ν(J) ≤ d} ∩ ⋂_j {ν(J_j) ≤ ν((J_j)_q)} misses S, since a point of S in it would have some ν(J_{j_0}) = 0 and hence ν(J) ≤ d − 1; and {ν(J) < d} misses S outright. Main Theorem II (one round of order reduction, exists_orderReductionRound) applied to (X̄ := X ∖ S, J|_{X̄}, ∅), whose max-ord is exactly d, gives a smooth blow-up sequence with regular irreducible centers of order d whose last weak transform has max-ord < d; by clause (ii) of Main Theorem II the centers lie over T ∖ S. The sequence extends across S (Hironaka/Resolution/Algebraic/Hir64/ExtendOpen.lean), the extension S₁ on X having S₁.pullback u = S̄ and centers in the ranges of the stage lifts, along which regularity, irreducibility, smoothness and the weak transforms transport (Hironaka/Scheme/BlowUpSequence/OpenTransport.lean). Along S̄ the weak transform of the product is the product of the weak transforms and each factor has constant order on each center; the constants are positive because the order of a weak transform at a point over the center never exceeds the order at the image (Hironaka's (B*) [Hir64, p. 143]; [Kol07, Lemma 61] read pointwise along a center of constant order, ord_weakTransform_le_ord_of_forall_ord_eq), the image lies in T ∖ S where every factor has order ≥ 1, and the orders add up to d on both sides. Over S the transforms are unchanged (ord_weakTransformSeq_of_stageMap_notMem), so on the end result the factors are simultaneously nontrivial only over X̄, where the product has order < d. Induction on d and concatenation (concat, with the per-stage clauses of Hironaka/Scheme/BlowUpSequence/ConcatClauses.lean) finish; at T_0 = ∅ the empty sequence does.

ℕ∞ bookkeeping #

theorem Hironaka.Sequence.sum_add_one_le_sum {ι : Type u_1} [Fintype ι] (a b : ι → ℕ∞) (hab : ∀ (i : ι), a i ≤ b i) (i₀ : ι) (ha : a i₀ = 0) (hb : 1 ≤ b i₀) :
∑ i : ι, a i + 1 ≤ ∑ i : ι, b i

ℕ∞ bookkeeping for the two "some factor is trivial" arguments: termwise a ≤ b with one a_{i₀} = 0 < 1 ≤ b_{i₀} gives ∑ a + 1 ≤ ∑ b.

Kollár's Lemma 61 pointwise along a center of constant order #

Hironaka's (B*) [Hir64, p. 143]; [Kol07, Lemma 61] read pointwise: along a smooth center Z on which I has constant order m at every point, the weak transform of I has order at most m = ord_{π q} I at every point q over the center. This is the argument of ord_controlledTransform_le_of_smooth with the global hypothesis max-ord I = m replaced by the constancy on Z (the only use of max-ord there was ord_{π q} I ≤ m). Off the center the blow-up is an isomorphism and the orders agree. Used for the positivity in condition (1) of [Hir64, Corollary 1].

Orders along a smooth sequence: monotonicity, the unchanged region, the product #

theorem Hironaka.Sequence.ord_weakTransformSeq_le_ord_stageMap {k : Type u} [Field k] [CharZero k] {X : AlgebraicGeometry.Scheme} (f : X ⟶ AlgebraicGeometry.Spec ↧k) (n : ℕ) [AlgebraicGeometry.SmoothOfRelativeDimension n f] (S : X.BlowUpSequence) :
S.IsSmooth f → ∀ (J : X.IdealSheafData), (∀ (j : ℕ) (hj : j < S.length), ∃ (c : ℕ), ∀ y ∈ (S.center ⟨j, hj⟩).support, (S.weakTransformSeq J ⟨j, ⋯⟩).ord y = ↑c) → ∀ (i : Fin (S.length + 1)) (z : ↥(S.stage i)), (S.weakTransformSeq J i).ord z ≤ J.ord ((S.stageMap i) z)

Along a smooth sequence whose induced ideals have constant order on every center, the order of the i-th induced ideal at any point is at most the order of the original ideal at its image; ord_weakTransform_le_ord_of_forall_ord_eq stage by stage.

theorem Hironaka.Sequence.ord_weakTransformSeq_of_stageMap_notMem {X : AlgebraicGeometry.Scheme} (S : X.BlowUpSequence) (J : X.IdealSheafData) (C : Set ↥X) :
(∀ (i : Fin S.length), ∀ y ∈ (S.center i).support, (S.stageMap i.castSucc) y ∈ C) → ∀ (i : Fin (S.length + 1)) (z : ↥(S.stage i)), (S.stageMap i) z ∉ C → (S.weakTransformSeq J i).ord z = J.ord ((S.stageMap i) z)

Over a point outside a closed set C containing the images of all the centers, the order of every induced ideal is the order of the original: each blow-up is an isomorphism there (ord_weakTransform_of_notMem). In the proof of [Hir64, Corollary 1] this is the observation that over S the transforms are unchanged.

theorem Hironaka.Sequence.weakTransformSeq_prod_of_forall_ord_eq {k : Type u} [Field k] [CharZero k] {X : AlgebraicGeometry.Scheme} (f : X ⟶ AlgebraicGeometry.Spec ↧k) (n : ℕ) [AlgebraicGeometry.SmoothOfRelativeDimension n f] (S : X.BlowUpSequence) :
S.IsSmooth f → (∀ (i : Fin S.length), IrreducibleSpace ↥(S.center i).subscheme) → ∀ {e : ℕ} (J : Fin e → X.IdealSheafData) (d : ℕ), (∀ (j : ℕ) (hj : j < S.length), ∀ y ∈ (S.center ⟨j, hj⟩).support, (S.weakTransformSeq (∏ l : Fin e, J l) ⟨j, ⋯⟩).ord y = ↑d) → ∀ (i : Fin (S.length + 1)), S.weakTransformSeq (∏ l : Fin e, J l) i = ∏ l : Fin e, S.weakTransformSeq (J l) i

Along a smooth sequence with irreducible centers on which the induced ideals of the product ∏ J_l have constant order d, the induced ideal of the product at every stage is the product of the induced ideals of the factors; weakTransform_prod_of_smooth stage by stage, the constancy of each factor's order on the center being ord_factors_const_on_center_of_smooth.

theorem Hironaka.Sequence.exists_ord_weakTransformSeq_eq_on_center {k : Type u} [Field k] [CharZero k] {X : AlgebraicGeometry.Scheme} (f : X ⟶ AlgebraicGeometry.Spec ↧k) (n : ℕ) [AlgebraicGeometry.SmoothOfRelativeDimension n f] (S : X.BlowUpSequence) (hsm : S.IsSmooth f) (hirr : ∀ (i : Fin S.length), IrreducibleSpace ↥(S.center i).subscheme) {e : ℕ} (J : Fin e → X.IdealSheafData) (d : ℕ) (hcen : ∀ (j : ℕ) (hj : j < S.length), ∀ y ∈ (S.center ⟨j, hj⟩).support, (S.weakTransformSeq (∏ l : Fin e, J l) ⟨j, ⋯⟩).ord y = ↑d) (j : ℕ) (hj : j < S.length) (l : Fin e) :
∃ (c : ℕ), ∀ y ∈ (S.center ⟨j, hj⟩).support, (S.weakTransformSeq (J l) ⟨j, ⋯⟩).ord y = ↑c

Condition (1) of [Hir64, Corollary 1] without positivity: along a smooth sequence with irreducible centers on which the induced ideals of ∏ J_l have constant order d, every factor's induced ideal has constant order on every center.

Along a smooth sequence the induced ideals of an ideal nonzero on every component are nonzero on every component.

A finite product of ideal sheaves nonzero on every component is nonzero on every component of a smooth k-scheme (the orders add and are finite).

theorem Hironaka.Sequence.isClosed_trivialLocus {k : Type u} [Field k] [CharZero k] {X : AlgebraicGeometry.Scheme} (f : X ⟶ AlgebraicGeometry.Spec ↧k) (n : ℕ) [AlgebraicGeometry.SmoothOfRelativeDimension n f] {e : ℕ} (J : Fin e → X.IdealSheafData) (d : ℕ) (hd : ∀ (x : ↥X), (∀ (j : Fin e), x ∈ (J j).support) → (∏ j : Fin e, J j).ord x ≤ ↑d) :
IsClosed {x : ↥X | ↑d ≤ (∏ j : Fin e, J j).ord x ∧ ∃ (j : Fin e), x ∉ (J j).support}

"We can prove that T, S and T − S are all closed" [Hir64, Corollary 1, proof], the case of S: in Hironaka's notation T := {ν(J) ≥ d}, the set S := T ∖ ⋂_j |J_j| is closed when d bounds ν(J) on ⋂_j |J_j|. The complement is open: around a point of order < d by upper semicontinuity, and around a point q of order ≥ d with all J_j nontrivial (so ν(J_q) = d) the open set {ν(J) ≤ d} ∩ ⋂_j {ν(J_j) ≤ ν((J_j)_q)} misses S, a point of S in it having some ν(J_{j₀}) = 0 and hence ν(J) ≤ d − 1.

One round of Hironaka's argument #

theorem Hironaka.Sequence.exists_corollary1_seq_of_forall_exists_notMem (X : AlgebraicGeometry.Scheme) {e : ℕ} (J : Fin e → X.IdealSheafData) (h : ∀ (x : ↥X), ∃ (j : Fin e), x ∉ (J j).support) :
∃ (S : X.BlowUpSequence), (∀ (i : Fin S.length), AlgebraicGeometry.IsRegular (S.center i).subscheme ∧ IrreducibleSpace ↥(S.center i).subscheme) ∧ (∀ (i : Fin S.length) (j : Fin e), ∃ (c : ℕ), 0 < c ∧ ∀ y ∈ (S.center i).support, (S.weakTransformSeq (J j) i.castSucc).ord y = ↑c) ∧ ∀ (x : ↥S.last), ∃ (j : Fin e), (S.weakTransformSeq (J j) (Fin.last S.length)).stalkIdeal x = ⊤

The trivial case of [Hir64, Corollary 1] ("if otherwise, the assertion is trivial"): when at every point some J_j is the unit ideal, the empty sequence does.

theorem Hironaka.Sequence.exists_round {k : Type u} [Field k] [CharZero k] (X : AlgebraicGeometry.Scheme) [X.Over (AlgebraicGeometry.Spec ↧k)] [AlgebraicGeometry.QuasiCompact (X ↘ AlgebraicGeometry.Spec ↧k)] [AlgebraicGeometry.IsSeparated (X ↘ AlgebraicGeometry.Spec ↧k)] (N : ℕ) [AlgebraicGeometry.SmoothOfRelativeDimension N (X ↘ AlgebraicGeometry.Spec ↧k)] {e : ℕ} (J : Fin e → X.IdealSheafData) (hJ : ∀ (j : Fin e), (J j).IsNonzeroEverywhere) (hne : {x : ↥X | ∀ (j : Fin e), x ∈ (J j).support}.Nonempty) (d : ℕ) (hd0 : 0 < d) (hmax : (∏ j : Fin e, J j).maxOrdAlong {x : ↥X | ∀ (j : Fin e), x ∈ (J j).support} = ↑d) :
∃ (S : X.BlowUpSequence), (∀ (i : Fin S.length), AlgebraicGeometry.IsRegular (S.center i).subscheme ∧ IrreducibleSpace ↥(S.center i).subscheme) ∧ (∀ (i : Fin S.length) (j : Fin e), ∃ (c : ℕ), 0 < c ∧ ∀ y ∈ (S.center i).support, (S.weakTransformSeq (J j) i.castSucc).ord y = ↑c) ∧ S.IsSmooth (X ↘ AlgebraicGeometry.Spec ↧k) ∧ ∀ (x : ↥S.last), (∀ (j : Fin e), x ∈ (S.weakTransformSeq (J j) (Fin.last S.length)).support) → (∏ j : Fin e, S.weakTransformSeq (J j) (Fin.last S.length)).ord x < ↑d

One round of Hironaka's argument for [Hir64, Corollary 1], as described in the module docstring, at the maximal order d ≥ 1 of ∏ J_j on ⋂_j |J_j| ≠ ∅: a smooth sequence on X with regular irreducible centers, every factor of constant positive order on every center (condition (1)), and ν(∏_j J_j(r)) < d at every point of the end result where all the J_j(r) are nontrivial.

Induction on d and assembly #

theorem Hironaka.Sequence.exists_corollary1_seq_of_maxOrdAlong_le {k : Type u} [Field k] [CharZero k] (d : ℕ) (X : AlgebraicGeometry.Scheme) [X.Over (AlgebraicGeometry.Spec ↧k)] [AlgebraicGeometry.QuasiCompact (X ↘ AlgebraicGeometry.Spec ↧k)] [AlgebraicGeometry.IsSeparated (X ↘ AlgebraicGeometry.Spec ↧k)] (N : ℕ) [AlgebraicGeometry.SmoothOfRelativeDimension N (X ↘ AlgebraicGeometry.Spec ↧k)] {e : ℕ} (_he : 0 < e) (J : Fin e → X.IdealSheafData) (_hJ : ∀ (j : Fin e), (J j).IsNonzeroEverywhere) :
(∏ j : Fin e, J j).maxOrdAlong {x : ↥X | ∀ (j : Fin e), x ∈ (J j).support} ≤ ↑d → ∃ (S : X.BlowUpSequence), (∀ (i : Fin S.length), AlgebraicGeometry.IsRegular (S.center i).subscheme ∧ IrreducibleSpace ↥(S.center i).subscheme) ∧ (∀ (i : Fin S.length) (j : Fin e), ∃ (c : ℕ), 0 < c ∧ ∀ y ∈ (S.center i).support, (S.weakTransformSeq (J j) i.castSucc).ord y = ↑c) ∧ ∀ (x : ↥S.last), ∃ (j : Fin e), (S.weakTransformSeq (J j) (Fin.last S.length)).stalkIdeal x = ⊤

[Hir64, Corollary 1] for every system whose product has order at most d on the common support ("either the integer d gets reduced, or (2) is achieved"): induction on d, one round (exists_round) followed by the induction hypothesis at the end result, the two sequences concatenated.

theorem Hironaka.Sequence.hironaka_corollary1 {k : Type u} [Field k] [CharZero k] (X : AlgebraicGeometry.Scheme) [X.Over (AlgebraicGeometry.Spec ↧k)] [AlgebraicGeometry.QuasiCompact (X ↘ AlgebraicGeometry.Spec ↧k)] [AlgebraicGeometry.IsSeparated (X ↘ AlgebraicGeometry.Spec ↧k)] (hN : ∃ (N : ℕ), AlgebraicGeometry.SmoothOfRelativeDimension N (X ↘ AlgebraicGeometry.Spec ↧k)) {e : ℕ} (he : 0 < e) (J : Fin e → X.IdealSheafData) (hJ : ∀ (j : Fin e), (J j).IsNonzeroEverywhere) :
∃ (S : X.BlowUpSequence), (∀ (i : Fin S.length), AlgebraicGeometry.IsRegular (S.center i).subscheme ∧ IrreducibleSpace ↥(S.center i).subscheme) ∧ (∀ (i : Fin S.length) (j : Fin e), ∃ (c : ℕ), 0 < c ∧ ∀ y ∈ (S.center i).support, (S.weakTransformSeq (J j) i.castSucc).ord y = ↑c) ∧ ∀ (x : ↥S.last), ∃ (j : Fin e), (S.weakTransformSeq (J j) (Fin.last S.length)).stalkIdeal x = ⊤

Hironaka's Corollary 1 [Hir64, Corollary 1, pp. 143–144], the trivialization of a system of coherent sheaves of ideals: a finite succession of monoidal transformations with non-singular irreducible centers such that (1) every weak transform J_j(i) has constant positive order along the center D(i), and (2) at every point of the end result at least one J_j(r) is the unit ideal. This is exists_corollary1_seq_of_maxOrdAlong_le at the (finite) maximal order of ∏ J_j on ⋂_j |J_j|. Hironaka's non-singularity of X is not a hypothesis: the proof runs the order reduction of Main Theorem II on X, which is smooth of a single relative dimension over the field k of characteristic zero (hN; the restriction is lifted in ComponentwiseCorollaries), and smoothness over the perfect field k gives the regularity of the local rings.