Alphabeta Math
Pipeline-generated
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

Higher-Dimensional Resolution of Singularities

1 · Prerequisites

2 · Summary

This page develops the characteristic-zero resolution of singularities after Hironaka in the form given by Włodarczyk's marked-ideal algorithm, over Conventions for the resolution development. The construction is carried out for marked ideals (I,E,μ), whose order function Order of an ideal sheaf at a point and simultaneous normal-crossings divisors Simple normal crossings divisors and simultaneous normal crossings position provide the geometry of the supports. Multiple test blow-ups, their controlled transforms and the equivalence relation on marked ideals are set up in Multiple test blow-ups, controlled transforms and resolutions of marked ideals and Equivalence of marked ideals, with the well-definedness of the transform calculus in Controlled transforms are well defined.

The core of the page is the calculus of derivative ideals Derivative ideals of an ideal sheaf and of a marked ideal, maximal order and tangent directions Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors, homogenized ideals The homogenized ideal of a marked ideal of maximal order and coefficient ideals The coefficient ideal of a marked ideal of maximal order, together with the glueing lemma Glueing of homogenized ideals along etale neighbourhoods that makes the choice of a hypersurface of maximal contact irrelevant. These tools feed the canonical resolution of marked ideals, Canonical resolution of marked ideals, the engine of the subject.

From the engine the page derives the main theorems in their classical form: principalization of ideals Canonical principalization of ideals in characteristic zero, weak embedded desingularization Weak embedded desingularization in characteristic zero with the Bravo–Villamayor full-transform strengthening Bravo-Villamayor strengthening of embedded desingularization, embedding independence Independence of the embedded desingularization from the ambient embedding, open restriction Open restrictions of the canonical desingularization, and resolution of singularities in characteristic zero Resolution of singularities in characteristic zero with its functoriality under smooth morphisms Resolution of singularities is functorial under smooth morphisms. The worked blowup computations of the companion page illustrate the theory on the quadric cone (higher-dimensional-resolution-of-singularities-examples).

3 · Logical flowchart

4 · Definitions, theorems and proofs

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

Order of an ideal sheaf at a point

Definition

Let X be a locally Noetherian scheme, I⊆OX a coherent ideal sheaf (Coherent module sheaves, Ideal sheaves) and x∈X with maximal ideal mx⊂OX,x (A local ring is a nonzero commutative ring with a unique maximal ideal). The order of I at x is ord⁡x(I):=max⁡{ n≥0:Ix⊆mxn }, with ord⁡x(I):=+∞ when Ix=0; the maximum is attained for Ix≠0 because ⋂nmxn=0 in the Noetherian local ring OX,x (The Krull intersection is the (1−a)-torsion submodule, and it vanishes in the Jacobson-radical case). For f∈OX(U) put ord⁡x(f):=ord⁡x(fOU); a germ is of multiplicity one at x if ord⁡x(f)=1. If in addition OX,x is regular — the only case used on this page — then f is part of a regular system of parameters of OX,x, so the zero scheme V(f) is regular at x of dimension dim⁡OX,x−1 (regular local quotient by parameter is regular, under AC The Axiom of Choice). The parameter assertion follows by extending the nonzero class of f in the finite-dimensional cotangent space to a basis and applying the finite-generator Nakayama argument. If x∈V(I) and s=dim⁡OX,x, then ord⁡x(I)≥1; ord⁡x is the multiplicity used throughout this page, and it differs from the Hilbert-Samuel multiplicity, which is not used here.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

Simple normal crossings divisors and simultaneous normal crossings position

Definition

Let X be a regular locally Noetherian scheme of pure dimension n (embedding dimension and regular local ring). A reduced effective Cartier divisor D⊆X (Effective cartier divisor, Cartier divisor) is a strict normal crossings divisor (an SNC divisor) if for every point p∈X, writing dp=dim⁡OX,p, there is a regular system of parameters u1,…,udp of OX,p and a subset J⊆{1,…,dp} such that the ideal of D in OX,p is generated by ∏j∈Juj; equivalently, every irreducible component of D is regular and the local equations of the components through p extend to a regular system of parameters. A finite family E={D1,…,Dm} of reduced effective Cartier divisors on X is in simultaneous SNC position if for every subset J⊆{1,…,m} the union of components of the divisors in J is an SNC divisor; equivalently, at every point p the union of all components of members of E passing through p is cut out by a subset of a regular system of parameters of OX,p. For the marked-ideal algorithm on this page, the irreducible components within each individual member of E must be pairwise disjoint, and no irreducible component is repeated in different members. Thus at a point each member contributes at most one distinct parameter equation. This is the additional boundary convention of Włodarczyk, Definition 2.1.1. Such a family E always carries a fixed total order, and the members of E play the role of the exceptional divisors of a multiple test blow-up.

LemmaStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Étale maps induce completion isomorphisms at equal-residue points

Statement

Let φ ⁣:X′→X be a morphism of locally Noetherian schemes that is étale at x′∈X′ (Étale morphism of schemes), with x=φ(x′), and let mx⊂OX,x and mx′⊂OX′,x′ be the maximal ideals (A local ring is a nonzero commutative ring with a unique maximal ideal). Then:

(1) φ is flat at x′ (Smooth morphism of schemes);

(2) mxOX′,x′=mx′;

(3) for every ideal I⊆OX,x and every n≥0, I⊆mx n  ⟺  IOX′,x′⊆mx′ n.

If the induced residue-field map κ(x)→κ(x′) is an isomorphism, then the induced map of maximal-adic completions O^X,x⟶O^X′,x′ is an isomorphism. This completion conclusion requires the residue-field hypothesis.

Facts & Assumptions

Given: A morphism φ ⁣:X′→X of locally Noetherian schemes, étale at x′∈X′ with x=φ(x′), and the maximal ideals mx⊆OX,x and mx′⊆OX′,x′.

[F1]

Étale morphism of schemes, Smooth morphism of schemes, Relative dimension of a smooth morphism at a point, Flat morphism of schemes: étaleness at x′ makes φ smooth and of relative dimension 0, hence flat, and its geometric fibre has local dimension zero.

[F2]

Stalks of the scheme-theoretic fibre, embedding dimension and regular local ring, Locally Noetherian and Noetherian schemes: the local ring of the fibre is OX′,x′/mxOX′,x′; it is a zero-dimensional regular Noetherian local ring and therefore a field. Indeed its maximal ideal n has n/n2=0; local Noetherianity makes n finitely generated, and the relation n=n2 gives a matrix M with entries in n such that I−M annihilates the generators. Since det⁡(I−M) is a unit, those generators vanish.

[F3]
[F4]

The I-adic completion of a module, Completion of a Noetherian local ring is local with the same residue field: the maximal-adic completion of a Noetherian local ring is the inverse limit of its quotients by powers of its maximal ideal.

Proof

1.1F1F2F3

Flatness and the fibre local ring. By [F1], φ is flat at x′. Put s=φ(x′). The local ring of the fibre at x′ is OX′,x′/msOX′,x′ by [F2]; it is regular because the geometric fibre is regular and has dimension zero by [F1], so it is a field by [F2]. Therefore mxOX′,x′=msOX′,x′ is a maximal ideal of OX′,x′, hence equals its unique maximal ideal mx′ by [F3]. This proves (1) and (2).

2.1step 1.1

Forward filtration detection. If I⊆mx n, extension of ideals and step 1.1 give IOX′,x′⊆mx nOX′,x′=(mxOX′,x′)n=mx′ n. Thus the forward implication in (3) holds.

2.2step 1.1

Reverse filtration detection. Put A=OX,x, B=OX′,x′, m=mx, and n=mx′. Assume IB⊆nN. If some a∈I were not in mN, choose the largest k<N with a∈mk. Its class in the finite-dimensional κ(x)-vector space mk/mk+1 is nonzero. Flatness in step 1.1 identifies (mk/mk+1)⊗κ(x)κ(x′) with nk/nk+1; extension of scalars along a field extension is injective on a finite-dimensional vector space, so the class of a remains nonzero there. But a∈IB⊆nN⊆nk+1, a contradiction. Thus every a∈I lies in mN, proving the reverse implication in (3).

3.1F1F4step 1.1∎

Completion when residue fields agree. Assume κ(x)→κ(x′) is an isomorphism. By step 1.1, mxOX′,x′=mx′, so the induced map on residue fields and the degree-zero associated graded pieces is an isomorphism. For every k≥0, flatness from [F1] identifies mxk/mxk+1⊗κ(x)κ(x′)≅mx′k/mx′k+1; since the residue fields agree, the map of associated graded pieces is an isomorphism in every degree. The exact sequences 0→mxq−1/mxq→OX,x/mxq→OX,x/mxq−1→0 and their analogues for OX′,x′ show by induction on q that the induced map on each finite quotient by the qth power is an isomorphism. Taking inverse limits using [F4] proves the asserted isomorphism of completions.

Remarks

  • The residue-field condition in the completion clause cannot be omitted. For a nontrivial finite separable extension L/K, the morphism Spec⁡L→Spec⁡K is étale (Finite field extensions and etaleness), while the completed local rings are L and K, and the induced completion map is the proper inclusion K↪L, hence is not an isomorphism. No assertion about abstract isomorphism of the two fields is needed. Włodarczyk's phrase “formal analytic isomorphism” in the proof of Lemma 2.4.1 is valid in its algebraically closed closed-point setting; the order argument for arbitrary étale points needs only (1)–(3).
  • Flatness and the associated-graded argument prove the filtration and equal-residue completion claims without a choice principle.
LemmaStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Extending an étale morphism to a smooth ambient neighbourhood

Statement

Assume the Axiom of Choice. Let K be a field of characteristic zero and let ιU ⁣:U↪AKm and ιV ⁣:V↪AKn be closed immersions of affine K-varieties of finite type (Closed immersions of schemes, Integral schemes). Let φ ⁣:U→V be étale at a K-rational point u0, and put v0=φ(u0). Choose affine coordinates on the two ambient spaces carrying u0 and v0 to their origins.

Then there is a closed subscheme X⊆AKn×KAKm containing the graph copy of U, smooth at (v0,u0), such that the projection Φ:=pr⁡1∣X ⁣:X→AKn is étale at (v0,u0) and restricts on U to ιV∘φ. Thus φ extends to a morphism of smooth ambient varieties étale at the chosen point.

After extending the ground field so the point is rational, the usual local factorization of a smooth morphism as an étale morphism followed by a projection gives the corresponding smooth ambient extension locally. This is the reduction used in the source's proof of functoriality for smooth morphisms.

Facts & Assumptions

Given: The Axiom of Choice; a characteristic-zero field K; closed immersions U↪AKm and V↪AKn of affine varieties of finite type; and a morphism φ ⁣:U→V étale at the K-rational point u0, with image v0.

[F1]

Étale maps induce completion isomorphisms at equal-residue points: since u0 and v0 are K-rational, their residue fields are both K, and the étale local map induces an isomorphism of completed local rings.

[F2]

Closed immersions of schemes: the graph map U→AKn×KAKm, u↦(ιV(φ(u)),ιU(u)), is a closed immersion; its ideal is finitely generated and contains the equations of the target V.

[F3]

Standard smooth presentations and locally standard smooth maps, Locally standard smooth iff flat with geometrically regular fibres, Étale morphism of schemes: if m equations in the y-coordinates have invertible m×m Jacobian at the point, their zero scheme is standard smooth of relative dimension 0 over AKn, hence its projection to AKn is étale there.

[F4]

Flat maps with geometrically regular fibres have standard smooth local presentations, Relative dimension of a smooth morphism at a point: a smooth germ of relative dimension r has a standard smooth local presentation with r free parameters; separating these parameters from the equations factors it locally as an étale germ to the product with Ar followed by projection.

[F5]

Jacobson-adic completion is faithfully flat: the maximal-ideal completion of a Noetherian local ring is faithfully flat, so equality of finite defining ideals can be checked after completion.

Proof

1.1F1F2

Embed the source by its graph. Use coordinates x1,…,xn on AKn and y1,…,ym on AKm, centered at v0 and u0. Write the equations of V as f1(x),…,fl(x) and choose further equations h1(x,y),…,hs(x,y) for the graph copy of U. Put S=O^V,v0. By [F1] the induced map S→O^U,u0 is an isomorphism. In the completed local ring of V×KAKm, identified with S⟦y1,…,ym⟧, the graph map therefore sends each yj to some gj∈S with zero residue, and its ideal is generated by y1−g1,…,ym−gm. The images of the hj generate this graph ideal, so their classes span its conormal space modulo the maximal-ideal multiple. The classes of yj−gj are a basis of that m-dimensional K-vector space. Hence some m of the hj have coefficient matrix on these generators invertible at the point; since each gj has zero residue, this matrix is their y-Jacobian there.

2.1F2step 1.1

Choose equations for the ambient extension. Select m of the equations hj, say hi1,…,him, whose y-linear parts form a basis of that vector space, and define X:=V(hi1,…,him)⊆AKn×KAKm. Every point of the graph copy of U satisfies these equations, so U is a closed subscheme of X. The determinant of the matrix (∂hia/∂yb)a,b=1m is nonzero at (v0,u0).

3.1F3F5step 2.1

The projection is étale. The presentation of X from step 2.1 has m variables y1,…,ym and m equations with invertible Jacobian minor, so it is standard smooth of relative dimension 0 over AKn at (v0,u0). By [F3], Φ ⁣:X→AKn is étale there. Consequently X is smooth at (v0,u0), and the projection restricts on U to ιV∘φ by construction. Moreover the germ of U equals that of X×AKnV: in S⟦y1,…,ym⟧ the selected equations generate the graph ideal, because their coefficient matrix on yj−gj is invertible in the complete local ring. Thus the two defining ideals agree after completion, and [F5] makes them agree in the local ring itself. Their quotient is a finite module, so it vanishes on a neighbourhood of the point. Shrinking X there therefore makes this square Cartesian.

4.1F4step 3.1∎

Smooth morphisms. Let W→S be smooth at a geometric point. By [F4], after a local standard smooth chart it factors as an étale germ W→S×Ar followed by projection. After extending the ground field so the point is rational, embed the two affine germs in affine spaces and apply steps 1.1–3.1 to the étale germ; composing the resulting ambient étale map with the projection gives a smooth ambient extension of the original germ.

Remarks

  • The source proof uses coordinates on An×Am, hence the ambient dimension is n+m. Its printed line saying X⊂Am is inconsistent with those coordinates and with its projection Am+n→An; the statement above follows the coordinate proof.
  • The ambient extension is proved for the source's affine-embedding and rational-point setting. The smooth-morphism consequence is used after passage to a geometric point; no claim is made here that an arbitrary non-rational point over the original field is rational.
RemarkRemark: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

Conventions for the resolution development

Remark

This page follows the source conventions of Field and A field's prime subfield is isomorphic to Q in characteristic zero and to Fp in characteristic p: K is a field of characteristic zero, i.e. its prime subfield is Q. A K-variety is an integral, separated scheme of finite type over K (Integral schemes, Separated morphism of schemes, Locally finite type and finite type morphisms); a smooth K-scheme is a scheme of finite type over K whose structure morphism is smooth (Smooth morphism of schemes); in characteristic zero a scheme of finite type over K is smooth over K exactly when all its local rings are regular, by the perfect-field geometric-regularity criterion (Field tests for geometric regularity, Locally standard smooth iff flat with geometrically regular fibres). The core resolution construction is over an algebraically closed field of characteristic zero, as in the source. Items with an explicitly broader field or characteristic range retain their stated hypotheses; the descent item and final theorems apply over arbitrary characteristic-zero fields. The positive-characteristic derivative converses require the perfect-field qualifications stated in their items. All blowups are blowups of regular closed subschemes of smooth K-schemes, and all divisors are effective Cartier divisors; SNC' abbreviates simple normal crossings' in the sense of Simple normal crossings divisors and simultaneous normal crossings position. Resolutions are stated for reduced or integral base schemes and are constructed from an ambient smooth scheme. The Axiom of Choice is assumed throughout this resolution development and is inherited from the published blowup and relative-Proj suppliers (The Axiom of Choice); no dependent-choice or other choice principle is used by this page's new arguments beyond what those suppliers already assume.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Order and simultaneous normal crossings are preserved by smooth morphisms

Statement

Assume AC (The Axiom of Choice), as inherited from the regular-local algebra suppliers.

Let φ ⁣:X′→X be a smooth morphism of smooth K-schemes (Smooth morphism of schemes).

(1) For every coherent ideal sheaf I⊆OX (Coherent module sheaves) and every x′∈X′ with x=φ(x′), ord⁡x′(φ∗I)=ord⁡x(I), where the order is that of Order of an ideal sheaf at a point.

(2) Assume in addition that X′ is of pure dimension, as required by the SNC interface. If E is a family of divisors in simultaneous SNC position on X (Simple normal crossings divisors and simultaneous normal crossings position) then the family φ−1(E) of scheme-theoretic inverse images of its members is a family of divisors in simultaneous SNC position on X′; the individual inverse images are reduced effective Cartier divisors (Effective cartier divisor).

This is the source's Lemma 2.4.1.

Facts & Assumptions

Given: A smooth morphism φ:X′→X of smooth K-schemes, an arbitrary point x′∈X′ with image x, and a coherent ideal sheaf I or a simultaneous SNC family on X. Assume the Axiom of Choice inherited from the regular-local algebra suppliers.

[A1]

The Axiom of Choice: AC is inherited from the regular-local parameter, regular-sequence, and associated-graded suppliers below; no residue-field equality is assumed.

[F1]

Smooth morphism of schemes and A local ring is a nonzero commutative ring with a unique maximal ideal: the induced map of Noetherian local rings A=OX,x→B=OX′,x′ is flat and local, and its fibre local ring B/mAB is regular. Both A and B are regular, since X and X′ are smooth over K.

[F2]

embedding dimension and regular local ring and regular system of parameters: a regular local ring of dimension d has a minimal maximal-ideal generating tuple of length d, called a regular system of parameters; its classes are a basis of the cotangent space.

[F3]

regular local rings are domains and cohen macaulay: a regular local ring is a domain and every regular system of parameters is a regular sequence.

[F4]

Localisation And Faithfully Flat Base Change Of Regular Sequences: a regular sequence remains regular after faithfully flat base change.

[F5]

quotient and lifting regularity across a regular element: if z is a nonzerodivisor in the maximal ideal of a Noetherian local ring R and R/(z) is regular, then R is regular and z∉mR2. In a regular local ring, quotienting by a parameter gives a regular local ring.

[F6]

associated graded ring of a regular local ring: a cotangent basis in a regular local ring identifies its maximal-adic associated graded ring with the polynomial algebra on the classes of that basis.

[F7]

Order of an ideal sheaf at a point and Coherent module sheaves: the order of an ideal stalk I⊆A is the supremum of the integers N≥0 with I⊆mAN, with value +∞ for the zero ideal.

[F8]

Simple normal crossings divisors and simultaneous normal crossings position and Effective cartier divisor: an SNC divisor is locally the reduced product of a subset of a regular system of parameters, and simultaneous SNC requires this for the union of every subfamily. A nonzerodivisor equation gives an effective Cartier divisor; a unit equation gives the empty effective divisor.

Proof

1.1F1given

Work at an arbitrary x′ and its image x, with A,B as in [F1], maximal ideals m,n, and residue fields κ=A/m, λ=B/n. The map A→B is faithfully flat: for any proper ideal J⊆A, locality gives JB⊆n, so B/JB≠0; any nonzero A-module contains a nonzero cyclic submodule A/J, whose injection remains injective after flat tensoring, so its tensor with B is nonzero. The closed-fibre local ring B/mB is regular by smoothness. This argument applies to nonclosed and nonrational points as well.

2.1F2F3F4F5step 1.1choose

Choose a regular system of parameters u1,…,ud of A. Its image in B is a regular sequence by [F3, F4]. Put Bi=B/(u1,…,ui)B; the terminal ring Bd=B/mB is regular. Backward induction using [F5] shows that every Bi is regular and that the class of ui+1 is outside the square of its maximal ideal. Thus ui+1∉(u1,…,ui)B+n2, and the images of all ui are linearly independent in n/n2. Extend them to a cotangent basis; its lifts generate n by the finite-generator Nakayama argument (if the quotient module equals its maximal-ideal multiple, a matrix I−M with unit determinant annihilates its generators). They therefore form a regular system of parameters of B. When d=0 the tuple is empty and the same conclusion is immediate.

3.1F6step 2.1algebra

By [F6], the parameter systems in step 2.1 identify gr⁡mA with κ[U1,…,Ud] and gr⁡nB with λ[U1,…,Ud,V1,…,Ve]. The induced graded map sends each Ui to the corresponding parameter class and extends the residue-field embedding κ→λ; it is therefore injective. In particular an element of mj∖mj+1 remains outside nj+1.

3.2F8step 2.1

At a point x on an SNC union, choose its distinct component equations u1,…,uc as part of a regular system of parameters of A, as in [F8]. Step 2.1, applied with that system, shows that their images are part of a regular system of parameters of B. Hence the inverse image union is locally the product of those same distinct parameter equations, so it has SNC at x′. This applies to every subfamily and every point over it, independently of its residue field or its position in the fibre, proving simultaneous SNC. Where no component occurs, the equation is a unit and its inverse image is empty.

4.1F1F7step 3.1algebra

For an ideal I⊆A and N≥0, the inclusion I⊆mN implies IB⊆nN because the map is local. Conversely, if IB⊆nN and a∈I∖mN, choose the largest j<N for which a∈mj; step 3.1 gives a∉nj+1, contradicting a∈IB⊆nN. Thus I⊆mN  ⟺  IB⊆nN for every N, and taking suprema proves (1), including unit ideals of order zero and zero ideals of infinite order. Flatness identifies the ideal pullback with its extended ideal.

5.1A1F3F5F8step 3.2algebra∎

Each pulled-back member is a product of distinct parameters in the regular local domain B, hence a nonzerodivisor. Each individual parameter generates a prime ideal, since its quotient is regular by [F5] and a domain by [F3]. Their distinct principal prime ideals have intersection equal to their product: divisibility by one prime parameter and the fact that it divides none of the others prove this successively. The product ideal is consequently radical. Thus each inverse image is a reduced effective Cartier divisor, completing (2). AC is used only through the parameter and associated-graded suppliers [A1]; no completion isomorphism or equality of residue fields is used.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

Marked ideals and their support

Definition

Let X be a smooth K-scheme of pure dimension n (Conventions for the resolution development) and let E be a finite family of reduced divisors on X in simultaneous SNC position (Simple normal crossings divisors and simultaneous normal crossings position). A marked ideal on X is a triple (I,E,μ) consisting of a coherent ideal sheaf I⊆OX, the family E, and an integer μ≥0; it is displayed as (I,μ) when E is understood. The support of (I,E,μ) is supp⁡(I,E,μ):={x∈X:ord⁡x(I)≥μ} (Order of an ideal sheaf at a point); it is in fact a closed subset of X, as shown below. The canonical-resolution existence theorem below requires μ≥1 and I nonzero at the generic point of every irreducible component of X. Merely requiring I≠0 globally does not suffice on a disconnected smooth scheme. For μ=0 the support is all of X, so support-clearing resolution by the canonical algorithm is not asserted. Writing (I,μ) for (I,E,μ) suppresses only E, never the order μ.

DefinitionDefinition: Literature-sourcedProof: Not applicableOpen item page →

Multiple test blow-ups, controlled transforms and resolutions of marked ideals

Definition

Let (I,E,μ) be a marked ideal on the smooth K-scheme X (Marked ideals and their support). A multiple test blow-up of (I,E,μ) is a finite sequence X=X0←σ1X1←⋯←Xr in which each σi ⁣:Xi→Xi−1 is specified either as an inserted isomorphism step or as the blowup (Blowup of a scheme along an ideal sheaf) of Xi−1 at a regular closed subscheme Ci−1⊆supp⁡(Ii−1,Ei−1,μ) (Closed immersions of schemes) that has SNC with Ei−1, together with the marked ideals defined inductively for a blowup step by (Ii,Ei,μ),Ii:=I(Di)−μ⋅σi∗(Ii−1),Ei:=σic(Ei−1)∪{Di}, where Di⊆Xi is the exceptional divisor (Exceptional subscheme of a blowup), I(Di) is its invertible ideal (Invertible sheaf of cartier divisor), I(Di)−μ its (−μ)-th tensor power (Invertible sheaves, Tensor product of sheaves of modules), and σic(Ei−1) is the family of strict transforms (Strict transform of a closed subscheme) ordered so that all old members precede Di. Empty members of the transformed boundary are omitted, with the order of all surviving labels retained. The transform σic(Ii−1,μ):=(Ii,μ) is the controlled transform of (Ii−1,μ); for a local section f∈Ii−1(U) with local equation yi of Di the section yi−μσi∗(f) is a controlled transform of f, well defined up to a unit. A blowup step retains the specified center and Di=σi−1(Ci−1), even when its underlying morphism is an isomorphism, as for a Cartier center. An inserted isomorphism step has empty Di, transports the ideal and ordered boundary, and appends no boundary member. A resolution of (I,E,μ) is a multiple test blow-up with supp⁡(Ir,Er,μ)=∅. An extension of a multiple test blow-up (Xi)0≤i≤m is a multiple test blow-up (Xj′)0≤j≤m′ with X0′=X, indices j0=0<j1<⋯<jm and isomorphisms forming an identification Xji′=Xi; the extended sequence is obtained from the original one by inserting isomorphisms, with every original blow-up retained in its original order. No new nontrivial blow-ups are inserted. This is the extension convention of the source, Definition 2.1.5.

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

Derivative ideals of an ideal sheaf and of a marked ideal

Definition

Assume AC (The Axiom of Choice) through the smooth-differential supplier. Let X be a smooth K-scheme (Smooth morphism of schemes) with sheaf of relative differentials ΩX/K (Sheaf of relative Kähler differentials, Existence and generators of Kähler differentials) and derivation sheaf Der⁡K(OX) (Derivation of an algebra, Derivations are maps out of Ω). For a coherent ideal sheaf I⊆OX (Coherent module sheaves) define the first derivative (extension) D(I) intrinsically as the ideal generated by I and all D(f), for local sections f of I and K-derivations D. The smooth-differential theorem Differentials of a smooth morphism makes ΩX/K finite locally free. Choose functions u1,…,un whose differentials form a basis on a neighbourhood and let ∂/∂ui be the dual derivations. Here n is the relative dimension over K, which need not equal the local-ring dimension at a nonclosed point. For generators f1,…,fs of I, the same ideal is generated by the fj and ∂fj/∂ui: Leibniz reduces derivatives of arbitrary ∑ajfj to these generators, and every derivation is a linear combination of the dual basis. This finite generator list proves coherence and independence of generators and differential coordinates. Set D0(I):=I and Di(I):=D(Di−1(I)). The ideal is generated by fj and their derivatives of order at most i and is independent of the chosen generators and coordinates; for a marked ideal put Di(I,μ):=(Di(I),μ−i) for 0≤i≤μ. In characteristic p>0, ordinary derivations need not lower order: on AK1 for a field K of characteristic p, the ideal I=(xp) satisfies D(I)=I.

DefinitionDefinition: Literature-sourcedProof: Not applicableOpen item page →

Equivalence of marked ideals

Definition

Two marked ideals (I,EI,μI) and (J,EJ,μJ) on the same smooth K-scheme X are equivalent, (I,EI,μI)≃(J,EJ,μJ), if: (1) EI=EJ as ordered families of divisors; (2) their supports agree; and (3) the multiple test blow-ups of the one are exactly the multiple test blow-ups of the other, and for every such blow-up (Xi) the induced supports agree, supp⁡(Ii,Ei,μI)=supp⁡(Ji,Ei,μJ) for every i (Multiple test blow-ups, controlled transforms and resolutions of marked ideals). The relation is reflexive, symmetric and transitive by definition, and the algorithm below replaces a marked ideal by equivalent ones at the steps marked in the source.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Controlled transforms are well defined

Statement

Assume AC (The Axiom of Choice), inherited from the regular-local and blowup suppliers. Let (I,E,μ) be a marked ideal on a smooth K-scheme X, let C⊆supp⁡(I,E,μ) be a regular closed subscheme with SNC with E, let σ ⁣:X′→X be the blowup of C with exceptional divisor D (Blowup of a scheme along an ideal sheaf, Exceptional subscheme of a blowup), and let IC⊆OX be the ideal sheaf of C and I(D)⊆OX′ the invertible ideal of D (Invertible sheaf of cartier divisor). Then I⊆ICμandσ∗I⊆I(D)μ. Consequently the controlled transform σc(I,μ):=(I(D)−μσ∗I,μ) is an ideal sheaf on X′ (Multiple test blow-ups, controlled transforms and resolutions of marked ideals), and for f∈I(U) the local section y−μσ∗(f), y a local equation of D, is well defined up to a unit and generates the controlled transform wherever f generates I.

Facts & Assumptions

Given: A marked ideal (I,E,μ) on a smooth K-scheme X, a regular closed subscheme C⊆supp⁡(I,E,μ) with SNC with E, the blowup σ ⁣:X′→X of C with exceptional divisor D, the ideal IC of C and the invertible ideal I(D) of D.

[F1]

Marked ideals and their support: the support is supp⁡(I,E,μ)={x:ord⁡x(I)≥μ}, and C⊆supp⁡(I,E,μ) means ord⁡x(I)≥μ for every x∈C.

[F2]

Order of an ideal sheaf at a point: ord⁡x(I)=max⁡{n:Ix⊆mx n}; equivalently, ord⁡x(I) is the minimum of ord⁡x(f) over local sections f of I at x.

[F3]

The pulled-back center ideal is the relative twist; the exceptional divisor is Cartier, Exceptional subscheme of a blowup, Invertible sheaf of cartier divisor: σ∗IC=I(D) is the invertible ideal of the exceptional divisor, generated locally by the equation y of D.

[F4]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals: the controlled transform of (I,μ) along σ is σc(I,μ)=(I(D)−μσ∗I,μ); for a local section f the section y−μσ∗(f) is called a controlled transform of f.

[F5]

regular local regular quotient ideal is parameter generated, regular local rings are domains and cohen macaulay: at a point x∈C the ideal IC is generated by parameters u1,…,uk that extend to a regular system of parameters u1,…,un of OX,x; in particular C is reduced at x, and a section not in IC has a nonvanishing value at some point of C.

[F6]

The pulled-back center ideal is the relative twist; the exceptional divisor is Cartier, Effective cartier divisor: the exceptional divisor D is an effective Cartier divisor with invertible ideal I(D) and local equation y a nonzerodivisor.

[F7]

Associated graded algebra of an ideal generated by a regular sequence (Stacks, Lemma 10.69.2): if P=(u1,…,uc) is generated by a regular sequence in R, then gr⁡PR=(R/P)[U1,…,Uc]. In particular each Pj/Pj+1 is a finite free R/P-module. The proof eliminates a homogeneous relation by induction on the sequence length and its degree, using nonzerodivisibility of the last generator modulo the preceding ones.

Proof

1.1F1F2F5F7

Fix x∈C, put R=OX,x and P=IC,x. By [F5], P is generated by an initial parameter sequence and R/P is a regular local domain. If f∈Ix were outside Pμ, choose the largest j<μ with f∈Pj. Its nonzero class in Pj/Pj+1 remains nonzero after localization at P, since [F7] makes this module free over the domain R/P. But RP=OX,η at the generic point η of the component of C through x, and its maximal ideal is PRP. Thus f has order j<μ at η, contradicting η∈C⊆supp⁡(I,μ). This proves Ix⊆Pμ; outside C the inclusion is automatic. If μ=0 it is immediate everywhere.

2.1F3step 1.1

Second inclusion. Pulling back the inclusion of step 1.1 along σ and using that inverse image commutes with ideal products, σ∗I⊆σ∗(ICμ)=(σ∗IC)μ=I(D)μ, the last equality by [F3].

3.1F3F4F6step 2.1∎

The controlled transform is defined and well posed. By step 2.1 the product I(D)−μσ∗I is an ideal sheaf on X′, namely the controlled transform of [F4]. If y is a local equation of D and f,f′ are local generators of I on an open set, then f′=uf for a unit u, so y−μσ∗(f′)=(σ∗u) y−μσ∗(f) differs by the unit σ∗u; and replacing y by the equation y′=vy of another local generator changes y−μσ∗(f) by v−μ, again a unit. Hence y−μσ∗(f) is well defined up to a unit and generates the controlled transform wherever f generates I, as asserted.

Remarks

  • The two inclusions admit the empty center reading: if C=∅ then IC=OX and both inclusions are trivial; the controlled transform is then σ∗I with σ an isomorphism, matching the definition of a multiple test blow-up extended by isomorphisms.
  • No choice beyond the published blowup interface is used in this item.
LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Iterated derivative ideals preserve support in the safe characteristic range

Statement

Assume AC (The Axiom of Choice) for the completed-local and smooth-coordinate arguments.

Let (I,μ) be a marked ideal on a smooth K-scheme X with I≠0 (Marked ideals and their support, Smooth morphism of schemes), and let 0≤i≤μ−1 (Derivative ideals of an ideal sheaf and of a marked ideal). In every characteristic, supp⁡(I,μ)⊆supp⁡(Di(I),μ−i), and, when K is perfect, supp⁡(I,μ) is closed. If K has characteristic zero, or is perfect of characteristic p>0 with μ<p (Field), then the inclusion is an equality. In these same characteristics, for μ≥1 the condition ord⁡x(I)≤μ for every x∈X is equivalent to Dμ(I)=OX.

Facts & Assumptions

Given: A marked ideal (I,μ) on a smooth K-scheme X and an integer 0≤i≤μ−1.

[F1]

Derivative ideals of an ideal sheaf and of a marked ideal: D0(I)=I, Di(I) is generated locally by generators f of I and their coordinate partial derivatives of order at most i; the recursive identity Di(Dj(I))=Di+j(I) holds in every characteristic.

[F2]

Order of an ideal sheaf at a point: ord⁡x(I)=max⁡{n:Ix⊆mx n}, with order at least μ equivalent to vanishing in OX,x/mxμ.

[F3]

Marked ideals and their support: supp⁡(I,μ)={x:ord⁡x(I)≥μ}.

[F4]

The regular-local associated-graded and completion theorems identify gr⁡mA with κ(x)[U1,…,Ud] and its completion with κ(x)⟦u1,…,ud⟧ when K is perfect. A coefficient field containing K is obtained by lifting a separating transcendence basis of κ(x)/K, then its finite separable algebraic generators by the coefficient-field adjunction lemmas. The parameter map is surjective by the Cohen presentation and injective by the associated-graded isomorphism. These are associated graded ring of a regular local ring, Completion of a Noetherian local ring is local with the same residue field, completion preserves regular local rings, A complete equicharacteristic Noetherian local ring is a power-series quotient, Transcendental residue elements adjoin across a maximal subfield, and Separable residue elements adjoin across a maximal subfield. Formal parameter derivatives restrict to derivations A→A^; since ΩA/K is finite free locally, universality identifies them with A^-linear combinations of the algebraic derivations (Differentials of a smooth morphism, Derivations are maps out of Ω). Iterated Leibniz therefore puts their order-r derivatives of f∈I in Dr(I)A^.

[F5]

On an étale chart to affine N-space, the infinitesimal Taylor map with coordinate increments t1,…,tN exists uniquely by Etale morphisms are the formally etale morphisms locally of finite presentation, modulo (t)μ. Its finitely many coefficients for a function f are the Hasse derivatives of orders <μ, regular functions on the chart. Over a perfect field, their residues all vanish at x exactly when f∈mxμ. Indeed the completed chart can be expressed using a separating residue-field coordinate system and the normal parameters in [F4]; Taylor substitution in the normal parameters detects every nonzero initial form of degree <μ. An invertible change of smooth coordinates gives invertible changes of these truncated Taylor coefficients. This reasoning concerns Hasse derivatives, and uses no factorial division.

Proof

1.1F1F2F3algebra

If Ix⊆mxμ, Leibniz shows D(mxa)⊆mxa−1 for every derivation D and a≥1: differentiate each product of a elements of mx. Iterating gives Di(I)x⊆mxμ−i. This proves the forward inclusion over any field, independently of perfection or factorials.

1.2F2F3F5

Assume K perfect. For finitely many local generators fj of I, take the finitely many Taylor coefficients in [F5] of orders <μ. Their simultaneous vanishing locus is exactly {x:Ix⊆mxμ}, so this set is closed on each chart and hence on X. This proves closedness in the stated perfect-field range, in every characteristic.

1.3F1F2F4

Reverse inclusion in the safe-order range. Assume char⁡K=0 or K perfect with char⁡K=p>μ. If x∈supp⁡(Di(I),μ−i) but ord⁡x(I)=j<μ, choose f∈Ix of order j and let fj be its nonzero initial form in [F4]. If j≤i, choose a monomial cUα of fj with ∣α∣=j and differentiate by ∂α; its initial constant term is cα!≠0, since j≤μ<p in positive characteristic. This puts a unit in Di(I)x, contradicting its order being at least μ−i≥1. If i<j, choose a monomial cUα of fj and a multiindex β≤α with ∣β∣=i. Then ∂βfj is nonzero: its selected coefficient is a product of falling factorials of integers at most j<p, so is nonzero in κ(x). Hence ∂βf has order exactly j−i<μ−i, contradicting the same support assumption. Thus the reverse inclusion holds in the stated range.

2.1F1F2F4step 1.1∎

The maximal-order criterion in the same range. Suppose first that ord⁡x(I)≤μ for every x. For a point with j:=ord⁡x(I)>0, choose f of order j and a monomial cUα in its initial form; ∣α∣=j≤μ, and the coefficient of ∂αf is cα!≠0 in characteristic zero or when p>μ. Thus Dμ(I)x=OX,x; the case j=0 is immediate since Ix=OX,x. Conversely, if Dμ(I)x=OX,x, at least one of its local generators ∂αf is a unit, with f∈Ix and ∣α∣≤μ. Since differentiation lowers order by at most ∣α∣, ord⁡x(f)≤∣α∣≤μ. This proves the equivalence stalkwise.

Remarks

Perfection is essential to the positive-characteristic converse: for K=Fp(a), f=xp−a on AK1 has order one at the closed point (f), but every ordinary K-derivative of f vanishes. Thus D(f)=(f) and the marking-one maximal-order criterion fails even though 1<p. The all-characteristic forward inclusion above remains valid. Closedness over imperfect fields is not established by this proof or used by the characteristic-zero development.

LemmaStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Etale pullback commutes with derivative ideals

Statement

Assume the Axiom of Choice.

Let φ ⁣:X′→X be an etale morphism of smooth K-schemes (Étale morphism of schemes) and let I⊆OX be a coherent ideal sheaf (Coherent module sheaves). Then for every i≥0 φ∗(Di(I))=Di(φ∗I), where on both sides the derivative ideals are taken over K (Derivative ideals of an ideal sheaf and of a marked ideal).

Facts & Assumptions

Given: The data in the Statement, with AC assumed through the cited smooth-differentials and étale suppliers.

[A1]

The Axiom of Choice: AC is used only through the explicitly AC-assuming suppliers [F4] and [F5].

[F1]

Derivative ideals of an ideal sheaf and of a marked ideal: D(I) is generated locally by local generators f of I and all first partial derivatives ∂f/∂ui in local coordinates u1,…,un; equivalently, globally, it is generated by the f and the sections D(f) for all K-derivations D of the structure sheaf.

[F2]

Derivations are maps out of Ω, Derivation of an algebra: for a K-algebra A and A-module M, Hom⁡A(ΩA/K,M)≅Der⁡K(A,M); for a locally free ΩA/K this gives Der⁡K(A,M)≅M⊗ADer⁡K(A).

[F3]

Transitivity sequence for differential modules: for K→A→B the sequence B⊗AΩA/K→ΩB/K→ΩB/A→0 is exact.

[F4]

Under AC, Etale morphisms are the formally etale morphisms locally of finite presentation says an étale morphism is formally étale; it is therefore formally unramified, and Formal unramifiedness iff Omega vanishes gives ΩB/A=0, or ΩX′/X=0 in sheaf notation.

[F5]

Under AC, Differentials of a smooth morphism identifies the rank of relative differentials with relative dimension. The composition rule in Smoothness survives base change and composition says reldim⁡X′/K(x′)=reldim⁡X/K(x)+reldim⁡X′/X(x′); since φ is étale, the last term is 0. Thus ΩX/K, ΩX′/K and the pullback of the former are locally free of the same rank at corresponding points.

[F6]

Locally free sheaves of finite rank: a surjection of finite locally free modules of the same rank is an isomorphism; this is checked after localising to free modules and reducing to linear algebra.

[F7]

Sheaf of relative Kähler differentials: ΩX/K is the sheaf of relative differentials, with its universal K-derivation; the pullback φ∗ΩX/K is the sheaf φ−1ΩX/K⊗φ−1OXOX′.

Proof

1.1F1

Affine-local reduction. The question is local on X′ and on X: it suffices to prove φ∗D(I)=D(φ∗I) on affine charts Spec⁡B→Spec⁡A of an étale ring map, and then to iterate for higher derivatives. So fix an étale map A→B and an ideal I⊆A; write φ also for the ring map.

1.2F3F4F5F6

The differential comparison. By [F4] one has ΩB/A=0, so the transitivity sequence of [F3] gives a surjection B⊗AΩA/K↠ΩB/K. Both sides are finite locally free by [F5]. Their ranks agree at corresponding points because the relative dimensions of X′/K and X/K agree by [F5]; no identification with the local-ring dimension is needed. By [F6] the comparison map B⊗AΩA/K→ΩB/K is an isomorphism.

2.1F2F7step 1.2

Derivations under the étale map. Dualising the isomorphism of step 1.2 and using [F2, F7], Der⁡K(B)≅Hom⁡B(ΩB/K,B)≅Hom⁡B(B⊗AΩA/K,B)≅Hom⁡A(ΩA/K,B)≅B⊗ADer⁡K(A). Concretely, every K-derivation D′ of B is a finite sum ∑ibiDi′ where for each i there is a K-derivation Di of A with Di′(φ(a))=φ(Di(a)) for all a, and conversely each Di gives such a Di′.

3.1A1F1step 2.1∎

The ideals agree. The ideal D(φ∗I) is generated by φ∗I together with all D′(g) for g∈φ∗I and K-derivations D′ of B ([F1]). Write a generator g of φ∗I=IB as ∑jbjφ∗fj with fj∈I. For D′=∑ibiDi′ as in step 2.1, the Leibniz rule gives D′(g)=∑i,jbiDi′(bj)φ∗fj+∑i,jbibjφ∗(Difj), which lies in φ∗D(I). Conversely φ∗D(I) is generated by the φ∗f and the φ∗(Dif)=Di′(φ∗f), which lie in D(φ∗I). Hence φ∗D(I)=D(φ∗I), and iteration gives φ∗Di(I)=Di(φ∗I) for every i≥0. AC [A1] is used only through [F4] and [F5]; the derivation transport and ideal-generation computations are choice-free.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Restriction of a marked ideal to a smooth subvariety and its blow-ups

Statement

Assume AC (The Axiom of Choice) for the regular-parameter and blowup suppliers.

Let (I,E,μ) be a marked ideal of maximal order on the smooth K-scheme X, and let S⊆X be a regular closed subscheme having SNC with E and not contained in supp⁡(I,E,μ) (Marked ideals and their support, Simple normal crossings divisors and simultaneous normal crossings position). For restriction to S, work componentwise and omit from the restricted boundary the members containing that component of S; retain the restrictions of the other members in their original order. Here SNC position for S means that its ideal is generated by a subset of parameters compatible with the boundary equations. This convention is required, for example, when S is itself a boundary stratum: a divisor containing S does not restrict to a Cartier divisor on it.

Then supp⁡(I,μ)∩S⊆supp⁡(I∣S,μ), where I∣S is the pullback ideal on S (Coherent module sheaves). If C⊆supp⁡(I,μ)∩S is a regular center with SNC with E, σ ⁣:X′→X is the blowup and S′⊆X′ is the strict transform of S (Strict transform of a closed subscheme), then σc((I,μ)∣S)=(σc(I,μ))∣S′. Moreover, for any multiple test blow-up (Xi) of (I,μ) all of whose centers lie in the strict transforms Si of S, the restrictions σi∣Si define a multiple test blow-up (Si) of (I,μ)∣S and [(I,μ)∣S]i=(Ii,μ)∣Si for every i.

Facts & Assumptions

Given: A marked ideal (I,E,μ) on a smooth K-scheme X whose support does not contain a smooth subvariety S⊆X that has SNC with E; a blowup σ ⁣:X′→X with center C⊆supp⁡(I,E,μ)∩S; the strict transform S′⊆X′ of S.

[F1]

Marked ideals and their support: the restriction of the marked ideal is (I,μ)∣S=(I⋅OS,μ), with support {x∈S:ord⁡x(IOS)≥μ}.

[F2]

Order of an ideal sheaf at a point: order is defined by containment of stalks in powers of the maximal ideal; for x∈S the maximal ideal of OS,x is the image of mX,x, so Ix⊆mX,xμ implies IxOS,x⊆mS,xμ; restriction can only raise the order.

[F3]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals, Exceptional subscheme of a blowup: the controlled transform of a section f∈I(U) is f′=y−μσ∗(f) for a local equation y of the exceptional divisor D; the controlled transform of the marked ideal is generated by the f′.

[F4]

Strict transform of a closed subscheme, Blowup of a scheme along an ideal sheaf: in coordinates x1,…,xk defining S and y1,…,yn−k along it at a point of C, with the center described by x1,…,xk,y1,…,ym, the chart of the blowup has coordinates xi′=xi/ym (i≤k), yj′=yj/ym (j<m), ym′=ym, yj′=yj (j>m), and the strict transform S′ is described by x1′=⋯=xk′=0 with ym′ a local equation of the exceptional divisor of S′→S.

[F5]

embedding dimension and regular local ring, Simple normal crossings divisors and simultaneous normal crossings position: S is smooth with SNC with E, so the coordinates can be chosen adapted both to S and to E; after omitting the members containing a component of S, the remaining restrictions are again a family in simultaneous SNC position. The same omission convention applies at every stage.

Proof

1.1F1F2

The first inclusion. Let x∈supp⁡(I,μ)∩S, so Ix⊆mX,xμ. Applying the ring map OX,x→OS,x gives IxOS,x⊆mX,xμOS,x=mS,xμ, so ord⁡x(I⋅OS)≥μ and x∈supp⁡((I,μ)∣S). Hence supp⁡(I,μ)∩S⊆supp⁡((I,μ)∣S).

1.2F3F4F5algebra

Work at a point of S′ over C. By the parameter-generation theorem, the center ideal is (x1,…,xk,y1,…,ym), where S is defined by the x's. On a chart indexed by an xi, saturation makes the strict transform of S empty. On a chart indexed by a=yj, it is defined by xi/a=0, and its chart is precisely the corresponding chart of Bl⁡CS. The exceptional equation on S′ is a∣S′, a nonzerodivisor. Thus for every generator f of I, restriction of a−μσ∗f to S′ equals (a∣S′)−μ(σ∣S′)∗(f∣S). This proves the transform identity on every nonempty chart without assuming a power-series expansion; the empty charts have no stalk to check. These identities glue, and the remaining restricted boundary has SNC by [F5].

2.1F1F5step 1.2∎

Iteration along a multiple test blow-up. Suppose (Xi)0≤i≤r is a multiple test blow-up of (I,μ) with every center Ci contained in the strict transform Si of S. By induction on i, step 1.2 applied to the restricted marked ideal (Ii,μ)∣Si and the blowup σi+1 with center Ci⊆Si gives [(I,μ)∣S]i+1=σi+1c((Ii,μ)∣Si)=((Ii+1,μ)∣Si+1); moreover Ci, being also a center for the restricted marked ideal with SNC with Ei∣Si by [F5], makes (Si) a multiple test blow-up of (I,μ)∣S with the same transform rule. The base case i=0 is the identity. This proves the final assertion for every length, and in particular the equality [(I,μ)∣S]i=(Ii,μ)∣Si stated in the lemma.

Remarks

  • The hypothesis that S is not contained in supp⁡(I,μ) is used only to keep the two suppressed-locus readings apart; the computation of step 1.2 uses no such hypothesis, and the first inclusion of step 1.1 is unconditional. The empty case S=∅ makes all statements vacuous.
  • The identity is the source's Lemma 2.10.3 and is the restriction calculus on which the coefficient-ideal lemmas below are built.
LemmaStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Derivative ideals under semilinear ground-field isomorphisms

Statement

Let K and K′ be fields of characteristic zero and let σ ⁣:K⟶∼K′ be a field isomorphism; both fields have prime subfield Q (Field, A field's prime subfield is isomorphic to Q in characteristic zero and to Fp in characteristic p). Let X be a smooth K-scheme and X′ a smooth K′-scheme (Smooth morphism of schemes). Suppose φ ⁣:X′→X is a σ-semilinear isomorphism, meaning that it is an isomorphism of Q-schemes and its pullback acts on the ground-field constants by σ (Morphisms of schemes).

Then for every coherent ideal sheaf I⊆OX and every i≥0, φ∗(DKi(I))=DK′i(φ∗I), where the derivative ideals on X and X′ are formed using K- and K′-derivations, respectively (Derivative ideals of an ideal sheaf and of a marked ideal).

Facts & Assumptions

Given: Fields K,K′ of characteristic zero, a field isomorphism σ ⁣:K→K′, smooth schemes X/K and X′/K′, a σ-semilinear isomorphism φ ⁣:X′→X, and a coherent ideal sheaf I⊆OX.

[F1]

Derivative ideals of an ideal sheaf and of a marked ideal: DK(I) is generated locally by I and the sections D(f) for K-derivations D; DKi is the i-fold iterate, and similarly over K′.

[F2]

Derivation of an algebra: a derivation is additive, satisfies the Leibniz rule, and is linear over the indicated ground field.

[F3]

Morphisms of schemes: on corresponding open sets the isomorphism induces inverse ring isomorphisms φ∗ and (φ∗)−1, and semilinearity means φ∗(c)=σ(c) for c∈K.

[F4]

Field, A field's prime subfield is isomorphic to Q in characteristic zero and to Fp in characteristic p: the prime subfields of K and K′ are both Q, and σ fixes that prime field.

Proof

1.1F2F3F4

Transport derivations. Let D be a K-derivation of OX(U) and define D′ on OX′(φ−1U) by D′(g)=φ∗(D((φ∗)−1g)). If c′=σ(c)∈K′, then semilinearity gives D′(c′g)=φ∗(cD((φ∗)−1g))=c′D′(g) because D(c)=0; the Leibniz rule follows by conjugating the Leibniz rule for D. Thus D′ is a K′-derivation. Conjugation by φ∗ is bijective, with inverse conjugation by (φ∗)−1.

2.1F1step 1.1∎

Derivative ideals agree. For each local section f∈I(U), one has D′(φ∗f)=φ∗(D(f)). As D varies, the bijection in step 1.1 identifies all K-derivative generators with all K′-derivative generators, so φ∗(DK(I))=DK′(φ∗I). Applying this identity successively to each derivative ideal gives φ∗(DKi(I))=DK′i(φ∗I) for every i≥0.

Remarks

  • Włodarczyk's Lemma 4.3.1 states this for varieties over one characteristic-zero field and an isomorphism over Q; the semilinear formulation above also permits relabelling the ground field along an isomorphism K≃K′.
  • In particular, this applies to the automorphisms of an algebraic closure in Galois descent; those automorphisms need not be linear over the algebraic closure.
DefinitionDefinition: Literature-sourcedProof: Not applicableOpen item page →

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors

Definition

Let (I,E,μ) be a marked ideal with I≠0 on the smooth K-scheme X (Marked ideals and their support). It is of maximal order if ord⁡x(I)≤μ for every x∈X; when μ is attained this is equivalent to supp⁡(I,E,μ)={x:ord⁡xI=μ}. For μ≥1, in characteristic zero or in perfect characteristic p>0 with μ<p, maximal order is also equivalent to Dμ(I)=OX (Field, Iterated derivative ideals preserve support in the safe characteristic range). For the tangent-direction construction require μ≥1. Put T(I):=Dμ−1(I) (Derivative ideals of an ideal sheaf and of a marked ideal). A tangent direction of (I,E,μ) on an open U⊆X is a section u∈T(I)(U) of multiplicity one (Order of an ideal sheaf at a point): ord⁡x(u)=1 for every x∈V(u), so V(u) is a regular hypersurface in U containing supp⁡(I,E,μ)∩U. Such a u is transversal to E at x if x∈V(u) and the class of u together with the classes of the distinct local boundary equations through x is linearly independent in mx/mx2, equivalently these equations and u extend together to a regular system of parameters of OX,x (Simple normal crossings divisors and simultaneous normal crossings position); in particular u cannot be a boundary parameter or lie in the span of the boundary classes. This hypothesis ensures that V(u) is transversal to the boundary and is used in restriction, completion-automorphism and glueing arguments.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Addition and multiplication of marked ideals

Statement

Let (X,E) be a smooth K-scheme with a fixed family E in simultaneous SNC position, and let (I,μI), (J,μJ), (I1,μ1),…,(Im,μm) be marked ideals on X with common E (Marked ideals and their support). All marks are nonnegative. For the sum operation below require every summand mark to be positive; the product operation permits zero marks. For assertion (1), assume the Axiom of Choice; assertion (2) and its proof are choice-free.

For positive marks define

(I,μI)+(J,μJ):=(IμJ+JμI, μIμJ),

and inductively

(I1,μ1)+⋯+(Im,μm):=(∑j=1mIj∏k≠jμk, ∏kμk).

For nonnegative marks define

(I,μI)⋅(J,μJ):=(IJ, μI+μJ).

(1) For any m≥1, the support of the sum is ⋂jsupp⁡(Ij,μj). Its multiple test blow-ups are exactly the simultaneous multiple test blow-ups of all summands, and controlled transforms commute with sums:

(I1,μ1)i+⋯+(Im,μm)i=[(I1,μ1)+⋯+(Im,μm)]i

at every stage i.

(2) The product satisfies

supp⁡(I,μI)∩supp⁡(J,μJ)⊆supp⁡(IJ,μI+μJ).

Every simultaneous multiple test blow-up of (I,μI) and (J,μJ) is a multiple test blow-up of their product, and

(Ii,μI)⋅(Ji,μJ)=[(I,μI)⋅(J,μJ)]i

at every such stage.

Under the AC hypothesis in (1), the sum is not associative on the nose, but its two bracketings are equivalent in the sense of Equivalence of marked ideals.

Facts & Assumptions

Given: The smooth K-scheme, common SNC boundary, marked ideals, and weight ranges stated above. Assertion (1) is under AC; assertion (2) has no choice assumption.

[A1]

The Axiom of Choice: AC is used in assertion (1) through the associated-graded theorem for regular local rings; no choice is used in assertion (2).

[F1]

Marked ideals and their support, Order of an ideal sheaf at a point: supp⁡(A,ν)={x:ord⁡x(A)≥ν}, with ord⁡x(A)=+∞ for the zero ideal and finite order attained for nonzero ideals.

[F2]

Tensor product of sheaves of modules: products and powers of ideal sheaves are formed by multiplying local sections.

[F3]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals: a multiple test blow-up has regular centers in the successive supports meeting the successive boundaries with SNC; its controlled transform is σc(A,ν)=(I(D)−νσ∗A,ν).

[F4]

Equivalence of marked ideals: two marked ideals with the same ordered boundary are equivalent when their supports and all multiple test blow-ups, with induced supports, agree.

[F5]

Smooth morphism of schemes, Geometrically regular algebras and geometrically regular fibres: for every x∈X, the local ring OX,x is a regular local ring, since X→Spec⁡K is smooth.

[F6]

Under AC, associated graded ring of a regular local ring identifies gr⁡mx(OX,x) with a polynomial algebra over its residue field; in particular, this associated-graded ring is a domain.

Proof

1.1A1F1F5F6

Order calculus. Work at x∈X, with R=OX,x and maximal ideal m. For ideals A,B⊆R, ord⁡x(A+B)=min⁡(ord⁡xA,ord⁡xB): containment of both ideals in A+B gives one inequality, and A+B⊆mn forces both into mn. Always ord⁡x(AB)≥ord⁡xA+ord⁡xB by multiplying ideal containments. For assertion (1), if A≠0 has finite order a, choose f∈A∖ma+1; its initial class in gr⁡mR is nonzero by [F1]. By [A1], [F5], and [F6], the associated-graded ring is a domain, so the initial class of fk is nonzero in degree ka; hence ord⁡x(Ak)=ka, since the reverse inequality follows from A⊆ma. The zero ideal has infinite order and its positive powers are zero, so the identity also holds there.

2.1F1step 1.1

Supports. Put μ=∏jμj and ej=∏k≠jμk, all positive. By step 1.1, ord⁡x(∑jIjej)=min⁡jejord⁡x(Ij). This is at least μ exactly when every ord⁡x(Ij)≥μj, giving the support intersection in (1). For (2), if x is in both factor supports, then ord⁡x(I)+ord⁡x(J)≥μI+μJ; the product lower bound in step 1.1 puts x in the product support. This proves the stated reverse-direction inclusion, including zero marks.

3.1F2F3step 2.1

Transform identities. For a blow-up with exceptional equation y and center in the support of the sum, step 2.1 places that center in every summand support. Writing Ij,i=y−μjσ∗Ij, we have y−μσ∗(∑jIjej)=∑j(y−μjσ∗Ij)ej=∑jIj,iej because ejμj=μ. For the product, at any simultaneous admissible center, y−(μI+μJ)σ∗(IJ)=(y−μIσ∗I)(y−μJσ∗J). Thus the corresponding controlled transforms agree. These are ideal-sheaf identities and use no choice.

4.1F3step 2.1step 3.1

Test blow-ups. The support equality in step 2.1 and transform identity in step 3.1 show inductively that a sequence is a multiple test blow-up of the sum exactly when each center is simultaneously admissible for every summand; all transformed supports agree at each stage. For the product, simultaneous admissibility puts each center in the intersection of factor supports and hence, by step 2.1, in the product support; the product transform identity then gives the induction that every simultaneous test sequence is a product test sequence.

5.1F4step 2.1step 4.1∎

Associativity of the sum. For either bracketing of three or more summands, step 2.1 identifies the initial supports and step 4.1 identifies the multiple test blow-ups and induced supports. The two bracketings therefore satisfy the equivalence criterion [F4], although their defining ideal sheaves need not be equal. This proves the final assertion.

DefinitionDefinition: AI-adaptedProof: Not applicableOpen item page →

Canonical resolutions with invariants of a marked ideal

Definition

Let (I,E,μ) be a marked ideal on a smooth K-scheme X of finite type (Marked ideals and their support). A canonical resolution of (I,E,μ) is a resolution (Xi)0≤i≤m, X0=X, together with functions inv⁡ ⁣:supp⁡(Ii,Ei,μ)→Q≥0×Q≥0∞,ν ⁣:supp⁡(Ii,Ei,μ)→Q≥0,ρ ⁣:supp⁡(Ii,Ei,μ)→Sub⁡(Ei), Q≥0∞ being the lexicographically ordered set of infinite sequences in Q≥0 with finitely many nonzero entries, such that for every i: (i) the centers Ci of the blow-ups are regular and are the locus where the pair (inv⁡,ρ) attains its maximum, in fact components of the maximal locus of inv⁡; (ii) the three invariants have finite ranges on each support; inv⁡ and the lexicographically ordered pairs (inv⁡,ν) and (inv⁡,ρ) are upper semicontinuous there; equivalently, the auxiliary functions ν and ρ are upper semicontinuous on every level stratum {inv⁡=a}; (iii) for x∈supp⁡(Ii+1,Ei+1,μ) with σi+1(x)∈Ci one has inv⁡(x)<inv⁡(σi+1(x)) or inv⁡(x)=inv⁡(σi+1(x)) and ν(x)<ν(σi+1(x)), while for σi+1(x)∉Ci one has inv⁡(x)=inv⁡(σi+1(x)), ν(x)=ν(σi+1(x)) and ρ(x)=ρ(σi+1(x)); (iv) for every etale morphism φ ⁣:X′→X the induced sequence φ∗(Xi) is an extension of the canonical resolution of φ∗(I,E,μ) and the invariants agree, inv⁡(φi(x′))=inv⁡(x′), ν(φi(x′))=ν(x′) and ρ(φi(x′))=ρ(x′). Order Sub⁡(Ei) by writing the labels of a subset in increasing order in the fixed total boundary order, padding with zeros, and comparing the resulting sequences lexicographically, with 0 below every divisor label. An empty subset is the all-zero sequence. This supplies the order used for relative upper semicontinuity of ρ and maxima of (inv⁡,ρ); it does not introduce a new choice of boundary order.

For an arbitrary étale morphism X′→X (Étale morphism of schemes), the pullback input may have a non-quasi-compact source. In condition (iv), the same notion is extended to these étale-local marked-ideal data: on every finite-type open chart of X′ use the preceding definition, and require a finite global sequence with a finite-range invariant family whose restrictions agree with those chart constructions up to inserted isomorphisms. The finite-type assumption remains on the original input X; no existence assertion is made for unrelated non-quasi-compact inputs. The canonical-resolution proposition proves this extension for arbitrary étale pullbacks, with sequence length bounded by that of the original resolution. Empty inverse centers contribute only inserted isomorphism steps, and invariant equalities refer to the corresponding stages.

The invariant inv⁡ takes values in the lexicographic order; the resolution is canonical in the sense that it is uniquely determined by the invariants, which are intrinsic.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Smooth base change of multiple test blow-ups

Statement

Assume AC (The Axiom of Choice), inherited from the order/SNC and blowup suppliers.

Let (I,E,μ) be a marked ideal on a smooth K-scheme X, let (Xi)0≤i≤r be a multiple test blow-up defining marked ideals (Ii,Ei,μ) (Multiple test blow-ups, controlled transforms and resolutions of marked ideals), and let φ ⁣:X′→X be a smooth morphism with X′ smooth of pure dimension (Smooth morphism of schemes). Put Xi′:=X′×XXi and (I′,E′,μ):=φ∗(I,E,μ). Then: (1) for every i the morphism φi ⁣:Xi′→Xi is smooth; (2) the induced sequence (Xi′)0≤i≤r is a multiple test blow-up of (I′,E′,μ) with Ii′=φi∗Ii and Ei′ the nonempty inverse images of the members of Ei with the induced order (empty members are omitted); (3) if (Xi) is a resolution of (I,E,μ) then (Xi′) is an extension of a resolution of (I′,E′,μ). The blow-up step uses flat base change of blowups (Flat base change for blowups, and failure without flatness): the pullback of σi+1 along the smooth, hence flat, morphism φi is the blowup of the inverse image center when that center is nonempty, and an isomorphism when it is empty.

Facts & Assumptions

Given: Assume AC. A marked ideal (I,E,μ) on a smooth K-scheme X, a multiple test blow-up (Xi)0≤i≤r defining marked ideals (Ii,Ei,μ), and a smooth morphism φ ⁣:X′→X with X′ smooth of pure dimension.

[A1]

The Axiom of Choice: AC is inherited through the smooth order/SNC supplier [F3] and the published blowup suppliers.

[F1]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals: each step σi+1 ⁣:Xi+1→Xi is either an isomorphism or the blowup of a regular center Ci⊆supp⁡(Ii,Ei,μ) in SNC position with Ei, and Ii+1=I(Di+1)−μσi+1∗Ii, Ei+1=σi+1c(Ei)∪{Di+1}.

[F2]

Flat base change for blowups, and failure without flatness, Universal property of the blowup: the base change of a blowup along a flat morphism is the blowup of the pulled-back ideal when the pulled-back center is nonempty, and an isomorphism otherwise; the exceptional divisor pulls back to the exceptional divisor, and φi+1∗I(Di+1)=I(Di+1′).

[F3]

Order and simultaneous normal crossings are preserved by smooth morphisms: smooth morphisms preserve the order of an ideal, ord⁡x′(φ∗I)=ord⁡φ(x′)(I), and, on pure-dimensional smooth source schemes, pull back families in simultaneous SNC position to families in simultaneous SNC position.

[F4]

Smoothness survives base change and composition, Smooth morphism of schemes: base changes of smooth morphisms along arbitrary morphisms are smooth; the fibre product of X′ with a smooth K-scheme over X is smooth over K.

[F5]

Strict transform of a closed subscheme, Exceptional subscheme of a blowup: for a flat base change of a blowup the strict transform of a divisor pulls back to the strict transform of its pullback, because the pullback of the exceptional divisor is the exceptional divisor and pullback is compatible with the open complement and scheme-theoretic closure.

Proof

1.1F1F4

The base case and the invariants of the induction. The pure dimension of X′ is preserved by blowing up regular centers: standard charts over positive-codimension centers retain the ambient dimension, whereas whole-component centers simply delete those components. For i=0 we have X0′=X′×XX=X′ and (I0′,E0′,μ)=φ∗(I,E,μ)=(φ∗I,φ−1E,μ) by definition of the pullback of a marked ideal; assertion (1) for i=0 is the smoothness of φ itself. We prove by induction on i that φi ⁣:Xi′→Xi is smooth, that (Xj′)j≤i is a multiple test blow-up of (I′,E′,μ), and that Ij′=φj∗Ij, Ej′=φj−1Ej with the induced order, omitting empty members.

1.2A1F1F2F3F4

The inductive step: center and blowup. Assume the induction hypothesis for i. If σi+1 is an inserted isomorphism step, identify Xi+1′ with Xi′ along that step and φi+1=φi; all assertions follow by transport. Otherwise, even if the blowup morphism is an isomorphism, let Ci⊆supp⁡(Ii,Ei,μ) be the center and put Ci′:=φi−1(Ci). By [F2] the base change Xi+1′:=Xi′×XiXi+1→Xi′ of the blowup is the blowup of Ci′ when Ci′≠∅ and an isomorphism otherwise; it is smooth over Xi+1 because φi is smooth and smoothness is stable under base change [F4]. The center Ci′ is regular (the fibre product of the smooth morphism φi with the regular closed subscheme Ci is smooth over K by [F4], hence regular over the characteristic-zero field), and it has SNC with Ei′=φi−1Ei by [F3].

2.1A1F3step 1.2

The inductive step: supports. For x′∈Ci′ with x=φi(x′) one has x∈Ci⊆supp⁡(Ii,μ) and hence, by [F3], ord⁡x′(φi∗Ii)=ord⁡x(Ii)≥μ; since Ii′=φi∗Ii this says Ci′⊆supp⁡(Ii′,μ), so the pulled-back sequence is admissible at step i+1.

3.1F2F5step 1.2step 2.1

The inductive step: transform identities. The exceptional divisor Di+1′ of the pulled-back blowup is φi+1−1(Di+1) and satisfies I(Di+1′)=φi+1∗I(Di+1) by [F2]. Therefore Ii+1′=I(Di+1′)−μ(σi+1′)∗Ii′=φi+1∗(I(Di+1)−μσi+1∗Ii)=φi+1∗Ii+1, using the commutativity of pullback with products and with the controlled transform of ideals; and Ei+1′=(σi+1′)c(Ei′)∪{Di+1′}=φi+1−1((σi+1)c(Ei)∪{Di+1})=φi+1−1Ei+1 by [F5]; the order is the induced one after omitting empty members. If the pulled-back center is empty, its exceptional inverse image is empty and the step transports the existing boundary; a nonempty Cartier-center blowup retains its nonempty exceptional divisor despite having an isomorphic underlying morphism. This closes the induction and proves assertions (1) and (2).

4.1A1F1F3step 3.1∎

Resolution case. Suppose (Xi) is a resolution of (I,E,μ), so supp⁡(Ir,μ)=∅. By the induction identity Ir′=φr∗Ir, and by [F3] every point x′ of Xr′ satisfies ord⁡x′(Ir′)=ord⁡φr(x′)(Ir) if φr(x′) lies in the locus where Ir is defined; since supp⁡(Ir,μ)=∅, no point satisfies ord⁡≥μ, so supp⁡(Ir′,μ)=∅ as well. Hence the steps of (Xi′) with nonempty centers form a resolution of (I′,E′,μ), and the full sequence (Xi′) is an extension of it in the sense of [F1], which is assertion (3).

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Controlled derivative transforms are contained in derivatives of the controlled transform

Statement

Let (I,μ) be a marked ideal on a smooth K-scheme, C⊆supp⁡(I,μ) a regular center with SNC with E, σ ⁣:X′→X the blowup with exceptional divisor D, and 0≤r≤μ (Multiple test blow-ups, controlled transforms and resolutions of marked ideals). Then σc(Dr(I),μ−r)⊆Dr(σc(I,μ)), i.e. the controlled transform of the r-th derivative ideal is contained in the r-th derivative ideal of the controlled transform (Derivative ideals of an ideal sheaf and of a marked ideal).

Facts & Assumptions

Given: A marked ideal (I,μ) on a smooth K-scheme X, a regular center C⊆supp⁡(I,μ) with SNC with E, the blowup σ ⁣:X′→X with exceptional divisor D and local equation y, and 0≤r≤μ.

[F1]

Derivative ideals of an ideal sheaf and of a marked ideal: D(A) is generated by the local sections of A and their first derivatives with respect to local coordinates; equivalently by the sections f∈A together with D(f) for K-derivations D of OX; Dr is the iterate and Dr(I,μ)=(Dr(I),μ−r).

[F2]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals, Controlled transforms are well defined: the controlled transform is σc(A,ν)=(I(D)−νσ∗A,ν), computed by f↦y−νσ∗(f) on sections; it is an ideal sheaf, well defined up to units.

[F3]

embedding dimension and regular local ring, Blowup of a scheme along an ideal sheaf: at a point of C choose local parameters u1,…,un with C described by u1=⋯=um=0 and y=um; the corresponding chart of the blowup has coordinates ui′=ui/y (i<m), ui′=ui (i>m), um′=um=y.

[F4]

Derivation of an algebra, Derivations are maps out of Ω: the K-derivations of the structure sheaf form a locally free sheaf; a derivation is determined by its values on local coordinates, and yσ∗(D) for a derivation D on X acts on σ∗(f) by yσ∗(Df).

[F5]

Iterated derivative ideals preserve support in the safe characteristic range: for 0≤j≤μ−1, supp⁡(I,μ)⊆supp⁡(Dj(I),μ−j) in every characteristic, so the center hypothesis transfers to the lower derivative ideal.

Proof

1.1F3F4

Use the adapted blowup coordinates of [F3]. The chain rule gives δ:=yσ∗D∈Der⁡K(OX′) for each coordinate derivation D on X, and δ(y)/y is regular: for normal directions different from the chart index δ(y)=0, for the chart-index direction δ(y)=y, and for the tangent directions δ(y)=0. Linear combinations have the same properties.

2.1F1F2step 1.1algebra

Put I′=y−μσ∗I. For a generator f∈I, write σ∗f=yμg with g∈I′. Leibniz gives y1−μσ∗(Df)=δ(g)+μ(δ(y)/y)g∈D(I′). The undifferentiated generator of D(I) transforms to y1−μσ∗f=yg∈I′. These two calculations, including the original ideal generators, prove σc(D(I),μ−1)⊆D(I′). No division by the characteristic is used.

3.1F1F5step 2.1∎

The case r=0 is equality. For 1≤r≤μ, [F5] makes the center admissible for (Dr−1(I),μ−r+1). Apply step 2.1 to this marked ideal and then use monotonicity of D and the induction hypothesis: σc(Dr(I),μ−r)⊆D(σc(Dr−1(I),μ−r+1))⊆Dr(I′). This completes the induction.

LemmaStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

A marked ideal is equivalent to its powers

Statement

Assume the Axiom of Choice. For every marked ideal (I,E,μ) on a smooth K-scheme and every integer k≥1,

(I,E,μ)≃(Ik,E,kμ)

in the sense of Equivalence of marked ideals.

Facts & Assumptions

Given: A marked ideal (I,E,μ) on a smooth K-scheme and an integer k≥1. We use AC only through the associated-graded theorem in [F3].

[A1]

The Axiom of Choice: assume AC for applying the associated-graded theorem to the regular local rings OX,x.

[F1]

Order of an ideal sheaf at a point and Marked ideals and their support: ord⁡x(I) is the largest n with Ix⊆mxn, and the support of (I,E,μ) is {x:ord⁡x(I)≥μ}; the zero ideal has order +∞.

[F2]

Smooth morphism of schemes, Geometrically regular algebras and geometrically regular fibres: since X→Spec⁡K is smooth, every OX,x is regular local.

[F3]

Under AC, associated graded ring of a regular local ring identifies gr⁡mx(OX,x) with a polynomial algebra over the residue field, so it is a domain.

[F4]

Addition and multiplication of marked ideals, clause (2): controlled transforms commute with products of marked ideals, including products with equal factors.

[F5]

Equivalence of marked ideals: equivalence means equality of the initial supports, of the multiple test blow-ups, and of their induced supports, with the same ordered boundary.

Proof

1.1A1F1F2F3

Exact order of powers. Fix x∈X, put R=OX,x and let m be its maximal ideal. If Ix=0, then every positive power is zero and both orders are +∞. Otherwise a=ord⁡x(I) is finite and attained by [F1]. Choose f∈Ix∖ma+1; its initial class in gr⁡mR is nonzero. By [A1], [F2] and [F3], this associated-graded ring is a domain, so the initial class of fk is nonzero in degree ka. Thus ord⁡x(Ik)≤ka, while Ix⊆ma gives the reverse inequality. Hence ord⁡x(Ik)=kord⁡x(I).

2.1F1step 1.1

Equal supports. By step 1.1, ord⁡x(I)≥μ if and only if ord⁡x(Ik)=kord⁡x(I)≥kμ. Therefore supp⁡(I,E,μ)=supp⁡(Ik,E,kμ), including μ=0.

3.1F3F4step 1.1step 2.1∎

Equal multiple test blow-ups. Induct on the sequence length. At each stage i, assume the two transforms are (Ii,Ei,μ) and (Iik,Ei,kμ). Step 1.1 gives equal supports for these marked ideals, so their admissible regular centers meeting Ei with SNC agree. If y is the exceptional equation for such a center, the controlled-transform product identity [F4] gives y−kμσ∗(Iik)=(y−μσ∗Ii)k, so the next transforms again have this form and their supports agree. The base case is step 2.1; induction works in both directions, so the test sequences and all induced supports coincide. By [F5], the marked ideals are equivalent.

DefinitionDefinition: Literature-sourcedProof: Not applicableOpen item page →

The homogenized ideal of a marked ideal of maximal order

Definition

Let (I,E,μ) be a marked ideal of maximal order with μ≥1 and put T(I)=Dμ−1(I) (Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors). The homogenized ideal is the ideal sheaf H(I):=I+D(I)⋅T(I)+⋯+Di(I)⋅T(I)i+⋯+Dμ−1(I)⋅T(I)μ−1, the products and sums being products and sums of ideal sheaves; the homogenized marked ideal is H(I,μ):=(H(I),E,μ). It satisfies H(I)⊇I and T(H(I))=T(I) for μ≥1. If μ>1 and K has characteristic zero or perfect characteristic p>μ, then D(H(I))⊆H(D(I),μ−1), where (D(I),μ−1) is again of maximal order, so the right-hand homogenization is defined. It is designed so that it looks the same from every tangent direction and is equivalent to (I,μ) (proved below).

DefinitionDefinition: Literature-sourcedProof: Not applicableOpen item page →

The coefficient ideal of a marked ideal of maximal order

Definition

Let X be a smooth K-scheme over a field K, and let (I,E,μ) be a marked ideal of maximal order on X with μ≥1 (Smooth morphism of schemes). The coefficient ideal is the marked ideal C(I,μ):=∑i=0μ−1(Di(I),μ−i), the sum being that of Addition and multiplication of marked ideals; explicitly C(I,μ)=(∑i=0μ−1Di(I)μ!/(μ−i),μ!). Under AC (The Axiom of Choice), in every characteristic it satisfies C(I,μ)≃(I,μ). Under the same AC assumption, if K has characteristic zero or perfect characteristic p>μ (Field), then for every regular closed subscheme S⊆X having SNC with E, supp⁡(I,μ)∩S=supp⁡(C(I,μ)∣S), and this identity persists under multiple test blow-ups whose centers lie in the strict transforms of S (proved below).

DefinitionDefinition: Literature-sourcedProof: Not applicableOpen item page →

The monomial part, the non-monomial part and the companion ideal

Definition

Let (I,E,μ) be a marked ideal on the smooth K-scheme X, I generically nonzero on every irreducible component and μ≥1 (Marked ideals and their support). Factoring out the monomial part, write I=M(I)⋅N(I), where M(I) is the monomial part of I with respect to E, i.e. the product of powers of the invertible ideal sheaves of the components of the members of E that divide I, and N(I) is the complementary part, divisible by no component of a member of E; for E=∅ one has M(I)=OX and N(I)=I. If supp⁡(I,E,μ)=∅, the marked ideal is already resolved in Step 2, and no companion ideal is assigned. Otherwise put ord⁡N(I):=max⁡{ord⁡x(N(I)):x∈supp⁡(I,E,μ)}. If ord⁡N(I)=0, then N(I) is a unit in a neighborhood of the support, so (I,μ) is locally the monomial marked ideal (M(I),μ) there. Route this case directly to Step 2b; it has no non-monomial companion. When (I,μ) is of maximal order, (M(I),μ) is an equivalent maximal-order input for Step 1; for arbitrary inputs this is the monomial branch, without a claim that it is a maximal-order companion.

Write r=ord⁡N(I). It is finite in the characteristic-zero setting of this page: the closed order-superlevel sets on the Noetherian X stabilize, and their intersection is empty because N(I) is generically nonzero on each smooth component (Order functions and normal-crossings strata are upper semicontinuous, Order of an ideal sheaf at a point). Work on the open neighbourhood X∖{x:ord⁡xN(I)>r} of the support. Here ord⁡xN(I)≤r everywhere; outside this neighbourhood there are no support points to resolve. This restriction is needed when the maximum was taken only over the support.

If r>0, define the companion ideal O(I,μ) by O(I,μ):=(N(I),ord⁡N(I))+(M(I),μ−ord⁡N(I))when ord⁡N(I)<μ, and to O(I,μ):=(N(I),ord⁡N(I)) when ord⁡N(I)≥μ; the sum is that of Addition and multiplication of marked ideals. In the sum case, 0<ord⁡N(I)<μ, so both marks are positive; the sum clause uses AC (The Axiom of Choice). On this neighbourhood the companion is of maximal order: in the sum case its Nμ−r summand has order at most r(μ−r), which is the sum marking, and in the other case N has order at most r. Its support is supp⁡(O(I,μ))=supp⁡(I,E,μ)∩{x:ord⁡x(N(I))=ord⁡N(I)}. The support formula follows from the marked-sum intersection rule and exact additivity ord⁡(MN)=ord⁡(M)+ord⁡(N) in the regular-local associated graded domain (associated graded ring of a regular local ring). A companion selects the maximal residual-order part of the support and need not be equivalent to the original input, even if that input is of maximal order. For example, on A2⨿A2 take I=(x2) on the first component and I=(u,v)2 on the second, mark 2, and boundary V(x) on the first component. The original maximal-order support is V(x)⨿{(0,0)}; its companion has support only {(0,0)} on the second component. Resolving O(I,μ) lowers the maximal order of the non-monomial part (Step 2a of the algorithm below).

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Order functions and normal-crossings strata are upper semicontinuous

Statement

Assume AC (The Axiom of Choice) and let K be perfect. Let X be a smooth K-scheme, I⊆OX a nonzero coherent ideal sheaf and E a finite family of divisors in simultaneous SNC position (Simple normal crossings divisors and simultaneous normal crossings position). (1) The function x↦ord⁡x(I) is upper semicontinuous: for every k≥1 the set {x:ord⁡x(I)≥k} is closed. If char⁡K=0 or char⁡K=p≥k, then this set equals V(Dk−1(I)) (Iterated derivative ideals preserve support in the safe characteristic range). (2) The function sE(x):=#{D∈E:x∈D} is upper semicontinuous and locally constant on the finite stratification of X by the sets of members of E through a point; for every k the set {x:sE(x)≥k} is a finite union of closed subsets, hence closed. (3) A finite lexicographic tuple of upper semicontinuous functions with locally finite ranges is upper semicontinuous; so is a finite maximum of such tuples. Application to the piecewise-defined resolution invariants requires the branchwise induction in the canonical-resolution proposition below (source Proposition 3.0.8), rather than following merely from a definition.

Facts & Assumptions

Given: A smooth K-scheme X, a nonzero coherent ideal sheaf I⊆OX, and a finite family E of divisors in simultaneous SNC position.

[F1]

Iterated derivative ideals preserve support in the safe characteristic range: for every k≥1, supp⁡(I,k)={x:ord⁡xI≥k} is closed over the perfect field K in every characteristic. In characteristic zero or characteristic p>k, it equals V(Dk−1(I)); the endpoint case p=k is proved directly in step 1.1.

[F2]

Simple normal crossings divisors and simultaneous normal crossings position: at every point p the components of the members of E through p are cut out by pairwise distinct elements of a regular system of parameters of OX,p; in particular each member D of E has a zero locus that is closed, and only finitely many members pass through a given point.

[F3]

Upper semicontinuity for a function to a linearly ordered set means that {x:f(x)≥a} is closed for every threshold a in that order. In particular, for integer- or rational-valued functions it is not enough to check integer thresholds unless the range is known to lie in a discrete sublattice.

[F4]

Locally Noetherian and Noetherian schemes, Coherent module sheaves, Iterated derivative ideals preserve support in the safe characteristic range: on a quasi-compact Noetherian open, for any coherent ideal J the closed order-superlevel sets {x:ord⁡x(J)≥k} descend with k and therefore stabilize, by the perfect-field closedness in [F1] (the zero ideal gives the whole open at every level). The order consequently has only finitely many finite values there, together with the value +∞ on the stable intersection. No derivative-ideal equality is used for this bound.

[F6]

Finite lexicographic assembly preserves upper semicontinuity under the local finite-range bounds just established. For two coordinates f,g with finite local ranges, the lexicographic superlevel set at (a,b) is {f>a}∪({f=a}∩{g≥b}). The first set is a finite union of closed superlevel sets; any limit point of the second either has f>a or has f=a and g≥b, since both superlevel sets are closed. Induction gives the result for every finite tuple. A finite maximum of such tuples is upper semicontinuous because its superlevel set is the union of the component superlevel sets.

Proof

1.1F1F3

The order function is upper semicontinuous and the derivative formula holds in the stated range. The superlevel set for k is closed by [F1] in every characteristic. If char⁡K=0 or p>k, the formula with V(Dk−1(I)) is [F1]. For the remaining safe endpoint p=k, first suppose ord⁡x(I)≥k. Every coordinate derivative of order r≤k−1 lowers order by at most r, so Dk−1(I)x⊆mx and x∈V(Dk−1(I)). If instead j:=ord⁡x(I)<k, choose f∈Ix with order j and a nonzero monomial cUα in its degree-j initial form. Then ∣α∣=j≤k−1 and ∂αf has nonzero residue cα! in κ(x), since every factor in α! is less than p=k. This derivative lies in Dk−1(I)x, so that ideal is a unit at x and x∉V(Dk−1(I)). Thus the formula holds also for p=k, proving assertion (1).

1.2F2F3

The normal-crossings count is upper semicontinuous. For a finite set J of members of E the set {x:x∈D for all D∈J}=⋂D∈JD is closed by [F2], and {x:sE(x)≥k} is the finite union of these intersections over the k-element subsets J; hence it is closed, and the function is upper semicontinuous. On the locally closed stratum where exactly the members of J pass through the point, sE is constant equal to ∣J∣; these finitely many strata cover X locally because E is finite and its members have SNC, so only finitely many subsets occur locally at a point [F2]. This is assertion (2).

2.1F3F4F6algebra∎

Work on an open neighbourhood where f,g have finite ranges; the order functions satisfy this bound by [F4]. For a lexicographic threshold (a,b), the superlevel set is {f>a}∪({f≥a}∩{g≥b}). Since f has finite range locally, {f>a} is a finite union of closed superlevel sets, and the second set is closed by upper semicontinuity. Thus this union is closed. Induction gives the same result for a finite tuple; the superlevel set of a finite maximum is the finite union of the corresponding closed sets. This proves (3). It makes no claim that an unspecified piecewise assembly has satisfied these hypotheses.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Derivative ideals of a maximal-order marked ideal have maximal order

Statement

Assume μ≥1 and either char⁡K=0 or K perfect with char⁡K=p>μ (Field). If (I,E,μ) is a marked ideal of maximal order, then for every 0≤i≤μ, the derivative marked ideal Di(I,μ)=(Di(I),μ−i) is of maximal order (Derivative ideals of an ideal sheaf and of a marked ideal).

Facts & Assumptions

Given: A field K, a maximal-order marked ideal (I,E,μ) with μ≥1, and either char⁡K=0 or K perfect with char⁡K=p>μ.

[F1]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors: In characteristic zero or perfect characteristic p>ν, for a marking ν≥1, maximal order is equivalent to Dν(A)=OX.

[F2]

Derivative ideals of an ideal sheaf and of a marked ideal: Di(I,μ)=(Di(I),μ−i) and the recursive definition gives Da(Db(I))=Da+b(I) in every characteristic.

Proof

1.1F1F2given

By maximality and the safe-characteristic hypothesis, Dμ(I)=OX by [F1]. For 0≤i≤μ, the recursive identity in [F2] gives Dμ−i(Di(I))=Dμ(I)=OX.

2.1F1F2step 1.1∎

If μ−i≥1, the field remains in characteristic zero or has p>μ≥μ−i, so [F1] implies that Di(I,μ) is of maximal order. If i=μ, its residual marking is zero and its ideal is OX, which is maximal order directly. This proves the assertion for every i.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Derivative ideals under a multiple test blow-up

Statement

Let (I,μ) be a marked ideal on a smooth K-scheme and let (Xi)0≤i≤k be a multiple test blow-up with controlled transforms (Ii,μ) (Multiple test blow-ups, controlled transforms and resolutions of marked ideals). Then (Xi) is also a multiple test blow-up of the marked ideal Dj(I,μ) for every 0≤j≤μ, and for all i [Dj(I,μ)]i⊆Dj(Ii,μ). Indeed the centers of (Xi) lie in the derivative supports by the all-characteristic forward inclusion in Iterated derivative ideals preserve support in the safe characteristic range, and the transform inclusion follows by induction on i using Controlled derivative transforms are contained in derivatives of the controlled transform.

Facts & Assumptions

Given: A marked ideal (I,μ) on a smooth K-scheme and a multiple test blow-up (Xi)0≤i≤k with controlled transforms (Ii,μ), and 0≤j≤μ.

[F1]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals: each step blows up a regular center Ci⊆supp⁡(Ii,μ) in SNC position with Ei, or is an isomorphism; the controlled transform is σc(A,ν)=(I(D)−νσ∗A,ν).

[F2]

Iterated derivative ideals preserve support in the safe characteristic range: in every characteristic, supp⁡(Ii,μ)⊆supp⁡(Dj(Ii),μ−j) for 0≤j<μ; for j=μ the target marking is zero and its support is all of Xi.

[F3]

Controlled derivative transforms are contained in derivatives of the controlled transform: for the single blow-up σi+1 with center in the support, σi+1c(Dj(A),μA−j)⊆Dj(σi+1c(A,μA)), and Dj is monotone on inclusions.

[F4]

Controlled transforms are well defined, Derivative ideals of an ideal sheaf and of a marked ideal: the controlled transform and the derivative ideal are well-defined ideal sheaves, so inclusions can be checked locally.

Proof

1.1F1F2

The centers are admissible for the derivative ideal. Let 0≤j≤μ and let (Dj(I,μ))i denote the i-th controlled transform of the marked ideal Dj(I,μ) along the same sequence. We prove by induction on i that the sequence (Xi) is a multiple test blow-up of Dj(I,μ) and that (Dj(I,μ))i⊆Dj(Ii,μ). For i=0 this is equality. Assume it for i. By [F2] applied at stage i, Ci⊆supp⁡(Ii,μ)⊆supp⁡(Dj(Ii),μ−j), and the latter is contained in supp⁡((Dj(I,μ))i,μ−j) by the induction inclusion (a smaller ideal has a larger order-superlevel support), so Ci is an admissible center for the derivative marked ideal and step i+1 is defined for it.

2.1F3F4step 1.1∎

The transform inclusion. With the notation of step 1.1, (Dj(I,μ))i+1=σi+1c((Dj(I,μ))i)⊆σi+1c(Dj(Ii,μ))⊆Dj(σi+1c(Ii,μ))=Dj(Ii+1,μ), using the induction inclusion and the monotonicity of the controlled transform in step 1, and [F3] in step 2. This completes the induction and proves both assertions.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Controlled transforms preserve maximal order on nonempty transformed schemes

Statement

Let (I,E,μ) be a marked ideal of maximal order (Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors) and let C⊆supp⁡(I,E,μ) be a regular center with SNC with E, with blowup σ ⁣:X′→X (Multiple test blow-ups, controlled transforms and resolutions of marked ideals). Then the controlled transform has order at most μ at every point of X′. If X′≠∅, it is nonzero and hence again of maximal order. If X′=∅ (possible for μ=0 and a whole-space center), the order bound is vacuous; the definition requiring a nonzero ideal does not apply.

Facts & Assumptions

Given: A marked ideal (I,E,μ) of maximal order, a regular center C⊆supp⁡(I,E,μ) with SNC with E, and the blowup σ ⁣:X′→X.

[F1]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors: maximal order means ord⁡x(I)≤μ at every point; since C⊆supp⁡(I,μ), equality holds at every x∈C. The derivative characterization is restricted to characteristic zero or the safe range p>μ and is not used here.

[F2]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals: a test blow-up has a regular center contained in the support, is an isomorphism off that center, and has controlled transform I(D)−μσ∗I, locally y−μσ∗(I) for an exceptional equation y.

Proof

1.1F1F2

If μ=0, maximal order forces I=OX, so the controlled transform is OX′, of order zero at every point. On a nonempty X′ it is nonzero; on the empty scheme the asserted order bound is vacuous. Now assume μ≥1. Off the exceptional divisor the blow-up is an isomorphism, so the controlled transform has order at most μ there. It remains to check points over C.

2.1F1F2step 1.1∎

Fix c∈C and write the regular center locally as J=IC=(x1,…,xq) in regular coordinates transverse to C, with additional coordinates along C. Since every point of C has ord⁡c(I)=μ, I⊆Jμ near C by Controlled transforms are well defined. The normal associated-graded ring is a polynomial ring over OC, by Associated graded algebra of an ideal generated by a regular sequence. At each c, equality of the order gives some f∈I whose initial transverse form F∈Sym⁡μ(J/J2)⊗κ(c) is nonzero. In any blow-up chart over c with exceptional equation y=xj, the restriction of y−μσ∗(f) to the exceptional fiber is the nonzero dehomogenization of F, a polynomial of degree at most μ. Its order at any point of that fiber is at most μ. Indeed, for a nonzero polynomial P of total degree d≤μ, choose a nonzero top-degree monomial cZα. The Hasse derivative of index α is the nonzero constant c. Hasse derivatives extend to the local polynomial ring by substituting Z↦Z+T and inverting denominators as formal series. Their higher Leibniz rule sends mN into mN−∣α∣, because at most ∣α∣ factors of a product can receive positive derivative index. Thus P∈md+1 would force the unit c into m, a contradiction. This establishes the degree bound at every prime and in every characteristic; the order of the full local section is no greater than that of its restriction. Thus the controlled transform has order at most μ at every point over C, and is nonzero when X′≠∅, since the zero ideal at any point has infinite order. This proves the stated maximal-order conclusion with its empty-scheme qualification.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Elementary properties of the homogenized ideal

Statement

Let (I,μ) be a marked ideal of maximal order with μ≥1 and let H(I) be its homogenization (The homogenized ideal of a marked ideal of maximal order). Then: (1) if μ=1 then H(I)=I; (2) H(I)=I+D(I)T(I)+⋯+Dμ−1(I)T(I)μ−1+… agrees with its truncation at order μ−1 as an element of the equivalence class of (I,μ); (3) Assume AC. Then H(I,μ)=(I,μ)+D(I,μ)(T(I),1)+⋯+Dμ−1(I,μ)(T(I),1)μ−1 up to marked equivalence for the operation of Addition and multiplication of marked ideals; (4) if μ>1 and K has characteristic zero or perfect characteristic p>μ, then D(H(I,μ))⊆H(D(I),μ−1); (5) T(H(I))=T(I).

Facts & Assumptions

Given: A marked ideal (I,μ) of maximal order with μ≥1, with T(I)=Dμ−1(I) and homogenization H(I)=∑i=0μ−1Di(I)T(I)i.

[F1]

The homogenized ideal of a marked ideal of maximal order: H(I)=I+D(I)T(I)+⋯+Dμ−1(I)T(I)μ−1 and T(I)=Dμ−1I.

[F2]

Derivative ideals of an ideal sheaf and of a marked ideal: Di(I)⊇Di−1(I), Di+j(I)=Dj(Di(I)), and Di(A)⊆Di(B) for A⊆B.

[F3]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors: in characteristic zero or perfect characteristic p>μ, maximal order implies Dμ(I)=OX. In every characteristic D(T(I))=Dμ(I) is an ideal subsheaf of OX.

[F4]

Addition and multiplication of marked ideals: sums and products of marked ideals and their controlled transforms are computed componentwise as in that item.

[A1]

The Axiom of Choice: AC is used in clause (3) through the iterated marked-sum theorem in [F4].

Proof

1.1F1F2

Clauses (1) and (2). If μ=1 then T(I)=D0I=I and the defining sum has the single term I, so H(I)=I, which is (1). For (2), extend the defining sum to all i≥0. For each i≥μ, the i-th term Di(I)T(I)i is contained in T(I)i because Di(I)⊆OX, and hence is contained in T(I)μ. Since T(I)=Dμ−1(I), the i=μ−1 term is exactly T(I)μ. Thus every later term is contained in the last retained term, and the full sum equals its truncation.

1.2F1F2

Clause (5). For a local product generator of a summand Di(I)T(I)i, applying a coordinate derivative of total order at most μ−1 gives sums of products with a0 derivatives on the first factor and a1,…,ai derivatives on the i tangent factors, where a0+∑ℓaℓ≤μ−1. The first factor lies in Di+a0(I). If i+a0≤μ−1, this is contained in T(I)=Dμ−1(I) because derivative ideals increase with their index. If i+a0>μ−1, then ∑ℓaℓ≤μ−1−a0<i, so at least one tangent factor is undifferentiated and the product contains a factor of T(I). In both cases the resulting product lies in T(I), proving Dμ−1(H(I))⊆T(I). The reverse inclusion follows from I⊆H(I) and monotonicity of derivative ideals. Thus T(H(I))=T(I), proving (5).

1.3A1F1F4

Clause (3). Each product Di(I,μ)(T(I),1)i has underlying ideal Ji=Di(I)T(I)i and mark μ. Their literal ideal sum is H(I) with mark μ. Its support is the intersection of the supports of (Ji,μ), since the order of an ideal sum is the minimum of the summand orders. At a common admissible center the controlled transform of the literal sum distributes termwise, because all marks are μ. Induction therefore identifies its test sequences and induced supports with the simultaneous ones for the summands. Under AC, [F4] gives exactly those supports and test sequences for the iterated marked sum. Thus the displayed operation represents H(I,μ) up to marked equivalence, proving (3); no literal equality between a sum of ideal powers and a power of an ideal sum is used.

1.4F1F2F3step 1.2∎

Clause (4). Assume μ>1 and K has characteristic zero or perfect characteristic p>μ. The maximal-order criterion in [F3] gives Dμ(I)=OX, so (D(I),μ−1) is maximal order and T(D(I))=Dμ−2(D(I))=T(I). Leibniz gives D(H(I))⊆∑i=0μ−1Di+1(I)T(I)i+∑i=1μ−1Di(I)T(I)i−1D(T(I)). Since D(T(I))=Dμ(I)=OX, the second sum is the sum of the terms Di(I)T(I)i−1 for 1≤i≤μ−1. In the first sum, the terms with 1≤i+1≤μ−1 are precisely the terms Dj(D(I))T(I)j of H(D(I),μ−1) after setting j=i; its remaining top term is Dμ(I)T(I)μ−1=T(I)μ−1, already the i=μ−1 term of the second sum. The second sum itself consists of the terms Di−1(D(I))T(I)i−1 of H(D(I),μ−1). Hence the derivative ideal is contained in that homogenized ideal, proving (4).

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Homogenization commutes with smooth pullback

Statement

Assume AC (The Axiom of Choice).

Let φ ⁣:X′→X be a smooth morphism of smooth K-schemes and let (I,E,μ) be a marked ideal of maximal order with μ≥1 on X with ordered SNC exceptional family E (Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors). Then φ∗(H(I))=H(φ∗I) (The homogenized ideal of a marked ideal of maximal order).

Facts & Assumptions

Given: Assume AC. Let φ ⁣:X′→X be a smooth morphism of smooth K-schemes and let (I,E,μ) be a marked ideal of maximal order with μ≥1 on X.

[A1]

The Axiom of Choice: AC is assumed through the derivative-transport and order-preservation suppliers [F1] and [F2].

[F1]

Etale pullback commutes with derivative ideals: for an étale morphism ψ one has ψ∗Di(A)=Di(ψ∗A) for all i.

[F2]

Order and simultaneous normal crossings are preserved by smooth morphisms: for a smooth morphism φ the order is preserved, ord⁡x′(φ∗A)=ord⁡φ(x′)(A), The standard smooth local presentation in Flat maps with geometrically regular fibres have standard smooth local presentations factors a smooth germ locally as an étale morphism followed by a projection.

[F3]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors: maximal order means ord⁡x(I)≤μ at every point in every characteristic, and T(I)=Dμ−1(I).

[F4]

The homogenized ideal of a marked ideal of maximal order: H(I)=∑i=0μ−1Di(I)T(I)i, and pullback of ideal products and sums is computed termwise.

Proof

1.1A1F1F2

Derivative ideals commute with smooth pullback. A projection π ⁣:X×Ar→X satisfies π∗Di(A)=Di(π∗A) because differentiating a pulled-back function in the X-directions gives the pulled-back derivatives and the new Ar-coordinate derivatives annihilate the pulled-back generators of A. For an étale morphism this is [F1]. A general smooth germ factors locally as an étale morphism after a projection by [F2], so for smooth φ one has φ∗Di(A)=Di(φ∗A) for every coherent ideal A and every i≥0.

2.1A1F2F3step 1.1

Maximal order is preserved. By [F2, F3], at every x′∈X′ one has ord⁡x′(φ∗I)=ord⁡φ(x′)(I)≤μ, so the pullback is of maximal order in every characteristic. Step 1.1 also gives T(φ∗I)=Dμ−1(φ∗I)=φ∗T(I).

3.1A1F4step 1.1step 2.1∎

Homogenization commutes. Using step 1.1 termwise, φ∗H(I)=∑i=0μ−1φ∗(Di(I)T(I)i)=∑i=0μ−1Di(φ∗I)(φ∗T(I))i=∑i=0μ−1Di(φ∗I)T(φ∗I)i=H(φ∗I), where step 2.1 identified T(φ∗I)=φ∗T(I).

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

An automorphism of the completed local ring matching two tangent directions preserves the homogenization

Statement

Assume AC (The Axiom of Choice) and char⁡K=0 (Field).

Let (I,E,μ) be a marked ideal of maximal order on the smooth K-scheme X and let u,v∈T(I,μ)x=Dμ−1(I)x be tangent directions at x∈supp⁡(I,E,μ) that are transversal to E (Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors). Then there is an automorphism φ^uv of the completed local scheme X^x:=Spec⁡O^X,x (The I-adic completion of a module, Completion of a Noetherian local ring is local with the same residue field) such that: (1) φ^uv∗(HI^x)=(HI^x); (2) φ^uv∗(E)=E; (3) φ^uv∗(u)=v; (4) the formal support, defined here as supp⁡(I^,μ):=V(T(I)R) is contained in the fixed-point set of φ^uv.

Facts & Assumptions

Given: Assume char⁡K=0. Let (I,E,μ) be a marked ideal of maximal order on the smooth K-scheme X, let x∈supp⁡(I,E,μ), let u,v∈T(I,μ)x=Dμ−1(I)x be tangent directions transversal to E, and let R=O^X,x with X^x=Spec⁡R.

[F1]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors, embedding dimension and regular local ring: T(I)=Dμ−1(I); u,v∈T(I)x have multiplicity one; since X is smooth at x the local ring is regular and u,v are each part of a regular system of parameters; transversality to E means the parameters can be chosen compatible with the local equations of the members of E through x.

[F2]

Completion of a Noetherian local ring is local with the same residue field, completion preserves regular local rings: R is a Noetherian regular local ring, faithful flat over OX,x, with maximal ideal mR; I^=IR, and the completed tangent ideal is T^=T(I)R.

[F3]

The homogenized ideal of a marked ideal of maximal order: H(I)=∑i=0μ−1Di(I)T(I)i, so its completion is H(I)^=∑iDi(I)^ T^ i.

[F4]

Field, Derivation of an algebra, Derivative ideals of an ideal sheaf and of a marked ideal: Write Di(I)^=Di(I)R; formal parameter derivatives fixing the coefficient field send this completed ideal into Di+1(I)R. To justify this, restrict a formal parameter derivation to OX,x. It is a K-derivation into R, and the finite-free differential module identifies it with an R-linear combination of the algebraic K-derivations (Differentials of a smooth morphism, Derivations are maps out of Ω). Leibniz also differentiates the multiplying coefficients in R, giving terms already in the lower derivative ideal. This supplies the Taylor containment without identifying the full algebraic differential module of a power-series ring with a finite module. In characteristic zero, for a continuous coordinate substitution uj↦uj+δj with all δj in an ideal T, formal Taylor expansion gives φ∗(f)=∑α∈Nn1α!(∂αf) δα, convergently in the maximal-adic topology; if f∈Di(I), then ∂αf∈Di+∣α∣(I).

[F5]

Simple normal crossings divisors and simultaneous normal crossings position, Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors: the local equations of the components of E through x span a subspace W⊆m/m2. Transversality says that each of u and v is independent of W, so the boundary equations together with either direction can be completed separately to a regular system of parameters; no common complementary parameters are asserted.

[F6]

Transcendental residue elements adjoin across a maximal subfield, Separable residue elements adjoin across a maximal subfield, A complete equicharacteristic Noetherian local ring is a power-series quotient, regular local rings are domains and cohen macaulay: R is a complete equicharacteristic regular local ring, and smoothness in characteristic zero makes κ(x)/K a finitely generated separably generated extension. Starting from the image of K in R, lift a separating transcendence basis and then the finite separable algebraic generators by these field-adjunction lemmas; this gives a coefficient field k⊂R containing K. The continuous parameter map k⟦U1,…,Un⟧→R is surjective by the Cohen presentation. Its kernel is zero: the regular-local associated-graded theorem identifies the graded map on the parameter classes with an isomorphism, and a nonzero series in the kernel would have a least nonzero homogeneous term mapping to zero, a contradiction. Thus any regular parameter system identifies R with a formal power-series ring, where substitution by another parameter system with invertible cotangent matrix is a continuous K-algebra automorphism.

Proof

1.1F1F5F6

Construction of the automorphism. Put V=m/m2, let U be the image of T in V, and let W⊆V be spanned by the local equations of the components of E through x. The classes uˉ,vˉ lie in U∖W. Choose a decomposition V=U⊕C with W=(W∩U)⊕WC, and choose AU∈GL⁡(U) fixing W∩U pointwise and sending uˉ to vˉ; this is possible because both classes are nonzero modulo W∩U. Then A=AU⊕idC fixes W pointwise and (A−id)(V)⊆U. Choose a regular system of parameters u=u1,u2,…,un with the boundary equations among u2,…,un. Set v1=v; keep each boundary parameter unchanged; for every other j choose δj∈T lifting A(uˉj)−uˉj and put vj=uj+δj. Then v1,…,vn is a regular system of parameters, each δj∈T, and all boundary equations are fixed. By [F6], substitution uj↦vj defines a continuous K-algebra automorphism φ^uv of R. It sends u to v, preserves E, and induces the identity on R/T. This proves (2), (3), and the congruence used below.

1.2F3F4

The homogenization is preserved. The substitution induces the identity modulo T^, so it sends T^ into itself. This inclusion is an equality: the ascending chain T^⊆φ^−1(T^)⊆φ^−2(T^)⊆⋯ stabilizes in the Noetherian ring R, and applying a suitable power of φ^ gives φ^(T^)=T^. For f∈Di(I)R and t∈T^ i, the degree-s Taylor terms of φ^(f) lie in Di+s(I)R T^ s by [F4]. Multiplication by φ^(t)∈T^ i puts them in Di+s(I)R T^ i+s. If i+s<μ, this is a summand of H(I)R; otherwise it lies in T^ μ, the last retained summand. The ideal H(I)R is closed in the maximal-adic topology, since completion of a finite quotient is the quotient of the completion (Completion commutes with finite quotients and induced submodules). Thus the convergent Taylor sum stays in H(I)R, proving φ^(H)⊆H. Its inverse also induces the identity modulo T^ and has every coordinate increment in T^; the same argument gives φ^−1(H)⊆H. Applying φ^ to this inclusion proves the reverse containment, hence (1).

2.1F1F2F6∎

Fixed points on the support. Every coordinate increment δj lies in T^, and the automorphism fixes the coefficient field. Thus for every prime p⊇T^, it induces the identity on R/p; every point of V(T^)=supp⁡(I^,μ) is fixed. This is assertion (4).

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

The coefficient ideal commutes with smooth pullback

Statement

Assume AC (The Axiom of Choice).

Let φ ⁣:X′→X be a smooth morphism of smooth K-schemes and let (I,E,μ) be a marked ideal of maximal order with μ≥1 on X (The coefficient ideal of a marked ideal of maximal order). Then φ∗(C(I,μ))=C(φ∗I,μ).

Facts & Assumptions

Given: Assume AC. Let φ ⁣:X′→X be a smooth morphism of smooth K-schemes and let (I,E,μ) be a marked ideal of maximal order with μ≥1 on X.

[A1]

The Axiom of Choice: AC is assumed through the smooth/étale supplier [F2] and the marked-sum supplier [F3].

[F1]

The coefficient ideal of a marked ideal of maximal order: C(I,μ)=∑i=0μ−1(Di(I),μ−i), the sum of marked ideals.

[F2]

Etale pullback commutes with derivative ideals, Order and simultaneous normal crossings are preserved by smooth morphisms: derivative ideals commute with étale pullback, Flat maps with geometrically regular fibres have standard smooth local presentations factors smooth germs locally as étale after a projection, and smooth pullback preserves orders; hence φ∗Di(I)=Di(φ∗I) for all i and φ∗(I,μ) is of maximal order with the same μ.

[F3]

Addition and multiplication of marked ideals: pullback of a sum of marked ideals is the sum of the pullbacks, since the operation is defined by ideal sums, products and orders, all of which are compatible with inverse image.

Proof

1.1A1F1F2

Termwise comparison. For a projection, derivatives in the base directions pull back, and derivatives in the new variables annihilate the pulled-back generators; Leibniz shows that derivatives of their variable-coefficient multiples give no additional generators. The étale comparison and local factorization in [F2] therefore give derivative-ideal equality for every smooth morphism. By [F2] the pullback of the i-th summand is φ∗(Di(I),μ−i)=(Di(φ∗I),μ−i), and these are exactly the summands of C(φ∗I,μ).

2.1A1F1F3step 1.1∎

Summing the summands. Pullback of a sum of marked ideals is the sum of the pullbacks by [F3], so summing the identity of step 1.1 over 0≤i≤μ−1 gives φ∗C(I,μ)=∑i(Di(φ∗I),μ−i)=C(φ∗I,μ), which is the assertion.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Giraud's tangent-direction lemma

Statement

Let (I,E,μ) be a marked ideal of maximal order on the smooth K-scheme X, let C⊊supp⁡(I,E,μ) be a regular center with SNC with E, and let u∈T(I)(U)=Dμ−1(I)(U) be a tangent direction of multiplicity one on an open U⊆X (Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors). Let σ ⁣:X′→X be the blowup of C with exceptional divisor D, and on an open U′⊆σ−1(U) where D has local equation y put u′:=y−1σ∗(u). Then: (1) u′∈Dμ−1(σc(I∣U,μ)∣U′), i.e. u′ is a tangent direction of the controlled transform; (2) u′ is of multiplicity one on U′; (3) V(u′) is the strict transform of V(u) restricted to U′ (Strict transform of a closed subscheme).

Facts & Assumptions

Given: A marked ideal (I,E,μ) of maximal order on the smooth K-scheme X, a regular center C⊊supp⁡(I,E,μ) with SNC with E, a tangent direction u∈T(I)(U)=Dμ−1(I)(U) of multiplicity one on an open U⊆X, the blowup σ ⁣:X′→X with exceptional divisor D, and an open U′⊆σ−1(U) on which D has local equation y, with u′=y−1σ∗(u).

[F1]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors, Iterated derivative ideals preserve support in the safe characteristic range: T(I)=Dμ−1(I), and the all-characteristic inclusion supp⁡(I,μ)⊆supp⁡(Dμ−1(I),1) implies every section of T(I) vanishes on the support; a tangent direction of multiplicity one is a section u∈T(I)(U) with ord⁡x(u)=1 for every x∈V(u).

[F2]

Controlled derivative transforms are contained in derivatives of the controlled transform: for 0≤r≤μ, σc(Dr(I),μ−r)⊆Dr(σc(I,μ)).

[F3]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals, Exceptional subscheme of a blowup: a controlled transform is computed by f↦y−(exponent)σ∗(f); the exceptional divisor D has invertible ideal generated by y.

[F4]

Order of an ideal sheaf at a point: multiplicity one at x means ord⁡x(u)=1, i.e. u∈mx∖mx2; in a regular local ring this means that u may be taken as the first member of a regular system of parameters.

[F5]

Strict transform of a closed subscheme, Blowup of a scheme along an ideal sheaf: the strict transform of V(u) is cut out on U′ by the saturation (σ∗(u):y∞); since u∈IC (it vanishes on C⊆V(u)) and ord⁡ considerations show y divides σ∗(u) exactly once in suitable charts, (σ∗(u):y∞)=(σ∗(u)/y)=(u′).

Proof

1.1F1F2F3

u′ is a tangent direction of the controlled transform. By [F3] the section u′=y−1σ∗(u) is the controlled transform of the section u of the marked ideal (Dμ−1(I),1); since the generator u lies in Dμ−1(I)(U) and the controlled transform of a generated ideal is generated by the controlled transforms of generators, u′∈σc(Dμ−1(I),1)(U′). By [F2] with r=μ−1 and the hypothesis C⊆supp⁡(I,μ), this ideal is contained in Dμ−1(σc(I,μ))(U′); hence u′ is a tangent direction of the controlled transform, which is (1).

1.2F1F4F5

Multiplicity one. Away from D, the map is an isomorphism and y is a unit, so a zero of u′ has the same order as the corresponding zero of u, namely one. At a zero of u′ on D, its image x lies in C. By [F1], C⊆V(u), and the class of u in mx/mx2 is nonzero by [F4]; hence u can be chosen as one of the regular parameters generating IC. Since u′=0, the point is on a blowup chart whose exceptional parameter y is a different normal parameter. There u′=u/y is a chart coordinate, so [F4] gives order one. Thus u′ has multiplicity one on U′, which is assertion (2).

2.1F5step 1.2∎

The zero locus is the strict transform. By [F5] the strict transform of V(u) on U′ is cut out by (σ∗(u):y∞)=(u′); hence V(u′) is the strict transform of V(u) restricted to U′, which is assertion (3).

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

The coefficient ideal is equivalent to the marked ideal

Statement

Assume AC (The Axiom of Choice).

Let (I,μ) be a marked ideal of maximal order with μ≥1 (The coefficient ideal of a marked ideal of maximal order). Then C(I,μ)≃(I,μ) in the sense of Equivalence of marked ideals.

Facts & Assumptions

Given: Assume AC. Let (I,E,μ) be a marked ideal of maximal order with μ≥1 and its coefficient ideal C(I,μ)=∑i=0μ−1(Di(I),μ−i).

[A1]

The Axiom of Choice: AC is used through the marked-sum support and test-sequence assertion [F2].

[F1]

The coefficient ideal of a marked ideal of maximal order: C(I,μ) is the sum of the marked ideals (DiI,μ−i) for 0≤i≤μ−1 in the sense of the addition operation.

[F2]

Addition and multiplication of marked ideals: the multiple test blow-ups of a sum are exactly the simultaneous multiple test blow-ups of its summands, and the controlled transforms of the sum are the sums of the controlled transforms.

[F3]

Derivative ideals under a multiple test blow-up: every multiple test blow-up of (I,μ) is a multiple test blow-up of each Di(I,μ), with [Di(I,μ)]k⊆Di(Ik,μ).

[F4]

Iterated derivative ideals preserve support in the safe characteristic range: in every characteristic and at every stage k, supp⁡(Ik,μ)⊆supp⁡(Di(Ik),μ−i) for 0≤i<μ.

[F5]

Equivalence of marked ideals: equivalence means equal E-data, equal supports and equal multiple test blow-ups with equal induced supports.

Proof

1.1A1F1F2F3

The multiple test blow-ups coincide. By [F2], a multiple test blow-up of the coefficient sum is a simultaneous multiple test blow-up of its summands; since the i=0 summand is (I,μ), every such sequence is a multiple test blow-up of (I,μ). Conversely, [F3] shows that every multiple test blow-up of (I,μ) is a multiple test blow-up of every derivative summand, hence of their sum. Thus the two families coincide in every characteristic.

2.1A1F2F3F4F5step 1.1∎

Equal supports at every stage. At any stage k, [F4] shows that the support of (Ik,μ) is contained in the support of every derivative summand. Their intersection therefore contains supp⁡(Ik,μ), using the transformed-derivative inclusion in [F3]; the reverse inclusion follows because the i=0 summand is exactly (Ik,μ). By [F2], this intersection is the support of the coefficient sum's controlled transform. Hence the two supports agree at every stage, and together with step 1.1 and [F5] this proves C(I,μ)≃(I,μ) in every characteristic.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

The homogenized ideal is equivalent to the marked ideal

Statement

Let (I,μ) be a marked ideal of maximal order with μ≥1 (Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors). Then: (1) (I,μ)≃(H(I),μ) in the sense of Equivalence of marked ideals; (2) Assume AC. For every multiple test blow-up (Xk) of (I,μ), the controlled transform H(I,μ)k is equivalent to the iterated marked sum (I,μ)k+[D(I,μ)]k⋅[(T(I),1)]k+⋯+[Dμ−1(I,μ)]k⋅[(T(I),1)]kμ−1 of Addition and multiplication of marked ideals. Its underlying ideal with mark μ is the literal sum of the controlled transforms of the homogenization summands.

Facts & Assumptions

Given: A marked ideal (I,μ) of maximal order with μ≥1, its tangent ideal T(I)=Dμ−1(I) and its homogenization H(I).

[F1]

The homogenized ideal of a marked ideal of maximal order: H(I)=∑i=0μ−1Di(I)T(I)i, where T(I)=Dμ−1(I).

[F2]

Marked ideals and their support, Order of an ideal sheaf at a point, Iterated derivative ideals preserve support in the safe characteristic range: in every characteristic, supp⁡(I,μ)⊆supp⁡(Di(I),μ−i) for 0≤i<μ, and supp⁡(I,μ)⊆supp⁡(T(I),1).

[F3]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals, Addition and multiplication of marked ideals: the homogenized ideal is a literal sum of ideals with common mark μ; the order of an ideal sum is the minimum of the summand orders, and for a common admissible sequence the controlled transform distributes over that sum and over products of the marked factors.

[F4]

Derivative ideals under a multiple test blow-up: every multiple test blow-up of (I,μ) is also a multiple test blow-up of each (Di(I),μ−i) and (T(I),1), with [Di(I,μ)]k⊆Di(Ik,μ).

[F5]

Equivalence of marked ideals, Multiple test blow-ups, controlled transforms and resolutions of marked ideals: equivalence requires equal supports and the same multiple test blow-ups, with equal induced supports at every stage.

[A1]

The Axiom of Choice: AC is required in assertion (2) for the iterated marked-sum support and test-sequence route.

Proof

1.1F1F2F3F4

Equality of supports at every stage. Fix a multiple test blow-up (Xk) of (I,μ). By [F4], each derivative factor and the tangent ideal have controlled transforms along this same sequence. At every stage, [F2] gives supp⁡(Ik,μ)⊆supp⁡(Di(Ik),μ−i) for i<μ, and also supp⁡(Ik,μ)⊆supp⁡(T(Ik),1). By [F4], the actual transformed derivative and tangent factors are contained in the corresponding derivative ideals of Ik; their product is thus contained in Di(Ik)T(Ik)i. The product order inequality therefore puts every point of supp⁡(Ik,μ) in the support of each transformed summand (Di(Ik)T(Ik)i,μ). The support of the literal sum is the intersection of the summand supports: for ideals Ji, ord⁡x(∑iJi)=min⁡iord⁡x(Ji). Hence supp⁡(Ik,μ)⊆supp⁡(H(I,μ)k). Conversely, the first summand of H is I, so Ik⊆H(I,μ)k and supp⁡(H(I,μ)k)⊆supp⁡(Ik,μ). This proves equality of supports at every stage.

2.1A1F3F4F5step 1.1∎

Equivalence and transform decomposition. Equality of supports at every stage in step 1.1 shows that the two marked ideals have the same admissible centers and the same induced supports along every multiple test blow-up. By [F5] they are equivalent, proving (1). For (2), the controlled transform of the literal ideal sum defining H(I) distributes termwise because all summands have mark μ; each transformed summand is the product of the transforms of (Di(I),μ−i) and (T(I),1)i by [F3, F4]. Thus its underlying ideal with mark μ is the literal sum displayed in the Statement. Under AC, the iterated marked sum has the intersection of the transformed summand supports and exactly their simultaneous test sequences by [F3]. The literal sum with common mark μ has that same support, and its transforms continue to distribute termwise at every common admissible center. Induction gives equal induced supports and the same test sequences, hence equivalence by [F5], proving (2).

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

The supports of a multiple test blow-up stay inside the strict transforms of a hypersurface of maximal contact

Statement

Let u∈T(I)(U) be a tangent direction of the maximal-order marked ideal (I,E,μ) on U (Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors) and let (Ui) be any multiple test blow-up of (I∣U,μ) (Multiple test blow-ups, controlled transforms and resolutions of marked ideals). Then for every i the support of the induced marked ideal is contained in the strict transform V(u)i of the hypersurface V(u): supp⁡(Ii,Ei,μ)⊆V(u)i. If u is transversal to E, so that the restricted boundary on V(u) is SNC, the ambient sequence induces a multiple test blow-up of the restricted marked ideal on V(u). Restriction is asserted under this boundary hypothesis; the support containment above does not require it.

Facts & Assumptions

Given: A marked ideal (I,μ) of maximal order on the smooth K-scheme X, a tangent direction u∈T(I)(U) of multiplicity one on an open U, and a multiple test blow-up (Ui) of (I∣U,μ) with controlled transforms (Ii,μ).

[F1]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors, Iterated derivative ideals preserve support in the safe characteristic range: T(I)=Dμ−1(I), and in every characteristic supp⁡(I,μ)⊆supp⁡(Dμ−1(I),1); hence every section u∈T(I) vanishes on supp⁡(I,μ). A tangent direction is such a section of multiplicity one.

[F2]

Giraud's tangent-direction lemma: along one blow-up of a center C⊊supp⁡(I,μ), the controlled transform u′=y−1σ∗(u) is again a tangent direction of multiplicity one with V(u′) the strict transform of V(u); iterating, at every stage i the section ui is a tangent direction of (Ii,μ) and V(ui) is the strict transform V(u)i.

[F3]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals, Strict transform of a closed subscheme: the multiple test blow-up is a sequence of blow-ups of regular centers with SNC with E, with controlled transforms of the marked ideal; strict transforms of divisors are defined by saturation and commute with the iteration.

[F4]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals, Strict transform of a closed subscheme: if a regular center C⊆V(u) of codimension at least two has local equations u,t2,…,tc with u a regular parameter, then on the blow-up chart indexed by u the strict transform of V(u) is empty, while on a chart indexed by tj it is cut out by u/tj. This follows by saturating the chart pullback of u, which is respectively the exceptional equation or tj(u/tj). A Cartier center has isomorphic blow-up.

[F5]

Controlled derivative transforms are contained in derivatives of the controlled transform: if C⊆supp⁡(I,μ) is any regular center with SNC boundary, then for every 0≤r≤μ the controlled transform of (Dr(I),μ−r) is contained in Dr of the controlled transform of (I,μ). The supplier's chartwise chain-rule proof applies to the non-strict containment, including C=supp⁡(I,μ).

Proof

1.1F1

Base case. By [F1], each point of supp⁡(I,μ) lies in supp⁡(Dμ−1(I),1), so the section u∈Dμ−1(I)(U) vanishes there. Hence supp⁡(I,μ)⊆V(u), which is the case i=0.

2.1F1F2F3F4F5step 1.1∎

Inductive step. Assume supp⁡(Ii,μ)⊆V(ui). The next center satisfies Ci⊆supp⁡(Ii,μ)⊆V(ui). If Ci is Cartier, work locally near it with a regular centre equation y. Since Ci is reduced and ui vanishes on it, ui=ay; multiplicity one of ui along the centre makes a a unit there. Although the blow-up is an isomorphism, the controlled transform divides by y: [F5] gives ui/y=a∈Dμ−1(Ii+1). Thus this derivative ideal is the unit ideal locally along the centre, and [F1] makes the new support empty. The strict transform of V(ui) is also empty there because V(ui)=Ci locally. Away from the centre, both transforms preserve the prior containment. If Ci is not Cartier and is a proper subset of the support, [F2] gives a tangent direction ui+1 whose zero scheme is the strict transform V(u)i+1. In the remaining case, work componentwise where Ci has codimension at least two. Since ui vanishes on Ci and has order one, it is one of local regular parameters generating the ideal of Ci. On the chart indexed by ui, the controlled transform ui+1=ui/y is 1, and [F4] says the strict transform of V(ui) is empty there. By [F5] with r=μ−1, the controlled transform of (Dμ−1(Ii),1) is contained in Dμ−1(Ii+1). Since ui∈Dμ−1(Ii), its controlled transform ui+1=y−1σi∗(ui) lies in that derivative ideal; [F5] applies because Ci⊆supp⁡(Ii,μ), including when equality holds. On the chart indexed by ui, y=ui, so ui+1=1; the derivative ideal is the unit ideal, and [F1] makes the transformed support empty there, as required by the empty strict-transform chart in [F4]. On a chart indexed by another center parameter tj, y=tj and ui+1=ui/tj is a chart coordinate; [F5] places it in Dμ−1(Ii+1), so [F1] puts the support in V(ui+1), the strict transform by [F4]. This proves the containment for this center as well. Applying [F1] at stage i+1 closes the induction for all stages. For the restricted-sequence conclusion, assume u transversal to the initial boundary. The distinct equations u and the boundary parameters then form part of one regular parameter system. Since each regular center lies in V(ui) and has SNC with the boundary, its ideal can be generated by ui and further parameters compatible with the boundary restrictions. The nonempty charts in [F4] restricted to V(ui) are precisely the charts of its center blowup, the restricted exceptional equation is the same y, and division by yμ commutes with restriction. These charts preserve SNC of the restricted boundary; where a component is deleted there is no stalk to check. Thus the restricted ideals and boundaries form the asserted multiple test blow-up.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

The coefficient ideal controls the support after restriction

Statement

Assume AC (The Axiom of Choice), μ≥1, and either char⁡K=0 or K perfect with char⁡K=p>μ (Field). Let (I,E,μ) be a marked ideal of maximal order on the smooth K-scheme X and let S⊆X be a regular closed subscheme with SNC with E and not contained in supp⁡(I,μ) (Marked ideals and their support). Then supp⁡(I,μ)∩S=supp⁡(C(I,μ)∣S) (the restriction of the coefficient ideal, The coefficient ideal of a marked ideal of maximal order). Moreover, if (Xi) is a multiple test blow-up of (I,μ) whose centers Ci are contained in the strict transforms Si of S (or disjoint from them), then the restrictions σi∣Si define a multiple test blow-up (Si) of C(I,μ)∣S, and supp⁡(Ii,μ)∩Si=supp⁡[C(I,μ)∣S]i; conversely every multiple test blow-up of C(I,μ)∣S is induced by one of (I,μ) with centers in the strict transforms of S.

Facts & Assumptions

Given: Assume AC. Let K be a field, let μ≥1 with char⁡K=0 or K perfect with char⁡K=p>μ, let (I,E,μ) be a maximal-order marked ideal whose support does not contain a regular closed subscheme S⊆X having SNC with E, and let (Xi) be a multiple test blow-up with centers contained in the strict transforms Si of S.

[A1]

The Axiom of Choice: AC is used through the coefficient-equivalence and marked-sum support suppliers [F1] and [F2].

[F1]

The coefficient ideal of a marked ideal of maximal order: C(I,μ)=∑i=0μ−1(Di(I),μ−i); by Addition and multiplication of marked ideals this is the sum operation, so its support is the intersection of the summands' supports and its controlled transforms are the sums of the controlled transforms.

[F2]
[F3]

Restriction of a marked ideal to a smooth subvariety and its blow-ups: supp⁡(I,μ)∩S⊆supp⁡((I,μ)∣S) and, along a multiple test blow-up with centers in the Si, [(I,μ)∣S]i=(Ii,μ)∣Si and σc((Ii,μ)∣Si)=(σc(Ii,μ))∣Si+1.

[F4]

Order of an ideal sheaf at a point, Derivative ideals of an ideal sheaf and of a marked ideal: in local coordinates x1,…,xk defining S and y1,…,yn−k along it, a local section f=∑αcαf(y)xα has cαf∣S=1α!∂αf∣S∈D∣α∣(I)∣S; hence x∈supp⁡(I,μ)∩S if and only if ord⁡x(cαf∣S)≥μ−∣α∣ for all f and all ∣α∣≤μ.

[F5]

Taylor expansions can be taken after faithful flat completion using the regular-local completed-parameter construction in Iterated derivative ideals preserve support in the safe characteristic range, [F4]. The coefficients of transverse degree ∣α∣<μ are ∂xαf/α! restricted to S, so they belong to D∣α∣(I)∣S. Only factorials with ∣α∣<μ<p occur in positive characteristic.

Proof

1.1A1F1F2F3F4

The two inclusions at the initial stage. The inclusion supp⁡(I,μ)∩S=supp⁡(C(I,μ))∩S⊆supp⁡(C(I,μ)∣S) is [F2, F3]; note that the summands of C(I,μ) have supports containing supp⁡(I,μ), and the restriction can only raise orders. Conversely, if x∈supp⁡(C(I,μ)∣S), then ord⁡x(Di(I)∣S)≥μ−i for all i≤μ−1, so in the notation of [F4] every coefficient cαf∣S has order at least μ−∣α∣; reading the Taylor development of f along S gives ord⁡x(f)≥μ for every local section, i.e. x∈supp⁡(I,μ)∩S. Hence the supports agree.

2.1A1F1F2F3F4F5step 1.1algebra

Track the coefficients of the original generators throughout the sequence. Let Jr,i be the transform on Si of (Dr(I)∣S,μ−r), and write a transformed generator as fi=∑αcα,ixiα in completed adapted coordinates. Initially cα,0∈J∣α∣,0 by [F5]. In a nonempty restricted blowup chart with exceptional equation a, xi+1=xi/a and fi+1=a−μσ∗fi, hence cα,i+1=a−(μ−∣α∣)σ∗cα,i. This is exactly the transform with mark μ−∣α∣, so the coefficient membership persists. At each stage keep the normal generators xi fixed while adapting the center by a change of parameters along Si in the completed coefficient ring. This is possible because the center is a regular subscheme of Si. Such a change transports the coefficient ideals and does not alter the stated membership; no change mixing normal and tangent parameters is needed. A point in the support of the transformed coefficient sum lies in every Jr,i support by [F1]; the persistent coefficient membership therefore gives ord⁡(cα,i)≥μ−∣α∣, and the Taylor expansion gives ord⁡(fi)≥μ. This proves the inclusion from restricted coefficient support to ambient support. Conversely, [F2] gives equality of ambient supports for the transformed C(I,μ) and I; restriction raises order by [F3], so ambient support on Si lies in the restricted coefficient support.

3.1A1F3F4step 2.1∎

The converse direction. Conversely, let (Si) be a multiple test blow-up of C(I,μ)∣S with centers Di⊆Si. Lifting the center Di to Xi and blowing up Xi there is legitimate because Di⊆supp⁡[C(I,μ)∣S]i=supp⁡(Ii,μ)∩Si⊆supp⁡(Ii,μ) by step 2.1, and its SNC position with the restricted boundary, together with the parameter equations defining Si, makes it SNC with the ambient boundary under the omission convention of [F3]; the blow-up of Xi at Di restricts to the blow-up of Si at Di along the strict transform, and the chart computation of [F4] read in this direction shows that the equality of supports of step 2.1 persists at every stage. Hence (Si) defines a multiple test blow-up (Xi) of (I,μ) with all centers contained in the strict transforms of S, which is clause (3).

LemmaStatement: AI-adaptedProof: AI-adaptedOpen item page →

Codimension-one components of a maximal-order support

Statement

Assume AC (The Axiom of Choice).

Let (I,E,μ) be a marked ideal of maximal order whose support supp⁡(I,E,μ) has a component of codimension one in the smooth K-scheme X (Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors). Every codimension-one component C of the support is regular and isolated from its other components. The labelled Cartier blowup at C has y−μσ∗I=OX′ near its exceptional divisor. Whenever C has SNC with E, this division is the controlled transform of an admissible multiple test blowup, and its support does not meet the exceptional divisor. Without that SNC hypothesis only the stated ideal-division calculation is asserted. More precisely, at a general point of C one has I=(uμ) for a local equation u of C, and if σ ⁣:X′→X is the blowup of C with exceptional divisor D given by y, then, on a neighbourhood of the inverse image of C, σ∗I=yμOX′ and hence the divided ideal is the unit ideal. No unit-ideal conclusion is asserted away from that neighbourhood. Blowing up a codimension-one regular center is an isomorphism, but it is a nontrivial transformation of the marked ideal because its pulled-back ideal is divided by the μ-th power of the exceptional ideal while the marking μ is unchanged.

Facts & Assumptions

Given: Assume AC. Let (I,E,μ) be a marked ideal of maximal order whose support has a codimension-one component C, and let σ ⁣:X′→X be the blowup of C.

[A1]

The Axiom of Choice: AC is used through the regular-local UFD supplier [F1].

[F1]

Regular local rings are unique factorization domains: every local ring of the smooth scheme X is a UFD under AC. Thus the height-one component prime at a point of C is generated by a prime element u.

[F2]

localisations of regular local rings are regular, one dimensional regular local rings are dvrs: the localization of a regular local ring at a height-one prime is a one-dimensional regular local ring, hence a DVR; its maximal ideal is generated by the local parameter u.

[F3]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors, Order of an ideal sheaf at a point: at a point of the marked support the order is at least μ, while maximal order gives an upper bound μ.

[F4]

Blowup of a scheme along an ideal sheaf, Effective cartier divisor: blowing up a Cartier divisor is an isomorphism and its exceptional ideal is generated locally by its equation y. If the center has SNC with E, Multiple test blow-ups, controlled transforms and resolutions of marked ideals makes the division y−μσ∗I the controlled transform of the marked ideal; otherwise we use this expression only as a local ideal quotient.

[F5]

embedding dimension and regular local ring, regular local quotient by parameter is regular: in a regular local ring, an element of order one is a regular parameter and its quotient is regular.

Proof

1.1A1F1F2F3F5

The local shape along the component. Fix x∈C, put R=OX,x, and let P=I(C)x be the height-one component prime. By [F1], P=(u) for a prime element u. Since C lies in the marked support and μ≥1, Ix⊆P. Let η be the generic point of C. By [F2], RP is a DVR with uniformizer u; since η lies in the support and I has maximal order, [F3] gives ord⁡RP(IP)=μ, so IP=(uμ). For any f∈Ix, this gives sf=uμg for some s∉P and g∈R. The element u is prime and does not divide s, so repeated cancellation gives f∈(uμ). Hence Ix=(uμ)J for an ideal J⊆R. If ord⁡x(u)≥2, then Ix⊆mx2μ⊆mxμ+1, contradicting maximal order at x. Thus ord⁡x(u)=1. If J were proper, then J⊆mx and Ix⊆mxμ+1, again a contradiction. Therefore Ix=(uμ). By [F5], u is a regular parameter and R/(u) is regular, so C is regular at x. Since x was arbitrary, C is regular, and locally along it the support is exactly V(u); hence no other support component meets C.

2.1F4step 1.1∎

The blowup is an isomorphism and resolves along the component. The regular component C is an effective Cartier divisor by step 1.1, so its blowup is an isomorphism. Locally along C, write y=u for its equation. Step 1.1 gives I=(uμ), hence σ∗I=(yμ) and y−μσ∗I=OX′ along the exceptional divisor. If C has SNC with E, this is an admissible marked-ideal transformation by [F4], so its controlled-transform support misses the divisor. The local quotient calculation itself needs no boundary hypothesis.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Glueing of homogenized ideals along etale neighbourhoods

Statement

Assume AC (The Axiom of Choice) and char⁡K=0 (Field).

Let (I,E,μ) be a marked ideal of maximal order on the smooth K-scheme X and let u,v∈T(I,μ) be tangent directions at x∈supp⁡(I,E,μ) transversal to E (Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors). Then there exist etale neighbourhoods φu,φv ⁣:X~→X of x (Étale morphism of schemes) with a common point x~, φu(x~)=φv(x~)=x, such that (1) φu∗(X,H(I),E,μ)=φv∗(X,H(I),E,μ); (2) φu∗(u)=φv∗(v); and, writing (X~,I~,E~,μ) for the common pullback: (3) for every y∈supp⁡(I~,E~,μ) one has φu(y)=φv(y); (4) for every multiple test blow-up (Xi) of (I,E,μ) the induced multiple test blow-ups φu∗(Xi) and φv∗(Xi) coincide (same centers), the induced homogenized marked ideals agree, and φu−1(V(u)i)=φv−1(V(v)i) for the strict transforms of the hypersurfaces of maximal contact.

Facts & Assumptions

Given: Assume AC and char⁡K=0. Let (I,E,μ) be a marked ideal of maximal order on the smooth K-scheme X, let x∈supp⁡(I,E,μ), and let u,v∈T(I,μ) at x be tangent directions transversal to E.

[A1]

The Axiom of Choice: AC is used through the completion automorphism and smooth-pullback suppliers [F3] and [F6].

[A2]

Field: The field K has characteristic zero, as required for the completed Taylor substitution in [F3].

[F1]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors: put V=mx/mx2, let U be the image of T(I)x in V, and let W be spanned by the local equations of E through x. The classes of u,v lie in U∖W. Choose V=U⊕C with W=(W∩U)⊕WC, take AU∈GL⁡(U) fixing W∩U pointwise and sending uˉ to vˉ, and set A=AU⊕idC. Then (A−1)V⊆U and A fixes W. Starting with u1=u, choose u2,…,ud completing the boundary equations to parameters; lift A(uˉi)−uˉi to δi∈T(I)x, taking δ1=v−u and δi=0 on boundary parameters. The functions vi=ui+δi form a second parameter system with v1=v, all boundary equations unchanged, and vi−ui∈T(I)x. Shrink the neighborhood so transversality persists at all support points.

[F2]

Relative Jacobian criterion with its presentation hypothesis, Standard smooth presentations and locally standard smooth maps, Étale equals flat and unramified in finite presentation, Étale morphism of schemes: a smooth chart at x supplies common residue-field coordinates together with local parameters as a full coordinate system to affine space. Keeping the residue-field coordinates fixed, replacing the local parameters by another system with invertible cotangent matrix again gives an étale chart; this follows from the invertible Jacobian criterion. Étale pullback preserves orders of ideals.

[F3]

An automorphism of the completed local ring matching two tangent directions preserves the homogenization, proof 1.2: the Taylor argument applies to any continuous parameter substitution whose increments lie in the completed tangent ideal, which fixes a coefficient field containing K and preserves the boundary equations. Indeed a Taylor term of degree s from Di(I)Ti lies in Di+s(I)Ti+s; for i+s≥μ it lies in Tμ⊆H(I). These ideals are closed in the maximal-adic topology, so the convergent sum lies in H(I). The substitution is the identity modulo T, hence preserves T; its inverse has the same property, proving equality for H(I). This calculation, rather than the mere existence assertion of the supplier, applies to the specific systems constructed in [F1].

[F4]

Completion of a Noetherian local ring is local with the same residue field: the completion map is faithfully flat, so for ideals on a Noetherian local ring equality after completion implies equality; The I-adic completion of a module identifies the completed stalks.

[F5]

Elementary properties of the homogenized ideal, Etale pullback commutes with derivative ideals, Homogenization commutes with smooth pullback, The homogenized ideal is equivalent to the marked ideal: homogenization commutes with smooth (in particular étale) pullback and (I,μ)≃(H(I),μ), so supports and admissible centers may be computed with H. Since T(H(I))=T(I) and derivative ideals commute with étale pullback, equality of homogenized pullbacks also gives equality of the two pulled-back tangent ideals.

[F6]

Smooth base change of multiple test blow-ups, Multiple test blow-ups, controlled transforms and resolutions of marked ideals: smooth base change of a multiple test blow-up is a multiple test blow-up of the pulled-back marked ideal, with the pulled-back centers and exceptional families.

[F7]

Strict transform of a closed subscheme, Controlled transforms are well defined: strict transforms of the hypersurfaces V(u),V(v) are computed by saturation with the exceptional equations, and the controlled transform of a generator is well defined up to a unit.

[F8]

Étale equals flat and unramified in finite presentation, An unramified morphism has an open diagonal: étale chart maps are unramified and hence have open diagonals. Since fu and fv agree on S=V(T(I)), the diagonal section in each base change S×ANU is open. Shrink the fibre product around (x,x) so each preimage of S is exactly this diagonal there.

Proof

1.1A1A2F1F2F8

Separate étale charts. Extend the parameter systems of [F1] by the same residue-field coordinates from one smooth chart. By [F2] these full coordinate systems define étale maps fu,fv:U→AN; their values at x coincide, since the residue coordinates are common and all local parameters vanish at x. Form Y=U×ANU and let φu,φv:Y→U be its projections. The pair (x,x) gives a point x~. The fiber-product equations give φu∗(ui)=φv∗(vi) for every local parameter, in particular φu∗(u)=φv∗(v), proving clause (2). Because vi−ui∈T(I)x, the restrictions of fu and fv to S=V(T(I)) agree. By [F8] the diagonal section in each base change of fu or fv over S is open; shrink Y around (x,x) so both inverse images of S are those diagonals. The restrictions X~ of the two projections are the required étale neighbourhoods.

1.2A1A2F1F2F5F8

The charts agree on the support. By construction, every coordinate difference between fu and fv lies in T(I), so their restrictions to S=V(T(I)) agree scheme-theoretically. The chosen open parts of the fibre product over S are the diagonal sections; consequently either projection of a point in the pulled-back support lies in S exactly when the other does, and then the two projections are equal. Étale preservation of order identifies both pulled-back supports with this common locus. This proves clause (3).

2.1A1A2F1F3F4step 1.1

Equality of the homogenized marked ideals. At the distinguished point x~ over x, the completed étale charts identify the two completed stalks with the same formal power-series ring. The residue-field coordinates are fixed and the parameter relation is the substitution ui↦vi (or its inverse), with every increment in T by [F1]. The calculation in [F3] therefore proves equality of the completed pullbacks of H(I) for this particular substitution. By [F4] the stalks themselves are equal. Coherence now permits shrinking X~ about x~ so the two ideal sheaves agree: the two finite quotient modules measuring either failure of containment vanish on a neighbourhood of this point. Boundary equations are fixed, so the ordered boundaries agree there as well. This shrinking preserves the open-diagonal construction and proves clause (1). No claim that H(I) is the unit ideal outside the marked support is needed.

3.1A1A2F5F6step 1.1step 1.2step 2.1

At the initial stage the projections agree scheme-theoretically on V(T), by the open-diagonal construction in steps 1.1–1.2; consequently their differences on every local function lie in the common pulled-back tangent ideal T0. Let Ti denote the controlled transform of (T,1) along the common sequence. Derivative-transform inclusion gives Ti⊆Dμ−1(Ii), so the transformed marked support is contained in V(Ti). Inductively the projections agree on V(Ti), and hence their inverse images of any reduced center contained in the support have equal ideal sheaves. Flat pullback preserves the intersections with this locus, so both base changes are the blowup of that same center. Equality of the homogenized pullbacks in step 2.1 and the controlled-transform rule then give equality of the homogenized marked ideals at the next stage.

4.1A1A2F3F6F7step 3.1algebra∎

Here is the quotient-coordinate check needed to close the induction in step 3.1. Write aj=φu∗zj and bj=φv∗zj for adapted center parameters. Their differences lie in Ti. On a common blowup chart let e be its exceptional equation; by definition Ti+1=e−1TiO. At a point of V(Ti+1), the chart denominators am/e and bm/e have the same residue because (am−bm)/e∈Ti+1. Thus whenever one is a unit the other is a unit, and the same chart works for both projections. The difference of ratio coordinates is aj/am−bj/bm=((aj−bj)/e)/(bm/e)−(aj/am)((am−bm)/e)/(bm/e), which lies in Ti+1. For unscaled coordinates the old difference lies in eTi+1⊆Ti+1. Thus the lifted maps agree on V(Ti+1); at each new point their completed comparison still has parameter increments in Ti+1, and preserved boundary equations up to units. This proves the support-agreement induction, rather than inferring it only from agreement on the base. Initially the hypersurfaces have equal pullbacks by clause (2); identical blowups and saturation by the same exceptional ideal preserve that equality at every stage. These are all assertions in clause (4), for homogenized transforms as in the source's Glueing Lemma.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Coefficient-ideal control with centres allowed off the subvariety

Statement

Assume AC (The Axiom of Choice), μ≥1, and either char⁡K=0 or K perfect with char⁡K=p>μ (Field). In the setting of The coefficient ideal controls the support after restriction, let (Xi) be a multiple test blow-up of (I,μ) whose centers Ci are either contained in the strict transforms Si of S or disjoint from them, for every i. Then the restrictions σi∣Si define a multiple test blow-up (Si) of C(I,μ)∣S and supp⁡(Ii,μ)∩Si=supp⁡[C(I,μ)∣S]i for every i.

Facts & Assumptions

Given: Assume AC. Let K be a field and let μ≥1 with char⁡K=0 or K perfect with char⁡K=p>μ; let (I,E,μ) be a marked ideal of maximal order whose support does not contain a smooth subvariety S⊆X having SNC with E; and let (Xi) be a multiple test blow-up all of whose centers Ci are either contained in the strict transform Si of S or disjoint from it.

[A1]

The Axiom of Choice: AC is used through the restriction-support supplier [F1].

[F1]

The coefficient ideal controls the support after restriction, Field: under AC and the stated characteristic bound, for centers contained in the strict transforms, (Si) is a multiple test blow-up of C(I,μ)∣S and supp⁡(Ii,μ)∩Si=supp⁡[C(I,μ)∣S]i.

[F2]

Restriction of a marked ideal to a smooth subvariety and its blow-ups: the restriction of a controlled transform is the controlled transform of the restriction along the strict transform, and the restriction of the marked ideal does not change when the blow-up center is disjoint from S.

[F3]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals: a blow-up with center disjoint from Si restricts to an isomorphism over Si and does not modify the restricted marked ideal.

Proof

1.1A1F1F2F3

The restriction sequence is a multiple test blow-up. At each step, if Ci⊆Si then the restricted center is admissible for C(I,μ)∣S and the transform rule is [F2]; if Ci∩Si=∅ then the step restricts to an isomorphism over Si and does not change the restricted marked ideal by [F3]. Hence the restrictions of the morphisms define a multiple test blow-up (with isomorphism steps allowed) of C(I,μ)∣S.

2.1A1F1F2F3step 1.1∎

Equality of supports persists. For steps with centers contained in Si the equality is [F1]; for steps with centers disjoint from Si, both sides are unchanged: supp⁡(Ii+1,μ)∩Si+1=supp⁡(Ii,μ)∩Si because the blow-up is an isomorphism near Si and the strict transform of S is identified with S, and supp⁡[C(I,μ)∣S]i+1=supp⁡[C(I,μ)∣S]i by [F3]. Induction over the steps gives the asserted identity at every stage.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Refined maximal-contact statement via the coefficient ideal

Statement

Assume AC (The Axiom of Choice), inherited from the blowup and strict-transform suppliers throughout.

Let (I,∅,μ) be a marked ideal of maximal order with μ≥1 whose support has codimension at least two at a point x∈supp⁡(I,μ), and let U∋x be an open neighbourhood on which a tangent direction u∈T(I)(U) exists and supp⁡(I,μ)∩U has codimension at least two throughout U (Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors). Let V(u) be the regular hypersurface defined by u (Strict transform of a closed subscheme). For every multiple test blow-up (Ui) of (I∣U,μ), write Ii for its controlled transforms on Ui. Then (1) the support supp⁡(Ii,μ) is contained in the strict transform V(u)i as a proper subset. If K has characteristic zero or perfect characteristic p>μ (Field), then in addition: (2) the sequence (V(u)i) is a multiple test blow-up of C(I,μ)∣V(u); (3) supp⁡(Ii,μ)∩V(u)i=supp⁡[C(I,μ)∣V(u)]i; (4) every multiple test blow-up of C(I,μ)∣V(u) defines a multiple test blow-up of (I∣U,μ) with centers in the strict transforms of V(u).

Facts & Assumptions

Given: Assume AC throughout. A field K, a maximal-order marked ideal (I,∅,μ) with μ≥1 whose support has codimension at least 2 at a point x∈supp⁡(I,μ), an open neighbourhood U∋x admitting a tangent direction u∈T(I)(U) with supp⁡(I,μ)∩U of codimension at least 2, the regular hypersurface V(u), and a multiple test blow-up (Ui) of (I∣U,μ). For clauses (2)-(4), assume also the safe characteristic range of the Statement.

[F1]

The supports of a multiple test blow-up stay inside the strict transforms of a hypersurface of maximal contact: for every i the support supp⁡(Ii,μ) is contained in the strict transform V(u)i of V(u).

[A1]

The Axiom of Choice: AC is inherited for clause (1) through the blowup and strict-transform suppliers and for clauses (2)-(4) through the coefficient-restriction supplier [F2].

[F2]

The coefficient ideal controls the support after restriction: under AC and characteristic zero or perfect characteristic p>μ, for a regular S with SNC with E whose support is not contained in supp⁡(I,μ), the restrictions along centers in Si give supp⁡(Ii,μ)∩Si=supp⁡[C(I,μ)∣S]i, and every multiple test blow-up of the restriction defines one of (I,μ) with centers in the Si.

[F4]

Strict transform of a closed subscheme, Chain dimension and the empty-space convention: a hypersurface V(u) is of codimension one, so it is not contained in the codimension-at-least-two support; the strict transform V(u)i is again a hypersurface.

Proof

1.1F1F4given

The support stays a proper subset of the hypersurface in every characteristic. By [F1], supp⁡(Ii,μ)⊆V(u)i for every i. At i=0 this inclusion is proper because the support has codimension at least two throughout U while V(u) is a hypersurface [F4]. If it is proper at stage i, then V(u)i∖supp⁡(Ii,μ) is a nonempty open subset disjoint from the next center, since every center lies in the marked support. The blow-up is an isomorphism over this open set, and the controlled transform agrees there with the unchanged ideal, so this nonempty open subset persists inside V(u)i+1 and remains outside supp⁡(Ii+1,μ). Thus the inclusion is proper at every stage.

2.1A1F2givenstep 1.1

Assume now that K has characteristic zero or perfect characteristic p>μ. Since V(u) is a regular hypersurface containing the initial support and is not contained in that support, it satisfies the hypotheses of [F2]. Applying [F2] to S=V(u) proves that (V(u)i) is a multiple test blow-up of C(I,μ)∣V(u) and gives the support identity of clause (3) at every stage.

3.1A1F2step 2.1∎

Under AC and the same characteristic condition, the converse clause of [F2] says that every multiple test blow-up of C(I,μ)∣V(u) is induced by a multiple test blow-up of (I∣U,μ) with centers in the strict transforms of V(u). The equality in clause (3) follows at every stage from [F2].

PropositionStatement: AI-adaptedProof: AI-adaptedOpen item page →

Canonical resolution of marked ideals

Statement

Assume AC (The Axiom of Choice).

Let K be an algebraically closed field of characteristic zero and let (I,E,μ) be a marked ideal with μ≥1 on a smooth K-scheme X of finite type (Marked ideals and their support), where I is not identically zero on any irreducible component of X (equivalently, its stalk at each component's generic point is nonzero). Then (I,E,μ) admits a canonical resolution: a resolution (Xi)0≤i≤m, X0=X, satisfying the four conditions of Canonical resolutions with invariants of a marked ideal: (1) there are successively defined invariants inv⁡, ν and ρ with values in Q≥0×Q≥0∞, Q≥0 and Sub⁡(Ei) respectively and finite ranges on each support; inv⁡ and the lexicographically ordered pairs (inv⁡,ν) and (inv⁡,ρ) are upper semicontinuous there, equivalently the auxiliary functions are upper semicontinuous on each fixed-inv⁡ stratum; (2) the centers Ci of the blow-ups are regular, are contained in the supports, have SNC with Ei, and are the locus where the pair (inv⁡,ρ) attains its maximum (components of the maximal locus of inv⁡); (3) after blowing up Ci, (a) for x∈supp⁡(Ii+1,Ei+1,μ) with σi+1(x)∈Ci either inv⁡(x)<inv⁡(σi+1(x)) or inv⁡(x)=inv⁡(σi+1(x)) and ν(x)<ν(σi+1(x)); (b) for σi+1(x)∉Ci the three invariants are unchanged; (4) for every etale morphism φ ⁣:X′→X the induced sequence φ∗(Xi) is an extension of the canonical resolution of φ∗(I,E,μ) and the invariants agree. The construction is an induction on dim⁡X and proceeds in two steps: Step 1 resolves marked ideals of maximal order by first replacing the ideal with the equivalent coefficient ideal of its homogenization, then moving marked ideals and normal-crossings strata apart (Step 1a), removing codimension-one components (Step 1ba) and reducing to hypersurfaces of maximal contact (Step 1bb), where the glueing lemma makes the invariant independent of the chosen tangent direction and the inductive hypothesis on the hypersurface provides the next center; Step 2 resolves a general marked ideal by decreasing the maximal order of its non-monomial part through the companion ideal (Step 2a), reducing to the monomial case (Step 2b), whose resolution is finite because ν takes values in the discrete set 1μZ≥0.

Facts & Assumptions

Given: Assume AC. Let K be an algebraically closed field of characteristic zero, let X be a smooth finite-type K-scheme, and let (I,E,μ) be a marked ideal with μ≥1 and I nonzero at the generic point of every component.

[A2]

The Axiom of Choice: AC is assumed for the marked-sum, completion, and regular-local UFD suppliers used in the construction.

[A1]

Encode each source invariant coordinate by the order embedding q↦(0,q) for q∈Q≥0 and ∞↦(1,0) into Q≥02. Apply this map once when a source coordinate is introduced. An invariant returned by a lower-dimensional or companion induction is already encoded; prepend the new encoded coordinates directly and never encode that returned tuple again. Thus, for example, the source values (s,∞,0,0,…) and (0,∞,0,0,…) are compared using the rational tuples ((0,s),(1,0),(0,0),(0,0),…) and ((0,0),(1,0),(0,0),(0,0),…), respectively. Flatten each displayed pair and each grouped tuple into a single rational sequence, identified with its first coordinate and its remaining infinite tail in the stated codomain. This preserves lexicographic comparisons and uses only rational coordinates.

[F1]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors, Iterated derivative ideals preserve support in the safe characteristic range: a marked ideal is of maximal order when Dμ(I)=OX; its support is V(Dμ−1I), and tangent directions are multiplicity-one sections of T(I)=Dμ−1I.

[F2]

The homogenized ideal is equivalent to the marked ideal, The coefficient ideal is equivalent to the marked ideal: (J,μ)≃(H(J),μ)≃C(H(J),μ); replacing a marked ideal by an equivalent one does not change its supports, its admissible centers or the resolution process.

[F3]

Glueing of homogenized ideals along etale neighbourhoods, An automorphism of the completed local ring matching two tangent directions preserves the homogenization: homogenized ideals look the same from all tangent directions, and the étale neighbourhoods constructed from two tangent directions identify the two pullbacks; hence invariants defined via different tangent directions agree.

[F4]

The coefficient ideal controls the support after restriction, Coefficient-ideal control with centres allowed off the subvariety, Refined maximal-contact statement via the coefficient ideal: restriction of the coefficient ideal to a smooth subvariety computes the support on that subvariety, and multiple test blow-ups of the restriction define ones of the ambient marked ideal with centers in the strict transforms.

[F5]

Codimension-one components of a maximal-order support: codimension-one maximal-order support components are smooth and isolated. Their labelled Cartier ideal division gives the unit ideal near the exceptional divisor; this is an admissible controlled transform when the component has SNC with the boundary.

[F6]

Order functions and normal-crossings strata are upper semicontinuous: the order function and normal-crossings count sE are upper semicontinuous; finite-range lexicographic assembly preserves this property. Upper semicontinuity of the algorithm's branchwise invariants must be proved in the induction below; it is not supplied by the definition.

[F7]

The monomial part, the non-monomial part and the companion ideal, Addition and multiplication of marked ideals: if the support is empty, Step 2 is terminal and no companion is used. If ord⁡N(I)=0, the marked ideal is handled directly by the monomial Step 2b; when the input is of maximal order, (M(I),μ) is an equivalent maximal-order input for Step 1. If ord⁡N(I)>0, the companion O(I,μ) is of maximal order with support supp⁡(I,μ)∩{x:ord⁡xN(I)=ord⁡N(I)}; its sum branch has positive marks and uses the AC clause of the addition lemma. It is considered on the open neighbourhood of the original support where ord⁡(N)≤r. A positive-residual-order companion is not asserted to be equivalent to the entire original input.

[F8]

Canonical resolutions with invariants of a marked ideal, Equivalence of marked ideals, Multiple test blow-ups, controlled transforms and resolutions of marked ideals: a canonical resolution is a resolution whose invariants satisfy conditions (1)–(4) of the definition; equivalence preserves supports and multiple test blow-ups at every stage.

[F9]

Locally finite type and finite type morphisms, Chain dimension and the empty-space convention: X is Noetherian of finite dimension, so an induction on dim⁡X is available, and closed subsets have finitely many irreducible components.

[F10]

Affine blowup standard charts and overlaps: for a center ideal generated by regular parameters, its chart indexed by a generator w has coordinate ring R[I/w]; the other center parameters divide by w, and the new exceptional divisor has equation w=0.

Proof

1.1A2F2F9

Induction setup. We argue by induction on d=dim⁡X, componentwise on the disjoint irreducible components of the smooth scheme. For d=0, X is a finite disjoint union of K-points; the generic-nonzero hypothesis makes I=OX on every component, so the support is empty and the trivial sequence is the canonical resolution. Assume the proposition for smooth schemes of dimension <d and marked ideals satisfying the stated positive-marking and componentwise generic-nonvanishing hypotheses; by [F2] we may replace any marked ideal of maximal order by its equivalent C(H(J,μ)), and we do so without changing the supports or the admissible centers.

1.2A2A1F5F6

Step 1a, eliminating strata contained in the support. At the start of each nonterminal Step 1a pass at stage i, compute and hold fixed for that pass si=max⁡{si(x):x∈supp⁡(Ji,Ei,μ)}, where si(x) counts strict transforms of the boundary family E fixed when this maximal-order Step 1 begins; new exceptional divisors created during Step 1a remain in the full Ei but are not added to this boundary count. If the support is empty, terminate; if si=0, finish Step 1a and proceed to Step 1b without restricting to H0=X. For si>0, use only irreducible components Hα,isi of intersections of exactly si boundary components that meet the support. These strata are smooth and have finitely many pairwise disjoint irreducible components; taking them componentwise ensures that each retained restricted ideal is generically nonzero unless the stratum belongs to the contained-support branch. Their intersections with the support are pairwise disjoint: a point in two distinct such pieces would lie on more than si boundary components. The ambient intersections can overlap away from the support, where the induced test blow-ups are identities. If Hα,isi⊆supp⁡(Ji,Ei,μ), blow up this stratum. For si≥2, the boundary count drops at every support point above it; for si=1, [F5] shows that the controlled transform has empty support above the codimension-one stratum. Assign points of this stratum the source invariant (si,∞,0,0,…), represented by ((0,si),(1,0),(0,0),(0,0),…) under [A1], with ν=0 and ρ=∅. This is lexicographically above the finite lower-dimensional invariant values on the other support pieces. Recompute s only at the start of the next pass.

1.3A2A1F5F10

Step 1b, codimension-one components. After Step 1a reaches s=0, no strict transform of the inherited Step 1 boundary family meets the support. If the support has codimension-one components, [F5] shows they are smooth and isolated from the remaining support. They are strict transforms of the original such components: maximal order makes the ideal order at the generic point of each earlier center exactly its marking, so division by that exceptional power leaves no new exceptional component in the support. Every earlier center is contained in an isolated component or disjoint from it. For a center contained in a smooth component C=(u=0), choose adapted regular parameters for that center including u and the equations of the exceptional divisors already transverse to C; this is possible because the center is regular and SNC with that boundary. In a blowup chart indexed by another center parameter w, the strict transform of C has equation u/w=0 and the new exceptional divisor has equation w=0; the chart indexed by u misses the strict transform. These coordinate charts preserve simultaneous SNC of C with the newly created boundary members, by induction over the preceding blowups. Since the inherited boundary no longer meets the support, C has SNC with the full current boundary. Its labelled Cartier blowup is therefore admissible under [F5]; assign them the source invariant (0,∞,0,…), represented under [A1] by ((0,0),(1,0),(0,0),…), with ν=0 and ρ=∅. This is above all finite lower-dimensional invariant values, so these components are the next centers. Blowing them up makes the controlled-transform support empty above them. The remaining support has codimension at least two.

1.4A2F6F7F8

Step 2b, the monomial case. If the support is empty, terminate. Otherwise, for a monomial ideal I=M(I)=∏DxDaD, set inv⁡=(0,0,…) and ν(x)=ord⁡x(I)/μ. At each point, let A be the inclusion-minimal threshold subset selected by the source's maximal ρ rule: ∑D∈AaD≥μ, while ∑D∈A∖{D0}aD<μ for every D0∈A. The maximal locus of (inv⁡,ρ) gives the regular SNC center locally defined by the divisors in this selected A. In a blow-up chart indexed by any Dj∈A, the new exceptional exponent is a=∑D∈AaD−μ<aDj, by the minimality condition with D0=Dj. For any surviving support point q above p∈C, the old strict boundary divisors through q are among those through p, except for the chart divisor Dj, whose old exponent is replaced by a<aDj. Hence ord⁡q(Mnew)≤a+∑D∋p, D≠DjaD<ord⁡p(M); divisors missing q only lower the sum. Thus at every support point above the center the total monomial order drops, so ν strictly drops while the monomial-case inv⁡ stays fixed; equivalently, (inv⁡,ν) strictly descends. There are finitely many initial maximal support components. Each chosen component is replaced by finitely many maximal components above it, and the maximum monomial order on each new component is strictly smaller, as shown by the chart inequality. This is a finitely branching replacement tree whose path length is bounded by the initial integer monomial-order maximum. It is finite, so the process terminates with empty support; the discreteness of ν∈1μZ≥0 provides this integer bound.

2.1A2F4F6step 1.2

Step 1ab, moving support away from maximal boundary strata. This branch occurs only for the pass-start value si>0, after all strata contained in the support have been blown up. Resolve each lower-dimensional restricted marked ideal on Hα,isi. These restrictions are generically nonzero: at the generic point of a retained component the ambient support is absent, and [F4] identifies the restricted support with its ambient intersection. The restricted support pieces are disjoint, so their canonical sequences glue; on ambient overlaps outside the support there are no centers and the induced maps are identities. By [F4], each restricted support equals the corresponding ambient support intersection, and the glued sequence is a multiple test blow-up of the ambient ideal. On a support piece use inv⁡(x)=((0,sE(x)),inv⁡Ji∣Hα,isi(x)), encoding the finite leading coordinate by [A1], with the restricted ν and ρ; the support piece determines the stratum uniquely. By induction, above each center the pair (inv⁡,ν) strictly descends, while off the center the induced map is the identity and the invariants are unchanged. Once all these restrictions are resolved, no support point remains on any maximal stratum from the pass-start boundary count, so the maximum boundary count drops before the next pass; recompute it only at that next pass. As the count is a nonnegative integer, finitely many passes reach si=0, and then Step 1b begins.

2.2A2A1F1F3F4F6F10step 1.3

Step 1b, hypersurfaces of maximal contact. When the inherited Step 1 boundary misses the support, first remove codimension-one support components as in 1.3. For the remaining codimension-at-least-two support, use the strict transforms of tangent hypersurfaces chosen on finitely many initial Step 1 charts, with the inherited boundary temporarily forgotten. These charts cover the surviving support: the original isolated codimension-one pieces have already been removed. By Derivative ideals under a multiple test blow-up, their controlled equations remain in the transformed tangent ideal, and The supports of a multiple test blow-up stay inside the strict transforms of a hypersurface of maximal contact puts the support in their strict transforms. They are simultaneously transverse to the newly created boundary: initially that boundary is empty; if a regular center lies in V(u) and is SNC with the existing boundary transverse to V(u), choose adapted center parameters including u. In a chart indexed by another center parameter w, the equations are u/w for the strict hypersurface and w for the new exceptional divisor, along with the surviving old boundary parameters; the u-chart has no hypersurface support. Thus transversality persists. The inherited boundary now misses the support, so these transformed hypersurfaces are transverse to the full boundary there, as [F3] requires. Write V(u) for such a transformed hypersurface. Its coefficient-ideal restriction is generically nonzero on each component, since [F4] identifies its support with a subset of ambient codimension at least two, hence a proper subset of every hypersurface component. By [F4], the inductive resolution on V(u) transfers to the ambient ideal; off the restricted support, this transfer is an isomorphism. Set inv⁡(x)=((0,0),inv⁡J∣V(u)(x)), with ν,ρ from the restriction. Apply [F3] to two initial tangent directions with the inherited boundary forgotten, and lift its étale comparisons along the preceding test sequence. The new exceptional divisors are preserved by these lifted comparisons; near the surviving support the inherited boundary is absent. Hence they identify the pulled-back homogenized ideals and the full current boundary, so the values, centers, and local resolutions agree on overlaps. Thus above every transferred center (inv⁡,ν) strictly descends by induction, and off the center all invariants are unchanged. The local sequences glue to the canonical resolution.

3.1A2A1F7F8step 2.2

Step 2, the companion reduction and monomial case. If supp⁡(I,E,μ)=∅, the resolution is complete. Otherwise factor I=M(I)N(I) and compute r=ord⁡N(I) as in [F7]. If r=0, N(I) is a unit on a neighbourhood of the support, so the marked ideal agrees there with (M(I),μ): resolve this neighbourhood by Step 2b and extend by the identity outside its support, without forming a companion. The maximal-order Step 1 may use (M(I),μ) only when the original input is already of maximal order. If r>0, form the maximal-order companion O(I,μ) and apply Step 1. For r<μ, a companion test sequence is simultaneously admissible for (N,r) and (M,μ−r), and their product transforms to (Ii,μ) by the product rule. For r≥μ, the controlled transform of I is the transformed (N,r) times a monomial factor, with each new exponent containing r−μ≥0. Thus every companion center is admissible for (I,μ) in either case. Write Ni∗ for the transform of (N,r); it remains of maximal order by Controlled transforms preserve maximal order on nonempty transformed schemes. Factoring its exceptional powers gives N(Ii), so ord⁡N(Ii)≤ord⁡Ni∗≤r. At a point with ord⁡N(Ii)=r, every exceptional factor removed from Ni∗ is a unit there, so ord⁡Ni∗=r. If that point is in the original marked support, it is in the transformed companion support by the sum-support rule (or directly when r≥μ). Consequently clearing the companion support lowers the maximum order of N(Ii) on the remaining original support, so the first source coordinate r/μ, represented by (0,r/μ) under [A1] in inv⁡=((0,r/μ),inv⁡O) on the transformed companion support, strictly drops between companion passes. This formula is used only where inv⁡O is defined; values at other original-support points are assigned by the successive-transfer construction below. Within a companion pass, Step 1 gives strict descent of (inv⁡O,νO) above every center; therefore (inv⁡,ν) strictly descends there, and off the center the induced map is the identity and all invariants are unchanged. Repeat only while r>0; the finite sequence of strictly decreasing nonnegative integer values of r ends with empty support or r=0. In the latter case, Step 2b has constant monomial-case inv⁡ and strictly lowers ν=ord⁡x(M(I))/μ above every center, so (inv⁡,ν) strictly descends until the support is empty.

4.1A2A1F6F7F8step 1.4step 2.1step 2.2step 3.1

Define the invariants on the entire support by successive assignment, rather than evaluating a companion invariant outside its domain. During a positive-r pass assign the tuple in 3.1 only on the transformed companion support; within Step 1 assign boundary-stratum and maximal-contact values only on their stated domains. Each point not yet assigned lies off every center in that pass, so it has a unique unchanged lift through the pass, with the same local ideal and boundary germs. After that pass continue the algorithm on the remaining support. When its residual-order stratum, boundary-count stratum, or final monomial branch is reached, assign the corresponding invariant there and transfer it back through these unchanged lifts. Termination proved in the preceding steps guarantees a finite first assignment for every point still in a support: otherwise an unassigned point would survive to the empty final support. Use the same transferred value at all earlier stages where the point was untouched. In particular a point of residual order below r receives its own later residual-order coordinate, not the current r/μ and not an undefined inv⁡O. The analogous convention fills lower boundary-count pieces during Step 1. Thus the active maximum is unchanged, and the invariants at points outside every intervening center are unchanged by construction.

5.1A2A1F3F4F6F7F8step 1.2step 1.4step 2.1step 2.2step 3.1step 4.1

Descent, termination, and canonicity. In a boundary pass the fixed pass-start count makes each maximal-stratum blow-up lower sE above its center; on restricted strata, induction strictly lowers (inv⁡,ν) above centers, and the ambient maps are identities off the support intersections. The codimension-one branch assigns the top source value (0,∞,0,…), rationally encoded as ((0,0),(1,0),(0,0),…); blowing up that component makes the controlled-transform support empty above it. In the hypersurface branch, induction strictly lowers (inv⁡,ν) above every transferred center and the maps are identities off the restricted support. In the companion branch, the first source coordinate r/μ (encoded as (0,r/μ)) strictly decreases between passes and the companion invariant pair strictly decreases within a pass. In the monomial branch, inv⁡ is constant and ν strictly decreases above each center by 1.4. Thus in every branch the pair (inv⁡,ν) strictly decreases above centers, while off centers the morphism is an isomorphism and all invariants are unchanged; each loop terminates by the decreasing boundary count, the decreasing integer r, induction in lower dimension, or the discrete nonnegative values of ν. The values and maximal centers are determined by boundary counts, orders, the threshold-subset rule and lower-dimensional canonical invariants. Their semicontinuity is proved in the next steps.

6.1A1F6F9step 5.1

Finite ranges and lexicographic assembly. Each stage has finitely many boundary components and finitely many maximal-contact charts. There are finitely many passes by step 5.1; in each dimension reduction the finite collection of restricted stage invariants has finite range by induction, and the tuple depth is uniformly bounded by the ambient dimension. Residual orders, boundary counts and the exponents of the finitely many monomial ideals occurring in these passes take finitely many values. Consequently all three invariants have finite ranges on each support; padding the bounded-depth rational tuples adds no values. For the finite-stratum assembly used below, let f be a finite-range upper semicontinuous function on a support S. For each attained value a, its level stratum Sa={f=a} is locally closed, since {f>a} is a finite union of closed superlevels. If F={x∈Sa:g(x)≥b} is closed in Sa, then its closure in S lies in {f≥a} and meets Sa exactly in F. Thus F‾⊆F∪{f>a}, and the lexicographic superlevel {(f,g)≥(a,b)}={f>a}∪F={f>a}∪F‾ is closed. For unattained a this superlevel is just {f>a}. Conversely, restricting pair superlevels to Sa gives relative superlevels of g.

7.1A1F3F4F5F6F7F8step 1.2step 1.3step 1.4step 2.1step 2.2step 3.1step 4.1step 6.1

Primary and auxiliary semicontinuity. On each residual-order or boundary-count stratum, unchanged lifts in step 4.1 identify the local ideal and boundary germs with their later pass. The leading residual-order and boundary-count functions have closed superlevels by [F6]. On a fixed leading level, the terminal infinity branch is closed and has maximal tail; distinct maximal boundary pieces are disjoint on the support, and their lower-dimensional primary tails are upper semicontinuous by induction. In the nonboundary branch, the isolated codimension-one pieces have the maximal infinity tail; on the remainder, the finitely many maximal-contact charts identify their primary tails étale-locally by [F3], so relative closed superlevels glue. Repeated application of step 6.1, with the leading coordinate as f and the lower-dimensional tail as g, proves upper semicontinuity of the entire primary invariant. Now fix its complete value a. In every recursive branch this fixes the leading coordinate and the entire lower-dimensional primary tail, so the inductive auxiliary semicontinuity on that lower primary level applies. Terminal infinity branches have constant ν=0, ρ=∅. On a monomial branch, ν is a finite sum of nonnegative exponents of boundary divisors through the point divided by the mark, hence has closed superlevels. Each eligible inclusion-minimal threshold subset contributes its closed boundary intersection; the maximum subset rule makes each ρ superlevel a finite union of these intersections. Boundary pieces are disjoint on the relevant support; maximal-contact charts agree by [F3]; unchanged lifts transport these relative-closed loci through earlier passes. Hence ν and ρ are upper semicontinuous on {inv⁡=a}. Applying step 6.1 to the primary invariant and each auxiliary function proves upper semicontinuity of (inv⁡,ν) and (inv⁡,ρ) on the entire support. In particular their finite ranges yield closed maximal loci; the branch constructions prove that the center selected by (inv⁡,ρ) is regular and has SNC with the boundary.

8.1A2A1F3F4F8step 5.1step 7.1∎

Étale canonicity. Under an étale morphism, the tracked Step 1 boundary components, orders, monomial and non-monomial factors, threshold subsets, companion supports and maximal-contact restrictions pull back compatibly. The glueing equality [F3] identifies maximal-contact choices, while [F4] and [F8] preserve restrictions, test sequences and equivalences. Thus the primary and both auxiliary values agree at corresponding points, the pair maxima give the pulled-back centers, and empty centers contribute only identity steps. The induced sequence is an extension of the canonical resolution with agreeing invariants. For an arbitrary étale source X′, take finite-type open charts: étale local finite presentation over the finite-type X gives such charts, and the preceding finite-type comparison applies on each. Use the pulled-back original finite sequence to align these chart sequences, inserting only empty-center isomorphisms. The sheaf ideals, boundary labels and invariant values agree on overlaps because they are pullbacks of the same original data. Their ranges are subsets of the original finite ranges. At a nonempty inverse center the original maximum is attained in the image and is exactly the maximum on the pullback; centers missed by the entire image give only inserted identities. Hence, after deleting globally empty-center steps, the glued sequence is canonical in the extended sense of the definition, with at most m blowups, even when X′ is not quasi-compact. Compatibility with further étale pullbacks follows by composition. Together with the descent and termination of step 5.1 and the semicontinuity of step 7.1, this proves all four stated conditions.

Remarks

Separate upper semicontinuity of the auxiliary functions on the whole support is not asserted. On X=AK2 take I=(x(x+y)), marking 1, and ordered boundary E={V(x)}. Its monomial part is (x) and its residual part is (x+y). At points of V(x) away from the origin, the residual order is zero, so successive assignment reaches Step 2b with ν=1 and ρ={V(x)}. At the origin the companion is (x+y,1) and Step 1a restricts it to V(x), giving (y,1) with empty restricted boundary. The lower-dimensional codimension-one branch assigns ν=0 and ρ=∅. Thus the superlevel locus {ν≥1} is V(x)∖{0}, which is not closed in the support V(x)∪V(x+y); the corresponding positive-ρ superlevel locus is the same. The source's Proposition 3.0.8 makes the same standalone upper-semicontinuity assertion, but its displayed branch formulas give this counterexample. Semicontinuity on fixed-primary-invariant strata does not imply the stated independent semicontinuity on the whole support. The corrected pair semicontinuity in the Statement retains closed superlevels for the center and descent comparisons while preserving the source algorithm and its invariant values.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Etale commutativity of the maximal-order resolution step

Statement

Assume AC (The Axiom of Choice).

In the setting of Step 1 of Canonical resolution of marked ideals, let φ ⁣:X′→X be an etale morphism (Étale morphism of schemes) and let (Xi)0≤i≤m be the canonical resolution of the maximal-order marked ideal (J,E,μ) constructed in Step 1. Then: (1) the induced sequence φ∗(Xi)0≤i≤m is an extension of the canonical resolution (Xj′)0≤j≤m′ of φ∗(J,E,μ); (2) for every x′∈supp⁡(φ∗(Ji,Ei,μ)) the invariants agree: inv⁡(x′)=inv⁡(φi(x′)),ν(x′)=ν(φi(x′)),ρ(x′)=ρ(φi(x′)).

Facts & Assumptions

Given: A maximal-order marked ideal (J,E,μ) on the smooth finite-type K-scheme X, its canonical resolution (Xi)0≤i≤m from Step 1 of Canonical resolution of marked ideals, and an étale morphism φ ⁣:X′→X.

[A1]

The Axiom of Choice: AC is assumed for the canonical-resolution consumer clauses and their cited AC-dependent construction suppliers.

[F1]

Canonical resolution of marked ideals: the resolution is constructed via the splitting sequence si0>⋯>sik of maximal numbers of divisors of E through the support; between consecutive indices, contained boundary strata are blown up directly in Step 1aa, and the remaining generically nonzero restrictions to the strata Hαs are resolved by lower-dimensional induction. In the non-boundary case, Step 1ba first removes isolated codimension-one support components directly; only the remaining codimension-at-least-two support is reduced to a hypersurface of maximal contact V(u) and lower-dimensional induction.

[F2]

The coefficient ideal controls the support after restriction, Smooth base change of multiple test blow-ups: étale base change of the marked ideal commutes with restriction to the strata and with multiple test blow-ups, and preserves supports; invariant equality for the lower-dimensional canonical sequences is provided by induction.

[F3]

Glueing of homogenized ideals along etale neighbourhoods, Refined maximal-contact statement via the coefficient ideal: the reduced problem on a hypersurface of maximal contact is independent of the choice of tangent direction, and the inverse image of a hypersurface of maximal contact is again one.

[F4]

Canonical resolutions with invariants of a marked ideal: inv⁡ and (inv⁡,ρ) have closed superlevel loci, the centers are the maxima of (inv⁡,ρ), and (inv⁡,ν) gives the descent comparison. Equality of all invariant values for the induced sequence follows once the centers and the reductions agree stage by stage.

[F5]

Smooth base change of multiple test blow-ups: étale morphisms are smooth, so the order/SNC calculation used in that supplier preserves orders of ideals and hence supports.

Proof

1.1A1F1F5

The splitting sequence of the pullback. Let s0′>⋯>sk′′ be the corresponding sequence of maximal divisor counts for the canonical resolution of φ∗(J,E,μ); since φ is flat, the inverse image of E has the same intersection pattern, so sj′≤si-values at corresponding stages. We prove by induction on the pairs (l,l′) of the two splitting sequences that the induced sequence φ∗(Xi) agrees with the canonical sequence of the pullback up to extension and that the invariants correspond.

2.1A1F2F4step 1.1

Case 1: the étale image misses the strata. If s(φ∗(Xil))<sil, then the centers blown up in (Xi)il≤i≤il+1 lie in strata that do not meet the image of φ, so their inverse images are empty and the induced morphisms are isomorphisms; the equality of marked ideals and invariants with the pullback at stage il is inherited from stage l.

3.1A1F1F2step 2.1

Case 2: the strata meet the image. If s(φ∗(Xil))=sil>0, the strata of the pullback are the inverse images of the strata of X. First apply the contained-stratum Step 1aa: whether a component is contained in the support is detected by its coefficient-ideal restriction support, so this condition pulls back. Such a component is a regular SNC boundary stratum; its blowup and controlled division commute with the étale base change, and both sides assign the same count/infinity primary value and the auxiliary values ν=0, ρ=∅. No induction is applied to its zero restricted ideal. After these contained components are removed, [F2] makes every retained restricted ideal generically nonzero: an identically zero component restriction would have whole marked support and hence would still be in the contained branch. Now φ∗(Jil∣Hαs)=Jjl′′∣(H′)αs; the resolution process is reduced on both sides to the restrictions, so by the inductive hypothesis in lower dimension (the strata have dimension <dim⁡X) the canonical resolutions correspond and the invariants satisfy inv⁡(x′)=inv⁡(φ(x′)), as do ν and ρ.

4.1A1F1F2F3step 3.1

Case 3: the non-boundary case. If s(φ∗(Xik))=sik=0, first apply Step 1ba of [F1]. Its codimension-one support components are regular and isolated; étale pullback preserves their codimension and their local equation J=(uμ). Their labelled Cartier-center blowups divide by uμ and give the unit ideal near them. Thus this branch corresponds on both sides, with an inserted isomorphism only when a center has empty inverse image. The remaining support has codimension at least two. Only now apply Step 1bb: its maximal-contact hypersurface has a generically nonzero restricted ideal, so the lower-dimensional canonical-resolution induction applies. Its inverse image is a hypersurface of maximal contact by [F3], and the induction gives matching centers and invariants; [F3] gives independence of the direction. The codimension-one branch carries the same top encoded invariant on both sides, and all other points receive the values from their later unchanged lifts as in the proposition's successive assignment.

5.1A1F4step 2.1step 3.1step 4.1∎

Conclusion. Combining the preceding cases over the finitely many elements of the splitting sequence, φ∗(Xi) is an extension of the canonical resolution of the pullback and the invariants agree on all supports; this proves (1) and (2).

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Canonical resolutions commute with embeddings of ambient smooth schemes

Statement

Assume AC (The Axiom of Choice).

Let K have characteristic zero (Field) and let φ ⁣:X↪X+ be a closed immersion of smooth K-schemes of finite type (Closed immersions of schemes) and let I⊆OX be a coherent ideal sheaf that is not identically zero on any irreducible component of X, with I+ the inverse-image ideal of φ∗I under the quotient map OX+→φ∗OX. For the marked ideals (I,∅,1) on X and (I+,∅,1) on X+ the canonical resolutions are related, at points x∈supp⁡(I,1), by inv⁡I+(x)=(0,1,0,0;0,1,0,0;… ;0,1,0,0;inv⁡I(x)),νI+(x)=νI(x),ρI+(x)=ρI(x), with (0,1,0,0) repeated k times, where k is the local codimension of the smooth immersion at x, and resolving (X+,I+,∅,1) is locally equivalent to resolving (X,I,∅,1) (Canonical resolutions with invariants of a marked ideal).

Facts & Assumptions

Given: A characteristic-zero field K, a closed embedding φ ⁣:X↪X′ of smooth finite-type K-schemes, and a marked ideal (I,∅,1) on X generically nonzero on every component. Let I′ be the inverse image of φ∗I under OX′→φ∗OX; locally the ideal of X is generated by regular parameters u1,…,uk∈I′.

[A1]

The Axiom of Choice: AC is assumed for the canonical-resolution consumer clauses and their cited AC-dependent construction suppliers.

[F1]

Closed immersions of schemes, Marked ideals and their support: the quotient map to φ∗OX defines I′ as the ideal of the same closed subscheme of X viewed inside X′. Locally it is generated by the equations u1,…,uk of X and lifts of generators of I. It is coherent because the affine smooth coordinate rings are Noetherian, its restriction to X is I, and its support for mark 1 is the image of supp⁡(I,1).

[F2]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors, Giraud's tangent-direction lemma, The supports of a multiple test blow-up stay inside the strict transforms of a hypersurface of maximal contact: the sections u1,…,uk are tangent directions of (I′,1) at points of X, and along the canonical resolution the supports are contained in the strict transforms of the hypersurfaces V(ui).

[F3]

Canonical resolution of marked ideals: the canonical resolution of a marked ideal is built by the algorithm of Steps 1–2, whose reductions depend only on intrinsic data (supports, orders, derivative ideals and the invariants).

Proof

1.1A1F1F2F3

The invariant of the embedded ideal. Running Steps 2a and 1bb of the algorithm for (I′,∅,1) at a point x∈X: in Step 2a the ideal is of maximal order with ord⁡N=1, so the companion ideal is I′ itself and the invariant receives the first coordinate 1; in Step 1bb (non-boundary case s(x)=0) the resolution passes to a hypersurface of maximal contact. Passing successively to the tangent directions u1,…,uk, each passage adjoins the source pair (1,0), encoded as the four rational coordinates (0,1,0,0) under [F3], to the invariant by [F2], and after k passages one arrives at the restriction to X, where the invariant of (I,1) appears. Hence inv⁡I′(x)=(0,1,0,0;… ;0,1,0,0;inv⁡I(x)) with (0,1,0,0) repeated k times, and analogously νI′(x)=νI(x) and ρI′(x)=ρI(x).

2.1A1F2F3step 1.1∎

The resolutions correspond. By [F3] the centers of the resolution of (I′,1) are the maximal loci of the invariants computed in step 1.1; because the first 4k encoded coordinates of inv⁡I′ are constant on the image of the resolution of (I,1), the maximal locus over X is the image of the maximal locus of inv⁡I, and blowing it up in X′ restricts to blowing it up in X. Iterating, the canonical resolution of (I′,1) restricts over X to the canonical resolution of (I,1), with the invariant comparison of step 1.1 (the prefixed constant pairs are removed on restriction).

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Canonical resolution under isomorphisms of the ground field

Statement

Assume AC (The Axiom of Choice).

Let K,K′ be fields of characteristic zero and let σ ⁣:K⟶∼K′ be a field isomorphism fixing the common prime field Q (Field, A field's prime subfield is isomorphic to Q in characteristic zero and to Fp in characteristic p). Let X/K and X′/K′ be smooth schemes and let φ ⁣:X′→X be a σ-semilinear isomorphism, as in Derivative ideals under semilinear ground-field isomorphisms. Let (I,E,μ) be a marked ideal on X with μ≥1 and with I not identically zero on any irreducible component (Marked ideals and their support), and suppose it has a canonical resolution (Xi), with induced marked ideals (Ii,Ei,μ) (Canonical resolution of marked ideals).

Then the induced sequence (Xi′):=(Xi×XX′) is a canonical resolution of φ∗(I,E,μ). The isomorphism φ lifts to semilinear isomorphisms φi ⁣:Xi′→Xi, and for every i, φi−1 ⁣(supp⁡(Ii,Ei,μ))=supp⁡(Ii′,Ei′,μ). On these corresponding supports the invariants agree: inv⁡(φi(x))=inv⁡(x),ν(φi(x))=ν(x),ρ(φi(x))=φi(ρ(x)).

Facts & Assumptions

Given: Characteristic-zero fields K,K′ and a semilinear field isomorphism σ ⁣:K→K′, smooth schemes X/K and X′/K′, a σ-semilinear isomorphism φ ⁣:X′→X, a marked ideal (I,E,μ) with μ≥1 and I generically nonzero on every component of X, its pullback φ∗(I,E,μ)=(φ∗I,φ−1E,μ), a canonical resolution (Xi)0≤i≤m of (I,E,μ) with induced marked ideals (Ii,Ei,μ), and the base changes Xi′:=Xi×XX′ with induced marked ideals (Ii′,Ei′,μ).

[A1]

The Axiom of Choice: AC is assumed for the canonical-resolution consumer clauses and their cited AC-dependent construction suppliers.

[F1]

Derivative ideals under semilinear ground-field isomorphisms, Derivative ideals of an ideal sheaf and of a marked ideal: for every coherent ideal sheaf A on X and every i≥0 one has φ∗(DKi(A))=DK′i(φ∗A); in particular φ∗(TK(I))=TK′(φ∗I).

[F2]

The homogenized ideal of a marked ideal of maximal order, The coefficient ideal of a marked ideal of maximal order: For a maximal-order input (J,ν) with ν≥1, H(J,ν)=∑i=0ν−1Di(J)T(J)i and C(J,ν)=∑i=0ν−1(Di(J),ν−i)=(∑i=0ν−1Di(J)ν!/(ν−i), ν!) are built from the derivative ideals by finite sums, products and powers of ideal sheaves, and are used only on maximal-order inputs.

[F3]

Marked ideals and their support, Order of an ideal sheaf at a point: for a marked ideal (J,ν) the support is {x:ord⁡x(J)≥ν}; the underlying scheme isomorphism φ induces local-ring isomorphisms carrying maximal ideals and ideal stalks to their pullbacks. Hence corresponding orders, supports, exceptional-divisor counts sE, and SNC conditions agree, independently of the ground-field semilinearity.

[F4]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals: (Xi) is a multiple test blow-up with Ii+1=I(Di+1)−μσi+1∗(Ii) and Ei+1=σi+1c(Ei)∪{Di+1}, the centers regular and in SNC position with Ei; a resolution is a multiple test blow-up with empty support.

[F5]

Marked ideals of maximal order, tangent directions and transversality to the exceptional divisors: a tangent direction of (J,ν) is a multiplicity-one section u of T(J)=Dν−1(J), and V(u) is the hypersurface of maximal contact containing the support.

[F6]

Canonical resolution of marked ideals, Canonical resolutions with invariants of a marked ideal, The monomial part, the non-monomial part and the companion ideal, Equivalence of marked ideals, The homogenized ideal is equivalent to the marked ideal, The coefficient ideal is equivalent to the marked ideal: the canonical resolution is produced by the algorithm of Steps 1-2; for a maximal-order input (J,ν), replacing it by the equivalent marked ideals H(J,ν) and C(H(J,ν)) does not change the supports, the admissible centers or the resolution process; each further reduction is determined by intrinsic data, namely the strata Hαs (intersections of members of E), the restrictions of marked ideals to them and to hypersurfaces of maximal contact, the monomial/non-monomial decomposition and the companion ideal, and the invariants inv⁡,ν,ρ assembled from sE, the order functions and the lower-dimensional invariants, with the center the maximal locus of the pair (inv⁡,ρ).

[F7]

Smooth base change of multiple test blow-ups, Smooth morphism of schemes: the underlying scheme isomorphism φ is smooth, so the base change (Xi′)=(Xi×XX′) is a multiple test blow-up of φ∗(I,E,μ) with Ii′=φi∗Ii and Ei′=φi−1Ei, where each φi ⁣:Xi′→Xi is an isomorphism.

[F8]

Order functions and normal-crossings strata are upper semicontinuous, Canonical resolution of marked ideals: divisor counts and order functions are upper semicontinuous by the former supplier; the latter supplies finite ranges and upper semicontinuity of inv⁡ and the lexicographic pairs (inv⁡,ν) and (inv⁡,ρ). Thus the center-ordering pair has a closed maximal locus.

Proof

1.1A1F1F2F3F6

Pullback of the derived objects at their proper inputs. By [F1], derivative ideals of every coherent ideal commute with the semilinear isomorphism. For a maximal-order marked ideal (J,η), finite sums, products and powers commute with this pullback, so [F2] gives φ∗H(J,η)=H(φ∗J,η) and the analogous identity for C(H(J,η)). For a general input (I,μ) first transport its monomial and non-monomial factors: local-ring isomorphisms preserve divisibility by each ordered boundary equation and preserve the maximum residual order on the corresponding supports. If that maximum is positive, the corresponding companions O(I,μ) are maximal-order inputs, and it is these companions to which the homogenization and coefficient identities apply. If the maximum is zero, the inputs are monomial near their supports and Step 2b applies directly; no homogenization of an unrestricted input is used.

2.1A1F3step 1.1

Pullback of orders, supports and strata. By [F3] one has ord⁡φ(x)(J)=ord⁡x(φ∗J) for every coherent ideal J on X, hence supp⁡(φ∗(J,ν))=φ−1(supp⁡(J,ν)), and sφ−1E(x)=sE(φ(x)); since the SNC condition is stalk-local, φ maps the strata Hαs of the algorithm for (I,E,μ) isomorphically onto the corresponding strata for φ∗(I,E,μ), and carries the restriction (I∣Hαs,μ) to (φ∗I∣φ−1Hαs,μ).

3.1A1F3F5F6F8step 1.1step 2.1

Commutation with the reductions of the algorithm. We prove the assertion by induction on d=dim⁡X. For d=0, every component is the spectrum of a finite separable field extension, and the hypothesis of generic nonvanishing makes the ideal the unit ideal on each component. The positive marking therefore gives empty support and the canonical sequence is the identity; its base change is again the identity by step 2.1. Assume now d≥1 and the assertion known in dimensions <d, and transport the algorithm of [F6] along the identification of steps 1.1-2.1: (a) for every maximal-order input reached in Step 1, the replacement of (J,ν) by the equivalent C(H(J,ν)) commutes with pullback by step 1.1; (b) the strata Hαs and the restricted coefficient ideals correspond by step 2.1. First remove components contained in the support by Step 1aa: their regular SNC stratum blowups and controlled division commute with the isomorphism, and both sides give the same count/infinity primary value with ν=0, ρ=∅. No lower-dimensional induction is applied to these zero restrictions. By the coefficient-ideal support identity in [F6], the remaining restrictions are generically nonzero on each retained component; their lower-dimensional canonical resolutions therefore commute with φ by induction. Once the inherited boundary is disjoint from the support, the isolated codimension-one components of Step 1ba correspond by their local equations and labelled Cartier division. Only on the remaining codimension-at-least-two support do we pass to Step 1bb; (c) a hypersurface of maximal contact V(u), u∈T(J), is carried to the hypersurface V(φ∗u) with φ∗u∈T(φ∗J) by steps 1.1-2.1, and the restriction of C(J,ν) to V(u) corresponds, so the induction hypothesis applies to the lower-dimensional marked ideal on V(u); (d) the monomial/non-monomial decomposition J=M(J)N(J) is read off from the vanishing orders of J along the members of E by [F3], hence is transported, and, when the residual maximum is positive, its companion satisfies φ∗(O(J))=O(φ∗J); when it is zero the local monomial Step 2b data correspond directly; (e) the invariants assembled in [F6] from sE, the order functions and the lower-dimensional invariants have equal values at corresponding points by steps 1.1-2.1 and the induction hypothesis, and equality of all values together with preservation of their orders carries the maximal locus of (inv⁡,ρ) on Xi to its exact preimage under φi on Xi′; its closedness is supplied by [F8]. Since the algorithm's centre at each stage is exactly that maximal locus [F6], the process for φ∗(I,E,μ) has centres φi−1(Ci) and produces the base-changed marked ideals (Ii′,Ei′,μ).

4.1A1F4F6F7step 2.1step 3.1∎

The canonical resolution of the pullback. By [F4, F7] the sequence (Xi′)=(Xi×XX′) is a multiple test blow-up of φ∗(I,E,μ) with Ii′=φi∗Ii, Ei′=φi−1Ei, and each φi is an isomorphism; by step 3.1 its centre at each stage is φi−1(Ci), which is the centre prescribed by the algorithm for the pullback, and the algorithm is determined by the intrinsic data [F6]. Hence (Xi′) is the canonical resolution of φ∗(I,E,μ). For x∈Xi′ with φi(x)∈supp⁡(Ii,Ei,μ) step 2.1 gives x∈supp⁡(Ii′,Ei′,μ), and step 3.1(e) gives inv⁡(φi(x))=inv⁡(x) and ν(φi(x))=ν(x); since φi carries the ordered family Ei′=φi−1Ei isomorphically onto Ei, the subsets of exceptional divisors through corresponding points match and ρ(φi(x))=φi(ρ(x)). This completes the induction and the proof.

Remarks

  • The source's Proposition 4.3.2 is stated for isomorphisms over Q that may act nontrivially on the ground field. The semilinear formulation above includes the Galois automorphisms of K‾/K used in Canonical resolutions over non-algebraically-closed ground fields.
  • The scalar extension used in the descent is algebraic and separable in characteristic zero, so every K-derivation of OX‾ is K‾-linear and the derivative ideals computed over K and over K‾ coincide; the lemma therefore applies to the Galois action on the base change.
  • The case of an empty support is the identity resolution and is covered by step 2.1.
LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Etale commutativity of the companion-ideal step

Statement

Assume AC (The Axiom of Choice).

In the setting of Step 2 of Canonical resolution of marked ideals, let φ ⁣:X′→X be an etale morphism and let (Xi)0≤i≤m be the canonical resolution of the marked ideal (I,E,μ). Then: (1) the induced sequence φ∗(Xi)0≤i≤m is an extension of the canonical resolution of φ∗(I,E,μ); (2) for every x′∈supp⁡(φ∗(Ii,Ei,μ)) the invariants agree: inv⁡(x′)=inv⁡(φi(x′)),ν(x′)=ν(φi(x′)),ρ(x′)=ρ(φi(x′)).

Facts & Assumptions

Given: A marked ideal (I,E,μ) with I≠0, its canonical resolution from Canonical resolution of marked ideals with the sequence of values ord⁡N(Ir0)>⋯>ord⁡N(Irk) read along Step 2a, and an étale morphism φ.

[A1]

The Axiom of Choice: AC is assumed for the canonical-resolution consumer clauses and their cited AC-dependent construction suppliers.

[F1]

Canonical resolution of marked ideals, The monomial part, the non-monomial part and the companion ideal: Step 2a resolves the companion ideal O(I,μ), which is of maximal order, and the resolution strictly decreases ord⁡N on the support; the process terminates either with empty support or in residual order zero, where I=M(I) on a neighbourhood of the support, which is handled by the discrete invariant ν of Step 2b.

[F2]

Addition and multiplication of marked ideals, The coefficient ideal is equivalent to the marked ideal: the decomposition I=M(I)N(I), the companion ideal and its support identity commute with the sum and product operations; pointwise residual orders are preserved by étale pullback. A global maximum on a nonsurjective étale image can be smaller; equal maxima are required only in the matching companion pass.

[F3]

Etale commutativity of the maximal-order resolution step: the canonical resolution of a maximal-order marked ideal commutes with étale morphisms, with equality of invariants.

[F4]

Etale pullback commutes with derivative ideals, Order and simultaneous normal crossings are preserved by smooth morphisms: étale pullback commutes with derivative ideals, and the monomial part pulls back to the monomial part with the same exponents; hence φ∗(N(I))=N(φ∗I) and ord⁡x′N(φ∗I)=ord⁡φ(x′)N(I) at corresponding points. The maxima over the two supports need not be equal.

Proof

1.1A1F1F2F4

The trichotomy. If the pulled-back support is empty, every center has empty inverse image, so the induced sequence consists of isomorphisms and the assertion is immediate. Otherwise, along Step 2a we compare ord⁡N(Irl) with ord⁡N(φ∗(Irl)); by [F4] the pointwise residual orders agree; the global maximum on the image can be smaller, so the pullback may omit a companion pass. If the value at stage rl exceeds the value of the pullback, the centers of (Xi)rl≤i<rl+1 lie in the locus where ord⁡N attains its maximal value, which does not meet the image of φ, so the induced morphisms are isomorphisms; if the values agree, the companion ideals correspond, φ∗(O(Irl))=O(φ∗Irl), and [F3] gives the commutativity of the maximal-order step together with equality of the invariants.

2.1A1F1F4step 1.1∎

The monomial end and conclusion. If the residual maximum is zero, restrict to the open neighbourhood of the support where N(Irk) is a unit. There Irk=M(Irk) is monomial, as is its pullback by [F4], and both resolutions are controlled by the invariant ρ on subsets of E; the ordered boundary labels and exponents identify the pointwise values of ρ and ν. If the image misses the current global maximum of ρ, the inverse center is empty and its blowup pulls back to an isomorphism. If it meets that maximum, the pulled-back center is precisely the maximal ρ locus on the pullback. Iterating these two cases gives the same nonempty centers and invariant values, with isomorphism steps inserted where a larger maximum is missed. All centers lie in the support, so these local monomial sequences extend by the identity off it and agree on overlaps. Values on lower residual-order strata skipped by earlier companion passes are transferred through their unchanged lifts from the first later applicable pass, as in the proposition; the same transfers commute with étale pullback. Assembling the finitely many stages proves (1) and (2).

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Canonical resolutions commute with smooth morphisms

Statement

Assume the Axiom of Choice (The Axiom of Choice).

Let (I,E,μ) be a marked ideal with μ≥1 and with I not identically zero on any irreducible component of the smooth finite-type K-scheme X, and let φ ⁣:X′→X be a smooth morphism of relative dimension n with X′ of finite type over K (Smooth morphism of schemes, Relative dimension of a smooth morphism at a point). Let (Xi) be the canonical resolution of (I,E,μ) (Canonical resolution of marked ideals). Then: (1) φ∗(Xi) is an extension of the canonical resolution of φ∗(I,E,μ), locally obtained by pulling back the centers on X×An along an étale factorization of φ; (2) for every x′∈supp⁡(φ∗(Ii,μ)) the invariants agree: inv⁡(φi(x′))=inv⁡(x′),ν(φi(x′))=ν(x′),ρ(φi(x′))=ρ(x′).

Facts & Assumptions

Given: Assume AC. A marked ideal (I,E,μ) with μ≥1 and generic nonvanishing on every component of a smooth finite-type K-scheme X, and a smooth morphism φ ⁣:X′→X of relative dimension n with X′ of finite type over K.

[A1]

The Axiom of Choice: AC is used through the coefficient-ideal smooth-pullback equality in [F4].

[F1]

Smooth morphism of schemes, Order and simultaneous normal crossings are preserved by smooth morphisms: smooth morphisms preserve orders, hence supports, and Flat maps with geometrically regular fibres have standard smooth local presentations factors a smooth germ locally as an étale morphism U′→X×An followed by the projection. At each point the local ring map is flat and local, hence faithfully flat: every proper ideal extends into the target maximal ideal, and flatness preserves injections of nonzero cyclic modules. Thus the pullback of a nonzero ideal stalk remains nonzero; in particular φ∗I is generically nonzero on every irreducible component of X′. The explicit finite-type hypothesis on X′ and the constant relative dimension over the pure-dimensional marked-ideal ambient X make X′ a smooth pure-dimensional finite-type ambient, satisfying the resolution proposition's hypotheses.

[F2]

Smooth base change of multiple test blow-ups: the pullback of the canonical resolution of I is a multiple test blow-up of φ∗I, and if the original resolves then so does the pullback.

[F3]

Canonical resolution of marked ideals, Etale commutativity of the maximal-order resolution step, Etale commutativity of the companion-ideal step: the canonical resolution is natural under étale morphisms, with equality of the invariants; its derivative operations commute with pullback, and its homogenization and coefficient operations do so on their maximal-order inputs, in particular on the positive-residual-order companions.

[F4]

The coefficient ideal commutes with smooth pullback, Homogenization commutes with smooth pullback: under AC, coefficient ideals commute with smooth pullback; homogenizations also commute with smooth pullback.

[F5]

Canonical resolutions with invariants of a marked ideal: the invariants determine the centers, so equality of invariants for the induced and the canonical resolution suffices for them to coincide up to extension.

Proof

1.1A1F1F2F3

The projection case: direct branches. For π ⁣:X×An→X, induct on the dimension of the base X, for all positive-marked, componentwise generically nonzero inputs. Dimension zero has empty support. Consider first a maximal-order pass of the algorithm in [F3]. Orders and boundary counts agree under projection by [F1], and boundary strata pull back to their products with An. In the contained-stratum Step 1aa, containment in the support is preserved and reflected by this surjective projection. The stratum is regular and SNC with the boundary; its blowup and controlled division commute with projection by [F2]. Both sides give it the same encoded count/infinity primary value, ν=0 and ρ=∅. Its restricted ideal may be zero, so no lower-dimensional induction is used here. In the nonboundary Step 1ba, the isolated codimension-one support components likewise pull back to their products. The proposition proves their SNC with the current boundary and their local equation J=(uμ); the labelled Cartier blowup divides by uμ and gives the unit ideal near the component on both sides. Its encoded infinity branch and zero auxiliary values therefore agree directly.

2.1A1F1F2F3F4F5step 1.1

The projection case: inductive branches and general inputs. After the contained strata are removed, each retained boundary-stratum restriction is generically nonzero; after the codimension-one components are removed, the coefficient restriction to each maximal-contact hypersurface is generically nonzero. Indeed, the support identities in the proposition identify these restricted supports with proper subsets of their respective components. Projection preserves these identities, and the products of the components remain generically outside the restricted supports. These restrictions have base dimension smaller than dim⁡X, so induction now compares their resolutions with their products, including all three invariants. The homogenization and coefficient operations commute with projection by [F4]; derivatives in the new coordinates add no generators, since local extended ideals are generated by functions from X and Leibniz's rule differentiates only their coefficients in those directions. The chosen maximal-contact hypersurfaces can therefore be their products, and [F3] ensures independence of those choices. For a general input, the monomial exponents, residual orders and threshold subsets are unchanged by projection, so companion passes reduce to the compared maximal-order passes and the monomial branch has identical values and centers. The proposition's successive assignment transports the remaining values through unchanged lifts on both sides. Thus at every stage inv⁡π∗I(z)=inv⁡I(π(z)), and likewise for ν and ρ; [F5] gives the same centers, while [F2] identifies their blowups and transforms. Hence the canonical sequence is (Xi×An).

3.1A1F1F2F3step 2.1∎

The étale case and conclusion. For an étale morphism ψ this is [F3]: the induced resolution is an extension of the canonical one and the invariants agree. For a general smooth φ with local factorization φ=π∘ψ, ψ étale and π the projection, the canonical resolution of φ∗I=ψ∗π∗I is obtained by first taking the projection case for π∗I and then the étale case for ψ; both steps preserve the invariants, so inv⁡(φ(x′))=inv⁡(π(ψ(x′)))=inv⁡(ψ(x′))=inv⁡(x′), and analogously for ν and ρ.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Canonical resolutions over non-algebraically-closed ground fields

Statement

Assume AC (The Axiom of Choice).

Let K be a field of characteristic zero, K‾ an algebraic closure (An algebraically closed field: every nonconstant polynomial has a root in the field), X a smooth K-scheme of finite type and (I,E,μ) a marked ideal on X with μ≥1 and I not identically zero on any irreducible component (Marked ideals and their support). Then (I,E,μ) admits a canonical resolution over K: base changing to K‾ and taking the canonical resolution of (IK‾,EK‾,μ) gives a Aut⁡K(K‾)-equivariant resolution, where Aut⁡K(K‾) denotes the automorphisms fixing K, which descends to a resolution of (I,E,μ) over K. This resolution commutes with smooth morphisms and, for mark-one inputs with empty boundary, has the ambient-embedding comparison of Canonical resolutions commute with embeddings of ambient smooth schemes, and is natural under isomorphisms of the ground field.

Facts & Assumptions

Given: A marked ideal (I,E,μ) with μ≥1 and generic nonvanishing on every component of a smooth finite-type K-scheme X, with K of characteristic zero, an algebraic closure K‾ of K, the base change (X‾,I‾,E‾,μ) over K‾, and the Galois group G=Aut⁡K(K‾).

[A1]

The Axiom of Choice: AC is assumed for the canonical-resolution consumer clauses and their cited AC-dependent construction suppliers.

[F1]

Canonical resolution of marked ideals: over the algebraically closed field K‾ the marked ideal (I‾,E‾,μ) admits a canonical resolution (X‾i)0≤i≤m with invariants satisfying the conditions of Canonical resolutions with invariants of a marked ideal.

[F2]

Canonical resolution under isomorphisms of the ground field: the canonical resolution is natural under semilinear isomorphisms of the ground field and schemes; for every σ∈G=Aut⁡K(K‾), the induced semilinear automorphism of X‾ transports the canonical resolution of (I‾,E‾,μ) to that of its pullback, which equals the same marked ideal because the original data are defined over K. Thus σ acts on the canonical resolution.

[F3]

For a G-stable ideal on XK‾, the finite-dimensional argument in step 2.1 below proves ideal descent. No quasi-coherent sheaf-descent theorem is inferred from the definitions of fields or Galois groups.

[F4]

Canonical resolutions commute with smooth morphisms, Canonical resolutions commute with embeddings of ambient smooth schemes: the canonical resolution over K‾ commutes with smooth morphisms; the closed-ambient-embedding comparison applies to the mark-one empty-boundary inputs of the cited embedding lemma.

[F5]

Galois fixed points recover finite-dimensional scalar extensions: a finite-dimensional semilinear space over a finite Galois extension L/K is L-spanned by its invariant vectors; for the canonical scalar extension of a finite-dimensional K-space, the fixed vectors are exactly that K-space.

[F6]

Assuming Choice, a base-field embedding extends across every algebraic extension: under AC each automorphism of a finite Galois subextension of K‾/K extends to a K-automorphism of K‾. The extension embedding is onto because its image is algebraically closed and K‾ is algebraic over that image.

[F7]

Field tests for geometric regularity, clause (3), and Locally standard smooth iff flat with geometrically regular fibres, field case: a finite-type K-algebra whose extension to K‾ is geometrically regular is geometrically regular, hence smooth over K.

Proof

1.1A1F1F2

Galois equivariance. The positive marking is unchanged by base change, and generic nonvanishing persists: the field extension K⊆K‾ is flat, and each generic point of a component of XK‾ lies over a generic point of a component of X, where I is the unit ideal. Thus the base-changed marked ideal satisfies the proposition's hypotheses. By [F2] each σ∈G maps the canonical resolution (X‾i) with its centers C‾i to the canonical resolution of the same marked ideal; by uniqueness of the canonical resolution (its invariants are intrinsic) this conjugate resolution agrees with the original one, so the centers C‾i and the invariant strata are G-stable.

2.1A1F2F5F6F7step 1.1algebra

Prove descent on an affine Spec⁡A⊆Xi defined over K, with ideal JK‾ of the invariant center. For f∈JK‾ choose a finite-dimensional K-subspace V⊆A containing its finitely many coefficient vectors, and a finite Galois subextension L/K containing its scalar coefficients (adjoin the finitely many roots of their minimal polynomials). The space W=JK‾∩(V⊗KL) is stable under Gal⁡(L/K) by [F6]. By [F5], f is an L-linear combination of vectors in WGal⁡(L/K)⊆V⊆A. These invariant vectors belong to J:=JK‾∩A, so JK‾=J(A⊗KK‾). The ideal J is finitely generated because A is Noetherian. Intersections with A commute with localization: if a localized class lies in the extended ideal, some power of its denominator multiplies its numerator into that ideal and hence into J. Thus these affine ideals glue uniquely. By [F7] their quotient rings define smooth centers Ci whose scalar extensions are the original C‾i.

3.1A1F1F2F4step 2.1∎

The resolution over K. The blowup construction commutes with the faithfully flat base change K→K‾ (Flat base change for blowups, and failure without flatness), so the sequence Xi+1=Bl⁡CiXi over K base changes to X‾i+1; by construction the supports satisfy supp⁡(Ii,μ)K‾=supp⁡(I‾i,μ) by the derivative-support equality in characteristic zero, since derivative ideals commute with separable algebraic scalar extension; at the last stage both are empty, hence supp⁡(Im,μ)=∅ over K and the sequence is a resolution of (I,E,μ) over K. The closed superlevels of inv⁡ and the lexicographic pairs (inv⁡,ν) and (inv⁡,ρ) are G-stable by step 1.1 and descend by the same ideal argument in step 2.1. Their finite ranges reconstruct unique functions over K: finite differences of primary superlevels determine {inv⁡=a}, and restricting the descended pair thresholds (a,b) to this stratum gives its auxiliary superlevels. These sets determine each auxiliary value, are relatively closed on that stratum, and pull back to the original level sets. The descended primary and pair superlevels are closed, preserving the center maxima and every pointwise descent comparison. Thus the resolution is canonical, determined by the intrinsic resolution over K‾; and it commutes with smooth morphisms and embeddings over K by applying [F4] after base change to K‾ and descending the equal center ideals by step 2.1. The blowups and their natural comparison maps are then defined over K by their Rees-algebra constructions. SNC of the boundary and center descends as well: the individual divisors and their intersection quotients are smooth after scalar extension, and the exact codimensions are preserved by field extension, giving the strict normal crossings criterion (Stacks, tag 0BIA).

TheoremStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Canonical principalization of ideals in characteristic zero

Statement

Assume AC (The Axiom of Choice).

Let K be a field of characteristic zero, X a smooth K-scheme of finite type and I⊆OX a coherent ideal sheaf that is not identically zero on any irreducible component of X (Coherent module sheaves). Then there is a canonical principalization of I: a sequence X=X0←X1←⋯←Xr=X~ of blowups of regular centers Ci−1⊆Xi−1 (Blowup of a scheme along an ideal sheaf) such that (a) the exceptional divisor Ei of the composite σi ⁣:Xi→X has only simple normal crossings and Ci−1 has SNC with Ei−1 (Simple normal crossings divisors and simultaneous normal crossings position); (b) the total transform σr∗I is the ideal of an effective Cartier divisor with simple normal crossings support E~ which is a natural combination of the irreducible components of Er. The morphism (X~,σr∗I)→(X,I) commutes with smooth morphisms and with embeddings of ambient smooth schemes, and is equivariant under every group action on X preserving I (not necessarily preserving K).

Facts & Assumptions

Given: A field K of characteristic zero, a smooth finite-type K-scheme X, and a coherent ideal sheaf I⊆OX that is not identically zero on any irreducible component of X.

[A1]

The Axiom of Choice: AC is assumed for the canonical-resolution consumer clauses and their cited AC-dependent construction suppliers.

[F1]

Canonical resolution of marked ideals, Canonical resolutions over non-algebraically-closed ground fields: on each of the finitely many disjoint open-and-closed pure-dimensional components of the smooth X, the marked ideal (I,∅,1) admits a canonical resolution σ ⁣:X~→X, a sequence of blowups of regular centers with SNC with the successive exceptional divisors, over K.

[F2]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals, Exceptional subscheme of a blowup: the controlled transform is Ii=I(Di)−1σi∗Ii−1; a controlled transform with empty support is the unit ideal, since a sheaf of ideals with nowhere-vanishing stalks is O.

[F3]

Canonical resolutions commute with smooth morphisms, Canonical resolutions commute with embeddings of ambient smooth schemes, Canonical resolution under isomorphisms of the ground field: within finite-type smooth ambient schemes, the marked-ideal construction commutes locally with smooth morphisms of constant relative dimension and with ambient embeddings; decomposing into the open relative-dimension loci gives the general smooth comparison, and with semilinear isomorphisms carrying the input ideal to its pullback.

[F4]

Simple normal crossings divisors and simultaneous normal crossings position, Blowups of finite type ideals are locally H-projective, and proper, Proper morphisms: blowups of regular centers are proper; strict transforms of exceptional divisors together with the new exceptional divisor form a family in simultaneous SNC position along the process.

[F5]

Noether normalisation yields module finiteness over a polynomial subring, In a nonzero commutative ring, every proper ideal is contained in a maximal ideal, A field finitely generated as a k-algebra is a finite extension of k, Lying over for integral ring maps: under AC a finite-type domain is finite over a polynomial subring; a nonzero finite-type algebra over a field has a maximal ideal with finite residue extension; integral inclusions satisfy lying over.

Proof

1.1A1F1F2F4

Principalization from the resolution. Let (Xi)0≤i≤r be the canonical resolution of (I,∅,1) from [F1] and let (Ii,1) be the controlled transforms. Since supp⁡(Ir,1)=∅, the ideal Ir is the unit ideal by [F2]; unwinding the transform rule we get σi∗Ii−1=I(Di)Ii at each step, so the total transform σr∗I is a product of powers of the exceptional divisors Di and their strict transforms, each a component of Er by [F4]. Hence σr∗I is the ideal of an effective Cartier divisor E~ with SNC support which is a natural combination of the irreducible components of Er, and by [F4] both the exceptional divisors and the centers are in SNC position: clauses (a) and (b).

1.2A1F5givenalgebrachoose

Intrinsic constants on a component. For an integral open-and-closed component Z of X, put R=Γ(Z,OZ) and let L be its elements algebraic over K. The minimal polynomial expresses the inverse of each nonzero such element as a polynomial in it, so L is a field. On an affine chart Spec⁡A⊆Z, [F5] makes A finite over a polynomial ring K[z]. Every finite subextension L0/K gives a subfield L0(z) of A⊗K[z]K(z), so [L0:K] is bounded by the module-generator count; choosing a maximal degree proves L/K finite, hence separable. Any field subring of R must lie in L: if it contains f transcendental over K, every f−q, q∈Q, is a unit. By [F5], the nonzero finite-type K(f)-algebra A⊗K[f]K(f) has a finite residue field E/K(f). Images bi of finitely many generators of A become integral over K[f,1/p] after clearing finitely many denominators, with p≠0. The algebra B=K[f,1/p,b1,…,bs] is integral over K[f,1/p] and receives a map from A. Choose q∈Q with p(q)≠0; lying over supplies a prime containing f−q in B, contradicting its being the image of a unit of A. Thus L is the unique largest field subring of R. Scheme automorphisms, including those permuting components, consequently induce isomorphisms of these intrinsic fields.

2.1A1F1F3step 1.1

Canonicity and smooth naturality. The resolution, hence the principalization, is determined by its canonical invariant centers. The smooth and ambient-embedding comparisons in [F3] transport these centers and the controlled-transform rules, so they transport the total-transform factorization from step 1.1 as well. This proves canonicity and both commutation clauses.

3.1A1F1F3step 1.1step 2.1step 1.2∎

Equivariance for ideal-preserving actions. Regard each component as an L-scheme. A smooth affine L-ambient presentation is smooth over K because L/K is finite separable. Every K-derivation annihilates L by its separable minimal polynomials, so the K- and L-derivative ideals agree; orders, boundary strata, homogenizations, coefficient ideals and companion ideals then agree, giving the same canonical sequence. An automorphism preserving I transports the componentwise marked ideals by a semilinear isomorphism of their intrinsic constant fields. By [F3] it therefore transports each canonical center, and lifts successively to the blowups. These lifts satisfy identity and composition: the blowup lifts induced from the ideal identifications are natural, and equivalently two lifts agree on the dense complement of the centers and hence on the reduced separated smooth resolution. Thus any ideal-preserving abstract group action lifts coherently, including actions not preserving K.

TheoremStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Weak embedded desingularization in characteristic zero

Statement

Assume AC (The Axiom of Choice), inherited from the blowup and canonical-resolution suppliers.

Let K be a field of characteristic zero, X a smooth K-scheme of finite type and Y⊆X a reduced closed subscheme (Integral schemes, Closed immersions of schemes); put IY for its ideal sheaf. Then there is a canonical embedded desingularization of Y in X: a sequence X=X0←X1←⋯←Xr=X~ of blowups of regular centers Ci−1⊆Xi−1 such that (a) the exceptional divisor Ei of the composite has only simple normal crossings and each center has SNC with Ei−1; (b) every center Ci is disjoint from the smooth locus Reg⁡(Y)⊆Yi, the strict transform of Y in Xi; (c) the strict transform Y~:=Yr is smooth and has only simple normal crossings with the exceptional divisor Er; (d) the construction is canonical and commutes with smooth morphisms and with embeddings of ambient smooth schemes. In particular the induced morphism Y~→Y is proper and birational on every irreducible component which is an isomorphism over the smooth locus of Y.

Facts & Assumptions

Given: A field K of characteristic zero, a smooth finite-type K-scheme X, a reduced closed subscheme Y⊆X with ideal sheaf IY, and the marked ideal (IY,∅,1), whose support is Y.

[F1]

Włodarczyk, Simple Hironaka Resolution, §4.7, Theorem 4.7.1 (pp. 25–26), together with §4.4 for descent to non-algebraically-closed fields, and Hauser, The Hironaka Theorem on Resolution of Singularities, §13 (pp. 386–387): the modified canonical algorithm for a reduced closed subscheme gives conditions (a)–(d), and in fact makes the irreducible strict transforms smooth and disjoint and gives the stronger full-transform factorization. It continues only until a strict-transform component would become the next center; the §4.7 induction shows that component is then regular and transverse to the exceptional divisor, after which the modified procedure ignores it and resolves the remaining components.

[F2]

Since a smooth scheme is regular, its irreducible components are disjoint open-and-closed components. The algorithm is applied componentwise; a component contained in Y is smooth and receives the identity sequence.

[F3]

Strict transform of a closed subscheme, Multiple test blow-ups, controlled transforms and resolutions of marked ideals: the strict transform is obtained by saturating the total pullback along exceptional components, while the controlled transform divides the total pullback by the marking at each blow-up. Residual exceptional factors can remain, so the controlled transform need not vanish exactly on the strict transform; the modified algorithm's stopping rule concerns the strict transform itself.

Proof

1.1F1F2F3

The modified canonical sequence. On each open-and-closed ambient component contained in Y, take the identity sequence by [F2]. On the remaining components, use the modified canonical procedure in [F1], not the full support-clearing resolution of (IY,1). Whenever a strict-transform component would next be chosen as a center, the source induction shows it is already smooth and transverse to the exceptional divisor; the modified rule removes that completed component from the active support and continues on the others. Thus no executed center meets Reg⁡(Y). By [F3], this stopping rule concerns the strict transform and does not assume that the controlled transform has no exceptional factors.

2.1F1F3step 1.1∎

The resulting sequence and its properties. The modified procedure in [F1] terminates with smooth strict transform having SNC with the exceptional divisor; its centers are regular and SNC with the exceptional boundary, and the construction is canonical and commutes with smooth morphisms and ambient embeddings. Since the centers avoid Reg⁡(Y), the composite is an isomorphism there. A composition of blowups of regular centers is proper and birational, so its restriction induces the stated proper morphism Y~→Y, birational on each irreducible component. This gives all clauses of the Statement while preserving the source's full-transform strengthening as an additional consequence.

TheoremStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Bravo-Villamayor strengthening of embedded desingularization

Statement

Assume AC (The Axiom of Choice).

Let Y⊆X be a reduced closed subscheme of a smooth K-scheme of finite type over a field K of characteristic zero, with decomposition Y=⋃iYi into irreducible components. Then there is a canonical resolution of Y in X by blowups of regular centers as in Weak embedded desingularization in characteristic zero such that, in addition, the strict transforms Y~i are smooth and pairwise disjoint and the full transform of Y has the form (σ~)∗(IY)=M((σ~)∗(IY))⋅IY~, where IY~ is the ideal sheaf of the disjoint union Y~=∐iY~i and M((σ~)∗(IY)) is the monomial part of the full transform with respect to the exceptional divisors. Componentwise, any irreducible component Xα of X contained in Y receives the identity sequence; on it set the monomial factor to OXα, so the factorization there is exactly 0=OXα⋅0. On every other ambient component the ideal of Y is generically nonzero and the monomial part has its usual meaning.

Facts & Assumptions

Given: A reduced closed subscheme Y=⋃iYi of a smooth finite-type K-scheme X over a field K of characteristic zero, with ideal sheaf IY, and the marked ideal (IY,∅,1) on components not contained in Y; components contained in Y are handled by the identity sequence.

[A1]

The Axiom of Choice: AC is assumed for the canonical-resolution consumer clauses and their cited AC-dependent construction suppliers.

[F1]

Canonical resolution of marked ideals, Canonical resolutions over non-algebraically-closed ground fields, The monomial part, the non-monomial part and the companion ideal supply the original Steps 1–2 on components where the ideal is generically nonzero with positive marking, including the monomial threshold-subset rule and the maximal-order reduction. They do not assert the modified 3/2 branch. That modification is constructed and checked in step 1.1 below, following the complete argument in Włodarczyk §4.7, pp. 25–26. Components of X contained in Y take the identity sequence.

[F2]

Weak embedded desingularization in characteristic zero supplies the embedded stopping convention and conditions (a)–(d). Its source-backed procedure uses the same §4.7 modification checked here; the full-transform identity still requires the local ideal argument of step 1.2.

[F3]

The codimension induction needed here is proved in step 1.2 below, following Włodarczyk, §4.7, pp. 25–26. The coefficient-restriction and tangent-support lemmas supply its reduction to a hypersurface; they do not themselves assert equality with the strict-transform ideal.

[F4]

Codimension-one components of a maximal-order support supplies smoothness, isolation and the local ideal-division calculation for codimension-one maximal-order support components. Their Cartier blowups are admissible controlled transforms only when the components have SNC with the full boundary; after Step 1a, Canonical resolution of marked ideals, Proof 1.3, supplies this SNC condition. Then those blowups remove the components from the support. Coefficient-ideal control with centres allowed off the subvariety keeps subsequent calculations on the remaining strict transforms after separation.

[F5]

Multiple test blow-ups, controlled transforms and resolutions of marked ideals: unwinding controlled transforms factors the total transform into exceptional monomial factors times the residual controlled ideal. Equality of that residual ideal with the strict-transform ideal is the additional assertion established below, not a consequence of the transform formula alone.

Proof

1.1A1F1F2

Construct and check the modified algorithm. Split the smooth ambient into its disjoint open-and-closed components Xfull contained in Y and Xrest; use the identity on the former. On the latter start with (IY,∅,1) and use the original Steps 1–2, with the following extra branch at every recursive mark-one input. After higher residual-order strata have been handled, where the residual order is at most one and M is nonunit, resolve (M,1) before the residual ideal, assigning source invariant (3/2,0,…) and the monomial ν,ρ; where M is a unit use the original residual branch. At mark one each minimal threshold subset is a single positive-exponent boundary divisor, so a maximal-ρ center is a union of disjoint regular components of that divisor. It lies in the marked support, is SNC with the full boundary, and its labelled Cartier blowup reduces its exponent by one. The residual factor N is unchanged, and no new non-monomial singularity is introduced. There are finitely many positive exponents, so these passes terminate; in higher recursive marks use the unchanged finite monomial procedure of [F1]. The rational 3/2 lies strictly between residual orders one and two, so it gives the intended priority on these closed boundary pieces without changing the higher-order passes. Successive assignment on untouched strata, as in [F1], preserves the primary invariant comparisons; the closed monomial strata give the chosen centers. No standalone upper semicontinuity of the auxiliary functions is used here. All centers are defined by ordered boundary equations, exponents and lower-dimensional invariant maxima; these data commute with smooth pullback and semilinear isomorphisms, and the maximal-contact gluing is unchanged. Under an ambient embedding the extra tangent parameters give the same constant prefixes before the induced recursive problem, so this modification commutes with that comparison as well. Thus the dimension induction proving canonicity and SNC centers for the original algorithm applies with this checked finite extra branch. The stopping convention of [F2] preserves the original smooth locus; any exceptional divisor encountered there would require an earlier center meeting that locus. Over nonclosed fields these intrinsic centers are Galois-stable and descend as in [F1].

1.2A1F1F2F3F4

Prove the source's local ideal claim by induction on the codimension c of an irreducible component Z of the support, whose ideal agrees generically with the active mark-one ideal J. At the stage with maximal source invariant (1,0;…;1,0;∞;0,…), with c copies of (1,0), the added monomial branch has already removed residual exceptional factors, and the active input is (J,1) of maximal order. Since H(J,1)=C(J,1)=(J,1), the boundary reduction and maximal-contact reductions act on this same ideal. For c=1, Codimension-one components of a maximal-order support gives J=IZ along the smooth isolated component. For c>1, choose a tangent parameter u∈J and let H=V(u). The coefficient-restriction lemma identifies the induced mark-one ideal and its sequence on H, where Z has codimension c−1. The induction gives JOH=IZ,H near the final strict transform. Since u∈J and Z⊆H, equality modulo (u) lifts to J=IZ in the ambient ring: both are inverse images of the same ideal under its quotient by (u). The smoothness and SNC position of Z also lift from the successive compatible parameter reductions. Thus when such a component would be the next center it is already smooth, isolated in the active support, and its active ideal is exactly its strict-transform ideal. Retain it and omit that blowup; repeat on the remaining components.

2.1A1F1F3F4step 1.2

The remaining components. After Y~1 is separated, all strict transforms of components of codimension r1 are isolated and are ignored in the further resolution; the process continues with the next codimension r2>r1 and the same argument shows that at the moment the strict transform of a codimension-r2 component becomes a center, the controlled transform near it is its ideal. Iterating over the finitely many codimensions and components separates the strict transforms Y~i, which are smooth and pairwise disjoint. Any remaining active support is disjoint from these completed strict transforms and is principalized by the canonical process of [F1].

3.1A1F2F3F5step 1.2step 2.1∎

The full transform. On Xrest, the controlled-transform rule [F5] separates the exceptional monomial factors from the residual ideal. By steps 1.2 and 2.1, that residual ideal is the ideal of each completed strict transform near it and is the unit ideal away from their union after the remaining principalization; these local identities glue. Thus the full transform factors as M((σ~)∗IY)⋅IY~, where the monomial part collects exceptional-divisor factors and the remaining factor is the ideal of the disjoint strict transforms. On Xfull the sequence is the identity, IY=0, and by convention the factorization is 0=OXfull⋅0. This proves the componentwise formula and the theorem.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Independence of the embedded desingularization from the ambient embedding

Statement

Assume AC (The Axiom of Choice).

Let U be an integral affine K-variety of finite type over a field K of characteristic zero and let φ1 ⁣:U↪X1 and φ2 ⁣:U↪X2 be two closed immersions into smooth affine K-schemes (Closed immersions of schemes). Let U~i⊆X~i be the canonical embedded desingularizations of U in Xi (Bravo-Villamayor strengthening of embedded desingularization). Then the induced desingularizations U~1→U and U~2→U are canonically isomorphic over U. More precisely, the amalgamated embeddings Ψ0,Ψ1,Ψ2 ⁣:U→A2n of any two presentations of U as a closed subvariety of affine space are related by automorphisms Φ1,Φ2 of A2n with ΦiΨ0=Ψi, and commutativity of embedded desingularization with ambient embeddings identifies the resulting resolutions over U.

Facts & Assumptions

Given: An affine K-variety U of finite type over a field K of characteristic zero and two closed embeddings φ1 ⁣:U↪X1, φ2 ⁣:U↪X2 into smooth affine K-varieties.

[A1]

The Axiom of Choice: AC is assumed for the canonical-resolution consumer clauses and their cited AC-dependent construction suppliers.

[F1]

Weak embedded desingularization in characteristic zero, Bravo-Villamayor strengthening of embedded desingularization: embedded desingularization of U↪X is compatible with closed embeddings of smooth ambient varieties: if X↪X′ is a closed embedding of smooth varieties and U′ denotes the closure extension, then the desingularization of U in X′ restricts to the desingularization in X.

[F2]

Weak embedded desingularization in characteristic zero, Bravo-Villamayor strengthening of embedded desingularization: for each closed embedding U↪X there is a canonical embedded desingularization U~⊂X~ of U in X.

[F3]

Closed immersions of schemes, Locally finite type and finite type morphisms: the embeddings φi may be composed with closed embeddings Xi↪An (for n large enough to accommodate finite generating lists for both affine coordinate rings) to obtain embeddings ψiφi ⁣:U→An; the coordinate functions of Xi express the two families of generators of K[U].

[F4]

The automorphism lemma (source Lemma 4.8.1). If g1,…,gn,h1,…,hn generate K[U] and Ψ0(x)=(g,h), Ψ1(x)=(g,0), Ψ2(x)=(0,h) are the three embeddings U→A2n, then choosing polynomials wi(h)=gi and vi(g)=hi (possible because the h's generate K[U]) gives automorphisms Φ1(x,y)=(x,y−v(x)) and Φ2(x,y)=(x−w(y),y) of A2n with ΦiΨ0=Ψi; both are polynomial automorphisms with polynomial inverse.

Proof

1.1A1F3F4

Reduction to a common ambient space. By [F3], compose the given embeddings with embeddings of Xi into a common An. Their coordinate functions give two generating lists g,h for K[U]. In A2n put Ψ0=(g,h), Ψ1=(g,0) and Ψ2=(0,h). By [F4], ΦiΨ0=Ψi, hence Φi−1Ψi=Ψ0. Thus after the coordinate inclusions An↪A2n, the inverse polynomial automorphisms carry each presentation to the same closed embedding Ψ0.

2.1A1F1F2F4step 1.1∎

Comparison of the resolutions. Use [F2] to resolve the common embedding Ψ0. The compatibility with closed smooth ambient embeddings in [F1] identifies the resolution induced by each Xi with that induced by its coordinate inclusion in A2n. Transport along Φi−1, using the naturality of the canonical construction under ambient automorphisms, then identifies it with the resolution of Ψ0. Both comparisons are over U, so their composite canonically identifies U~1→U with U~2→U.

LemmaStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Open restrictions of the canonical desingularization

Statement

Assume AC (The Axiom of Choice).

In the setting of Independence of the embedded desingularization from the ambient embedding, let V→U be an open immersion of integral affine K-varieties (Open immersions of schemes) and let res⁡U ⁣:U~→U, res⁡V ⁣:V~→V be the canonical desingularizations. Then there is an open immersion V~↪U~ lifting V→U such that V~→res⁡U−1(V) is an isomorphism over V.

Facts & Assumptions

Given: An open embedding V↪U of affine K-varieties of finite type over a field K of characteristic zero, with canonical desingularizations V~→V and U~→U.

[A1]

The Axiom of Choice: AC is assumed for the canonical-resolution consumer clauses and their cited AC-dependent construction suppliers.

[F1]

Weak embedded desingularization in characteristic zero, Canonical resolutions commute with smooth morphisms: embedded desingularization commutes with smooth ambient morphisms; an open immersion is étale, hence smooth of relative dimension zero.

[F2]

Weak embedded desingularization in characteristic zero, Independence of the embedded desingularization from the ambient embedding: the canonical desingularization of an affine variety is obtained by an embedded desingularization in any smooth affine ambient variety and is independent of that choice.

[F3]

Open immersions of schemes, Locally finite type and finite type morphisms: in the principal case V=Uf for a regular function f; one may choose a closed embedding U↪X into a smooth affine X and a function F on X restricting to f, so that Uf↪XF is a closed immersion and XF↪X is an open immersion of smooth affine schemes.

Proof

1.1A1F1F2F3

The principal case. Assume first V=Uf with f∈K[U]. By [F3] choose a smooth affine ambient X and F∈K[X] with F∣U=f; the open immersions XF↪X and Uf↪U are étale, so [F1] gives a canonical identification of the embedded desingularization of Uf in XF with the restriction of the embedded desingularization of U in X; by [F2] this is the required open embedding of the canonical desingularizations over Uf.

2.1A1F1F2step 1.1∎

The general case. For an arbitrary affine open V↪U, the complement is defined by finitely many functions; covering V by principal opens Ug contained in V and applying step 1.1 to each, the identifications glue because the desingularizations are canonical and their restrictions to the intersections agree by the same argument. This yields an open embedding V~↪U~ lifting V↪U, with V~→res⁡U−1(V) an isomorphism over V.

TheoremStatement: Literature-sourcedProof: AI-adaptedOpen item page →

Resolution of singularities in characteristic zero

Statement

Assume the Axiom of Choice (The Axiom of Choice).

Let K be a field of characteristic zero and let Y be an integral separated scheme of finite type over K (a K-variety, Integral schemes, Locally finite type and finite type morphisms). Then there exists a resolution of singularities of Y: a smooth K-scheme Y~ together with a proper birational morphism res⁡Y ⁣:Y~→Y (Proper morphisms, Birational morphisms of integral finite-type schemes), constructed on affine charts by restricting compositions of blowups of regular centers in smooth ambient schemes, which is an isomorphism over the smooth locus of Y, and which is canonically determined by Y. The resolution is functorial for smooth morphisms: for every smooth morphism Y′→Y of K-varieties there is a natural lifting Y~′→Y~ which is again smooth, making the square commute up to the canonical identification Y~′=Y~×YY′. It is equivariant under every group action on Y, whether or not the action preserves K.

Facts & Assumptions

Given: The Axiom of Choice and a K-variety Y of finite type over a field K of characteristic zero (integral, separated), covered by finitely many affine open subschemes.

[F1]

Weak embedded desingularization in characteristic zero, Bravo-Villamayor strengthening of embedded desingularization: every closed embedding of an affine chart into a smooth affine variety admits a canonical embedded desingularization.

[F2]

Independence of the embedded desingularization from the ambient embedding: the canonical desingularization of an affine variety is independent of the chosen smooth ambient embedding.

[F3]

Open restrictions of the canonical desingularization: for an open embedding V↪U the canonical desingularizations restrict compatibly: V~↪U~ identifies V~ with the restriction of U~ over V.

[F4]

Canonical resolutions over non-algebraically-closed ground fields, A field's prime subfield is isomorphic to Q in characteristic zero and to Fp in characteristic p: the construction is carried out over K after the characteristic-zero conventions of the page; no algebraic closedness is required.

[F5]

Proper morphisms, Birational morphisms of integral finite-type schemes: blowups are proper, and the local embedded resolutions in [F1] induce proper birational morphisms on their strict transforms.

[F6]

Extending an étale morphism to a smooth ambient neighbourhood: at a rational point an étale germ extends to an étale map of smooth ambient neighbourhoods, with the source germ equal to the inverse image of the target germ. Smooth germs factor locally as an étale germ followed by projection.

[F7]

Canonical resolutions commute with smooth morphisms and the embedded procedure in [F1]: marked ideals, their invariants and centers commute with smooth ambient pullback, up to omission of empty centers; the modified embedded procedure stops or ignores a completed strict-transform component by its regularity and transversality, which are also preserved by smooth pullback.

[F8]

Canonical resolution under isomorphisms of the ground field: the canonical marked-ideal construction is natural under semilinear isomorphisms. The embedded stopping rule in [F1] is invariant under these isomorphisms as well.

[F9]

In a nonzero commutative ring, every proper ideal is contained in a maximal ideal, A field finitely generated as a k-algebra is a finite extension of k, Lying over for integral ring maps, Noether normalisation yields module finiteness over a polynomial subring: under AC a nonzero finite-type algebra over a field has a maximal ideal with finite residue extension; integral inclusions have lying over; a finite-type domain is module-finite over a polynomial subring.

Proof

1.1F1F2given

Local construction. Choose a finite affine cover Y=⋃iUi with closed embeddings Ui↪Xi into smooth affine K-schemes. By [F1] each strict transform U~i is smooth and gives a proper birational morphism to Ui which is the identity over its smooth locus. By [F2] this resolution is independent of its ambient embedding.

1.2F9algebrachoose

An intrinsic field of constants. Put R=Γ(Y,OY), embedded in the function field using any nonempty affine chart. The elements of R algebraic over K form a field L: sums and products remain algebraic, and the minimal polynomial of a nonzero algebraic element expresses its inverse as a polynomial in that element. For a nonempty affine chart Spec⁡A, [F9] makes A finite over K[z1,…,zd]. If L0/K is any finite subextension of L, the zi remain algebraically independent over L0 and L0(z1,…,zd) embeds in the finite-dimensional algebra A⊗K[z]K(z). Thus [L0:K] is bounded by the number of module generators of A. Among these degrees choose a maximal one; adjoining any further element of L cannot increase it, so L/K is finite, hence separable.

2.1F2F3F5step 1.1

Gluing. On Ui∩Uj, [F3], applied on affine open subcharts, identifies the two restrictions. These identifications satisfy the cocycle condition: each is the identity over the dense smooth locus, and two morphisms from an integral reduced scheme to a separated scheme agreeing on a dense open agree everywhere (on affine target charts the difference of each pair of pulled-back functions vanishes in the function field). Gluing gives Y~→Y. Smoothness and finite type are local on these charts; properness is local on the target and follows from [F5]. Birationality and the isomorphism over the smooth locus follow from step 1.1. The same unique comparisons show independence of the affine cover.

2.2F2F4F6F7step 1.1

Étale comparison. First work over an algebraic closure of K. Geometric charts can be reduced with several irreducible components. The proofs of [F2] and [F6] still apply to them: the former uses only coordinate generators and polynomial ambient automorphisms, and the latter uses the completed local-ring isomorphism and graph equations, with no integrality assumption. Comparisons are unique on a reduced source by checking equality on the dense smooth open of every component. At each closed point of an étale morphism V→U, choose affine charts and use [F6] to realize it as a Cartesian restriction of an étale ambient map X′→X. Its defining ideal is the pullback of that of U. By [F7] the ambient centers and marked transforms pull back stage by stage. Blowups commute with flat base change: their Rees algebras pull back because flatness preserves every inclusion of an ideal power. Strict transforms commute as well: saturation by an exceptional equation is a filtered union of kernels of multiplication maps, all preserved by flat pullback. Thus the modified embedded procedure has the same strict transforms and stopping decisions after pullback, omitting only empty centers, and its final strict transform is U~×UV. By [F2] this is the canonical resolution of V. Closed points cover the comparison loci by open neighbourhoods, since a finite-type scheme over an algebraically closed field has a closed point in every nonempty closed subset. By [F4] the canonical centers over K are the descended geometric centers; equality of their ideals and the resulting comparison descend by faithful flatness. This proves the étale comparison over K.

2.3F9step 1.2algebrachoose

Every field subring of R lies in L. Otherwise it contains an element f transcendental over K. Its prime field is Q, so every f−q, q∈Q, is a unit in R and in A. The nonzero finite-type K(f)-algebra A⊗K[f]K(f) has a maximal ideal with finite residue field E/K(f) by [F9]. Write bi∈E for the images of finite K[f]-algebra generators of A. Clearing denominators in their monic equations over K(f) gives a nonzero polynomial p such that B=K[f,1/p,b1,…,bs] is integral over K[f,1/p], and there is a map A→B. Choose q∈Q with p(q)≠0. Lying over supplies a prime of B above (f−q), contradicting the image of the unit f−q∈A. Hence every field subring is contained in L, so L is the unique largest field subring of R and is preserved by every scheme automorphism of Y.

3.1F2F3F6F7step 2.1step 2.2

Smooth comparison. For U×Ar→U, pull back a chosen embedded presentation along X×Ar→X. The same center, blowup and strict-transform calculations of step 2.2 give the canonical resolution U~×Ar. A smooth germ factors as an étale map to U×Ar followed by projection, so step 2.2 and this product case give Y~′≅Y~×YY′ locally for every smooth Y′→Y. The comparisons glue by the uniqueness argument of step 2.1, and their composite for two smooth maps is the same canonical comparison. The lifted map to Y~ is smooth because it is the base change of Y′→Y.

4.1F1F2F4F8step 2.1step 3.1step 1.2step 2.3∎

Equivariance. Regard Y as an L-variety using L⊆R; it is still integral, separated and finite type. Its canonical resolution over L equals the one over K: choose smooth affine L-ambient presentations, which are smooth over K because L/K is finite separable, and use [F2]. In these ambients every K-derivation kills L (differentiate the separable minimal polynomial), so the K- and L-derivative ideals coincide. Orders, boundary components, homogenized and coefficient ideals, monomial parts and companion ideals therefore coincide, as do the invariant centers and the modified embedded stopping rule. Each scheme automorphism of Y is semilinear over its induced automorphism of L by step 2.3; [F8] then transports each center and lifts it through the blowups and gluing. The lift is unique by step 2.1, so lifts preserve identity and composition and yield a group action on Y~. This proves equivariance even when the original action does not preserve K, and step 3.1 proves the promised smooth functoriality.

LemmaStatement: Literature-sourcedProof: AI-adaptedjudge pass (gpt-6.1-sol)Open item page →

Resolution of singularities is functorial under smooth morphisms

Statement

Assume the Axiom of Choice (The Axiom of Choice).

Let K be a field of characteristic zero, let Y be an integral finite-type K-scheme and let φ ⁣:Y′→Y be a smooth morphism of K-varieties, with both Y and Y′ connected of finite type over K (Smooth morphism of schemes, Integral schemes). Then the canonical resolutions of Resolution of singularities in characteristic zero are compatible with φ: the natural morphism φ~ ⁣:Y~′→Y~,Y~′=Y~×YY′, is smooth and makes the square with res⁡Y and res⁡Y′ commute, and the identification is canonical. In particular a smooth morphism is resolved by the base change of the resolution of its target, and, for y~∈Y~ with image y∈Y, the fibre of φ~ at y~ is Yy′×κ(y)κ(y~).

Facts & Assumptions

Given: The Axiom of Choice; a smooth morphism φ ⁣:Y′→Y of K-varieties of finite type over a field K of characteristic zero; and the canonical desingularizations res⁡Y′ ⁣:Y~′→Y′ and res⁡Y ⁣:Y~→Y of Resolution of singularities in characteristic zero.

[F1]

Resolution of singularities in characteristic zero, proof steps 2.2–3.1: the canonical resolutions are compatible with smooth base change; the proof uses the earlier ambient-extension and marked-ideal smooth-commutation results, the marked-ideal construction independently of this consequence.

[F2]

Smoothness survives base change and composition: smooth morphisms remain smooth under base change.

Proof

1.1F1given

Apply the already proved smooth comparison [F1] to φ. It gives the canonical isomorphism Y~′≅Y~×YY′ over Y′. Composing it with the first projection defines φ~ and makes the required square Cartesian, hence commutative. Naturality for composites and identities follows from the canonical comparisons in [F1].

2.1F1F2step 1.1∎

The projection Y~×YY′→Y~ is the base change of φ, so it is smooth by [F2]. For y~↦y, its fibre is Y′×YSpec⁡κ(y~)=Yy′×κ(y)κ(y~) by associativity of fibre products. Thus the base change resolves the source and has the stated fibres.

5 · Examples, counterexamples and false statements

None yet.

Sources