Derivative ideals #
On a smooth scheme the first derivative D(I) of an ideal sheaf I is the ideal generated by
all derivatives of elements of I, and Dʳ⁺¹(I) := D(Dʳ(I)); in local coordinates
D(f₁, …, fₛ) = (fᵢ, ∂fᵢ/∂xⱼ) [Kol07, Definition 73]. Here, for a regular local ring R ⊇ ℚ
with a coordinate structure c (Hironaka/Algebra/Local/Coords.lean), the derivative ideal is
defined with the fixed derivations c.pderiv i; the spanning condition c.SpansDerivations (the
∂ᵢ span all derivations) identifies it with Kollár's coordinate-free definition (Dimg_eq_D).
Main declarations #
RegularCoords.D,RegularCoords.Dpow, andRegularCoords.markedD, the marked derivativeDʳ(I, m) = (Dʳ I, m - r)of [Kol07, Definition 73, (73.3)].le_D,D_le_iff,pderiv_mem_D,D_mono,Dpow_mono,Dpow_mono_left,le_Dpow;D_span,Dpow_span(generators and iterated partials);D_sup,D_iSup;D_changeCoords(independence of the coordinates);D_mul_le(the product rule of [Kol07, Lemma 74(2)] forr = 1);D_maximalIdeal_pow,Dpow_maximalIdeal_pow;Dimg_eq_D.Dpow_eq_top_of_ord_le:ord I ≤ m → Dᵐ(I) = R, Kollár's remark that theDʳ(I)form an ascending chain ending in𝒪_X[Kol07, Definition 73].- The local forms of [Kol07, Lemma 74]:
Dpow_Dpow(iteration, (1)),Dpow_mul_le(the product rule, (2)),le_ord_iff_le_ord_Dpow(the order under derivatives, (3), characteristic zero only),D_map(compatibility with a ring map compatible with the coordinates, the local content of (4)); and the Euler identityeuler_sub_nsmul_membehind (3).
That the top-order partials alone generate Dʳ(I) in characteristic zero (Kollár's
parenthetical remark in Definition 73) is proved in Hironaka/Algebra/Local/TopPartials.lean.
The first derivative ideal [Kol07, Definition 73, (73.1)]: the ideal generated by I and
all ∂ᵢ f, f ∈ I.
Instances For
I ≤ D(I) (built into the definition).
D is monotone.
Higher derivative ideals [Kol07, Definition 73, (73.2)]: D⁰(I) = I,
Dʳ⁺¹(I) = D(Dʳ(I)).
Instances For
Dʳ is monotone in the ideal.
The ascending chain I ⊆ D(I) ⊆ D²(I) ⊆ ⋯ of [Kol07, Definition 73]: Dʳ(I) ≤ Dʳ'(I) for
r ≤ r'.
D(⟨S⟩) is generated by S and the ∂ⱼ s, s ∈ S (Leibniz: ∂ⱼ(a s) = a ∂ⱼ s + s ∂ⱼ a):
Kollár's D(f₁, …, fₛ) = (fᵢ, ∂fᵢ/∂xⱼ) [Kol07, Definition 73].
The iterated partial derivative along a list of indices, ∂_{i₁} ⋯ ∂_{iₖ} f.
Equations
- c.iterPderiv l f = List.foldr (fun (i : Fin n) (g : R) => (c.pderiv i) g) f l
Instances For
The iterated partials of order at most r of the elements of S.
Equations
- c.iterPderivSet S r = ⋃ (l : List (Fin n)), ⋃ (_ : l.length ≤ r), c.iterPderiv l '' S
Instances For
Dʳ(⟨S⟩) is generated by the iterated partials of order at most r of the elements of S
[Wlo05, Definition 2.6.1]; compare the parenthetical remark in [Kol07, Definition 73].
D commutes with finite sums of ideals.
D commutes with arbitrary sums of ideals.
The derivative ideal does not depend on the coordinate structure (Kollár's definition [Kol07, Definition 73] is coordinate-free).
The product rule for derivative ideals, D(IJ) ≤ D(I) J + I D(J): [Kol07, Lemma 74(2)] for
r = 1, [BM08, Lemma 3.2(2)].
D(𝔪^(a+1)) ≤ 𝔪^a, by the product rule.
D(𝔪^a) ≤ 𝔪^(a-1) for a ≥ 1 (and D(𝔪^0) = R, D_maximalIdeal_pow_zero).
Dʳ(𝔪^a) ≤ 𝔪^(a-r) for r ≤ a.
"If max-ord I ≤ m, then Dᵐ(I) = 𝒪_X" [Kol07, Definition 73], in the local form
ord I ≤ m → Dᵐ(I) = R, conditional on the order statement (3) of [Kol07, Lemma 74], which is
proved below (le_ord_iff_le_ord_Dpow); Dpow_eq_top_of_ord_le is the unconditional form.
The marked derivative [Kol07, Definition 73, (73.3)]: Dʳ(I, m) := (Dʳ(I), m - r) for
r ≤ m.
Instances For
Kollár's image definition #
Kollár defines D(I) as the image of Der_X × I → 𝒪_X over all derivations
[Kol07, Definition 73, (73.1)], and notes I ⊂ D(I) through f = ∂(x f) - x ∂f. With the fixed
∂ᵢ the inclusion is built in; the two definitions agree once the ∂ᵢ span the derivations
(SpansDerivations) and a coordinate exists.
Kollár's definition of the derivative ideal: the ideal generated by δ f for all
ℚ-derivations δ and all f ∈ I [Kol07, Definition 73, (73.1)]; it does not refer to
coordinates.
Equations
- IsLocalRing.Dimg I = ⨆ (δ : Derivation ℚ R R), Ideal.span (⇑δ '' ↑I)
Instances For
When the ∂ᵢ span the derivations and there is at least one coordinate, Kollár's image
definition agrees with D; the inclusion I ≤ Dimg I is Kollár's f = ∂₁(x₁ f) - x₁ ∂₁ f
[Kol07, Definition 73].
Kollár's Lemma 74 #
Kollár: "Higher derivatives have the usual properties" [Kol07, Lemma 74], stated without proof
(likewise [Wlo05, Lemma 2.6.2] and [BM08, Lemma 3.2]). Item (1) is iteration; (2) is Leibniz;
(3) rests on the Euler identity ∑ᵢ xᵢ ∂ᵢ f ≡ a f (mod 𝔪^(a+1)) for f ∈ 𝔪^a, and on a
being a unit in characteristic zero; (4) in local form is a computation on generators.
Dʳ (Dˢ I) = Dʳ⁺ˢ I [Kol07, Lemma 74(1)].
The product rule Dʳ (I J) ≤ ∑_{i ≤ r} Dⁱ I · Dʳ⁻ⁱ J [Kol07, Lemma 74(2)].
D (J^k) ≤ D J · J^(k-1) for k ≥ 1 (the product rule as used in the proof of
[Kol07, Proposition 99]).
The product rule for a nonempty finite product of ideals.
The product rule for a nonempty finite product of powers with positive exponents. (A sum
over the k with cₖ ≥ 1 would be false when every exponent vanishes, since D R = R; the
restriction to the support of the exponent vector is left to the caller.)
The easy inclusion of [Kol07, Lemma 74(3)]: m ≤ ord I implies m - r ≤ ord (Dʳ I) for
r ≤ m, from Dʳ(𝔪^m) ≤ 𝔪^(m-r).
The Euler identity #
The Euler operator ∑ᵢ xᵢ ∂ᵢ.
Instances For
E h ≡ h (mod 𝔪²) for h ∈ 𝔪 (Euler's identity in degree one).
The Euler identity: E f ≡ a f (mod 𝔪^(a+1)) for f ∈ 𝔪^a, a ≥ 1.
An element of 𝔪^a ∖ 𝔪^(a+1), a ≥ 1, has a partial derivative outside 𝔪^a (the
characteristic-zero fact behind [Kol07, Lemma 74(3)], "char. 0 only!", and
[Wlo05, Lemma 2.6.2]). If all ∂ᵢ f ∈ 𝔪^a then E f ∈ 𝔪^(a+1), so a f ∈ 𝔪^(a+1) by the
Euler identity, and a is a unit.
D 0 = 0.
For r < m, m ≤ ord I ↔ m - r ≤ ord (Dʳ I): the local form of
cosupp(I, m) = cosupp(Dʳ(I), m − r) [Kol07, Lemma 74(3)], [Wlo05, Lemma 2.6.2],
[BM08, Lemma 3.2(3)].
If ord I ≤ m then Dᵐ I = R [Kol07, Definition 73] (from le_ord_iff_le_ord_Dpow).
The derivative of a power of a coordinate: D ⟨xᵢ^m⟩ = ⟨xᵢ^(m-1)⟩ for m ≥ 1 (m is a unit
in characteristic zero).
Dʳ ⟨xᵢ^m⟩ = ⟨xᵢ^(m-r)⟩ for r ≤ m.
The local content of "D(f^* I) = f^*(D(I)) for smooth f" [Kol07, Lemma 74(4)],
[Wlo05, Lemma 2.6.5], [BM08, Lemma 3.2(4)]: for φ : R → R' compatible with the derivations
(∂'_{e i} ∘ φ = φ ∘ ∂ᵢ, and ∂'ⱼ ∘ φ = 0 off the image of e), D' (φ(I) R') = φ(D I) R'.
If every coordinate derivation of R' is induced along φ by one of R
(∂'ⱼ ∘ φ = φ ∘ ∂_{e j}), then D' (φ I) ≤ φ (D I): the local content of Kollár's
(DI)|_S ⊃ D(I|_S) for the restriction to a smooth subvariety (the discussion preceding
[Kol07, Definition 83]). For the restriction to R̄ = R/⟨x₁⟩ with coordinates x₂, …, xₙ the
extra generators ∂₁ f only enlarge the right side.
D'ʳ (φ I) ≤ φ (Dʳ I) under the hypothesis of D_map_le.