Iterated derivatives of a transform along a divided coordinate #
In the chart of the blow-up of P = (x₀, …, x_r) dividing by x_r
(Hironaka/Algebra/Local/TransformDeriv.lean), [Kol07, (75.1)] says that for a divided coordinate
x_h, h < r, differentiation with respect to x_h commutes with the transform:
∂'_h π⁻¹_*(f, m+1) = π⁻¹_*(∂_h f, m). Iterating j times gives, for f ∈ Pᵐ and j ≤ m,
∂'_h^j π⁻¹_*(f, m) = π⁻¹_*(∂_h^j f, m − j) (iterate_chartDerivRing_transformElem), and hence,
for I ≤ Pᵐ and f ∈ I, the j-fold ∂'_h-derivative of the generator π⁻¹_*(f, m) of
π⁻¹_*(I, m) lies in π⁻¹_*(D^j I, m − j), because ∂_h^j f ∈ D^j I
(iterate_chartDerivRing_transformElem_mem).
This is the observation of the proof of [Kol07, Theorem 88] that, "by (75.1)", differentiation
commutes with the birational transform for ∂/∂x₁ and ∂/∂y₁. It is stated here for the
generators π⁻¹_*(f, m), f ∈ I, of the transform and not for all its elements: the ideal
generated by the j-fold ∂'_h-derivatives of all elements of π⁻¹_*(I, m) need not lie in
π⁻¹_*(D^j I, m − j) (HironakaExamples/Theorem88Counterexample.lean). The statement on
generators suffices for the inclusion ⊆ of [Kol07, (88.1)], proved in
Hironaka/Algebra/Local/TransformDerivNormalForm.lean.
Transformed elements of equal elements with equal marks are equal (the membership proofs are irrelevant).
[Kol07, (75.1)] iterated (the proof of [Kol07, Theorem 88]): in the chart dividing by x_r,
for h < r, f ∈ Pᵐ and j ≤ m, ∂'_h^j π⁻¹_*(f, m) = π⁻¹_*(∂_h^j f, m − j).
For I ≤ Pᵐ, f ∈ I and j ≤ m, the j-fold ∂'_h-derivative of the generator
π⁻¹_*(f, m) lies in the transform of D^j I:
∂'_h^j π⁻¹_*(f, m) = π⁻¹_*(∂_h^j f, m − j) ∈ π⁻¹_*(D^j I, m − j).