The tuned ideal I + D(I)² at m = 2: the general facts #
[Kol07, Proposition 99] for the maximal coefficient ideal W_s(I) at m = 2,
s = m! = 2 = 1·lcm(2): W_2(I) = I + D(I)² (W_two_eq of
Hironaka.Resolution.Algebraic.Tuning.Assemble, in the sheaf vocabulary) is MC-invariant (99.5) and
D-balanced (99.8). Here these facts are proved for the intrinsic derivative ideal Ideal.derivative k of Hironaka.Derivative on any k-algebra, with no hypothesis on I, the form the
polynomial-ring examples HironakaExamples.Balanced.Example11 and Example106 (Kollár's Examples
11 and 106) instantiate:
D(I + D(I)²) = D(I)—⊇by monotonicity;⊆sinceD(D(I)²) ⊆ D(D(I))·D(I) + D(I)·D(D(I)) ⊆ D(I)by the Leibniz rule (derivative_mul_le) (derivative_sup_derivative_sq);- hence
D(I + D(I)²)² = D(I)² ⊆ I + D(I)²:I + D(I)²is D-balanced with respect to2([Kol07, Definition 83]; the clausei = 0is trivial) (isDBalanced_sup_derivative_sq) and MC-invariant with respect to2,MC(W)·D(W) = D(W)·D(W) ⊆ W([Kol07, Definition 79]:MC(W) = D^{2−1}(W) = D(W)) (mcInvariant_sup_derivative_sq).
The same facts in coordinates on a regular local ring are RegularCoords.isMCInvariant_W and
RegularCoords.isDBalanced_W of Hironaka.Local; nothing here redefines W.
D(I + D(I)²) = D(I) (the computation behind [Kol07, Proposition 99] at m = 2).
[Kol07, Proposition 99 (8)] at m = 2, s = 2, for the intrinsic derivative ideal: I + D(I)²
is D-balanced with respect to 2, since D(I + D(I)²)² = D(I)² ⊆ I + D(I)².
[Kol07, Proposition 99 (5)] at m = 2, for the intrinsic derivative ideal: I + D(I)² is
MC-invariant with respect to 2, MC(W)·D(W) = D(W)·D(W) ⊆ W (MC(W) = D^{2−1}(W) = D(W)).