Maximal coefficient ideals #
[Kol07, Definition 98]: for an ideal I with m = max-ord I, the maximal coefficient ideal of
order s is W_s(I) = ∑ { ∏ⱼ (Dʲ I)^{cⱼ} : ∑ⱼ (m − j) cⱼ ≥ s }, the sum over all exponent
vectors c : {0, …, m} → ℕ of weight at least s. Here R is a regular local ring containing ℚ
with coordinates (RegularCoords), Dʲ its iterated derivative ideals
(Hironaka/Algebra/Local/Derivative.lean), and the parameter m is explicit: c.W m s I, with the
weight wt m c. The module proves the local forms of [Kol07, Proposition 99] and of its Claim 99.9:
W_0(I) = R;Iᵏ ≤ W_s(I)form k ≥ s(pow_le_W);MC_m(I)ˢ ≤ W_s(I)(MC_pow_le_W);W_1(I) = MC_m(I)form ≥ 1(W_one);W_s(I) = Rwhenord I < m(W_eq_top_of_ord_lt);- (99.1)
W_{s+1}(I) ≤ W_s(I)and (99.2)W_s(I) · W_t(I) ≤ W_{s+t}(I)(W_succ_le,W_mul_le); - (99.3)
D(W_{s+1}(I)) = W_s(I)forord I ≤ m,m ≥ 1(D_W_succ: the inclusion≤by the product rule; the reverse by Kollár's argument with an elementx₁ ∈ MC_m(I)of order1and a derivation∂with∂ x₁ = 1, showingx₁^{s−t} W_t(I) ≤ D(W_{s+1}(I))by induction ont), and its iterateDⁱ(W_s(I)) = W_{s−i}(I)(Dpow_W); - the order:
ord W_s(I) = swhenord I = m ≥ 1(ord_W; fors = m!this is [Kol07, Theorem 54.2 (i)]), ands ≤ ord W_s(I) ↔ m ≤ ord Ifors ≥ 1(le_ord_W_iff: the cosupports of(W_s(I), s)and(I, m)agree); - (99.4)
MC_s(W_s(I)) = W_1(I) = MC_m(I)(MC_W) and (99.5)W_s(I)is MC-invariant (isMCInvariant_W); - Claim 99.9 in the weight language (
exists_le_wt_eq: every exponent vector of weight≥ (r + m − 1) L_m, withL_m = lcm(2, …, m), has a sub-vector of weight exactlyr L_m) and its consequences (99.6)W_s(I) · W_t(I) = W_{s+t}(I)fors = r L_mandt ≥ (m − 1) L_m(W_mul_W_eq), (99.7)W_s(I)ʲ = W_{js}(I)fors = r L_m,r ≥ m − 1(W_pow), and (99.8)W_s(I)is D-balanced for suchs(isDBalanced_W).
Conventions: indices are zero-based; MC_m(I) = D^{m−1}(I) is the maximal-contact ideal of
Hironaka/Algebra/Local/MaximalContact.lean; D-balanced and MC-invariant are IsDBalanced
(Hironaka/Algebra/Local/Balanced.lean) and IsMCInvariant
(Hironaka/Algebra/Local/MaximalContact.lean).
Used for the maximal coefficient ideal sheaf and Theorem 100
(Hironaka/Resolution/Algebraic/Tuning/Sheaf.lean,
Hironaka/Resolution/Algebraic/Tuning/Cosupport.lean,
Hironaka/Resolution/Algebraic/Tuning/MCTuned.lean,
Hironaka/Resolution/Algebraic/Tuning/Corollary101.lean,
Hironaka/Resolution/Algebraic/Tuning/Parameter.lean), for the tuning parameter
(Hironaka/Algebra/Local/TuningParam.lean) and on manifolds
(Hironaka/Manifold/IdealSheaf/TuningLemmas.lean).
The weight ∑ⱼ (m − j) cⱼ of an exponent vector c : {0, …, m} → ℕ
([Kol07, Definition 98]).
Instances For
Claim 99.9 #
L_m = lcm(2, …, m), with L_1 = 1 ([Kol07, Claim 99.9]).
Equations
- IsLocalRing.Lcm m = (Finset.Icc 2 m).lcm id
Instances For
[Kol07, Claim 99.9] in the weight language of Definition 98 (Kollár's variable uᵢ of degree
i is D^{m−i} I, dᵢ = c_{m−i}, deg d = wt_m c; the index m, of weight 0, is ignored):
every exponent vector e with wt e ≥ (r + m − 1) L_m has a sub-vector e' ≤ e with
wt e' = r L_m. Proof as Kollár's: write eⱼ = bⱼ (L_m/wⱼ) + fⱼ with wⱼ fⱼ < L_m
(wⱼ = m − j); if ∑ bⱼ ≥ r pick d ≤ b with ∑ d = r and e'ⱼ = dⱼ L_m/wⱼ; otherwise
wt e < (r − 1) L_m + m L_m, a contradiction.
The maximal coefficient ideal of order s ([Kol07, Definition 98]),
W_s(I) = ∑_{wt c ≥ s} ∏ⱼ (Dʲ I)^{cⱼ}, with the parameter m (Kollár's max-ord I) explicit.
Equations
Instances For
W_0(I) = R.
MC_m(I)ˢ ≤ W_s(I): the exponent vector c_{m−1} = s has weight s (Kollár's
x₁^{s+1} ∈ W_{s+1}(I) in the proof of [Kol07, Proposition 99]).
W_1(I) = D^{m−1} I = MC_m(I) for m ≥ 1 (the proof of (99.4) in
[Kol07, Proposition 99]).
If ord I < m then W_s(I) = R for every s, since MC_m(I) = R ([Kol07, Lemma 74 (3)]
locally).
Proposition 99 (1)–(3) #
[Kol07, (99.1)]: W_{s+1}(I) ≤ W_s(I).
[Kol07, (99.2)]: W_s(I) · W_t(I) ≤ W_{s+t}(I).
The product rule for one generator: D((Dʲ I)^k) ≤ W_{(m-j)k - 1}(I).
The product rule for a product of generators (Kollár's "follows from the product rule" in the
proof of [Kol07, Proposition 99]): D(∏ (Dˡ I)^{eₗ}) ≤ W_{wt(e) − 1}(I).
The inclusion ⊆ of [Kol07, (99.3)], by the product rule: D(W_{s+1}(I)) ≤ W_s(I).
An element of 𝔪 ∖ 𝔪² has a coordinate partial that is a unit (the contrapositive of
mem_maximalIdeal_sq_of_pderiv_mem, Hironaka/Algebra/Local/Coords.lean): Kollár's ∂/∂x₁ in the
proof of [Kol07, Proposition 99] is u⁻¹ ∂ᵢ for such an i, with ∂ᵢ x₁ = u.
Kollár's "Pick x₁ ∈ MC(I) that has order 1" (the proof of [Kol07, Proposition 99]): for
ord I = m ≥ 1 there are x₁ ∈ MC_m(I) of order 1 and a derivation ∂ with ∂ x₁ = 1 mapping
every ideal J into D J, namely ∂ = u⁻¹ ∂ᵢ for an i with ∂ᵢ x₁ = u a unit. No hypothesis
that the ∂ᵢ span the derivations is needed.
(The proof of [Kol07, Proposition 99].) x₁ˢ = (s+1)⁻¹ ∂(x₁^{s+1}) ∈ D(W_{s+1}(I)).
(The proof of [Kol07, Proposition 99].) x₁^{s−t} W_t(I) ≤ D(W_{s+1}(I)) for t ≤ s, by
induction on t.
[Kol07, (99.3)]: D(W_{s+1}(I)) = W_s(I) for ord I ≤ m, m ≥ 1.
(99.3) iterated, Kollár's "applying (3) repeatedly": Dⁱ(W_s(I)) = W_{s−i}(I) for i ≤ s.
The order of W_s(I) #
Every generator of W_s(I) lies in 𝔪^{wt(e)} when I ≤ 𝔪ᵐ, since Dʲ I ≤ 𝔪^{m−j} (the
estimate in the proof of [Kol07, Theorem 100]).
If m ≤ ord I then s ≤ ord (W_s(I)) (the estimate in the proof of
[Kol07, Theorem 100]).
If ord I = m ≥ 1 then ord (W_s(I)) ≤ s, since Dˢ(W_s(I)) = W_0(I) = R ((99.3)
iterated) and an ideal whose s-th derivative ideal is the unit ideal has order ≤ s
([Kol07, Lemma 74 (3)] locally).
If ord I = m ≥ 1 then ord (W_s(I)) = s; for s = m! this is [Kol07, Theorem 54.2 (i)],
locally.
For s ≥ 1, cosupp (W_s(I), s) = cosupp (I, m), locally: s ≤ ord (W_s(I)) ↔ m ≤ ord I;
no hypothesis ord I ≤ m is needed.
Proposition 99 (4)–(5) #
[Kol07, (99.4)]: MC_s(W_s(I)) = W_1(I) = MC_m(I) for s ≥ 1, ord I ≤ m, m ≥ 1 (the
mark of W_s(I) is s, its order by ord_W).
[Kol07, (99.5)] (Kollár: "Together with (2) and (3), this implies (5)"): W_s(I) is
MC-invariant with respect to s: MC_s(W_s) · D(W_s) = W_1 · W_{s−1} ≤ W_s.
Proposition 99 (6)–(8) #
[Kol07, (99.6)] (Kollár: "(6) is implied by (99.9)"): W_s(I) · W_t(I) = W_{s+t}(I) for
s = r L_m and t ≥ (m − 1) L_m.
[Kol07, (99.7)]: W_s(I)ʲ = W_{js}(I) for s = r L_m, r ≥ m − 1.
(The proof of (99.8) in [Kol07, Proposition 99].) For s = r L_m, r ≥ m − 1, ord I ≤ m
and i < s: (Dⁱ W_s(I))ˢ = W_{s−i}(I)ˢ ≤ W_{s(s−i)}(I) = W_s(I)^{s−i}.
[Kol07, (99.8)]: W_s(I) is D-balanced with respect to s for s = r L_m, r ≥ m − 1,
ord I ≤ m (the mark s is the order of W_s(I) by ord_W when s ≥ 1, that is, r ≥ 1).