Bourbaki

Exercise 23, § 3

Commutative Algebra, VII

Bourbaki marks this one as harder than the rest.

Let BB be a Krull ring, EE its field of fractions, Δ\Delta a derivation of EE such that Δ(B)⊂B\Delta(B) \subset B, KK the subfield of EE which is the kernel of Δ\Delta and AA the Krull ring B∩KB \cap K; suppose that KK has characteristic p>0p > 0; one has Ep⊂KE^p \subset K and Bp⊂AB^p \subset A, so that BB is the integral closure of AA in EE, and the canonical homomorphism i:C(A)→C(B)i : C(A) \to C(B) is defined (§ 1, No. 10). We denote by UU the group of invertible elements of BB.

a) If b∈Eb \in E is such that div⁡B(b)\operatorname{div}_B(b) is the canonical image of a divisor of D(A)D(A), now prove that one has Δb/b∈B\Delta b/b \in B (note that for every prime ideal P\mathfrak{P} of height 1 of BB, there exists b′∈Kb' \in K such that vP(b)=vP(b′)v_{\mathfrak{P}}(b) = v_{\mathfrak{P}}(b'), and observe that BPB_{\mathfrak{P}} is stable under Δ\Delta). Let LL be the additive subgroup of BB formed by the Δb/b\Delta b/b (“logarithmic derivatives”) which belong to BB (for b∈Eb \in E or b∈Bb \in B, which amounts to the same thing, and b≠0b \neq 0), and let L′L' be the subgroup of LL formed by the Δu/u\Delta u/u, where u∈Uu \in U. Now prove that, for every divisor d∈D(A)d \in D(A) such that the image of dd is a principal divisor in D(B)D(B), the class mod. L′L' of Δb/b\Delta b/b for every bb such that i(d)=div⁡B(b)i(d) = \operatorname{div}_B(b) depends only on the class of dd in C(A)C(A), and deduce a canonical injective homomorphism φ\varphi of Ker⁡(i)\operatorname{Ker}(i) into L/L′L/L'.

(b) Prove that if Δ(B)\Delta(B) is not contained in any prime ideal of height 1 of BB and if [E:K]=p[E : K] = p, φ\varphi is bijective. (Reduce to proving that if b∈Eb \in E is such that Δb/b∈B\Delta b/b \in B and if P\mathfrak{P} is a prime ideal of height 1 in BB such that vP(b)v_{\mathfrak{P}}(b) is not multiple of pp, then e(P/p)=1e(\mathfrak{P}/p) = 1, where p=P∩Ap = \mathfrak{P} \cap A; for this, deduce from the assumption that if tt is a uniformizer of BPB_{\mathfrak{P}}, one has Δt/t∈BP\Delta t/t \in B_{\mathfrak{P}}, so that PBP\mathfrak{P}B_{\mathfrak{P}} is stable under Δ\Delta, and that Δ\Delta therefore defines by passing to quotients a derivation Δˉ\bar{\Delta} of the residue field k=BP/PBPk = B_{\mathfrak{P}}/\mathfrak{P}B_{\mathfrak{P}}; prove that Δˉ≠0\bar{\Delta} \neq 0, and deduce that f(P/p)=pf(\mathfrak{P}/p) = p.

(c) Let kk be a field of characteristic 2, BB the polynomial ring k[X,Y,Z]k[X, Y, Z], Δ\Delta the derivation of E=k(X,Y,Z)E = k(X, Y, Z) such that Δ(X)=Y4\Delta(X) = Y^4, Δ(Y)=X2\Delta(Y) = X^2, Δ(Z)=XYZ\Delta(Z) = XYZ; the field KK kernel of Δ\Delta is such that [E:K]=4[E : K] = 4. One has Δ(Z)/Z∈B\Delta(Z)/Z \in B, but prove that div⁡B(Z)\operatorname{div}_B(Z) is not the canonical image of a divisor of D(A)D(A). (Reason by contradiction by assuming that e(P/p)=1e(\mathfrak{P}/p) = 1 for P=BZ\mathfrak{P} = BZ, p=P∩Ap = \mathfrak{P} \cap A; there would then be in AA a uniformizer of BPB_{\mathfrak{P}}, necessarily of the form a(X,Y,Z)Za(X, Y, Z)Z with b=a(X,Y,0)≠0b = a(X, Y, 0) \neq 0; deduce that one would have Δb/b=−XY\Delta b/b = -XY and obtain a contradiction by calculating Δ(−XY)\Delta(-XY)).

Solution

No solution has been written yet.

Machine translation of the English, by gpt-5-6, gpt-5-mini. Not checked by a person. That model is a cut down version of the one the rest of this language was written by.