Let M1,M, be two finitely generated A-modules, and f:M1→M2 an injective homomorphism. Then the kernel of f⊗1;M1⊗AB→M2⊗AB is pseudo-zero.
Let p be a prime ideal of A of height ≤1. Then (Mi⊗AB)p=(Mi)p⊗ApBp (i=1,2) (Chapter 11, § 2, no. 7, Proposition 18) and (f⊗1B)p=fp⊗1Bp; the hypothesis that f is injective implies that so is f, (Chapter II, § 2, no. 4, Theorem 1); on the other hand, in view of the choice of p, Ap is a principal ideal domain and Bp a finitely generated torsion-free Ap-module and hence free; we conclude that fp⊗1Bp is itself injective. If I=Ker(f⊗1), then Ip=Ker((f⊗1)p) (Chapter 11, § 2, no. 4, Theorem 1); therefore Ip=0, whence a fortiori Ip=(Ip)p=0 for p∣p, which proves the lemma (no. 4, Definition 2).
We return now to the proof of (ii). For every finitely generated torsion A-module M, write ϕ(M)=χB(M⊗AB); it follows from (i) that, if M is pseudo-zero, ϕ(M)=0. On the other hand, consider an exact sequence of finitely generated torsion A-modules :
0→M1→M2→M3→0. It follows from Lemma 4 that there is an exact sequence of
B-modules:
0→I→M1⊗AB→M2⊗AB→M3⊗AB→0 where
I is pseudo-zero. Using no. 5, Corollary to Proposition 10, we therefore obtain
ϕ(M2)=ϕ(M1)+ϕ(M3). We therefore conclude from Proposition 11 of no. 5 that there exists a homomorphism
θ:D(A)→D(B) such that
ϕ(M)=θ(χA(M)) for every finitely generated torsion
A-module
M. To prove that
θ=i, it suffices to show that
ϕ(A/p)=i(p) for all
p∈P(A); now
(A/p)⊗AB=B/pB and, for all
p∈P(B),
(B/pB)p=Bp/pBp; the last module is 0 if
p does not lie above
p; if on the contrary
p∣p,
Bp/pBp is a
Bp-module of length
e(p/p) by definition of the ramification index (Chapter VI, § 8, no. 1); therefore
χB(B/pB)=∑p∣pep/p.13=i(p), which completes the proof of (ii).
Formula (24) is immediate, for
(M⊗AB)⊗BL=M⊗AL=(M⊗AK)⊗KL
and the rank of (M⊗AK)⊗KL over L is equal to the rank of M⊗AK over K. To show (23), consider a free submodule H of M such that Q=M/H is a torsion A-module. Applying Lemma 4 as above, we obtain an exact sequence of B-modules:
0→I→H⊗AB→M⊗AB→Q⊗AB→0
where I is pseudo-zero. It therefore follows from no. 7, Proposition 16(ii) and (v) and Corollary 1 to Proposition 16 that
cB(M⊗AB)=cB(Q⊗AB)=−cB(χB(Q⊗AB))=−cB(i(χA(Q)))
by virtue of (ii); but by definition of the homomorphism i:C(A)→C(B), cB(i(χA(Q)))=i(cA(χA(Q)))=−i(cA(M)), which completes the proof of (23).
Remarks (1) If M is a reflexive A-module, M⊗AB is not necessarily reflexive (exercise 6). However it is so when B is a flat A-module (no. 2, Proposition 8). (2) Let C be a third integrally closed Noetherian domain such that B⊂C and C is a finitely generated B-module (and hence also a finitely generated A-module). Then we have the transitivity equations:
(25)
NC/A=NB/A∘NC/B, (26)
iC/A=iC/B∘iB/A
which follow immediately from the transitivity equations for the ramification indices and the residue class degrees (Chapter VI, § 8, no. 1, Lemma 1).
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.
Nothing.