a) The A-module I is a Matlis module. For every integer m⩾0, Im is the a-submodule of I consisting of the elements annihilated by mAm.
b) For every A-module M of finite length, the a-linear mapping θM:T(M)→HomA(M,I) is bijective.
c) For every a-linear mapping f:M→N between A-modules of finite length, one has θM∘T(f)=HomA(f,1I)∘θN.
Let us prove c). Let n be an integer and M, N A-modules of finite length annihilated by mAn. Let f:M→N be an a-linear mapping. For u in T(N) and x in M, one has by FD 3)
Let us prove b). First consider the particular case M=A/mAn. Then T(M) is equal by definition to In. If a is an element of A, of class aˉ in M, one has φM,aˉn=a1M, whence T(φM,aˉn)=a1In; thus θM:In→HomA(A/mAn,I) is the canonical isomorphism which sends an element x of In to the mapping aˉ↦ax. This proves b) in this case.
Now suppose given an exact sequence
PuNvM→0
of A-modules of finite length annihilated by mAn. Consider the diagram
it is commutative by the beginning of the proof, and its rows are exact by FD 4). We deduce that θM is bijective if θP and θN are so. Applying this to a presentation
(A/mAn)r⟶(A/mAn)s⟶M⟶0
of the A/mAn-module M, we deduce that θM is bijective for every A-module of finite length annihilated by mAn, whence b).
Let us prove a). It follows from the foregoing applied to the A-module A/mAn that In is the set of elements of I which are annihilated by mAn. By FD 5) the A-module I1=T(κA) is isomorphic to κA; by Prop. 2 of No. 2, it is enough for us to prove that, for every integer n⩾0, the canonical A-linear mapping
β:In+1/In⟶HomA(mAn/mAn+1,I)
is bijective. Now from the exact sequence
0⟶mAn/mAn+1uA/mAn+1pn+1,nA/mAn⟶0,
one deduces an exact sequence
0⟶Inin+1,nIn+1T(u)T(mAn/mAn+1)⟶0.
By composing T(u) with θmAn/mAn+1, one thus obtains a surjective homomorphism
γ:In+1⟶HomA(mAn/mAn+1,I)
with kernel In. By c), γ is the composite of the arrows θIn+1:In+1→HomA(A/mAn+1,I) and Hom(u,1):HomA(A/mAn+1,I)→HomA(mAn/mAn+1,I); since θIn+1 is the linear mapping associated with the multiplication A/mAn+1×In+1→I, the isomorphism In+1/In→HomA(mAn/mAn+1,I) deduced from γ coincides with β, which completes the proof.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.