(3) Applying Proposition 16 to the case n=1, M1=M, P=A, then H=M∗, the dual of M, and it is seen that for a K-derivation (d0,d,d′) of (A,M,M) is derived a K-derivation (d0,d∗,d∗) of (A,M∗,M∗) such that
(38) d0⟨m,m∗⟩=⟨dm,m∗⟩+⟨m,d∗m∗⟩
for m∈M and m∗∈M∗. The K-linear mapping of M⊕M∗ into itself which is equal to d on M and to d∗ on M∗ then satisfies condition (29) and there is therefore a K-derivation D of the A-algebra T(M⊕M∗), which reduces to d0 on A, to d on M and to d∗ on M∗. The restriction dJI of D to the sub-A-module TJI(M) of T(M⊕M∗) (§5, no. 6) is a K-endomorphism of TJI(M) such that (d0,dJI,dJI) is a K-derivation of (A,TJI(M),TJI(M)). Moreover, for i∈I,j∈J, if we write I′=I−{i},J′=J−{j}, it is immediately verified that for the contraction cji (§5, no. 6)
cji(dJI(z))=dJ′I′(cji(z))for all z∈TJI(M).
(4) Let Mi (1≤i≤3) be three A-modules and, for each i, suppose that (d0,di,di) is a derivation of (A,Mi,Mi); applying Proposition 16 again for n=1, a derivation (d0,dij,dij) of (A,Hij,Hij), where Hij=HomA(Mi,Mj), is derived for each ordered pair (i,j). With this notation, for u∈HomA(M1,M2) and v∈HomA(M2,M3),
d13(v∘u)=(d23v)∘u+v∘(d12u) as is immediately verified from the definitions.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.