Let E be a graded bigebra (resp. skew graded bigebra). For every element x of degree 1 in E, the left and right inner products by x are derivations (resp. antiderivations) (§10, no. 2) of the algebra E∗gr.
In the notation of no. 6, for every homogeneous element x of degree 1 in a graded bigebra (resp. a skew graded bigebra) E,
c(x)=x⊗1+1⊗x, by Proposition 5 of no. 3 and the fact that
c is a homomorphism of degree 0. Suppose first that
E is a graded bigebra. For all
y∈E, by definition
⟨(uv)&x,y⟩=⟨uv,xy⟩=m((u⊗v)(c(xy))) and since
c is an
algebra homomorphism,
c(xy)=c(x)c(y). Let
c(y)=∑isi⊗ti with
si and
ti in
E; therefore
c(xy)=i∑(xsi)⊗ti+i∑si⊗(xti). Hence
⟨(uv)&x,y⟩=∑iu(xsi)v(ti)+∑iu(si)v(xti). But we may write
i∑u(xsi)v(ti)=m(((u&x)⊗v)(c(y)))=⟨(u&x)v,y⟩ and similarly
i∑u(si)v(xti)=m((u⊗(v&x))(c(y)))=⟨u(v&x),y⟩, whence, returning to the notation
i(x) for the inner product.
(i(x))(uv)=((i(x))(u))v+u((i(x))(v)) which proves that
i(x) is a
derivation in
E∗gr.
Suppose now that E is a skew graded bigebra, that u∈Ep∗, v∈Eq∗ and y∈Er; then we can write
c(y)=0≤j≤r∑(i∑sij⊗ti,r−j) where the
sij and
tij belong to
Ej; by definition of the product in
Eg⊗AE, then
c(xy)=c(x)c(y)=0≤j≤r∑(i∑(xsij)⊗ti,r−j+(−1)ji∑sij⊗(xti,r−j)) whence this time
⟨(uv)&x,y⟩=0≤j≤r∑(i∑u(xsij)v(ti,r−j)+(−1)ji∑u(sij)v(xti,r−j)).
Then also ∑0≤j≤r(∑iu(xsij)v(ti,r−j))=⟨(u⊔x)v,y⟩. On the other hand, u(sij)=0 unless j=−p and hence we may also write
0≤j≤r∑(−1)j(i∑u(sij)v(xti,r−j))=(−1)p⟨u(v⊔x),y⟩. We therefore conclude that
(i(x))(uv)=((i(x))(u))v+(−1)pu((i(x))(v)), in other words
i(x) is an
antiderivation in
E∗gr. The assertions relating to the left inner product by an element
x of degree 1 in
E are proved similarly.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.