Bourbaki

Remark

Tag 0668, permanent. Cite it as [Bourbaki, Tag 0668].

Algebra, chapter III, § 10. Derivations

(1) The notion of derivation can be defined for non-graded modules by agreeing to give these modules the trivial graduation.

(2) If only ε\varepsilon-derivations of given degree δ\delta are considered, the commutation factor ε\varepsilon may be disposed of as follows: the bilinear mapping λ2:A×B′→B′′\lambda_2 : A \times B' \to B'' is modified by replacing it by the bilinear mapping λ2′:A×B′→B′′\lambda'_2 : A \times B' \to B'' such that, for every homogeneous aa in AA and all b′∈B′b' \in B',

λ2′(a,b′)=εδ,deg⁡(a)λ2(a,b′). \lambda'_2(a, b') = \varepsilon_{\delta, \deg(a)} \lambda_2(a, b').
Then dd is a derivation relative to the bilinear mappings μ,λ1,λ2′\mu, \lambda_1, \lambda'_2.

The general definition of ε\varepsilon-derivations given above is especially used in two cases:

Case (I): A=B,A′=B′,A′′=B′′A = B, A' = B', A'' = B'' and the three bilinear mappings μ,λ1,λ2\mu, \lambda_1, \lambda_2 are equal to the same mapping.

Case (II): A=A′=A′′,B=B′=B′′A = A' = A'', B = B' = B'', so that (for μ\mu) AA is a graded algebra and the two K-bilinear mappings.

λ1:B×A→B,λ2:A×B→B \lambda_1 : B \times A \to B, \quad \lambda_2 : A \times B \to B
are such that the corresponding K-linear mappings B⊗KA→B,A⊗KB→BB \otimes_K A \to B, A \otimes_K B \to B are graded of degree 0. An ε\varepsilon-derivation of degree δ\delta of AA into BB is then a graded K-linear mapping d:A→Bd : A \to B of degree δ\delta, such that for every homogeneous xx in AA and every y∈Ay \in A, we have the relation
d(xy)=(dx)y+εδ,deg⁡(a)x(dy). d(xy) = (dx)y + \varepsilon_{\delta, \deg(a)} x(dy).

Consider in particular in case (II) the case where A is a unital associative K-algebra and λ1\lambda_1 and λ2\lambda_2 are the external laws of an (A, A)-bimodule (§ 4, no. 3, Definition 3). This holds notably when A and B are two unital associative K-algebras, a unital homomorphism of graded K-algebras ρ:A→B\rho : A \to B is given and an (A, A)-bimodule structure is considered on B defined by the two external laws

λ1:(b,a)↦bρ(a),λ2:(a,b)↦ρ(a)b \lambda_1 : (b, a) \mapsto b \rho(a), \qquad \lambda_2 : (a, b) \mapsto \rho(a)b

for a∈A,b∈Ba \in A, b \in B.

Cases (I) and (II) have the following case in common: consider a graded K-algebra A, take B=AB = A, mappings (7) both being multiplication on A. We then speak of an ε\varepsilon-derivation (or (K, ε\varepsilon)-derivation) of the graded algebra A: it is a graded K-linear mapping of A into itself, of degree δ\delta, satisfying (8) for every homogeneous xx in A and all y∈Ay \in A. In particular if A is a graded ring, considered as an (associative) Z\mathbf{Z}-algebra, we speak of the ε\varepsilon-derivation of the ring A.

Let A be a unital commutative associative K-algebra and B an A-module; when we speak of a derivation of A into B, it will always be understood that we mean with the A-bimodule structure on B derived from its A-module structure; then the formula

d(xy)=x(dy)+y(dx)forx∈A,y∈A d(xy) = x(dy) + y(dx) \quad \text{for} \quad x \in A, y \in A

holds for such a derivation d:A→Bd : A \to B.

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Nothing.

Cites

Transcribed from the printed edition, 1998, Springer. Not translated.