Bourbaki

§ 3. Graded algebras

The graduations considered in this paragraph will have as set of degrees a commutative monoid written additively whose identity element is denoted by 0.

1. Graded algebras

Definition 1

Let Δ\Delta be a commutative monoid, AA a graded ring of type Δ\Delta* (II, § 11, no. 2), (Aλ)λ∈Δ(A_\lambda)_{\lambda \in \Delta} its graduation and E an A-algebra. A graduation (Eλ)λ∈Δ(E_\lambda)_{\lambda \in \Delta} *of type Δ\Delta on the additive group E is said to be compatible with the A-algebra structure on E if it is compatible both with the A-module and with the ring structure on E, in other words, if, for all λ,μ\lambda, \mu in Δ\Delta,

(1)

AλEμ⊂Eλ+μ A_\lambda E_\mu \subset E_{\lambda + \mu}
(2)
EλEμ⊂Eλ+μ. E_\lambda E_\mu \subset E_{\lambda + \mu}.

The A-algebra E, with this graduation, is then called a graded algebra of type Δ\Delta over the graded ring A.

When the graduation on A is trivial (that is (II, § 11, no. 1) A0=A,Aλ={0}A_0 = A, A_\lambda = \{0\} for λ≠0\lambda \neq 0), condition (1) means that the EλE_\lambda are sub-A-modules of E. This leads to the definition of the notion of graded algebra of type Δ\Delta over a non-graded commutative ring AA: AA is given the trivial graduation of type Δ\Delta and the above definition is applied.

When we consider graded AA-algebras EE with a unit element ee, it will always be understood that ee is of degree 0 (cf. Exercise 1).

It follows that if an invertible element x∈Ex \in E is homogeneous and of degree pp, its inverse x−1x^{-1} is homogeneous and of degree −p-p: it suffices to decompose x−1x^{-1} as a sum of homogeneous elements in the relations x−1x=xx−1=ex^{-1}x = xx^{-1} = e.

Let EE and E′E' be two graded algebras of type Δ\Delta over a graded ring AA of type Δ\Delta. An AA-algebra homomorphism u:E→E′u : E \to E' is called a graded algebra homomorphism if u(Eλ)⊂Eλ′u(E_\lambda) \subset E'_\lambda for all λ∈Δ\lambda \in \Delta (where (Eλ)(E_\lambda) and (Eλ′)(E'_\lambda) denote the respective graduations of EE and E′E'); where EE and E′E' are associative and unital and uu is unital, this condition means that uu is a graded ring homomorphism (II, § 11, no. 2).

Let EE be a graded AA-algebra of type NN. EE is identified with a graded AA-algebra of type Z\mathbf{Z} by writing En={0}E_n = \{0\} for n<0n < 0.

Remark

Definition 1 can also be interpreted by saying that EE is a graded AA-module and that the AA-linear mapping

m:E⊗AE→E m : E \otimes_A E \to E

defining the multiplication on EE (§ 1, no. 3), is homogeneous of degree 0 when E⊗AEE \otimes_A E is given its graduation of type Δ\Delta (II, § 11, no. 5).

To define a graded AA-algebra structure of type Δ\Delta on the graded ring AA, with EE as underlying graded AA-module, therefore amounts to defining for each ordered pair (λ,μ)(\lambda, \mu) of elements of Δ\Delta a Z\mathbf{Z}-bilinear mapping

mλμ:Eλ×Eμ→Eλ+μ m_{\lambda \mu} : E_\lambda \times E_\mu \to E_{\lambda + \mu}

such that for every triple of indices (λ,μ,ν)(\lambda, \mu, \nu) and for α∈Aλ, x∈Eμ, y∈Eν\alpha \in A_\lambda,\ x \in E_\mu,\ y \in E_\nu,

α.mμν(x,y)=mλ+μ,ν(αx,y)=mμ,λ+ν(x,αy). \alpha . m_{\mu \nu}(x, y) = m_{\lambda + \mu, \nu}(\alpha x, y) = m_{\mu, \lambda + \nu}(x, \alpha y).

Example

(1) Let BB be a graded ring of type Δ\Delta; if BB is given its canonical Z\mathbf{Z}-algebra structure (§ 1, no. 1, Example 2), BB is a graded AA-algebra (Z\mathbf{Z} being given the trivial graduation).

(2) Let AA be a graded commutative ring of type Δ\Delta and MM a graded AA-module of type Δ\Delta. Suppose that all the elements of the monoid Δ\Delta are cancellable, which allows (II, § 11, no. 6) us to define on HomgrA(M,M)=EndgrA(M)\mathrm{Homgr}_A(M, M) = \mathrm{Endgr}_A(M) a graded AA-module structure of type Δ\Delta; as this graduation is compatible with the ring structure on EndgrA(M)\mathrm{Endgr}_A(M) (II, § 11, no. 6), it defines a unital graded AA-algebra structure on the AA-algebra EndgrA(M)\mathrm{Endgr}_A(M).

(3) Algebra of a magma. Let SS be a magma and ϕ:S→Δ\phi : S \to \Delta a homomorphism. For all λ∈Δ\lambda \in \Delta, we write Sλ=ϕ−1(λ)S_\lambda = \phi^{-1}(\lambda); then SλSμ⊂Sλ+μS_\lambda S_\mu \subset S_{\lambda + \mu}. Let AA be a graded commutative ring of type Δ\Delta and (Aλ)λ∈Δ(A_\lambda)_{\lambda \in \Delta} its graduation; we shall define a graded AA-algebra structure on the algebra E=A(S)E = A^{(S)} of the magma SS (§ 2, no. 6). To this end, let EλE_{\lambda} denote the additive subgroup of EE generated by the elements of the form α.s\alpha.s such that α∈Aμ,s∈Sν\alpha \in A_{\mu}, s \in S_{\nu} and μ+ν=λ\mu + \nu = \lambda. As the SλS_{\lambda} are pairwise disjoint, EE is the direct sum of the AμSνA_{\mu}S_{\nu} and hence also the direct sum of the EλE_{\lambda} and it is immediate that the EλE_{\lambda} satisfy conditions (1) and (2) and therefore define on EE the desired graded AA-algebra structure. If SS admits an identity element ee, it may also be supposed that ϕ(e)=0\phi(e) = 0. A particular case is the one where the graduation of the ring AA is trivial; then EλE_{\lambda} is the sub-AA-module of EE generated by SλS_{\lambda}. More particularly, if we take S=N(I),Δ=NS = \mathbf{N}^{(I)}, \Delta = \mathbf{N} and ϕ\phi the mapping such that ϕ((ni))=∑i∈Ini\phi((n_i)) = \sum_{i \in I} n_i, the ring AA having the trivial graduation, a graduation is thus obtained on the polynomial algebra A[Xi]i∈IA[X_i]_{i \in I}, for which the degree of a homogeneous polynomial ≠0\neq 0 is the total degree defined in § 2, no. 9 (cf. § 6, no. 6).

We now take SS to be the free monoid Mo(B)Mo(B) of a set BB (I, § 7, no. 2) and ϕ\phi the homomorphism Mo(B)→NMo(B) \to \mathbf{N} which associates with each word its length. Thus a graded AA-algebra structure is obtained on the free associative algebra of the set BB (§ 2, no. 7; cf. § 5, no. 5).

2. Graded subalgebras, graded ideals of a graded algebra

Let EE be a graded algebra of type Δ\Delta over a graded ring AA of type Δ\Delta. If FF is a sub-AA-algebra of EE which is a graded sub-AA-module, then the graduation (Fλ)(F_{\lambda}) on FF is compatible with its AA-algebra structure, since Fλ=F∩EλF_{\lambda} = F \cap E_{\lambda}; in this case FF is called a graded subalgebra of EE and the canonical injection F→EF \to E is a graded algebra homomorphism.

Similarly, if aa is a left (resp. right) ideal of EE which is a graded sub-AA-module, then Eλaμ⊂aλ+μE_{\lambda}a_{\mu} \subset a_{\lambda+\mu} (resp. aλEμ⊂aλ+μa_{\lambda}E_{\mu} \subset a_{\lambda+\mu}), since aλ=a∩Eλa_{\lambda} = a \cap E_{\lambda}; then aa is called a graded ideal of the algebra EE. If bb is a graded two-sided ideal of EE the quotient graduation on the module E/bE/b is compatible with the algebra structure on E/bE/b and the canonical homomorphism E→E/bE \to E/b is a graded algebra homomorphism.

If u:E→E′u : E \to E' is a graded algebra homomorphism, Im⁡(u)\operatorname{Im}(u) is a graded subalgebra of E′E', Ker⁡(u)\operatorname{Ker}(u) is a graded two-sided ideal of EE and the bijection E/Ker⁡(u)→Im⁡(u)E/\operatorname{Ker}(u) \to \operatorname{Im}(u) canonically associated with uu is a graded algebra isomorphism.

Proposition 1

Let AA be a graded commutative ring of type Δ\Delta, EE a graded AA-algebra of type Δ\Delta and SS a set of homogeneous elements of EE. Then the sub-AA-algebra (resp. left ideal, right ideal, two-sided ideal) generated by SS is a graded subalgebra (resp. graded ideal).*.

The subalgebra of EE generated by SS is the sub-AA-module generated by the finite products of elements of SS, which are homogeneous; similarly, the left (resp. right) ideal generated by SS is the sub-AA-module generated by the elements of the form u1(u2(…(uns))…)u_1(u_2(\ldots(u_n s))\ldots) (resp. (…((sun)un−1)…)u2(\ldots((s u_n) u_{n-1})\ldots) u_2 ) u1u_1) where s ∈ S and the u_j ∈ E are homogeneous (n arbitrary) and these products are homogeneous, whence in this case the conclusion by virtue of II, § 11, no. 3, Proposition 2); finally the two-sided ideal generated by S is the union of the sequence (\mathfrak{I}_n)_{n \geq 1}, where \mathfrak{I}_1 is the left ideal generated by S and \mathfrak{I}_{2n} (resp. \mathfrak{I}_{2n+1}) the right (resp. left) ideal generated by \mathfrak{I}_{2n-1} (resp. \mathfrak{I}_{2n}), which completes the proof.

3. Direct limits of graded algebras

Let (A_\alpha, \phi_{\beta \alpha}) be a directed direct system of graded commutative rings of type \Delta (II, § 11, no. 3, Remark 3) and for each \alpha let E_\alpha be a graded A_\alpha-algebra of type \Delta; for \alpha \leq \beta let f_{\beta \alpha}: E_\alpha \to E_\beta be an A_\alpha-homomorphism of graded algebras and suppose that f_{\gamma \alpha} = f_{\gamma \beta} \circ f_{\beta \alpha} for \alpha \leq \beta \leq \gamma; then we shall call (E_\alpha, f_{\beta \alpha}) a directed direct system of graded algebras of type \Delta over the directed direct system (A_\alpha, \phi_{\beta \alpha}) of graded commutative rings of type \Delta. Then we know (II, § 11, no. 3) that E = \lim \to E_\alpha has canonically a graded module structure of type \Delta over the graded ring A = \lim \to A_\alpha and a multiplication such that E^\lambda E^\mu \subset E^{\lambda + \mu} (where (E^\lambda) denotes the graduation on E); then this multiplication and the graded A-module structure on E define on E a graded A-algebra structure of type \Delta; E, with this structure, is called the direct limit of the direct system (E_\alpha, f_{\beta \alpha}) of graded algebras. The canonical homomorphisms E_\alpha \to E are taken A_\alpha-homomorphisms of graded algebras. Moreover, if F is a graded A-algebra of type \Delta and (u_\alpha) a direct system of A_\alpha-homomorphisms u_\alpha: E_\alpha \to F, u = \lim \to u_\alpha is an A-homomorphism of graded algebras.

Exercises

See the exercises for § 3.

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