Bourbaki

§ 3. Tensor products

1. Tensor product of two modules

Let G1,G2G_1, G_2 be two Z\mathbf{Z}-modules; a mapping uu of the set G=G1×G2G = G_1 \times G_2 into a Z\mathbf{Z}-module is called biadditive (or Z\mathbf{Z}-bilinear) if u(x1,x2)u(x_1, x_2) is "additive with respect to x1x_1 and with respect to x2x_2"; to be precise, this means that, for x1,y1x_1, y_1 in G1G_1, x2,y2x_2, y_2 in G2G_2,

u(x1+y1,x2)=u(x1,x2)+u(y1,x2) u(x_1 + y_1, x_2) = u(x_1, x_2) + u(y_1, x_2)
u(x1,x2+y2)=u(x1,x2)+u(x1,y2). u(x_1, x_2 + y_2) = u(x_1, x_2) + u(x_1, y_2).

Note that this implies in particular that u(0,x2)=u(x1,0)=0u(0, x_2) = u(x_1, 0) = 0 for all x1∈G1,x2∈G2x_1 \in G_1, x_2 \in G_2.

Let A be a ring, E a right A-module and F a left A-module. We are going to consider the universal mapping problem (Set Theory, IV, § 3, no. 1) where Σ\Sigma is the species of Z\mathbf{Z}-module structure (the morphisms then being Z\mathbf{Z}-linear mappings, in other words, additive group homomorphisms) and the α\alpha-mappings are the mappings ff of E×FE \times F into a Z\mathbf{Z}-module G which are Z\mathbf{Z}-bilinear and further satisfy, for all x∈E,y∈Fx \in E, y \in F and λ∈A\lambda \in A

(1)(1)
f(xλ,y)=f(x,λy). f(x \lambda, y) = f(x, \lambda y).

We show that this problem admits a solution. For this we consider the Z\mathbf{Z}-module C=Z(E×F)C = \mathbf{Z}^{(E \times F)} of formal linear combinations of the elements of E×FE \times F with coefficients in Z\mathbf{Z} (§ 1, no. 11), a basis of which can be considered to consist of the ordered pairs (x,y)(x, y), where x∈Ex \in E and y∈Fy \in F. Let D be the sub-Z\mathbf{Z}-module of C generated by the elements of one of the following types:

(2)(2)
{(x1+x2,y)−(x1,y)−(x2,y)(x,y1+y2)−(x,y1)−(x,y2)(xλ,y)−(x,λy) \begin{cases} (x_1 + x_2, y) - (x_1, y) - (x_2, y) \\ (x, y_1 + y_2) - (x, y_1) - (x, y_2) \\ (x \lambda, y) - (x, \lambda y) \end{cases}

where x,x1,x2x, x_1, x_2 are in EE, y,y1,y2y, y_1, y_2 are in FF and λ\lambda is in AA.

Definition 1

The tensor product of the right A-module E and the left A-module F*, denoted by E⊗AFE \otimes_A F or E⊗AFE \otimes_A F (or simply E⊗FE \otimes F if no confusion is to be feared) is the quotient Z\mathbf{Z}-module C/DC/D (the quotient of the Z\mathbf{Z}-module C of formal linear combinations of elements of E×FE \times F with coefficients in Z\mathbf{Z}, by the submodule DD generated by the elements of one of the types (2)). *For x∈Ex \in E and y∈Fy \in F, the element of E⊗AFE \otimes_A F which is the canonical image of the element (x,y)(x, y) of C=Z(E×F)C = \mathbf{Z}^{(E \times F)} is denoted by x⊗yx \otimes y and called the tensor product of xx and yy.

Cited by 3

Cited by § 3. Tensor products, Remark, Proposition 1

The mapping (x,y)↦x⊗y(x, y) \mapsto x \otimes y of E×FE \times F into E⊗AFE \otimes_A F is called canonical. It is a Z\mathbf{Z}-bilinear mapping which satisfies conditions (1).

We show that the tensor product E⊗AFE \otimes_A F and the above canonical mapping form a solution of the universal mapping problem posed earlier. To be precise:

Proposition 1

(a) *Let gg be a Z\mathbf{Z}-linear mapping of E⊗AFE \otimes_A F into a Z\mathbf{Z}-module GG. The mapping (x,y)↦f(x,y)=g(x⊗y)(x, y) \mapsto f(x, y) = g(x \otimes y) of E×FE \times F into GG is Z\mathbf{Z}-bilinear and satisfies conditions (1).

Cites 3 · Cited by 8

Cites II § 1. Modules, Proposition 17, II § 1, Definition 1

Cited by § 3. Tensor products, Corollary, Proposition 2, Proposition 3, Proposition 8, II § 4. Relations between tensor products and homomorphism modules, Proposition 7, II § 6, Proposition 16, II § 7

(b) Conversely, let ff be a Z\mathbf{Z}-bilinear mapping of E×FE \times F into a Z\mathbf{Z}-module GG satisfying conditions (1). Then there exists one and only one Z\mathbf{Z}-linear mapping gg of E⊗AFE \otimes_A F into GG such that f(x,y)=g(x⊗y)f(x, y) = g(x \otimes y) for x∈E,y∈Fx \in E, y \in F.

If ϕ\phi denotes the canonical mapping of E×FE \times F into E⊗AFE \otimes_A F, then f=g∘ϕf = g \circ \phi; whence (a). To show (b), we note that, in the notation of Definition 1, ff extends to a Z\mathbf{Z}-linear mapping fˉ\bar{f} of CC into GG (§ 1, no. 11, Proposition 17). By virtue of relations (1), fˉ\bar{f} is zero for all the elements of CC of one of the types (2) and hence on DD. There therefore exists a Z\mathbf{Z}-linear mapping gg of C/D=E⊗AFC/D = E \otimes_A F into GG such that fˉ=g∘ψ\bar{f} = g \circ \psi, where ψ:C→C/D\psi : C \to C/D is the canonical homomorphism (§ 1, no. 8, Remark). The uniqueness of gg is immediate since E⊗AFE \otimes_A F is generated, as a Z\mathbf{Z}-module, by the elements of the form x⊗yx \otimes y.

Proposition 1 defines a canonical isomorphism of the Z\mathbf{Z}-module of Z\mathbf{Z}-bilinear mappings ff of E×FE \times F into GG, satisfying conditions (1), onto the Z\mathbf{Z}-module HomZ(E⊗AF,G)\mathrm{Hom}_{\mathbf{Z}}(E \otimes_A F, G).

When A=ZA = \mathbf{Z}, conditions (1) are automatically satisfied for every Z\mathbf{Z}-bilinear mapping ff and the submodule DD of CC is already generated by the elements of the first two types in (2).

If now we return to the general case and E′E' and F′F' denote the underlying Z\mathbf{Z}-modules of EE and FF respectively, the above remark and Definition 1 show immediately that the Z\mathbf{Z}-module E⊗AFE \otimes_A F can be canonically identified with the quotient of the Z\mathbf{Z}-module E′⊗ZF′E' \otimes_{\mathbf{Z}} F' by the sub-Z\mathbf{Z}-module generated by the elements of the form (xλ)⊗y−x⊗(λy)(x \lambda) \otimes y - x \otimes (\lambda y), where xx runs through EE, yy runs through FF and λ\lambda runs through AA.

Corollary 1

Let HH be a Z\mathbf{Z}-module and h:E×F→Hh : E \times F \to H a Z\mathbf{Z}-bilinear mapping satisfying conditions (1) and such that HH is generated by h(E×F)h(E \times F). Suppose that for every Z\mathbf{Z}-module GG and every Z\mathbf{Z}-bilinear mapping ff of E×FE \times F into GG satisfying (1)* there exists a Z\mathbf{Z}-linear mapping g:H→Gg : H \to G such that f=g∘hf = g \circ h. Then, if ϕ\phi denotes the canonical mapping of E×FE \times F into E⊗AFE \otimes_A F, there exists one and only one isomorphism θ\theta of E⊗AFE \otimes_A F onto HH such that h=θ∘ϕh = \theta \circ \phi.

Cites 1 · Cited by 1

Cites E IV § 3. Universal mappings

Cited by § 3. Tensor products

The hypothesis that h(E×F)h(E \times F) generates HH implies the uniqueness of gg; the corollary is then just the general uniqueness property of a solution of a universal mapping problem (Set Theory, IV, § 3, no. 1).

Corollary 2

Let* E0E^0 (resp. F0F^0) denote the module EE (resp. FF) considered as a left (resp. right) module over the opposite ring A0A^0; then there exists one and only one Z\mathbf{Z}*-module isomorphism* σ:E⊗AF→F0⊗A0E0\sigma : E \otimes_A F \to F^0 \otimes_{A^0} E^0 such that σ(x⊗y)=y⊗x\sigma(x \otimes y) = y \otimes x for x∈Ex \in E and y∈Fy \in F ("commutativity" of tensor products).

Cited by 3

Cited by § 3. Tensor products, Corollary, III § 1. Algebras

By definition of the A0A^0-module structures on E0E^0 and F0F^0, the mapping (x,y)↦y⊗x(x, y) \mapsto y \otimes x of E×FE \times F into F0⊗A0E0F^0 \otimes_{A^0} E^0 is Z\mathbf{Z}-bilinear and satisfies conditions (1), whence the existence and uniqueness of the Z\mathbf{Z}-linear mapping σ\sigma. Similarly a Z\mathbf{Z}-linear mapping τ:F0⊗A0E0→E⊗AF\tau : F^0 \otimes_{A^0} E^0 \to E \otimes_A F is defined such that τ(y⊗x)=x⊗y\tau(y \otimes x) = x \otimes y and clearly σ\sigma and τ\tau are inverse isomorphisms.

Remark

The tensor product of non-zero modules may be zero: for example, taking the two Z\mathbf{Z}-modules E=Z/2ZE = \mathbf{Z}/2\mathbf{Z} and F=Z/3ZF = \mathbf{Z}/3\mathbf{Z}, 2x=02x = 0 and 3y=03y = 0 for all x∈Ex \in E and y∈Fy \in F; therefore, in E⊗ZFE \otimes_{\mathbf{Z}} F,

x⊗y=3(x⊗y)−2(x⊗y)=x⊗(3y)−(2x)⊗y=0 x \otimes y = 3(x \otimes y) - 2(x \otimes y) = x \otimes (3y) - (2x) \otimes y = 0
for all xx and yy (cf. no. 6, Corollary 4 to Proposition 6).

Cites 2

Cites Proposition 6, Corollary 4

2. Tensor product of two linear mappings

Let AA be a ring, E,E′E, E' two right AA-modules, F,F′F, F' two left AA-modules and u:E→E′u : E \to E' and v:F→F′v : F \to F' two AA-linear mappings. It is easily verified that the mapping

(x,y)↦u(x)⊗v(y) (x, y) \mapsto u(x) \otimes v(y)
of E×FE \times F into E′⊕AF′E' \oplus_A F' is Z\mathbf{Z}-bilinear and satisfies conditions (1) of no. 1. By Proposition 1 of no. 1 there thus exists one and only one Z\mathbf{Z}-linear mapping w:E⊗AF→E′⊗AF′w : E \otimes_A F \to E' \otimes_A F' such that
w(x⊗y)=u(x)⊗v(y) w(x \otimes y) = u(x) \otimes v(y)
for x∈E,y∈Fx \in E, y \in F. This mapping is denoted by u⊗vu \otimes v (when no confusion can arise) and is called the tensor product of the linear mappings uu and vv.

It follows immediately from (3) that (u,v)↦u⊗v(u, v) \mapsto u \otimes v is a Z-bilinear\mathbf{Z}\text{-bilinear} mapping called canonical

Hom⁡A(E,E′)×Hom⁡A(F,F′)→Hom⁡Z(E⊗AF,E′⊗AF′). \operatorname{Hom}_A(E, E') \times \operatorname{Hom}_A(F, F') \to \operatorname{Hom}_{\mathbf{Z}}(E \otimes_A F, E' \otimes_A F').

There corresponds to it by Proposition 1 of no. 1 a Z\mathbf{Z}-linear mapping called canonical

(4)

Hom⁡A(E,E′)⊗ZHom⁡A(F,F′)→Hom⁡Z(E⊗AF,E′⊗AF′) \operatorname{Hom}_A(E, E') \otimes_{\mathbf{Z}} \operatorname{Hom}_A(F, F') \to \operatorname{Hom}_{\mathbf{Z}}(E \otimes_A F, E' \otimes_A F')
which associates with every element u⊗vu \otimes v of the tensor product the linear mapping u⊗v:E⊗AF→E′⊗AF′u \otimes v : E \otimes_A F \to E' \otimes_A F'. Note that the canonical mapping (4) is not necessarily injective nor surjective. The notation u⊗vu \otimes v can therefore lead to confusion and it will be necessary for the context to indicate whether it denotes an element of the tensor product or a linear mapping.

Further, let E′′E'' be a right AA-module, F′′F'' a left AA-module and u′:E′→E′′u' : E' \to E'', v′:F′→F′′v' : F' \to F'' AA-linear mappings; it follows from (3) that (5)

(u′∘u)⊗(v′∘v)=(u′⊗v′)∘(u⊗v). (u' \circ u) \otimes (v' \circ v) = (u' \otimes v') \circ (u \otimes v).

3. Change of ring

Proposition 2

Let A,BA, B be two rings, ρ:B→A\rho : B \to A a ring homomorphism and EE (resp. FF) a right (resp. left) AA-module. Then there exists one and only one Z\mathbf{Z}-linear mapping* (6)

ϕ:ρ∗(E)⊗Bρ∗(F)→E⊗AF \phi : \rho_*(E) \otimes_B \rho_*(F) \to E \otimes_A F
*such that, for all x∈Ex \in E and y∈Fy \in F, the image under ϕ\phi of the element x⊗yx \otimes y of ρ∗(E)⊗Bρ∗(F)\rho_*(E) \otimes_B \rho_*(F) is the element x⊗yx \otimes y of E⊗AFE \otimes_A F; this Z\mathbf{Z}-linear mapping is surjective.

Cites 1 · Cited by 1

Cites Proposition 1

Cited by Corollary

We consider the mapping (x,y)↦x⊗y(x, y) \mapsto x \otimes y of ρ∗(E)×ρ∗(F)\rho_*(E) \times \rho_*(F) into E⊗AFE \otimes_A F; it is Z\mathbf{Z}-bilinear and, for all β∈B\beta \in B, by definition (xρ(β))⊗y=x⊗(ρ(β)y)(x \rho(\beta)) \otimes y = x \otimes (\rho(\beta)y), hence conditions (1) of no. 1 hold, whence the existence and uniqueness of ϕ\phi (no. 1, Proposition 1). The latter assertion follows from the fact that the elements x⊗yx \otimes y generate the Z\mathbf{Z}-module E⊗AFE \otimes_A F.

The mapping (6) is called canonical.

Corollary

Let J\mathfrak{J} be a two-sided ideal of AA such that J\mathfrak{J} is contained in the annihilator of EE and in the annihilator of FF, so that EE (resp. FF) has a canonical left (resp. right) (A/J)(A/\mathfrak{J})-module structure* (§1§ 1, no. 12). Then the canonical homomorphism (6)

ϕ:E⊗AF→E⊗A/JF \phi : E \otimes_A F \to E \otimes_{A/\mathfrak{J}} F
*corresponding to the canonical homomorphism ρ:A→A/J\rho : A \to A/\mathfrak{J} is the identity.

Cites 1 · Cited by 1

Cites Proposition 2

Cited by Corollary 3

For all αˉ∈A/J\bar{\alpha} \in A/\mathfrak{J}, all x∈Ex \in E and all y∈Fy \in F, xαˉ=xαx \bar{\alpha} = x \alpha (resp. αˉy=αy\bar{\alpha} y = \alpha y) for all α\alpha such that ρ(α)=αˉ\rho(\alpha) = \bar{\alpha}. If C=Z(E×F)C = \mathbf{Z}^{(E \times F)}, the submodule of CC generated by the elements (xα,y)−(x,αy)(x \alpha, y) - (x, \alpha y) is then equal to the submodule generated by the elements (xαˉ,y)−(x,αˉy)(x \bar{\alpha}, y) - (x, \bar{\alpha} y).

With the hypotheses and notation of Proposition 2, let E′E' be a right BB-module, F′F' a left BB-module and consider two semi-linear mappings u:E′→Eu : E' \to E, v:F' \to F relative to the homomorphism ρ:B→A\rho : B \to A; uu (resp. vv) can be considered as a BB-linear mapping E′→ρ∗(E)E' \to \rho_*(E) (resp. F′→ρ∗(F′)F' \to \rho_*(F')), whence a Z\mathbf{Z}-linear mapping w:E′⊗BF′→ρ∗(E)⊗Bρ∗(F)w : E' \otimes_B F' \to \rho_*(E) \otimes_B \rho_*(F) such that w(x′⊗y′)=u(x′)⊗v(y′)w(x' \otimes y') = u(x') \otimes v(y') for x′∈E′, y′∈F′x' \in E',\ y' \in F'; composing the canonical mapping (6) with this mapping, a Z\mathbf{Z}-linear mapping w′:E′⊗BF′→E⊗AFw' : E' \otimes_B F' \to E \otimes_A F is hence obtained such that w′(x′⊗y′)=u(x′)⊗v(y′)w'(x' \otimes y') = u(x') \otimes v(y') for x′∈E′, y′∈F′x' \in E',\ y' \in F'; this is the mapping which will normally be denoted by u⊗vu \otimes v if no confusion can arise. Clearly (u,v)↦u⊗v(u, v) \mapsto u \otimes v is a Z\mathbf{Z}-bilinear mapping

Hom⁡B(E′,ρ∗(E))×Hom⁡B(F′,ρ∗(F))→Hom⁡Z(E′⊗BF′,E⊗AF). \operatorname{Hom}_B(E', \rho_*(E)) \times \operatorname{Hom}_B(F', \rho_*(F)) \to \operatorname{Hom}_\mathbf{Z}(E' \otimes_B F', E \otimes_A F).

Moreover, if CC is a third ring, σ:C→B\sigma : C \to B a homomorphism, E′′E'' a right CC-module, F′′F'' a left CC-module, u′:E′′→E′u' : E'' \to E' and v′:F′′→F′v' : F'' \to F' semi-linear mappings relative to σ\sigma, then

(u∘u′)⊗(v∘v′)=(u⊗v)∘(u′⊗v′). (u \circ u') \otimes (v \circ v') = (u \otimes v) \circ (u' \otimes v').

4. Operators on a tensor product; tensor products as multimodules

With the hypotheses and notation of no. 1, for every endomorphism uu (resp. vv) of the AA-module EE (resp. FF), u⊗1Fu \otimes 1_F (resp. 1E⊗v1_E \otimes v) is an endomorphism of the Z\mathbf{Z}-module E⊗AFE \otimes_A F; it follows immediately from (5) (no. 2) that the mapping u↦u⊗1Fu \mapsto u \otimes 1_F (resp. v↦1E⊗vv \mapsto 1_E \otimes v) is a ring homomorphism

End⁡A(E)→End⁡Z(E⊗AF) \operatorname{End}_A(E) \to \operatorname{End}_\mathbf{Z}(E \otimes_A F)

(resp. End⁡A(F)→End⁡Z(E⊗AF)\operatorname{End}_A(F) \to \operatorname{End}_\mathbf{Z}(E \otimes_A F)); moreover,

(u⊗1F)∘(1E⊗v)=(1E⊗v)∘(u⊗1F)=u⊗v (u \otimes 1_F) \circ (1_E \otimes v) = (1_E \otimes v) \circ (u \otimes 1_F) = u \otimes v

and therefore (§ 1, no. 14) E⊗AFE \otimes_A F has a canonical left bimodule structure with respect to the rings End⁡A(E)\operatorname{End}_A(E) and End⁡A(F)\operatorname{End}_A(F).

This being so, suppose given on EE a ((Bi′);A,(Cj′))((B'_i); A, (C'_j))-multimodule structure and on FF a (A,(Bh′′);(Ckc′′))(A, (B''_h); (C''_{k_c}))-multimodule structure (§ 1, no. 14); it amounts to the same to say that ring homomorphisms Bi′→End⁡A(E)B'_i \to \operatorname{End}_A(E), Cj′0→End⁡A(E)C'_j{}^0 \to \operatorname{End}_A(E) are given with pairwise permutable images, and ring homomorphisms Bh′′→End⁡A(F)B''_h \to \operatorname{End}_A(F), Ckc′′0→End⁡A(F)C''_{k_c}{}^0 \to \operatorname{End}_A(F) with pairwise permutable images. If these homomorphisms are composed respectively with the canonical homomorphisms End⁡A(E)→End⁡Z(E⊗AF)\operatorname{End}_A(E) \to \operatorname{End}_\mathbf{Z}(E \otimes_A F) and End⁡A(F)→End⁡Z(E⊗AF)\operatorname{End}_A(F) \to \operatorname{End}_\mathbf{Z}(E \otimes_A F) defined above, it is seen (taking account of (7)) that ring homomorphisms

Bi′→End⁡Z(E⊗AF),Cj′0→End⁡Z(E⊗AF)Bh′′→End⁡Z(E⊗AF),Ckc′′0→End⁡Z(E⊗AF) \begin{align*} B'_i &\to \operatorname{End}_\mathbf{Z}(E \otimes_A F), & C'_j{}^0 &\to \operatorname{End}_\mathbf{Z}(E \otimes_A F) \\ B''_h &\to \operatorname{End}_\mathbf{Z}(E \otimes_A F), & C''_{k_c}{}^0 &\to \operatorname{End}_\mathbf{Z}(E \otimes_A F) \end{align*}

are defined with pairwise permutable images; in other words, there has been defined on E⊗AFE \otimes_A F a ((Bi′),(Bh′′);(Cj′),(Ckc′′))((B'_i), (B''_h); (C'_j), (C''_{k_c}))-multimodule structure; it is this multimodule which is also called the tensor product (relative to AA) of the

((B'_i); A, (C'_j))-multimodule E and the (A, (B''_h); (C''_k))-multimodule F. This multimodule is the solution of a universal mapping problem analogous to that considered in no. 1; to be precise:

Proposition 3

Let G be a ((B'_i), (B''_h); (C'_j), (C''_k))-multimodule.

Cites 1 · Cited by 2

Cites Proposition 1

Cited by § 3. Tensor products, Scholium

(a) Let g be a linear mapping of the multimodule E ⊗A\otimes_A F into G. The mapping f:(x,y)↦g(x⊗y)f : (x, y) \mapsto g(x \otimes y) of E ×\times F into G is Z\mathbf{Z}-bilinear and satisfies relations (1) of no. 1 and the conditions

{f(μi′x,y)=μi′f(x,y),f(xvj′,y)=f(x,y)vj′f(x,μh′′y)=μh′′f(x,y),f(x,yvk′′)=f(x,y)vk′′ \begin{cases} f(\mu'_i x, y) = \mu'_i f(x, y), & f(x v'_j, y) = f(x, y) v'_j \\ f(x, \mu''_h y) = \mu''_h f(x, y), & f(x, y v''_k) = f(x, y) v''_k \end{cases}

for all x∈E,y∈F,μi′∈Bi′,vj′∈Cj′,μh′′∈Bh′′,vk′′∈Ck′′,i,j,h,kx \in E, y \in F, \mu'_i \in B'_i, v'_j \in C'_j, \mu''_h \in B''_h, v''_k \in C''_k, i, j, h, k arbitrary.

(b) Conversely, let f be a Z\mathbf{Z}-bilinear mapping of E ×\times F into G satisfying conditions (1) (no. 1) and (8). Then there exists one and only one linear mapping g of the multimodule E ⊗A\otimes_A F into the multimodule G such that f(x,y)=g(x⊗y)f(x, y) = g(x \otimes y) for x∈E,y∈Fx \in E, y \in F.

Assertion (a) follows immediately from the definition of the multimodule structure on E ⊗A\otimes_A F, since for example (x⊗y)vj′=(xvj′)⊗y(x \otimes y)v'_j = (xv'_j) \otimes y. To prove (b), we note first that Proposition 1 of no. 1 gives the existence and uniqueness of a Z\mathbf{Z}-linear mapping g such that g(x⊗y)=f(x,y)g(x \otimes y) = f(x, y) for x∈E,y∈Fx \in E, y \in F; all that is needed is to verify that g is linear for the multimodule structures. As the elements x⊗yx \otimes y generate the Z\mathbf{Z}-module E ⊗A\otimes_A F, it suffices to verify the relations g(μ′(x⊗y))=μi′g(x⊗y)g(\mu'(x \otimes y)) = \mu'_i g(x \otimes y) and their analogues; but this follows immediately from the formula g(x⊗y)=f(x,y)g(x \otimes y) = f(x, y) and relations (8).

Scholium

An element of E ⊗A\otimes_A F may in general be written in several ways in the form ∑i(xi⊗yi)\sum_i (x_i \otimes y_i), where xi∈Ex_i \in E and yi∈Fy_i \in F; but to define a linear mapping g of the multimodule E ⊗A\otimes_A F into a multimodule G, there is no need to verify that, if ∑i(xi⊗yi)=∑j(xj′⊗yj′)\sum_i (x_i \otimes y_i) = \sum_j (x'_j \otimes y'_j), then ∑ig(xi⊗yi)=∑jg(xj′⊗yj′)\sum_i g(x_i \otimes y_i) = \sum_j g(x'_j \otimes y'_j); it suffices to be given g(x⊗y)g(x \otimes y) for x∈Ex \in E and y∈Fy \in F and to verify that (x,y)↦g(x⊗y)(x, y) \mapsto g(x \otimes y) is Z\mathbf{Z}-bilinear and satisfies conditions (1) (no. 1) and (8).

Let E' be a ((B'_i), A, (C'_j))-multimodule, F' an (A, (B''_h); (C''_k))-multimodule and u:E→E′,v:F→F′u : E \to E', v : F \to F' linear mappings of multimodules; it follows immediately from the definitions (no. 2) that u⊗vu \otimes v is a linear mapping of the multimodule E ⊗A\otimes_A F into the multimodule E' ⊗A\otimes_A F'.

With E always denoting a right A-module, let sAd_sA_d denote the ring A considered as an (A, A)-bimodule (§ 1, no. 14, Example 1); by the above, the tensor product E ⊗A(sAd)\otimes_A (_sA_d) has a canonical right A-module structure such that (x⊗λ)μ=x⊗(λμ)(x \otimes \lambda)\mu = x \otimes (\lambda \mu) for x∈E,λ∈A,μ∈Ax \in E, \lambda \in A, \mu \in A. The mapping (x,λ)↦xλ(x, \lambda) \mapsto x \lambda of E ×(sAd)\times (_sA_d) into E is Z\mathbf{Z}-bilinear and satisfies conditions (1) (no. 1) and (8) (where, in the latter, the Bi′,Cj′B'_i, C'_j and Bh′′B''_h are absent and the family (Ck′′)(C''_k) reduces to AA); hence (Proposition 3), there exists an AA-linear mapping gg (called canonical) of E⊗A(sAd)E \otimes_A (sA_d) into EE such that g(x⊗λ)=xλg(x \otimes \lambda) = x\lambda for x∈E,λ∈Ax \in E, \lambda \in A.

Cites 2

Cites Example, II § 1, Proposition 3

Proposition 4

If EE is a right AA-module, the mapping h:x↦x⊗1h : x \mapsto x \otimes 1 of EE into E⊗A(sAd)E \otimes_A (sA_d) is a right AA-module isomorphism, whose inverse isomorphism gg is such that g(x⊗λ)=xλg(x \otimes \lambda) = x\lambda for x∈E,λ∈Ax \in E, \lambda \in A.

Cited by 16

Cited by AC I § 2. Flat modules, Proposition 7, II § 11, Remark, Corollary 1, Corollary 3, Corollary 6, Lemma 3, II § 4, Proposition 4, II § 4, Corollary, II § 5, Proposition 2, II § 5, Proposition 6, II § 5, Proposition 26, II § 7, III § 1. Algebras, Proposition 3, III § 11, Proposition 13, III § 7, Definition 1, VIII § 6

If gg is the canonical mapping, g∘hg \circ h is the identity mapping 1E1_E and h∘gh \circ g coincides with the identity mapping of E⊗A(sAd)E \otimes_A (sA_d) onto itself for elements of the form x⊗yx \otimes y, which generate the latter Z\mathbf{Z}-module; hence the conclusion.

We shall normally write E⊗AAE \otimes_A A instead of E⊗A(sAd)E \otimes_A (sA_d) and shall often identify E⊗AAE \otimes_A A with EE by means of the above canonical isomorphisms. Observe that, if EE also has a (left or right) BB-module structure which is compatible with its right AA-module structure, gg and hh are also isomorphisms for the BB-module structures on EE and E⊗AAE \otimes_A A (and hence multinode module isomorphisms).

Now let FF be a left AA-module; (sAd)⊗AF(sA_d) \otimes_A F (also denoted by A⊗AFA \otimes_A F) then has a canonical left AA-module structure and as in Proposition 4 a canonical isomorphism is defined of A⊗AFA \otimes_A F onto FF mapping λ⊗x\lambda \otimes x to λx\lambda x, and its inverse isomorphism is x↦1⊗xx \mapsto 1 \otimes x.

In particular there exists a canonical isomorphism of the (A,A)(A, A)-bimodule (sAd)⊗A(sAd)(sA_d) \otimes_A (sA_d) onto sAdsA_d which maps λ⊗μ\lambda \otimes \mu to λμ\lambda \mu.

5. Tensor product of two modules over a commutative ring

Let CC be a commutative ring; for every CC-module EE, the module structure on EE is compatible with itself (§ 1, no. 14). If EE and FF are two CC-modules, the considerations of no. 4 then allow us to define two CC-module structures on the tensor product E⊗CFE \otimes_C F, respectively such that γ(x⊗y)=(γx)⊗y\gamma(x \otimes y) = (\gamma x) \otimes y and such that γ(x⊗y)=x⊗(γy)\gamma(x \otimes y) = x \otimes (\gamma y); but as, by Definition 1 of no. 1, in this case (γx)⊗y=x⊗(γy)(\gamma x) \otimes y = x \otimes (\gamma y), these two structures are the same. Henceforth when we speak of E⊗CFE \otimes_C F as a CC-*module*, we shall mean with the structure thus defined, unless otherwise stated. The canonical isomorphism.

σ:F⊗CE→E⊗CF \sigma : F \otimes_C E \to E \otimes_C F

(no. 1, Corollary 2 to Proposition 1) is then a CC-module isomorphism.

It follows from this definition that, if (aλ)λ∈L(a_\lambda)_{\lambda \in L} (resp. (bμ)μ∈M(b_\mu)_{\mu \in M}) is a generating system of the CC-module EE (resp. FF), (aλ⊗bμ)(a_\lambda \otimes b_\mu) is a generating system of the CC-module E⊗CFE \otimes_C F; in particular, if EE and FF are finitely generated CC-modules, so is E⊗CFE \otimes_C F.

For every CC-module GG, the Z\mathbf{Z}-bilinear mappings ff of E×FE \times F into GG for which

f(γx,y)=f(x,γy)=γf(x,y)for x∈E,y∈F,γ∈C f(\gamma x, y) = f(x, \gamma y) = \gamma f(x, y) \quad \text{for } x \in E, y \in F, \gamma \in C

are then called C-bilinear and form a C-module denoted by L2(E,F;G)\mathscr{L}_2(E, F; G); Proposition 3 (no. 4) defines a canonical C-module isomorphism (cf. § 1, no. 14, Remark 1).

L2(E,F;G)→HomC(E⊗CF,G). \mathscr{L}_2(E, F; G) \to \mathrm{Hom}_C(E \otimes_C F, G).

Let E', F' be two C-modules and u:E→E′u : E \to E', v:F→F′v : F \to F' two C-linear mappings; then (no. 4) u⊗vu \otimes v is a C-linear mapping of E⊗CFE \otimes_C F into E′⊗CF′E' \otimes_C F'. Further, it is immediate that (u,v)↦u⊗v(u, v) \mapsto u \otimes v is a C-bilinear mapping of HomC(E,E′)×HomC(F,F′)\mathrm{Hom}_C(E, E') \times \mathrm{Hom}_C(F, F') into HomC(E⊗CF,E′⊗CF′)\mathrm{Hom}_C(E \otimes_C F, E' \otimes_C F'); hence there canonically corresponds to it a C-linear mapping, called canonical:

HomC(E,E′)⊗CHomC(F,F′)→HomC(E⊗CF,E′⊗CF′) \mathrm{Hom}_C(E, E') \otimes_C \mathrm{Hom}_C(F, F') \to \mathrm{Hom}_C(E \otimes_C F, E' \otimes_C F')

which associates with every element u⊗vu \otimes v of the tensor product

HomC(E,E′)⊗CHomC(F,F′) \mathrm{Hom}_C(E, E') \otimes_C \mathrm{Hom}_C(F, F')

the linear mapping u⊗vu \otimes v. Note that the canonical mapping (11) is not necessarily injective nor surjective (§ 4, Exercise 2).

Remark

(1) Let A, B be two commutative rings, ρ:B→A\rho : B \to A a ring homomorphism and E and F two A-modules; then the canonical mapping (6) of no. 3 is a B-linear mapping

ρ∗(E)⊗ρ∗(F)→ρ∗(E⊗AF). \rho_*(E) \otimes \rho_*(F) \to \rho_*(E \otimes_A F).

(2) What has been said in this no. can be generalized to the following case: let E be a right A-module, F a left A-module, C a commutative ring and ρ:C→A\rho : C \to A a homomorphism of C into A such that ρ(C)\rho(C) is contained in the centre of A (cf. III, § 1, no. 3). We may then consider the C-modules ρ∗(E)\rho_*(E) and ρ∗(F)\rho_*(F) and the hypothesis on ρ\rho implies that these C-module structures are compatible respectively with the A-module structures on E and F (§ 1, no. 14). The tensor product E⊗AFE \otimes_A F is thus (by virtue of no. 4) given two C-module structures such that γ(x⊗y)=(xρ(γ))⊗y\gamma(x \otimes y) = (x \rho(\gamma)) \otimes y and

γ(x⊗y)=x⊗(ρ(γ)y) \gamma(x \otimes y) = x \otimes (\rho(\gamma)y)

respectively for γ∈C\gamma \in C, x∈Ex \in E, y∈Fy \in F and Definition 1 (no. 1) shows also that these two structures are identical. If E' (resp. F') is a right (resp. left) A-module and u:E→E′u : E \to E', v:F→F′v : F \to F' are two A-linear mappings, then u⊗v:E⊗AF→E′⊗AF′u \otimes v : E \otimes_A F \to E' \otimes_A F' is C-linear for the C-module structures just defined; the mapping (u,v)↦u⊗v(u, v) \mapsto u \otimes v:

HomA(E,E′)×HomA(F,F′)→HomC(E⊗AF,E′⊗AF′) \mathrm{Hom}_A(E, E') \times \mathrm{Hom}_A(F, F') \to \mathrm{Hom}_C(E \otimes_A F, E' \otimes_A F')

is C-bilinear (for the C-module structures on HomA(E,E′)\mathrm{Hom}_A(E, E') and HomA(F,F′)\mathrm{Hom}_A(F, F')) defined in § 1, no. 14, Remark 1), whence we also derive a C-linear mapping, called *canonical

(13) Hom⁡A(E,E′)⊗CHom⁡A(F,F′)→Hom⁡C(E⊗AF,E′⊗AF′)\operatorname{Hom}_A(E, E') \otimes_C \operatorname{Hom}_A(F, F') \to \operatorname{Hom}_C(E \otimes_A F, E' \otimes_A F').

(3) Let A be an integral domain and K its field of fractions. If E and F are two vector K-spaces, the canonical mapping

(E[A])⊗A(F[A])→E⊗KF (E_{[A]}) \otimes_A (F_{[A]}) \to E \otimes_K F

(no. 3 and § 1, no. 13) is bijective. It suffices (no. 4) to prove that if ff is an A-bilinear mapping of E×FE \times F into a vector K-space G, ff is also K-bilinear. Now, for all α≠0\alpha \neq 0 in A.

αf(α−1x,y)=f(x,y)=αf(x,α−1y) \alpha f(\alpha^{-1}x, y) = f(x, y) = \alpha f(x, \alpha^{-1}y)
whence
f(α−1x,y)=f(x,α−1y)=α−1f(x,y) f(\alpha^{-1}x, y) = f(x, \alpha^{-1}y) = \alpha^{-1}f(x, y)
since G is a vector K-space.

Cites 3

Cites II § 1. Modules, III § 1. Algebras, Definition 1

6. Properties of E⊗AFE \otimes_A F relative to exact sequences

Proposition 5

Let* E, E', E'' be right A-*modules*, F a left A-*module* and

(14)

E′→uE→vE′′→0 E' \xrightarrow{u} E \xrightarrow{v} E'' \to 0

an exact sequence of linear mappings.* Writing uˉ=u⊗1F,vˉ=v⊗1F,\bar{u} = u \otimes 1_F, \bar{v} = v \otimes 1_F, *the sequence

Cited by 8

Cited by Proposition 7, II § 11, Corollary, Proposition 6, Exercise 6, § 4, Proposition 14, II § 7, VIII § 11. Grothendieck Groups, Proposition 9, VIII § 11, Proposition 11, VIII § 6

(15)

E′⊗AF→uˉE⊗AF→vˉE′′⊗AF→0 E' \otimes_A F \xrightarrow{\bar{u}} E \otimes_A F \xrightarrow{\bar{v}} E'' \otimes_A F \to 0

of Z\mathbf{Z}-*homomorphisms* is exact.

By virtue of no. 2, formula (5), vˉ∘uˉ=(v∘u)⊗1F=0\bar{v} \circ \bar{u} = (v \circ u) \otimes 1_F = 0; the image H=uˉ(E′⊗F)H = \bar{u}(E' \otimes F) is contained in the kernel L=Ker⁡(vˉ)L = \operatorname{Ker}(\bar{v}); by passing to the quotient, we therefore derive from vˉ\bar{v} a Z\mathbf{Z}-linear mapping ff of the cokernel M=(E⊗F)/HM = (E \otimes F)/H of uˉ\bar{u} into E′′⊗FE'' \otimes F; it must be proved that ff is bijective and it will hence suffice to define a Z\mathbf{Z}-linear mapping g:E′′⊗F→Mg : E'' \otimes F \to M such that g∘fg \circ f and f∘gf \circ g are the identity mappings.

Let x′′∈E′′x'' \in E'', y∈Fy \in F; by hypothesis there exists x∈Ex \in E such that v(x)=x′′v(x) = x''. We show that, if x1,x2x_1, x_2 are two elements of E such that v(x1)=v(x2)=x′′v(x_1) = v(x_2) = x'' and ϕ:E⊗F→M\phi : E \otimes F \to M is the canonical mapping, then ϕ(x1⊗y)=ϕ(x2⊗y)\phi(x_1 \otimes y) = \phi(x_2 \otimes y). It suffices to prove that if v(x)=0v(x) = 0 then ϕ(x⊗y)=0\phi(x \otimes y) = 0, which follows from the fact that x=u(x′)x = u(x') with x′∈E′x' \in E', whence x⊗y=u(x′)⊗y=uˉ(x′⊗y)∈Hx \otimes y = u(x') \otimes y = \bar{u}(x' \otimes y) \in H. If (x′′,y)(x'', y) is mapped to the unique value of ϕ(x⊗y)\phi(x \otimes y) for all x∈Ex \in E such that v(x)=x′′v(x) = x'', a mapping is defined of E′′×FE'' \times F into M; this mapping is Z\mathbf{Z}-bilinear and satisfies conditions (1) (no. 1), since v(xλ)=x′′λv(x \lambda) = x'' \lambda and (xλ)⊗y=x⊗(λy)(x \lambda) \otimes y = x \otimes (\lambda y) for x∈Ex \in E; hence there is a Z\mathbf{Z}-linear mapping gg of E′′⊗FE'' \otimes F into M such that g(x′′⊗y)=ϕ(x⊗y)g(x'' \otimes y) = \phi(x \otimes y) for y∈F,x∈Ey \in F, x \in E and x′′=v(x)x'' = v(x). This definition further proves that f∘gf \circ g coincides with the identity mapping for the elements of

E′′⊗FE'' \otimes F of the form x′′⊗yx'' \otimes y and hence f∘gf \circ g is the identity mapping of E′′⊗FE'' \otimes F; on the other hand, for x∈Ex \in E and y∈Fy \in F, f(ϕ(x⊗y))=v(x)⊗yf(\phi(x \otimes y)) = v(x) \otimes y by definition, hence g(f(ϕ(x⊗y)))=ϕ(x⊗y)g(f(\phi(x \otimes y))) = \phi(x \otimes y) and, as the elements of the form ϕ(x⊗y)\phi(x \otimes y) generate M, g∘fg \circ f is the identity mapping of M.

Corollary

Let F, F', F'' be left A-modules, E a right A-module and*

F′→sF→tF′′⟶0 \begin{array}{ccccc} F' & \xrightarrow{s} & F & \xrightarrow{t} & F'' \longrightarrow 0 \end{array}
(16) an exact sequence of linear mappings. Writing \bar{s} = lE \otimes s, \bar{t} = l_E \otimes t, the sequence of \mathbf{Z}-homomorphisms*
E⊗AF′→sˉE⊗AF→tˉE⊗AF′′⟶0 \begin{array}{cccccc} E \otimes_A F' & \xrightarrow{\bar{s}} & E \otimes_A F & \xrightarrow{\bar{t}} & E \otimes_A F'' & \longrightarrow 0 \end{array}
(17) *is exact.

Cites 3 · Cited by 2

Cites Proposition 1, Corollary 2, Proposition 5

Cited by Lemma 1, AC I § 2, Proposition 6

When E (resp. F) is considered as a left (resp. right) A^0-module, F \otimes_{A^0} E is identified with E \otimes_A F and there are analogous identifications for F' \otimes_{A^0} E and F'' \otimes_{A^0} E (no. 1, Corollary 2 to Proposition 1); the corollary then follows immediately from Proposition 5.

Remark

Note that in general, if E' is a submodule of a right A-module E and j : E' \to E the canonical injection, the canonical mapping

j⊗1F:E′⊗F→E⊗F j \otimes 1_F : E' \otimes F \to E \otimes F
is not necessarily injective. In other words, for an exact sequence
0⟶E′→uE→vE′′⟶0 \begin{array}{ccccccc} 0 & \longrightarrow & E' & \xrightarrow{u} & E & \xrightarrow{v} & E'' \longrightarrow 0 \end{array}
(18) it cannot in general be concluded that the sequence
0⟶E′⊗F→uE⊗F→vE′′⊗F⟶0 \begin{array}{ccccccc} 0 & \longrightarrow & E' \otimes F & \xrightarrow{u} & E \otimes F & \xrightarrow{v} & E'' \otimes F \longrightarrow 0 \end{array}
(19) *is exact.

Take for example A = \mathbf{Z}, E = \mathbf{Z}, E' = 2\mathbf{Z}, F = \mathbf{Z}/2\mathbf{Z}. As E' is isomorphic to E, E' \otimes F is isomorphic to E \otimes F which is itself isomorphic to F (no. 4, Proposition 4). But for all x' = 2x \in E' (where x \in E) and all y \in F, j(x') \otimes y = (2x) \otimes y = x \otimes (2y) = 0, since 2y = 0, and the canonical image of E' \otimes F in E \otimes F reduces to 0.

In other words, care must be taken to distinguish, for a submodule E' of E and an element x \in E', between the element x \otimes y "calculated in E' \otimes F'" and the element x \otimes y "calculated in E \otimes F" (in other words, the element j(x) \otimes y).

We shall study later, under the name of flat modules, the modules F such that the sequence (19) is exact for every exact sequence (18) (Commutative Algebra, I, § 2).

Cites 1

Cites Proposition 4

Proposition 6

Given two exact sequences (14) and (16), the homomorphism v \otimes t : E \otimes_A F \to E'' \otimes_A F'' is surjective and its kernel is equal to*

Im⁡(u⊗1F)+Im⁡(lE⊗s) \operatorname{Im}(u \otimes 1_F) + \operatorname{Im}(l_E \otimes s)

Cites 2 · Cited by 4

Cites Proposition 5, Corollary

Cited by Corollary 3, II § 11, Remark, Exercise 1, § 1, Proposition 3, III § 5

Now v⊗t=(v⊗1F′′)∘(1E⊗t)v \otimes t = (v \otimes 1_{F''}) \circ (1_E \otimes t) (no. 2, formula (5)) and v⊗tv \otimes t is therefore surjective, being the composition of two surjective homomorphisms by virtue of Proposition 5 and its Corollary. On the other hand, for z∈E⊗Fz \in E \otimes F to be in the kernel of v⊗tv \otimes t, it is necessary and sufficient that (1E⊗t)(z)(1_E \otimes t)(z) belong to the kernel of v⊗1F′′v \otimes 1_{F''}, that is, by virtue of (15) to the image of

u⊗1F′′:E′⊗F′′→E⊗F′′. u \otimes 1_{F''}: E' \otimes F'' \to E \otimes F''.

But as the homomorphism t:F→F′′t: F \to F'' is surjective, so is

1E′⊗t:E′⊗F→E′⊗F′′ 1_{E'} \otimes t: E' \otimes F \to E' \otimes F''

by the Corollary to Proposition 5, hence the condition on zz reduces to the existence of an a∈E′⊗Fa \in E' \otimes F such that

(lE⊗t)(z)=(u⊗t)(a). (l_E \otimes t)(z) = (u \otimes t)(a).

Let b=z−(u⊗1F)(a)b = z - (u \otimes 1_F)(a); then (lE⊗t)(b)=0(l_E \otimes t)(b) = 0, and by virtue of (17), bb belongs to the image of lE⊗sl_E \otimes s, which proves the proposition.

In other words:

Corollary 1

Let E′E' be a submodule of a right AA-module EE, F′F' a submodule of a left AA-module FF and Im⁡(E′⊗AF)\operatorname{Im}(E' \otimes_A F) and Im⁡(E⊗AF′)\operatorname{Im}(E \otimes_A F') the sub-Z\mathbf{Z}-modules of E⊗AFE \otimes_A F, the respective images of the canonical mappings E′⊗AF→E⊗AFE' \otimes_A F \to E \otimes_A F,

E⊗AF′→E⊗AF. E \otimes_A F' \to E \otimes_A F.

Then there is a canonical Z\mathbf{Z}-module isomorphism

Cited by 2

Cited by Proposition 1, III § 4, Proposition 8, IV § 6

(20)π:(E/E′)⊗A(F/F′)→(E⊗AF)/(Im⁡(E′⊗AF)+Im⁡(E⊗AF′))(20)\quad \pi : (E/E') \otimes_A (F/F') \to (E \otimes_A F)/(\operatorname{Im}(E' \otimes_A F) + \operatorname{Im}(E \otimes_A F'))

*such that, for ξ∈E/E′\xi \in E/E', η∈F\eta \in F, π(ξ⊗η)\pi(\xi \otimes \eta) is the class of all elements x⊗y∈E⊗AFx \otimes y \in E \otimes_A F such that x∈ξx \in \xi and y∈ηy \in \eta.

Note that when EE is a ((Bi′);A,(Cj′))((B'_i); A, (C'_j))-multimodule, FF a (A,(Bh′′);(Ck′′))(A, (B''_h); (C''_k))-multimodule and E′E' and F′F' submultimodules of EE and FF respectively, the isomorphism (20) is an isomorphism for the ((Bi′),(Bh′′);(Cj′),(Ck′′))((B'_i), (B''_h); (C'_j), (C''_k))-multimodule structures of the two sides (no. 3).

Corollary 2

Let aa be a right ideal of AA, FF a left AA-module and aFaF the sub-Z\mathbf{Z}-module of FF generated by the elements of the form λx\lambda x, where λ∈a\lambda \in a and x∈Fx \in F. Then there is a canonical Z\mathbf{Z}-module isomorphism

Cited by 1

Cited by Proposition 7, II § 11

(21)π:(A/a)⊗AF→F/aF(21)\quad \pi : (A/a) \otimes_A F \to F/aF

*such that, for all λˉ∈A/a\bar{\lambda} \in A/a and all x∈Fx \in F, π(λˉ⊗x)\pi(\bar{\lambda} \otimes x) is the class mod. aFaF of λx\lambda x, where λ∈λˉ\lambda \in \bar{\lambda}.

In particular, for A=ZA = \mathbf{Z}, it is seen that for every integer nn and every Z\mathbf{Z}-module FF, (Z/nZ)⊗ZF(\mathbf{Z}/n\mathbf{Z}) \otimes_{\mathbf{Z}} F is canonically identified with the quotient Z\mathbf{Z}-module F/nFF/nF.

Corollary 3

Let AA be a commutative ring, aa an ideal of AA and EE and FF two AA-modules such that aa is contained in the annihilator of FF. Then the (A/a)(A/a)-modules E⊗AFE \otimes_A F and (E/aE)⊗A/aF(E/aE) \otimes_{A/a} F are canonically isomorphic.

Cites 1 · Cited by 1

Cites Corollary

Cited by Proposition 1, AC III § 5

FF and E⊗AFE \otimes_A F are annihilated by aa and hence have canonical (A/a)(A/a)-module structures (§1§ 1, no. 12) and if we write E′=aEE' = aE, then Im⁡(E′⊗AF)=0\operatorname{Im}(E' \otimes_A F) = 0; then there is a canonical isomorphism (20) of E⊗AFE \otimes_A F onto (E/aE)⊗AF(E/aE) \otimes_A F and the latter is itself identical with (E/aE)⊗A/aF(E/aE) \otimes_{A/a} F (no. 3, Corollary to Proposition 2).

Corollary 4

Let a,ba, b be two ideals in a commutative ring CC; the CC-module (C/a)⊗C(C/b)(C/a) \otimes_C (C/b) is then canonically isomorphic to C(a+b)C(a + b).

Cited by 1

Cited by Remark

7. Tensor products of products and direct sums

Let (Eλ)λ∈L(E_\lambda)_{\lambda \in L} be a family of right AA-modules, (Fμ)μ∈M(F_\mu)_{\mu \in M} a family of left AA-modules and consider the product modules C=∏λ∈LEλ,D=∏μ∈MFμC = \prod_{\lambda \in L} E_\lambda, D = \prod_{\mu \in M} F_\mu. The mapping ((xλ),(yμ))↦(xλ⊗yμ)((x_\lambda), (y_\mu)) \mapsto (x_\lambda \otimes y_\mu) of C×DC \times D into the product Z\mathbf{Z}-module

∏(λ,μ)∈L×M(Eλ⊗AFμ) \prod_{(\lambda, \mu) \in L \times M} (E_\lambda \otimes_A F_\mu)
is Z\mathbf{Z}-bilinear and obviously satisfies conditions (1) (no. 1). Thus there exists (no. 1, Proposition 1) a Z\mathbf{Z}-linear mapping, called canonical
f:(∏λ∈LEλ)⊗A(∏μ∈MFμ)→∏(λ,μ)∈L×M(Eλ⊗AFμ) f : \left( \prod_{\lambda \in L} E_\lambda \right) \otimes_A \left( \prod_{\mu \in M} F_\mu \right) \to \prod_{(\lambda, \mu) \in L \times M} (E_\lambda \otimes_A F_\mu)
such that f((xλ)⊗(yμ))=(xλ⊗yμ)f((x_\lambda) \otimes (y_\mu)) = (x_\lambda \otimes y_\mu).

When C=RL,D=SM,RC = R^L, D = S^M, R (resp. SS) being a right (resp. left) AA-module, the canonical mapping (22) associates with every tensor product u⊗vu \otimes v, where uu is a mapping of LL into RR and vv a mapping of MM into SS, the mapping (λ,μ)↦u(λ)⊗v(μ)(\lambda, \mu) \mapsto u(\lambda) \otimes v(\mu) of L×ML \times M into R⊗ASR \otimes_A S; even in this case the canonical mapping (22) is in general neither injective nor surjective (Exercise 3; cf. Corollary 3 to Proposition 7).

When the EλE_\lambda are ((Bi′);A,(Cj′))((B'_i); A, (C'_j))-multimodules and the FμF_\mu (A,(Bh′′);(Ckc′′))(A, (B''_h); (C''_{k_c}))-multimodules, the homomorphism (22) is also a homomorphism for the ((Bi′),(Bh′′);(Cj′),(Ckc′′))((B'_i), (B''_h); (C'_j), (C''_{k_c}))-multimodule structures of the two sides.

Consider now the submodule E=⨁λ∈LEλE = \bigoplus_{\lambda \in L} E_\lambda (resp. ⨁μ∈MFμ\bigoplus_{\mu \in M} F_\mu) of CC (resp. DD); the canonical injections E→C,F→DE \to C, F \to D define canonically a Z\mathbf{Z}-linear mapping E⊗AF→C⊗ADE \otimes_A F \to C \otimes_A D which, composed with the mapping (22), gives a Z\mathbf{Z}-linear mapping gg of E⊗AFE \otimes_A F into ∏λ,μ(Eλ⊗AFμ)\prod_{\lambda, \mu} (E_\lambda \otimes_A F_\mu) such that

g((xλ)⊗(yμ))=(xλ⊗yμ); g((x_\lambda) \otimes (y_\mu)) = (x_\lambda \otimes y_\mu);

moreover, as the families (xλ)(x_\lambda) and (yμ)(y_\mu) have finite support, so does (xλ⊗yμ)(x_\lambda \otimes y_\mu) and hence finally gg is a canonical homomorphism

g:(⨁λ∈LEλ)⊗A(⨁μ∈MFμ)→⨁(λ,μ)∈L×M(Eλ⊗AFμ), g : \left( \bigoplus_{\lambda \in L} E_\lambda \right) \otimes_A \left( \bigoplus_{\mu \in M} F_\mu \right) \to \bigoplus_{(\lambda, \mu) \in L \times M} (E_\lambda \otimes_A F_\mu),

which is a multimodule homomorphism under the same conditions as (22).

To prove this it suffices to define a Z\mathbf{Z}-linear mapping hh of the direct sum G=⨁(λ,μ)∈L×M(Eλ⊗AFμ)G = \bigoplus_{(\lambda, \mu) \in L \times M} (E_\lambda \otimes_A F_\mu) into E⊗AFE \otimes_A F such that g∘hg \circ h and h∘gh \circ g are the identity mappings. But, to define a Z\mathbf{Z}-linear mapping of GG into E⊗AFE \otimes_A F, it suffices (§ 1, no. 6, Proposition 6) to define a Z\mathbf{Z}-linear mapping

hλμ:Eλ⊗AFμ→E⊗AF h_{\lambda \mu} : E_\lambda \otimes_A F_\mu \to E \otimes_A F

for every ordered pair (λ,μ)(\lambda, \mu), and we take hλμ=iλ⊗jμh_{\lambda \mu} = i_\lambda \otimes j_\mu, where iλ:Eλ→Ei_\lambda : E_\lambda \to E and jμ:Fμ→Fj_\mu : F_\mu \to F are the canonical injections. Then clearly h∘gh \circ g coincides with the identity mapping for the elements of the form (∑λxλ)⊗(∑μyμ)\left( \sum_\lambda x_\lambda \right) \otimes \left( \sum_\mu y_\mu \right) which generate the Z\mathbf{Z}-module E⊗AFE \otimes_A F; similarly g∘hg \circ h coincides with the identity mapping for the elements of the form ∑λ,μ(xλ⊗yμ)\sum_{\lambda, \mu} (x_\lambda \otimes y_\mu) which generate the Z\mathbf{Z}-module GG, since for each ordered pair (λ,μ)(\lambda, \mu) the products

xλ⊗yμ(xλ∈Eλ,yμ∈Fμ) x_\lambda \otimes y_\mu \quad (x_\lambda \in E_\lambda, y_\mu \in F_\mu)

generate the Z\mathbf{Z}-module Eλ⊗AFμE_\lambda \otimes_A F_\mu. Whence the proposition.

Let uλ:Eλ→Eλ′u_\lambda : E_\lambda \to E'_\lambda, vμ:Fμ→Fμ′v_\mu : F_\mu \to F'_\mu be AA-homomorphisms; clearly the diagram

(⨁λEλ)⊗A(⨁μFμ)⟶⨁λ,μ(Eλ⊗A(Fμ))(⨁uλ)⊗(⨁vμ)↓↓(⨁(uλ⊗vμ))(⨁λEλ′)⊗A(⨁μFμ′)⟶⨁λ,μ(Eλ′⊗AFμ′) \begin{array}{ccc} \left( \bigoplus_\lambda E_\lambda \right) \otimes_A \left( \bigoplus_\mu F_\mu \right) & \longrightarrow & \bigoplus_{\lambda, \mu} (E_\lambda \otimes_A (F_\mu)) \\ (\bigoplus u_\lambda) \otimes (\bigoplus v_\mu) \downarrow & & \downarrow (\bigoplus (u_\lambda \otimes v_\mu)) \\ \left( \bigoplus_\lambda E'_\lambda \right) \otimes_A \left( \bigoplus_\mu F'_\mu \right) & \longrightarrow & \bigoplus_{\lambda, \mu} (E'_\lambda \otimes_A F'_\mu) \end{array}

is commutative.

Corollary 1

If the left AA-module FF admits a basis (bμ)μ∈M(b_\mu)_{\mu \in M}, every element of E⊗AFE \otimes_A F can be written uniquely in the form ∑μ(xμ⊗bμ)\sum_\mu (x_\mu \otimes b_\mu), where xμ∈Ex_\mu \in E and the family (xμ)(x_\mu) has finite support. The Z\mathbf{Z}-module E⊗AFE \otimes_A F is isomorphic to E(M)E^{(M)} considered as a Z\mathbf{Z}-module.

Cites 2 · Cited by 9

Cites Proposition 4, Proposition 7

Cited by Lemma 10, AC I § 2, Corollary 3, Corollary 4, Lemma 3, II § 4, Proposition 4, II § 5, Corollary, II § 7, Proposition 6, III § 4, Proposition 7, III § 4, Proposition 5, VIII § 5

The basis (bμ)(b_\mu) defines an isomorphism of FF onto ⨁μ∈MAbμ\bigoplus_{\mu \in M} Ab_\mu, whence there is an isomorphism E⊗AF→⨁μ∈M(E⊗AAbμ)E \otimes_A F \to \bigoplus_{\mu \in M} (E \otimes_A Ab_\mu) by virtue of Proposition 7; as ξ ↦ ξb_μ is an isomorphism of A_s onto Ab_μ, x ↦ x ⊗ b_μ is an isomorphism of E onto E ⊗_A (Ab_μ) by virtue of Proposition 4 of no. 4, whence the corollary.

If E is a ((B'_i); A, (C'_j))-multimodule, the canonical isomorphism E ⊗_A F → E^{(M)} is a ((B'_i); (C'_j))-multimodule isomorphism.

In particular, if also E admits a basis (a_λ)_{λ ∈ L}, every z ∈ E ⊗_A F may be written in one and only one way in the form ∑λ,μ(aλξλμ)⊗bμ\sum_{\lambda, \mu} (a_\lambda \xi_{\lambda\mu}) \otimes b_\mu, where the ξ_{λμ} belong to A (and form a family of finite support); the mapping z↦(ξλμ)(λ,μ)∈L×Mz \mapsto (\xi_{\lambda\mu})_{(\lambda, \mu) \in L \times M} is an isomorphism of E ⊗_A F onto A^{(L × M)} for the \mathbf{Z}-module structures (and even the module structures over the centre of A). More particularly:

Corollary 2

If E and F are two free modules over a commutative ring C and (a_λ) (resp. (b_μ)) is a basis of the C-module E (resp. F), then (a_λ ⊗ b_μ) is a basis of the C-module E ⊗_C F.

Cited by 9

Cited by II § 10. Matrices, § 3. Tensor products, Corollary 7, Remark, II § 7, III § 1. Algebras, Proposition 4, III § 4, Remark, III § 5, Theorem 1, III § 7, Proposition 1, VIII § 12

By an abuse of language, the basis (a_λ ⊗ b_μ) is sometimes called the tensor product of the bases (a_λ) and (b_μ).

Remark (1). Let E be a free right A-module, F a free left A-module, (a_λ)_{λ ∈ L} a basis of E and (b_μ)_{μ ∈ M} a basis of F. Every element z ∈ E ⊗_A F may be written uniquely as ∑λaλ⊗yλ\sum_\lambda a_\lambda \otimes y_\lambda, where y_λ ∈ F, and also uniquely as ∑μxμ⊗bμ\sum_\mu x_\mu \otimes b_\mu, where x_μ ∈ E. If we write yλ=∑μηλμbμy_\lambda = \sum_\mu \eta_{\lambda\mu} b_\mu, xμ=∑λaλξλμx_\mu = \sum_\lambda a_\lambda \xi_{\lambda\mu}, where the ξ_{λμ} and η_{λμ} belong to A, then ξλμ=ηλμ\xi_{\lambda\mu} = \eta_{\lambda\mu} for all (λ, μ), for

∑λ(aλ⊗(∑μηλμbμ))=∑λ,μ((aληλμ)⊗bμ)=∑μ((∑λaληλμ)⊗bμ). \sum_\lambda \left( a_\lambda \otimes \left( \sum_\mu \eta_{\lambda\mu} b_\mu \right) \right) = \sum_{\lambda, \mu} ((a_\lambda \eta_{\lambda\mu}) \otimes b_\mu) = \sum_\mu \left( \left( \sum_\lambda a_\lambda \eta_{\lambda\mu} \right) \otimes b_\mu \right).

Corollary 3

Let (E_λ)_{λ ∈ L} be a family of right A-modules and F a free (resp. finitely generated free) left A-module. Then the canonical mapping (22)

Cites 3 · Cited by 2

Cites Corollary 1, Proposition 7, Proposition 4

Cited by § 3. Tensor products, Proposition 15, II § 7

(∏λ∈LEλ)⊗AF→∏λ∈L(Eλ⊗AF) \left( \prod_{\lambda \in L} E_\lambda \right) \otimes_A F \to \prod_{\lambda \in L} (E_\lambda \otimes_A F)

is injective (resp. bijective).

If (b_μ) is a basis of F, every element of (∏λ∈LEλ)⊗AF\left( \prod_{\lambda \in L} E_\lambda \right) \otimes_A F can be written uniquely as z=∑μ((xλ(μ))⊗bμ)z = \sum_\mu ((x_\lambda^{(\mu)}) \otimes b_\mu) (Corollary 1); to say that its canonical image is zero means that, for all λ ∈ L, ∑μ(xλ(μ)⊗bμ)=0\sum_\mu (x_\lambda^{(\mu)} \otimes b_\mu) = 0, hence xλ(μ)=0x_\lambda^{(\mu)} = 0 for all λ ∈ L and all μ (Corollary 1) and therefore z=0z = 0.

Showing that the canonical mapping is bijective when F admits a finite basis is immediately reduced, by virtue of Proposition 7, to the case where

F = A_s; but then the two sides are canonically identified with ∏λ∈LEλ\prod_{\lambda \in L} E_\lambda (no. 4, Proposition 4) and after these identifications the canonical mapping (22) becomes the identity.

Corollary 4

Let A be a ring with no divisor of zero, E a free right A-module and F a free left A-module. Then the relation x ⊗\otimes y = 0 in E ⊗A\otimes_A F implies x = 0 or y = 0.

Cites 1 · Cited by 1

Cites Corollary 1

Cited by Proposition 1, VII § 4

Let (aλ)(a_\lambda) be a basis of E, (bμ)(b_\mu) a basis of F and let x=∑λaλξλ,y=∑μημbμx = \sum_\lambda a_\lambda \xi_\lambda, y = \sum_\mu \eta_\mu b_\mu; then x⊗y=∑λ,μ((aλξλημ)⊗bμx \otimes y = \sum_{\lambda, \mu} ((a_\lambda \xi_\lambda \eta_\mu) \otimes b_\mu and the relation x⊗y=0x \otimes y = 0 implies ξλημ=0\xi_\lambda \eta_\mu = 0 for every ordered pairs of indices (λ,μ)(\lambda, \mu) (Corollary 1). Hence, if x≠0x \neq 0, that is ξλ≠0\xi_\lambda \neq 0 for at least one λ\lambda, it follows that ημ=0\eta_\mu = 0 for all μ\mu, whence y=0y = 0.

Corollary 5

Let E be a right A-module, F a left A-module, M a submodule of E and N a submodule of F. If M is a direct factor of E and N a direct factor of F, the canonical homomorphism M⊗AN→E⊗AFM \otimes_A N \to E \otimes_A F is injective and the image of M⊗ANM \otimes_A N under this homomorphism is a direct factor of the Z\mathbf{Z}-module E⊗AFE \otimes_A F.

Cites 1 · Cited by 4

Cites Proposition 7

Cited by Lemma 1, AC I § 2, Corollary 7, Proposition 14, II § 7, Proposition 2, III § 11

This follows immediately from Proposition 7.

Note that if E is a ((Bi′);A,(Cj′))((B'_i); A, (C'_j))-multimodule and F an (A,(Bh′′);(Ck′′))(A, (B''_h); (C''_k))-multimodule and M and N direct factors in these multimodules, M⊗NM \otimes N is a direct factor of the ((Bi′),(Bh′′);(Cj′),(Ck′′))((B'_i), (B''_h); (C'_j), (C''_k))-multimodule E⊗FE \otimes F.

Corollary 6

Let P be a projective left A-module and E, F two right A-modules. For every injective homomorphism u:E→Fu : E \to F, the homomorphism*

u⊗1P:E⊗AP→F⊗AP u \otimes 1_P : E \otimes_A P \to F \otimes_A P
*is injective.

Cites 4 · Cited by 4

Cites Proposition 4, II § 2, Proposition 7, Corollary 1, II § 1, Proposition 4

Cited by Proposition 7, II § 11, Corollary, II § 4, Proposition 9, VIII § 11, Proposition 11, VIII § 6

There exists a left A-module Q such that L=P⊕QL = P \oplus Q is free (§ 2, no. 2, Proposition 4) and u⊗1Lu \otimes 1_L is identified (Proposition 7) with

(u⊗1P)⊕(u⊗1Q);(u \otimes 1_P) \oplus (u \otimes 1_Q);
it therefore suffices to prove the corollary when P is free (§ 1, no. 6, Corollary 1 to Proposition 7). The same argument reduces the problem to the case where P=AsP = A_s, which follows immediately from no. 4, Proposition 4.

Corollary 7

Let C be a commutative ring. If E and F are two projective C-modules, E⊗CFE \otimes_C F is a projective C-module.

Cites 2

Cites Corollary 2, Corollary 5

This follows immediately from Corollary 5 and the fact that the tensor product of two free C-modules is a free C-module (Corollary 2).

Remark 2

Under the hypotheses of Proposition 7, let Eλ′E'_\lambda be a submodule of EλE_\lambda, Fμ′F'_\mu a submodule of FμF_\mu and let E′=⨁λ∈LEλ′E' = \bigoplus_{\lambda \in L} E'_\lambda, F′=⨁μ∈MFμ′F' = \bigoplus_{\mu \in M} F'_\mu. Let Im(E′⊗AF′)(E' \otimes_A F') (resp. Im(E'_λ ⊗_A F'_μ)) denote the image of E' ⊗_A F' (resp. E'_λ ⊗_A F'_μ) in E ⊗_A F (resp. E_λ ⊗_A F_μ) under the canonical mapping; then the isomorphism (23) identifies the sub-Z\mathbf{Z}-modules

Im⁡(E′⊗AF′)and⨁(λ,μ)∈I×MIm⁡(Eλ′⊗AFμ′); \operatorname{Im}(E' \otimes_A F') \quad \text{and} \quad \bigoplus_{(\lambda, \mu) \in I \times M} \operatorname{Im}(E'_\lambda \otimes_A F'_\mu);

this follows immediately from the commutativity of the diagram

(⨁λEλ′)⊗A(⨁μFμ′)⟶(⨁λEλ)⊗A(⨁μFμ)↓↓⨁λ,μ(Eλ′⊗AFμ′)⟶⨁λ,μ(Eλ⊗AFμ) \begin{array}{ccc} \left( \bigoplus_\lambda E'_\lambda \right) \otimes_A \left( \bigoplus_\mu F'_\mu \right) & \longrightarrow & \left( \bigoplus_\lambda E_\lambda \right) \otimes_A \left( \bigoplus_\mu F_\mu \right) \\ \downarrow & & \downarrow \\ \bigoplus_{\lambda, \mu} (E'_\lambda \otimes_A F'_\mu) & \longrightarrow & \bigoplus_{\lambda, \mu} (E_\lambda \otimes_A F_\mu) \end{array}

where the vertical arrows are the canonical isomorphisms.

Cites 1

Cites Proposition 7

8. Associativity of the tensor product

Proposition 8

Let A, B be two rings, E a right A-module, F an (A, B)-bimodule and G a left B-module. Then E ⊗_A F is a right B-module, F ⊗_B G a left A-module and there exists one and only one Z\mathbf{Z}-linear mapping

Cites 1 · Cited by 6

Cites Proposition 1

Cited by Proposition 8, AC I § 2, § 3. Tensor products, Proposition 3, II § 5, Proposition 6, II § 5, Example, III § 4, Definition 1, VIII § 6

ϕ:(E⊗AF)⊗BG→E⊗A(F⊗BG) \phi : (E \otimes_A F) \otimes_B G \to E \otimes_A (F \otimes_B G)

such that ϕ((x⊗y)⊗z)=x⊗(y⊗z)\phi((x \otimes y) \otimes z) = x \otimes (y \otimes z) for x∈E, y∈F, z∈G;x \in E,\ y \in F,\ z \in G; moreover this Z\mathbf{Z}-linear mapping is bijective ("associativity" of the tensor product).

The right B-module structure on E ⊗_A F and left A-module structure on F ⊗_B G have been defined in no. 4. The uniqueness of ϕ\phi is obvious since the elements (x⊗y)⊗z(x \otimes y) \otimes z generate the Z\mathbf{Z}-module (E⊗AF)⊗BG(E \otimes_A F) \otimes_B G. To show the existence of ϕ\phi, we note that, for all z∈G, hz:y↦y⊗zz \in G,\ h_z : y \mapsto y \otimes z is an A-linear mapping of the left A-module F into the left A-module F ⊗_B G. We write gz=1E⊗hzg_z = 1_E \otimes h_z, which is therefore a Z\mathbf{Z}-linear mapping of E ⊗_A F into E ⊗_A (F ⊗_B G) and consider the mapping (t,z)↦gz(t)(t, z) \mapsto g_z(t) from (E⊗AF×G(E \otimes_A F \times G into E⊗A(F⊗BG)E \otimes_A (F \otimes_B G); as hz+z′=hz+hz′h_{z + z'} = h_z + h_{z'} for z∈G, z′∈Gz \in G,\ z' \in G, it is immediate that the above mapping is Z\mathbf{Z}-bilinear. Further, we show that for all μ∈B, gμz(t)=gz(tμ)\mu \in B,\ g_{\mu z}(t) = g_z(t \mu); it is obviously sufficient to do this for t=x⊗yt = x \otimes y where x∈Ex \in E and y∈Fy \in F; now and

gμz(x⊗y)=x⊗(y⊗μz) g_{\mu z}(x \otimes y) = x \otimes (y \otimes \mu z)
gz((x⊗y)μ)=gz(x⊗yμ)=x⊗(yμ⊗z). g_z((x \otimes y)\mu) = g_z(x \otimes y\mu) = x \otimes (y\mu \otimes z).

Proposition 1 (no. 1) then proves the existence of a Z\mathbf{Z}-linear mapping

ϕ:(E⊗AF)⊗BG→E⊗A(F⊗BG) \phi : (E \otimes_A F) \otimes_B G \to E \otimes_A (F \otimes_B G)

such that ϕ(t⊗z)=gz(t)\phi(t \otimes z) = g_z(t), hence ϕ((x⊗y)⊗z)=x⊗(y⊗z)\phi((x \otimes y) \otimes z) = x \otimes (y \otimes z). Similarly a Z\mathbf{Z}-linear mapping

ψ:E⊗A(F⊗BG)→(E⊗AF)⊗BG \psi : E \otimes_A (F \otimes_B G) \to (E \otimes_A F) \otimes_B G

is defined such that ψ(x⊗(y⊗z))=(x⊗y)⊗z\psi(x \otimes (y \otimes z)) = (x \otimes y) \otimes z and clearly ψ∘ϕ\psi \circ \phi and ϕ∘ψ\phi \circ \psi are the identity mappings of (E⊗AF)⊗BG(E \otimes_A F) \otimes_B G and E⊗A(F⊗BG)E \otimes_A (F \otimes_B G) respectively, since they reduce to the identity mappings on generating systems of these Z\mathbf{Z}-modules.

It is immediate that, if EE is a ((Ci′);A,(Dj′))((C'_i); A, (D'_j))-multimodule, FF an (A,(Ch′′);B,(Dk′′))(A, (C''_h); B, (D''_k))-multimodule and GG a (B,(Cl′′′);(Dm′′′))(B, (C'''_l); (D'''_m))-multimodule, the canonical isomorphism defined in Proposition 8 is a ((Ci′),(Ch′′),(Cl′′′);(Dj′),(Dk′′),(Dm′′′))((C'_i), (C''_h), (C'''_l); (D'_j), (D''_k), (D'''_m))-multimodule isomorphism. In particular, if CC is a commutative ring and E,F,GE, F, G three CC-modules, there is a canonical CC-module isomorphism

(E⊗CF)⊗CG→E⊗C(F⊗CG). (E \otimes_C F) \otimes_C G \to E \otimes_C (F \otimes_C G).

We shall see below that, under certain conditions, the definition of tensor product can be generalized to a family of multimodules, which will in particular give us under the hypotheses of Proposition 8 a Z\mathbf{Z}-module E⊗AF⊗BGE \otimes_A F \otimes_B G, which is canonically isomorphic to each of the Z\mathbf{Z}-modules (E⊗AF)⊗BG(E \otimes_A F) \otimes_B G and E⊗A(F⊗BG)E \otimes_A (F \otimes_B G) and with which the latter will be identified.

9. Tensor product of families of multimodules

Let (Gλ)λ∈L(G_\lambda)_{\lambda \in L} be a family of Z\mathbf{Z}-modules; a mapping uu of the set G=∏λ∈LGλG = \prod_{\lambda \in L} G_\lambda into a Z\mathbf{Z}-module is said to be multiadditive (or Z\mathbf{Z}-*multilinear*) if (xλ)↦u((xλ))(x_\lambda) \mapsto u((x_\lambda)) is additive with respect to each of the variables xλx_\lambda; to be precise, this means that, for all μ∈L\mu \in L and every element (aλ)∈∏λ≠μGλ(a_\lambda) \in \prod_{\lambda \neq \mu} G_\lambda, canonically identifying GG with Gμ×∏λ≠μGλG_\mu \times \prod_{\lambda \neq \mu} G_\lambda,

u(xμ+yμ,(aλ))=u(xμ,(aλ))+u(yμ,(aλ))for xμ,yμ in Gμ. u(x_\mu + y_\mu, (a_\lambda)) = u(x_\mu, (a_\lambda)) + u(y_\mu, (a_\lambda)) \quad \text{for } x_\mu, y_\mu \text{ in } G_\mu.

This implies in particular that u((xλ))=0u((x_\lambda)) = 0 if one of the xλx_\lambda is zero.

We consider also the universal mapping problem where Σ\Sigma is the species of Z\mathbf{Z}-module structure and the α\alpha-mappings are the multiadditive mappings of GG into a Z\mathbf{Z}-module. A solution is still obtained by considering the Z\mathbf{Z}-module C=Z(G)C = \mathbf{Z}^{(G)} of formal linear combinations of elements of GG with coefficients in Z\mathbf{Z} and the sub-Z\mathbf{Z}-module DD of CC generated by the elements of the form

(xμ+yμ,(zλ)λ≠μ)−(xμ,(zλ)λ≠μ)−(yμ,(zλ)λ≠μ) (x_\mu + y_\mu, (z_\lambda)_{\lambda \neq \mu}) - (x_\mu, (z_\lambda)_{\lambda \neq \mu}) - (y_\mu, (z_\lambda)_{\lambda \neq \mu})

where μ∈L,xμ∈Gμ,yμ∈Gμ\mu \in L, x_\mu \in G_\mu, y_\mu \in G_\mu and the zλ∈Gλz_\lambda \in G_\lambda (λ≠μ\lambda \neq \mu) are arbitrary. The quotient Z\mathbf{Z}-module C/DC/D is called the tensor product (over Z\mathbf{Z}) of the family (Gλ)λ∈L(G_\lambda)_{\lambda \in L} of Z\mathbf{Z}-*modules* and is denoted by ⨂λ∈LGλ\bigotimes_{\lambda \in L} G_\lambda; for every element (xλ)λ∈L(x_\lambda)_{\lambda \in L} of GG which is an element of the canonical basis of CC, ⨂λ∈Lxλ\bigotimes_{\lambda \in L} x_\lambda denotes the canonical image of this element in C/D. It follows from the above definitions that the mapping ϕ:(xλ)↦⨂λ∈Lxλ\phi : (x_\lambda) \mapsto \bigotimes_{\lambda \in L} x_\lambda of G into ⨂λ∈LGλ\bigotimes_{\lambda \in L} G_\lambda is Z\mathbf{Z}-multilinear and that, for every Z\mathbf{Z}-multilinear mapping ff of G into a Z\mathbf{Z}-module H, there exists one and only one Z\mathbf{Z}-linear mapping g:⨂λ∈LGλ→Hg : \bigotimes_{\lambda \in L} G_\lambda \to H such that f=g∘ϕf = g \circ \phi; the ordered pair (⨂λ∈LGλ,ϕ)\left( \bigotimes_{\lambda \in L} G_\lambda, \phi \right) thus resolves the universal mapping problem in question.

Let (Gλ′)λ∈L(G'_\lambda)_{\lambda \in L} be another family of Z\mathbf{Z}-modules and, for all λ∈L\lambda \in L, let vλ:Gλ→Gλ′v_\lambda : G_\lambda \to G'_\lambda be a Z\mathbf{Z}-linear mapping (in other words a homomorphism of commutative groups). Then the mapping

(xλ)↦⨂λ∈Lvλ(xλ) (x_\lambda) \mapsto \bigotimes_{\lambda \in L} v_\lambda(x_\lambda)
of G into ⨂λ∈LGλ′\bigotimes_{\lambda \in L} G'_\lambda is Z\mathbf{Z}-multilinear and hence defines canonically a Z\mathbf{Z}-linear mapping of ⨂λ∈LGλ\bigotimes_{\lambda \in L} G_\lambda into ⨂λ∈LGλ′\bigotimes_{\lambda \in L} G'_\lambda denoted by ⨂λ∈Lvλ\bigotimes_{\lambda \in L} v_\lambda and such that
(⨂λ∈Lvλ)(⨂λ∈Lxλ)=⨂λ∈Lvλ(xλ). \left( \bigotimes_{\lambda \in L} v_\lambda \right) \left( \bigotimes_{\lambda \in L} x_\lambda \right) = \bigotimes_{\lambda \in L} v_\lambda(x_\lambda).
In particular, we consider, for some μ∈L\mu \in L, an endomorphism θ\theta of GμG_\mu; we denote by θ~\tilde{\theta} the endomorphism of ⨂λ∈LGλ\bigotimes_{\lambda \in L} G_\lambda equal to ⨂λ∈Lvλ\bigotimes_{\lambda \in L} v_\lambda where vμ=θv_\mu = \theta and vλ=1Gλv_\lambda = 1_{G_\lambda} for λ≠μ\lambda \neq \mu.

Then we suppose given a set Ω\Omega, a mapping

c:ω↦(ρ(ω),σ(ω)) c : \omega \mapsto (\rho(\omega), \sigma(\omega))
of Ω\Omega into L×LL \times L and, for all ω∈Ω\omega \in \Omega, an endomorphism pωp_\omega of Gρ(ω)G_{\rho(\omega)} and an endomorphism qωq_\omega of Gσ(ω)G_{\sigma(\omega)}; there correspond to them two endomorphisms p~ω\tilde{p}_\omega and q~ω\tilde{q}_\omega of P=⨂λ∈LGλP = \bigotimes_{\lambda \in L} G_\lambda. Let R be the sub-Z\mathbf{Z}-module of P generated by the union of the images of the endomorphisms p~ω−q~ω\tilde{p}_\omega - \tilde{q}_\omega when ω\omega runs through Ω\Omega. The quotient Z\mathbf{Z}-module P/RP/R is called the tensor product of the family (Gλ)λ∈L(G_\lambda)_{\lambda \in L} relative to c, p, q and is denoted by ⨂(c,p,q)Gλ\bigotimes_{(c, p, q)} G_\lambda; composing the canonical homomorphism P→P/RP \to P/R with the mapping ϕ:G→⨂λ∈LGλ\phi : G \to \bigotimes_{\lambda \in L} G_\lambda defined above, we obtain a Z\mathbf{Z}-multilinear mapping ϕ(c,p,q):G→⨂(c,p,q)Gλ\phi_{(c, p, q)} : G \to \bigotimes_{(c, p, q)} G_\lambda and write ϕ(c,p,q)((xλ))=⨂(c,p,q)xλ\phi_{(c, p, q)}((x_\lambda)) = \bigotimes_{(c, p, q)} x_\lambda or simply ⨂(c)xλ\bigotimes_{(c)} x_\lambda. The ordered pair consisting of ⨂(c,p,q)xλ\bigotimes_{(c, p, q)} x_\lambda and ϕ(c,p,q)\phi_{(c, p, q)} resolves the following universal mapping problem: let pˉω\bar{p}_\omega (resp. qˉω\bar{q}_\omega) denote the mapping
(xρ(ω),(xλ)λ≠ρ(ω))↦(pω(xρ(ω)),(xλ)λ≠ρ(ω)) (x_{\rho(\omega)}, (x_\lambda)_{\lambda \neq \rho(\omega)}) \mapsto (p_\omega(x_{\rho(\omega)}), (x_\lambda)_{\lambda \neq \rho(\omega)})
(resp. (xσ(ω),(xλ)λ≠σ(ω))↦(qω(xσ(ω)),(xλ)λ≠σ(ω))(x_{\sigma(\omega)}, (x_\lambda)_{\lambda \neq \sigma(\omega)}) \mapsto (q_\omega(x_{\sigma(\omega)}), (x_\lambda)_{\lambda \neq \sigma(\omega)})) of G into itself. Then Σ\Sigma is taken to be the species of Z\mathbf{Z}-module structure and the α-mappings the Z\mathbf{Z}-multilinear mappings uu of GG into a Z\mathbf{Z}-module which further satisfy the conditions

(26)(26)
u∘pˉω=u∘qˉω u \circ \bar{p}_\omega = u \circ \bar{q}_\omega

for all ω∈Ω\omega \in \Omega. The proof is obvious from the above definitions.

This construction recovers in particular that of E⊗AFE \otimes_A F described in no. 1: in this case we take L={1,2}L = \{1, 2\}, G1=EG_1 = E, G2=FG_2 = F, Ω=A\Omega = A; further, for all ω∈A\omega \in A, we must have ρ(ω)=1\rho(\omega) = 1, σ(ω)=2\sigma(\omega) = 2, ω\omega is the endomorphism x↦xωx \mapsto x\omega of the Z\mathbf{Z}-module EE and qωq_\omega the endomorphism y↦ωyy \mapsto \omega y of the Z\mathbf{Z}-module FF.

Let (Gλ′)λ∈L(G'_\lambda)_{\lambda \in L} be a second family of Z\mathbf{Z}-modules; keeping the mapping cc the same, suppose given for all ω∈Ω\omega \in \Omega, an endomorphism pω′p'_\omega of Gρ(ω)′G'_{\rho(\omega)} and an endomorphism qω′q'_\omega of Gσ(ω)′G'_{\sigma(\omega)}. For all λ∈L\lambda \in L, then let vλ:Gλ→Gλ′v_\lambda : G_\lambda \to G'_\lambda be a Z\mathbf{Z}-linear mapping such that, for all ω∈Ω\omega \in \Omega,

(27)(27)
vρ(ω)∘pω=pω′∘vρ(ω)andvσ(ω)∘qω=qω′∘vσ(ω) v_{\rho(\omega)} \circ p_\omega = p'_\omega \circ v_{\rho(\omega)} \quad \text{and} \quad v_{\sigma(\omega)} \circ q_\omega = q'_\omega \circ v_{\sigma(\omega)}

(in other words, for all λ∈L\lambda \in L, vλv_\lambda is a morphism for the laws of action on GλG_\lambda (resp. Gλ′G'_\lambda) defined on the pξp_\xi and qηq_\eta (resp. pξ′p'_\xi and qη′q'_\eta) with ξ\xi and η\eta such that ρ(ξ)=λ\rho(\xi) = \lambda and σ(η)=λ\sigma(\eta) = \lambda). Then the mapping

u:(xλ)↦⨂(c,p′,q′)vλ(xλ) u : (x_\lambda) \mapsto \bigotimes_{(c, p', q')} v_\lambda(x_\lambda)
of GG into ⨂(c,p′,q′)Gλ′\bigotimes_{(c, p', q')} G'_\lambda is Z\mathbf{Z}-multilinear and satisfies conditions (26) and hence defines a Z\mathbf{Z}-linear mapping of ⨂(c,p,q)Gλ\bigotimes_{(c, p, q)} G_\lambda into ⨂(c,p′,q′)Gλ′\bigotimes_{(c, p', q')} G'_\lambda, which we shall denote simply by ⨂(c)vλ\bigotimes_{(c)} v_\lambda if no confusion can arise.

We shall now give an "associativity" property for the general tensor products thus defined. Let (Li)1≤i≤n(L_i)_{1 \leq i \leq n} be a finite partition of LL; for every index ii, let Ωi\Omega_i denote the subset of Ω\Omega consisting of the elements such that ρ(ω)∈Li\rho(\omega) \in L_i and σ(ω)∈Li\sigma(\omega) \in L_i; clearly the Ωi\Omega_i are pairwise disjoint; we set Ω′=Ω−(⋃iΩi)\Omega' = \Omega - \left( \bigcup_i \Omega_i \right). For each index ii, c(i)c^{(i)} will denote the mapping ω↦(ρ(ω),σ(ω))\omega \mapsto (\rho(\omega), \sigma(\omega)) of Ωi\Omega_i into Li×LiL_i \times L_i; for ω∈Ωi\omega \in \Omega_i, we write pω(i)p^{(i)}_\omega and qω(i)q^{(i)}_\omega instead of pωp_\omega and qωq_\omega. Then for each ii there is a "partial" tensor product

Fi=⨂(c(i),p(i),q(i))Gλ. F_i = \bigotimes_{(c^{(i)}, p^{(i)}, q^{(i)})} G_\lambda.

We shall further make the following "permutability" hypothesis:

(P) If ω∈Ω′\omega \in \Omega', pωp_\omega (resp. qωq_\omega) permutes with each of the endomorphisms pξp_\xi and qηq_\eta of Gρ(ω)G_{\rho(\omega)} (resp. Gσ(ω)G_{\sigma(\omega)}) such that ξ∉Ω′\xi \notin \Omega', η∉Ω′\eta \notin \Omega' and ρ(ω)=ρ(ξ)=σ(η)\rho(\omega) = \rho(\xi) = \sigma(\eta) (resp. σ(ω)=ρ(ξ)=σ(η)\sigma(\omega) = \rho(\xi) = \sigma(\eta)).

For each ω∈Ω′\omega \in \Omega', let ii be the index such that ρ(ω)∈Li\rho(\omega) \in L_i; then consider the family (vλ)λ∈Li(v_\lambda)_{\lambda \in L_i} where vρ(ω)=pωv_{\rho(\omega)} = p_\omega and vλ=1Gλv_\lambda = 1_{G_\lambda} for λ≠ρ(ω)\lambda \neq \rho(\omega); hypothesis (P) implies that the family (vλ)(v_\lambda) satisfies conditions (27) (where p′p' and pp must be replaced by p(i)p^{(i)}, q′q' and qq by q(i)q^{(i)}, ω\omega by an element ξ\xi running through Ωi\Omega_i); thus an endomorphism ⨂(c(i))vλ=rω\bigotimes_{(c^{(i)})} v_\lambda = r_\omega is derived of the Z\mathbf{Z}-module FiF_i. Similarly, an endomorphism sωs_\omega is defined of the Z\mathbf{Z}-module FjF_j starting with qωq_\omega, jj being the index such that σ(ω)∈Lj\sigma(\omega) \in L_j; finally let d(ω)=(i,j)d(\omega) = (i, j). Then we can define the tensor product ⨂(d,r,s)Fi\bigotimes_{(d, r, s)} F_i and the corresponding canonical mapping

ϕ(d,r,s):∏i=1nFi→⨂(d,r,s)Fi. \phi_{(d, r, s)} : \prod_{i=1}^n F_i \to \bigotimes_{(d, r, s)} F_i.

On the other hand, for each ii, the canonical mapping

ψi=ϕ(c(i),p(i),q(i)):∏λ∈LiGλ→Fi; \psi_i = \phi_{(c^{(i)}, p^{(i)}, q^{(i)})} : \prod_{\lambda \in L_i} G_\lambda \to F_i;

using the associativity of the product of sets, a Z\mathbf{Z}-multilinear mapping ψ=ϕ(d,r,s)∘(ψi)\psi = \phi_{(d, r, s)} \circ (\psi_i) of GG into ⨂(d,r,s)Fi\bigotimes_{(d, r, s)} F_i is derived. We show that the ordered pair (⨂(d,r,s)Fi,ψ)\left( \bigotimes_{(d, r, s)} F_i, \psi \right) is a solution of the same universal problem as (⨂(c,p,q)Gλ,ϕ(c,p,q))\left( \bigotimes_{(c, p, q)} G_\lambda, \phi_{(c, p, q)} \right), whence will follow the existence of a unique Z\mathbf{Z}-module isomorphism

θ:⨂(c,p,q)Gλ→⨂(d,r,s)Fi \theta : \bigotimes_{(c, p, q)} G_\lambda \to \bigotimes_{(d, r, s)} F_i

such that ψ=θ∘ϕ(c,p,q)\psi = \theta \circ \phi_{(c, p, q)} (Set Theory, IV, § 3, no. 1).

By induction on nn, the proof is reduced to the case n=2n = 2; for simplicity we write F1⊗(d)F2F_1 \otimes_{(d)} F_2 and y1⊗(d)y2y_1 \otimes_{(d)} y_2 instead of ⨂(d,r,s)Fi\bigotimes_{(d, r, s)} F_i and ⨂(d,r,s)yi\bigotimes_{(d, r, s)} y_i. Consider the mapping from GG to F1⊗(d)F2F_1 \otimes_{(d)} F_2

h:(xλ)→(⨂(c(1))xλ)⊗(d)(⨂(c(2))xλ). h : (x_\lambda) \to \left( \bigotimes_{(c^{(1)})} x_\lambda \right) \otimes_{(d)} \left( \bigotimes_{(c^{(2)})} x_\lambda \right).

It is obviously Z\mathbf{Z}-multilinear; we show that it satisfies conditions (26) for all ω∈Ω\omega \in \Omega. This is obvious if ω∈Ω1\omega \in \Omega_1 or ω∈Ω2\omega \in \Omega_2; otherwise, supposing, in order to fix the ideas, that ρ(ω)∈L1\rho(\omega) \in L_1 and σ(ω)∈L2\sigma(\omega) \in L_2, the values of h∘pˉωh \circ \bar{p}_\omega and h∘qˉωh \circ \bar{q}_\omega at (xλ)(x_\lambda) are respectively

(rω(⨂(c(1))xλ))⊗(d)(⨂(c(2))xλ)and(⨂(c(1))xλ)⊗(d)(sω(⨂(c(2))xλ)) \left( r_\omega \left( \bigotimes_{(c^{(1)})} x_\lambda \right) \right) \otimes_{(d)} \left( \bigotimes_{(c^{(2)})} x_\lambda \right) \quad \text{and} \quad \left( \bigotimes_{(c^{(1)})} x_\lambda \right) \otimes_{(d)} \left( s_\omega \left( \bigotimes_{(c^{(2)})} x_\lambda \right) \right)

which are also equal by definition of F1⊗(d)F2F_1 \otimes_{(d)} F_2.

This being so, let uu be a Z\mathbf{Z}-multilinear mapping of GG into a Z\mathbf{Z}-module HH, satisfying conditions (26); we shall define a Z\mathbf{Z}-linear mapping v:F1⊗(d)F2→Hv : F_1 \otimes_{(d)} F_2 \to H such that u=v∘hu = v \circ h and that will prove our assertion (repeating the argument of no. 1, Corollary 1 to Proposition 1). For all z2=(xλ)λ∈L2z_2 = (x_\lambda)_{\lambda \in L_2} we consider the "partial" linear mapping of ∏λ∈L1Gλ\prod_{\lambda \in L_1} G_\lambda into HH

u(.,z2):(xλ)λ∈L1↦u((xλ)λ∈L1,z2)=u((xλ)λ∈L). u(., z_2): (x_\lambda)_{\lambda \in L_1} \mapsto u((x_\lambda)_{\lambda \in L_1}, z_2) = u((x_\lambda)_{\lambda \in L}).

Clearly it is Z\mathbf{Z}-multilinear and satisfies conditions (26) for ω∈Ω1\omega \in \Omega_1; by definition there thus exists a Z\mathbf{Z}-linear mapping y1↦w1(y1,x2)y_1 \mapsto w_1(y_1, x_2) of F1F_1 into HH such that

w1(⨂(c(1))xλ,z2)=u((xλ)λ∈L1,z2) w_1\left( \bigotimes_{(c^{(1)})} x_\lambda, z_2 \right) = u((x_\lambda)_{\lambda \in L_1}, z_2)

We next consider the mapping

u2:(xλ)λ∈L2↦w1(.,(xλ)λ∈L2) u_2: (x_\lambda)_{\lambda \in L_2} \mapsto w_1(., (x_\lambda)_{\lambda \in L_2})

of ∏λ∈L2Gλ\prod_{\lambda \in L_2} G_\lambda into HomZ(F1,H)\mathrm{Hom}_\mathbf{Z}(F_1, H); it is obviously Z\mathbf{Z}-multilinear and satisfies conditions (26) for ω∈Ω2\omega \in \Omega_2, by virtue of the hypothesis on uu and relations (28) and (29) and taking account of the fact that the elements of the form ⨂(c(1))xλ\bigotimes_{(c^{(1)})} x_\lambda generate the Z\mathbf{Z}-module F1F_1. Hence there is a Z\mathbf{Z}-linear mapping

w2:F2→HomZ(F1,H) w_2: F_2 \to \mathrm{Hom}_\mathbf{Z}(F_1, H)

such that

w2(⨂(c(2))xλ)=u2((xλ)λ∈L2) w_2\left( \bigotimes_{(c^{(2)})} x_\lambda \right) = u_2((x_\lambda)_{\lambda \in L_2})

or also

(w2(⨂(c(2))xλ))(⨂(c(1))xλ)=u((xλ)λ∈L). \left( w_2\left( \bigotimes_{(c^{(2)})} x_\lambda \right) \right)\left( \bigotimes_{(c^{(1)})} x_\lambda \right) = u((x_\lambda)_{\lambda \in L}).

We now consider, for y1∈F1,y2∈F2y_1 \in F_1, y_2 \in F_2, the element of HH

w(y1,y2)=(w2(y2))(y1). w(y_1, y_2) = (w_2(y_2))(y_1).

Clearly ww is a Z\mathbf{Z}-bilinear mapping of F1×F2F_1 \times F_2 into HH. We show further that, for all ω∈Ω′\omega \in \Omega', (assuming, to fix the ideas, that ρ(ω)∈L1\rho(\omega) \in L_1 and σ(ω)∈L2\sigma(\omega) \in L_2)

w(rω(y1),y2)=w(y1,sω(y2)). w(r_\omega(y_1), y_2) = w(y_1, s_\omega(y_2)).

It suffices to verify this relation when y1y_1 (resp. y2y_2) is of the form ⨂(c(1))xλ\bigotimes_{(c^{(1)})} x_\lambda (resp. ⨂(c(2))xλ\bigotimes_{(c^{(2)})} x_\lambda), since these elements generate the Z\mathbf{Z}-module F1F_1 (resp. F2F_2). But by definition, rω(⨂(c(1))xλ)=⨂(c(1))xλ′r_\omega\left( \bigotimes_{(c^{(1)})} x_\lambda \right) = \bigotimes_{(c^{(1)})} x'_\lambda, where xρ(ω)′=pω(xρ(ω))x'_{\rho(\omega)} = p_\omega(x_{\rho(\omega)}) and xλ′=xλx'_\lambda = x_\lambda for λ ≠ ρ(ω) in L₁; similarly sω(⨂(c(2))xλ)=⨂(c(2))xλ′′s_\omega\left( \bigotimes_{(c^{(2)})} x_\lambda \right) = \bigotimes_{(c^{(2)})} x''_\lambda, where xσ(ω)′′=qω(xσ(ω)′)x''_{\sigma(\omega)} = q_\omega(x'_{\sigma(\omega)}) and xλ′′=xλx''_\lambda = x_\lambda for λ ≠ σ(ω) in L₂; using (30) and (31), relation (32) then follows from (26). Hence there exists a Z\mathbf{Z}-linear mapping v of F₁ ⊗ F₂ into H such that v(y1⊗y2)=w(y1,y2)v(y_1 \otimes y_2) = w(y_1, y_2) and it then follows from (30) and (31) that v∘h=uv \circ h = u.

The most important special case of the general tensor product defined above is the following: we start with a family (Ai)1≤i≤n−1(A_i)_{1 \leq i \leq n-1} of rings and a family (Ei)1≤i≤n(E_i)_{1 \leq i \leq n}, where E1E_1 is a right A1A_1-module, EnE_n is a left An−1A_{n-1} module and for 2≤i≤n−12 \leq i \leq n-1, EiE_i is an (Ai−1,Ai)(A_{i-1}, A_i)-bimodule. Then the above definition is applied as follows: L is the set [1,n][1, n], Gi=EiG_i = E_i, Ω is the set the sum of the AiA_i (1≤i≤n−11 \leq i \leq n-1). For ω∈Ai\omega \in A_i (1≤i≤n−11 \leq i \leq n-1), take ρ(ω)=i\rho(\omega) = i, σ(ω)=i+1\sigma(\omega) = i + 1, pωp_\omega is the endomorphism x↦xωx \mapsto x\omega of the ZZ-module EiE_i and qωq_\omega the endomorphism y↦ωyy \mapsto \omega y of the ZZ-module Ei+1E_{i+1}; the corresponding tensor product is denoted by

E1⊗A1E2⊗A2E3⊗⋯⊗An−2En−1⊗An−1En E_1 \otimes_{A_1} E_2 \otimes_{A_2} E_3 \otimes \cdots \otimes_{A_{n-2}} E_{n-1} \otimes_{A_{n-1}} E_n

(a notation where the AiA_i may occasionally be suppressed) and the elements ⨂(c,p,q)xi\bigotimes_{(c, p, q)} x_i of this tensor product, for a family (xi)(x_i) such that xi∈Eix_i \in E_i for 1≤i≤n1 \leq i \leq n, may be written x1⊗x2⊗⋯⊗xnx_1 \otimes x_2 \otimes \cdots \otimes x_n if no confusion can arise; an analogous notation is used for a ZZ-linear mapping ⨂(c)vi\bigotimes_{(c)} v_i. Hypothesis (P) holds for every partition of [1,n][1, n], since the EiE_i are bimodules for 2≤i≤n−12 \leq i \leq n-1. When n=3n = 3, we have thus defined the ZZ-module E⊗AF⊗BGE \otimes_A F \otimes_B G alluded to in no. 8, and recovered Proposition 8 (no. 8).

When each of the EiE_i is a multimodule (with, for 2≤i≤n−12 \leq i \leq n-1, Ai−1A_{i-1} one of the rings operating on the left and AiA_i one of the rings operating on the right and analogous conditions for i=1i = 1 and i=ni = n), as in no. 4, a multimodule structure is defined on E1⊗A1E2⊗⋯⊗An−1EnE_1 \otimes_{A_1} E_2 \otimes \cdots \otimes_{A_{n-1}} E_n with respect to all the rings except the AiA_i which operate on the EiE_i (1≤i≤n1 \leq i \leq n).

In particular, let C be a commutative ring, (Ei)1≤i≤n(E_i)_{1 \leq i \leq n} a family of C-modules. By giving E1E_1 and EnE_n two C-module structures identical with the given structure and EiE_i for 2≤i≤n−12 \leq i \leq n-1 three C-module structures identical with the given structure, we define on the tensor product

E1⊗CE2⊗CE3⊗⋯⊗CEn−1⊗CEn E_1 \otimes_C E_2 \otimes_C E_3 \otimes \cdots \otimes_C E_{n-1} \otimes_C E_n

nn C-modules structures which are compatible with one another and which are in fact identical, since for all γ∈C\gamma \in C and (xi)∈∏i=1nEi(x_i) \in \prod_{i=1}^n E_i, by definition

(γx1)⊗x2⊗⋯⊗xn=x1⊗(γx2)⊗⋯⊗xn=⋯=x1⊗x2⊗⋯⊗(γxn). (\gamma x_1) \otimes x_2 \otimes \cdots \otimes x_n \\ = x_1 \otimes (\gamma x_2) \otimes \cdots \otimes x_n = \cdots = x_1 \otimes x_2 \otimes \cdots \otimes (\gamma x_n).

When we speak of the tensor product (34) as a C-module, we shall always mean with this structure, unless otherwise mentioned, and the tensor product (34) is also denoted by ⨂1≤i≤nEi\bigotimes_{1 \leq i \leq n} E_i if no confusion can arise. For every C-module G, the Z\mathbf{Z}-multilinear mappings of ∏i=1nEi\prod_{i=1}^n E_i into G which, for every index i, satisfy the relation

f(x1,…,xi−1,γxi,xi+1,…,xn)=γf(x1,…,xn) f(x_1, \ldots, x_{i-1}, \gamma x_i, x_{i+1}, \ldots, x_n) = \gamma f(x_1, \ldots, x_n)

for γ∈C\gamma \in \mathbf{C} and (xi)∈∏iEi(x_i) \in \prod_i E_i are then called C-multilinear and form a C-module denoted by Ln(E1,…,En;G)\mathscr{L}_n(E_1, \ldots, E_n; G); the universal property of the tensor product (34) then allows us to define a canonical C-module isomorphism

Ln(E1,…,En;G)→Hom⁡C(E1⊗CE2⊗⋯⊗CEn,G) \mathscr{L}_n(E_1, \ldots, E_n; G) \to \operatorname{Hom}_\mathbf{C}(E_1 \otimes_\mathbf{C} E_2 \otimes \cdots \otimes_\mathbf{C} E_n, G)

which associates with every C-multilinear mapping f the C-linear mapping g such that

f(x1,…,xn)=g(x1⊗x2⊗⋯⊗xn). f(x_1, \ldots, x_n) = g(x_1 \otimes x_2 \otimes \cdots \otimes x_n).

A C-multilinear mapping of E1×⋯×EnE_1 \times \cdots \times E_n into C\mathbf{C} is also called an nn-linear form.

Let (Fi)1≤i≤n(F_i)_{1 \leq i \leq n} be another family of C-modules; for every system of n C-linear mappings ui:Ei→Fi,u1⊗u2⊗⋯⊗unu_i : E_i \to F_i, u_1 \otimes u_2 \otimes \cdots \otimes u_n (also denoted by ⨂1≤i≤nui\bigotimes_{1 \leq i \leq n} u_i) is a C-linear mapping of

E1⊗CE2⊗⋯⊗CEn into F1⊗CF2⊗⋯⊗CFn. E_1 \otimes_\mathbf{C} E_2 \otimes \cdots \otimes_\mathbf{C} E_n \text{ into } F_1 \otimes_\mathbf{C} F_2 \otimes \cdots \otimes_\mathbf{C} F_n.

Further, (u1,…,un)↦u1⊗u2⊗⋯⊗un(u_1, \ldots, u_n) \mapsto u_1 \otimes u_2 \otimes \cdots \otimes u_n is a C-multilinear mapping of ∏iHom⁡C(Ei,Fi)\prod_i \operatorname{Hom}_\mathbf{C}(E_i, F_i) into

Hom⁡C(E1⊗CE⊗⋯⊗CEn,F1⊗CF2⊗⋯⊗CFn). \operatorname{Hom}_\mathbf{C}(E_1 \otimes_\mathbf{C} E \otimes \cdots \otimes_\mathbf{C} E_n, F_1 \otimes_\mathbf{C} F_2 \otimes \cdots \otimes_\mathbf{C} F_n).

Hence there corresponds canonically to the latter mapping a C-linear mapping called canonical

Hom⁡C(E1,F1)⊗CHom⁡C(E2,F2)⊗⋯⊗CHom⁡C(En,Fn) \operatorname{Hom}_\mathbf{C}(E_1, F_1) \otimes_\mathbf{C} \operatorname{Hom}_\mathbf{C}(E_2, F_2) \otimes \cdots \otimes_\mathbf{C} \operatorname{Hom}_\mathbf{C}(E_n, F_n)
→Hom⁡C(E1⊗CE2⊗⋯⊗CEn,F1⊗CF2⊗⋯⊗CFn) \to \operatorname{Hom}_\mathbf{C}(E_1 \otimes_\mathbf{C} E_2 \otimes \cdots \otimes_\mathbf{C} E_n, F_1 \otimes_\mathbf{C} F_2 \otimes \cdots \otimes_\mathbf{C} F_n)

generalizing that defined in no. 5 for n=2n = 2.

The general associativity property seen earlier can be specialized here as follows. Given a partition (Jk)1≤k≤m(J_k)_{1 \leq k \leq m} of the interval {1,n}\{1, n\} of N\mathbf{N}, let, for each kk, FkF_k be the tensor product Ei1⊗CEi2⊗⋯⊗CEirE_{i_1} \otimes_\mathbf{C} E_{i_2} \otimes \cdots \otimes_\mathbf{C} E_{i_r}, where (i1,…,ir)(i_1, \ldots, i_r) is the strictly increasing sequence of elements of JkJ_k. Then there is a canonical C-module isomorphism (called "associativity isomorphism")

F1⊗CF2⊗⋯⊗CFm→E1⊗CE2⊗⋯⊗CEn F_1 \otimes_C F_2 \otimes \cdots \otimes_C F_m \to E_1 \otimes_C E_2 \otimes \cdots \otimes_C E_n

which, in the above notation, maps the tensor product

y1⊗y2⊗⋯⊗ym,where yk=xi1⊗xi2⊗⋯⊗xir, y_1 \otimes y_2 \otimes \cdots \otimes y_m, \quad \text{where } y_k = x_{i_1} \otimes x_{i_2} \otimes \cdots \otimes x_{i_r},

to the tensor product x1⊗x2⊗⋯⊗xnx_1 \otimes x_2 \otimes \cdots \otimes x_n (where xi∈Eix_i \in E_i for all ii).

In particular, if π\pi is a permutation of {1,n}\{1, n\}, writing Jk={π(k)}J_k = \{\pi(k)\} for 1≤k≤n1 \leq k \leq n, we obtain a canonical ("commutativity") isomorphism

Eπ(1)⊗CEπ(2)⊗⋯⊗CEπ(n)→E1⊗CE2⊗⋯⊗CEn E_{\pi(1)} \otimes_C E_{\pi(2)} \otimes \cdots \otimes_C E_{\pi(n)} \to E_1 \otimes_C E_2 \otimes \cdots \otimes_C E_n

which maps xπ(1)⊗xπ(2)⊗⋯⊗xπ(n)x_{\pi(1)} \otimes x_{\pi(2)} \otimes \cdots \otimes x_{\pi(n)} to x1⊗x2⊗⋯⊗xnx_1 \otimes x_2 \otimes \cdots \otimes x_n. We shall often identify the various tensor products which correspond to one another under these canonical isomorphisms.

For 1≤i≤n1 \leq i \leq n, suppose that EiE_i admits a basis (bλi(i))λi∈Li(b^{(i)}_{\lambda_i})_{\lambda_i \in L_i}; by induction on nn, it follows from no. 7, Corollary 2 to Proposition 7 that the family (bλ1(1)⊗bλ2(2)⊗⋯⊗bλn(n))(b^{(1)}_{\lambda_1} \otimes b^{(2)}_{\lambda_2} \otimes \cdots \otimes b^{(n)}_{\lambda_n}), where (λ1,…,λn)(\lambda_1, \ldots, \lambda_n) runs through ∏1≤i≤nLi\prod_{1 \leq i \leq n} L_i, is a basis of ⨁1≤i≤nEi\bigoplus_{1 \leq i \leq n} E_i, sometimes called the tensor product of the bases (bλ1(1))(b^{(1)}_{\lambda_1}) in question.

Remark

(1) The above remarks concerning the case of modules over a commutative ring generalize as in no. 5, Remark 2 when there is a tensor product E1⊗A1E2⊗⋯⊗An−1EnE_1 \otimes_{A_1} E_2 \otimes \cdots \otimes_{A_{n-1}} E_n where the rings AiA_i are not necessarily commutative and where, for each ii, there is a homomorphism ρi:C→A\rho_i : C \to A of the same commutative ring CC such that: (1) ρi(C)\rho_i(C) is contained in the centre of AiA_i; (2) for 2≤i≤n−12 \leq i \leq n - 1, the CC-module structures on EiE_i obtained using the homomorphisms ρi−1\rho_{i-1} and ρi\rho_i coincide. Then a CC-*module* structure is obtained on E1⊗A1E2⊗⋯⊗An−1EnE_1 \otimes_{A_1} E_2 \otimes \cdots \otimes_{A_{n-1}} E_n and canonical mappings analogous to (13) (no. 5), which will be left to the reader to describe.

(2) Let A,BA, B be two rings, EE a right AA-module, E′E' a left AA-module, FF a right BB-module and F′F' a left BB-module. The Z-bilinear\mathbf{Z}\text{-bilinear} mappings of (E⊗AE′)×(F⊗BF′)(E \otimes_A E') \times (F \otimes_B F') into a Z-module\mathbf{Z}\text{-module} GG are then in one-to-one correspondence with the Z-multilinear\mathbf{Z}\text{-multilinear} mappings ff of E×E′×F×F′E \times E' \times F \times F' into GG satisfying the conditions

{f(xλ,x′,y,y′)=f(x,λx′,y,y′)f(x,x′,yμ,y′)=f(x,x′,y,μy′) \begin{cases} f(x \lambda, x', y, y') = f(x, \lambda x', y, y') \\ f(x, x', y \mu, y') = f(x, x', y, \mu y') \end{cases}

for λ∈A,μ∈B,x∈E,x′∈E′,y∈F,y′∈F′\lambda \in A, \mu \in B, x \in E, x' \in E', y \in F, y' \in F'. The general constructions given in this no. reduce the proof of this to defining a canonical Z-module\mathbf{Z}\text{-module} isomorphism between (E⊗AE′)⊗Z(F⊗BF′)(E \otimes_A E') \otimes_{\mathbf{Z}} (F \otimes_B F') and E⊗AE′⊗ZF⊗BF′E \otimes_A E' \otimes_{\mathbf{Z}} F \otimes_B F', which follows from the associativity property of tensor products of the form (33).

Exercises

See the exercises for § 3.

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