Bourbaki

Example

Tag 067J, permanent. Cite it as [Bourbaki, Tag 067J].

Algebra, chapter III, § 11. Cogebras, products of multilinear forms, inner products and duality

(1) The canonical isomorphism A→A⊗AAA \to A \otimes_A A (II, § 3, no. 5) defines an A-cogebra structure on A.

(2) Let E be a cogebra, cc its coproduct and σ\sigma the canonical automorphism of the A-module E⊗AEE \otimes_A E such that σ(x⊗y)=y⊗x\sigma(x \otimes y) = y \otimes x for x∈E,y∈Ex \in E, y \in E; the A-linear mapping σ∘c\sigma \circ c defines a new cogebra structure on E. With this structure E is called the opposite cogebra to the given cogebra E.

(3) Let B be an A-*algebra* and let m:B⊗AB→Bm : B \otimes_A B \to B be the A-linear mapping defining multiplication on B (§ 1, no. 3). The transpose tm^t m is then an A-linear mapping of the dual B∗B^* of the A-module B into the dual (B⊗AB)∗(B \otimes_A B)^* of the A-module B⊗ABB \otimes_A B. If also B is a finitely generated projective A-module, the canonical mapping μ:B∗⊗AB∗→(B⊗AB)∗\mu : B^* \otimes_A B^* \to (B \otimes_A B)^* is an A-module isomorphism (II, § 4, no. 4); the mapping c=μ−1∘tmc = \mu^{-1} \circ ^t m is then a coproduct defining a cogebra structure on the dual B∗B^* of the A-module B.

(4) Let X be a set, A(X)A^{(X)} the A-module of formal linear combinations of elements of X with coefficients in A (II, § 1, no. 11) and (ex)x∈X(e_x)_{x \in X} the canonical basis of A(X)A^{(X)}. An A-linear mapping c:A(X)→A(X)⊗AA(X)c : A^{(X)} \to A^{(X)} \otimes_A A^{(X)} is defined by the condition c(ex)=ex⊗exc(e_x) = e_x \otimes e_x and a canonical cogebra structure is thus obtained on A(X)A^{(X)}.

(5) Let M be an A-module and T(M)T(M) the tensor algebra of M (§ 5, no. 1); by II, § 3, no. 9 there exists one and only one A-linear mapping cc of the A-module T(M)T(M) into the A-module T(M)⊗AT(M)T(M) \otimes_A T(M) such that, for all n≥0n \geq 0,

c(x1x2…xn)=∑0≤p≤n(x1x2…xp)⊗(xp+1…xn) c(x_1 x_2 \ldots x_n) = \sum_{0 \leq p \leq n} (x_1 x_2 \ldots x_p) \otimes (x_{p+1} \ldots x_n)

for all xi∈Mx_i \in M (x1x2…xnx_1 x_2 \ldots x_n denotes the product in the algebra T(M)T(M)). Thus T(M)T(M) is given a cogebra structure.

(6) Let M be an A-module and S(M)S(M) the symmetric algebra of M (§ 6, no. 1); the diagonal mapping Δ:x↦(x,x)\Delta : x \mapsto (x, x) of M into M×MM \times M is an A-linear mapping to which there therefore corresponds a homomorphism S(Δ)S(\Delta) of the A-algebra S(M)S(M) into the A-algebra S(M×M)S(M \times M) (§ 6, no. 2, Proposition 3). On the other hand, in § 6, no. 6 we defined a canonical graded algebra isomorphism h:S(M×M)→S(M)⊗AS(M)h : S(M \times M) \to S(M) \otimes_A S(M); by composition we therefore obtain an A-algebra homomorphism

c=h∘S(Δ):S(M)→S(M)⊗AS(M), c = h \circ S(\Delta) : S(M) \to S(M) \otimes_A S(M),

thus defining on S(M)S(M) a cogebra structure. For all x∈Mx \in M, by definition S(Δ)(x)=(x,x)S(\Delta)(x) = (x, x) and the definition of hh given in § 6, no. 6 shows that

h((x,x))=x⊗1+1⊗x. h((x, x)) = x \otimes 1 + 1 \otimes x.

It follows that cc is the unique algebra homomorphism such that, for all x∈Mx \in M,

(4)

c(x)=x⊗1+1⊗x. c(x) = x \otimes 1 + 1 \otimes x.

As cc is an algebra homomorphism, it follows that, for every sequence (xi)1≤i≤n(x_i)_{1 \leq i \leq n} of nn elements of MM,

(5)

c(x1x2…xn)=∏i=1n(xi⊗1+1⊗xi) c(x_1 x_2 \ldots x_n) = \prod_{i=1}^n (x_i \otimes 1 + 1 \otimes x_i)
=∑(xi1…xip)⊗(xj1…xjn−p) = \sum (x_{i_1} \ldots x_{i_p}) \otimes (x_{j_1} \ldots x_{j_{n-p}})

the summation on the right hand side of (5) being taken over all ordered pairs of strictly increasing sequences (in some cases empty)

i1<i2<⋯<ip,j1<j2<⋯<jn−p i_1 < i_2 < \cdots < i_p, \quad j_1 < j_2 < \cdots < j_{n-p}

of elements of {1,n}\{1, n\}, whose sets of elements are complementary. The element c(x1x2…xn)c(x_1 x_2 \ldots x_n) is an element of total degree nn in S(M)⊗AS(M)S(M) \otimes_A S(M) and its component of bidegree (p,n−p)(p, n-p) is

(6)

∑σ(xσ(1)…xσ(p))⊗(xσ(p+1)…xσ(n)) \sum_{\sigma} (x_{\sigma(1)} \ldots x_{\sigma(p)}) \otimes (x_{\sigma(p+1)} \ldots x_{\sigma(n)})

where the summation is taken over all permutations σ∈Sn\sigma \in S_n which are increasing in each of the intervals {1,p}\{1, p\} and {p+1,n}\{p+1, n\}.

(7) Let MM be an AA-module and proceed with the exterior algebra Λ(M)\Lambda(M) as with S(M)S(M) in Example 6; the diagonal mapping Δ:M→M×M\Delta : M \to M \times M this time defines a homomorphism Λ(Δ)\Lambda(\Delta) of the AA-algebra Λ(M)\Lambda(M) into the AA-algebra Λ(M×M)\Lambda(M \times M) (§ 7, no. 2, Proposition 2); on the other hand there is a canonical graded algebra isomorphism

h:Λ(M×M)→Λ(M)g⊗AΛ(M) h : \Lambda(M \times M) \to \Lambda(M)^g \otimes_A \Lambda(M)

(§ 7, no. 7, Proposition 10), whence by composition there is an algebra homomorphism c=h∘Λ(Δ):Λ(M)→Λ(M)g⊗AΛ(M)c = h \circ \Lambda(\Delta) : \Lambda(M) \to \Lambda(M)^g \otimes_A \Lambda(M), which can be considered as an AA-module homomorphism Λ(M)→Λ(M)⊗AΛ(M)\Lambda(M) \to \Lambda(M) \otimes_A \Lambda(M) and which therefore defines on Λ(M)\Lambda(M) a cogebra structure. It can be proved as in Example 6 that cc is the unique algebra homomorphism such that, for all x∈Mx \in M,

(7)

c(x)=x⊗1+1⊗x, c(x) = x \otimes 1 + 1 \otimes x,

whence, for every sequence (xi)1≤i≤n(x_i)_{1 \leq i \leq n} of elements of MM,

c(x1∧x2∧⋯∧xn)=(x1⊗1+1⊗x1)∧…∧(xn⊗1+1⊗xn) c(x_1 \wedge x_2 \wedge \cdots \wedge x_n) = (x_1 \otimes 1 + 1 \otimes x_1) \wedge \ldots \wedge (x_n \otimes 1 + 1 \otimes x_n)
where the product on the right hand side is taken in the algebra
⋀(M)g⊗A⋀(M); \bigwedge(M)^{\mathfrak{g}} \otimes_A \bigwedge(M);
to calculate this product, consider, for every ordered pair of strictly increasing sequences i1<i2<⋯<ip, j1<j2<⋯<jn−pi_1 < i_2 < \cdots < i_p,\ j_1 < j_2 < \cdots < j_{n-p} of elements of [1,n][1, n], whose sets of elements are complementary, the product y1y2…yny_1 y_2 \ldots y_n, where yih=xih⊗1y_{i_h} = x_{i_h} \otimes 1 (1≤h≤p1 \leq h \leq p) and yjk=1⊗xjky_{j_k} = 1 \otimes x_{j_k} (1≤k≤n−p1 \leq k \leq n - p) and the sum is taken over all these products. As the graded algebra ⋀(M)g⊗A⋀(M)\bigwedge(M)^{\mathfrak{g}} \otimes_A \bigwedge(M) is anticommutative and the elements xi⊗1x_i \otimes 1 and 1⊗xi1 \otimes x_i are of total degree 1, by § 4, no. 6, Lemma 3 and Lemma 1,
(8)c(x1∧x2∧⋯∧xn)(8)\quad c(x_1 \wedge x_2 \wedge \cdots \wedge x_n)
=∑(−1)v(xi1∧⋯∧xip)⊗(xj1∧⋯∧xjn−p) = \sum (-1)^v (x_{i_1} \wedge \cdots \wedge x_{i_p}) \otimes (x_{j_1} \wedge \cdots \wedge x_{j_{n-p}})
vv being the number of ordered pairs (h,k)(h, k) such that jk<ihj_k < i_h and the summation being taken over the same set as in (5). The element c(x1∧⋯∧xn)c(x_1 \wedge \cdots \wedge x_n) is of total degree nn in ⋀(M)g⊗A⋀(M)\bigwedge(M)^{\mathfrak{g}} \otimes_A \bigwedge(M) and its homogeneous component of bi-degree (p,n−p)(p, n - p) is equal to
(9)∑σεσ(xσ(1)∧⋯∧xσ(p))⊗(xσ(p+1)∧⋯∧xσ(n))(9)\quad \sum_{\sigma} \varepsilon_\sigma (x_{\sigma(1)} \wedge \cdots \wedge x_{\sigma(p)}) \otimes (x_{\sigma(p+1)} \wedge \cdots \wedge x_{\sigma(n)})
the summation being taken over permutations σ∈Sn\sigma \in S_n which are increasing in each of the intervals [1,p][1, p] and [p+1,n][p + 1, n].

When in future we speak of A(x),T(M),S(M)A^{(x)}, T(M), S(M) or ⋀(M)\bigwedge(M) as cogebras, we shall mean, unless otherwise mentioned, with the cogebra structures defined in Examples 4, 5, 6 and 7 respectively.

(8) Let E,FE, F be two AA-cogebras and c,c′c, c' their respective coproducts. Let τ:(E⊗AE)⊗A(F⊗AF)→(E⊗AF)⊗A(E⊗AF)\tau : (E \otimes_A E) \otimes_A (F \otimes_A F) \to (E \otimes_A F) \otimes_A (E \otimes_A F) denote the associativity isomorphism such that τ((x⊗x′)⊗(y⊗y′))=(x⊗y)⊗(x′⊗y′)\tau((x \otimes x') \otimes (y \otimes y')) = (x \otimes y) \otimes (x' \otimes y') for x,x′x, x' in EE and y,y′y, y' in FF. Then the composite linear mapping

E⊗AF→c⊗c′(E⊗AE)⊗A(F⊗AF)→τ(E⊗AF)⊗A(E⊗AF) E \otimes_A F \xrightarrow{c \otimes c'} (E \otimes_A E) \otimes_A (F \otimes_A F) \xrightarrow{\tau} (E \otimes_A F) \otimes_A (E \otimes_A F)
defines a cogebra structure on the AA-module E⊗AFE \otimes_A F, called the tensor product of the cogebras EE and FF.

Let EE be a cogebra and Δ\Delta a commutative monoid. A graduation (Eλ)λ∈Δ(E_\lambda)_{\lambda \in \Delta} on the AA-module EE is said to be compatible with the coproduct cc of EE if cc is a graded homomorphism of degree OO of the graded AA-module EE into the graded AA-module (of type Δ\Delta) E⊗AEE \otimes_A E, in other words (II, § 11, no. 5) if

(10)c(Eλ)⊂∑μ+ν=λEμ⊗AEν.(10)\quad c(E_\lambda) \subset \sum_{\mu + \nu = \lambda} E_\mu \otimes_A E_\nu.

In what follows, we shall most often limit our attention to graduations of type N\mathbf{N} compatible with the coproduct; a cogebrà with such a graduation will also be called a graded cogebrà. If F is another graded cogebrà, a graded cogebrà morphism ϕ:E→F\phi : E \to F is by definition a cogebrà morphism (Definition 2) which is also a graded homomorphism of degree 0 of graded A-modules.

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

Cited by

Cites

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