(i) In the associative algebra E∗=HomA(E,A), the product of n elements u1,u2,…,un of degree 1 is given by
u1u2…un=mn∘(u1⊗u2⊗⋯⊗un)∘cn. (ii)
Suppose also that the cogebra E is graded. Then, in the graded associative algebra E∗gr=HomgrA(E,A), the product of n elements u1,u2,…,un of degree 1 is given by u1u2…un=mn∘(u1⊗u2⊗⋯⊗un)∘δn where δn:E→E⊗n is the linear mapping which maps each x∈E to the component of cn(x) of multidegree (1,1,…,1).
Formula (26) is just the definition of the product in E∗ for n=2; to prove it by induction on n, observe that
u1u2…un =m∘((u1u2…un−1)⊗un)∘c =m∘((mn−1∘(u1⊗u2⊗⋯⊗un−1)∘cn−1)⊗un)∘c =m∘(mn−1⊗1A)∘(u1⊗u2⊗⋯⊗un−1⊗un)∘(cn−1⊗1E)∘c =mn∘(u1⊗u2⊗⋯⊗un)∘cn by virtue of (24), (25), II, § 3, no. 3, formula (5) and the relation
un=1A∘un∘1E∗.
When E is graded and the elements ui∈E∗gr homogeneous of degree 1, then by definition for homogeneous elements xi∈E,
(u1⊗u2⊗⋯⊗un)(x1⊗x2⊗⋯⊗xn)=0
unless all the xi are of degree 1, whence formula (27).
It follows from formulae (3), (5) and (7) of no. 1 and formula (24) that when E is taken to be one of the three graded cogebras T(M), S(M) and Λ(M), we obtain respectively by induction on n (using the fact that the coproduct is a graded homomorphism of degree 0), for x1,x2,…,xn in M:
when E=T(M),
δn(x1x2⋯xn)=x1⊗x2⊗⋯⊗xn
when E=S(M),
δn(x1x2⋯xn)=σ∈Sn∑xσ(1)⊗xσ(2)⊗⋯⊗xσ(n)
when E=Λ(M),
δn(x1x2⋯xn)=σ∈Sn∑εσ⋅xσ(1)⊗xσ(2)⊗⋯⊗xσ(n).
It suffices to note, for example when E=Λ(M), that in the expression
cn(x1x2⋯xn)=(cn−1⊗1E)(∑(−1)v(xi1…xip)⊗(xj1…xjn−p)) arising from formula (8) of no. 1, the only terms which can give a term of multidegree
(1,1,…,1) are those for which
n−p=1 and hence
δn(x1x2⋯xn) is the term of multidegree
(1,1,…,1) in the sum
i=1∑n(−1)n−icn−1(x1…xi−1xi+1…xn)⊗xi and this term is necessarily equal to
i=1∑n(−1)n−iδn−1(x1…xi−1xi+1…xn)⊗xi, whence the result by the induction hypothesis.
Using Lemma 1, the product in T(M)∗gr of n linear forms x1∗,x2∗,…,xn∗ of M∗ is given by
⟨x1∗x2∗…xn∗,x1x2…xn⟩=i=1∏n⟨xi∗,xi⟩ for
xi∈M (
1≤i≤n); the product of these
n forms in
S(M)∗gr is given by
⟨x1∗x2∗…xn∗,x1x2…xn⟩=σ∈Sn∑(i=1∏n⟨xσ(i)∗,xi⟩);
finally, the product of these forms in ⋀(M)∗gr is given by
⟨x1∗x1∗…xn∗,x1x2…xn⟩=det(⟨xi∗,xj⟩). In each of the three cases, we have respectively
θT(x1∗⊗x2∗⊗…⊗xn∗)=x1∗x2∗…xn∗ θS(x1∗x2∗…xn∗)=x1∗x2∗…xn∗ θΛ(x1∗∧x2∗∧⋯∧xn∗)=xn∗xn−1∗…x1∗=(−1)n(n−1)/2x1∗x2∗…xn∗ and hence we deduce from (28), (29) and (30) the relations
(28 bis)⟨θT(x1∗⊗x2∗⊗⋯⊗xn∗),x1⊗x2⊗⋯⊗xn⟩=i=1∏n⟨xi∗,xi⟩ (in other words
θT restricted to
T2(M∗) is just the canonical homomorphism of II, § 4, no. 4)
(29 bis)⟨θS(x1∗x2∗…xn∗),x1x2…xn⟩=σ∈Sn∑(i=1∏n⟨xσ(i)∗,xi⟩) (30 bis)⟨θΛ(x1∗∧x2∗∧⋯∧xn∗),x1∧x2∧⋯∧xn⟩ =(−1)n(n−1)/2det(⟨xi∗,xj⟩).
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.
Nothing.