Let (ei)i∈I be an orthonormal basis of the hilbertian space E. For every α in N(I), put
zα=i∈I∏eiαi/(αi!)1/2.
Then (zα)α∈N(I) is an orthonormal basis of S^(E).
Let E0 be the vector subspace of E generated by the vectors ei for i ranging over I. Then the zα form a basis of the vector space S(E0) (A, III, § 6, No. 6). But E0 is dense in E, and the multilinear mapping (x1,…,xn)↦x1…xn from E×⋯×E into S(E) is continuous for all n⩾1; hence S(E0) is dense in S(E). It is now enough to prove that the family of the zα is orthonormal. First observe that S^n(E) and S^m(E) are orthogonal for n=m. Hence it is enough to prove the formula
⟨zα∣zβ⟩={10if α=βif α=β
when ∣α∣=∑i∈Iαi and ∣β∣=∑i∈Iβi are equal to the same integer n.
Consider a partition (Pi)i∈I of the set {1,2,...,n} such that Card Pi=αi for all i∈I. Put xk=ei if k belongs to Pi, then x1...xn=∏i∈Ieiαi. Similarly we define (Qi)i∈I and yk in such a way that Card Qi=βi and y1...yn=∏i∈Ieiβi. Since the ei are mutually orthogonal, we have ⟨xk∣yσ(k)⟩=0 except if there exists an indice i∈I such that k∈Pi and σ(k)∈Qi. By formula (15), we then have ⟨x1...xn∣y1...yn⟩=0 unless there exists a permutation σ∈Sn such that σ(Pi)=Qi for all i∈I, which implies that α=β. Then ⟨zα∣zβ⟩=0 for α=β. The same argument proves that ∥x1...xn∥2 is equal to the number of the σ∈Sn such that σ(Pi)=Pi for all i∈I, hence equal to ∏i∈Iαi!. We get ∥zα∥=1, and the proposition is proved.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.
Nothing.