Bourbaki

Proposition 14

Tag 068H, permanent. Cite it as [Bourbaki, Tag 068H].

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

(i) The orthogonal of MzM_z in E∗E^* is the set of x∗∈E∗x^* \in E^* such that x∗⊥z=0x^* \perp z = 0.

(ii) The subspace MzM_z associated with z is the image of Λp−1(E∗)\Lambda^{p-1}(E^*) under the mapping λz:u∗↦u∗⊥z\lambda_z : u^* \mapsto u^* \perp z of Λp−1(E∗)\Lambda^{p-1}(E^*) into E.

Let N denote the image of λz\lambda_z. For x∗∈E∗x^* \in E^* and u∗∈Λp−1(E∗)u^* \in \Lambda^{p-1}(E^*),

⟨θΛ(x∗),u∗⊥z⟩=⟨θΛ(u∗∧x∗),z⟩=(−1)p−1⟨θΛ(x∗∧u∗),z⟩=(−1)p−1⟨θΛ(u∗),x∗⊥z⟩. \langle \theta_{\Lambda}(x^*), u^* \perp z \rangle = \langle \theta_{\Lambda}(u^* \wedge x^*), z \rangle = (-1)^{p-1} \langle \theta_{\Lambda}(x^* \wedge u^*), z \rangle \\ = (-1)^{p-1} \langle \theta_{\Lambda}(u^*), x^* \perp z \rangle.
Therefore, for x∗x^* to be orthogonal to N, it is necessary and sufficient that x∗⊥zx^* \perp z be orthogonal to θΛ(Λ(E∗))\theta_{\Lambda}(\Lambda(E^*)). Now, the latter condition is equivalent to saying that x∗⊥z=0x^* \perp z = 0; for let (eλ)λ∈L(e_{\lambda})_{\lambda \in L} be a basis of E; giving L a total ordering, it has been seen (§ 7, no. 8, Theorem 1) that the eJe_J, for J running through the set F(L)\mathscr{F}(L) of finite subsets of L, form a basis of Λ(E)\Lambda(E); it then follows from formula (30) of no. 5 that the elements θ∧(eJ∗)\theta_{\wedge}(e_J^*) are, to within a sign, the coordinate forms on ∧(E)\wedge(E) relative to the basis (eJ)(e_J); whence our assertion.

The orthogonal of N therefore consists of the x∗∈E∗x^* \in E^* such that x∗⊥z=0x^* \perp z = 0 and the conclusion of (i) will therefore follow from (ii).

We show first that N⊂MzN \subset M_z. Let M be a vector subspace of E such that z∈∧(M)z \in \wedge(M) and let j:M→Ej : M \to E be the canonical injection; let μz\mu_z denote the mapping v∗↦v∗⊥zv^* \mapsto v^* \perp z of ∧p−1(M∗)\wedge^{p-1}(M^*) into M; it follows from formula (60) of no. 7 that there is a canonical factorization

λz:∧p−1(E∗)→∧p−1(tj)∧p−1(M∗)→μzM→jE \lambda_z : \wedge^{p-1}(E^*) \xrightarrow{\wedge^{p-1}(t_j)} \wedge^{p-1}(M^*) \xrightarrow{\mu_z} M \xrightarrow{j} E

which proves that N⊂MN \subset M and hence N⊂MzN \subset M_z by definition of MzM_z. It remains to verify that N=MzN = M_z. Suppose the converse: there would then exist a basis (ei)1≤i≤n(e_i)_{1 \leq i \leq n} of MzM_z and an element x∗∈E∗x^* \in E^* such that ⟨x∗,e1⟩=1,⟨x∗,ej⟩=0\langle x^*, e_1 \rangle = 1, \langle x^*, e_j \rangle = 0 for 2≤j≤n2 \leq j \leq n and such that x∗x^* is orthogonal to N and hence x∗⊥z=0x^* \perp z = 0. We write z=∑HaHeHz = \sum_H a_H e_H, where the sum is taken over the subsets of {1,n}\{1, n\} with pp elements. By (68) (no. 9),

x∗⊥eH=0if 1∉Hx∗⊥e{1}∪H=eHif H⊂{2,n} x^* \perp e_H = 0 \quad \text{if } 1 \notin H \\ x^* \perp e_{\{1\} \cup H} = e_H \quad \text{if } H \subset \{2, n\}

which shows that the relation x∗⊥z=0x^* \perp z = 0 implies aH=0a_H = 0 for 1∈H1 \in H. But this is impossible, for zz would then belong to ∧p(M′)\wedge^p(M'), where M′M' is the subspace of M generated by e2,…,ene_2, \ldots, e_n.

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

Cited by

Nothing.

Cites

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