Bourbaki

Definition 6

Tag 05NJ, permanent. Cite it as [Bourbaki, Tag 05NJ].

Algebra, chapter II, § 2. Modules of linear mappings. Duality

Given an isomorphism u of an A-module E onto an A-module F, the transpose of the inverse isomorphism of u (equal to the inverse isomorphism of the transpose of u) is called the contragredient isomorphism of u and denoted by u~\tilde{u}.

The isomorphism u~\tilde{u} is thus characterized by the relation (19)

⟨u(x),u~(y∗)⟩=⟨x,x∗⟩for x∈E,x∗∈E∗. \langle u(x), \tilde{u}(y^*) \rangle = \langle x, x^* \rangle \quad \text{for } x \in E, x^* \in E^*.

If v:F→Gv : F \to G is an isomorphism, the contragredient isomorphism of v∘uv \circ u is v~∘u~\tilde{v} \circ \tilde{u}.

In particular, the mapping u↦u~u \mapsto \tilde{u} is an isomorphism of the linear group GL(E)\mathbf{GL}(E) onto a subgroup of the linear group GL(E∗)\mathbf{GL}(E^*).

Let σ:A→B\sigma : A \to B be an isomorphism of a ring AA onto a ring BB, EE a left AA-module, FF a left BB-module and u:E→Fu : E \to F a semi-linear mapping (§ 1, no. 13) relative to σ\sigma. Let σ−1\sigma^{-1} be the inverse isomorphism of σ\sigma; for all y∗∈F∗y^* \in F^*, the mapping x↦⟨u(x),y∗⟩σ−1x \mapsto \langle u(x), y^* \rangle^{\sigma^{-1}} of EE into AA is a linear form; if it is also denoted by tu(y∗){}^t u(y^*), a mapping tu:F∗→E∗{}^t u : F^* \to E^* is defined which is also called the transpose of the semi-linear mapping uu; it is thus characterized by the identity

⟨u(x),y∗⟩=⟨x,tu(y∗)⟩σ \langle u(x), y^* \rangle = \langle x, {}^t u(y^*) \rangle^\sigma

for x∈E,y∗∈F∗x \in E, y^* \in F^*. It is immediately verified that tu{}^t u is a semi-linear mapping relative to σ−1\sigma^{-1}. If vv denotes the mapping uu considered as an A-linear mapping of EE into σ∗(F)\sigma_*(F) (§ 1, no. 13), we may write u=ϕ∘vu = \phi \circ v, where ϕ\phi is the identity mapping σ∗(F)→F\sigma_*(F) \to F, considered as a semi-linear mapping relative to σ\sigma. It is immediate that tu=tv∘tϕ{}^t u = {}^t v \circ {}^t \phi and (tϕ,σ−1)({}^t \phi, \sigma^{-1}) is a di-isomorphism of F∗F^* onto (σ∗(F))∗(\sigma_*(F))^* relative to the isomorphism σ−1\sigma^{-1}; this relation allows us immediately to extend the properties of transposes of linear mappings to transposes of semi-linear mappings.

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.