Bourbaki

Proposition 7

Tag 09HJ, permanent. Cite it as [Bourbaki, Tag 09HJ].

Integration, VII, p. 70, § 3. Applications and examples

Let D+∗D_+^* be the group of diagonal matrices with diagonal elements >0> 0. The mapping (U,D,T)↦UDT(U, D, T) \mapsto UDT is a homeomorphism of U(n,K)×D+∗×T1(n,K)\mathbf{U}(n, \mathbf{K}) \times D_+^* \times \mathbf{T}_1(n, \mathbf{K}) onto GL(n,K)\mathbf{GL}(n, \mathbf{K}).

Let (ε1,…,εn)(\varepsilon_1, \ldots, \varepsilon_n) be the canonical basis of KnK^n. Let X∈GL(n,K)X \in \mathbf{GL}(n, K). Then the X⋅εi=fiX \cdot \varepsilon_i = f_i form a basis of E. To this basis (fi)(f_i) one can associate a basis (ei)(e_i) as in Lemma 10. Let UU be the matrix of the unitary automorphism of E that transforms εi\varepsilon_i into eie_i. Then

U−1⋅fi=ε1αi1+ε2αi2+⋯+εiαii U^{-1} \cdot f_i = \varepsilon_1 \alpha_{i1} + \varepsilon_2 \alpha_{i2} + \cdots + \varepsilon_i \alpha_{ii}

with αii>0\alpha_{ii} > 0 for i=1,2,…,ni = 1, 2, \ldots, n. Thus X=UCX = UC, where CC is the matrix

(α11α21⋯αn10α22⋯αn2⋮⋮⋱⋮00⋯αnn). \begin{pmatrix} \alpha_{11} & \alpha_{21} & \cdots & \alpha_{n1} \\ 0 & \alpha_{22} & \cdots & \alpha_{n2} \\ \vdots & \vdots & \ddots & \vdots \\ 0 & 0 & \cdots & \alpha_{nn} \end{pmatrix}.

Moreover, UU and CC depend continuously on X by Lemma 10. On the other hand, formula (8) shows that CC may be put in the form DTDT with D∈D+∗D \in \mathbf{D}_+^*, T∈T1(n,K)T \in \mathbf{T}_1(n, K), DD and TT depending continuously on CC. The uniqueness of the decomposition X=UDTX = UDT follows from the uniqueness property of Lemma 10.

The homeomorphism of Prop. 7 is called the Iwasawa decomposition of GL(n,K)\mathbf{GL}(n, K).

The group G=D+∗⋅T1(n,K)G = \mathbf{D}_+^* \cdot \mathbf{T}_1(n, K) is the set of upper triangular matrices over K whose diagonal elements are >0> 0. Let us identity the element (zij)(z_{ij}) of this group with the element

((zii)1≤i≤n,(zij)1≤i<j≤n)∈(R+∗)n×Kn(n−1)/2. ((z_{ii})_{1 \leq i \leq n}, (z_{ij})_{1 \leq i < j \leq n}) \in (\mathbf{R}_+^*)^n \times K^{n(n-1)/2}.

Arguing exactly as in Example 4, one finds as right Haar measure on this group the measure (when K=RK = \mathbf{R})

(∏i=1nzii−i)⋅(⨂i=1ndzii)⊗(⨂i<jdzij). \left( \prod_{i=1}^n z_{ii}^{-i} \right) \cdot \left( \bigotimes_{i=1}^n dz_{ii} \right) \otimes \left( \bigotimes_{i < j} dz_{ij} \right).

Then applying Prop. 13 of §2, No. 9, one sees that if GL(n,K)\mathbf{GL}(n, K) is identified with U(n,K)×G\mathbf{U}(n, K) \times G by the mapping (U,S)↦US(U, S) \mapsto US, a Haar measure on GL(n,K)\mathbf{GL}(n, K) is given by (when K=RK = \mathbf{R})

(∏i=1nzii−i)⋅α⊗(⨂i=1ndzii)⊗(⨂i<ndzij), \left( \prod_{i=1}^n z_{ii}^{-i} \right) \cdot \alpha \otimes \left( \bigotimes_{i=1}^n dz_{ii} \right) \otimes \left( \bigotimes_{i < n} dz_{ij} \right),

where α\alpha denotes a Haar measure on U(n,K)\mathbf{U}(n, K).

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 volume. Not translated.