Bourbaki

Theorem 1

Tag 00QG, permanent. Cite it as [Bourbaki, Tag 00QG].

Algebra, VIII, p. 455, Appendix 2. Determinants over a Noncommutative Field

Suppose n⩾2n\geqslant 2.

a) The subgroup SLn(D)\mathbf{S}\mathbf{L}_n(D) of GLn(D)\mathbf{G}\mathbf{L}_n(D) is generated by the matrices Bij(λ)B_{ij}(\lambda ), where ii and jj are distinct integers in the interval [1,n][1, n] and λ\lambda runs through D.

b) Suppose n⩾3n\geqslant 3 or Card(D) ⩾3\geqslant 3. The derived group of GLn(D)\mathbf{G}\mathbf{L}_n(D) is equal to SLn(D)\mathbf{S}\mathbf{L}_n(D).

c) Suppose n⩾3n\geqslant 3 or Card(D) ⩾4\geqslant 4. The derived group of SLn(D)\mathbf{S}\mathbf{L}_n(D) is equal to SLn(D)\mathbf{S}\mathbf{L}_n(D).

A) Denote by T the subgroup of GLn(D)\mathbf{G}\mathbf{L}_n(D) generated by the matrices Bij(λ)B_{ij}(\lambda ). By Example 1 of VIII, p. 454, we have det(Bij(λ)B_{ij}(\lambda )) =1= 1, and therefore T⊂SLn(D)T\subset \mathbf{S}\mathbf{L}_n(D). To prove that these two groups are equal, it then suffices, by Example 3 of loc. cit., to prove that every matrix of the form diag(1,...,1,d, . . . ,1, d) with π(d)=1\pi (d) = 1 belongs to T. The matrix diag(1,...,1,d, . . . ,1, d) belongs to the image of the homomorphism U↦(In−200U)U\mapsto \begin{pmatrix} I_{n-2} & 0 \\ 0 & U \end{pmatrix} from GL2(D)\mathbf{G}\mathbf{L}_2(D) to GLn(D)\mathbf{G}\mathbf{L}_n(D); it therefore suffices to consider the case n=2n= 2. Since the kernel of π\pi is the derived group of D∗D^*, we may assume d=uvu−1v−1d=uvu^{-1}v^{-1} with u,vu, v in D∗D^*. Our assertion then follows from the equalities

(28) (100d)=(u−100u)(v−100v)(vu00u−1v−1)\begin{pmatrix} 1 & 0 \\ 0 & d \end{pmatrix} = \begin{pmatrix} u^{-1} & 0 \\ 0 & u \end{pmatrix}\begin{pmatrix} v^{-1} & 0 \\ 0 & v \end{pmatrix}\begin{pmatrix} vu & 0 \\ 0 & u^{-1}v^{-1} \end{pmatrix}

and

(29) (s00s−1)=(1s01)(101−s−11)(1−101)(101−s1)\begin{pmatrix} s & 0 \\ 0 & s^{-1} \end{pmatrix} = \begin{pmatrix} 1 & s \\ 0 & 1 \end{pmatrix}\begin{pmatrix} 1 & 0 \\ 1-s^{-1} & 1 \end{pmatrix}\begin{pmatrix} 1 & -1 \\ 0 & 1 \end{pmatrix}\begin{pmatrix} 1 & 0 \\ 1-s & 1 \end{pmatrix}

for s∈D∗s\in D^*.

B) By construction, SLn(D)\mathbf{S}\mathbf{L}_n(D) is the kernel of a homomorphism from GLn(D)\mathbf{G}\mathbf{L}_n(D) to a commutative group, so it contains the derived group (GLn(D),GLn(D))(\mathbf{G}\mathbf{L}_n(D),\mathbf{G}\mathbf{L}_n(D)) of GLn(D)\mathbf{G}\mathbf{L}_n(D). By the above, we have

SLn(D)⊃(GLn(D),GLn(D))⊃(SLn(D),SLn(D)) \mathbf{S}\mathbf{L}_n(D)\supset (\mathbf{G}\mathbf{L}_n(D),\mathbf{G}\mathbf{L}_n(D))\supset (\mathbf{S}\mathbf{L}_n(D),\mathbf{S}\mathbf{L}_n(D))

To prove c), it suffices to prove that the matrices Bij(λ)B_{ij}(\lambda ) are commutators in SLn(D)\mathbf{S}\mathbf{L}_n(D).

Suppose n⩾3n\geqslant 3. If i,j,ki,j,k are distinct integers in the interval [1,n][1, n] and μ,ν\mu, \nu are elements of D, then we have

Bij(μν)=Bik(μ)−1Bkj(ν)−1Bik(μ)Bkj(ν)(30) B_{ij}(\mu\nu ) =B_{ik}(\mu)^{-1}B_{kj}(\nu )^{-1}B_{ik}(\mu)B_{kj}(\nu ) \tag{30}

By taking μ=1\mu= 1 and ν=λ\nu =\lambda, we see that the matrix Bij(λ)B_{ij}(\lambda ) is a commutator of elements of SLn(D)\mathbf{S}\mathbf{L}_n(D).

Now suppose n=2n= 2. Let uu and vv be elements of D with u≠0u\not= 0. We have the relations

(31) (1v−uvu01)=(u00u−1)(1−v01)(u−100u)(1v01)\begin{pmatrix} 1 & v-uvu \\ 0 & 1 \end{pmatrix} = \begin{pmatrix} u & 0 \\ 0 & u^{-1} \end{pmatrix}\begin{pmatrix} 1 & -v \\ 0 & 1 \end{pmatrix}\begin{pmatrix} u^{-1} & 0 \\ 0 & u \end{pmatrix}\begin{pmatrix} 1 & v \\ 0 & 1 \end{pmatrix} and

(32) (10v−uvu1)=(u−100u)(10−v1)(u00u−1)(10v1)\begin{pmatrix} 1 & 0 \\ v-uvu & 1 \end{pmatrix} = \begin{pmatrix} u^{-1} & 0 \\ 0 & u \end{pmatrix}\begin{pmatrix} 1 & 0 \\ -v & 1 \end{pmatrix}\begin{pmatrix} u & 0 \\ 0 & u^{-1} \end{pmatrix}\begin{pmatrix} 1 & 0 \\ v & 1 \end{pmatrix}.

We have det (u00u−1)\begin{pmatrix} u & 0 \\ 0 & u^{-1} \end{pmatrix} = 1, so the matrices B12(v−uvu)B_{12}(v-uvu) and B21(v−uvu)B_{21}(v-uvu) are commutators of elements of SLn(D)\mathbf{S}\mathbf{L}_n(D).

Suppose that the field D has at least four elements. Let λ\lambda be an element of D. If λ\lambda is equal to 0, 1, or −1-1, choose an arbitrary element uu in D−{0,1,−1}D-\{0,1,-1\}; otherwise, set u=λu=\lambda. In both cases, uu is a nonzero element of D, it commutes with λ\lambda, and we have u2≠1u^2\not= 1. Set v=λ(1−u2)−1v=\lambda (1-u^2)^{-1}. We have uv=vuuv=vu and therefore v−uvu=v(1−u2)=λv-uvu=v(1-u^2) =\lambda. It then follows from relations (31) and (32) that the matrices B12(λ)B_{12}(\lambda ) and B21(λ)B_{21}(\lambda ) are commutators in SLn(D)\mathbf{S}\mathbf{L}_n(D). This completes the proof of c).

C) It remains to prove that SL2(D)\mathbf{S}\mathbf{L}_2(D) is the derived group of GL2(D)\mathbf{G}\mathbf{L}_2(D) when D has three elements. By A), the group SL2(D)\mathbf{S}\mathbf{L}_2(D) is generated by the matrices B12(1)=(1101)B_{12}(1) = \begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix} and B21(1)=(1011)B_{21}(1) = \begin{pmatrix} 1 & 0 \\ 1 & 1 \end{pmatrix}, and these matrices are commutators of elements of GL2(D)\mathbf{G}\mathbf{L}_2(D) because we have B21(1)=tB12(1)B_{21}(1) =^tB_{12}(1) and

(33) (1101)=(−1001)(1101)(−1001)−1(1101)−1\begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix} = \begin{pmatrix} -1 & 0 \\ 0 & 1 \end{pmatrix}\begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix}\begin{pmatrix} -1 & 0 \\ 0 & 1 \end{pmatrix}^{-1}\begin{pmatrix} 1 & 1 \\ 0 & 1 \end{pmatrix}^{-1}.

Bourbaki sets the proof directly after the statement, and the corpus does not separate them, so what is above is the statement and its proof together.

Cited by

Cites

Transcribed from the printed edition, 2023, Springer Nature. Not translated.