Let g be a Lie algebra over K and M a g-module. The g-module structure on M and the trivial g-module structure on K define a g-module structure on the K-module N=L(M,M;K) of bilinear forms on M. Then
(xN⋅β)(m,m′)=−β(xM⋅m,m′)+β(m,xM⋅m′)
for all x∈g,m,m′ in M,β∈N. If β is a given element of N, the set of x∈g such that xN⋅β=0 is a subalgebra of g.
Let M be a K-module and β a bilinear form on M. By the above, the set of x∈gl(M) such that
β(x⋅m,m′)+β(m,x⋅m′)=0
for all m∈M and m′∈M is a Lie subalgebra of gl(M). Suppose that K is a field, M is finite-dimensional and β is non-degenerate. Then every x∈gl(M) admits a left adjoint x∗ (relative to β) which is everywhere defined and the subalgebra in question is the set of x∈gl(M) such that x∗=−x. By this process we can construct two important examples of Lie algebras:
(a) Take M=Kn and
β((ξ1,…,ξn),(η1,…,ηn))=ξ1η1+⋯+ξnηn.
We canonically identify gl(Kn) with Mn(K). Then the Lie algebra obtained is the Lie algebra of skew-symmetric matrices. (When K=R, this algebra is the Lie algebra of the orthogonal group O(n,R)).
(b) Take M=K2m and
β((ξ1,…,ξ2m),(η1,…,η2m))=ξ1ηm+1−η1ξm+1+⋯+ξmη2m−ηmξ2m. The matrix of
β with respect to the canonical basis of
K2m is the matrix
(0−ImIm0). Let
U=(ACBD) be the matrix with respect to the canonical basis of
K2m of an element
u of
gl(M) (
A,B,C,D lying in
Mm(K)). By formula (50) of
Algebra, Chapter IX, § 1, no. 10,
u∗ has with respect to the same basis the matrix
(0Im−Im0)(tAtBtCtD)(0−ImIm0)=(tD−tC−tBtA). The condition
u∗=−u is therefore equivalent to the conditions
D=−tAB=tBC=tC. When K=R, the Lie algebra obtained is the Lie algebra of the symplectic group Sp(2m,R).
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.