Bourbaki

Exercise 12, § 3

Lie Groups and Lie Algebras, I, 88

Bourbaki marks this one as harder than the rest.

Suppose that KK is a field. Let gg be a Lie KK-algebra and MM a gg-module. Let Cp(g,M)C^p(g, M) denote the space of alternating linear mappings of gpg^p into MM. We write C0(g,M)=MC^0(g, M) = M and, for p<0p < 0, Cp(g,M)={0}C^p(g, M) = \{0\}. Let C∗(g,M)C^*(g, M) be the direct sum of the Cp(g,M)C^p(g, M). The elements of C∗(g,M)C^*(g, M) are called the cochains of gg with values in MM; those of Cp(g,M)C^p(g, M) are said to be of degree pp. For all y∈gy \in g, we denote by i(y)i(y) the endomorphism of C∗(g,M)C^*(g, M) which maps each subspace Cp(g,M)C^p(g, M) into Cp−1(g,M)C^{p-1}(g, M) and which, for p>0p > 0, is given by the formula:

(i(y)f)(x1,…,xp−1)=f(y,x1,…,xp−1). (i(y)f)(x_1, \ldots, x_{p-1}) = f(y, x_1, \ldots, x_{p-1}).

We have i(y)2=0i(y)^2 = 0.

(a) The adjoint representation of gg and the representation of gg on MM define a representation of gg on the space of multilinear mappings of gpg^p into MM. Show that Cp(g,M)C^p(g, M) is stable under this representation. Let θ\theta be the representation of gg on C∗(g,M)C^*(g, M) thus defined. Show that:

θ(x)i(y)−i(y)θ(x)=i([x,y]) \theta(x)i(y) - i(y)\theta(x) = i([x, y])

for all x,yx, y in gg.

(b) Show that there exists one and only one endomorphism dd of C∗(g,M)C^*(g, M), mapping Cp(g,M)C^p(g, M) into Cp+1(g,M)C^{p+1}(g, M), such that

di(y)+i(y)d=θ(y) di(y) + i(y)d = \theta(y)

for all y∈gy \in g. (Argue by induction on the degree of the cochains.) Show that, for f∈Cp(g,M)f \in C^p(g, M):

df(x1,x2,…,xp+1)=∑i<j(−1)i+jf([xi,xj],x1,…,x^i,…,x^j,…,xp+1) df(x_1, x_2, \ldots, x_{p+1}) = \sum_{i < j} (-1)^{i+j}f([x_i, x_j], x_1, \ldots, \hat{x}_i, \ldots, \hat{x}_j, \ldots, x_{p+1})
+∑i(−1)i+1(xi)Mf(x1,…,x^i,…,xp+1) + \sum_i (-1)^{i+1}(x_i)_M f(x_1, \ldots, \hat{x}_i, \ldots, x_{p+1})

(where the symbol ^ over a letter means that it is omitted).

(c) Show that, for all y∈gy \in g,

dθ(y)=θ(y)d. d \theta(y) = \theta(y)d.

(Show first, using (2) and (3), that dθ(y)−θ(y)dd \theta(y) - \theta(y)d commutes with every i(x)i(x). Then argue by induction on the degree of the cochains.)

(d) Show that d2=0d^2 = 0. (Show first, using (3) and (4), that d2d^2 commutes with every i(x)i(x). The argue by induction on the degree of the cochains.)

(e) The restriction of dd to Cp(g,M)C^p(g, M) has a kernel Zp(g,M)Z^p(g, M) whose elements are called cocycles of degree pp with values in MM. The restriction of dd to Cp−1(g,M)C^{p-1}(g, M) has image Bp(g,M)B^p(g, M), whose elements are called coboundaries of degree pp with values in MM. Then Bp(g,M)⊂Zp(g,M)B^p(g, M) \subset Z^p(g, M). The quotient space

Zp(g,M)/Bp(g,M)=Hp(g,M) Z^p(g, M)/B^p(g, M) = H^p(g, M)

is called the cohomology space of gg of degree pp with values in MM. The direct sum of the Hp(g,M)H^p(g, M) is denoted by H∗(g,M)H^*(g, M). Show that H0(g,M)H^0(g, M) is identified with the set of invariants of MM. Let ϕ\phi be a homomorphism of the gg-module MM into the GG-module NN. For all f∈Cp(g,M)f \in C^p(g, M), ϕ∘f∈Cp(g,N)\phi \circ f \in C^p(g, N), so that ϕ\phi can be extended to a K-linear mapping ϕ′:C∗(g,M)→C∗(g,N)\phi': C^*(g, M) \to C^*(g, N). Show that ϕ′∘d=d∘ϕ′\phi' \circ d = d \circ \phi'. Deduce that ϕ′\phi' defines a homomorphism

ϕ~:H∗(g,M)→H∗(g,N) \tilde{\phi}: H^*(g, M) \to H^*(g, N)

which is said to be associated with ϕ\phi.

(f) Let LL be a sub-gg-module of MM and NN the quotient gg-module M/LM/L. The canonical homomorphisms L→iM→pNL \xrightarrow{i} M \xrightarrow{p} N define homomorphisms

C∗(g,L)→i′C∗(g,M)→p′C∗(g,N),Hn(g,L)→inHn(g,M)→pnHn(g,N). \begin{array}{cccccc} C^*(g, L) & \xrightarrow{i'} & C^*(g, M) & \xrightarrow{p'} & C^*(g, N), \\ H^n(g, L) & \xrightarrow{i^n} & H^n(g, M) & \xrightarrow{p^n} & H^n(g, N). \end{array}

Show that i′i' is injective and has image the kernel of p′p' and that p′p' is surjective. For all c∈Hn(g,N)c \in H^n(g, N), let z∈Zn(g,N)z \in Z^n(g, N) be a representative of cc and a∈Cn(g,M)a \in C^n(g, M) be such that p′(a)=zp'(a) = z; show that da∈Zn+1(g,L)da \in Z^{n+1}(g, L) and that its class in Hn+1(g,L)H^{n+1}(g, L) depends only on cc; denote this class by δnc\delta^n c. Show that the sequence of homomorphisms:

0⟶H0(g,L)→i0H0(g,M)→p0H0(g,N)→δ0⟶H1(g,L)→i1H1(g,M)→p1H1(g,N)→δ1⋯ \begin{array}{cccccccccc} 0 & \longrightarrow & H^0(g, L) & \xrightarrow{i^0} & H^0(g, M) & \xrightarrow{p^0} & H^0(g, N) & \xrightarrow{\delta^0} \\ & & \longrightarrow & H^1(g, L) & \xrightarrow{i^1} & H^1(g, M) & \xrightarrow{p^1} & H^1(g, N) & \xrightarrow{\delta^1} & \cdots \end{array}

is an exact sequence.

(g) With the notation of (f), the exact sequence:

0→L(N,L)→L(N,M)→L(N,N)→0 0 \to \mathscr{L}(N, L) \to \mathscr{L}(N, M) \to \mathscr{L}(N, N) \to 0

defines an exact sequence

H0(g,L(N,M))→H0(g,L(N,N))→H1(g,L(N,L)). H^0(g, \mathscr{L}(N, M)) \to H^0(g, \mathscr{L}(N, N)) \to H^1(g, \mathscr{L}(N, L)).

The identity mapping of N is an invariant u of L(N,N)\mathscr{L}(N, N) and hence an element of H0(g,L(N,N))H^0(g, \mathscr{L}(N, N)); let c be its image in H1(g,L(N,L))H^1(g, \mathscr{L}(N, L)). Then for the existence in M of a sub-g-module supplementary to L it is necessary and sufficient that c=0c = 0. (This condition means that u is the image of an element of H0(g,L(N,M))H^0(g, \mathscr{L}(N, M)), that is a homomorphism v of the g-module N into the g-module M; then v(N)v(N) is the desired supplement.)

(h) Show that Z1(g,g)Z^1(g, g) (where g is considered as a g-module by means of the adjoint representation) is identified with the vector space of derivations of g and that B1(g,g)B^1(g, g) is identified with the vector space of inner derivations of g.

(i) Let a and b be two Lie K-algebras with b commutative and b→λg→μab \xrightarrow{\lambda} g \xrightarrow{\mu} a an extension of a by b. For all x∈gx \in g, the restriction of adgx\mathrm{ad}_g x to b depends only on the class of x modulo b, whence there is an a-module structure on b. Let v be a K-linear mapping of a into g such that μ∘v\mu \circ v is the identity mapping of a. For x,yx, y in a, we write f(x,y)=[vx,vy]−v([x,y])f(x, y) = [vx, vy] - v([x, y]). Show that f∈Z2(a,b)f \in Z^2(a, b) and that the class c of f in H2(a,b)H^2(a, b) does not depend on the choice of v. For two extensions of a by b, which define the same a-module structure on b, to be equivalent, it is necessary and sufficient that the corresponding class c∈H2(a,b)c \in H^2(a, b) be the same. For the extension to be inessential, it is necessary and sufficient that c=0c = 0. If B is an a-module and c∈H2(a,B)c \in H^2(a, B), there exists an extension of a by B (considered as a commutative Lie algebra) defining the given a-module structure on B and the given element c of H2(a,B)H^2(a, B).

(j) Let g be a finite-dimensional Lie algebra over K, U its enveloping algebra, h an ideal of g, β\beta an invariant bilinear form on g whose restriction to h is non-degenerate, (yi)1≤i≤n(y_i)_{1 \leq i \leq n} and (yi′)1≤i≤n(y'_i)_{1 \leq i \leq n} two bases of h such that

β(yi,yj′)=δij,t=∑i=1nyiyi′∈U,M a g-module and ρ the endomorphism of C∗(g,M) which maps each Cp(g,M) into Cp−1(g,M) and which, for p>0, is defined by: \beta(y_i, y'_j) = \delta_{ij}, \quad t = \sum_{i=1}^n y_i y'_i \in U, \quad M \text{ a } g\text{-module and } \rho \text{ the endomorphism of } C^*(g, M) \text{ which maps each } C^p(g, M) \text{ into } C^{p-1}(g, M) \text{ and which, for } p > 0, \text{ is defined by:}
(ρf)(x1,…,xp−1)=∑k=1n(yk)Mf(yk′,x1,…,xp−1). (\rho f)(x_1, \ldots, x_{p-1}) = \sum_{k=1}^n (y_k)_M f(y'_k, x_1, \ldots, x_{p-1}).

Finally let Γ\Gamma be the endomorphism of C∗(g,M)C^*(g, M) extending tMt_M which maps each cochain f of degree > 0 onto the cochain tM∘ft_M \circ f. Show that ρd+dρ=Γ\rho d + d \rho = \Gamma. (If for x∈gx \in g we write:

[x,yk]=∑l=1naklyl,[x,yk′]=∑l=1nakl′yl′, [x, y_k] = \sum_{l=1}^n a_{kl} y_l, \qquad [x, y'_k] = \sum_{l=1}^n a'_{kl} y'_l,

show first that akl=−alk′a_{kl} = -a'_{lk}.) Deduce that Γd=dΓ\Gamma d = d \Gamma. If M is simple and finite-dimensional over K, β\beta is the associated bilinear form and dim h is not divisible by the characteristic of K, show that Hp(g,M)={0}H^p(g, M) = \{0\} for all pp. (Using Proposition 12, show that Γ\Gamma is an automorphism of C∗(g,M)C^*(g, M) and hence induces automorphisms of Zp(g,M)Z^p(g, M) and Bp(g,M)B^p(g, M). For

f∈Zp(g,M),Γf=(dρ+ρd)f=dρf∈Bp(g,M), f \in Z^p(g, M), \quad \Gamma f = (d\rho + \rho d)f = d\rho f \in B^p(g, M),
whence Zp(g,M)=Bp(g,M)Z^p(g, M) = B^p(g, M).)

Solution

No solution has been written yet.

Transcribed from the printed volume. Not translated.