LECTURE 7: LIE GROUP ACTIONS ON SYMPLECTIC MANIFOLDS
Contents 1. A crash course on Lie groups and Lie algebras 1 2. A crash course on Lie group actions 4 3. Symplectic actions 7 4. Lie algebra cohomology 9
NOTE: The proofs of the results stated in first two sections below can be found in my Lie group course notes at http://staff.ustc.edu.cn/˜ wangzuoq/Lie13/Lie.html
1. A crash course on Lie groups and Lie algebras ¶ Lie groups. Roughly speaking, • A group is a set with simple algebraic structure (multiplication and inverse) • A manifold is a set with nice geometry (locally looks like Rn) • A smooth manifold is a manifold on which one can do analysis and a Lie group is an organic integration of all these structures: Definition 1.1. A Lie group G is a smooth manifold with a group structure, so that the group multiplication map (g1, g2) 7→ g1g2 is smooth. Remark. Suppose G is a Lie group. • One can prove that the inversion map g 7→ g−1 is automatically smooth. • Any element a ∈ G gives rise to three natural diffeomorphisms on G, namely the left multiplication
La : G → G, g 7→ ag, the right multiplication
Ra : G → G, g 7→ ga, and the conjugation −1 ca : G → G, g 7→ aga . 1 2 LECTURE 7: LIE GROUP ACTIONS ON SYMPLECTIC MANIFOLDS
Example. Any vector space is a Lie group under vector addition. Example. The circle S1 ⊂ C and the punctured complex plane C∗ ⊂ C are Lie groups under complex multiplication. Example. The real torus Tn = S1 × · · · × S1 ⊂ Cn and the complex torus (C∗)n are Lie groups. Example. The general linear group GL(n, R) of all nonsingular matrices is a Lie group under matrix multiplication. Definition 1.2. A Lie group homomorphism ϕ : G → H between two Lie groups is a smooth map that preserves the group multiplication, i.e. ϕ(g1g2) = ϕ(g1)ϕ(g2) for all g1, g2 ∈ G. Definition 1.3. A subgroup ι : H,→ G of a Lie group G is called a Lie subgroup if it is a Lie group with respect to the induced group operation, and the inclusion map ι is a smooth immersion. Note that we don’t require H to be embedded submanifold of G. For example, one can easily construct a dense curve in T2 which is a Lie subgroup. Definition 1.4. A closed Lie subgroup of G is a Lie subgroups H of G that is also an embedded submanifold. One can show that any closed Lie subgroup must be closed subset of G. Con- versely, Theorem 1.5 (Cartan’s closed subgroup theorem). Any closed subgroup H of a Lie group G is a closed Lie subgroup. This is a very powerful tool in determine whether a group is a Lie group. ¶ Lie algebras associated to Lie groups. By definition a Lie algebra is a vector space g together with an anti-symmetric bilinear bracket [·, ·]: g × g → g which satisfies the Jacobi identity [[X,Y ],Z] + [[Y,Z],X] + [[Z,X],Y ] = 0. Lie algebras arises naturally as the linearization of a Lie group. There are several equivalent ways to describe the Lie algebra g associated to a given Lie group G. • First one can regard g as a Lie subalgebra of (Vect(G), [·, ·]). Definition 1.6. A vector field X on G is called left invariant if ∀a ∈ G,
dLa(Xg) = Xag. Fact: If X,X0 are left invariant vector fields on G, so is [X,X0]. In other words, the space of left invariant vector fields on G form a Lie subalgebra (whose dimension equals the dimension of G) of the (infinitely dimensional) Lie algebra of all vector fields on G. LECTURE 7: LIE GROUP ACTIONS ON SYMPLECTIC MANIFOLDS 3
Definition 1.7. The Lie algebra g = Lie(G) of a Lie group G is g = {all left invariant vector fields on G}, where the Lie bracket is the usual Lie bracket between vector fields. • Note that left invariant vector fields on G are in one-to-one correspondence with vectors in TeG: any vector Xe ∈ TeG determines uniquely a left invari- ant vector field X on G via Xa = dLa(Xe). So as a vector space
g = TeG.
The Lie bracket of two vectors X,Y ∈ TeG is defined to be [X,Y ] := ad(X)Y,
where ad : TeG → End(TeG) is defined as follows: For each g ∈ G, the conjugation map cg : G → G maps e to e. So its differential at e gives us a linear map
Adg = (dcg)e : TeG → TeG. In other words, we get a map (the adjoint representation of G)
Ad : G → Aut(TeG), g 7→ Adg.
Note that Ad(e) is the identity map in Aut(TeG). Moreover, since Aut(TeG) is an open subset in the linear space End(TeG), its tangent space at Id can be identified with End(TeG) in a natural way. Taking derivative again at e, we get (the adjoint representation of the Lie algebra g)
ad : TeG → End(TeG). • One can prove that any left invariant vector field X on G is complete. So the flow φt = exp(tX) exists for all t ∈ R and is a one-parameter subgroup of G: exp(tX) exp(sX) = exp((t + s)X). Conversely, from any one-parameter subgroup φ of G one can construct a d left-invariant vector field on G through the vector Xe = dt |t=0φt. So g = {all one-parameter subgroups of G}. The Lie bracket between exp(tX) and exp(tY ) can be defined to be the one-parameter subgroup generated by the vector
∂ ∂ exp(tX) exp(sY ) exp(−tX). ∂t t=0 ∂s s=0 So we associate to each Lie group G a Lie algebra g. Using the one-parameter subgroup exp(tX) associated to X one gets a natural map exp : g → G from the Lie algebra g to G. This is called the exponential map. It is a very important tool in Lie theory. 4 LECTURE 7: LIE GROUP ACTIONS ON SYMPLECTIC MANIFOLDS
Example. The Lie algebra of GL(n, R) is gl(n, R), the space of all n × n matrices, with Lie bracket [A, B] = AB − BA. the exponential map is given explicitly by 1 1 exp : gl(n, ) → GL(n, ),A 7→ eA = I + A + A2 + A3 + ··· . R R 2! 3! ¶ Standard facts. Here we list some standard facts from Lie theory:
• The differential dϕe : g = TeG → TeH = h of any Lie group homomorphism ϕ : G → H at e is a Lie algebra homomorphism. • The exponential map exp : g → G is a local diffeomorphism near 0, and is natural in the sense that for any Lie group homomorphism ϕ : G → H,
ϕ ◦ expg = exph ◦dϕe. [Draw a commutative diagram!] • The Lie algebra h of a subgroup H of G is automatically a Lie subalgebra of g. Explicitly,
h = {X ∈ g | expg(tX) ∈ H for all t ∈ R}. • This is a one-to-one correspondence between connected Lie subgroups of G and Lie subalgebras of g. • Any continuous homomorphism between Lie groups is smooth. • Suppose G is connected and simply connected. Then any Lie algebra homo- morphism ρ : g → h lifts to a Lie group homomorphism ϕ : G → H so that ρ = dϕ. • Any finitely dimensional Lie algebra is the Lie algebra of a unique connected and simply connected Lie algebra G. • Any connected abelian Lie group is of the form Rk × Tl.
2. A crash course on Lie group actions ¶ Lie group actions. Definition 2.1. A smooth action of a Lie group G on a smooth manifold M is a group homeomorphism τ : G → Diff(M) so that the evaluation map ev : G × M → M, (g, m) = τ(g)(m) is smooth.
For simplicity we will denote τ(g)(m) by g · m. Example. S1 acts on R2(= C) by rotations (scalar multiplications). Example. GL(n, R) acts on Rn by linear transformations. Example. The flow of any complete vector field on M is a smooth R-action on M. Conversely, any smooth R-action is the flow of a complete vector field. Example. G acts on G by left multiplication, right multiplication and by conjugation. LECTURE 7: LIE GROUP ACTIONS ON SYMPLECTIC MANIFOLDS 5
Example. The adjoint action of G on g is
Ad : G → Aut(g), g 7→ Adg,
where Adg : g → g is the differential of the conjugation c(g): G → G, a 7→ gag−1 at a = e. For the case G = GL(n, R), the adjoint action is given explicitly by −1 AdC (X) = CXC for C ∈ GL(n, R) and X ∈ gl(n, R). Example. The coadjoint action of G on g∗ is defined so that ∗ hAdgξ, Xi = hξ, Adg−1 Xi holds for all ξ ∈ g∗ and X ∈ g. [Check this is an action.] Definition 2.2. Let M and N be smooth manifolds with smooth G-action. A smooth map f : M → N is called equivariant if it commutes with the given G- actions, i.e. f(g · m) = g · f(m) for all g ∈ G and all m ∈ M.
¶ Orbits and stabilizers. Let τ : G → Diff(M) be a smooth action. Definition 2.3. The orbit of G through m ∈ M is G · m = {g · m | g ∈ G} ⊂ M, and the stabilizer (or isotropic subgroup) of m ∈ M is
Gm = {g ∈ G | g · m = m}. Obviously if m, m0 lie in the same G-orbit, i.e. m0 = g · m for some g ∈ G, then G · m = G · m0. So M can be decomposed into disjoint union of G-orbits. We define M/G = the space of G-orbits, equipped with the quotient topology. In general, this quotient topology might be very bad. For example, the quotient space of the R>0 action on R (by multiplica- tions) has non-Hausdorff topology. However, if one put suitable assumptions on the action, this quotient can have a nice geometric structure. Definition 2.4. A smooth G-action on M is
• free if Gm = {e} for all m ∈ M. • locally free if Gm is discrete for all m ∈ M. • transitively if G · m = M.
The following definition is useful. 6 LECTURE 7: LIE GROUP ACTIONS ON SYMPLECTIC MANIFOLDS
Definition 2.5. An action τ of a Lie group G on M is proper if the action map F : G × M → M × M, (g, m) = (g · m, m) is proper, i.e. the pre-image of any compact set is compact.
¶ Infinitesimal actions. Now let τ : G → Diff(M) be a Lie group action. For any X ∈ g, one can define a vector field XM on M via
d XM (m) = exp(tX) · m. dt t=0 This gives us a map dτ : g → Vect(M),X 7→ XM . Definition 2.6. We call dτ the infinitesimal action of g on M associated to the G-action τ, and call XM the induced vector field on M of X ∈ g. One can show that the infinitesimal action dτ is a Lie algebra anti-homomorphism, i.e. [XM ,YM ] = −[X,Y ]M . Moreover, it is natural in the sense
τ(exp(tX)) = exp(tXM ). ¶ Important facts. Again we list some important facts for Lie group actions. • G · m is an immerse submanifold. • Gm is a Lie subgroup of G with Lie algebra
gm = {X ∈ g | XM (m) = 0.} • If G is compact, then any G-action is proper. • If G acts on M properly, then Gm is compact. • If G acts on M properly, then the quotient M/G is Hausdorff. • If G acts on M properly, then each orbit G · m is an embedded submanifold of M, with tangent space
Tm(G · m) = {XM (m) | X ∈ g}. • If G acts on M properly and freely, then the orbit space is a manifold and the quotient map π : M → M/G is a submersion. • In particular, for any closed subgroup H ⊂ G, acting on G via h · g = gh−1, the quotient G/H is a smooth manifold. • If G acts on M smoothly, then the map
φ : G/Gm → G · m, Gm · m 7→ g · m
is a diffeomorphism between the quotient manifold G/Gm and the orbit G·m. LECTURE 7: LIE GROUP ACTIONS ON SYMPLECTIC MANIFOLDS 7
• Suppose G acts on M properly. Then for any subgroup H ⊂ G, the H-fixed point set M H = {m ∈ M | h · m = m for all h ∈ H} is a disjoint union of closed submanifolds of M.
¶ The slice theorem.
Now suppose G acts properly on M. Then for any g ∈ Gm, g · m = m. Taking derivative one gets the isotropy action of Gm on TmM via
g · v = dτg(v).
Obviously the tangent space Tm(G · m) is a subspace of TmM which is invariant under the isotropy action of Gm. By choosing a Gm-invariant inner product on TmM one can get a Gm-invariant decomposition
TmM = Tm(G · m) ⊕ W.
Choose D to be a sufficiently small disc (with respect to some Gm-invariant inner product, which always exists because Gm is a compact Lie group) in W so that Gm also acts on D. Theorem 2.7 (The slice theorem). Let G be a Lie group acts properly on a smooth manifold M. Then for each m ∈ M, there exists a G-equivariant diffeomorphism from the disc bundle G ×Gm D onto a G-invariant neighborhood of the orbit G · m in
M, whose restriction to the zero section G ×Gm {0} = G/Gm is the diffeomorphism F : G/Gm → G · m above.
Note that if the G-action at m is free, then G ×Gm D = G × D. Remark. In general if P is a principal G-bundle over M so that G acts from right on P properly and freely, and suppose G also acts linearly on a vector space W , then G acts (from left) on the product P × W by g · (p, w) = (p · g−1, g · w). This action is proper and free, and the quotient P × W/G is a vector bundle over M = P/G. As a consequence, locally above a small open set
U ⊂ G/Gm, the disc bundle G ×Gm D looks like U × D.
3. Symplectic actions ¶ Symplectic actions. Now let (M, ω) be a symplectic manifold, G a Lie group, and τ : G → Diff(M) a smooth action of G on M. The following definition is natural: Definition 3.1. The action τ is called a symplectic action if τ(G) ⊂ Symp(M, ω). 8 LECTURE 7: LIE GROUP ACTIONS ON SYMPLECTIC MANIFOLDS
In other words, for any g ∈ G, τ(g) ∈ Symp(M, ω) is a symplectomorphism. Example. Consider the S1 action on (S2, dθ ∧ dz) by rotations around the z-axis. In coordinates (θ, z) this action is explicitly given by t · (θ, z) = (θ + t, z). This is a symplectic action since for any fixed t, ∗ τt ω = d(θ + t) ∧ dz = dθ ∧ dz. 1 2 Example. Similarly the S action on (T , dθ1 ∧ dθ2) via
t · (θ1, θ2) = (θ1 + t, θ2) is a symplectic action.
2n Example. By definition the natural action of Sp(2n) on (R , Ω0) is symplectic. Proposition 3.2. A smooth action τ : G → Symp(M, ω) is symplectic if and only if for each X ∈ g, the induced vector field XM is a symplectic vector field. Proof. If the action is symplectic, then for any X ∈ g,
d ∗ d ∗ LXM ω = exp(tXM ) ω = τ(exp(tX)) ω = 0, dt t=0 dt t=0 i.e. XM is a symplectic vector field.
Conversely, suppose XM is a symplectic vector field for any X ∈ g. Then by definition exp(tXM ) are symplectomorphisms. In other words, for any t and X, τ(exp(tX)) ∈ Symp(M, ω). Since any element in G is a product of elements in this form, the conclusion follows. ¶ Hamiltonian actions. It is natural to define Hamiltonian actions as those symplectic actions whose generating vector fields are Hamiltonian functions. Since for each X ∈ g there is a generating vector field XM , it is natural to suitable choose Hamiltonian functions µX for these vector fields so that there depends nicely on X, e.g. are linear in X ∈ g. Gluing these together, one gets a map µ∗ : g → C∞(M),X → µX , or alternately, a map µ from M to g∗ so that µX (m) = hµ(m),Xi. Aiming at the applications in the next several lectures, we would like the maps µ∗ and µ be nice, for example the µ∗ should be a Lie algebra homomorphism, and µ should be a G-equivariant map. (And, we will prove next time that in fact these two conditions are equivalent.) LECTURE 7: LIE GROUP ACTIONS ON SYMPLECTIC MANIFOLDS 9
Definition 3.3. A symplectic action τ of G on a symplectic manifold (M, ω) is called weakly-Hamiltonian if there exists a map µ : M → g∗ so that for any X ∈ g, X dµ = ιXM ω, where µX (·) = hµ(·),Xi is the component of µ along X. A weakly Hamiltonian action is called Hamiltonian if the map µ : G → g∗ is equivariant with respect to the G-action on M and the adjoint G-action on g∗, i.e. ∗ µ ◦ τg = Adg ◦ µ. We shall call µ the moment map, and call (M, ω, G, µ) a Hamiltonian G-space. Remark. Consider the simple case G is an abelian group. Then g = Rn and g∗ = Rn, ∗ ∗ with trivial Lie bracket. Then the coadjoint action of G on g is trivial: Adg = Id for all g. It follows that the equivariance condition on µ is reduced to the simple condition: µ is G-invariant. Example. The S1-action on S2 by rotations described above is a Hamiltonian action with moment map µ(θ, z) = z, while the S1-action on T2 by rotations in the first variable is not a Hamiltonian action.
4. Lie algebra cohomology Student presentation