Classification of homogeneous symplectic manifolds (Kirillov-Kostant-Souriau)

Intuition Beginner

A coadjoint orbit is built from a group acting on the dual of its Lie algebra, and it always carries a symplectic form. That is the forward story. This unit tells the reverse story: among all phase spaces with the most symmetry, the orbits are not just examples. They are the whole list.

The "most symmetry" condition is that the group can slide any point of the phase space onto any other point, while respecting the symplectic pairing. A space like that is called homogeneous.

The headline is short. Up to a shift by a charge and a possible covering, every homogeneous symplectic phase space is a coadjoint orbit.

Visual Beginner

The picture shows one big space on the left, an arrow labelled by a group action, and a flat orbit sheet on the right inside the dual space. The arrow says: collapse the symmetric space onto an orbit.

The dashed loop on the orbit marks the charge shift: sometimes the orbit lives at an offset point, not through the origin. The covering label reminds us the match can wrap several times.

Worked example Beginner

Take the round sphere of radius one. The rotation group moves any point of the sphere to any other point, so the sphere is homogeneous.

Give the sphere its area form. This is a symplectic form, because the sphere is two dimensional and the area form never vanishes. Rotations preserve area, so they respect the pairing.

Now match this to the reverse story. The sphere is exactly the orbit of the rotation group acting on the three dimensional dual space, sitting at radius one. The area form is the orbit's own symplectic form.

What this tells us: a very symmetric phase space, the sphere, was a coadjoint orbit all along.

Check your understanding Beginner

Formal definition Intermediate+

Let $(M,\omega )$ be a connected symplectic manifold and let a Lie group $G$ act on $M$ by symplectomorphisms. The prerequisites used here are the coadjoint orbit and KKS form 05.03.01, the Souriau cocycle and affine coadjoint action 05.04.07, and the moment map 05.04.01. The sign convention is the geometric Hamiltonian convention $\omega (X_H,-)=dH$.

The action is Hamiltonian when it admits a moment map $J:M\to {\mathfrak{g}}^{\ast }$ whose components $J_{\xi }(x)=⟨J(x),\xi ⟩$ satisfy ${\iota }_{X_{\xi }}\omega =d{J}_{\xi }$ for every $\xi \in \mathfrak{g}$, where $X_{\xi }$ is the fundamental vector field generated by $\xi$. The action is transitive when, for every pair $x,y\in M$, some $g\in G$ carries $x$ to $y$.

A connected symplectic manifold carrying a transitive Hamiltonian $G$-action is called a homogeneous symplectic $G$-manifold. The moment map need not be equivariant for the linear coadjoint action; its defect is measured by the Souriau one-cocycle $\theta :G\to {\mathfrak{g}}^{\ast }$, $\theta (g)=J(g\cdot x)-{\mathrm{Ad}}_g^{\ast }J(x)$ 05.04.07, whose class lives in $H^1(G,{\mathfrak{g}}^{\ast })$ with infinitesimal partner in $H^2(\mathfrak{g},\mathbb{R})$.

The affine coadjoint action twisted by $\theta$ is $g\cdot \mu ={\mathrm{Ad}}_g^{\ast }\mu +\theta (g)$. Its orbits are the affine coadjoint orbits, and the moment map $J$ is equivariant for this affine action by construction of $\theta$ ^{[Souriau 1970]}.

Key theorem with proof Intermediate+

Theorem (Kirillov-Kostant-Souriau exhaustion). Let $(M,\omega )$ be a homogeneous symplectic $G$-manifold with moment map $J:M\to {\mathfrak{g}}^{\ast }$ and Souriau cocycle $\theta$. Then $J$ is a local diffeomorphism onto a single affine coadjoint orbit ${\mathcal{O}}_{\mu }^{\theta }\subset {\mathfrak{g}}^{\ast }$, equivariant for the affine action $g\cdot \mu ={\mathrm{Ad}}_g^{\ast }\mu +\theta (g)$, and the pullback $J^{\ast }{\omega }_{\mathrm{KKS}}^{\theta }$ of the twisted KKS form equals $\omega$. Consequently $J:M\to {\mathcal{O}}_{\mu }^{\theta }$ is a covering map, and $M$ is, up to that covering, the affine coadjoint orbit with its twisted KKS symplectic form. When $\theta =0$ the orbit is an ordinary coadjoint orbit of $G$.

Proof.

Step 1 — The image is one affine orbit. Because the action is transitive and $J$ is affine-equivariant, $J(M)=J(G\cdot x)=G\cdot J(x)$ under the affine action, which is exactly the affine orbit ${\mathcal{O}}_{\mu }^{\theta }$ through $\mu =J(x)$. So $J$ maps $M$ onto a single affine coadjoint orbit.

Step 2 — The differential is injective. Fix $x\in M$. A tangent vector at $x$ has the form $X_{\xi }(x)$ for some $\xi \in \mathfrak{g}$, by transitivity. The infinitesimal affine action sends $\xi$ to the vector ${\mathrm{ad}}_{\xi }^{\ast }\mu +{\theta }^{\prime }(\xi )$ at $\mu =J(x)$, where ${\theta }^{\prime }$ is the infinitesimal cocycle 05.04.07. The defining identity ${\iota }_{X_{\xi }}\omega =d{J}_{\xi }$ gives, for any $\eta \in \mathfrak{g}$,

$${\omega }_x(X_{\xi },X_{\eta })=d{J}_{\eta }(X_{\xi })=⟨d{J}_x(X_{\xi }),\eta ⟩=⟨{\mathrm{ad}}_{\xi }^{\ast }\mu +{\theta }^{\prime }(\xi ),\eta ⟩.$$

If $d{J}_x(X_{\xi })=0$, then ${\omega }_x(X_{\xi },X_{\eta })=0$ for all $\eta$, hence for all tangent vectors at $x$ by transitivity. Non-degeneracy of $\omega$ forces $X_{\xi }(x)=0$. Therefore $d{J}_x$ is injective on $T_{x}M$.

Step 3 — Local diffeomorphism, matching dimensions. The same computation shows $T_{\mu }{\mathcal{O}}_{\mu }^{\theta }$ is spanned by the vectors ${\mathrm{ad}}_{\xi }^{\ast }\mu +{\theta }^{\prime }(\xi )$, so $d{J}_x$ is a linear surjection onto $T_{\mu }{\mathcal{O}}_{\mu }^{\theta }$. Being injective and surjective, $d{J}_x$ is an isomorphism, and $\dim M=\dim {\mathcal{O}}_{\mu }^{\theta }$. Hence $J$ is a local diffeomorphism.

Step 4 — The form is the pullback of the twisted KKS form. The twisted KKS form on the affine orbit is fixed at $\mu$ by

$${\omega }_{\mathrm{KKS}}^{\theta }{|}_{\mu }({\mathrm{ad}}_{\xi }^{\ast }\mu +{\theta }^{\prime }(\xi ),{\mathrm{ad}}_{\eta }^{\ast }\mu +{\theta }^{\prime }(\eta ))=⟨\mu ,[\xi ,\eta ]⟩+{\theta }^{\prime }(\xi )(\eta ),$$

which is precisely the affine cocycle correction of the orbit form 05.03.01, 05.04.07. The displayed identity of Step 2 is the same expression evaluated through $J$, so $(J^{\ast }{\omega }_{\mathrm{KKS}}^{\theta }{)}_x(X_{\xi },X_{\eta })={\omega }_x(X_{\xi },X_{\eta })$ for all generators, giving $J^{\ast }{\omega }_{\mathrm{KKS}}^{\theta }=\omega$.

Step 5 — Covering. A surjective local diffeomorphism from a connected manifold onto a homogeneous space, equivariant for transitive group actions on both sides, is a covering map: the fibres are the cosets of the discrete subgroup $G_x/G_{\mu }^{\circ }$ relating the stabiliser of $x$ to the identity component of the affine stabiliser of $\mu$, and these cosets have constant cardinality. Thus $J:M\to {\mathcal{O}}_{\mu }^{\theta }$ is a covering, and $M$ is identified with the affine coadjoint orbit up to that cover. When $\theta =0$ the affine action reduces to ${\mathrm{Ad}}^{\ast }$ and ${\mathcal{O}}_{\mu }^{\theta }$ is an ordinary coadjoint orbit of $G$ 05.03.01. $\square$

Bridge. This theorem builds toward the orbit method and its appearance again in geometric quantisation 05.11.05, where the same affine orbits are quantised into representations. The central insight is that transitivity plus the symplectic pairing leaves no freedom beyond a charge and a cover, and this is exactly why coadjoint orbits are the universal classical phase spaces. The forward construction of 05.03.01 generalises here into a converse, and the cocycle correction is dual to the central-extension data of 05.04.07: putting these together, the foundational reason elementary systems are classified by orbits is that homogeneity forces the moment map to be an immersion onto one.

Exercises Intermediate+

Advanced results Master

The cohomological content of the theorem is the whole story of when the charge shift can be removed. The Souriau cocycle defines a class in $H^1(G,{\mathfrak{g}}^{\ast })$, whose infinitesimal shadow is the antisymmetric two-cocycle $c(\xi ,\eta )={\theta }^{\prime }(\xi )(\eta )$ in $Z^2(\mathfrak{g},\mathbb{R})$ 05.04.07. The orbit is an ordinary coadjoint orbit of $G$ exactly when $[c]$ vanishes in $H^2(\mathfrak{g},\mathbb{R})$; otherwise it is a coadjoint orbit of the central extension $\tilde{G}$ classified by $[c]$.

For a semisimple Lie group, Whitehead's lemma gives $H^2(\mathfrak{g},\mathbb{R})=0$, so every cocycle is a coboundary and can be absorbed by translating the moment map. Every homogeneous symplectic manifold of a semisimple group is therefore a covering of a genuine coadjoint orbit, with no charge to carry ^{[Kostant 1970]}. The phenomenon of an unremovable charge is intrinsically a feature of non-semisimple groups.

The Galilei group is the standard witness. Its second cohomology is one dimensional, generated by the Bargmann cocycle with bracket defect $[K_i,P_j]=m{\delta }_{ij}$ 05.04.07. The parameter $m$ is the mass, and it cannot be set to zero by any coboundary. The homogeneous symplectic manifold of a free non-relativistic particle is thus a mass-shifted affine orbit, an orbit of the Bargmann central extension rather than of the bare Galilei group. The orbit is not at the origin of the dual algebra but at the charge offset set by $m$.

These examples set up the moral that the next layer reads off. Coadjoint orbits, possibly of a central extension, are not a catalogue of nice symplectic manifolds. They are the complete inventory of the most symmetric classical phase spaces, the spaces Souriau named elementary systems ^{[Souriau 1970]}.

Synthesis. The exhaustion theorem is the converse that completes the forward construction of 05.03.01: where that unit showed a coadjoint orbit is symplectic, this one shows the most symmetric symplectic manifolds are exactly those orbits, up to a cover and a charge. This is exactly why the orbit method works, and it generalises the sphere example into a structural law; the cocycle correction is dual to the central-extension data of 05.04.07, so the charge of an elementary system is a cohomology class. Putting these together, the foundational reason coadjoint orbits recur from classical mechanics through geometric quantisation 05.11.05 is the central insight that homogeneity pins the moment map to be a covering onto one affine orbit, and the bridge is the moment map 05.04.01 itself, recognised here as a local symplectomorphism rather than merely an equivariant map.

Full proof set Master

Proposition. Let $(M,\omega )$ be a homogeneous symplectic $G$-manifold with equivariant moment map $J$ (cocycle $\theta =0$). Then $J(M)$ is a single coadjoint orbit and $\omega =J^{\ast }{\omega }_{\mathrm{KKS}}$.

By equivariance $J(g\cdot x)={\mathrm{Ad}}_g^{\ast }J(x)$, so transitivity gives $J(M)={\mathrm{Ad}}_G^{\ast }J(x)={\mathcal{O}}_{\mu }$, one coadjoint orbit through $\mu =J(x)$. For generators $\xi ,\eta \in \mathfrak{g}$ the moment identity gives ${\omega }_x(X_{\xi },X_{\eta })=d{J}_{\eta }(X_{\xi })=⟨{\mathrm{ad}}_{\xi }^{\ast }\mu ,\eta ⟩=⟨\mu ,[\xi ,\eta ]⟩$, which is the value of $J^{\ast }{\omega }_{\mathrm{KKS}}$ on the same vectors 05.03.01. Since fundamental fields span every tangent space by transitivity, $\omega =J^{\ast }{\omega }_{\mathrm{KKS}}$.

Proposition. The infinitesimal cocycle $c(\xi ,\eta )={\theta }^{\prime }(\xi )(\eta )$ is an antisymmetric two-cocycle on $\mathfrak{g}$.

The defect of equivariance differentiates to $c(\xi ,\eta )=\{J_{\xi },J_{\eta }\}-J_{[\xi ,\eta ]}$, the failure of the moment map to be a Lie-algebra homomorphism into the Poisson algebra. Antisymmetry follows from antisymmetry of the Poisson bracket and the bracket on $\mathfrak{g}$. The cocycle condition

$$c([\xi ,\eta ],\zeta )+c([\eta ,\zeta ],\xi )+c([\zeta ,\xi ],\eta )=0$$

follows from the Jacobi identity for the Poisson bracket applied to $J_{\xi },J_{\eta },J_{\zeta }$, since the $J_{[\cdot ,\cdot ]}$ terms cancel cyclically. Hence $c\in Z^2(\mathfrak{g},\mathbb{R})$, and its class $[c]\in H^2(\mathfrak{g},\mathbb{R})$ obstructs equivariance.

Connections Master

• The forward direction this unit reverses is the coadjoint orbit and its KKS form 05.03.01: there a group orbit in ${\mathfrak{g}}^{\ast }$ is shown to be symplectic, and here that construction is shown to exhaust the homogeneous case.

• The charge shift and the affine action come from the Souriau cocycle 05.04.07, whose class in $H^1(G,{\mathfrak{g}}^{\ast })$ decides whether the orbit sits at the origin or at a mass-like offset such as the Bargmann central charge.

• The moment map 05.04.01 is the very map proved here to be a local symplectomorphism onto an affine orbit, so this theorem is a structural statement about moment maps of transitive actions.

• Laterally, geometric quantisation of coadjoint orbits and the Borel-Weil theorem 05.11.05 quantise exactly the orbits this unit identifies, turning the universal classical phase spaces into the building blocks of representation theory.

Historical & philosophical context Master

Souriau introduced the classification of homogeneous symplectic manifolds in his 1970 treatise, where he argued that the elementary systems of physics, the free particle and the spinning particle among them, are precisely the orbits of a symmetry group acting affinely on the dual of its Lie algebra ^{[Souriau 1970]}. The affine twist by a cocycle was his device for capturing mass and other charges that the bare symmetry group cannot see, and it is this twist that distinguishes his account from the purely equivariant orbit picture.

Kostant gave a parallel development in the same years, framing the central extensions and the integrality of the symplectic form within the program of geometric quantisation ^{[Kostant 1970]}. Together with Kirillov's orbit method for nilpotent groups, the result became the structural foundation for the claim that classical phase spaces and unitary representations are two readings of one geometric object: the coadjoint orbit, possibly of a central extension.

Bibliography Master

@book{Souriau1970Structure,
  author = {Souriau, Jean-Marie},
  title = {Structure des syst\`emes dynamiques},
  publisher = {Dunod},
  year = {1970}
}

@incollection{Kostant1970Quantization,
  author = {Kostant, Bertram},
  title = {Quantization and unitary representations},
  booktitle = {Lectures in Modern Analysis and Applications III},
  series = {Lecture Notes in Mathematics},
  volume = {170},
  publisher = {Springer},
  year = {1970},
  pages = {87--208}
}

@book{Kirillov2004Orbit,
  author = {Kirillov, A. A.},
  title = {Lectures on the Orbit Method},
  series = {Graduate Studies in Mathematics},
  volume = {64},
  publisher = {American Mathematical Society},
  year = {2004}
}