Symplectic Geometry

⚡ Key Takeaways

Symplectic geometry is the study of smooth manifolds equipped with a closed, nondegenerate 2-form $\omega$ (a symplectic form). One-line statement: it is the geometry of phase space, the natural arena for Hamiltonian mechanics, where area — not length — is the fundamental measured quantity. Key result: locally all symplectic manifolds look identical (Darboux's theorem), so there are no local invariants; all interest is global. Applies to: classical and quantum mechanics, integrable systems, mirror symmetry, low-dimensional topology. Origins: Lagrange and Hamilton (1808–1834); name coined by Hermann Weyl (1939); modern era launched by Gromov (1985).

💡 The Core Insight

Symplectic structure measures oriented area in conjugate planes, not lengths. A Riemannian metric is a symmetric nondegenerate 2-tensor; a symplectic form is the antisymmetric nondegenerate counterpart. Antisymmetry forces even dimension and bans local invariants, while closedness ( $d\omega=0$ ) is exactly the integrability condition that makes the form arise from a flow and guarantees the conservation laws of mechanics. The slogan: dynamics = a function (the Hamiltonian) + a symplectic form, and the form converts the differential of the function into a vector field.

DefinitionSymplectic Vector Space

A symplectic vector space is a finite-dimensional real vector space $V$ with a bilinear form $\omega : V \times V \to \mathbb{R}$ that is

antisymmetric: $\omega(u,v) = -\omega(v,u)$ , and
nondegenerate: if $\omega(u,v)=0$ for all $v\in V$ , then $u=0$ .

1

\quad \omega(u,v) = -\,\omega(v,u), \qquad \omega^\flat : V \xrightarrow{\ \sim\ } V^*,\quad \omega^\flat(u) = \omega(u,\,\cdot\,).

Derivation hint: nondegeneracy is precisely the statement that $\omega^\flat$ is injective; in finite dimensions injectivity equals bijectivity, giving the isomorphism with $V^*$ . Nondegeneracy forces $\dim V = 2n$ to be even, and there exists a symplectic (Darboux) basis $e_1,\dots,e_n,f_1,\dots,f_n$ with

2

\quad \omega(e_i,e_j)=\omega(f_i,f_j)=0,\qquad \omega(e_i,f_j)=\delta_{ij}.

Derivation hint: pick any nonzero $e_1$ ; by nondegeneracy find $f_1$ with $\omega(e_1,f_1)=1$ ; the $\omega$ -orthogonal complement of $\mathrm{span}(e_1,f_1)$ is again symplectic of dimension $2n-2$ ; induct.

DefinitionSymplectic Manifold

A symplectic manifold is a pair $(M^{2n},\omega)$ where $\omega\in\Omega^2(M)$ satisfies

3

\quad d\omega = 0 \quad(\text{closed}), \qquad \omega^n = \underbrace{\omega\wedge\cdots\wedge\omega}_{n} \neq 0 \ \text{ everywhere (nondegenerate)}.

Derivation hint: nondegeneracy of the bilinear form at each point is equivalent to the top power $\omega^n$ being a nowhere-vanishing volume form, since $\omega^n = n!\,e_1^*\wedge f_1^*\wedge\cdots$ in a Darboux frame. The form $\tfrac{1}{n!}\omega^n$ is the Liouville volume; it canonically orients $M$ . In local Darboux coordinates $(q^1,\dots,q^n,p_1,\dots,p_n)$ ,

4

\quad \omega = \sum_{i=1}^{n} dp_i \wedge dq^i .

Derivation hint: this is the manifold version of [2]; Darboux's theorem (below) guarantees such coordinates exist in a neighborhood of every point.

DefinitionHamiltonian Vector Field & Poisson Bracket

Given a smooth function $H:M\to\mathbb{R}$ (the Hamiltonian), nondegeneracy lets us solve for the unique vector field $X_H$ :

5

\quad \iota_{X_H}\omega = dH, \qquad\text{i.e.}\qquad \omega(X_H,\,\cdot\,) = dH.

Derivation hint: $\omega^\flat$ from [1] is invertible, so $X_H = (\omega^\flat)^{-1}(dH)$ ; this is well-defined precisely because $\omega$ is nondegenerate. The Poisson bracket of two functions is

6

\quad \{F,G\} = \omega(X_F,X_G) = X_F(G) = -X_G(F).

Derivation hint: substitute $\iota_{X_F}\omega = dF$ ; then $\omega(X_F,X_G)=dF(X_G)= X_G(F)$ and antisymmetry of $\omega$ gives the sign relations. In Darboux coordinates these become Hamilton's equations and the canonical bracket:

7

\quad \dot q^i = \frac{\partial H}{\partial p_i},\qquad \dot p_i = -\frac{\partial H}{\partial q^i},\qquad \{F,G\}=\sum_i\Big(\frac{\partial F}{\partial q^i}\frac{\partial G}{\partial p_i}-\frac{\partial F}{\partial p_i}\frac{\partial G}{\partial q^i}\Big).

Derivation hint: write $X_H=\sum a^i\partial_{q^i}+b_i\partial_{p_i}$ , apply [5] with $\omega=\sum dp_i\wedge dq^i$ , and match components of $dH$ .

TheoremDarboux's Theorem

Let $(M^{2n},\omega)$ be a symplectic manifold. Around every point there is a coordinate chart in which $\omega = \sum_i dp_i\wedge dq^i$ . Consequently symplectic manifolds have no local invariants: any two are locally symplectomorphic.

TheoremLiouville's Theorem

The flow $\phi_t$ of any Hamiltonian vector field $X_H$ preserves the symplectic form, $\phi_t^*\omega=\omega$ , and hence the Liouville volume $\omega^n/n!$ . Moreover $H$ is conserved: $\{H,H\}=0$ .

Proof

We prove $\phi_t^*\omega=\omega$ using Cartan's magic formula $\mathcal L_X = d\circ\iota_X + \iota_X\circ d$ .

Step 1. Compute the Lie derivative of $\omega$ along $X_H$ :

8

\quad \mathcal L_{X_H}\omega = d(\iota_{X_H}\omega) + \iota_{X_H}(d\omega).

Derivation hint: Cartan's formula applied directly to the 2-form $\omega$ .

Step 2. The second term vanishes because $\omega$ is closed: $d\omega=0$ . The first term vanishes because $\iota_{X_H}\omega = dH$ by [5], and $d(dH)=0$ . Hence

9

\quad \mathcal L_{X_H}\omega = d(dH) + 0 = 0.

Derivation hint: $d^2=0$ kills the exact part; closedness kills the other — both hypotheses on $\omega$ are used.

Step 3. Since $\dfrac{d}{dt}\phi_t^*\omega = \phi_t^*(\mathcal L_{X_H}\omega) = 0$ and $\phi_0=\mathrm{id}$ , we get $\phi_t^*\omega=\omega$ for all $t$ . Taking the $n$ -th wedge power gives $\phi_t^*(\omega^n)=\omega^n$ , i.e. volume preservation.

Step 4. Conservation of energy: $\dfrac{dH}{dt}=X_H(H)=\{H,H\}=\omega(X_H,X_H)=0$ by antisymmetry. $\blacksquare$

∎

WR

William Rowan Hamilton

1805–1865

The Irish mathematician and physicist reformulated Newtonian mechanics in 1834–1835 into the form bearing his name, replacing $n$ second-order equations by $2n$ first-order ones in position $q$ and momentum $p$ . Hamilton's deep aim was an analogy between optics and mechanics ("the characteristic function"), which prefigured both the Hamilton–Jacobi equation and, a century later, wave mechanics: Schrödinger explicitly used Hamilton's optical–mechanical analogy. Hamilton never wrote the form $\omega$ , but his canonical equations are the local expression [7] of the symplectic structure. The coordinate-free distillation of his work into closed nondegenerate 2-forms came only with the 20th-century differential-geometers.

🧠 Intuition

Think of a Riemannian metric as a device for measuring lengths and angles of single vectors; think of a symplectic form as a device for measuring the signed area of the parallelogram spanned by pairs of vectors. Because area changes sign when you swap the two sides, the form is antisymmetric. The pairing $(q,p)$ is the key: in $\mathbb R^2$ with $\omega=dp\wedge dq$ , the symplectic "area" of position and momentum directions is exactly the area swept in phase space. Hamilton's equations say a particle moves so as to follow the level sets of energy rotated 90° by $\omega$ : $dH$ points "uphill" in energy, and $X_H=\omega^{-1}dH$ rotates it to point along the energy level. That is why energy is conserved — the flow is always tangent to its own energy surface. Liouville's theorem is then the statement that this 90°-rotated flow is incompressible: a blob of initial conditions in phase space deforms wildly but never changes its volume. This is the geometric heart of statistical mechanics.

Figure

Hamiltonian flow rotates the energy gradient by 90°, keeping motion along energy level sets while preserving phase-space area

ExampleThe Harmonic Oscillator — Phase-Space Flow

Given: Phase space $M=\mathbb R^2$ with coordinates $(q,p)$ , symplectic form $\omega = dp\wedge dq$ , and Hamiltonian

H(q,p)=\tfrac{1}{2}\big(p^2 + \omega_0^2\, q^2\big),

with angular frequency $\omega_0 = 2\ \mathrm{rad/s}$ .

Step 1 — find $X_H$ . Using [7]: $\dot q = \partial H/\partial p = p$ , $\dot p = -\partial H/\partial q = -\omega_0^2 q$ . So $X_H = p\,\partial_q - \omega_0^2 q\,\partial_p$ .

Step 2 — solve the flow. Differentiating: $\ddot q = \dot p = -\omega_0^2 q$ , giving $q(t)=A\cos(\omega_0 t+\varphi)$ , $p(t)=-A\omega_0\sin(\omega_0 t+\varphi)$ . With initial $q(0)=1\ \mathrm m$ , $p(0)=0\ \mathrm{kg\,m/s}$ (unit mass): $A=1$ , $\varphi=0$ .

Step 3 — verify conservation. $H=\tfrac12(0 + 4\cdot 1)=2\ \mathrm J$ , and along the flow $H(t)=\tfrac12(A^2\omega_0^2\sin^2+\omega_0^2 A^2\cos^2)=\tfrac12\omega_0^2A^2=2\ \mathrm J$ constant. ✓

Step 4 — phase-space area. The orbit is the ellipse $p^2+\omega_0^2q^2=\omega_0^2A^2$ , with semi-axes $a=A=1$ (in $q$ ) and $b=A\omega_0=2$ (in $p$ ). Liouville area enclosed:

\mathrm{Area}=\pi a b = \pi\cdot1\cdot2 = 2\pi.

Boxed answer:

\boxed{\,q(t)=\cos(2t)\ \mathrm m,\quad p(t)=-2\sin(2t)\ \mathrm{kg\,m/s},\quad H=2\ \mathrm J,\quad \text{enclosed area}=2\pi\ \mathrm{J\,s}\,}

The action $J=\oint p\,dq = \mathrm{Area} = 2\pi$ is the adiabatic invariant.

ExampleCotangent Bundle of

S^1

— The Pendulum's Canonical Form

Given: Configuration space $Q=S^1$ (angle $\theta$ ), so phase space is $T^*Q$ with coordinates $(\theta, p_\theta)$ . The tautological 1-form is $\lambda = p_\theta\,d\theta$ and the canonical symplectic form is $\omega = d\lambda$ .

Step 1 — compute $\omega$ . $\omega = d(p_\theta\,d\theta) = dp_\theta\wedge d\theta$ . This matches [4] with $n=1$ .

Step 2 — pendulum Hamiltonian. $H(\theta,p_\theta)=\dfrac{p_\theta^2}{2m\ell^2} - mg\ell\cos\theta$ . Take $m=1\ \mathrm{kg}$ , $\ell=1\ \mathrm m$ , $g=10\ \mathrm{m/s^2}$ .

Step 3 — Hamilton's equations. $\dot\theta = \partial H/\partial p_\theta = p_\theta$ (since $m\ell^2=1$ ), and $\dot p_\theta = -\partial H/\partial\theta = -mg\ell\sin\theta = -10\sin\theta$ .

Step 4 — separatrix energy. The unstable equilibrium is at $\theta=\pi$ , $p_\theta=0$ , with energy $H=-mg\ell\cos\pi = +10\ \mathrm J$ . The stable equilibrium at $\theta=0$ has $H=-10\ \mathrm J$ . Orbits with $H<10\ \mathrm J$ oscillate (bounded $\theta$ ); orbits with $H>10\ \mathrm J$ rotate (circulating).

Boxed answer:

\boxed{\,\omega = dp_\theta\wedge d\theta,\quad \dot\theta=p_\theta,\quad \dot p_\theta = -10\sin\theta,\quad E_{\text{separatrix}} = +10\ \mathrm J\,}

The cylinder topology of $T^*S^1$ is what allows the qualitatively distinct libration vs rotation regimes.

⚠️ Common mistake

"A symplectic form is just an antisymmetric metric." No — there is a structural chasm. A metric has local invariants (curvature, geodesics) and a rich isometry-rigidity. A symplectic form, by Darboux's theorem, is locally unique: it has no curvature, no local invariants at all. All symplectic information is global/topological. Conflating the two leads to looking for nonexistent "symplectic curvature."

"Closedness $d\omega=0$ is a technicality." It is the crux. Drop it and you get an almost symplectic structure: Hamiltonian flows would no longer preserve $\omega$ (Step 2 of the proof fails), energy conservation and Liouville's theorem collapse, and there is no Darboux normal form. Closedness is what makes mechanics work.

"Every even-dimensional manifold is symplectic." False — there are topological obstructions. A closed symplectic manifold must have $H^2_{\mathrm{dR}}(M)\neq 0$ with $[\omega]^n\neq 0$ , ruling out $S^4$ and $S^{2n}$ for $n\ge 2$ .

Counterexample

$S^{2n}$ is not symplectic for $n\geq 2$ . Suppose $S^4$ admitted a symplectic form $\omega$ . Then $[\omega]\in H^2_{\mathrm{dR}}(S^4)$ . But $H^2_{\mathrm{dR}}(S^4)=0$ , so $[\omega]=0$ , i.e. $\omega=d\alpha$ is exact. Then $\omega^2 = d(\alpha\wedge\omega)$ is also exact, so by Stokes $\int_{S^4}\omega^2 = 0$ . Yet $\omega^2$ is a nowhere-vanishing volume form, forcing $\int_{S^4}\omega^2\neq 0$ — contradiction. Note $S^2$ is symplectic (its area form works) precisely because $H^2_{\mathrm{dR}}(S^2)\neq 0$ . The dimension-2 sphere is the lucky exception.

connection

Symplectic geometry connects outward in many directions:

To Riemannian Geometry: a symplectic form $\omega$ together with a compatible almost-complex structure $J$ and metric $g$ form a compatible triple, $\omega(u,v)=g(Ju,v)$ ; when $J$ is integrable and $\omega$ is its Kähler form, one gets Kähler Manifolds, the overlap of symplectic, complex, and Riemannian geometry.
To Lie Groups via the moment map: a Hamiltonian $G$ -action produces $\mu:M\to\mathfrak g^*$ , and Symplectic Reduction (Marsden–Weinstein) builds new symplectic manifolds $\mu^{-1}(0)/G$ — the geometric engine of gauge theory's phase-space reduction.
To Poisson Geometry: symplectic manifolds are exactly the nondegenerate Poisson manifolds; the bracket [6] makes $C^\infty(M)$ a Poisson algebra, generalizing to singular settings.
To Contact Geometry: the odd-dimensional sibling, governing energy hypersurfaces and the boundaries of symplectic manifolds.
To Geometric Quantization and Deformation Quantization: the formal bridge from the Poisson bracket to the quantum commutator, $[\hat F,\hat G] = i\hbar\,\widehat{\{F,G\}}+O(\hbar^2)$ .

🔬 Where this appears

⏳ History

- 1808–1810 — Lagrange introduces the "Lagrange bracket" while studying perturbations of planetary orbits; this is the first appearance of the symplectic form on phase space, born from celestial mechanics.

- 1809 — Poisson defines the Poisson bracket [6] (in coordinates), capturing the algebraic structure of conserved quantities.

- 1834–1835 — Hamilton recasts mechanics into canonical form [7], unifying optics and dynamics through the principal function.

- 1838 — Liouville proves volume preservation of Hamiltonian flow, foundational for statistical mechanics.

- 1880s–1890s — Poincaré develops qualitative dynamics, integral invariants, and the geometric viewpoint on the three-body problem; the "last geometric theorem" (1912) anticipates symplectic fixed-point results.

- 1939 — Hermann Weyl coins the word "symplectic" in The Classical Groups, deliberately calquing the Greek root of "complex" (συμπλεκτικός, "plaiting together") to avoid confusion with the complex group.

- 1971 — Marsden and Weinstein (and independently work building on Kostant–Souriau) formalize the moment map and symplectic reduction, linking symmetry to phase-space quotients.

- 1985 — Mikhail Gromov introduces $J$ -holomorphic curves and proves the non-squeezing theorem, igniting modern symplectic topology and revealing the field's deep rigidity.

- 1988–1990s — Andreas Floer constructs Floer homology, resolving cases of the Arnold conjecture and connecting symplectic geometry to gauge theory and, soon after, to mirror symmetry in string theory.

open-problem

The Arnold conjecture. Vladimir Arnold conjectured that a Hamiltonian symplectomorphism of a closed symplectic manifold $(M^{2n},\omega)$ must have at least as many fixed points as a function on $M$ has critical points — in the nondegenerate case, at least $\sum_i \dim H_i(M)$ . Floer homology proved many cases, and the degenerate (cup-length) and most general forms remain partially open, especially over arbitrary coefficients and in infinite-dimensional or non-compact settings.

Symplectic capacities and packing. Gromov non-squeezing shows the radius is a symplectic invariant; this spawned the theory of symplectic capacities. Whether all normalized capacities coincide on convex domains (the Viterbo conjecture, with its connection to the Mahler conjecture in convex geometry) is open and actively researched.

Existence of symplectic structures. On open manifolds the $h$ -principle (Gromov) gives flexibility, but on closed manifolds existence is governed by subtle topology; deciding which closed manifolds admit symplectic (but not Kähler) structures remains hard, with constructions (Thurston, Gompf) and obstructions (Seiberg–Witten via Taubes) only partially mapping the landscape.

Quantization. There is no functor turning every symplectic manifold into a Hilbert space compatibly with all classical observables (Groenewold–van Hove no-go theorem); the precise scope of geometric and deformation quantization is still being delineated.

Formal Statement

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also