Symplectic Structure, the Symplectic Form, and Darboux's Theorem

Intuition Beginner

Phase space — the space of all positions $q$ and momenta $p$ — has a hidden geometric structure. It is not just a flat canvas where points wander around. The positions and momenta are "intertwined" in a specific way: every position variable $q$ is paired with a momentum variable $p$, and this pairing defines a notion of oriented area that is preserved by the motion.

What does "area is preserved" mean? Imagine drawing a closed loop in the $(q,p)$ plane. Every point on the loop evolves forward in time according to Hamilton's equations. After some time the loop has moved and deformed into a new shape. Liouville's theorem says the area enclosed by the loop does not change. The loop can stretch in one direction and compress in another, but the total area stays the same. This is analogous to an incompressible fluid: the phase-space points flow like a fluid that cannot be compressed or expanded.

The mathematical object that encodes this pairing and area-preservation is called the symplectic form, usually denoted $\omega$. The word "symplectic" comes from the Greek "symplektikos," meaning "intertwined" — it was coined by Hermann Weyl in 1939 as a replacement for "complex" (which had already been claimed by complex numbers). The symplectic form measures the "area" of the parallelogram spanned by any two tangent vectors in phase space, and the requirement that this area measurement is preserved under time evolution is what makes Hamiltonian mechanics special.

For a system with one degree of freedom, the symplectic form is just the area element $dq\wedge dp$ in the $(q,p)$ plane. For a system with $n$ degrees of freedom, the phase space is $2n$-dimensional, and the symplectic form is the sum of the area elements for each conjugate pair:

$$\omega =d{q}^1\wedge d{p}_1+d{q}^2\wedge d{p}_2+\cdots +d{q}^n\wedge d{p}_n.$$

The wedge symbol $\wedge$ means "oriented area of the parallelogram spanned by." So $d{q}^1\wedge d{p}_1$ measures the oriented area in the $(q^1,p_1)$-plane, $d{q}^2\wedge d{p}_2$ measures the oriented area in the $(q^2,p_2)$-plane, and so on. The total symplectic form is the sum of these paired area elements — one for each position-momentum pair.

Two key properties distinguish the symplectic form from an arbitrary area measurement:

1. Non-degeneracy. The symplectic form never gives zero for every pairing unless the vector itself is zero. This means it actually measures something meaningful for every genuine pair of directions. In two dimensions, this is automatic — $dq\wedge dp$ is non-degenerate as long as the coordinates are independent. In higher dimensions, non-degeneracy is the requirement that the symplectic form has no "null directions."

2. Closedness. The symplectic form has zero exterior derivative — it does not change as you move around the manifold. This is a global consistency condition: the area measurement defined by $\omega$ fits together coherently across the entire phase space. For the standard form $\omega =d{q}^i\wedge d{p}_i$ on Euclidean phase space, closedness is obvious because the coefficients are constant.

The combination of non-degeneracy and closedness makes $\omega$ a symplectic form, and a manifold equipped with such a form is a symplectic manifold. The fundamental theorem about symplectic manifolds — Darboux's theorem — says something remarkable: every symplectic manifold looks locally the same. There are no local invariants. Around any point, you can always find coordinates $(q^1,\dots ,q^n,p_1,\dots ,p_n)$ in which $\omega$ takes the standard form $d{q}^i\wedge d{p}_i$.

This is very different from Riemannian geometry, where the curvature tensor provides genuine local invariants — you can detect the difference between a flat plane and a sphere by measurements in an arbitrarily small neighbourhood. In symplectic geometry, there is no symplectic curvature. A symplectic manifold has no local geometry beyond the standard pairing. All the interesting structure is global. This is why symplectic geometry is sometimes called "topology with a symplectic flavour" — the rigidity comes from global properties, not local ones.

Visual Beginner

Figure: A closed curve in the $(q,p)$ plane at time $t=0$ (solid blue) and its evolved image at time $t=T$ (dashed orange) under Hamiltonian flow. The two curves enclose the same area. The curve can deform — stretching along $q$ while compressing along $p$, or vice versa — but the area $\text{Area}=dq\wedge dp$ is invariant. This is Liouville's theorem in its simplest form: the symplectic area enclosed by any loop is a constant of the motion.

Figure: The Darboux theorem illustrated. On the left, an abstract 2-dimensional symplectic manifold with a point $x$ highlighted. On the right, a neighbourhood of $x$ mapped to the standard $(q,p)$ plane by the Darboux chart, with the symplectic form in the chart being exactly $dq\wedge dp$. The map is not an isometry (there is no metric) and not unique, but it always exists.

Worked example Beginner

Consider the harmonic oscillator with Hamiltonian $H=p^2/(2m)+\frac{1}{2}k{q}^2$. The phase space is the $(q,p)$ plane, and the symplectic form is $\omega =dq\wedge dp$.

Hamilton's equations are $\dot{q}=p/m$ and $\dot{p}=-kq$. These define a vector field on the plane: at each point $(q,p)$, the flow moves with velocity $(\dot{q},\dot{p})=(p/m,-kq)$. The solutions are ellipses: $q(t)=A\cos ({\omega }_{0}t+\varphi )$, $p(t)=-mA{\omega }_0\sin ({\omega }_{0}t+\varphi )$, where ${\omega }_0=\sqrt{k/m}$.

Take a rectangle in the $(q,p)$ plane with corners at $(0,0)$, $(q_0,0)$, $(q_0,p_0)$, $(0,p_0)$. Its area is $q_0{p}_0$. Under the Hamiltonian flow, each corner moves along its own elliptical orbit. At time $t=\pi /(2{\omega }_0)$, the point $(q_0,0)$ has moved to $(0,-m{\omega }_0{q}_0)$ and the point $(0,p_0)$ has moved to $(p_0/(m{\omega }_0),0)$. The rectangle has rotated and deformed into a new shape, but the enclosed area is still $q_0{p}_0$ — the same as at $t=0$.

This area preservation is not a coincidence. It holds for any Hamiltonian system, not just the harmonic oscillator. The symplectic form $\omega$ guarantees it. In fact, the area of any region evolved under any Hamiltonian flow is constant in time. This is Liouville's theorem, and it is a direct consequence of the symplectic structure of phase space.

Check your understanding Beginner

Formal definition Intermediate+

Let $M$ be a smooth manifold of dimension $2n$. A symplectic form on $M$ is a differential 2-form $\omega \in {\Omega }^2(M)$ satisfying two conditions:

1. Closedness: $d\omega =0$.
2. Non-degeneracy: for every $x\in M$ and every nonzero tangent vector $v\in T_{x}M$, there exists $w\in T_{x}M$ such that ${\omega }_x(v,w)\ne 0$.

Equivalently, non-degeneracy says that the linear map ${\omega }_x^{♭}:T_{x}M\to T_x^{\ast }M$ defined by $v\mapsto {\omega }_x(v,\cdot )$ is an isomorphism at every point. The pair $(M,\omega )$ is called a symplectic manifold.

The canonical example is the cotangent bundle $T^{\text{astQ}}$ of any smooth manifold $Q$ with local coordinates $(q^1,\dots ,q^n)$ on $Q$ and conjugate momenta $(p_1,\dots ,p_n)$ on the fibres. The canonical symplectic form is

$$\boxed{\omega =d{q}^i\wedge d{p}_i=d{q}^1\wedge d{p}_1+\cdots +d{q}^n\wedge d{p}_n.}$$

In matrix form, with the ordered basis $(d{q}^1,\dots ,d{q}^n,d{p}_1,\dots ,d{p}_n)$, the symplectic form is represented by the $2n\times 2n$ matrix

$$J=\left(\begin{array}{cc}0& -I_n\\ I_n& 0\end{array}\right),$$

where $I_n$ is the $n\times n$ identity matrix. This is the standard symplectic matrix. Non-degeneracy is immediate: $\det J=1$. Closedness is immediate: the coefficients of $\omega$ in Darboux coordinates are constant, so $d\omega =0$.

The Hamiltonian vector field

Given a symplectic manifold $(M,\omega )$ and a smooth function $H:M\to \mathbb{R}$ (the Hamiltonian), the Hamiltonian vector field $X_H$ is defined by the equation

$$\boxed{\omega (X_H,\cdot )=dH,}$$

or equivalently ${\iota }_{X_H}\omega =dH$, where $\iota$ denotes interior product. Non-degeneracy of $\omega$ guarantees that $X_H$ exists and is unique: the isomorphism ${\omega }^{♭}:TM\to T^{\text{astM}}$ maps $X_H$ to $dH$.

In Darboux coordinates $(q^i,p_i)$ on $T^{\text{astQ}}$, the equation $\omega (X_H,\cdot )=dH$ gives

$$X_H=\frac{\partial H}{\partial p_i}\frac{\partial }{\partial q^i}-\frac{\partial H}{\partial q^i}\frac{\partial }{\partial p_i}.$$

The integral curves of $X_H$ satisfy Hamilton's equations ${\dot{q}}^i=\partial H/\partial p_i$, ${\dot{p}}_i=-\partial H/\partial q^i$. The symplectic form is therefore the geometric object that turns a Hamiltonian function into a system of equations of motion.

The Poisson bracket from the symplectic form

The symplectic form induces a bilinear operation on smooth functions. Given $f,g\in C^{\infty }(M)$, define

$$\{f,g\}:=\omega (X_f,X_g)=X_f[g]=-X_g[f].$$

In Darboux coordinates this becomes the familiar Poisson bracket

$$\{f,g\}=\frac{\partial f}{\partial q^i}\frac{\partial g}{\partial p_i}-\frac{\partial f}{\partial p_i}\frac{\partial g}{\partial q^i}.$$

The Jacobi identity for the Poisson bracket follows from the closedness of $\omega$ ($d\omega =0$). This is the first indication that closedness is not just a technical condition — it is the condition that makes the Poisson bracket a Lie bracket, and hence makes the algebra of observables a Lie algebra.

Canonical transformations and symplectomorphisms

A symplectomorphism is a diffeomorphism $\Phi :(M,\omega )\to (M,\omega )$ that preserves the symplectic form: ${\Phi }^{\ast }\omega =\omega$. In classical mechanics, symplectomorphisms are called canonical transformations 09.05.01. The Hamiltonian flow ${\Phi }_t$ generated by any Hamiltonian function $H$ is a one-parameter family of symplectomorphisms:

$$\frac{d}{dt}{\Phi }_t^{\ast }\omega ={\Phi }_t^{\ast }(L_{X_H}\omega )={\Phi }_t^{\ast }(d{\iota }_{X_H}\omega +{\iota }_{X_H}d\omega )={\Phi }_t^{\ast }(d(dH)+0)=0.$$

This is the geometric proof that Hamiltonian flow preserves $\omega$. Since ${\omega }^n/n!$ is the Liouville volume form on a $2n$-dimensional phase space, preservation of $\omega$ implies preservation of phase-space volume — Liouville's theorem.

Counterexamples to common slips

• A symplectic manifold must be even-dimensional. Non-degeneracy of $\omega$ means ${\omega }^n$ is nowhere vanishing, which requires $\dim M=2n$. There is no symplectic structure on an odd-dimensional manifold.

• Non-degeneracy is not the same as positive-definiteness. A Riemannian metric is a non-degenerate symmetric 2-tensor; a symplectic form is a non-degenerate antisymmetric 2-form. Symplectic geometry has no notion of length or angle. It measures oriented areas, not distances.

• Closedness is essential. A non-degenerate 2-form that is not closed is called an "almost symplectic" structure. Without closedness, the Poisson bracket need not satisfy the Jacobi identity, and the Hamiltonian flow need not preserve the form. Closedness is what makes the machinery of canonical transformations and conservation laws work.

Key theorem with proof Intermediate+

Theorem (Darboux). Let $(M,\omega )$ be a $2n$-dimensional symplectic manifold. For any point $x\in M$, there exists a coordinate neighbourhood with coordinates $(q^1,\dots ,q^n,p_1,\dots ,p_n)$ centred at $x$ in which

$$\omega =\sum _{i=1}^{n}d{q}^i\wedge d{p}_i.$$

These coordinates are called Darboux coordinates (or canonical coordinates). Moreover, the coordinates are not unique — any symplectomorphism of the Darboux chart produces another valid Darboux chart.

The significance is profound: symplectic manifolds have no local invariants. Every symplectic manifold of dimension $2n$ looks the same in a sufficiently small neighbourhood of any point. This contrasts sharply with Riemannian geometry, where the curvature tensor provides genuine local information.

Proof (Moser's homotopy method, 1965). Choose any local coordinates $(x^1,\dots ,x^{2n})$ centred at $x$ such that ${\omega }_x=0$ is satisfied at the origin and $\omega$ takes the standard form at $x$ itself: $\omega (x)={\sum }_{i=1}^{n}d{x}^i\wedge d{x}^{n+i}$ (this can always be arranged by a linear change of coordinates). Define the constant 2-form ${\omega }_0={\sum }_{i=1}^{n}d{x}^i\wedge d{x}^{n+i}$ and consider the family ${\omega }_t=(1-t){\omega }_0+t\omega$ for $t\in [0,1]$.

Since ${\omega }_0$ and $\omega$ agree at $x$, the form ${\omega }_t$ is non-degenerate in a neighbourhood of $x$ for all $t\in [0,1]$ (non-degeneracy is an open condition). Compute $d{\omega }_t/dt=\omega -{\omega }_0$. Both $\omega$ and ${\omega }_0$ are closed, so $d(\omega -{\omega }_0)=0$. By the Poincare lemma (on a contractible neighbourhood), there exists a 1-form $\alpha$ with $d\alpha =\omega -{\omega }_0$. Define the time-dependent vector field $X_t$ by ${\iota }_{X_t}{\omega }_t=-\alpha$ (non-degeneracy guarantees this is well-defined). Let ${\Phi }_t$ be the flow of $X_t$. Then

$$\frac{d}{dt}({\Phi }_t^{\ast }{\omega }_t)={\Phi }_t^{\ast }\left(L_{X_t}{\omega }_t+\frac{d{\omega }_t}{dt}\right)={\Phi }_t^{\ast }(d{\iota }_{X_t}{\omega }_t+d\alpha )={\Phi }_t^{\ast }(-d\alpha +d\alpha )=0.$$

So ${\Phi }_t^{\ast }{\omega }_t$ is constant: ${\Phi }_1^{\ast }\omega ={\Phi }_0^{\ast }{\omega }_0={\omega }_0$. The map ${\Phi }_1$ is the desired Darboux coordinate change. $\square$

Worked example: the 2-sphere

The 2-sphere $S^2$ with its standard area form $\omega =\sin \theta d\theta \wedge d\phi$ is a symplectic manifold. This is a 2-dimensional symplectic form: closed (all 2-forms on a 2-manifold are closed) and non-degenerate ($\sin \theta \ne 0$ away from the poles).

Darboux's theorem says we can find coordinates $(q,p)$ in which $\omega =dq\wedge dp$. Define $p=-\cos \theta$ (so $dp=\sin \theta d\theta$) and $q=\phi$. Then $\omega =\sin \theta d\theta \wedge d\phi =dp\wedge dq=-dq\wedge dp$, which is the standard symplectic form up to an overall sign (absorbed by relabelling). The Darboux coordinates $(q,p)=(\phi ,-\cos \theta )$ cover the sphere minus the poles, which is the maximal domain for any Darboux chart — no single Darboux chart can cover the entire sphere because $S^2$ is compact.

This example illustrates two points: (a) symplectic manifolds need not be cotangent bundles (the sphere is not a cotangent bundle of any manifold), and (b) the Darboux chart is generally not global.

Exercises Intermediate+

Lean formalization Intermediate+

lean_status: none. Mathlib has Mathlib.Geometry.Symplectic which defines symplectic forms as closed non-degenerate 2-forms, provides the symplectic group $\mathrm{Sp}(2n)$, symplectic linear maps, and the symplectic linear algebra infrastructure (Lagrangian subspaces, isotropic subspaces). What is missing for the physical Hamiltonian-mechanics application is: (a) the correspondence between a Hamiltonian function $H$ and its vector field $X_H$ via the interior-product equation ${\iota }_{X_H}\omega =dH$, (b) the proof that the flow of $X_H$ preserves $\omega$ (Cartan's homotopy formula applied to the symplectic form), (c) Darboux's theorem with its Moser-trick proof, (d) the induced Poisson bracket $\{f,g\}=\omega (X_f,X_g)$ and the proof that it satisfies the Jacobi identity (which requires $d\omega =0$), and (e) the identification of symplectomorphisms with canonical transformations in the sense of classical mechanics. These are the key results that connect the algebraic symplectic theory already in Mathlib to the dynamical systems of Hamiltonian mechanics.

Advanced results Master

Symplectomorphisms and the symplectic group

A symplectomorphism $\Phi :(M_1,{\omega }_1)\to (M_2,{\omega }_2)$ is a diffeomorphism satisfying ${\Phi }^{\ast }{\omega }_2={\omega }_1$. The group of symplectomorphisms of $(M,\omega )$ is denoted $\mathrm{Symp}(M,\omega )$. It is an infinite-dimensional Lie group whose Lie algebra consists of Hamiltonian vector fields (modulo constants, since $H$ and $H+c$ define the same $X_H$ on a connected manifold).

At the linear level, the symplectic group $\mathrm{Sp}(2n,\mathbb{R})$ consists of matrices $A$ satisfying $A^{T}JA=J$. This is the linear symplectic group — the symmetry group of the standard symplectic form on ${\mathbb{R}}^{2n}$. It is a non-compact Lie group of dimension $n(2n+1)$. Key properties:

• $\mathrm{Sp}(2n,\mathbb{R})\subset \mathrm{GL}(2n,\mathbb{R})$ is defined by the constraint $A^{T}JA=J$, which gives $n(2n+1)$ independent equations on the $4n^2$ entries of $A$, leaving $n(2n+1)$ free parameters.
• $\mathrm{Sp}(2,\mathbb{R})=\mathrm{SL}(2,\mathbb{R})$ — in two dimensions, area-preserving linear maps are exactly symplectic linear maps.
• Every symplectic matrix has determinant $+1$: $\mathrm{Sp}(2n,\mathbb{R})\subset \mathrm{SL}(2n,\mathbb{R})$. Symplectic maps are volume-preserving, but the converse is false for $n\ge 2$.
• The eigenvalues of $A\in \mathrm{Sp}(2n,\mathbb{R})$ come in reciprocal pairs: if $\lambda$ is an eigenvalue, so is ${\lambda }^{-1}$. This constrains the dynamics of linear Hamiltonian systems.

The relationship between the linear and nonlinear theories is that the derivative $D_x\Phi$ of a symplectomorphism $\Phi$ at any point $x$ is a symplectic linear map: $D_x\Phi \in \mathrm{Sp}(2n,\mathbb{R})$.

Moser's trick and its consequences

The proof of Darboux's theorem given above uses Moser's homotopy method, introduced by Jurgen Moser in 1965. The method has become one of the central tools of symplectic geometry. The idea is to interpolate between two structures (e.g., $\omega$ and ${\omega }_0$) along a path, and construct a time-dependent vector field whose flow carries one to the other.

Three major applications:

1. Darboux's theorem (local normal form): the homotopy ${\omega }_t=(1-t){\omega }_0+t\omega$ between the standard form and a given symplectic form, on a small neighbourhood.

2. Moser's stability theorem (global rigidity on compact manifolds): if ${\omega }_t$ is a family of cohomologous symplectic forms on a compact manifold, all non-degenerate along the path, then there is an isotopy ${\Phi }_t$ with ${\Phi }_t^{\ast }{\omega }_t={\omega }_0$. This shows that the symplectic structure on a compact manifold cannot be deformed within a fixed cohomology class.

3. Weinstein's neighbourhood theorem: a compact Lagrangian submanifold $L$ in a symplectic manifold $(M,\omega )$ has a neighbourhood symplectomorphic to a neighbourhood of the zero section in $(T^{\text{astL}},{\omega }_{\mathrm{can}})$. This is proved by a Moser-type argument comparing the given symplectic form with the pullback of the canonical form on $T^{\text{astL}}$.

The unifying principle is that non-degeneracy (an open condition) allows one to solve for a vector field from a 1-form, and closedness ($d\omega =0$) ensures the resulting flow preserves the form. This is the symplectic analogue of the Poincare lemma for volume forms.

Coadjoint orbits and the Kirillov-Kostant-Souriau form

Let $G$ be a Lie group with Lie algebra $\mathfrak{g}$ and dual ${\mathfrak{g}}^{\ast }$. The coadjoint action of $G$ on ${\mathfrak{g}}^{\ast }$ is ${\mathrm{Ad}}_g^{\ast }(\mu )=\mu \circ {\mathrm{Ad}}_{g^{-1}}$ for $\mu \in {\mathfrak{g}}^{\ast }$. The orbits of this action, called coadjoint orbits, carry a natural symplectic structure.

For $\mu \in {\mathfrak{g}}^{\ast }$ and $\xi \in \mathfrak{g}$, the infinitesimal coadjoint action is ${\xi }_{{\mathfrak{g}}^{\ast }}(\mu )=-{\mathrm{ad}}_{\xi }^{\ast }(\mu )=\mu \circ {\mathrm{ad}}_{\xi }$. Define the Kirillov-Kostant-Souriau symplectic form on the coadjoint orbit ${\mathcal{O}}_{\mu }$ by

$${\omega }_{\mu }({\xi }_{{\mathfrak{g}}^{\ast }},{\eta }_{{\mathfrak{g}}^{\ast }})=\mu ([\xi ,\eta ])$$

for $\xi ,\eta \in \mathfrak{g}$, evaluated at the point $\mu$. This is well-defined (independent of the choice of representatives), closed (a general fact about homogeneous symplectic forms), and non-degenerate (the kernel of ${\omega }_{\mu }$ at $\mu$ is the annihilator of ${\mathfrak{g}}_{\mu }$, the stabiliser Lie algebra, which is zero by the orbit-stabiliser theorem).

Coadjoint orbits are the natural phase spaces for many mechanical systems. The rigid body, for example, has its phase space identified with a coadjoint orbit of $\mathrm{SO}(3)$ in $\mathfrak{so}(3{)}^{\ast }\cong {\mathbb{R}}^3$ — the sphere $|\mathbf{L}{|}^2=\mathrm{const}$ in angular-momentum space. The Kirillov-Kostant-Souriau form on this sphere is (up to scale) the standard area form, and the Euler equations for the rigid body are Hamilton's equations on this coadjoint orbit.

The orbit method in representation theory (Kirillov 1962) relates the irreducible unitary representations of $G$ to the coadjoint orbits of $G$: each orbit corresponds to a representation, and the symplectic structure on the orbit determines the representation's character. This is one of the deepest connections between symplectic geometry and pure mathematics.

Marsden-Weinstein reduction

The Marsden-Weinstein reduction theorem (1974) provides a systematic way to reduce the dimension of a Hamiltonian system by exploiting symmetries. Let $(M,\omega )$ be a symplectic manifold with a Hamiltonian $G$-action and an equivariant momentum map $\mathbf{J}:M\to {\mathfrak{g}}^{\ast }$. For a regular value $\mu \in {\mathfrak{g}}^{\ast }$, define:

1. The level set ${\mathbf{J}}^{-1}(\mu )\subset M$.
2. The reduced space $M_{\mu }={\mathbf{J}}^{-1}(\mu )/G_{\mu }$, where $G_{\mu }$ is the stabiliser of $\mu$.

Theorem (Marsden-Weinstein). $M_{\mu }$ is a symplectic manifold with symplectic form ${\omega }_{\mu }$ defined by ${\pi }^{\ast }{\omega }_{\mu }={\iota }^{\ast }\omega$, where $\iota :{\mathbf{J}}^{-1}(\mu )\hookrightarrow M$ is the inclusion and $\pi :{\mathbf{J}}^{-1}(\mu )\to M_{\mu }$ is the quotient map.

The dimension formula is $\dim M_{\mu }=\dim M-2\dim G_{\mu }$. Each symmetry reduces the phase-space dimension by twice the dimension of the symmetry group (or its stabiliser).

The standard example is the reduction of the $N$-body problem by translational symmetry. The total linear momentum $\mathbf{P}={\sum }_i{\mathbf{p}}_i$ is the momentum map for the action of ${\mathbb{R}}^3$ by translations. Fixing $\mathbf{P}=0$ and quotienting by translations reduces the phase space from $6N$ dimensions to $6N-6$ dimensions — the centre-of-mass motion is removed, leaving only the relative coordinates and momenta.

For the rigid body, Marsden-Weinstein reduction with $G=\mathrm{SO}(3)$ acting on $T^{\ast }\mathrm{SO}(3)$ with momentum map $\mathbf{J}=\mathbf{L}$ (angular momentum) and $\mu ={\mathbf{L}}_0$ fixed, reduces the 6-dimensional cotangent bundle to a 2-dimensional coadjoint orbit — the sphere $|\mathbf{L}{|}^2=|{\mathbf{L}}_0{|}^2$ in ${\mathbb{R}}^3$. The reduced dynamics are the Euler equations on this sphere.

Symplectic topology and Gromov's non-squeezing theorem

The absence of local invariants (Darboux's theorem) might suggest that symplectic manifolds have no rigidity at all. Gromov's 1985 non-squeezing theorem shows this is spectacularly false at the global level.

Theorem (Gromov's non-squeezing theorem). Let $(M,\omega )$ be a symplectic manifold. There exists a symplectic embedding of the ball $B^{2n}(r)$ into the cylinder $B^2(R)\times {\mathbb{R}}^{2n-2}$ if and only if $r\le R$.

In plain language: you cannot symplectically squeeze a ball of radius $r$ into a cylinder of smaller radius $R<r$, even though you can easily do so as a volume-preserving map. The constraint comes not from volume but from the symplectic structure — specifically, from the fact that symplectic maps must preserve the area of certain 2-dimensional projections.

This theorem introduced the notion of symplectic capacity: a function $c$ assigning to each symplectic manifold $(M,\omega )$ a non-negative number (or infinity) satisfying:

1. Monotonicity: if there is a symplectic embedding $(M_1,{\omega }_1)\hookrightarrow (M_2,{\omega }_2)$, then $c(M_1)\le c(M_2)$.
2. Conformality: $c(M,\alpha \omega )=|\alpha |c(M,\omega )$.
3. Normalisation: $c(B^{2n}(1))=\pi =c(Z^{2n}(1))$, where $Z^{2n}(R)=B^2(R)\times {\mathbb{R}}^{2n-2}$ is the cylinder.

Gromov's theorem is equivalent to the statement that symplectic capacities exist and are non-vacuous (i.e., $c(B^{2n}(1))=c(Z^{2n}(1))$). The first symplectic capacity, the Gromov width, is $w_G(M)=\sup \{\pi r^2:B^{2n}(r)\hookrightarrow M\text{ symplectically}\}$.

The non-squeezing theorem has a physical interpretation due to Hermann and Uhlmann. For a classical system, the symplectic form measures the area in each canonical $(q^i,p_i)$ plane. Gromov's theorem says that no Hamiltonian evolution can reduce this area below a certain threshold. In the semiclassical limit, this area is quantised in units of Planck's constant $\hslash$ — the non-squeezing theorem is the classical precursor of the uncertainty principle $\Delta q^i\Delta p_i\ge \hslash /2$.

Connections to quantum mechanics

The symplectic form is the classical structure that underlies the canonical commutation relations of quantum mechanics 12.02.01. The correspondence is:

Classical	Quantum
Phase space $(M,\omega )$	Hilbert space $\mathcal{H}$
Observable $f\in C^{\infty }(M)$	Self-adjoint operator $\hat{f}$
Poisson bracket $\{f,g\}$	Commutator $[\hat{f},\hat{g}]/(i\hslash )$
Symplectomorphism ${\Phi }^{\ast }\omega =\omega$	Unitary transformation $U^{\text{astU}}=I$
Hamiltonian flow ${\Phi }_t$	Unitary evolution $e^{-iHt/\hslash }$

The Poisson bracket $\{f,g\}=\omega (X_f,X_g)$ satisfies the Jacobi identity because $d\omega =0$. The canonical commutation relation $[{\hat{q}}^i,{\hat{p}}_j]=i\hslash {\delta }_j^i$ is the quantisation of $\{q^i,p_j\}={\delta }_j^i$. The symplectic form defines the classical bracket structure; quantisation promotes this bracket to a commutator of operators.

Geometric quantisation (Kostant 1970, Souriau 1970) constructs the Hilbert space directly from the symplectic manifold by choosing a compatible complex structure (a "polarisation") and defining quantum states as square-integrable sections of a line bundle $L\to M$ with curvature $\omega /\hslash$. The prequantum line bundle exists if and only if $[\omega ]/2\pi \hslash \in H^2(M,\mathbb{Z})$ — the cohomology class of the symplectic form must be integral (in units of $\hslash$). This is the origin of the quantisation of charge, flux, and other topological quantities.

The deformation quantisation programme (Bayen, Flato, Fronsdal, Lichnerowicz, Sternheimer 1978) takes a different route: it deforms the commutative algebra $C^{\infty }(M)[[\hslash ]]$ by replacing pointwise multiplication with a non-commutative star product $f\star g=fg+(i\hslash /2)\{f,g\}+O({\hslash }^2)$. The symplectic form provides the leading-order deformation. Kontsevich's 1997 proof of the formality conjecture shows that such a star product exists on any Poisson manifold (not just symplectic), establishing deformation quantisation in full generality and earning him the Fields Medal.

Lagrangian submanifolds

A Lagrangian submanifold $L$ of a $2n$-dimensional symplectic manifold $(M,\omega )$ is an $n$-dimensional submanifold on which $\omega$ vanishes: $\omega {|}_L=0$. The dimension is maximal: any submanifold on which $\omega$ vanishes has dimension at most $n$, and if it has dimension exactly $n$, it is Lagrangian.

Examples:

• The zero section $\{(q,0)\}\subset T^{\text{astQ}}$ is Lagrangian: $\omega =d{q}^i\wedge d{p}_i$ restricts to zero because $p=0$.
• The graph of a closed 1-form $\alpha$: $L_{\alpha }=\{(q,\alpha (q))\}\subset T^{\text{astQ}}$ is Lagrangian if and only if $d\alpha =0$. The function $S$ with $\alpha =dS$ is the generating function of the Lagrangian.
• The diagonal $\Delta =\{(q,q)\}\subset {\mathbb{R}}^n\times {\mathbb{R}}^n$ with the symplectic form $\omega =d{q}_1^i\wedge d{p}_{1,i}-d{q}_2^i\wedge d{p}_{2,i}$ is Lagrangian in the product.

Lagrangian submanifolds are the symplectic geometric objects that correspond to states in quantum mechanics (via the WKB approximation) and to solutions of the Hamilton-Jacobi equation 09.04.02. The intersection theory of Lagrangian submanifolds (Floer homology, Fukaya categories) is one of the most active areas of modern symplectic topology.

Synthesis. The symplectic form $\omega$ is the fundamental geometric structure of Hamiltonian mechanics. It defines the Hamiltonian vector field $X_H$ via ${\iota }_{X_H}\omega =dH$, the Poisson bracket via $\{f,g\}=\omega (X_f,X_g)$, and the volume form via ${\omega }^n/n!$. Darboux's theorem ensures that this structure has no local invariants — all symplectic manifolds look locally like standard phase space — while Gromov's non-squeezing theorem reveals profound global rigidity. The Marsden-Weinstein reduction theorem and the theory of coadjoint orbits connect symplectic geometry to symmetry reduction in mechanics and to representation theory in pure mathematics. The quantisation correspondence — Poisson bracket to commutator, symplectomorphism to unitary map, Lagrangian submanifold to quantum state — places the symplectic form at the boundary between classical and quantum physics.

Full proof set Master

Proposition 1 (Closedness implies the Jacobi identity for the Poisson bracket). Let $(M,\omega )$ be a symplectic manifold and $\{f,g\}=\omega (X_f,X_g)$ the induced Poisson bracket. If $d\omega =0$, then $\{f,\{g,h\}\}+\{g,\{h,f\}\}+\{h,\{f,g\}\}=0$ for all $f,g,h\in C^{\infty }(M)$.

Proof. The Jacobiator is

$$J(f,g,h)=\{f,\{g,h\}\}+\text{cyclic}=X_f[\{g,h\}]+X_g[\{h,f\}]+X_h[\{f,g\}].$$

Using $\{g,h\}=X_g[h]$, the first term is $X_f[X_g[h]]$. Similarly, the second is $X_g[X_h[f]]$ and the third is $X_h[X_f[g]]$. Now use the identity $[X_f,X_g]=X_{\{f,g\}}$ (the map $f\mapsto X_f$ is a Lie algebra homomorphism), which follows from $d\omega =0$:

$${\iota }_{[X_f,X_g]}\omega =L_{X_f}{\iota }_{X_g}\omega -{\iota }_{X_g}{L}_{X_f}\omega =d(X_f[{\iota }_{X_g}\omega ])-{\iota }_{X_g}d({\iota }_{X_f}\omega )=d(X_f[dg])-{\iota }_{X_g}d(df).$$

Since $d(df)=0$ and $X_f[dg]=\{f,g\}$: ${\iota }_{[X_f,X_g]}\omega =d\{f,g\}={\iota }_{X_{\{f,g\}}}\omega$. Non-degeneracy gives $[X_f,X_g]=X_{\{f,g\}}$.

Now $X_f[X_g[h]]-X_g[X_f[h]]=[X_f,X_g][h]=X_{\{f,g\}}[h]=\{\{f,g\},h\}$. So

$$J(f,g,h)=X_f[X_g[h]]+X_g[X_h[f]]+X_h[X_f[g]].$$

Permuting the identity $X_f[X_g[h]]-X_g[X_f[h]]=\{\{f,g\},h\}$ cyclically and adding:

$$J(f,g,h)=\frac{1}{2}\left(\{\{f,g\},h\}+\{\{g,h\},f\}+\{\{h,f\},g\}\right)+\frac{1}{2}J(f,g,h).$$

This self-referential equation gives $J=0$ if and only if $J=0$, so a direct proof is needed. Instead, use the coordinate expression: in Darboux coordinates, $\{f,g\}=(\partial f/\partial q^i)(\partial g/\partial p_i)-(\partial f/\partial p_i)(\partial g/\partial q^i)$. The Jacobi identity follows by direct computation — all second-derivative terms cancel pairwise. This works precisely because the coefficients of the Poisson bivector are constant in Darboux coordinates, which is a consequence of $d\omega =0$ via Darboux's theorem. $\square$

Proposition 2 (Linear Darboux theorem). Let $V$ be a $2n$-dimensional real vector space equipped with a non-degenerate antisymmetric bilinear form $\omega$. Then there exists a basis $\{e_1,\dots ,e_n,f_1,\dots ,f_n\}$ of $V$ such that $\omega (e_i,f_j)={\delta }_{ij}$, $\omega (e_i,e_j)=0$, $\omega (f_i,f_j)=0$.

Proof. Pick any nonzero $e_1\in V$. Since $\omega$ is non-degenerate, there exists $f_1$ with $\omega (e_1,f_1)=1$. Let $W=\mathrm{span}\{e_1,f_1\}$ and $W^{\perp }=\{v:\omega (v,w)=0\text{ for all }w\in W\}$. Non-degeneracy implies $V=W\oplus W^{\perp }$ and $\omega {|}_{W^{\perp }}$ is non-degenerate. By induction on $n$, the result follows. $\square$

This is the linear algebra prerequisite for Darboux's theorem: at each point $x\in M$, the tangent space $T_{x}M$ is a symplectic vector space, and Proposition 2 gives a symplectic basis. The nonlinear content of Darboux's theorem is that this linear normal form can be extended to a neighbourhood, which requires the Moser argument (closedness $d\omega =0$).

Proposition 3 (Energy conservation from symplectic structure). Let $(M,\omega )$ be a symplectic manifold and $H\in C^{\infty }(M)$ a Hamiltonian. Then $H$ is constant along the flow of $X_H$.

Proof. $dH/dt=X_H[H]=\{H,H\}=\omega (X_H,X_H)=0$ by antisymmetry of $\omega$. $\square$

This is the most economical proof of energy conservation in Hamiltonian mechanics: it requires only the antisymmetry of the symplectic form. No coordinate calculation, no chain rule, no assumption about the form of $H$. Energy conservation is a structural consequence of the symplectic framework.

Proposition 4 (Weinstein's Lagrangian neighbourhood theorem). Let $(M,\omega )$ be a symplectic manifold and $L\subset M$ a compact Lagrangian submanifold. There exists a neighbourhood $U$ of $L$ in $M$ and a symplectomorphism $\Phi :U\to V\subset T^{\text{astL}}$ where $V$ is a neighbourhood of the zero section, such that $\Phi {|}_L$ is the identification of $L$ with the zero section.

Proof (sketch). The key step is to construct a diffeomorphism from a tubular neighbourhood of $L$ to a neighbourhood of the zero section in $T^{\text{astL}}$. The normal bundle of $L$ in $M$ is naturally isomorphic to $T^{\text{astL}}$ because $L$ is Lagrangian: for $v\in T_{x}L$ and $n\in N_{x}L$ (the normal space), the map $n\mapsto \omega (n,\cdot ){|}_{T_{x}L}$ identifies $N_{x}L\cong T_x^{\text{astL}}$. This gives a diffeomorphism $\Psi$ from a tubular neighbourhood to $T^{\text{astL}}$. The symplectic form pulls back to ${\Psi }^{\ast }\omega$ on the tubular neighbourhood, which agrees with the canonical form ${\omega }_{\mathrm{can}}$ on the zero section. A Moser argument (interpolating between ${\Psi }^{\ast }\omega$ and ${\omega }_{\mathrm{can}}$, which are cohomologous on the neighbourhood) produces the symplectomorphism $\Phi$. $\square$

Connections Master

• Legendre transform 09.04.01 — the Legendre transform $\mathbb{F}L:TQ\to T^{\text{astQ}}$ identifies the Lagrangian symplectic structure on $TQ$ with the canonical symplectic structure on $T^{\text{astQ}}$. The symplectic form $\omega =d{q}^i\wedge d{p}_i$ is the pullback of the Lagrangian symplectic form by the inverse Legendre transform.

• Hamilton's equations 09.04.02 — Hamilton's equations ${\dot{q}}^i=\partial H/\partial p_i$, ${\dot{p}}_i=-\partial H/\partial q^i$ are the integral curves of the Hamiltonian vector field $X_H$ defined by ${\iota }_{X_H}\omega =dH$. The symplectic form is the geometric structure that converts a function $H$ into a dynamical system.

• Canonical transformations 09.05.01 — canonical transformations are symplectomorphisms: diffeomorphisms $\Phi$ with ${\Phi }^{\ast }\omega =\omega$. The generating-function machinery of canonical transformations is a computational technique for constructing symplectomorphisms.

• Liouville's theorem 09.04.06 — phase-space volume preservation follows from symplectic preservation (${\Phi }^{\ast }\omega =\omega$ implies ${\Phi }^{\ast }({\omega }^n)={\omega }^n$). Liouville's theorem is a corollary of the symplectic structure, not an independent result.

• Geometric mechanics 09.09.01 pending — the symplectic form is the foundation on which the entire edifice of geometric mechanics is built: the momentum map, Marsden-Weinstein reduction, coadjoint orbits, and the Hamiltonian description of field theories all require the symplectic framework.

• Quantum mechanics 12.02.01 — the canonical commutation relations are the quantisation of the Poisson bracket, which is defined by the symplectic form. Geometric quantisation, deformation quantisation, and the path integral all take the symplectic manifold as their classical starting point.

• Symplectic topology — Gromov's non-squeezing theorem, Floer homology, and the Arnold conjecture on fixed points of symplectomorphisms are the central results of symplectic topology, a field that studies the global rigidity properties of symplectic manifolds. The Arnold conjecture (proved in many cases by Floer 1988 and later by Fukaya-Ono 1999) states that a Hamiltonian symplectomorphism on a compact symplectic manifold has at least as many fixed points as a Morse function has critical points — a deep link between symplectic dynamics and topology.

Historical and philosophical context Master

The symplectic structure of mechanics has roots in the work of Lagrange, who in 1808 discovered that the Poisson bracket of two functions satisfies the identity that now bears his name (the Jacobi identity for the Poisson bracket was discovered independently by Jacobi around 1840). Lagrange did not have the language of differential forms, but his calculation showed that the bracket structure $\{f,g\}$ has a special algebraic property that makes it compatible with the equations of motion.

Poincare, in his landmark 1899 paper "Les methodes nouvelles de la mecanique celeste," introduced the integral invariants that bear his name and recognised that Hamiltonian flows preserve a 2-form — the first statement of symplectic preservation, though Poincare used the language of integral invariants rather than differential forms. The Poincare-Cartan integral invariant ${\oint }_{\gamma }{p}_{i}d{q}^i=\mathrm{const}$ along any closed curve evolved by the Hamiltonian flow is equivalent to the statement that the flow is a symplectomorphism.

Elie Cartan, in his 1922 book "Lecons sur les invariants integraux," gave the modern formulation using exterior differential calculus. Cartan introduced the Liouville (Poincare-Cartan) 1-form $\theta =p_{i}d{q}^i$ and showed that its exterior derivative $\omega =-d\theta =d{q}^i\wedge d{p}_i$ is preserved by the Hamiltonian flow. Cartan's work established the symplectic form as a geometric object and the exterior calculus as the natural language for mechanics.

The term "symplectic" was coined by Hermann Weyl in his 1939 book "The Classical Groups" as a replacement for "complex" (from the Latin complexus = intertwined), which by then was firmly associated with complex numbers. Weyl chose the Greek cognate "symplektikos" to preserve the etymological meaning. The mathematical study of symplectic manifolds as independent geometric objects — not just as the phase spaces of specific mechanical systems — began with the work of Kirillov, Kostant, and Souriau in the early 1960s, who identified the symplectic structure of coadjoint orbits and its connection to representation theory.

Darboux's theorem was originally proved by Gaston Darboux in 1882 in the context of Pfaffian systems (the theory of differential 1-forms). Darboux showed that a non-degenerate closed 2-form can always be brought to the standard form $d{q}^i\wedge d{p}_i$ by a coordinate change. The modern proof using Moser's homotopy method (1965) is simpler and more conceptual: it replaces Darboux's explicit iterative coordinate construction with a flow argument.

Alan Weinstein's work in the 1970s (Lagrangian neighbourhood theorem, neighbourhood theorem for isotropic submanifolds, the extension of Darboux's theorem to various geometric settings) established symplectic geometry as a field in its own right, with its own questions and techniques independent of mechanics. Marsden and Weinstein's 1974 reduction theorem connected symplectic geometry systematically to the theory of symmetry reduction in mechanics, providing a unified framework for problems from the rigid body to fluid dynamics to gauge field theory.

Mikhael Gromov's 1985 paper "Pseudoholomorphic curves in symplectic manifolds" revolutionised the field by introducing the technique of pseudoholomorphic curves and proving the non-squeezing theorem. Gromov showed that symplectic manifolds have a rigidity that cannot be detected by volume-preserving maps alone — the symplectic structure imposes constraints beyond the topological ones. This insight spawned the entire field of symplectic topology, which studies the global rigidity and flexibility properties of symplectic manifolds using techniques from algebraic topology, analysis, and partial differential equations.

The philosophical point is that the symplectic form encodes the "intertwining" of positions and momenta — the pairing that makes Hamiltonian mechanics work. This pairing is not a convention or a coordinate choice; it is a geometric structure intrinsic to the phase space of any Hamiltonian system. Darboux's theorem says this structure has no local variability (all symplectic manifolds are locally the same), while Gromov's theorem says it has global rigidity (symplectic maps cannot squeeze balls into cylinders). The interplay between this local flexibility and global rigidity is the central theme of symplectic geometry.

Bibliography Master

• Darboux, G., "Sur le probleme de Pfaff," Bull. Sci. Math. 6 (1882), 14–36, 49–68. (Original proof of Darboux's theorem.)

• Poincare, H., Les methodes nouvelles de la mecanique celeste, Vol. 3 (Gauthier-Villars, 1899). (Integral invariants, symplectic preservation.)

• Cartan, E., Lecons sur les invariants integraux (Hermann, 1922). (Exterior calculus formulation of symplectic mechanics.)

• Weyl, H., The Classical Groups: Their Invariants and Representations (Princeton University Press, 1939). (Coining of "symplectic.")

• Kirillov, A. A., "Unitary representations of nilpotent Lie groups," Russian Math. Surveys 17 (1962), 53–104. (Orbit method, coadjoint orbits.)

• Kostant, B., "Quantization and unitary representations," Lecture Notes in Math. 170 (1970), 87–208. (Geometric quantisation, Kirillov-Kostant-Souriau form.)

• Souriau, J.-M., Structure des systemes dynamiques (Dunod, 1970). (Symplectic mechanics, geometric quantisation.)

• Moser, J., "On the volume elements on a manifold," Trans. Amer. Math. Soc. 120 (1965), 286–294. (Moser's homotopy method, stability theorem.)

• Marsden, J. E. & Weinstein, A., "Reduction of symplectic manifolds with symmetry," Rep. Math. Phys. 5 (1974), 121–130. (Marsden-Weinstein reduction.)

• Weinstein, A., "Symplectic manifolds and their Lagrangian submanifolds," Advances in Math. 6 (1971), 329–346. (Lagrangian neighbourhood theorem.)

• Arnold, V. I., Mathematical Methods of Classical Mechanics, 2nd ed. (Springer GTM 60, 1989), §37–44.

• Gromov, M., "Pseudoholomorphic curves in symplectic manifolds," Invent. Math. 82 (1985), 307–347. (Non-squeezing theorem, pseudoholomorphic curves.)

• McDuff, D. & Salamon, D., Introduction to Symplectic Topology, 3rd ed. (Oxford, 2017).

• Marsden, J. E. & Ratiu, T. S., Introduction to Mechanics and Symmetry, 2nd ed. (Springer TAM 17, 1999).

• Abraham, R. & Marsden, J. E., Foundations of Mechanics, 2nd ed. (Addison-Wesley, 1978).

• Goldstein, H., Poole, C. P. & Safko, J., Classical Mechanics, 3rd ed. (Pearson, 2002), Ch. 8–9.

• Landau, L. D. & Lifshitz, E. M., Mechanics, 3rd ed. (Course of Theoretical Physics Vol. 1, Pergamon, 1976), §40–46.

• Floer, A., "Morse theory for Lagrangian intersections," J. Differential Geom. 28 (1988), 513–547. (Floer homology.)

• Kontsevich, M., "Deformation quantization of Poisson manifolds," Lett. Math. Phys. 66 (2003), 157–216. (Formality conjecture, deformation quantisation.)

• Hofer, H. & Zehnder, E., Symplectic Invariants and Hamiltonian Dynamics (Birkhauser, 1994). (Symplectic capacity theory.)