Classical Mechanics: Geometric Formalism

Classical Mechanics » Geometric Formalism

Geometric mechanics reformulates Lagrangian and Hamiltonian mechanics in coordinate-free terms: configuration spaces are manifolds, phase space is a symplectic manifold, dynamics is the flow of a vector field built from the Hamiltonian, and symmetries are Lie group actions. This page develops that language (symplectic forms, Poisson brackets, Liouville’s theorem, Lagrangian submanifolds, integrable systems, momentum maps and reduction, connections and geometric phases) and shows where each idea pays off in physics and computation. It assumes the coordinate treatment in Lagrangian & Hamiltonian Mechanics and basic familiarity with manifolds and differential forms.

Why geometry

Newton’s equations are written in a particular coordinate system. Lagrange’s and Hamilton’s work in any coordinates, which suggests that the physics lives in structures that do not depend on coordinates at all. Identifying those structures pays off in several ways:

  • Conservation laws become invariance of geometric objects under a flow.
  • Constraints become submanifolds, and symmetries become group actions that can be quotiented out.
  • The qualitative theory of dynamics (invariant tori, area-preserving maps, KAM theory, chaos) is a study of how a flow deforms geometric structure.
  • Numerical methods that preserve the geometry (symplectic integrators) outperform those that do not over long times.
  • The same structures carry over to quantum mechanics (Poisson bracket to commutator, Berry phase) and field theory (gauge connections, multisymplectic geometry).
flowchart TD
    Q["Configuration manifold Q<br/>points q"]
    Q --> TQ["Tangent bundle TQ<br/>states (q, q-dot)<br/>Lagrangian L : TQ to R"]
    Q --> TsQ["Cotangent bundle T*Q<br/>states (q, p)<br/>Hamiltonian H : T*Q to R"]
    TQ -- "Legendre transform<br/>p = dL/d(q-dot)" --> TsQ
    TsQ --> S["Symplectic manifold (M, omega)<br/>omega = -d theta"]
    S --> F["Hamiltonian flow<br/>preserves omega and volume"]
    S --> G["Lie group symmetry G<br/>momentum map J : M to g*"]
    G --> R["Reduced space J^-1(mu) / G_mu<br/>fewer degrees of freedom"]
    S -. "drop non-degeneracy" .-> P["Poisson manifold<br/>(rigid body, fluids, Lie-Poisson)"]
    R -. "often lands in" .-> P

Symplectic manifolds

The symplectic form

A symplectic manifold is a pair $(M, \omega)$, where $\omega$ is a 2-form that is

  • closed, $d\omega = 0$, and
  • non-degenerate: if $\omega(v, w) = 0$ for all $w$, then $v = 0$.

Non-degeneracy forces $\dim M = 2n$ to be even, matching $n$ coordinates and $n$ momenta. The basic example is the phase space $M = T^*Q$ of a configuration manifold $Q$. It carries a canonical tautological (Liouville) 1-form $\theta$, which in any cotangent coordinates $(q^i, p_i)$ reads $\theta = \sum_i p_i\, dq^i$, and the canonical symplectic form

\[\omega = -d\theta = \sum_{i=1}^{n} dq^i \wedge dp_i .\]

Because $\omega$ is exact, it is automatically closed.

Sign conventions. Authors differ. Marsden–Ratiu and Abraham–Marsden use $\omega = dq \wedge dp$ together with $\iota_{X_H}\omega = dH$, as this page does. Arnold uses $\omega = dp \wedge dq$ together with $\iota_{X_H}\omega = -dH$. Both give the same Hamilton’s equations. Mixing the two conventions flips the sign of the equations of motion, a common source of errors.

Darboux’s theorem

Darboux’s theorem states that every symplectic manifold is locally standard: around every point there are coordinates in which $\omega = \sum_i dq^i \wedge dp_i$. Riemannian geometry has a local invariant, curvature, that distinguishes one metric from another; symplectic geometry has none beyond the dimension. All interesting symplectic structure is global: topology, the cohomology class of $\omega$ on a closed manifold, and rigidity phenomena such as non-squeezing (below). For mechanics this means that once canonical coordinates are chosen, all the physics is contained in the Hamiltonian function.

Hamiltonian vector fields

Non-degeneracy makes $\omega$ an isomorphism between tangent vectors and 1-forms, $v \mapsto \iota_v \omega = \omega(v, \cdot)$. A function $H : M \to \mathbb{R}$ therefore determines a unique Hamiltonian vector field $X_H$ through

\[\iota_{X_H}\, \omega = dH .\]

Writing $X_H = a^i \partial_{q^i} + b_i \partial_{p_i}$ gives $\iota_{X_H}\omega = a^i\, dp_i - b_i\, dq^i$. Matching this with $dH = \frac{\partial H}{\partial q^i} dq^i + \frac{\partial H}{\partial p_i} dp_i$ yields Hamilton’s equations:

\[X_H = \sum_{i=1}^{n} \left( \frac{\partial H}{\partial p_i}\, \frac{\partial}{\partial q^i} - \frac{\partial H}{\partial q^i}\, \frac{\partial}{\partial p_i} \right), \qquad \dot{q}^i = \frac{\partial H}{\partial p_i}, \quad \dot{p}_i = -\frac{\partial H}{\partial q^i} .\]

Poisson brackets

The symplectic form induces the Poisson bracket of functions,

\[\{f, g\} = \omega(X_f, X_g) = \sum_{i=1}^{n} \left( \frac{\partial f}{\partial q^i}\, \frac{\partial g}{\partial p_i} - \frac{\partial f}{\partial p_i}\, \frac{\partial g}{\partial q^i} \right) .\]

It is bilinear, antisymmetric, and a derivation in each argument (Leibniz rule). Because $d\omega = 0$, it also satisfies the Jacobi identity ${f, {g, h}} + {g, {h, f}} + {h, {f, g}} = 0$. The evolution of any observable is

\[\frac{df}{dt} = \{f, H\} + \frac{\partial f}{\partial t} ,\]

so a time-independent $f$ is conserved exactly when ${f, H} = 0$. By the Jacobi identity, the bracket of two conserved quantities is also conserved (Poisson’s theorem). Dirac’s correspondence ${f, g} \to \frac{1}{i\hbar}[\hat f, \hat g]$ is the starting point of canonical quantization. Groenewold–van Hove shows it cannot hold exactly for all observables, which is why deformation and geometric quantization exist.

Poisson manifolds

Many important systems have a Poisson bracket but no symplectic form. A Poisson manifold is a manifold with a bracket obeying all the properties above, which may be degenerate. Degenerate brackets have Casimir functions $C$ with ${C, f} = 0$ for every $f$. Casimirs are conserved by every Hamiltonian flow, and their level sets foliate the manifold into symplectic leaves, on each of which the dynamics is ordinary symplectic mechanics.

The standard example is the free rigid body in body-frame angular momentum $\Pi \in \mathbb{R}^3 \cong \mathfrak{so}(3)^*$, with the Lie–Poisson bracket

\[\{F, G\}(\Pi) = -\,\Pi \cdot \left(\nabla F \times \nabla G\right), \qquad H = \frac{1}{2}\left(\frac{\Pi_1^2}{I_1} + \frac{\Pi_2^2}{I_2} + \frac{\Pi_3^2}{I_3}\right) .\]
The equation $\dot F = {F, H}$ reproduces Euler’s equations $\dot\Pi = \Pi \times \Omega$ with $\Omega_i = \Pi_i/I_i$. The Casimir $ \Pi ^2$ confines motion to spheres, which are the symplectic leaves. Intersecting these spheres with energy ellipsoids gives the familiar rigid-body orbits and explains the instability of rotation about the intermediate axis (Rigid Body Dynamics). Ideal fluids, plasmas, and many reduced models have the same Lie–Poisson structure.

Phase-space flow

The flow is symplectic

The flow $\varphi_t$ of $X_H$ preserves $\omega$:

\[\varphi_t^{*}\, \omega = \omega .\]

The proof is one line using Cartan’s formula $\mathcal{L}_X = d\,\iota_X + \iota_X\, d$:

\[\mathcal{L}_{X_H}\, \omega = d(\iota_{X_H}\, \omega) + \iota_{X_H}(d\omega) = d(dH) + 0 = 0 .\]

The first term vanishes because $d^2 = 0$ and the second because $\omega$ is closed. A Hamiltonian flow is therefore a one-parameter family of symplectomorphisms, which is the coordinate-free meaning of “canonical transformation.” The converse holds locally: any vector field whose flow preserves $\omega$ is locally Hamiltonian, because $\mathcal{L}_X \omega = d(\iota_X \omega) = 0$ means $\iota_X\omega$ is closed, and closed forms are locally exact.

The figure shows the flow for the simplest nonlinear example, the pendulum. Each curve is a level set of $H$. The flow carries phase-space regions along these curves, shearing them (inner librations have shorter periods than outer ones) without changing their area.

Phase portrait of the simple pendulum: closed librations around the origin, the separatrix through the unstable points at plus and minus pi, and open rotations above and below q p −π π rotation (E > 2mgl) separatrix (E = 2mgl) libration filled dot: stable equilibrium (center) · open dots: unstable equilibrium (saddle), the same point q = ±π
Level sets of the pendulum Hamiltonian $H = p^2/2ml^2 + mgl(1 - \cos q)$. The flow moves clockwise around the center; the dashed separatrix divides oscillation from full rotation.

Liouville’s theorem

Because the flow preserves $\omega$, it also preserves the top exterior power, the Liouville volume form

\[\Omega = \frac{(-1)^{n(n-1)/2}}{n!}\, \omega^{\wedge n} = dq^1 \wedge \cdots \wedge dq^n \wedge dp_1 \wedge \cdots \wedge dp_n .\]

Hence $\varphi_t^{*}\Omega = \Omega$: phase-space volume is conserved (Liouville’s theorem). An ensemble of initial conditions moves like an incompressible fluid. It may stretch and filament, drastically so in chaotic systems, but its volume never changes. The density $\rho$ obeys the Liouville equation

\[\frac{\partial \rho}{\partial t} + \{\rho, H\} = 0 ,\]

so $\rho$ is constant along trajectories. This is the foundation of classical statistical mechanics: any function of $H$ (and of other conserved quantities) is a stationary density, which leads to the microcanonical and canonical ensembles (Statistical Mechanics).

Poincaré integral invariants

The flow also preserves a hierarchy of lower-dimensional invariants. For each $k = 1, \ldots, n$, the integral of $\omega^{\wedge k}$ over any $2k$-dimensional surface carried along by the flow is constant. For $k = 1$, Stokes’ theorem turns this into a loop integral,

\[\oint_{\gamma_t} \sum_i p_i\, dq^i = \text{const}\]

for any closed loop $\gamma_t$ transported by the flow. This is the Poincaré–Cartan invariant. It is the origin of action variables $I = \frac{1}{2\pi}\oint p\, dq$ and of adiabatic invariance.

Symplectic rigidity: non-squeezing

Symplectic maps preserve more than volume. Gromov’s non-squeezing theorem (1985) states that a ball of radius $r$ in $\mathbb{R}^{2n}$ can be symplectically embedded in a cylinder ${q_1^2 + p_1^2 < R^2}$ only if $r \leq R$. A volume-preserving map could squeeze the ball into the cylinder by stretching it along the other directions; a symplectic map cannot. Physically, no Hamiltonian evolution can shrink the spread of an ensemble in any single conjugate pair $(q_i, p_i)$ below its initial projected area. This is a classical analogue of the uncertainty principle. Non-squeezing started modern symplectic topology. It also shows that “symplectic” is strictly stronger than “volume-preserving” when $n \geq 2$.

Lagrangian submanifolds and generating functions

A submanifold $L \subset M$ of dimension $n$ (half the phase-space dimension) is Lagrangian if $\omega$ vanishes on it, $\omega _L = 0$. Examples:
  • The zero section ${p = 0}$ and each fiber ${q = q_0}$ of $T^*Q$.
  • The graph of an exact 1-form, $L = {(q, \nabla S(q))}$. Here $\theta _L = dS$, so $\omega _L = -d(dS) = 0$. Locally, every Lagrangian submanifold that projects nicely onto $Q$ has this form for some generating function $S$.
  • The graph of a symplectomorphism $\phi : M \to M$, viewed as a Lagrangian submanifold of $M \times M$ with the form $\omega \ominus \omega$. This is why every canonical transformation has a generating function (of types $F_1$ to $F_4$) locally.

Hamilton–Jacobi theory is the study of Lagrangian submanifolds that are invariant under the flow. If $L = \operatorname{graph}(\nabla S)$ lies in an energy surface $H = E$, then

\[H\!\left(q, \frac{\partial S}{\partial q}\right) = E ,\]

and trajectories on $L$ are the characteristics of this PDE. Where the projection of $L$ to $Q$ folds over, $S$ becomes multivalued and the projected trajectories focus on caustics. In the semiclassical wavefunction $\psi \approx A\, e^{iS/\hbar}$, these folds produce the Maslov phase corrections that turn Bohr–Sommerfeld quantization into $\oint p\, dq = 2\pi\hbar\,(n + \tfrac{1}{2})$.

Integrable systems

A Hamiltonian system on a $2n$-dimensional symplectic manifold is Liouville integrable if it has $n$ conserved quantities $F_1 = H, F_2, \ldots, F_n$ that are functionally independent and in involution, ${F_i, F_j} = 0$.

Arnold–Liouville theorem. If a level set $M_f = {F_i = f_i}$ of an integrable system is compact and connected, then:

  1. $M_f$ is diffeomorphic to an $n$-torus $T^n$, and it is a Lagrangian submanifold;
  2. a neighborhood of it has action–angle coordinates $(I, \theta)$, with $\omega = \sum_i d\theta^i \wedge dI_i$, in which $H = H(I)$;
  3. the motion is linear on each torus, $\theta(t) = \theta(0) + \omega(I)\, t$ with $\omega(I) = \partial H/\partial I$.

The actions are computed as loop integrals of the tautological form over a basis of cycles $\gamma_i$ on the torus:

\[I_i = \frac{1}{2\pi} \oint_{\gamma_i} \sum_j p_j\, dq^j .\]

Examples include the harmonic oscillator, the Kepler problem, the Euler and Lagrange tops, geodesic flow on an ellipsoid, and infinite-dimensional systems such as the KdV equation. Integrability is exceptional: a generic perturbation destroys it. What survives is described by KAM theory, and what breaks becomes chaos; see Chaos & Nonlinear Dynamics.

Symmetry, momentum maps, and reduction

Momentum maps

Let a Lie group $G$ act on $(M, \omega)$ by symplectomorphisms. Each element $\xi$ of the Lie algebra $\mathfrak{g}$ generates a vector field $\xi_M$. The action has a momentum map $J : M \to \mathfrak{g}^*$ if each $\xi_M$ is Hamiltonian, with Hamiltonian function $\langle J, \xi \rangle$:

\[\iota_{\xi_M}\, \omega = d\langle J, \xi \rangle .\]

Noether’s theorem in this language: if $H$ is $G$-invariant, then ${ \langle J, \xi\rangle, H} = -\,dH(\xi_M) = 0$, so $J$ is conserved by the flow. The momentum map collects all of the symmetry’s conserved charges into one object.

Symmetry group acting on $T^\mathbb{R}^3$ (or $T^\mathbb{R}^{3N}$) Momentum map $J$ Conserved quantity
Translations $\mathbb{R}^3$ $\sum_a \mathbf{p}_a$ Linear momentum
Rotations $SO(3)$ $\sum_a \mathbf{q}_a \times \mathbf{p}_a$ Angular momentum
Galilean boosts (with time) $\sum_a (m_a \mathbf{q}_a - t\,\mathbf{p}_a)$ Center-of-mass motion
Phase rotation $U(1)$ of an oscillator $\tfrac{1}{2}(q^2 + p^2)$ Action (particle number after quantization)
Time translation (extended phase space) $H$ Energy

Reduction

Conserved momenta allow the symmetric degrees of freedom to be eliminated. Marsden–Weinstein (symplectic) reduction fixes the momentum at a value $\mu$ and quotients by the subgroup $G_\mu$ that preserves it:

\[M_\mu = J^{-1}(\mu) / G_\mu .\]

Under regularity conditions $M_\mu$ is a symplectic manifold of dimension $\dim M - \dim G - \dim G_\mu$, and the dynamics of $H$ descends to it.

flowchart LR
    A["Phase space M<br/>dim 2n, symmetry G"] -->|"fix momentum<br/>J = mu"| B["Level set J^-1(mu)<br/>invariant under the flow"]
    B -->|"quotient by the<br/>isotropy group G_mu"| C["Reduced space M_mu<br/>symplectic, smaller"]
    C -->|"solve the reduced dynamics,<br/>then reconstruct"| D["Full motion =<br/>reduced motion + drift along G<br/>(dynamic + geometric phase)"]

Reduction is the geometric form of the textbook practice of using conserved quantities to lower the order of the equations:

  • Central-force motion. Reducing by rotations at fixed angular momentum $\ell$ leaves one-dimensional radial motion in the effective potential $V(r) + \ell^2/2mr^2$.
  • Rigid body. Reducing $T^SO(3)$ by $SO(3)$ leaves $\mathfrak{so}(3)^$ with its Lie–Poisson bracket. The reduced symplectic spaces are the coadjoint orbits, spheres $ \Pi = \text{const}$, which carry the Kirillov–Kostant–Souriau symplectic form.
  • Restricted three-body problem. Moving to the rotating frame removes the uniform rotation and leaves the Jacobi integral.
  • Gauge theories. Gauss’s law is the momentum-map constraint $J = 0$ of the gauge group, and the physical phase space is the reduced space.

When a reduced solution is lifted back to the full space, the motion along the group orbit contains a dynamic part and a geometric part. The geometric part depends only on the path in the reduced space. This is the mechanical origin of geometric phases, discussed next.

Connections and mechanical gauge theory

When the configuration space is a bundle over a shape space, for example the orientation of a deformable body over its internal shape, mechanics acquires a gauge structure. A connection 1-form $A$ splits every motion into a vertical part (a rigid motion along the fiber) and a horizontal part (a change of shape). For a free body with zero angular momentum, conservation of momentum is exactly the statement that the motion is horizontal, and $A$ is the mechanical connection determined by the inertia tensor. Its curvature

\[F = dA + A \wedge A\]

measures the failure of horizontal loops to close. A closed loop in shape space generally produces a net displacement along the fiber, the holonomy, equal to the integral of the curvature over the enclosed region of shape space (to leading order for small loops).

This is how a falling cat rights itself with zero angular momentum: it executes a cyclic change of shape, and the curvature of the mechanical connection converts that loop into a net rotation. Astronauts and divers use the same mechanism to reorient in flight. Microorganisms swimming at low Reynolds number exploit the analogous connection defined by viscous drag (Shapere and Wilczek, 1989): only the geometry of the stroke matters, not its speed. The mathematics is that of Yang–Mills gauge theory, with the shape-space loop playing the role of a Wilson loop.

Geometric phases

Berry phase

A quantum eigenstate $ n(R)\rangle$ of a Hamiltonian that depends on slowly varying parameters $R$, carried adiabatically around a closed loop $C$ in parameter space, returns to itself up to a phase. Besides the dynamical phase $-\frac{1}{\hbar}\int E_n\, dt$, it acquires the Berry phase
\[\gamma_n = \oint_C \mathbf{A}_n \cdot d\mathbf{R}, \qquad \mathbf{A}_n(\mathbf{R}) = i\,\langle n(\mathbf{R}) | \nabla_{\mathbf{R}}\, n(\mathbf{R}) \rangle ,\]

where $\mathbf{A}n$ is the (real) Berry connection. By Stokes’ theorem the phase is the flux of the Berry curvature $\mathbf{F}_n = \nabla{\mathbf{R}} \times \mathbf{A}_n$ through any surface $S$ bounded by $C$:

\[\gamma_n = \iint_S \mathbf{F}_n \cdot d\mathbf{S} .\]

The phase depends only on the geometry of the loop, not on how fast it is traversed, provided the traversal is adiabatic. A spin-$s$ particle in a magnetic field whose direction traces a loop gets $\gamma = -m_s\, \Omega_C$, where $\Omega_C$ is the solid angle the loop subtends and $m_s$ is the spin component along the field. Berry curvature is central in modern condensed matter physics: it underlies the anomalous and quantum Hall effects and the topological invariants (Chern numbers) of band structures (Condensed Matter).

Hannay angle

The classical counterpart, found by Hannay (1985), concerns integrable systems with slowly varying parameters. The actions $I$ are adiabatic invariants. After a closed loop in parameter space, the angle variables have advanced by the dynamical amount $\int \omega(I, R)\, dt$ plus an extra geometric shift, the Hannay angle:

\[\Delta\theta_{\text{H}} = -\frac{\partial}{\partial I} \oint_C \mathbf{A}^{\text{cl}}(I, \mathbf{R}) \cdot d\mathbf{R}, \qquad \mathbf{A}^{\text{cl}} = \left\langle p\, \nabla_{\mathbf{R}}\, q \right\rangle_{\theta} ,\]

where $\langle \cdot \rangle_\theta$ denotes the average over the torus with fixed $I$. Berry showed that the two are related semiclassically, $\Delta\theta_{\text{H}} = -\partial \gamma_n / \partial n$ with $I = (n + \tfrac12)\hbar$: the classical angle shift is the rate of change of the quantum phase with quantum number.

Foucault pendulum

The best-known classical geometric phase is the Foucault pendulum. As the Earth rotates, the pendulum’s swing plane is parallel-transported around the circle of latitude $\phi$. Relative to the ground, the swing plane rotates clockwise in the Northern Hemisphere by

\[\Delta\theta = 2\pi \sin\phi\]

per sidereal day. Equivalently, parallel transport around the latitude circle rotates a tangent vector by the solid angle enclosed by the circle, $2\pi(1 - \sin\phi)$, which differs from the rotation above by exactly $2\pi$. No torque acts on the swing plane; the rotation is pure holonomy of the sphere’s curvature. The same geometry explains the rotation of light’s polarization along a helically coiled optical fiber (Tomita–Chiao) and the spin precession of particles in slowly rotating fields.

Beyond symplectic mechanics

Several extensions handle systems that ordinary symplectic mechanics does not:

Structure Handles Key idea
Poisson and Lie–Poisson manifolds Rigid bodies, ideal fluids, plasmas, reduced systems Degenerate brackets with Casimirs; dynamics on symplectic leaves
Contact geometry Time-dependent Hamiltonians, certain dissipative systems, equilibrium thermodynamics Odd-dimensional manifolds with a maximally non-integrable 1-form, e.g. $dS - p\, dq$
Nonholonomic mechanics Rolling without slipping, skates, wheeled robots Velocity constraints imposed by the Lagrange–d’Alembert principle; the flow is generally not symplectic and not variational
Port-Hamiltonian and Dirac structures Interconnected, controlled, and dissipative engineering systems Energy flow through ports; widely used in control theory and robotics
Multisymplectic geometry Classical field theories A form of degree $n+1$ on a jet bundle; covariant Hamiltonian field theory and multisymplectic integrators
Presymplectic geometry (Dirac–Bergmann) Gauge theories and singular Lagrangians Degenerate $\omega$; constraint algorithm and reduction by gauge orbits

Summary of objects

Object Local expression Role
Configuration manifold $Q$ coordinates $q^i$ Possible positions
Tangent bundle $TQ$ $(q^i, \dot q^i)$ Domain of the Lagrangian
Cotangent bundle $T^*Q$ $(q^i, p_i)$ Hamiltonian phase space
Tautological 1-form $\theta$ $\sum_i p_i\, dq^i$ Action integrand; Poincaré–Cartan invariant
Symplectic form $\omega$ $-d\theta = \sum_i dq^i \wedge dp_i$ Turns $dH$ into $X_H$; defines Poisson brackets and area
Hamiltonian vector field $X_H$ $\iota_{X_H}\omega = dH$ The dynamics
Poisson bracket ${f, g} = \omega(X_f, X_g)$ Evolution $\dot f = {f, H}$; classical limit of commutators
Liouville volume $\Omega$ $dq^1 \wedge \cdots \wedge dp_n$ Conserved phase-space measure
Lie derivative $\mathcal{L}_X$ $d\,\iota_X + \iota_X\, d$ on forms Rate of change along a flow
Momentum map $J$ $\iota_{\xi_M}\omega = d\langle J, \xi\rangle$ Noether charges
Connection $A$, curvature $F$ $F = dA + A \wedge A$ Holonomy, geometric phases

See also