Integrable Systems

Integrable systems are dynamical systems possessing the maximal number of independent conserved quantities in involution, allowing exact solution by quadratures rather than chaos. Key result: the Liouville–Arnold theorem guarantees that an

n

-degree-of-freedom Hamiltonian system with

n

Poisson-commuting integrals confines motion to invariant tori and is solvable via action–angle variables. Applies to: classical mechanics, soliton equations (KdV, sine-Gordon, NLS), spin chains, random matrices, and algebraic geometry. Who/when: Liouville (1855), Arnold (1963), and the inverse-scattering revolution of Gardner–Greene–Kruskal–Miura (1967) and Lax (1968).

An integrable system is a Hamiltonian system on a $2 n$ -dimensional symplectic manifold that admits $n$ functionally independent first integrals in mutual involution under the Poisson bracket. This structure is exceptional: generic Hamiltonian systems are chaotic, yet integrable ones are exactly solvable, their phase space foliated by invariant tori on which motion is linear (quasi-periodic). The theory bridges classical mechanics, partial differential equations, Lie groups, and algebraic geometry through unifying objects like the Lax pair and the $r$ -matrix. A surprising fact: certain nonlinear PDEs such as the Korteweg–de Vries equation, despite being infinite-dimensional, behave as integrable systems with infinitely many conservation laws, and their solitary-wave solutions (solitons) pass through one another unchanged, retaining only a phase shift.

The Core Insight

Integrability means hidden linearity. Beneath an apparently complicated nonlinear flow lies a coordinate system — the action–angle variables — in which the equations of motion become trivial: the actions are constant and the angles advance at constant frequencies. The conserved quantities are not accidental; they generate commuting flows whose simultaneous level sets are tori, and the dynamics is just uniform straight-line motion wound around these tori. The same principle reappears infinite-dimensionally as the isospectral deformation: an integrable PDE is the compatibility condition for a linear spectral problem whose spectrum never changes in time.

Formal Statement

Hamiltonian System and Poisson Bracket

Let $(M, ω)$ be a smooth symplectic manifold of dimension $2 n$ , with closed nondegenerate two-form $ω$ . A function $H \in C^{\infty} (M)$ (the Hamiltonian) determines a vector field $X_{H}$ via

ι_{X_{H}} ω = d H, \dot{f} = {f, H} .

⟳

X_{H}

is defined by contracting

ω

with it to recover

d H

; nondegeneracy of

ω

guarantees existence and uniqueness of

X_{H}

The Poisson bracket of $f, g \in C^{\infty} (M)$ is

{f, g} = ω (X_{f}, X_{g}) = i = 1 \sum n (\frac{\partial f}{\partial q ^{i}} \frac{\partial g}{\partial p _{i}} - \frac{\partial f}{\partial p _{i}} \frac{\partial g}{\partial q ^{i}}),

in canonical Darboux coordinates $(q^{i}, p_{i})$ , where $ω = \sum_{i} d p_{i} \land d q^{i}$ .

⟳evaluate

ω (X_{f}, X_{g})

in Darboux coordinates using

X_{f} = \sum_{i} (\partial_{p_{i}} f \partial_{q^{i}} - \partial_{q^{i}} f \partial_{p_{i}})

Liouville Integrability

A Hamiltonian system on $(M^{2 n}, ω)$ is completely integrable (in the Liouville sense) if there exist $n$ functions $F_{1} = H, F_{2}, \dots, F_{n} \in C^{\infty} (M)$ that are

in involution: ${F_{i}, F_{j}} = 0$ for all $i, j$ ;
functionally independent: $d F_{1} \land \dots \land d F_{n} \neq = 0$ on a dense open set.

The map $F = (F_{1}, \dots, F_{n}) : M \to R^{n}$ is the moment map of the resulting $R^{n}$ action.

Liouville–Arnold Theorem

Let $F = (F_{1}, \dots, F_{n})$ be the integrals of a completely integrable system, and let $c \in R^{n}$ be a regular value with $M_{c} = F^{- 1} (c)$ compact and connected. Then:

$M_{c}$ is diffeomorphic to the $n$ -torus $T^{n} = R^{n} / Z^{n}$ ;
there exist action–angle coordinates $(I_{k}, θ_{k})$ in a neighbourhood of $M_{c}$ , with $ω = \sum_{k} d I_{k} \land d θ_{k}$ , the $I_{k}$ functions of $F$ alone and $θ_{k}$ angle coordinates on the torus;
the dynamics is linear:

\dot{I}_{k} = 0, \dot{θ}_{k} = ω_{k} (I) = \frac{\partial H}{\partial I _{k}},

so the flow is quasi-periodic with frequencies $ω_{k} (I)$ .

⟳the commuting flows of

X_{F_{1}}, \dots, X_{F_{n}}

define a free transitive

R^{n}

action on

M_{c}

; compactness forces the stabilizer to be a lattice, giving the torus. Actions arise as

I_{k} = \frac{1}{2 π} \oint_{γ_{k}} \sum_{i} p_{i} d q^{i}

over the homology basis

γ_{k}

Why It Holds / Derivation

Sketch of the Liouville–Arnold Theorem

Step 1 — Commuting flows. Since

{F_{i}, F_{j}} = 0

, the Hamiltonian vector fields satisfy

[X_{F_{i}}, X_{F_{j}}] = - X_{{F_{i}, F_{j}}} = 0

. Hence the flows

ϕ_{i}^{t}

X_{F_{i}}

commute pairwise, giving a smooth action of

R^{n}

on the level set

M_{c}

. Step 2 — Tangency and dimension. Each

X_{F_{i}}

is tangent to

M_{c}

because

X_{F_{i}} (F_{j}) = {F_{j}, F_{i}} = 0

. Functional independence makes the

n

vector fields linearly independent everywhere on

M_{c}

, so they span its tangent space. Thus

dim M_{c} = n

and

M_{c}

is a Lagrangian submanifold (

ω ∣_{M_{c}} = 0

). Step 3 — Group action structure. The

R^{n}

action is locally free; its orbit through any point is open. By connectedness one orbit is all of

M_{c}

, so

M_{c} ≅ R^{n} /Λ

for a stabilizer subgroup

Λ

. Compactness forces

Λ

to be a full-rank lattice

≅ Z^{n}

, so

M_{c} ≅ T^{n}

. Step 4 — Constructing actions. Choose a basis

γ_{1}, \dots, γ_{n}

H_{1} (M_{c}; Z)

. Define

I_{k} = \frac{1}{2 π} \oint_{γ_{k}} θ_{Liouville}, θ_{Liouville} = i \sum p_{i} d q^{i} .

⟳

d θ_{Liouville} = ω

, and since

M_{c}

is Lagrangian,

θ_{Liouville}

restricts to a closed form whose periods over

γ_{k}

are deformation-invariant (Stokes), making

I_{k}

functions of

c

only.

Step 5 — Conjugate angles. A generating function

S (q, I)

produces angles

θ_{k} = \partial S / \partial I_{k}

that increase by

2 π

around

γ_{k}

and satisfy the canonical relations. In these coordinates

H = H (I)

alone, yielding

\dot{I}_{k} = 0

\dot{θ}_{k} = \partial H / \partial I_{k}

■

Joseph Liouville | 1809–1882

Liouville, a towering figure of nineteenth-century French analysis, proved in 1855 the theorem that bears his name: a system with as many independent integrals in involution as degrees of freedom can be integrated "by quadratures" — reduced to a finite sequence of integrations and algebraic operations. His insight that involution is the crucial algebraic condition (not mere abundance of integrals) is the seed of the entire modern theory. Liouville also founded the Journal de Mathématiques Pures et Appliquées and made foundational contributions to transcendental number theory, Sturm–Liouville theory, and conformal geometry.

Building Intuition

Picture phase space as a high-dimensional ocean. A generic Hamiltonian flow stirs it chaotically — nearby trajectories diverge exponentially and trajectories fill regions densely. Integrability is the special, fragile circumstance in which the ocean is perfectly layered into nested doughnut surfaces (tori). On each torus, the motion is like a point moving along a fixed straight line drawn on the torus's "unrolled" square — it never leaves, never spreads, just winds around forever at fixed speeds. If the frequencies $ω_{k}$ are rationally related, the orbit closes into a loop; if irrationally related, it densely covers the torus (quasi-periodic motion). The $n$ conserved quantities are the "addresses" of the tori, and the action variables are precisely the cross-sectional sizes of the doughnut.

Invariant tori foliating phase space; trajectories wind quasi-periodically. The highlighted curve is a single orbit on one torus.

Worked Examples

The Planar Harmonic Oscillator (2 DOF)

Given:

H = \frac{1}{2} (p_{x}^{2} + p_{y}^{2}) + \frac{1}{2} (ω_{x}^{2} x^{2} + ω_{y}^{2} y^{2})

, with

ω_{x} = 2 s^{- 1}

ω_{y} = 3 s^{- 1}

, unit mass. Step 1 — Identify integrals. The energies of each direction separately,

F_{1} = \frac{1}{2} (p_{x}^{2} + ω_{x}^{2} x^{2}), F_{2} = \frac{1}{2} (p_{y}^{2} + ω_{y}^{2} y^{2}),

are conserved and

{F_{1}, F_{2}} = 0

(they depend on disjoint variables). With

n = 2

independent involutive integrals, the system is integrable. Step 2 — Action variables. For a 1-DOF oscillator of energy

E

and frequency

ω

I = \frac{1}{2 π} \oint p d q = E / ω

. Take

E_{x} = 8 J

E_{y} = 9 J

I_{x} = \frac{E _{x}}{ω _{x}} = \frac{8}{2} = 4 J\cdots, I_{y} = \frac{E _{y}}{ω _{y}} = \frac{9}{3} = 3 J\cdots .

⟳the phase-space orbit is an ellipse of area

2 π E / ω

; the action is the enclosed area divided by

2 π

Step 3 — Frequencies.

H = ω_{x} I_{x} + ω_{y} I_{y}

, so

\dot{θ}_{x} = ω_{x} = 2

\dot{θ}_{y} = ω_{y} = 3

. The frequency ratio

2 : 3

is rational, so the orbit closes (a Lissajous figure). Answer:

I_{x} = 4 J\cdots, I_{y} = 3 J\cdots; closed Lissajous orbit, period 2 π s

The Korteweg–de Vries One-Soliton

Given: the KdV equation

u_{t} + 6 u u_{x} + u_{xxx} = 0

, an infinite-dimensional integrable system. We test the one-soliton solution

u (x, t) = \frac{c}{2} sech^{2} (\frac{c}{2} (x - c t)),

with speed

c = 4

(so amplitude

c /2 = 2

). Step 1 — Soliton profile. The peak amplitude is

A = c /2 = 2

and the width scales as

w \sim 2/ c = 1

. Taller solitons are narrower and faster — amplitude and speed are locked. Step 2 — A conserved quantity (mass). KdV conserves

M = \int_{- \infty}^{\infty} u d x

. For the soliton,

M = \int_{- \infty}^{\infty} \frac{c}{2} sech^{2} (\frac{c}{2} ξ) d ξ = \frac{c}{2} \cdot \frac{2}{c} \cdot 2 = 2 c .

⟳substitute

η = \frac{c}{2} ξ

; use

\int sech^{2} η d η = tanh η

, giving total

2

Step 3 — Evaluate. With

c = 4

M = 24 = 4

. The momentum

P = \frac{1}{2} \int u^{2} d x

and energy are likewise conserved; KdV has infinitely many such integrals in involution, the hallmark of infinite-dimensional integrability. Answer:

M = 4 (conserved mass), amplitude = 2, speed = 4

Common Misconceptions

"Having many conserved quantities makes a system integrable." False — involution is essential. A system can possess

n

independent integrals that do not Poisson-commute, in which case the level set need not be a torus and the Liouville–Arnold construction fails. Conversely, "superintegrable" systems (like the Kepler problem) have more than

n

integrals, but only

n

of them can be mutually in involution. Integrability is about the right algebraic relations among conserved quantities, not merely their abundance. "Integrability is generic or stable." It is exceptionally fragile. The KAM theorem shows that under small perturbations most tori survive but a positive-measure set near resonances breaks up, and the system becomes chaotic on those regions. Integrable systems form a measure-zero, infinite-codimension subset of all Hamiltonian systems.

The Double Pendulum

The planar double pendulum has $n = 2$ degrees of freedom and one obvious integral, the energy $H$ . But there is no second independent integral in involution: numerical experiments and the Melnikov method reveal homoclinic tangles and positive Lyapunov exponents. The phase space is not foliated by tori; instead it contains a chaotic sea interlaced with islands of stability. This shows that having one fewer integral than required is enough to destroy integrability entirely — the system is non-integrable and exhibits deterministic chaos.

Connections

Integrability sits at a crossroads of several theories.

The classical Lax pair formulation, $\dot{L} = [L, P]$ , links integrability to Lie Algebras and isospectral flow: the eigenvalues of $L$ are conserved, generating the integrals.
The classical $r$ -matrix encodes involution via ${L_{1}, L_{2}} = [r, L_{1} + L_{2}]$ and connects to Quantum Groups and the Yang–Baxter equation in the quantum case.
Algebro-geometric solutions of soliton equations live on Jacobian varieties of Riemann Surfaces, linking integrable PDEs to Algebraic Geometry through the Its–Matveev theta-function formulas.
The breakdown of integrability is governed by KAM Theory and the Hamiltonian Mechanics perturbation series.
Quantum integrable spin chains relate to Statistical Mechanics via transfer matrices and the Bethe ansatz.

Real-World Applications

Soliton fiber optics. The nonlinear Schrödinger equation governing pulse propagation in optical fibers is integrable; soliton pulses preserve their shape over transoceanic distances, underpinning high-capacity telecommunications.
Shallow-water and tsunami modeling. The KdV and Kadomtsev–Petviashvili equations describe weakly nonlinear long waves; soliton interactions explain observed wave patterns and bores in channels and estuaries.
Plasma physics. Ion-acoustic solitons and Langmuir waves in plasmas obey integrable model equations, relevant to fusion confinement and space-plasma turbulence.
Bose–Einstein condensates. The Gross–Pitaevskii (NLS-type) equation supports matter-wave solitons observed experimentally in ultracold atomic gases.
Random matrix theory & number theory. Painlevé transcendents — integrable ODEs — give exact eigenvalue gap distributions (Tracy–Widom law) and appear in conjectures on the Riemann zeta zeros.
Celestial mechanics. Integrable approximations (Kepler, Euler, and Lagrange tops) are the unperturbed backbone of perturbation theory for planetary and spacecraft orbits.
Quantum spin chains and condensed matter. The integrable Heisenberg XXX/XXZ chains, solved by the Bethe ansatz, model magnetism in one-dimensional materials and inform quantum-computing qubit arrays.

Historical Development

1834 — John Scott Russell observes the "wave of translation" (a soliton) in the Union Canal, riding alongside it on horseback for over a mile.
1855 — Liouville proves that systems with $n$ involutive integrals are integrable by quadratures.
1888–1889 — Sofia Kovalevskaya discovers her integrable spinning top, the last classical case, using complex-analytic (Painlevé) methods.
1895 — Korteweg and de Vries derive the KdV equation, capturing Russell's solitary wave analytically.
1954–1963 — Kolmogorov, Arnold, and Moser develop KAM theory, explaining the persistence and destruction of tori under perturbation.
1965 — Zabusky and Kruskal coin the term soliton after numerical experiments showing particle-like wave collisions in KdV.
1967 — Gardner, Greene, Kruskal, and Miura invent the inverse scattering transform, solving KdV exactly via the Schrödinger spectral problem.
1968 — Peter Lax introduces the Lax pair, recasting integrability as isospectral evolution and unifying the field.
1971–1979 — Zakharov, Shabat, Faddeev, Takhtajan, and the Leningrad school extend ISM to NLS, sine-Gordon, and develop the classical/quantum $r$ -matrix and algebraic Bethe ansatz.

Subtleties & Limits

What does "integrable" mean precisely, and how do we recognize it? There is no single accepted definition: Liouville integrability, algebraic integrability,

C

-integrability, and the Painlevé property all capture different aspects, and they do not coincide. Determining whether a given system is integrable remains genuinely hard — there is no decision procedure. The classification of integrable PDEs is open; symmetry approaches (Shabat, Mikhailov) classify only restricted families. In the quantum realm, the relationship between classical and quantum integrability is subtle: some quantum-integrable models lack a clean classical limit, and the completeness of the Bethe ansatz is proven only in special cases. Finally, the role of integrable structures in gauge theory and the AdS/CFT correspondence (integrability of planar

N = 4

super Yang–Mills) is a vast, active, and not fully understood frontier.

Integrable Systems

The Core Insight

Formal Statement

Hamiltonian System and Poisson Bracket

Liouville Integrability

Liouville–Arnold Theorem

Why It Holds / Derivation

Sketch of the Liouville–Arnold Theorem

Joseph Liouville | 1809–1882

Building Intuition

Invariant tori foliating phase space; trajectories wind quasi-periodically. The highlighted curve is a single orbit on one torus.

Worked Examples

The Planar Harmonic Oscillator (2 DOF)

The Korteweg–de Vries One-Soliton

Common Misconceptions

The Double Pendulum

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also

Before this

Where this leads