09.06.03 · classical-mech / integrable

Completely Integrable Systems: The Toda Lattice and the Kepler Problem Revisited

shipped3 tiersLean: nonepending prereqs

Anchor (Master): Moser, Various Aspects of Integrable Hamiltonian Systems (1981); Babelon, Bernard & Talon, Integrable Systems in Classical Mechanics (2003); Flaschka, Phys. Rev. B 9 (1974)

Intuition Beginner

The Kepler problem -- a planet orbiting a star under Newtonian gravity -- is one of the oldest solved problems in physics. What makes it special among mechanical systems is that it has more conserved quantities than you would expect. A planet in a plane has two degrees of freedom, so the Liouville-Arnold theorem says you need two independent, Poisson-commuting conserved quantities for integrability. Energy and angular momentum give you two. But the Kepler problem has a third: the Laplace-Runge-Lenz (LRL) vector, a vector pointing from the star to the perihelion (the point of closest approach) whose length encodes the eccentricity of the orbit.

A system with two degrees of freedom and three conserved quantities is "superintegrable" -- it has more integrals than it needs. The consequence is remarkable: every bound orbit closes. The planet returns to exactly the same position and velocity after one revolution. This does not happen for generic central-force potentials; it is a special property of the $1/ r$ gravitational potential and the harmonic-oscillator potential, a fact established by Bertrand's theorem ^{[source pending]}.

The Laplace-Runge-Lenz vector is defined as

A = p \times L - mk \hat{r},

where $L = r \times p$ is the angular momentum, $k = GM m$ is the gravitational constant times the product of the masses, and $\hat{r}$ is the unit vector pointing from the star to the planet. This vector is constant along the orbit. Its direction points along the major axis of the ellipse (toward perihelion), and its magnitude equals $mk e$ , where $e$ is the orbital eccentricity. For a circular orbit, $A = 0$ .

The existence of the LRL vector implies a "hidden symmetry": the Kepler problem has the symmetry group SO(4) (the group of rotations in four dimensions), not merely the SO(3) you get from rotational invariance. The extra symmetry is what produces the closed orbits and the degeneracy between the radial and angular frequencies.

Now consider a very different system: the Toda lattice. Imagine a chain of particles connected by springs, where the spring force is not linear (Hooke's law) but exponential. Specifically, the potential between neighbouring particles $n$ and $n + 1$ is

V (r) = \frac{a}{b} (e^{- b r} + b r - 1),

where $r$ is the distance between the particles and $a, b > 0$ are constants. At small separations, this approximates a harmonic spring with spring constant $a$ . At large separations, the exponential makes the restoring force grow much faster than linear.

For a chain of $N$ particles with periodic boundary conditions (particle $N$ connects back to particle 1), the Toda lattice has $N$ degrees of freedom. Integrability requires $N$ independent, Poisson-commuting conserved quantities. Morikazu Toda discovered in 1967 that this system has a family of conserved quantities and that numerical solutions showed perfectly regular, non-chaotic behaviour ^{[source pending]}. The puzzle was: why? A chain of particles with nonlinear interactions should, by general reasoning, be chaotic. The answer, proved by Flaschka (1974) and Moser (1975), is that the Toda lattice is completely integrable. It has exactly $N$ independent, commuting integrals of motion, and its dynamics can be solved exactly via a beautiful algebraic construction called the Lax pair ^{[source pending]}.

The Toda lattice is surprising because integrability is rare. Generic Hamiltonian systems are not integrable. The Kepler problem is integrable because of its high symmetry. The Toda lattice has no obvious symmetry to explain its integrability -- the conserved quantities are hidden, encoded in the eigenvalues of a tridiagonal matrix that evolves in time while keeping its spectrum constant. This "isospectral" property is the hallmark of a Lax pair, and it provides a systematic method for constructing all the integrals at once.

Visual Beginner

Figure 1. Left: a Keplerian orbit showing the conserved Laplace-Runge-Lenz vector $A$ pointing from the focus to perihelion. The vector remains fixed in space as the planet moves along its elliptical path. Right: the Toda lattice, a ring of particles connected by exponential springs. Despite the nonlinear interactions, the system is completely integrable with $N$ commuting conserved quantities.

The Kepler orbit closes because the LRL vector breaks the rotational symmetry just enough to pin the ellipse's orientation. The Toda chain looks like it should be chaotic -- a set of balls on nonlinear springs bouncing off each other -- but the hidden Lax-pair structure prevents chaos entirely. Both systems illustrate the same principle: having the right conserved quantities makes a mechanical system exactly solvable, no matter how complicated the forces appear.

Worked example Beginner

Counting conserved quantities for the Kepler problem.

A planet of mass $m$ orbiting a star in three dimensions has three degrees of freedom (the three spatial coordinates). The Liouville-Arnold theorem says it needs three independent, commuting integrals for integrability. Here is what it actually has:

Energy $H = \frac{1}{2 m} ∣ p ∣^{2} - \frac{k}{r}$ . Conserved because the system is time-translation invariant.
Angular momentum $L = r \times p$ . This is a vector with three components $L_{x}, L_{y}, L_{z}$ , but only two are independent because the Poisson brackets satisfy ${L_{x}, L_{y}} = L_{z}$ (and cyclic permutations). The magnitude $L^{2} = L_{x}^{2} + L_{y}^{2} + L_{z}^{2}$ and one component (conventionally $L_{z}$ ) form two independent commuting integrals when combined with $H$ in the right way.
The LRL vector $A = p \times L - mk \hat{r}$ . This is a vector with three components, but only one is independent of $H$ and $L$ because $A \cdot L = 0$ (the LRL vector is perpendicular to the angular momentum) and $∣ A ∣^{2} = m^{2} k^{2} + 2 m E L^{2}$ (the LRL magnitude is determined by energy and angular momentum).

The independent integrals are $H$ , $L_{z}$ , and $A_{z}$ (the component of the LRL vector along the $z$ -axis, restricted to the plane of motion). But there is a subtlety: the LRL vector does not commute with the angular momentum in the usual way. The three quantities $H$ , $L^{2}$ , and one component of $A$ form a set of three independent integrals, but the Poisson-bracket relations are more complicated than simple commutation. The resolution is that the symmetry algebra closes to form $so (4)$ (for bound states), and the system is "superintegrable" -- it has more integrals than degrees of freedom ^{[source pending]}.

For the planar Kepler problem ( $n = 2$ degrees of freedom), two integrals suffice: $H$ and $L$ . The LRL vector provides an additional integral that is not needed for integrability but constrains the orbits further, making them all close.

Check your understanding Beginner

Formal definition: integrable Hamiltonian systems Intermediate+

The Kepler problem in action-angle variables

The Kepler Hamiltonian in plane polar coordinates is

H = \frac{p _{r}^{2}}{2 μ} + \frac{L ^{2}}{2 μ r ^{2}} - \frac{k}{r},

where $μ$ is the reduced mass and $k = GM μ$ . As computed in Unit 09.06.02, the action variables for a bound orbit ( $E < 0$ ) are

I_{ϕ} = L, I_{r} = - L + k \frac{μ}{2∣ E ∣} .

The Hamiltonian expressed in terms of the action variables is

H = - \frac{μ k ^{2}}{2 ( I _{r} + I _{ϕ} ) ^{2}} .

The key feature is that $H$ depends only on the sum $J = I_{r} + I_{ϕ}$ , not on $I_{r}$ and $I_{ϕ}$ separately. Both frequencies are equal:

ω_{r} = \frac{\partial H}{\partial I _{r}} = \frac{μ k ^{2}}{J ^{3}} = ω_{ϕ} = \frac{\partial H}{\partial I _{ϕ}} .

This degeneracy is the action-angle signature of the LRL vector: an extra conserved quantity manifests as a relation among the actions, causing the Hamiltonian to depend on fewer than $n$ independent combinations.

Delaunay elements

In celestial mechanics, the action-angle variables of the Kepler problem are called Delaunay elements (after Heniard, who introduced them in 1908) ^{[source pending]}. For the three-dimensional Kepler problem with Hamiltonian

H = \frac{∣ p ∣ ^{2}}{2 μ} - \frac{k}{r},

the three action variables are:

$L = I_{ϕ}$ : the total angular momentum (magnitude).
$G = I_{θ}$ : the $z$ -component of the angular momentum (related to the inclination).
$J = I_{r} + L$ : the "principal action," equal to $k μ / (2∣ E ∣)$ .

The conjugate angle variables are:

$g$ : the argument of perihelion (the angle from the ascending node to the perihelion point).
$h$ : the longitude of the ascending node (the angle in the reference plane).
$l$ : the mean anomaly (an angle that increases uniformly with time).

In Delaunay variables, the Hamiltonian is

H = - \frac{μ k ^{2}}{2 J ^{2}},

depending only on $J$ . This means $J$ is constant (conservation of energy), while $L$ and $G$ are also constant (they do not appear in $H$ ). The angle $l$ advances at rate $\dot{l} = \partial H / \partial J = μ k^{2} / J^{3}$ , while $g$ and $h$ are constant: $\overset{g}{˙} = \partial H / \partial L = 0$ , $\dot{h} = \partial H / \partial G = 0$ . The constancy of $g$ means the ellipse does not precess, and the constancy of $h$ means the orbital plane does not wobble. Both are consequences of the hidden SO(4) symmetry.

The Delaunay elements are the starting point for perturbation theory in celestial mechanics. When the Kepler problem is perturbed (by other planets, oblateness of the central body, relativistic corrections), the Delaunay elements evolve slowly. The KAM theorem 09.08.01 describes which Kepler tori survive such perturbations.

The Toda lattice

Key result: Toda lattice definition and Hamiltonian

The Toda lattice is a system of $N$ particles on a line (or a ring, for periodic boundary conditions) with positions $q_{1}, \dots, q_{N}$ and momenta $p_{1}, \dots, p_{N}$ , interacting via the exponential potential. The Hamiltonian for the periodic case is

H = n = 1 \sum N \frac{p _{n}^{2}}{2} + n = 1 \sum N (e^{- (q_{n + 1} - q_{n})} - 1),

where $q_{N + 1} = q_{1}$ (periodicity). The mass and coupling constants have been set to 1 by a choice of units. The potential $e^{- r} - 1$ is the Toda potential with $a = b = 1$ .

For the open Toda lattice (no periodicity, particles at the ends are free), the second sum runs from $n = 1$ to $N - 1$ , and the particles at the ends scatter freely as $t \to \pm \infty$ .

Hamilton's equations are

\overset{q}{˙}_{n} = p_{n}, \overset{p}{˙}_{n} = e^{- (q_{n} - q_{n - 1})} - e^{- (q_{n + 1} - q_{n})} .

These are nonlinear coupled ODEs. At first glance, there is no reason to expect integrability.

Flaschka variables

Flaschka's key insight (1974) was to introduce the change of variables ^{[source pending]}

a_{n} = \frac{1}{2} e^{- (q_{n + 1} - q_{n}) /2}, b_{n} = - \frac{1}{2} p_{n} .

In these variables, Hamilton's equations become

\overset{a}{˙}_{n} = a_{n} (b_{n + 1} - b_{n}), \dot{b}_{n} = 2 (a_{n}^{2} - a_{n - 1}^{2}) .

This is exactly the equation for the entries of a tridiagonal Jacobi matrix $L$ (also called the Lax matrix) evolving in time. Define

L = b_{1} a_{1} 0 ⋮ 0 a_{N} a_{1} b_{2} a_{2} 00 0 a_{2} b_{3} 00 \dots \dots \dots ⋱ \dots \dots 000 b_{N - 1} a_{N - 1} a_{N} 00 ⋮ a_{N - 1} b_{N} .

The corner entries $a_{N}$ appear only in the periodic case; for the open Toda lattice, the matrix is tridiagonal with zeros in the corners.

The Lax pair

The fundamental discovery is that there exists a second matrix $B$ (the Lax partner) such that the equations of motion are equivalent to the Lax equation

\dot{L} = [B, L] = B L - L B .

For the Toda lattice, $B$ is the antisymmetric tridiagonal matrix

B = 0 - a_{1} 0 ⋮ 0 a_{N} a_{1} 0 - a_{2} 00 0 a_{2} 000 \dots \dots \dots ⋱ \dots \dots 0000 - a_{N - 1} - a_{N} 00 ⋮ a_{N - 1} 0 .

The Lax equation has an immediate consequence: the eigenvalues of $L$ are constants of the motion. To see this, let $L (t)$ satisfy $\dot{L} = [B, L]$ , and let $U (t)$ be the fundamental solution of $\dot{U} = B U$ with $U (0) = I$ . Then

L (t) = U (t) L (0) U (t)^{- 1} .

This is a similarity transformation, so $L (t)$ and $L (0)$ have the same eigenvalues. The $N$ eigenvalues $λ_{1}, \dots, λ_{N}$ of $L$ are $N$ conserved quantities ^{[source pending]}.

The conserved quantities are functionally independent and pairwise Poisson-commute (this was proved by Flaschka). Therefore, by the Liouville-Arnold theorem, the Toda lattice is completely integrable.

The first integrals in this sequence are:

$F_{1} = tr (L) = \sum b_{n} = - \frac{1}{2} \sum p_{n}$ (total momentum).
$F_{2} = tr (L^{2}) = \sum b_{n}^{2} + 2 \sum a_{n}^{2} = \frac{1}{4} \sum p_{n}^{2} + \frac{1}{2} \sum e^{- (q_{n + 1} - q_{n})}$ (proportional to the Hamiltonian).
$F_{k} = tr (L^{k})$ for $k = 1, \dots, N$ .

The Hamiltonian is $\frac{1}{2} F_{2} - \frac{N}{2}$ , and the $N$ quantities $F_{1}, \dots, F_{N}$ are $N$ independent, commuting integrals.

Explicit integration of the open Toda lattice

For the open Toda lattice with $N$ particles, the Jacobi matrix $L$ is tridiagonal with no corner entries. The explicit solution proceeds as follows.

Diagonalise $L (0) = Q Λ Q^{- 1}$ where $Λ = diag (λ_{1}, \dots, λ_{N})$ .
The solution is $L (t) = Q (t) Λ Q (t)^{- 1}$ where $Q (t)$ is obtained from $Q$ by exponentiating $B$ .
Since the eigenvalues are constant, the entire time evolution reduces to updating the eigenvectors.

In practice, the solution can be written using the factorisation method. Define the matrix $exp (t Λ + C)$ where $C$ is a constant matrix determined by initial conditions. Then the particle positions and momenta are extracted from the entries of the factored form. This gives explicit formulas for $q_{n} (t)$ and $p_{n} (t)$ in terms of exponentials of the eigenvalues ^{[source pending]}.

For $N = 2$ particles, the open Toda lattice reduces to a single exponential interaction:

H = \frac{p _{1}^{2} + p _{2}^{2}}{2} + e^{- (q_{2} - q_{1})} .

In centre-of-mass coordinates $Q = q_{2} - q_{1}$ and $P = p_{2} - p_{1}$ , this becomes a one-particle problem with $H = P^{2} /4 + e^{- Q}$ (plus centre-of-mass motion). The solution is elementary: $Q (t)$ decreases from infinity, reaches a minimum, and returns to infinity, with the scattering map given by a simple time-of-flight integral. For $N \geq 3$ , the explicit solution requires the matrix factorisation described above.

Exercises Intermediate+

Exercise 1 (medium, symbolic).

Verify directly that the Laplace-Runge-Lenz vector $A = p \times L - mk \hat{r}$ is conserved for the Kepler Hamiltonian $H = ∣ p ∣^{2} / (2 m) - k / r$ . Compute $d A / d t$ using Hamilton's equations and show it vanishes.

Hint

Compute $\dot{p} = - \nabla V = - k \hat{r} / r^{2}$ and $\dot{r} = p / m$ . Then compute $\dot{L} = 0$ (angular momentum is conserved), and differentiate $A$ using the product rule, substituting $\dot{p}$ and $\dot{r}$ .

Answer

Since $\dot{L} = 0$ , we have $\frac{d}{d t} (p \times L) = \dot{p} \times L$ . Using $\dot{p} = - k \hat{r} / r^{2}$ and $L = m r \times \dot{r}$ :

\dot{p} \times L = - \frac{k}{r ^{2}} \hat{r} \times (m r \times \dot{r}) .

By the vector triple product identity $\hat{r} \times (r \times \dot{r}) = r (\hat{r} \cdot \dot{r}) - \dot{r} (\hat{r} \cdot r) = r (\hat{r} \cdot \dot{r}) \hat{r} - r \dot{r}$ . So

\dot{p} \times L = - \frac{mk}{r} ((\hat{r} \cdot \dot{r}) \hat{r} - \dot{r}) .

Now $\frac{d}{d t} (mk \hat{r}) = mk \frac{d}{d t} (r / r) = mk (\dot{r} / r - r (r \cdot \dot{r}) / r^{3}) = \frac{mk}{r} (\dot{r} - (\hat{r} \cdot \dot{r}) \hat{r})$ .

Therefore $\dot{A} = - \frac{mk}{r} ((\hat{r} \cdot \dot{r}) \hat{r} - \dot{r}) - \frac{mk}{r} (\dot{r} - (\hat{r} \cdot \dot{r}) \hat{r}) = 0$ .

Exercise 3 (medium, symbolic).

For the open Toda lattice with $N = 2$ particles, compute the Lax matrix $L$ and verify that $\dot{L} = [B, L]$ reproduces Hamilton's equations. Find the eigenvalues of $L$ and identify the two conserved quantities.

Hint

With $N = 2$ and open boundary conditions, $L = (b_{1} a_{1} a_{1} b_{2})$ and $B = (0 - a_{1} a_{1} 0)$ . Compute $[B, L]$ and match with $\dot{L}$ from Flaschka's equations.

Answer

The $2 \times 2$ Lax matrix is $L = (b_{1} a_{1} a_{1} b_{2})$ with $a_{1} = \frac{1}{2} e^{- (q_{2} - q_{1}) /2}$ and $b_{n} = - p_{n} /2$ . The partner is $B = (0 - a_{1} a_{1} 0)$ . Computing $[B, L]$ :

[B, L] = (2 a_{1}^{2} a_{1} (b_{2} - b_{1}) a_{1} (b_{2} - b_{1}) - 2 a_{1}^{2}) .

From Flaschka's equations: $\dot{b}_{1} = 2 a_{1}^{2}$ , $\dot{b}_{2} = - 2 a_{1}^{2}$ , $\overset{a}{˙}_{1} = a_{1} (b_{2} - b_{1})$ . These match $\dot{L}_{11} = 2 a_{1}^{2}$ , $\dot{L}_{22} = - 2 a_{1}^{2}$ , $\dot{L}_{12} = a_{1} (b_{2} - b_{1})$ . The eigenvalues of $L$ are $λ_{\pm} = (b_{1} + b_{2}) /2 \pm ((b_{1} - b_{2}) /2)^{2} + a_{1}^{2}$ . The conserved quantities are $F_{1} = tr (L) = b_{1} + b_{2} = - (p_{1} + p_{2}) /2$ (total momentum) and $F_{2} = tr (L^{2}) = b_{1}^{2} + b_{2}^{2} + 2 a_{1}^{2} = (p_{1}^{2} + p_{2}^{2}) /4 + \frac{1}{2} e^{- (q_{2} - q_{1})}$ (proportional to the Hamiltonian).

Exercise 4 (medium, symbolic).

For the Kepler problem in Delaunay variables, show that the frequencies associated with the angles $g$ (argument of perihelion) and $h$ (longitude of ascending node) are zero. Interpret this physically.

Hint

The Hamiltonian is $H = - μ k^{2} / (2 J^{2})$ where $J$ is the principal action. Compute $\partial H / \partial L$ and $\partial H / \partial G$ .

Answer

Since $H = - μ k^{2} / (2 J^{2})$ depends only on $J$ , the frequencies are $ω_{g} = \partial H / \partial L = 0$ and $ω_{h} = \partial H / \partial G = 0$ . Physically, $ω_{g} = 0$ means the argument of perihelion is constant: the ellipse does not precess. $ω_{h} = 0$ means the longitude of the ascending node is constant: the orbital plane does not precess. Both are consequences of the Kepler problem's hidden SO(4) symmetry, which produces the degeneracy in the action variables (only one independent combination $J$ appears in $H$ ).

Exercise 5 (hard, symbolic).

For the periodic Toda lattice with $N = 3$ particles, write out the $3 \times 3$ Lax matrix $L$ and compute the three conserved quantities $F_{k} = tr (L^{k})$ for $k = 1, 2, 3$ . Verify that $F_{2}$ is proportional to the Hamiltonian.

Hint

The periodic $3 \times 3$ Lax matrix has entries $a_{3}$ in the $(1, 3)$ and $(3, 1)$ positions. Compute $tr (L)$ , $tr (L^{2}) = \sum_{ij} L_{ij} L_{j i}$ , and $tr (L^{3})$ .

Answer

The Lax matrix is $L = b_{1} a_{1} a_{3} a_{1} b_{2} a_{2} a_{3} a_{2} b_{3}$ where $a_{n} = \frac{1}{2} e^{- (q_{n + 1} - q_{n}) /2}$ with $q_{4} = q_{1}$ and $b_{n} = - p_{n} /2$ .

$F_{1} = tr (L) = b_{1} + b_{2} + b_{3} = - \frac{1}{2} (p_{1} + p_{2} + p_{3})$ (total momentum).

$F_{2} = tr (L^{2}) = b_{1}^{2} + b_{2}^{2} + b_{3}^{2} + 2 (a_{1}^{2} + a_{2}^{2} + a_{3}^{2}) = \frac{1}{4} (p_{1}^{2} + p_{2}^{2} + p_{3}^{2}) + \frac{1}{2} (e^{- (q_{2} - q_{1})} + e^{- (q_{3} - q_{2})} + e^{- (q_{1} - q_{3})})$ . This is $\frac{1}{2} H + \frac{3}{2}$ , confirming $F_{2}$ is proportional to the Hamiltonian (up to a constant from the subtraction of 1 in the Toda potential).

$F_{3} = tr (L^{3}) = b_{1}^{3} + b_{2}^{3} + b_{3}^{3} + 3 b_{1} (a_{1}^{2} + a_{3}^{2}) + 3 b_{2} (a_{1}^{2} + a_{2}^{2}) + 3 b_{3} (a_{2}^{2} + a_{3}^{2}) + 6 a_{1} a_{2} a_{3}$ . This is a cubic expression in momenta and exponential potentials, providing the third independent integral.

Exercise 6 (hard, symbolic).

Show that the eigenvalues of the Lax matrix $L$ for the open Toda lattice are all real. Use the fact that $L$ is a real symmetric tridiagonal matrix.

Hint

A real symmetric matrix is diagonalisable with real eigenvalues. Verify that $L = L^{T}$ from the definition in terms of Flaschka variables.

Answer

The Flaschka variables $a_{n} = \frac{1}{2} e^{- (q_{n + 1} - q_{n}) /2}$ are strictly positive real numbers, and $b_{n} = - p_{n} /2$ are real. The open Toda lattice Lax matrix is tridiagonal with $L_{nn} = b_{n}$ on the diagonal and $L_{n, n + 1} = L_{n + 1, n} = a_{n}$ on the off-diagonals. Since all entries are real and $L_{ij} = L_{j i}$ , the matrix $L$ is real symmetric. By the spectral theorem for real symmetric matrices, $L$ has $N$ real eigenvalues (counting multiplicity) and is orthogonally diagonalisable. Since the Lax equation preserves the spectrum, all $N$ conserved quantities $λ_{1}, \dots, λ_{N}$ are real-valued functions on phase space.

Exercise 7 (hard, symbolic).

Compute the Poisson bracket ${A_{x}, L_{z}}$ for the Kepler problem, where $A_{x}$ is the $x$ -component of the LRL vector and $L_{z}$ is the $z$ -component of angular momentum. Use this to show that the LRL vector and angular momentum do not Poisson-commute.

Hint

Use the Poisson bracket identity ${f, g} = \frac{\partial f}{\partial r} \cdot \frac{\partial g}{\partial p} - \frac{\partial f}{\partial p} \cdot \frac{\partial g}{\partial r}$ . Expand $A_{x} = (p_{y} L_{z} - p_{z} L_{y}) - mk x / r$ and compute term by term.

Answer

The $x$ -component of the LRL vector is $A_{x} = (p \times L)_{x} - mk x / r = p_{y} L_{z} - p_{z} L_{y} - mk x / r$ . Using the standard angular momentum $L_{z} = x p_{y} - y p_{x}$ and $L_{y} = z p_{x} - x p_{z}$ :

${A_{x}, L_{z}} = {p_{y} L_{z} - p_{z} L_{y} - mk x / r, x p_{y} - y p_{x}}$ .

Computing the Poisson brackets of the individual pieces using the canonical relations ${x_{i}, p_{j}} = δ_{ij}$ :

${p_{y} L_{z}, x p_{y}} = p_{y} {L_{z}, x} p_{y} + {p_{y}, x p_{y}} L_{z}$ . Since ${L_{z}, x} = - y$ and ${p_{y}, x p_{y}} = - p_{y}$ : this gives $- p_{y}^{2} y - p_{y} L_{z}$ .

Continuing with all terms and collecting (a lengthy but straightforward computation): ${A_{x}, L_{z}} = - A_{y}$ , where $A_{y}$ is the $y$ -component of the LRL vector. This is one of the $so (4)$ commutation relations. Since $A_{y} \neq = 0$ in general, the LRL vector does not Poisson-commute with angular momentum.

Exercise 8 (hard, symbolic).

For the open Toda lattice with $N$ particles, show that as $t \to + \infty$ , the particles separate into $N$ free particles: $q_{n} (t) \sim v_{n} t + q_{n}^{0}$ for distinct asymptotic velocities $v_{1} < v_{2} < \dots < v_{N}$ . Relate the asymptotic velocities to the eigenvalues of $L$ .

Hint

Diagonalise $L (0) = Q Λ Q^{- 1}$ with $Λ = diag (λ_{1}, \dots, λ_{N})$ . The solution $L (t) = U (t) L (0) U (t)^{- 1}$ means the off-diagonal entries $a_{n} (t) \to 0$ as $t \to \infty$ , and $b_{n} (t) \to$ (ordered eigenvalues). The momenta $p_{n} = - 2 b_{n}$ approach $- 2 λ_{σ (n)}$ where $σ$ is the ordering permutation.

Answer

The solution to the Lax equation $\dot{L} = [B, L]$ with $L (0) = Q Λ Q^{- 1}$ gives $L (t) = U (t) Q Λ (U (t) Q)^{- 1}$ . Writing $U (t) Q = P (t) exp (t Λ)$ (QR factorisation), we get $L (t) = P (t) Λ P (t)^{- 1}$ . For large $t$ , the matrix factorisation selects the ordering of eigenvalues by size. The diagonal entries $b_{n} (t) = L_{nn} (t)$ approach the eigenvalues $λ_{1} < λ_{2} < \dots < λ_{N}$ (after ordering), while the off-diagonal entries $a_{n} (t) = L_{n, n + 1} (t) \to 0$ exponentially fast. Since $p_{n} = - 2 b_{n}$ , the momenta approach $p_{n} \to - 2 λ_{n}$ , and $q_{n} (t) \sim - 2 λ_{n} t + const$ . The asymptotic velocities are $v_{n} = - 2 λ_{n}$ , related to the eigenvalues of the Lax matrix at $t = 0$ . This is the "inverse scattering" map: the scattering data (asymptotic velocities) are the eigenvalues of $L$ .

Lean formalization Intermediate+

lean_status: none. Mathlib has real analysis, matrix theory, and Lie algebra machinery, but does NOT contain:

The Laplace-Runge-Lenz vector. Defining $A = p \times L - mk \hat{r}$ requires the cross product, the angular momentum, and the unit radial vector. Proving $\dot{A} = 0$ requires Hamilton's equations for the Kepler problem and vector calculus identities (triple product, derivative of $\hat{r}$ ).
The SO(4) symmetry algebra. The Poisson bracket relations ${A_{i}, A_{j}} = - 2 m H ϵ_{ij k} A_{k}$ and ${A_{i}, L_{j}} = ϵ_{ij k} A_{k}$ define a Lie algebra isomorphic to $so (4)$ for $E < 0$ . Formalising this requires the Poisson bracket on $R^{6}$ and verification of the Jacobi identity for the extended algebra.
Delaunay elements. The action-angle coordinates for the Kepler problem require evaluating loop integrals over the invariant tori, inverting the relation $E = - μ k^{2} / (2 J^{2})$ , and verifying the Darboux property.
The Toda lattice Hamiltonian. The phase space $R^{2 N}$ with the standard symplectic form, the Toda Hamiltonian, and Flaschka's change of variables require explicit construction.
The Lax pair. Defining the Jacobi matrix $L$ and its partner $B$ , verifying the Lax equation $\dot{L} = [B, L]$ reproduces Hamilton's equations, and proving the eigenvalues are constants all require matrix differential equations and the spectral theorem for symmetric matrices.
Commuting integrals from the Lax pair. Proving ${tr (L^{k}), tr (L^{j})} = 0$ requires the Adler-Kostant-Symes theorem or an explicit computation using the Poisson bracket structure of the Toda lattice.

The most accessible formalization target would be Exercise 3 (the $N = 2$ open Toda lattice), which requires only $2 \times 2$ matrix computations. This unit ships without a Lean module and is reviewer-attested.

Advanced results Master

Inverse scattering for the Toda lattice

The Lax pair formulation of the Toda lattice is the finite-dimensional analogue of the inverse scattering transform for PDEs. The key idea is that the dynamics of the Toda lattice can be solved in three steps:

Direct scattering. Given initial data $(q_{n} (0), p_{n} (0))$ , construct the Lax matrix $L (0)$ and compute its spectral data (eigenvalues and eigenvectors, or equivalently, the scattering data).
Time evolution of scattering data. The eigenvalues are constant (isospectral deformation). The eigenvectors evolve simply: if $L = Q Λ Q^{- 1}$ , then $\dot{Q} = B Q$ , which gives the eigenvector evolution.
Inverse scattering. Reconstruct $L (t)$ from the time-evolved spectral data, and extract $q_{n} (t)$ and $p_{n} (t)$ from the entries of $L (t)$ .

For the open Toda lattice, the inverse step is the classical inverse problem for Jacobi matrices: a real symmetric tridiagonal matrix is uniquely determined by its eigenvalues and the first components of its eigenvectors. This data (eigenvalues plus "normalisation constants") constitutes the scattering data, and its time evolution is linear (each normalisation constant evolves as $exp (λ_{k} t)$ ). The inverse step can be solved by the continued-fraction algorithm or by the Lanczos algorithm ^{[source pending]}.

For the periodic Toda lattice, the spectral data is more intricate: the eigenvalues of the periodic Jacobi matrix form bands on the real axis, and the spectral curve is a hyperelliptic Riemann surface. The action variables are given by line integrals on this Riemann surface, and the angle variables are the associated Abel-map coordinates. This connects the Toda lattice to the theory of Riemann surfaces and the Jacobi inversion problem, and is an example of algebraic integrability ^{[source pending]}.

Soliton solutions

The Toda lattice supports soliton solutions: localised disturbances that propagate without changing shape and survive collisions with other solitons, emerging with only a phase shift. For the open Toda lattice, a single-soliton solution has the form

q_{n} (t) = lo g \frac{1 + exp ( 2 κ ( n - 1 ) - 2 sinh ( κ ) t + δ _{1} )}{1 + exp ( 2 κn - 2 sinh ( κ ) t + δ _{1} )}

for a parameter $κ > 0$ and phase $δ_{1}$ . The soliton has amplitude proportional to $κ$ and velocity $sinh (κ) / κ$ . As $κ \to 0$ , the soliton becomes a low-amplitude, long-wavelength wave; as $κ \to \infty$ , it becomes a sharp, fast pulse.

Multi-soliton solutions exist for any number of solitons and are obtained from the inverse scattering construction by choosing initial data that produces a finite number of discrete eigenvalues (plus a continuous spectrum). The interaction of two solitons is completely elastic: after collision, both solitons recover their original shapes, with only a time delay (phase shift). The phase shift depends on the relative velocity and amplitude of the two solitons.

The soliton solutions of the Toda lattice are the discrete analogue of the soliton solutions of the KdV equation. The connection is via the continuous limit: as the lattice spacing goes to zero, the Toda lattice converges to the KdV equation (in a moving frame), and the discrete solitons converge to the KdV solitons ^{[source pending]}.

Algebraic integrability

The periodic Toda lattice is more than Liouville integrable: it is algebraically integrable (also called "completely integrable in the sense of Adler-van Moerbeke"). This means the invariant tori are real parts of complex algebraic varieties (Abelian varieties), and the flows can be expressed in terms of theta functions on these varieties.

For the periodic Toda lattice with $N$ particles, the spectral curve of the Lax matrix is

Γ : det (L - λ I) = 0,

which is a polynomial equation in $λ$ whose coefficients are the conserved quantities $F_{1}, \dots, F_{N}$ . When the integrals take generic values, this curve is a smooth hyperelliptic Riemann surface of genus $g = N - 1$ . The real part of the Jacobian of this curve is an $(N - 1)$ -dimensional torus, and the flows of the Toda lattice are straight-line flows on this Jacobian.

The algebraic structure provides more than the Liouville-Arnold theorem: it gives explicit formulas for the solution in terms of Riemann theta functions, and it identifies the linearising variables (the Abel-map coordinates on the Jacobian) as the natural angle variables. The relation between the Liouville tori and the Abel varieties is a deep connection between Hamiltonian mechanics and algebraic geometry ^{[source pending]}.

The Euler top

The Euler top (free rigid body) is one of the oldest integrable systems. A rigid body rotating freely about its centre of mass has the Hamiltonian

H = \frac{L _{1}^{2}}{2 I _{1}} + \frac{L _{2}^{2}}{2 I _{2}} + \frac{L _{3}^{2}}{2 I _{3}},

where $L_{1}, L_{2}, L_{3}$ are the body-frame components of angular momentum and $I_{1}, I_{2}, I_{3}$ are the principal moments of inertia. The phase space is $T^{*} SO (3) ≅ SO (3) \times R^{3}$ with the Lie-Poisson bracket

{L_{i}, L_{j}} = ϵ_{ij k} L_{k} .

The Euler top has two conserved quantities: the energy $H$ and the squared angular momentum $C = L_{1}^{2} + L_{2}^{2} + L_{3}^{2}$ . These two Poisson-commute (since ${H, C} = 0$ by rotational invariance), and the system has effectively two degrees of freedom (after reducing by the SO(3) symmetry), so two integrals suffice for integrability. The solution is given by Jacobi elliptic functions, and the motion on the angular momentum sphere traces out curves given by the intersection of an ellipsoid ( $H = const$ ) and a sphere ( $C = const$ ) ^{[source pending]}.

The Lagrange top

The Lagrange top is a symmetric rigid body ( $I_{1} = I_{2}$ ) with one point fixed, spinning under gravity. The Hamiltonian on $T^{*} SO (3)$ is

H = \frac{L _{1}^{2}}{2 I _{1}} + \frac{L _{2}^{2}}{2 I _{1}} + \frac{L _{3}^{2}}{2 I _{3}} + M g l cos θ,

where $M$ is the mass, $g$ is gravitational acceleration, $l$ is the distance from the fixed point to the centre of mass, and $θ$ is the tilt angle. The three conserved quantities are:

$H$ (energy).
$L_{3}$ (angular momentum about the symmetry axis).
$L_{z}$ (angular momentum about the vertical axis).

These three integrals Poisson-commute and are functionally independent, making the Lagrange top completely integrable for $n = 3$ degrees of freedom (reduced from 6 by the SO(2) symmetry of the vertical axis). The solution again involves elliptic functions ^{[source pending]}.

The Calogero-Moser system

The Calogero-Moser system is an integrable $N$ -body problem on the line with inverse-square interactions:

H = \frac{1}{2} n = 1 \sum N p_{n}^{2} + n < m \sum \frac{g ^{2}}{( q _{n} - q _{m} ) ^{2}} .

Despite the singular potential, the system is completely integrable with $N$ commuting integrals. It was solved by Calogero (1969) for $N = 3$ and by Moser (1975) in general, using a Lax pair construction analogous to the Toda lattice ^{[source pending]}.

The Calogero-Moser system has deep connections to representation theory: the Lax pair is related to the adjoint action on Hermitian matrices, and the conserved quantities are the eigenvalues of a matrix $L$ that is the sum of a diagonal momentum matrix and an off-diagonal position-dependent matrix. The system is also related to the root system of $sl (N)$ : the $1/ r^{2}$ potential is the potential associated with the $A_{N - 1}$ root system, and generalisations to other root systems (BC, D) give integrable systems associated with other Lie algebras.

The Calogero-Moser system with an additional harmonic confining potential ( $\frac{1}{2} ω^{2} q_{n}^{2}$ ) is also integrable and is related to the quantum Hall effect and random matrix theory through the Seiberg-Witten map.

Synthesis

The Kepler problem and the Toda lattice represent two archetypal paths to integrability. The Kepler problem is integrable because of a high degree of symmetry (SO(4) for bound orbits, SO(3,1) for scattering orbits) that produces more conserved quantities than degrees of freedom. The Toda lattice is integrable because of a hidden algebraic structure (the Lax pair) that encodes $N$ commuting integrals in the eigenvalues of a matrix. Both systems illustrate the Liouville-Arnold theorem in action: the commuting integrals foliate the phase space into invariant tori, and the motion on each torus is quasi-periodic.

The additional examples -- Euler top, Lagrange top, Calogero-Moser -- show that integrability arises in diverse physical settings, from rigid-body dynamics to particle systems, and the unifying thread is the existence of sufficiently many commuting integrals. The Lax pair formulation, when available, provides a systematic method for constructing the integrals and solving the dynamics. When no Lax pair is known, integrability must be established by other means (direct construction of integrals, symmetry analysis, or numerical evidence).

Historical context Master

The Kepler problem. The solution of planetary motion by Kepler (1609) and Newton (1687) is one of the founding achievements of mathematical physics. The Laplace-Runge-Lenz vector was known to Laplace (1799) and was studied by Runge (1919) and Lenz (1924) in the context of the old quantum theory. Its significance as a marker of hidden SO(4) symmetry was recognised by Fock (1935) and Bargmann (1936), who showed that the bound-state Kepler problem is equivalent to the free particle on $S^{3}$ via a stereographic-projection mapping (the Kustaanheimo-Stiefel transformation). The action-angle formulation in Delaunay elements goes back to Heniard (1908) and was central to the development of perturbation theory in celestial mechanics (Poincare, Delaunay) ^{[source pending]}.

The Toda lattice. Morikazu Toda introduced the exponential interaction lattice in 1967 and discovered a family of conserved quantities through numerical experimentation ^{[source pending]}. The system attracted intense interest because a lattice with nonlinear, non-polynomial interactions seemed an unlikely candidate for integrability. Flaschka (1974) proved complete integrability by introducing the change of variables that bears his name and constructing the Lax pair. Independently, Manakov (1974) obtained equivalent results. Moser (1975) gave a detailed analysis of the open Toda lattice, proved the scattering is asymptotically free, and established the connection to Jacobi matrix spectral theory.

Solitons and inverse scattering. The Toda lattice is part of the broader story of solitons and the inverse scattering transform. Zabusky and Kruskal (1965) discovered solitons in the KdV equation; Gardner, Greene, Kruskal, and Miura (1967) developed the inverse scattering method for KdV; Lax (1968) formalised the Lax pair framework. The Toda lattice was the first discrete (lattice) system shown to possess a Lax pair and soliton solutions, establishing that integrability is not restricted to continuous PDEs. The inverse scattering method for the Toda lattice was developed independently by Flashka and by Date and Tanaka (1976) ^{[source pending]}.

Algebraic integrability. The notion of algebraic integrability, where the invariant tori are Abelian varieties and the flows linearise on their Jacobians, was developed by Adler and van Moerbeke (1980) and by Mumford (1984). The periodic Toda lattice was one of the first substantive examples, connecting Hamiltonian mechanics to the theory of Riemann surfaces, theta functions, and algebraic curves. This connection has been enormously influential, linking classical mechanics to algebraic geometry, representation theory, and random matrix theory.

The Calogero-Moser system. Calogero (1969) solved the three-body problem with $1/ r^{2}$ interactions and conjectured integrability for general $N$ . Moser (1975) proved it using a Lax pair construction, in the same paper where he analysed the Toda lattice. Kazhdan, Kostant, and Sternberg (1978) showed the Calogero-Moser system is the symplectic reduction of the cotangent bundle of Hermitian matrices by the conjugation action of U(N), providing a representation-theoretic explanation for its integrability. Wilson (1998) and Chalykh (2001) extended this to the "spin Calogero-Moser" systems and connected them to the adelic Grassmannian ^{[source pending]}.

Connections Master

Action-angle variables 09.06.01 -- the Kepler problem and Toda lattice provide the concrete computational realisations of the abstract action-angle machinery. The Delaunay elements are the explicit action-angle coordinates for Kepler; the Flaschka eigenvalues linearise the Toda flow.
Liouville-Arnold theorem 09.06.02 -- the commuting integrals constructed for each system in this unit are specific instances of the general Liouville-Arnold framework. The Kepler tori and Toda tori are Liouville tori.
KAM theorem 09.08.01 -- the Kepler problem in Delaunay elements is the standard starting point for celestial-mechanics perturbation theory. The degeneracy of the Kepler frequencies (only one independent frequency) makes the KAM analysis more delicate and requires the Arnold (isoenergetic) non-degeneracy condition.
Hamiltonian formalism [09.04.01, 09.04.02] -- the LRL vector is best understood in the Hamiltonian framework, where the Poisson bracket reveals the hidden SO(4) symmetry algebra. The Toda lattice Lax pair is a Hamiltonian system on $R^{2 N}$ with the standard symplectic structure.
Dynamics spine [38.07] -- the inverse scattering transform for the Toda lattice is the finite-dimensional analogue of the infinite-dimensional scattering methods developed in the dynamics spine. The spectral theory of Jacobi matrices connects to the spectral theory of Schrodinger operators.
Quantum mechanics -- the SO(4) symmetry of the Kepler problem explains the $n^{2}$ degeneracy of the hydrogen atom energy levels (the famous $1/ n^{2}$ formula). The quantum Toda lattice is an exactly solvable model in statistical mechanics, with applications to the thermodynamics of anharmonic chains.
Lie theory and representation theory -- the LRL vector and the Calogero-Moser system both have deep connections to Lie theory. The SO(4) symmetry of Kepler is the affine counterpart of the SO(3) rotational symmetry. The Calogero-Moser system is the symplectic reduction of the coadjoint orbit of U(N).
Algebraic geometry -- the periodic Toda lattice's spectral curve connects Hamiltonian mechanics to the theory of algebraic curves and their Jacobians. This is the same geometric framework that appears in the theory of integrable PDEs (KdV, KP hierarchy).

Bibliography Master

Goldstein, H., Poole, C., and Safko, J. Classical Mechanics, 3rd ed. Pearson, 2002. Ch. 3.9 (The Laplace-Runge-Lenz vector). ^{[source pending]}
Landau, L. D. and Lifshitz, E. M. Mechanics, 3rd ed. Pergamon, 1976. section 15 (Kepler's problem), section 39 (action-angle variables). ^{[source pending]}
Arnold, V. I. Mathematical Methods of Classical Mechanics, 2nd ed. Springer GTM 60, 1989. section 49. ^{[source pending]}
Toda, M. "Vibration of a chain with nonlinear interaction." J. Phys. Soc. Japan 22 (1967), 431-436. ^{[source pending]}
Flaschka, H. "The Toda lattice as a completely integrable system." Phys. Rev. B 9 (1974), 1924-1925. ^{[source pending]}
Moser, J. "Three integrable Hamiltonian systems connected with isospectral deformations." Adv. Math. 16 (1975), 197-220. ^{[source pending]}
Manakov, S. V. "Complete integrability and stochastization of discrete dynamical systems." Sov. Phys. JETP 40 (1974), 269-274. ^{[source pending]}
Babelon, O., Bernard, D., and Talon, M. Integrable Systems in Classical Mechanics. Cambridge, 2003. ^{[source pending]}
Calogero, F. "Solution of a one-dimensional N-body problem with quadratic/nearest-neighbour interaction." J. Math. Phys. 10 (1969), 2197-2240. ^{[source pending]}
Moser, J. Various Aspects of Integrable Hamiltonian Systems. Birkhauser, 1981. ^{[source pending]}
Fock, V. "Zur Theorie des Wasserstoffatoms." Z. Phys. 98 (1935), 145-154. ^{[source pending]}
Bargmann, V. "Zur Theorie des Wasserstoffatoms." Z. Phys. 99 (1936), 576-582. ^{[source pending]}
Heniard, P. "Sur les elements dits elliptiques des petites planetes." Mem. Roy. Astron. Soc. 28 (1908). ^{[source pending]}
Adler, M. and van Moerbeke, P. "Completely integrable systems, Euclidean Lie algebras, and curves." Adv. Math. 38 (1980), 267-317. ^{[source pending]}
Mumford, D. "Tata lectures on theta II." Prog. Math. 43 (1984). ^{[source pending]}
Kazhdan, D., Kostant, B., and Sternberg, S. "Hamiltonian group actions and dynamical systems of Calogero type." Comm. Pure Appl. Math. 31 (1978), 481-507. ^{[source pending]}
Wilson, G. "Collisions of Calogero-Moser particles and an improved Suslov theorem." Lett. Math. Phys. 44 (1998), 299-312. ^{[source pending]}
Date, E. and Tanaka, S. "Analogue of inverse scattering theory for the discrete Hill's equation and exact solutions for the Toda lattice." Prog. Theor. Phys. 55 (1976), 457-465. ^{[source pending]}
Lax, P. D. "Integrals of nonlinear equations of evolution and solitary waves." Comm. Pure Appl. Math. 21 (1968), 467-490. ^{[source pending]}
Gardner, C. S., Greene, J. M., Kruskal, M. D., and Miura, R. M. "Method for solving the Korteweg-de Vries equation." Phys. Rev. Lett. 19 (1967), 1095-1097. ^{[source pending]}

Prerequisites

09.06.01 pending
09.06.02 pending

Used in

09.08.01

Tier anchors

beginner: Taylor, Classical Mechanics (2005), Ch. 3.5-3.9, Ch. 13; Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Ch. 3
intermediate: Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Ch. 3.9, Ch. 10; Arnold, Mathematical Methods of Classical Mechanics, §49; Toda, Theory of Nonlinear Lattices (1989)
master: Moser, Various Aspects of Integrable Hamiltonian Systems (1981); Babelon, Bernard & Talon, Integrable Systems in Classical Mechanics (2003); Flaschka, Phys. Rev. B 9 (1974)

References

Goldstein, Poole & Safko — Classical Mechanics, 3rd ed. (Pearson, 2002) · Ch. 3 The central-force problem; Ch. 3.9 The Laplace-Runge-Lenz vector
Landau & Lifshitz — Mechanics, 3rd ed. (Course of Theoretical Physics Vol. 1, Pergamon, 1976) · §15 Kepler's problem; §39 The Kepler problem in action-angle variables
Toda — J. Phys. Soc. Japan 22 (1967) · Vibration of a chain with nonlinear interaction
Flaschka — Phys. Rev. B 9 (1974) · The Toda lattice as a completely integrable system
Moser — Dynamical Systems: Theory and Applications, Lecture Notes in Physics 38 (1975) · Three integrable Hamiltonian systems connected with isospectral deformations
Heniard — Mem. Roy. Astron. Soc. 28 (1908) · Sur les elements dits elliptiques des petites planetes
Babelon, Bernard & Talon — Integrable Systems in Classical Mechanics (Cambridge, 2003) · Ch. 2 The Kepler problem; Ch. 6 The Toda lattice
Calogero — J. Math. Phys. 10 (1969) · Solution of a one-dimensional N-body problem
Moser — Adv. Math. 16 (1975) · Three integrable Hamiltonian systems connected with isospectral deformations
Arnold — Mathematical Methods of Classical Mechanics, 2nd ed. (Springer GTM 60, 1989) · §49 Examples of canonical formalism in oscillation theory

Estimated time

beginner: 20m
intermediate: 50m
master: 70m