Kepler Billiards: Dynamics and Integrability

• Kepler billiards are mechanical systems where free-flight segments follow Keplerian orbits under a central Kepler–Coulomb potential instead of straight lines.
• They encompass models like Boltzmann’s line-wall and conic-boundary systems, where walls focused at the Kepler center yield distinct integrability properties.
• Their dynamics span from fully integrable regimes with conserved geometric invariants to chaotic behavior in perturbed and high-energy settings, impacting ergodicity debates.

Searching arXiv for recent and foundational papers on Kepler billiards to ground the article in published work. A Kepler billiard is a mechanical billiard in which the free-flight segments are not straight lines but Keplerian orbits: between impacts the particle moves under a Kepler–Coulomb potential, and at the boundary it reflects elastically. In the planar literature this includes Boltzmann’s line-wall model, billiards bounded by focused conics, and billiards in strictly convex domains with the attraction center in the interior; in higher-dimensional and constant-curvature settings the same term is used for reflective or closely related refractive systems whose interior arcs are Keplerian (Plum et al., 2023, Takeuchi et al., 2021, Baranzini et al., 11 Jul 2025).

1. Mechanical definition and principal model classes

In the general framework of natural mechanical systems, a mechanical billiard is specified by a manifold, a metric, a force function, and a reflection wall; motion follows the mechanical flow in the interior and undergoes elastic reflection at the wall. For the planar Kepler problem the force function is $U=s/r$, so the Hamiltonian is

$H(q,p)=\frac{|p|^2}{2}-\frac{s}{|q|},$

with both attractive and repulsive signs allowed in some treatments (Takeuchi et al., 2021). The reflection law is the standard specular one: the tangential component of momentum is preserved and the normal component changes sign.

One important model is the Kepler specialization of Boltzmann’s billiard system. There the particle moves in the plane under the attractive Kepler potential

$V(r)=-\frac{\alpha}{2r}, \qquad \alpha>0,$

and reflects on the line

$y=\gamma, \qquad \gamma>0,$

which does not pass through the force center. The unreflected motion satisfies

$$\dot r^2+r^2\dot\phi^2 = 2E+\frac{\alpha}{r}, \qquad r^2\dot\phi=C,$$

and the billiard dynamics is obtained by concatenating Kepler arcs with elastic reflections. Energy is preserved, whereas angular momentum generally changes because the wall is not centered at the origin (Plum et al., 2023).

A second major class consists of billiards bounded by conics with one focus at the Kepler center. In that setting the wall may be a focused parabola, focused ellipse, focused hyperbola, or a line, and the system remains a mechanical billiard in the Kepler field rather than a free Birkhoff billiard (Takeuchi et al., 2021). A third class uses a strictly convex bounded planar domain $\Omega$ with an interior attraction center $c$; the billiard map is then defined on the boundary phase space by following the Kepler flow from one impact to the next. In that convex-domain setting the analysis in the literature is concentrated on sufficiently large positive energy (Baranzini et al., 11 Jul 2025).

2. Orbit geometry and the billiard map

The geometry of a Kepler billiard is inherited from the conic geometry of the Kepler problem. For negative energy, bounded planar Kepler trajectories are ellipses; for zero energy they are parabolas; for positive energy they are hyperbolas (Zhao, 18 Jul 2025, Baranzini et al., 11 Jul 2025). In the line-wall Boltzmann model at fixed $E<0$, each free-flight segment is part of an ellipse focused at the origin, and the billiard map jumps from one such ellipse to another after reflection (Plum et al., 2023).

For the planar Kepler case with fixed negative energy, the orbital equation is

$$r=\frac{p}{1+e\cos(\theta-g)}, \qquad p=\frac{2C^2}{\alpha}, \qquad e=\sqrt{1+\frac{8EC^2}{\alpha^2}}.$$

Here $C$ is angular momentum and $H(q,p)=\frac{|p|^2}{2}-\frac{s}{|q|},$0 is the argument of periapsis. In the Boltzmann model this leads to a billiard map

$H(q,p)=\frac{|p|^2}{2}-\frac{s}{|q|},$1

defined on the space of orbit arcs rather than on the usual wall coordinates. The canonical coordinates $H(q,p)=\frac{|p|^2}{2}-\frac{s}{|q|},$2 carry symplectic form

$H(q,p)=\frac{|p|^2}{2}-\frac{s}{|q|},$3

and after fixing energy and reducing by the flow the orbit space carries the reduced symplectic form

$H(q,p)=\frac{|p|^2}{2}-\frac{s}{|q|},$4

The billiard reflection preserves this reduced $H(q,p)=\frac{|p|^2}{2}-\frac{s}{|q|},$5-form, so the billiard map preserves area in $H(q,p)=\frac{|p|^2}{2}-\frac{s}{|q|},$6-space (Plum et al., 2023).

At zero energy the geometry changes from elliptic to parabolic. In the normalized Hamiltonian

$H(q,p)=\frac{|p|^2}{2}-\frac{s}{|q|},$7

the level $H(q,p)=\frac{|p|^2}{2}-\frac{s}{|q|},$8 is the parabolic regime, and free trajectories are parabolas focused at the Kepler center. This is why zero-energy convexity is expressed in terms of intersections with focused parabolas rather than with straight lines (Zhao, 18 Jul 2025).

High-energy convex-domain Kepler billiards admit another geometric description. For large $H(q,p)=\frac{|p|^2}{2}-\frac{s}{|q|},$9, the Jacobi lengths of direct and indirect Kepler arcs satisfy

$V(r)=-\frac{\alpha}{2r}, \qquad \alpha>0,$0

$V(r)=-\frac{\alpha}{2r}, \qquad \alpha>0,$1

Thus direct arcs approximate Euclidean chords, while indirect arcs approximate broken paths through the singular center. This is the geometric basis of the “punctured Birkhoff” picture used in high-energy non-integrability results (Baranzini et al., 11 Jul 2025).

3. Integrability, conserved quantities, and focal geometry

The central structural fact is that the pure Kepler line-wall billiard is integrable rather than ergodic. The Boltzmann system was historically proposed as an illustration of the ergodic hypothesis, but later work showed that the Kepler case is integrable; the 2023 analysis treats this case as the distinguished benchmark against which nonzero $V(r)=-\frac{\alpha}{2r}, \qquad \alpha>0,$2 perturbations are compared (Plum et al., 2023). A conceptual explanation is given by projective dynamics: the line-wall Kepler billiard is projectively related to a spherical Kepler–Coulomb system, and the extra first integral is the energy of the corresponding spherical problem pulled back to the planar phase space (Zhao, 2020).

In the line-wall model this additional conserved quantity is written as

$V(r)=-\frac{\alpha}{2r}, \qquad \alpha>0,$3

where $V(r)=-\frac{\alpha}{2r}, \qquad \alpha>0,$4 is angular momentum and $V(r)=-\frac{\alpha}{2r}, \qquad \alpha>0,$5 is a component of the Laplace–Runge–Lenz vector; in the broader confocal formulation it appears as

$V(r)=-\frac{\alpha}{2r}, \qquad \alpha>0,$6

The literature summarized in later work presents this same quantity as common to integrable Kepler billiards and certain partially averaged systems from secular three-body theory, with a common projective-dynamical origin in two-center geometry (Zhao, 2020, Pinzari et al., 24 Apr 2025).

Conic boundaries focused at the Kepler center supply a broad class of integrable examples. The conformal Hooke–Kepler correspondence shows that any focused parabola, focused ellipse, focused hyperbola, and line can serve as an integrable Kepler reflection wall on the appropriate fixed energy level, with the Gallavotti–Jauslin invariant providing the additional first integral (Takeuchi et al., 2021). This extends the original line-wall result to a large focused-conic family.

For these focused-conic billiards, integrability has a sharp focal reformulation. If $V(r)=-\frac{\alpha}{2r}, \qquad \alpha>0,$7 are the successive Kepler flight conics and $V(r)=-\frac{\alpha}{2r}, \qquad \alpha>0,$8 their second foci, then all $V(r)=-\frac{\alpha}{2r}, \qquad \alpha>0,$9 lie on a fixed circle centered at the second focus $y=\gamma, \qquad \gamma>0,$0 of the wall conic: $y=\gamma, \qquad \gamma>0,$1 The same work shows that the directrices of the flight conics form a one-parameter family whose envelope is again a conic, and it interprets the additional first integral geometrically as the radius of this circle of second foci (Jaud et al., 2023).

In the planar elliptic integrable case, this focal description becomes a Poncelet-type theorem. For nonzero energy, the lines joining consecutive second orbital foci are all tangent to a fixed circle, called the foci-caustic circle. The dynamics then linearizes on an elliptic curve

$y=\gamma, \qquad \gamma>0,$2

and $y=\gamma, \qquad \gamma>0,$3-periodicity is characterized by Cayley-type determinant conditions. This makes the planar elliptic Kepler billiard a Poncelet system in the auxiliary geometry of second foci rather than in the physical trajectory plane (Jaud et al., 9 Nov 2025).

4. Rigidity and classification results

The integrable examples are accompanied by strong rigidity theorems. At zero energy, the natural class is that of $y=\gamma, \qquad \gamma>0,$4-convex tables: smooth closed connected simple curves containing the center and intersecting each focused parabola in at most two points. In this setting the classification theorem states that if a planar zero-energy Kepler billiard in a $y=\gamma, \qquad \gamma>0,$5-convex table admits an additional $y=\gamma, \qquad \gamma>0,$6-first integral on the zero-energy level, then the boundary is an ellipse focused at the Kepler center (Zhao, 18 Jul 2025).

The proof uses Levi-Civita regularization. After multiplying the Hamiltonian by $y=\gamma, \qquad \gamma>0,$7 on $y=\gamma, \qquad \gamma>0,$8 and applying

$y=\gamma, \qquad \gamma>0,$9

the zero-energy Kepler problem becomes free Euclidean motion with Hamiltonian

$$\dot r^2+r^2\dot\phi^2 = 2E+\frac{\alpha}{r}, \qquad r^2\dot\phi=C,$$0

at energy $$\dot r^2+r^2\dot\phi^2 = 2E+\frac{\alpha}{r}, \qquad r^2\dot\phi=C,$$1. Focused parabolas become lines, $$\dot r^2+r^2\dot\phi^2 = 2E+\frac{\alpha}{r}, \qquad r^2\dot\phi=C,$$2-convex domains become strictly convex centrally symmetric domains, and the billiard reflection law is preserved because the square map is conformal. The classification is then reduced to the Bialy–Mironov rigidity theorem for centrally symmetric Birkhoff billiards (Zhao, 18 Jul 2025).

A second rigidity result concerns high energies in real-analytic strictly convex planar domains. Except for possibly one placement of the attraction center in a non-elliptic table, analytic integrability at sufficiently large positive energy occurs only when the table is an ellipse and the center is located at one of its foci. If the boundary is an ellipse, the exceptional positions are exactly the two foci (Baranzini et al., 11 Jul 2025). The argument constructs symbolic dynamics by shadowing chains of punctured Birkhoff-type trajectories and derives positive topological entropy, which precludes analytic first integrals of the first return map.

This high-energy theory introduces focal points of the second kind. For a real-analytic boundary there are at most two such points; if there are exactly two, the boundary is a non-circular ellipse and the two points are its foci (Baranzini et al., 11 Jul 2025). This gives a partial affirmative answer to a Keplerian analogue of the Birkhoff–Poritsky conjecture.

5. Extensions beyond the basic planar model

The planar theory admits several nontrivial extensions. In three dimensions, the Kustaanheimo–Stiefel transformation relates the $$\dot r^2+r^2\dot\phi^2 = 2E+\frac{\alpha}{r}, \qquad r^2\dot\phi=C,$$3-dimensional Kepler problem to the $$\dot r^2+r^2\dot\phi^2 = 2E+\frac{\alpha}{r}, \qquad r^2\dot\phi=C,$$4-dimensional isotropic harmonic oscillator. Under this correspondence, $$\dot r^2+r^2\dot\phi^2 = 2E+\frac{\alpha}{r}, \qquad r^2\dot\phi=C,$$5-invariant centered quadrics in $$\dot r^2+r^2\dot\phi^2 = 2E+\frac{\alpha}{r}, \qquad r^2\dot\phi=C,$$6-space project to integrable spatial Kepler billiard walls in $$\dot r^2+r^2\dot\phi^2 = 2E+\frac{\alpha}{r}, \qquad r^2\dot\phi=C,$$7: a plane, a centered sphere, a spheroid, a sheet of a two-sheeted circular hyperboloid, or a paraboloid, with the Kepler center as a focus in the latter three cases. The corresponding spatial Kepler billiards are integrable because they descend from integrable oscillator billiards (Takeuchi et al., 2023).

A broader higher-dimensional construction uses Lagrange’s integrable extension of Euler’s two-center problem in Euclidean space, on the sphere, and in hyperbolic space. In that setting any finite combination of quadrics focused at the Kepler centers yields an integrable mechanical billiard, and the same holds in arbitrary dimension $$\dot r^2+r^2\dot\phi^2 = 2E+\frac{\alpha}{r}, \qquad r^2\dot\phi=C,$$8 by central projection to the constant-curvature models. These systems include one-center Kepler billiards, two-center Kepler billiards, and their spherical and hyperbolic analogues (Takeuchi et al., 2023).

There are also related refractive systems. In the planar refraction model with piecewise potential

$$\dot r^2+r^2\dot\phi^2 = 2E+\frac{\alpha}{r}, \qquad r^2\dot\phi=C,$$9

trajectories alternate inner Kepler hyperbolae with outer harmonic ellipses, and the interface obeys the generalized Snell law

$\Omega$0

For a circular interface the first return map is integrable,

$\Omega$1

and perturbations of the circle are treated by KAM, Poincaré–Birkhoff, and Aubry–Mather theory (Blasi et al., 2021). This is not an elastic Kepler billiard in the strict sense, but it is a closely related piecewise-Kepler boundary dynamics.

A further $\Omega$2-dimensional extension studies reflective and refractive Keplerian billiards in a bounded star-shaped domain $\Omega$3. The reflective model concatenates hyperbolic Kepler arcs with elastic reflections; the refractive model alternates an inner Keplerian regime with an outer harmonic one. Central configurations on the boundary produce homothetic equilibrium trajectories, principal curvatures determine their linear stability, and for sufficiently large inner energy the first return map contains an invariant subsystem topologically conjugate to the Bernoulli shift on two symbols (Blasi, 2024).

6. Dynamical regimes, numerics, and the ergodicity question

The modern theory separates sharply between integrable Kepler billiards and nearby or higher-energy regimes that display chaos. In the Boltzmann family with potential

$\Omega$4

the case $\Omega$5 is the pure Kepler billiard and is integrable. Numerical experiments at

$\Omega$6

show periodic behavior of the mapping trajectory for $\Omega$7, and the associated discretized Koopman operator has large multiplicity for the eigenvalue $\Omega$8, with invariant subsets consisting of periodic trajectories. For $\Omega$9 the system still appears quasi-periodic, while at $c$0 mixed behavior appears, and at $c$1 one orbit seems to cover the whole allowed region densely and the discretized Koopman problem seems to have only one simple eigenvalue near $c$2, suggesting possible ergodicity (Plum et al., 2023).

A recurring historical contrast follows from these results. Boltzmann proposed his line-wall central-force billiard as an illustration of the ergodic hypothesis, but the pure Kepler realization is not ergodic: it is integrable and carries large families of invariant sets (Plum et al., 2023). This is the main correction to the older intuition. At the same time, the nonzero-$c$3 deformations, the high-energy convex-domain models, and the higher-dimensional reflective or refractive systems show that Keplerian billiards can also support symbolic dynamics, positive topological entropy, and topological chaos (Baranzini et al., 11 Jul 2025, Blasi, 2024).

Numerical evidence reinforces this distinction. In the high-energy real-analytic convex-domain theory, even the exceptional non-elliptic tables possessing a focal point of the second kind show chaotic behavior in simulations, together with invariant curves near the boundary. The published theory does not classify that exceptional case as integrable; it leaves it as the only case not excluded by the current symbolic-dynamics argument (Baranzini et al., 11 Jul 2025).

Taken together, these developments give a precise modern picture. A Kepler billiard is not a single model but a family of mechanical billiards organized by Keplerian interior motion, conic and projective geometry, and regularization theory. In some geometries—most notably focused conic walls, zero-energy $c$4-convex ellipses, and the focal elliptic tables of the planar integrable theory—the dynamics is rigidly integrable. In others—especially high-energy non-elliptic domains, perturbed Boltzmann systems, and several higher-dimensional reflective or refractive models—the same Keplerian framework produces chaotic subsystems rather than ergodic or Liouville-regular motion.