Kepler Billiards: Dynamics & Integrability

• Kepler billiards are mechanical systems where a point particle follows Keplerian conic trajectories and reflects elastically on a smooth, closed boundary.
• Integrability emerges in special cases, notably with elliptical boundaries focused at the Kepler center, as shown via Levi–Civita regularization and rigidity theorems.
• Dynamics span from exact integrability at zero energy to high-energy chaotic behavior, highlighting complex interactions between geometry and central-force potentials.

Searching arXiv for papers on Kepler billiards and related integrability results. Kepler billiards are mechanical billiard systems in which a point particle moves between impacts under a Kepler potential and undergoes elastic reflection at a prescribed boundary. In the planar one-center setting, the underlying Hamiltonian is $H(q,p)=\tfrac12|p|^2-\mu/|q|$ with $\mu>0$, and the reflection law preserves the tangential component of momentum while reversing the normal component (Zhao, 18 Jul 2025). The subject combines celestial mechanics, billiard dynamics, projective and conformal regularization, and rigidity theory. Recent work has clarified both the classical integrable examples—primarily conic boundaries focused at the Kepler center—and strong non-integrability results, including a zero-energy classification that identifies ellipses with the force center at a focus as the only smooth integrable planar tables under a natural $K$-convexity hypothesis (Zhao, 18 Jul 2025).

1. Definition and basic dynamical structure

A Kepler billiard in the plane consists of a smooth, simple, closed curve $\Gamma\subset \mathbb R^2$ bounding a domain, together with a distinguished Kepler center $O$, typically placed at the origin, and the Hamiltonian

$$H(p,q)=\frac{|p|^2}{2}-\frac{\mu}{|q|}, \qquad \mu>0,$$

defined on $\mathbb R^2\setminus\{O\}$ (Zhao, 18 Jul 2025). Between reflections, the particle follows a Keplerian conic arc: ellipse, parabola, or hyperbola depending on the energy. At the boundary, the impact is elastic: the angle of incidence equals the angle of reflection, equivalently the tangential component of velocity is preserved and the normal component is reversed (Zhao, 18 Jul 2025).

For positive-energy formulations used in high-energy rigidity, the center may be written as $c\in\Omega$ and the Hamiltonian as

$H(q,p)=\tfrac12|p|^2-\frac{\mu}{|q-c|},$

with the billiard map encoded on $\partial\Omega\times(0,\pi)$ by arc-length and incidence-angle coordinates; in that formulation the map is area-preserving (Baranzini et al., 11 Jul 2025). This is the natural Keplerian analogue of the Birkhoff billiard map.

The zero-energy level $\mu>0$0 is distinguished because all free-flight arcs are parabolas focused at the Kepler center (Zhao, 18 Jul 2025). In contrast, for fixed negative energy, the free flights are elliptic arcs with a common semi-major axis $\mu>0$1 in the focused-conic integrable setting (Jaud et al., 2023). This energy dependence is central to the modern theory: zero energy is amenable to exact regularization into free motion, while high positive energies admit asymptotic comparison with punctured Birkhoff billiards (Zhao, 18 Jul 2025, Baranzini et al., 11 Jul 2025).

2. Classical integrable examples with conic boundaries

The standard integrable Kepler billiards arise when the reflection wall is a conic section with one focus at the Kepler center. In Cartesian coordinates, a focused ellipse or hyperbola may be written as

$\mu>0$2

with foci $\mu>0$3 and $\mu>0$4 in the elliptic case (Jaud et al., 2023). In the broader conformal formulation, the wall is a single branch of a conic with one focus at the origin, and every ellipse or hyperbola focused at the origin defines an integrable Kepler billiard (Takeuchi et al., 2021).

These systems possess a second first integral beyond energy. One formulation uses angular momentum $\mu>0$5 and the Runge–Lenz vector $\mu>0$6. For a focused conic wall, the invariant can be written as

$\mu>0$7

where $\mu>0$8 is the appropriate component of the Runge–Lenz vector (Jaud et al., 2023). An equivalent Cartesian expression is

$\mu>0$9

which is preserved both by the Kepler flow and by elastic reflections at the focused conic (Takeuchi et al., 2021).

A geometric hallmark of integrability is the “second-focus circle.” If $K$0 are the successive Kepler ellipses between reflections, each with one focus at the Kepler center and second focus $K$1, then all $K$2 lie on a fixed circle centered at the second focus $K$3 of the reflecting conic: $K$4 (Jaud et al., 2023). This converts the extra first integral into an explicitly geometric invariant. The same paper shows that the family of flight ellipses is confined between two confocal conics $K$5, providing caustic-type envelope curves, and analyzes the envelope of the directrices of the flight ellipses (Jaud et al., 2023).

A plausible implication is that focused conics furnish the canonical Liouville-integrable model for one-center Kepler billiards in much the same way that confocal quadrics do for free Birkhoff billiards. That interpretation is consistent with later rigidity results identifying ellipses as the only smooth integrable planar zero-energy tables under natural convexity assumptions (Zhao, 18 Jul 2025).

3. Zero-energy regime and Levi–Civita regularization

At zero energy, the planar Kepler billiard admits a particularly transparent regularization through the Levi–Civita map

$K$6

combined with the rescaled Hamiltonian

$K$7

which becomes

$K$8

on the regularized side (Zhao, 18 Jul 2025). Under this transformation, focused parabolic Kepler arcs in the $K$9-plane lift to straight line segments in the $\Gamma\subset \mathbb R^2$0-plane, and the reflection law is conformally preserved (Zhao, 18 Jul 2025).

The boundary $\Gamma\subset \mathbb R^2$1 lifts to the two-sheeted preimage

$\Gamma\subset \mathbb R^2$2

which is centrally symmetric. If every focused parabola meets $\Gamma\subset \mathbb R^2$3 in at most two points—the $\Gamma\subset \mathbb R^2$4-convexity condition—then every line meets $\Gamma\subset \mathbb R^2$5 in at most two points, which implies strict convexity of $\Gamma\subset \mathbb R^2$6 (Zhao, 18 Jul 2025). Thus the zero-energy Kepler billiard becomes a standard Birkhoff billiard in a centrally symmetric strictly convex domain.

This reduction is the foundation of the zero-energy classification theorem. If the original Kepler billiard admits a nontrivial $\Gamma\subset \mathbb R^2$7 first integral independent of the Hamiltonian, then the pulled-back Birkhoff billiard also admits a nontrivial $\Gamma\subset \mathbb R^2$8 first integral (Zhao, 18 Jul 2025). The problem is thereby transferred from a singular central-force billiard to an ordinary free billiard.

The special role of zero energy had already appeared in conformal correspondence results relating Kepler and Hooke dynamics by the square map $\Gamma\subset \mathbb R^2$9, but the 2025 zero-energy theorem makes this regularization the core of a rigidity argument rather than merely a source of examples (Takeuchi et al., 2021, Zhao, 18 Jul 2025).

4. Zero-energy rigidity and classification

The principal rigidity result states that if $O$0 is a smooth $O$1, simple, closed curve satisfying the $O$2-convexity condition that every Kepler parabolic orbit focused at the origin meets $O$3 in at most two points, and if the zero-energy Kepler billiard inside $O$4 admits a $O$5 first integral not depending only on the Hamiltonian, then $O$6 must be an ellipse with one focus at the Kepler center (Zhao, 18 Jul 2025).

The proof is a direct translation of a theorem of Bialy–Mironov via the Levi–Civita map. Once the lifted boundary $O$7 is shown to be a strictly convex, centrally symmetric, $O$8 curve, the Bialy–Mironov rigidity theorem applies: a Birkhoff billiard in $O$9 bounded by a $$H(p,q)=\frac{|p|^2}{2}-\frac{\mu}{|q|}, \qquad \mu>0,$$0, strictly convex, centrally symmetric curve and admitting a nontrivial $$H(p,q)=\frac{|p|^2}{2}-\frac{\mu}{|q|}, \qquad \mu>0,$$1 first integral near the boundary must have elliptical boundary (Zhao, 18 Jul 2025). Since $$H(p,q)=\frac{|p|^2}{2}-\frac{\mu}{|q|}, \qquad \mu>0,$$2 is then an ellipse in the $$H(p,q)=\frac{|p|^2}{2}-\frac{\mu}{|q|}, \qquad \mu>0,$$3-plane, its image under $$H(p,q)=\frac{|p|^2}{2}-\frac{\mu}{|q|}, \qquad \mu>0,$$4 is an ellipse in the original $$H(p,q)=\frac{|p|^2}{2}-\frac{\mu}{|q|}, \qquad \mu>0,$$5-plane, and one focus is the origin (Zhao, 18 Jul 2025).

This settles the zero-energy case of the conjectural classification that the only integrable planar Kepler billiards are the classical focused conics (Zhao, 18 Jul 2025). The result is exact rather than perturbative, requires only smoothness and $$H(p,q)=\frac{|p|^2}{2}-\frac{\mu}{|q|}, \qquad \mu>0,$$6-convexity, and does not depend on analyticity. It also sharpens the geometric meaning of zero-energy integrability: any nontrivial first integral forces a hidden free-billiard structure whose only compatible smooth geometry is elliptic.

A common misconception is that the presence of a Kepler potential should enlarge the class of integrable boundaries because Kepler motion itself is maximally superintegrable. The zero-energy classification shows the opposite in the smooth planar setting: the combination of Kepler arcs and elastic reflection is sufficiently rigid that integrability collapses to the focused ellipse (Zhao, 18 Jul 2025).

5. High-energy non-integrability and the Keplerian Birkhoff conjecture

A complementary line of work treats analytic integrability at high positive energies. For a strictly convex planar domain $$H(p,q)=\frac{|p|^2}{2}-\frac{\mu}{|q|}, \qquad \mu>0,$$7 with real-analytic boundary, there exists $$H(p,q)=\frac{|p|^2}{2}-\frac{\mu}{|q|}, \qquad \mu>0,$$8 such that for all $$H(p,q)=\frac{|p|^2}{2}-\frac{\mu}{|q|}, \qquad \mu>0,$$9, if $\mathbb R^2\setminus\{O\}$0 is not an ellipse then for every attractive center $\mathbb R^2\setminus\{O\}$1 except at most one point, the Kepler billiard at energy $\mathbb R^2\setminus\{O\}$2 does not admit an analytic first integral; if $\mathbb R^2\setminus\{O\}$3 is an ellipse, then exactly the two foci are analytically integrable at all energies (Baranzini et al., 11 Jul 2025).

The mechanism is symbolic dynamics. The generating functions for large $\mathbb R^2\setminus\{O\}$4 split into “direct” and “indirect” branches,

$\mathbb R^2\setminus\{O\}$5

$\mathbb R^2\setminus\{O\}$6

with further clockwise and anti-clockwise branches near antipodal configurations (Baranzini et al., 11 Jul 2025). In the high-energy limit, the Kepler billiard is approximated by a “punctured Birkhoff billiard” in which line segments may pass through the center. From this limit one constructs three kinds of chaotic building blocks: those based on first-kind points, nondegenerate punctured triangles, and degenerate punctured triangles (Baranzini et al., 11 Jul 2025).

In each case, one obtains an invariant set on which the induced dynamics is semi-conjugate to the full two-symbol Bernoulli shift. Positive topological entropy follows, excluding analytic integrability (Baranzini et al., 11 Jul 2025). This gives a partial affirmative answer to a Keplerian analogue of the classical Birkhoff–Poritsky conjecture.

The same work introduces focal points of the second kind, defined by the property that every billiard ray through $\mathbb R^2\setminus\{O\}$7 returns to $\mathbb R^2\setminus\{O\}$8 after two elastic bounces. In a real-analytic, strictly convex table there can be at most two such points; if there are exactly two, the boundary must be an ellipse and they are its classical foci; in a non-elliptic analytic table there is at most one (Baranzini et al., 11 Jul 2025). The existence of an infinite-dimensional family of non-elliptic analytic tables with a single focal point of the second kind explains why the theorem allows at most one exceptional center rather than none (Baranzini et al., 11 Jul 2025).

Taken together with the zero-energy rigidity theorem, these results delineate two complementary regimes. At zero energy, smooth integrability forces an ellipse with a focal center (Zhao, 18 Jul 2025). At sufficiently high positive energy, analytic integrability is again essentially restricted to ellipses with the center at a focus, up to at most one exceptional center in the non-elliptic case (Baranzini et al., 11 Jul 2025).

6. Geometric invariants, projective correspondences, and extensions

The extra first integral $\mathbb R^2\setminus\{O\}$9 reappears in several broader contexts, linking Kepler billiards with projective dynamics and secular celestial mechanics (Pinzari et al., 24 Apr 2025). In that formulation, $c\in\Omega$0 denotes angular momentum, $c\in\Omega$1 a component of the Laplace–Runge–Lenz vector, and $c\in\Omega$2 the focus–wall distance. The same quantity arises in the two-center problem and in a partially averaged system in the secular three-body problem, providing a structural bridge between billiard integrability and classical perturbation theory (Pinzari et al., 24 Apr 2025).

Projective methods also produce integrable Kepler billiards on surfaces of constant curvature. For $c\in\Omega$3, $c\in\Omega$4, or $c\in\Omega$5, with Hamiltonian

$c\in\Omega$6

and reflecting wall taken from a confocal family, the resulting billiard is Liouville-integrable with independent integrals $c\in\Omega$7 and $c\in\Omega$8 (Pinzari et al., 24 Apr 2025). This extends the planar focused-conic picture to spherical and hyperbolic geometries.

A parallel generalization concerns higher-dimensional space forms and two-center mechanical billiards. In $c\in\Omega$9, $H(q,p)=\tfrac12|p|^2-\frac{\mu}{|q-c|},$0, and $H(q,p)=\tfrac12|p|^2-\frac{\mu}{|q-c|},$1, billiards with two Kepler centers and finite unions of confocal quadrics as walls admit a complete involutive set of integrals $H(q,p)=\tfrac12|p|^2-\frac{\mu}{|q-c|},$2 (Takeuchi et al., 2023). Although these are two-center rather than one-center systems, they show that the confocal-quadric paradigm persists in higher-dimensional mechanical billiards.

In dimension three, quaternionic regularization via the Kustaanheimo–Stiefel transformation identifies integrable one-center zero-energy Kepler billiards with $H(q,p)=\tfrac12|p|^2-\frac{\mu}{|q-c|},$3-invariant Hooke billiards in four dimensions. The resulting admissible reflection walls in $H(q,p)=\tfrac12|p|^2-\frac{\mu}{|q-c|},$4 are planes, centered spheres, spheroids with focus at the origin, one sheet of a two-sheeted circular hyperboloid with focus at the origin, and paraboloids with focus at the origin (Takeuchi et al., 2023). This gives a geometric explanation for the rotationally invariant families of known three-dimensional integrable Kepler billiards.

These extensions suggest that the integrable core of the theory is governed less by the specific Newtonian singularity than by regularization-compatible confocal geometry. That is an inference rather than an explicit theorem, but it is consistent with the repeated appearance of conformal, projective, and symplectic reductions across the planar, curved, and higher-dimensional settings (Pinzari et al., 24 Apr 2025, Takeuchi et al., 2023, Takeuchi et al., 2023).

7. Poncelet-type geometry, variants, and chaotic regimes

The integrable planar elliptic Kepler billiard has a refined Poncelet-type structure. For non-zero-energy orbits in a billiard bounded by an ellipse whose one focus is the Kepler center, the lines joining consecutive second foci are tangent to a fixed circle (Jaud et al., 9 Nov 2025). The dynamics can be linearized on an elliptic curve, with one reflection corresponding to translation by a fixed divisor class, and $H(q,p)=\tfrac12|p|^2-\frac{\mu}{|q-c|},$5-periodicity is governed by Cayley-type determinant conditions (Jaud et al., 9 Nov 2025). This places elliptic Kepler billiards within the algebro-geometric tradition of Poncelet porisms and elliptic-curve linearization.

A distinct variant replaces reflection by refraction. In a two-zone model with an inner Kepler region and outer harmonic region, trajectories alternate Keplerian hyperbolae and harmonic ellipses, with the interface obeying a generalized Snell law

$H(q,p)=\tfrac12|p|^2-\frac{\mu}{|q-c|},$6

(Blasi et al., 2021). For a circular interface the return map is integrable with a global action $H(q,p)=\tfrac12|p|^2-\frac{\mu}{|q-c|},$7, and under small perturbations KAM tori and Aubry–Mather sets persist (Blasi et al., 2021). Although these are not reflective Kepler billiards in the strictest sense, they show how Keplerian flight segments remain compatible with twist-map methods once the boundary interaction is modified.

Three-dimensional non-integrable behavior has also been studied directly. For reflective and refractive Keplerian billiards in $H(q,p)=\tfrac12|p|^2-\frac{\mu}{|q-c|},$8, the existence of two distinct non-antipodal nondegenerate central configurations on the boundary implies topological chaos for sufficiently large inner energy, via a symbolic dynamics built from homothetic radial trajectories (Blasi, 2024). A triaxial ellipsoid provides such a chaotic example, despite the complete integrability of the zero-potential ellipsoidal billiard (Blasi, 2024).

These developments clarify a broad point. Integrable Kepler billiards are not generic mechanical billiards with a central potential; they are special constructions tied to focused conics, confocal families, or exact regularization schemes. Outside those structures, the dominant phenomena are positive entropy, symbolic dynamics, broken invariant curves, and KAM/Aubry–Mather remnants rather than Liouville integrability (Baranzini et al., 11 Jul 2025, Blasi, 2024, Blasi et al., 2021).