A Correspondence between Billiards and Geodesics

From a geometric viewpoint, billiard trajectories and geodesics are related by mutual approximation results. In one direction, it is known that every geodesic curve in the boundary of a smooth convex body can be approximated by a sequence of billiard trajectories inside of it. We establish the other direction by proving that, for Riemannian billiard tables (under mild assumptions), there exists a family of fold-type surfaces such that every sequence of geodesic segments on these surfaces has a subsequence that converges to a billiard trajectory in the table. In particular, this is true for convex Euclidean tables. We also describe a more general class of tables to which this result applies and present explicit non-Euclidean examples.

💡 Research Summary

The paper establishes a two‑way approximation relationship between billiard trajectories inside a smooth convex domain and geodesic curves on the domain’s boundary. The forward direction—already known—states that any geodesic on the boundary of a C²‑smooth convex body K ⊂ ℝⁿ can be approximated by a sequence of billiard trajectories whose initial directions converge to the tangent of the geodesic. The authors briefly re‑prove this result (Theorem 1.1) for completeness.

The novel contribution is the converse: given a family of geodesic segments on a family of “fold” hypersurfaces that flatten onto a piece of the billiard table, the limit of these geodesics, as the folds become flat, is a billiard trajectory in the original table. The key construction is a parameterized family of hypersurfaces M_{p,λ} (λ∈(0,1)) defined in a neighbourhood of a boundary point p∈∂K. As λ→0⁺ the hypersurfaces converge in the Hausdorff sense to K∩U, while each M_{p,λ} retains a uniform lower curvature bound κ. This uniform bound allows the use of comparison geometry (Hessian comparison theorem) and the notion of κ‑quasigeodesics.

A κ‑quasigeodesic is a Lipschitz curve whose distance function to any point satisfies a differential inequality derived from the model space of constant curvature κ. In manifolds without boundary, κ‑quasigeodesics coincide with genuine geodesics. With boundary, additional “polar” conditions on the left and right tangent vectors at boundary points are required; these precisely encode the law of reflection (angle of incidence equals angle of reflection).

The main technical steps are:

1. Construction of the folds – Theorem 3.11 shows that the family {M_{p,λ}} flattens onto K∩U with Hausdorff convergence and uniform curvature bound.

2. Uniform limit of quasigeodesics – For any decreasing sequence λ_k→0, pick arclength‑parameterized geodesic segments γ_k on M_{p,λ_k} with γ_k(0)=p. Because each M_{p,λ_k} satisfies the same curvature lower bound, each γ_k is a κ‑quasigeodesic (Proposition 2.15). Proposition 2.18 guarantees that a subsequence converges uniformly to a continuous curve γ in K.

3. Identification as a billiard trajectory – Theorem 3.6 proves that any κ‑quasigeodesic whose left and right tangents at boundary points are polar is exactly a billiard trajectory. The polar condition follows from the construction of the folds and the curvature bound, so the limit curve γ satisfies the reflection law and therefore is a genuine billiard path.

The authors formulate two explicit theorems. Theorem 1.2 treats the Euclidean case: for any convex Euclidean billiard table with C² boundary, any sequence of geodesic segments on the flattening folds has a subsequence converging to a billiard trajectory. Theorem 1.3 extends this to general Riemannian billiard tables under the hypothesis (H) that a family of folds exists with a uniform curvature lower bound.

Section 4 provides concrete examples. In flat Euclidean space the folds are ellipsoids M_λ = {x²+y²+z²/λ²=1} collapsing onto a planar disk, reproducing the classical picture. In constant negative curvature (hyperbolic space) and constant positive curvature (spherical space) analogous families of hypersurfaces are constructed, showing that the correspondence holds in non‑Euclidean geometries as well. The authors also discuss cases where the hypothesis fails (e.g., unbounded curvature or highly twisted folds), indicating the limits of the theory.

Overall, the paper bridges billiard dynamics and Riemannian geometry by showing that billiard trajectories are precisely the limits of geodesic segments on appropriately flattened manifolds. This dual approximation enriches the understanding of billiard systems, opens avenues for studying billiards in curved spaces, and suggests further investigations into curvature bounds, non‑smooth boundaries, and higher‑dimensional generalizations.