Liao, Li and Yang (New Astronomy 96 (2022) 101850) combined clean numerical simulation with artificial-neural-network (ANN) regression to continue known planar periodic orbits of the gravitational three-body problem to arbitrary masses, and inferred for the studied BHH-satellite family a 'nearly largest' mass domain S∗S^* beyond which, they claimed, no periodic orbits of that family exist. The open questions left by that formulation, corresponding to the authors' own list of unresolved issues, are: whether the empirical mass domains are definitive or artifacts of search resolution; whether the mass-to-initial-condition continuation is single-valued or admits bifurcations; whether the assembled massive orbit databases contain a generalized Kepler-like empirical law for arbitrary masses; whether the program extends to three-dimensional families; and whether periodic triple-star systems beyond the Broucke–Hadjidemetriou–Hénon (BHH) family exist in nature. Li and Liao (2025) have since reported 10,059 new three-dimensional periodic orbits of the general three-body problem.

Background

The Newtonian three-body problem asks for the motion of three point masses attracting each other gravitationally. Poincaré (1890) showed that no additional analytic first integral (a conserved quantity beyond energy, momentum, and angular momentum) exists, so the general motion is chaotic and cannot be written in closed form. Periodic solutions — trajectories for which the three bodies exactly return to their initial positions and velocities after a finite period TT — are therefore the main analytical handle on the problem; Poincaré called them the only opening through which to penetrate it, and modern dynamics treats them as the skeleton organizing the chaotic phase space.

Only three families were known in the first three centuries: the collinear and equilateral solutions of Euler and Lagrange; the BHH family (Broucke, Hadjidemetriou, and Hénon, 1970s), a computer-discovered family of planar orbits with zero angular momentum in which the three bodies pass through a collinear configuration; and the figure-eight choreography (Moore 1993; proven to exist by Chenciner and Montgomery 2000), in which three equal masses chase each other along a single figure-eight curve; its discovery used braid-theoretic methods, which classify a planar periodic orbit by the way the bodies' trajectories intertwine in space-time. Šuvakov and Dmitrašinović (2013) then used the shape-space sphere (a two-dimensional reduced configuration space for equal masses with zero angular momentum) to find 13 new equal-mass families, and Li and Liao (2017), using clean numerical simulation, reported more than six hundred further families. Clean numerical simulation (CNS) is a validation strategy in which the equations of motion are integrated in arbitrary-precision arithmetic at successively higher precision until the trajectory is verified to be independent of numerical noise — essential in a chaotic system where round-off error otherwise grows exponentially. Subsequent work extended the catalogs to unequal masses (Li, Jing and Liao 2018) and to linearly stable non-hierarchical unequal-mass orbits (Li, Li and Liao 2021); all these solutions are compiled in a public database (github.com/sjtu-liao/three-body).

The source paper (Liao, Li and Yang 2022) added a machine-learning layer: an artificial neural network (ANN), a trainable nonlinear regression model, learns the mapping from the two free masses (m1,m2)(m_1,m_2) (with m3=1m_3=1 fixed by scaling) to the initial conditions (x1,v1,v2)(x_1,v_1,v_2) and period TT of a periodic orbit of a given family, and predicts seeds that are then corrected by a Newton–Raphson scheme under CNS. Iterating prediction and correction grew the mass domain of the BHH-satellite family from a small initial patch to the domain S∗=S1∪S2∪S3∪S4S^* = S_1\cup S_2\cup S_3\cup S_4; at the last expansion step only about 7.1% of candidate masses on a grid of spacing Δm=0.01\Delta m = 0.01 yielded periodic orbits, from which the authors concluded that they had reached the boundary beyond which no periodic orbits of that family exist — the 'nearly largest' mass domain. Applying the same pipeline to relatively periodic orbits (orbits that close up to a rotation of the configuration rather than returning exactly) yielded 35,895 solutions in a mass domain about 1000 times larger than the initial one. Linear stability (whether infinitesimal perturbations of the orbit grow or stay bounded, decided by the Floquet multipliers of the variational equations) was predicted by a separate ANN classifier.

Two further threads frame the remaining questions. First, empirical regularities: for equal masses, the period of a periodic orbit at fixed energy was found numerically to obey a Kepler-like third law along each topological family (Janković and Dmitrašinović 2016 for the BHH family; Sun 2018 conjectured such a law for n-body periodic orbits), but no analogous law is known for arbitrary unequal masses. Second, observation: all periodic triple-star systems observed so far belong to the BHH family (near-resonant hierarchical configurations), so the catalogs' thousands of other families are unconfirmed in nature.

Problem Statement

The source paper converted the discovery of planar periodic orbits of the general three-body problem into a mass-space mapping problem: continue each known family to arbitrary masses (m1,m2)(m_1,m_2) and delineate the mass domain in which it exists. The open problem is to convert that numerical, heuristic map into definitive knowledge of the periodic-orbit structure of the three-body problem in mass space, and to determine its physical realization. Concretely, for the planar problem: (a) decide, for each family for which a near-maximal mass domain S∗S^* has been empirically identified (in the source paper, the BHH-satellite family), whether S∗S^* is the true existence domain — i.e., that no periodic orbit of that family exists for masses beyond it — or whether the boundary is an artifact of search resolution, candidate-grid spacing, and ANN extrapolation capability; (b) decide whether the mapping from masses to the initial conditions and period of a periodic orbit of a given family is single-valued, as all computed examples so far suggest, or whether bifurcations occur — multiple distinct periodic solutions of the same family at the same masses — and if so, locate and characterize them; (c) determine whether the assembled massive databases of periodic orbits (tens of thousands of high-precision solutions across families and masses) contain a generalized Kepler-like empirical law — a quantitative relation among period, energy, angular momentum, and the masses valid for arbitrary unequal masses — and state it or establish that none exists. For the three-dimensional problem, whose existence half is now settled (10,059 new three-dimensional periodic orbits were reported in 2025), two further sub-questions are: (d1) map the mass domains of three-dimensional families for arbitrary unequal masses, and (d2) establish a topological classification of three-dimensional periodic orbits, for which no braid-group or shape-sphere analogue currently exists. (e) Determine observationally whether periodic triple-star systems in nature belong to families other than the BHH family. All sub-questions serve the single objective of establishing, with certainty rather than numerical suggestion, the structure of the set of periodic solutions of the three-body problem as a function of the masses and its realization in physical stellar systems. The propositions (a), (b), (c), and (e) are independently standing sub-claims, each refutable on its own: a definitive resolution of any one of them — a proof, a counterexample, or a certified numerical or observational determination — resolves the task, and resolving several settles more of the objective. (d1) and (d2) are jointly posed sub-questions rather than independently standing ones: each covers arbitrary unequal masses over every three-dimensional family identified, so that the resulting mass-domain map and topological classification are established as complete.

References

  1. S. Liao, X. Li, Y. Yang, Three-body problem — from Newton to supercomputer plus machine learning, New Astronomy 96 (2022) 101850. DOI: 10.1016/j.newast.2022.101850; arXiv:2106.11010. https://doi.org/10.1016/j.newast.2022.101850
  2. X. Li, S. Liao, Discovery of 10,059 new three-dimensional periodic orbits of general three-body problem, Science China Physics, Mechanics & Astronomy, advance online publication 29 June 2026 (journal DOI 10.1007/s11433-026-3056-8); preprint arXiv:2508.08568 (2025). https://arxiv.org/abs/2508.08568
  3. X. Li, S. Liao, A review on periodic orbits of the general planar three-body problem, Science China Physics, Mechanics & Astronomy 68 (8) (2025) 289501. DOI: 10.1007/s11433-024-2686-6. https://doi.org/10.1007/s11433-024-2686-6
  4. V. Vasiljević, B. Raonić, V. Dmitrašinović, An island of linearly stable non-hierarchical unequal mass periodic three-body orbits — A harbinger of circumbinary exoplanets?, New Astronomy 100 (2023) 101969. DOI: 10.1016/j.newast.2022.101969. https://doi.org/10.1016/j.newast.2022.101969
  5. M. R. Janković, V. Dmitrašinović, Angular momentum and topological dependence of Kepler's third law in the Broucke-Hadjidemetriou-Hénon family of periodic three-body orbits, Physical Review Letters 116 (2016) 064301. DOI: 10.1103/PhysRevLett.116.064301. https://arxiv.org/abs/1604.08358
  6. B. H. Sun, Kepler's third law of n-body periodic orbits in a Newtonian gravitation field, Science China Physics, Mechanics & Astronomy 61 (2018) 054721. DOI: 10.1007/s11433-017-9154-0. https://doi.org/10.1007/s11433-017-9154-0
  7. M. Šuvakov, V. Dmitrašinović, Three classes of Newtonian three-body planar periodic orbits, Physical Review Letters 110 (2013) 114301. DOI: 10.1103/PhysRevLett.110.114301. https://doi.org/10.1103/PhysRevLett.110.114301
  8. C. Moore, Braids in classical dynamics, Physical Review Letters 70 (1993) 3675–3679. DOI: 10.1103/PhysRevLett.70.3675. https://doi.org/10.1103/PhysRevLett.70.3675
  9. A. Chenciner, R. Montgomery, A remarkable periodic solution of the three-body problem in the case of equal masses, Annals of Mathematics 152 (2000) 881–901. DOI: 10.2307/2661357. https://doi.org/10.2307/2661357
  10. X. Li, S. Liao, More than six hundred new families of Newtonian periodic planar collisionless three-body orbits, Science China Physics, Mechanics & Astronomy 60 (2017) 129511. DOI: 10.1007/s11433-017-9078-5. https://doi.org/10.1007/s11433-017-9078-5
  11. X. Li, Y. Jing, S. Liao, Over a thousand new periodic orbits of a planar three-body system with unequal masses, Publications of the Astronomical Society of Japan 70 (2018) 64. DOI: 10.1093/pasj/psy057. https://doi.org/10.1093/pasj/psy057
  12. X. M. Li, X. C. Li, S. J. Liao, One family of 13315 stable periodic orbits of non-hierarchical unequal-mass triple systems, Science China Physics, Mechanics & Astronomy 64 (2021) 219511. DOI: 10.1007/s11433-020-1624-7. https://doi.org/10.1007/s11433-020-1624-7
  13. I. Hristov, R. Hristova, V. Dmitrašinović, K. Tanikawa, Three-body periodic collisionless equal-mass free-fall orbits revisited, Celestial Mechanics and Dynamical Astronomy 136 (2024) 7. DOI: 10.1007/s10569-023-10177-w. https://doi.org/10.1007/s10569-023-10177-w
  14. I. Hristov, R. Hristova, T. Puzynina, Z. Sharipov, Z. Tukhliev, Numerical search for three-body periodic free-fall orbits with central symmetry (2025), arXiv:2503.00432. https://arxiv.org/abs/2503.00432
  15. X. Li, Y. Tao, X. Li, S. Liao, A numerical scheme to obtain periodic three-body orbits with finite angular momentum, New Astronomy 119 (2025) 102407. DOI: 10.1016/j.newast.2025.102407. https://doi.org/10.1016/j.newast.2025.102407
  16. N. Kollias, N. Matzakos, Physics-informed neural networks for discovering periodic orbits in the gravitational three-body problem (2026), arXiv:2607.23501. https://arxiv.org/abs/2607.23501