Seifert's conjecture on the multiplicity of brake orbits in potential wells

The conjecture formulated by H. Seifert in 1948 states that a natural Hamiltonian system $H(q,p)=\frac{1}{2}a^{ij}(q)p_{i}p_{j}+V(q)$ at a regular energy level $E$ whose potential well $\{q:V(q)\le E\}$ is homeomorphic to the closed $N$-dimensional disk possesses at least $N$ geometrically distinct brake orbits — periodic solutions whose momentum vanishes twice per period. The bound is sharp (anisotropic harmonic oscillators with irrational frequency ratios have exactly $N$), a claimed set of real-analytic counterexamples was retracted in 2012, and the peer-reviewed literature establishes the conjecture in dimension $2$ (Giambò–Giannoni–Piccione, 2015) and in the even-and-convex case for all $N$ (Liu–Zhang, 2014). A claimed complete proof by Giambò–Giannoni–Piccione has circulated as an unpublished preprint since 2020 and is described by the most recent independent literature (2026) as announced rather than established. The surviving open core is the unconditional case $N\ge3$: prove or disprove that every such system has at least $N$ geometrically distinct brake orbits.

Background

A natural Hamiltonian system is a mechanical system whose Hamiltonian function splits into kinetic and potential energy,

$$H(q,p)=\frac{1}{2}\sum_{i,j=1}^{N}a^{ij}(q)\,p_{i}p_{j}+V(q),$$

where $(q,p)$ are canonical position–momentum coordinates on phase space $\mathbb{R}^{2N}$, the coefficient matrix $(a^{ij}(q))$ is symmetric, positive definite and of class $C^{2}$, and $V$ is the potential energy. A brake orbit is a nonconstant periodic solution of Hamilton's equations whose momentum vanishes at some instant — after a shift of the time origin, $p(0)=p(T)=0$ for some $T>0$, so that the momentum vanishes twice per period: since the kinetic energy is even in $p$, the motion after time $T$ retraces the same arc in configuration space, like a ball thrown upward in a gravitational well that rises, stops, and falls back along its path. Fix an energy $E>\inf V$ that is a regular value of $V$ (meaning $\mathrm{d}V\neq0$ everywhere on the level set $V^{-1}(E)$) and assume that the closed sublevel $\Omega_{E}=\{q:V(q)\le E\}$, called the potential well, is homeomorphic to the closed $N$-dimensional disk. Seifert proved in 1948, under analyticity hypotheses, that such a system has at least one brake orbit, and in a footnote he suggested that $N$ geometrically distinct brake orbits — orbits with distinct images in configuration space — ought to exist. This remark is now called Seifert's conjecture on brake orbits.

The conjectured bound is optimal: the anisotropic harmonic oscillator $H(q,p)=\frac{1}{2}|p|^{2}+\frac{1}{2}\sum_{i}\lambda_{i}^{2}q_{i}^{2}$ with irrational squared ratios $(\lambda_{i}/\lambda_{j})^{2}$ has, at every positive energy, exactly $N$ brake orbits (those along the principal axes of the ellipsoid $\{\sum_{i}\lambda_{i}^{2}q_{i}^{2}=E\}$) and no other periodic solutions, so no stronger lower bound can hold in general.

By the Maupertuis variational principle, trajectories of fixed energy $E$ are, up to reparameterization, geodesics of the Jacobi metric $(E-V)\,g$, a metric conformal to the configuration metric that degenerates at the boundary of the well; brake orbits correspond to orthogonal geodesic chords (OGCs), that is, geodesic segments in the closed well that meet the boundary orthogonally at both endpoints — a correspondence established rigorously by Giambò–Giannoni–Piccione in 2005. Existence of one brake orbit is classical beyond the analytic category, through work of Weinstein on convex systems and of Gluck–Ziller. The multiplicity statement, however, has been confirmed only under additional structure. Bolotin–Kozlov and Gluck–Ziller obtained at least $N$ brake orbits under strong ad hoc assumptions on the energy integral that force distinct minimax critical levels; Szulkin proved at least $N$ for star-shaped energy hypersurfaces satisfying an extra geometric condition; Kozlov's 1985 survey records this early state of the problem. For convex symmetric systems, Long–Zhang–Zhu proved at least two brake orbits in bounded convex symmetric domains and Zhang–Liu obtained at least $[(n+1)/2]+1$ brake orbits on compact convex symmetric reversible hypersurfaces; the full conjecture in the even-and-convex case for every $N$ was proved by Liu–Zhang (2014). Giambò–Giannoni–Piccione (2015) proved the conjecture in dimension $N=2$ without any assumption on the potential, and (2018) established the lower bound for Riemannian $N$-disks without orthogonal–tangent chords (geodesic chords meeting the boundary orthogonally at one endpoint and tangentially at the other), a class that includes small perturbations of rotationally symmetric (radial) systems. The difficulty in a general nonconvex, nonsymmetric well is that the standard variational machinery breaks down: once a curve being shortened becomes tangent to the boundary, arbitrarily small perturbations leave the domain, so the Palais–Smale compactness condition (the standard property that sequences of approximate critical points of the action functional admit convergent subsequences, on which minimax deformation arguments rest) fails, and one must instead run a nonsmooth critical point theory for 'geodesics with obstacle' in the degenerate Jacobi metric. Two further episodes frame the current status: in 2012 Giambò–Giannoni–Piccione announced real-analytic counterexamples (disk-shaped wells carrying a single brake orbit) but withdrew the preprint, stating that its crucial technical lemma was incorrect; and in 2020–2022 the same authors posted a claimed proof of the full conjecture (at least $N$ orthogonal geodesic chords in strongly concave Riemannian $N$-disks — disks whose boundary has positive-definite second fundamental form in the exterior normal direction — plus a reduction step for the degenerate Jacobi case), which as of 2026 remains unpublished. Rademacher (2026) describes the unconditional result as 'announced', and Liu's 2019 survey records the conjecture as open for $N>2$ under the original hypotheses.

Problem Statement

Determine the truth of Seifert's conjecture on brake orbits in its original generality. Let $N\ge1$ and let $H(q,p)=\frac{1}{2}\sum_{i,j=1}^{N}a^{ij}(q)\,p_{i}p_{j}+V(q)$ be a natural Hamiltonian of class $C^{2}$ (equivalently, a natural Lagrangian system on an $N$-dimensional configuration manifold), let $E>\inf V$ be a regular value of $V$, and suppose the closed potential well $\{q:V(q)\le E\}$ is homeomorphic to the closed $N$-disk $\mathbb{D}^{N}$. Must the system admit at least $N$ geometrically distinct brake orbits of energy $E$ — nonconstant periodic solutions of energy $E$ whose momentum vanishes (two brake orbits being distinct when their images in configuration space differ)? Prove this, or refute it by exhibiting an admissible $C^{2}$ (or analytic) potential, with a regular energy level whose well is a topological $N$-disk, that carries fewer than $N$ geometrically distinct brake orbits. The bound cannot be improved if the conjecture holds, since anisotropic harmonic oscillators with irrational squared frequency ratios have exactly $N$ brake orbits and no other periodic solutions. In the peer-reviewed literature the case $N=2$ is settled affirmatively (Giambò–Giannoni–Piccione 2015), as are the even-and-convex, symmetric-convex, and near-radial classes; the open core is the unconditional case $N\ge3$, with no symmetry, convexity, analyticity, or radial-perturbation hypothesis.

References

1. H. Seifert, 'Periodische Bewegungen mechanischer Systeme', Mathematische Zeitschrift 51 (1948) 197–216. DOI: 10.1007/BF01291002. https://doi.org/10.1007/BF01291002
2. A. Weinstein, 'Periodic orbits for convex Hamiltonian systems', Annals of Mathematics (2) 108 (1978) 507–518. DOI: 10.2307/1971185. https://doi.org/10.2307/1971185
3. S. V. Bolotin and V. V. Kozlov, 'Libration in systems with many degrees of freedom', Journal of Applied Mathematics and Mechanics 42 (1978) 256–261. DOI: 10.1016/0021-8928(78)90141-7. https://doi.org/10.1016/0021-8928(78)90141-7
4. H. Gluck and W. Ziller, 'Existence of periodic motions of conservative systems', in E. Bombieri (ed.), Seminar on Minimal Submanifolds (Annals of Mathematics Studies 103), Princeton University Press (1983), pp. 65–98. DOI: 10.1515/9781400881437-004. https://doi.org/10.1515/9781400881437-004
5. V. V. Kozlov, 'Calculus of variations in the large and classical mechanics', Russian Mathematical Surveys 40 (1985) 37–71. DOI: 10.1070/RM1985v040n02ABEH003557. https://doi.org/10.1070/RM1985v040n02ABEH003557
6. A. Szulkin, 'An index theory and existence of multiple brake orbits for star-shaped Hamiltonian systems', Mathematische Annalen 283 (1989) 241–255. DOI: 10.1007/BF01446433. https://doi.org/10.1007/BF01446433
7. R. Giambò, F. Giannoni and P. Piccione, 'Orthogonal geodesic chords, brake orbits and homoclinic orbits in Riemannian manifolds', Advances in Differential Equations 10 (2005) 931–960. DOI: 10.57262/ade/1355867824. https://doi.org/10.57262/ade/1355867824
8. Y. Long, D. Zhang and C. Zhu, 'Multiple brake orbits in bounded convex symmetric domains', Advances in Mathematics 203 (2006) 568–635. DOI: 10.1016/j.aim.2005.05.005. https://doi.org/10.1016/j.aim.2005.05.005
9. R. Giambò, F. Giannoni and P. Piccione, 'Existence of orthogonal geodesic chords on Riemannian manifolds with concave boundary and homeomorphic to the N-dimensional disk', Nonlinear Analysis 73 (2010) 290–337. DOI: 10.1016/j.na.2010.03.019. https://doi.org/10.1016/j.na.2010.03.019
10. R. Giambò, F. Giannoni and P. Piccione, 'Potential wells with a unique brake orbit. Counterexamples to a conjecture by H. Seifert', arXiv:1203.5198 (2012), withdrawn 19 July 2012; stated withdrawal reason: 'crucial technical result (Lemma 2.2) is incorrect'. https://arxiv.org/abs/1203.5198
11. C. Liu and D. Zhang, 'Seifert conjecture in the even convex case', Communications on Pure and Applied Mathematics 67 (2014) 1563–1604. DOI: 10.1002/cpa.21525. https://doi.org/10.1002/cpa.21525
12. D. Zhang and C. Liu, 'Multiple brake orbits on compact convex symmetric reversible hypersurfaces in R^{2n}', Annales de l'Institut Henri Poincaré C, Analyse Non Linéaire 31 (2014) 531–554. DOI: 10.1016/j.anihpc.2013.03.010. https://doi.org/10.1016/j.anihpc.2013.03.010
13. R. Giambò, F. Giannoni and P. Piccione, 'Multiple brake orbits in m-dimensional disks', Calculus of Variations and Partial Differential Equations 54 (2015) 2553–2580. DOI: 10.1007/s00526-015-0875-5. https://doi.org/10.1007/s00526-015-0875-5
14. R. Giambò, F. Giannoni and P. Piccione, 'Multiple orthogonal geodesic chords in nonconvex Riemannian disks using obstacles', Calculus of Variations and Partial Differential Equations 57 (2018), no. 5, article 117. DOI: 10.1007/s00526-018-1394-y. https://doi.org/10.1007/s00526-018-1394-y (preprint: arXiv:1807.00887)
15. R. Giambò, F. Giannoni and P. Piccione, 'Multiple orthogonal geodesic chords and a proof of Seifert's conjecture on brake orbits', arXiv:2002.09687 (v1 2020, v3 2022; unpublished preprint). https://arxiv.org/abs/2002.09687
16. C. Liu, 'Multiplicity of Brake Orbits on a Fixed Energy Surface', chapter in Index Theory in Nonlinear Analysis, Springer, Singapore (2019), pp. 275–291. DOI: 10.1007/978-981-13-7287-2_10. https://doi.org/10.1007/978-981-13-7287-2_10
17. D. Corona, 'A multiplicity result for orthogonal geodesic chords in Finsler disks', Discrete and Continuous Dynamical Systems 41 (2021) 5329–5357. DOI: 10.3934/dcds.2021079. https://doi.org/10.3934/dcds.2021079
18. L. Asselle, X. Hu, A. Portaluri and L. Wu, 'Non-minimality and instability of brake orbits for natural Lagrangians on Riemannian manifolds', arXiv:2307.06105 (2023; v2 2026). https://arxiv.org/abs/2307.06105
19. I. Beach, 'Short Simple Orthogonal Geodesic Chords on a 2-Disk with Convex Boundary', arXiv:2507.07797 (2025). https://arxiv.org/abs/2507.07797
20. I. Beach, H. Contreras-Peruyero, E. Griffin, R. Rotman and C. Searle, 'Lengths of Orthogonal Geodesic Chords on Riemannian Manifolds', arXiv:2509.19620 (2025). https://arxiv.org/abs/2509.19620
21. H.-B. Rademacher, 'Periodic orbits of reversible Lagrangian systems without self-intersections and Mañé genericity', arXiv:2602.17833 (2026). https://arxiv.org/abs/2602.17833