Classical Mechanics
Bertrand's Theorem: Why Only 1/r and r² Forces Give Closed Orbits
Out of the infinite family of central forces you could write down — anything of the form F ∝ rⁿ, plus every logarithm, exponential, and Yukawa screening term in between — exactly two make every bound orbit trace and re-trace the same closed curve forever. Joseph Bertrand proved this in a single dense page in 1873: only the inverse-square gravitational/Coulomb force (F = −k/r²) and the linear Hooke's-law spring force (F = −k·r) guarantee that a particle returns precisely to its starting point after one radial cycle, with no precession.
Bertrand's theorem states that among all central-force potentials admitting stable circular orbits, only V(r) = −k/r and V(r) = ½k·r² produce orbits that close for all bound initial conditions. Every other attractive potential produces rosette-shaped orbits that slowly precess and, in general, never exactly close.
- TypeTheorem in classical mechanics (central-force dynamics)
- Proved byJoseph Louis François Bertrand, 1873
- The two forcesF = −k/r² (inverse-square) and F = −k·r (Hooke)
- Key conditionApsidal angle Φ constant and a rational multiple of π
- Apsidal anglesΦ = π (Kepler) and Φ = π/2 (oscillator)
- Near-circular formulaΦ = π / √(3 + n) for F ∝ rⁿ
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
The physical setup: closed orbits and the apsidal angle
A particle of mass m moving in a central potential V(r) conserves both energy E and angular momentum L. Its radial motion decouples into a one-dimensional problem governed by the effective potential V_eff(r) = V(r) + L²/(2mr²), where the second term is the centrifugal barrier. Bound motion oscillates between a minimum radius r_min (pericenter) and a maximum r_max (apocenter) — the two apsides.
As r cycles once from r_min to r_max and back, the particle also sweeps through some azimuthal angle. The apsidal angle Φ is the angle swept between successive pericenter and apocenter — exactly half a full radial period. An orbit closes only if, after some whole number of radial oscillations, the total azimuthal sweep is a whole number of full turns. Formally, the orbit closes if and only if Φ = π·(p/q) with p and q integers — a rational multiple of π. If Φ is irrational relative to π, the orbit densely fills an annulus and never repeats.
The derivation: perturbing a circular orbit
Bertrand's 1873 argument starts by demanding that every bound orbit close, which is far stronger than closing one orbit. First consider a nearly circular orbit at radius r₀. Small radial perturbations oscillate at frequency ω_r while the particle revolves at ω_θ. Expanding V_eff to second order gives the ratio
- Φ = π·(ω_θ/ω_r) = π / √(3 + r₀·V''(r₀)/V'(r₀))
For a power law F = −k·rⁿ, this collapses to the clean result Φ = π/√(3 + n). For the orbit to close, Φ must be constant across all radii r₀ — so 3 + n must be constant, forcing a pure power law, and √(3 + n) must be rational. Bertrand then pushes to second order in the perturbation amplitude to handle non-circular orbits. That higher-order term vanishes only for two exponents: n = −2 (giving Φ = π/√1 = π) and n = +1 (giving Φ = π/√4 = π/2). All other rational choices fail once the orbit is appreciably eccentric.
Key quantities and a worked example
The two survivors have exact, radius-independent apsidal angles:
- Kepler (n = −2): Φ = π = 180°. Pericenter and apocenter sit exactly opposite each other — one radial cycle equals one full revolution, so the ellipse closes after a single loop.
- Oscillator (n = +1): Φ = π/2 = 90°. The particle hits pericenter and apocenter twice each per revolution; the ellipse is centered on the force center and closes after one revolution containing two radial cycles.
Worked check: take a generic attractive power law with n = 0 (a constant-magnitude force, linear potential V ∝ r). Then Φ = π/√3 ≈ 1.8138 rad ≈ 103.92°. Since 103.92°/180° = 0.5774 is irrational, successive pericenters advance by 2Φ − 360° ≈ −152.2° each cycle — a steadily precessing rosette that never repeats. Contrast Mercury's tiny general-relativistic precession of 43 arc-seconds per century (3.3×10⁻⁵ of a turn): that is the correction that turns the ideal closed Kepler ellipse into a slowly open rosette, a direct violation of the pure-Newtonian Bertrand condition.
How it's observed, tested, and applied
Bertrand's theorem is not merely formal — it explains a striking observational fact: planetary orbits in the Solar System are, to excellent approximation, fixed closed ellipses, because gravity is inverse-square. Any deviation from 1/r² immediately shows up as apsidal precession, making orbital tracking a sensitive probe of the force law.
- Mercury's perihelion: the observed 43″/century anomaly (after subtracting planetary tugs) is the classic test — it signals that the true relativistic potential is not pure 1/r, so the orbit does not perfectly close.
- Binary pulsars: systems like PSR B1913+16 show periastron advance of ~4.2° per year, a colossal Bertrand-violation used to confirm general relativity.
- Atomic and lab analogues: the isotropic 3D harmonic oscillator (F = −k·r) models trapped ions and small-amplitude molecular vibrations, whose closed elliptical orbits underlie clean normal-mode spectra.
- Accelerator and plasma design: knowing that only two potentials give resonance-free closed motion guides the choice of confining fields.
Comparison with related regimes and cousins
Bertrand's theorem sits alongside several nearby results that are easy to conflate:
- Kepler's laws describe the shape of one orbit under 1/r²; Bertrand asks the deeper question of which force laws make all orbits close.
- The Laplace–Runge–Lenz vector is the hidden conserved quantity behind the Kepler ellipse's non-precession; a similar extra symmetry (an SU(3) / Fradkin tensor) protects the oscillator. Bertrand's two potentials are exactly the two with such dynamical symmetry beyond angular momentum.
- Newton's theorem of revolving orbits shows adding an inverse-cube term multiplies the apsidal angle by a constant — precession you can dial in.
- Isochrone potentials (Hénon) have a period depending only on energy, not L; the two Bertrand potentials are the isochrones that also close.
Crucially, the inverse-cube force (n = −3) gives Φ → undefined and unstable circular orbits — the boundary case where centrifugal and force terms scale identically.
Significance, famous cases, and open questions
Bertrand's theorem is celebrated because it turns a vague aesthetic observation — 'why are the planets' orbits such tidy ellipses?' — into a rigorous uniqueness statement: nature had essentially no choice. Given that gravity is inverse-square, the closed ellipse is mandatory, not a coincidence, and any measured precession is real new physics.
- Historical note: Bertrand published the result in Comptes Rendus in 1873; his original proof was terse, and mathematically complete versions (via second-order perturbation, and later fully non-perturbative and global topological proofs in the 2000s) followed for over a century.
- Modern reach: the theorem has been extended to curved spaces (spheres and hyperbolic planes), where the analogues of the Kepler and oscillator potentials remain the only closed-orbit forces — a result tied to the Higgs oscillator and superintegrability.
- Open threads: classifying all superintegrable systems, and understanding closed-orbit conditions in relativistic and quantum settings (where 'closed' becomes 'degenerate energy levels'), remains active. The oscillator's and Kepler's exceptional degeneracies are precisely why the hydrogen atom and the 3D harmonic oscillator are the two exactly solvable central-force quantum problems.
| Force law | Exponent n (F ∝ rⁿ) | Apsidal angle Φ (near-circular) | Orbit shape | Closed? |
|---|---|---|---|---|
| Inverse-square (Kepler/Coulomb) | n = −2 | π (= 180°) | Ellipse, focus at center | Yes — all bound orbits |
| Hooke's law (isotropic oscillator) | n = +1 | π/2 (= 90°) | Ellipse, center at center | Yes — all bound orbits |
| Inverse-cube attraction | n = −3 | undefined (no stable circular orbit) | Spiral / unstable | No |
| Constant-magnitude force (linear potential) | n = 0 | π/√3 ≈ 103.9° | Precessing rosette | No (irrational multiple of π) |
| Yukawa / screened Coulomb | not a pure power law | varies with radius | Precessing rosette | No |
Frequently asked questions
What does Bertrand's theorem actually say?
It says that among all central-force potentials that admit stable circular orbits, only two produce bound orbits that close for every initial condition: the inverse-square force F = −k/r² (Newtonian gravity and the Coulomb force) and the linear Hooke's-law force F = −k·r (the isotropic harmonic oscillator). Every other central force produces orbits that generally precess and never exactly retrace themselves.
Why do only 1/r² and r forces give closed orbits?
An orbit closes only if its apsidal angle Φ (the azimuthal angle swept between pericenter and apocenter) is a rational multiple of π and is the same at every radius. For a power law F ∝ rⁿ, the near-circular apsidal angle is Φ = π/√(3+n). Requiring this to stay constant and rational even for eccentric orbits (a second-order condition) singles out exactly n = −2 (Φ = π) and n = +1 (Φ = π/2).
What is the apsidal angle?
The apsidal angle Φ is the angle a particle sweeps around the force center as it moves from its closest approach (pericenter) to its farthest point (apocenter). It equals half of one radial oscillation. For the Kepler problem Φ = 180°, so pericenter and apocenter are exactly opposite; for the harmonic oscillator Φ = 90°.
How does Bertrand's theorem relate to Kepler's laws?
Kepler's first law states that a single planet's orbit under 1/r² gravity is a fixed ellipse. Bertrand's theorem is the deeper 'why': it proves that 1/r² is one of only two force laws for which this closure holds universally. If gravity were even slightly different from inverse-square, planetary ellipses would precess and Kepler's neat picture would break down.
Does Mercury's orbit violate Bertrand's theorem?
In a sense, yes. Mercury's perihelion advances by 43 arc-seconds per century beyond what Newtonian planetary perturbations explain. That precession means Mercury's orbit does not perfectly close, which tells us the true gravitational potential near the Sun is not exactly 1/r — it is the general-relativistic correction, later explained by Einstein in 1915.
What happens for other force laws, like inverse-cube?
They fail to give universally closed orbits. The inverse-cube force (n = −3) is a critical case: its centrifugal and attractive terms scale identically, so no stable circular orbit exists and trajectories spiral. Intermediate laws like a constant-magnitude force (linear potential, n = 0) give Φ = π/√3 ≈ 103.9°, an irrational multiple of π, producing rosette orbits that precess forever without ever closing.