Restricted Few-Body Dynamics, Resonances, and Stability

The circular restricted three-body problem is the simplest orbital model in which an exact integral coexists with resonances, unstable manifolds, and chaos. Two primaries move on circular Newtonian orbits; a third body has negligible mass and therefore feels the primaries without affecting them. Linear stability near an equilibrium answers only a local, infinitesimal question within this restricted model.

Required background. Classical and Variational Mechanics provides action principles, canonical variables, and mechanical conservation laws; Newtonian Gravity provides the gravitational potential, Poisson equation, and inverse-square force law. Helpful background. Calculus and Linear Algebra provides multivariable differentiation, linear maps, inner products, and basis changes; Ordinary Differential Equations provides initial-value problems, linearization, and asymptotic solution methods.

Equations and the Jacobi integral

Let the primary masses be $m_1$ and $m_2$, separated by $a$, with angular frequency

$$n^2=\frac{G(m_1+m_2)}{a^3}.$$

Use a frame centered on the binary barycenter and rotating with angular velocity $n\widehat{\mathbf z}$, so the primaries are fixed on the $x$-axis at $x_1=-am_2/(m_1+m_2)$ and $x_2=am_1/(m_1+m_2)$. The rotating frame is a computational choice, not a new force law.

In the orbital plane, define distances $r_1,r_2$ from the test body to the primaries and

$$\Omega(x,y) =\frac{Gm_1}{r_1}+\frac{Gm_2}{r_2} +\frac{n^2}{2}(x^2+y^2).$$

The equations of motion are

$$\ddot x-2n\dot y=\frac{\partial\Omega}{\partial x}, \qquad \ddot y+2n\dot x=\frac{\partial\Omega}{\partial y}.$$

The velocity-dependent Coriolis terms do no work. Multiplying by $\dot x,\dot y$, adding, and integrating gives Jacobi’s integral,

$$C=2\Omega-(\dot x^2+\dot y^2).$$

It is not the inertial-frame mechanical energy, because the rotating coordinates depend explicitly on time. Nevertheless, it restricts configuration space: motion is possible only where $2\Omega\ge C$. Boundaries $2\Omega=C$ are zero-velocity curves; they constrain transit but do not specify the trajectory inside an allowed region. Battin 2008, Lecture 25, “The Restricted Problem of Three Bodies” derives the same equations and integral before constructing these regions.

Lagrange equilibria and linear stability

Equilibria satisfy $\nabla\Omega=0$ with zero rotating-frame velocity. There are five: three collinear points $L_1,L_2,L_3$ and two equilateral points $L_4,L_5$. Linearize with $\mathbf q=\mathbf q_L+\boldsymbol\xi$ and seek $\boldsymbol\xi\propto e^{\lambda t}$. The eigenvalues of the resulting gyroscopic system determine spectral growth or oscillation.

To evaluate the triangular points, measure length in units of $a$ and time in units of $n^{-1}$, and set

$$\mu=\frac{m_2}{m_1+m_2}\le\frac12,$$

so the primaries have mass fractions $1-\mu$ and $\mu$. At either $L_4$ or $L_5$, the dimensionless effective-potential Hessian has

$$\Omega_{xx}=\frac34, \qquad \Omega_{yy}=\frac94, \qquad \Omega_{xy}=\pm\frac{3\sqrt3}{4}(1-2\mu).$$

Substitution of $e^{\lambda t}$ into the two linearized equations gives

$$\lambda^4 +\left(4-\Omega_{xx}-\Omega_{yy}\right)\lambda^2 +\Omega_{xx}\Omega_{yy}-\Omega_{xy}^2=0,$$

and therefore

$$\lambda^4+\lambda^2+\frac{27}{4}\mu(1-\mu)=0.$$

Writing $z=\lambda^2$, its roots are

$$z_\pm =\frac{-1\pm\sqrt{1-27\mu(1-\mu)}}{2}.$$

For $0\lt\mu\le\tfrac12$, the roots $z_\pm$ are distinct, real, and negative precisely on the nondegenerate Routh-stable side

$$27\mu(1-\mu)\lt1, \qquad \mu\lt\mu_{\rm R} =\frac12\left(1-\sqrt{\frac{23}{27}}\right) \simeq0.03852.$$

At equality, $z_+=z_-=-1/2$: the frequencies coincide. This is not part of the bounded linear-stability region. For each repeated eigenvalue $\lambda=\pm i/\sqrt2$, the two displacement equations have a one-dimensional eigenspace, so the first-order system has a Jordan block and admits terms proportional to $t e^{\lambda t}$. Purely imaginary eigenvalues alone do not exclude this secular growth.

The collinear points instead have an exponentially growing linear mode. The general characteristic equation, its collinear instability, and the triangular-point reduction through the Routh condition are derived in Richard Fitzpatrick, Celestial Mechanics, Chapter 9, Eqs. (9.73)–(9.89).

This result is conditional. It assumes circular primaries, a massless third body, planar infinitesimal perturbations for the stated calculation, and no nongravitational forces. A realistic mission design, such as the proposed Lynx observatory, instead uses an unstable Sun–Earth $L_2$ halo orbit with regular station-keeping, as described in Hopkins et al. 2020, §4.5, p. 13, PDF; a spacecraft is not simply parked at the collinear equilibrium point.

Resonance and the limits of local stability

A mean-motion resonance occurs when integer combinations of orbital frequencies nearly vanish,

$$k_1n_1+k_2n_2+\cdots\simeq0, \qquad k_A\in\mathbb Z.$$

The associated resonant angle evolves slowly, so terms that would average away can accumulate. A resonance may protect trajectories through phase locking or enlarge excursions. When calculated widths of neighboring nonlinear resonances overlap in a specified perturbation model, the overlap criterion can mark a region in which stochastic behavior and chaotic transport are permitted; it does not make every trajectory there chaotic. Wisdom reviews this criterion and applies and tests it for the small-mass-ratio planar circular restricted problem in Wisdom 1980, §§II–IV, pp. 1122–1130. Proximity to a commensurability alone, without the width calculation, does not establish overlap or chaotic transport.

An exact Jacobi constant constrains a trajectory to one invariant surface but does not by itself prove regular motion. Establishing chaos for specified parameters and initial data requires an additional argument or converged diagnostic, such as a return map, long-time Lyapunov analysis, or resolved stable-and-unstable-manifold intersections. A positive local Lyapunov exponent over a finite interval does not alone establish escape or global instability.

What linear stability does not prove

• Purely imaginary linear eigenvalues exclude exponential growth at first order; they do not automatically prove nonlinear Lyapunov stability. Resonances and higher-order terms can decide the result.
• Linear instability does not mean immediate collision or escape. Stable and unstable manifolds organize long-lived transit orbits near a saddle.
• The Jacobi constant forbids some passages but says nothing about which allowed passage is taken.
• “Stable Lagrange point” is shorthand tied to a model and perturbation class. Finite primary eccentricity, inclination, radiation pressure, other bodies, and station-keeping alter the dynamics.
• Numerical boundedness over a finite integration is evidence on that time interval, not a theorem of eternal stability. Convergence and shadowing checks are essential near chaotic regions.

The Jacobi integral and zero-relative-velocity boundaries used to delimit the accessible region are derived in Battin 2008, Lecture 25, “The Restricted Problem of Three Bodies”.

Transfer task

At fixed Jacobi constant $C$, what happens to an allowed region as $C$ is increased?

Solution

The kinetic-energy identity is

$$\dot x^2+\dot y^2=2\Omega-C\ge0.$$

Increasing $C$ makes the inequality harder to satisfy, so the allowed configuration-space region shrinks. As $C$ is increased through a critical value $2\Omega$ at a Lagrange point, the corresponding zero-velocity neck closes; as $C$ is decreased through that value, the neck opens. This topological change permits or forbids transit, but does not determine whether a particular initial condition uses the neck.

The next chapter, Special Relativity and Minkowski Spacetime, replaces Newtonian absolute time with invariant causal geometry. Orbital stability in a relativistic spacetime must then be rebuilt from its own conserved quantities and perturbation equations.

References

• Richard H. Battin, MIT 16.346 Lecture 25, “The Restricted Problem of Three Bodies”, Jacobi equations, Jacobi integral, and zero-relative-velocity surfaces, Fall 2008.
• Richard Fitzpatrick, Celestial Mechanics, Chapter 9, “Stability of Lagrange Points,” Eqs. (9.73)–(9.89), University of Texas at Austin.
• Hopkins, R. C., A. R. Schnell, S. G. Sutherlin, R. J. Suggs, T. M. Boswell, A. Dominguez, P. D. Capizzo, M. Baysinger, J. C. Garcia, and L. L. Fabisinski. Orbit Selection for the Proposed Lynx Observatory Mission. NASA/TM—2020–220554, January 2020. §4.5, p. 13, PDF.
• Jack Wisdom, “The Resonance Overlap Criterion and the Onset of Stochastic Behavior in the Restricted Three-Body Problem,” The Astronomical Journal 85 (1980): 1122–1133. doi:10.1086/112778.

Further reading

• MIT OpenCourseWare 16.07, Lecture L18, “Exploring the Neighborhood: the Restricted Three-Body Problem”, stability and applications of the Lagrange points, Fall 2009.