Appearance
11.6 — Lagrange Points, Gravity Assists and Chaos
The two-body problem was solved completely in Chapter 11.3. Add a third body and there is no general solution, and there never will be.
This is not a failure of technique. Bruns and Poincaré proved in the 1880s and 90s that the three-body problem has no solution in terms of the standard functions of mathematics. Poincaré, investigating it for a prize competition, found something worse: the motion can be genuinely chaotic, and in the process he founded the whole field.
But special cases can be solved, and they turn out to be extremely useful.
The restricted three-body problem
Simplify by assuming:
Two massive bodies in circular orbits about their common centre of mass — the Earth and Sun, say.
A third body of negligible mass — a spacecraft, which is influenced by them and does not influence them.
Work in a rotating frame that turns with the two large bodies, so they appear stationary.
In that frame two fictitious forces appear (Chapter 1.5): centrifugal, pointing outward, and Coriolis, acting on moving objects.
The effective potential combines gravity from both bodies with the centrifugal term:
U_{\text{eff}} = -\frac{Gm_1}{r_1}-\frac{Gm_2}{r_2}-\frac{1}{2}\omega^2 d^2
where d is the distance from the rotation axis.
Equilibrium points are where \nabla U_{\text{eff}} = 0. Lagrange found in 1772 that there are exactly five.
The five Lagrange points
L1 — between the two bodies. The Sun's pull is partly cancelled by the Earth's, so an object there orbits the Sun with the same period as the Earth despite being closer.
L2 — beyond the smaller body. Here both pulls add, providing the extra centripetal force needed to orbit at a larger radius with the Earth's period.
L3 — on the far side of the larger body, opposite the smaller.
L4 and L5 — at the corners of equilateral triangles with the two bodies, 60° ahead and behind.
Deriving L1 and L2
For L1, at distance r from the Earth on the Sun side, the balance is:
\frac{GM_\odot}{(R-r)^2}-\frac{GM_E}{r^2} = \omega^2(R-r)
with \omega^2 = GM_\odot/R^3 from the Earth's own orbit.
For r \ll R, expand. The standard result is the Hill radius:
\boxed{r \approx R\left(\frac{M_E}{3M_\odot}\right)^{1/3}}
Worked example: Sun–Earth L1.
\frac{M_E}{3M_\odot} = \frac{5.972\times10^{24}}{3(1.989\times10^{30})} = \frac{5.972\times10^{24}}{5.967\times10^{30}} = 1.0008\times10^{-6}
Cube root: \log_{10}(1.0008\times10^{-6}) = -5.99965, divided by 3 is -1.99988, giving 1.0003\times10^{-2}.
r = 1.496\times10^{11}\times1.0003\times10^{-2} = 1.496\times10^{9}\ \text{m} = 1.50\ \text{million km}
About four times the Moon's distance. The accepted value for Sun–Earth L1 is 1.4811\times10^{6} km, and the small difference comes from the approximations dropped.
L2 is at the same distance on the far side, 1.5017\times10^{6} km out.
Stability
L1, L2 and L3 are unstable saddle points. Displace a spacecraft slightly and it drifts away, with a characteristic timescale of about 23 days for Sun–Earth L1.
So spacecraft there need station-keeping, typically a few m/s per year, and they do not sit exactly at the point. They fly halo orbits — large loops around the point, tens of thousands of kilometres across — which are easier to maintain and, for L2, keep the spacecraft out of the Earth's shadow so its solar panels work.
L4 and L5 are stable, and this is the surprising result.
They sit at maxima of the effective potential, which should make them maximally unstable. The Coriolis force saves them. An object drifting away from L4 acquires a velocity, the Coriolis force deflects that velocity sideways, and the object curls back into a stable loop around the point.
The condition for stability is a mass ratio:
\frac{M_1}{M_2} > 24.96
Sun–Jupiter is 1047, comfortably stable. Earth–Moon is 81, also stable.
Real objects at Lagrange points
Jupiter's Trojans. Over 12,000 known asteroids are catalogued at Jupiter's L4 and L5, and the true population above 1 km is estimated at about a million — comparable to the main asteroid belt.
They have been there for billions of years, which is the observational proof of L4/L5 stability.
The naming convention is charming and strictly observed: L4 asteroids are named after Greek heroes and L5 after Trojan heroes, with two mistakes made early on — Hektor sits in the Greek camp and Patroclus in the Trojan camp, and both are stuck there.
Neptune has about 30 known Trojans, Mars has several, and Earth has two confirmed: 2010 TK7 and 2020 XL5.
And Saturn's moons Tethys and Dione each have two co-orbital companions at their L4 and L5 points.
Spacecraft at Lagrange points
| Point | Mission | Why |
|---|---|---|
| Sun–Earth L1 | SOHO, DSCOVR, ACE | Uninterrupted view of the Sun; early warning of solar wind |
| Sun–Earth L2 | JWST, Gaia, Euclid, Planck | Sun, Earth and Moon all in one direction, so one shield blocks all three |
| Earth–Moon L2 | Queqiao relay | Line of sight to the lunar far side |
Why L2 is so valuable for astronomy is worth spelling out.
A telescope in low Earth orbit sees the Earth filling a large part of the sky, has its thermal environment swing as it passes in and out of shadow every 45 minutes, and must be pointed away from the Earth, Sun and Moon in turn.
At L2, all three are in the same direction. A single sunshield blocks all of them, and the telescope can be cooled passively.
JWST's five-layer sunshield keeps the mirrors at about 40 K — with no refrigerator, purely by radiating to deep space (Chapter 3.7). The Sun-facing side is at about 360 K, so the shield holds a 320-degree difference across five layers of aluminised Kapton, each 25 μm thick.
And L2 is 1.5 million km away, which means no servicing mission is possible. Hubble was serviced five times; JWST cannot be touched. Everything had to work first time, including 344 single-point failures during deployment.
SOHO at L1 has an uninterrupted view of the Sun and sits 1.5 million km upstream in the solar wind, giving about an hour of warning before a coronal mass ejection reaches Earth. That hour is what power grid operators use to protect transformers.
Gravity assists
A spacecraft flying past a planet can gain or lose speed relative to the Sun, without using any fuel.
The apparent paradox: in the planet's frame the encounter is a hyperbolic flyby, and the spacecraft leaves with exactly the speed it arrived with. Energy is conserved in that frame.
So where does the gain come from?
The bookkeeping, done properly
The trick is that the planet's frame is moving.
In the planet's frame: the spacecraft arrives at \vec{v}_\infty and leaves at \vec{v}_\infty', with |\vec{v}_\infty| = |\vec{v}_\infty'|. Only the direction changed.
In the Sun's frame: add the planet's velocity \vec{v}_p to both.
\vec{v}_{\text{in}} = \vec{v}_\infty+\vec{v}_p, \qquad \vec{v}_{\text{out}} = \vec{v}_\infty'+\vec{v}_p
Since \vec{v}_\infty' points differently from \vec{v}_\infty, the two sums have different magnitudes.
Maximum gain is a 180° turn, giving:
\Delta v = 2v_p
Worked example: the slingshot analogy. Throw a ball at 10 m/s at a wall approaching you at 5 m/s. In the wall's frame it arrives at 15 m/s and leaves at 15 m/s. In your frame it leaves at 15+5 = 20 m/s. The ball gained 10 m/s and the wall lost an unmeasurably small amount of momentum.
Where the energy comes from: the planet. Momentum and energy are conserved exactly, and the planet's orbit changes — by an utterly negligible amount because of the mass ratio.
Compute it for Voyager 2 at Jupiter. The spacecraft gained about 10 km/s and has a mass of 722 kg. Jupiter's mass is 1.898\times10^{27} kg:
\Delta v_{\text{Jupiter}} = \frac{m_{\text{sc}}\Delta v_{\text{sc}}}{M_J} = \frac{(722)(10^{4})}{1.898\times10^{27}} = 3.8\times10^{-21}\ \text{m/s}
Jupiter slowed by four billion-billionths of a millimetre per second. Over the age of the solar system that would move it about 10^{-4} m.
Gravity assists are free in every practical sense, and they are the reason the outer solar system is reachable at all.
Voyager 2
The Grand Tour exploited an alignment of Jupiter, Saturn, Uranus and Neptune that occurs once every 175 years.
| Encounter | Date | Speed gain |
|---|---|---|
| Jupiter | July 1979 | +10 km/s |
| Saturn | August 1981 | +5 km/s |
| Uranus | January 1986 | +2 km/s |
| Neptune | August 1989 | - (used to bend towards Triton) |
Without assists, reaching Neptune directly would have taken about 30 years and a far larger rocket. Voyager 2 did it in 12.
And it is still transmitting, 47 years after launch, from over 20 billion kilometres, on a signal received at about 10^{-19} watts.
Other examples:
Cassini used Venus twice, Earth once and Jupiter once over seven years to reach Saturn.
MESSENGER needed six assists over six and a half years to reach Mercury, because arriving at Mercury means shedding enormous orbital energy.
Parker Solar Probe uses seven Venus flybys to lose energy and spiral closer to the Sun, reaching 6.9 million km and 690,000 km/h — the fastest human-made object.
Juno performed an Earth flyby that gained 7.3 km/s, and during it the tracking showed a small unexplained velocity anomaly. The "flyby anomaly" has appeared in several missions at the level of millimetres per second, and no accepted explanation exists. It is one of the few genuinely open puzzles in solar system dynamics.
Chaos
Poincaré's discovery, made while competing for a prize offered by King Oscar II of Sweden in 1889.
He submitted an entry, won, and then found an error while preparing it for publication. Correcting it revealed that the restricted three-body problem has orbits that never repeat and depend so sensitively on initial conditions that long-term prediction is impossible. He paid to have the printed copies destroyed — more than the prize money — and the corrected version founded chaos theory.
Sensitive dependence means two nearby trajectories separate exponentially:
\delta(t) \approx \delta_0e^{\lambda t}
\lambda is the Lyapunov exponent, and 1/\lambda is the timescale on which prediction fails.
The solar system's Lyapunov time is about 5 million years.
What that means concretely. An uncertainty of 1 metre in a planet's position today grows to:
\delta = 1\times e^{100/5} = e^{20} = 4.9\times10^{8}\ \text{m}
Half a million kilometres after 100 million years. Over the age of the solar system the uncertainty exceeds the size of the orbits entirely.
\boxed{\text{The solar system is chaotic and nobody can predict where the planets will be in 100 million years.}}
But it is stable in a weaker and more useful sense. Numerical integrations by Laskar and others, running thousands of simulations over 5 billion years, find:
The outer planets are essentially stable.
Mercury is the problem. In about 1 % of simulations its eccentricity grows enough over billions of years that it either collides with Venus, falls into the Sun, or is ejected. In a few of those cases it destabilises the inner system and Earth is affected.
A 1 % chance of catastrophe over 5 billion years, which is reassuring on any human timescale and is a genuine result about the world.
KAM theory — Kolmogorov, Arnold and Moser, from the 1950s and 60s — explains why the system is as stable as it is: for sufficiently small perturbations, most orbits survive as slightly deformed versions of the unperturbed ones. Only orbits near resonances are destroyed, which is exactly what the Kirkwood gaps in the asteroid belt show (Chapter 11.2).
And chaos is exploited, not just endured.
The Interplanetary Transport Network uses the fact that near the unstable Lagrange points, tiny impulses produce large changes in trajectory. Weak stability boundary transfers ride these low-energy pathways between bodies at a fraction of the Hohmann cost.
Japan's Hiten spacecraft, in 1991, reached lunar orbit this way after its primary mission failed and it lacked the fuel for a conventional transfer. Edward Belbruno designed the trajectory, which took five months instead of three days and used almost no fuel.
GRAIL used a similar low-energy transfer, taking 3.5 months to the Moon and saving enough fuel to justify the delay.
Where this shows up in your life
Space weather warnings come from spacecraft at L1, giving about an hour of notice before a solar storm.
Every deep space image you have seen — from Webb, Planck, Gaia, Herschel — came from L2.
GPS accuracy depends on modelling perturbations from the Sun, Moon and Earth's non-spherical field.
And chaos theory, which started here, now underlies weather forecasting, population dynamics, cardiac arrhythmia models, and the reason a forecast beyond about ten days is impossible in principle.
What the next chapter fixes
Every manoeuvre in this Part has been priced in delta-v and none of it has been paid for. Getting delta-v means throwing mass backwards, and the relationship between the fuel carried and the speed gained is exponential — which is the single hardest fact in spaceflight. Chapter 11.7 derives the rocket equation, works out why staging is not optional, explains specific impulse and gravity turns, and shows why reaching orbit is a problem of speed rather than of height.