Gravity assist
A Socratic walk-through of gravity assist — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #How can a spacecraft leave a planet faster than it arrived, when that planet's gravity pulls back just as hard on the way out?
A probe falls toward Jupiter, whips around it, and departs with more speed than it had coming in. That looks like theft. Gravity is a conservative force: whatever it gives you falling in, it takes back climbing out, and the books should balance to zero.
So either something supplies energy, or the question contains a hidden assumption. Which is it?
Reasoning it through
REASONING #Press on the word "faster". Faster than what? A speed is never a property of an object; it is a relation between an object and something we have chosen to hold still. The question quietly picks two different somethings and treats them as one.
Take the planet as the thing held still. In Jupiter's frame the gravitational field around it is unchanging, and energy in an unchanging field is conserved exactly. The probe arrives from far away at some speed, dives, swings, and climbs back out to the same great distance. It must leave at precisely the speed it arrived. Not approximately — identically. Relative to Jupiter, a flyby gives you nothing at all. Only the direction has changed.
So the paradox is already dissolved, and what is left is a different and better question: if the speed relative to the planet is untouched, why does the speed relative to the Sun change?
Because Jupiter is moving. The probe's velocity in the Sun's frame is the sum of two vectors: Jupiter's velocity, and the probe's velocity relative to Jupiter. The flyby cannot alter the length of the second vector, but it rotates it — and rotating one vector in a sum changes the length of the sum. That is the whole mechanism. Not energy from gravity; geometry in a moving frame.
Put numbers on it, chosen for clarity rather than drawn from a real mission. Let Jupiter travel at 13 km/s (its orbital speed, recalled as about 13.1). Let a probe approach head-on at 10 km/s in the Sun's frame. Relative to Jupiter it is closing at 10 + 13 = 23 km/s. Now imagine the extreme case, a perfect reversal: it departs at 23 km/s the other way relative to Jupiter, which in the Sun's frame is 23 + 13 = 36 km/s. It gained 26 km/s, exactly twice the planet's orbital speed. That doubling is the ceiling of the whole technique, and it depends only on how fast the planet moves — which is the tell that the gravity is not the source.
Then where did 26 km/s come from? From Jupiter. Momentum given to the probe is momentum taken from the planet. Take an 800 kg probe (a round figure) and Jupiter's mass of 1.898 x 10^27 kg (recalled): the planet's velocity changes by 800 x 26,000 / 1.898 x 10^27, which is about 1 x 10^-20 m/s. Undetectable by any instrument that will ever exist — and exactly the amount the ledger requires. Jupiter's orbit shrinks by an unmeasurable sliver, and the probe carries away the difference.
Two tests kill the folk account that "gravity gives you free energy". First, arithmetic: set the planet's orbital speed to zero and the ceiling above becomes zero. A planet at rest relative to the Sun offers precisely nothing, no matter how deep and violent the pass, because the two frames now coincide. Second, a decisive case: pass ahead of a planet instead of behind it and the sign flips — you lose heliocentric speed. Missions do this on purpose; MESSENGER used Venus and Mercury flybys to shed speed so Mercury could capture it (recalled). Nothing that behaves like a free battery can be run backwards on demand.
The load-bearing claim is the first one: speed relative to the planet is unchanged. It is directly falsifiable. Track the probe by Doppler and compare its planet-relative speed at equal distances inbound and outbound. Find a difference that no thruster firing, no atmospheric drag, no solar pressure accounts for, and the two-body account is wrong.
Which is worth saying plainly, because that measurement has not always come out clean.
The analogy
THE ANALOGY #Throw a tennis ball at 10 m/s toward a train coming at you at 13 m/s. In the train's frame the ball arrives at 23 m/s and bounces back at 23 m/s — the train's frame sees no gain. On the ground you watch the same ball leave at 23 + 13 = 36 m/s, and it will hurt.
the ball actually touches the train and reverses completely, whereas a probe never touches anything and is only turned through a partial angle by a smooth gravitational curve — so a real assist collects a fraction of the doubling, not all of it; and a train is held on rails by the Earth, whereas Jupiter genuinely recoils, which is where the missing energy goes.
Clarifying the model
THE MODEL #Two neighbouring pieces point here and should be reconciled. Rocket staging says gravity assists "substitute for propellant" — true in the delta-v ledger, but not like for like: an assist changes velocity without spending mass, yet you cannot choose its size, direction or date, since the planet's position dictates all three. That is exactly why interplanetary launch windows are so unforgiving. Orbital rendezvous establishes that a burn changes the orbit on the far side from where you burn; an assist is the same conservation bookkeeping, except that the reaction mass is a planet.
Now the honest gaps. The clean claim that planet-relative speed is exactly conserved holds in a two-body idealisation, and the real situation is at least three bodies: the Sun's gravity acts throughout, so the "patched conic" treatment is an approximation and real trajectory design integrates the full problem numerically. And empirically, several Earth flybys in the 1990s and 2000s showed unexplained velocity residuals of a few millimetres per second — the flyby anomaly, still argued over, recalled here rather than looked up. That is about one part in a million of a typical assist, so it does not touch the mechanism; but a claim of exact conservation should not pretend the measurement came out exact.
A picture of it
THE PICTURE #How to readThe horizontal axis is how far the planet's gravity turns the probe — zero at the left for a distant miss, 180 degrees at the right for the unreachable perfect reversal. The flat line is the departure speed measured against the planet: 23 km/s whatever the turn, which is the conservation law doing its work. The rising line is the departure speed measured against the Sun, computed from the same two vectors by the cosine rule, running from 10 km/s to the 36 km/s ceiling. The gap between the lines is the entire gravity assist, and it is a gap between two frames, not two energies.
What became clearer
WHAT CLEARED #Nothing gives the probe energy except the planet, and the planet gives it by moving. Relative to the planet the encounter is perfectly symmetric and gains nothing; relative to the Sun the same encounter rotates a vector inside a sum and changes its length. The paradox came entirely from switching frames mid-sentence — and once the frame is named, the missing energy is sitting in a change to Jupiter's orbit too small ever to be seen.
Where to go next
ONWARD #- The Oberth effect: why a burn deep in a gravity well buys more energy than the same burn far from one, which is a genuine energy gain rather than a frame effect.
- How a chain of assists is planned when each one constrains the arrival conditions of the next.
Key terms
TERMS #| Term | What it means |
|---|---|
| Hyperbolic excess speed | the speed a probe retains relative to a planet once effectively free of it; unchanged by the flyby. |
| Turning angle | how far the encounter rotates the probe's planet-relative velocity; larger for a closer, slower pass. |
| Flyby anomaly | small unexplained velocity residuals observed in some Earth flybys, of order millimetres per second. |
Every term the collection defines is gathered in the glossary.