Catching up by slowing down, which cost Gemini 4 its fuel

On 3 June 1965, James McDivitt tried to fly Gemini 4 up to the spent second stage of its own launch vehicle, a few hundred metres away and clearly visible. He pointed at it and thrust. It receded. He thrust harder, and it receded faster.

After several attempts he had spent something like half the mission’s manoeuvring propellant and the booster was further away than when he started. The attempt was abandoned.

Nothing was wrong with the spacecraft or with the piloting. The manoeuvre was impossible in the form attempted, because in orbit thrusting toward a target ahead is the way to fall behind it. Reaching something in front means slowing down.

Why thrusting forward loses ground #

The paradox has one cause, and it is the harmonic law.

Fire prograde and the orbit’s energy rises, so its semi-major axis rises, so its period rises. The spacecraft is now on a bigger, slower orbit. Half a revolution later it is at a higher altitude than the target — and behind where it would have been, because its period is longer and it has fallen back in phase.

The immediate effect is the opposite: for the first few minutes after a prograde burn the spacecraft does close on a target directly ahead, because it is momentarily moving faster. That is exactly the trap. The instantaneous behaviour agrees with intuition, and the behaviour over a revolution reverses it.

Fire retrograde and everything inverts. The energy falls, the orbit shrinks, the period shortens, and the spacecraft comes round faster than the target. It is lower, and moving faster, and it gains ground each lap.

There is no separate control over speed and altitude in orbit. There is one quantity — the orbital energy, which appears as the semi-major axis in the vis-viva relation — and a burn changes it. Everything else follows.

The phasing manoeuvre, and its arithmetic #

The correct approach is to change the period deliberately, wait, and change it back.

Drop the chaser into an orbit a small fraction lower. Its semi-major axis falls, and the harmonic law gives the new period as $(a'/a)^{3/2}$ of the old. The angle gained per revolution is $360°(1 - P'/P)$, and the number of revolutions required is the angle to be closed divided by that.

The trade is between propellant and time, and it is unusually clean. A deeper drop gains more angle per lap and costs more $\Delta v$; a shallower one is cheap and slow. Real missions choose according to what is scarce — an uncrewed cargo vehicle takes two days and a low budget, while a crewed vehicle on a “fast rendezvous” profile reaches the station in about three hours by launching into a very precisely timed phase angle and paying for a deep phasing orbit.

That last point is the one that makes fast rendezvous possible at all. The launch itself is the cheapest phasing manoeuvre available: lifting off at exactly the right moment puts the spacecraft at exactly the right phase angle, and there is nothing to close. The three-hour profile is not a propulsion achievement; it is a timing achievement, and it became possible when the station’s orbit could be predicted precisely enough to plan a launch window minutes wide.

The view from the target #

The manoeuvre becomes far more legible in a frame attached to the target, and the equations of motion in that frame have a closed-form solution that every rendezvous is flown on.

Write the chaser’s position relative to the target in a rotating frame — $x$ along the velocity vector, $z$ toward the Earth. For separations small compared with the orbit radius, the linearised equations are the Clohessy–Wiltshire equations, and their solutions are worth knowing because none of them behaves the way an intuition trained on aircraft expects.

A chaser at the same altitude with a small velocity difference along-track drifts steadily. A chaser at a different altitude, with matched velocity, drifts along-track at a constant rate — higher means drifting backwards, lower means drifting forwards — which is the phasing manoeuvre in local coordinates. And a chaser given a small radial impulse does not go up: it traces an ellipse around its starting point, twice as wide as it is tall, and returns to where it began after one orbit.

That last solution is the useful one. A football orbit — a closed relative ellipse — lets a spacecraft loiter near a target indefinitely without station-keeping, and it is what a vehicle waiting for a docking window flies.

The equations also explain why approaches are made from particular directions. Approaching along the velocity vector — a V-bar approach — requires continuous thrust, because any closing velocity along-track immediately changes the altitude and starts a drift. Approaching from directly below — an R-bar approach — is naturally stable: drifting toward the target reduces the altitude difference, which reduces the drift, so the approach is self-damping. Gravity does the braking. R-bar approaches are standard for docking with crewed stations for exactly this reason, and they also point the thrusters away from the station.

The braking that has to be planned #

Closing the gap is the easy half. Arriving is the half that consumes the propellant, and the reason is that a rendezvous is not finished when the two vehicles are in the same place.

Two spacecraft at the same point with different velocities are about to be at different points. So the terminal phase has to null three quantities at once — the relative position, the relative velocity, and the relative attitude — and each of them is controlled by a different set of thrusters on a vehicle whose mass is changing as it fires.

The standard structure is a sequence of computed impulses rather than a continuous approach. From the end of the phasing orbit the chaser flies a series of Lambert transfers: given where the target will be at a chosen time, solve for the conic that arrives there, burn onto it, coast, and correct. Each leg is planned to end at a defined waypoint — a station-keeping point a few hundred metres out, then one at a few tens of metres — where the relative velocity is deliberately brought to zero and everything is checked before proceeding.

Those hold points are not caution for its own sake. A relative velocity of 10 cm/s covers 360 metres in an hour, so a small unnulled residual becomes a large excursion on the timescale a docking takes; and the closer the vehicles are, the less time there is to notice. The braking gates get closer together as the range shrinks for exactly that reason.

The propellant spent on the terminal phase is comparable to the phasing manoeuvre despite covering a thousandth of the distance, which is the clearest statement of where the difficulty lies.

What was actually flown #

The Gemini programme existed largely to make rendezvous routine, because Apollo could not work without it, and the sequence of attempts is a clean record of a technique being learned.

Gemini 4, June 1965, failed as described. The crew had no rendezvous radar and no onboard computer capable of the solution, and the flight plan had underestimated the problem.

Gemini 5, August 1965, carried a rendezvous evaluation pod but a fuel-cell problem cut the attempt short.

Gemini 6A and 7, December 1965, succeeded. Wally Schirra flew a phasing approach from below, closed to within 30 centimetres, and station-kept for over five hours. The manoeuvre used about 50 kg of propellant.

Gemini 8, March 1966, achieved the first docking — and then a stuck thruster set the joined vehicles tumbling at one revolution per second, and Armstrong separated and used the re-entry control system to stop it, which ended the mission.

The measurement that matters here is the propellant. Gemini 4’s failed attempt consumed roughly 42% of its orbital manoeuvring propellant; Gemini 6A’s successful rendezvous, closing a far larger gap, used a fraction of that. The difference is entirely the manoeuvre chosen, and it is the most direct demonstration available that orbital intuition has to be learned rather than felt. The propellant it cost is an exponential in the velocity change, which is why a wasted 40% is not recoverable.

Modern rendezvous is automatic and routine, and the numbers reflect it. A Soyuz or a Progress on the fast profile reaches the station in two orbits; the approach is flown by relative GPS and laser ranging, and the terminal phase is a sequence of computed impulses executed to centimetres per second — each one a difference of two vis-viva speeds, like every other burn. The physics has not changed since 1965 — only the ability to solve it in real time on board.

The same problem, without a spacecraft #

The phase-angle arithmetic is not about rendezvous. It is about two bodies on different orbits coming back into the same relative configuration, and that quantity has a name and a much wider use.

The synodic period is the time between successive identical alignments of two bodies:

$$\frac{1}{T_{\text{syn}}} = \left|\frac{1}{T_1} - \frac{1}{T_2}\right|,$$

which is exactly the phasing calculation with the roles relabelled. For a chaser and a target in nearly the same orbit, $T_{\text{syn}}$ is the time to close a full 360°, and the phasing manoeuvre is a deliberate shortening of it.

The same expression governs when a mission to Mars can launch — 25.6 months for Earth and Mars — and when a planet is at opposition, and how often two moons of Jupiter line up. It also fixes the difference between a sidereal and a solar day, which is the synodic period of the Earth’s rotation against its own orbit.

There is a limiting case worth noticing. As two periods approach each other, the synodic period runs to infinity: two bodies on nearly identical orbits take nearly forever to lap one another. That is why a spacecraft in the same orbit as its target, a few kilometres away, will stay a few kilometres away essentially indefinitely — and it is why rendezvous requires a deliberate manoeuvre rather than patience. It is also the reason co-orbital bodies can persist: two objects sharing an orbit interact so slowly that the interaction has time to be subtle.

The same arithmetic, used to avoid a meeting #

The phasing calculation runs in reverse, and in that direction it is the busiest application of orbital mechanics there is.

Two catalogued objects in similar orbits have a synodic period, and every synodic period brings them back into the same relative configuration — so a conjunction that nearly happened will nearly happen again. Predicting when means the same period-difference arithmetic, applied to objects that are not trying to meet.

The scale of the problem is set by that. Around forty thousand objects larger than 10 cm are tracked, and the number of close-approach screenings run per day is in the hundreds of thousands. Most are dismissed immediately; a few hundred a day are flagged for analysis; a handful a year result in a manoeuvre.

The uncertainty that decides is the along-track one, and it is the sixth orbital element again. The two objects’ positions across the encounter are known to metres; their positions along their orbits are known to hundreds of metres or worse, because atmospheric drag depends on solar activity that cannot be predicted a week ahead. A collision probability is therefore an integral over an error ellipsoid that is enormously elongated in one direction — and the counterintuitive consequence is that better tracking can make a computed probability rise, because shrinking the ellipsoid concentrates it.

The same reversal, one rung larger #

The rule that reaching something ahead means slowing down is not confined to two vehicles a few hundred metres apart. It governs the whole geometry of interplanetary flight, and there it produces a result that sounds wrong even to people who accept the rendezvous version.

Reaching an inner planet requires losing energy. Venus is closer to the Sun than the Earth is, so an orbit that reaches it is smaller, so it has less energy, so the departure burn has to be retrograde with respect to the Earth’s motion around the Sun. A spacecraft bound for Venus is thrown backwards along the Earth’s orbit.

The extreme case is the Sun itself, and it is the most expensive destination in the solar system. To fall into the Sun a spacecraft must cancel essentially all of the Earth’s orbital velocity — about thirty kilometres a second — whereas escaping the solar system entirely from the Earth’s orbit needs only the difference between escape speed and the speed already possessed, some twelve. Leaving is cheaper than arriving, by a factor of two and a half, and the reason is the same one that stranded Gemini 4: the intuition that a nearby destination is an easy one has no purchase on a system where position is controlled through energy.

Missions to the inner solar system are therefore flown the way a phasing orbit is flown, by removing energy in stages rather than all at once, and the stages are gravity assists rather than burns. A spacecraft aiming for a close solar orbit flies past Venus repeatedly, each pass taking a little of its angular momentum, and spends years doing what no propulsion system it could carry would achieve.

The structural similarity to the phasing orbit is exact. In both cases the direct route is unaffordable, the affordable route requires going the apparently wrong way first, and the currency being spent is time.

The arithmetic has one more property that decides how rendezvous is scheduled, and it is easiest to see at a lead angle nobody would choose.

That is the sense in which phasing is a slow manoeuvre rather than an expensive one. The Δv is trivial — a few tens of metres a second, against the nine kilometres a second it took to reach orbit at all — and the cost is measured in revolutions, consumables and crew time. A mission that wants to arrive quickly has to arrange the phase angle before launch, which is what makes a rendezvous launch window a moment rather than an interval.

The two ways out are both about the phase angle rather than about the manoeuvre. Launch when the target is at the right angle ahead, and the phasing is short; or accept a longer chase and launch when the pad and the weather allow. The station’s four-orbit and thirty-four-orbit approach profiles are exactly those two choices, and the fast one exists only because the launcher’s guidance became precise enough to place the vehicle in the correct phase directly. Nothing about the orbital mechanics changed; the ability to hit a narrow window did. It is a rare case of an operational improvement that came entirely from precision rather than from performance: the same vehicle, the same propellant, the same orbital mechanics, and a journey shortened from two days to six hours because the insertion could be placed within a few degrees of where it was aimed. Everything the fast profile saves is time the old profile spent removing an error nobody had to make. The slow profile is still flown whenever the launcher cannot guarantee the phase, and it remains the fallback when anything goes wrong on the way up.

Where the model stops #

Circular, coplanar orbits. Any difference in orbital plane must be removed separately, and a plane change is expensive — proportional to the orbital speed, so about 130 m/s per degree in low Earth orbit. Rendezvous missions therefore launch into the target’s plane, which is what makes the launch window instantaneous rather than merely narrow.

Impulsive burns. The phasing arithmetic assumes instantaneous velocity changes. A long burn spreads the change over an arc and the closed-form answer needs correcting.

Two bodies. The Earth’s equatorial bulge makes the two orbital planes precess at slightly different rates if their altitudes differ, so a phasing orbit slowly opens a plane difference that must be paid for. It also means the target’s own orbit drifts between planning and execution.

No drag. At station altitudes the atmosphere is thin and not absent, and the target’s own orbit decays by tens of metres a day. Rendezvous planning uses a predicted target orbit, not a measured one, and the prediction is the largest error source.

The figures have a limitation that is central rather than incidental. They are drawn in the inertial frame, where the manoeuvre is a pair of nested orbits and the geometry is clear — and no crew has ever experienced that view. What is seen from the spacecraft is the relative motion, in which the target appears to drift backwards and downwards as the chaser closes from below, along a curved path with no obvious relation to thrusting. The inertial picture is the one that explains the manoeuvre and the relative picture is the one that has to be flown, and the whole difficulty of rendezvous is holding both.

A closing note on why the manoeuvre remained counterintuitive even after it was understood. Every pilot’s intuition is trained in a regime where thrust and position are directly coupled — push forward, go forward. In orbit the coupling runs through the energy and takes half a revolution to appear, so the feedback a pilot learns from is delayed past the point of usefulness. Training for rendezvous is largely training to distrust the first few seconds of every input.

The ladder from here #

Later rungs on this anchor: the Clohessy–Wiltshire equations derived, and their solution set. The football orbit and natural motion circumnavigation. V-bar and R-bar approaches, and why one is self-damping. Phasing orbits optimised against time and propellant. Launch windows for rendezvous, and why they are instantaneous. Fast rendezvous profiles. Lambert’s problem applied to the terminal phase. Relative navigation — radar, laser, and relative GPS. Docking mechanisms and contact dynamics. And the rendezvous problem for a target that is tumbling and uncooperative, which is the current frontier and the reason debris removal is hard.

Buzz Aldrin’s 1963 doctoral thesis was titled Line-of-Sight Guidance Techniques for Manned Orbital Rendezvous, and its dedication reads: “In the hopes that this work may in some way contribute to their exploration of space, this is dedicated to the crew members of this country’s present and future manned space programs.” He joined that crew the following year and flew the rendezvous on Gemini 12 himself, by hand, after the radar failed.