A manoeuvre that has never been flown once

A spacecraft arriving at Mars on a transfer from Earth comes in on a hyperbola with a couple of kilometres a second of excess speed. To stay, it has to lose that excess and a good deal besides — a burn of one to two kilometres a second, paid at the worst speed there is to pay it, which through the exponential in the rocket equation is a large fraction of the mass that left Earth.

The atmosphere will do it for nothing. One pass at the right depth removes exactly the right energy, the vehicle leaves on an ellipse, and a small burn at apoapsis raises the periapsis clear of the air. The propellant saved is most of the orbit-insertion budget.

It has never been done. The slow version — dipping repeatedly over months, taking a little out each time — has been flown at Venus and at Mars several times. The single-pass version has been studied since the 1960s, designed for a dozen missions, and flown by none, and the reason is one number.

The energy a pass removes #

The arithmetic is short and the exponential in it is everything.

A vehicle passing near periapsis through an exponential atmosphere meets a column of air whose integral along the path has a closed form. Because the density falls with a scale height $H$ and the path curves around a body of radius $r_p$, the effective path length is $\sqrt{2\pi r_p H}$ — for Mars, about three hundred kilometres — and the column is that times the density at periapsis.

The drag impulse is the column divided by the ballistic coefficient, so the speed after the pass is

$$v_1 = v_0\exp\!\left(-\frac{\rho(h_p)\sqrt{2\pi r_p H}}{2\beta}\right),$$

and the remaining energy gives the outgoing orbit by vis-viva.

Two features of that expression decide the whole manoeuvre.

The dependence on periapsis is exponential, through $\rho(h_p)$. A scale height on Mars is eleven kilometres, so a one-kilometre error in periapsis is a nine per cent error in the density met and therefore in the energy removed.

The dependence on the vehicle is through $\beta$ alone. A heavier or slimmer vehicle needs to go deeper. That is the one parameter a designer controls, and it does not change the shape of the curve.

What a corridor is #

The vehicle has to end up in a useful orbit, and “useful” is a tolerance. Too shallow and it leaves on a hyperbola, still escaping. Too deep and it loses too much energy — either arriving on an orbit too low to be useful, or not leaving at all.

Between those is the corridor: the range of entry conditions that produce an acceptable outcome. It is conventionally quoted as a range of flight-path angle at the entry interface, which is the aim point a hyperbolic approach is navigated to, and it is equivalent to a range of periapsis altitude.

For a returning capsule the corridor is set by a deceleration limit at the steep end and a skip-out at the shallow one, and it is a degree and a half wide. For aerocapture the requirement is tighter, because the outcome is not “survive” but “arrive at a stated apoapsis”, and an apoapsis tolerance of ten per cent is a very narrow band of periapsis.

The figures put it at a couple of hundred metres for a ballistic vehicle at Mars. That is a demanding target for an approach navigated from a hundred million kilometres away — and it is not the binding constraint.

The factor of two #

The binding constraint is that the curve’s position is uncertain by more than the corridor’s width.

The density at a given altitude is known to a factor of two at Mars, and worse in the lower thermosphere where an aerocapture pass goes. That is not a navigation error; it is a statement about the atmosphere, and no amount of tracking removes it.

A factor of two in density is a factor of two in the energy removed at a given altitude. Recovering the intended energy means moving the aim point by a scale height times the logarithm of two — about seven and a half kilometres at Mars.

Seven and a half kilometres of aim-point uncertainty against a two-hundred-metre corridor is a ratio of thirty. A vehicle that aims where the nominal atmosphere says will miss by more than the corridor allows, in one direction or the other, most of the time.

That is the whole reason the manoeuvre has not been flown, and it is worth being clear that it is not a failure of engineering. The corridor is narrow because the physics is exponential and the atmosphere is uncertain because it is an atmosphere.

Two ways out, and both have been chosen #

Faced with an uncertainty larger than the tolerance, there are two responses, and the missions that have flown took the second.

Fly the pass rather than aim it. A vehicle with lift can raise or lower its trajectory during the pass by rolling, which changes the vertical component of the lift. A guidance law measuring the deceleration in real time can infer the density it is actually meeting and adjust, so the vehicle targets the outgoing energy rather than the periapsis. That converts the problem from open-loop aiming to closed-loop control, and the corridor widens by roughly a factor of the lift-to-drag ratio.

That is the design every aerocapture study since the 1970s has assumed, and it requires a vehicle with lift, an inertial measurement unit, and a guidance computer running a real-time energy prediction. All three exist. None has been flown for this purpose, because the failure mode is losing the mission rather than losing a manoeuvre.

Or take many small bites. Aerobraking dips to an altitude where the energy removed per pass is a small fraction of what is needed, and repeats for months. Each pass is shallow enough that a factor of two in density is a factor of two in a small number, and the accumulated error is corrected by adjusting the periapsis between passes.

Aerobraking has been flown at Venus once and at Mars four times, saving of order a kilometre a second each. It costs months of flight time and continuous operations, and it puts the spacecraft through hundreds of heating cycles.

The two are the same manoeuvre at different values of one parameter, and the choice between them is a choice between a risk taken once and a cost paid slowly.

Where it would be worth most #

The saving scales with the arrival speed, and that sorts the destinations.

At Mars the saving is of order a kilometre a second on an insertion budget of one to two — substantial, and aerobraking already captures most of it.

At Venus the arrival speed is higher and the atmosphere is thick, so the saving is larger and the heating is severe.

At Titan the combination is the most favourable anywhere: a thick extended atmosphere, low gravity, and a long approach. Studies have put the saving there at several kilometres a second, which is the difference between a mission that can carry an orbiter and one that cannot.

At Neptune and Uranus the case is strongest of all and the risk is worst. An arrival at an ice giant carries eight to ten kilometres a second of hyperbolic excess after a decade of cruise, so a propulsive insertion is prohibitive and aerocapture may be the only way to put an orbiter there at all, since the exponential in the rocket equation prices the alternative out. The corridor there is wider in absolute terms — kilometres rather than hundreds of metres, because the scale height is five times larger — and the aim-point shift is wider in the same proportion, so the ratio improves and does not resolve. And the atmospheres of both are known only from occultations and one flyby each, which is precisely the factor-of-two case the figures show failing.

The destinations where the manoeuvre is necessary are the destinations where the atmosphere is least known, and that inverse relation is not an accident: both follow from being far away and rarely visited.

What the vehicle has to survive #

The corridor is the navigation problem and the heat shield is the engineering one, and they pull in opposite directions.

The kinetic energy removed from the orbit goes into the gas, and a large fraction of that returns to the vehicle as heat. The convective heating rate at the stagnation point scales roughly as $\sqrt{\rho/R_n}\,v^3$, so the two ways of making the manoeuvre easier both make it hotter: going deeper removes more energy per pass and raises the density term, and arriving faster raises the cube.

At an ice giant, where the arrival speed is eight to ten kilometres a second above escape, radiative heating from the shock layer becomes comparable with the convective term and then exceeds it — because the shock-heated hydrogen radiates, and the radiation scales far more steeply with speed than the convection does.

That produces a design bind with no obvious resolution.

A low ballistic coefficient — a large light vehicle — decelerates high up in thin air, which keeps the heating low. It also makes the vehicle large, which is expensive to launch and hard to package.

A high ballistic coefficient is compact and goes deep, which is where the heating is.

And a high lift-to-drag ratio, which is what widens the corridor, requires a slender shape, which is exactly the shape with a small nose radius and therefore the highest stagnation heating.

The three requirements — a wide corridor, a survivable heat load, and a packageable vehicle — are mutually opposed, and the studies that come closest use inflatable or deployable decelerators to get a low ballistic coefficient without a large launch volume. None has flown at another planet.

What has been measured, and it is the wrong quantity #

Every aerobraking campaign has produced a large dataset of exactly the measurements aerocapture needs, and the data have not settled the question.

An aerobraking spacecraft measures the density along hundreds of passes, by accelerometer and by the change in its own orbit. Those are the best measurements of Mars’s upper atmosphere there are, and they have mapped its structure, its variability with season and with dust, and the amplitude of the waves in it.

What they show is that the variability is large. Pass-to-pass density at the same altitude varies by tens of per cent, driven by atmospheric waves whose amplitude depends on the topography beneath and on the season. Dust storms change the density at aerobraking altitudes by factors.

So the measurement campaign did not convert the factor of two into a small number. It confirmed that the number is a real variability rather than a modelling shortfall, and a real variability is not removed by a better model.

The consequence is that closed-loop guidance is not an optional refinement. A vehicle aiming open-loop at a density known to a factor of two cannot succeed; one measuring the density as it goes can. The data campaign established which of those two statements is true.

Where the model stops #

A single exponential is not an atmosphere. The closed form used here assumes one scale height, and a real atmosphere’s scale height varies with altitude as the temperature and composition change. A pass spans several scale heights, so the column integral has to be done through a real profile — which shifts the aim point and does not change the corridor’s width, since the width comes from the local logarithmic derivative.

Drag alone is not the physics of a pass. A lifting vehicle’s trajectory depends on its lift as well as its drag, and the two enter differently: drag removes energy and lift changes the depth reached, so a full treatment is a two-dimensional trajectory integration rather than an impulse.

And the heating is not in any of this. The energy removed from the orbit goes somewhere, and most of it goes into the shock layer and thence into the heat shield. The peak heating rate scales as the density to the half power and the speed cubed, so a deeper pass is quadratically worse — and for an ice-giant arrival the heating, not the corridor, is what limits the design.

Why aerobraking is not a small aerocapture #

The two are described above as the same manoeuvre at different values of one parameter, and the difference in operations is larger than that suggests.

An aerobraking campaign is flown as a control loop with a human in it. Each pass removes a per cent or so of the energy needed; the orbit is tracked afterwards; the density met is inferred; and the periapsis for the next pass is adjusted by a small burn. The loop runs hundreds of times over months, and the atmosphere’s variability appears as noise on a converging process rather than as a single throw.

That is why the factor of two is survivable there. A pass that meets twice the expected density removes twice the expected energy, which is twice a small number, and the next pass is adjusted. A pass that meets a tenth removes a tenth, and the campaign takes longer.

What aerobraking costs is not propellant but time and attention: months of continuous operations, a spacecraft in an eccentric orbit with a periapsis inside the atmosphere, and hundreds of thermal cycles on structures designed for vacuum. It also delays the science by the length of the campaign, which for an orbiter with a limited lifetime is a real cost.

The trade is therefore between a risk concentrated into six minutes and a cost spread over six months, and every mission so far has preferred the second — which is a statement about institutional risk tolerance as much as about engineering.

The general shape #

The structure worth taking away is about when an uncertainty is fatal, and it is a comparison rather than a magnitude.

A factor of two in a density sounds large and is not, on its own, disqualifying: an aerobraking campaign lives with it comfortably. What makes it fatal for aerocapture is that the tolerance is a tenth of a scale height while the uncertainty is a whole one.

An uncertainty matters when it exceeds the tolerance, and both have to be quoted in the same units before anything can be concluded. That comparison is what the hero figure is: a corridor width and an aim-point shift, on one axis, in kilometres.

The same comparison decides whether a conjunction warrants a manoeuvre — a position error against a miss distance — and whether an ephemeris is good enough to predict an occultation, and whether a measurement can distinguish two models. In each case the useful question is not how well something is known but how well it needs to be, and the answer is a ratio.

When the ratio is thirty, no amount of care with the numerator helps. The response has to change the denominator, which for aerocapture means measuring the atmosphere during the pass rather than before it.

Still open: the first flight #

Everything needed for a single-pass aerocapture exists. The guidance algorithms have been developed and tested in simulation for fifty years; the vehicles that fly entry corridors at Earth and at Mars have the lift and the sensors; the navigation delivers the approach.

What has not happened is a mission willing to try it, and the reason is a straightforward asymmetry. A propulsive insertion that underperforms leaves a spacecraft in a poor orbit. An aerocapture that goes shallow leaves it on a hyperbola, departing, with no second chance; one that goes deep destroys it.

The proposals that come closest are small dedicated demonstrations — a cheap vehicle sent to Mars or to Venus for no purpose but to perform the manoeuvre and report. Several have been designed and none funded, because a demonstration mission competes for money against missions that return science.

So the manoeuvre sits where it has sat since the 1960s: worth a kilometre a second or more, understood in every detail, and waiting for somebody to be the first.