The exponential that decides what can be flown

Every other vehicle pushes against something. A car pushes on the road, a ship on the water, an aeroplane on the air. A rocket has nothing to push on, so it carries its own reaction mass and throws it backwards.

That single difference produces the defining difficulty of spaceflight, and it is not the energy. It is that the propellant has to be accelerated too. The last kilogram of propellant to be burned has been carried the whole way, and accelerating it required burning more propellant, which also had to be carried. The bookkeeping compounds, and the result is not a proportionality but an exponential.

$$\Delta v = v_e \ln\frac{m_0}{m_f}.$$

The velocity change is the logarithm of the mass ratio. Inverted, the mass ratio is the exponential of the velocity change, and that is the whole of the difficulty.

Reading the curve #

The relation is easier to read as a propellant fraction than as a mass ratio, because the fraction shows the wall.

For hydrogen and oxygen, with an exhaust speed of 4.44 km/s: buying 4.44 km/s of $\Delta v$ needs a mass ratio of $e$, so 63% of the vehicle is propellant. Buying 8.9 km/s needs $e^2$, so 86%. Buying 13.3 needs $e^3$, so 95%. Each additional exhaust-speed’s worth of $\Delta v$ removes another 63% of what is left over.

The numbers a mission actually needs sit right where the curve is steepening. Reaching low Earth orbit takes about 9.4 km/s including losses; with hydrogen and oxygen, that is a propellant fraction of 88%. With kerosene and oxygen, at 2.55 km/s exhaust speed, it is 97.5% — which means the structure, the engines, the tanks, the avionics and the payload together must fit in the remaining 2.5%.

That is why launch vehicles look the way they do. They are, structurally, thin-walled pressurised balloons; an empty Atlas booster could not stand up without being pressurised, and several were crushed by their own weight when the pressure was lost on the ground.

Where the exhaust speed comes from #

Only one number in the equation is a property of the engine, and improving it is worth far more than improving anything else.

The exhaust speed is set by how much energy per unit mass the propellant releases and how efficiently a nozzle converts that into directed motion. For a chemical rocket the ceiling is thermodynamic: the combustion temperature is limited by what the chamber survives, and the exhaust speed scales as $\sqrt{T/M}$ with $M$ the mean molecular mass of the products. Light products are better, which is why hydrogen and oxygen — producing water, molecular mass 18 — beats kerosene and oxygen, which produces a mixture averaging nearer 22.

Engineers usually quote specific impulse rather than exhaust speed: $I_{sp} = v_e/g_0$, in seconds. The division by standard gravity is a historical convention that makes the number the same whether the mass is measured in pounds or kilograms, and it has the unfortunate side effect of making a velocity look like a time. An $I_{sp}$ of 452 seconds is an exhaust speed of 4.43 km/s.

The ceiling for chemistry is around 480 seconds and has been approached. Beyond it the options change the energy source rather than the chemistry: a nuclear thermal rocket heats hydrogen with a reactor and reaches about 900 seconds, and an ion thruster accelerates ions electrically and reaches 3,000 or more. The cost of the latter is thrust — an ion engine’s is measured in tens of millinewtons — so it can only be used where a manoeuvre may take months.

Staging, which is how the exponential is beaten #

The equation has one exploitable feature: the mass ratio is the ratio of what is there at the start to what is there at the end, and what is there at the end can be reduced by throwing away the empty tanks.

Consider a vehicle needing 9.4 km/s with kerosene and oxygen. As a single stage, the propellant fraction is 97.5%, which leaves 2.5% for everything else — beyond any structure that has ever been built with a useful payload. Split it into two stages of 4.7 km/s each and the arithmetic changes completely: each stage needs a mass ratio of $e^{1.84} = 6.3$, or 84% propellant, which is entirely ordinary. The first stage’s structure is discarded before the second stage begins, so it never has to be accelerated the rest of the way.

That is the whole reason every orbital launch vehicle since 1957 has staged. It is not an efficiency measure; it is what makes the problem possible at all with materials that exist.

The gain diminishes. Three stages beat two by a smaller margin than two beat one, four by a smaller margin again, and each stage adds engines, separation mechanisms and failure modes. Two or three is the practical optimum, and the calculation that decides between them is a straightforward optimisation over how to divide the total $\Delta v$.

What was actually flown #

The equation is exact, and every real launch spends more than it says. The gap is worth accounting for, because it is a third of the budget.

An orbital velocity of 7.8 km/s becomes a required $\Delta v$ of about 9.4. The extra 1.6 km/s is not orbital mechanics at all, and it divides into three parts.

Gravity loss, the largest, is around 1.2 km/s. While the vehicle is climbing, gravity is removing 9.8 m/s of velocity every second, and the propellant burned to replace it buys nothing. It is minimised by accelerating hard and turning horizontal as early as the atmosphere permits, which is exactly what a launch trajectory looks like — nearly vertical for twenty seconds, then a long pitch-over.

Drag loss is 0.1 to 0.3 km/s. Small, because the vehicle is only in dense air briefly, and it drives the design of the vehicle’s shape and its speed profile through maximum dynamic pressure.

Steering loss is what is spent thrusting in a direction other than straight along the velocity vector.

Set against them is one gain: launching eastward collects the Earth’s rotation, worth 0.46 km/s at the equator and 0.41 at Cape Canaveral, and nothing at the poles. That is a substantial fraction of a stage, and it is why launch sites are near the equator and why almost everything launches east.

The flown numbers are checkable. The Saturn V lifted 2,970 tonnes off the pad to place 45 tonnes on a trans-lunar trajectory, a mass ratio of 66 — the price of a transfer whose cost was known before rockets existed. The Space Shuttle’s ratio was about 22 to low orbit, and its low payload fraction — 1.5% of lift-off mass — was the direct consequence of returning the orbiter, whose 78 tonnes of dry mass had to be accelerated to orbital velocity and were part of the final mass in the equation.

The payload fraction, which is what is actually being bought #

The propellant fraction says what has to be spent. The number a mission cares about is what is left, and it is small enough to be worth stating plainly.

Of a rocket’s lift-off mass, roughly 85–90% is propellant, 8–12% is structure and engines, and 1.5–4% reaches orbit as payload. The Falcon 9 delivers about 4% expendable and rather less when the booster is recovered; the Saturn V managed about 4% to low orbit; the Space Shuttle managed 1.5%.

The reason the structural fraction matters so much is that it sits inside the exponential’s final mass along with the payload. A stage’s dry mass and its cargo are indistinguishable to the equation — both are mass that must be accelerated to the end — so a kilogram saved from a tank is a kilogram gained in payload, exactly. That equivalence is why launch vehicle engineering is dominated by mass reduction to a degree that looks obsessive from outside, and why the structural coefficient, the ratio of dry mass to propellant mass, is the single figure of merit for a stage. Bringing it from 0.10 to 0.08 raises the payload of a typical two-stage vehicle by something like a third.

It also explains the awkward economics of reuse. Recovering a booster requires reserving propellant for the return and adding legs, grid fins and thermal protection — all of which are dry mass in the stage that has to be accelerated. The payload penalty is 30–40% for a downrange landing and more for a return to the launch site. Whether that is worth paying is a question about how many times the hardware flies, and it is entirely outside the equation; what the equation supplies is the price.

The generalisation: it is not about rockets #

The equation follows from conservation of momentum applied to a vehicle whose mass changes, and it appears wherever that happens.

The derivation is short. Over a small interval, expelling $dm$ of mass at exhaust speed $v_e$ relative to the vehicle changes the vehicle’s momentum by $v_e\,dm$, so $m\,dv = -v_e\,dm$. Integrating gives the logarithm. Nothing about combustion or nozzles enters, and the same expression applies to a squid, to a person throwing bricks off a raft, and to a galaxy expelling gas.

The astrophysical instance matters. A star losing mass in a wind is a rocket: as it sheds material carrying angular momentum and linear momentum, its own motion changes. For a binary in which one component is losing mass, the resulting acceleration is calculable by exactly this equation, and it is one of the mechanisms that turns a bound pair into an unbound one. A supernova that ejects its envelope asymmetrically gives the remnant neutron star a “kick” of several hundred kilometres per second, which is why pulsars move so much faster than ordinary stars — and what makes them the remnants of the shortest-lived stars — and the kick is a rocket effect with the equation unchanged.

Running the argument the other way explains why interstellar travel is hard in a way that has nothing to do with engineering. Reaching even 10% of the speed of light — 30,000 km/s — with an exhaust speed of 30 km/s requires a mass ratio of $e^{1000}$, which exceeds the number of atoms in the universe by several hundred orders of magnitude. The only escapes are to raise the exhaust speed toward $c$, which means the exhaust must be relativistic particles or light itself, or to stop carrying the reaction mass — a solar sail, a laser-pushed sail, a ramjet collecting interstellar hydrogen. Every serious proposal for interstellar flight is an attempt to get out from under this one equation.

Why the currency is velocity and not energy #

Mission budgets are quoted in kilometres per second rather than in joules, and the choice is not arbitrary — it reflects which quantity the rocket equation makes additive.

Energy is the natural currency for almost every other transport problem. It is not for this one. The kinetic energy a burn adds depends on the speed the vehicle already has, through $2v\,\Delta v + \Delta v^2$, so the same propellant buys different amounts of energy at different points of a trajectory — which is the Oberth effect, and which makes energy a poor unit of account.

Velocity change is additive in a way energy is not. A mission needing 3.2 km/s to escape and 0.6 for a Mars transfer needs 3.8, and the rocket equation converts that total into one mass ratio. The propellant cost of a sequence of manoeuvres depends only on the sum of their $\Delta v$ values, never on the order they are performed in or on where in the gravity field each happens.

That is why the delta-v map exists as a graph with additive edges, and why “how much $\Delta v$ does it have” is the single number that characterises a spacecraft. It is also why the Oberth effect is described as a way of buying more energy for the same $\Delta v$ rather than of buying more $\Delta v$: the propellant is unaffected, and only what the velocity is worth has changed.

There is a corollary worth stating because it decides how missions are argued about. Because the mass ratio is exponential in the velocity change and only linear in the payload, a mission that needs slightly more velocity is far more expensive than one that needs slightly more payload — so a design argument about capability is usually an argument about a few hundred metres per second rather than about a few hundred kilograms, and the two are not interchangeable however similar the percentages look.

It is also why the argument for a reusable vehicle is an argument about the base rather than about the exponent. Reuse does not change how much propellant a mission needs; it changes how many times the hardware carrying that propellant can be paid for, which is an economic quantity rather than a physical one and is the only term in the whole business that is not fixed by chemistry.

The curve is worth reading at both ends of the range that has ever been flown, because the exponential’s whole character changes across it.

Where the model stops #

Instantaneous mass flow, constant exhaust speed. Real engines throttle and their exhaust speed varies with ambient pressure — a nozzle optimised for vacuum performs badly at sea level and vice versa, which is why first and second stages have visibly different nozzle shapes.

No external forces. The equation describes a vehicle in free space. Gravity, drag and steering losses are all outside it, and together they are 20% of a launch budget.

Impulsive use. The equation itself does not assume a short burn, but the orbital calculations that supply the $\Delta v$ usually do. A low-thrust spiral needs substantially more $\Delta v$ than an impulsive transfer between the same orbits, because it fights gravity for months.

A single vehicle. Refuelling in orbit breaks the whole framing: it resets the mass ratio, and a vehicle that departs from a fuelled depot is solving a different and much easier problem than one that departs from the ground. The same is true of a gravity assist, which supplies velocity with no propellant at all.

The figure has a limitation that matters for reading it. It plots propellant fraction, which flattens against 1 and makes the high-$\Delta v$ end look like a gentle approach to a limit. On a mass-ratio axis the same curve is a vertical wall: 95% propellant is a ratio of 20, 99% is 100, and 99.9% is 1,000. The fraction is the honest quantity for a designer, because it is what has to fit in the vehicle, and it systematically understates how violently the requirement is growing.

That asymmetry is also why the field’s effort has gone where it has. A better engine raises the exhaust speed, which enters the exponent, and a lighter tank lowers the dry mass, which enters the base — and the first is worth more than the second by whatever factor the velocity change divided by the exhaust speed happens to be. For a launch from the ground that factor is around three, which is why chemical propulsion has been pushed to within a few per cent of what its chemistry allows while structural mass fractions have improved comparatively little.

And the staging arithmetic at a larger budget, since staging is the only way the exponential is beaten and its returns diminish.

One more reading covers the budget an outer-planet mission would need without any gravity assist at all.

The exponential is the one constraint in spaceflight that no engineering has ever softened, and every architectural decision in the field — staging, assists, aerobraking, electric propulsion — is an attempt to be somewhere else on the curve rather than to change its shape.

The ladder from here #

Later rungs on this anchor: the derivation from momentum conservation. Specific impulse, and the convention that makes it a time. Exhaust speed from thermodynamics, and the $\sqrt{T/M}$ scaling. Nozzle design and expansion ratios. Optimal staging, and the calculus that divides the $\Delta v$. Gravity, drag and steering losses computed. The launch site latitude bonus. Electric propulsion, and the thrust-versus-efficiency trade. Nuclear thermal and nuclear pulse. Aerobraking as a way of getting $\Delta v$ for free. Solar and laser sails. And the relativistic rocket equation, where the mass ratio involves a hyperbolic tangent and the situation gets worse rather than better.

Tsiolkovsky derived the equation in 1897, working as a provincial schoolteacher in Kaluga and publishing in a journal nobody in the field read. Goddard rederived it independently, and so did Oberth. Three people arrived at the same result within thirty years, each unaware of the others, because it is what conservation of momentum says and there is no other answer available.