LUNAROPS · OPERATIONAL UPLINK 100% UPTIME 1,247d POSTS 893 JEFF.MOON@LUNAROPS.DEV UTC --:--:--

How Rocket Engines Actually Work

20 min read Aviation & Aerospace
rocket-enginespropulsionaerospacestaged-combustionspecific-impulsespaceflight
Contents

A rocket engine does exactly one thing, and it does it without any help from the outside world: it throws mass backwards as fast as possible. Everything else - the combustion chambers, the turbopumps spinning at tens of thousands of RPM, the regeneratively cooled nozzle walls glowing just short of melting, the labyrinthine plumbing that has driven engineering teams to the edge of insanity - exists to serve that single goal. Thrust is reaction. You accelerate exhaust one way, the vehicle accelerates the other way, and unlike a jet engine that breathes air, a rocket carries both its fuel and its oxidizer because it has to work in vacuum where there is no air to breathe. That self-sufficiency is the whole reason rockets can leave the planet, and it is also the source of nearly every hard problem in the field. Carrying your own oxidizer means your tanks are enormous, your mass fraction is hostile, and every fraction of a percent of efficiency you can wring out of the exhaust velocity translates into payload that actually reaches orbit instead of falling back into the ocean. This post is about how that efficiency is generated, why it is so hard to generate, and the deeply unglamorous plumbing decisions that separate a rocket that flies from a rocket that detonates on the test stand.

The Combustion Chamber: Where Chemistry Becomes Pressure

The heart of a chemical rocket is the combustion chamber, and its job is to turn a controlled chemical reaction into a reservoir of very hot, very high-pressure gas. Fuel and oxidizer are injected through an injector plate - often hundreds or thousands of tiny orifices arranged to atomize and mix the propellants - and they ignite, releasing chemical energy as heat. The result is gas at temperatures that routinely exceed 3,000 K and at pressures that range from a few tens of bar in simple engines to over 300 bar in the most aggressive designs.

That pressure is the currency of the whole machine. A higher chamber pressure lets you build a smaller, lighter engine for a given thrust, and it raises the pressure ratio available to the nozzle, which is what ultimately sets how fast you can accelerate the exhaust. The SpaceX Raptor runs a chamber pressure around 300 bar; the Space Shuttle Main Engine (RS-25) ran around 200 bar; a simple pressure-fed engine might sit at 10 to 20 bar. Every bar of chamber pressure is hard-won, because the higher you push it, the harder everything upstream and downstream has to work to contain and exploit it.

The chamber is also where one of the field’s most feared phenomena lives: combustion instability. The gases inside are not burning quietly. Small pressure oscillations can couple with the acoustic modes of the chamber and with the combustion process itself, and if the feedback is positive, the oscillation grows. At its worst, high-frequency instability can destroy an engine in milliseconds, with pressure spikes that tear injectors apart and burn through chamber walls. The F-1 engine that powered the Saturn V Apollo missions famously took years of empirical work - baffles, redesigned injectors, deliberately detonating bombs inside the chamber to see if it could recover - before its combustion was tamed. There is still no fully predictive theory that lets you design instability out on the first try. It remains, in part, a dark art.

The de Laval Nozzle: Why a Constriction Makes Things Faster

The single most counterintuitive and most important component of a rocket engine is the nozzle, and specifically the converging-diverging shape invented by Gustaf de Laval. Hot gas in the chamber has high pressure but relatively low velocity. The nozzle’s job is to trade that pressure for velocity - to convert thermal and pressure energy into directed kinetic energy of the exhaust - and it does this with a shape that defies everyday intuition.

        Combustion          Throat           Diverging
         chamber          (sonic, M=1)         section
       +-----------+         |  |          /-------------+
       |           |         |  |         /               \
  ===> | hot, slow |  ====>  |  |  ====> /  supersonic,     \  ====>  EXHAUST
  ===> | high P    |  ====>  |  |  ====> \  cold, fast       /  ====>
       |           |         |  |         \               /
       +-----------+         |  |          \-------------+
        subsonic         M = 1 here        M > 1, expanding
        M < 1

In the converging section, the gas is subsonic, and like water in a narrowing pipe it speeds up as the area shrinks. At the narrowest point - the throat - the flow reaches exactly Mach 1, the local speed of sound. Here is the magic: once the flow is supersonic, the rules invert. In the diverging section, where the area increases, a supersonic flow keeps accelerating. This is because in compressible supersonic flow, density drops faster than area grows, so to conserve mass the velocity must climb. The expanding bell takes gas leaving the throat at Mach 1 and accelerates it to Mach 3, 4, or higher, while its pressure and temperature plummet. The exhaust leaves a large vacuum-optimized nozzle moving on the order of 3,000 to 4,500 m/s.

The expansion ratio - the area of the nozzle exit divided by the area of the throat - is tuned to the environment. At sea level, the atmosphere pushes back, and if you over-expand the flow below ambient pressure the exhaust separates from the walls and you lose performance, so sea-level nozzles are short and stubby. In vacuum there is nothing to push back, so you want an enormous expansion ratio to extract every bit of velocity, which is why upper-stage and vacuum-optimized engines have those huge bell extensions. The RS-25 has an expansion ratio around 69:1; vacuum engines like the RL10 or Raptor Vacuum push far higher.

The Thrust Equation and Exhaust Velocity

Thrust comes from two sources, and the full equation makes both explicit:

F = mdot * Ve + (Pe - Pa) * Ae

The first term, mdot * Ve, is the momentum thrust: the mass flow rate of exhaust multiplied by its exhaust velocity. This is the dominant term and it is the heart of the matter - throw more mass per second, or throw it faster, and you get more thrust. The second term, (Pe - Pa) * Ae, is the pressure thrust: the difference between the exhaust pressure at the nozzle exit Pe and the ambient pressure Pa, multiplied by the exit area Ae. When the nozzle is perfectly expanded, Pe equals Pa and this term vanishes; a perfectly expanded nozzle is the most efficient at that altitude. Because ambient pressure falls as a rocket climbs, no fixed nozzle is optimal everywhere, and engineers compromise.

It is often useful to fold both terms together into an effective exhaust velocity c, so that F = mdot * c. The whole game of engine performance is maximizing c. And here is the deep insight: exhaust velocity is fundamentally set by the energy per unit mass of the propellant and by how cleanly you can convert that energy into directed flow. Lighter exhaust molecules move faster at a given temperature, which is exactly why hydrogen-rich exhaust is so prized - low molecular weight means high velocity for the same energy.

Specific Impulse: The Real Figure of Merit

If you only learn one number about a rocket engine, learn its specific impulse. Specific impulse, Isp, measures how much thrust you get per unit weight of propellant consumed per second. In its cleanest form:

Isp = Ve / g0      (Ve = effective exhaust velocity, g0 = 9.81 m/s^2)

Measured this way, Isp comes out in seconds, which is a historical convenience that makes the number independent of the unit system. An Isp of 350 seconds means the engine could hold one kilogram of thrust for 350 seconds while burning one kilogram of propellant. The deeper meaning is simpler: specific impulse is just exhaust velocity in disguise. Multiply Isp by g0 and you recover the effective exhaust velocity directly. An engine with 450 seconds of vacuum Isp is producing roughly 4,400 m/s of effective exhaust velocity.

Why does this matter more than raw thrust? Because thrust gets you off the pad, but specific impulse determines how much velocity change - how much delta-v - you can extract from a given mass of propellant. A high-thrust, low-Isp engine is a brute that burns through propellant fast. A high-Isp engine is an efficient miser that makes every kilogram of propellant count. The tension between these two is one of the central trade-offs in the whole discipline, and we will return to it.

The Rocket Equation and Why Staging Exists

Konstantin Tsiolkovsky’s rocket equation is the unforgiving law that governs everything:

dv = Ve * ln(m0 / mf)

Your achievable velocity change dv equals the effective exhaust velocity Ve times the natural logarithm of the mass ratio - initial mass m0 (fully fueled) divided by final mass mf (empty). That logarithm is the cruelty. To double your delta-v at fixed exhaust velocity, you do not double your propellant; you square your mass ratio. Reaching orbit needs roughly 9.4 km/s of delta-v once gravity and drag losses are included. With a good kerolox Ve around 3,000 m/s, the equation demands a mass ratio of about e^(9400/3000), which is over 20 to 1. That means more than 90 percent of your liftoff mass must be propellant, leaving a sliver for structure, engines, and payload.

This is why staging exists. As a rocket burns propellant, it keeps hauling the now-empty tankage and engines that no longer hold anything useful. That dead mass kills your mass ratio. Staging solves it brutally and effectively: you build the rocket in sections, and when a section’s propellant is gone, you throw the whole section away - tanks, engines, structure, everything. The remaining vehicle now has a far better mass ratio for the rest of the climb, and the delta-v of each stage adds. Staging is the only practical way to beat the logarithm hard enough to reach orbit with chemical propellants. It is also why a smaller, lighter, cheaper engine can be the right choice for an upper stage even if it produces less thrust - up there, Isp is king and thrust-to-weight matters less.

Propellant Combinations and Their Trade-offs

The choice of propellant is one of the most consequential decisions in vehicle design, and there is no free lunch. The headline numbers are specific impulse and density, and they pull in opposite directions.

Propellant Type Vacuum Isp (s) Density Key trade-off
LOX / LH2 (hydrolox) Cryogenic 450-465 Very low Highest Isp; huge bulky tanks, deep cryo, hydrogen leaks
LOX / RP-1 (kerolox) Cryo + refined kerosene 340-360 High Dense and storable-ish fuel; coking and soot, mid Isp
LOX / CH4 (methalox) Cryogenic 360-380 Medium Clean-burning, reusable-friendly, ISRU on Mars; mild cryo
N2O4 / UDMH (hypergolic) Storable 320-340 High Ignites on contact, infinite restarts; toxic and corrosive
APCP (solid) Solid composite 250-285 Very high Dead simple, huge thrust, cannot throttle or shut off

Hydrolox (liquid oxygen and liquid hydrogen) gives the best specific impulse of any flown chemical combination because hydrogen’s tiny molecular weight produces blistering exhaust velocity. The RL10 and RS-25 both burn it. But liquid hydrogen is fiendish: it must be kept below 20 K, it is so low-density that the tanks are enormous (which adds dry mass and drag), and it leaks through seals that would hold any other fluid. Hydrolox shines on upper stages where Isp dominates and the bulk penalty hurts least.

Kerolox (liquid oxygen and RP-1, a refined kerosene) trades some Isp for far higher density. Dense propellant means smaller tanks and a better thrust-to-weight engine, which is exactly what you want on a first stage clawing off the pad. The Merlin and the F-1 are kerolox. The downside is that kerosene coking - carbon deposits from incomplete combustion - fouls cooling passages and complicates reuse.

Methalox (liquid oxygen and liquid methane) is the modern darling, used by Raptor and BE-4. It sits between kerolox and hydrolox on Isp and density, burns far cleaner than kerosene (which helps reusability), needs only mild cryogenics, and can in principle be manufactured on Mars from atmospheric CO2 and water. It is an excellent all-around compromise.

Hypergolics like N2O4 and UDMH or MMH ignite spontaneously on contact, which means no ignition system and reliable restarts - perfect for spacecraft maneuvering and deep-space stages. The price is toxicity and corrosiveness so severe that handling them requires hazmat suits. Solids, finally, are rubber-like blocks of fuel and oxidizer cast together; they are simple, dense, and produce enormous thrust, but once lit they cannot be throttled or shut down, which is why they serve as boosters rather than primary controllable stages.

Regenerative Cooling and Turbopumps

Two pieces of supporting hardware make high-performance liquid engines possible, and both are marvels.

The first is regenerative cooling. Combustion gas at 3,500 K would melt any chamber wall in seconds, so before the propellant is burned it is routed through tiny channels milled or brazed into the chamber and nozzle walls. The cold propellant - often the cryogenic fuel - absorbs heat from the wall, keeping the metal survivable, and arrives at the injector preheated, which actually improves combustion. The wall is cooled by the very fluid about to be burned. It is an elegant closed loop, and it is why those bell nozzles have the characteristic ribbed or tubular construction: those are the cooling channels.

The second is the turbopump, and it is the single hardest piece of machinery in the engine. Feeding a high-pressure chamber means delivering propellant at a pressure higher than the chamber - sometimes well over 300 bar - at flow rates measured in hundreds of kilograms per second. You cannot do that with tank pressure alone without absurdly heavy tanks. So you use a pump, and to spin the pump you use a turbine driven by hot gas. The RS-25’s high-pressure fuel turbopump spins past 35,000 RPM and develops over 70,000 horsepower from a unit you could carry in your arms. It pumps liquid hydrogen, which is cryogenic, low-density, and prone to cavitation, while the turbine that drives it runs in hot combustion gas a few centimeters away. The thermal gradients, the bearing loads, the seals separating cryogenic fuel from blazing turbine gas - this is the part of rocketry that engineers describe, only half joking, as the hardest plumbing problem in engineering. How you generate the gas to drive that turbine is precisely what distinguishes the engine power cycles, which is where the real design philosophy lives.

The Engine Power Cycles, Honestly

Every pump-fed engine must answer one question: where does the hot gas that spins the turbopump come from, and what happens to it afterward? The answers form a ladder of increasing performance and increasing pain.

GAS-GENERATOR (open cycle)          STAGED COMBUSTION (closed cycle)
                                    
   fuel    oxidizer                    fuel    oxidizer
     |        |                          |        |
     +---+----+                          +---+----+
         |                                   |
     [gas gen] --> turbine                [preburner] --> turbine
         |            |                       |              |
    (overboard      pumps                (fuel-rich         pumps
     exhaust,        |                    hot gas)           |
     wasted)         v                       |               v
                  CHAMBER                     +----------> CHAMBER
                     |                        (turbine exhaust
                  NOZZLE                       fed INTO chamber,
                                               nothing wasted)
                                                   |
                                                NOZZLE

Pressure-fed is the simplest possible answer: there is no turbopump at all. High-pressure gas in the tanks pushes the propellant into the chamber directly. This is dead reliable and cheap, but it caps chamber pressure at whatever your tanks can safely hold, which is low, and it forces heavy thick-walled tanks. It is used for small thrusters, some upper stages, and hypergolic spacecraft engines where simplicity and restartability beat raw performance.

Gas-generator (open cycle) is the workhorse. A small fraction of the propellant - a few percent - is burned in a separate small combustor called a gas generator, producing relatively cool gas that spins the turbopump turbine. After doing its job, that turbine exhaust is simply dumped overboard, often through the side or down the nozzle as a visible darker exhaust. Burning propellant just to throw it away is a direct Isp penalty, but the architecture is simple and robust because the turbine drive loop is decoupled from the main chamber. The Merlin and the mighty F-1 are gas-generator engines. You pay a few percent of efficiency for a great deal of design sanity.

Expander cycle is a beautiful trick that wastes nothing and burns nothing extra. Instead of a gas generator, it uses the heat picked up by the regenerative cooling channels. The cryogenic fuel, flashed to gas as it cools the chamber walls, is expanded through the turbine to drive the pumps, and then fed into the chamber and burned. No separate combustor, no dumped propellant. The RL10 is the canonical expander-cycle engine, and it achieves a superb vacuum Isp near 450 seconds. The catch is a hard ceiling: the cycle can only generate as much turbine power as the chamber walls can transfer as heat, and heat transfer scales with surface area while thrust scales with volume. So expander-cycle engines do not scale to large thrust. They are upper-stage specialists.

Staged combustion (closed cycle) is where performance gets serious. Instead of dumping the turbine drive gas, you burn all the propellant in a preburner first - producing a large volume of fuel-rich (or oxidizer-rich) hot gas - run that entire flow through the turbine to drive the pumps, and then route the turbine exhaust into the main chamber where the remaining propellant is added and combustion completes. Nothing is thrown away; every gram of propellant ends up contributing to thrust. This buys several percent of Isp over a gas generator and enables very high chamber pressures. The RS-25 and the Russian RD-180 are staged-combustion engines. The cost is brutal: the turbine now runs in the path of nearly all the propellant, the plumbing is densely coupled, and oxidizer-rich hot gas - as in the RD-180 - is so chemically aggressive that the Soviets spent decades developing metallurgy that would not spontaneously ignite in it, a feat Western engineers long believed impossible.

Full-flow staged combustion is the summit. There are two preburners, one fuel-rich and one oxidizer-rich, and the entire fuel flow and entire oxidizer flow each pass through their own turbine before meeting in the main chamber. Every molecule of propellant goes through a turbine, both pumps are driven by gas of their own propellant type (so the seals never separate fuel from oxidizer - a major reliability win), and turbine temperatures can be kept lower because so much mass is flowing. The result is the highest practical chamber pressure and Isp of any cycle. SpaceX’s Raptor is the first full-flow staged-combustion engine to fly operationally, achieving around 300 bar chamber pressure. The price is the most complex plumbing ever flown: two preburners, two turbopumps tightly coupled through shared chamber dynamics, and a control problem that gave the program enormous trouble before it converged.

The Trade-offs, and Why Not Everyone Uses Full-Flow

Cycle Example engines Relative Isp Complexity What it buys
Pressure-fed Apollo SPS, many thrusters Low Very low Reliability, restart, no pumps
Gas-generator Merlin, F-1, RS-68 Good Moderate Robust, decoupled, cheap
Expander RL10, Vinci Very high Moderate No waste, but small thrust only
Staged combustion RS-25, RD-180 Very high High High Pc, closed-cycle efficiency
Full-flow staged Raptor Highest Extreme Max Pc, max Isp, reusability margins

The obvious question is: if full-flow staged combustion is the best, why does anyone build anything else? The answer is that performance is not the only axis. Higher-performance cycles are dramatically harder to develop, more expensive, heavier in their plumbing, and far more prone to the catastrophic failure modes that make engine development a graveyard of programs.

The first trade is thrust-to-weight versus Isp. A first-stage booster fighting gravity off the pad cares enormously about thrust-to-weight - it needs to lift its own weight plus the entire stack right now, and every second spent at low acceleration is delta-v lost to gravity. Dense propellant and a simpler, lighter, higher-thrust engine often win there even at a small Isp cost. An upper stage, already moving fast and operating in vacuum, cares about Isp above almost everything, which is why high-Isp hydrolox expander engines like the RL10 dominate that niche despite being useless as boosters.

The second trade is performance versus development risk and cost. Combustion instability, turbopump bearing failures, oxidizer-rich gas attacking turbine blades, the thermal nightmare of a turbine running centimeters from cryogenic pumps - these problems get worse, not better, as you climb the cycle ladder. Staged combustion concentrates enormous energy in tightly coupled hardware where a small instability can cascade into total destruction. The RS-25 took years and an enormous budget to mature. Raptor went through a long, public, failure-strewn development before it became reliable. A gas-generator kerolox engine like Merlin, by contrast, can be built, tested, and iterated quickly and cheaply, and SpaceX flew it hundreds of times by leaning into that simplicity. Sometimes the rational engineering answer is to give up a few percent of Isp to get an engine that you can actually build, test, mass-produce, and reuse without it killing the program.

Throttling adds another wrinkle. Deep throttling - running an engine well below full power for landing or for limiting acceleration - stresses the combustion stability and the turbopump operating range, and it is much harder in high-pressure closed cycles. The Merlin and Raptor both throttle deeply for landing, and achieving that took serious work to keep combustion stable across the range. An engine that flies once does not need it; a reusable engine that lands itself absolutely does, and that requirement reshapes the whole design.

Verdict

A rocket engine is a study in compounding constraints. The nozzle physics is elegant and the equations are simple enough to fit on a napkin - F = mdot*Ve + (Pe-Pa)*Ae, Isp = Ve/g0, dv = Ve*ln(m0/mf) - but the moment you try to maximize the one number that matters, exhaust velocity, you walk straight into the hardest machinery problems in engineering. Higher chamber pressure demands turbopumps that are barely on the right side of physically possible. Closed power cycles wring out the last few percent of efficiency at the cost of plumbing so tightly coupled that the engine can tear itself apart if the combustion so much as hiccups. Better propellants improve Isp but punish you with cryogenics, leaks, toxicity, or coking.

There is no universally best engine, only the best engine for a job. A first stage wants thrust-to-weight and accepts a gas-generator kerolox or methalox engine. An upper stage wants Isp and reaches for an expander-cycle hydrolox engine. A reusable, Mars-bound architecture pushes toward full-flow staged methalox and pays the development cost in years and explosions. The genius of the field is not in any single component but in the ruthless honesty with which good engineers weigh efficiency against the brutal reality of building something that has to contain a controlled explosion, survive thousands of degrees and hundreds of bar, and do it again next week. For the systems that point all this energy in the right direction, see modern avionics. The rocket equation does not negotiate, and neither does the plumbing.

Sources

Comments