Fun With Science / Globe Deconstruction / Rockets in Vacuum / Thrust, Measured in Flight
Rockets firing in real vacuum, audited by people who had no stake in this argument.
Claim #2 says a rocket cannot produce thrust in vacuum because thrust needs something to push against. The main answer makes the physics case, and its §9 gathers corroboration at three scales — but all of it except the missile intercepts is measured on the ground. This page closes that gap. Engines have been fired in actual space, and their thrust measured there, many times, by engineers doing ordinary engineering. The useful part is that several of those measurements were made in physically unrelated ways at the same time, so they can be checked against each other rather than taken on trust. None of the work below was done to settle a debate about whether rockets work. It was done because mission planners needed the number.
Audited repeatedlyBudgets close
What the flight record shows
In 1993 NASA published a cross-check of four independent ways of measuring one ion engine's thrust in orbit — an onboard accelerometer, the change in the spacecraft's orbit, the angular momentum of its reaction wheels, and electrical telemetry calibrated on a ground thrust stand. They agree to within a few per cent. Deep Space 1's thrust was measured from the Doppler shift of its radio signal to better than 0.5 mN and plotted directly against ground-chamber data. Dawn's published propellant consumption and its measured velocity change close via the rocket equation to about 3%. There is a real vacuum-chamber artefact in all this, and it is measured too: roughly one per cent, in the direction that makes chambers flatter engines slightly.
A vacuum chamber can always be accused of not being empty enough. It is a fair worry in principle — and §11 of the main page concedes that chamber sizing is exactly where an amateur version of this test goes wrong. Flight removes the objection rather than arguing with it: an engine on an interplanetary trajectory is operating in a vacuum many orders of magnitude harder than any chamber, with no wall anywhere near it, and it either produces the thrust its designers expected or the spacecraft does not arrive.
The second thing flight gives you is independent instrumentation. On the ground a thrust stand is one instrument, and if you distrust it you are stuck. In orbit the same firing can be measured by several unrelated physical routes at once. That is the material below.
The single most useful document here is a NASA technical memorandum from 1993, written to compare the ways in-flight thrust had been determined on ion-propulsion flights up to that point. Its purpose was housekeeping — engineers wanting to know which method to trust for future missions. It has nothing to say about anybody's cosmology, which is precisely what makes it worth reading.
SERT II (launched 1970, a 1,524 kg spacecraft in a roughly 1,000 km orbit) carried a mercury ion thruster. Its thrust was determined three ways in flight, and a fourth on the ground:
| how it was measured | thrust | what it depends on |
|---|---|---|
| Electrical telemetry, corrected by a ground thrust stand | 26.8 mN | beam current and voltage; a pendulum calibration in a chamber |
| Onboard accelerometer (MESA) | 27.4 mN | a mass on a spring, in space |
| Change in the spacecraft's orbit | 28 mN | Newtonian gravity and 34 days of tracking |
Sovey & Rawlin, NASA TM-106283 (1993). The paper reports the orbit-derived figure as 3.7–4.5% higher than the telemetry-plus-ground-stand value, and the accelerometer as within 3% of it.
The orbital route is the one worth dwelling on, because it uses nothing but geometry and gravity. The spacecraft's altitude rose 20 km over 34 days of thrusting from an initial orbit radius of 7,400 km. For a continuously thrusting satellite the rate of change of orbit size is fixed by the tangential force, the mass and the gravitational parameter — so the thrust follows from the orbit change alone. Working it independently from those published figures gives a tangential component of 5.1 mN, against the paper's 4.9 mN; with the thrust vector 10° off, that is a total of about 30 mN against their 28. The residual is rounding in the inputs. The point is not the third decimal place — it is that a reader can take three published numbers and recover the thrust of a rocket firing in space without trusting any thrust measurement at all.
ETS-III, a 382 kg Japanese satellite with a 5 cm ion thruster, gives the same picture on different hardware built by a different agency: electrical telemetry plus ground calibration gave 1.76 and 1.87 mN for its two thrusters, reaction-wheel diagnostics gave 1.92 and 1.99 mN, and raising the orbit 5.7 km over 110 hours of thrusting gave 2.05 mN — about 9.6% above the electrical figure.
Modern deep-space missions do not carry thrust sensors at all. Dawn's chief engineer put it plainly: “Because the thrust is so gentle, there are no sensors on board that directly measure it.” What is measured instead is the frequency of the spacecraft's radio carrier, which shifts with its velocity. Velocity change is read off the Doppler shift; thrust follows from that and the vehicle's mass.
This matters for the argument, because it means the thrust figure is not a manufacturer's specification being reported back. It is a physical measurement of the spacecraft's motion, taken from Earth, using an effect anyone can verify on a passing ambulance. On Deep Space 1 the JPL team stated the precision directly: Doppler residuals gave the thrust “with an uncertainty of less than 0.5 mN” on an engine producing 20–92 mN. Dawn's navigation could detect velocity changes “smaller than 0.5 millimetres per second.”
Dawn is the cleanest worked case, because the two halves of the sum were published separately by the same team and can be put together by anyone.
Put those in the rocket equation — Δv = ve ln(m0/m1) — and 71.7 kg thrown from 1,215 kg at an exhaust velocity of 3,100 × 9.81 m/s predicts 1.85 km/s. The measured figure was 1.80. The budget closes to within 3%, with the residual sitting comfortably inside the rounding on every input.
The same sum over the whole mission is consistent rather than precise, and is worth stating that way: 401 kg of xenon and 11.14 km/s of measured Δv from a 1,218 kg spacecraft implies a mission-average specific impulse of about 2,845 s, which is where it should sit for a mission that ran near full power early and heavily throttled later at Ceres. Dawn's own propellant bookkeeping was itself cross-checked two ways — integrating flow-orifice pressure against tank pressure and temperature — and the two agreed “to within approximately 5 kg, or approximately one percent of the total xenon used.”
A sceptic's best line against all the ground-based evidence is that a vacuum chamber still contains gas, and the engine is quietly pushing against it. That is not a silly worry, and the propulsion literature agrees with it — which is why the same 1993 memorandum quantifies it rather than dismissing it.
Residual chamber gas is drawn into a thruster and adds to its propellant flow. The measured size of that effect: about 1% of the input flow rate for a 5 kW xenon thruster tested at 1.3 mPa, which is why the paper recommends test pressures below 2 mPa for flight qualification. So the artefact exists, it has been measured, and it runs in the direction a sceptic would expect — the chamber slightly helps.
The flight data shows the same sign. Deep Space 1's engine, measured by Doppler in interplanetary space, came out marginally worse than the ground chamber had predicted: specific impulse about 60 s low at the start of the mission, efficiency 2 to 2.5 percentage points low, thrust up to 1.6 mN below the calculated value at intermediate power. The JPL team plotted flight against ground thrust-balance data in the same figure and reported that they “agree well… at low powers, but are lower at intermediate powers.”
The claim's structure is that mainstream physics has misapplied Newton's third law and that this was “abused to sell humanity on the idea of space travel.” That framing needs the error to be undetected. But the propulsion literature is not a set of confident assertions — it is a long record of engineers measuring the same quantity several ways because they did not trust any single method, publishing the disagreements between those methods to the nearest per cent, and quantifying the one artefact that would flatter their own results.
An engine whose thrust depended on ambient gas would have failed this audit immediately, in 1970, on SERT II, where the orbit itself was the measuring instrument. It did not. Nor is there a version of the claim that survives the sign of the facility effect: if a chamber's residual gas helped, then flight in harder vacuum should be worse — and it is, by about two per cent, which is what the papers report and is nowhere near enough for the claim.