The Flyby Anomaly
UnsolvedSummary
Spacecraft on their way to the outer solar system sometimes swing past the Earth to steal a little of its orbital energy. It is a routine manoeuvre and the mathematics is very well understood.
Between 1990 and around 2005, several spacecraft came out of these flybys going very slightly faster, or very slightly slower, than they should have been. The discrepancies were tiny, a few millimetres per second, and the tracking is good enough to see them, and nobody could account for them.
Then it stopped.
The registry files this record as Unsolved, and notes that it is unsolved in a particular and slightly unsatisfying way. Nobody explained it. It simply stopped happening, and everybody moved on.
What is documented
The manoeuvre. A gravity assist. A spacecraft passes close to a planet, and the encounter changes its speed and direction relative to the Sun.
The anomalies. Several Earth flybys produced small unexplained changes in speed.
Galileo, in 1990, came out roughly four millimetres per second fast. NEAR, in 1998, came out about thirteen millimetres per second fast, which was the largest. Cassini, in 1999, came out about two millimetres per second slow. Rosetta, in 2005, came out about two millimetres per second fast.
These are extremely small numbers. They are also, in each case, larger than the uncertainty in the tracking, which is what made them a problem.
The formula. In 2008, John Anderson and colleagues published in Physical Review Letters an empirical relation which fitted the observed anomalies rather well. It related the size of the anomaly to the angles at which the spacecraft came in and went out.
It was a formula, not a mechanism. It described the effect and did not explain it.
And then nothing. Later flybys did not show it.
Rosetta's subsequent Earth flybys, in 2007 and 2009, produced no anomaly, or nothing that could be distinguished from zero. Juno's Earth flyby in 2013, carefully analysed, showed no anomaly of the predicted size.
Leading explanations
An unmodelled systematic effect in the earlier analyses. Leading, and unproven.
The trajectory of a spacecraft is not computed by a single equation. It is the output of an enormous model, which has to account for the Earth's gravity field in fine detail, for the atmosphere at altitude, for solar radiation pressure, for thermal radiation from the spacecraft's own instruments, for the motion of the tracking stations on a rotating and deforming Earth, and for relativistic corrections.
Any of those can go wrong at the millimetre-per-second level. Several of them have improved substantially since 1990.
The most natural reading of the evidence is that the earlier flybys were analysed with poorer models, and the anomalies were artefacts of that, and the later flybys did not show anything because by then the models were better.
But nobody has demonstrated it. No specific error has been identified and shown to produce the observed effects with the observed pattern. The Anderson formula's fit to the early data remains unexplained.
New physics. Proposed at the time, and never supported. The registry notes AR-0029, which is the obvious precedent: the Pioneer anomaly also looked like new physics for thirty years, was pursued seriously and properly, and turned out to be the spacecraft's own waste heat.
The flyby anomaly may be a cousin. Nobody has done for it what Turyshev did for Pioneer.
What the popular version gets wrong
"The flyby anomaly proves there is unknown physics." It proves that four spacecraft in the 1990s and early 2000s did not come out of their flybys where the models of the time said they would.
"It was explained." It was not. It stopped occurring, and the field's attention moved on, and those are not the same thing.
"It was disproved." Also not. The early data is still there and the discrepancies in it are still unaccounted for.
Current status
Unsolved, and probably permanently.
The registry keeps this record because of what it demonstrates about how science actually disposes of things, which is not always tidy.
The flyby anomaly was not solved. It was not refuted. It simply ceased to reproduce, and a phenomenon that stops reproducing stops being interesting, and the people who were working on it went and worked on something that was still happening.
That is a perfectly reasonable thing to do. It also means that somewhere in the archived tracking data of four spacecraft, from thirty years ago, there is a small unexplained number, and there probably always will be.
Sources
- Anderson, J. D., Campbell, J. K., Ekelund, J. E., Ellis, J. and Jordan, J. F. (2008). "Anomalous Orbital-Energy Changes Observed during Spacecraft Flybys of Earth." Physical Review Letters 100, 091102.
- Tracking data and post-flyby analyses for Galileo (1990, 1992), NEAR (1998), Cassini (1999), Rosetta (2005, 2007, 2009) and Juno (2013).
- Subsequent analyses reporting null results for the later flybys.
Last reviewed: July 2026. Records are provisional. Where the evidence changes, the entry changes. Found an error? Tell us.