Watch the Moon night after night for a month and it will not behave like an object moving at a fixed rate. Some nights it slides further east against the stars; on others it seems to dawdle.
The geared bronze device recovered from a shipwreck discovered off the Greek island of Antikythera in 1900 accounted for exactly this irregularity; its makers found a way to build the wobble into metal.
The problem the builders were staring at
The Moon really does speed up and slow down in its orbit. As NASA explains, it moves along an elliptical orbit at a speed that changes with its distance from Earth, pulling ahead of or falling behind a fictitious Moon moving at a constant rate. Ancient astronomers represented this irregular motion as the lunar anomaly.
The builders therefore faced a mechanical problem. Ordinary concentric circular gears transmit fixed ratios: turn the input steadily and the output also turns steadily. But the lunar pointer needed to race ahead during part of its cycle and fall behind during another, repeating the pattern once each anomalistic month.
The clever bit of hardware
The solution, reconstructed from X-ray scans by Tony Freeth and his colleagues in a 2006 paper in Nature, used two gears known as k1 and k2. They were mounted on axes that were close together but not identical. A pin fixed to one gear engaged with a radial slot in the other.
As k1 turned, the pin forced k2 to follow. Because their axes were offset, however, the pin slid along the slot and altered the driven gear’s angular rate. The researchers’ supplementary notes say the arrangement “introduces a small quasi-sinusoidal variation in k2’s rotation rate.” They also show that the geometry reproduces Hipparchus’s lunar theory using methods that were, in principle, available in ancient Greece.
This remains a reconstruction from broken and corroded remains, not a photograph of the complete machine in operation. But the pin-and-slot device is visible in the scans, and the reconstructed gearing gives the variation the period of the anomalistic month. A 2025 preprint, citing the 2006 reconstruction, puts the intended lunar variation at about 6.53 degrees either side of the mean position.
Why an off-centre pin gives you a wave
Think about what the offset does over one full turn. When the geometry gives the pin greater leverage, a small movement of the driver sweeps the second gear through a larger angle. Half a turn later, the same input produces a smaller angular movement. The output moves ahead of a uniform rotation, falls behind it and catches up again.
Plot the difference between this output and a perfectly steady one and the curve rises, peaks, falls, dips and returns. It is not an exact modern sine generator, which is why the researchers call the motion quasi-sinusoidal, but the resemblance is real. Nobody in the workshop needed to write down a sine function. They embodied a periodic harmonic variation in bronze.
Where it actually sits in history
The mechanism was constructed around the end of the second century BC, when Greek astronomy already used sophisticated geometry. The historians J J O’Connor and E F Robertson note that Hipparchus produced the first known table of chords around 140 BC. For a circle of unit radius, a chord is directly related to a sine, so the mathematical world behind the mechanism was not a blank.
The first surviving tables that are recognisably sine tables came later. The same history credits Aryabhata, around AD 500, with tables of half-chords equivalent to sines. That places them roughly six centuries after the mechanism, not two millennia. Modern notation and the treatment of sine as a function came much later still, but the honest framing is that the device mechanically realised a quasi-sinusoidal variation before surviving sine tables, not before anyone possessed relevant mathematics.
Its engineering remains extraordinary without exaggeration. The 2006 authors describe it as “technically more complex than any known device for at least a millennium afterwards.” About thirty bronze gears survive, and nothing comparably sophisticated appears in the surviving record until medieval astronomical clocks.
Building a truth before you can write it
What keeps pulling me back to the mechanism is the order in which the knowing happened. Its designers worked from an astronomical theory of the Moon’s uneven motion and found a linkage that reproduced that motion. The formal language we now use to describe a sinusoidal curve did not yet exist, even though Greek chord geometry supplied closely related mathematics.
A machine can carry a piece of understanding that its makers could not have written in modern symbols. The Antikythera mechanism carried that understanding through two thousand years of silence, until X-rays exposed a pin sliding in a slot and researchers recognised the harmonic variation its builders had cut into bronze.
