The Antikythera Mechanism: The 2,000-Year-Old Machine That Shouldn't Have Existed

Around Easter 1900, Greek sponge divers waiting out a storm off the small island of Antikythera, between Crete and the Peloponnese, found a Roman-era wreck at about 45 metres. Over the next year divers brought up bronze and marble statues, glassware and pottery. They also brought up a lump of corroded bronze and rotted wood about the size of a shoebox, which went to the National Archaeological Museum in Athens with everything else. In 1902 the archaeologist Valerios Stais noticed a gear wheel in one of the broken pieces.

Illustration: the Antikythera mechanism's corroded bronze gears.

What survives today is 82 fragments. The ship itself sank around 70 to 60 BCE, carrying luxury goods from the eastern Mediterranean, probably towards Rome. The device it carried is now understood to be a hand-cranked astronomical calculator, and nothing else of its complexity survives from the ancient world.

Reading a machine you can't open

For most of the twentieth century the fragments were too corroded to read. The physicist and historian Derek de Solla Price spent two decades on them, and in 1974 published Gears from the Greeks, based partly on gamma-ray images taken in Athens. He argued that the gears modelled the motions of the sun and moon. Many scholars were sceptical, because no ancient text described gearing of that kind.

The decisive work came in 2005, when the Antikythera Mechanism Research Project brought an eight-tonne X-ray CT scanner to the museum. The scans showed gear teeth inside the corrosion and, just as important, thousands of tiny Greek letters on the inner surfaces, effectively a user's manual. Surface imaging by Hewlett-Packard picked up more text on the outer faces. Tony Freeth and his colleagues published the first full reading of the gear trains in Nature in 2006. They counted 30 surviving gears, the largest about 13 centimetres across.

A reconstruction; its dials tracked the Metonic and Saros cycles, among others. Photo: Moravec / Wikimedia Commons, CC BY-SA 4.0

What turning the handle showed

The mechanism sat in a wooden box about the size of a thick book, with a crank on one side. The front had a large dial with two rings, one marked with the twelve signs of the zodiac and one with the 365 days of the Egyptian calendar. Pointers showed the positions of the sun and moon, and a small half-silvered ball turned to show the phase of the moon.

The back held two spiral dials. The upper one followed the Metonic cycle: 235 lunar months add up almost exactly to 19 solar years, so the moon's phases return to the same calendar dates after that interval. The lower one followed the Saros cycle of 223 lunar months, about 18 years and 11 days, after which a similar sequence of eclipses repeats. Glyphs on the Saros dial marked the months when an eclipse could be expected. In 2008 the same team found that a subsidiary dial, labelled with the names of games including the Olympics at Olympia, tracked the four-year cycle of the major athletic festivals.

The month names on the Metonic dial are Corinthian. That points to a Corinthian colony, such as Syracuse in Sicily or one of the cities in northwestern Greece, as the place the calendar was set up for. It has fed speculation about a link with Archimedes, who lived in Syracuse and, according to Cicero, built a geared sphere showing the motions of the sun, moon and planets. There is no direct evidence that he had anything to do with this one. Rhodes, home of the astronomer Hipparchus and later of the philosopher Posidonius, is another candidate.

The front almost certainly showed the five planets visible to the naked eye as well, though the gears for them are lost. In 2021 Freeth's group at University College London proposed a full reconstruction of this planetary display, using the cycle periods inscribed on the front cover to work out gear trains that would fit in the space available.

When it was made, and what it doesn't tell us

Dating the mechanism is harder than dating the wreck. Freeth's 2006 paper put its construction between about 150 and 100 BCE, based on the style of the lettering. In 2014 Christián Carman and James Evans worked backwards from the eclipse predictions and argued that the Saros dial begins in 205 BCE. That would be the starting date of the dial's cycle, not necessarily the year the device was built.

Nothing quite like it appears again in the surviving record until the astronomical clocks of medieval Europe. Richard of Wallingford began his astronomical clock at St Albans around 1327, and Giovanni de' Dondi's astrarium was finished in Padua in 1364. That gap of well over a thousand years is the real puzzle. The quality of the gearing suggests a workshop with experience, not a single inspired attempt, and Cicero's description shows that geared models of the heavens were known in his day. If other mechanisms existed, they were almost certainly made of bronze, and bronze was usually melted down and reused. This one survived only because it sank.

The fragments are on display at the National Archaeological Museum in Athens.

Sources

1. T. Freeth, Y. Bitsakis, X. Moussas, J. H. Seiradakis et al., "Decoding the ancient Greek astronomical calculator known as the Antikythera Mechanism", Nature 444 (2006): 587–591. https://doi.org/10.1038/nature05357
2. T. Freeth, A. Jones, J. M. Steele and Y. Bitsakis, "Calendars with Olympiad display and eclipse prediction on the Antikythera Mechanism", Nature 454 (2008): 614–617. https://doi.org/10.1038/nature07130
3. T. Freeth, D. Higgon, A. Dacanalis, L. MacDonald et al., "A model of the cosmos in the ancient Greek Antikythera Mechanism", Scientific Reports 11 (2021): 5821. https://doi.org/10.1038/s41598-021-84310-w
4. C. C. Carman and J. Evans, "On the epoch of the Antikythera mechanism and its eclipse predictor", Archive for History of Exact Sciences 68 (2014): 693–774. https://doi.org/10.1007/s00407-014-0145-5
5. Derek de Solla Price, Gears from the Greeks: The Antikythera Mechanism, a Calendar Computer from ca. 80 B.C., Transactions of the American Philosophical Society 64, part 7 (1974). https://doi.org/10.70249/9780871693006