Antikythera mechanism

The Antikythera mechanism is the remains of an ancient Greek machine made of bronze gears inside a wooden case. By turning a small hand crank, the user could move a set of pointers forwards or backwards in time and read off astronomical information: where the Sun and Moon were in the zodiac, what the phase of the Moon was, which month it was in several calendars, and whether an eclipse was expected.

It is often called the oldest known analogue computer. That does not mean it worked like a modern electronic computer. It was “analogue” because it represented the cycles of the sky by the turning of wheels: the ratio between the numbers of teeth on its gears reproduced the ratios between astronomical periods. Once the gears were set, the calculation happened mechanically.

The object is named after the small Greek island of Antikythera, between Crete and the Peloponnese, off which it was found in a sunken ship. Only about a third of it survives, in 82 fragments, but modern imaging has allowed researchers to reconstruct much of how it worked.

Around Easter 1900, sponge divers from the island of Symi, led by Captain Dimitrios Kontos, stopped at Antikythera to wait for good winds. Diving to about 45 metres, one of them found a heap of bronze and marble statues on the seabed. With help from the Greek navy, divers recovered many objects between late 1900 and 1901: statues such as the bronze Youth of Antikythera, a bronze head known as “the Philosopher,” marble figures, pottery, glassware, jewellery and coins.

The mechanism came up in 1901 as a corroded lump of bronze and wood and was taken to the National Archaeological Museum in Athens with the other finds. Museum staff were busy reassembling the statues, and the lump was not treated; as it dried, the corroded bronze cracked and shrank. On 17 May 1902 the archaeologist Valerios Stais, with his cousin the politician Spyridon Stais, noticed a gear wheel embedded in one of the pieces.

The ship itself was a Roman-era cargo vessel that sank in the second quarter of the first century BCE. Coins found by Jacques Cousteau’s team in 1976 were minted between about 76 and 67 BCE, and the pottery points to the same period, so the wreck is usually dated to about 70–60 BCE. Where the ship came from and where it was going, perhaps towards Italy, remain uncertain.

The front of the mechanism had a large round dial with two rings. The inner ring was marked with the twelve signs of the zodiac, each taking up an equal 30-degree sector. The outer ring was a calendar with month names from the Egyptian calendar written in Greek letters; three of them, Pachon, Payni and Epiphi, survive.

Pointers moved around this dial to show the position of the Sun (which also marked the date) and of the Moon. The Moon pointer did not move at a steady speed: it sped up and slowed down, just as the real Moon does because of its elliptical orbit. A small ball, half light and half dark, turned to show the Moon’s phase.

Around the dials ran a parapegma, a star calendar listing the risings and settings of bright stars during the year, keyed to letters on the zodiac scale. Inscriptions on the front cover also describe the five planets known in antiquity, Mercury, Venus, Mars, Jupiter and Saturn, which strongly suggests that the front once showed their positions too, though the gears for them are lost.

The back of the mechanism had two large spiral dials and several smaller ones. The upper spiral, of five turns, showed the Metonic cycle: 235 lunar months that fit almost exactly into 19 solar years. This let people keep a calendar based on the Moon in step with the seasons. The month names on it turned out to belong to a Corinthian-type calendar.

The lower spiral, of four turns, showed the Saros cycle of 223 lunar months, about 18 years and 11 days, after which eclipses repeat in a similar pattern. Small symbols (glyphs) in the cells marked months when a solar or lunar eclipse could be expected, with the time of day. A small dial beside it, the exeligmos dial, showed how many hours to add, since each Saros is about eight hours longer than a whole number of days.

Another small dial followed a four-year cycle of Panhellenic athletic festivals: the Olympic, Pythian, Isthmian and Nemean games, as well as the lesser Naa games of Dodona and the Halieia of Rhodes. It is the only pointer that turned anticlockwise. Its presence shows that the mechanism was about human time, festivals and calendars, not only astronomy.

The crank turned a large four-spoked gear (known to researchers as b1) once per year. From it, trains of smaller gears drove the other pointers at speeds set by the ratios of their teeth. For example, the gear train for the Metonic dial gives a period of about 6,939.5 days for the full spiral, very close to the modern value of 6,939.69 days; the Saros train gives about 1,646.3 days per turn of the pointer against a modern 1,646.33.

The most remarkable feature is the lunar gearing. Two gears are mounted one on top of the other on a turntable; one has a pin that engages a slot in the other. Because the two axes are slightly offset, the output speeds up and slows down once per lunar month. This pin-and-slot device mechanically reproduces the theory of lunar motion associated with the astronomer Hipparchus, and the turntable itself rotates slowly to model the gradual shift of the Moon’s orbit, a cycle of about 8.88 years. It is the earliest known use of epicyclic gearing.

The gear teeth are small equilateral triangles cut by hand, with an average spacing of about 1.6 millimetres, on wheels about 1.4 millimetres thick. Their irregularity shows they were made with hand tools. Such hand-made gearing would have introduced play and friction that limited the precision of the finest corrections, but the design itself is exceptionally sophisticated.

The mechanism combines two astronomical traditions. From Babylonian astronomy came long period relations, whole-number cycles such as the Saros and the relations used for the planets, discovered over centuries of observation recorded on clay tablets. The 2006 study showed that the eclipse predictions were based on Babylonian arithmetic cycles.

From Greek astronomy came geometrical models that explained irregular motions with circles and epicycles, such as the lunar theory of Hipparchus. The mechanism turned these models into moving parts, making a physical model of a Greek cosmos with the Earth at the centre.

In 2016 researchers read numbers on the front cover inscription giving cycles for Venus (462 years) and Saturn (442 years). The 2021 reconstruction by Tony Freeth’s team at University College London argued that these periods could have been derived with a method of combining period relations, which they linked to mathematics associated with Plato’s Academy, and used them to design compact gear trains for all five planets.

No maker’s name survives. The inscriptions are in Greek, the astronomy is Hellenistic, and the workmanship suggests a practised tradition rather than a single lucky invention: the mechanism is too refined to be the first of its kind, so there must have been predecessors that are now lost.

Several places have been proposed. Rhodes was a centre of astronomy and engineering, home to Hipparchus and, later, to the Stoic philosopher Posidonius, whom Cicero credits with building a similar device; the Games dial includes the Rhodian Halieia, and the star calendar fits latitudes around Rhodes. The Corinthian-type calendar and the mention of the Naa games at Dodona point to a user in northwestern Greece, perhaps in a Corinthian colony in Epirus. The 2008 study also raised the possibility of Syracuse, Archimedes’ city, but later work showed the calendar cannot be Syracuse’s. A compromise view suggests a design tradition from Rhodes, with this particular example made for a client in Epirus.

Dates also vary. Derek de Solla Price proposed about 87 BCE; others argue for 150–100 BCE; a 2014 study placed the start date of the eclipse dial in 205 BCE, and a 2022 proposal put the initial calibration at 178 BCE. A calibration date is not necessarily a construction date. Everyone agrees only that it was made before the ship sank, about 70–60 BCE.

Ancient writers mention devices that showed the motions of the heavens. In his dialogue On the Republic (54–51 BCE), Cicero describes two sphere-like models made by Archimedes that the Roman general Marcellus brought back after the siege of Syracuse in 212 BCE, one of which could show the Sun, Moon and planets. Cicero also says that his friend Posidonius had recently built a device whose revolutions reproduced the daily motions of the Sun, Moon and five planets.

Pappus of Alexandria, writing centuries later, says Archimedes wrote a lost book On Sphere-Making. Later Latin and Greek authors, from Lactantius to Proclus, continued to mention Archimedes’ sphere. These texts show that such instruments were known and admired, although the Antikythera mechanism itself cannot be identified with any of them.

Among surviving objects, machines of similar complexity appear again only about 1,500 years later, in the fourteenth-century European astronomical clocks of Richard of Wallingford and Giovanni de’ Dondi. Whether any continuous tradition of geared devices linked antiquity to these clocks is uncertain, since the intervening evidence is sparse.

After Stais’s identification, many scholars doubted that the object could be ancient, because it seemed far too complex. The German philologist Albert Rehm was the first to argue clearly that it was an astronomical calculator. Serious study resumed when the historian of science Derek J. de Solla Price took up the problem in 1951; with the Greek physicist Charalampos Karakalos he made X-ray and gamma-ray images in 1971, and in 1974 he published Gears from the Greeks, proposing a first overall reconstruction.

In 2002 the mechanical engineer Michael Wright built the first working model incorporating the known gears and his own proposed planetary display; he also discovered the Moon-phase mechanism. From 2005 the Antikythera Mechanism Research Project, with a Cardiff University team led by Mike Edmunds working alongside Greek colleagues, used high-resolution surface imaging and microfocus X-ray computed tomography to see inside the fragments. Their results, published in Nature in 2006 and 2008, doubled the number of deciphered inscriptions and clarified the back dials.

Later work has focused on the lost front gearing. In 2021 the UCL Antikythera Research Team published a complete proposed model of the planetary display, arguing that earlier reconstructions did not fit all the evidence. Other researchers continue to test alternatives, for example about the number of divisions on the front calendar ring.

The mechanism changed our picture of ancient technology. Before its decoding, historians assumed that fine geared machinery began with medieval clockwork. It shows that Hellenistic craftsmen could miniaturise complex gear trains and that astronomers wanted to make their theories visible and usable.

It also shows how ancient science worked across cultures: Babylonian numerical astronomy, Greek geometrical models and Egyptian and Greek calendars were brought together in a single object. Its games dial and calendars remind us that astronomy served social life, from fixing festivals and religious dates to, possibly, astrology and teaching.

Popular culture sometimes treats the mechanism as a mystery that needs aliens, time travel or lost super-civilisations to explain it. Scholars see it instead as the product of a known intellectual world whose other examples happen not to have survived; bronze was usually melted down and reused, so a shipwreck preserved what ordinary history destroyed. It has been featured in documentaries and, in fictionalised form, in the film Indiana Jones and the Dial of Destiny (2023).

The back dials are now well understood, but the front remains contested because most of its gearing is missing. The space between the main gear and the front plate, features on the four-spoked wheel and the inscriptions all imply further gears. Wright’s 2002 model added a “true Sun” pointer and pointers for all five planets. Evans, Carman and Thorndike (2010) instead interpreted irregular spacing of the zodiac scale as evidence for an eccentric solar display and suggested simpler dials for the planets. Freeth and Jones (2012) proposed compact planetary trains using pin-and-slot devices similar to the lunar mechanism; the 2021 UCL model, with a total of dozens of gears and eight coaxial outputs, derives the gear counts from period relations with small prime factors, sharing gears between planets.

Reconstructions must satisfy tight physical constraints: the thickness of the surviving plates, the positions of bearings and posts, and the texts of the back-cover “user manual,” which describes a display in rings with planets as “little spheres” and the Sun as a “little golden sphere.” Critics note that any full model involves assumptions about lost parts, so none can be proved; the models are best seen as hypotheses that fit the evidence more or less well.

The front calendar ring is another case. It was long assumed to have 365 divisions for the Egyptian solar year, but statistical analyses of the surviving holes beneath the ring (Budiselic and colleagues in 2020, followed by independent studies) suggest about 354 intervals, pointing to a lunar calendar. If correct, this changes how the front dial was used, although the Egyptian month names remain puzzling.

Dating the mechanism combines several kinds of evidence. The wreck gives a terminus ante quem: its coins (minted about 76–67 BCE) and pottery place the sinking around 70–60 BCE, while a radiocarbon date from hull timber (220 BCE ± 43 years) is thought to reflect old wood. Astronomical dating tries to find the moment at which the dials were calibrated: Carman and Evans argued in 2014 that the Saros dial’s eclipse pattern best fits a start in 205 BCE; others prefer 204 or 178 BCE. Such dates identify an epoch built into the design, which could have been copied from an older model.

Provenance arguments depend on calendars and festivals. The 2008 Nature paper identified all twelve month names of the Metonic calendar as Corinthian, which pointed to Corinthian colonies in northwestern Greece or Syracuse. Paul Iversen’s 2017 study of the calendar and the games dial argued that the prototype came from Rhodes but that this example was adapted for a client in Epirus, using an Epirote calendar, and was made no earlier than a generation before the wreck, a view supported by Alexander Jones in the same year.

Methodologically, the mechanism is a showcase for collaboration between archaeology, epigraphy, history of astronomy, engineering and imaging science. Much of the text was invisible until X-ray CT allowed researchers to read letters millimetres high inside corroded layers. At the same time, the history of research is a caution: confident reconstructions have repeatedly been revised when new readings appeared, and some fragments may still await identification in museum storage, as Fragment F did until 2005.