Antikythera mechanism

• 1. Introduction

  ›

  The Antikythera mechanism sat unrecognized in an Athens museum for two years before anyone understood what it was. It had come up from the sea floor in 1901, a corroded lump of bronze and cracked wood. Sponge divers had salvaged it from a shipwreck off the Greek island of Antikythera. Museum staff were busy with far more eye catching finds pulled from the same wreck.

  Then in 1902, Greek politician Spyridon Stais noticed something odd in one of the rock like fragments. A gear wheel was embedded inside it. That single wheel belonged to what is now considered the oldest known analogue computer. The device could predict the positions of the Sun and Moon and forecast eclipses decades in advance. It could even track the four year cycle of Greek athletic games, the ancient Olympiad.

  How did people working more than two thousand years ago build something this precise? Who designed it, and why did machines of similar complexity not appear again until the fourteenth century?

• 2. Salvage And Skepticism

  ›

  Captain Dimitrios Kontos and a crew of sponge divers from the island of Symi found the wreck in early 1900. They worked with the Hellenic Royal Navy through 1901. The wreck lay 45 metres deep off Point Glyphadia, on Antikythera. Divers raised bronze and marble statues, pottery, unique glassware, jewellery, and coins. No one knows how the mechanism ended up aboard that cargo ship.

  Once ashore, the bronze had already corroded into atacamite, which cracked and shrank the fragments out of their original shape. Archaeologist Valerios Stais first guessed the gear bearing fragment was an astronomical clock. Most scholars at the time judged the device too advanced for its era, calling it prochronistic. German philologist Albert Rehm disagreed and argued it was an astronomical calculator.

  Interest in the object then stalled for decades. It was not revived until 1951, when Yale historian Derek J. de Solla Price took up the case. In 1971, Price and Greek physicist Charalampos Karakalos produced X-ray and gamma-ray images of all 82 fragments. Price published his findings in 1974.

  Further dives at the wreck site in 2012 and 2015 turned up more art objects and a second, possibly related ship. Those searches also recovered a bronze disc stamped with a bull, its four ears pierced with holes. Some once suspected it was a cog from the mechanism itself. Little evidence supports that; it was more likely a decorative fitting from a piece of furniture.

• 3. Tracing Its Makers

  ›

  Scholars generally date the mechanism's construction to the late second or early first century BC. Proposed dates for the finished device range from 205 BC to about 87 BC. It must have been built before the shipwreck, which is dated to roughly 70-60 BC. In 2022, one team proposed the device was first calibrated on the 23rd of December 178 BC, while other researchers argue 204 BC fits the evidence better.

  One theory traces its origin to Corinth or one of its colonies in northwest Greece or Sicily, based on the calendar carved into the Metonic Spiral. Syracuse was a Corinthian colony and the home of Archimedes, which raised the possibility of a connection to his school of mathematics. A 2017 study confirmed the calendar is of the Corinthian type but ruled out Syracuse specifically.

  A separate theory points to Pergamon, home of the Library of Pergamum, second in importance only to the Library of Alexandria. That theory rests on coins that Jacques Cousteau recovered from the wreck site in the 1970s.

  A third theory points to Rhodes, a busy trading port and center of astronomy, based on Rhodian style vases found aboard the ship. Rhodes was home to the astronomer Hipparchus, active from about 140-120 BC, and the mechanism's lunar theory matches his work. Researchers have even argued the astronomical events on its parapegma fit best for latitudes of 33.3 to 37.0 degrees north, a band that includes Rhodes.

  A 2014 study proposed a start-up date around 200 BC, tying the Saros Dial to the new moon of the 28th of April 205 BC. A 2017 study by researcher Iversen suggested the prototype came from Rhodes. This particular unit, he argued, was modified for a client from Epirus in northwestern Greece. He dated its construction to no more than a generation before the shipwreck, a conclusion researcher Jones supported the same year.

• 4. Eighty-Two Fragments

  ›

  The mechanism survived as a single encrusted lump before it fractured into three main pieces. Conservation work has since split those into 82 separate fragments. Only seven of them are considered mechanically significant, holding most of the surviving gears and inscriptions. Four fragments contain gears at all; the rest carry only text.

  The largest surviving gear measures about 13 centimetres across and originally carried 223 teeth. Fragment A alone holds 27 gears, including the large four spoked b1 gear visible on its front, plus the socket where the crank once turned. Fragment B preserves part of the Metonic spiral; only 49 of its 235 calendar cells survive there in whole or in part. Fragment D is stranger still, a small unidentified gear that some researchers believe once tracked the position of Jupiter. Fragment F was not found until 2005, and still carries traces of the mechanism's wooden housing.

  In that same year, a Cardiff University team led by Mike Edmunds used computer X-ray tomography to look inside the crust covered fragments. Their scans revealed 37 meshing bronze gears in total, enough to track the Moon and Sun through the zodiac and to model the Moon's uneven orbit. All these pieces are now kept at the National Archaeological Museum in Athens, alongside reconstructions built to show how the gears once meshed together.

• 5. Reading The Dials

  ›

  A single hand crank, now lost, once drove the entire mechanism through a crown gear connected to its largest wheel. One full turn of that crank advanced the date pointer by about 78 days. Turning it set every interlocking gear in motion at once, computing the Sun and Moon's positions, the Moon's phase, eclipses, and multiple calendar cycles simultaneously.

  The front dial carried two rings: a fixed inner scale marking the zodiac in 30 degree sectors, and a movable outer ring for the Egyptian calendar. That Egyptian calendar ran twelve 30 day months plus five extra days, but ignored leap years entirely. Left uncorrected, it drifted a full zodiac sign out of step roughly every 120 years.

  For over a century, researchers assumed the front dial's outer ring of holes represented a 365 day solar calendar. A 2020 study by Budiselic and colleagues challenged that, finding statistical evidence for 354 holes instead, pointing to a lunar calendar. Researchers Woan and Bayley independently calculated 354 to 355 holes and concluded that '365 holes is not plausible.' Researchers Malin and Dickens put the figure at 352.3, plus or minus 1.5. That makes the odds of exactly 365 holes less than one in ten thousand.

  On the back, five separate dials tracked longer cycles. The Metonic dial followed 235 lunar months across five spiral turns, a span nearly matching 19 solar years. That let the mechanism reconcile lunar and solar timekeeping. A smaller dial, once called the Olympiad dial, was later renamed the Games dial after researchers found it did not actually track Olympiad years at all. Its closest four year match was a different competition cycle, the Halieiad, and its pointer was the only one on the whole mechanism that moved anticlockwise.

  The Saros dial, the largest on the back, spanned 223 lunar months and could forecast eclipses down to the day and hour. Glyphs packed into 51 of its 223 monthly cells marked 38 lunar eclipses and 27 solar ones. A smaller Exeligmos dial then corrected for the extra eight hours built into each Saros cycle, so eclipse timing could be read precisely.

  A tiny ball on the Moon pointer, painted half white and half black, showed the Moon's phase as it turned. That small ball needed a differential gear, an arrangement summing two separate rotations into one, a solution no other known device of its age attempted.

• 6. Rebuilding And Doubt

  ›

  In 2002, mechanical engineer Michael Wright built the first working physical model based on the known gearing. He added pointers for Mercury, Venus, Mars, Jupiter, Saturn, and a hypothetical 'true sun' indicator alongside the known mean sun and lunar pointers.

  In 2010, researchers Evans, Carman, and Thorndike proposed something simpler: individual dials for each planet rather than full mechanical tracking. They pointed to uneven spacing in the front dial's inscriptions as evidence for an off-centre sun indicator. That would remove the need to simulate the solar anomaly mechanically. Freeth and Jones published their own compact solution in 2012, though theirs exists only as a 3D computer model, never built physically.

  In March 2021, the Antikythera Research Team at University College London, led by Tony Freeth, published a fresh reconstruction of the entire device. They found gears that could be shared across different planetary trains by using small prime factors, 7 and 17, in their calculations. The team concluded that none of the earlier models were compatible with all the currently known data.

  In 2025, another research team reached a starker conclusion: manufacturing error in the original gears was too great for the mechanism to have worked at all. They cautioned that the scans underlying that judgement could themselves be wrong about the extent of the imperfections.

  Freeth and Jones separately tested their simulated mechanism's accuracy and found the Mars pointer could be off by as much as 38 degrees at certain points. That error came from the limits of Greek planetary theory itself, not from mistakes in the gear ratios. No one could fix it until Ptolemy's Almagest, published around 160 AD, introduced a refinement called the equant. Even that would not be fully resolved until Johannes Kepler's laws of planetary motion arrived in 1609 and 1619.

  The gear teeth themselves were hand cut as triangles from bronze blanks, and not all of them are even. That friction likely blurred many of the finer corrections its design intended, however sophisticated the theory behind it.

• 7. Other Ancient Machines

  ›

  Cicero's De re publica, written between 54 and 51 BC, describes two similar devices, both built by Archimedes. The Roman general Marcus Claudius Marcellus brought them to Rome after Archimedes died at the siege of Syracuse in 212 BC. Marcellus kept one himself and placed the other in the Temple of Virtue.

  In Cicero's dialogue, a character recalls Gaius Sulpicius Gallus demonstrating the device around 129 BC. Gallus explained how a single revolution reproduced the unequal motions of the Sun, Moon, and five planets. Cicero's narrator, Philus, marvels that Archimedes 'must have possessed a genius superior to any thing we usually conceive to belong to our nature.' Cicero also mentions a third device, built more recently by his friend, the philosopher Posidonius. The Greek writer Pappus of Alexandria later credited Archimedes with a lost manuscript on the subject, titled On Sphere-Making.

  Such machines were not confined to Greece and Rome. A bronze combination lock found in Athens' Kerameikos district, from the Augustan or Hadrianic period, worked on a form of mechanical logic. It blocked its bolt from retracting until two independent rotary dials lined up correctly.

  In the ninth century, the Banu Musa brothers compiled the Book of Ingenious Devices for the Caliph of Baghdad, describing more than a hundred mechanisms. Around 1000, the scientist al-Biruni described a similar geared calendar, and a surviving thirteenth century astrolabe contains a comparable clockwork device.

  In the eleventh century, Chinese polymath Su Song built a mechanical clock tower with a rotating armillary sphere that tracked stars and planets. It arose from an entirely separate engineering tradition, thousands of miles from the Aegean.

• 8. From Museum To Movie Screen

  ›

  Reconstructions of the mechanism now sit in museums scattered across the world. The American Computer Museum in Bozeman, Montana, the Children's Museum of Manhattan, and the Musee des Arts et Metiers in Paris each display their own. So do museums in Kassel, Germany, in Olympia, Greece, in Athens, and at the Western Australian Museum.

  In 2010, hobbyist Andy Carol built a functioning replica entirely out of Lego, later featured in a short film in 2011. The YouTube channel Clickspring has documented building a replica using only tools, machining techniques, and materials available in ancient Greece.

  Television has told the story too. National Geographic aired 'Star Clock BC' on the 20th of January 2011, and BBC Four followed with The Two-Thousand-Year-Old Computer in 2012. American audiences saw the same program on NOVA on the 3rd of April 2013, retitled Ancient Computer. Filmmaker Tony Freeth produced his own documentary, The World's First Computer, in 2012. On the 17th of May 2017, Google marked the discovery's 115th anniversary with a Google Doodle.

  The 2023 film Indiana Jones and the Dial of Destiny fictionalizes the mechanism entirely, calling it Archimedes' Dial. In the story, Archimedes built it as a temporal mapping system, one a former Nazi scientist sought to obtain.