The Eye of the Sahara and Atlantis

The structure lies in the north-west of the Taoudenni basin, in the middle of the Mauritanian Adrar plateaus, where it shows at the surface as a slightly elliptical depression about 40 km across, ringed by ridges⁠[19]. The rocks are a Late Proterozoic to Ordovician sequence of sandstone, limestone, dolomite and mudstone. They dip outward at 10 to 20 degrees, and erosion has cut the resistant quartzite beds into three nested cuestas⁠[1] that stand 20 to 30 m above the hollows between them⁠[5]. The plateau around the structure stands about 200 m above the desert sand, the outer rim reaches about 485 m above sea level, and sand is spreading onto its southern side⁠[14].

At the centre a shelf of limestone and dolomite encloses a siliceous breccia about 3 km across and at least 40 m thick. Basalt ring dikes, carbonatite dikes, kimberlite and alkaline volcanic rocks cut the dome and are not found outside it. The outer ring dike, about 50 m wide, lies 7 to 8 km from the centre; the inner, about 20 m wide, about 3 km⁠[1]. Two rhyolite centres are the eroded remains of maars, craters left by steam explosions⁠[5].

Geologists have argued over its origin since the 1950s⁠[19]. Space crews have photographed it since the Gemini missions⁠[1]; the picture at the head of this page was taken from the International Space Station on 17 December 2011⁠[2]. The International Union of Geological Sciences lists it as a geological heritage site, a concentric alkaline magmatic complex⁠[19]. The caravan town of Ouadane, one of four ksour founded in the 11th and 12th centuries for the trans-Saharan trade and on the World Heritage List since 1996, lies to the south-west⁠[20].

Théodore Monod found accumulations of pre-Acheulean and Acheulean tools in the depression of the outer ring, along Wadi Akerdil and Wadi Bamouéré, beside outcrops of Chinguetti quartzite that supplied the stone⁠[21]⁠[8]. Robert Vernet calls the Richât, east of Ouadane, probably the largest concentration of pebble tools in the Sahara. More than 100 pebble tools at Akerdil were drawn where they lay in 1982. In 1999 G. Prince found a site, named after Monod, with 113 fresh artefacts, mostly choppers and flakes, in a sandy palaeosol; a refit and the evenness of the knapping led him to judge it undisturbed and older than the early Acheulean⁠[7].

Missions under Henry de Lumley worked in the Adrar from 1999 to 2006. Vernet notes that they produced publications but no study of the material, no substantial stratigraphy and no absolute dates⁠[7]. Their sediment study dated the torrential alluvium of the outer ring: eleven AMS radiocarbon dates on shells and calcite cements fall between about 15,000 and 7,730 years BP, in the African Humid Period, and palaeomagnetism places most of the deposits in the present normal-polarity epoch, so older Pleistocene sediment is unlikely to survive on the slopes. The authors read the large tools as made on the spot, concentrated by wind erosion and then cemented by calcite, and expect more to lie buried in the outer alluvial plain⁠[8].

Later people left engravings. At Tililit, in the Richat, they are of several styles or periods: elephants, giraffes, ostriches, antelopes, cattle, armed men, horsemen, a stylised chariot and Libyco-Berber inscriptions. Chariots are engraved at Oued Tenzer on the eastern edge. None is dated⁠[22]. No walls, houses, pottery or engineering works have been reported from the structure⁠[15]⁠[23].

The origin of the structure is settled in outline. It is an eroded dome raised by an igneous intrusion; its rings are the edges of resistant beds, not built or dug features. Magnetic and gravity models show circular anomalies on the gabbro ring dykes above a deep mafic body and favour an intrusion over an impact⁠[24]. The carbonatite and the hydrothermal breccia are Cretaceous, about 99 and 98 million years old⁠[1]. What is not settled is the age of the gabbros. Matton and Jébrak read them as part of one complex, inferring 104 to 94 million years for the whole suite⁠[5]. Abdeina and colleagues obtained no robust age from them, but their modelling and chemistry point to sills of the Central Atlantic Magmatic Province, 230 to 200 million years old, domed about 100 million years later⁠[6].

Plato's Timaeus and Critias are the only independent source for Atlantis; every later ancient account found for this review goes back to him. In the Timaeus the island lies in the ocean in front of the Pillars of Heracles, larger than Libya and Asia together; its kings held Libya as far as Egypt; after a war with Athens it was swallowed by the sea in one grievous day and night, leaving a shoal of mud⁠[11]. In the Critias Poseidon cuts belts of sea and land around Cleito's hill, two of land and three of sea. Their widths are 3, 3, 2, 2 and 1 stades around an island 5 stades across, and a wall runs 50 stades beyond the largest ring. Around the city is a plain of 3,000 by 2,000 stades that faces south, is sheltered from the north by mountains, and is enclosed by a ditch 10,000 stades long. The island had a great many elephants⁠[10]. The translator gives the stade as about 600 feet, so the rings and island together, 27 stades, come to about 5 km⁠[10].

Readers in antiquity disagreed about whether it was true. Strabo praises Posidonius for allowing that the story may not be fiction, rather than saying that "its inventor caused it to disappear, just as did the Poet the wall of the Achaeans"⁠[25]; elsewhere Strabo gives that remark about the wall as Aristotle's, so read together the two passages make the sceptic Plato's own pupil⁠[26]. Proclus, writing in the fifth century AD, says that Crantor, the first commentator on Plato, took the story as plain history, and cites Marcellus's Ethiopic History for islands in the outer sea whose people remembered the Atlantic island⁠[27]. Plutarch has Solon hear the story from priests of Heliopolis and Saïs and leave his poem on it unfinished, and Plato die before finishing his own⁠[28]. Philo, Pliny, Tertullian, Arnobius and Ammianus repeat it as land or an island swallowed by the Atlantic; Philo and Tertullian measure it against Africa and Asia, and Pliny adds "if we are to believe Plato"⁠[29]⁠[30]⁠[31]⁠[32]⁠[33]. Aelian preserves a tale by Theopompus of a continent beyond the ocean and its people the Meropes, and calls its author "an egregious Romancer"⁠[34].

Two ancient texts tie names like Atlantis to north-west Africa, and neither is about Plato's island. Herodotus, a generation before Plato, puts a people called the Atlantes on the salt ridge of inner Libya, named after Mount Atlas, which the people there call "the pillar of heaven"; beyond the ridge, which runs to the Pillars of Heracles, Libya is desolate and waterless, without rain or forest⁠[35]. Diodorus, in the first century BC, tells of the Atlantians, the most civilised people of western Libya, living by the ocean near Mount Atlas, who were conquered by Amazons from the island Hespera in the marsh Tritonis; the marsh later vanished in an earthquake. His Loeb editor traces the story to a mythical romance by Dionysius Skytobrachion of the mid-second century BC⁠[36]. Where Lake Tritonis lay was disputed in antiquity: the Herodotean tradition points to the Gulf of Gabès, other writers to the Lesser Syrtis, between the Syrtes or Cyrenaica, and only Mela, and perhaps Ptolemy, to an inland lake such as the great Tunisian chotts⁠[37]. Pliny calls Mount Atlas "the most fabulous locality even in Africa"⁠[38].

No map of Atlantis survives from Greek or Roman antiquity; every ancient account is prose. The earliest drawn map of Atlantis found for this review is Athanasius Kircher's of 1665. It sets the island in the middle of the Atlantic between Africa and Spain on one side and America on the other, with south at the top⁠[18]. The nearest thing to a Greek map of the region is a modern reconstruction of the world Herodotus described. It marks the Atlantes, Mount Atlas and the ridge running to the Pillars, and has no Atlantis⁠[39].

Most archaeologists and classical philologists read Atlantis as Plato's invention, a story made to carry his argument about the ideal state⁠[40]. Alan Cameron calls it "virtually demonstrable" that Plato made the story up, and argues that neither Crantor nor Posidonius shows that ancient readers believed it⁠[41]. Christopher Gill reads it as the earliest narrative fiction in Greek literature⁠[42]⁠[43]; Pierre Vidal-Naquet treats it as a fiction made to say something about the Athenian state and traces how later writers reused it⁠[44]; Heinz-Günther Nesselrath's 2002 lecture is titled Plato and the Invention of Atlantis⁠[45].

The Sahara was not always desert. A dust record off Cap Blanc, Mauritania, shows the African Humid Period between about 14,800 and 5,500 years ago, when the Sahara was "nearly completely vegetated" and held "numerous perennial lakes". An arid interval matching the Younger Dryas interrupts it, and both its start and its end took only decades to centuries⁠[16]. A synthesis of records across Africa gives the same span and finds that the end, abrupt in places, came later the nearer a site lay to the equator⁠[46]. Leaf waxes in cores off the western Sahara put the Green Sahara at about 11,000 to 5000 years ago, with rain reaching 31°N and a drier pause about 8000 years ago⁠[47]. In the eastern Sahara the drying was gradual, with desert by about 2700 years ago⁠[48]. In wet phases rivers, lakes and inland deltas joined into networks across the desert⁠[49], and large rivers reached the Mauritanian coast⁠[50]. Such green phases have come and gone more than 230 times in 8 million years⁠[51]. In the Adrar itself, small diatom lakes stood on a wadi floodplain at Tamga between 9630 and 7550 radiocarbon years BP⁠[17], and the outer-ring sediments of the Richat, of 15,000 to 7,730 BP, are torrential stream alluvium⁠[8]. Vernet judges that the Sahara was probably empty of people during the last glacial maximum, between about 28,000 and 11,500 calibrated years BP⁠[7]. World sea level then stood as much as 134 m below today's and was still rising through the Younger Dryas and on to about 8,200 years ago⁠[52].

The Richat is the latest of several attempts to find Atlantis in Africa. In 1883 Étienne-Félix Berlioux identified it with the land of the Moroccan Atlas, putting its capital where the Sous valley once held water⁠[53]. After his West African expedition of 1908 to 1910, Leo Frobenius argued that Atlantis had a sphere of influence reaching North Africa and that "the greater Atlantis" could still be found⁠[54]. Pierre Benoit's novel L'Atlantide (1919) put a surviving Atlantis in the Ahaggar and gave its characters "the hypothesis of the Saharan sea" to argue over⁠[55]. In 1927 Paul Borchardt, after Butavand, placed Atlantis between the Chott el Djerid and the Gulf of Gabès and put the Pillars of Heracles at a temple of Hercules at Gabès; Albert Herrmann then identified the Chott el Djerid with both Lake Tritonis and Atlantis, dated it to the 13th or 14th century BC and cut Plato's figures by a factor of thirty⁠[56]⁠[57]. Sarantitis's Pillars at the Gulf of Gabès are in the same place as Borchardt's⁠[58].

The Atlantis reading began with George S. Alexander and Natalis Rosen, who noticed the likeness to Plato's city in 2006, visited the structure in 2008 and put a film of it online in 2010⁠[9]. George Sarantitis made the same identification at an Atlantis conference in 2008, moving Plato's Pillars of Heracles to the Gulf of Gabès and dating Atlantean wars to 9600 and 8600 BC; books by Carlos Bisceglia (2021) and David Edwards (2022) followed⁠[58]. The idea reached a mass audience in September 2018, when a video by Jimmy Corsetti's channel Bright Insight passed half a million views in two days and was picked up by the press⁠[12]⁠[13].

At its strongest the case runs like this. The Richat is a great set of concentric rings, and Plato's city was a set of rings around a central island. Add Plato's rings and island (27 stades) to the wall 50 stades beyond them on either side and the city is 127 stades across, about 23.5 km, which is the diameter of the Richat's main rings on Google Maps. There are uplands to the north and open ground to the south, as Plato's plain faced south and was sheltered from the north. The video dates the destruction to about 11,600 years ago, at the Younger Dryas⁠[12]⁠[23]⁠[13]. Rivers did cross western Mauritania in humid periods⁠[50], elephants are engraved in the Richat itself⁠[22], and Corsetti has argued that salt in the structure shows the sea once covered it⁠[59].

Each match has a reply from specialists. The rings are ridges of tilted rock exposed by erosion of a dome about 100 million years old, not dug channels of water⁠[1], and they are neither complete nor discrete⁠[23]. The structure is at least 40 km across and NASA gives 45 km, so 127 stadia matches only an inner ring and only by ignoring the outermost⁠[2]⁠[15]; Plato's rings and island alone would span about 5 km⁠[10]. His plain of 3,000 by 2,000 stades with its 10,000-stade ditch has no counterpart on the ground, and the ocean lies west of the Richat, not south⁠[10]⁠[23]. The Mauritanian rivers were natural drainage to the Atlantic coast⁠[50]. Elephants are no clue to place: Herodotus, a generation before Plato, already put them in western Libya⁠[60], and the Tililit engravings are undated and include horsemen and a chariot⁠[22].

The sinking is the hardest part to fit. Plato's island was swallowed by the sea⁠[11]; the Richat's rim stands about 485 m above sea level some 500 km inland⁠[14]⁠[2]. Twelve thousand years ago world sea level was lower than today, not higher⁠[52], and the dated sediments of the outer ring from that time are stream and wind deposits⁠[8]. The Younger Dryas lasted about 1,200 years and was felt mostly at high northern latitudes⁠[23]. Plato also puts Atlantis outside Libya, in the Atlantic beyond the Pillars of Heracles, with Libya among the lands it ruled⁠[11]. On the ground there are Palaeolithic stone tools and undated rock art, but no city, walls, pottery or canal⁠[8]⁠[15].

The climate argument is the strongest of the matches, and it is partly right. Plato's Egyptian priest tells Solon that Athens was founded 9,000 years before, which puts the war at about 9600 BC, some 11,600 years ago⁠[11]. That falls inside the African Humid Period as the Mauritanian dust record dates it⁠[16]⁠[46], and streams were laying down gravel in the Richat's outer ring across that span⁠[8]. The Richat was a wetter place within human memory. What the record shows, though, is streams and small lakes, not a sea or a ring lake. The dated deposits in the rings are torrential alluvium⁠[8]; the nearest dated lake beds found for this review are small floodplain lakes in the south-east Adrar⁠[17]; and Plato's rings were belts of sea joined to the ocean by a channel, not a lake fed by rivers⁠[10]. The Younger Dryas, to which the 2018 video ties the destruction⁠[23], is an arid break in the Mauritanian record⁠[16], and one leaf-wax record does not start the Green Sahara until about 11,000 years ago⁠[47]. The wet phase also went on for thousands of years after Plato's date, ending about 5,500 years ago⁠[46], and by Herodotus's day inner Libya was known as waterless⁠[35]. Green phases have recurred hundreds of times⁠[51], so a wet Sahara at Plato's date is not in itself a sign of a link.

Nor does any ancient writer reviewed here put Plato's Atlantis on the African mainland. Those who repeat the story set it in or under the Atlantic⁠[29]⁠[30]⁠[31]⁠[33], and those who took it as history looked for traces in the outer sea⁠[27]. The Atlantes of Herodotus are a people named after Mount Atlas⁠[35]. Diodorus's Atlantians come closest to a North African Atlantis, a civilised people by the ocean beside a marsh lost to an earthquake, but his source was a romance written two centuries after Plato, he never links them to Plato's island, and the ancients could not agree where Tritonis was⁠[36]⁠[37]. Kircher's map, the earliest found, puts Atlantis in the open ocean⁠[18].

Two readings of the geology itself remain in play. The impact-crater reading, based on the shape and a reported coesite, is ruled out⁠[4]⁠[3]. Within the igneous reading, the gabbro rings may belong to the Cretaceous complex⁠[5] or be older sills folded by it⁠[6]; an early mapping of them as subvertical dikes sits awkwardly with the sill model⁠[1].

For the geology, a robust radiometric age on the gabbros, from a method less disturbed than plagioclase 40Ar/39Ar, and field study of whether they cut the bedding or lie along it, would decide between one Cretaceous event and two events 100 million years apart⁠[6]⁠[5].

For the archaeology, the Richat needs what Vernet says the whole Mauritanian Palaeolithic lacks: an excavated, dated stratigraphy⁠[7]. A stratified dig at an undisturbed pebble-tool site such as Prince's, with absolute dates, would show whether the pebble tools are older than the Acheulean. Giresse and colleagues expect further assemblages buried in the outer alluvial plain⁠[8]. The engravings and any tumuli need recording and dating.

For Atlantis, the identification makes predictions that can be checked: built walls, harbour works or canals on the rings; occupation layers, hearths or burials of about 9600 BC; and marine sediments or shorelines of that age at the structure⁠[13]⁠[23]. None has been reported. A systematic survey of the rings and the centre, with test pits and a dated core through the inner depressions, would settle it on the ground.

For the wet Sahara, a dated core through the young sediments of the ring depressions, examined for diatoms, lake clays and freshwater shells, would show whether water ever stood in the rings or only ran through them, and when. The Tamga lakes show what such deposits look like in the Adrar⁠[17].