The Archimedes Palimpsest

A Byzantine prayer book of 174 parchment folios, finished by April 1229 and probably made in Jerusalem, written on parchment recycled from at least five older books. One of them was a tenth-century copy of Archimedes. Three of the treatises underneath exist in no other copy anywhere: On Floating Bodies in Greek, The Method of Mechanical Theorems, and the Stomachion. Ten further folios carry the orator Hyperides, previously known only from papyrus scraps.

It was catalogued in Constantinople in 1899 as an unremarkable service book, recognised by Johan Ludvig Heiberg in 1906, photographed by him, and then lost for most of a century. In the meantime a leaf had already been removed by Constantine Tischendorf and sold to Cambridge, four forged Evangelist portraits had been painted over the text to raise the price, and three folios had been cut out. It was sold at Christie's New York at 2pm on 29 October 1998 for a hammer price of USD 2,000,000, USD 2.2 million with the premium, over an injunction sought by the Greek government and the Greek Orthodox Patriarchate of Jerusalem. The underbidder was the Greek consul general, at USD 1.9 million.

The anonymous buyer put it on deposit at the Walters Art Museum, paid for a decade of conservation and imaging including X-ray fluorescence work at a synchrotron, and on the tenth anniversary of the sale released the entire data set under a Creative Commons Attribution licence. In March 2026 one of the three missing folios was identified in a museum store at Blois, where it had sat unrecognised since 1973.

It is a prayer book, and calling it the Archimedes Palimpsest is a little confusing

The object is a medieval parchment manuscript of 174 folios which is, as it now stands, a Byzantine prayer book written in Greek, technically a euchologion. It was completed by April 1229 and was probably made in Jerusalem. The project that has studied it for a quarter of a century says in its own first sentence that calling it the Archimedes Palimpsest is a little confusing, and that is the correct place to start.

To make the prayer book, the scribes took parchment from books that already existed, washed and scraped off the writing, cut the leaves, turned them ninety degrees and folded them into new quires. One of the books they cannibalised contained at least seven treatises by Archimedes, copied in the second half of the tenth century, almost certainly at Constantinople. The prayer book is therefore both the destroyer of the older text and the sole reason it survived at all: nobody kept tenth-century mathematics for its own sake, and everybody kept a service book.

The economics are worth stating plainly. Parchment was expensive. The scholar who identified the manuscript's lost folio 123 puts the material cost of a codex of this size at the skins of about fifty animals. Recycling was not vandalism; it was accountancy.

Three of the treatises underneath exist nowhere else on earth

The Archimedes undertext carries The Equilibrium of Planes, Spiral Lines, The Measurement of the Circle, Sphere and Cylinder, On Floating Bodies, The Method of Mechanical Theorems and the Stomachion. Of these, three matter beyond measure. There is no other surviving copy of On Floating Bodies in Greek, the language Archimedes wrote in. There is no version in any language, anywhere, of The Method of Mechanical Theorems or of the Stomachion.

The Method is the text in which Archimedes explains how he found his results before he proved them, which is the one thing the rest of the ancient mathematical corpus never tells you. It is also, by roughly four hundred years, the earliest surviving Archimedes manuscript, and the most important evidence for the diagrams he actually drew.

The Archimedes book supplied most of the pages. Others were used too. Ten folios come from a codex of the fourth-century BC Attic orator Hyperides, who before this discovery was known only from papyrus fragments and from quotations in other writers. Further folios carry a commentary on Aristotle's Categories, and more come from books that have still not been identified.

1846: the first person to take a leaf out of it was a scholar

The manuscript was at the monastery of Mar Saba in the Judaean desert by the sixteenth century, and had moved to the Metochion of the Holy Sepulchre in Constantinople by about 1840. The Biblical scholar Constantine Tischendorf visited the Metochion in the early 1840s and published his Travels in the East in 1846, reporting that he had found nothing of particular interest there except a palimpsest containing some mathematics.

He found it interesting enough. One leaf of the Archimedes Palimpsest was sold to Cambridge University Library in 1876 out of Tischendorf's estate. It is now Cambridge University Library MS Add. 1879.23, and it was not recognised as belonging to this manuscript until Nigel Wilson identified it in 1968, a hundred and twenty years after it was removed.

Tischendorf did not know the book contained Archimedes. Neither did anybody else. He took the leaf because it was an old palimpsest, and the theft is only visible now because the scholarship caught up with it.

A catalogue entry in 1899, and a Danish philologist who knew what he was reading

In 1899 A. Papadopoulos-Kerameus catalogued the library of the Metochion. The palimpsest is MS 355 in that catalogue, and he transcribed a few lines of the undertext without knowing whose they were. Those lines reached Johan Ludvig Heiberg of Copenhagen, the world authority on Archimedes, who recognised them as coming from Sphere and Cylinder.

Heiberg went to Constantinople in 1906, read what he could with the naked eye and a magnifying glass, and had the manuscript photographed under white and ultraviolet light. He established that it carried seven treatises, including the unique witnesses. He incorporated the results into his new edition of the complete works of Archimedes, published between 1910 and 1915, in which this manuscript is Codex C. His photographs are now in the Royal Danish Library and for most of a century they were the only documentary trace of the book.

Heiberg last studied it in 1908. After that it left the Metochion, and how it left is not known.

The missing decades, and an owner the family did not name

It reappeared in a French private collection. The Greek Orthodox Patriarchate of Jerusalem alleged that it had been stolen from a convent library in the 1920s. The selling family's account was that it had belonged to them since the 1920s; the court records establish only that it was in the French collection at least since the 1960s. The project's own history page says flatly that it is not known how the manuscript left the Metochion.

There is now a further name in the frame. Victor Gysembergh, writing in 2026 about the leaf he identified at Blois, states that the illumination on it was executed around 1942 at the instigation of its owner Salomon Guerson, in a desperate attempt to raise its value in order to sell it and escape antisemitic persecution. That is an account of a specific person holding the book, or parts of it, in occupied France, and it sits at odds with the simpler version in which the manuscript passes quietly from Marie Louis Sirieix to his daughter Anne Guersan.

This record does not adjudicate between those accounts. It records that the chain between 1908 and the 1960s is not established by documents, that a court found in 1999 that there was little evidence of the original acquisition, and that the person who put the object on the market in 1998 was Anne Guersan.

Four Evangelists, painted onto the book after 1929 to raise its price

Sometime in the twentieth century, four paintings of the Evangelists in Byzantine style were added to the manuscript, over the prayer book text and therefore over the Archimedes text beneath it. John Lowden showed they were made after 1929, because they were copied from a publication of that year, Manuscrits Grecs de la Bibliotheque Nationale.

They are forgeries in the most damaging possible sense. They were painted to make a scruffy prayer book look like an illuminated one, and each of them destroyed a page of the only surviving copy of the text underneath. The Blois leaf, recovered in 2026, carries a fifth such painting, the prophet Daniel between two lions. That leaf was one of three that Heiberg had photographed and that had since gone missing, which means the pages were removed and decorated to be sold separately.

The counterfeiter's calculation was correct on its own terms. An illuminated Byzantine manuscript was worth more in 1942 than a palimpsest of unreadable mathematics. It took multispectral imaging and a synchrotron to reverse the transaction.

2pm, 29 October 1998, and the figure everybody quotes is the wrong one

The manuscript was sold at Christie's in New York. The project's own account is precise: at 2pm on 29 October 1998 the book was sold to an anonymous collector for USD 2,000,000. The press reported USD 2.2 million; that is the figure in Archaeology magazine and in the Washington Post headline of 30 October 1998.

Both numbers are right, and they are not the same number. USD 2,000,000 is the price at which the lot was knocked down. USD 2.2 million is what the buyer paid including the buyer's premium. The gap is exactly ten per cent, which matches the premium Christie's New York was charging, but no source consulted here states the split in those words, so the identification of USD 2.2m as the premium-inclusive total is DERIVED and labelled as such. Anyone writing that the Archimedes Palimpsest made 2.2 million dollars at hammer is repeating a rounding of a rounding.

One more caution about the date. The project's own history page says the sale was on 28 October and its digital page says 2pm on 29 October. The same institution's website carries both. The catalogue and the contemporary press support 29 October.

There was an underbidder, and it was a government. Greece's consul general in New York bid up to USD 1.9 million before dropping out. After the sale the Greek culture minister Evangelos Venizelos said he was 'very proud and moved because my country raised a great amount of money to buy a treasure which was stolen'. An underbid at 1.9 against a hammer of 2.0 is the tightest possible measure of what the object was worth: one more increment and the Hellenic Republic would have owned it.

The injunction failed, and so did the case, on a point of choice of law

The day before the sale the Greek government and the Greek Orthodox Patriarchate of Jerusalem sought an injunction against Christie's, arguing the book was stolen. Judge Kimba Wood ruled just hours before the auction that the French sellers had clear title, and the sale went ahead. Christie's position, stated at the time by its spokeswoman Vredy Lytsman, was that it had checked the history of ownership and concluded the French family had legal title.

The litigation continued. Greek Orthodox Patriarchate of Jerusalem v. Christie's, Inc., No. 98 Civ. 7664 (KMW), 1999 WL 673347 (S.D.N.Y. 30 August 1999), ended in summary judgment for the defendants. The court held that French law applied to the action. Its observations on laches were made in the alternative, and one of them is worth quoting for what it says about diligence: the court noted that the Patriarchate had retained counsel to prevent the Christie's auction, which showed it could have obtained help to search for the palimpsest at any point in the preceding seventy years.

That is the reasoning that decides most of these cases, and it is why a claimant's silence is more dangerous to its title than a possessor's bad faith. The opinion is unreported; it exists on Westlaw and Lexis and in the docket, not in the Federal Supplement.

Mould, and glue tougher than the parchment it was protecting

The buyer deposited the manuscript at the Walters Art Museum in Baltimore in January 1999 for conservation, imaging and study, and has funded the work ever since. The conservator was Abigail Quandt, Senior Conservator of Rare Books and Manuscripts.

The book had to be taken apart, because the undertext runs through the gutter: the older leaves were bound so that the Archimedes text disappears into the spine. Disbinding began in February 2000 and finished in November 2004. Part of the spine had been glued in the second half of the twentieth century with a modern adhesive similar to wood glue, which is tougher than the parchment it was applied to and extremely difficult to remove. That single intervention, made by somebody trying to hold the book together, is the reason the disbinding took nearly five years.

The other damage is mould. The Canadian Conservation Institute took a core sample the size of a pinhead and found that the parchment collagen is reasonably sound in general and disintegrating where the mould has been at work. Comparison with Heiberg's 1906 photographs shows whole areas of text that were legible then and are gone now. The manuscript that reached Baltimore was in materially worse condition than the one a Danish professor read with a magnifying glass ninety years earlier.

Ultraviolet, pseudocolour, and finally a synchrotron

The imaging campaign ran for years and used everything available. Multispectral photography in visible, ultraviolet and infrared bands, processed into pseudocolour composites that separate the two inks, recovered most of the text. Where a forged painting sat on top of the page, no optical method could help, because the paint is opaque.

X-ray fluorescence solved that. Mapping the iron in the original ink lets the undertext be read through the illumination, because the metal in a thirteenth-century ink is still there under the twentieth-century paint. The X-ray fluorescence work was done at the Stanford Linear Accelerator Center. The imaging as a whole was produced by the Rochester Institute of Technology, Equipoise Imaging and Boeing LTS, with the synchrotron work at SLAC.

Gysembergh notes that these methods are now substantially better than they were at the start of the 2000s, and that a new campaign on the Blois leaf might also recover passages that the Baltimore project could not read.

On the tenth anniversary of the sale, the owner gave everything away

At 2pm on 29 October 2008, ten years to the hour after he bought it, the anonymous owner released the core data of the project on the web. The release is deliberately not a website: no graphical interface, no proprietary software, standards-conformant files, images with embedded metadata and checksums, and transcriptions spatially mapped onto the images. The stated reasoning is that interfaces come and go and an archive should not depend on one.

The licence is stated in the release documentation: the data is released for use under Creative Commons Attribution 3.0 Unported. The only thing asked in return is that copies of any published articles based on the data be sent to the Curator of Manuscripts at the Walters Art Museum. The project's own note on the licence reads: take it, courtesy of the private owner whose generosity is only exceeded by his desire to remain anonymous.

That is why this record can be illustrated at all. A privately owned manuscript, bought at auction over the objection of two governments, is the most freely reproducible object in this archive, because its owner decided it should be.

March 2026: folio 123 turns up in a museum store in Blois

Three of the folios Heiberg photographed had gone missing and were treated as lost. In October 2025 Victor Gysembergh, discussing with colleagues the fact that Blois had once housed part of the French royal libraries, typed the words into Arca, the digital library of the Institut de recherche et d'histoire des textes, more or less as a joke. A record appeared: a parchment in the Musee des beaux-arts de Blois, inventory 73.7.52. The script looked familiar, one face carried a geometrical figure, and the legible text was a passage of On the Sphere and the Cylinder. Set against Heiberg's photographs, the match was conclusive. It is folio 123, carrying propositions 39 to 41 of the first book.

The leaf had been bequeathed to the museum in 1973 by the collector Andre Frank. How he acquired it is not known. The museum had catalogued and digitised it without knowing what it was, which is precisely why it could be found. Its other face carries the painting of Daniel between two lions.

Two folios are still missing. Gysembergh's argument is that all three were probably cut out in France around 1942, that one survived in a provincial museum store and was identified in seconds by a database search, and that the other two may well have taken comparable routes. He plans multispectral imaging and X-ray fluorescence mapping on the Blois leaf.

The owner has never been named, and that is the last open item

The buyer of 1998 has never been identified. He is described in the project's own material only as the private owner, and the widely circulated identifications rest on a first initial and an industry. This record does not repeat them.

What is documented is the behaviour, and the behaviour is unusual enough to state without a name attached. He bought a book that two governments had tried to stop being sold, placed it immediately on deposit at a public museum, paid for a decade of conservation and imaging, and then released the entire dataset under a licence that lets anybody, including a commercial publisher, use it for anything. The manuscript is still on deposit at the Walters.