In June 2026, a research team announced that it had read the surviving length of a papyrus scroll left buried by Vesuvius for almost 2,000 years. The scroll, PHerc. 1667, remained closed throughout the process. Its layers were carbonised, compressed and too fragile to separate by hand, but X-ray scans allowed researchers to reconstruct a writing surface containing about twenty-two columns of Greek. The surviving text appears to discuss Stoic ethics. The passage adds evidence for Stoic thought to a collection long associated with Epicurean philosophy. Its author remains unidentified.
PHerc. 1667 came from the Villa of the Papyri in Herculaneum, a prosperous Roman town near Naples destroyed in the eruption of 79 CE. Excavators in the eighteenth century found hundreds of papyrus rolls inside the villa, some still tightly wound and others broken or partly opened. The rolls contained ancient works otherwise lost or known only in fragments. The volcanic heat had converted the papyrus to carbon, preserving its form while leaving it extremely brittle. Attempts to pull apart the fused layers could destroy the inked surface before anyone recorded what it said.
Researchers must reconstruct a roll’s layers before they can read its ink. X-rays record the internal structure of the papyrus; computational tools trace the crumpled layers and create images as though the sheets had been laid flat. Machine-learning methods help reveal weak indications of ink in that data. Finally, specialists in ancient papyri inspect the results, distinguish Greek letters from damage, establish possible readings and decide what the evidence supports. Any reading must remain traceable to the scanned material, and specialists must still interpret the Greek.
The Roman villa that preserved an ancient library
Herculaneum lay on the Bay of Naples west of Vesuvius. During the eruption of 79 CE, swift currents of hot ash and gas engulfed the town and left it buried under deep volcanic deposits. The calendar date is still debated. Some scholars defend the traditional 24 August date derived from Pliny the Younger, while geological and archaeological evidence has been used to argue for an autumn eruption. The eruption destroyed Herculaneum and carbonised the papyri regardless of the precise calendar date. A 2023 study of Herculaneum’s pyroclastic surges examines the early sequence of the volcanic surges and the extreme conditions endured by people on the ancient shoreline.
The Villa of the Papyri was an extensive seaside residence, furnished with colonnades, gardens, elaborate rooms and an exceptional collection of sculpture. These remains indicate a wealthy owner. Historians have frequently linked the property to Lucius Calpurnius Piso Caesoninus, a Roman politician, Julius Caesar’s father-in-law and a patron of Philodemus, whose philosophical works were discovered there. The recovered philosophical texts support that association, although no surviving inscription names the owner. Ownership may also have changed between the first century BCE and the eruption of 79 CE.
Royal excavators approached the buried building through tunnels in the middle of the eighteenth century. Workers reached it around 1750 and recovered much of the papyrus material in 1752–1754. The passages were cramped and poorly lit, and the dark cylinders did not immediately look like books. Some could be mistaken for pieces of charcoal or burned timber until their layered structure and surviving writing became apparent. Karl Weber prepared plans documenting the residence. The search for valuable antiquities often came at the expense of recording archaeological context.
Estimates of the library’s size depend on what is being counted. According to the Herculaneum Society’s account, excavators recovered approximately 1,800 fragments that may represent around 800 original volumes. Other catalogues quote larger totals because one ancient roll can survive in several pieces, and a fragment, museum inventory number and complete book are different units. It remains the largest library to survive from the Greco-Roman world. The excavated rooms are only part of the villa, however, so these numbers describe what was recovered, not necessarily everything once stored there. Additional chambers and a possible Latin collection are hypotheses rather than discovered libraries.
Naples holds most of the recovered material, especially in the Herculaneum papyri collection of the National Library. Other scrolls reached Oxford, Paris and further institutions through royal gifts and later transfers. Their separate locations matter for modern investigations. Individual rolls have different degrees of damage and possibly different ink compositions, and conservation decisions rest with the institutions caring for them. An imaging technique that reveals writing clearly in one object may achieve little with another. Each scan must therefore account for an individual object’s condition and conservation requirements.
What the ancient texts have already taught us
Opened or partly opened Herculaneum papyri have already transformed the study of Greek philosophy. Many contain works by Epicurus and later followers of his school. Epicurus founded his philosophical community in Athens around 300 BCE and taught that a good life depended on understanding nature, reducing harmful desires and escaping needless fear. For Epicurus, pleasure included freedom from bodily pain and mental distress, supported by friendship and careful choices. Most of his own extensive writings are lost, which makes later witnesses and surviving papyrus books especially important for recovering the school’s arguments.
Philodemus of Gadara, a first-century BCE poet and philosopher, brought Epicurean ideas into the Greek-speaking intellectual world of Roman Italy. His writings found at Herculaneum address rhetoric, poetry, music, the gods and failures of character, often through arguments with other philosophical schools. These writings preserve philosophical arguments in greater detail than later summaries alone can provide. Even a short additional passage can reveal the reasoning an author used to support a position, how he answered critics or what earlier texts he expected his readers to know.
Although Epicurean works dominate the known collection, other philosophical traditions may also have been represented. The excavated sample reflects where particular rolls were discovered, what survived and which objects could be opened or read. Individual passages should not be assigned to Philodemus merely because they come from Herculaneum or address ethics. An attribution needs support from language, argument, parallel texts or direct information such as a preserved author and title. The Stoic material identified in 2026 provides a specific reason to examine each newly read text on its own evidence.
Why the scrolls could not simply be unrolled
A papyrus book was assembled from strips cut from the papyrus plant, placed in crossing layers, pressed together and joined to make a long sheet. Text usually ran in columns across that sheet. A reader gradually unrolled one portion while winding up the part already read. The Herculaneum examples retained their cylindrical form, but their original flexibility vanished. Carbonised fibres, warped sheets and pressure between layers made the rolls brittle. Pulling on one sheet could tear away the surface on which the writing lay or break the book into fragments.
Early scholars could sometimes obtain writing only by sacrificing parts of a scroll. Some rolls were cut apart and their exposed layers copied or scraped away to reveal those beneath. Antonio Piaggio, an eighteenth-century Vatican priest experienced with manuscripts, designed a machine using threads, weights and carefully regulated tension to draw outer layers apart slowly. The machine helped recover important texts from rolls in sufficiently good condition. Other scrolls were too brittle or too tightly fused to unwind safely. Their surviving transcripts preserve valuable words, but they often reflect a process that damaged or permanently changed the physical object.
Photography offered safer ways to examine sheets that had already been exposed. Infrared and multispectral imaging record the responses of ink and carbonised papyrus to wavelengths beyond ordinary visible light. Marks almost impossible to discern on a dark surface can appear more clearly under these conditions. The method can recover text without scraping an accessible sheet. Yet light-based photography cannot reach writing buried under hundreds of layers in a closed roll.
Carbon-based ink creates a further problem for X-ray imaging. Many ancient scribes used carbon-black ink, which resembles the carbon-rich papyrus after the eruption. Conventional X-ray images respond to differences in material density and absorption, but carbon-based ink and a carbonised sheet can behave so similarly that the letters nearly disappear from the scan. The internal layers may be visible while the writing is not. Metallic ingredients or contamination can produce stronger contrast in some cases. Since the ink recipe is not known to have been uniform across the collection, researchers cannot count on that advantage for every scroll.
How virtual unwrapping turns an X-ray scan into a page
Computed tomography, or CT scanning, supplies the three-dimensional image needed for virtual unwrapping. Multiple X-ray projections of an object are processed into a digital volume. In medical scanning, those data show internal anatomy; in a carbonised papyrus roll, they reveal extremely thin sheets twisted, compressed and folded against one another. Synchrotrons produce powerful, controllable beams that can capture the minute differences needed to follow those sheets. The Diamond Light Source in Oxfordshire and the European Synchrotron Radiation Facility in Grenoble have provided important scans for the Herculaneum work.
The scanned layers then have to be separated in software. Software must recognise the boundaries of a papyrus layer, follow its spiral path through the roll and distinguish it from neighbouring turns. A sheet may be torn, crushed, bent backward or lost inside a damaged area. This operation is called segmentation or surface reconstruction. Researchers represent the traced sheet as a mesh, a connected arrangement of points and tiny surfaces in three dimensions. They then map that mesh onto a flat plane while trying to maintain the spacing and position of whatever letters were originally written on it.
The resulting image resembles an opened sheet even though the real object has not moved. Flattening can also introduce mistakes. A program might cross from one layer to another, join surfaces that were never adjacent or distort a fold enough to suggest false writing. The 2026 research preprint describing the complete reading of PHerc. 1667 describes how the flattened pictures were linked to their precise locations in the three-dimensional scroll volume. That link gives specialists a means of testing a doubtful shape against the underlying X-ray data. This allows scholars to verify that the flattened image corresponds to the physical papyrus.
Once the surface has been followed, researchers must identify ink. Machine-learning systems examine small variations within the X-ray data, including surface features potentially left by ink deposits. Training examples can come from exposed fragments where actual writing can be checked, supplemented by carefully evaluated scans. The models help bring signals into view that would be extremely hard to recognise unaided. A March 2026 study of surface topography tested whether the minute relief of opened papyri differs between inked and blank areas, an approach particularly useful when the chemical contrast is weak. The test helps identify the physical features that may indicate ink.
Ink detection aims to identify physical traces of writing, not to supply missing sentences. Their immediate task is to locate probable marks on the reconstructed surface. A papyrologist then has to decide whether those marks constitute letters, how uncertain strokes should be recorded, where the text is physically missing and what the words mean. Greek grammar and historical context help interpret a reading after the physical evidence has been examined; they should not be substituted for that evidence. The separation helps prevent plausible-looking machine output from being mistaken for ink supported by the scan.
The approach grew out of years of work before the Vesuvius Challenge. In 2016, Brent Seales and colleagues published a non-invasive reading of the damaged En-Gedi scroll. They obtained Hebrew biblical writing from a damaged object that could not safely be unfolded, establishing a workflow for non-invasive reading. The Herculaneum scrolls were more difficult because the carbonised papyrus and many of the inks have such similar X-ray properties. Reading the Greek scrolls required improved scanning, surface reconstruction and ink-detection models adapted to the material.
The Vesuvius Challenge: from one word to Greek paragraphs
In March 2023, Seales and technology entrepreneurs Nat Friedman and Daniel Gross launched the Vesuvius Challenge. They made scans, software and carefully defined research problems available to a large community, offering monetary prizes for advances that could be measured. Independent teams could test layer tracing, ink detection and display methods on the same data. Successful methods and code could be compared and improved publicly, rather than leaving every technical problem to a single laboratory working through it alone.
A major early result arrived in October 2023. Luke Farritor recovered the Greek word porphyras, meaning purple, from a sealed papyrus roll. Youssef Nader independently developed legible readings, while Julian Schilliger helped advance the tools for reconstructing layers. The three later combined their work. The Vesuvius Challenge’s February 2024 grand-prize report records the reviewers’ assessment of four passages each at least 140 characters long, with further columns containing more than 2,000 characters altogether. Farritor, Nader and Schilliger received the $700,000 grand prize for their combined achievement.
The winning readings came from PHerc. Paris 4, held by the Institut de France. The surviving passage considers pleasure, including food, music, fragrances and whether desirable things are available. Those topics fit discussions familiar from Epicurean philosophy, but identifying the author requires more than subject matter alone. The technical advance was the recovery of connected Greek prose without mechanically opening the roll. The recovered portion represented only part of an ancient book. It was nevertheless long enough for specialists to see sustained writing rather than a handful of isolated possible letters.
Releasing scans and tools allowed other teams to build on the competition’s results. Participants could use images of already opened fragments to develop ink-detection methods, test improved surface tracing and compare results produced by different teams. The first-letters announcement records an early phase of the advance, followed by further prize reports and released data. A proposed reading can therefore be challenged by returning to the underlying scan and trying another interpretation. Open methods do not prevent mistakes, but they make it easier for papyrologists and computational researchers to find and correct them.
Oxford’s PHerc. 172 revealed a book by Philodemus
A separate advance came from the Bodleian Libraries in Oxford, which hold three Herculaneum scrolls given to them by Ferdinand IV of Naples in the early nineteenth century. One of these, PHerc. 172, was scanned at the I12 beamline of Diamond Light Source in July 2024. Its ink showed more pronounced X-ray contrast than was usual for these scrolls. A relatively dense ingredient such as lead was suggested as one possible explanation. The Bodleian’s February 2025 announcement stressed that the precise ink composition had not been established, so that possibility remained a hypothesis rather than a confirmed chemical analysis.
Early virtual unwrapping revealed portions of several columns, including diatropē, a Greek term that the Bodleian translated as disgust. The word showed that the scans contained detectable writing, while the author and title remained unknown. Researchers needed to follow the text and look for the kind of closing information that ancient books sometimes carried. A colophon, the information written at the end of a roll, could preserve an author’s name, a title and a book number, supplying stronger identification than an isolated word concerning ethics.
In May 2025, scholars identified this kind of evidence in PHerc. 172. The recovered writing gave Philodemus’s name and the title On Vices, apparently accompanied by a book number. The work belonged to a larger treatment of moral faults and their opposing virtues. Its identification also posed questions: a first book had been associated with the vice of flattery, but the new material did not align with every expectation drawn from previously known witnesses. The Diamond Light Source’s June 2026 report confirmed the identification and described how much further the readable writing from the Oxford scroll had grown.
Ancient philosophical treatises were often divided into books, each of which might occupy its own papyrus roll. Identifying the author and title of PHerc. 172 therefore gives scholars a fixed point for reconstructing a larger work. They can compare the new material with other Herculaneum copies, ancient references and quotations preserved by later writers. A colophon identifies a work but leaves the damaged passages to be deciphered. Possible book numbers must be weighed against the rest of the series, and a secure title does not make every uncertain letter or sentence recoverable.
June 2026: the surviving length of PHerc. 1667 was read
On 25 June 2026, the Vesuvius Challenge announced the first complete digital unwrapping and continuous scholarly reading of the surviving portion of a sealed Herculaneum roll, PHerc. 1667. The claim concerns only the portion of the roll that survives. Earlier physical attempts to open it had removed or destroyed outer layers, leaving an inner core about eight centimetres high from a roll originally much taller. Scanning could not restore the destroyed outer sheets. The achievement was reading the writing that remained, along its surviving length, while keeping the remaining object closed.
According to the official June 2026 announcement, the reconstructed surface extends for approximately 1.4 metres and contains the surviving lower parts of about twenty-two Greek columns. The team combined scans from the European Synchrotron Radiation Facility with digital surface tracing, ink enhancement and careful review by papyrologists. Damaged areas still contain missing fibres and incomplete writing. The team reported a continuous reading of the surviving portion under its published criteria, while acknowledging illegible letters and the lost sections of the ancient roll.
Its subject matter also broadens what scholars know of the villa’s collection. Its vocabulary and arguments suggest a Stoic context rather than the Epicurean teaching that dominates much of the known library. Stoicism stressed the disciplined use of reason and cultivation of virtue as the basis of the good life, differing from Epicurean accounts of pleasure and relief from disturbance. The surviving columns discuss human nature, impulse, practical wisdom and moral judgement. One mentions Aristocreon, nephew and student of the Stoic philosopher Chrysippus. That reference supports a Hellenistic intellectual setting, possibly a composition from the second century BCE, but the author remains unidentified. The reference to Chrysippus’s circle cannot establish that Chrysippus wrote the text.
The Stoic material expands the range of known texts but offers no proof of a second library, and the Epicurean holdings remain substantial. Roman readers could own books that defended their own views, criticised opponents or supplied arguments for study. A work from another philosophical school opens fresh possibilities for comparing the villa’s texts with Stoic material preserved only through later quotations and summaries. It also warns against assigning the remaining unopened rolls to a single school before their words have been examined.
Other 2026 results and the limits of the technology
The 2026 technical paper describes two additional results from 2026. In PHerc. Paris 4, improved scans revealed deposits associated with ink directly within the three-dimensional X-ray volume. Researchers could compare this physical indication with writing predicted by earlier machine-learning techniques. Agreement offers a separate check on the reconstructed letters. In PHerc. 139, the team also identified an author and title: Philodemus’s On Gods, Book 8. The finding helps reconstruct the contents of a larger philosophical work on gods and Epicurean theology.
The 2026 results do not make the remaining scrolls automatically readable. Each roll presents different layer distortions, damage and ink composition. Some marks may be recognised directly through their three-dimensional physical properties; elsewhere the team relies on models that enhance far weaker signals along reconstructed surfaces. Entire portions of the original sheet may no longer exist. In a scholarly transcript, the difference between a clear ink stroke, an uncertain letter and words supplied through editorial reconstruction has to remain explicit.
Other techniques can add information about exposed papyri. A 2025 experimental study of pulsed thermography experimented with pulsed light and thermal imaging, measuring how different parts of damaged papyrus respond to heat. On accessible sheets, those responses could help make faint writing visible and inform conservation work. Thermal imaging of exposed sheets cannot replace X-ray reconstruction of the layers inside a sealed roll. It is useful for fragments, for testing physical properties and for deciding how fragile objects should be studied and cared for.
Reading an ancient text also requires specialists who can interpret the reconstructed Greek. Papyrologists must also apply their knowledge of Greek language, handwriting, the arrangement of book-rolls and philosophical arguments. An abundance of candidate ink marks leaves substantial work in establishing the text and its argument. Specialists publish doubtful readings with indications of damage and uncertainty, allowing later researchers to revise them as more evidence appears. Public datasets and preprints help others test the results, but a preprint has a different status from a completed critical edition or a peer-reviewed article. Some readings and interpretations are likely to change as the work continues.
How much of the Herculaneum library remains unread?
Hundreds of unopened rolls and rolled fragments remain candidates for study. Precise counts vary because a scroll core, a detached fragment, an inventory entry and an ancient book-roll are not interchangeable. A scan is the beginning of the reading process. Researchers still have to reconstruct its surfaces, detect ink and check the text, and fragile objects must be handled according to the priorities of the institutions that care for them. Progress depends as much on skilled conservation and papyrology as on computational capacity.
The University of Kentucky announced in July 2025 that Seales and European partners had secured about €11.5 million for UnLost, an ERC Synergy project intended to combine computer vision with papyrology. Its goal is to develop methods that can be applied systematically to the Herculaneum collection rather than solving each roll as a separate experiment. The university’s project announcement lists Federica Nicolardi and Vincent Christlein among its principal investigators. The grant funds a defined research programme; how many ancient books it will recover remains unknown.
Archaeologists have not fully excavated the Villa of the Papyri. Further investigation might find additional book storage, more scrolls or evidence about the people who assembled the library. Excavation beneath a modern settlement and deep volcanic material can be costly, technically demanding and damaging, so an imagined untouched second library must remain a possibility rather than an available collection of lost books. For now, the surviving rolls in Naples and other collections provide material that can be studied without new excavation. Their writing can increasingly be examined without opening or further damaging the surviving rolls.
The discoveries since 2023 trace how much virtual unwrapping has advanced. A word identified inside a sealed scroll in 2023 became connected passages of Greek in 2024. The identification of Philodemus’s On Vices in 2025 attached new writing to a known ancient author and title. In 2026, PHerc. 1667 yielded a continuous reading of its surviving length, while the recovery of Philodemus’s On Gods, Book 8 provided another reliable point of reference. The lost outer layers cannot be restored, and many closed rolls will present different technical problems. Even within those limits, researchers can now investigate books that earlier generations could protect only by leaving them unopened.
Discussion
Comments