What Do the Plaster Casts of Pompeii Tell Us? — Genes, Temperatures and X-rays Reread the Year 79

AD 79, Two Cities Buried

In AD 79, roughly 1,950 years ago, Mount Vesuvius erupted on the Bay of Naples in southern Italy. At the time, this region belonged to the Roman Empire, and people lived in the city of Pompeii southeast of the volcano and the small coastal city of Herculaneum at its foot. The eruption proceeded in two stages. First, pumice, a light volcanic rock full of holes, poured down like rain, and then a pyroclastic flow swept down the mountain (DOI: 10.1017/s0016756819001560). A pyroclastic flow is a mixture of volcanic ash and hot gas that flows swiftly down along the ground. Both cities were buried just like that, only to be rediscovered in the 18th century by people digging in the ground. At Herculaneum, the theater was uncovered first in 1710 while digging a well (DOI: 10.1002/9781119682691.ch22).

In Pompeii, volcanic ash hardened around bodies, and the spaces where the bodies rotted away remained as empty cavities inside the ash (PMID 39515325). In 1863, the Italian archaeologist Giuseppe Fiorelli, who led the Pompeii excavations, made casts of these empty spaces for the first time. A plaster cast is made by pouring plaster into the human-shaped cavity left in the volcanic ash and letting it harden, casting that person's final posture. So far, 104 people have been cast this way (DOI: 10.1017/s0068246220000264).

Fiorelli's plaster casts made Pompeii's dead visible for anyone to see, and at a time when new tools like the camera and nascent sciences were emerging together, Pompeii easily entered the imagination of 19th-century people (DOI: 10.1093/acprof:osobl/9780199569366.003.0011). An adult wearing a gold bracelet with a child on their lap was commonly called a mother and child, and two people who died locked in an embrace were commonly called sisters (PMID 39515325).

This article follows how far science has re-read three questions regarding the victims and artifacts of Pompeii and Herculaneum. Who were they to one another, what killed them, and what was written in the books they left behind?

The Adult Wearing the Bracelet Was a Male

Sometimes, bones remain inside plaster casts. Ancient DNA is genetic material extracted and read from thousands-of-year-old bones and fragments remaining in teeth (PMID 39515325). A 2024 study obtained genome-wide ancient DNA from bones embedded inside plaster casts, reading their sex, biological relationships with one another, and ancestry.

The results did not match the stories attached to the plaster casts. The adult wearing a gold bracelet and holding a child on their lap was genetically an adult male, and had no biological relationship with that child. At least one of the two people embraced and called sisters was genetically male (PMID 39515325). When the genes of the five people in the plaster casts were read, two relationship stories attached to their postures did not match.

Based on these results, the authors conclude that today's assumptions about gender-based behavior may not be a reliable lens for viewing past materials (PMID 39515325). The connections—that someone holding a child is a mother and that people hugging are sisters—were added by people who looked at those postures.

The scope of what these results speak to is five people out of 104 cast in plaster. And the plaster casts themselves are objects that have been handled by human hands. In a 2015 investigation peering into 26 plaster casts with CT or X-rays, the casts had been considerably touched up during the creation process. Bones were often removed before pouring the plaster, and other materials not originally present were sometimes introduced (DOI: 10.1017/s0068246220000264). The researchers who initiated this investigation write that they sought to re-examine existing interpretations that relied solely on appearance and circumstantial evidence. The appearance of the plaster casts seen in museums today contains both the moment of 79 AD and the 19th- to 20th-century hands of restoration.

In another house, genes supported parts of the old estimations. Thirteen bones from a house in Pompeii were considered a family based on archaeology, anthropology, and traces of disease. Reading DNA segments inherited only from mothers and short tandem repeats, six of the 13 people were close blood relatives (DOI: 10.1111/j.1469-1809.2009.00520.x). Six out of the 13 people were confirmed to be close blood relatives, and the relationships of the remaining seven people were not determined by this result.

Regarding ancestors, the emphasis of interpretation differs for the same comparison. A 2022 study analyzed two people from the "House of the Craftsmith" in Pompeii to read the first Pompeian genome, and that genome showed a strong affinity with central Italians from the Roman Empire era. The authors view that there was a noticeable genetic homogeneity in the Italian Peninsula around that time (PMID 35618734). The 2024 study writes that all Pompeians with genome-wide data inherited most of their ancestry from recent immigrants who crossed over from the Eastern Mediterranean, and that the same pattern is seen in ancient genomes of Rome from the same era, emphasizing the cosmopolitan nature of the Roman Empire (PMID 39515325). Both studies used people from the Rome area of the same era as comparison groups, but one interprets by putting homogeneity first, while the other puts migration first. Pompeians with genome-wide data are still few, so it is difficult to determine the history of all Pompeian residents with this small number of genomes.

One study that attempted to obtain DNA from bones in Herculaneum examined four bones bearing signs of having received heat of about 300°C. DNA was visible through staining in three people's bone cell loci, but the only sample actually amplified and read was for one person, and even then, only a short segment determining sex (DOI: 10.3906/biy-1702-48). The authors conclude that if burial conditions are favorable, bones exposed to about 300°C should still be considered candidates for DNA analysis. The traces of heat remaining on the bones lead in themselves to the next question: what killed these people?

The Heat of Herculaneum

In the 1980s, hundreds of bones emerged from the old shoreline of Herculaneum. People had crowded into rooms lining the beach and the waterfront, and were buried in their final postures by the first pyroclastic surge that rolled in (PMID 31782750). Among these bones uncovered in the spring of 1982, those belonging to 139 individuals were studied first (DOI: 10.1201/9781351077750-13). Because the bones of Pompeii victims trapped inside plaster casts could not be accessed and victims from other sites were scarce, anthropological research on the victims was conducted primarily using the bones from Herculaneum (PMID 31782750). A great deal of scattered human bone also remains in Pompeii after excavation, and a significant portion of the plaster casts had their bones removed before the plaster was poured (DOI: 10.1017/s0068246220000264); thus, the weight of research leaning toward Herculaneum was a matter of accessibility rather than the quantity of bones.

A pyroclastic surge is created when a column of eruption soaring into the sky collapses under its own weight. In the eruption of 79 CE, the eruption column alternated between stability and collapse multiple times, during which six pyroclastic surges were generated and destroyed the Roman cities around the volcano (DOI: 10.1130/g32092.1).

The temperature of the deposits left behind by those flows can be measured. By measuring the magnetism acquired as some 200 fragments of lava and roof tiles carried along in the flow cooled, researchers found that the temperature of the deposits ranged from 180 to 380°C and was mostly 240 to 340°C. The temperature dropped when air became mixed in or external water intruded (DOI: 10.1029/2002jb002251). This value is the temperature of the place where the flow stopped and accumulated, and it is not the same as the temperature of the air a person breathed in that exact moment.

In Herculaneum, located right below the volcano, the temperature and force of the flow were too great for anyone to survive (PMID 33654154). The bones bore traces of this. The bones recovered from 12 rooms on the waterfront had red and black mineral residues clinging to them, and those residues had also permeated the ash filling the interiors of the skulls. Large amounts of iron and iron oxides were detected in the residues, leading the research team to view this as the final product of iron-containing proteins in blood breaking down from heat, and to propose the hypothesis that body fluids and soft tissues rapidly vaporized upon exposure to extreme heat (PMID 30256793). The paper's title itself uses the word "Hypothesis".

Inside the skull of one male victim, pieces of brain and spinal cord that had turned into glass were found. Under an electron microscope, the structures of nerve cells remained inside them. The research team explains that the tissue turned to glass as the temperature dropped rapidly after sudden exposure to hot volcanic ash, and thanks to that, the cell structures were locked in place (PMID 33022024). This is an observation derived from a single individual.

The proteins inside the bones showed an unexpected sequence. When 12 bone samples from victims of the two cities were analyzed, the proteins in the bones from Pompeii were more degraded than those from Herculaneum, which had experienced higher temperatures. The research team interprets this as occurring because in Pompeii, soft tissues slowly decayed underground through repeated cycles of drying and wetting, which further damaged the bones, whereas in Herculaneum, the tissues vanished quickly due to intense heat and the bones were subsequently buried in waterlogged soil (PMID 35624181). How well bones survive depends not only on the heat at the moment of death, but also on the underground conditions over the 2,000 years since.

In Herculaneum, multiple lines of evidence point to extreme heat. Vaporization is a hypothesis and the glassified brain comes from a single person, but a 2021 model's flow intensity and temperature estimates point in the same direction.

The Answer Varies in Pompeii

50% of Pompeii's victims were found in the streets during the phase when pyroclastic flows were descending. A study that newly surveyed the strata covering Pompeii found traces of two brief pauses during that second phase, and suggests that these pauses may explain why people were found in the streets (DOI: 10.1017/s0016756819001560). Then what caused these people to die? Pompeii is about 10km away from the crater, and at this distance, the answers differ depending on the study.

The first answer is heat. A 2010 study combining field surveys, laboratory analysis, numerical simulations, and experiments concluded that the main cause of death for victims in the Pompeii area, previously thought to be asphyxiation by volcanic ash, was heat. It states that even at a location 10km away from the crater, exposure to flows of at least 250°C was enough to cause instant death even if one was inside a building (DOI: 10.1371/journal.pone.0011127). The authors added that assuming the damage was minor just because only thin deposits remained at the edges of the flow could lead to underestimating the danger.

The second answer is time. A 2021 study calculated with a new model and suggests that the flow that reached Pompeii was weak in force and low in temperature. It notes that the lack of signs of trauma on the bodies supports this. Under such conditions, people could have survived if the flow had passed within a few minutes, but the calculated duration of the flow was 17 minutes, leading to the conclusion that in that amount of time, simply inhaling the ash floating in the flow was fatal (PMID 33654154). The authors suggest that far from the volcano, the duration of the flow is the key that divides life and death.

The third answer came from plaster casts. A 2023 study examined six plaster casts around the Nola Gate and one from the Suburban Baths in Pompeii using portable X-ray equipment to measure the elemental composition inside the bones and plaster. Although analysis was difficult because the bones retained both changes caused by high temperatures and contamination from the plaster, the authors conclude that they nevertheless support the side that asphyxiation is a more likely cause of death. The same paper notes first that research into the cause of death of these victims has not yet reached a conclusion (PMID 37610984).

When applying the two explanations to a single person who collapsed at the same distance, the difference becomes clear. In the explanation of the 2010 study, that person died the moment a flow exceeding 250°C reached them, and the result would have been the same even if they had taken refuge inside a building (DOI: 10.1371/journal.pone.0011127). In the model of the 2021 study, the flow surrounding that person was low in both temperature and force, and the ash that continued for 17 minutes was the problem (PMID 33654154). The two explanations draw different timetables for the same person at the same distance.

A 2023 review dealing with the bones of volcanic victims summarizes that while hot flows have generally been accepted as the main cause of death in the eruption of 79 AD, interpretations vary depending on the distance from the volcano and the spot within the same site where the victims were found (DOI: 10.1002/9781119682691.ch22). Whether heat or asphyxiation from inhaling ash killed the victims of Pompeii still yields divided answers depending on the study. The timetable of the final moments of each individual can only be narrowed down by reading the discovery location and the strata together, and it has not yet been unified into a single answer.

A Book That Crumbles When Opened

From a Roman villa in Herculaneum, the "Villa of the Papyri," roughly 1,800 charcoal-like scrolls emerged. It is the only library to survive from the ancient world. Inside are many writings by Greek philosophers, particularly Philodemus of Gadara, a first-century BCE Epicurean philosopher (DOI: 10.1038/srep27227). Herculaneum papyri is the name given to the scrolls and their fragments that came out of this villa, burned to a char by the heat of the 79 CE eruption.

The scrolls were found in 1752, and for over 250 years since, they have resisted attempts to unroll them by hand (DOI: 10.1093/9780197835210.003.0058). The charred scrolls are extremely fragile, so trying to open them inevitably damages or breaks them (DOI: 10.1038/ncomms6895). Thus, the scrolls that are still rolled up have been kept untouched for a long time (DOI: 10.7347/ar-2023-p100), and hundreds of completely closed scrolls still remain today (DOI: 10.1093/9780197835210.003.0058). Even the ones already unrolled have not all been read. On the back of one unrolled scroll, hidden Greek writing was revealed by short-wave infrared imaging more than 220 years after it was found (DOI: 10.1126/sciadv.aav8936).

It might seem that taking X-ray pictures of closed scrolls would work, but there is a stumbling block. The ink is made of carbon, and the scroll has also turned into carbon, so the two cannot be distinguished in X-ray photos (DOI: 10.1109/iccit60459.2023.10441240). The ink of Herculaneum has long been considered a carbon ink widely used in antiquity, though studies showing a significant amount of lead in the ink of some fragments are revising that common belief (DOI: 10.1163/9789004444805_005). The parchment scroll from En-Gedi near the Dead Sea, written with an iron-based ink, was read without unrolling in 2016, revealing it to be a copy of Leviticus (DOI: 10.30687/mag/2724-3923/2021/03/002), but Herculaneum's carbon ink did not show up through the same method—namely, the difference in the degree to which X-rays are absorbed.

The first breakthrough came from the imaging method. Phase-contrast tomography is a method to capture 3D interior images by tracking, instead of how much X-rays are absorbed when passing through an object, how much the waves of X-rays shift out of step as they pass through. In 2015, this method showed for the first time several letters hidden inside rolled-up papyri without opening them (DOI: 10.1038/ncomms6895).

Between seeing letters and reading text remains the task of unrolling. Virtual unrolling is the work of following the rolled layers one by one inside the 3D X-ray image of a scroll and flattening them out with a computer. In 2016, it was reported that phase-contrast tomography and new virtual unrolling algorithms found the largest amount of Greek text from inside a closed scroll up to that point, and deciphered part of it (DOI: 10.1038/srep27227). Still, until 2021, the text recovered from inside the scrolls was limited to small portions, and difficulties in the stages of identifying and flattening the layers were blocking the path to unrolling large amounts (DOI: 10.1038/s41598-020-80458-z).

The Machine Finds the Ink

In 2019, a single Greek letter written in carbon ink was revealed on an actual Herculaneum papyrus fragment using machine learning (DOI: 10.30687/mag/2724-3923/2021/03/002). Ink detection is the process of having a machine learning model, trained on fragments with known answers, figure out where ink is located in X-ray 3D images. The correct answers come from fragments that have already fallen apart. Since some fragments show writing under infrared light, the model is trained to find ink that was invisible in the X-rays, using the positions of letters on those fragments as the correct answer (DOI: 10.1109/iccit60459.2023.10441240).

An international competition was launched around this. The Vesuvius Challenge is an international competition aimed at reading closed Herculaneum scrolls using machine learning and computer vision, and within that framework, the scroll PHercParis 4 stored at the Académie des Inscriptions et Belles-Lettres in Paris was virtually unwrapped (DOI: 10.7347/ar-2023-p100).

In June 2026, a further report was released. It is a preprint that has not yet undergone peer review. It describes how high-resolution phase-contrast imaging taken at the European Synchrotron Radiation Facility, improved unwrapping computations, and machine learning were used together to virtually unwrap and read a single scroll named PHerc. 1667 from start to finish, meeting predetermined scope criteria and papyrologist review standards. The authors write that this is the first Herculaneum papyrus to be fully digitally unwrapped and read enough for academic research without being physically opened. In the same report, it is noted that shooting conditions for PHercParis 4 were adjusted so that ink is directly visible in single-slice images, and that for PHerc. 139, the title and author attribution were found, identifying it as Book 8 of Philodemus's On Piety (DOI: 10.48550/arxiv.2606.29085, preprint).

Laying out this progression in stages reveals the scope. That preprint also notes that what previous studies resurrected from closed scrolls were local readings or limited surface areas (DOI: 10.48550/arxiv.2606.29085). There was individual letters in 2015, text in multiple parts in 2016, the virtual unwrapping of a single scroll through a competition in 2023, and an entire scroll in the 2026 preprint. The report of reading a single scroll from start to finish is still a pre-peer-review preprint, many scrolls have not yet been read, and even those that have been read are mostly partial sections.

Who Audits a Book That Cannot Be Opened?

There is no way to open and compare the original text against words read from a closed scroll by machine. A 2021 paper calls text of this kind "virtual from birth" because its subject can never be opened to check the results. The author suggests that scholarly editions of such scrolls must record the role played by the machine, and that a new editorial method combining the humanities and sciences is needed to transparently reveal how the text came into being (DOI: 10.30687/mag/2724-3923/2021/03/002).

Homework remains for the readers, too. A peer-reviewed preprint posted to a data repository in 2026 points out that current state-of-the-art ink readers are neural networks trained on labor-intensive ground-truth labeling, that characters vary in how evenly they are read, and that they offer little explanation for why they made a given judgment. The author proposes a method of extracting ink signals through fixed calculations without training, as a supplementary signal to be used alongside neural network readers (DOI: 10.5281/zenodo.20766168).

Attempts to find other clues continue as well. A 2026 study showed that within data collected from already-unrolled papyri, high-resolution measurements of surface elevation alone contain signals that can distinguish between locations with and without ink. However, the phenomenon of raised letters was not present everywhere, and the performance was inconsistent when a model trained by excluding a single papyrus fragment was applied to that fragment. The authors call these results a proof of concept (DOI: 10.1038/s41598-026-58467-1).

Because text read by a machine cannot be opened and verified against the original, the process by which that text was produced must remain alongside it. The 2021 proposal for editorial methods and the 2026 preprint's emphasis on scope standards and papyrologist review criteria point toward the same question: on what grounds do we trust machine-read text?

What Those Stopped on the Same Day Left Behind

The victims of Herculaneum are a community that died together on the very same day. That is why they serve as material that asks not only about death, but also about life. When researchers restored the diet by separating amino acids one by one from the bone proteins of 17 adults and measuring the ratios of atoms with slightly different weights, the ratio of eating sea food to land food was clearly different between men and women. The research team suggests that this result implies access to food differed by sex (DOI: 10.1126/sciadv.abg5791).

The fact that they died at the exact same moment is also used to test measuring tools. Radiocarbon dating is a method that measures the radioactive carbon remaining in something that was once alive, counting backward to find how much time has passed since it stopped taking in new carbon. Nine people and one sheep who died together in Herculaneum died on the same day, so their actual dates should be identical. However, the dates measured by this method came out different from one another, and the difference grew larger alongside the chemical indicators of bone pointing to sea food, which was explained by the share of old carbon from the ocean entering the body. The authors simply warn that without knowing how nutrients are incorporated into the body, the accuracy of such dietary reconstruction drops significantly (DOI: 10.1002/ajpa.22352).

The tools introduced in this part were covered one by one in the primer series. The genetic relationships of the five people inside the plaster casts are read with the tool from "What Genes Can Tell Us — An Old Book Copied Down the Generations", the heat and diet left in the bones of Herculaneum with the tool from "What Bones Tell — One Person's Diary, Split Between Teeth and Bone", and the mismatched dates of the people who died on the same day with the tool from "Dating with Carbon — Radiocarbon Dates, Read with an Hourglass and Tree Rings". How and why to trust the scrolls read by machines returns to the question posed by "Reading Historical Records — Grandmother's Diary and That Day's Newspaper", namely what must be checked before using a single record as evidence.

To sum up, what has been confirmed and what remains open are divided. The relationship stories of two pairs among the five people in the plaster casts did not match the genes, evidence is gathering at Herculaneum that extreme heat made survival impossible, and letters and passages were read in the charred scrolls. The cause of death for the Pompeii victims varies depending on the study, reading an entire single scroll is at the preprint stage, and many scrolls have not been read yet while even the ones that have been read are mostly just partial sections. Science has separated which parts were data and which parts were our interpretation in the stories attached to the plaster casts.

Source Papers

• "Ancient DNA challenges prevailing interpretations of the Pompeii plaster casts." (2024) — PMID 39515325 — Genome-scale ancient DNA and strontium data from five people's bones inside the plaster casts. The adult in "Mother and Child" is genetically male and has no biological relation to the child; the "Sisters" pair includes at least one male. Most ancestors were recent migrants from the eastern Mediterranean. Concludes that modern gender assumptions may not be reliable as a lens to read the past.
• "Bioarchaeological and palaeogenomic portrait of two Pompeians that died during the eruption of Vesuvius in 79 AD." (2022) — PMID 35618734 — Two people from the "House of the Craftsman"; Pompeii's first ancient genome shows affinity with central Italians of the Roman Imperial period, genetic homogeneity of the Italian peninsula, and spinal tuberculosis.
• "Ancient DNA and Family Relationships in a Pompeian House" (2009) — DOI: 10.1111/j.1469-1809.2009.00520.x — Mitochondrial DNA and STR analysis of 13 bones from a single house, showing six are close biological relatives.
• "Recovery and amplification of ancient DNA from Herculaneum victims killed by the 79 AD Vesuvius hot surges" (2017) — DOI: 10.3906/biy-1702-48 — Four bones exposed to roughly 300°C; staining confirmed DNA in three, but amplification succeeded only for the amelogenin region of one sample.
• "INSIDE THE CASTS OF THE POMPEIAN VICTIMS: RESULTS FROM THE FIRST SEASON OF THE POMPEII CAST PROJECT IN 2015" (2020) — DOI: 10.1017/s0068246220000264 — Fiorelli's first plaster cast from 1863; 104 people so far, with CT and X-ray examination of 26 casts revealing substantial touching up, bone removal, and insertion of other materials.
• "Cities of the Dead" (2011) — DOI: 10.1093/acprof:osobl/9780199569366.003.0011 — Fiorelli's 1863 plaster casts and the 19th-century British audience's imagined Pompeii.
• "Pompeian hiatuses: new stratigraphic data highlight pauses in the course of the ad 79 eruption at Pompeii" (2020) — DOI: 10.1017/s0016756819001560 — Pumice fall and pyroclastic density current phases; two brief pauses during the second phase may explain why 50% of victims were found in the streets.
• "Volcanoes, Bones, and Heat" (2023) — DOI: 10.1002/9781119682691.ch22 — Discovery of the theater during a well excavation in 1710; hot flows are generally accepted as the primary cause of death, but interpretations vary depending on the street and discovery location.
• "The Herculaneum victims of the 79 AD Vesuvius eruption: a review." (2019) — PMID 31782750 — Hundreds of bodies on the beach and in beachfront rooms in the 1980s; Pompeii's bones were locked inside plaster casts, concentrating research efforts on Herculaneum.
• "The Skeletons of Herculaneum, Italy" (2018) — DOI: 10.1201/9781351077750-13 — Discovered in the spring of 1982, studying 139 people.
• "Column collapse and generation of pyroclastic density currents during the A.D. 79 eruption of Vesuvius: The role of pyroclast density" (2011) — DOI: 10.1130/g32092.1 — Repeated stabilization and collapse of the eruption column, resulting in six pyroclastic density currents.
• "Temperatures of the A.D. 79 pyroclastic density current deposits (Vesuvius, Italy)" (2004) — DOI: 10.1029/2002jb002251 — Thermoremanent magnetization of roughly 200 lava and tile fragments; deposit temperatures ranged from 180 to 380°C (mostly 240 to 340°C), with air inflow and external water lowering the temperature.
• "The impact of pyroclastic density currents duration on humans: the case of the AD 79 eruption of Vesuvius." (2021) — PMID 33654154 — Herculaneum was unsurvivable; Pompeii experienced a weaker, lower-temperature flow that lasted 17 minutes, making re-inhalation fatal. Duration is the key at a distance.
• "A hypothesis of sudden body fluid vaporization in the 79 AD victims of Vesuvius." (2018) — PMID 30256793 — Large amounts of iron and iron oxide in the red and black residue from bones in 12 beachfront rooms, suggesting a hypothesis of rapid vaporization of body fluids and soft tissues.
• "Preservation of neurons in an AD 79 vitrified human brain." (2020) — PMID 33022024 — Preservation of nerve cell structures in the vitrified brain and spinal cord of a male victim from Herculaneum.
• "Molecular signatures written in bone proteins of 79 AD victims from Herculaneum and Pompeii." (2022) — PMID 35624181 — Proteomic analysis of 12 bones; Pompeii's bone proteins were more degraded than those of Herculaneum despite Herculaneum experiencing higher temperatures (slow decay and wet-dry cycles versus rapid loss and constant waterlogged burial).
• "Lethal Thermal Impact at Periphery of Pyroclastic Surges: Evidences at Pompeii" (2010) — DOI: 10.1371/journal.pone.0011127 — Reinterpreting the cause of death from asphyxiation to heat; at a distance of 10 km, a flow of at least 250°C causes instant death even inside buildings. Warns against underestimating peripheral risks.
• "The casts of Pompeii: Post-depositional methodological insights." (2023) — PMID 37610984 — Portable XRF of six plaster casts from the Nola Gate and one from the Suburban Baths; analysis is difficult due to high-temperature alterations and plaster contamination, but supports asphyxiation as a highly probable cause of death. Studies on the cause of death have not reached a final conclusion.
• "Virtual unrolling and deciphering of Herculaneum papyri by X-ray phase-contrast tomography" (2016) — DOI: 10.1038/srep27227 — Roughly 1,800 charred papyri, Philodemus, yielding the largest volume of Greek text decoded up to that point from closed scrolls.
• "Historical Unfolding and Virtual Unwrapping of the Papyri from the Villa dei Papiri" (2026) — DOI: 10.1093/9780197835210.003.0058 — Discovered in 1752, over 250 years of physical restoration attempts, with hundreds of completely closed scrolls.
• "Revealing letters in rolled Herculaneum papyri by X-ray phase-contrast imaging" (2015) — DOI: 10.1038/ncomms6895 — Phase-contrast tomography reveals letters inside rolled scrolls for the first time, avoiding the damage caused by opening them.
• "UNA NUOVA STAGIONE PER LA PAPIROLOGIA ERCOLANESE: LA VESUVIUS CHALLENGE E LO SVOLGIMENTO VIRTUALE DEI ROTOLI DELLA VILLA DEI PAPIRI" (2023) — DOI: 10.7347/ar-2023-p100 — Rolled scrolls have long been untouched for preservation; virtual unfolding of PHercParis 4 within the framework of the Vesuvius Challenge.
• "Ancient Greek text concealed on the back of unrolled papyrus revealed through shortwave-infrared hyperspectral imaging" (2019) — DOI: 10.1126/sciadv.aav8936 — Shortwave-infrared hyperspectral imaging reveals Greek text hidden on the back of an unrolled papyrus after more than 220 years.
• "Ink in Herculaneum: A Survey of Recent Perspectives" (2021) — DOI: 10.1163/9789004444805_005 — The conventional view of carbon ink is being revised due to the detection of lead in the ink of some fragments.
• "Ink Detection from Carbonized Herculaneum Papyri using Deep Learning" (2023) — DOI: 10.1109/iccit60459.2023.10441240 — An ink detection model using fragments readable by infrared as ground truth data, addressing the lack of X-ray contrast between carbon ink and carbonized papyrus.
• "Scholarly Editing and AI: Machine Predicted Text and Herculaneum Papyri" (2021) — DOI: 10.30687/mag/2724-3923/2021/03/002 — Virtual unfolding of the En-Gedi scroll (iron ink) in 2016, machine-learning-assisted reading of a single carbon ink Greek letter in 2019; editions of texts that are "virtual from birth" must document the machine's role.
• "A computational platform for the virtual unfolding of Herculaneum Papyri" (2021) — DOI: 10.1038/s41598-020-80458-z — Recovered text is limited to small portions; surface segmentation and flattening hinder the unfolding of large volumes.
• "Complete virtual unwrapping and reading of a rolled Herculaneum papyrus" (2026) — DOI: 10.48550/arxiv.2606.29085 (preprint) — Complete virtual unfolding and reading of PHerc. 1667 under scope and papyrological review criteria, direct visualization of ink in PHercParis 4, and identification of PHerc. 139 as Philodemus's On Piety, Book 8.
• "Volumetric Texture Analysis for Unsupervised Ink Detection in Herculaneum Papyri" (2026) — DOI: 10.5281/zenodo.20766168 (data repository report, pre-peer-review) — Highlights per-character reading discrepancies and lack of explainability in supervised readers, proposing unsupervised complementary signals.
• "Ink detection from surface topography of the Herculaneum papyri" (2026) — DOI: 10.1038/s41598-026-58467-1 — Ink signals derived purely from surface topography; general character relief is absent, and transfer between papyrus layers is uneven. Proof of concept.
• "High-resolution dietary reconstruction of victims of the 79 CE Vesuvius eruption at Herculaneum by compound-specific isotope analysis" (2021) — DOI: 10.1126/sciadv.abg5791 — Diet reconstruction using amino acid isotopes from 17 adults; the ratio of marine to terrestrial food differs significantly between males and females, suggesting gender-based differences in food access.
• "Evaluating marine diets through radiocarbon dating and stable isotope analysis of victims of the AD79 eruption of vesuvius" (2013) — DOI: 10.1002/ajpa.22352 — Radiocarbon dating discrepancies between nine people who died together and a single sheep are due to old carbon from the marine reservoir. Without knowing the degree of nutrient mixing, the accuracy of dietary reconstruction is severely compromised.