💡 Scientists found that lead in ancient ink absorbs up to 25 times more X-rays than burned papyrus. A September 2026 PLOS ONE study shows that scanning the Herculaneum scrolls for lead, then CT-imaging those that contain it, could let researchers read 2,000-year-old text without ever physically opening the scrolls.
- Lead in ancient ink absorbs up to 25 times more X-rays than carbonized papyrus, making previously invisible text stand out dramatically in scans.
- A relatively inexpensive handheld X-ray fluorescence scanner can screen scrolls for lead first, limiting costly CT imaging to the most promising candidates.
- Around 1,800 Herculaneum scrolls and fragments survive - the only known intact ancient library from antiquity, many still unread.
- This method only helps scrolls whose ink contained lead: many ancient inks were purely carbon-based and would not benefit from this approach.
- Real progress continues: researchers read part of scroll PHerc. 1667 in 2026, and AI recovered 15 columns from a scroll in 2024.
What Are the Herculaneum Scrolls?
In 79 CE, Mount Vesuvius buried the Roman town of Herculaneum under 65 to 70 feet of volcanic rock and ash. Among the ruins was something extraordinary: a private library of roughly 1,800 papyrus scrolls, now known as the Herculaneum scrolls. The eruption carbonized them, turning them into brittle, coal-like tubes that look more like charred logs than books.
They are the only known intact library to survive from antiquity. The collection is centered on the works of the Epicurean philosopher Philodemus, but historians believe it could also hold lost texts by Aristotle, Sophocles, or Sappho. Most remain sealed, unread for two millennia. Physical unrolling attempts have often destroyed the scrolls entirely.
What Did the PLOS ONE Study Actually Find?
A team led by Douglas Seiler of UC Berkeley, alongside physicist Jacob Michael LaManna of the National Institute of Standards and Technology (NIST) and David Kreimer, published their findings on September 16, 2026 in PLOS ONE. Their key insight: some ancient inks contained lead, and lead shows up dramatically in X-ray scans.
To verify this, the team recreated the problem from scratch. They wrote on Egyptian papyrus using ink mixed with measured concentrations of lead, then burned the scrolls in a low-oxygen furnace to replicate what Vesuvius did. X-ray fluorescence detected lead at every concentration tested. X-ray CT scanning combined with software adapted from lithium-ion battery research allowed the team to virtually unroll the model scrolls and read words they had written themselves.
The numbers are striking: lead absorbs up to 25 times more X-rays than burned papyrus. At concentrations as low as 25 micrograms per square centimeter, the text stood out clearly. Letters with lead-based ink, as one account described, "lit up like a Christmas tree" under imaging.
How Does Lead in Ink Help X-Rays Read Burned Text?
The core problem with reading carbonized scrolls has always been contrast. Both the papyrus and carbon-based ink are mostly carbon, so standard X-rays see them as nearly the same material. There is not enough difference to distinguish letters from background.
Lead is chemically different. It is dense and absorbs X-rays efficiently, which is why lead aprons are used in medical imaging. When ancient scribes mixed lead compounds into their ink - as some did in the Greco-Roman world - that lead survived the fire. Two thousand years later, it still provides the contrast an X-ray scanner needs.
The practical workflow the researchers propose has three stages. First, use an inexpensive handheld X-ray fluorescence (XRF) scanner to quickly screen scrolls for lead. Those that test positive move to high-resolution CT scanning. Custom software then digitally separates the layers of the scroll and reconstructs the text, something a human hand could never do without destroying the document.
What This Means for Lost Ancient Texts
The Vesuvius Challenge, a crowdsourced effort that launched in 2023 with prizes for reading Herculaneum text, has already produced results. In 2024, AI recovered 15 complete columns from a single sealed scroll - the first substantial reading in centuries. In 2026, researchers read part of scroll PHerc. 1667. The new lead-detection method builds directly on that momentum.
If a meaningful share of the 1,800 surviving scrolls contain lead-based ink, researchers now have an inexpensive screening method to find the most promising candidates before committing to expensive CT imaging time. That changes the economics of the effort: instead of imaging every scroll at high cost, institutions can run a fast, cheap screen first and prioritize accordingly.
For anyone who cares about language, history, or how knowledge travels across time, the stakes are real. Philosophers and scientists of the ancient world wrote things we have never read, not because the scrolls are gone, but because they survived in a form we could not open. That situation is slowly changing. AI's role in language work has grown fast in recent years, and this is a striking example of what AI-assisted text recovery looks like at the frontier.
It is also a reminder that breakthroughs often arrive through cross-disciplinary borrowing: the software that virtually unrolled these ancient scrolls was originally built to map lithium-ion batteries.
The Honest Limits: What This Method Cannot Do
Not all ancient scrolls used lead-containing ink. Many Greco-Roman scribes wrote with carbon-black lampblack, which leaves no X-ray-distinguishable trace. This method only helps the subset of scrolls whose scribes happened to use lead compounds. Screening will reveal which ones qualify, but a significant number may not.
Even for scrolls that do contain lead, reading the text is not automatic. The CT scan produces a dense three-dimensional volume of data. Separating the tightly packed, fragile layers of a carbonized scroll, then assembling readable letters from those layers, still requires sophisticated algorithms and considerable human effort. The 15 columns recovered in 2024 took a coordinated team months of AI training and refinement.
This is a screening and imaging tool, not a decipherment tool. Once text is recovered, the actual work of reading, interpreting, and contextualizing ancient Greek is a separate challenge - one that still needs classical scholars. Recovering text from a burned scroll and understanding what it says are two different problems. Understanding how language is processed by the brain is hard enough with living speakers; ancient dialects with no living community add another layer.
The study itself also used recreated model scrolls, not originals. The real test comes when researchers apply XRF scanners to the actual Herculaneum collection and see how many scrolls screen positive for lead at detectable levels.
What to Watch Next
The immediate next step is running handheld XRF scanners across the actual Herculaneum collection, held in institutions in Italy, France, and the UK. If a meaningful portion screens positive for lead, the case for prioritized CT imaging becomes compelling and cost-effective.
The Vesuvius Challenge model - open prizes, crowdsourced AI development, public data releases - has already moved faster than traditional academic timelines. That model is likely to continue as more scroll data becomes publicly available for algorithm training.
For the broader picture: the bottleneck is no longer "we cannot access the text." It is shifting toward "we need better computational tools to interpret what we find." That is a more tractable problem, and it is being worked on by researchers from physics, computer science, chemistry, and classical studies simultaneously.
FAQ
What are the Herculaneum scrolls and why are they important?
The Herculaneum scrolls are roughly 1,800 papyrus scrolls buried by Mount Vesuvius in 79 CE. They are the only known intact ancient library to survive from antiquity, believed to include works by Philodemus and possibly lost texts by Aristotle or Sophocles. Most remain sealed and unread.
How does X-ray imaging help read burned scrolls?
Standard X-rays cannot distinguish carbon-based ink from burned papyrus because both are similar materials. If the ink contained lead, however, lead absorbs up to 25 times more X-rays than papyrus, creating enough contrast to image text inside a sealed, fragile scroll without physically opening it.
Does this X-ray lead method work on all ancient scrolls?
No. It only works for scrolls written with lead-containing ink. Many ancient scribes used pure carbon-black ink, which leaves no X-ray contrast. The method's first step is screening scrolls with a handheld scanner to identify which ones contain lead before attempting expensive CT imaging.
When was this research published and who conducted it?
The study was published on September 16, 2026 in PLOS ONE. It was led by Douglas Seiler (UC Berkeley affiliate) and Jacob Michael LaManna (National Institute of Standards and Technology), with funding from the US Department of Commerce.
What is the Vesuvius Challenge and how does this fit in?
The Vesuvius Challenge launched in 2023 as a crowdsourced effort to read Herculaneum scrolls using AI. In 2024, a team recovered 15 columns of text from a sealed scroll using AI. The lead-detection method described in this 2026 study is designed to identify the most promising scrolls for that kind of imaging and AI decipherment pipeline.
Source(s): ScienceDaily (2026); EurekAlert (2026); Archaeology Magazine (2026)
About the author
Dao Huy (Lucas) is a professional translator working across English, Vietnamese, Chinese, and French, with over 7 years of experience in technical, legal, and IP translation. He follows the frontier of science and technology out of genuine curiosity, because understanding how humans record, transmit, and recover knowledge across time is inseparable from the work of translation itself.
He offers English-Vietnamese translation services covering technical documents, patents, legal materials, and software localization. If you have a project that needs accurate, professional translation, he invites you to request a quote at daohuy.com.
Écrit par Đào Huy (Lucas), traducteur vietnamien et spécialiste de la localisation (anglais, chinois et français vers le vietnamien). See translation services →