A reed pen dipped in black ink draws lines on a sheet of Egyptian papyrus. Lead nitrate, a salt that dissolves in water, has been mixed into the ink. The sheet is rolled up and shut in a steel container. The container goes into a high-temperature furnace with little or no oxygen. Out comes a black cylinder 4 centimetres across and 11 long. Every attempt to unroll it breaks it into pieces.
It is a Herculaneum scroll made in the laboratory. It was built by Douglas Seiler, an affiliate of the University of California, Berkeley, and five colleagues from Berkeley, the US National Institute of Standards and Technology, University College London and the Rutherford Appleton Laboratory in England. The paper appeared in PLOS One on 16 September 2026, in the public domain. ScienceDaily ran the university press release on 26 September.

The real scrolls were found in 1752 in the Villa of the Papyri at Herculaneum. The eruption of Vesuvius in 79 CE carbonised them at temperatures estimated between 400 and 900 degrees Celsius. It is the only known intact library from classical antiquity, and almost all of the scrolls are in the National Library in Naples. Most of the texts are Epicurean philosophy, many by Philodemus of Gadara. Hundreds of scrolls are still rolled up, and early attempts to open them by hand destroyed many.
Rolled scrolls are studied with X-ray tomography, the technique behind a CT scan. The best scans are made at a synchrotron, an accelerator that produces very intense X-ray beams. The Vesuvius Challenge, an international competition led by Brent Seales and Seth Parker, analyses those images with artificial intelligence and has already read portions of text. The problem is contrast. The ancient ink is carbon-based, and burnt papyrus is carbon too. Both absorb X-rays almost the same way, and the letters merge into the background.
Lead is a heavy element and stops X-rays far better than carbon. An earlier study cited in the paper found lead in the letters of two Herculaneum fragments using X-ray fluorescence. The method strikes the sample and measures the radiation each element gives back at its own energy. In the two fragments the lead amounted to 84 and 16 micrograms per square centimetre, give or take 5. Where it came from is still unclear. The authors suggest an ingredient of the ink or an additive used to dry it.

The team first made test pieces. On one piece of papyrus the top two lines are written without lead. The next two carry 200 micrograms of lead per square centimetre, the rest progressively less, down to 25. After baking, the piece goes under the X-rays. The two lead-free lines vanish. All the others show, even the one with the least lead.
Then the whole scroll. NIST’s tomography system is a laboratory instrument, not a synchrotron. It took 1,001 images over a full turn, 5 seconds each, with the tube at 30 kilovolts. A purpose-written program follows the spiral along curves drawn on the images and lays the scroll flat on screen. In the enlarged detail, with no enhancement, three English words can be read, «the children of».

The second part of the test is about choosing scrolls. A handheld fluorescence unit, of the kind tuned to find lead in paint, was held almost against the carbonised scroll and found lead at several points. On the test pieces another handheld unit gave a signal down to 20 micrograms. On the lead-free piece there was no signal.
Hence the authors’ proposal. Screen the library’s scrolls with a handheld unit on site, and send those with lead in their letters to the synchrotron first. And use purpose-built scrolls, with a text known before baking, to train reading software without risking real papyri. The method has not yet been tried on the Herculaneum scrolls.
Some limitations are stated by the authors, others show in the numbers. The model scroll is 4 centimetres across, a real one about ten, and unrolled it often reaches 10 metres. The X-ray test went down to 25 micrograms of lead, more than the 16 of the poorer ancient fragment. The furnace temperature is not given in the text. The authors call for a faster, or automated, way to screen scrolls with the handheld unit. The work relied on NIST equipment and staff, and the authors declare no competing interests. The paper carries no statement on the use of artificial intelligence.

The press release adds details the paper does not contain. It mentions high school students writing lines from Star Wars and the Bible, software first built for lithium-ion batteries, and lead absorbing up to 25 times the X-rays of burnt papyrus. The text of the paper contains none of the three.
The Berkeley scroll gave back three words. How many of the Naples scrolls carry lead in their ink is still unknown.

References
Seiler D., LaManna J.M., McOsker M., Kreimer D., Daugherty M.C., Dopke J., «A model carbonized papyrus scroll opens a novel path to identifying readable scrolls of the Herculaneum Library», PLOS One, 21, 9, e0353485, 16 September 2026, open access under the CC0 public domain dedication. doi: 10.1371/journal.pone.0353485
Study data, National Institute of Standards and Technology repository. doi: 10.18434/mds2-4227
University of California, Berkeley, press release by Robert Sanders, 16 September 2026, «To Read 2,000-Year-Old Burned Scrolls, Scientists Burn Their Own», republished by ScienceDaily on 26 September 2026 as «Scientists Burned Their Own Scrolls to Unlock a 2,000-Year-Old Mystery».