Lead in ancient ink could make Herculaneum’s sealed scrolls easier to read

The story

A laboratory-made papyrus scroll has shown how trace lead in ancient ink could help researchers recover writing from the carbonized Herculaneum papyri without physically opening them. In a peer-reviewed PLOS ONE study, researchers wrote on modern papyrus with inks containing known amounts of lead, rolled and carbonized the material, and then used X-ray methods to detect and reconstruct the hidden text. The result is a proof of concept, not a reading of an intact ancient scroll, but it proposes a practical way to decide which artifacts are most likely to yield clear images.

The problem dates to the eruption of Mount Vesuvius in 79 CE. Heat and volcanic material carbonized scrolls in the Villa of the Papyri at Herculaneum, preserving their physical form while turning both papyrus and conventional soot-based ink into nearly indistinguishable carbon. Attempts to unroll the brittle objects can destroy them. Modern projects therefore scan the scrolls and computationally flatten their layers, but X-rays often struggle to separate carbon ink from carbonized writing material because the two absorb radiation in similar ways.

Douglas Seiler and collaborators at UC Berkeley, the U.S. National Institute of Standards and Technology, University College London and the UK's Rutherford Appleton Laboratory tested whether metal in the ink could provide a stronger signal. They prepared lead-nitrate mixtures at different concentrations, wrote known passages on Egyptian papyrus and carbonized the finished scroll in a low-oxygen furnace. The model became dark, tightly packed and extremely fragile, reproducing several of the practical obstacles presented by the ancient collection.

The lead changed what the instruments could see. The team reports that lead-bearing lines absorbed substantially more X-rays than the surrounding papyrus, with ink containing as little as 25 micrograms of lead per square centimeter remaining detectable. A laboratory CT system created thousands of projection images, custom reconstruction software assembled them into a three-dimensional volume, and a virtual-unrolling program mapped the curved layers onto a flat surface. In part of the reconstruction, the words ‘the children of’ could be recognized without additional image enhancement.

A second result may be more immediately useful for conservation teams. The researchers detected lead in the carbonized model with a handheld X-ray fluorescence instrument. That suggests museums could first screen unopened scrolls for elemental signatures using relatively accessible equipment, then reserve expensive, high-resolution synchrotron tomography for candidates most likely to contain high-contrast writing. The study proposes this as a two-stage workflow rather than scanning every object with the most demanding system.

There is a credible reason to test the idea on the real collection. Earlier research detected lead in letters on two Herculaneum fragments, and other Roman-period papyri contain metal-bearing inks. The metal may have been an intentional ingredient, a drying aid or an impurity; its origin is not essential to the imaging strategy. What matters is whether enough lead survived in particular scrolls to create contrast. The authors do not claim that every papyrus contains it, and they call their semi-quantitative model results preliminary support for testing intact material.

Independent reporting has emphasized both the promise and the boundary. Nature described the experiment as a demonstration that lead detection can expose writing in burned, rolled papyri, while Chemical & Engineering News highlighted the contrast advantage and the proposed XRF-first workflow. Those reports also preserve the crucial distinction between a controlled replica and an ancient artifact whose ink chemistry, layer geometry and heat history are not fully known.

The model scrolls have value even when an original contains no lead. Their text is known in advance, so researchers can compare what a reconstruction algorithm recovers with the actual ground truth. Varying ink recipes, lead concentrations, carbonization conditions and layer shapes could produce standardized datasets for evaluating virtual-unrolling and machine-learning systems. That is safer than repeatedly adjusting equipment or software on objects that cannot be replaced if damaged.

INNOVOX analysis: the most important innovation is not simply adding a brighter marker to a scan. It is the combination of triage, imaging and validation into a repeatable pipeline. Handheld XRF could rank candidates, CT or synchrotron tomography could acquire detailed volumes, and synthetic scrolls could test whether algorithms recover letters accurately rather than merely generating plausible shapes. This approach may convert a cultural-heritage challenge from a sequence of one-off demonstrations into a measurable engineering process.

What comes next will determine whether the technique changes access to the Herculaneum library. Researchers need permission to screen intact scrolls, map where lead occurs and compare concentrations with the model's detection thresholds. Promising objects would then require high-resolution scans and careful virtual unrolling. Success on even a subset could reveal works of philosophy, literature or science that disappeared from every other surviving tradition. For now, the study supplies an encouraging instrument strategy and a safer testing platform—not a guarantee that sealed ancient texts will become readable.

INNOVOX analysis

The advance does not replace machine-learning ink detection; it could make that task more selective and better grounded. A low-cost elemental screen could identify lead-bearing scrolls before scarce synchrotron time is committed, while model scrolls provide known text for testing virtual-unrolling algorithms without risking irreplaceable originals.

What to watch

Watch for systematic XRF screening of the Herculaneum collection, confirmation of lead concentrations in intact ancient scrolls, synchrotron scans of promising candidates and algorithm tests that compare lead-assisted images with conventional carbon-ink recovery.