To read a Roman scroll nobody can open, scientists burned a new one first

To improve the reading of scrolls too fragile to open, researchers made a new one and carbonised it. Its text and ink were known in advance, giving the team a controlled test of imaging and virtual unrolling for material resembling the Herculaneum papyri.[1][2]

The experiment, reported in PLOS ONE in September 2026 by researchers associated with Berkeley and the US National Institute of Standards and Technology, starts with a physical problem that has frustrated computer scientists for years. Carbon-based ink and carbonised papyrus can look too similar in X-ray imagery. A system may be capable of virtually unrolling a scanned scroll and still have almost no visible letters to recognise. What if a tiny amount of metal in the ink made the writing stand out?[1][2]

Why a scroll can survive and still be unreadable

The eruption carbonised papyrus by exposing it to intense heat in conditions that left some rolls physically intact but extraordinarily fragile. Archaeologists once tried to unroll such material mechanically and often turned it into fragments or dust. Modern computed tomography offers a gentler route: record many X-ray views of the solid roll and reconstruct its interior geometry in three dimensions.

But geometry is only half the challenge. The software must find the tightly compressed layers and follow them in a virtual unrolling process, accounting for warped and irregular thickness. Then it must detect ink against the surrounding papyrus. If both are largely carbon, the difference may be too faint for ordinary CT to separate cleanly. Machine-learning methods have helped recognise subtle texture and shape changes, but a better physical signal would simplify the task before prediction begins.

A replica with words the researchers already knew

Douglas Seiler and collaborators acquired modern papyrus and traditional writing materials, added controlled amounts of lead-containing material to ink, wrote passages, rolled the sheets and carbonised them in a low-oxygen container. The text was known in advance. That mattered because a researcher assessing an imaging algorithm needs ground truth: a way to tell whether a reconstructed letter was actually there rather than an attractive statistical guess.

With X-ray imaging, lead-bearing marks were much easier to distinguish from the blackened substrate. Berkeley reported absorption contrast reaching as much as 25 times the surrounding charred papyrus under the tested conditions. The investigators could reconstruct and virtually unroll the laboratory scroll and read its writing. In the report, ink containing lead at concentrations as low as 25 micrograms per square centimetre was detectable with the tested instrumentation.[1][2]

Screening first, expensive imaging second

A handheld X-ray fluorescence instrument detected lead in the laboratory roll. The proposed next use is screening preserved rolls for heavier elements, so scarce tomography time can be directed toward specimens whose ink may produce stronger contrast.[1][2]

AI-assisted methods have already helped recover writing from real Herculaneum papyri. This experiment contributes a different resource: replicas with known text and controlled ink composition, allowing imaging and recognition methods to be tested against a reliable answer.[1][2]

The next step is to assess ink composition in surviving rolls and test whether the imaging advantage carries over. Ancient materials and contamination vary, so the laboratory result provides a route to examine rather than a uniform assumption about the collection.

Better physical contrast makes the computational task easier before an algorithm proposes a letter. Materials analysis, virtual unrolling and classical scholarship each supply a different part of the route from a closed roll to readable text.[1][2]

Report an error or suggest a correction