This appeared last week:

Digital archaeology

The decoding of ancient Roman scrolls is speeding up

More data, and a more powerful particle accelerator, should pay dividends

May 28th 2025|Herculaneum

IF YOU WANTED to read an ancient Roman scroll, you might reach for a dictionary, and perhaps a magnifying glass. You would probably not think of using a particle accelerator. But that is what is required to unravel the papyrus scrolls found in Herculaneum, a Roman town buried by the eruption of Mount Vesuvius in 79AD. Even then, success is far from guaranteed: since 2023 researchers attempting to unravel the scrolls have been stuck on the first few. Now, armed with more data and a more powerful particle accelerator, they expect to make more rapid headway.

The scrolls in question, stored in a library in a Roman villa that is thought to have belonged to the father-in-law of Julius Caesar, were carbonised by scorching gases that engulfed the town during the volcanic eruption that also buried the nearby town of Pompeii. All attempts to unroll them physically, starting in the 18th century, caused them to disintegrate. So instead researchers have been unrolling them virtually, through computer analysis of high-resolution 3d X-ray scans—which is where the particle accelerator comes in.

Such virtual unrolling is a two-stage process pioneered by W. Brent Seales, a computer scientist at the University of Kentucky. The first stage, called segmentation, involves tracing the edges of the rolled-up papyrus sheet inside the 3D scan, and then extracting 2D images of the scroll’s surface. The second stage, ink detection, analyses the resulting images to distinguish the ink of the scroll’s text from the papyrus background. This is particularly tricky for the Herculaneum scrolls, which were written in carbon-based ink, so there is very little contrast against the background of carbonised papyrus.

Dr Seales thought artificial-intelligence techniques might be able to help. In 2023 he launched a contest, called the Vesuvius Challenge, along with Nat Friedman and Daniel Gross, two technology entrepreneurs who provided backing. A few X-ray scans were made available online, and a community of thousands of enthusiasts has subsequently developed a range of software tools to speed up the fiddly processes of segmentation and ink detection. In late 2023 the project achieved a breakthrough when the first passages of text, in Greek, were extracted from scans of a scroll called “Banana Boy”. Three computer-science students shared a $700,000 reward for doing so. (The scroll’s nickname refers to its banana-like shape rather than its content, which appears to be a previously unknown philosophical work.)

At the time, Mr Friedman predicted that entire scrolls would be decoded by the end of 2024. But progress has been significantly slower than anticipated. “I think I entered 2024 a little cocky,” he admits. One problem was that improving the segmentation software turned out to be unexpectedly difficult. But Mr Friedman now thinks the main obstacle was the quality of the original X-ray scans.

Banana Boy, which belongs to a Parisian museum, was one of four scrolls that had been scanned at the Diamond Light Source (DLS), a particle accelerator in Oxfordshire. A so-called synchrotron light source, it accelerates electrons to almost the speed of light in a storage ring 562 metres in circumference. As the electrons are steered around the ring, they emit electromagnetic radiation, the frequency of which can be carefully tuned, so as to produce X-rays. The resulting powerful beams are then used for various scientific purposes—such as scanning ancient texts.

Of the four originally scanned scrolls, however, Banana Boy is the only one in which ink has been detected. The scan of a second scroll was not as good, Mr Friedman says. Two other, smaller scrolls also seemed to contain very little ink. One possibility is that they were unfinished works, and so were mostly blank. But it is also possible, says Dr Seales, that chemical treatment of those scrolls in the 1980s, during efforts to unwrap them physically, could have affected the ink. Having found text only in Banana Boy, says Mr Friedman, “We were banging our heads against the other three scrolls.”

Then the winds changed. During 2024 the team secured permission to scan a fifth scroll, kept at the Bodleian Library in Oxford, at the DLS. For the first time, individual letters were directly visible in the scans after the segmentation step, probably because this scroll was written with a different type of ink. Finding ink in another scroll was heartening, Mr Friedman says.

This month two volunteer researchers were awarded a $60,000 prize for detecting the scroll’s title—the first time the specific work on a Herculaneum scroll has been identified. It turned out to be “On Vices” by Philodemus, a philosopher who lived in the town (and the likely author of Banana Boy, too). Mr Friedman says longer fragments of text are now being found within the scroll. In April the team scanned another 20 scrolls at the DLS, flown by private jet from the Victor Emmanuel III National Library in Naples. This deluge of new data will help make the segmentation and ink-detection algorithms much better.

At the same time the team has secured a boost in scanning power. This month they undertook a further six-day scanning campaign using the Extremely Brilliant Source (EBS) at the European Synchrotron Radiation Facility in Grenoble, the world’s brightest synchrotron. The EBS can produce X-rays 10trn times brighter than those used in medical imaging—and with twice the maximum energy of the DLS, the EBS can perform scans more quickly. To determine how best to use this more powerful machine, the researchers spent the first three days trying out different scanning parameters, akin to adjusting the settings on a photo scanner to get the best results.

Increasing the incident energy of the individual X-rays in the beam (measured in kiloelectron volts, or keV) produces sharper images, but too much energy reduces the contrast and makes features harder to distinguish, says Dr Seales. Adjusting the so-called propagation distance between the item being scanned and the detector, can also affect sharpness and contrast. A third parameter is the spatial resolution, defined as the width of each volumetric pixel in the scan, measured in microns (millionths of a metre). Scanning at two-micron resolution produces far more detail than at eight microns, but the resulting digital files are 64 times larger.

Over a series of scans, all these parameters were varied in turn. The conclusion, says Dr Seales, was to use X-rays with an incident energy of 110keV (higher than the 53keV used at the DLS); a propagation distance of one metre (a longer distance made the contrast worse); and to scan at four microns and then downsample to eight microns, to get good sharpness at a smaller file size. Having established these settings, the team spent three days scanning a further 20 scrolls. The resulting scans, says Mr Friedman, are easily the best so far. “It’s a step change for us—we think it’s a game changer,” he says.

In particular, there is fine detail even in compressed regions, where layers of papyrus are very close together. This should help make the segmentation process easier and more accurate. Mr Friedman thinks reading entire scrolls by the end of this year is now feasible. “Nothing is going to stop me—we are going to solve this,” he insists. The next step is to “triage” the new scans to find the scrolls that can be read most easily, says Dr Seales. Further improvements are no doubt possible in the scanning process, he suggests: in future it may make sense to do a high-resolution scan that is optimised for segmentation, and then a lower-resolution one with more contrast for ink detection.

Eventually, the team aims to scan all 300 surviving unwrapped scrolls. The ultimate hope is that extracting text from the scrolls, and revealing previously unknown books from antiquity, will provide the justification for a full excavation of the villa in Herculaneum, which may contain thousands more scrolls. Gaining access to a lost library of that size “would be the largest discovery in human history”, says Dr Seales. For now, the villa remains under wraps in a quiet hollow next to the ancient town. But elsewhere, vast energies are being unleashed to uncover its secrets.

Here is the link:

https://www.economist.com/science-and-technology/2025/05/28/the-decoding-of-ancient-roman-scrolls-is-speeding-up

What fantastic news!!! I hope I am around to know what some of them say that is interesting!

David.