Hidden Library: How Science Is Virtually Unwrapping the Charred Scrolls of Herculaneum

University of Kentucky/Brent Seales
University of Kentucky/Brent Seales / University of Kentucky/Brent Seales
facebooktwitterreddit

Brent Seales called them Fat Bastard and Banana Boy. They were two charred, highly fragile relics that had survived the Mount Vesuvius volcanic eruption of 79 CE, which doused residents of Pompeii and neighboring Herculaneum in a searing blast of destructive gas and volcanic matter. Herculaneum was buried under 80 feet of ash that eventually became solid rock.

Entombed for centuries, the city was rediscovered in the mid-1700s. Incredibly, the library of Herculaneum (known as the Villa dei Papiri) was still filled with over 1800 scrolls, solidified into dark husks. The words inside—religious text, scientific observation, poetry—could provide unprecedented insight into human history. Yet unraveling them has proved difficult. The papyri are so damaged and rigid from lack of moisture that they suffer from a kind of archaeological rigor mortis. And unlike the paralysis that seizes the body upon death, this condition is permanent. Delicate attempts to open the scrolls by hand have been destructive. For a long time, it seemed as if the secrets of the texts would remain locked away for good.

But as Seales stared at the two hardened masses in front of him in 2009, he didn’t share that pessimism. A professor of computer science at the University of Kentucky, he believed that the manual unwrapping that had long failed could be replaced by virtual unwrapping—the digital opening of the texts using computer tomography (CT) scanning and software to penetrate inside the rolled-up scrolls, revealing layers once thought invisible to the eye.

“It’s the only library from antiquity that we have,” Seales tells Mental Floss. “All the knowledge that seems lost, your imagination can run wild.”

 
 

Seales first grew curious about the role of digital manipulation in 1995, when he was invited to assist the British Library in London in scanning and preserving Beowulf. Its 1000-year-old pages had been damaged by fire and warped by the passage of time, imperfections that 2D scans left intact. The use of special software and a 3D visualization, Seales realized, could make it possible to actually flatten the pages and restore smeared copy.

The idea of capturing and manipulating visual data came from Seales's experience in medical imaging, where CT scans can peer inside the body in a noninvasive manner. What if, Seales wondered, the same principle could be applied to the study of fragile documents? What if a relic could be examined in the way a radiologist can visualize, say, the lungs? "That was the eureka moment," he says.

A CT scan of a damaged scroll, with layers visible (L). The red outline is digitally reconstructed in a process called "segmentation" (R).

Seales believed he could use these diagnostic tools to virtually rebuild manuscripts, and returned to the British Library in 2000 to examine other warped documents. After taking images using a prototype of a machine that achieved 3D scans without physical contact, he wrote software that smoothed out the buckled and bunched pages. He likens it to a computer mimicking the tug of gravity, or reversing the direction of a billowing flag. The technique worked—he was able to achieve realistic, flat versions of centuries-old damaged pages.

But Seales believed he could set his ambitions higher: to not only virtually repair a damaged page, but peer inside the Herculaneum scrolls without the risk of causing additional harm. Like many scholars before him, the allure of Herculaneum's vast repository of knowledge had captured his curiosity.

However, the idea of subjecting the scrolls to even minimal handling was something few would consider. Only the Institut de France—one of four major holders of the scrolls—would entertain the idea, and it took four long years to convince them of the possibilities. In 2009, they finally granted permission to Seales's team to scan two Herculaneum scrolls they had in their possession. Officially, the scrolls were categorized as P.Herc.Paris 3 and P.Herc.Paris 4. Seales nicknamed them Fat Bastard and Banana Boy.

The easiest way to imagine the first part of his process is to visualize a sheet of dough that is covered with small red letters and then rolled up. Seen from its edges, the wrap displays its layers and colored pieces, though no observer could possibly identify sentences from that perspective. By slicing the roll into cross-sections as small as 14 microns thick (human hairs are around 75 microns) in a process known as volumetric scanning, Seales can then use geometric "mesh" to reassemble them into a readable surface, depicting the paper so it appears to be as flat as the day it was first written on.

In 2009, the technique allowed Seales to peer inside a closed Herculaneum scroll for the first time, revealing a fibrous labyrinth of data that initially looked like coiled string.

“We saw this amazing structure,” Seales says. But that's where things went wrong.

Seales had believed that trace metals commonly found in the ink of the period could be isolated by the imaging, separating them from the page once the scroll was unraveled and rendering the script legible. But so little of the metals were present that it didn't allow him to identify letters. Nor could Seales distinguish the carbon in the papyrus from the carbon in the ink, which rendered them indistinguishable from one another. The software also wasn't prepared to process the terabytes of data from the scan. While he technically had been able to look inside the scrolls, there was no functional way to determine what he was seeing.

Over the next several years, “Seales Stymied” became something of a headline in academic circles. That ignored the larger point: Seales had proven it was possible to retrieve images from inside the Herculaneum scrolls. It was now a matter of how best to visualize and process it.

 
 

The Herculaneum scans pushed Seales and his team to renovate their software, an act made easier by Seales’s sabbatical work as a visiting scientist at Google’s Cultural Institute in 2012 and 2013. “The interns helped me with the algorithms,” he says, which was a major perk of working for one of the world’s most concentrated and talented assembly of programmers.

His software was vastly improved by the time Seales was approached in 2014 by Pnina Shor, the curator of the Dead Sea Scrolls Project at the Israel Antiquities Authority. Shor had heard of Seales’s work and wanted to know if he could take a look at some CT scan data she had gathered from a 3-inch stick of parchment found in En-Gedi, Israel, in 1970. There was probably ink, but it was obscured by the folds and twists of the parchment.

The En-Gedi scroll's layers are tightly wound (L). Special software is able to isolate one layer to look for text (R).

Seales looked at the scans and applied his process for virtual unwrapping. He used a step he called "texturing," which identifies density differences and other data on the paper that indicate where ink has been applied and assigns a value to that point. Logging the information on individual voxels—the 3D equivalent of pixels—he's able to reassemble them so they appear as a familiar letter shape. The data is then flattened so it resembles an unrolled sheet.

The En-Gedi scroll was made from animal skin, which Seales says is better for contrast against the ink than papyrus, and also benefited from resolution that was twice as good as what he used in 2009. He sent his findings to Shore in 2015; she wrote him back an email humming with excitement. Seales didn’t know what he had uncovered—he doesn’t read Hebrew—but Shor did: It was the first two chapters of the Book of Leviticus, the earliest example of Bible text after the Dead Sea Scrolls themselves.

“When we saw the results we almost fainted,” Shor told reporters. “We had been certain it was just a shot in the dark.”

The fully unwrapped En-Gedi scroll revealed writing that had not been seen in centuries.

Shor’s willingness to embrace new technology helped reveal text locked away for centuries. Conservators are notoriously cautious when it comes to handling such delicate relics—even though Seales never touches one personally, since curators are responsible for getting scrolls in and out of CT scanners. Only recently has Seales been able have more productive conversations at the Officina dei Papiri at the National Library of Naples in Italy, where the bulk of the Herculaneum scrolls are kept, and the University of Oxford. (The Institut de France and the British Library also hold Herculaneum scrolls.)

He remains optimistic that the method used for the En-Gedi material will work for the Herculaneum collection. At a conference this past March, he and members of his team presented new findings showing success in determining the column structure of one text (17 characters per line), as well as reading specific letters—and even entire names. Part of the breakthrough comes from high-powered x-ray beams like the one housed at Diamond Light Source in the UK, which are proving potent enough to isolate the trace amounts of lead in the ink.

 
 

The progress can seem glacial, but Seales has nonetheless gone from imaging a wrapped papyrus to isolating a clearly defined letter. Next, he hopes, will come sentences, possibly isolated by artificial intelligence software he's currently writing.

But even with permission, Seales’s pursuit of a viewable Herculaneum fragment is still dependent on funding. “I sometimes cringe when I see people say, ‘Seales has been working on this for two decades, unable to figure out the problem,’” he says. “Funding comes and goes.” Commercial applications for his software and methodology—like bone scanning or even virtual colonoscopy—could one day underwrite the academic work.

With access, cooperation, and a little luck, he remains optimistic we’ll eventually be able to uncover the knowledge long buried by Mount Vesuvius—time capsules that are slowly revealing their secrets, one micron at a time.

All images courtesy of University of Kentucky/Brent Seales.