In the aftermath of the eruption of Mount Vesuvius in AD 79, something extraordinary happened inside a Roman villa at Herculaneum. A library was destroyed. But it wasn't completely destroyed.
The papyrus scrolls inside were exposed to extreme heat, carbonising them into blackened, brittle cylinders. The writing survived in some cases, but the books became so fragile that opening them could destroy the very information researchers were trying to recover.
For almost two thousand years, the knowledge inside remained effectively sealed.
Now scientists are doing something that sounds completely backwards. They are burning scrolls on purpose.
Not the ancient ones. They are making modern replicas, filling them with known writing, rolling them up and deliberately carbonising them.
Then they scan the wreckage.
The reason is simple: if researchers can understand exactly what happens to writing when a scroll burns, they may be able to work backwards from the damaged objects at Herculaneum and recover texts that have been invisible since the Roman world disappeared beneath Vesuvius.
The library that shouldn't still exist
The Villa of the Papyri at Herculaneum contained what is regarded as the only known surviving intact library from antiquity. More than 1,800 scrolls and fragments are known from the collection, although hundreds remain rolled and unreadable.
The eruption that destroyed the city effectively transformed the books into charcoal.
That sounds catastrophic. But in another sense, it was the reason anything survived at all. The intense heat carbonised the papyrus rather than simply allowing it to rot away.
The result was a bizarre archaeological paradox: the destruction preserved the information, but also made the information almost impossible to access.
Some scrolls were physically opened in the past.
The results were mixed. Attempts to separate the fragile layers could tear, crush or otherwise destroy the material. And once a scroll has been damaged, there is no second opportunity to open it.
That makes conventional archaeology a poor solution. The books need to be read without being opened.
The problem with the ink
At first glance, modern imaging seems like the obvious answer. If you cannot open something, scan it.
X-ray computed tomography can penetrate solid objects and produce three-dimensional representations of their internal structures.
That approach has already transformed the Herculaneum project.
Researchers can digitally reconstruct the tightly packed layers of a scroll and then attempt to virtually unroll them using computer algorithms.
There is just one enormous problem. The papyrus was carbonised. And much of the ink was carbon-based too.
Inside an X-ray scan, the ink and the carbonised papyrus can therefore look remarkably similar. The writing is physically there. The scanner is looking at it. But the contrast may be so weak that the letters effectively disappear.
It is rather like trying to read black writing on black paper in the dark. Except the paper is wrapped around itself hundreds of times.
Then came the artificial intelligence
This is where the story becomes considerably stranger. Researchers have developed increasingly sophisticated software capable of mapping the internal layers of rolled papyrus.
The process involves reconstructing the three-dimensional geometry of the scroll and digitally flattening individual layers. Artificial intelligence can then search those reconstructed surfaces for subtle patterns associated with writing.
It sounds almost like science fiction. But it has already worked.
The Vesuvius Challenge helped accelerate the development of these techniques. In 2024, AI-assisted work recovered 15 columns from a Herculaneum scroll, revealing an Epicurean philosophical text concerning perception and pleasure.
In 2026, another damaged scroll, PHerc. 1667, yielded evidence of a text apparently connected with Stoic philosophy.
These successes demonstrate something important. The information does not necessarily have to be physically extracted from the object. It can potentially be reconstructed computationally.
But the problem remains: How do you teach a computer what invisible writing looks like?
So scientists made their own ancient scrolls
Douglas Seiler and collaborators approached the problem from an unusual direction. Instead of experimenting directly on priceless ancient manuscripts, they made their own.
The team obtained papyrus and reed pens and produced scrolls containing known writing. Students even wrote passages from Star Wars, the Bible and the television series The Outer Limits onto some of the material.
Then the researchers rolled the papyrus up. And burned it.
The objective wasn't to reproduce every detail of the Vesuvius eruption. It was to create controlled examples where researchers knew exactly what had been written before the material was carbonised.
That gives them something the ancient scrolls cannot: a ground truth. They know what the hidden writing says. They can therefore compare the original text with what the scanner and algorithms recover.
If the computer gets something wrong, researchers know precisely where it failed.
The clue may be metal
The breakthrough investigated in the new study concerns something surprisingly mundane.
Lead.
The researchers found that adding lead to ink creates a much stronger X-ray signal than the surrounding carbonised papyrus, as lead absorbs X-rays far more strongly than carbon.
According to the researchers, lead-containing ink can produce a signal as much as 25 times stronger than the charred papyrus surrounding it.
That changes the problem completely. Instead of asking a computer to find an extremely faint difference between carbon and carbon, researchers may be able to look for something that behaves very differently under X-rays.
The letters could effectively light up. And that means some scrolls may be much easier to read than others.
A scanner could tell them which ones
Perhaps the most useful part of the research isn't the expensive CT scanner. It could be the humble handheld X-ray fluorescence device.
The researchers demonstrated that a relatively inexpensive XRF scanner can detect lead in their carbonised model scrolls.
That creates a possible screening system.
Instead of subjecting every ancient scroll to extremely detailed imaging, researchers could first scan the exterior for evidence of lead. Scrolls showing promising signatures could then be prioritised for high-resolution X-ray CT and computational reconstruction.
That would be a significant change in strategy.
Rather than asking: “Can we read this scroll?” researchers could first ask: “Does this scroll contain the chemical signature that gives us the best chance of reading it?”
But there is a catch
The discovery doesn't mean that the Herculaneum library can suddenly be opened like a digital archive. The experiments demonstrate that lead can make writing much easier to detect in carbonised model scrolls.
They do not demonstrate that large numbers of surviving Herculaneum scrolls actually contain enough lead for the technique to work.
The ancient ink compositions vary. The condition of individual scrolls varies. The geometry of each scroll varies. And even when writing can be detected, virtually unrolling a badly distorted scroll remains a difficult computational problem.
Researchers therefore still have to establish how common useful lead-containing inks actually are among the surviving manuscripts.
That is the next investigation.
There may be more than one way to see invisible writing
Interestingly, the lead discovery is not happening in isolation. Another 2026 study has been investigating something completely different: the physical surface of the papyrus itself.
Researchers used three-dimensional optical profilometry and deep-learning techniques to investigate whether the microscopic topography of carbonised papyrus can distinguish areas that contain ink from areas that do not.
The idea is fascinating. Perhaps the writing doesn't merely differ chemically from the papyrus. Perhaps applying ink physically changes the microscopic surface.
The researchers found that three-dimensional surface morphology contained enough information to distinguish inked and uninked regions in their dataset.
However, they stress that this is a proof of concept based on mechanically opened material, not yet a ready-made method for reading sealed scrolls.
So researchers are effectively attacking the same problem from multiple directions; chemical contrast, surface morphology, and x-ray tomography.
Each technique may reveal a different part of the puzzle.
What if the library contains books we don't know existed?
This is where the archaeological implications become enormous. Much of the surviving literature from the ancient Greek and Roman world is known because later writers copied earlier works. Anything that wasn't copied could simply have disappeared.
But the Herculaneum scrolls are different. They are not medieval copies of ancient books. They are physical manuscripts from the ancient world.
They potentially preserve texts that survived nowhere else.
That means opening a previously unread scroll isn't merely recovering another copy of a known book.
It could mean discovering an unknown work, an unknown author, an unknown philosophical argument, or even a different version of a familiar historical event. We don't know. And that is precisely the point.
The unread scrolls therefore represent information whose contents are still genuinely unknown.
The strange importance of a burned copy of Star Wars
There is something almost absurd about the experimental method. Somewhere in a laboratory is a carbonised papyrus scroll containing words from Star Wars.
It was deliberately destroyed so scientists could learn how to read something written nearly two thousand years ago.
But scientifically, that seemingly ridiculous object is extremely valuable.
Researchers know exactly what it contains. They know where every letter was written. And they know what happened to it during the experiment.
They can scan it, virtually unroll it, and ask an algorithm to reconstruct the writing. They can then compare the result against the original.
That is almost impossible with an ancient manuscript. There is no original digital copy of a lost Roman book sitting in a laboratory waiting to tell researchers whether the algorithm got it right.
The modern replica provides that missing control.
And the machines themselves are evolving
There is another intriguing connection. Some of the software being adapted for virtual unrolling was not originally designed for ancient manuscripts.
Researchers have adapted computational techniques used for analysing the layered structures of lithium-ion batteries.
That may sound unrelated. But mathematically, the problems have something in common. Both involve complicated layered structures that need to be reconstructed from imaging data.
The computer doesn't need to know that one object is a battery and the other is a Roman book. It needs to understand geometry.
That is one of the less obvious consequences of modern archaeology.
The technology used to recover ancient knowledge may come from fields that have nothing to do with archaeology; physics, materials science, medical imaging, and AI machine learning. All are now converging on a library buried by a volcano in AD 79.
What else is hiding inside?
There are still hundreds of unopened or unread Herculaneum objects. And the number matters less than the possibility.
Suppose only a small proportion contain ink that responds strongly to X-rays. That could still represent a substantial number of potential texts. And even if most turn out to be damaged beyond recovery, a handful of successful readings could change what we know about ancient intellectual life.
The philosophical works already recovered demonstrate what is possible.
But there is no guarantee that every future discovery will be philosophical. There could be poetry, letters, technical material, historical writing, scientific observations, administrative documents, or texts from authors whose names have completely disappeared from history.
At this point, nobody knows.
The paradox of preservation
The Herculaneum scrolls present archaeology with a remarkable paradox. The eruption destroyed the library. But it also preserved it.
The heat destroyed the papyrus as ordinary writing material. But it stopped the manuscripts from decomposing. The carbonisation made them almost impossible to open. But modern imaging may allow researchers to read them precisely because the pages remain physically present inside those blackened cylinders.
The destruction created the preservation. And the preservation created the problem.
Now science is trying to reverse the second part without undoing the first.
We may eventually read books that history forgot
For centuries, the Herculaneum scrolls were essentially sealed time capsules.
People could see that books were there. They could hold them. They could study their physical remains. But the words themselves remained trapped.
That is changing. And the most interesting part isn't simply the use of AI. It is the methodology.
Scientists are deliberately recreating the destruction of ancient books so they can understand how information survived that destruction. They are manufacturing controlled disasters and burning known texts.
They scan the wreckage. They are teaching algorithms how to recognise the faintest traces. And they are looking for chemical fingerprints that might make hidden letters dramatically easier to detect.
If the approach works on enough surviving scrolls, the result could be something extraordinary. Not a reconstruction based on later copies, nor a historian's guess. Not a translation preserved through centuries of repetition either. But the actual words written by people who lived nearly two thousand years ago.
Words that were there all along. Hidden inside the blackened remains of a library that should have been lost forever.
The books survived the fire. Now scientists are trying to make them speak.
Read more on:
Italy's Roman Underground Structures: Hidden Cities Beneath the Eternal Landscape
The Roman Shipwrecks of the Mediterranean: How Much Did Ancient Trade Really Move?
Mysterion: The Hidden Heart of Ancient Greek Religion
Comments
Post a Comment