Using light tech to reveal Herculaneum scrolls
Synchrotron light technology and digital software tools will be used to decode and study 2,000 year-old scrolls. Shardell Joseph finds out more.
The irreparably damaged Herculaneum scrolls from Institut de France are being virtually unwrapped and decoded using the UK’s national synchrotron – Diamond Light Source. The research team from the University of Kentucky, USA, will use both the Diamond Light Source and digital software to decode two complete scrolls and four fragments from the ancient Herculaneum artifacts.
Working to digitally restore and read vast amounts of material in the invisible library – irreparably damaged manuscripts – the team aims to use the same technology as when they successfully visualised the never-before-seen writing inside the ancient Hebrew scrolls from En Gedi.
Buried and carbonised by the deadly eruption of Mount Vesuvius in 79AD, the scrolls are considered to be excessively damaged, extremely fragile, and contain difficult-to-detect ink.
All six items will be scanned at the science facility, Diamond, in Oxforshire, UK, which will provide the data needed to develop the next iteration of the team’s virtual unwrapping software pipeline, a machine-learning algorithm that will enable the visualisation of carbon ink.
Working in the same manner as a giant microscope, the synchrotron produces bright light through harnessing the power of electrons, allowing scientists to study a broad range of objects including fossils, jet engines, viruses and vaccines.
‘Diamond Light Source accelerates particles around an inner ring, and tangential to that ring is emitting different wavelengths of energy,’ University of Kentucky Department of Computer Science Technical Project Manager, Seth Parker, told Materials World.
According to Diamond, the synchrotron can accelerate electrons to near light speed, giving off light 10 billion times brighter than the sun. By using the bright beams known as beamlines, the machine is 10,000 times more powerful than a traditional microscope.
Once the scrolls have been transported to the UK, they will be inserted into the I12 Beamline or joint, engineering, environmental and processing (JEEP) – a high energy X-ray beamline for imaging, diffraction and scattering, which operates at photon energies of 53-150keV.
‘The reason we are at Diamond is because the X-ray tomography we can do with the I12 Beamline is much brighter than any other than any other lab computed tomography (CT) machine that you might be able to get,’ Parker said.
‘A medical CT scanner or a micro CT scanner that you can get in your research lab are very good, but they are not very bright sources. We need that brightness because the objects that we're dealing with are relatively very large, and the resolution that we need is relatively very small. Diamond is one of the only tomography options that lets us quickly acquire extremely high-resolution scans of these objects.’
Restoring the scrolls
According to Parker, the researchers have recently finished the beamline process at Diamond, where they acquired at least 15TB of data. The next step for the team is to run the custom software on the data using a multi-step process.
‘The first thing that we try to do is model the internal structure of the object that we are dealing with – we call this process segmentation. Once we get the model of the internal structure, then we look for ink.
‘That's the most important part of the Herculaneum thing right now is that detecting the ink is very, very difficult even in tomography.’
When the scrolls were written, a carbon-based ink was used. Parker explained how this presents a rather large challenge for decoding the writing as the carbonates make it harder to detect. Because the papyrus material and the ink used are both carbon-based, it becomes difficult to differentiate between the two.
‘Typically, when people are using CT scans to try to see an object they are relying on the fact that the two different objects that touch each other are a different material. The particular challenge with Herculaneum is decarbonisation of the scroll itself,’ said Parker.
The team has used an approach to try to separate the carbonated ink from the decarbonised papyrus, which they have tested on an extremely small fragment.
‘We have developed a custom machine-learning approach that does not look for those continuation differences in the CT data, but instead we train it on getting examples of carbon ink on papyrus. We use that as a predictor inside the Herculaneum scrolls to say, this is where we think the ink is, and this is where we don't think it is,’ Parker said.
‘The third stage is to take all that stuff and plot that out into an image so you can just look at the pages of the scroll. Essentially, we can scale-up our machine learning, run it on this data, and it should show us the text that we are seeing visibly on the surface.’
Read more about Diamond Light Source on page 41.