Virtual unrolling and deciphering of Herculaneum papyri by X-ray phase-contrast tomography

A collection of more than 1800 carbonized papyri, discovered in the Roman ‘Villa dei Papiri’ at Herculaneum is the unique classical library survived from antiquity. These papyri were charred during 79 A.D. Vesuvius eruption, a circumstance which providentially preserved them until now. This magnificent collection contains an impressive amount of treatises by Greek philosophers and, especially, Philodemus of Gadara, an Epicurean thinker of 1st century BC. We read many portions of text hidden inside carbonized Herculaneum papyri using enhanced X-ray phase-contrast tomography non-destructive technique and a new set of numerical algorithms for ‘virtual-unrolling’. Our success lies in revealing the largest portion of Greek text ever detected so far inside unopened scrolls, with unprecedented spatial resolution and contrast, all without damaging these precious historical manuscripts. Parts of text have been decoded and the ‘voice’ of the Epicurean philosopher Philodemus is brought back again after 2000 years from Herculaneum papyri.

performed 4 . Between 1999 and 2002 multispectral imaging (MSI) of all unrolled papyri was made with important results 5 . This approach showed that the writing could be clearly imaged and distinguished from the papyrus substrate with excellent spatial resolution, given the different characteristic absorption spectra of ink and papyrus and without damaging or destroying the manuscripts. This technique is now used extensively in papyrological and codicological research.
Shortly after MSI was introduced, other non-destructive techniques were applied to both papyrus scrolls and fragments, for use in reading both unrolled and rolled-up papyri. These additional techniques included, amongst other, Scanning Electron Microscopy (SEM)/EDX, Particle Induced X-ray Emission (PIXE), Computer Tomography (CT) and Nano-Computer Tomography (Nano CT), High Resolution Digital Radiography (HRDR), Thermographic Imaging (TI), X-ray Fluorescence (XRF) and Micro-CT analysis (MCT) 6 . Most recently, X-ray phase-contrast tomography (XPCT) was applied to a Herculaneum papyrus roll and a Herculaneum papyrus fragment owned by the Institute of France (Paris). This feasibility testexperiment showed the potential benefit of applying XPCT to papyri, but neither virtual unrolling nor extensive text reading was achieved 7 .

Materials
The materials characterized in this study are: a) two lengthwise entire papyrus rolls (PHerc. 375 and PHerc. 495), with dimensions 7,5 cm (diameter) x 16,5 cm (height) and 7,5 cm (diameter) x 17,5 cm (height), respectively; b) a phantom sample of papyrus roll was realized using sixteen high quality sheets of bare papyrus paper originated from Sicilian papyrus plants and manufactured to reproduce the above papyrus rolls in both geometry and material. 2 cm in diameter x 10 cm in length.

Sequence of numbers and
Greek letters were written on the paper sheets using homemade ink. The latter was manufactured following the elemental composition used for writing on Herculanum papyri available in the literature 4 . Subsequently the paper sheets were first rolled up and then combusted -under vacuum chamber and CO 2 flow -at 350 C, a procedure which was designed to reproduce the physical condition of heat and no oxygen, the most similar to the volcanic flow of Vesuvius eruption in 79 AD.
The images of the phantom sample and of the distribution of complex components inside are shown in Figure 1S. The latter illustrates the capacity and reliability to image the contrast between the location of different fibres components and texts.
In Movie 1S one can appreciate the complex internal structure of papyrus roll PHerc.
495. Because of catastrophic events it is now composed of layers bent and twisted in random directions. The video shows a virtual slicing of the papyrus in the longitudinal direction, i.e. sections parallel to the long axis of the papyrus. Papyrus slices, followed during the video, represent the internal structure at different depths. Each slice has a thickness of about 50 microns and an area that covers almost all papyrus length times its diameter. Layers in the outer regions are well 'sealed', due to the thermal shock. This movie exploit the remarkable contrast and spatial resolution of XPCT experiment to its full potential.
Thanks to the XPCT technique, we discovered that PHerc. 495 contained a large number of grains of sand. This result revealed the history this precious scroll underwent, after being written and stored in the Library of Herculaneum. Before the thermal shock caused by the eruption of Vesuvius, it is believed that a catastrophic phenomenon, such as an earthquake, occurred. During this event PHerc. 495 might have been covered with sand, which penetrated inside. As the volcanic eruption took place, the thermal shock 'sealed' the roll in its outer surface for the subsequent centuries.
Movie 2S shows a virtual slicing of the papyrus into radial direction. The sequence of images in the movie represents the sections (each about 50 microns thickness) parallel to the transversal area of the papyrus at different height. The white grains are particles of high density, i.e. the sand grains. The video shows that sand comes from the outside and it is distributed in depth.
The history of the papyri revealed to us through the XCPT analysis, is fascinating.
Most exciting is the possibility to read and decipher portions of text hidden for almost two thousand years inside the bookrolls. The 'virtual unrolling' has revealed valuable writings by, most likely, the Epicurean philosopher and poet Philodemus of Gadara, with remarkable contrast and legibility. Both the texts and the handwriting 'brought to light' in the bookrolls allowed us to guess the author of both rolls and the subject matter of one of them. Movie 3S shows the virtual opening of the most intact portion of such ancient text ever found.

History and characteristics of the bookrolls
According to the Catalogo dei papiri ercolanesi 6  Distribution of sand particles in PHerc. 495 can be further seen in movie S2. Both images and movies reveal that the small particles of sand are visible in the outer layers of the scroll as well as in its bulk. This is a strong evidence that the small particles penetrated into the bookroll before these were actually exposed to thermic shock waves. These results allow the conclusion that PHerc. 375 was already better preserved than PHerc. 495 already before it was eventually exposed to the thermic shock waves of the volcanic eruption.

Textual reconstruction
The virtual unrolling of the bookrolls revealed sequences of letters, words and, for the first time, expressions and textual portions as well as a textual sign (coronis). Just some instances of what was found are reported in Figure 3, together with a possible textual reconstruction. Please note that these reconstructions are partly conjectural and that alternative readings are also possible. Due to the large number of letters and texts detected inside the bookrolls we were also able to reconstruct a complete Greek alphabet for each roll. These are displayed in Figure 5Sa Figure 3A. In the first case, we have three certain letters (ene) followed by a vertical stroke slightly reaching below the line, which is very likely to be a rho. Textual portions. For PHerc. 375 we have chosen a clearly decipherable text with an unexpectedly high contrast and legibility ( Figure 3C). At line 1, we read the sequence of letters οκτο (okto). In an author like Philodemus (by far the most represented in the Herculanean library) this textual sequence does not seem to exist. Conversely, the sequence οκτ does exist and gives place to the words ὀκτώ, ἀποκτείνω and δέδοκται, middle perfect from δοκέω, 'think', 'seem'. All three possibilites must be excluded when adding an -ο (omicron) after οκτ-. However, the third one stands provided that we inflect it in the impersonal middle pluperfect ἐδέδοκτο,'it had seemed', 'it had been resolved', a quite spread form in Greek (126 occurrences in the TLG). At line 2, the succession of letters ϲτοιοι (stoioi) is not possible as such in Greek. Yet, by dividing ϲτοι from οι, we can either have a syntagm like ἕκα]ϲτοι οἱ̣ , 'each of', or, as alternatives, an adjective in the plural masculine nominative such as χρηϲτοί or also a superlative like πλεῖϲτοι followed by either a definite article (οἱ) or a relative pronoun (οἵ) agreed with the preceding adjective or also any word beginning with οι-. At line 3, parts of an omicron and a large eta are distinctly legible. As for PHerc. 495, among the various textual portions that we were able to identify, we have decided to present here one which may give us an idea of what is being discussed (Figure 3G). At the end of line 1, we read π̣ η̣ ν (pē n) or τ̣ η̣ ν (tē n). At line 2, we find the succession νπολ̣ ι (npoli) that can either admit a single word like, e.g., ϲυνπολίτηϲ, 'fellow-citizen', or ϲυνπολιτεύω, 'live as fellow-citizens', or two different words whereof the latter must be something like πόλιϲ, 'city-state', πολιτικόϲ, 'civil' or 'political', etc. In effect, in the following line we clearly read π̣ όλει̣ ϲ (poleis), 'city-states', followed by ὁ̣ π[ (hop), which may either belong to a conjunction like ὅπωϲ, 'as', 'how', 'that', or an adjective like ὁποῖοϲ, 'of what sort', or a pronoun like ὅπερ‚ 'what', or also an adverb like ὅπου, '(some)where' or even be divided into ὁ or ὅ followed by any word beginning with π-. In any case, the discussion seems to focus on civil or political matters. Please note that the different size and the deformation of some letters in this as well as other images is due to the optical distorsion produced by the (even extremely) irregular surface of the papyrus substrate.
Signs. We also detected inside PHerc. 495 a typical sign used by scribes in antiquity for signalling specific articulations of the text (Figure 3H

Palaeographical description
Given the large number of texts identified, it was possible to analyse the handwriting exhibited by each roll. While being aware that the irregular surface of the layers on which the letters lie may well cause optical distortions, we could make a detailed paleographical description. In this respect, we have to do with two similar, though not identical, hands. Both represent very formal round bookhands with a geometrical shaping and no modular contrast. The letters, which are ca. 2-3 mm high in an average, are accurately written and are endowed with a few decorative elements.
Bilinearity is strictly followed, with the exception of phi and psi. In neither case were ligatures detected. In particular, in PHerc. 375 alpha has the middle stroke either horizontal or ascending, epsilon has the middle bar outreaching the arch and either detached from or attached to it, eta is either epigraphical or with an ascending middle stroke joining the right vertical at the top, kappa has a rounded vertical stroke and is made with either very long arms or in a narrower way with the lower arm fairly suspended above the line, my is made in four strokes and is very angular, zeta and xi are epigraphical, pi is very large and has slightly divergent verticals which appear rounded at the end, tau is made in two strokes and is also quite large, ypsilon and omega are made in two strokes, chi has the descending stroke with symmetrically hooked finials. In PHerc. 495, alpha is sometimes performed more cursively, delta and lambda have the right stroke projecting above, epsilon has the middle bar often outreaching the arch and always attached to it, zeta has slightly undulated horizontals, theta is ogival and sometimes shows the middle bar in a higher position, kappa has either the lower or both arms rounded at the end, my is in four strokes and is very angular as in PHerc. 375, xi, pi and tau have the horizontal stroke gently rounded on the left, pi shows the verticals blobbed at the end and the right one often concave and suspended above the line, ypsilon is made in three strokes with a wide chalice and a short vertical, omega is either as in PHerc. 375 or shows slightly angular curves. On the basis of these features, it is possible to date tentatively these hands to the middle and the third quarter of I century BC, respectively. Useful parallels are, for PHerc. bearing the same number. One may wonder why this is possible. There may be two different answers to this question: either the book was written in two different hands (as is elsewhere attested in the Herculaneum library) or, more likely, was wrongly associated with these unrolled portions at some time after its partial unrolling. In the latter case, the roll would obviously contain a different text as that handed down by the unrolled portions in question.