Scanning Herculaneum Scrolls at Scale
This week, we quadrupled the library of scanned scrolls from Herculaneum
Before last week, along with our partners at EduceLab and Diamond Light Source, we had scanned and released five intact Herculaneum scrolls. This past week, we scanned twenty!
Scanning a scroll is a large coordinated effort that requires the collaboration of many partners. To scan twenty in a single session, we are building on years of foundational work from teams at EduceLab at the University of Kentucky, the Biblioteca Nazionale, the University of Naples Federico II, and Diamond Light Source, along with their unique mindset of openness that has enabled our progress to date.
Following recent strides in ink detection and virtual unrolling, we are on a mission to scan all 300 known scrolls. This week, we made substantial progress towards that goal.
The primary focus of our recent scan session at Diamond Light Source was to test the “productionization” of scanning scrolls. This session was months in the making, and required huge changes to the pipeline for scanning. Many stages in the preparation and scanning process, from the scroll photogrammetry to the 3D printed case design to the scan process itself, were significantly streamlined and automated.
These software improvements of course are only a part of the work required. Our papyrological team and the library staff of the Biblioteca Nazionale were critical to the success of this scaled effort. In addition to making the scrolls available, they also contributed excellent work with photogrammetry, mounting, transportation, and consultation on the content within. We are incredibly thankful to have them as partners.
Preparation
Scanning a scroll is a multi-step task that starts with a process known as photogrammetry. The scroll is placed on a rotating platform with a known scale, and images are captured as it rotates. This gives us the ability to create a 3D model of exactly the right size and shape to safely hold and transport a scroll.
Once we have accurate 3D models, we design and 3D print the form-fitting case in which the scroll, nestled in a blanket of Teflon Relic Wrap, will be placed. It will remain here for the entire transport and scanning process.
Once a scroll is wrapped and placed in a case, it moves to a larger Pelican case with foam cut outs to ensure it does not shift throughout transport. Stickers — carefully chosen by Federica and Giorgio — keep track of which case has which scrolls.
Scanning
Once transported to the facility, the individual scroll cases are carefully removed from their larger travel cases and allowed to settle in the vertical position before imaging. This is to prevent any shifting that may occur during the scanning process, which can take a number of hours.
During this time, the team received safety training by our wonderful beamline scientist Leigh, seen here on the left giving some equipment a stern look, lest it forget who the boss is.
The first step is of course to mount the scroll. The scroll is placed on the lab jack and securely fastened — Mike is explaining why the lab jack is necessary (detailed this below)
The Experiment Hutch, the lead-lined room in which the high energy x-rays that allow us to scan the scroll work their magic, is quite dangerous. To open the shutter and allow the beam to enter the room, the team must go through a multi-step search process to clear the room and close the door. A 30 second alarm sequence is triggered before the shutter can open.
With the shutter open and the beam on, the scanning can begin! The scroll, now on a rotating platform known as the “sample stage”, is scanned in “grids”. Because we have relatively large samples to be captured at high resolution in entirety, we employ a scanning technique that some staff on-site have begun to refer to as the “double-double” (otherwise known as a grid-scan).
The scroll height sometimes actually exceeds the vertical motor travel available, so we use a manual lab jack to scan tall scrolls in multiple parts.
The scroll is centered within the beam, rotated 180 degrees, then translated slightly off-center, lowered, and rotated another 180. In this way we create separate grids of slightly overlapping captures, stitched together to form our full projection. This process is the result of many hours of hard work from EduceLab’s Dr. Seth Parker in collaboration with DLS.
The result of the scan is a projection image of all the data in the plane of the beam. Though pretty, it requires one more step before it is useful for our final ink detection or segmentation tasks.
This is accomplished in the reconstruction process, using a technique called filtered back propagation, which seeks to find the values of voxels at specific locations within the scan data through some fancy applications of Fourier transforms along sliced projection data. After many months of work, a few technical issues at the beamline, and a ill-planned sleep schedule — we finally have our long-sought z-slices.
We’re still working on the reconstruction of all these scans. The image above was a test slice. This process will take some time, after which we will get this data out to our community as quickly we are able. We’d like to again thank our partners at Biblioteca Nazionale, Diamond Light Source, EduceLab, and of course our team and community. This won’t be our last scan session this year - having tested out the productionization, we are ready for more!
Nice, as a build123d contributor I am glad to see it used here. I hope it has been a pleasant experience!
Great news. You and your team's dedication of time and effort to this rescue of history and permanent contribution to humanity's self-knowledge is greatly appreciated by your many silent onlookers. Every update is appreciated, even of frustrating difficulties and the techiest of accomplishments.