Palm-leaf manuscripts reveal traces of their preparation through X-rays and machine learning

For centuries, religious texts, literary works, and scientific knowledge were preserved on fragile palm leaves across South and Southeast Asia. The leaves underwent meticulous preparation before inscribing, with methods varying by region and leaving distinct material signatures. Today, determining how they were processed and their origins remains a challenge.

An international research team has investigated what structure and chemical composition of these manuscripts can reveal about their production. They combined X-ray analyses at PETRA III at DESY and at the ESRF in Grenoble, France, with machine learning. The study examined 160 material entries, including 85 manuscripts and 75 modern reference materials originating from different collections and regions across South and Southeast Asia.

The team's nondestructive approach, which uses X-ray techniques, provides information about the leaves' internal structure and chemical composition without removing any samples from fragile historical objects. By analyzing the measurements using machine learning, the researchers discovered that the two palm genera can be distinguished based on their material properties.

Processing significantly impacted the differences between the palm genera, with these differences becoming more pronounced after treatment. While geographical provenance was more challenging to establish, systematic material differences suggest material fingerprints related to regional processing practices. However, some provenance groups were represented by few samples, and the influence of aging and preservation conditions on material properties was only partially assessed.

This study offers manuscript scholars an additional way to investigate historical objects, providing nondestructive, quantitative, and cross-collection sources of evidence for cultural and manuscript studies when combined with historical and codicological evidence.