Classification of papyrus fragments by chemometric analysis of near-infrared spectra

Abstract

Papyrus has been used for millennia to record information, for sophisticated works of art as well as mundane notes. The collection, identification, and translation of papyrus fragments therefore opens a gateway into the past. To aid the efforts to access the history recorded in papyri, we investigated the suitability of NIR spectroscopy to perform two tasks: One is to support the authentication of ancient papyri, by differentiation of papyri that were manufactured more recently and subjected to accelerated ageing to resemble the originals. The other is the extensive task to piece together papyrus fragments into readable texts again. In museums around the world, more than 100,000 ancient papyrus fragments still wait for their proper assembly, deciphering and publication. The papyrus writing-ground was analysed by near-infrared (NIR) spectroscopy, and the spectra were evaluated using principal component analysis (PCA), hierarchical cluster analysis (HCA), partial least squares discriminant analysis (PLS-DA), and self-organizing maps (SOM). Cluster analysis and PLS-DA proved to be useful tools for distinguishing modern papyri from ancient papyri which were provided by collections in Vienna and Leipzig. Neither natural nor accelerated ageing affected the classification. A PLS-DA classification model, constructed from NIR spectra of 89 model scores, detected recent Papyri samples with 100 % sensitivity and specificity, even after accelerated ageing. The identification of groups of fragments of ancient papyri based on NIR spectra and chemometry is not straightforward. HCA, which focuses on the differences between samples, only grouped the fragments of 4 out of 20 papyri correctly. SOM, which rather focuses on the similarities, grouped 6 sets of fragments correctly. An automated grouping of fragments remains difficult, since the fragments themselves are heterogeneous while similarities between unrelated ancient papyri can be large.