Residues on unglazed Roman pottery: extraction of organic spectral signatures via ATR-FTIR and machine learning

: The study of organic residues on archaeological objects offers invaluable insights into past human behaviours, including diet, trade, storage, and ritual practices. Here, we present an innovative analytical approach that combines attenuated total reflectance Fourier transform infra-red spectroscopy (ATR-FTIR) with machine learning techniques to extract spectral signatures for identifying organic residues within complex inorganic matrix, primarily depositional materials in archaeological contexts. The methodology was applied to the study of six unglazed ceramic artefacts coming from a sealed context of a taberna looking onto the Forum area of the archaeological site of Peltuinum Roman city, central Italy. The probable commercial nature of the taberna and the porous, unglazed pottery supported the high probability of organic residue preservation. The analysis successfully identified lipidic residues in two objects: a few African type D sealed earth plate fragments with black staining and a few common pottery olla fragments. These residues were notable for the absence of proteins and polysaccharides. Infra-red spectroscopy together with machine learning extracts organic spectral features at 2921, 2853, 1737 cm-1, indicative of resins, suggesting its use in food preparation or ritual activity. Furthermore, additional markers at 2951, 2918, 2852, 1701, 1558, 1462, 1247, and 1160 cm-1 were attributed to a degraded oil, likely associated with ancient culinary or ritual practices. The findings underscore the value of a non-destructive methodology, enhanced by machine learning, for the detection of organic residues in complex archaeological contexts, especially where molecular signals are masked by interference from archaeological soils. This advanced approach provides a reproducible and chemical-free strategy for the detection of organic spectral features in ancient pottery, significantly enhancing the reliability and efficiency of archaeological organic residue analysis.