VNIR–SWIR reflectance spectroscopy as a nondestructive technique for compositional determination of archaeological talc samples with a machine learning approach

Abstract

AbstractThe material from which an archaeological piece is made provides a great deal of information regarding the society in which it was made; thus, any misidentification can lead to erroneous conclusions. The uniqueness of many of these pieces hinders their mineralogical analysis because the pieces cannot be damaged for sampling; therefore, errors in the classification of these materials are common. In the present study, we evaluate the suitability of the spectroradiometry technique in the analysis of two archaeological talc pieces. Both came from the Iron Age archaeological site of Peña del Castro (La Ercina, León) in the northwestern Iberian Peninsula. We compared the spectral curves of these 2 pieces with the spectral curves from 50 talc geological samples from different geographical sources, combining bulk and powdered samples. Our results show that spectral curves enabled the identification of the talc material in the powdered and bulk samples. Moreover, the absence of the talc characteristic features in other soft materials also used in antiquity enabled the detection of incorrect identification of the classified pieces. Even though our results cannot be used to define any absorption feature characteristic to establish the provenance of the material, in the present study, spectral analysis has been shown to be suitable as a nondestructive technique to mineralogically identify archaeological pieces.