The study of the provenance of ancient copper objects on a scientific basis embraces a number of challenging issues that can be summarised in the following question: is it possible to link a copper object to the minerals that were used to produce it? In this respect this work reflects one of the most important problem related to this question, namely to find out if Cu-minerals coming from different historical Cu-ores can be traced as a function of their provenance, giving information about the origin of the metal used by the chalcolithic metalworkers.

To aid metal provenance studies, a database of fully characterized Alpine copper mineralization is being developed as the fundamental reference frame for metal extraction and diffusion in the past. In the early stages of the project, some of the most well known copper deposits in the Western Alps were selected and compared with very different minerogenetic deposits from the French Queyras (Saint Veran) and the Ligurian Apennines (Libiola, Monte Loreto).

The fully characterized samples were then analysed by ICP-QMS (Inductively Coupled Plasma-Quadrupole Mass Spectrometry). The abundances of about 60 minor and trace elements, including the rare earths, were measured in all samples. Furthermore, the feasibility of the routine reliable measurement of the 65Cu/63Cu isotope ratio and its eventual use as a possible ore tracer was tested.

Principal Component Analysis (PCA) was initially employed for this purpose. However, the PCA models were able to separate only a few numbers of different Cu-ores according to their geochemical and isotopic features. A complete description of all the investigated Cu-ores was obtained using a different approach which is based on Partial Least Squares Regression -— Discriminant Analysis (PLS-DA). The advantage of this method is related to the regression coefficients that the calculation provides. They allow establishing which variables mainly characterize a Cu-ore with respect to the others. A variable selection was thus performed both using PLS-DA, and studying the correlation loadings calculated from the PCA models.

As we will show, the combination of both methods, PCA and PLS-DA, allows separating the investigated Cu-ores efficiently. Finally, applications of these models to ancient copper objects provenance determination will be also treated.