Cluster analysis of XRPD data in ancient ceramics: What for?

• Clustering of samples according to mineralogical composition
• Definition of most representative samples on which to perform detailed analysis
• The applicability of this method was successfully tested on ancient pottery.
• XRPD clustering was used as a tool to follow the firing dynamics of clay materials.

Cluster analysis of X-ray powder diffraction (XRPD) data was tested on datasets of ancient ceramics, to explore the possibility of using this method for automatic comparison of large numbers of samples, since the interpretation and comparison of their diffraction patterns are time-consuming processes. For this purpose, several case studies were examined, representing coarse and fine-paste pottery, materials from various regions (Italy, Syria, Iran) and of various ages (Bronze Age, Iron Age, 17th century AD). The method turned out to be suitable for defining homogeneous groups of samples for which similar raw materials/paste and firing conditions were used. This multivariate approach also allowed us to identify the most representative samples of each cluster, and was also very useful in selecting a small number of samples on which to perform detailed analyses, such as quantitative phase analysis, out of an initially large number of samples. In addition, statistical comparison of a series of firing experiments showed that cluster analysis of XRPD data is a powerful method for identifying when mineralogical changes occur and for studying the dynamics of firing in terms of mineral changes. Together with these advantages, the limitations and drawbacks of this statistical method, applied to analysis of XRPD data of ancient ceramics, were evaluated.