Quantitative Chemical Analysis of Archaeological Slag Material Using Handheld X-ray Fluorescence Spectrometry

Abstract

Handheld X-ray fluorescence spectrometry (XRF) has seen a dramatic increase in use for archaeological projects. The attraction of the technique is its portable and nondestructive nature. In many cases, the archaeological artefacts in question cannot be destructively sampled, or the piece itself cannot be sent to an analytical laboratory. One of the current research interests associated with the Sagalassos project is the study of the Roman iron industry. Previously exported iron slag and ore from the site and the surrounding area was subjected to laboratory chemical analyses. These indicated that different ores were being utilized in the production of iron in different periods. In order to further the project the slag material still in the site depot needed to be analyzed. However, recent legislative changes mean that materials can only be analyzed on-site. Since samples could not be taken and destructive chemical analysis was no longer feasible, a portable, nondestructive technique was required. Handheld XRF can easily provide qualitative data, but these data are only comparable to other handheld XRF qualitative data, from the same device. Quantitative data gathering is possible, but can be more problematic, particularly when the material in question is heterogeneous in nature. A calibration file was created using the manufacturer's software and "in-house" standards made from the pre-quantified samples of iron slag available in the laboratory. In order to make the calibration as robust as possible, the composition of the standards was analyzed statistically to determine which of these created bias and leverage for specific elements. These standards were then omitted from the calibration for that element. The calibration was tested in the laboratory using samples of iron slag previously analyzed with wet chemistry, and the results indicated that most sample analyses showed <30% error. Results with a >30% error were found in samples which contained very low or very high quantities of the analyzed element, i.e., outside the limits of the calibration. The handheld XRF and the associated calibration file were then used to provide a semi-quantified chemical characterization of the samples in the field depot.