# Development and optimisation of ex situ portable X-ray fluorescence spectroscopy for heterogenous post-metallurgical sites

## TL;DR

This study improves portable X-ray fluorescence for measuring metals in soil and slag, showing some preparation steps can be skipped to save time and money.

## Contribution

The study demonstrates that certain pre-processing steps are unnecessary for accurate pXRF measurements at heterogeneous post-metallurgical sites.

## Key findings

- Pre-processing steps like sieving and grinding improved pXRF measurement precision and accuracy.
- Organic matter had negligible impact on pXRF accuracy and precision.
- BOS slag and soil created bimodal metal distributions and a negative correlation for titanium.

## Abstract

Portable X-ray fluorescence spectroscopy (pXRF) is widely used for rapid measurement of metals in soils, however, thorough evaluation of common pre-processing methods and their effectiveness is limited. This study addresses processing methods using samples collected at a highly heterogeneous post-metallurgical site containing, basic oxygen steelmaking (BOS) slag and soil; the former being an important source of potentially toxic and valuable elements. Impact of pre-treatment processes, (sieving, drying, grinding, sample vessel, and ignition) on the accuracy of pXRF measurements was compared against reference ICP-MS measurements.Of the twelve elements detected, four showed qualitative (Cr and Fe r2 ≥ 0.60, RSD ≤ 30%) or quantitative (Mn and Ca r2 ≥ 0.70, RSD ≤ 20%) measurements for raw samples. Improving to six elements after pre-processing (Sr qualitative, and Pb, Cr, Mn, Ca, Fe quantitative). Sieving and grinding improved precision (average RSD fell by 7.17 and 8.37% respectively), while drying and grinding enhanced accuracy (average r2 increased by 0.03 and 0.10 respectively). This study provides the first evidence that organic matter does not significantly impact pXRF accuracy or precision (average r2 and RSD changed by zero and − 0.32%, respectively). The two distinct matrices (BOS slag and soil) on-site resulted in a bimodal concentration distribution and a negative correlation for Ti. Importantly, this research proposes that not all common pre-processing steps are necessary to generate high-quality data due to their negligible impact on accuracy or precision (such as incineration to remove organic matter), thereby increasing the speed and reducing the cost of data collection.

The online version contains supplementary material available at 10.1007/s10653-025-02574-x.