Overcoming matrix effects on the determination of phosphorus concentration in sewage sludge ash using laser-induced breakdown spectroscopy hand-held instrumentation

Sewage sludge ash (SSA), derived from the incineration of wastewater treatment sludge, typically contains phosphorus concentrations ranging from 4 % to 12 % by weight. This significant P content makes SSA a promising secondary resource, particularly for applications such as fertilizer production. In this study, we explore the feasibility of using handheld Laser-Induced Breakdown Spectroscopy (LIBS) for the rapid and direct determination of phosphorus content in SSA samples. The proposed method enables rapid and accurate quantification of phosphorus with minimal sample preparation and demonstrates strong resilience to matrix effects, which often compromise the reliability of conventional LIBS analysis. The approach is based on a Convolutional Neural Network (CNN), designed to produce a single calibration model capable of addressing the wide variability in phosphorus concentration typically observed in SSA samples. The innovative aspect of this work is the complete separation of the training stage of the CNN, which is done using simple synthetic reference samples, from the validation, which involves actual SSA samples collected from a waste-to-energy power plant, previously characterized using standard laboratory methods. This procedure allows to select, among the many spectral features that can be used for modelling the training set, only the ones that are proven to effectively work for the determination of phosphorous concentration in the SSA samples, which have a much more complex composition with respect to the synthetic training samples. In addition to presenting this novel methodology, the study also includes a discussion of alternative approaches reported in the literature for matrix-independent quantitative LIBS analysis of phosphorus. This comparative overview highlights the advantages of the proposed method for in-situ analysis of SSA.