Abstract
A probabilistic–deterministic design of experiments (PDDoE) approach was employed to optimize laser-induced breakdown spectroscopy (LIBS) parameters for the quantitative determination of minor components in lead-based alloys. The PDDoE optimization identified 18 J laser pump lamp energy, 1 µs delay, and 1 µs exposure as optimal conditions, minimizing spectral dispersion (5–8%) and ensuring stable plasma formation. The acquired spectra were subsequently processed in an R-based automated workflow, where Linear, Lasso, and Ridge regression models were used to establish quantitative relationships between normalized line intensities and atomic absorption spectroscopy (AAS) reference data. The resulting models demonstrated high accuracy (R2 = 0.97 for Sn, 0.985 for Sb, 0.982 for Bi, 0.919 for As, and 0.905 for Ag), with prediction errors (RMSE) below 10% and limits of quantification (LOQ) under 0.05 wt.%. Principal component analysis (PCA) applied to 43 historical (19th–20th century) and technogenic samples (19th–20th century) allowed us to isolate clusters of Pb–Sb alloys corresponding to secondary accumulator materials, alongside a diffuse group of nearly pure Pb specimens containing variable minor impurities. The combined PDDoE–LIBS–R analytical framework provides a reproducible, non-destructive, and chemometrically validated methodology for the quantitative characterization and classification of archeological and industrial lead alloys.