Particle swarm optimization for gamma spectra fitting in neutron activation analysis: a comparative study on selenium quantification in wheat

This study presents the application of a Particle Swarm Optimization (PSO) algorithm for the deconvolution of complex gamma-ray spectra encountered in instrumental neutron activation analysis (INAA). The method addresses the limitations of conventional least-squares fitting for asymmetric, overlapping, or low-intensity peaks by employing flexible multi-Gaussian peak models. The robustness and accuracy of the PSO-based approach were rigorously evaluated through its application to a real-world analytical challenge: the determination of selenium in wheat samples from eight Algerian regions. Using the same raw spectral data originally processed with the commercial Genie 2000 software, we performed an independent re-analysis. The PSO algorithm demonstrated excellent fitting performance (R2 > 0.995) and yielded selenium concentrations in strong agreement with the reference values, with relative differences consistently within ±7%. Crucially, the PSO method exhibited enhanced stability in complex spectral regions and reduced sensitivity to initial parameter guesses. This work validates the PSO-driven multi-Gaussian fitting as a reliable, user-independent alternative for precise peak-area determination in gamma spectrometry, offering significant potential for improving accuracy in trace element analysis across various fields.