A Gaussian-mixture-model based peak splitting method in X-ray fluorescence for high concentration ratio solution

Abstract

Simultaneous detection of multi-heavy metal ions in zinc sulfate solution facilitates the zinc hydrometallurgy process control. X-ray fluorescence spectrometry is widely used for composition detection due to its high detection speed and no sample pre-treatment. However, the high matrix ion-to-trace impurity ion ratio in zinc sulfate causes severe spectral overlap, hindering pure peak area acquisition—a vital feature in ion concentration detection. To address this, a peak-splitting method based on the Gaussian mixture model was proposed. Initially, two ion mixing spectra were obtained with optimized reagents. Subsequently, Gaussian models were established for the characteristic peak regions with varying distributions. Finally, the proposed model was combined with Chaotic Sparrow Search Algorithm to handle the overlapping peaks. The experimental results demonstrated superior stability and additivity compared to other methods. All coefficients of determination between single peak areas and ion concentrations exceeded 0.99, confirming the method's efficacy in resolving heavily masked overlapping peaks.