Ag+ modified paper-based SERS combined with SiPLS for quantitative detection of soluble As3+ in aqueous realgar solutions.

The detection of soluble arsenic in realgar and its preparations is crucial for toxicity evaluation. Therefore, surface-enhanced Raman spectroscopy (SERS) combined with machine learning was applied for the rapid detection of soluble As3+ in realgar and realgar-containing preparations. The substrate was prepared via interfacial self-assembly of silver nanoparticles and modified with Ag + to enhance arsenic adsorption and improve detection sensitivity. Quantitative models were constructed based on the analysis of As3+ standard solutions ranging from 10 to 1000 ppb, utilizing five different chemometric methods. Amongst the algorithms, the SiPLS model produced a lower Root Mean Square Error (RMSE) of 9.125 and a coefficient of determination (R2) of 0.999, indicating a more accurate fit between predicted and measured values. Based on the established model, the leaching concentrations of As3+ were determined in realgar and four realgar-containing preparations (Children's Qizhen Pills, Children's Qiying Pills, Children's Huadu Powder, and Liushen Pills) after 12 h of aqueous immersion. The mean As3+ concentrations from three batches of realgar were 862, 948, and 807 ppb, with relative standard deviations (RSDs) of 4.9 %, 5.2 %, and 5.1 %, respectively. For the four prepasrations, the As3+ concentrations were 138, 96, 86, and 117 ppb, with RSDs of 7.1 %, 6.2 %, 8.3 %, and 8.5 %, respectively. This method is rapid, accurate, non-destructive, and easy to implement, showing potential for quantitative analysis and quality control of soluble arsenic in realgar and related formulations.