SERS Detection and Efficient Identification of Heroin and Its Metabolites Based on Au/SiO2 Composite Nanosphere Array

The spread of high-risk opioid heroin has caused severe harm to national stability, social economy, and people's life and property safety. Efficient and accurate detection/identification methods for heroin and its metabolites are significant in combating drug crimes, dealing with drug-related cases and anti-drug campaigns. Surface-enhanced Raman spectroscopy (SERS) has the merits of fast detection speed, simple operation, high sensitivity, fingerprint identification and non-destructivity, which can realize efficient and portable detection of drugs. If combined with pattern recognition, it can improve the efficiency of data processing and avoid the occurrence of human misjudgment. Thereby the purpose of automatic and accurate classification and identification can be achieved. In this work, to achieve sensitive detection and efficient identification of trace heroin and its metabolites in solution, a method combining SERS measurement based on Au-coated SiO2 composite nanosphere array (Au/SiO2 NSA) and pattern recognition is proposed. Firstly, gas-liquid interface self-assembly and magnetron sputtering deposition prepare Au/SiO2 NSA with good SERS activity and signal reproducibility. Employing such an array as SERS substrate (chip), combined with a portable Raman spectrometer, the high-efficiency detection of heroin and its main active metabolites (6-monoacetylmorphine (6-MAM) and morphine) in water solution is successfully achieved with the detection limit of 10(-4) mg.mL(-1). Next, to perform qualitative/quantitative identification of heroin and its metabolites, SERS spectral data are processed via hierarchical cluster analysis (HCA), principal component analysis (PCA) and support vector machine (SVM). When classifying heroin, 6-MAM and morphine, on the foundation of the accurate classification of them by both HCA and PCA, the PCA-SVM models based on radial basis function, linear kernel function, polynomial kernel function or sigmoid kernel function all can 100% qualitatively identify them. When adopting the PCA-SVM model to analyze heroin. 6-MAN and morphine quantitatively, the accuracy of quantitatively distinguishing different concentrations of heroin can reach 90.1% using the SVM model based on radial basis function. Via the SVM model of the linear kernel function, the accuracies in discriminating different concentrations of 6-MAM and morphine are 84.8% and 70.2%, respectively. This work provides a high-quality substrate (chip) with practical value for sensitive detection and accurate identification based on SERS but also puts forward a feasible approach for efficient classification and identification of heroin and its metabolites.