<b>Rapid quantitative analysis of methamphetamine by portable near-infrared </b><b>spectroscopy and PLS modelling</b>

For rapid on-site detection of methamphetamine, a portable NIRspectrometer was used to model and analyze the collected samples.Transmission spectra (908-1676 nm) of 359 methamphetamine crystal samples and 187 methamphetamine tablet samples were collected, first-order derivatives and the S-G smoothing method were applied for spectral preprocessing. The partial least squares (PLS) method was used for modeling, and the quantitative models for methamphetamine crystals were 2.10%, 2.12% and 2.19% for RMSECV, RMSEV and RMSEP, respectively; in contrast, RMSECV, RMSEV, and RMSEP of methamphetamine tablets were 0.78%, 0.80%, and 0.82%, respectively. The results of the internal cross-validation and the methodology review showsthat the measurements of the modelled results are consistent with the true values, and that the model has good stability and can be used for accurate field measurements. Therefore, real-time sample analysis using portable NIR technology combined with chemometric modeling in crime scene investigation has become an inexpensive, rapid and reproducible analytical method, which helps the police to determine the quality of drugs in a timely manner for case analysis.为实现甲基苯丙胺（methamphetamine）的现场快速检测，本研究采用便携式近红外（Near Infrared, NIR）光谱仪对采集的样本开展建模与分析工作。共采集359份甲基苯丙胺晶体样本与187份甲基苯丙胺片剂样本的透射光谱（波段范围908~1676 nm），并通过一阶导数法与萨维茨基-戈莱（Savitzky-Golay, S-G）平滑法完成光谱预处理。本研究采用偏最小二乘（Partial Least Squares, PLS）法构建定量模型：甲基苯丙胺晶体定量模型的校正交叉验证均方根误差（Root Mean Square Error of Cross Validation, RMSECV）、验证集均方根误差（Root Mean Square Error of Validation, RMSEV）与预测集均方根误差（Root Mean Square Error of Prediction, RMSEP）分别为2.10%、2.12%与2.19%；与之相比，甲基苯丙胺片剂样本的RMSECV、RMSEV及RMSEP分别为0.78%、0.80%与0.82%。内部交叉验证与方法学验证结果显示，模型预测结果与真实值高度吻合，且模型具备良好的稳定性，可用于精准的现场检测。综上，将便携式近红外技术结合化学计量学建模应用于犯罪现场调查中的实时样本分析，已成为一种低成本、快速且可复现的分析手段，能够协助警方及时判定涉案毒品的质量以支撑案件研判。