Robust chemical identification of drugs and explosives in seconds with portable NIR spectroscopy, data science, and forensic expert knowledge

Within law enforcement, homeland security, border patrols, and customs, there is an ongoing need for on-scene chemical identification of drugs of abuse and explosives. Ideally, the analysis is rapid, non-invasive, easy to conduct, and highly robust (i.e. no false positive or false negative test results). Of the many mobile, portable, and hand-held analysis technologies available, NIR (near infrared) spectroscopy offers many attractive features, including speed of analysis, miniaturization, and non-invasive analysis. Currently, portable technologies for chemical identification of drugs and explosives are considered to be of an indicative nature only requiring confirmation in the forensic laboratory with more selective and robust techniques like GC-MS. A true paradigm shift would be established if portable methods would have the robustness and selectivity to unequivocally identify common drugs of abuse and explosives and generate forensic evidence in seconds that would also be admissible in court. In a collaboration between the University of Amsterdam, the Dutch Police, the Dutch company TIPb, and several other partners, a NIR based platform has been developed with the aim of flawless identification of frequently occurring drugs of abuse and organic and inorganic energetic materials. The spectroscopic data is generated in seconds with a miniaturized MEMS NIR sensor that records a highly reproducible reflectance spectrum (1350-2550 nm) of powders and pills. A dedicated data analysis strategy involves compact reference libraries with pure compound spectra, data pre-processing, candidate selection using Linear Discriminant Analysis (LDA), and mixture composition analysis using Net Analyte Signal (NAS). By applying (expert) decision rules and quality criteria, false positive outcomes can be eliminated at the cost of a limited number of inconclusive outcomes. False negative rates can be kept low but can never be reduced to zero because of sensitivity limits.