Fast Identification of Hazardous Liquids Based on Raman Spectroscopy

Fast and accurate identification of unknown hazardous fluids are of pivotal interest in public security and safety. Raman spectroscopy is a fast and sensitive non-contacting measurement technology. Its virtues have it has become one of the important technologies in the public security field in recent years. In this study, the Raman spectra of forty-two dangerous and common liquids were measured: five chemical warfare agents (including sarin, soman, tabun, VX, and mustard gas), and their fifteen precursors, hydrolysates ( including dimethyl hydrogen phosphite, trimethyl phosphite, triethyl phosphite, ethyl methylphosphonochloridoate, methylphosphonic dichloride, methylphosphonic difluoride, chlorosarin, bis ( diethylamino ) phosphoryl chloride, 2-(diethylamino) ethanethiol, thiodiglycol, isopropyl alcohol, 3, 3-dimethyl-2-butanol, methylphosphonic acid, isopropyl methylphosphonate, and pinacolyl methylphosphonate), five chemical warfare agents simulants ( including trimethyl phosphate, triethyl phosphate, tributyl phosphate, dimethyl methylphosphonate, and diisopropyl methylphosphonate) , fifteen toxic industrial compounds (including o-xylene, m-xylene, anisole, chlorobenzene, ethyl acetate, vinyl acetate, benzyl acetate, methanol, ethanol, 1-butanol, acetonitrile, acetone, hexane, 1 , 1 ,1-trichloroethane, and carbon tetrachloride), gasoline, and water. Raman spectroscopy detection method for these compounds using a portable Raman spectrometer equipped with a 785 nm excitation laser was developed to obtain high SNR scattering spectrum data. Structural assignments to Raman bands observed in the spectrum were also proposed. Six pattern recognition algorithms, including linear discriminant analysis ( LDA) , quadratic discriminant analysis ( QDA) k-nearest neighbor ( kNN) , naive bayesian ( NB) , classification tree (CT) , and support vector machine (SVM) were studied to identify and classify Raman spectrum data. The results show that support vector machine and linear discriminant analysis model could realize the fast identification with a high recognition accuracy rate of 100%. However, considering the influence of non-standard spectrum, instrument conditions, and changes in the external environment on support vector machine recognition results, the linear discriminant analysis model seemed superior in identifying the aforementioned dangerous liquids. Meanwhile, all the testing process can be completed within 1 similar to 2 minutes without loss of samples. It can be concluded from this study that the combination of Raman spectroscopy with fingerprint characteristics and pattern recognition algorithm can be used for rapid screening of unknown compounds. Moreover, this method provides solutions for timely detection of customs clearance, guarantee of logistics security, and emergency response to chemical terrorist incidents.