Optimized Data Analysis Algorithm for On-Site Chemical Identification Using a Hand-Held Attenuated Total Reflection Fourier Transform Infrared (ATR FT-IR) Spectrometer. Issue 12 (December 2013)

• Record Type: Journal Article
• Authors: Ron, Izhar; Zaltsman, Amalia; Kendler, Shai
• Is Part Of: Applied spectroscopy. Volume 67:Issue 12(2013)
• Page Start: 1395
• Page End: 1400
• Publication Date: 2013-12
• Subjects: Attenuated total reflection Fourier transform infrared -- ATR FT-IR -- Chemical identification -- Receiver operating characteristics; Spectrum analysis -- Periodicals; 543.505
• DOI: 10.1366/12-06951
• Languages: English
• ISSNs: 0003-7028

Abstract

On-site identification of organic compounds in the presence of interfering materials using a field-portable attenuated total reflection Fourier transform infrared (ATR FT-IR) spectrometer is presented. Identification is based on an algorithm that compares the analyte's infrared absorption spectrum with the reference spectra. The comparison is performed at several predetermined frequencies, and a similarity value (distance) between the measured and the reference spectra is calculated either at each frequency individually, or, alternatively, the average distance for all frequencies is calculated. The examined frequencies are selected to give the best contrast between the target materials of interest. In this study, the algorithm was optimized to identify three common chemical warfare agents (CWAs): O -ethyl S-[2-(diisopropylamino)ethyl] methylphosphonothioic acid (VX), sarin (GB), and sulfur mustard (bis(2-chloroethyl) sulfide) (HD), in the presence of field-related interfering materials (fuels, water, and dust). Receiver operating characteristics analysis was performed in order to determine the probabilities for detection (PD) and for false alerts (PF). Challenging the algorithm with a set of data that contains mixtures of CWAs and interfering materials resulted in PD of 90% and PF of 0%, 0%, and 1% for VX, GB, and HD, respectively, using the average distance approach, which was found to be much more effective than analyzing each frequency individually. This finding was validated for all possible combinations of 2–7 peaks per material. It is suggested that this algorithm provides a reliable mean for the identification of a predetermined set of target analytes and interfering materials.