We report the use of a low-cost dispersive Raman instrument with charge-coupled-device (CCD) detection, near-infrared (near-IR) diode laser excitation, and remote fiber-optic sampling to analyze mock petroleum samples which contain high benzene, toluene, ethylbenzene, and xylene (BTEX) concentrations. Partial least-squares regression (PLSR) analysis is used to correlate the individual xylene isomer concentrations to the Raman spectral signal without the use of an internal standard. The resulting PLSR model is used to predict the concentration of individual xylene isomers, and it is found that, at a 95% confidence level, samples containing between ~1.5 and 15% xylene isomer can be predicted to better than ±0.1% for <i>meta</i>- and <i>para</i>-xylene, and to ±0.15% for <i>ortho</i>-xylene. The use of PLS model leverage plots provides a facile statistical method by which to identify Raman spectra which involve diode laser mode hops or significant fiber backscatter.
John B. Cooper, Philip E. Flecher, Thomas M. Vess, and William T. Welch, "Remote Fiber-Optic Raman Analysis of Xylene Isomers in Mock Petroleum Fuels Using a Low-Cost Dispersive Instrument and Partial Least-Squares Regression Analysis," Appl. Spectrosc. 49, 586-592 (1995)