Iterative least-squares fit procedures for the identification of organic vapor mixtures by Fourier-transform infrared spectrophotometry.
Xiao-H; Levine-SP; D'Arcy-JB
Anal Chem 1989 Dec; 61(24):2708-2714
Iterative least squares fitting procedures for identifying the components of mixtures containing trace quantities of organic vapors by Fourier transform infrared spectrophotometry (FTIR) were described. The procedures assumed a linear baseline over the width of the measured spectral peak and utilized a library of reference FTIR spectra. The error variance in the absorbance of the suspected components was computed. From this, the estimated concentration and its 95% confidence interval (CI) was computed. Compounds whose CIs did not include zero were kept and formed the basis of a set used for the next iteration. This process was continued until all compounds thought to be present in the mixture were accounted for. Two basic approaches were used. The set reduction method (SRM) attempted to eliminate compounds from a mixture by determining the worst fit levels and comparing them with a library of reference spectra. The set building method (SBM) attempted to build a set of possible compounds by fitting the reference spectra one at a time to the sample spectrum. These methods were applied to three mixtures containing five to 11 compounds having concentrations of approximately 2 parts per million in air. The spectral regions examined were 2,900 to 3,200/centimeter (cm) in the carbon/hydrogen stretch region and 650 to 1,350/cm in the fingerprint region. Both the SRM and SBM procedures were applied using the entire spectral region (general window approach) or a portion of the spectral region specific for each compound (specific window approach). The sensitivities of the method were: SRM general window approach, 81%; SRM using a single iteration, 91%; SRM specific window approach, 91%; and SBM specific window approach, 94%. The specificity of these methods was 96, 92, 94, and 97%, respectively. The authors conclude that iterative least squares fit methods are very useful for interpreting the infrared spectra of mixtures.
NIOSH-Publication; NIOSH-Grant; Grants-other; Organic-vapors; Infrared-spectrophotometry; Chemical-composition; Mathematical-models; Chemical-analysis; Statistical-analysis; Industrial-hygiene
Environmental & Indust Health School of Public Health II 1420 Washington Heights Ann Arbor, MI 48109-2029
Other Occupational Concerns; Grants-other
University of Michigan at Ann Arbor, Ann Arbor, Michigan