Chemometrics in Aerosol Analysisquantitative Analysis of Silica Dust Mixtures by Multivariate Calibration Applied to Infrared Spectroscopy

Dust mixtures of crystalline and amorphous silica cannot be quantified by infrared spectroscopic (i.r.) methods alone, due to a severe overlap of the primary spectral SiO₂ band in the most characteristic infrared region, i.e. 850–750 cm⁻¹. This problem has been solved with the use of multivariate calibration by partial least-squares regression (PLSR) applied to i.r. spectroscopy. The use of these techniques has provided a method for the simultaneous quantification of amorphous silica in mixtures with quartz and cristobalite. A calibration model was established by standard materials and the prediction error are 25,26 and 24 μg for the three compounds, respectively. Based on the synthetic mixtures analysed in this laboratory study, the present technique, i.e. multivariate calibration with PLSR applied to i.r., may offer a robust, rapid and low cost method for quantification of complex dust mixtures.