Application of Multivariate Calibration Techniques to Quantitative Analysis of Bandpass-Filtered Fourier Transform Infrared Interferogram Data

Multivariate calibration models are developed that allow quantitative analysis of short segments of Fourier transform infrared (FT-IR) interferogram data. Before the interferogram segments are submitted to partial least-squares (PLS) regression analysis, a bandpass digital filter is applied to isolate a narrow range of frequencies that correspond to an absorption band of the target analyte. This adds frequency selectivity to the analysis, thereby overcoming the principal obstacle to the direct use of interferogram data for quantitative analysis. With the optimization of the frequency response function of the filter, as well as the position and length of the interferogram segment employed, calibration models are developed that compare well with those computed with conventional absorbance spectra. This methodology is demonstrated by developing calibration models for determining glucose in an aqueous buffer matrix over the physiologically relevant concentration range of 1–20 mM. Through the use of a time-domain filter designed to isolate the modulated interferogram frequencies corresponding to the glucose C–H combination band at 4400 cm⁻¹, a three-factor PLS calibration model is computed on the basis of interferogram points 601–850. This model is characterized by standard errors of calibration (SEC) and prediction (SEP) of 0.3311 and 0.6950 mM, respectively. The best model obtained in a thorough analysis of the corresponding absorbance spectra was also based on three PLS factors. This model was characterized by values of SEC and SEP of 0.2396 and 0.6115, respectively. In addition to achieving similar calibration and prediction results to the spectral-based model, the interferogram-based method has the advantage of requiring no background measurement of the sample matrix. Furthermore, since the analysis is based on only a 250-point segment of the interferogram, a reduction in the instrumentation and data collection requirements is realized.