GC/FTIR data have bean used to construct functional group-specific chromatograms. The classification of the GC peaks is based on the interferometric data. Because no Fourier transforms are involved, the algorithm is computationally fast enough to be performed on-line during data collection. The Gram-Schmidt reconstruction was used to locate the eluting compounds in the GC data, and pattern recognition techniques were used to classify the compounds. A linear learning machine and composite segment were compared, and the composite segment reconstruction was found to be superior.
A method is presented for the direct analysis of interferometric data from gas chromatography Fourier transform infrared spectroscopy (GC/FTIR). A synthetic interferogram is initially produced which represents the characteristic absorption features of a particular functional group or compound class. A zero displacement correlation is performed between this test interferogram and each sample interferogram from the GC data. The presence of the desired functionality in the GC effluent is indicated by a small value of the resulting cumulative sum. A “correlogram” which emulates the response from a chemically specific GC detector is obtained by plotting the cumulative sum from each sample correlation. Synthetic interferograms representing infrared absorption bands which are truly specific for a particular functionality yield the best results.
The combined use of gas chromatography and Fourier transform infrared spectroscopy (GC/FTIR) provides a powerful approach to the identification of complex mixtures. The capability of this method is increased when the retention information implicit in the experiment is utilized in structure elucidation. Two methods employing deuterated retention markers are presented for constructing separate chromatograms of both the sample components and the set of retention markers from a single GC/FTIR experiment; this allows accurate calculation of retention index values.
Various methods are described for the reconstruction of GC/ IR chromatograms from single scan interferometric data. The methods are compared on the basis of signal-to-noise estimate calculations and on the basis of estimates of the total number of computations required per interferogram processed. Some improvements over previously described methods have been devised. It has been concluded that the use of an improved form of the Gram-Schmidt orthogonalization reconstruction or of a Euclidean distance reconstruction results in optimal signal-to-noise with respect to computation time required.
