Comprehensive two-dimensional gas chromatography and chemometrics for the analysis of complex mixtures.

摘要

The main goal of this dissertation is to enhance the analysis capabilities of comprehensive two-dimensional (2-D) gas chromatography (GC x GC) by applying chemometrics to extract component signals in the presence of significant interference, noise, or both. This is accomplished through three projects.; The first project quantifies the theoretical analysis enhancement provided by the use of the generalized rank annihilation method (GRAM) for the analysis of component signals, i.e., peaks, that are unresolved in GC x GC separations. Monte Carlo simulations, modeled after real GC x GC data, are used to determine the conditions where the use of GRAM results in the successful analysis of unresolved peaks. This information is then used on Monte Carlo simulations of 2-D chromatograms. Ultimately, it is determined that the use of GRAM increases the average number of analyzable peaks by a factor of two for 2-D chromatograms that are 67 percent occupied by randomly distributed peaks. The use of GRAM should increase the number of analyzable peaks for all forms of comprehensive 2-D separations.; The second project extends the use of GRAM analysis to the GC x GC separation of real world complex mixtures by the use of the standard addition method and bilinear data alignment. Standard addition and bilinear data alignment are used to correct peak width variations and retention time shifts that would otherwise negate the utility of GRAM. The use of GRAM, in part, reduces the separation time of three isomers in jet fuel by a factor of five compared to a single column GC separation. The use of bilinear data alignment improves GRAM quantification accuracy and precision by a factor of four. In addition, 2-D bilinear data alignment is introduced.; The last project demonstrates the signal enhancement provided by GRAM. It substantially improves the quantitative precision and accuracy of GC x GC compared to peak integration. In the case of a 2.7 part per million by mass propylbenzene sample, GRAM analysis is 2.6 times more precise and 4.2 times more accurate than integration. The GC x GC limit of detection is also lowered by a factor of three.