The application and development of chemometric methods for the analysis of plastic recycling and Mo(CO)(6) photolysis products.

摘要

Chemometrics is a collection of diverse methods that employ a variety of multivariate statistical methods, neural networks, and mathematical transforms to extract information from chemical data. Several chemometric techniques have been developed and applied to two chemical problems: the analysis of products from plastic recycling and Mo(CO){dollar}sb6{dollar} photolysis.; A catalyzed method for thermal cracking has been reported. The products for the thermal cracking of polyethylene are alkanes, monounsaturated alkenes (alkenes) and dienes. Three automated methods for monitoring the catalyzed cracking of polyethylene by gas chromatography/mass spectrometry were developed. The methods employ three different expert systems: a knowledge-based expert system, a univariate rule building expert system and a fuzzy multivariate expert system. The chemometric methods were evaluated for their ability to correctly classify products of polyethylene cracking by their degree of unsaturation. The two rule-building expert systems performed better than the knowledge-based expert system. The chemometric detectors were then used for mathematically resolving chromatograms of plastic cracking products.; In the second application, two-dimensional (2D)-correlation analysis was used to visualize the correlations in multivariate data by applying the complex cross-correlation function. It is shown that 2D-correlation analysis can be used both as a qualitative and quantitative tool. The previously reported range scaling method for correcting differences in intensities between signals should be avoided. Instead, the data channel should be scaled by its standard deviation before calculating the two-dimensional spectra.; The applicability of 2D-correlation spectroscopy to kinetic data is demonstrated. The photodecomposition of Mo(CO){dollar}sb6{dollar} was used as a model kinetic system. The effects of the steady-state and time-average methods for static spectrum removal were evaluated. The time-averaged method provided 2D-correlation spectra that were more easily interpreted. The influence of noise was evaluated with the synthetic data. The methods of static spectral removal are shown to be affected differently by the presence of noise. The time average method is shown to be more sensitive to noise. A Fourier filtering method was devised to reduce the effect of noise.