Mixture analysis using reduced resolution direct sampling mass spectrometry and chemometrics

The need for field portable instrumentation has driven significant research into the development of small, portable chemical analysis systems. Direct sampling mass spectrometry (DSMS) techniques show potential for field applications due to the strong characteristics of mass spectrometric techniques. A characteristic of DSMS analysis is the simultaneous introduction of sample analytes into the mass spectrometer during analysis, which may pose problems if the samples contain analytes with overlapping mass spectra. Multivariate chemometric analysis techniques were shown to improve quantitative analysis due to the technique's ability to detect and utilize subtle differences in the mass spectra. The multivariate analysis techniques of partial least squares regression (PLS), trilinear partial least squares regression (Tri-PLS), and parallel factor analysis (PARAFAC) were shown to be useful during quantitative analysis of mixtures of toluene, ethyl benzene and cumene. The temporal profile differences of the three analytes were utilized by PARAFAC to provide improved quantitation results in the presence of unmodeled interferences, while PLS and Tri-PLS exhibited increased errors in quantitation. Recognizing the need for smaller mass spectrometers, researchers are currently developing miniature mass spectrometers that may be utilized in field portable instruments. A common characteristic of the miniature mass spectrometers is the reduction in mass resolution, which results in increased spectral overlap and potential problems during DSMS analysis. DSMS data from four data sets of analytes with overlapping spectra were synthetically reduced in resolution and quantitative analysis of the reduced resolution data showed that, in general, the quantitative performance did not significantly degrade with resolution, and the optimum quantitative analysis technique (i.e. PLS or Tri-PLS) was analyte and mixture dependant. Qualitative discriminant analysis of a mixture of seven halogenated solvents, at reduced resolutions, showed that the linear techniques of discriminant PLS and discriminant Tri-PLS outperformed the nonlinear back propagation neural network algorithm. In general, discriminant analysis performance did not significantly degrade as the resolution was reduced, and the optimum analysis technique was analyte and resolution dependant. The lack of significant degradation in quantitative and discriminant analysis performance shows that miniature mass spectrometers will be useful for future DSMS based portable instruments.