Numerical pattern recognition analysis of CO atmospheric simulation experiments

A technique entitled Hybrid Linear Pattern Analysis (HLPA), which represent s a combination of model-based and pattern recognition-based approaches to the analysis of spectroscopic data, is introduced and applied to the analys is of time-resolved infrared emission spectra of ground electronic state (X (1)Sigma(+)) CO obtained in atmospheric simulation experiments. The spectra are highly congested and consist of incompletely resolved, overlapping v' - v" = 1 emission bands from v' = 1 up to at least v' = 12. The analysis of the time dependence of the emission intensity in the various vibrational b ands had been stymied by a severe optical opacity effect in the v' = 1 --> 0 emission, which is difficult to simulate; thus, conventional least-square s fitting could not be used confidently to determine the time-dependent emi ssion intensity of this band, or that of at least three other emission band s that overlap strongly with it. The HLPA technique permits an alternate ap proach in which the v' = 1 --> 0 emission band is considered to be an unkno wn pattern that is identified by the Extended Cross Correlation (XCC) patte rn recognition technique (J. Chem. Phys. 1997, 107, 8349). The intensity pr ofiles of the other bands, however, can be predicted accurately based on th e experimental parameters, and this knowledge is used in conjunction with t he results of the XCC to determine the time dependence of all of the vibrat ional bands, and the intensity profile of the v = 1 --> 0 emission band.