Coherent Anti-Stokes Raman Spectroscopy CARS Measurements in a Laminar Flame Coal Combustor

Page 11

1.0 ~------------------------------------------------~ 0.8 ~ 0.6 en c:: (l) E ~ 0.4 ~ ¢) a: 0.2 0.0 1800 1900 2000 2100 2200 2300 2400 Raman Shift (em -1 ) Figure 7. Example of spectra showing different levels of enhanced nonresonant signal. The lower spectrum was taken in a gas flame and contains no noresonant enhancement, while the two upper spectra were taken in a coal flame and show appreciable nonresonant enhancement. investigated in this study were obtained by fitting the nitrogen spectra to a computer generated nitrogen library for a 60% nitrogen mole fraction. Spectra were normalized by a Gaussian fit of the Stokes (dye) laser profile and then fit on a square root of intensity basis. The dye profile was fit with a Gaussian curve in order to smooth over the effect of a small nitrogen Signal that could not be completely removed. This small nitrogen signal in the dye profile was most likely due to difficulties in completely voiding the reactor of nitrogen while the dye profile was being taken, and from a small CARS signal that might have been generated in the reactor, but outside the diagnostic volume. Had the nitrogen-adulterated dye profile been used to normalize the data, unacceptable errors would have been introduced into the resutts. This artificial Gaussian dye profile was therefore used to normalize all of the experimental spectra. The Sandia fitting program was modified to allow subtractk>n of a variable fraction of the Stokes laser profile in order to fit spectra from coal-seeded flames which contained additional background from enhancement of the nonresonant signal. The spectra were fit over the range 2200-2335 cm-1 although most of the resonant transitions with significant strength at combustion temperatures lie between 2250- 2335 cm-1. Extending the fitting range to include 2200-2250 cm-1 allowed the appropriate level of background subtraction to be determined, thus making temperature fitting more accurate. Figure 8 illustrates this point by showing the same spectrum fitted with different levels of background subtracted. With the proper subtractk>n level, residuals were small and constant over the fitting range while fits with 11