OCR Text |
Show OH AND TEMPERATURE DETERMINATIONS O H is an important flame species, a key member of the O-H-OH radical pool that controls flame speed and many other primary aspects of combustion. O H measurements have been standard in our laboratory and others for many years, and w e have developed O H LIF into a sensitive thermometer for low pressure flames. 15 Rotational temperatures in O H are determined within ±30K in flames such as these. The variation of temperature with height above the burner is then used as input to the flame code, obviating any need for quantitative estimation of heat losses in the burner system. At the same time we make OH concentration profiles. There is sufficient OH, approaching 1 % mole fraction, in our low pressure flames that w e can calibrate directly by absorption experiments. The O H profile is then normalized to this value, a simpler technique than the calibration used for C H . Agreement with predicted absolute values from GRI-Mech is excellent, overlapping with experimental noise. Profiles are shown in Ref. 13. However, there is a disturbing tendency of the model to predict the leading edge of the rising O H profile to appear too far from the burner, regardless of flame stoichiometry. This is a different pattern than observed for C H , suggesting a different cause. W e do not yet understand the discrepancy. It may be inadequacies in the transport included in Premix, or the fact that the flame is not truly one dimensional. NO MEASUREMENTS FOR PROMPT-NO Sensitivity analysis is a useful way to view flame chemistry codes and comparisons with experiment. Sensitivity in the chemical mechanism is the change in some predicted quantity, e.g., N O concentration, with respect to some input variable, e.g., the rate coefficient for the C H + N 2 reaction. The sensitivity of prompt N O to this rate coefficient is nearly unity, that is, there is a one-to-one mapping of our ability to predict this pollutant. With good predictions of C H by GRI-Mech in the Lean and Standard Flames, and known absolute concentrations in the Rich Flames, w e can examine the value of this rate coefficient in the present version of GRI-Mech. The NO measurements are made using the A-X system near 226 nm, at a fixed height of 2 c m in the flame. In this case w e can calibrate the system using standard addition of the stable N O gas (diluted in nitrogen). W e use values similar to those encountered in the flame, replacing some of the nitrogen flow with a gas containing N O diluted in nitrogen. The actual flame concentrations are not exactly the same as measured by the flowmeters, due to dilution in the flame gases; this reduces the apparent amounts by 3 to 9%, depending on flame. Reburn in the rich flames (see below) removes another 1 % of the calibration N O . These small corrections are made via model predictions, using iterative procedures. Measured native concentrations of NO in the five flames are given in Table 1, with uncertainties representing 1 standard deviation of replicate measurements. They are compared with GRI-Mech predictions for prompt N O , which constitutes 5 7 % , 6 4 % , and 9 1 % of the total N O in our Lean, Standard, and Rich flames, respectively. In all cases the mechanism significandy underpredicts the prompt N O , in contrast to its ability to predict C H . This suggests the value of the rate coefficient for the C H + N 2 reaction in the current version of GRI-Mech is too low. Sensitivity analysis shows this is the only reaction which can be altered without disturbing the good agreement for the measured C H concentration. W h e n that value is multiplied by the factor listed as k'/k in Table 1, the model and experiment match. (This is not simply the ratio of measured to modeled N O , because some of the N O is thermal and the sensitivity coefficients vary from flame to flame.) A more thorough examination 16 using values of the C H concentration measured in the rich flame makes the agreement yet better. That 6 |