- Email: [email protected]
N2 flames
Quantitative LIF Measurements and Modeling of Nitric Oxide in High-Pressure C,H,/O,/N, Flames JOHN R REISEL* and NORMAND M. LAURENDEAU Flame Diagnostics Laboratory, School of Mechanical Engineering, Purdue Universiry, West Lafayette, IN 47907-1288
We have obtained quantitative LIF measurements of NO concentration in the post-flame zone of a series of flat, laminar, premixed C,H,/O,/N, flames at pressures ranging from 1 to 11.9 atm. The internal consistency of the data was verified by demonstrating good agreement between measurements based on two different NO excitation lines. The temperatures of the flames were 1600-1850 K, indicating that most of the NO produced in the flames is prompt-NO. The results show that the equivalence ratio corresponding to the peak NO concentration at a given pressure shifts towards leaner conditions with increasing pressure. We have also modeled the flames using two chemical kinetics schemes: (1) the Glarborg-Miller-Kee mechanism as modified by Drake and Blint (GMK-DB), and (2) the Miller-Melius hydrocarbon mechanism, combined with the nitrogen kinetics of Drake and Blint (MIME-DB). Both models are qualitatively inaccurate with respect to the measured pressure shift, particularly at pressures below 6.1 atm. Both models also tend to underpredict the NO concentrations, although the GMK-DB model is acceptable in many cases.
INTRODUCTION High-pressure combustion applications, such as gas-turbine engines, are a major source of nitric oxide (NO) emissions. As the environmental problems caused by high NO emissions grow, it has become imperative to reduce NO emissions from combustion processes. The achievement of this goal by combustion designers requires, among other things, a thorough understanding of the chemical kinetics involved in the production of NO at high pressure. Such understanding, in turn, mandates accurate in situ measurements of NO concentration. Quantitative measurements of NO concentration can be obtained using both physical techniques, such as probe-sampling [l-31, and optical techniques. such as laser-induced fluorescence (LIF) [4-91. Probe-sampling combined with chemiluminescent detection is advantageous since it possesses a lower detection limit, is easier to use, and is less expensive than laser-based methods. However, a physical probe can disrupt the flow field, potentially altering the concentrations of radical species; moreover, such probes may not be able to
*Corresponding author. FLAME 101: 141-152 (19951 Copyright 0 1995 by The Combustion Institute Published by Elsevier Science Inc.
withstand the harsh conditions of practical combustion environments [lo]. These disadvantages can be overcome by employing optical techniques. Optical procedures allow for remote sensing of numerous species in a variety of environments. Unlike sampling probes, optical methods generally do not alter the combustion process; in addition, many combustors are more readily adaptable to optical access than to physical sampling probes. Finally, precise spatial resolution is achievable through the use of optical methods. Previously, we have demonstrated the feasibility of making quantitative LIF measurements of NO in C,H,/O,/N, flames up to 14.6 atm [ll]. These results showed that the equivalence ratio corresponding to the peak [NO] at a given pressure shifts towards leaner conditions with increasing pressure. In addition, the measured pressure shift was successfully modeled using two chemical kinetics schemes [12, 131; the model from Drake and Blint [13] also provided good quantitative agreement with the measured NO concentrations. In this paper, we extend the previous work by presenting LIF measurements of NO in flat, laminar, C,H,/O,/N, flames at pressures up to 11.9 atm. By comparing NO measurements obtained from two different excitation lines, we verify the internal consistency of the LIF
COMBUSTIONAND
OOlO-2180/95/$9.50 SSDI 0010-2180(94)00195-X
142
J. R. REISEL AND N. M. IAURENDEAU FROM LASER
procedure. We also investigate whether the shifting behavior is limited to paraffinic fuels (like ethane), or if, as anticipated, it is a more universal phenomenon. In addition, by modeling the flames using two different reaction mechanisms, we evaluate the ability of current chemical kinetics schemes to predict [NO] in high-pressure C,H,/O,/N, flames. One of these is the Glarborg-Miller-Kee model as modified by Drake and Blint (GMK-DB) [13, 141; the other is the hydrocarbon scheme of Miller and Melius [15] combined with the nitrogen kinetics of the GMK-DB model (MIME-DB). Comparisons between the predicted results and the LIF measurements provide a test of the ability of these mechanisms to predict the effects of high pressure on NO emissions from premixed ethylene flames. LASER-INDUCED METHODOLOGY
FLUORESCENCE i.;q
Figure 1 shows a schematic diagram of the experimental apparatus used in the LIF measurements. Excitation of NO was achieved via one of two spectral lines: the Q,(26.5) line and the Rr(18.5) line, both within the y(O,O) band of NO. The Q,(26.5) line (h = 225.6 nm> was chosen because (1) its Boltzmann fraction is insensitive to temperature fluctuations over the range of temperatures of our flames, and (2) other species, such as 0,, do not interfere spectrally with this NO absorption line [9]. The Rr(18.5) line (A = 225.5 nm) was chosen as an alternative transition based on its separation from other spectral lines as well as the lack of spectral interference from other species. The excitation wavelengths were generated by employing the second harmonic (A = 532 nm> of a Quanta-Ray DCR3G Nd:YAG laser to pump a PDL-2 dye laser, which produces laser radiation at N 572 nm. The output of the dye laser was frequency doubled in a Quanta-Ray Wavelength Extender (WEX-1) and the doubled-dye beam was mixed with the first harmonic of the Nd:YAG laser, producing - 1 mJ/pulse at - 225.5 nm for these measurements. After leaving the laser system, the beam was directed towards a 2.5-cm-diameter, watercooled, sintered-bronze McKenna flat-flame burner. The burner is located inside the high-
EiE
I-.-.-._._.-----.I CoMPUTER 1
Fig. 1. Schematic diagram of experimental apparatus: A, trigger photodiode; B, G, beam-splitter; C, lOOO-mm focal-length lens; D, K, beam steering assembly; E, aperture; F, pressure vessel; H, beam dump; I, power-monitoring photodiode; J, 200-mm focal length lens; L, 300-mm focal-length lens; M, 1/2-m monochromator; N, PMT.
facility described by pressure combustion Carter et al. [16]. The pressure vessel has four optical ports, two of which are used for directing the laser beam through the facility. The spot size produced by the optical arrangement is - 250 pm. After exiting the vessel, the beam passed through a fused silica plate which directed a portion of the beam towards a UVsensitive photodiode. This photodiode was employed to monitor the beam energy, which was required for normalization of the fluorescence signal. For fluorescence detection, we made use of an optical port perpendicular to the laser entrance and exit ports. The fluorescence was focused on the entrance slit of a 4-m monochromator. The detector was an RCA lP28B photomultiplier tube specially wired for temporal resolution of the fluorescence signal [171. The broadband fluorescence signal was de-
LIF MEASUREMENTS
OF NO IN C,H,/O,/N,
tected at N 236 nm, corresponding to N 3 nm of the y(O,l) band of NO. To account for pressure shifting of the Q,(26.5) and the R,(18.5) absorption lines, the laser and collection optics were tuned for the maximum NO fluorescence signal at each pressure. A 500-ps window at the peak of the fluorescence pulse was sampled using a Stanford Research Systems SR255 fast sampler. The image of the entrance slit over the burner was 80 pm X 6.7 mm. Each data point was averaged over 600 laser shots. When performing a linear LIF measurement, one must be concerned with the effects of both laser power fluctuations and quenching variations on the fluorescence signal. One way to avoid such effects is to employ laser-saturated fluorescence &SF), which has been previously applied with broadband detection to NO at atmospheric pressure [9]. However, as discussed by Reisel and Laurendeau 1111, saturated conditions are difficult to maintain for NO measurements above 3 atm. For linear fluorescence measurements, corrections for laser power fluctuations can be made by normalizing the fluorescence signal using the measured laser power. Quenching variations could be handled in a similar manner; however, the measurement of quenching rate coefficients is not a trivial task. By comparing measurements obtained using both linear and saturated LIF, we have previously found that quenching variations at a given pressure are not significant for our range of flame conditions [9]. To confirm that this result is also true for the C,H,/O,/ N, flames of this study, we have calculated quenching rate coefficients in the post-flame zone using equilibrium concentrations and the quenching cross-sections from Drake and Ratcliffe [181. The quenching rate coefficient per unit pressure, Q/P, can be calculated from
where k is Boltzmann’s constant, T is the temperature, Xi is the mole fraction of quenching species i, a, is the quenching crosssection of NO with species i, and pi is the reduced mass between species i and NO. Only
FLAMES
143
the species studied by Drake and Ratcliffe [18] were considered for the calculations (N,, O,, H,O, CO,, CO, C,H,, H,, NO, H, OH, and 0); these include the dominant quenching species in the post-flame zone. The quenching variation proves to be relatively small (5 15%) over most of the flame conditions for which linear fluorescence is required (P 2 6.1 atm) in the C,H,/O,/N, flames. A few of the flames with $J 2 1.4 have quenching rate coefficients which vary from the calibration flame by N 20%; however, even this difference is less than the uncertainty in the measurements. Greater differences would arise if the linear fluorescence measurements were extended to flames having larger equivalence ratios (4 = 1.8); however, measurements at these equivalence ratios were only performed at lower pressures, for which LSF was employed. NITRIC OXIDE CHEMICAL KINETICS NO is produced through three main chemical pathways [13]: (1) the Zeldovich, or thermalNO, mechanism [193, (2) the N,O-intermediate mechanism [20,21], and (3) the prompt-NO mechanism [22]. The amount of NO formed through each pathway depends on the temperature, pressure, and equivalence ratio of the flame. The lower temperatures and rich flames of this investigation favor the prompt-NO mechanism. The controlling reaction for prompt-NO is thought to be [12] CH + N, * HCN + N.
(RI)
Since reaction Rl demonstrates a hydrocarbon linkage, the amount of prompt-NO must depend on the type of fuel. In fact, Bachmaier et al. 1231 demonstrated that the amount of prompt-NO formed at a given equivalence ratio varies with fuel type. Most of the experimental flames in this study were investigated through computer modeling. The modeling of the chemical kinetics was performed using the Sandia steady, laminar, one-dimensional, premixed flame code [24]. In addition, the CHEMKIN-II computer program library 1251 was used to process the reaction mechanism into a form which is appropriate
144
J. R. REISEL AND N. M. LAURENDEAU
for use by the flame code. The thermodynamic and transport properties, required by the Sandia flame code for calculation of the species concentration profiles, were provided by a thermodynamic property data base [26] and a transport property data base [27]. Previously, Reisel et al. [9] found that most of the NO produced in atmospheric-pressure versions of similar low-temperature C,H,/ 0,/N, flames is produced in the flamefront rather than in the post-flame region. Due to the similar temperatures of the C,H,/ O,/ N, flames of this study, we expect to again have little post-flame thermal-NO production. As discussed by Reisel and Laurendeau [ill, modeling the flamefront NO in such flames generally requires consideration of all three reaction mechanisms. In particular, while the promptNO mechanism does not contribute to NO production in the postflame zone (where our measurements are taken), its inclusion remains necessary since most of the NO produced in the flamefront is related to the prompt-NO pathway. The primary goal of the kinetics modeling effort is to assess the feasibility of using current kinetic schemes to simulate the effects of pressure on NO concentration. To do this, we must employ the coupled species-energy equations to determine the relevant temperature profiles as the rapid temperature rise in the thin flamefront is not easily measured at high pressures. A burner surface temperature of 300 K is used as a boundary condition to mimic heat loss to the burner. While the calculated temperature profiles will not agree precisely with the actual temperature profiles (leading to some potential errors in quantitative agreement), the calculated postflame temperatures agree sufficiently with the measured postflame temperatures to allow for accurate assessment of the pressure trends, and for a reasonable assessment of quantitative capabilities. Two mechanisms have been used as the chemical kinetics scheme for the computer model. Both are based on the comprehensive mechanism assembled by Glarborget al. 1141. The first mechanism (GMK-DB) is taken from the set of elementary reactions listed by Drake and Blint [13]. This reaction mechanism considers 49 species and over 200 reactions. Drake
and Blint [13] adopted most of the reaction mechanism from Glarborg et al. [141; however, they made a few modifications. These include the introduction of pressure dependency into four unimolecular reactions, the addition of a C,H, .reaction mechanism, and the introduction of rate parameters for reaction (Rl) based on measurements in a high temperature shock tube [28]. The rate parameters at pressures of 1.0, 3.05, 6.1, and 9.15 atm are given by Drake et al. [3]. The rate parameters at 11.9 atm were provided by Drake and Blint [29]. The second mechanism (MIME-DB) is a combination of the hydrocarbon mechanism assembled by Miller and Melius [15] and the nitrogen kinetics of the GMK-DB model. The Miller-Melius mechanism was designed to model rich combustion of aliphatic fuels such as ethylene and acetylene. The mechanism contains most of the Miller-Bowman 1121 mechanism for modeling small hydrocarbon species, and adds many larger hydrocarbon compounds. Due to the size of this mechanism, the coupled species-energy equations were not solved for the complete mechanism. Rather, only the hydrocarbon kinetics were solved using the coupled species-energy solution; the entire mechanism was then solved using the temperature profile from this partial solution. A sample case was run with the full mechanism using the energy solution, and the temperature profile from the reduced and full mechanisms were found to be nearly identical (differing by less than 5 K>. RESULTS AND DISCUSSION Using the experimental apparatus described above, quantitative LIF measurements of NO were performed in flat, laminar, high-pressure C,H,/O,/N, flames. Data were obtained at five pressures over a range of 1.0-11.9 atm. The purity of the C,H, was greater than 99.9%. The temperatures of these flames, as measured with radiation-corrected Pt-Pt/lO%Rh thermocouples in the post-flame region, ranged from 1600 to 1850 K (precision f30 K, accuracy f75 K). A few of the temperature measurements are listed in Table 1. All of the flames had a dilution ratio (~r,@02) of 3.1. The total flow rates were held constant at each
LIF MEASUREMENTS
OF NO IN C,H,/O,/N,
TABLE 1 Radiation-Corrected Thermocouple Measurements (K) of Selected C,H,/O,/N, Flames” Pressure (atm) 4
1.00
3.05
6.10
9.15
11.9
0.90 1.20 1.35 1.40
1625 1810 1840
1650 1785
1640 1740 1740 -
1635 1710 1725
1645 1690 1695
1750
’ The precision is + 30 K and the accuracy is f 75 K. indicates no measurement at this condition.
pressure, and were as follows: 3.50 slpm (1.0 atm), 6.18 slpm (3.05 atm), 9.10 slpm (6.10 atm), 10.95 slpm (9.15 atm), and 12.75 slpm (11.9 atm). To avoid possible errors due to changes in the quenching environment, it would be desirable to apply laser-saturated fluorescence (LSF) to the NO measurements [9]. However, as discussed by Reisel and Laurendeau [ll], the broadband LIF signal resulting from excitation of the Q,(26.5) transition of NO does not remain saturated above 3 atm; in fact, the LIF measurements display a linear variation of fluorescence signal with laser power at pressures greater than 6 atm. However, as discussed above, the quenching variation with equivalence ratio at a given pressure is small, thus allowing for quantitative measurements of NO despite neglecting the variation in quenching rate coefficient. The measurements discussed below were obtained with a laser energy of N 1 mJ/pulse; these represent well-saturated LIF measurements at 1 atm, partially-saturated measurements at 3.05 atm, and linear LIF measurements at 6.1, 9.15, and 11.9 atm. At 3.05 atm, the sensitivity of the fluorescence signal to laser power is N .65 (as compared with 1 for linear fluorescence.) Linear corrections were made for laser power fluctuations at pressures above 6.1 atm, nonlinear corrections were made at 3.05 atm, and no corrections were made at 1 atm. The measurements were calibrated using the following procedure, applied at each pressure. Measurements of the fluorescence voltage from the burned-gas region of the 4 = 0.90 flame were obtained for several different levels of
FLAMES doped NO. We assumed that the doped NO does not react through the flame, and that the amount of NO found in this flame is small compared to the amount of doped NO. The former assumption is supported by computer modeling, which indicates that the burned-gas NO concentration is equal to the doped NO concentration to within 5% for these lean flames. The data from several doping conditions, when plotted as fluorescence signal versus doped [NO], form a straight line. The slope of this line was used to obtain a fluorescence voltage calibration, which was then applied to the fluorescence signal measured in the undoped 4 = 0.90 flame. The observed linear relationship further indicates that the NO undergoes little reaction in this flame. NO concentrations in the other flames at the same pressure were determined from the measured fluorescence signal using the NO concentration versus signal voltage calibration determined in the 4 = 0.90 flame. As discussed previously, while the concentrations of the major quenching species CO, and H,O change with equivalence ratio, the variation in the quenching rate coefficient with equivalence ratio is less than the uncertainty of the measurement at a given pressure. Since only small changes occur in the quenching environment at a given pressure, corrections for any variation in quenching between flames were deemed unnecessary. Quenching corrections approaching +20% would only occur for those LIF measurements at p 2 6.1 atm and 4 > 1.4. Thus, by calibrating at each pressure, it is not necessary to make corrections for changes in the quenching environment due to pressure, or for changes in the optical alignment which result from maximizing the NO fluorescence signal at each pressure. This calibration procedure also allows us to neglect changes in the spectral linewidth due to pressure broadening, variations in the spectral line overlap with pressure, and changes in rotational energy transfer with pressure. The results of the LIF measurements are shown in Fig. 2, and most of the results are tabulated in Table 2. The measurements were taken in the postflame zone, 3 mm above the burner surface at each pressure except at 1
146
J. R. REBEL
AND N. M. LAURENDEAU TABLE 2
Measured NO Number Densities ( X 10-l 3 cm- 3, in the C,H,/Oa/N, Flames of This Study c
140
1
‘b.6
ti
-*-:11.9atm
Pressure (atm)
1,
0.6
1.0
1.2
1.4
1.6
1.6
2.0
Equivalence Ratio
Fig. 2. LIF measured NO concentrations in high-pressure C,H,/O,/N, flames. The measurements were taken 5 mm above the burner for the 1.0 atm flames, and 3 mm above the burner for the high-pressure flames. The dilution ratio was 3.1 for all flames, and the total flow rates were 3.50 slpm (1.0 atm), 6.18 slpm (3.05 atm), 9.1 slpm (6.10 atm), 10.95 slpm (9.15 atm), and 12.75 slpm (11.9 atm). The uncertainty shown is the estimated accuracy of &25%. The precision of the measurements is < 7.5%.
atm, for which the measurements were taken at 5 mm above the burner surface. The trends in the data are very similar to those found for previous LIF measurements in C,H,/O,/N, flames Ill]. The increase in NO number density with pressure is primarily only a result of the increase in the total number of particles with pressure: the peak NO mole fraction at each pressure is _ 30-40 ppm. As the equivalence ratio increases, the [NO] rises steadily through a rich peak and then rapidly decreases. In addition, the peak [NO] at a given pressure shifts towards leaner conditions with increasing pressure. The latter feature was anticipated, based upon the analysis of Reisel and Laurendeau [ll] for ethane flames, which suggested that the shift was primarily due to the reaction CH, + OH ti CH + H,O,
(R2)
d
1.00
3.05
6.10
9.15
11.9
0.70 0.80 0.90 0.95
0.71 1.1 -
5.8 8.0
17.3 24.8 -
35.0 55.8 -
12.6 21.4 35.5 45.5 54.3 57.3 56.4 46.8 32.8 14.6 6.4
35.8 65.8 78.3 89.6 90.7 77.5 51.9 30.8 11.3 7.1 -
91.2 113.7 137.5 140.4 133.8 95.0 59.2 34.7 21.3 16.8
26.7 44.7 82.0 99.6 117.8 137.8 145.3 130.3 102.6 68.4 45.0 30.9 23.4 -
1.00 1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70 1.80
1.6 2.4 3.9 5.6 8.1 10.5 11.5 12.4 12.3 12.0 9.3
1.90
2.0
-
-
-
-
’ The precision is f 7.5%, and the estimated accuracy is *25%. - indicates no measurement at this condition.
which promotes CH production at leaner conditions with increasing pressure. In the region of maximum CH production, the forward reaction rate dominates (N 40 times greater). In general, CH is produced from CH, through reactions with OH, 0, and H. As the pressure increases, the near-stoichiometric flames become more “efficient,” relative to the richer flames, at producing CH from CH,. This improved efficiency is due primarily to the increasing importance of reaction R2. The [Ol and [H] decrease steadily with increasing pressure, as does the [OH] in moderately rich flames. However, the [OH] does not decrease as rapidly with pressure in slightly rich flames. Consequently, the peak [NO] shifts towards stoichiometric conditions with increasing pressure. The similar behavior of ethane and ethylene flames suggests that reaction R2 is the basis for a universal NO phenomenon during high-pressure hydrocarbon combustion. To verify the consistency of the LIF method, we compared measurements of [NOI made us-
LIF MEASUREMENTS
OF NO IN C,H,/O,/N,
ing two different excitation lines. Sample results for this study are presented in Fig. 3. Here, we performed the LIF measurements using both the Q,(26.5) line and the Ri(18.5) line. Based on Boltzmann fraction calculations, the fluorescence from the Ri(18.5) line should depend more strongly on temperature than that from the QJ26.5) line; therefore, the R,(18.5) line is a less desirable line to use for an LIF measurement. Figure 3 presents the results of the measurements from 3.05 to 9.15 atm; the results were similar at 1 atm and at 11.9 atm. As can be seen from the comparative measurements, there appears to be little if any difference attributable to the excitation transition. The measurements using one line fall within the uncertainty of the other. The worst agreement appears in the 3.05-atm flames. This difference may be attributable to the partially saturated fluorescence behavior that exists at this pressure; i.e., different degrees of saturation may exist for the two lines. Even with this possible discrepancy, the two measurements
140
t :3.05atn -A-:6.10 atn +:9.15atn
0.8
1.0
1.2
1.4
1.6
Equivalence Ratio
Fig. 3. Comparisons between LIF measurements of NO concentration obtained using two different spectral lines for excitation at three pressures. The results found using the Q,(26.5) line agree well with those found using the R&18.5) line.
FLAMES
147
are still within acceptable accuracy, indicating that the LIF measurements are essentially independent of the chosen excitation line. The results from the computer modeling are shown in Figs. 4 and 5. Figure 4 presents the results for [NO] in the post-flame zone using the GMK-DB model; Fig. 5 contains the results using the MIME-DB model. Both models follow trends similar to the measurements; however, they tend to underpredict both the peak [NO] and its corresponding equivalence ratio. One noticeable difference in the qualitative behavior of the two models is that for the GMK-DB scheme, the regions of decreasing [NO] on the rich side fall on approximately the same curve at all pressures. However, for the MIME-DB model, the regions fall on different curves at different pressures, which is similar to the behavior observed in the measurements. This similarity may be important for future chemical kinetics modifications. In general, the predicted temperatures were in good agreement ( + 50 K) with the measured temperatures. The MIME-DB temperatures tended to be N lo-50 K lower than the GMK-DB temperatures. The only significant deviations between the measured and modeled temperatures occurred in the lean flames (for which the measured temperatures are N 100150 K lower than the GMK-DB predicted temperatures), and in the rich flames at atmospheric pressure (for which the measured temperatures are N 150 K higher than the GMKDB predicted temperatures). The measured temperatures of the rich flames at atmospheric pressure were high in comparison to the same flames at greater pressures; from previous work [9, 111, we would expect approximately equal temperatures. The higher temperatures at atmospheric pressure could be due to catalytic effects on the uncoated thermocouple; these effects may be more significant at atmospheric pressure since the flame front is located higher above the burner. Nevertheless, these deviations are not significant enough to affect the comparative trends between the measurements and the predictions. Figure 6 compares plots of the equivalence ratio corresponding to the peak [NO], and the equivalence ratios corresponding to the halfpeak [NO], at each pressure, as found by both
148
REISEL 140
C,H,: GMK-DB Model
--t
40
: 1 .OO atm
C,H,: MIME-DB Model
-w-:3.05atm --V-ft
AND N. M. LAURENDEAU
: 6.10 atm
--C:9.15atm
I \ I \
+:l.OOatm -w-:3.05atm --V-- : 6.10 atm + : 9.15 atm
-*-:11.9atm n 25 -
J
0.8
1.2
1.4
Equivalence
Ratio
1.0
1.6
I .a
1 .o
1.2 Equivalence
1.4
1.6
1.0
Ratio
Fig. 4. NO concentrations found for the experimental flames by solving the coupled species-energy equations using the GMK-DB reaction mechanism. The concentrations are at the same heights as the measurements shown in Fig. 2.
Fig. 5. NO concentrations found for the experimental flames by solving the coupled species-energy equations using the MIME-DB reaction mechanism. The concentrations are at the same heights as the measurements shown in Fig. 2.
modeling and measurements. The two models follow very similar qualitative trends, which show a shift in the curves towards leaner conditions with increasing pressure. While these trends are similar to the measured trends, the measurements give consistently higher equivalence ratios than the predictions for both the peak and half-peak [NO] values. The agreement is especially poor at lower pressures, with improved agreement at higher pressures. Figures 7 and 8 offer direct comparisons of the measured and modeled NO concentrations at 3.05 and 9.15 atm, respectively. In general, poor qualitative agreement is obtained for flames richer than C#J = 1.2 at lower pressures (P I 6.1 atm). Better qualitative agreement is obtained at higher pressures (P 2 9.15 atm). Figures 7 and 8 also show that both models under-predict the measured NO concentrations, particularly in the moderately rich flames. For the GMK-DB model at lower pressures, this difference is directly due to a significant underprediction of the equivalence ratio corresponding to the peak [NO] (see Fig. 6). For flames at P I 6.1 atm, the GMK-DB model
adequately predicts the [NO] up to an equivalence ratio of N 1.2, but then underpredicts the [NO] at higher equivalence ratios (see Fig. 7). Because the calculated [NO] peaks at a lower equivalence ratio, the quantitative agreement becomes progressively poorer in richer flames. At higher pressures (see Fig. 81, the GMK-DB model demonstrates good qualitative agreement with the measurements, but continues to demonstrate a quantitative underprediction of [NO]. The GMK-DB model also provides good predictions of the [NO] in highly rich flames, i.e., well above the equivalence ratio corresponding to the peak [NO]. In general, then, the GMK-DB model appears to be useful for predicting [NO] for 4 < 1.2 at all pressures, and for all equivalence ratios at P 2 9.15 atm. The improved agreement at higher pressures results from the shift in the equivalence ratio corresponding to the peak [NO] towards stoichiometric conditions for both the measurements and predictions. The MIME-DB model provides a qualitative behavior similar to that of the GMK-DB model, but predicts a much lower [NOI (see Figs. 7
LIF MEASUREMENTS
OF NO IN C,H,/O,/N,
149
FLAMES
1.6 LocationofPeak[NO] 1.6-
@
1:
i‘:.;---1_.,
I
1.0 -
1.6
t
1.6 1.4 1.2
.
Locationof Haif-Peak[NO]
kg.._
1.0
.
--
0.6 : 0
I
2
,,,I,
4
6
0’
1 0
1
I
I
6
10
,
I
1.0
1
Pressure (atm)
’ 1.6
1.4
Ratio
Fig. 7. Comparison of the measured and predicted NO concentrations for the flames at 3.05 atm. The results from both the GMK-DB and MIME-DB models are shown.
LIF
-C - : GMK-DB Model ---t : MIME-DB Model
160
CC
and 8). The MIME-DB model includes most of the same kinetics for smaller hydrocarbons as the mechanism of Miller and Bowman [12]. Because larger hydrocarbons are not formed in high concentrations for most of the flames of this investigation (since the equivalence ratio is below sooting conditions), we may apply results from a previous analysis [ 111 of prompt-NO formation from the Miller-Bowman scheme. In particular, Reisel and Laurendeau [ 1l] found that the inclusion of
1.2
Equivalence
12
Fig. 6. Comparisons of experimental and predicted locations on the [NO] versus 4 curves. Both the predictions from the GMK-DB and the MIME-DB model are shown. The top plot represents the equivalence ratio at peak [NO] for each pressure. The bottom plot represents the locations corresponding to the half-maximum [NO] on the rich and lean sides of the [NO] versus 4 curves.
CH + H,O tf CH,O + H
/
0.8
140
f
i
'E
00 7 0 _
T
P=
9.15 atm
i
/
120 ; ,!I;
i
',
20 i
(R3) 0L-i
in the Miller-Bowman model depletes a large amount of CH from the flame; the removed CH is then unable to form NO via reaction Rl. A test was performed in which six rich flames at different pressures were modeled using the MIME-DB mechanism, both with and without
0.6
-1
0.8
1 .o Equivalence
1.2
1.4
1.6
Ratio
Fig. 8. Comparison of the measured and predicted NO concentrations for the flames at 9.15 atm. The results from both the GMK-DB and MIME-DB models are shown.
150
J. R. REISEL AND N. M. LAURENDEAU
reaction R3. The [NO] increased 45%-90% with this reaction removed; while the resulting [NO] is still smaller than that from the GMKDB model, the two predictions are in better agreement. Therefore, the inclusion of reaction R3 is the primary reason for the difference in the quantitative prediction between the two models. However, reaction R3 most likely belongs in the mechanism; therefore, its inclusion requires additional kinetic modifications to compensate for the resulting reduction in NO formation. The effect of fuel type on the variation of [NO] with equivalence ratio can be seen in Fig. 9. Figure 9 is a comparison of the equivalence ratios corresponding to the peak [NO] and the half-peak [NO] as found for the C,H, flames of this study and the C,H, flames of Reisel and Laurendeau [ll]. As noted previously, fuel type can affect the amount of prompt-NO formed for a given equivalence ratio. From
Fig. 9, it is clear that high NO levels occur in leaner flames for C&H, as compared to C,H,. The difference between the two fuels is more pronounced at lower pressures. All of the characteristic equivalence ratios show improved agreement with increasing pressure, as the equivalence ratios corresponding to the peak [NO] approach unity for both fuels at higher pressures. Bachmaier et al. [23] similarly found that prompt-NO formation peaks at a higher equivalence ratio for C,H, compared to C,H, flames at atmospheric pressure. The present work demonstrates that this difference resulting from fuel type is considerably reduced at higher pressures. An important question is why, for p I 6.1 atm, both models are predicting the peak NO concentration to be in a leaner flame than that found from the measurements. Recall that the [NO] is underpredicted in moderately rich flames at lower pressures, but that the quantitative agreement is much better for 4 < 1.2. In the following discussion, we present a possible explanation for this behavior; however, additional experimental data will be needed to confirm our hypothesis. Recall that much of the difference between the GMK-DB and MIME-DB models is due to the inclusion of reaction R3 in the MIME-DB model. Other than that, the two models behave in roughly the same manner (much like the GMK-DB model and the Miller-Bowman [12] model for ethane flames [ll]); thus, we will deem conclusions which are valid for one model to be valid for the other. Due to the large percentage of prompt-NO formed in the rich flames, we would anticipate that the predicted inaccuracy in [NO] would be caused by the hydrocarbon kinetics. To determine a model’s accuracy at predicting species profiles, one must usually rely on low-pressure data (as flames at high pressure are located too near the burner surface to adequately resolve the spatial profiles of many radicals). Applying results from lowpressure flames to high-pressure flames may not be very accurate; however, we will assume that such an extrapolation should lead to approximately correct results. Bernstein et al. 1301 found that the MillerBowman mechanism [ 121 accurately predicts the location of the peak CH concentration
1.81
I Location of Peak [NO]
1.6
H
‘.,I,
CT
0
: C2H4
-
t
,
,
, , , , , , , 1
2
4
6
8
10
12
E 5
3 2
s
2.0
I
I
w
1.8
Location of Hail Peak [NO]
1.6
0
2
4
6 Pressure
8
10
12
(atm)
Fig. 9. Comparison of the measured equivalence ratios corresponding to peak and half peak [NO] for the C,H,/O,/N, flames of this study, and the C2H,/02/N2 flames of Reisel and Laurendeau [ll]. Both sets of flames have a dilution ratio of 3.1 and the same volumetric flow rates.
LIF MEASUREMENTS
OF NO IN C,H,/O,/N,
in stoichiometric, low-pressure C2H4/02/Ar flames. Due to the similarities between the Miller-Bowman mechanism and the MIME-DB mechanism, one would expect similar behavior from the MIME-DB mechanism. Miller et al. [31] compared measured [CH] profiles with predictions from the Miller-Melius [151 mechanism in low-pressure C,H,/O,/Ar flames at several stoichiometries. They again found that the measured and modeled [CHI profiles agreed well in a near stoichiometric flame, much like the results of Bernstein et al. [30] for the C,H, flame. However, the calculated [CHI profile peaked closer to the burner than that for the measured profile in richer flames (4 > 1.6); the discrepancy increased with increasing equivalence ratio. If this behavior applies to C,H, flames as well, one would expect that in very rich flames, the predicted [CHI profile would peak substantially nearer the burner surface than the actual [CH] profile. Such behavior may lead to an underprediction of the amount of prompt-NO in richer flames, for the residence time of CH at high temperatures would be smaller for the model than for the experiment. Hence, the calculated equivalence ratio corresponding to the peak [NO] would be at a lower equivalence ratio than that indicated by the measurements. On the other hand, the near-stoichiometric flames would have similar predicted and experimental [CHI profiles, which explains the good predictions for [NO] at 4 < 1.2. For flames at higher pressure, the overall qualitative agreement is better because the measured equivalence ratio corresponding to the peak [NO] has shifted far enough towards stoichiometric conditions to satisfy 4 < 1.2. CONCLUSIONS In summary, we have obtained LIF measurements of NO concentration in laminar, premixed, flat C,H,/O,/N, flames at pressures ranging from 1.0 to 11.9 atm. The temperatures of these flames were between 1600 and 1850 K. NO measurements obtained from the excitation of two different spectral lines were found to give very similar results. As expected from previous work, the equivalence ratio corresponding to the peak [NO] at a given pres-
FLAMES
151
sure shifts towards stoichiometric conditions with increasing pressure. Both the GMK-DB and MIME-DB mechanisms tend to underprediet the [NO] in these flames, although the GMK-DB model offers more quantitative accuracy than the MIME-DB model. Qualitatively, both models exhibit similar behavior; the predictions are fairly poor at lower pressures, and better at higher pressures. This behavior is mostly caused by the under-prediction of [NO] in moderately rich flames at lower pressures, which causes inaccurate prediction of the equivalence ratio corresponding to the peak [NO] at a given pressure. The agreement is better at higher pressures because the equivalence ratio at the peak [NO] approaches stoichiometric conditions for both the measurements and predictions. The same rationale also explains the reduced influence of fuel type at higher pressures. The authors would like to thank the NASALewis Research Center for funding this investigation. REFERENCES 1. Heberling, P. V., Sixteenth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1977, pp. 159-168. 2. Leonard, G. L., and Correa, S. M., in Fossil Fuel Combustion Symposium 1990 (Singh, S. N., Ed.), ASME/ PD, Vol. 30, 1990, pp. 69-74. 3 Drake, M. C., Ratcliffe, J. W., Blint, R. J., Carter, _. C. D., and Laurendeau N. M., Twenty-Third Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1991, pp. 387-395. 4. Morley, C., Eighteenth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1981, pp. 23-31. .5 Morley, C., Cornbust. Flame 47:67-81 (1982). 6. Chou, M.-S., Dean, A. M., and Stem, D., J. Chem. Phys. 78:5962-5970 (1983). 7. Cattolica, R. J., Cavolowsky, J. A., and Mataga, T. G., Twenty-Second Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1989, pp. 1165-1173. 8. Heard, D. W., Jeffries, J. B., Smith, G. P., and Crosley, D. R., Cornbust. Flame 88:137-148 (1992). 9. Reisel, J. R., Carter, C. D., Laurendeau, N. M., and Drake, M. C., Combust. Sci. Technol. 91:271-295 (1993). 10. Miller, J. A., and Fisk, G. A., Chem. Eng. News 65122-46 (1987).
11. Reisel, J. R., and Laurendeau, N. M., Combust. Sci. Technol., 98:137-160 (1994).
152
J. R. REISEL
12. Miller, J. A., and Bowman, C. T., Prog. Ener. Combust. Sci. 15:287-338 (1989). 13. Drake, M. C., and Blint, R. J., Combust. Sn’. Technol. 75:261-285 (19911. 14. Glarborg, P., Miller, J. A., and Kee, R. J., Combust. Flame 65:177-202 (1986). 15. Miller J. A., and Melius C. F., Combust. Flame 91:21-39 (1992). 16. Carter, C. D., King, G. B., and Laurendeau, N. M., Rev. Sci. Instrum. 60:2606-2609 (1989). 17. Harris, J. M., Lytle, F. E., and McCain, T. C., Anal. Chem. 48:2095-2098 (1976). 18. Drake, M. C., and Ratcliffe, J. W., J. Chem. Whys. 98:3850-3865 (1993). 19. Zeldovich, J., Acta Physiochem. URSS 21:577-628
24. Kee, R. J., Grcar, J. F., Smooke, M. D., and Miller, J. A., Sandia National Laboratories, SAND85-8240, 1985. 25. Kee, R. J., Rupley, F. M., and Miller, J. A., Sandia National Laboratories, SAND89-8009, 1989. 26. Kee, R. J., Rupley, F. M., and Miller, J. A., Sandia National Laboratories, SAND87-8215, 1987. 27. Kee, R. J., Dixon-Lewis Jr., G., Wamatz, J., Coltrin, M. E., and Miller, J. A., Sandia National Laboratories, SAND86-8246, 1986. 28. Dean, A. J., Davidson, D. F., Hanson, R. K., and Bowman, C. T., Western States Section/The Combustion Institute, Paper 88-91 (1988). 29. Drake, M. C., and Blint, R. J., personal communication (1992). 30. Bernstein, J. S., Fein, A., Choi. J. B., Cool, T. A., Sausa, R. C., Howard, S. L., Locke, R. J., and Miziolek, A. W., Cornbust. Flame 92~85-105 (1993). 31. Miller: J. A., Volponi, J. V., Durant Jr., J. L., Goldsmith, J. E. M., Fisk, G. A., and Kee, R. J., Twenty-
(1946).
Wolfrum, J., Chem. Zng. Tech. 44:656-659 (1972). 21. Malte, P. C., and Pratt, D. T., Combust. Sci. Technol.
20.
9:221-231
(1974).
22.
Fenimore, C. P., Thirteenth Symposium (Internation) on Combustion, The Combustion Institute, Pittsburgh, 1973, pp. 373-379. 23. Bachmaier, F., Eberius, K. H., and Just, Th., Combust. Sci. Technol. 7177-84 (1973).
AND N. M. LAURENDEAU
Third Symposium (International) on Combustion, The
Combustion Institute, Pittsburgh, 1991, pp. 187-194. Received 18 March 1994; revised 21 July 1994
We have obtained quantitative LIF measurements of NO concentration in the post-flame zone of a series of flat, laminar, premixed C,H,/O,/N, flames at pressures ranging from 1 to 11.9 atm. The internal consistency of the data was verified by demonstrating good agreement between measurements based on two different NO excitation lines. The temperatures of the flames were 1600-1850 K, indicating that most of the NO produced in the flames is prompt-NO. The results show that the equivalence ratio corresponding to the peak NO concentration at a given pressure shifts towards leaner conditions with increasing pressure. We have also modeled the flames using two chemical kinetics schemes: (1) the Glarborg-Miller-Kee mechanism as modified by Drake and Blint (GMK-DB), and (2) the Miller-Melius hydrocarbon mechanism, combined with the nitrogen kinetics of Drake and Blint (MIME-DB). Both models are qualitatively inaccurate with respect to the measured pressure shift, particularly at pressures below 6.1 atm. Both models also tend to underpredict the NO concentrations, although the GMK-DB model is acceptable in many cases.
INTRODUCTION High-pressure combustion applications, such as gas-turbine engines, are a major source of nitric oxide (NO) emissions. As the environmental problems caused by high NO emissions grow, it has become imperative to reduce NO emissions from combustion processes. The achievement of this goal by combustion designers requires, among other things, a thorough understanding of the chemical kinetics involved in the production of NO at high pressure. Such understanding, in turn, mandates accurate in situ measurements of NO concentration. Quantitative measurements of NO concentration can be obtained using both physical techniques, such as probe-sampling [l-31, and optical techniques. such as laser-induced fluorescence (LIF) [4-91. Probe-sampling combined with chemiluminescent detection is advantageous since it possesses a lower detection limit, is easier to use, and is less expensive than laser-based methods. However, a physical probe can disrupt the flow field, potentially altering the concentrations of radical species; moreover, such probes may not be able to
*Corresponding author. FLAME 101: 141-152 (19951 Copyright 0 1995 by The Combustion Institute Published by Elsevier Science Inc.
withstand the harsh conditions of practical combustion environments [lo]. These disadvantages can be overcome by employing optical techniques. Optical procedures allow for remote sensing of numerous species in a variety of environments. Unlike sampling probes, optical methods generally do not alter the combustion process; in addition, many combustors are more readily adaptable to optical access than to physical sampling probes. Finally, precise spatial resolution is achievable through the use of optical methods. Previously, we have demonstrated the feasibility of making quantitative LIF measurements of NO in C,H,/O,/N, flames up to 14.6 atm [ll]. These results showed that the equivalence ratio corresponding to the peak [NO] at a given pressure shifts towards leaner conditions with increasing pressure. In addition, the measured pressure shift was successfully modeled using two chemical kinetics schemes [12, 131; the model from Drake and Blint [13] also provided good quantitative agreement with the measured NO concentrations. In this paper, we extend the previous work by presenting LIF measurements of NO in flat, laminar, C,H,/O,/N, flames at pressures up to 11.9 atm. By comparing NO measurements obtained from two different excitation lines, we verify the internal consistency of the LIF
COMBUSTIONAND
OOlO-2180/95/$9.50 SSDI 0010-2180(94)00195-X
142
J. R. REISEL AND N. M. IAURENDEAU FROM LASER
procedure. We also investigate whether the shifting behavior is limited to paraffinic fuels (like ethane), or if, as anticipated, it is a more universal phenomenon. In addition, by modeling the flames using two different reaction mechanisms, we evaluate the ability of current chemical kinetics schemes to predict [NO] in high-pressure C,H,/O,/N, flames. One of these is the Glarborg-Miller-Kee model as modified by Drake and Blint (GMK-DB) [13, 141; the other is the hydrocarbon scheme of Miller and Melius [15] combined with the nitrogen kinetics of the GMK-DB model (MIME-DB). Comparisons between the predicted results and the LIF measurements provide a test of the ability of these mechanisms to predict the effects of high pressure on NO emissions from premixed ethylene flames. LASER-INDUCED METHODOLOGY
FLUORESCENCE i.;q
Figure 1 shows a schematic diagram of the experimental apparatus used in the LIF measurements. Excitation of NO was achieved via one of two spectral lines: the Q,(26.5) line and the Rr(18.5) line, both within the y(O,O) band of NO. The Q,(26.5) line (h = 225.6 nm> was chosen because (1) its Boltzmann fraction is insensitive to temperature fluctuations over the range of temperatures of our flames, and (2) other species, such as 0,, do not interfere spectrally with this NO absorption line [9]. The Rr(18.5) line (A = 225.5 nm) was chosen as an alternative transition based on its separation from other spectral lines as well as the lack of spectral interference from other species. The excitation wavelengths were generated by employing the second harmonic (A = 532 nm> of a Quanta-Ray DCR3G Nd:YAG laser to pump a PDL-2 dye laser, which produces laser radiation at N 572 nm. The output of the dye laser was frequency doubled in a Quanta-Ray Wavelength Extender (WEX-1) and the doubled-dye beam was mixed with the first harmonic of the Nd:YAG laser, producing - 1 mJ/pulse at - 225.5 nm for these measurements. After leaving the laser system, the beam was directed towards a 2.5-cm-diameter, watercooled, sintered-bronze McKenna flat-flame burner. The burner is located inside the high-
EiE
I-.-.-._._.-----.I CoMPUTER 1
Fig. 1. Schematic diagram of experimental apparatus: A, trigger photodiode; B, G, beam-splitter; C, lOOO-mm focal-length lens; D, K, beam steering assembly; E, aperture; F, pressure vessel; H, beam dump; I, power-monitoring photodiode; J, 200-mm focal length lens; L, 300-mm focal-length lens; M, 1/2-m monochromator; N, PMT.
facility described by pressure combustion Carter et al. [16]. The pressure vessel has four optical ports, two of which are used for directing the laser beam through the facility. The spot size produced by the optical arrangement is - 250 pm. After exiting the vessel, the beam passed through a fused silica plate which directed a portion of the beam towards a UVsensitive photodiode. This photodiode was employed to monitor the beam energy, which was required for normalization of the fluorescence signal. For fluorescence detection, we made use of an optical port perpendicular to the laser entrance and exit ports. The fluorescence was focused on the entrance slit of a 4-m monochromator. The detector was an RCA lP28B photomultiplier tube specially wired for temporal resolution of the fluorescence signal [171. The broadband fluorescence signal was de-
LIF MEASUREMENTS
OF NO IN C,H,/O,/N,
tected at N 236 nm, corresponding to N 3 nm of the y(O,l) band of NO. To account for pressure shifting of the Q,(26.5) and the R,(18.5) absorption lines, the laser and collection optics were tuned for the maximum NO fluorescence signal at each pressure. A 500-ps window at the peak of the fluorescence pulse was sampled using a Stanford Research Systems SR255 fast sampler. The image of the entrance slit over the burner was 80 pm X 6.7 mm. Each data point was averaged over 600 laser shots. When performing a linear LIF measurement, one must be concerned with the effects of both laser power fluctuations and quenching variations on the fluorescence signal. One way to avoid such effects is to employ laser-saturated fluorescence &SF), which has been previously applied with broadband detection to NO at atmospheric pressure [9]. However, as discussed by Reisel and Laurendeau 1111, saturated conditions are difficult to maintain for NO measurements above 3 atm. For linear fluorescence measurements, corrections for laser power fluctuations can be made by normalizing the fluorescence signal using the measured laser power. Quenching variations could be handled in a similar manner; however, the measurement of quenching rate coefficients is not a trivial task. By comparing measurements obtained using both linear and saturated LIF, we have previously found that quenching variations at a given pressure are not significant for our range of flame conditions [9]. To confirm that this result is also true for the C,H,/O,/ N, flames of this study, we have calculated quenching rate coefficients in the post-flame zone using equilibrium concentrations and the quenching cross-sections from Drake and Ratcliffe [181. The quenching rate coefficient per unit pressure, Q/P, can be calculated from
where k is Boltzmann’s constant, T is the temperature, Xi is the mole fraction of quenching species i, a, is the quenching crosssection of NO with species i, and pi is the reduced mass between species i and NO. Only
FLAMES
143
the species studied by Drake and Ratcliffe [18] were considered for the calculations (N,, O,, H,O, CO,, CO, C,H,, H,, NO, H, OH, and 0); these include the dominant quenching species in the post-flame zone. The quenching variation proves to be relatively small (5 15%) over most of the flame conditions for which linear fluorescence is required (P 2 6.1 atm) in the C,H,/O,/N, flames. A few of the flames with $J 2 1.4 have quenching rate coefficients which vary from the calibration flame by N 20%; however, even this difference is less than the uncertainty in the measurements. Greater differences would arise if the linear fluorescence measurements were extended to flames having larger equivalence ratios (4 = 1.8); however, measurements at these equivalence ratios were only performed at lower pressures, for which LSF was employed. NITRIC OXIDE CHEMICAL KINETICS NO is produced through three main chemical pathways [13]: (1) the Zeldovich, or thermalNO, mechanism [193, (2) the N,O-intermediate mechanism [20,21], and (3) the prompt-NO mechanism [22]. The amount of NO formed through each pathway depends on the temperature, pressure, and equivalence ratio of the flame. The lower temperatures and rich flames of this investigation favor the prompt-NO mechanism. The controlling reaction for prompt-NO is thought to be [12] CH + N, * HCN + N.
(RI)
Since reaction Rl demonstrates a hydrocarbon linkage, the amount of prompt-NO must depend on the type of fuel. In fact, Bachmaier et al. 1231 demonstrated that the amount of prompt-NO formed at a given equivalence ratio varies with fuel type. Most of the experimental flames in this study were investigated through computer modeling. The modeling of the chemical kinetics was performed using the Sandia steady, laminar, one-dimensional, premixed flame code [24]. In addition, the CHEMKIN-II computer program library 1251 was used to process the reaction mechanism into a form which is appropriate
144
J. R. REISEL AND N. M. LAURENDEAU
for use by the flame code. The thermodynamic and transport properties, required by the Sandia flame code for calculation of the species concentration profiles, were provided by a thermodynamic property data base [26] and a transport property data base [27]. Previously, Reisel et al. [9] found that most of the NO produced in atmospheric-pressure versions of similar low-temperature C,H,/ 0,/N, flames is produced in the flamefront rather than in the post-flame region. Due to the similar temperatures of the C,H,/ O,/ N, flames of this study, we expect to again have little post-flame thermal-NO production. As discussed by Reisel and Laurendeau [ill, modeling the flamefront NO in such flames generally requires consideration of all three reaction mechanisms. In particular, while the promptNO mechanism does not contribute to NO production in the postflame zone (where our measurements are taken), its inclusion remains necessary since most of the NO produced in the flamefront is related to the prompt-NO pathway. The primary goal of the kinetics modeling effort is to assess the feasibility of using current kinetic schemes to simulate the effects of pressure on NO concentration. To do this, we must employ the coupled species-energy equations to determine the relevant temperature profiles as the rapid temperature rise in the thin flamefront is not easily measured at high pressures. A burner surface temperature of 300 K is used as a boundary condition to mimic heat loss to the burner. While the calculated temperature profiles will not agree precisely with the actual temperature profiles (leading to some potential errors in quantitative agreement), the calculated postflame temperatures agree sufficiently with the measured postflame temperatures to allow for accurate assessment of the pressure trends, and for a reasonable assessment of quantitative capabilities. Two mechanisms have been used as the chemical kinetics scheme for the computer model. Both are based on the comprehensive mechanism assembled by Glarborget al. 1141. The first mechanism (GMK-DB) is taken from the set of elementary reactions listed by Drake and Blint [13]. This reaction mechanism considers 49 species and over 200 reactions. Drake
and Blint [13] adopted most of the reaction mechanism from Glarborg et al. [141; however, they made a few modifications. These include the introduction of pressure dependency into four unimolecular reactions, the addition of a C,H, .reaction mechanism, and the introduction of rate parameters for reaction (Rl) based on measurements in a high temperature shock tube [28]. The rate parameters at pressures of 1.0, 3.05, 6.1, and 9.15 atm are given by Drake et al. [3]. The rate parameters at 11.9 atm were provided by Drake and Blint [29]. The second mechanism (MIME-DB) is a combination of the hydrocarbon mechanism assembled by Miller and Melius [15] and the nitrogen kinetics of the GMK-DB model. The Miller-Melius mechanism was designed to model rich combustion of aliphatic fuels such as ethylene and acetylene. The mechanism contains most of the Miller-Bowman 1121 mechanism for modeling small hydrocarbon species, and adds many larger hydrocarbon compounds. Due to the size of this mechanism, the coupled species-energy equations were not solved for the complete mechanism. Rather, only the hydrocarbon kinetics were solved using the coupled species-energy solution; the entire mechanism was then solved using the temperature profile from this partial solution. A sample case was run with the full mechanism using the energy solution, and the temperature profile from the reduced and full mechanisms were found to be nearly identical (differing by less than 5 K>. RESULTS AND DISCUSSION Using the experimental apparatus described above, quantitative LIF measurements of NO were performed in flat, laminar, high-pressure C,H,/O,/N, flames. Data were obtained at five pressures over a range of 1.0-11.9 atm. The purity of the C,H, was greater than 99.9%. The temperatures of these flames, as measured with radiation-corrected Pt-Pt/lO%Rh thermocouples in the post-flame region, ranged from 1600 to 1850 K (precision f30 K, accuracy f75 K). A few of the temperature measurements are listed in Table 1. All of the flames had a dilution ratio (~r,@02) of 3.1. The total flow rates were held constant at each
LIF MEASUREMENTS
OF NO IN C,H,/O,/N,
TABLE 1 Radiation-Corrected Thermocouple Measurements (K) of Selected C,H,/O,/N, Flames” Pressure (atm) 4
1.00
3.05
6.10
9.15
11.9
0.90 1.20 1.35 1.40
1625 1810 1840
1650 1785
1640 1740 1740 -
1635 1710 1725
1645 1690 1695
1750
’ The precision is + 30 K and the accuracy is f 75 K. indicates no measurement at this condition.
pressure, and were as follows: 3.50 slpm (1.0 atm), 6.18 slpm (3.05 atm), 9.10 slpm (6.10 atm), 10.95 slpm (9.15 atm), and 12.75 slpm (11.9 atm). To avoid possible errors due to changes in the quenching environment, it would be desirable to apply laser-saturated fluorescence (LSF) to the NO measurements [9]. However, as discussed by Reisel and Laurendeau [ll], the broadband LIF signal resulting from excitation of the Q,(26.5) transition of NO does not remain saturated above 3 atm; in fact, the LIF measurements display a linear variation of fluorescence signal with laser power at pressures greater than 6 atm. However, as discussed above, the quenching variation with equivalence ratio at a given pressure is small, thus allowing for quantitative measurements of NO despite neglecting the variation in quenching rate coefficient. The measurements discussed below were obtained with a laser energy of N 1 mJ/pulse; these represent well-saturated LIF measurements at 1 atm, partially-saturated measurements at 3.05 atm, and linear LIF measurements at 6.1, 9.15, and 11.9 atm. At 3.05 atm, the sensitivity of the fluorescence signal to laser power is N .65 (as compared with 1 for linear fluorescence.) Linear corrections were made for laser power fluctuations at pressures above 6.1 atm, nonlinear corrections were made at 3.05 atm, and no corrections were made at 1 atm. The measurements were calibrated using the following procedure, applied at each pressure. Measurements of the fluorescence voltage from the burned-gas region of the 4 = 0.90 flame were obtained for several different levels of
FLAMES doped NO. We assumed that the doped NO does not react through the flame, and that the amount of NO found in this flame is small compared to the amount of doped NO. The former assumption is supported by computer modeling, which indicates that the burned-gas NO concentration is equal to the doped NO concentration to within 5% for these lean flames. The data from several doping conditions, when plotted as fluorescence signal versus doped [NO], form a straight line. The slope of this line was used to obtain a fluorescence voltage calibration, which was then applied to the fluorescence signal measured in the undoped 4 = 0.90 flame. The observed linear relationship further indicates that the NO undergoes little reaction in this flame. NO concentrations in the other flames at the same pressure were determined from the measured fluorescence signal using the NO concentration versus signal voltage calibration determined in the 4 = 0.90 flame. As discussed previously, while the concentrations of the major quenching species CO, and H,O change with equivalence ratio, the variation in the quenching rate coefficient with equivalence ratio is less than the uncertainty of the measurement at a given pressure. Since only small changes occur in the quenching environment at a given pressure, corrections for any variation in quenching between flames were deemed unnecessary. Quenching corrections approaching +20% would only occur for those LIF measurements at p 2 6.1 atm and 4 > 1.4. Thus, by calibrating at each pressure, it is not necessary to make corrections for changes in the quenching environment due to pressure, or for changes in the optical alignment which result from maximizing the NO fluorescence signal at each pressure. This calibration procedure also allows us to neglect changes in the spectral linewidth due to pressure broadening, variations in the spectral line overlap with pressure, and changes in rotational energy transfer with pressure. The results of the LIF measurements are shown in Fig. 2, and most of the results are tabulated in Table 2. The measurements were taken in the postflame zone, 3 mm above the burner surface at each pressure except at 1
146
J. R. REBEL
AND N. M. LAURENDEAU TABLE 2
Measured NO Number Densities ( X 10-l 3 cm- 3, in the C,H,/Oa/N, Flames of This Study c
140
1
‘b.6
ti
-*-:11.9atm
Pressure (atm)
1,
0.6
1.0
1.2
1.4
1.6
1.6
2.0
Equivalence Ratio
Fig. 2. LIF measured NO concentrations in high-pressure C,H,/O,/N, flames. The measurements were taken 5 mm above the burner for the 1.0 atm flames, and 3 mm above the burner for the high-pressure flames. The dilution ratio was 3.1 for all flames, and the total flow rates were 3.50 slpm (1.0 atm), 6.18 slpm (3.05 atm), 9.1 slpm (6.10 atm), 10.95 slpm (9.15 atm), and 12.75 slpm (11.9 atm). The uncertainty shown is the estimated accuracy of &25%. The precision of the measurements is < 7.5%.
atm, for which the measurements were taken at 5 mm above the burner surface. The trends in the data are very similar to those found for previous LIF measurements in C,H,/O,/N, flames Ill]. The increase in NO number density with pressure is primarily only a result of the increase in the total number of particles with pressure: the peak NO mole fraction at each pressure is _ 30-40 ppm. As the equivalence ratio increases, the [NO] rises steadily through a rich peak and then rapidly decreases. In addition, the peak [NO] at a given pressure shifts towards leaner conditions with increasing pressure. The latter feature was anticipated, based upon the analysis of Reisel and Laurendeau [ll] for ethane flames, which suggested that the shift was primarily due to the reaction CH, + OH ti CH + H,O,
(R2)
d
1.00
3.05
6.10
9.15
11.9
0.70 0.80 0.90 0.95
0.71 1.1 -
5.8 8.0
17.3 24.8 -
35.0 55.8 -
12.6 21.4 35.5 45.5 54.3 57.3 56.4 46.8 32.8 14.6 6.4
35.8 65.8 78.3 89.6 90.7 77.5 51.9 30.8 11.3 7.1 -
91.2 113.7 137.5 140.4 133.8 95.0 59.2 34.7 21.3 16.8
26.7 44.7 82.0 99.6 117.8 137.8 145.3 130.3 102.6 68.4 45.0 30.9 23.4 -
1.00 1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70 1.80
1.6 2.4 3.9 5.6 8.1 10.5 11.5 12.4 12.3 12.0 9.3
1.90
2.0
-
-
-
-
’ The precision is f 7.5%, and the estimated accuracy is *25%. - indicates no measurement at this condition.
which promotes CH production at leaner conditions with increasing pressure. In the region of maximum CH production, the forward reaction rate dominates (N 40 times greater). In general, CH is produced from CH, through reactions with OH, 0, and H. As the pressure increases, the near-stoichiometric flames become more “efficient,” relative to the richer flames, at producing CH from CH,. This improved efficiency is due primarily to the increasing importance of reaction R2. The [Ol and [H] decrease steadily with increasing pressure, as does the [OH] in moderately rich flames. However, the [OH] does not decrease as rapidly with pressure in slightly rich flames. Consequently, the peak [NO] shifts towards stoichiometric conditions with increasing pressure. The similar behavior of ethane and ethylene flames suggests that reaction R2 is the basis for a universal NO phenomenon during high-pressure hydrocarbon combustion. To verify the consistency of the LIF method, we compared measurements of [NOI made us-
LIF MEASUREMENTS
OF NO IN C,H,/O,/N,
ing two different excitation lines. Sample results for this study are presented in Fig. 3. Here, we performed the LIF measurements using both the Q,(26.5) line and the Ri(18.5) line. Based on Boltzmann fraction calculations, the fluorescence from the Ri(18.5) line should depend more strongly on temperature than that from the QJ26.5) line; therefore, the R,(18.5) line is a less desirable line to use for an LIF measurement. Figure 3 presents the results of the measurements from 3.05 to 9.15 atm; the results were similar at 1 atm and at 11.9 atm. As can be seen from the comparative measurements, there appears to be little if any difference attributable to the excitation transition. The measurements using one line fall within the uncertainty of the other. The worst agreement appears in the 3.05-atm flames. This difference may be attributable to the partially saturated fluorescence behavior that exists at this pressure; i.e., different degrees of saturation may exist for the two lines. Even with this possible discrepancy, the two measurements
140
t :3.05atn -A-:6.10 atn +:9.15atn
0.8
1.0
1.2
1.4
1.6
Equivalence Ratio
Fig. 3. Comparisons between LIF measurements of NO concentration obtained using two different spectral lines for excitation at three pressures. The results found using the Q,(26.5) line agree well with those found using the R&18.5) line.
FLAMES
147
are still within acceptable accuracy, indicating that the LIF measurements are essentially independent of the chosen excitation line. The results from the computer modeling are shown in Figs. 4 and 5. Figure 4 presents the results for [NO] in the post-flame zone using the GMK-DB model; Fig. 5 contains the results using the MIME-DB model. Both models follow trends similar to the measurements; however, they tend to underpredict both the peak [NO] and its corresponding equivalence ratio. One noticeable difference in the qualitative behavior of the two models is that for the GMK-DB scheme, the regions of decreasing [NO] on the rich side fall on approximately the same curve at all pressures. However, for the MIME-DB model, the regions fall on different curves at different pressures, which is similar to the behavior observed in the measurements. This similarity may be important for future chemical kinetics modifications. In general, the predicted temperatures were in good agreement ( + 50 K) with the measured temperatures. The MIME-DB temperatures tended to be N lo-50 K lower than the GMK-DB temperatures. The only significant deviations between the measured and modeled temperatures occurred in the lean flames (for which the measured temperatures are N 100150 K lower than the GMK-DB predicted temperatures), and in the rich flames at atmospheric pressure (for which the measured temperatures are N 150 K higher than the GMKDB predicted temperatures). The measured temperatures of the rich flames at atmospheric pressure were high in comparison to the same flames at greater pressures; from previous work [9, 111, we would expect approximately equal temperatures. The higher temperatures at atmospheric pressure could be due to catalytic effects on the uncoated thermocouple; these effects may be more significant at atmospheric pressure since the flame front is located higher above the burner. Nevertheless, these deviations are not significant enough to affect the comparative trends between the measurements and the predictions. Figure 6 compares plots of the equivalence ratio corresponding to the peak [NO], and the equivalence ratios corresponding to the halfpeak [NO], at each pressure, as found by both
148
REISEL 140
C,H,: GMK-DB Model
--t
40
: 1 .OO atm
C,H,: MIME-DB Model
-w-:3.05atm --V-ft
AND N. M. LAURENDEAU
: 6.10 atm
--C:9.15atm
I \ I \
+:l.OOatm -w-:3.05atm --V-- : 6.10 atm + : 9.15 atm
-*-:11.9atm n 25 -
J
0.8
1.2
1.4
Equivalence
Ratio
1.0
1.6
I .a
1 .o
1.2 Equivalence
1.4
1.6
1.0
Ratio
Fig. 4. NO concentrations found for the experimental flames by solving the coupled species-energy equations using the GMK-DB reaction mechanism. The concentrations are at the same heights as the measurements shown in Fig. 2.
Fig. 5. NO concentrations found for the experimental flames by solving the coupled species-energy equations using the MIME-DB reaction mechanism. The concentrations are at the same heights as the measurements shown in Fig. 2.
modeling and measurements. The two models follow very similar qualitative trends, which show a shift in the curves towards leaner conditions with increasing pressure. While these trends are similar to the measured trends, the measurements give consistently higher equivalence ratios than the predictions for both the peak and half-peak [NO] values. The agreement is especially poor at lower pressures, with improved agreement at higher pressures. Figures 7 and 8 offer direct comparisons of the measured and modeled NO concentrations at 3.05 and 9.15 atm, respectively. In general, poor qualitative agreement is obtained for flames richer than C#J = 1.2 at lower pressures (P I 6.1 atm). Better qualitative agreement is obtained at higher pressures (P 2 9.15 atm). Figures 7 and 8 also show that both models under-predict the measured NO concentrations, particularly in the moderately rich flames. For the GMK-DB model at lower pressures, this difference is directly due to a significant underprediction of the equivalence ratio corresponding to the peak [NO] (see Fig. 6). For flames at P I 6.1 atm, the GMK-DB model
adequately predicts the [NO] up to an equivalence ratio of N 1.2, but then underpredicts the [NO] at higher equivalence ratios (see Fig. 7). Because the calculated [NO] peaks at a lower equivalence ratio, the quantitative agreement becomes progressively poorer in richer flames. At higher pressures (see Fig. 81, the GMK-DB model demonstrates good qualitative agreement with the measurements, but continues to demonstrate a quantitative underprediction of [NO]. The GMK-DB model also provides good predictions of the [NO] in highly rich flames, i.e., well above the equivalence ratio corresponding to the peak [NO]. In general, then, the GMK-DB model appears to be useful for predicting [NO] for 4 < 1.2 at all pressures, and for all equivalence ratios at P 2 9.15 atm. The improved agreement at higher pressures results from the shift in the equivalence ratio corresponding to the peak [NO] towards stoichiometric conditions for both the measurements and predictions. The MIME-DB model provides a qualitative behavior similar to that of the GMK-DB model, but predicts a much lower [NOI (see Figs. 7
LIF MEASUREMENTS
OF NO IN C,H,/O,/N,
149
FLAMES
1.6 LocationofPeak[NO] 1.6-
@
1:
i‘:.;---1_.,
I
1.0 -
1.6
t
1.6 1.4 1.2
.
Locationof Haif-Peak[NO]
kg.._
1.0
.
--
0.6 : 0
I
2
,,,I,
4
6
0’
1 0
1
I
I
6
10
,
I
1.0
1
Pressure (atm)
’ 1.6
1.4
Ratio
Fig. 7. Comparison of the measured and predicted NO concentrations for the flames at 3.05 atm. The results from both the GMK-DB and MIME-DB models are shown.
LIF
-C - : GMK-DB Model ---t : MIME-DB Model
160
CC
and 8). The MIME-DB model includes most of the same kinetics for smaller hydrocarbons as the mechanism of Miller and Bowman [12]. Because larger hydrocarbons are not formed in high concentrations for most of the flames of this investigation (since the equivalence ratio is below sooting conditions), we may apply results from a previous analysis [ 111 of prompt-NO formation from the Miller-Bowman scheme. In particular, Reisel and Laurendeau [ 1l] found that the inclusion of
1.2
Equivalence
12
Fig. 6. Comparisons of experimental and predicted locations on the [NO] versus 4 curves. Both the predictions from the GMK-DB and the MIME-DB model are shown. The top plot represents the equivalence ratio at peak [NO] for each pressure. The bottom plot represents the locations corresponding to the half-maximum [NO] on the rich and lean sides of the [NO] versus 4 curves.
CH + H,O tf CH,O + H
/
0.8
140
f
i
'E
00 7 0 _
T
P=
9.15 atm
i
/
120 ; ,!I;
i
',
20 i
(R3) 0L-i
in the Miller-Bowman model depletes a large amount of CH from the flame; the removed CH is then unable to form NO via reaction Rl. A test was performed in which six rich flames at different pressures were modeled using the MIME-DB mechanism, both with and without
0.6
-1
0.8
1 .o Equivalence
1.2
1.4
1.6
Ratio
Fig. 8. Comparison of the measured and predicted NO concentrations for the flames at 9.15 atm. The results from both the GMK-DB and MIME-DB models are shown.
150
J. R. REISEL AND N. M. LAURENDEAU
reaction R3. The [NO] increased 45%-90% with this reaction removed; while the resulting [NO] is still smaller than that from the GMKDB model, the two predictions are in better agreement. Therefore, the inclusion of reaction R3 is the primary reason for the difference in the quantitative prediction between the two models. However, reaction R3 most likely belongs in the mechanism; therefore, its inclusion requires additional kinetic modifications to compensate for the resulting reduction in NO formation. The effect of fuel type on the variation of [NO] with equivalence ratio can be seen in Fig. 9. Figure 9 is a comparison of the equivalence ratios corresponding to the peak [NO] and the half-peak [NO] as found for the C,H, flames of this study and the C,H, flames of Reisel and Laurendeau [ll]. As noted previously, fuel type can affect the amount of prompt-NO formed for a given equivalence ratio. From
Fig. 9, it is clear that high NO levels occur in leaner flames for C&H, as compared to C,H,. The difference between the two fuels is more pronounced at lower pressures. All of the characteristic equivalence ratios show improved agreement with increasing pressure, as the equivalence ratios corresponding to the peak [NO] approach unity for both fuels at higher pressures. Bachmaier et al. [23] similarly found that prompt-NO formation peaks at a higher equivalence ratio for C,H, compared to C,H, flames at atmospheric pressure. The present work demonstrates that this difference resulting from fuel type is considerably reduced at higher pressures. An important question is why, for p I 6.1 atm, both models are predicting the peak NO concentration to be in a leaner flame than that found from the measurements. Recall that the [NO] is underpredicted in moderately rich flames at lower pressures, but that the quantitative agreement is much better for 4 < 1.2. In the following discussion, we present a possible explanation for this behavior; however, additional experimental data will be needed to confirm our hypothesis. Recall that much of the difference between the GMK-DB and MIME-DB models is due to the inclusion of reaction R3 in the MIME-DB model. Other than that, the two models behave in roughly the same manner (much like the GMK-DB model and the Miller-Bowman [12] model for ethane flames [ll]); thus, we will deem conclusions which are valid for one model to be valid for the other. Due to the large percentage of prompt-NO formed in the rich flames, we would anticipate that the predicted inaccuracy in [NO] would be caused by the hydrocarbon kinetics. To determine a model’s accuracy at predicting species profiles, one must usually rely on low-pressure data (as flames at high pressure are located too near the burner surface to adequately resolve the spatial profiles of many radicals). Applying results from lowpressure flames to high-pressure flames may not be very accurate; however, we will assume that such an extrapolation should lead to approximately correct results. Bernstein et al. 1301 found that the MillerBowman mechanism [ 121 accurately predicts the location of the peak CH concentration
1.81
I Location of Peak [NO]
1.6
H
‘.,I,
CT
0
: C2H4
-
t
,
,
, , , , , , , 1
2
4
6
8
10
12
E 5
3 2
s
2.0
I
I
w
1.8
Location of Hail Peak [NO]
1.6
0
2
4
6 Pressure
8
10
12
(atm)
Fig. 9. Comparison of the measured equivalence ratios corresponding to peak and half peak [NO] for the C,H,/O,/N, flames of this study, and the C2H,/02/N2 flames of Reisel and Laurendeau [ll]. Both sets of flames have a dilution ratio of 3.1 and the same volumetric flow rates.
LIF MEASUREMENTS
OF NO IN C,H,/O,/N,
in stoichiometric, low-pressure C2H4/02/Ar flames. Due to the similarities between the Miller-Bowman mechanism and the MIME-DB mechanism, one would expect similar behavior from the MIME-DB mechanism. Miller et al. [31] compared measured [CH] profiles with predictions from the Miller-Melius [151 mechanism in low-pressure C,H,/O,/Ar flames at several stoichiometries. They again found that the measured and modeled [CHI profiles agreed well in a near stoichiometric flame, much like the results of Bernstein et al. [30] for the C,H, flame. However, the calculated [CHI profile peaked closer to the burner than that for the measured profile in richer flames (4 > 1.6); the discrepancy increased with increasing equivalence ratio. If this behavior applies to C,H, flames as well, one would expect that in very rich flames, the predicted [CHI profile would peak substantially nearer the burner surface than the actual [CH] profile. Such behavior may lead to an underprediction of the amount of prompt-NO in richer flames, for the residence time of CH at high temperatures would be smaller for the model than for the experiment. Hence, the calculated equivalence ratio corresponding to the peak [NO] would be at a lower equivalence ratio than that indicated by the measurements. On the other hand, the near-stoichiometric flames would have similar predicted and experimental [CHI profiles, which explains the good predictions for [NO] at 4 < 1.2. For flames at higher pressure, the overall qualitative agreement is better because the measured equivalence ratio corresponding to the peak [NO] has shifted far enough towards stoichiometric conditions to satisfy 4 < 1.2. CONCLUSIONS In summary, we have obtained LIF measurements of NO concentration in laminar, premixed, flat C,H,/O,/N, flames at pressures ranging from 1.0 to 11.9 atm. The temperatures of these flames were between 1600 and 1850 K. NO measurements obtained from the excitation of two different spectral lines were found to give very similar results. As expected from previous work, the equivalence ratio corresponding to the peak [NO] at a given pres-
FLAMES
151
sure shifts towards stoichiometric conditions with increasing pressure. Both the GMK-DB and MIME-DB mechanisms tend to underprediet the [NO] in these flames, although the GMK-DB model offers more quantitative accuracy than the MIME-DB model. Qualitatively, both models exhibit similar behavior; the predictions are fairly poor at lower pressures, and better at higher pressures. This behavior is mostly caused by the under-prediction of [NO] in moderately rich flames at lower pressures, which causes inaccurate prediction of the equivalence ratio corresponding to the peak [NO] at a given pressure. The agreement is better at higher pressures because the equivalence ratio at the peak [NO] approaches stoichiometric conditions for both the measurements and predictions. The same rationale also explains the reduced influence of fuel type at higher pressures. The authors would like to thank the NASALewis Research Center for funding this investigation. REFERENCES 1. Heberling, P. V., Sixteenth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1977, pp. 159-168. 2. Leonard, G. L., and Correa, S. M., in Fossil Fuel Combustion Symposium 1990 (Singh, S. N., Ed.), ASME/ PD, Vol. 30, 1990, pp. 69-74. 3 Drake, M. C., Ratcliffe, J. W., Blint, R. J., Carter, _. C. D., and Laurendeau N. M., Twenty-Third Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1991, pp. 387-395. 4. Morley, C., Eighteenth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1981, pp. 23-31. .5 Morley, C., Cornbust. Flame 47:67-81 (1982). 6. Chou, M.-S., Dean, A. M., and Stem, D., J. Chem. Phys. 78:5962-5970 (1983). 7. Cattolica, R. J., Cavolowsky, J. A., and Mataga, T. G., Twenty-Second Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, 1989, pp. 1165-1173. 8. Heard, D. W., Jeffries, J. B., Smith, G. P., and Crosley, D. R., Cornbust. Flame 88:137-148 (1992). 9. Reisel, J. R., Carter, C. D., Laurendeau, N. M., and Drake, M. C., Combust. Sci. Technol. 91:271-295 (1993). 10. Miller, J. A., and Fisk, G. A., Chem. Eng. News 65122-46 (1987).
11. Reisel, J. R., and Laurendeau, N. M., Combust. Sci. Technol., 98:137-160 (1994).
152
J. R. REISEL
12. Miller, J. A., and Bowman, C. T., Prog. Ener. Combust. Sci. 15:287-338 (1989). 13. Drake, M. C., and Blint, R. J., Combust. Sn’. Technol. 75:261-285 (19911. 14. Glarborg, P., Miller, J. A., and Kee, R. J., Combust. Flame 65:177-202 (1986). 15. Miller J. A., and Melius C. F., Combust. Flame 91:21-39 (1992). 16. Carter, C. D., King, G. B., and Laurendeau, N. M., Rev. Sci. Instrum. 60:2606-2609 (1989). 17. Harris, J. M., Lytle, F. E., and McCain, T. C., Anal. Chem. 48:2095-2098 (1976). 18. Drake, M. C., and Ratcliffe, J. W., J. Chem. Whys. 98:3850-3865 (1993). 19. Zeldovich, J., Acta Physiochem. URSS 21:577-628
24. Kee, R. J., Grcar, J. F., Smooke, M. D., and Miller, J. A., Sandia National Laboratories, SAND85-8240, 1985. 25. Kee, R. J., Rupley, F. M., and Miller, J. A., Sandia National Laboratories, SAND89-8009, 1989. 26. Kee, R. J., Rupley, F. M., and Miller, J. A., Sandia National Laboratories, SAND87-8215, 1987. 27. Kee, R. J., Dixon-Lewis Jr., G., Wamatz, J., Coltrin, M. E., and Miller, J. A., Sandia National Laboratories, SAND86-8246, 1986. 28. Dean, A. J., Davidson, D. F., Hanson, R. K., and Bowman, C. T., Western States Section/The Combustion Institute, Paper 88-91 (1988). 29. Drake, M. C., and Blint, R. J., personal communication (1992). 30. Bernstein, J. S., Fein, A., Choi. J. B., Cool, T. A., Sausa, R. C., Howard, S. L., Locke, R. J., and Miziolek, A. W., Cornbust. Flame 92~85-105 (1993). 31. Miller: J. A., Volponi, J. V., Durant Jr., J. L., Goldsmith, J. E. M., Fisk, G. A., and Kee, R. J., Twenty-
(1946).
Wolfrum, J., Chem. Zng. Tech. 44:656-659 (1972). 21. Malte, P. C., and Pratt, D. T., Combust. Sci. Technol.
20.
9:221-231
(1974).
22.
Fenimore, C. P., Thirteenth Symposium (Internation) on Combustion, The Combustion Institute, Pittsburgh, 1973, pp. 373-379. 23. Bachmaier, F., Eberius, K. H., and Just, Th., Combust. Sci. Technol. 7177-84 (1973).
AND N. M. LAURENDEAU
Third Symposium (International) on Combustion, The
Combustion Institute, Pittsburgh, 1991, pp. 187-194. Received 18 March 1994; revised 21 July 1994