Sei sulla pagina 1di 11

Comparisons of laser-saturated, laser-induced, and planar laser-induced fluorescence measurements of nitric oxide in a lean direct-injection spray flame

Clayton S. Cooper, Rayavarapu V. Ravikrishna, and Normand M. Laurendeau

We report quantitative, spatially resolved laser-saturated uorescence LSF , linear laser-induced uorescence LIF , and planar laser-induced uorescence PLIF measurements of nitric oxide NO concentration in a preheated, lean direct-injection spray ame at atmospheric pressure. The spray is produced by a hollow-cone, pressure-atomized nozzle supplied with liquid heptane, and the overall equivalence ratio is unity. NO is excited by means of the Q2 26.5 transition of the 0, 0 band. LSF and LIF detection are performed in a 2-nm region centered on the 0, 1 band. PLIF detection is performed in a broad 70-nm region with a peak transmission at 270 nm. Quantitative radial NO proles obtained by LSF are presented and analyzed so as to correct similar LIF and PLIF proles. Excellent agreement is achieved among the three uorescence methodologies. 1998 Optical Society of America OCIS codes: 120.0120, 300.2530, 280.2470.

1. Introduction

With the onset of stringent NOx emissions standards over the past few decades, the attention of the combustion diagnostics community has turned to the quantitative measurement of nitric oxide NO concentrations in practical combustion devices. In particular, many researchers have probed liquid-fueled spray ames to understand better the complex ow elds associated with this type of turbulent reactive ow. Spray ames pose a particularly difcult challenge to diagnosticians who seek quantitative laserinduced uorescence LIF measurements of species concentrations. Inefcient atomization of the fuel can introduce large Mie scattering interferences, particularly in the near-eld spray region. Because of the large molecular weights of typical hydrocarbon fuels and fuel fragments, uorescence from unburned hydrocarbons UHCs and polycyclic aromatic hydrocarbons PAHs often poses a challenge to selective detection of the LIF signal from the species of inter-

The authors are with the Flame Diagnostics Laboratory, School of Mechanical Engineering, Purdue University, Room 31, 1288 Mechanical Engineering, West Lafayette, Indiana 47907-1288. Received 23 October 1997; revised manuscript received 10 March 1998. 0003-6935 98 214823-11$15.00 0 1998 Optical Society of America

est. Moreover, the electronic quenching rate coefcient must be known or estimated at each measured point if a LIF-based procedure is to produce quantitative results. The most prominent optical method for the qualitative measurement of species concentrations in spray ames has been planar laser-induced uorescence PLIF . A limited number of research groups have been actively pursuing species-concentration measurements in spray ames. PLIF is a variation of LIF that involves a two-dimensional sheet of laser irradiance for species excitation plus uorescence detection with a CCD camera coupled with spectral lters. The problems associated with quantifying the LIF measurements are well recognized and are delineated below in the context of previous PLIF investigations. Allen and Hanson1 extended PLIF from gaseous ames to a turbulent n-heptaneair spray ame in 1986 by employing an air-atomized siphon nozzle coupled with a coaxial air ow at atmospheric pressure. Hydroxyl measurements were attempted by use of 283-nm laser excitation in an effort to explore the impact of turbulent mixing on combustion chemistry. They found that Mie scattering from liquid droplets can pose an interference problem when broadband detection schemes are used. As with traditional LIF, the uncertainty in the relative OH measurements arose from variations in the quenching rate coefcient across the ame structure and
20 July 1998 Vol. 37, No. 21 APPLIED OPTICS 4823

from changes in the Boltzmann population fraction with variations in temperature. The Boltzmann fraction for the ground level of a particular excitation transition can be chosen to be fairly independent of temperature over a range of operating conditions; however, the quenching rate coefcient often cannot be sufciently modeled, and assumptions must be made to interpret the PLIF images correctly. In their case the OH results were qualitative owing to the absence of a calibration procedure and to the uncertainties in the quenching rate coefcient. Allen and Hanson2 then utilized PLIF and planar multiphoton dissociation to image the relative distributions of OH , CH , and fuel-vapor concentration in the same n-heptane spray ame. PLIF was used to obtain qualitative OH and CH spatial distributions with 283- and 426-nm excitation, respectively, while planar multiphoton dissociation was used to monitor C2H2 with 193-nm excitation. Acetylene was labeled a marker for the distribution of fuel vapor in regions near the reaction interface in nonpremixed ames. Allen and Hanson2 subsequently identied another problem that limits the PLIF technique: broad spectral uorescence features arising from the excitation of UHCs and PAHs. In 1990 Andresen et al.3 imaged concentration distributions of NO and O2 with 193-nm excitation, plus OH and iso-octane fuel with 248-nm excitation, in a modied internal combustion engine cylinder. Andresen et al.3 coupled high-power excimer lasers and PLIF to produce qualitative measurements of NO and iso-octane in chamber environments that reached 30 atm. Within the combustion environment of an engine the composition and temperature often change drastically within millimeters; therefore no consistent calibration was possible to interpret quantitatively the PLIF images. Koch et al.4 introduced UV laser-ash photography coupled with LIF and laser-induced predissociative uorescence in 1993 to detect spatial distributions of NO, O2, OH, and fuel for conventional spray ames supplied with either heptane or iso-octane. NO and O2 were again excited with 193-nm radiation, while OH and iso-octane fuel were excited with 248-nm radiation. The burners were designed for oil furnaces and contained an air-swirling bafe plate that incorporated various inlets to mix the fuel and air before combustion. Koch et al.4 encountered interference problems caused by elastically scattered light and broad uorescence from hydrocarbons. These problems were overcome by subtraction of on- and off-resonance images and by spectral ltering. Koch et al.4 attempted calibration for only NO by simulating exhaust-gas conditions in a heated cell and doping various concentrations of NO into the cell while monitoring the signal produced by their experimental apparatus. However, the local gas composition can vary considerably for highly turbulent combustion, thus introducing signicant uncertainties if the images are to be interpreted quantitatively. In 1994 Allen et al.5 obtained qualitative OH images in a heptaneair spray ame burning at 0.1 0.8
4824 APPLIED OPTICS Vol. 37, No. 21 20 July 1998

MPa. They tested both solid- and hollow-cone spray nozzles. In that study the effect of interfering PAH uorescence was assessed by use of a spectrometer to separate the uorescence spectrally into individual features. Excitation of OH was achieved by use of the P1 8 transition at 285.67 nm. At atmospheric pressure Allen et al.5 discovered a laser-induced signal that exhibited features at 350, 400, and 450 nm on a quasi-continuum background. The strength and spectral characteristics of this broad background were observed to be independent of excitation wavelength within a 5-nm region centered on the P1 8 transition, thus indicating a broad absorbing species such as a large-molecular-weight hydrocarbon. As the pressure rose, the identiable spectral features became nondiscernable and exhibited an increase in uorescence strength that scaled as approximately P2. In an extension of their 1994 study, Allen et al.6 performed similar OH imaging in ethanol ames and further assessed the effects of PAH interferences produced by 283-nm laser excitation. As reported by Allen et al.6 and based on Haynes7 and Lam et al.,8 the most prevalent aromatic hydrocarbons, in order of decreasing concentration, are probably benzene, toluene, phenyacetylene, styrene, and naphthalene. Calvert9 showed that the room-temperature absorption spectra of these species display a common, broad, semiresolved feature at 250 nm. Moreover, Beretta et al.10 experimentally veried this broad absorption region for the major PAHs at ame conditions. Beretta et al.10 also found that the major PAHs within the fuel-rich zone of an atmospheric methane diffusion ame generally display maximum emission wavelengths that are longer than 280 nm. These PAH characteristics are consistent with the interference features encountered in the study by Allen et al.,6 namely, detection of uorescence in spectral regions greater than 280 nm. Allen et al.6 suggested that excitation near 226 nm may produce more severe laser attenuation and hence PAH uorescence. To test this suggestion, Upshulte et al.11 next obtained qualitative PLIF images of NO, O2, and fuel vapor by employing excitation wavelengths of 226 and 308 nm. Measurements were made for ethanol fuel in the same high-pressure spray-ame combustor used by Allen et al.6 As expected, a broad interference signal attributed to PAH was discovered and assigned to a nominal 5% of the NO signal at atmospheric pressure. Perhaps the most utilitarian application of PLIF imaging for spray ames was performed by Locke et al.12 These researchers employed PLIF to image OH concentrations in a high-pressure 10 14 atm combustor supplied with Jet-A fuel 0.59 0.83 kg s by means of lean direct injection LDI 0.41 0.53 with preheated air 811 866 K . Although only the qualitative distribution of OH radicals was assessed in the reacting ow, the combustor was designed to simulate actual gas-turbine conditions. The authors found that elastically scattered light and PAH uorescence were not evident in the down-

stream regions of their LDI-based combustor. This is a signicant contribution to the combustion diagnostics community, as quantitative measurements at the practical conditions exemplied by Locke et al.12 are a goal of spray-ame optical diagnostics. Cessou and Stepowski13 recently reported absolute concentrations from PLIF images of hydroxyl in an atmospheric spray ame produced by a coaxial airblast injector supplied with methanol. The OH uorescence 315 nm was spectrally discriminated from elastic scattering 283 nm produced by the methanol drops, although Raman scattering from the drops required spatial ltering by use of a novel suppression technique developed by Cessou and Stepowski.13 The OH calibration procedure used an absorption line and its associated linear uorescence signal at locations of peak OH concentration in a reference laminar diffusion ame.14 This procedure, however, cannot account for potential quenching effects away from the regions of peak OH in spray ames. Although the research presented above offers signicant data sets for the chosen ow elds, barriers to quantitative measurements have not been completely overcome by the current technology. Some of these barriers include 1 spatial variations in the electronic quenching rate coefcient, 2 interferences from other species owing to broadband detection, and 3 absorption of the laser sheet as it passes through the control volume. In an effort to make PLIFconcentration images quantitative, Partridge15 reported a procedure whereby qualitative PLIF images in gaseous ames can be scaled by a small number of laser-saturated uorescence LSF measurements so as to make the images quantitative within the error bars of the more accurate LSF point data. The quantitative nature of the LSF measurements arises from the uorescence being considerably less sensitive to both laser irradiance and the electronic quenching rate coefcient. This study is concerned with a comparison of quantitative LSF measurements of NO16 and qualitative LIF and PLIF measurements of NO in a swirling liquid-fueled spray ame that incorporates a hollowcone, pressure-atomized nozzle. The spray ame is based on the LDI conguration, which is of current importance to gas-turbine combustion. Because of the nature of this particular ame, an enhancement procedure similar to that of Partridge15 proves to be successful in quantifying the LIF and PLIF measurements to within the accuracy bars of the LSF data. To our knowledge, the present study represents the rst effort in the combustion diagnostics community to utilize the LSF technique to quantify LIF and PLIF measurements of NO in a turbulent, liquid-fueled spray ame.
2. Experimental Setup and Methodology A. Lean Direct-Injection Module

Fig. 1. Cutaway drawing of the LDI burner. The burner is constructed of stainless steel with a fuel tube entering coaxially at the bottom of the burner and air entering perpendicular to the axis. The air is passed through packed glass beads to attain purely vertical ow and then directed through a helical-vane swirler. Fuel is injected immediately after the swirler by use of a small pressure-atomized nozzle.

The burner utilized here is based on a LDI design and typies that used in the primary zone of advanced gas-turbine combustors. The stainless-steel LDI

module 16.5 cm 3.8 cm diameter accommodates a fuel tube 6.4-mm diameter that enters the module coaxially at the bottom see Fig. 1 . For operation at atmospheric pressure, a 60 helical swirler 22.9-mm diameter is mounted at the top of the fuel delivery tube. The swirler itself is tapped to allow a Delavan 6-mm peanut nozzle ow number, 0.4 with a 62 spray cone to be directly threaded into the swirler. The nozzle is positioned vertically relative to the converging diverging orice 12.7-mm diameter at 40 . The depth of the nozzle below this orice is adjustable by means of a slide-through tting located at the bottom of the module. The air is preheated to 475 K, delivered perpendicular to the module axis, and passed through packed glass beads 1.5-mm diameter that ll the module cavity 22.9-mm diameter . The glass beads ensure purely vertical ow of the air entering the air swirler. The fuel delivery system incorporates a four-gallon stainless-steel pressure vessel rated at 5170 kPa. The stored heptane is pressurized with nitrogen and metered by a rotameter ow controller. The air is provided from a building compressor. The air ow is adjusted with a metering valve and monitored with a Hastings Model HFM-230 fast-response thermal mass-ow meter. Preheating is achieved with two in-line air heaters controlled with voltage regulators. The maximum preheat air temperature is limited by boiling within the fuel tube, which leads to vapor lock in the injector.
20 July 1998 Vol. 37, No. 21 APPLIED OPTICS 4825

B. Laser-Induced FluorescenceLaser-Saturated Fluorescence Facility

Following the methods described by Reisel et al.17 and Cooper and Laurendeau,16 excitation of NO is achieved by means of the Q2 26.5 line of its 0, 0 band at 225.58 nm. The fundamental 1064 nm from a Spectra-Physics injection-seeded GCR-4 Nd: YAG laser is frequency doubled 532 nm and used to pump a tunable PDL-3 dye laser. Pyrromethene 580 dye is employed to generate the pulsed dye-laser fundamental at 572.54 nm.18 A WEX-2 wavelength extender is used to frequency double the dye-laser output to 286.27 nm, followed by frequency mixing with the Nd:YAG fundamental to produce the required 225.58-nm laser beam. The system is equipped with a FabryPerot wavelength stabilization system to control pulsed dye-laser drift.19 The maximum energy obtained for the mixed beam is 4 mJ pulse. The uorescence from the probe volume is monitored by use of a 3 4-m monochromator with a Hamamatsu Model R106-HA photomultiplier tube PMT specially wired for high temporal resolution. For saturation to be ensured, the LSF signal is limited to only the temporal 500 ps and spatial peaks of the uorescence signature.17 For the LSF experiments the entrance slit was 68 m wide and 1 mm tall, which denes a probe volume cross-sectional area 68 m wide along the diameter of the laser beam and 1 mm long along the axis of the laser beam. For our 240- m beam waist this slit setting spatially integrates only the center 26% of the beam, thus avoiding uorescence wing effects. However, the collection scheme is still sensitive to depth-of-eld wing effects.17,20 For the LIF experiments the entrance-slit width was opened to match the beam waist. In this manner the entire spatial uorescence was captured, which ensured that the linear wings were included and thus minimized the focused portion along the center line that could tend toward partial saturation. The exit-slit width was set to 1.818 mm, dening an integrated spectral region of 2-nm FWHM for both the LSF and the LIF experiments. This setting was chosen based on extensive studies by previous researchers.17,21 Because of the relatively close spectral spacing between the excitation 226 nm and detection 236 nm wavelengths, the 0, 1 band of NO cannot be entirely isolated from the spectral wings of the Mie scattering signal produced by the spray. To account for the resulting superposition of the uorescence and scattering signals, we utilized a background subtraction method.16 The inherent Mie scattering portion produced by the on-line excitation is separately determined by the employment of an off-line excitation wavelength 225.53 nm shifted from the NO transition 225.58 nm . A subtraction method is then implemented to convert the uorescence and interference signals to measured NO concentrations. A differential doping calibration was performed in a 0.8 C2H6 O2 N2 NO ame stabilized on a watercooled McKenna burner to produce a linear calibra4826 APPLIED OPTICS Vol. 37, No. 21 20 July 1998

Fig. 2. Spectral bandpass for PLIF detection.

tion plot. Experiments have shown that the background signal is overwhelmingly comprised of Mie scattering and that possible uorescence interferences from UHCs or PAHs are negligible within the 0, 1 detection window.16 On the other hand, any uorescence from large molecules such as UHCs or PAHs that might exist in fuel-rich regions would invariably exhibit broad absorption and uorescence signatures and thus would be superimposed onto the more prominent Mie scattering signal. The off-line excitation signal for these types of molecule would be comparable to that for on-line excitation and so would be subtracted out in the background correction process. The on-line NO uorescence signal is averaged over 600 laser shots, whereas the off-line background signal is averaged over 300 shots.
C. Planar Laser-Induced Fluorescence Facility

The same excitation transition employed for the LSFLIF measurements of NO is used for the PLIF measurements, namely, the Q2 26.5 line of its 0, 0 band at 225.58 nm. The generated sheet of laser irradiance is 550 m wide by 18 mm tall.15 A Princeton Instruments Model ICCD-576TC-RG proximity-focused intensied CCD detector incorporating a 578 pixel 384 pixel CCD with 23- m square pixels was utilized for detection of the NO uorescence. We focused the uorescence on the intensied CCD detector by utilizing an aberration-corrected, ve-fused-silica-element, 105-mm-focal-length f 4.5 lens, such that each pixel corresponded to a 77- m square in the image plane of the ame. A lter set consisting of a wide-band interference lter 92-nm FWHM spectrally centered at 250 nm coupled with 6 mm of a UG5 colored glass lter was used to reduce the Mie scattering from the liquid droplets. This lter scheme was selected based on a study by previous researchers21 and provides an 70-nm FWHM spectral window with a peak transmission at 270 nm see Fig. 2 . This scheme collects uorescence primarily from the NO 0, 2 and 0, 3 vibrational bands, as opposed to the restricted 0, 1 band used in the LIF and LSF measurements. An intensied CCD thermoelectric cooler was used in conjunction with an external water chiller

circulator to reduce the temperature of the CCD to 32 C. A pulse generator Princeton Instruments Model FG-100 was used to produce a gate of 30 ns to the intensied CCD. Operation of the intensied CCD and supporting hardware was controlled by a detector controller Princeton Instruments Model ST130 . The user interface to the intensied CCD system was provided by Princeton Instruments CSMA software, which was also used for all image analysis and reduction. For the data presented here the on-chip integration time was 4 min, which corresponded to 2400 laser shots. All images were corrected for dark noise, ame luminosity, Mie scattering plus laser-induced interferences, and laser-sheet nonuniformity.15
D. Operating Conditions
Fig. 3. Normalized uorescence signal versus normalized laser irradiance for the saturated and linear regimes of the LSFLIF experiments. The data for the linear regime are independent of those displayed for the saturated regime. The solid curves are curve ts to the data to accent the saturated and linear uorescence regimes.

The LDI burner was operated at an overall equivalence ratio of unity with heptane fuel supplied to the nozzle at 0.3 g s and air at 4.52 g s. Although this condition does not model lean operation of the LDI module, it was chosen here to provide ample NO in the combustion products. The air is preheated to 475 K to assist in vaporization and mixing of the fuel. The nozzle is located 2.54 mm below the burner throat. At this point the air has a velocity of 48 m s based on the mass-ow rate and exit area, resulting in a Reynolds number of 17,000. Owing to the intense mixing, the ame is essentially nonsooting and blue, with only an occasional soot tip forming and quickly disappearing.
3. Laser-Saturated Fluorescence Measurements

band of NO. This fraction is dened by xresonant. Finally, the NO signal can be computed as follows: SNO,On-line SPMT,On-line 1 xextinction 1 1 SMie,Off-line POn-line POff-line (3)

xresonant . xextinction

Following Cooper and Laurendeau,16 we expect that the on-line 225.58 nm signal measured by the PMT consists of NO uorescence SNO,On-line and Mie scattering SMie,On-line , i.e., SPMT,On-line SNO,On-line SMie,On-line. (1)

To reduce the PMT signal to the NO fraction of the collected radiation, we tune the pulsed dye laser to an off-line wavelength 225.53 nm and collect the resultant signal. When we correct for variations in laser power P between the on-line and off-line measurements, the on-line Mie signal becomes SMie,Off-line POff-line

SMie,On-line

POn-line.

(2)

Both S and P are also corrected to account for the electronic background of the boxcar data-acquisition system. The nal NO signal must account for the extinction of uorescence by the liquid-droplet environment. The fraction of the signal attenuated is dened by xextinction. In addition, the off-line signal contains a small resonant excitation of the 0, 0

The extinction fraction varies with axial height and radial position. For our worst-case conditions, xextinction 7%. The resonant excitation fraction is constant at xresonant 5%. The NO signal is converted to an absolute number density by means of a calibration performed in a 0.8 C2H6 O2 N2 NO ame stabilized on a watercooled McKenna burner. Calibration in a reference medium is justied by the fact that the saturated uorescence signal is substantially independent of the electronic quenching rate coefcient. For our LSF experiments the sensitivity of the uorescence signal to both laser irradiance and the quenching rate coefcient were determined to be 25%. Figure 3 displays a typical curve that relates the uorescence signal to the laser irradiance for both the saturated and linear regimes of the LIF scheme. Varying amounts of NO were doped into the calibration ame and the resulting PMT signal recorded to produce a linear calibration plot parts per million per volt . To convert to a number density calibration plot, we used a thermocouple to measure the local temperature of the calibration ame. The resulting NO proles are plotted in Fig. 4 for three axial heights of 5, 10, and 20 mm above the burner. Accuracy bars are shown at 10-mm increments. Typical accuracies at the 95% condence limit range from 12% to 30%, with larger accuracy bars near the Mie scattering regions of the ame. The NO proles are distributed almost symmetrically
20 July 1998 Vol. 37, No. 21 APPLIED OPTICS 4827

includes the laser-coupled excited level and the collisionally coupled upper rovibronic levels17: NTss N1 1 gu gu gl
sat

1 Qe Ru,c Rc,u Ru,c

(4) where NTss represents the steady-state value of the population in the upper manifold, 1 is the fraction of the original population remaining in the lasercoupled levels, gu and gl are the representative degeneracies of the upper and lower laser-coupled levels, and sat is the degree of saturation. For broadband detection the uorescence signal is proportional to the total upper manifold population NTss. The effect of RET is represented in Eq. 4 through the inverse collisional branching ratio Qe Ru,c and the ratio of RET rate coefcients into and out of the laser-coupled excited level Rc,u Ru,c, where here the subscript u represents the laser-coupled excited level and the subscript c represents the manifold of collisionally coupled upper rovibronic levels.17 To assess the RET effects encapsulated by the collisional branching ratio, we numerically modeled the NO molecule as a harmonic-oscillator rigid rotor with a rotational constant of B 1.70 cm 1 for the lower set of rovibronic levels and a rotational constant of B 1.99 cm 1 for the upper set. The laser-coupled rovibronic levels are denoted by J 27 in both v 0 and the v 0 vibrational levels of the X2 ground 2 excited electronic states, respectively. and A Spin splitting and -doubling of the rotational levels are ignored. Twenty additional rotational levels are coupled to both the lower and the upper laser-coupled levels with quantum numbers ranging from J 17 to J 37. Each rovibronic level is assumed to be collisionally coupled to three levels both immediately below and above. The total RET rate coefcient at a ame temperature of 1750 K is estimated as 4 109 s 1, based on the study by Mallard et al.25 The state-to-state RET rate coefcients for downward transitions to J 3 are assumed equal, whereas the upward rotational relaxation rate coefcients are calculated based on the principle of detailed balancing. The required spectroscopic parameters and the electronic quenching rate coefcient were evaluated based on previous research.26,27 Complete details of this model are available in Ref. 23. From this dynamic analysis the ratio Rc,u Ru,c was estimated as 0.093 and was found to be essentially independent of the magnitude of the total RET rate coefcient in the upper manifold. Using Eq. 4 to deduce the sensitivity of the broadband LSF signal to the inverse collisional branching ratio Qe Ru,c, we nd that Qe Ru,c Qe Ru,c

Fig. 4. LSF measurements for the LDI burner under the following operating parameters: 1 mfuel 0.3 g s, 2 1, 3 dnozzle 2.54 mm, and 4 Tair preheat 475 K.

about the centerline. This symmetry is due to the intense recirculation of combustion products and air into the center of the ame structure.22 Past the axial distance of 20 mm, the peak NO concentration decreases owing to continued entrainment of outside air. The effects of electronic quenching and rotational energy transfer RET on our ability to measure NO concentrations quantitatively when utilizing LSF can be addressed by considering a generic four-level model.23 By calibrating in a reference ame and changing ame environments, we potentially change both the rotational relaxation and the electronic quenching rate coefcients. The amount by which these changes affect the upper-state population and hence the uorescence yield depends on the molecular dynamics governing the excitation and detection processes. One concern arises from the large irradiances used in typical LSF measurements. During stimulated absorption from the lower lasercoupled level to an excited upper level, it is possible that the depopulated lower level is repopulated by means of RET from neighboring ground rovibronic levels; thus more molecules could actually be excited than predicted by a simple two-level model. It is, however, the balance of repopulation into the directly excited ground rovibronic level and depopulation out of the directly excited upper rovibronic level that constitutes the validity of the LSF technique, as interpreted by the balanced cross-rate model.24 To understand the limitations of the LSF technique, one must consider an expression for the total upper-level population in a four-level model, which
4828 APPLIED OPTICS Vol. 37, No. 21 20 July 1998

SQe

Ru,c

Qe Ru,c

Rc,u Ru,c

Rc,u Ru,c

. 1

(5)

For typical conditions in the calibration ame the sensitivity of the broadband LSF signal to the collisional quenching ratio is found to be 51%. This is a signicant dependence that could limit the utility of the LSF measurements. However, the validity of the broadband LSF technique has been previously addressed through experimentation. Carter and Laurendeau28 compared narrow-band and broadband LSF measurements of hydroxyl concentration and found excellent comparisons with a maximum deviation of 10% between the measurements for a wide range of ame conditions 0.61.6, P 110 bar . Reisel et al.17 compared broadband LSF measurements with broadband linear LIF measurements of NO concentration in a variety of laminar, at, premixed C2H6 O2 N2 ames at atmospheric pressure 0.6 1.6 and determined that the collisional branching ratio is fairly insensitive to the equivalence ratio. Since broadband LIF measurements and narrow-band LSF measurements can be shown to be essentially independent of RET effects,23 their research suggests that a change in the electronic quenching rate coefcient parallels that for the rotational relaxation rate coefcient, thus validating the utility of the LSF method for 0.6 1.6. Typical changes in the electronic quenching rate coefcient among lean ames are less than 25%, so without further knowledge of the RET rate coefcients, we postulate a change of 25% in the inverse collisional branching ratio between the calibration ame and the LDI test ame. Based on the above sensitivity discussion, broadband detection would yield an 13% error in the calibrated uorescence yield. This error is consistent with previous comparisons between LSF and linear LIF measurements of NO concentration at atmospheric pressure.17
4. Laser-Induced Fluorescence Measurements

Fig. 5. LIF measurements for the LDI burner under the following operating parameters: 1 mfuel 0.3 g s, 2 1, 3 dnozzle 2.54 mm, and 4 Tair preheat 475 K.

fraction fB and the electronic quenching rate coefcient Qe relative to the calibration ame is given by NO
LDI,absolute

fB Tcal fB TLDI

Qe,LDI NO Qe,cal

LDI,relative

(7)

Similar to the LSF measurements, the reduction of linear LIF data to NO number densities can be accomplished by the employment of SPMT,On-line POn-line SMie,Off-line 1 xresonant , POff-line

SNO,LIF

(6)

where the signal measured by the PMT and the power measured by the photodiode were already corrected for extinction. The NO signal is again converted to an absolute number density by means of a calibration performed in a 0.8 C2H6 O2 N2 NO ame stabilized on a water-cooled McKenna burner. The quenching rate coefcient at the calibration point in this ame is predicted by modeling to be 6.86 108 s 1.27 The uncorrected LIF measurements are inversely proportional to the quenching rate coefcient, thus limiting the data interpretation to a qualitative nature without an accurate estimate of the local quenching rate coefcient. The NO concentration corrected for variations in the Boltzmann population

where the subscripts cal and LDI refer to the calibration and LDI ame, respectively, and NO LDI,relative represents NO concentration relative to the calibration ame. The transition chosen for excitation, namely, the NO Q2 26.5 line of the 0, 0 band, is fairly insensitive 5% to population variations in the range of 1000 2000 K. One method of estimating the electronic quenching rate coefcient is to model numerically the combustion environment. Given sufcient knowledge of the local temperature, major species number densities, and quenching cross sections, we can easily estimate the local quenching rate coefcient.29 However, the local quenching rate coefcient cannot be accurately modeled in a complex turbulent spray ame. Uncorrected LIF measurements do, however, yield relative NO concentrations throughout the ame within the errors delimited by gradients in the electronic quenching rate coefcient. The uncorrected LIF measurements of NO concentration are shown in Fig. 5 for axial heights of 5, 10, and 20 mm. Accuracy bars are shown at 10-mm increments and range from 12% to 25%. Our rst observation is that the prole shape of the LIF measurements is quite similar to that for LSF see Fig. 4 . The magnitude of the LIF measurements, however, is 34% greater than that of the LSF measurements. Possible errors caused by the effects of
20 July 1998 Vol. 37, No. 21 APPLIED OPTICS 4829

values for H2O and N2, yields a quenching rate coefcient of 6.2 108 s 1 at a ame temperature of 1800 K. With this rate coefcient used, the LIF experiments would remain 20% larger than those for LSF. Reisel et al.17 compared LSF and LIF measurements of NO in premixed and nonpremixed C2H6 O2 N2 ames, respectively, and found discrepancies of 515% between the two measurements; somewhat greater discrepancies should be expected owing to additional interferences from Mie scattering.
5. Planar Laser-Induced Fluorescence Measurements

Fig. 6. Comparison of the LSF measurements and the enhanced LIF measurements based on a scaling factor of 0.75. Note the excellent correlation that signies a uniform electronic quenching rate coefcient.

RET dynamics on the LSF measurements are not sufcient to account for this deviation. We therefore invoke the more quantitative nature of the LSF measurements with respect to electronic quenching and presume that they accurately represent NO concentrations in the LDI ame. Thus we can enhance the qualitative nature of the LIF measurements by suitably scaling the NO proles in Fig. 5. Assuming that the local quenching rate coefcient does not vary considerably throughout the LDI ame because of the intense recirculation, this experimentally based correction procedure utilizes a single point to scale the LIF proles.15 The centerline point at the 10-mm axial height location was arbitrarily chosen to perform this scaling and yields a LIF scaling factor of 0.75. The nal results are illustrated in Fig. 6, with the accuracy bars of the LSF measurements shown. The excellent correlation between the LSF and LIF proles implies that the local quenching rate coefcient in this turbulent ame is essentially constant and equal to 5.1 108 s 1. The intense recirculation zone characteristic of highly swirling ames is responsible for the well-mixed nature of this ame. Probe measurements within the recirculation zone made with a Horiba CO2 infrared analyzer and a Rosemount paramagnetic O2 analyzer indicate that this stoichiometrically fed ame is actually burning at an equivalence ratio of 0.8 owing to entrainment of ambient air. Employing the mole fractions of the major species measured above, along with calculated
4830 APPLIED OPTICS Vol. 37, No. 21 20 July 1998

We have shown that the local quenching rate coefcient in the LDI ame is fairly uniform owing to an efcient recirculation and blending of the product species. This fact permits us to consider PLIF as an additional tool by which to explore the NOconcentration eld. The excellent LIFLSF comparison offers the possibility of performing PLIF-based measurements and quantifying such measurements with a single calibrated point in the NO eld. If the detected uorescence is not plagued by Mie scattering or laser-induced interferences, i.e., O2, PAH, or UHC uorescence, then the opportunity exists to make quantitative images of NO concentration. These images are spatially correct owing to the uniform electronic quenching rate coefcient; however, they remain qualitative without an accurate estimate of this rate coefcient. The procedure for converting the PLIF images to NO concentration is similar to that of LIF. We utilized an on-line wavelength 225.58 nm resonant with the Q2 26.5 transition to excite NO molecules. An image corresponding to the on-chip summation of 2400 laser shots is then recorded. The laser is tuned next to an off-line wavelength 225.53 nm , and a similar image is recorded. The data are then processed as follows: 1 the ame luminosity is subtracted from the initial on- and off-line images by the employment of a similar image with no laser beam passing through the probe volume; 2 these on- and off-line images are normalized to the distribution of energy in the laser sheet by a 100-shot image that records Raleigh scattering in air; and 3 the normalized off-line image is subtracted from the normalized on-line image. A differential doping calibration similar to that for both the LSF and the LIF measurements was performed in a lean 0.8 C2H6 O2 N2 NO ame stabilized on a water-cooled McKenna burner. A 1 mm 1 mm region along the centerline at h 7 mm was utilized to obtain a calibration slope. For direct comparison of the PLIF data with data obtained by LSF and LIF, 1-mm squares were averaged throughout the image and horizontal stripes were extracted corresponding to h 5, 10, and 20 mm. The larger sampling volume of the PLIF measurements compared with that of the LIF and LSF measurements was chosen to increase the signal-to-noise ratio. These stripes of data were then corrected for laser beam and uorescence extinction, just as for the LIF

Fig. 7. Comparison of the uncorrected LIF measurements and the uncorrected PLIF measurements. The PLIF measurements generally fall within the accuracy bars of the LIF data, emphasizing the utility of the PLIF technique for LDI spray ames. PLIF measurements were limited to 18 mm owing to the camera eld of view.

Fig. 8. Comparison of the LSF measurements and the enhanced PLIF measurements based on a scaling factor of 0.75.

measurements. However, the PLIF data were not corrected explicitly for variations in the quenching rate coefcient. A comparison of the 1-mm spatially resolved data with the uncorrected LIF measurements reveals an excellent correlation, as illustrated in Fig. 7. Considering the potential for possible interferences from O2 and other uorescing species caused by broadband detection, the results are quite remarkable. The PLIF measurements lie within the accuracy bars of the LIF data at the majority of points. The measurements at the 20-mm axial height occurred in an intense portion of the laser sheet and were partially saturated. A correction factor was applied to these data based on accompanying PLIF measurements to determine the degree of saturation. The uncertainty in this procedure may account for the discrepancy between the LIF and PLIF data at this axial height. The success of the PLIF excitation scheme lies in the subtraction technique, which was designed to remove both the O2 interferences30 and the Mie scattering background16 within the 2-nm detection window typically used for the LIFLSF measurements in our laboratory. For the broad detection window used in these PLIF measurements, off-line wavelength excitation continues to model the Mie background accurately; however, an accurate representation of any O2 interferences cannot be guaranteed without further research. The minimal 0.05-nm spacing between the off- and on-line wave-

lengths also ensures that off-line excitation accounts favorably for any PAH and UHC interferences owing to their broad absorption features.10 Considering the nearly invariant electronic quenching rate coefcient in this ame, the PLIF data should accurately scale to within the accuracy bars of the more quantitative LSF measurements. We again chose the centerline point at the 10-mm axial height to perform this scaling, which yields a PLIF scaling factor of 0.75, the same as that for the LIF scaling. The excellent results are illustrated in Fig. 8. Because of the unique quenching attributes of the LDI ame, we found that the PLIF technique is a viable method for measuring NO variations throughout this type of spray ame. For the image to be quantied, a single calibrated point must be used to scale the proles, owing to a lack of accurate turbulence models for this combustion environment and thus to insufcient information on the local electronic quenching rate coefcient.
6. Summary and Future Research

Quantitative LSF measurements of NO concentration have been obtained in a LDI burner fueled with liquid heptane at atmospheric pressure. We have extended the current technology for LSF measurements of NO to a liquid-droplet environment by utilizing a correction procedure for Mie scattering and possible PAH and UHC uorescence. Quenchinguncorrected measurements of NO based on similar LIF or PLIF techniques have been shown to be 34% greater than those for LSF. Since accurate modeling of the combustion environment and thus the elec20 July 1998 Vol. 37, No. 21 APPLIED OPTICS 4831

tronic quenching rate coefcient cannot be performed for such a complex turbulent ow eld, the LIF and PLIF proles have been corrected for collisional effects based on a single-point scaling with the more quantitative LSF data. The excellent agreement among the LSF, LIF, and PLIF results demonstrates that the local electronic quenching rate coefcient is approximately uniform for the LDI-based spray ame. The goal of this study has been to develop a LIF technique capable of measuring quantitative NO concentrations in a high-pressure version of the LDIbased spray ame. Considering the excellent prole comparisons among the three independent LIF-based techniques presented here, we might conclude that qualitative LIF or PLIF measurements of NO at high pressure could be scaled in a similar fashion by use of a single calibrated point to produce quantitative measurements. However, this tentative conclusion depends on the continued uniform electronic quenching rate coefcient at higher pressures. Moreover, other problems may arise at higher pressures, namely, signicant O2 interferences and probable soot production within the ames. On the other hand, our atmospheric study has presented one successful method LSF of quantitatively calibrating quenching-uncorrected LIFPLIF measurements. At higher pressures the current scaling procedure based on a single LSF measurement may instead require a single chemiluminescent analyzer measurement. This research was funded by the National Aeronautics and Space Administration Lewis Research Center through General Electric Aircraft Engines, Cincinnati, Ohio. We thank our project technical monitors, Chi Lee of NASA and Claude Chauvette of General Electric. The design of the LDI module was aided by Phil Lee of Fuel Systems Textron and Mark Durbin of General Electric. Bob Tacina of NASA provided the helical swirlers. C. S. Cooper acknowledges support by a National Defense Science and Engineering Graduate Fellowship.
References
1. M. G. Allen and R. K. Hanson, Planar laser-induceduorescence monitoring of OH in a spray ame, Opt. Eng. 25, 1309 1311 1986 . 2. M. G. Allen and R. K. Hanson, Digital imaging of species concentration elds in spray ames, in Twenty-First Symposium International on Combustion The Combustion Institute, Pittsburgh, Pa., 1986 , pp. 17751762. 3. P. Andresen, G. Meijer, H. Schluter, H. Voges, A. Koch, W. Hentschel, W. Oppermann, and E. Rothe, Fluorescence imaging inside an internal combustion engine using tunable excimer lasers, Appl. Opt. 29, 23922404 1990 . 4. A. Koch, A. Chryssostomou, P. Andresen, and W. Bornsheuer, Multi-species detection in spray ames with tunable excimer lasers, J. Appl. Phys. B 56, 165176 1993 . 5. M. G. Allen, K. R. McManus, and D. M. Sonnenfroh, PLIF imaging measurements in high-pressure spray ame combustion, AIAA paper 94-2913, presented at the 30th Joint Propulsion Conference, June 1994 American Institute of Aeronautics and Astronautics, New York, 1994 . 4832 APPLIED OPTICS Vol. 37, No. 21 20 July 1998

6. M. G. Allen, K. R. McManus, D. M. Sonnenfroh, and P. H. Paul, Planar laser-induced-uorescence imaging measurements of OH and hydrocarbon fuel fragments in high-pressure sprayame combustion, Appl. Opt. 34, 6287 6300 1995 . 7. B. S. Haynes, Soot and hydrocarbons in combustion, in Fossil Fuel Combustion, W. Bartok and A. Sarom, eds. Wiley, New York, 1991 , pp. 289 326. 8. F. W. Lam, J. P. Longwell, and J. B. Howard, The effect of ethylene and benzene addition on the formation of polycyclic aromatic hydrocarbons and soot in a jet-stirred plug-ow combustor, in Twenty-Third Symposium International on Combustion The Combustion Institute, Pittsburgh, Pa., 1990 , pp. 14771484. 9. J. G. Calvert and J. N. Pitts, Jr., Photochemistry Wiley, New York, 1966 . 10. F. Beretta, V. Cincotti, A. DAlessio, and P. Menna, Ultraviolet and visible uorescence in the fuel pyrolysis regions of gaseous diffusion ames, Combust. Flame 61, 211218 1985 . 11. B. L. Upschulte, M. G. Allen, and K. R. McManus, Fluorescence imaging of NO and O2 in a spray ame combustor at elevated pressures, in Twenty-Fourth Symposium International on Combustion The Combustion Institute, Pittsburgh, Pa., 1996 , pp. 2779 2786. 12. R. J. Locke, Y. R. Hicks, R. C. Anderson, K. A. Ockunzzi, and G. L. North, Two-dimensional imaging of OH in a lean burning high pressure combustor, AIAA paper 95-0173, presented at the 33rd Aerospace Sciences Meeting & Exhibit, January 1995 American Institute of Aeronautics and Astronautics, New York, 1995 . 13. A. Cessou and D. Stepowski, Planar laser induced uorescence measurements of OH in the stabilization stage of a spray jet ame, Combust. Sci. Technol. 118, 361381 1996 . 14. D. Stepowski and A. Garo, Local absolute OH concentration measurement in a diffusion ame by laser induced uorescence, Appl. Opt. 24, 2478 2480 1985 . 15. W. P. Partridge, Experimental assessment and enhancement of planar laser-induced uorescence measurements of nitric oxide in an inverse diffusion ame, Ph.D. dissertation Purdue University, West Lafayette, Ind., 1996 . 16. C. S. Cooper and N. M. Laurendeau, Quantitative lasersaturated uorescence measurements of nitric oxide in a heptane spray ame, Combust. Sci. Technol. 127, 363382 1997 . 17. J. R. Reisel, C. D. Carter, and N. M. Laurendeau, Lasersaturated uorescence measurements of nitric oxide in laminar, at, C2H6 O2 N2 ames at atmospheric pressure, Combust. Sci. Technol. 91, 271295 1993 . 18. W. P. Partridge, Jr., N. M. Laurendeau, C. C. Johnson, and R. N. Steppel, Performance of Pyrromethene 580 and 597 in a commercial Nd:YAG-pumped dye-laser system, Opt. Lett. 19, 1630 1632 1994 . 19. C. S. Cooper and N. M. Laurendeau, Effect of pulsed dye-laser wavelength stabilization on spectral overlap in atmospheric NO uorescence studies, Appl. Opt. 36, 52625265 1997 . 20. C. D. Carter, G. B. King, and N. M. Laurendeau, Saturated uorescence measurements of the hydroxyl radical in laminar high-pressure C2H6 O2 N2 ames, Appl. Opt. 31, 15111522 1992 . 21. W. P. Partridge, M. S. Klassen, Jr., D. D. Thomsen, and N. M. Laurendeau, Experimental assessment of O2 interferences on laser-induced uorescence measurements of NO in highpressure, lean premixed ames by use of narrow-band and broadband detection, Appl. Opt. 35, 4890 4904 1996 . 22. K. Lee and B. Chehroudi, Structure of a swirl-stabilized spray ame relevant to gas turbines and furnaces, J. Propul. Power 11, 1110 1117 1995 . 23. C. S. Cooper, Quantitative laser-induced uorescence mea-

surements of NO concentration in a lean direct-injection spray ame at atmospheric pressure, M.S. thesis Purdue University, West Lafayette, Ind., 1997 . 24. R. P. Lucht, D. W. Sweeney, and N. M. Laurendeau, Balanced cross-rate model for saturated molecular uorescence in ames using a nanosecond pulse length laser, Appl. Opt. 19, 32953300 1980 . 25. W. G. Mallard, J. H. Miller, and K. C. Smyth, Resonantly enhanced two-photon photoionization of NO in an atmospheric ame, J. Chem. Phys. 76, 34833492 1982 . 26. J. R. Reisel, C. D. Carter, and N. M. Laurendeau, Einstein coefcients for rotational lines of the 0, 0 band of the NO A2 X2 system, J. Quant. Spectrosc. Radiat. Transfer 47, 4354 1992 .

27. D. D. Thomsen, Purdue University, West Lafayette, Ind. 479071288 personal communication, 1997 . 28. C. D. Carter and N. M. Laurendeau, Wide- and narrow-band saturated uorescence measurements of hydroxyl concentration in premixed ames from 1 bat to 10 bar, Appl. Phys. B 58, 519 528 1994 . 29. P. H. Paul, J. A. Gray, J. L. Durant, Jr., and J. W. Thoman, Collisional quenching corrections for laser-induced uorescence measurement of NO A2 , AIAA J. 32, 1670 1675 1994 . 30. D. D. Thomsen, F. F. Kuligowski, and N. M. Laurendeau, Background corrections for laser-induced-uorescence measurements of nitric oxide in lean, high-pressure, premixed methane ames, Appl. Opt. 36, 3244 3252 1997 .

20 July 1998

Vol. 37, No. 21

APPLIED OPTICS

4833

Potrebbero piacerti anche