CARS Temperature Measurements in a Droplet Stream Flame

-This paper describes measurmcnts of the local temperature field near a combusting stream of methanol droplets using coherent anti-Stokes Raman scattering (CARS). Synchronizing the CARS measurements wilh the droplets in the stream allows correlations between the temperature field and the droplet position. These measurcmenls describe the thermal structure of the droplet stream flame and can help validale detailed spray combustion models. The diameter of the droplels in the stream is 120 Jim, which is comparable to the droplet size in a practical spray. The droplet spacing is approximately l.Omm (8 droplet diameters). The CARS measurements indicate a flame surrounding the droplet stream located I mm (8 droplet diameters) from the stream centerline. The temperature profile shows a local minimum tem perature a the stream centerline. Moving radially outward from the centerline. the temperature rises at the flame (about 2000 K), and then falls gradually to room temperature. The local minimum temperature is more pronounced on lines through droplel centers than on lines between adjacent droplets. The rcsulls also show that the temperature profile between adjacent droplets becomes more uniform as the droplets move downstream.


INTRODUCTION
Experimental studies of the combustion of a single droplet, a droplet array, or a droplet stream are valuable because these combustion processes can duplicate essential features of realistic spray flames while avoiding the complex behavior of the spatial and temporal distributions of droplet size and velocity present in full sprays. In addition, some basic spray combustion models have been developed for combusting droplet streams Sirignano, 1988, 1989) and combusting droplet arrays (Chiang and Sirignano, 1990) where experimental verification can be useful.
The applicability to spray flames of previous single droplet and droplet stream combustion experiments is limited by larger droplet sizes than the size found in practical sprays (e.g., ~I mm in Miyasaka and Law, 1981 ; ~I mm in Xiong et al., 1984;:::: 2 mm in Brzostowski et al., 1979;~ l -4 mm in Brzostowski et al., 1981 ), and by experimental methods that preclude the measurement of temperature and concentration (Sangiovanni and Kesten, 1976;Sangiovanni and Dodge, 1978;Twardus and Brzostowski, 1978;Koshland and Bowman , 1984;Quieroz and Yao, 1989). Recently, Montgomery et al. (1990) reported temperature and concentration measurements in a multiple droplet stream flame . However, they measured the temperature field with a thermocouple so they could not obtain lhe local temperature field near individual droplets.
Measuring the thermal field in a droplet stream flame is difficult because liquid droplets affect most common temperature probes, and because the combustion regime has very steep gradients over a small spatial scale. Fortunately, coherent anti-Stokes Raman scattering (CARS) has the capability for spatially resolved, non-intrusive temperature measurements in droplet and particle laden flows (Eckbreth and Hall, 1979;Beiting, 1985;Dunn-Rankin et al., 1990). The coherent nature of the CARS signal ensures strong discrimination from the background, even with droplets or particles in the flow. 97 CARS is a nonlinear four optical wave mixing process that can be used to interrogate the energy state of an ensemble of molecules. The use of CARS as a diagnostic tool has developed with the availability of high power lasers. In the spatially resolved (or BOXCARS) configuration, the CARS signal is generated by the interation of three incident laser beams (two pumping beams and one Stokes beam) within the CARS probe volume through the third order nonlinear susceptibility of the medium under investigation. When the frequency difference between the pumping beam and the Stokes beam coincides with the frequency of a Raman-active mode of the medium (normally a vibrational-rotational transition), the CARS signal is resonantly enhanced. From quantum mechanics, the CARS energy spectrum is a predictable function of the medium temperature. Therefore, the temperature can be determined by comparing the experimental CA RS spectrum to a spectrum calculated using temperature as a fitting parameter. Complete reviews of the CA RS process (Druet and Taran, 1981;Valentini, 1985) and a discussion of CA RS as a combustion diagnostic (Eckbreth, 1988) are available in the literature.
This paper demonstrates the capabilities of CA RS to measure temperature near combusting droplets, and presents measurements of the temperature field near drop· lets in a droplet stream flame. The goal of the present work is to correlate the temperature field with the droplet position in a combusting droplet stream, and thereby provide the temperature field surrounding a burning droplet and the behavior of this field as the droplet moves downstream. The results from these experiments provide insight into the structure and behavior of spray flames, and can validate detailed numerical models (e.g., Delplanque and Rangel, 1991 ).

EXPERIMENTAL APPARATUS
The experimental apparatus uses a monodisperse droplet generator and hydrogen ignition source to generate a droplet stream flame. Temperature measurements utilize a mobile CARS system in the BOXCARS configuration, and a frequency divider circuit synchronizes the CARS measurements with the droplet generation process. rn addition, a schlieren system, using the CARS laser as a synchronized light source, provides qualitative images of the droplet stream and flame. Further details of the experimental apparatus appear in the following subsections.

CARS System
The CARS system is set up on a two level optical cart, and the optical layout on the cart is shown in Figure I. The 1064nm laser beam from a JO Hz injection seeded Nd:Y AG laser (Spectra-Physics DCR-3) is frequency doubled through SHG 1 to provide a primary 532 nm laser beam. The primary 532 nm laser beam splits at the 30/70 beamsplitter BSI. The 30% leg pumps the dye laser amplifier DC2. The 50/50 beamsplitter BS4 splits the 70% leg into the two pumping beams of the BOXCARS. The residual I 064 nm radiation from the first frequency doubler passes the harmonic separator HSI and is doubled by SHG2 to provide a secondary 532 nm laser beam. The secondary beam pumps the dye laser oscillator DCI. We use a 0.55 x I 0-4 molar mixture of Rhodamine 640 dye (Exciton) in methanol for both the oscillator and the amplifier stages of the dye laser. This mixture produces a broadband Stokes beam centered at 607 nm for the BOXCARS system. The transmitting optics include mirrors, a prism delay loop (P3, P4), telescope lenses to collimate the laser beams (L3, L4, LS, L6), and the 250 mm focal length projection lens (L 7). The projection lens    Table I is below Table 2 on a moveable cart.
Mirrors M2, M3. M4. and MS represent periscopes to bring laser light from Table I to Table 2. focuses the two 532 nm pumping beams and the dye laser beam to a probe volume in the center of the droplet stream flame. The beam crossing angle is 3°. A second 250 mm focal length lens (L8) collects the CARS signal and passes it through filters to a spectrometer (SPEX 1702/04). An intensified diode array (Princeton Instruments IRY-1024 /PDA) records the CARS spectrum al the exit of the spectrometer.

Drop/er Stream Flame
A vibrating orifice droplet generator, similar to the one described by Berglund and Liu (1973), generates a steady monodisperse droplet stream. A hydrogen diffusion flame at the tip of a 27 gauge needle ignites and pilots the droplet stream flame. Both the droplet generator and the hydrogen flame reside on a three dimensional traverse mechanism which allows a full spatial mapping of the temperature field . The distance between the droplet generator and the hydrogen flame remains unchanged during the traverse, maintaining constant boundary and initial conditions for the combusting droplet stream. A schematic of the droplet generator is shown in Figure 2. Pressurized nitrogen drives the liquid fuel out of the reservoir and through a 50 µm diameter orifice. The orifice vibrates with the piezoelectric crystal driven by square waves from a function generator. At particular frequencies, the crystal vibration breaks up the fuel jct issuing from the orifice into a monodisperse droplet stream . For these experiments, we use a nitrogen pressure of 20 psig, methanol liquid, and a vibration Schematic showing the measurement locations and the nondimcnsional spatial variables used to present 1 he results.

Electronics
A frequency dividing circuil synchronizes lhe Y AG laser firing wilh the droplet generator. The synchronization ensures that the CA RS probe volume remains in the same position relative to the droplets for each sequence of measurements. The synchronous measurement allows the correlation of the d roplet position with the temperature field. A function generator sends 20 V peak-to-peak square waves to the piezoelectric crystal to vibrate the drop generator orifice. The TTL output of the function generator is input to the 12 bit frequency divider, and the o utput from the divider triggers the Nd: Y AG flash lamps. In our experiments, the divider drives the laser al 9.6 Hz, which is close to the I 0 Hz design operating frequency of the laser. The laser can run between 8.5-11.5 Hz without significant power loss. Figure 3 shows how the schlieren syslem employs the Nd:YAG laser as light source. As shown in the figure, the laser light comes from BS2afterM11 is removed. Because the laser beam has a "doughnut" intensity profile, an iris picks off a I mm diameter uniform intensity spot from the expanded beam. A prism and mirror direct the light from this spot through a beam expanding lens (25 mm focal length, 25 mm diameter). An aehromat collimating lens (350 mm focal length, 63 mm diameter) projects the

Experiment Resolution
From visual observation, the droplet stream flame is only about 2 mm thick. To effectively map the temperature field of such small dimension requires high spatial resolution. The droplets act as edge markers for measuring the diameter of the probe volume. After measuring the size of the droplets photographically, we translate the CA RS probe volume across the droplets while recording the region producing an audible and visible laser induced breakdown. The total traverse distance where breakdown occurs less the droplet diameter is twice the CARS probe diameter. Using this method, the CA RS probe volume is I 00 ± 15 J'm. The length of the CARS probe volume is measured by observing the resonant CARS signal while translating a stack of glass slides through the beam overlap region. The extreme points where the resonant signal appears and disappears marks the extent of the CARS probe volume. Using this method, our probe volume is approximately 2 mm long, with most of the signal generated in the central 1.5 mm region. Figure 4 shows the relative sizes and locations of the CARS probe volume, the droplet, and the flame. To obtain accurate temperature measurements, we traverse the droplet generator perpendicular to the long axis of the CARS probe volume as indicated in Figure 4. Between the droplet surface and the reaction zone (<I mm from the droplet) the CARS probe volume is large relative to the spatial dimensions of the region. Near the droplet, therefore, the temperature measurements represent a spatially averaged temperature. Because CARS is a non-linear process, however, the spatial average is not simply a density weighted average. Instead, a sum of contributions from the third-order susceptibility at different temperatures produce a single composite CARS spectrum. T his composite spectrum is then tit to a single temperat ure. Calculations using the Sandia CARS Code indicate that for a 2 mm CA RS probe length the spatial averaging smooths out sharp temperature peaks near the Oame and produces significant temperature errors only very near the droplet ( < 2 diameters). The spatial averaging effects diminish with increasing distance from the droplets and are negligible at radial distances larger than 15 drop diameters. Similar effects of CA RS spatial averaging in a turbulent flame sheet have been reported by Shepherd et al. ( 1990).
. The droplets in the stream flame move downstream at 10-15 m/s. The Nd:YAG laser pulse, however. is only 15 ns. Consequently, the droplets are effectively motionless during each CA R S temperature measurement. Multiple exposure (500 laser shots) backlit droplet images indicate that, relative to the CARS measurement volume, the droplet position fluctuates less than I 00 µm in the axial direction and less than 20 µm perpendicular to the droplet stream.

EXPERIMENT SUMMARY
The experimental conditions for the results reported in this paper are: methanol fuel, a 50 µm diameter orifice in the droplet generator, a liquid fuel of 20 psig, a hydrogen flow rate of 15 cc/min to the ignition pilot flame, and a droplet generator vibration frequency of 9.8 kHz. These conditions produced 120µm diameter droplets spaced 1.0 mm apart. We measure the droplet size and spacing with a microscope and reticle. The droplet spacing measurement is verified by focusing the CARS beams onto one droplet, and then recording the vertical traverse necessary to focus the beams onto the adjacent droplet. The digital divide circuit fired the Y AG laser at 9.6 Hz. The measurement location nearest the orifice is 70 mm downstream of the hydrogen flame. Thermocouple measurements indicate that at this distance the hydrogen flame does not have significant effects on the droplet flame temperature. The results include 4 radial temperature profiles. Figure 5 shows the geometric relation between the droplet stream flame and these measurement locations, where r is the radial coordinate from the droplet center outward. The nondimensional coordinate Y measures distance along the droplet stream axis, and is the distance from the droplet center divided by the droplet spacing ( Y = y/ H ). For Y = 0, corresponding to the plane of the droplet center, measurements begin IOOµm from the droplet center to avoid laser induced droplet breakdown. There are 16 points in each temperature profile. To improve the signal-to-noise ratio, each CARS measurement represents an average of 5 laser shots, and each data point is an average of 100 measurements.

Spectrum Fitting
To extract temperature from CARS measurements requires a spectral fitting procedure that matches experimental CARS spectra to theoretical CARS spectra using temperature as a fitting parameter. Our fitting procedure employs a library of cheoretical nitrogen CARS spectra over the range from 250-3000 K in 50 K increments. The Sandia CARS code CA RSFT (Palmer, 1989) generates the spectral library by assuming a n instrument function deduced from the fit to a room temperature nitrogen CA RS spectrum measured in the CARS system. The CARSFT code has been extensively used a nd validated with CARS measurements in combustion systems (e.g ., Farrow et al., 1984;Lucht er al., 1987;Boyack, 1990). Preparing the experimental CARS spectra for fitting requires a background subtraction, normalization by the dye laser profile, and taking the square root of the result. A modified version of the Sandia quick fitting code FTCARS (Palmer, 1990) fi nds the temperature whose associated theoretical spectrum mostly closely matches the experimental spectrum. The quick fitting program linearly interpolates between temperatures represented in the spectral library to find the least squares best fit between measured and calculated spectra. Rejecting spectra with poor signal-to-noise ratio coupled with visual checks of the fit for several spectra at each measurement location helps ensure reliable experimental results. With this procedure, the absolute temperature error is less than IOO K, and the standard error of the temperature at each measurement location is Jess than 25 K.
The methanol concentration varies significantly from the droplet surface to the flame front. Because hydrocarbons have large non-resonant susceptibility (Farrow, er al., 1987), this variation can affect the CA RS temperature fit. The fit can be improved by allowing the nonresonant susceptibility to vary as a fi tting parameter (Hall a nd Boedeker, 1984). value of nitrogen to the best fit value changes the predicted temperature by less than 75K. Figure 6 shows a typical spectral fit for measurements less than 9 mm from the stream centerline. The small scale random fluctuations come from noise in the detector. While the fit is very good, the first hot band is slightly more pronounced in the measured spectrum than in the calculated spectrum, and the fundamental band is slightly narrower in the measurement than in the calculation. This vibrationally hot and rocacionally cool behavior was noted by Snelling et al. ( 1989), who ascribed the phenomenon to stimulated Raman pumping to the vibrationally excited v = I level.
In their analysis, Snelling et al. found that stimulated Raman pumping affected high temperature spectra by less than 50 K. However, to ensure that stimulated Raman pumping does not affect the droplet stream flame CARS measurements, we take a series of spectra at a single point in the flame while varying the laser power. The temperature fits to these spectra differ by less than 50 K , indicating that stimulated Raman pumping does not affect high temperature measurements significantly. For measurements far from the droplet, where the temperature is less than 1400 K, the stimulated Raman pumping is more pronounced, making accurate analysis of the cool spectra difficult. Figure 7 shows two examples of coo! spectra with significant stimulated Raman pumping. While it is possible to fit only the fundamental band of these spectra to get the temperature, such fits can be unreliable, particularly for temperatures above 800 K . Consequently. this paper does not report measurements in the cooler regions of the flame . Reduced laser power coupled with improved signal collection in future experiments should help reduce the stimulated Raman pumping effect.
local Temperature Field Characteristics only to the side of the droplet stream, not between droplets, indicating collective combustion. ln similar experiments with larger droplets, Twardus and Brzustowski (1978) also concluded that their droplet stream burned collectively, and not as individual droplets.
It is interesting that the Y = 0 case in Figure 8 indicates that the temperature profile is concave down between the droplet surface and the flame . This qualitative shape agrees with the classical description of individual droplet combustion in Glassman (1987) and with detailed computations of droplet combustion by Rangel and Fernandez-Pcllo ( 1984). Figure 11 compares the shape of the temperature profiles at Y = 0 and Y = 0.5 by overlaying the smooth splines from Figures 9(a) and 9(b). The comparison shows that the maximum flame temperature is nearly the same for the on-droplet and between-droplet cases. The between-droplet profile is broader and more uniform, but the location of the maximum temperature coincides with the maximum of the on-droplet profile. This result suggests that at the 70 mm location the flame stand-off distance from the centerline does not depend on the location of the droplet. Figures 12 and 13 are similar to Figures 8 and 9, except that the measurements are taken 100 mm downstream of the ignition source. At this location, the scatter in the data is higher, and the overall temperatures are slightly higher than at the 70 mm location, but the shape of the profiles at the two locations are similar. Figure 14 compares the Y = 0.5 profiles between the 70mm and lOOmm measurement location.
The downstream profile has slightly higher temperature, a more uniform shape, and a maximum temperature farther from the droplet stream axis. These results suggest that the flame broadens as it progresses downstream . The schlieren photograph of Figure I 0 also indicates a slight broadening of the flame . The increase in temperature, and the more uniform, broader temperature profile, probably results from combustion heat release convected downstream.

SUMMARY
This paper describes a synchronous CA RS apparatus for measuring the temperature profile in a droplet stream flame. Using this apparatus we report temperature profile measurements in a droplet stream flame along lines perpendicular to the direction of the droplet travel. Profiles passing through the droplet position, and halfway between adjacent droplets are included. The principal findings of these experiments are: • Temperatures in the between-droplet profile reveal a local minimum temperature at the droplet stream axis. This minimum, and qualitative schlieren photographs, suggest that the droplets burn collectively, rather than as individual droplets. • The on-droplet profiles show a steeply rising temperature from near the droplet surface to a maximum temperature at the flame sheet, followed by a temperature decrease to room temperature. • As the droplets move downstream, the flame temperature increases slightly and the temperature profile becomes broader and more uniform. The data scatter is higher in the downstream measurements suggesting unsteady behavior of the droplet stream flame.
Further work requires improved performance of the CARS system and synchronization system to reduce the fluctuations in the droplet stream that limit the spatial resolution of the measurements. With these improvements, it will be possible to measure the temperature profile between drops along the stream axis, and lo reveal further the thermal structure of the droplet stream flame.