Paper # 070DI-0092 Topic: Diagnostics 1. Introduction

Transcription

1 Paper # 070DI-0092 Topic: Diagnostics 8 th U. S. National Combustion Meeting Organized by the Western States Section of the Combustion Institute and hosted by the University of Utah May 19-22, 2013 Quantification and Accuracy of a CMOS-Based Raman/Rayleigh Scattering Imaging System for High-Speed Measurements in Flames Kathryn Gabet 1 Frederik Fuest 1 Jeffrey Sutton 1 1 Department of Mechanical and Aerospace Engineering, The Ohio State University, 201 W 19 th Ave Columbus, OH In this paper we will describe recent work in our laboratory towards the quantification of a high-speed (> 10 khz) combined 1D Raman-Rayleigh scattering imaging system utilizing CMOS-based cameras. While our previous work has demonstrated the ability to acquire high-speed Raman/Rayleigh scattering images using a pulse burst laser system (Gabet et al., 2010), further study of the acquisition system is necessary for quantitative results. For the majority of high-speed imaging experiments, CMOS cameras are used because conventional CCD cameras cannot operate at sufficiently high acquisition rates to capture the full range of temporal scales and fluctuations in turbulent flows. Unlike CCD cameras, which typically have uniform and linear pixel response, each pixel on CMOS cameras has a unique response which needs to be characterized individually (Patton et al., 2011; Weber et al., 2011). In addition, CMOS cameras are known to exhibit increased levels of noise, particularly when coupled with an image intensifier. Careful examination and calibration of CMOS-based acquisition systems is of particular importance to understand their limitations and accuracy for low-signal applications such as Raman scattering. This paper will focus on quantifying the precision and accuracy of Raman/Rayleigh scattering measurements of major species, temperature, and mixture fraction using our CMOS-based 1D Raman/Rayleigh system in a series of near-adiabatic H 2 /air flames and turbulent H 2 /N 2 jet flames. A detailed analysis of the spectral response and signal-to-noise ratio (SNR) of major species (H 2, N 2, H 2 O, and O 2 ) and temperature is presented. The ability to measure single-shot scalar values accurately in turbulent flames is assessed by comparing scalar results in the DLR H3 (50% N 2 /50% H 2 Re=10,000) turbulent jet flame to previous work (Meier et al., 1996). The ultimate goal of our research is to measure the time-varying profiles of all major combustion species and deduce temporally resolved mixture fraction profiles in turbulent combustion environments. 1. Introduction Advancements in energy-conversion systems including increases in efficiency and decreases in harmful emission output will require a better understanding of chemical and physical processes occurring within turbulent combustion environments. Towards this goal, a number of standard, laboratory-scale turbulent flames appropriate for fundamental experimental and computational investigation were identified by the International Workshop on Measurement and Computation of Turbulent Nonpremixed Flames ( TNF Workshop, n.d.). Since the first TNF Workshop in 1996, numerous studies on these flames have provided a wealth of information on the underlying chemistry and physics of turbulent combustion. Of particular relevance to the current study are the complete statistical descriptions of mixture fraction, temperature, and major species concentration collected through the use of combined spontaneous Raman scattering (SRS) and Rayleigh scattering. These measurements (both single-point and line-imaging measurements) have helped provide valuable insight into species transport, mixing, and turbulence-chemistry interaction in turbulent flames (Barlow & Frank, 1998; Barlow et al., 2009; Barlow & Karpetis, 2005; Fuest et al., 2011; Geyer et al., 2005; Karpetis & Barlow, 2002; Meier et al., 2000; Meier et al., 1996; Wang et al., 2007; Wehr et al., 2007). Due to the relatively low signal strength of Raman scattering, ultra-high pulse energies typically are required and the use of solid state Nd:YAG or flashlamp-pumped dye lasers are necessary. These lasers typically are limited to pulse repetition rates of < 30 Hz, limiting the single-shot data available through the TNF to temporally uncorrelated statistical descriptions of the turbulent flames. Turbulent combustion processes are highly unsteady in both time and space. Therefore, in order to explore the interplay between turbulence and flame chemistry more thoroughly, temporally and spatially correlated species,

2 temperature, and mixture fraction measurements are highly desired. Achieving appropriate temporal resolution requires that the scalar measurements are taken at a sampling rate much faster than the characteristic time scales of the turbulent processes being studied (e.g. >>1kHz). While recent advances in diode-pumped, solid-state (DPSS) lasers have made it possible for a subset of diagnostics such as PIV or OH PLIF to be extended to khz-rate applications they do not output pulse energies sufficient for SRS measurements (Böhm et al., 2011; Thurow et. al, 2013). Using pulse burst technology (Thurow et al., 2004; Wu et al., 2000), our laboratory has demonstrated the ability to generate a series of high-energy pulses in rapid succession at repetition rates >10 khz suitable for low-signal applications such as planar Rayleigh scattering and 1D Raman/Rayleigh scattering imaging (Patton et al., 2011; Patton et al., 2012, Gabet et al., 2011, Fuest et al., 2012). While our ability to collect high-speed SRS signals in turbulent non-reacting flows was demonstrated using a previous generation of the pulse burst laser system (Gabet et al., 2010), a new high-energy pulse burst laser system (HEPBLS) provides ultra-high pulse energies (up to 1 Joule per pulse at 532 nm) sufficient to make khz-rate SRS measurements in turbulent combustion systems (Fuest et al., 2012). With the new HEPBLS laser technology in place to enable khz-rate SRS, the focus of this paper is to characterize the high-speed acquisition system necessary for such measurement; in particular, focus will be placed on assessing the ability to make quantitative measurements of temperature, major species concentrations, and mixture fraction using complementary metal oxide semiconductor (CMOS)-based camera systems. Traditional CCD-based cameras typically are limited to acquiring a few images per second, making them unsuitable for temporally resolved image acquisition. In this study, CMOS-based camera systems are used, which allow for recording rates of several thousand frames per second. The primary difference in the two camera architectures lies in how information is read from the sensor. On a CCD sensor, each pixel converts photons to electrons and shifts its charge in a serial manner to the neighboring pixel acting as a shift register. The last capacitor in the array dumps the accumulated charge into an amplifier where the charge is converted to a voltage, sampled through an A/D converter, and stored in memory. The singular charge-to-voltage conversion creates a uniform response throughout the entire sensor and reduces camera read noise. In contrast, each pixel of a CMOS chip converts charge to voltage directly and then reads out the information through A/D conversion individually. While this parallel conversion allows for much higher acquisition rates, it often leads to variations between irradiance and pixel count throughout the CMOS sensor, which manifests itself as sensor non-uniformity (Thurow et al., 2013; Weber et al., 2011). In addition to this non-uniformity, CMOS sensors suffer from lower quantum efficiencies (as a portion of each pixel contains the voltage converting hardware), lower resolution (as use at ultra-high speeds limits the maximum number of active pixels on the sensor), and smaller available dynamic range (to the authors knowledge, commercial CMOS cameras are limited to 12 bit). Finally, it is noted that if a high-speed image intensifier is coupled to the CMOS camera, additional issues such as photocathode depletion at high acquisition rates, non-local imaging effects in regions with strong gradients, and spurious noise arise ( Weber et al., 2011). Thus, the quantification of Raman/Rayleigh signals using CMOS-based detection is not as straightforward as in the case of low-repetition rate acquisition using well-characterized scientific-grade CCD cameras. This paper aims to investigate the uncertainty in Raman/Rayleigh scattering measurements due to instrument and shot noise, as well as the limitations of CMOS-based detection for quantitative khz-rate measurements of major species, temperature, and mixture fraction in turbulent flames. Results from near-adiabatic, laminar H 2 -air flames and turbulent H2/N2 flames will be presented. In the case of the turbulent flames, the current measurements will be compared to previous measurements from the German Aerospace Center (DLR) Stuttgart available through the TNF database.(meier et al., 1996) 2. Methods Raman and Rayleigh Scattering Line Imaging System The Raman and Rayleigh line imaging system at Ohio State is similar in design to those used at Sandia National Laboratories (Barlow et al., 2009; Fuest et al., 2011; Karpetis & Barlow, 2005) and DLR Stuttgart (Meier et al., 2000; Wehr et al., 2007). Though the collection optics and acquisition system can be used with any laser source, results discussed in this paper were obtained using a commercially available Q-switched Nd:YAG laser at a 10Hz repetition rate with pulse energies averaging approximately 700 mj. Future work will involve the use of the new HEPBLS (Fuest et al., 2012) for 10-kHz measurements. To avoid optical breakdown in the measurement volume, the 8-ns laser beam was stretched to an approximate width of 90 ns using a two-leg pulse stretcher similar to that described in (Kojima & Nguyen, 2002). The overlapping of the beam within the pulse stretcher was ensured through the use of gimbal mounts controlled by motorized linear actuators (Newport NSA12) on the final mirror of each leg. The stretched beam was then focused into the measurement volume with a 750 mm plano-convex spherical lens, which is also motorized in the 2

3 vertical direction in order to control beam stability throughout testing. The final spot size was approximately 135 µm 1 (full width at half max) or 300 µm (1/e 2 ). The system (shown in Figure 1a) collects and focuses the Raman- and Rayleigh-scattered light through a pair of 150 mm-diameter achromats (Qioptic Linos Phototonics) into a single spectrometer in order to ensure simultaneous measurements along the same spatial profile for temperature and species concentrations. The scattered light is passed through a polarizing filter to minimize stray light and collected into the spectrometer through a 7-mm tall, adjustablewidth slit. The high-throughput transmission imaging spectrometer is similar to that described in Barlow et al. (2009), but has been modified to take advantage of the short gating using the CMOS camera system. The spectrometer (shown in detail in Figure 1b) consists of five commercially available camera lenses and numerous filtering optics. After the first lens, which is used to collimate the scattered light, a dichroic beam splitter is used to reflect the Rayleigh light onto the first CMOS camera. The Raman-scattered light is transmitted through the dichroic beam splitter and filtered through two 532 laser-line filters (Semrock) to remove any residual Rayleigh signal before passing through a volume phase holographic (VPH) transmission grating (Kaiser Optical, 1200 lines/mm, 22º incident/refracted angles, >85% efficiency from 550 nm to 680 nm) which spectrally disperses the species-specific Raman-shifted light onto an intensified (LaVision HS-IRO) CMOS camera (Vision Research, Phantom 710). In the experiments shown here, the effective magnification is 1.02 and 1.36 on the Rayleigh and Raman channels, respectively. Data was processed along a 3.6mm line segment in order to ensure similar accuracy along the line. Figure 1 Raman/Rayleigh Experimental Setup. a) Experimental setup for 1D Raman/Rayleigh scattering imaging and b) a detailed schematic and optical layout of the high-throughput, imaging spectrometer. Note that in this experiment, the 532 nm band-pass filter (4) was removed and a second 532 nm notch filter was placed in the Raman optical path to ensure the removal of all Rayleigh light from the collected Raman signals. Laminar Flames A set of well-characterized, near-adiabatic flames were used as calibration sources for the Raman/Rayleigh system (procedures discussed below) and as a way to assess the shot and detector noise contributions to the accuracy of the measurements. The flames were stabilized above the surface of a 25.4 mm x 25.4 mm matrix ( Hencken ) burner (Prucker et al., 1994) with a 35 SLPM nitrogen coflow surrounding the flame to shield the probe volume from dust and disturbances. The H 2 /air flames had volumetric air flow rate of 33.4 SLPM and H 2 flow rates ranging from 2-27 SLPM corresponding to measured equivalence ratios ranging from Φ=0.2 to 1.9. The line measurements were taken 18mm above the surface, where heat loss to the burner surface is minimized and the conditions are nearly described by adiabatic and equilibrium. Turbulent Flames The turbulent jet flame chosen for this study is the DLR H3 flame, which serves as one of the standard test cases within the TNF workshop. The H3 flame is a 50% N 2 /50% H 2 mixture by volume (flame DLR H3) that issues from a circular tube with a 7.75 mm diameter (D) into an annular co-flowing stream of air (30 cm x 30 cm, 0.3 m/s). The mixture issues from the fuel tube at 34.8m/s, corresponding to a Reynolds number of 10,000 based on tube diameter. The 1 Though the system resolution is 135 µm, data was processed with a spatial resolution of 300 µm in order to better replicate data from (Meier et al., 1996). 3

4 H 2 and N 2 flow were measured with Alicat mass flow controllers, which should be accurate to 1-2%, though this has not been calibrated in our lab. 1D imaging measurements were performed at axial distances of x/d = 10 (centered at radial positions, r = 0.0, 1.0, 2.0, 4.0, 6.0, 8.0, 10.0, 12.0, 14.0, 16.0, 18.0, and 20.0 mm), x/d = 20 (r = 0.0, 1.5, 3.0, 4.5, 6.0, 7.5, 9.0, 10.5, 12.0, 13.5, 15.0, 16.5, 18.0, 19.5, 21.0, and 22.5 mm), and x/d = 40 (r = 0.0, 5.0, 10.0, 15.0, 20.0, 25.0, 30.0, 35.0, 40.0, 45.0, 50.0, and 55.0mm). Data Evaluation The Rayleigh and Raman signals from the major species (H 2, O 2, N 2, H 2 O) and temperature were calibrated in cold flows and in laminar flat Hencken flames (described above). A dark image background taken with the lens caps on was subtracted from both cameras before their signals were corrected for detector nonuniformity, software binned to correspond to Sandia hardware binning (F. Fuest et al., 2011) to allow the use of the Sandia polynomials, and corrected for laser pulse energy fluctuations via pulse energy measurements taken with a photodiode, although this correction is a subject of uncertainty at this point and will be improved in the future with a new high-speed energy meter. The Raman signals were corrected for background signal and crosstalk from other species using the hybrid method described in (Fuest et al., 2011) where the concentrations of major species are determined through a matrix inversion process relying on a combination of theoretically calculated, temperature-dependent spectra and a set of experimentally determined calibration factors whose temperature dependences is determined by a set of empirical polynomials created at Sandia National Labs. It should be noted at this point, that corrections for sensor nonlinearity have not been made, although the authors feel this is another important contributor to the current uncertainties. The species concentrations are then used to estimate the Rayleigh cross section needed for calculating the temperature. This process is then iterated upon until the Rayleigh or perfect gas temperatures (as determined from the Raman scattering measurements) converge. For the presented results, the mixture fraction (ξ), is determined by a modified form of the Bilger formulation appropriate for hydrogen-based flames shown in (1). (YH -Y H,ox )/2w H -(YO -Y O,ox )/w O ξ= (1) (Y -Y )/2w -(Y -Y )/w H,fuel H,ox H O,fuel O,ox O Precision and accuracy of the combined Rayleigh/Raman systems are reported separately. While the precision is generally dependent on the specific experimental setup, the accuracy (consider the mean value of a variable) depends more on the quality of the calibration, temporal drifts of system alignment during measurement, and uncertainty due to crosstalk corrections (Barlow et al., 2009; Meier et al., 2000, 1996). Since the CMOS-camera-based images are software binned rather than binned on chip, it could be argued that less temporal drift occurs as it can be corrected at any time. This leaves quality of the calibration and uncertainty due to crosstalk corrections, both of which are dependent upon calibration procedure. As the data reduction methods used in this study are very similar to those at Sandia National Labs and the Sandia calibration curves are used, accuracy of measurements in this study due to the data reduction procedures (and in the absence of SNR Flat Flame Condition Relative RMS (%) 50 Φ= Φ= Φ= Φ= T ray 57 Φ= phi 12 Φ=.33 8 Table 1. Table of Representative Signal-to-Noise Ratios. Signal-to-noise ratios are taken using the high-speed system at Ohio State in H 2 /air Hencken flames with spatial resolution of 300µm and averaged over 300 shots. Figure 2. Ramses Fit to Spectral Data. Shown above are a comparison between the measured (green) spectral profiles at a single spatial position and calculated spectra (red) summed for N 2, O 2, H 2 O, and H 2 from Ramses 3.0. a) Data taken in air at 300K. b) Data taken in a near-adiabatic H 2 /air flame with φ=0.33. c) Data taken in a near-adiabatic H 2 /air flame with φ=

5 system-specific factors that are the focus of this paper) should be similar to those obtained in the Sandia system. However, it should be noted that since the calibration curves are experimentally determined at Sandia, they implicitly have system-specific characteristics that are different than in the OSU system. Thus, there is likely to be uncertainty in the current measurements based on using the Sandia-derived calibration factors. The precision of the current measurements should be lower than previously reported data with high-resolution, low-noise, CCD cameras at low repetition rates because of the use of the high-speed, intensified camera system. The focus of this paper is not to compare data quality with the Sandia system, but to determine the feasibility and limitations of data collected with a high-speed imaging system such that khz-rate SRS measurements can be reduced to accurate scalar values. In this study, we report the measurement precision in a similar manner as in previous studies (Barlow et al., 2009); that is, we report the scalar RMS fluctuation divided by the scalar mean in the series of laminar, flat flames. 3. Results and Discussion Near-adiabatic Flame Calibrations The hybrid method for data reduction described in Fuest et al. (2011) requires RAMSES lookup tables which are theoretically calculated spectra using the Rayleigh-channel line shape in air. The Ramses spectra shown in Figure 2 were generated by summing the Ramses-calculated spectra for individual species that had been scaled to fit corresponding peaks in the measured Raman spectra. The calculated spectra are plotted with a solid red line while the measured Raman spectra (minus a dark background) are shown in the green dotted lines. While the calculated spectra generally fit the shape of the measured Raman spectra, there is an artificial baseline in the experimental data. Though it is present in all the spectral data, it is easiest to see in the rich flame spectra shown in Figure 2c. While its origin is unclear, this baseline is likely due to non-localized gradient blurring from the intensifier, as described in (Gordon et al., 2009). Using the calculated Ramses tables and the Sandia polynomials, the hybrid matrix for data reduction was created by adjusting species-specific calibration and crosstalk factors to match known laminar flame conditions in the Hencken burner. Since flame conditions can vary based on room temperature, pressure, and flow controllers, the calibration factors are adjusted to match the overall shape of the adiabatic flame curve and air, rather than a specific flow condition. Results from the laminar, near-adiabatic flames plotted versus fuel-to-air ratio are shown in Figure 3. Figure 3a shows the mean species mole fraction and temperature data averaged over 3.6 mm of the probe volume and 300 laser pulses, Figure 3b shows the data scatter (instantaneous measurements), and for reference, Figure 3c shows the scatter from data taken in the same Hencken burner at Sandia National Laboratories. As noted in Fig. 3, the mean values match the expected adiabatic flame temperatures and species mole fractions very well; however, the laminar flame results exhibit noticeable scatter. In addition, while the scatter using the high-speed imaging system at Ohio State is larger than that in the Sandia data, this is expected. The system at Sandia uses 1.8 J/pulse and high-resolution, low-noise, CCD cameras with hardware binning. The data is not meant as a comparison, but rather a proof of our ability to process the Raman data using the hybrid reduction approach. Note that this particular data set (taken at Sandia) and processed at Ohio State has similar precision as compared to reported values from Sandia (Barlow et al., 2009). As one example, the temperature precision at T ~ 2200 K is calculated as 0.73% as compared to 0.75% in a methane flat flame as reported in Barlow et al. (2009). The signal-to-noise rations (SNR) and relative RMS values for various scalars acquired using the high-speed imaging system are shown in Table 1. A focus of near-term future work will be to retake this data with a new energy-monitoring system (a possible contributor to the scatter) and correct the data for sensor nonlinearity, both of which should increase the signal-to-noise (precision). Characterization of Turbulent Flames In order to assess the capabilities of our high-speed Raman system to acquire instantaneous, 1D line images in turbulent flames, we performed single-shot Raman/Rayleigh line measurements in the DLR H3 flame, which serves as a standard turbulent jet flame within the TNF database (DLR H3). The 50% N 2 /50% H 2 by volume flame issues from a tube (D=7.75 mm) at a velocity of 34.6 m/s, which corresponds to a Reynolds number of 10,000. The H3 flame has been the subject of several experimental studies at DLR, including detailed Raman/Rayleigh scattering measurements for major species concentration and temperature. A scatterplot of the major species mole fractions and temperature versus mixture fraction measured at Ohio State (red) and DLR Stuttgart (blue) is plotted in Figure 4a (Meier et al., 1996). For comparison, the adiabatic flame calculations are also plotted (black line). Overall the high-speed system measurements in the H 2 /N 2 flame match both the adiabatic flame data and the point-based DLR measurements reasonably well, though there is increased scatter in the Ohio State data. This is expected as the relative RMS values in the high-speed CMOSbased system are larger than those reported in the PMT-based system at the DLR. While some outliers fall above the adiabatic calibration curve, the majority of the scatter falls within the range of error when both the accuracy and precision errors are taken into account. The accuracy errors for the turbulent flow are expected to be slightly higher than for laminar flows, as the mass flow controllers have not been calibrated. 5

6 Figure 5 shows conditionally averaged values of temperature and major species mole fractions plotted versus mixture fraction measured at Ohio State (red) and DLR Stuttgart (blue). For comparison, the adiabatic flame calculations also are plotted (black line). As in the scatter plots, the OSU values agree quite well with both the DLR data and adiabatic flame curves, with a few notable exceptions. The largest discrepancies between the DLR and OSU data occur on the nitrogen line nearest to the fuel exit. While this discrepancy is within error (only about 4%) and could simply be a calibration error, it disappears further from the fuel exit. One possible source of this error is the fuel composition. Since values of hydrogen are high near the fuel exit, it is possible that the fuel mixture is comprised of slightly more than 50% hydrogen (and less than 50% of N 2 ). This would also account for slightly higher than adiabatic temperature values in the scatter, as a higher hydrogen concentration in the fuel would raise the expected adiabatic flame temperature. Another noticeable difference between the DLR and OSU data is a slight baseline in oxygen values at higher mixture fraction. While the DLR data essentially approaches zero at higher mixture fraction, the OSU data has a minimum of approximately 1% oxygen. This error is likely due to the false baseline present in the Raman spectra. Future work will examine the root causes of these errors including gradient blurring due to the intensifier, crosstalk between O 2 and rotational hydrogen, and sensor nonlinearity. Figure 6 shows plots of the mean values and RMS of mole fraction, temperature, and mixture fraction versus radial positions at axial positions of x/d=10, 20, and 40. Means (solid red symbols) and RMS (open red symbols) taken in the high-speed Raman/Rayleigh system at Ohio State along the center 3.6 mm (binned over 300 µm increments) are plotted along with the point data taken at the DLR Stuttgart (blue symbols). It is seen, that in general, the line measurements from the high-speed imaging system agree very well with the DLR point-based measurements in both mean and RMS value. While there is a 9% error between the two systems in the mixture fraction under the richest conditions, this seems to be well within the range of uncertainty of the two systems. It is noted that this discrepancy between the two data sets diminishes with increasing axial position, thus giving credence to the proposition that image blurring due to the intensifier is a large factor contributing to the uncertainty in the measurements. At axial positions further downstream, length scales increase and spatial gradients are less, thus pixel-to-pixel intensity values are less subject to blurring errors. 6

7 XO 2 O XN 2 TRay (K) Fuel-to-Air Ratio (φ) Fuel-to-Air Ratio (φ) Fuel-to-Air Ratio (φ) Figure 3. Measured Temperature and Mole Fractions in Near-Adiabatic Flat Flames. Shown in the left column are mean values (red symbols) for the temperature and mole fractions measured in the near-adiabatic H 2 /air flames averaged over 300 images along a 2.3 mm line segment (limited to match available data taken at Sandia), 18 mm above the surface of the burner. Scatter plots of the single-shot values are plotted in the middle column. For comparison, the right column shows scatter plots of data taken at the Combustion Research Facility at Sandia National Laboratories using the same OSU Hencken burner and processed at Ohio State (blue symbols). Measured data is plotted versus calculated adiabatic flame conditions (black line). 7

8 XO 2 O XN 2 TRay (K) Mixture Fraction (ξ) Mixture Fraction (ξ) Figure 4. Scatter Plots of Major Species Mole fraction and Temperature vs Mixture Fraction in the H3 Flame. Measurements were taken along the jet at x/d = 20 (r = 0.0, 1.5, 3.0, 4.5, 6.0, 7.5, 9.0, 10.5, 12.0, 13.5, 15.0, 16.5, 18.0, 19.5, 21.0, and 22.5 mm). Shown in the left-hand column are measurements taken in the high-speed Raman (red symbols) and adiabatic flame conditions (black lines). The stoichiometric mixture fraction (ξ stoich =.304) is marked with a vertical dotted line. For comparison, the same measurements taken at DLR (blue symbols) are shown in the right-hand column. 8

9 XO 2 O XN 2 TRay (K) Mixture Fraction (ξ) Mixture Fraction (ξ) Mixture Fraction (ξ) Figure 5. Conditionally averaged values of Major Species Mole Fraction and Temperature in the H3 flame. Measurements are shown for x/d = 10, 20, 40 with values measured in the high-speed Raman system plotted in red and measurements taken at DLR plotted in blue. The single-shot results are conditionally averaged within distinct mixture fraction invervals of.02. OSU measurements are taken in 300µm segments along 3.6 mm of the 1D line. The adiabatic flame conditions are shown as black lines. 9

10 XO 2 O XN 2 Mixture Fraction (ξ) TRay (K) Radius (mm) Radius (mm) Radius (mm) Figure 6. Radial Profiles of the Mean Values and RMS Fluctuations of Major Species Mole Fraction and Temperature in the H3 flame. Measurements are shown for x/d = 10, 20, 40 with values measured in the high-speed Raman system plotted in red and measurements taken at DLR plotted in blue. The solid markers represent the mean, while the open symbols represent the RMS fluctuations. OSU measurements are taken in 300µm segments along 3.6 mm of the 1D line and averaged over 300 laser pulses. DLR data is also averaged over 300 data points. The stoichiometric mixture fraction (ξ stoich =.304), marked with a horizontal dotted line, occurs at r = 8.6 mm for x/d = 10, r = 11.3 mm for x/d = 20, and does not occur x/d =

11 4. Conclusions Progress has been made towards the goal of quantitatively measuring all major combustion species and temperature at khz acquisition rates using combined 1D Raman/Rayleigh scattering. Measurements taken in a turbulent hydrogen jet flame (DLR H3) at 10 Hz utilizing the high-speed CMOS-based collection at Ohio State have shown strong agreement with point measurements taken at DLR Stuttgart (Meier et al., 1996) and available through the TNF database. Values of major species (H 2, N 2, H 2 O, and O 2 ) mole fraction and temperature show strong agreement in both mean and RMS along a 3.6mm line segment. While more calibration of the high-speed measurement system is needed at khz rates, future research will focus on integrating the high-energy pulse-burst laser system (HEPBLS) with the Raman/Rayleigh collection optics. Acknowledgements Support from the Air Force Office of Scientific Research (Dr. Chiping Li, Program Monitor) is greatly appreciated. The authors would like to thank R.S. Barlow, his research group, and the staff at the CRF for allowing the use of their facilities and helping with data collection for system calibrations and comparisons. The authors also kindly acknowledge collaboration and conversations M.J. Dunn and D. Geyer. K.N. Gabet acknowledges support from the Department of Energy (DOE) Office of Science Graduate Fellowship Program administered by the Oak Ridge Institute for Science and Education for the DOE. ORISE is managed by Oak Ridge Associated Universities (ORAU) under DOE contract number DE AC05 06OR All opinions expressed in this paper are the author s and do not necessarily reflect the policies and views of DOE, ORAU, or ORISE. References Barlow, R. S., & Frank, J. H. (1998). Effects of turbulence on species mass fractions in methane/air jet flames. Symposium (International) on Combustion, 27(1), Retrieved from Barlow, R., Wang, G., Anselmofilho, P., Sweeney, M., & Hochgreb, S. (2009). Application of Raman/Rayleigh/LIF diagnostics in turbulent stratified flames. Proceedings of the Combustion Institute, 32(1), Elsevier Inc. Retrieved August 23, 2011, from Barlow, Robert S., & Karpetis, A. N. (2005). Scalar length scales and spatial averaging effects in turbulent piloted methane/air jet flames. Proceedings of the Combustion Institute, 30(1), Retrieved September 14, 2011, from Böhm, B., Heeger, C., Gordon, R. L., & Dreizler, A. (2011). New Perspectives on Turbulent Combustion: Multi- Parameter High-Speed Planar Laser Diagnostics. Flow, Turbulence and Combustion, 86(3-4), Retrieved October 7, 2011, from Fuest, F., Barlow, R. S., Geyer, D., Seffrin, F., & Dreizler, a. (2011). A hybrid method for data evaluation in 1-D Raman spectroscopy. Proceedings of the Combustion Institute, 33(1), Elsevier Inc. Retrieved August 22, 2011, from Fuest, Frederik, Papageorge, M. J., Lempert, W. R., & Sutton, J. a. (2012). Ultrahigh laser pulse energy and power generation at 10 khz. Optics letters, 37(15), Retrieved from Gabet, K. N., Jiang, N., Lempert, W. R., & Sutton, J. a. (2010). Demonstration of high-speed 1D Raman scattering line imaging. Applied Physics B, 101(1-2), 1 5. Retrieved October 7, 2011, from Gabet, K. N., Patton, R. a., Jiang, N., Lempert, W. R., & Sutton, J. a. (2012). High-speed CH2O PLIF imaging in turbulent flames using a pulse-burst laser system. Applied Physics B, 106(3), Retrieved December 19, 2012, from 11

12 Geyer, D., Kempf, a., Dreizler, a., & Janicka, J. (2005). Scalar dissipation rates in isothermal and reactive turbulent opposed-jets: 1-D-Raman/Rayleigh experiments supported by LES. Proceedings of the Combustion Institute, 30(1), Retrieved September 14, 2011, from Gordon, R. L., Heeger, C., & Dreizler, a. (2009). High-speed mixture fraction imaging. Applied Physics B, 96(4), Retrieved October 5, 2011, from Karpetis, A.N., & Barlow, R. S. (2002). Measurements of scalar dissipation in a turbulent piloted methane/air jet flame. Proceedings of the Combustion Institute, 29(2), Retrieved from Karpetis, Adonios N., & Barlow, R. S. (2005). Measurements of flame orientation and scalar dissipation in turbulent partially premixed methane flames. Proceedings of the Combustion Institute, 30(1), Retrieved July 15, 2011, from Kojima, J., & Nguyen, Q.-V. (2002). Laser pulse-stretching with multiple optical ring cavities. Applied optics, 41(30), Retrieved from Meier, W., Barlow, R.., Chen, Y.-L., & Chen, J.-Y. (2000). Raman/Rayleigh/LIF measurements in a turbulent CH4/H2/N2 jet diffusion flame: experimental techniques and turbulence chemistry interaction. Combustion and Flame, 123(3), Retrieved from Meier, W., Prucker, S., Cao, M.-H., & Stricker, W. (1996). Characterization of Turbulent H2/N2/Air Jet Diffusion Flames by Single-Pulse Spontaneous Raman Scattering. Combustion Science and Technology, 118, Patton, R. a., Gabet, K. N., Jiang, N., Lempert, W. R., & Sutton, J. a. (2011). Multi-kHz mixture fraction imaging in turbulent jets using planar Rayleigh scattering. Applied Physics B, 106(2), Retrieved November 8, 2012, from Prucker, S., Meier, W., & Stricker, W. (1994). A flat flame burner as calibration source for combustion research: Temperatures and species concentrations of premixed H2/air flames. Review of Scientific Instruments, 65(9), Retrieved from Thurow, B., Jiang, N., & Lempert, W. (2013). Review of ultra-high repetition rate laser diagnostics for fluid dynamic measurements. Measurement Science and Technology, 24(1), Retrieved November 20, 2012, from Thurow, B., Jiang, N., & Samimy, M. (2004). Narrow-linewidth megahertz-rate pulse-burst laser for high-speed flow diagnostics. Applied optics, 43(26), Retrieved October 7, 2011, from TNF Workshop. (n.d.).. Retrieved October 18, 2011, from Wang, G., Karpetis, A. N., & Barlow, R. S. (2007). Dissipation length scales in turbulent nonpremixed jet flames. Combustion and Flame, 148(1-2), Retrieved September 22, 2011, from Weber, V., Brübach, J., Gordon, R. L., & Dreizler, a. (2011). Pixel-based characterisation of CMOS high-speed camera systems. Applied Physics B, 103(2), Retrieved September 15, 2011, from 12

13 Wehr, L., Meier, W., Kutne, P., & Hassa, C. (2007). Single-pulse 1D laser Raman scattering applied in a gas turbine model combustor at elevated pressure. Proceedings of the Combustion Institute, 31(2), Retrieved September 22, 2011, from Wu, P., Lempert, W. L., & Miles, R. B. (2000). Megahertz Pulse-Burst Laser and Visualization of Shock- Wave/Boundary-Layer Interaction. AIAA Journal, 38(4), Retrieved from 13