1. INTRODUCTION
Driven by the worldwide interest in high-speed transportation and hypersonic vehicles, related research on hypersonic aerodynamics has been expanding and undergoing multiple advances. Ground test facilities enable simulation of flight conditions over a range of flight altitudes and velocities. For the successful design of hypersonic vehicles, an accurate quantitative understanding of the flow around the vehicle is important. Hypersonic flight involves complex fluid mechanical phenomena, including ionization, dissociation, excitation, and non-equilibrium chemistry. For the development and validation of high-fidelity computational fluid dynamics simulations, accurate and reliable instrumentation is vital to characterize ground test facilities and to understand flow-fields around test articles.
The harsh flow environment in hypersonic conditions requires non-intrusive measurement techniques for off body flow measurements [1]. To date, a variety of non-intrusive thermodynamic property measurement techniques have been developed, including coherent anti-Stokes Raman scattering (CARS) [2–7], coherent Rayleigh Brillouin scattering [8–10], laser absorption spectroscopy [11–13], spontaneous Raman scattering [14–16], and quantitative laser-induced fluorescence [17,18]. These approaches can complement surface measurements such as pressure and temperature sensitive paints [19,20]. A comprehensive review of non-intrusive measurement techniques for hypersonic flow characterization can be found in Ref. [21].
In comparison to common old ordinary Raman spectroscopy (COORS), coherent techniques such as CARS provide significantly stronger signal levels and have been widely adopted. For low-density applications, colinear CARS is usually applied in order to provide sufficient signal strength. Colinear CARS is a short path integrated measurement. That aspect together with the quadratic nature of the integrated signal and associated line mixing and background interference makes extracting spatial information in complex environments difficult. The linearity of COORS simplifies data analysis and the geometry of the scattering enables 1D sampling, providing a profile across a region of interest.
Table 1. Free-Stream Equilibrium Chemistry Flow Condition
Fig. 1. Nozzle exit static pressure trace (blue) and detection window (magenta) during runs #448 (solid) and #461 (dashed). Note that the detection window is between 2 and 3 ms, identified by camera monitor signals.
COORS is generally used for vibrational spectroscopy [14,22,23]. However, due to its very low scattering cross sections, it is often unsuitable for measuring low-pressure gas properties such as species concentrations, gas density, or temperature [21,24]. The greater cross sections of rotational Raman compared to vibrational Raman are factors enhancing measurements in low-density environments. Rotational Raman scattering has been used mostly in plasma discharges [25,26] or for combustion research [27–29], but has seen limited applications to aerodynamics facilities. Examples of the spontaneous Raman scattering for gas flow diagnostics include works by Sharma et al. [14] where they detected vibrational Raman in an expanding flow in an electric arc-driven shock tube, by Locke et al. [15], and by Wernet et al. [16] where the rotational Raman was used to measure gas temperature of supersonic flow. Additionally, Pilverdier et al. [30] used vibrational Raman in a small-scale shock tube to obtain point temperature measurements and identified a range of experimental difficulties including strong background emission. Ramos et al. [31] used rotational Raman in a hypersonic free jet. However, these efforts were limited to small-scale facilities, and the diagnostic capability remained as a point measurement. To date, to the best of our knowledge, no demonstration of spatially extended measurements in large-scale, near-model hypersonic flow diagnostics have been made.
In hypersonic flow facilities, where the runtime can be as short as a few milliseconds, it is highly desirable to extract as much information as possible from each run [32], since potential discrepancies between runs can hinder accurate flow characterization. The cost of running facilities multiple times is also a non-negligible factor. Therefore, diagnostics that can capture spatial profiles at high repetition rates are highly desirable.
Recent advancements in light scattering detection with paired volume Bragg grating (VBG) filters by Bak et al. [33] have enabled highly efficient scattered light measurement over an extended 1D length. This technique was developed at the Aerospace Laboratory for Lasers, Electromagnetics, and Optics (ALLEMO) at Texas A&M University [34] and was successfully applied for rotational Raman scattering and Thomson scattering in various plasma environments such as glow discharge [33], plasma filament [35], nanosecond pulsed discharge [36], and hollow cathode discharges [37]. It is capable of resolving measurements over a several-millimeter (tunable) path length [33] with a spatial resolution determined by the imaging setup. Its 1D capability, high optical efficiency, and strong background suppression suggested that this approach could make COORS a viable option for hypersonic flow characterization where low densities and high temperatures have previously been limitations.
In this work, we present the first, to the best of our knowledge, 1D rotational Raman spectrum capturing the air density and temperature distribution across a bow shock of hypersonic Mach 6 flow, encompassing both free-stream and post-shock regions, during a single run of the hypervelocity expansion tunnel. Data were taken at a 100 kHz repetition rate, enabling the removal of contaminated images and multi-image averaging over the approximately 1 ms run time.
2. EXPERIMENTAL SETUP AND METHODS
A. Hypervelocity Expansion Tunnel at Texas A&M University and Test Conditions
A hypersonic flow was generated in the Hypervelocity eXpansion Tunnel (HXT) at the National Aerothermochemistry and Hypersonic Laboratory (NAHL) at Texas A&M University. The dimensions of the HXT test section are 1.8m(width)×1.5m(height) with a free-jet flow diameter of 0.98 m. HXT can replicate Mach numbers as high as 15 for “true-flight” enthalpy and as high as Mach 23 for a cold flow. The accessible testing condition envelope and further details of the HXT can be found in Ref. [32].
In this work, a blunt wedge model having a 7° half-angle and 5 mm radius of curvature leading edge was used as a test article. A Mach 6 condition was repeated for two runs, and each run was measured by different detector approaches—run (1) a single frame accumulation of multiple laser shots at 100 kHz during a portion of the run and run (2) shot by shot single frames at 100 kHz over the entire run. Details of the detector setups are given in Section 2.B. To estimate the flow condition in the HXT, a code based on 1D normal shock and isentropic flow equations was used [32]. Table 1 shows estimated free-stream conditions, where M∞ is a free-stream Mach number, h0 is a stagnation enthalpy, ρ∞ is a free-stream static density, T∞ is a static temperature, and Re∞ is a Reynolds number. Figure 1 presents a static pressure comparison during runs #448 (blue solid) and #461 (blue dashed) where COORS detection time windows (represented by magenta boxes between 2 and 3 ms) are overlapped.
B. 1D Common Old Ordinary Raman Scattering Setup
Figure 2 illustrates a schematic of the HXT test section with the model and the optical setup of 1D COORS. A list of equipment and optics used is given in Table 2. The equipment trigger block diagram and transmission efficiencies of optical elements are provided in Supplement 1. The total transmission efficiency through all optical elements is ηt∼0.3.
Fig. 2. Schematic of the experimental setup.
Bursts of laser pulses generated from a burst-mode 532 nm Nd:YAG laser (QuasiModo; Spectral Energies Inc.) are used as a probe beam. The bursts run at a 100 kHz repetition rate. The burst duration is variable up to 1.5 ms at the 100 kHz repetition setting, and the individual pulse duration is ∼8ns. The laser is single mode, having a line bandwidth of <100MHz, according to the manufacturer. The probing laser has a diameter of 7 mm. It was introduced vertically from a viewport on the top of the HXT, focused by a f=1000mm plano-convex lens, and positioned approximately 1.8 mm upstream from the blunt leading edge model. The scattered light is collected at the collection angle of 90° through a side viewport on the HXT test section.
The optic system is essentially two sets of image relay systems. The field of view is from the front leading edge tip to the free stream along the laser that is perpendicular to the flow. The object plane where the laser beam lies was relayed to the outside of the test section by 3′′ diameter achromatic f1=850mm and f2=300mm doublet lenses L1 and L2. An adjustable slit was placed at this image plane, acting as a spatial filter to limit stray light into the detector. The relayed image by the first set of the lenses was once again relayed into the spectrometer slit plane by a set of achromatic doublet lenses L3 and L4, a 1′′ diameter f3=50mm, and a 2′′ diameter f4=75mm. The total image magnification of the system was 0.53.
The second optical relay stage provides the collimated optical path between L3 and L4 where VBG narrow linewidth spectral blocking filters are placed. Note that the collimated part is important for a proper rejection of unwanted light by the VBG filters, which are angular-spectral filters. The VBG filters spectrally filter light, in this case stray light and Mie/Rayleigh scattering near 532 nm. Coordinated angle tuning of the two VBG filters was made in a way to extend the measurable region to generate a filtered 1D window. Sample laser images with/without the VBG filters are provided in Supplement 1. Further details of the principle and method can be found in Ref. [33]. A total of approximately 10 mm of laser Rayleigh scattering was successfully blocked by two filters, thus providing the Raman detection length of ∼10mm. The typical transmission efficiency of VBGs at non-filtered wavelength is ∼0.9.
Note that the VBG filters have a square aperture of 11mm×11mm, which acts as an aperture stop in the optical system. To minimize signal loss being physically blocked at the VBGs, it is important to properly choose parameters of L2 and L3. The VBG aperture area is equivalent to a circular aperture by a diameter ∅VBG=12.4mm. The chosen L2 has f-number f/#=300mm75mm=4, while L3 choice is matched to the f-number by the VBG f/#=f3∅VBG=50mm12.4mm≈4.
In front of the spectrometer, a half-wave plate, two long-pass filters, and two short-pass filters were placed. The half-wave plate is to maximize the signal throughput on the grating which has a polarization-dependent efficiency. The long-pass and short-pass filters are to block broadband fluorescence from the hypersonic flow behind the shock.
Fig. 3. Spectra images taken with an emiCCD camera (30 shots on chip accumulation). (a) 1D COORS of room temperature air at 50 Torr (calibration) and (b) 1D COORS for HXT run#448.
Fig. 4. Spectra images taken with a high-speed camera (a single frame-single laser shot, selected 30 frames average). (a) 1D COORS of room temperature air at 50 Torr (calibration) and (b) 1D COORS for HXT run #461.
The slit opening of the spectrometer was µ200µm. The spectrometer was configured with 2400 grooves/mm holographic VIS grating. The center wavelength was detuned to 538.5 nm for the following reasons: 1) to capture rotational Raman scattering from transitions at higher rotational states at high temperature, 2) to capture Stokes transitions which give a stronger signal, and 3) to minimize the residual Rayleigh scattering and strong Mie scattering at 532 nm from particulates existent in the flow. It should be noted that even with this detuning measure, direct laser interactions with particles produces very bright broadband radiation, causing horizontal streaks across the entire spectral domain at the particle location.
Two detector configurations were used to capture the spectrum: 1) a single frame-multiple laser shots detection (on-chip accumulation) and 2) a single frame-single laser shot detection (high-speed). For the first configuration an emiCCD camera (PI-MAX4:1024EMB; Princeton Instruments) was used for the system. The emiCCD camera frame rate is limited, only allowing a single frame reading for a run; however, it provides the most sensitive detecting power, which is beneficial to detect weak signals. The camera setup parameters are a gate opening of 13 ns and a gain of ×2,000. A total of 30 laser pulses (175 mJ per pulse) were accumulated for one raw spectrum image. For the second configuration, a high-speed camera (HPV-X2; Shimadzu Corporation) with an intensifier (HiCATT 18, P46, GaAsP; Lambert Instruments) is used. While the sensitivity of this system is less than the emiCCD camera, it allows the capture of individual laser shots and, thus, time-resolved measurements. The camera runs at twice the frequency (200 kHz) of the pulsed burst laser repetition rate (100 kHz, 243 mJ per pulse) to capture a background frame between laser pulses. The exposure time of the camera was 200 ns (minimum allowed) per frame, and the exposure time of the intensifier was 30 ns (minimum allowed).
Gas density and temperature were extracted from rotational Raman scattering spectra fitted with a theoretical curve. The number density calibration was made with room temperature air at 50 Torr filling of the HXT test section.
3. RESULTS AND DISCUSSION
Figure 3 show 1D COORS spectra images obtained using a system with an emiCCD camera (30 shots on-chip accumulation), where Fig. 3(a) presents the calibration image acquired with room temperature air at 50 Torr and Fig. 3(b) shows the image obtained during HXT run #448. The vertical dimension in the image is the 1D sample, and the horizontal dimension is the Stokes side of the rotational Raman spectrum. The spectrum is clipped to avoid camera saturation from residual Rayleigh, so the first few rotational lines are missing. The center line of the wedge model is positioned at y=0mm, and the laser beam passes ≈1.8mm in front of the leading edge. Both free-stream and post-shock properties were successfully obtained. Several horizontal streaks are notable, which are attributed to scattering from dust particulates in the flow. Despite this, most of the resolved 1D regions remain uncontaminated, providing reliable Raman spectra for extracting thermal profiles. This highlights the advantages of the 1D measurement technique. It should be noted that in some runs, excessive scattering from particulates caused over-saturation in the images, preventing the extraction of reliable spectral information. Due to the integration time lasting for a good part of the run and no pulse to pulse background subtraction, the accumulation approach also captures some residual natural emission behind shocks, which could make spectral analysis challenging for higher enthalpy runs.
Fig. 5. 1D distributions of (a) neutral density nn and (b) rotational temperature Trot for Mach 6 hypersonic flow (HXT runs #448 and #461), and comparison with calculated post-shock values under assumptions of calorically perfect gas (C.P.) or chemical equilibrium (CEA). (c) Locations of the wedge model, shock-wave, probing laser. A shock shape (solid blue) was obtained by the empirical relations [38]. The black dashed lines in all figures indicate the shock boundary y location along the probing laser based on the shock-shape.
Figure 4 shows 1D COORS spectra images obtained using a system with a high-speed camera (single frame-single laser shot; 30 frames average) where Fig. 4(a) presents the calibration image obtained with room temperature air at 50 Torr and Fig. 4(b) shows the image obtained during HXT run #461. The difference in the y range compared to Fig. 3 is due to the differences in camera sensor dimensions. A total of 128 frames were captured at 200 kHz, with half synchronized to the pulsed burst laser shots. To be consistent with the other detector system where 30 shots were accumulated, a total of 30 frames with laser shots (without strong particulate encounters and therefore no streaks) were selected and averaged. Frames immediately following each of the selected frames were also averaged as well and subtracted as background. By avoiding contaminated frames, a clean 1D COORS image was successfully obtained, clearly showing the boundary between the free-stream and the post-shock region.
Figure 5 shows the extracted 1D distributions of (a) neutral number density nn and (b) rotational temperature Trot for HXT runs #448 and #461 and comparison with calculated post-shock values under assumptions of a calorically perfect gas (C.P.) or chemical equilibrium (CEA), and Fig. 5(c) presents the wedge model, shock-wave shape and position, and probing laser position in the y−z plane. The measured values were obtained by fitting line spectra at each y location. Sample line spectra and fitting results are provided in Supplement 1. The neutral density is the sum of N2 and O2 number densities. The error bars represent the 95% confidence interval for the fitting. The calculated values were obtained for Mach 5.9 flow using the measured free-stream parameters and shock angles from the empirical relations by Billig [38]. The black dashed lines in all figures indicate the vertical shock boundary location along the probing laser. A noticeable discrepancy is the presence of several overestimated density values in run #448 compared to those in run #461, specifically near y=0.6,1.0,3.0, and 11.2 mm. These locations correspond to where the streaks were present. Although the intensity offset caused by the streaks was considered during the fitting process, some residual intensity effects appear to have led to these overestimations. Another difference between the two runs is the shock location in the y direction, which is likely due to a slight discrepancy in the probing laser position relative to the test model leading edge between the two runs. For example, an approximate 0.2 mm axial location difference could result in a ∼1mm difference in the vertical y location of the shock. Additionally, fluctuation of shock motion is also accountable. Quantifying the shock motion during the tunnel run will be a valuable addition in future work. Ignoring the outlier data points, the results from both runs show excellent agreement in both rotational temperature and neutral density. Averaging values from y=5mm to y=9mm yields free-stream nn,0=5.0±0.7×1023/m3 and Trot,0=206±37K for run #448, while nn,0=4.9±0.3×1023/m3 and Trot,0=183±17K for run #461. In the post-shock region, both nn,1 and Trot,1 exhibit increasing trends toward the model axis (at y=0mm), changing from nn,1=1.8 to 2.2×1024/m3 and from Trot,1=1300 to 1800 K. Compared to the calculated post-shock values, the measured nn,1 is ∼20−30% lower, and Trot,1 is closely matched. While the measured temperatures are only governed by the Raman spectra line profiles, the density results depend on the absolute measured intensities. We consider photocathode sensitivity depletion due to overexposure by particle scattering as the most likely contributor to the underestimation. Especially in the post-shock region where strong particulate scatterings were often encountered, we observe localized sensitivity depletion during the run. Such depletion was confirmed for frames recorded after frames encountering particulate scattering, which led to underestimated density compared to nearby locations, i.e., y=1.2mm and y=1.7mm. Several countermeasures such as reducing tunnel impurities or further shifted Stokes sideband windowing could be considered to improve the density measurement.
To assess time-resolved measurement capability, we carried out single-frame data analysis. Figure 6 presents sample single-frame spectra images taken by the high-speed camera system. Figures 6(a)–(d) show frames #22, 24, 26, and 28 (µt=110−140µs), and Figs. 6(e)–6(h) show frames #82, 84, 86, and 88 (µt=410−440µs). Note that frame #0 corresponds to µt=0µs. In a single frame, the number of photons scattered is sparse, so the sporadic distribution is notable. For the frames #26 and 86, weak broad overlaying intensity distributions (in the free-stream and the post-shock, respectively) are present, which is considered to be from a particulate scattering that was not too harsh to over-saturate the image. This affects mostly density analysis, leading to overestimation. For temporal analysis, to improve the signal-to-noise ratio, vertical (spatial; -y=4.36-9.29mm for the free-stream region and -y=0.62-3.05mm for the post-shock region) and horizontal (spectral; 0.2 nm per bin) binning were performed before the fitting process.
Fig. 6. Sample single-frame spectra images for Mach 6 hypersonic flow (HXT run #461). (a)–(d) Frames #22, 24, 26, and 28 (-µt=110-140µs) and (e)–(h) frames #82, 84, 86, and 88 (-µt=410-440µs). Frame #0 corresponds to µt=0µs.
Figure 7 plots extracted temporal variation of gas properties with the expected post-shock value (averaged over the post-shock region) under assumptions of calorically perfect gas (solid red) or chemical equilibrium (dotted red). Figure 7(a) shows neutral density nn, and Fig. 7(b) shows rotational temperature Trot. The free-stream region properties were presented by the blue circles, while the post-shock properties were by the magenta triangle. The error bars represent the 95% confidence interval for the fitting. Sample line spectra and fitting results are provided in Supplement 1. For the density analysis, the calibration factor was obtained for each binned spatial region, free-stream, and post-shock, from the image, Fig. 4(a). The presented data points correspond to the total 30 frames that were used in obtaining Fig. 4. Note that the particulate scattering in the individual frames as in Figs. 6(c) and 6(g) caused outlying density points (e.g., at t=130 and µ430µs). Especially in a single-frame analysis, the influence of such scattering can be considerable, so such frames were identified and excluded (gray points). On the contrary, it is noteworthy that the particulate scattering less affects temperature results. Temporal averages in the free-stream and post-shock regions were obtained and indicated as the dashed line.
Fig. 7. (a) Neutral density nn and (b) rotational temperature Trot as a function of time for Mach 6 hypersonic flow (HXT run #461). Calculated post-shock values under assumptions of calorically perfect gas (C.P.) or chemical equilibrium (CEA) are compared.
In the free stream, the temporal average of ⟨nn,0⟩=4.8±1.2×1023/m3 and ⟨Trot,0⟩=187±47K, and in the post-shock region, ⟨nn,1⟩=1.5±0.2×1024/m3 and ⟨Trot,1⟩=1380±223K are obtained, where brackets denote the temporal average of a variable’s data points with its standard deviation (the fitting uncertainty is not included). In the post-shock region, the underestimated measured density compared to the calculated values is notable again, which is likely due to the sensitivity depletion as mentioned. Meanwhile, the temperature shows remarkable agreement. The apparent overall temporal fluctuations are mostly attributed to low signal-to-noise ratios in single-frame analysis. Especially in the free-stream region, due to the low temperature and a limited spectral range that captures the rotational Raman lineshape, the relative uncertainty of Trot is large. Another factor is the energy fluctuation of burst laser pulses, which was approximately 6% based on a photodiode signal. When it comes to the fluctuation of the flow itself, note that, in previous work measuring single-shot velocity fluctuations in the HXT, the velocity behind the oblique shock front had about 5% standard deviation of fluctuations attributed to the acoustic noise propagating through the tunnel [39]. The NiiFTI method, used for the velocimetry, was demonstrated previously with a near 0.5% measurement uncertainty in a hypersonic environment [40]. To improve the accuracy of temporal measurement, larger aperture collection optics and laser pulse energy calibration per individual burst laser shot are advised. Nonetheless, the temporal analysis results still agree with the spatial analysis results within errors. The demonstrated temporal analysis indicates that 1D COORS can be utilized for high-speed time-resolved measurements and can be a method for characterization of tunnel flow properties and temporal dynamics.
4. CONCLUSION
We have successfully demonstrated 1D COORS on Mach 6 hypersonic flow run in the hypervelocity expansion tunnel at the National Aerothermochemistry and Hypersonic Laboratory at Texas A&M University. We have compared two detector systems. The system with the emiCCD camera provides a wider resolvable region with better sensor resolutions and sensitivity, while the nature of accumulation suffers from the particulate scatterings which result in streaks over the spectral domain. Despite this, the 1D capability still provides information at numerous spatial locations. With the high-speed camera, individual frames per pulsed burst laser shot were captured at double the laser pulse repetition rate. By selecting frames without particle generated streaks, we obtained a clean 1D COORS 30 pulse averaged spectral image. 1D spatial distributions of gas rotational temperature and neutral density across the shock were successfully measured. Apart from the outliers in a few local points caused by particulate scattering, both detecting methods showed excellent agreement in rotational temperature and neutral density across free-stream and post-shock regions.
The results presented demonstrate the promising robustness of 1D COORS, a new diagnostic capability for measuring spatial profiles in challenging ground test environments including those having very short run times. The demonstrated capability to suppress background and capture broad dynamic ranges of thermal properties, coupled with temporally and spatially resolved high-speed measurement capability, showcases the robustness of 1D COORS, unveiling a new paradigm for the use of spontaneous Raman scattering for gas diagnostics.
