1. Introduction

In the combustion research field, Raman scattering has been widely applied as a non-intrusive combustion diagnostic technique [1]. Applied by itself or combined with Rayleigh scattering, the line-imaging of Raman scattering can provide high accuracy and high precision 1D measurements of temperature and major species (e.g., N$_{2}$, O$_{2}$, CO, CO$_{2}$, H$_{2}$, H$_{2}$O, and CHs) concentrations in hydrogen and hydrocarbon flames. Due to the multi-species nature of Raman scattering and the potential for single-shot measurements, the application of spontaneous Raman scattering has been successful in the research of turbulence-chemistry interaction. The application of Raman scattering measurements is not limited to the ambient pressure lab-scale flames [2–5], but also expanded to gas turbine model high-pressure combustor [6], and supersonic reacting flow [7]. To process the Raman scattering signal into species concentrations, the temperature dependence of Raman spectra must be account for. In the conventional "matrix inversion approach" [8,9] Raman data set is collected by a CCD camera with pre-defined on-chip binning Raman channels. The signal of each species channel consists of the spectrally integrated Raman signal, crosstalk from Raman spectra of other species, and other interferences (e.g., LIF and flame luminosity). The collected signal of each location is a vector $S$, which is the composition of the signal from each binning channel. The signal vector $S$ is related to the species number density vector $N$ with a matrix through the equation $S=C(T)N$. In the temperature dependent matrix, the diagonal element $c_{ii}(T)$ is the integrated Raman response of species $i$ inside its binning channel, whereas the off-diagonal element $c_{ij}(T)$ is the crosstalk from species $j$ to the channel of species $i$. The conventional approach to obtain the temperature dependent matrix requires an extensive empirical calibration. The calibration experiment needs to be conducted under several well defined temperature points and expanded to the intermediate value by polynomial fitting. With the knowledge of $C(T)$ matrix, the temperature and species concentrations can be derived. For an unknown flame, an initial temperature for iteration is guessed or obtained from Rayleigh scattering signal with a cross section of a guessed composition. By the inverse of the $C(T)$ matrix under guessed temperature to the signal vector $S$, the number density vector $N$ can be derived. Given the local number density of all the major species, the temperature can be updated by the ideal gas law or a new generated Rayleigh cross section. The iterative computation will be conduct until the convergence of temperature.

This matrix inversion approach makes the data processing of Raman scattering straightforward, however, there are still complications in the experiment. After the dispersion of grating, the line-imaging will experience a curvature distortion [10]. Except for applying specially designed optics to remove distortion [11], this kind of "bowing effect" is an intrinsic feature of imaging with grating. The bowing distortion will arise a shift to the Raman spectra. In the turbulent flames, the beam steering caused by the density gradient will introduce a pulse to pulse movement of the laser beam. Imaging after the grating, this movement will be in the spectral direction on the camera. Because the Raman binning channels are pre-defined and fixed, the movement is equivalent to a spectral shift of the Raman spectra. For every probe region these variations need to be accounted for in the $C(T)$ matrix, which is unattainable by only applying an empirical calibration. To get more accurate correction, Fuest et al. reported a "hybrid method" of 1D Raman scattering data processing [12]. The conversion matrix is calibrated at only a few temperature conditions, the temperature dependence is derived from calculated Raman spectral libraries. Then the shift of Raman spectra caused by beam steering and "bowing effect" can be simulated by moving the binning range on the spectral libraries. But the hybrid approach requires detailed temperature dependence information of Raman spectra. For most of the major species in the combustion applications, the temperature dependent Raman spectra can be calculated by using RAMSES (RAMan Scattering Experiment Simulator) code [13]. The Raman spectra of more complex molecules such as hydrocarbon fuels are not straightforward to be simulated. To extend the Raman measurement to heavy hydrocarbon fuels, the detailed Raman spectra of several kinds of hydrocarbon fuel (CH$_{4}$, C$_{2}$H$_{4}$, DME, etc.) were reported [14].

Recently, ammonia is increasingly considered as a clean carbon-free fuel. But there are still several challenges of ammonia combustion such as low flammability and high NOx emissions that are needed to be overcome [15]. Raman scattering is a powerful diagnostic tool that is capable of enabling a better understanding of ammonia combustion. However, to our knowledge, the application of Raman scattering in the ammonia flames is scarce. As mentioned above, the high accuracy Raman measurement requires the detailed temperature dependence of Raman spectra. The focus of this paper is the research of high temperature ammonia Raman spectra to explore the application of Raman scattering in ammonia flames. Although the Raman spectroscopy research of ammonia can be traced back to the early 1930s [16]. The spectra of liquid phase ammonia [17], aqueous ammonia solution [18], and gas phase [16] were widely researched. Under the gas phase, the Raman cross section of the $\nu _{1}$ [19] and $\nu _{2}$ vibrational mode [20] were measured, also the relative weak $\nu _{3}$ mode was resolved [21]. Nevertheless, most of these works of gas phase ammonia Raman spectra were reported decades ago. From the modern view, the quality and resolution of these spectra are inferior. Further, to the best of our knowledge, there are no experimental Raman spectra of ammonia reported in the temperature range of combustion research. In this paper, we acquired the Raman spectra of ammonia under temperature ranging from 299 K to 760 K with high dispersion (0.01 nm/pixel) to explore the Raman scattering in ammonia flames. To build up a quantified data library, the temperature of the probe region was measured with non-intrusive approach. The Raman spectra of ammonia under different temperatures were normalized by the number density derived from local temperature. The polarization property of different Raman transition bands are also critical for the application of the polarization separation approach to remove the interference from laser induced fluorescence [22]. So the polarization anisotropy of ammonia Raman scattering was also studied by acquiring the spectra along with two perpendicular polarization directions.

2. Experimental set-up

Figure 1 shows the experimental set-up employed in this work to obtain Raman spectra of ammonia at elevated temperatures. A 3.5 kW electric flow heater (Kanthal) controls the temperature of a flow of nitrogen (100 SLPM). A K-type thermocouple (TC1), inserted in the heater, monitored the gas temperature, and stabilized it with a closed-loop controller. An additional K-type thermocouple (TC2) monitored the nitrogen temperature at the exit of the heater. A Swagelok T-fitting was installed at the exit of the heater to mix a flow of gaseous ammonia (6 SLPM) with the heated nitrogen. A 20 cm insulated pipe guaranteed effective mixing of the two gases. The mixture flowed through a gas rectifier mesh to provide a more uniform velocity profile and reduce entrainment of ambient air and exited to the probe volume region through a 1-inch diameter tube. Nitrogen dilution greatly reduced the ammonia Raman signal but it was necessary. Ammonia is a highly toxic gas, and the 100 SLPM needed to avoid air entrainment, was deemed a safety risk, which was mitigated with high N$_{2}$ dilution. In addition, ammonia is corrosive, and no sufficient data were available to determine if the electric heater was compatible with it. Finally, ammonia dissociates at temperatures above 800 K, and even for gas exit temperatures below <800 K, near the heater surface, a higher temperature can be experienced leading to partial dissociation of the ammonia and making quantitative measurements more challenging.

 

Fig. 1. Experimental set-up of high temperature NH$_{3}$ Raman spectroscopy

Download Full Size | PDF

The left side of Fig. 1 shows the optical set-up. A narrow-band continuous wave laser (Coherent, Inc. Verdi G18) at 532 nm provided the Raman excitation source. The maximum power output was 18 Watt, and the linewidth was narrower than 30 GHz. A zero-order half-wave plate rotates the polarization of the laser beam to the direction normal to the plane defined by the beam propagation direction and the observation direction. A 500 mm lens focused the laser beam to the probe volume and a mirror deflected to a beam dump. The relatively long focal length provides a larger Rayleigh range, increasing the available probe volume. The distance from the laser beam to the outlet of heated flow was set to around 20 mm to avoid strong scattering from the metal surfaces, but sufficiently close to the exit to ensure a homogeneous mixture over several millimeters.

The Raman signal was collected by a Nikon micro lens (105 mm, f$\#$=2.8) at the minimum working distance. After the image plane, the signal beam was re-collimated by another DSLR lens (Samyang 135 mm, f$\#$=2.2), making the arrangement of other optical components more convenient and reducing signal losses. The anisotropy study required an independent collection of the two polarization components of the Raman spectra. This was achieved through a wire grid polarizer, placed after the collimating optics, to suppress one component of the polarization. A periscope rotated the Raman line image by 90$^{\circ }$, aligning it to the slit direction, and lowered the collimated beam to the entrance slit of the spectrometer. A 532 nm notch filter (Edmund Optics, OD 6), placed at the entrance of the spectrometer, suppressed interference from the Rayleigh scattering and the stray light. Because the grating efficiency is sensitive to the signal polarization, we added a half-wave plate (Thorlabs, WPH20ME-633) to preserve the polarization that maximizes the diffraction efficiency independently of the incoming signal polarization. Finally, a Nikon DSLR lens (50 mm, f$\#$=1.2) images the Raman signal to the slit on the spectrometer set to an opening of 10$\mu$m to maximize spectral resolution. The spectrometer employed in this work is a 320 mm focal length, astigmatism free Schmidt-Czerny-Turner spectrometer (Princeton Instruments, IsoPlane 320) with an f$\#$=4.6 aperture ratio. Two diffraction gratings were employed in this work: 600 lp/mm and 2400 lp/mm for low and high resolution respectively. An EMCCD camera (Princeton Instruments, ProEM:1600$\times$200), operated in low noise mode recorded the dispersed Raman spectra. The 1600 pixels in the spectral direction covered 3056 cm$^{-1}$ and 467 cm$^{-1}$ under low and high dispersion configuration respectively. The 200 pixels covered a probe volume of 8.6 mm. At a distance of 20 mm from the exit of the outlet, an 8.6 mm probe volume centered along the nozzle axis, is in the potential core of the jet, and temperature and composition are homogeneous. The laser beam size variation along the probe region estimated using the focusing depth function [23] is smaller than 0.7$\%$. Hardware binning along the spatial direction is highly desirable, as it can greatly increase the signal-to-noise ratio (SNR) without resorting to complex multi-pass arrangement as those used in Ref. [24,25]. In the previous experimental set-up for measurements of Raman spectra, hardware binning was limited to a fraction of the measurement volume [14], by optical aberrations introduced by the spectrometer, such as astigmatism and the bowing effect. The spectrometer employed in this work features an astigmatism free design, that removes these aberrations and allows hardware binning over the entire spectral region with no loss of resolution. Preliminary Rayleigh scattering measurements with no binning, showed negligible spectral shift and defocusing along the spatial direction, confirming that hardware binning would only increase the SNR, without compromising the spectral resolution. All the Raman measurements presented in this work were performed with full hardware binning in the spatial direction.

3. Results and discussion

In the experiment, the temperature of the heater was varied from 300 K to 900 K with a 50 K step. Under each setting temperature, the spectra were recorded after the temperature reading from TC2 was stable. For each temperature settings, 5 sets of data were acquired. The first dataset consists of high-dispersion N$_{2}$ spectra and provides a measurement of the gas temperature at the probe volume. The second dataset consists of low-dispersion (600 lp/mm grating) spectra of N$_{2}$ and NH$_{3}$ and it provides the relative Raman cross section of ammonia with respect to nitrogen and monitors the NH$_{3}$/N$_{2}$ ratio in the flow. The third dataset contains the high-resolution ammonia spectra, and it is the main objective of the work. The first and second datasets are repeated after the collection of ammonia spectra to verify the stability of the experimental set-up. Before the experiment, we performed a wavelength and intensity response calibration for the spectrometer with the IntelliCal calibration tool from Princeton Instruments to minimize the experimental error. For each dataset, background and bad pixel data were removed firstly, then we applied a median filter to the data sets to remove the random spikes in the spectra.

3.1 Probe region temperature measurements

In this work, the gas temperature at the probe volume is obtained from spectral fitting of the high-resolution N$_{2}$ spectra. A library of theoretical spectra of N$_{2}$ as a function of temperature was obtained by convolution of "stick spectra" generated by the RAMSES [13] code with an empirically determined instrument function. The Voigt shape instrument function was obtained by fitting the ambient temperature experimental spectrum with convoluted theoretical spectra with fixed temperature and floating Voigt shape parameters. A least square fitting algorithm between the experimental spectra and the spectra in the theoretical library provides the best-fit temperature reported. Figure 2(b) shows an example of the N$_{2}$ spectra fitting. The theoretical and experimental spectra have a good agreement for the Q, S, and O-branch of the N$_{2}$ ro-vibrational Raman transitions. The enlarged figure indicates that even for the very weak S-branch ($<1\%$ intensity of Q-branch) the fitting quality is satisfactory with residuals below 0.1$\%$ of maximum intensity. The maximum fitting residual is smaller than 2$\%$. Figure 2(a) compares results from the spectral fitting of the data taken before and after the acquisition of the ammonia spectra and the discrepancy is always smaller than $1.6\%$ of the mean value. The error bars in Fig. 2(a) represent $\pm$5 times standard deviations of temperature to make the figure more readable. Within each case, the maximum standard deviation of T_fit is smaller than $0.4\%$ of the mean value. For the remaining of this work, the probe volume temperature is defined as the mean value of T_fit before and after the acquisition of ammonia spectra.

 

Fig. 2. Measurements of probe region temperature. (a). Comparison between temperature measurement results before and after NH$_{3}$ Raman spectra acquisition. (b) Typical spectra fitting result with partial enlarged figure of S-branch

Download Full Size | PDF

3.2 High temperature Raman spectroscopy of ammonia

Because the NH$_{3}$ was indirectly heated by hot N$_{2}$, the mole fraction of NH$_{3}$ also affects the Raman intensity. To improve the accuracy of quantitative Raman spectra, the mole fraction of NH$_{3}$ was also monitored under all the temperature conditions. Low dispersion spectra, covering the spectral range of O$_{2}$, N$_{2}$, and NH$_{3}$ simultaneously were collected before and after the ammonia spectra measurement. Density normalized low-dispersion spectra are provided in the supplementary Dataset 1(Ref. [26]). The O$_{2}$, N$_{2}$, and NH$_{3}$ integrated Raman signals were defined as the integral in the range from 1495 cm$^{-1}$ to 1705 cm$^{-1}$, 2004 cm$^{-1}$ to 2620 cm$^{-1}$ and 3095 cm$^{-1}$ to 3870 cm$^{-1}$ respectively. For all temperatures tested, the O$_{2}$ signal was negligible, demonstrating that air entrainment in the probe volume was negligible. The ratio between the integrated N$_{2}$ and NH$_{3}$ signals provides a measurement of the ammonia mole fraction, neglecting temperature changes of the relative Raman cross section. The mean N$_{2}$/NH$_{3}$ ratio is 3.99 and for all case, the fluctuation of this ratio is within 1.5 $\%$ of the mean value with the exception of two outliers (out of 13) where a deviation of $\sim$4.0 $\%$ was observed, possibly caused by a drift of the mass-flow controllers. The N$_{2}$/NH$_{3}$ signal ratio provides a scaling coefficient to scale the intensity of the high-resolution spectra to account for the small changes in the ammonia mole fraction among the different test cases.

To keep the SNR roughly constant, during the acquisition of high resolution data, we used a variable exposure time for different temperature cases. The exposure time scales linearly with the inverse of the setting temperature with a 0.05 s/K slope. After acquisition, the spectra of different temperature cases were first normalized to the same exposure time and then corrected to the same effective number density by using the measured temperature (from spectral fitting) and ammonia mole fraction (from the low-dispersion N$_{2}$/NH$_{3}$). Under the high resolution configuration (2400 lp/mm grating, 0.01 nm/pixel), limited by the size of the sensor the spectra of NH$_{3}$ cannot be covered within one frame. The spectra presented in Fig. 3 are the glued spectra of two acquisitions centered at 3365 cm$^{-1}$ and 3691 cm$^{-1}$ respectively.

 

Fig. 3. Overall Raman spectra of ammonia within the temperature range from 299 K to 760 K.

Download Full Size | PDF

Ammonia is a symmetric top molecule with symmetrical pyramidal structure (point group $C_{3\nu }$). Nine parameters are required to fully describe the ro-vibrational transition of NH$_{3}$. The six vibrational degrees of freedom are labeled as the symmetric stretch($\nu _{1}$, 3336 cm$^{-1}$), the out-of-plane bending($\nu _{2}$, 932 cm$^{-1}$), the asymmetric stretch ($\nu _{3}$, 3443 cm$^{-1}$, doubly degenerated) and the asymmetric bending ($\nu _{4}$, 1626 cm$^{-1}$, doubly degenerated) [27]. For the degenerated modes $\nu _{3}$ and $\nu _{4}$, two additional quantum numbers, $l_{3}$ and $l_{4}$, are suffixed. Quantum numbers $J$ and $K$ represent the total angular momentum and its projection on the molecular symmetry axis respectively. The ninth quantum number $\Gamma$ is the total symmetry of NH$_{3}$ in the $C_{3\nu }$ point group. All the four vibrational modes are Raman active. Figure 3 shows that, in the spectral range of this work (3150 cm$^{-1}$ to 3810 cm$^{-1}$), $\nu _{1}$ mode, $\nu _{3}$ mode and the overtone of $\nu _{4}$ mode of NH$_{3}$ ( 3216 cm$^{-1}$) are detectable. The selection rules of transitions depend on the vibrational mode, such that within the symmetric vibrational mode, $\Delta J=0,\pm 1, \pm 2$ and $\Delta K=0$ are allowed, for the asymmetric vibrational mode, $\Delta J=0,\pm 1, \pm 2$ and $\Delta K=\pm 1,\pm 2$ are allowed. So the permitted transitions from each initial rotational state ($J$, $K$) are very rich, and the spectral lines strongly blend with each other. To study the detailed temperature dependence of NH$_{3}$ Raman spectra, data are plotted over a reduced spectral range (zoomed-in views) matching the $\nu _{1}$ (Fig. 4) and the $\nu _{3}$ and $2\nu _{4}$ (Fig. 5) vibrational bands. The original data of Fig. 4 and Fig. 5 are also available in the supplementary Dataset1 (Ref. [26]).

 

Fig. 4. Detailed structure of $\nu _{1}$ mode ranging from 299 K to 760 K

Download Full Size | PDF

As shown in Fig. 4, under low temperature the main peak of the $\nu _{1}$ mode is dominant, the Q-branch of low rotational states are the main components of this strong peak. The laser linewidth and the resolution of the spectrometer do not allow resolving the detailed structure of this main peak. However, we observe a shift of the peak of $\sim$2 cm$^{-1}$ toward lower Raman shift with increasing the temperature from 299 K to 760 K. To help the understanding of this shift, we referred to the infrared spectra of NH$_{3}$ [28], and labeled the line position of $\rm ^{Q}$Q transitions of (3,3) and (6,6) states in Fig. 4(a). The quantum numbers ($J$, $K$) of the initial states are labeled on the top, whereas the final states are on the bottom. Therefore, we attribute the blue shift of main Raman peaks to the increasing population of the states with higher rotational quantum number $J$. Under high temperature, the intensity of the two wings of the $\nu _{1}$ band also increase. Figure 4(b), shows the detailed structure of the left wing of $\nu _{1}$ mode under different temperatures. Because of the strongly blended lines, the peak structures resolved here are difficult to be assigned to specific transitions. According to the line list provided by TROVE program [29] and the hot ammonia infrared spectra in the BYTE database [30], the left side wing spectra overlap with $\rm ^{Q}$Q branch transitions of high $J$ states. The probable transitions with the wavelength near the peak structure are labeled on Fig. 4(b). The high temperature spectra indicate that, the Raman intensity of high $J$ states increase with the temperature. As presented in Fig. 4(c), the right side wing also increases with the temperature. The relative intensity of the right wing is lower than the left side. The line blending is more serious, under high temperature, the transition peaks can barely be resolved. We cannot assign these transitions to the spectral peaks without the knowledge of the polarizability of NH$_{3}$.

 

Fig. 5. Detailed structure of 2$\nu _{4}$ (a) and $\nu _{3}$ (b) band Raman spectra under different temperature

Download Full Size | PDF

The detailed high temperature Raman spectra of the overtone of $\nu _{4}$ mode are presented in Fig. 5(a). The strong lines of 2$\nu _{4}$ mode are relatively sparse when compared to the $\nu _{1}$ mode. For the lack of available theoretical Raman spectra of the 2$\nu _{4}$ band, it is hard to assign these peak structures in the 2$\nu _{4}$ band to specific transitions. Nevertheless, the temperature behavior of the 2$\nu _{4}$ band can be resolved in Fig. 5(a). The Raman peaks labeled as "A" type are sensitive to the temperature and decay rapidly, whereas, the intensity of the "B" type peaks are insensitive to the temperature. The relatively weaker "C" type peak increases with temperature, which shows an opposite temperature dependence with "A" type peaks. Under 760 K, the "C" type peak increase to a similar intensity of the "B" type peak around 3230 cm$^{-1}$. These temperature dependent behavior may be caused by the thermal population between different rotational states. The relative weaker $v_{3}$ band are presented in Fig. 5(b). The Raman shift of the $\nu _{3}$ band is mainly between 3400 cm$^{-1}$ to 3810 cm$^{-1}$. This band overlaps the Raman spectra of water at elevated temperatures, posing an additional challenge to the diagnostics of ammonia flames. The spectra are in agreement with the work published before [21], and we have identified eight strong $\rm ^{S}$S branches and two $\rm ^{R}$R branch transitions. Similar with the $\nu _{1}$ and 2$\nu _{4}$ band, transitions from low rotational states are sensitive to the temperature and decrease rapidly with the increasing of temperature. Meanwhile, as labeled in Fig. 5(b), the transitions starting from states $J=7$ to $J=9$ show low sensitivity to the temperature. Overall the density normalized signal in the $v_{3}$ band decreases with increasing temperature, leading to a reduction of the crosstalks of ammonia onto the water Raman channel under high temperature.

The strong temperature dependence of the ammonia Raman spectra makes it suitable for temperature measurements, in particular for the fuel side of diffusion flames using undiluted ammonia. The experimental spectra can be used to obtain temperature from spectral fitting, but this approach is often limited by the low SNR in practical configurations. For this reason, we analyzed the integrated the spectra over Raman channels associated to each detected mode. To boost the SNR, the integrated spectra of each band can be acquired by hardware spectral binning, while maintaining the temperature sensitivity. The spectral integrated range of $\nu _{1}$, 2$\nu _{4}$, and $\nu _{3}$ band were set as 3280-3370 cm$^{-1}$, 3165-3280 cm$^{-1}$, and 3370-3810 cm$^{-1}$ respectively. Figure 6 shows the integrated intensity ratio 2$\nu _{4}$/$\nu _{1}$ and $\nu _{3}$/$\nu _{1}$ as a function of temperature. The error bars represent the standard deviation computed from 15 measurements taken at each temperature condition. The ratio 2$\nu _{4}$/$\nu _{1}$ increases almost linearly with the temperature, with a coefficient of determination of fitting ($R^{2}$) equal to 0.9985. The ratio of 2$\nu _{4}$/$\nu _{1}$ linear increases $33\%$ from 299 K to 760 K, and the standard deviation of each temperature condition ranged from $0.4\%$ to $1.6\%$. This presents a good thermometry sensitivity for further applications. In addition the 2$\nu _{4}$ and the $\nu _{1}$ bands are spectrally separated from the Raman spectra of other species of interest in NH$_{3}$/Air flames, making their intensity ratio a suitable thermometry approach in the reactants and in the preheat zone. The $\nu _{3}$/$\nu _{1}$ ratio also shows a linear behavior with temperature but with a lower $R^{2}$ (0.797). The $\nu _{3}$/$\nu _{1}$ has a lower SNR than the 2$\nu _{4}$/ $\nu _{1}$ intensity ratio, leading to larger standard deviations from $1.0\%$ to $5.9\%$. In addition, the $\nu _{3}$ overlaps the Raman spectra of H$_{2}$O at elevated temperatures, making it unsuitable for measurements in the combustion environment.

 

Fig. 6. Ratio 2$\nu _{4}$/$\nu _{1}$, and $\nu _{3}$/$\nu _{1}$ under different temperature

Download Full Size | PDF

3.3 Polarization of characteristic ammonia Raman spectra

We also studied the polarization characteristic of NH$_{3}$ Raman. As mentioned above, the Raman signal was excited by a vertically polarized laser beam and collected both in the vertical and horizontal directions. To minimize the grating efficiency difference, a broadband half-wave plate was used to rotate the horizontal signal back to vertical again. The collected Raman signal in the vertical and horizontal directions are presented in Fig. 7. According to the work of Placzek [31], when the Raman transitions are excited with linearly polarized light, the value of the depolarization ratio of symmetric vibration $\rho _{\perp }/\rho _{\parallel }$ should be between 0 and 0.75. As shown in Fig. 7(a), the $\nu _{1}$ band and 2$\nu _{4}$ band are strongly polarized. After integration within the Raman channels associated to each band, as defined in the previous section, the depolarization ratio of $\nu _{1}$ and 2$\nu _{4}$ band are $\sim$0.03 and $\sim$0.1 respectively. The depolarization ratio of the 2$\nu _{4}$ overtone band is higher because it is a combination of symmetric and asymmetric transitions. Because the $\nu _{3}$ band is an asymmetric vibration, the depolarization ratio of $\nu _{3}$ band takes the maximum value 0.75, which is verified in Fig. 7(b). The $\nu _{3}$ band ro-vibrational transitions presented here have the depolarization ratio around 0.75, the fluctuation of $\rho _{\perp }/\rho _{\parallel }$ is caused by the inferior SNR of the horizontally polarized signal. With the help of this characteristic, the $\rm ^{O}$O branch transition (6,6)–(4,4) of $\nu _{3}$ mode can be resolved in Fig. 7(a). In applying Raman spectroscopy to ammonia flames, the major difficulties are the strong flame luminosity, a broadband laser-induced fluorescence, attributed prevalently to NH2 [32] , and the spectral overlap (crosstalk) between the Raman signal of ammonia and water. The polarization characteristics of the spectra reported here imply that the polarization separation method [22] is well suited to the study of ammonia flames. By subtracting the signal collected in two normal polarization directions, the depolarized fluorescence interference and the flame luminosity are removed, while the strong polarized $\nu _{1}$ and 2$\nu _{4}$ bands are preserved. Given the high-depolarization ratio of the $\nu _{3}$ band, the polarization separation greatly reduces the interference of ammonia to the H$_{2}$O Raman channel.

 

Fig. 7. Polarization of characteristic $\nu _{1}$ and 2$\nu _{4}$ band (a) and $\nu _{3}$ band (b) of ammonia Raman spectra

Download Full Size | PDF

4. Conclusion

We collected high resolution experimental Raman spectra of ammonia within the temperature range from 299 K to 760 K. The temperature of the probe region was measured by fitting the spontaneous Raman spectra of N$_{2}$ to a library of theoretical spectra. Under the ideal gas law assumption, the Raman spectra were normalized to the same number density. The detailed temperature behavior of the $\nu _{1}$, $\nu _{3}$ and 2$\nu _{4}$ band are reported and discussed. The relative intensity of the 2$\nu _{2}$/$\nu _{1}$ and $\nu _{3}$/$\nu _{1}$ mode under different temperature were also studied. The ratio between the 2$\nu _{4}$ band and $\nu _{1}$ band shows a strong linear relationship to the temperature, which has the potential to be developed as a thermometry approach in the fuel side of ammonia flames. The negative slope of $\nu _{3}$/$\nu _{1}$ to the temperature and the strong depolarization of $\nu _{3}$ band are also conducive to the application of Raman scattering on ammonia flames. In addition, the experimental data are provided in the supporting information for the application in combustion research and model validation of quantum mechanical calculation.