Collinear FAST CARS for Chemical Mapping of Gases

remote spectroscopic mapping of gas emissions and leaks. Abstract: We examine the concentration dependence of the Coherent Anti-Stokes Raman Scattering (CARS) signal obtained for gas mixtures at various conditions using the Femtosecond Adaptive Spectroscopic Technique (FAST). We use the CARS signal of the Q-branch vibrational oscillation of molecular oxygen (1556 cm − 1 ) to conﬁrm the quadratic dependence of the coherent signal on the number of molecules in a test volume. In addition, we demonstrate multi-shot FAST CARS imaging of a gas ﬂow in free space by raster-scanning the area of interest.


Introduction
Both spontaneous and coherent Raman scattering effects have been employed in spectroscopic systems applied to trace detection and identification of chemicals [1,2] and as a method to measure concentrations of various components in gases [3][4][5], liquids and solids [6,7]. In spontaneous Raman scattering, the signal is directly proportional to the number of molecules interacting with a single beam of input light. By contrast, coherent anti-Stokes Raman scattering (CARS) is a nonlinear process where the molecules are first put into a state of coherent oscillations, resulting in an increased probability for a probe pulse to scatter and produce Stokes or anti-Stokes shifted photons [8]. Therefore, CARS-based techniques possess an inherent ability to offer higher signal strength and faster collection speed compared to approaches based on spontaneous Raman scattering [9].
With the development of ultrashort pulsed laser sources, CARS imaging opened the possibility to study the dynamics of such rapidly changing systems as gas flows [10]. The great flexibility of the CARS technique makes it a popular instrument for performing thermometry measurements for various gases: nitrogen and oxygen [11][12][13], carbon dioxide [14], as well as methane and hydrogen [15]. CARS techniques allow concentration and temperature measurements in harsh environments and at high temperature during the combustion process [16,17]. Moreover, one-dimensional [18,19], as well as two-dimensional [20] single-shot thermometry and mapping of reactants and products in gas flows can be performed by different experimental approaches. The collinear geometry configuration for all beams aids to simplify optical alignment and signal collection, with automatically satisfied phase-matching conditions in gases (Figure 1b). The wavelength (FWHM bandwidth) of the pump, Stokes pulses and probe beams are set to 556nm (13.6 nm), 610 nm (14.1 nm) and 806 nm (0.7 nm), with 3.1 μJ, 2.6 μJ and 0.9 μJ pulse energies respectively. When the probe delay is varied, the full hybrid CARS spectrogram can be observed (for example, see the graphical abstract for this paper). In present experiments, the probe delay with respect to the pump/Stokes pulses is fixed at an optimum values of 2.1 ps. The beams are combined by two dichroic beamsplitters (BS1,2) and focused by a convex lens L1 (f = 100 mm) in the homemade gas cell (Borosilicate glass, 1 mm thickness, 25.2 cm 3 , 8 cm long) filled with N2 and air gas mixture. The three beams and the generated CARS signal are collimated by another lens (L2, f = 100 mm) and filtered by a set of neutral density (FW2AND Thorlabs, Newton, NJ, USA) and bandpass filters (FF01-732/68 Semrock, Rochester, NY, USA) (ND+BPF). After passing through the filters, the beam is focused by the lens L3 (f = 100 mm) on the entrance slit of the spectrometer: a spectrograph (Chromex Spectrograph 250is, Albuquerque, NM, USA), which has a liquid-nitrogen-cooled CCD camera (CCD: uncoated Spec-10:400B, Princeton Instruments, Trenton, NJ, USA) attached.
The spectral resolution of the experimental setup is limited by two factors: resolution of the spectrometer and the probe pulse spectral width. The spectrometer resolution is determined by its slit size and was set to 0.12 cm −1 for all measurements, thus the total spectral resolution during experiments was mainly limited by the width of the probe pulse (≈11 cm −1 ). The probe pulse duration In this work, we experimentally demonstrate that the hybrid CARS signal obtained for molecular oxygen scales as the square of the number of molecules in two scenarios. We should note that the determination of molecular concentration based on the quadratic dependence of the CARS signal intensity on the number of molecules has been performed before using ns-CARS [34] and hybrid CARS [35] techniques. However, here we focus on demonstrating that this dependence can be maintained at different experimental conditions without prior calibration of specific gases composition or concentration. In addition, we provide a simple yet clear illustration of how the hybrid CARS technique can be exploited for imaging and sensing of a gas flow escaping from a nozzle by performing multi-shot measurements along with a raster-scanning an area in a collinear configuration. In the future, the same ideas can be applied to remote detection and sensing of trace amounts of gases [36]. Figure 1c shows the experimental setup for our oxygen concentration measurements. As a laser source, we employ a Ti:sapphire regenerative amplifier (1 kHz repetition rate, 0.73 mJ/pulse, Legend, Coherent, Santa Clara, CA, USA) with two equally pumped optical parametric amplifiers (OPAs: OPerA-VIS/UV and OPerA-SFG/UV Coherent, Santa Clara, CA, USA). The outputs of the two OPAs are used as the pump and Stokes pulses (<130 fs). A small fraction of the amplifier output passes through a homemade pulse shaper with an adjustable slit and is used as a probe beam. The adjustable slit width allows us to select a narrow spectral band (∆ν ≈ 11 cm −1 ) at 806 nm, which corresponds to about 2 ps pulse duration. The pump and the probe beams pass through delay stages (DS1,2), so that the probe pulse is time delayed with respect to the pump and Stokes pulses. Choosing time delay for the probe pulse along with its pulse shaping, provide a clear CARS signal with suppressed non-resonant background from the four-wave mixing signal generated by the three beams.

Experimental Setup
The collinear geometry configuration for all beams aids to simplify optical alignment and signal collection, with automatically satisfied phase-matching conditions in gases ( Figure 1b). The wavelength (FWHM bandwidth) of the pump, Stokes pulses and probe beams are set to 556nm (13.6 nm), 610 nm (14.1 nm) and 806 nm (0.7 nm), with 3.1 µJ, 2.6 µJ and 0.9 µJ pulse energies respectively. When the probe delay is varied, the full hybrid CARS spectrogram can be observed (for example, see the graphical abstract for this paper). In present experiments, the probe delay with respect to the pump/Stokes pulses is fixed at an optimum values of 2.1 ps. The beams are combined by two dichroic beamsplitters (BS1,2) and focused by a convex lens L1 (f = 100 mm) in the homemade gas cell (Borosilicate glass, 1 mm thickness, 25.2 cm 3 , 8 cm long) filled with N 2 and air gas mixture. The three beams and the generated CARS signal are collimated by another lens (L2, f = 100 mm) and filtered by a set of neutral density (FW2AND Thorlabs, Newton, NJ, USA) and bandpass filters (FF01-732/68 Semrock, Rochester, NY, USA) (ND+BPF). After passing through the filters, the beam is focused by the lens L3 (f = 100 mm) on the entrance slit of the spectrometer: a spectrograph (Chromex Spectrograph 250is, Albuquerque, NM, USA), which has a liquid-nitrogen-cooled CCD camera (CCD: uncoated Spec-10:400B, Princeton Instruments, Trenton, NJ, USA) attached.
The spectral resolution of the experimental setup is limited by two factors: resolution of the spectrometer and the probe pulse spectral width. The spectrometer resolution is determined by its slit size and was set to 0.12 cm −1 for all measurements, thus the total spectral resolution during experiments was mainly limited by the width of the probe pulse (≈11 cm −1 ). The probe pulse duration (≈2 ps) constrains the temporal resolution. The spatial resolution in transverse to the beams propagation direction is mainly limited by a beam focal spot diameter (estimated at ≈20 µm). However, the axial spatial resolution, i.e., in the direction of the beams propagation, can be potentially limited by the fact that the CARS signal is generated in the region up to 6 times larger than the Rayleigh range (≈0.4 mm for our beams configuration) [8] in the collinear beams configuration. This fact does not limit the concentration dependence measurements since the gas cell length is significantly larger than the estimated length of this region (2.4 mm).

Results and Discussion
In the first set of experiments we study the CARS signal of O 2 molecules by filling the gas cell with air at different pressures. As the first step towards optimizing the beam alignment and time delays for pump/probe pulses, we collect the CARS spectrum showing rotational-vibrational structure of oxygen molecule (Figure 2), where several peaks can be clearly distinguished. The signal was collected for 2.2 s at atmospheric pressure at 1 kHz. The main peak of the spectrum (1555.6 cm −1 ) corresponds to the Q-branch (∆J = 0) vibrational transition from the vibrational ground level of the oxygen molecule. Smaller peaks on the left and on the right to the Q-branch transition represent O-(∆J = −2) and S-(∆J = +2) transitions correspondingly [37]. oxygen molecule. Smaller peaks on the left and on the right to the Q-branch transition represent O-(∆J = −2) and S-(∆J = +2) transitions correspondingly [37]. Thereafter, we use the maximum CCD count value from the region of Raman shift 1556 ± 1 cm −1 , i.e., the intensity of the Q-branch vibrational line to examine the dependence of the CARS signal on gas concentration and pressure. Before calculating the maximum intensity, the zero-pressure background signal was subtracted from the data. We find using the maximum intensity to be more convenient and yet sufficiently precise for studying the dependence of the CARS signal on molecular concentration, when the absolute concentration of gas species is irrelevant. However, it is worth mentioning, that the fitting curves parameters for integrated intensity data points differ from the peak fittings by less than one standard deviation (4.6%).
During these measurements laser power fluctuations at 1 kHz repetition rate were below 0.5% and the signal was integrated for 5.2 s. The long integration time aids in significantly reducing the signal intensity fluctuations, but forces us to use at least ND = 3.0 filter at ambient conditions in order to avoid saturation of the CCD.
First, the cell is filled with a gas mixture of ambient air and pure nitrogen ( Figure 3, solid circles). We assume the oxygen presence in air to be 21% and vary the partial pressures of air and nitrogen while keeping the total pressure in the cell constant. These measurements allow us to study the dependence of CARS signal on the concentration of O2 with different amounts of background molecules (N2). In this case, the CARS signal scales as a square of the relative oxygen concentration.
Next, we fill the cell with ambient air at various pressures, keeping the gas composition and ratio of oxygen molecules to background molecules constant. We control the total pressure inside the gas cell by a ball valve, and perform measurements for the range from 0.01 bar to 1.07 bar. The experimental data with a fit curve are shown in Figure 3 (hollow diamonds), where the signal is proportional to the square of the gas pressure. Hence, we conclude that in both cases of constant gas mixture at different pressures and varied gas compositions at constant pressure, the CARS signal is proportional to the square of the number of O2 molecules. Thereafter, we use the maximum CCD count value from the region of Raman shift 1556 ± 1 cm −1 , i.e., the intensity of the Q-branch vibrational line to examine the dependence of the CARS signal on gas concentration and pressure. Before calculating the maximum intensity, the zero-pressure background signal was subtracted from the data. We find using the maximum intensity to be more convenient and yet sufficiently precise for studying the dependence of the CARS signal on molecular concentration, when the absolute concentration of gas species is irrelevant. However, it is worth mentioning, that the fitting curves parameters for integrated intensity data points differ from the peak fittings by less than one standard deviation (4.6%).
During these measurements laser power fluctuations at 1 kHz repetition rate were below 0.5% and the signal was integrated for 5.2 s. The long integration time aids in significantly reducing the signal intensity fluctuations, but forces us to use at least ND = 3.0 filter at ambient conditions in order to avoid saturation of the CCD.
First, the cell is filled with a gas mixture of ambient air and pure nitrogen ( Figure 3, solid circles). We assume the oxygen presence in air to be 21% and vary the partial pressures of air and nitrogen while keeping the total pressure in the cell constant. These measurements allow us to study the dependence of CARS signal on the concentration of O 2 with different amounts of background molecules (N 2 ). In this case, the CARS signal scales as a square of the relative oxygen concentration.
Next, we fill the cell with ambient air at various pressures, keeping the gas composition and ratio of oxygen molecules to background molecules constant. We control the total pressure inside the gas cell by a ball valve, and perform measurements for the range from 0.01 bar to 1.07 bar. The experimental data with a fit curve are shown in Figure 3 (hollow diamonds), where the signal is proportional to the square of the gas pressure. Hence, we conclude that in both cases of constant gas mixture at different pressures and varied gas compositions at constant pressure, the CARS signal is proportional to the square of the number of O 2 molecules. In the next set of experiments, we replace the gas cell with a 1 mm round nozzle connected to a cylinder with N2 gas and pointed in the direction perpendicular to the beams propagation. Hereafter, we obtain the CARS signal for O2 molecules in ambient air in front of the nozzle in the vicinity of the focal plane of the beam (Figure 4), where zero of the X-axis corresponds to the nozzle surface and zero of the Y-axis to the center of the nozzle. After we set the gas pressure such that the gas flow from the nozzle remains constant (estimated at ≈4.5 m/s), we move the nozzle in a direction transverse to the axis of the beam propagation direction (Y and X axes in the figure). Hence, by obtaining the CARS signal for O2 at various nozzle positions it becomes possible to visualize the flow of nitrogen from the nozzle. One can see that the nitrogen flow stays almost symmetric about the zero of the Y-axis as it propagates away from the nozzle, where the slight slope can be due to a tilt of the nozzle. The CARS signal from oxygen in the central part of the flow increases very slowly, and the nitrogen diffuses into the surrounding air. We then add a thin metal plate (50 × 20 × 0.5 mm) in front of the nozzle to examine the resulting flow disturbance ( Figure 5). The plate is placed at a distance 5.9 mm away from the nozzle surface in such a manner that the top half of the nozzle is blocked by the plate. The thin plate acts as an In the next set of experiments, we replace the gas cell with a 1 mm round nozzle connected to a cylinder with N 2 gas and pointed in the direction perpendicular to the beams propagation. Hereafter, we obtain the CARS signal for O 2 molecules in ambient air in front of the nozzle in the vicinity of the focal plane of the beam (Figure 4), where zero of the X-axis corresponds to the nozzle surface and zero of the Y-axis to the center of the nozzle. After we set the gas pressure such that the gas flow from the nozzle remains constant (estimated at ≈4.5 m/s), we move the nozzle in a direction transverse to the axis of the beam propagation direction (Y and X axes in the figure). Hence, by obtaining the CARS signal for O 2 at various nozzle positions it becomes possible to visualize the flow of nitrogen from the nozzle. One can see that the nitrogen flow stays almost symmetric about the zero of the Y-axis as it propagates away from the nozzle, where the slight slope can be due to a tilt of the nozzle. The CARS signal from oxygen in the central part of the flow increases very slowly, and the nitrogen diffuses into the surrounding air. In the next set of experiments, we replace the gas cell with a 1 mm round nozzle connected to a cylinder with N2 gas and pointed in the direction perpendicular to the beams propagation. Hereafter, we obtain the CARS signal for O2 molecules in ambient air in front of the nozzle in the vicinity of the focal plane of the beam (Figure 4), where zero of the X-axis corresponds to the nozzle surface and zero of the Y-axis to the center of the nozzle. After we set the gas pressure such that the gas flow from the nozzle remains constant (estimated at ≈4.5 m/s), we move the nozzle in a direction transverse to the axis of the beam propagation direction (Y and X axes in the figure). Hence, by obtaining the CARS signal for O2 at various nozzle positions it becomes possible to visualize the flow of nitrogen from the nozzle. One can see that the nitrogen flow stays almost symmetric about the zero of the Y-axis as it propagates away from the nozzle, where the slight slope can be due to a tilt of the nozzle. The CARS signal from oxygen in the central part of the flow increases very slowly, and the nitrogen diffuses into the surrounding air. We then add a thin metal plate (50 × 20 × 0.5 mm) in front of the nozzle to examine the resulting flow disturbance ( Figure 5). The plate is placed at a distance 5.9 mm away from the nozzle surface in such a manner that the top half of the nozzle is blocked by the plate. The thin plate acts as an We then add a thin metal plate (50 × 20 × 0.5 mm) in front of the nozzle to examine the resulting flow disturbance ( Figure 5). The plate is placed at a distance 5.9 mm away from the nozzle surface in such a manner that the top half of the nozzle is blocked by the plate. The thin plate acts as an impenetrable barrier/obstacle for the gas and laser beams. The flow is disturbed and no longer symmetric; points with zero CARS signal mark the plate location. Moreover, the signal decreases in front of the plate as the nitrogen flow is partially redirected along the surface, while another portion of the flow is deflected by the plate downward. Right behind the barrier, the signal is restored to its value in ambient air since the nitrogen flow cannot penetrate through the plate; i.e., the air in this region stays undisturbed.
Appl. Sci. 2017, 7, 705 6 of 8 impenetrable barrier/obstacle for the gas and laser beams. The flow is disturbed and no longer symmetric; points with zero CARS signal mark the plate location. Moreover, the signal decreases in front of the plate as the nitrogen flow is partially redirected along the surface, while another portion of the flow is deflected by the plate downward. Right behind the barrier, the signal is restored to its value in ambient air since the nitrogen flow cannot penetrate through the plate; i.e., the air in this region stays undisturbed.

Conclusions
We have experimentally demonstrated the quadratic dependence of the CARS signal of the Qbranch vibrational transition of molecular oxygen on the number of O2 molecules in a gas mixture. We examined two cases. In the first, the mixture was prepared with various amount of background molecules but at constant total pressure. In the second case, the mixture composition remained unchanged while the pressure was varied. No significant difference between these cases was found as both signals showed quadratic dependence on the number of molecules. Furthermore, we have illustrated how CARS can be used for the visualization of gas flow in a simple, free-space configuration, both with a plate barrier and without it. We believe this method is applicable for performing gas flow images for any molecules with any Raman-active modes, as long as the CARS signal can be retrieved with a suppressed non-resonant background.

Conclusions
We have experimentally demonstrated the quadratic dependence of the CARS signal of the Q-branch vibrational transition of molecular oxygen on the number of O 2 molecules in a gas mixture. We examined two cases. In the first, the mixture was prepared with various amount of background molecules but at constant total pressure. In the second case, the mixture composition remained unchanged while the pressure was varied. No significant difference between these cases was found as both signals showed quadratic dependence on the number of molecules. Furthermore, we have illustrated how CARS can be used for the visualization of gas flow in a simple, free-space configuration, both with a plate barrier and without it. We believe this method is applicable for performing gas flow images for any molecules with any Raman-active modes, as long as the CARS signal can be retrieved with a suppressed non-resonant background.